Professional Documents
Culture Documents
DOI 10.1007/s10853-007-1952-8
Abstract The mechanical properties of biological materi- years, partly due to the emergence of physical science
als have been the focal point of extensive studies over the past based approaches in the biological sciences, leading to
decades, leading to formation of a new research field that cross-disciplinary investigations of materials, structures,
intimately connects biology, chemistry and materials sci- diseases as well as the development of new treatment and
ence. Significant advances have been made in many disci- diagnostics methods.
plines and research areas, ranging throughout a variety of This special issue of the Journal of Materials Science
material scales, from atomistic, molecular up to continuum entitled Nano- and micromechanical properties of hier-
scales. Experimental studies are now carried out with archical biological materials: Linking mechanics, chemis-
molecular precision, including investigations of how try and biology constitutes a snapshot of the current
molecular defects such as protein mutations or protein status, new directions and research opportunities in this
knockout influence larger length- and time-scales. Simula- field. The focus of the collection of papers is on experi-
tion studies of biological materials now range from electronic mental, computational and theoretical efforts that contrib-
structure calculations of DNA, molecular simulations of ute to a more quantitative understanding of biological
proteins and biomolecules like actin and tubulin to contin- materials and the interaction of biological materials with
uum theories of bone and collagenous tissues. The integration their environment, across various scales. This includes the
of predictive numerical studies with experimental methods deformation and fracture behavior of biological materials,
represents a new frontier in materials research. The field is at with a particular focus on nanoscale features and materials,
a turning point when major breakthroughs in the under- and its relationship with diseases.
standing, synthesis, control and analysis of complex biolog- The issue also contains contributions that focus on the
ical systems emerge. Here we provide a brief perspective of synthesis and growth of biological tissues and materials, in
the state of this field and outline new research directions. particular such methods that advance the ability to control
nano- and micro-structural features more quantitatively.
Individual contributions emphasize on linking the chemical
or molecular, and mesoscopic structures of these materials
Introduction to macroscopic engineering properties, across various
scales, including the impact of genetic mutations, solvent
The field of mechanical properties of biological materials conditions and other chemical stimuli on the macroscopic
has seen an exciting development over the past several behavior of the material. Specific topics include elasticity,
deformation and fracture of biological and biomimetic
materials including bone, tissues and scaffolding materials,
M. J. Buehler (&) interactions of cells with materials under mechanical load-
Laboratory for Atomistic and Molecular Mechanics, Department ing, mechanics of hierarchical biological materials as well
of Civil and Environmental Engineering, Massachusetts Institute
as the mechanics of single molecules. Development of
of Technology, 77 Massachusetts Ave. Room 1-272, Cambridge,
MA, USA experimental protocols to study the human stratum corne-
e-mail: mbuehler@MIT.EDU um, addressing the role of misfolded proteins in Alzheimers
123
8766 J Mater Sci (2007) 42:87658770
disease, development of theories of the properties of viral influence of tissue remodelling become more evident. The
capsids and elucidation of the statistical mechanics princi- dominance of specific mechanisms is controlled by geo-
ples of stretching biopolymers in geometric confinement metrical parameters as well as the structural arrangement of
represent a broad range of topics. the protein elementary building blocks, across many hier-
archical scales, from nano to macro (Fig. 1).
It is known from other fields in materials science that
Materials science of biological materials: challenges and nano- or microscopic structures control the macroscopic
opportunities material behavior: For example, grain size reduction or
confinement leads to an increase of the strength of crystal-
Historically, the use of classes of materials has been used line metals [36]. Deformation maps have been proposed to
to classify stages of civilizations, ranging from stone age characterize material properties for engineering applications
more than 300,000 years ago, to the bronze age, and pos- [7]. Discovering similar insight for biological structures and
sibly the silicon age in the late 20th and early 21st century. materials represents and important frontier of research. A
However, a systematic analysis of materials in the context particularly challenging question is the elucidation of the
of linking chemical and physical concepts with engineering significance and role of nanostructures for macroscopic
applications has not been achieved until very recently. For properties, that is, carrying out sensitivity analyses that show
instance, 50 year ago, E. Orowan, M. Polanyi and G.I. how small-scale features influence larger scale properties.
Taylor have discovered dislocations, a concept proposed A major trait of biological materials is the occurrence of
theoretically in 1905 by V. Volterra. It was discovered that hierarchies and, at the molecular scale, the abundance of
dislocations represent the fundamental mechanism of weak interactions (e.g. H-bonds). The presence of hierar-
plastic deformation of metals [1, 2]. Remarkably, it was not chies in biological materials may be vital to take advantage
until dislocations and other nano- and microscropic of molecular and sub-molecular features, often character-
mechanisms have been understood theoretically that major ized by weak interactions, and multiply their properties so
breakthroughs have been possible that utilize this knowl- that they become visible at larger scales, in order to provide
edge, to enable building airplanes, cars, space shuttles and a link between structural organization and function [8].
more recently, nanodevices, through synthesis of ultra- Utilization of weak interactions makes it possible to pro-
strong and heat resistant materials, for instance. duce strong materials at moderate temperatures and thus
Perhaps, today we stand at another cross-road: Biological with limited energy use. An important distinction between
materials and systems are vital elements of life, and there- traditional and biological materials is the geometrical
fore, a rigorous understanding of the matter that makes life occurrence of defects. While defects are often distributed
work is in reach. This may enable us eventually to inte- randomly over the volume in crystalline materials, biolog-
grate concepts from living systems into materials design, ical materials consist of an ordered structure that reaches
seamlessly. Optical, mechanical and electrical properties at down to the nano-scale. In many biological materials,
ultra-small material scales, their control, synthesis and defects are placed with atomistic or molecular precision,
analysis as well as their theoretical description represent and may play a major role in the material behavior observed
major scientific and engineering challenges and opportuni- at larger scales. These features have been observed in bone,
ties. However, just like in the case of more conventional nacre, collagenous tissue or cellular protein networks.
materials, these breakthroughs will probably only be The mechanical properties of biological materials have
accessible provided that the fundamentals are understood wide ranging implications for biology. In cells for instance,
very well. Characterization of the materials found in biology mechanical sensation is used to transmit signals from the
within a rigorous materials science approach is aimed environment to the cell nucleus in order to control tissue
towards the elucidation of these fundamental principles of formation and regeneration [9, 10]. The structural integrity
assembly, deformation and fracture of these materials. and shape of cells is controlled by the cells cytoskeleton,
Deformation and fracture properties are intimately which resembles an interplay of complex protein structures
linked to the atomic microstructure of the material. and signaling cascades arranged in a hierarchical fashion
Whereas crystalline materials show mechanisms such as [9]. Bone and collagen, providing structure to our body, or
dislocation spreading or crack extension, biological mate- spider silk, used for prey procurement, are examples of
rials feature molecular unfolding or sliding, with a partic- materials that have incredible elasticity, strength and
ular significance of rupture of chemical bonds such as robustness unmatched by many man-made materials,
hydrogen bonds, covalent cross-links or intermolecular mainly attributed to its structural formation with molecular
entanglement. Much different mechanisms operate at larger precision [1119]. The transfer of concepts observed in
length scales, where the interaction of molecules with cells biology into technological applications and new materials
and of cells with one another, different tissue types and the design remains a big challenge, with potential huge payoff.
123
J Mater Sci (2007) 42:87658770 8767
Fig. 1 Overview over different material scales, from nano to macro, mation and fracture mechanisms, it is crucial to elucidate atomistic
here exemplified for collagenous tissue [1115]. Biological materials and molecular mechanisms at each scale, and to appreciate the cross-
such as collagen, skin, bone, spider silk or cytoskeletal networks in scale interaction of these mechanisms. Our ability to synthesize,
cells feature complex, hierarchical structures. The macroscopic characterize and control such materials in their native environment or
mechanical material behavior is controlled by the interplay of in technological applications depends critically on the theoretical
properties throughout various scales. In order to understand defor- foundation of its mechanical behavior
123
8768 J Mater Sci (2007) 42:87658770
123
J Mater Sci (2007) 42:87658770 8769
while retaining a system theoretical perspective [2830], may enhance the quality of life of millions of people,
maybe eventually leading to a merger between system through advances in the medical sciences as well as
biology and materials science. through improvements of the living environment. A cur-
Similar to dislocation mechanics for metal plasticity, rently pressing question is the development of new tech-
what is the theoretical framework for biological materials nologies to address the energy problem. Advances may be
and structures? It is possible that statistical theories may possible by utilization of bacteria to produce and process
evolve into the theoretical language of nanomechanics. fuel from crops, or by enabling the synthesis of materials at
Atomistic simulations of complex protein structures with reduced processing temperature.
explicit solvents are often prohibitive, and coarse-graining Nanoscience and nanotechnology enable us to make
techniques are often used. However, how effective are structures at the ultimate scale (self assembly, recombinant
coarse-graining techniques? Can we indeed average out DNA, utilization of motor proteins for nano-machines and
over atomistic or mesoscale structures? How important are many others). This will perhaps lead to novel complex
atomistic features at macroscale? What are the best structural materials, designed from nano to macro. The
numerical strategies to simulate the role of water in very theoretical progress in understanding hierarchical biologi-
small confinement? How does confined water influence the cal materials will facilitate to use an extended physical
mechanics of natural and biological materials? space, through the use of multiple hierarchies, in an effi-
Progress in these various challenging fields will proba- cient and controlled manner, that is, lead to a bottom-up
bly occur specific to problems and applications, perhaps in structural design on the sub-macroscopic scale, instead of
those have most impact in medical or economic fields. trial-and-error approaches. For example, the extended
Eventually, we must generalize our insight into the hierarchical design space might serve as means to realize
formulation of a holistic theory that extends the current new physical realities that are not accessible to a single
nomenclature, theory and experimental thinking. These scale, such as material synthesis at moderate temperatures,
efforts will provide the scientific and engineering funda- or fault tolerant hierarchical assembly pathways [33],
mentals to develop and maintain the infrastructures to which enable biological systems to overcome the limita-
enable and evolve modern civilization. Materialsand tions to particular chemical bonds (soft) and chemical
materials sciencewill surely play a seminal role in these elements (organic) present under natural conditions [27].
developments. The increased understanding of the hierarchical design
laws might further enable the development and application
of new organic and organic-inorganic multi-featured
Future directions, challenges and impact composites (such as assemblies of carbon nanotubes and
proteins or polymer-protein composites [3436]), which
Over the last centuries, engineers have developed will mainly consist of chemical elements that appear in our
understanding of how to create complex man-made environment in an almost unlimited amount (C, H, N, O,
structures out of a diverse range of constituents, at var- S). These materials might consequently help to solve
ious scales (machines, buildings, airplanes, nuclear humans energy and resource problems (e.g. fossil
reactors and many others). Increased development and resources, iron etc.), and allow us to manufacture nano-
research funding into these areas of research will lead to materials, which will be produced in the future by tech-
breakthroughs not only on the fundamental sciences, but niques like recombinant DNA [31, 37, 38] or peptide self-
also in technological applications. Research in the area assembly [32, 39, 40], techniques where the borders
of mechanics of biological materials will extend our between materials, structures and machines vanish.
ability to carry out structural engineering, as used for Applications of these new materials and structures are
buildings or bridges today, to the ultimate scalenano- new biomaterials, new polymers, new composites, engi-
scale, and may be a vital component of the realization of neered spider silk, new scaffolding tissues, improved
nanotechnology. understanding of cell-ECM interactions, cell mechanics,
A better understanding of the mechanics of biological hierarchical structures and self-assembly. In addition to the
and natural materials, integrated within complex techno- long-term impact in biology, bioengineering and medicine,
logical systems will make it possible to combine living and this research may eventually contribute to our theoretical
non-living environments to develop sustainable technolo- understanding of how structural features at different scales
gies. New materials technologies such as protein-based interact with one another. In light of the extended physical
materials produced by recombinant DNA techniques rep- design space discussed above, this may transform engi-
resent new frontiers in materials design and synthesis [31, neering approaches not only for materials applications,
32]. These questions have high impact in the understanding but also in manufacturing, transportation or designs of
and design of environmentally friendly technologies and networks.
123
8770 J Mater Sci (2007) 42:87658770
Acknowledgements Id like to sincerely thank all authors and 17. Doyle J (2007) Nature 446:860
contributors to this special issue for submitting excellent illustrations 18. Kitano H (2002) Nature 420(6912):206
of exciting research work in this field to this special issue. I am 19. Kitano H (2002) Science 295(5560):1662
delighted to have such a strong collection of papers from outstanding 20. Smith BL et al (1999) Nature 399(6738):761
contributors, from many different countries, throughout a range of 21. Prater CB, Butt HJ, Hansma PK (1990) Nature 345(6278):839
scientific disciplines. 22. Sun YL et al (2004) J Biomech 37(11):1665
23. Dao M, Lim CT, Suresh S (2003) J Mech Phys Solids 51(11
12):2259
24. Tai K, Ulm FJ, Ortiz C (2006) Nanogranular origins of the
References
strength of bone. Nano Lett 6:2520
25. Lim CT et al (2006) Mater Sci Eng C-Biomimetic Supramol Syst
1. Taylor GI (1934) Proc Roy Soc A 145:362 26(8):1278
2. Hirth JP, Lothe J (1982) Theory of dislocations. Wiley-Inter- 26. Goddard WA (2006) A Perspective of materials modeling In: Yip
science, Hoboken, NJ S (ed) Handbook of materials modeling. Springer, New York
3. Wolf D et al (2003) Z Metallk 94:1052 27. Ackbarow T, Buehler MJ Hierarchical coexistence of universality
4. Yip S (1998) Nature 391:532 and diversity controls robustness and multi-functionality in
5. von Blanckenhagen B, Gumbsch P, Arzt E (2001) Modelling protein materials (unpublished)
Simul Mater Sci Eng 9:157 28. Bell GI (1978) Science 200(4342):618
6. Nieh TG, Wadsworth J (1991) Scripta Met 25(4):955 29. Lakes R (1993) Nature 361(6412):511
7. Frost HJ, Ashby MF (1982) Deformation-mechanism maps. 30. Bustamante C et al (1994) Science 265(5178):1599
Pergamon Press, Oxford 31. Langer R, Tirrell DA (2004) Nature 428(6982):487
8. Fraser P, Bickmore W (2007) Nature 447(7143):413 32. Zhao XJ, Zhang SG (2007) Macromol Biosci 7(1):13
9. Alberts B et al (2002) Molecular biology of the cell. Taylor & 33. Holland JH (1995) Hidden orderhow adaptation builds
Francis, Boca Raton complexity. Helix Books, Reading, MA
10. Engler AJ et al (2006) Cell 126(4):677 34. Cui XQ et al (2007) J Phys Chem C 111(5):2025
11. Buehler MJ (2007) Nanotechnology 18:295102 35. Hule RA, Pochan DJ (2007) MRS Bull 32(4):5
12. Buehler MJ (2006) P Natl Acad Sci USA 103(33):12285 36. Winey KI, Vaia RA (2007) MRS Bull 32(4):5
13. Fratzl P et al (2004) J Mater Chem 14(14):2115 37. Petka WA et al (1998) Science 281(5375):389
14. An KN, Sun YL, Luo ZP (2004) Biorheology 41(34):239 38. Smeenk JM, et al (2005) Angew Chem Int Ed 44(13):1968
15. Ramachandran GN, Kartha G (1955) Nature 176:593 39. Zhao XJ, Zhang SG (2006) Chem Soc Rev 35(11):1105
16. Currey JD (2002) Bones: structure and mechanics. Princeton 40. Mershin A et al (2005) Nat Biotechnol 23(11):1379
University Press, Princeton, NJ
123
J Mater Sci (2007) 42:87718787
DOI 10.1007/s10853-007-1719-2
Received: 3 January 2007 / Accepted: 23 March 2007 / Published online: 10 July 2007
Springer Science+Business Media, LLC 2007
Abstract Vimentin coiled-coil alpha-helical dimers are and continuum results help to interpret experimental
elementary protein building blocks of intermediate fila- studies that have provided evidence for the significnifi-
ments, an important component of the cells cytoskeleton cance of vimentin intermediate filaments for the large-
that has been shown to control the large-deformation deformation regime of eukaryotic cells. We conclude that
behavior of eukaryotic cells. Here we use a combination of vimentin dimers are superelastic, highly dissipative protein
atomistic simulation and continuum theory to model tensile assemblies.
and bending deformation of single alpha-helices as well
as coiled-coil double helices of the 2B segment of the
vimentin dimer. We find that vimentin dimers can be
extended to tensile strains up to 100% at forces below
50 pN, until strain hardening sets in with rapidly rising Introduction: the structure of vimentin intermediate
forces, approaching 8 nN at 200% strain. We systemati- filaments and role in eukaryotic cells
cally explore the differences between single alpha-helical
structures and coiled-coil superhelical structures. Based on Together with beta sheets, alpha helical (AH) structures are
atomistic simulation, we discover a transition in deforma- the most abundant secondary structures found in proteins.
tion mechanism under varying pulling rates, resulting in These two patterns are particularly common because they
different strength criteria for the unfolding force. Based on result from hydrogen bonding between the NH and C=O
an extension of Bells theory that describes the dependence groups in the polypeptide backbone. An alpha helix is
of the mechanical unfolding force on the pulling rate, we generated when a single polypeptide chain twists around on
develop a fully atomistically informed continuum model of itself stabilized by hydrogen bonds (H-bond) made
the mechanical properties of vimentin coiled-coils that is between every fourth residue, linking the O of peptide i to
capable of predicting its nanomechanical behavior over a the N of peptide i + 4 in the residue chain. Consequently,
wide range of deformation rates that include experimental at each convolution, three H-bonds are found in parallel
conditions. This model enables us to describe the arrangement that stabilize the helical configuration [1].
mechanics of cyclic stretching experiments, suggesting a Another particularly stable configuration, found for the
hysteresis in the forcestrain response, leading to energy first time in keratin intermediate filaments (IFs) about
dissipation as the protein undergoes repeated tensile 50 years ago, are alpha helical coiled-coils, where the
loading. We find that the dissipated energy increases primary structure reveals a pronounced seven residue
continuously with increasing pulling rate. Our atomistic periodicity (abcdefg)n, called a heptad repeat. Within this
repeat, positions a and d are preferably occupied
with nonpolar residues [2] such as LEU, ALA, VAL or
T. Ackbarow M. J. Buehler (&) ILE. The hydrophobic residuesconsequently concen-
Laboratory for Atomistic and Molecular Mechanics, Department
trated on one sideare the reason for the coiled-coil
of Civil and Environmental Engineering, Massachusetts Institute
of Technology, Cambridge, MA 02139, USA structure. In order to avoid contact with surrounding water
e-mail: mbuehler@MIT.EDU molecules, AHs assemble into coiled-coils by wrapping
123
8772 J Mater Sci (2007) 42:87718787
around each other and clustering the hydrophobic side The five types of IFs and the three assembly groups are
chains inside [1]. Additionally, interhelical and intrahelical listed in Table 2. The most widely distributed of all IF
salt bridges contribute to coiled-coil thermodynamic sta- proteins is vimentin, typically expressed in leukocytes,
bility and are expected to enhance resistance to mechanical blood vessel endothelial cells, some epithelial cells, and
stretch in a nonstereospecific manner [3]. mesenchymal cells such as fibroblasts [1].
Coiled-coils are also the building blocks of vimentin
intermediate filament (IF) dimers, which are composed of a The role of IF networks in the cytoskeleton
head, a tail, and an extremely elongated central rod-do-
main. A schematic of the vimentin dimer structure is IFs, in addition to microtubules (MTs) and microfilaments
shown in Fig. 1a. The rod-like structure is 310 residues (MFs) are one of the three major components of the
long and consists of four coiled-coil alpha helices (1A, 1B, cytoskeleton in eukaryotic cells [11]. The cytoskeleton
2A, 2B) divided by linkers (L1, L12, L2) [2, 4, 5, 7, 8]. plays a critical role in determining the shape and the
Interestingly, all helices in the rod domain have different mechanical properties of the cell, and is vital for numerous
lengths. additional functions such as cell motility or protein
The lengths of each of the components are absolutely synthesis [1113].
conserved for all types of IFs in different types of cells. Like many other biological materials, IFs are hierar-
Additionally, either ends of the rod as well as the position chical structures with highly specific features at nanoscale.
of the stutter in the 2B segment are highly conserved (a Vimentin IF dimers, shown schematically in Fig. 1a, are
definition of various biological terms and concepts is pro- the elementary building blocks of IFs and thus represent
vided in Table 1) [2, 4, 5, 9]. The stutter is an irregularity the first level of hierarchy. Through following the different
in the periodic sequence, where four extra residues are steps of assembly (Fig. 1b) [14, 15], dimers associate to
inserted into the continuous heptad repeat, what results in fibrils, which build the second level of hierarchy. In vivo,
an almost parallel run of both helices without interrupting these fibers can reach a length of up to several lm and
the coiled-coil geometry. For all types of IFs, the stutter is consist of 16 dimers in cross-section. The third level of
spaced precisely six heptads away from the C-terminal end hierarchy are three-dimensional IF-networks inside the
of coil 2B [4]. In contrast to the conserved regions, the cell, reinforcing the plasma membrane [4, 12, 16]. Inside
head and tail domain are greatly diverse for all types of IFs the network, IF-associated proteins such as plectins gen-
[2, 4]. erate the connection between individual filaments.
The cells IF networks are connected with other cells
networks and with the extracellular matrix at the plasma
membrane [1]. This architecture guarantees that tensile and
shear loads applied to the tissue can be carried by IF net-
works.
The fact that IFs span from the cell nucleus to the cell
membrane, and interact with IF networks of other cells
suggest that IFs play an important role in transmitting
mechanical signals from the plasma membrane to the nu-
cleus, where a specific response can be triggered by
mechanical stimulation [10, 17].
Plakin-type cross-bridging proteins, also known as cyt-
olinkers (e.g. plectins or desmoplakins) link all three
Fig. 1 Geometry of vimentin intermediate filament, from atomistic cytoskeletal networks (MTs, MFs and IFs). These proteins
to macromolecular structure. Subplot (a): The dimers, approximately
45 nm long, are the elementary building blocks of vimentin
attach the IFs to MTs, MFs or adhesion complexes [18].
intermediate filaments. A dimer consists of a head, tail (plotted in However, to date little is known about these interaction or
red) and an elongated rod domain which is divided into four alpha- the mechanical functions of plakin-proteins and their
helical coiled coils (1A, 1B, 2A, 2B) connected through linkers L1, ability to store or/and dissipate any elastic energy during
L12, L2 (also red) [2]. The molecular dynamics simulations described
in this paper are performed on alpha-helices, placed in the 2B
internal contraction or external deformation [2, 11].
segment (yellow). Subplot (b): Intermediate filaments assembly In contrast to MTs and MFs, IFs do not participate in the
hierarchically into fibrils. Dimers, the elementary building blocks, dynamic functions of the cytoskeleton. Further, they do not
assembly through a half-staggered, anti-parallel overlap into tetra- support active transport of motor proteins such as myosin
mers, which associate laterally into 13 nm thick unit-length-filaments
(ULFs). In the next step, ULFs assembly longitudinally into more
and kinsesin, due to the missing polarity in the protein
compact filaments (10 nm in diameter) with a length of more than structure (in contrast to MT and MF networks) [2]. They do
240 nm [2, 46] not participate in any cell movement [1]. These evidences
123
J Mater Sci (2007) 42:87718787 8773
Table 1 Summary of different biological terms and concepts used in this paper
Cytoskeleton: A composite inside the cell consisting of three different networks: Actin filaments, microtubules and intermediate
filaments. The intermediate filament network is in the focus of this paper. These networks connect the nucleus
(nuclear membrane) with the plasma membrane and are furthermore responsible for the organization inside the cell.
Intermediate filaments One of the three components of the cytoskeleton; mainly responsible for the large deformation behavior of the cell.
(IFs):
Cross bridging Cross bridging proteins form connections inside each cytoskeletal network as well as connections between different
proteins: networks.
Dimer: A dimer is the elementary building block of an IF fiber. This protein consists of a head domain, a tail domain and an
extremely elongated coiled-coil rod. A coiled-coil is a superhelix that consists of two alpha helices that twist around
each other.
Assembly: Individual IF dimers assemble systematically and hierarchically into filaments (Fig. 1b). Two dimers build a tetramer,
two tetramers build an octamer and four octamers build a unit length filament (ULF). Once this level of assembly is
reached, ULFs ally longitudinally into long fibers.
Residue: The primary structure of a protein consists of a sequence of amino acids. One residue is thus one amino acid in the
polypeptide backbone.
Conserved structure: A structure is conserved when parts of the residue sequence are similar or do not very at all between the different
species (e.g. human and fish). For example, certain parts of the IF sequence are very similar between different species
as well as inside the IF protein family (vimentin, desmin, keratin, etc.). Conserved structures often signify a
particular amino acid sequence that has proven to be particularly suitable for a specific task, and has thus been kept
identical during the evolutionary process.
Table 2 Different types of intermediate filaments, its location in the cell and its assembly group. Intermediate filaments are classified into five
different types. Most of them appear in the cytoskeleton, except lamins that are found in the nucleoskeleton. Keratins, in contrast to the other IF
types, assemble into heterodimers, consisting of one acidit and one basic keratin [6, 10]
# Type Location Assembly group
underline the specific static-mechanical role of IFs. These with equal weight concentrations were sheared and the
static-mechanical properties are the focus of the work de- deformation response (strain) was measured. It is apparent
scribed in this paper. that vimentin gels are capable of sustaining large defor-
mation at large forces. In contrast, actin filaments rupture at
Mechanical properties of IFs compared to other low strains but large forces, and MTs break at moderately
cytoskeletal components large strains, but small forces. This underlines the signifi-
cance of vimentin IFs to carry large forces at large defor-
A great diversity of mechanical properties enables the vi- mation.
mentin IFs to satisfy their specific mechanical role in cells, The stiff behavior of MTs and MFs may be one reason
such as to guarantee their structural integrity or their shape. for much smaller breaking strains of approximately 60%
It has been hypothesized that IFs are critical to provide for MT and 20% for MF in networks of equal weight
strength to the cell under large deformation, and to absorb concentration (see Fig. 2) [7]. In contrast, IFs feature much
large amounts of energy upon a certain load by unfolding higher extensibility. It has been suggested that a higher
[19, 20]. This represents a means to reinforce the cell in flexibility of IFs at small strains (compared with MFs)
extreme deformations so that cells can withstand dramatic results in a lower mechanical resistance during cell
loads and deformations [11, 19, 20]. movement, mainly performed by MFs. It has also been
IFs exhibit a highly nonlinear stressstrain relationship shown that the rigidity of circulating lymphocytes and fi-
with a high resistance against rupture, also known as strain broblasts primary depends on vimentin IFs, whereas MTs
hardening. Figure 2 depicts experimental results [7] of the play a minor role [11, 2123]. The different mechanical
stressstrain response of gels composed of the three cyto- properties of the cytoskeletal networks clearly indicate that
skeletal proteins. In these rheological experiments, gels the cytoskeleton is a composite with a range of mechanical
123
8774 J Mater Sci (2007) 42:87718787
123
J Mater Sci (2007) 42:87718787 8775
characteristics similar to the one observed for myosin. indicating a significant biological function. The goal is to
Additionally, a pulling rate dependence of the coiled-coil develop a structure-function relationship on the protein
structure was found, and it was suggested that the non- level and to link these properties with biological cellular
equilibrated system (due to the very high pulling rates in functions of vimentin.
MD) is the reason for this behavior. The outline of the paper is as follows. In Theoretical
Furthermore, some simulations were carried out on concepts: Modified Bell theory, we present a modification
single alpha helices [32, 33]. By applying tensile loads, to Bells classical model that enables us to describe the
unfolding of the helix was observed after a short steep dependence of protein unfolding forces as a function of
increase in force. Thereby, the 16 residues long helix in pulling rates. In Atomistic modeling methods, we briefly
[32] started to unfold at both ends simultaneously, in introduce our atomistic modeling procedure. In Compu-
contrast to the 20-residue long amino acid chain modeled tational results, we report results of atomistic modeling for
in [33], which unfolded systematically from the side where a variety of boundary conditions. Section Atomistically
force was applied. informed continuum model is dedicated to development of
a continuum model, based on the atomistic simulation
Research strategy and outline results reported in Computational results. In the last Sect.
Discussion and conclusion we conclude with an extensive
Experiments on entire cells have provided strong evidence discussion of our results in light of biological function and
for the importance of mechanical properties of vimentin earlier experimental results.
IFs on the large deformation behavior of cells (see e.g.
Figs. 2 and 3). However, it remains unclear if the origin of
these effects lies at the molecular protein level, at the level Theoretical concepts: Modified Bell theory
of dimers, or if it is a consequence of larger-scale structural
features of the vimentin protein network [4, 24]. Mechanical loading of protein structures can result in
Only few alpha-helical structures have been analyzed in severe changes in the protein structure, inducing unfolding
experiments or in simulations. Further, until now, neither of the protein. Typically, a variety of unfolding processes
any systematic studies on the difference between single exist for a given protein structure, each of which has a
AHs and coiled-coils were carried out, nor the dependence specific reaction pathway and an associated energy barrier.
on the pulling velocity was analyzed in detail. Addition- The different unfolding modes in the protein can be
ally, no explanations about the highly conserved assembly understood as the interplay between different unfolding
of AHs into coiled-coils were suggested that consider processes with different activation barriers operating at
mechanical aspects. A structure-property link for the three different activation distances.
deformation regimes and associated strength models has Several theories exist that describe competing processes
not been reported. Earlier MD simulations were carried out due to mechanically induced instabilities of protein struc-
at extremely large strain rates, and no direct link between tures. These concepts stem primarily from the field of
simulation and experiment has been reported. physical chemistry [3640]. Most of them are derived from
Here we perform a series of atomistic studies of tensile a theory originally postulated by Bell in 1978 [41].
and bending deformation of vimentin to arrive at a detailed In Bells theory, the off rate v is the product of a natural
understanding of the mechanical behavior of these mole- vibration frequency, x0, of the bond in vacuum and the
cules under small and large deformation and at different quasi-equilibrium likelihood of reaching the transition state
pulling velocities. We systematically investigate the pull- with an energy barrier Eb that is reduced by mechanical
ing rate dependence of the mechanical properties of this energy f xb, where f is the applied force along the coor-
particular structure. Development of a theoretical model dinate x, and xb is the distance between the equilibrated
enables us to develop a rigorous understanding of the state (minimum of the well) and the transition state.
pulling rate dependence, which can be used to extrapolate The off rate, also known from chemical reaction kinet-
our results to pulling rates that are comparable to those ics, is given by
applied in experiment. This leads to a direct link between
our simulations and results obtained in experiments. Eb f xb
v x0 exp : 1
Moreover, the continuum model enables us to predict the kb T
behavior of the protein under cyclic loading and under
varying pulling rates. The off rate describes how often a bond is broken per
We focus on the coiled-coil structure in the 2B segment unit time and equals to the reciprocal of the lifetime of a
of the vimentin rod domain, since this part of the vimen- bond. The natural vibration frequency of a bond is
tin protein is highly conserved across various species, x0 1 13 s1 [41].
123
8776 J Mater Sci (2007) 42:87718787
However, Eq. 1 does not describe the dependence of the A strategy to determine the dependence of the unfolding
speed at which a bond breaks due to the applied pulling force f on pulling speed, associated mechanisms and energy
force. Instead, it only provides an estimate of the time scale barriers is to use atomistic modeling, as described in the
at which the bond will be broken. following sections.
In order to overcome this limitation, we modify Eq. 1
based on the following idea: The speed v at which a bond is
broken equals to the distance that needs to be overcome in Atomistic modeling methods
order to break the bond (xb) divided by the time for the
bond breaking. Consequently, v is the product of v xb. Atomistic modeling
This leads to the following equation for the bond breaking
speed: Here we employ atomistic simulation to provide a bridge
between microscopic length- and time scales such as
Eb f xb quantum chemistry, and macroscopic scales such as con-
v x0 xb exp : 2
kb T tinuum mechanics. We use classical molecular dynamics
(MD).
This equation can be rewritten in the following way: Our MD simulations are carried out with the program
NAMD [42] using the CHARMM22 force field [43]. The
f xb
v v0 exp ; 3 CMARMM22 force field is a reasonable model for atom-
kb T istic interactions within proteins and between different
with v0 as the natural bond breaking speed (when no load is proteins (including covalent bonds, H-bonds, electrostatic
applied), defined as: interactions and vdW interactions).
To apply the forces to the molecule that induce defor-
Eb mation, we use steered molecular dynamics (SMD) [42].
v0 x0 xb exp : 4 To apply load, Ca atoms at one end are fixed and the force
kb T
is applied on the Ca atom at the other end. The SMD
This modified framework enables us to study the technique is equivalent to attaching one end of a virtual
dependence between the unfolding force and the bond harmonic spring to the end of the system and pulling at the
breaking speed or to calculate the force at which a bond virtual atom on the other end of the spring [16]. The SMD
breaks, at a certain pulling rate. We can rewrite Eq. 3 in the method thus mimics an experiment where one end of the
following way: molecule is fixed (e.g. on a gold surface), while the other
end is pulled at with the AFM cantilever tip. Using this
kb T kb T technique, different loading conditions (e.g. tensile and
f v ln v ln v0 a ln v b; 5
xb xb bending) can be realized.
The force experienced by the virtual atom is given by
where a kb T=xb and b - kb T=xb ln m0 . Equa- [16]:
tion 5 predicts that the unfolding force depends logarith-
mically on the pulling speed. We note that it contains two F kv t x: 6
parameters a and b, which can be calculated exactly from
the parameters xb and Eb for a certain temperature. Here, x is the displacement of the pulled atom, m the
Equation 5 now provides an immediate link between the pulling velocity, t the time step, and k is the spring con-
pulling rate and the pulling force that is necessary to lover stant.
the energy barrier in such a way that the bond can be For the tensile loading simulations described in this
broken with the applied velocity; increasing the pulling rate paper, the SMD spring constant is k = 10 kcal/mol/A 2.
means increasing the probability of bond rupture. This is Different pulling rates (v) used for the simulations are
due to lowering of the energy barrier at the transition point, indicated in the corresponding sections.
because of the applied force f. We emphasize that typically In the bending simulations, we chose a fixed pulling rate
several unfolding mechanisms exist, each of which is v = 0.000,002 A /fs, with a spring constant k = 0.01 kcal/
characterized uniquely by the pair xb and Eb. mol/A 2.
We note that multi-transition state energy landscapes By monitoring the applied force (F) and the position of
were already predicted in [37]. However, the driving the atom that is pulled (x) over the simulation time, we
parameter in the theory discussed in [37] was the loading obtain force-versus-displacement data that is used to derive
rate (the increase in force over time), and not the pulling the mechanical properties such as bending stiffness or the
rate as done here. Youngs modulus [44].
123
J Mater Sci (2007) 42:87718787 8777
Molecular strain is defined as e x x0 =x0 , where coiled-coil structure before the differences between the
x0 is the initial, undeformed length, and x is the current, single AH and the coiled-coil are described.
deformed length of the protein structure. In the first regime (I), the helical structure is stretched
Due to the time scale limitations of MD to several homogeneously. At the angular point (indicated as AP
nanoseconds, there is typically a huge difference between in Fig. 4, and marked with x in the plot), the structure
simulation and experiment with respect to pulling rates. begins to unfold by rupture of H-bonds, characterized by a
Experimental rates are six to eight magnitudes smaller than significant change in slope in the forcestrain plot, leading
in MD simulations, which requires additional consideration to regime (II). The AP thus represents the critical force
in order to interpret MD results in light of experimental necessary to induce a structural change in the protein,
findings. corresponding to the unfolding force.
Tensile and bending simulations were performed at a We observe that unfolding is initiated at the end where
temperature of 300 K (NVT ensemble), with temperature the load is applied. In the case of the coiled-coil, each of
control using a Berendsen thermostat. The time step used the two alpha helices unfolds individually, while the sec-
in all atomistic simulations discussed in this article is 1 fs. ondary superhelical arrangement remains intact, until
We use Visual Molecular Dynamics (VMD) for visu- strains reach more than 100%. Unfolding of each AH
alization of protein structures [45]. structure is characterized by sequential breaking of H-
bonds. During this regime, the force remains approxi-
mately constant while the entire protein is unfolded at
Initial protein structures
strains approaching 150%.
Once the complete helix is unfolded, the slope increases
We take structures obtained from X-ray diffraction exper-
continuously while the secondary superhelical configura-
iments and stored in the Protein Data Bank (PDB) as the
tion is lost at strains larger than approximately 150%,
starting point for our atomistic simulations.
eventually leading to stretching of covalent bonds in the
The first structure taken from the PDB is a 52 residue
long coiled-coil from the 2B segment (residues 355-406;
PDB ID 1GK6). The second structure is a single alpha
helix from the same coiled-coil. To obtain the geometry of
a single alpha helix, we extract one of the helices from the
1GK6 PDB file. Hydrogen atoms and charges are assigned
according to pH 7.
The coiled-coil part considered in this paper is colored
in yellow, as shown in Fig. 1a. The structure obtained from
the PDB is solved completely in a water skin that
encompasses the entire protein.
We perform energy minimization and finite temperature
equilibration of all structures simulated before the protein
is loaded by applying the SMD technique.
Computational results
Fig. 4 Tensile experiments of vimentin proteins. Subplot (a) depicts
Tensile deformation of single AHs and coiled-coils a schematic of the applied load (left: coiled-coil, right: single AH).
Subplot (b): The forcestrain curves of a single AH-structure and an
alpha-helical coiled-coil, both at a pulling rate of 5 m/s. To enable
Figure 4a shows the loading case for both structures con- better comparison of both curves, the force of the coiled-coil is
sidered. Figure 4b depicts the force versus strain response divided by the number of helices. The first regime (I) consists of a
of a single AH and the coiled-coil structure, both carried steep increase in force until a strain of approximately 25% (referred to
as angular point (AP)) for the coiled-coil and 13% for the single AH.
out at identical loading rates of 5 m/s. We note that the The first regime is followed by the second regime (II) during which
force in Fig. 4b is normalized by the number of AHs in the unfolding of the alpha-helices occurs. The forces at the AP are much
structure (that is, one for the single AH, and two for the higher for a single AH than for the coiled-coil. In the third regime
coiled-coil). (III), a non-linear increase in strain by stretching the backbone is
observed. Thereby the single AH has a much steeper and earlier
For both cases, we observe three distinct regimes (IIII)
increase in force than the coiled-coil structure. The differences
characterizing the stretching dynamics of the protein. between the single AH and the coiled-coil structure are summarized
These three regimes are now characterized first for the in Table 3
123
8778 J Mater Sci (2007) 42:87718787
protein backbone, giving rise to rapidly increasing forces at Table 3 Comparison of the mechanics of single AH with the coiled-
large deformation in regime (III). coil in different dimensions. The data were derived from simulations
with a pulling rate of 5 m/s. It clearly indicates that the single AH has
This unfolding sequence (first unfolding of individual
much more irregularities and instabilities compared to the coiled-coil,
AHs, then unfolding of the superhelical structure) suggests suggesting a higher mechanical stability of the coiled-coil structure
that in the coiled-coil case, the individual AHs represent
Dimension Single AH Coiled-coil
the weakest link in the structure.
Figure 5a and b depict snapshots of the AH coiled-coil Equilibrated H-bond length in A 3.08 0.29 2.97 0.16
and the single AH during the three regimes. Slope of the first regime in pN/A 186.1 94.9
Even though the unfolding curves of single AHs and Force per AH at angular point in pN 930 670
coiled coil structures are qualitatively similar, there are Strain at angular point 13% 20%
several significant quantitative differences. Breaking of H-bonds Simultaneous Sequential
First, the slope of the single AH in the first regime is Beginning strain hardening at 125% 145%
almost two times steeper and leads to much higher forces at Strain hardening Very steep Continuous
the angular point. The angular point appears at strains of
13% in the single AH structure, compared with 25% for the
coiled-coil. Second, unfolding of the single AH begins at
the residue where the pulling force is applied, but is fol-
lowed by immediate additional rupture of H-bonds at res-
idue 369 and 383 inside the protein. Third, strain hardening
of the single AH sets in at 20% lower strains; 125% in the
single AH case versus 145% in the coiled coil. The strain
hardening increases much steeper in the case of a single
AH. The reason for this difference is likely due to the
uncoiling process of the superhelical structure that is
missing in the pure AH structure. The differences between
the two structures are summarized in Table 3.
To investigate the pulling rate dependence of vimentin
mechanics, we have carried out pulling experiments with
Fig. 6 Forcestrain curves of a coiled-coil alpha-helical structure at
systematically varying pulling rates. Figure 6 depicts different pulling rates. We describe the behavior with the developed
forcestrain curves for different pulling velocities, clearly continuum model and extrapolate to pulling rates, used AFM
illustrating a strong pulling velocity (or, equivalently strain experiments. For pulling velocities of smaller than 5 m/s the strains,
where the regime changes take place are very similar to those found in
experiments and are 20% for the angular point and 120% for the
beginning of the strain hardening. The straight lines refer to the
prediction of our continuum model, for various pulling rates
123
J Mater Sci (2007) 42:87718787 8779
123
8780 J Mater Sci (2007) 42:87718787
Fig. 10 Snapshots of the unfolding dynamics. Breaking of H-bonds hardening (regime (III)). Oxygen atoms are in red, nitrogen atoms in
is followed by immediate unfolding of the convolution. However, blue. The H-bonds are represented by the light blue lines. The
after the unfolding, the convolution is not straightened completely, direction of the pulling force is upwards
which does not happen until higher forces are applied during strain
123
J Mater Sci (2007) 42:87718787 8781
deformation is primarily characterized by stretching of an individual H-bond is broken, as it is known that Eb for
covalent bonds. H-bonds is between 4 and 5 kcal/mol.
This mechanisms ceases to operate when v [ vcr , due to
Pulling rate dependence: change in mechanism disappearance of the unfolding wave. The most significant
consequence of the disappearance of the unfolding wave is
We believe there exist at least two different modes of that rather than inducing large local strains, the entire
unfolding of coiled-coil structures. The first mode is molecule is stretched homogeneously. Therefore, the H-
characterized by generation of a localized unfolding wave bonds that stabilize the helical structure are strained
as described in detail in Second regime (II): protein equally under the applied load.
unfolding. We first analyze this deformation mode. For rupture to occur, three H-bonds representing a
As pointed out in Theoretical concepts: Modified Bell complete convolution is broken simultaneously, at a ran-
theory, the accessible mechanisms and associated param- dom location in the coiled-coil. Assuming that a single H-
eters Eb and xb are an immediate consequence of the bonds has a rupture energy of 5 kcal/mol with rupture
atomistic structure and atomic interactions. The results of distance of 1 A , the total rupture energy is Eb = 15 kcal/
MD simulations shown in Fig. 6 are now used to extract mol and xb = 1 A . Estimates for breaking distance and
activation barrier and activation distance associated with energetics of breaking H-bonds are taken from the analysis
this mode of deformation. reported in [46].
Figure 11 plots the results of angular point forces from Figure 11 plots the force at the AP as a function of
simulations with varying pulling rates (data extracted from pulling speed, including the predictions for both deforma-
Fig. 6). In agreement with the theoretical prediction, we tion modes (continuous lines). The extended Bell theory
observe a logarithmic dependence of the force at the AP on combined with the predicted change in unfolding mode
the pulling rate. (rupture of one bond at high pulling rates versus rupture of
By directly fitting the results from atomistic simulation three parallel bonds at slow pulling rates) creates a rea-
to Eqs. 4 and 5, we obtain Eb = 5.6 kcal/mol and sonable link between the very small pulling rates applied in
xb = 0.17 A . These values indicate that at the angular point, experiments (nm/s) and several magnitudes higher veloci-
ties used in MD (m/s), and thus enables to predict the
unfolding force (force at the angular point) as a function of
the pulling speed.
Interestingly, the pulling speed corresponding to van-
ishing unfolding wave speed lies in close proximity to the
intersection of the two modes of deformation (see circle in
Fig. 11). This provides additional, independent prove for
the change in mechanism, and suggests that the second
mechanismhomogeneous rupture of three H-bonds in a
convolutionis correct.
We note that these concepts also make sense from a
biological point of view: Nature forms three H-bonds in
parallel, instead of forming a single, much stronger bond,
as three H-bonds are energetically easier to create. How-
ever, this concept only makes sense if the three H-bonds
are rupturing at the same time, which indeed appears to be
Fig. 11 Theoretical predictions for the unfolding force at the angular the case under smaller strain rates comparable to those
point (AP) compared with simulation and experiment. Simulation
directly proves existence of the first mechanism (local unfolding,
present under physiological conditions.
brown data points), with Eb = 5.6 kcal/mol and xb = 0.17 A (results Not much is known about the specific pulling velocities
obtained directly from fitting to the simulation data). The local that appear under physiological conditions. MT are rear-
unfolding mode is characterized by a deformation wave that is ranged with pulling velocities in the magnitude of lm/min
linearly proportional to the pulling speed (see Fig. 9). The
deformation wave vanishes at a pulling speed of vcr = 0.161 m/s
[47], close to pulling rates used in the experiments reported
(see thicker, dotted line), suggesting that below 0.161 m/s, H-bonds in [6, 29]. However, pulling rates sensed by the IF net-
are strained homogeneously. This gives rise to a different deformation works inside cells due to external load (e.g. during physical
mode characterized by simultaneous breaking of three H-bonds activity such as running) may lie in the regime between
(homogeneous rupture), so that Eb = 15 kcal/mol and xb = 1 A .
The continuous red line indicates the theoretical prediction based on
experiment and simulation.
this theory. Experimental results [6, 29] agree well with the Figure 12 provides an illustration of the competition
theoretical prediction between two mechanisms, each characterized by a pair of
123
8782 J Mater Sci (2007) 42:87718787
f= 500 pN
f= 1000 pN we assume that the single AH has a circular area, the
10
change factor of 4.8 would equal to an increase in radius of
E = E b f xb
about 50%, which is close to the geometrical change in
5 going from single AH to coiled-coil geometry.
With this result, the persistence length can be estimated
0 using the following expression:
-5 EI
-0.2 0.0 0.2 0.4 0.6 0.8 1.0
nP ; 9
kB T
distance x_b from equilibrium in A
where kB is the Boltzmann constant and T the absolute
Fig. 12 Fitting Bells theory with experimental and simulation temperature in Kelvin (K). We estimate a persistence
results, we predict the following energy landscape for coiled-coil
structure for different forces f at angular point. The equation
length of 63 nm for the coiled-coil and of 13 nm for the
indicates, how the energy is calculated. The equilibrium energy and single helix.
distance is set to 0. The first transition state appears 0.17 A away from However, it was reported that the persistence length of
the equilibrium, the second transition state appears at 1 A . We can see
AHs is approximately 1 nm [48] and thus about one
that at forces of 700 pN or lesswhich we predict to appear in
experimentsthe second transition state is dominating. At higher
magnitude smaller. We explain the difference between
force (dotted lines), the first transition state, observed in simulation experiment and simulation by the strain rate dependence,
has higher energies compared to the second transition state which was already found for tensile loads and was also
123
J Mater Sci (2007) 42:87718787 8783
reported for bending deformation of tropocollagen mole- curve of an experiment of cycling loading of a coiled-coil
cules [49]. We address the detailed analysis of this vimentin structure.
dependence in future work. In modeling cyclic loading, we assume that the relaxa-
tion curve for the coiled-coil structure equals to the one at
pulling rates in the order of nm/s (quasistatic pulling rate, in
Atomistically informed continuum model
accordance with experimental observations that show
refolding processes at time scales of seconds [29]). Similar
Based on the simulation data for varying pulling rates and
results were shown in experiments of cyclic loading for the
the analysis based on Bells theory, we generate a continuum
coiled-coil structure of myosin, even though different levels
model, consisting of linear functions for the first and the
of assemblies were considered [29]. Our results show
second regime and a harmonic function for the third regime.
qualitative agreement with these experimental studies.
During the first two regimes, the slope increases linearly
Figure 14b depicts the dissipation energy per loading
with increasing strain rates, while in the third regime the
cycle as a function of the pulling rate. This plot clearly
curvature is proportional to the square of the pulling rate.
indicates, as predicted by [4, 24], that vimentin interme-
As mentioned above, we find that the force at the angular
diate filaments are a stress buffering protein for higher
points depends logarithmically on the pulling rate, fol-
deformation rates.
lowing the predictions by the theoretical model.
c11 and xs Fa 1 ss21 c4 =c3 s2 . in addition to the stretching of the backbone. Some of the
Note that Fa and xa are the force and the displacement at differences between AH structures and coiled-coils are
the angular point, Fs and xs are the force and the dis- summarized in Table 3.
placement at the beginning of the third regime. The We also have observed that the unfolding of the single
parameters ci are summarized in Table 4. AH structure at large pulling velocities is much less con-
Results of this continuum model and comparison with trolled, leading to simultaneous rupture of H-bonds at
MD simulation results are shown in Fig. 6 (the continuous several residues at the angular point (AP), in contrast to the
lines are the predictions by the continuum model). forcestrain curve of the coiled-coil structure, where a
systematic unfolding appears by propagation of an
Example application: cyclic loading experiment of unfolding wave.
vimentin coiled-coils We thus conclude that coiled-coils are mechanically
more stable than single AH structures. This is exemplified
The continuum model given in Eq. 10 enables us to de- by the smooth change from the first to the second regime at
scribe the behavior of this structure even if the strain rate is the AP, the continuous unfolding during the second regime
varying during the stress application or cyclic loading is and the less steep increase in force during the third regime.
applied. Figure 14a shows a prediction of the forcestrain This could also explain why such structures appear in cells
123
8784 J Mater Sci (2007) 42:87718787
Table 4 Parameters of the continuum model derived from the curve fitting to match atomistic results within a hierarchical multi-scale scheme
Numerical parameters and its units
c1 c2 c3 c4 C5 c6 c7 c8 c9 c10 c11 c12 c13
220 3.60 237 19.66 57.71 17.40 173.52 6.3 2.4E7 4.55 2.7 41 1.17
pN pN
pN/A pN
A
A 2
pNfs/A pN/A 3
pNfs2/A 2
pNfs/A
pN/A pN pN
123
J Mater Sci (2007) 42:87718787 8785
behavior of IFs [24]. Glass micro-beam force transducer based on its diameter), we can roughly calculate the average
apparatus were used to measure tensile properties of single area per dimer to be 480 A 2, and use this ratio for comparing
keratin IF from hagfish threads [19, 20]. Youngs Modulus between our model and experiment. Our
In recent experiments [12], filaments were manipulated model predicts a modulus of approximately 25 MPa, which
with the AFM cantilever perpendicular to the filament axis is in good agreement with experiment (10 MPa). Addi-
with forces between 3040 nN and velocities of 0.4 lm/s. tionally, the second regime in experiment was demonstrated
It was found that the breaking extension at the rupture point to begin with the opening of the AH (also known as the alpha
is up to 3.6 times of the initial length and the average to beta transition). This was also observed in the simulation
extension over all experiments was 2.6 (160% strain). at the beginning of the second regime.
Assuming that there are 16 dimers in one filament, the Comparing our results with those from experiments, we
applied force per dimer in this experiment was of about have shown that the specific mechanical properties of vi-
1,8002,500 pN. mentin IFs at least partly originate from the protein level.
By fitting these forces and velocities in our continuum Among others, the particular molecular nanomechanics
model, we estimate protein strains of 150160%. At this explains how the large extensibility known from filament
strain, the coiled-coil structure is in the third regime (the rupture experiments is possible and determined the slip-
regime of strain hardening), but still far away from rup- page between dimers as the weaker link in filaments.
turing. The strength of covalent peptide bonds was reported Our results let us suggest that the first regime does not
to be in the order of 8 nN (and 16 nN for the coiled-coil appear due to the entropic elasticity of the dimer head and
[50]). Therefore, in addition to the protein extension, slip tail, as hypothesized in [20], but due to the stretching of the
between proteins may occur, which may lead to additional H-bonds. As long as H-bonds are not brokenwhich
strains of up to 50%. We derived this number by taking the happens at the angular pointdeformation is completely
length of a tetramer (approximately 60 nm; two dimer, half elastic. This type of elasticity was observed in experiment
staggered overlap) and by assuming that after slip has taken and additionally supports our hypothesis.
place, the dimers are arranged sequentially, leading to a
maximal length of 90 nm (two times 45 nm). Relating the Linking results to other coiled-coil structures
difference in length (30 nm) to the initial length (60 nm)
we receive a strain due to slippage of 50%. Consequently, The forcestrain curves observed in our simulations are
the breaking strain being the sum of stretching and slip, can similar to MD simulations carried out on different AHs and
be estimated to be on the order of 200%. This is in rather coiled-coils [3033], as well as to previous experimental
good agreement with experiment (160% average, and studies on other alpha-helical coiled-coil structures such as
maximum of up to 260%). myosin [29] or desmin intermediate filaments [6], even
We conclude that yield in vimentin filaments appears though the force levels are much larger due to the very high
due to the slip between proteins, as it is the weakest part in pulling velocities.
the assembled system. This idea further explains the highly Due to the similarity in the curve shape, the similar
reduced diameter of the ruptured filaments being a result of strain levels in experiment and simulation, as well as the
the unfolded proteins. This suggests that both mechanisms similar force levels in simulations for different coiled-coil
(extension and slippage) appear during failure of IFs. Up structures, we conclude that the secondary structure, rein-
until now this was only suggested [20, 24]. Future AFM forced by H-bonds is mainly responsible for the mechani-
studies could be used to determine the forces at which slip cal behavior during tensile deformation, as long as point
appears, which will help to understand the mechanical mutations do not destroy the structure. It was already
properties of vimentin on the filament level. suggested earlier that the mechanisms underlying the
Tensile tests were recently performed on wet keratin IFs mechanics of proteins are very simple, even if the number
in hagfish threads [19, 20]. The curves observed in these of possible amino-acid configurations is extremely high
experiments contained four regimes, whereof the first three [51]. Our observations corroborate this notion.
regimes appear very similar to the regimes that we observed
in our simulations. Not only the shape of the curves, but also Interpretation of results in light of biological function
the strains where the changes between different regimes
take place fit rather closely. For example, the angular point The mechanical properties of metazoan cells, its rigidity at
appeared in experiment at a strain of 34% (our simulation high stresses and its integrity are mainly caused by vi-
25%), and the beginning of the third regime was in exper- mentin IFs. However, up to now, it was not known exactly,
iment at 100% of strain (our simulation 120%). from which level of hierarchy the specific mechanical
Additionally, if we assume that each thread (diameter of properties arise. In this work we have shown which
10 nm) has in average 16 dimers (a reasonable assumption mechanical properties appear on the protein level and
123
8786 J Mater Sci (2007) 42:87718787
proved the suggestion that coiled-coils mainly contribute to neously. This clearly supports the idea of creating parallel
the mechanical behavior, discovered on the filament level bonds. The coiled-coil configuration reinforces the protein
as well as on the IF network level. additionally and makes it much more stable.
We hypothesize that Nature utilizes coiled-coils as a
simple construct in order to realize superelasticity in bio- Outlook
logical materials. The deformation behavior of coiled-coils
is not only elastic but also contains a long plateau at con- Future studies could be focused to develop a more com-
stant force (Figs. 4 and 6). Thereby, the very high and plete understanding of vimentin dimers, the role of the
reversible deformation of coiled-coils is realized by a head and tail domain, the reason for the different lengths of
second, stress induced phase, which equals to the rupture of the conserved coiled-coil segments in the rod domain, and
the H-bonds followed by an immediate unfolding. Once the the function of the stutter. In addition, a thorough theo-
tensile load is reduced, the second phase becomes unstable, retical understanding of the driving forces during assembly
or in other word, the structure folds back to the more stable and the interaction of assembled IFs is still missing.
helical shape by reforming the H-bonds. This behavior is Furthermore, the time and length scale, which can be
reminiscent of a shape memory alloy. realized with MD is still comparably small. In order to
There is another twist to this feature: The second phase overcome the limitations of length and time scale, multi-
may be used simultaneously as the security belt of the scale modeling techniques could be used in future work,
cell, as covalent bonds rupture at forces of three magni- linking the atomistic scale through mesoscale, such as
tudes higher than H-bonds. Additionally, the transition coarse graining, to the continuum scale. By using such
from the plateau regime to the stretching of the covalent techniques, it may be possible to shed light on the different
bonds is very smooth so that no shock waves appear that levels of hierarchies and thus contribute to the under-
could lead to uncontrolled rupture. Therefore, coiled-coils standing of the protein mechanics of vimentin alpha-helical
may be considered as the elementary building blocks of IFs coiled-coils. Techniques such as nudged elastic band
that enable cells to withstand dramatic loads, large defor- (NEB) may be used to determine the energetic pathway of
mation and very high deformation rates. rupturing three H-bonds simultaneously.
In coiled-coil structures, H-bonds are apparently not The combination of new theories, experimental tech-
used to generate mechanically robust structures, as can be niques on the nano-scale in addition to modern simulation
seen by the existence of regime (II) with low unfolding approaches, might be the key in understanding hierarchical
forces. Instead, they provide a means of enabling low-force biological materials and thus help to heal diseases or help
unfolding and simple refolding into helical structures. designing new multifunctional materials such as biological
Strength originates from the superhelical structure and actuators.
from the covalent chemistry present in the protein back- We hope that our theoretical studies could motivate new
bone. experiments on single alpha helical as well as coiled-coil
The hysteresis (Fig. 14) and the strong strain rate proteins. In particular, experiments are needed that produce
dependence (Fig. 6) provide further evidence that coiled- forceextension curves for systematically varying pulling
coils represent strain buffering elements in cells, with the rates. Such results can be used directly to elucidate energy
possibility to dissipate great amounts of energy as it barriers of different mechanisms, and may be used to
undergoes repeated stretching and relaxation cycles. At the compare quantitatively with the theoretical concepts and
same time, at small deformations (smaller than approxi- results reported here.
mately 20%) and small pulling velocities that appear dur-
Acknowledgements TA acknowledges the support of the German
ing cell movement, the resistance of the coiled-coil is National Academic Foundation and the Dr.-Juergen-Ulderup Foun-
completely elastic, and thus do not dissipate energy. dation. This research was partly supported by the Army Research
The comparison of our results with those from simula- Office (ARO), grant number W911NF-06-1-0291, program officer Dr.
tions and experiments on other coiled-coils, underline the Bruce LaMattina. We acknowledge fruitful discussions with Professor
Harald Herrmann (University of Heidelberg, Germany) and Professor
important role of H-bonds as a key in the realization of the Laurent Kreplak (University of Basel, Switzerland).
specific mechanical properties of coiled-coils. Our results
suggest that Nature creates three H-bonds in parallel (three
for single AHs, and six for coiled-coils, where three are References
present in each AH). This is energetically much easier to
realize than single bonds with strength equal to three H- 1. Alberts B et al (2002) In: Molecular biology of the cell. Taylor &
Francis
bonds. Our calculations predict that during pulling veloc-
2. Strelkov SV, Herrmann H, Aebi U (2003) BioEssays 25:243
ities applied in experiments, and most likely during those 3. Burkhard P, Kammerer RA, Steinmetz MO, Bourenkov GP, Aebi
appearing in vivo, three parallel H-bonds break simulta- U (2000) Structure 8:223
123
J Mater Sci (2007) 42:87718787 8787
4. Herrmann H, Aebi U (2004) Annu Rev Biochem 73:749 27. Omary MB, Coulombe PA, Irwin McLean WH (2004) N Engl J
5. Strelkov SV, Herrmann H, Geisler N, Wedig T, Zimbelmann R, Med 351:2087
Aebi U, Burkhard P (2002) EMBO J 6:1255 28. Schietke R, Broehl D, Wedig T, Muecke N, Herrmann H, Magin
6. Kiss B, Karsai A, Kellermayer MSZ (2006) J Struct Biol 155:327 TM (2006) Eur J Cell Biol 85:1
7. Janmey PA, Euteneuer U, Traub P, Schliwa M (1991) J Cell Biol 29. Schwaiger I, Sattler C, Hostetter D, Rief M (2002) Nat Mater
113:155 1:232
8. Mucke N, Wedig T, Burer A, Marekov L, Steinert P, Langowski 30. Akkermans RL, Warren CPB (2004) Phil Trans R Soc Lond
J, Aebi U, Herrmann H (2004) J Mol Biol 340:97 362:1783
9. Smith TA, Strelkov SV, Burkhard P, Aebi U, Parry DAD (2002) J 31. Root DD, Yadavalli VK, Forbes JF, Wang K (2006) Biophys J
Struct Biol 137:128 90:2852
10. Wilson KL, Zastrow MS, Lee KK (2001) Cell 104:647 32. Cieplak M, Hoang TX, Robbins MO (2002) Proteins: Struct
11. Wang N, Stamenovic D (2003) J Muscle Res Cell Motil 23:535 Funct Genet 49:104
12. Mucke N, Kreplak L, Kirmse R, Wedig T, Herrmann H, Aebi U, 33. Rohs R, Etchebest C, Lavery R (1999) Biophys J 76:2760
Langowski J (2004) J Mol Biol 355:2342 34. Bornschloegl T, Rief M (2006) PRL 96:118102
13. Helfand BT, Chang L, Goldman RD (2004) J Cell Sci 117:133 35. Mitsui J, Nakajima K, Arakawa H, Hara M, Ikai A (2000) Bio-
14. Smith TA, Hempstead PD, Palliser CC, Parry DAD (2003) Pro- chem Biophys Res Commun 272:55
teins 50:207 36. Hanke F, Kreuzer HJ (2006) Phys Rev 74:031909
15. Kreplak L, Aebi U, Herrmann H (2004) Exp Cell Res 301:77 37. Evan E, Ritchie K (1997) Biophys J 72:1541
16. Strelkov SV, Schumacher J, Burkhard P, Aebi U, Herrmann H 38. Dudko OK, Hummer G, Szabo A (2006) PRL 96:108101
(2004) J Mol Biol 343:1067 39. Wiita AP, Ainavarapu SRK, Huang HH, Fernandez JM (2006)
17. Moir RD, Spann TP (2001). Cell Mol Life Sci 58:1748 PNAS 103:7222
18. Coulombe PA, Bousquet O, Ma L, Yamada S, Wirtz D (2000) 40. Gilli P, Bertolasi V, Pretto L, Gilli G (2006) J Mol Struct 790:40
Cell Biol 10:420 41. Bell GI (1978) Science 200:618
19. Fudge DS, Gosline JM (2004) Proc R Soc Lond 271:291 42. Lu H, Isralewitz B, Krammer A, Vogel V, Schulten K (1998)
20. Fudge DS, Gardner KH, Forsyth VT, Riekel C Gosline JM (2003) Biophys J 75:662
Biophys J 85:2015 43. MacKerell AD et al (1998) J Phys Chem 102:3586
21. Eckes B, Dogic D, Colucci-Guyon E, Wang N., Maniotis A, 44. Courtney TH (1990) In: Mechanical behaviour of materials.
Ingber D (1998) J Cell Sci 111:1897 McGraw-Hill
22. Brown MJ, Hallam JA, Colucci-Guyon E, Shaw S (2001) J 45. Humphrey W, Dalke A, Schulten K (1996) J Mol Graphs 14:33
Immunol 166:6640 46. Sheu SY, Yang DY, Selzle HL, Schlag EW (2003) PNAS
23. Nieminen M, Henttinen T, Merinen M, Marttila- Ichihara F, 100:12683
Eriksson JE, Jalkanen S (2006) Nat Cell Biol 8:156 47. Inoue S, Salmon ED (1995) Mol Biol Cell 6(12):1619
24. Kreplak L, Bar H, Leterrier JF, Herrmann H, Aebi U (2005) J 48. Papadopoulos P et al (2006) Biomacromolecules 7:618
Mol Biol 354:569 49. Buehler MJ, Wand SY (in press) Entropic elasticity controls
25. Storm C., Pastore JJ, MacKintosh FC, Lubensky TC, Janmey PA nanomechanics of single tropocollagen molecules
(2005) Nature 435:191 50. Buehler MJ (2007) J Mech Mater Struct (in print)
26. Guzman C, Jeney S, Kreplak L, Kasas S, Kulik AJ, Aebi U, Forro 51. Baker D (2000) Nature 405:39
L (2006) J Mol Biol 360:623
123
J Mater Sci (2007) 42:87888794
DOI 10.1007/s10853-007-1918-x
Received: 21 December 2006 / Accepted: 5 June 2007 / Published online: 10 July 2007
Springer Science+Business Media, LLC 2007
Abstract Absence of fibrillin 2 (Fbn2), a non-collage- osteoclasts and bone formation by osteoblasts, however,
nous bone protein, causes a connective tissue disorder when the activity levels of osteoclasts exceeds osteoblasts
called congenital contractural arachnodactyly (CCA) and uncoupling takes place causing a significant loss in bone
has been associated with decreased bone mineral density. mass and increased fragility. The reduced bone mineral
Nanoindentation and Raman microspectroscopy have been density (BMD) is the main cause of the most common bone
used to compare the mechanical and chemical properties of disease, namely osteopenia/osteoporosis, which is charac-
cortical bone from femora of Fbn2/ deficient mice and terized by the microarchitectural deterioration of bone
their wild-type controls (Fbn2+/+). It was found that tissue and an increase in fracture risk [1]. Both osteoblastic
Fbn2/ bones have significantly lower hardness and and osteoclastic activity are closely linked to the presence
elastic modulus compared to Fbn2+/+ bones, especially in of non-collagenous bone matrix proteins. Although the
the mid-cortical section. The Raman analysis showed little organic phase of bone is 90% type I collagen, there are still
difference with genotype except for a decrease in type-B considerable amounts of the non-collagenous proteins
carbonate substitution in the endosteal region of Fbn2/ present including fibrillin 1 and 2. Investigators have pro-
bones. The results indicate that Fbn2 plays a direct role in posed that the overall mix of non-collagenous proteins
determining the mechanical properties of bone. makes a significant contribution to the characteristic
properties of bone, for instance its ability to mineralize
compared to other tissues [2, 3]. It has also been reported
that the non-collageous proteins have a role in bone turn-
over regulation [4] and contribute to the structural integrity
Introduction of bone.
Fibrillin 1 and 2 are extracellular glycoproteins that
Bone is a composite tissue comprised of a mineral phase, constitute the major structural components of connective
an organic matrix and water. Bone mass in adults tissue microfibrils. Their absence in bones can have dramatic
is maintained by coupling between bone resorption by effects on the bones properties. The connective tissue
disorders known as Marfans syndrome and congenital
N. B. Kavukcuoglu A. B. Mann contractural arachnodactyly are caused by mutations in
Materials Science & Engineering, Rutgers University, fibrillin 1 (Fbn1) and fibrillin 2 (Fbn2), respectively [5, 6].
Piscataway, NJ 08854, USA In addition to other manifestations, these two heritable
disorders of the connective tissues have been associated with
E. Arteaga-Solis S. Lee-Arteaga F. Ramirez
Child Health Institute, UMDNJ-RWJMS, New Brunswick, a reduction in bone mineral density [710]. The findings that
NJ 08903, USA fibrillin-rich microfibrils are widely distributed in the
developing mouse and human skeleton has been interpreted
A. B. Mann (&)
to suggest key roles of these macromolecular aggregates in
Biomedical Engineering, Rutgers University, Piscataway,
NJ 08854, USA bone formation, growth and mineralization [1113].
e-mail: abmann@rci.rutgers.edu Furthermore, the fact that fibrillins are highly expressed by
123
J Mater Sci (2007) 42:87888794 8789
differentiating osteoblasts may also indicate a function in Rev. 1985). The left femur from the mice of each genotype
matrix adherence through interaction with the avb3 integrin was used for this study. The bones were prepared for
[11, 14, 15]. Finally, fibrillin-rich microfibrils have been nanoindentation testing as described by Roy et al [25].
associated recently with modulating the activity of TGF-b, Briefly, this involved dehydration in graded alcohol solu-
one of the major growth factors involved in bone physiology tions (70100%) and mounting in a low temperature cure
[16]. Consistent with these earlier lines of indirect evidence, epoxy (SPI supplies, West Chester, PA). Dehydration can
unpublished data indicates that both fibrillin-1 and fibrillin-2 affect the mechanical properties of bone, however all the
deficient mice are osteopenic due to distinct alterations in the bones where prepared in the same way so these effects will
balance of bone remodeling [17; Arteaga-Soils et al. be present in bones from both genotypes. There is a small
(unpublished)]. possibility that one genotype is affected more by dehy-
In this study Raman spectroscopy has been used to dration than the other, but this was not investigated. After
investigate how the absence of Fbn2 affects the chemistry mounting, the included femora were sectioned transversely
of mouse bones, while nanoindentation has been used to at mid-shaft using a diamond wafering saw. The surfaces
examine the bones mechanical properties. Raman Spec- were ground with silicon carbide paper of decreasing grit
troscopy is a non-destructive way of analyzing the size (400, 600 and then 1,200 particles per inch) followed
molecular structure of the mineral components of bone at by polishing with diamond paste down to 1/2 then 1/4 lm
the microscopic level. Raman can characterize both the grit size. After the polishing, all specimens were cleaned
organic and inorganic component of bone by examining the ultrasonically to remove surface debris. Raman studies do
vibrational spectra of the chemical bonds. It provides not require any specific kind of sample preparation [19], so
quantitative information on the changes in the mineral and the same bones used for nanoindentation testing were used.
matrix compositon as well as the nature and amounts of It should be noted that embedding samples into epoxy has
substituents in the mineral. These advantages of Raman been shown to have no effect on the Raman data [26]. The
Spectroscopy have led to it gaining in importance as a locations for the nanoindentation and Raman spectroscopy
method to study bone [1820] and, in particular, to measurements are shown in Fig. 1.
examine compositional changes due to aging and diseases,
and the link between bone chemistry and mechanical Raman microspectroscopy
deformation.
Nanoindentation has become the standard tool for A Renishaw inViaTM Raman microscope was used for this
characterizing the mechanics of materials at small length study with a 785 nm laser beam and a grating of 1,200 1/mm.
scales because of its ability to probe a surface and map its The laser was focused on each point of interest through a
properties with a resolution smaller than 1 lm. Due to Leica DMLM, 50X/0.75 NA objective providing an
small dimensions of mouse bones, this technique provides approximately 2 lm spot size. All spectral acquisitions were
an effective measurement method for their mechanical performed in the 3502,000 cm1 wavenumber range and
properties [2123]. In the present study the average prop- the spots were spaced every 10 lm in the x-(transverse)
erties (mechanical and chemical) and the location depen- direction (see Fig. 1). Depending on the thickness of the
dent (intra-bone) variations in the properties were cortex 1520 measurements were taken. Test locations were
investigated and compared between cortical femora bones marked by microindents prior to testing. Signal-to-noise
from a number of knockout (Fbn2/) and wild-type ratio and fluorescence background were minimized by
(Fbn2+/+) mice. employing 18 s of exposure time and 3 times of accumula-
tion (three separate spectra) at each measurement location.
The Wire 2 software provided by Renishaw was used to
Materials and methods remove the background using cubic spline interpolation at
each acquisition. A single spectrum for each sample point
Sample preparation was obtained by averaging the three accumulations. The
software was then used to analyze the peaks after Gaussian
The study used 3-month old female 129/SvEv Fbn2/ Lorentzian curve fitting. The intensities of the following
mouse models (n = 5) and their age, sex and background peaks were measured: phosphate m1 at ~958 cm1 (PO3 4
matched wild-type controls (n = 6). Creation of Fbn2/ symmetric stretching band), carbonate m1 at ~1,071 cm1
mice has been already described [24]. The research was (CO23 symmetric stretching band), CH2 wag at ~1,450 cm
1
approved by the Rutgers and UMDNJ-RWJMS Animal (CH bending band), amide ~1,667 cm1 (C=O stretching
Care and Facilities Committees (Protocol # I 06012-2) band), amide III at ~1243 cm1 (in-phase combination of
and performed according to the NIH guidelines for the care the NH bending and CN stretching) and the width (full
and use of laboratory animals (NIH Publication #85-23 width at half maximum, FWHM) of phosphate m1 at around
123
8790 J Mater Sci (2007) 42:87888794
Nanoindentation
ion [18, 2830]. In the present study the type-B carbonate Statistical analysis
substitution was quantified by the ratio of the peak inten-
sities for CO2 1 3 1
3 (~1,071 cm )/PO4 m1 (~958 cm ). Type- For each sample the Raman and nanoindentation data
A substitution, however, could not be quantified because were average so a single value for each property, such as
the wavenumber for the vibrational signal for this substi- hardness, was obtained. These values were then averaged
tution (~1,108 cm1) was very weak and could not be for all of the samples of each genotype to give the mean
123
J Mater Sci (2007) 42:87888794 8791
value of the property for the genotype. ANOVA was then Results
used to examine the statistical significance of differences
between the genotypes. Nanoindentation
The Raman spectra and the nanoindentation across the
transverse axis of each knock-out and wild-type bone were First, the mean values of H and E across the entire section of
examined to give the intra-bone variations for each geno- cortical bone, plus the standard deviations and the standard
type. Mechanical and chemical properties were calculated errors for both genotypes (Fbn2/ and Fbn2+/+) were
at each measurement location for each individual sample. calculated (see Table 1). ANOVA revealed that bones from
The nanoindentation data collected across the cortical Fbn2/ mice exhibited significantly lower hardness and
thicknesses of each bone were then divided into seven elastic modulus values compared to bones from Fbn2+/+
equal sections. Sections I and II are the bone adjacent to mice (p < 0.05). The average H and E of Fbn2/ mouse
periosteal (outer) surface. Sections IIIV are the mid- bones were 1.15 0.01 GPa and 29.3 0.3 GPa, respec-
cortical and sections VIVII are the bone adjacent to the tively. For Fbn2+/+, H was 1.23 0.01 GPa and E was
endosteal (inner) surface (Fig. 1). For the Raman spectral 31.8 0.3 GPa.
measurements because of the larger spot size of the micro- The intra-bone variations were plotted as bar charts to
Raman compared to the nanoindent size and the lower show the variation of H and E with genotype across the
number of measurements (1520) across the cortical transverse axis of the cortical bone (Fig. 3). Individual
thickness it was divided into three equal sections specimens from the same genotype exhibited similar vari-
(Iperiosteal, IImid-cortical and IIIendosteal). The ations with respect to location. The results show that
mechanical and chemical properties were averaged over Fbn2/ mice bones have lower H and E at locations IIVII
each section for each specimen. The mean values of the compared to the Fbn2+/+ mice bones. This difference
properties for each section were then averaged over all the between the genotypes was statistically significant
bones of each genotype. ANOVA was conducted to assess (p < 0.05) in the mid-cortical region, sections IIIV. The H
the significance of the differences between the knock-out and E values of bones from Fbn2/ mice both decreased
and wild-type mouse bones in each transverse section. All between the periosteal (sections I and II) and mid-cortical
tests used 95% minimum level of confidence and p 0.05 (sections IIIV), then increased slightly towards the end-
was considered statistically significant. osteal (sections VI and VII) part of the bones cross-section.
Table 1 Average bone parameters and standard errors derived from nanoindentation and Raman spectra of bones from Fbn2/ mice and
Fbn+/+ mice
Genotype Hardness, H Elastic modulus, E Mineral: organic Carbonate substitution Crystallinity
(GPa) (GPa) PO3
4 m1/amide I CO2 3
3 /PO4 m1 1/PO3
4 m1 (FWHM)
Fbn2/ 1.15 0.01 29.3 0.3 15.02 0.2 0.110 0.0005 0.0530 0.0002
Fbn2+/+ 1.23 0.01 31.8 0.3 14.63 0.3 0.115 0.0023 0.0529 0.0002
123
8792 J Mater Sci (2007) 42:87888794
Raman analysis
123
J Mater Sci (2007) 42:87888794 8793
endosteal region (III) (Fig. 4a). Type-B carbonate substi- the type-B carbonate substitution of Fbn2/ was signifi-
tution was significantly lower in Fbn2/ mouse bones at cantly lower than that of Fbn2+/+ bones.
the near endosteal region compared to Fbn2+/+ mouse Overall the observed effects of Fbn2 deficiency on the
bones (p < 0.05), as seen in Fig. 4b. The crystallinity mechanical properties of mouse bones indicate that the
exhibited the greatest difference in the mid-cortical where it protein plays a direct role in determining the mechanics of
was higher for Fbn2/ mouse bones compared to Fbn2+/+ the bones. The Raman data shows that the difference with
mouse bones (see region II on Fig. 4c), though this was of genotype in the apatite phase of the bone matrix is not of
low statistical significance. statistical significance except for a decrease in carbonate
Overall the Raman data showed only minor differences substitution in the absence of Fbn2 in the endosteal region.
between the apatite phase of the bones from the two All the other differences in the Raman data were of low
genotypes. The only difference that was of statistical sig- statistical significance. Based on this it seems unlikely that
nificance was for carbonate substitution in the endosteal the differences in mechanical properties due to the absence
region. of Fbn2 were due to changes in the apatite phase.
Discussion Conclusions
This study has investigated the role of fibrillin 2 (Fbn2), on This study has demonstrated that at the nano-scale Fbn2, a
the mechanical and chemical properties of cortical bone non-collagenous bone protein, is critical in determining the
from mouse femora. Specifically, the average and location mechanical properties and, to a much lesser extent, the
dependent differences between Fbn2/ and Fbn2+/+ chemical properties of bones. In bones from Fbn2+/+
mice bones have been investigated. mice, hardness and elastic modulus were higher, especially
Fbn2 deficiency caused a statistically significant decline in the mid-cortical region compared to bones from Fbn2/
in the average mechanical properties as measured by nano- mice. Typically the mechanical properties of bone would
indentation of mouse bones from sex and age matched be expected to increase with increasing crystallinity, but
knock-out and wild-type genotypes. Further to this, the this is not seen in these results as crystallinity appeared to
location dependent analyses revealed that the difference is be higher in the mid-cortical region of the Fbn2/ bones,
predominantly in the mid-cortical sections where the bones though this was of low statistical significance. This sug-
from Fbn2/ mice exhibited significantly lower H and E gests that the fibrillin (microfibrils) in the wildtype mouse
compared to the Fbn2+/+ mouse bones. The endosteal and bones may be contributing directly to its mechanical
periosteal showed no statistically significant difference in properties. The decreased mechanical properties did to a
mechanical properties with genotype. The sections of bone limited extent correspond to a decrease in carbonate sub-
showing the most significant changes in mechanical prop- stitution in the apatite phase. However, the most significant
erties with genotype correspond to the sections showing decrease in carbonate substitution was observed only in the
differences in crystallinity, though the crystallinity differ- endosteal region where the mechanical properties did not
ences were of low statistical significance (possibly due to the show a statistically significant difference with genotype.
size of the sample set). The effects of genotype on In summary, the mechanical properties of mouse corti-
mechanical properties were mostly confined to the mid- cal bone show a clear dependence on the presence of fi-
cortical section which comprises the mature bone, while the brillin 2. This is most noticeable in the mature mid-cortical
differences in mechanical properties with genotype in the bone and it does not appear to be due to changes in the
periosteal and endosteal regions where much smaller. This bones mineral (apatite) phase. Thus, it is concluded that
suggests that mature lamellar bone has properties which fibrillin 2 and, consequently, microfibrils play a direct role
show a different dependence on Fbn2 than immature woven in determining the mechanical properties of bone.
bone. Specifically, it appears that the absence of Fbn2
reduces the mechanical properties in the fully mineralized Acknowledgments The authors would like to thank Prof. David
Denhardt of Rutgers University, NJ and Prof. Nejat Guzelsu of
matrix. This finding is consistent with a previous suggestion
UMDNJ, NJ for helpful discussions. Financial support for this
based solely on a gene expression data of a structural role for research has been provided by NSF, ACS PRF, DOD DURIP,
microfibrils in the mature bone [11, 13, 33]. It also agrees NJCHE, NIH and the Rutgers University Busch Bequest.
with unpublished data indicating an increase in bone fragility
in Fbn2/ mice (Arteaga-Soils et al. (unpublished)). References
Another interesting finding of this study was the
decreased type-B carbonate substitution in Fbn/ mouse 1. Ammann P (2003) Therapie 58:403
bones. This was mostly seen in the endosteal region where 2. Gundberg CM (2003) Osteoporosis Int 14:S37
123
8794 J Mater Sci (2007) 42:87888794
3. Roach HI (1994) Cell Biol Int 18:617 on the musculoskeletal aspects of marfan syndrome: meeting
4. Kaartinen MT, Pirhonen A, LinnalaKankkunen A, Maenpaa PH report and state of the science, p n/a
(1997) J Biol Chem 272:22736 18. Akkus O, Adar F, Schaffler MB (2004) Bone 34:443
5. Putnam EA, Zhang H, Ramirez F, Milewicz DM (1995) Nat 19. Morris MD, Finney WF (2004) Spectrosc Int J 18:155
Genet 11:456 20. McCreadie BR, Morris MD, Chen T-C, SudhakerRao D, Finney
6. Quondamatteo F, Reinhardt DP, Charbonneau NL, Pophal G, WF, Widjaja E, Goldstein SA (2006) Bone 39:1190
Sakai LY, Herken R (2002) Matrix Biol 21:637 21. Akhter MP, Fan Z, Rho JY (2004) Calcified Tissue Int 75:416
7. Carter N, Duncan E, Wordsworth P (2000) Rheumatology 39:307 22. Chen Q, Rho JY, Fan Z, Laulederkind SJF, Raghow R (2003)
8. Dallas SL, Keene DR, Bruder SP, Saharinen J, Sakai LY, Mundy Calcified Tissue Int 73:387
GR, Bonewald LF (2000) J Bone Mineral Res 15:68 23. Silva MJ, Brodt MD, Fan Z, Rho J-Y (2004) J Biomech 37:1639
9. Giampietro PF, Peterson M, Schneider R, Davis JG, Raggio C, 24. Arteaga-Solis E, Gayraud B, Lee SY, Shum L, Sakai L, Ramirez
Myers E, Burke SW, Boachie-Adjei O, Mueller CM (2003) F (2001) J Cell Biol 154:275
Osteoporosis Int 14:559 25. Roy ME, Rho JY, Tsui TY, Evans ND, Pharr GM (1999)
10. Kohlmeier L, Gasner C, Bachrach LK, Marcus R (1995) J Bone J Biomed Mater Res 44:191
Mineral Res 10:1550 26. Carden A, Morris MD (2000) J Biomed Optics 5:259
11. Zhang H, Hu W, Ramirez F (1995) J Cell Biol 129:1165 27. Boskey A, Mendelsohn R (2005) J Biomed Optics 10(3):031102
12. Plantin P, Durigon M, Boileau C, Le Parc JM (2000) Annales De 28. Freeman JJ, Wopenka B, Silva MJ, Pasteris JD (2001) Calcified
Pathologie 20:115 Tissue Int 68:156
13. Keene DR, Sakai LY, Burgeson RE (1991) J Histochem Cyto- 29. Penel G, Leroy G, Rey C, Bres E (1998) Calcified Tissue Int
chem 39:59 63:475
14. Ramirez F, Pereira L (1999) Int J Biochem Cell Biol 31:255 30. Yerramshetty JS, Lind C, Akkus O (2006) Bone 39:1236
15. Sakamoto H, Broekelmann T, Cheresh DA, Ramirez F, Rosen- 31. Akkus O, Polyakova-Akkus A, Adar F, Schaffler MB (2003)
bloom J, Mecham RP (1996) J Biol Chem 271:4916 J Bone Mineral Res 18:1012
16. Neptune ER, Frischmeyer PA, Arking DE, Myers L, Bunton TE, 32. Oliver WC, Pharr GM (1992) J Mater Res 7:1564
Gayraud B, Ramirez F, Sakai LY, Dietz HC (2003) Nature Genet 33. Kitahama S, Gibson MA, Hatzinikolas G, Hay S, Kuliwaba JL,
33:407 Evdokiou A, Atkins GJ, Findlay DM (2000) Bone 27:61
17. Jones KB, Sponseller PD, Erkula G, Sakai L, Ramirez F, Dietz
HC, Kost-Byerly S, Bridwell KH, Sandell L (2006) Symposium
123
J Mater Sci (2007) 42:87958803
DOI 10.1007/s10853-007-1914-1
Received: 17 January 2007 / Accepted: 4 June 2007 / Published online: 10 July 2007
Springer Science+Business Media, LLC 2007
123
8796 J Mater Sci (2007) 42:87958803
mechanical response of collagen when mineral has min- compared to wet bone. These indicate that water has sig-
eralized at specific locations is not known. This is very nificant influence on overall mechanical response of bone.
difficult to obtain from experimental techniques due to However, the role of water on the nanostructure of bone,
small size of mineralized hydroxyapatite crystals where the collagen is interacting with hydroxyapatite
(50 nm 25 nm 3 nm) and collagen molecule (300 nm) (HAP) nanocrystals, is not well understood. Here, we
[2831]. In this present study we have obtained the attempt to understand the mechanical response of wet
mechanical response of collagen molecule in proximity of collagen (solvated collagen) and compare it with dry col-
mineral using steered molecular dynamics (SMD) simula- lagen (unsolvated collagen) when interacting with HAP
tions. In our prior work on a biological hybrid system surface. This has been done using SMD simulations.
nacre, [32], we have observed marked influence of prox- In order to perform simulations, nano and molecular
imity of mineral on the mechanics of the protein molecule structure of collagen-HAP system should be understood
at organic-inorganic interfaces. clearly. At the nanoscale, bone consists of HAP mineral-
Also, it has been observed from experiments that water ized at specific locations on collagen molecules [7].
affects the overall mechanical response of bone. Bone has Between the collagen molecules, HAP mineralizes in
about 10 % of water in body environment (Table 1) [28, specific zones called as hole zones (Figs. 1 and 2) [44,
3335] and with loss of water, bone exhibits different 45]. The mineralized HAP in these zones is hexagonal in
mechanical properties [3643]. It has been observed that structure with its c-axis aligned parallel to collagen fibers
strain-at-fracture and energy-to-fracture decreases, whereas [46]. Collagen exhibits helical structure whereas its ends
tensile strength, stiffness, and hardness increases with loss are non-helical and are known as N- or C-telopeptides
of water [3641]. Nyman et al. have shown that toughness [4749]. Thus, collagen interacts with HAP through these
of bone decreases when water is removed from collagen, telopeptides as shown in Fig. 2. We have constructed the
whereas both the strength and toughness decrease when model in a similar way as shown in Fig. 2. The main focus
water is lost from mineral phase [42]. Currey et al. have of the present modeling study with SMD simulations is to
shown that water affects the viscoelastic behavior of bone evaluate the role of mineral (HAP) and water on the
[43]. Dry bone exhibits lower anelastic deformation as load-deformation behavior of collagen molecule in close
proximity of HAP.
Table 1 Bone Composition (adapted from reference 19)
Components Weight percentage Simulation Details
Hydroxyapatite 60
NAMD [50] has been used to perform MD and SMD
Collagen 20
simulations using CHARMm (Chemistry at Harvard
Water 9
Macromolecular mechanics) force field [51] and VMD [52]
Ions and Non-collagenous proteins 11
has been used for all the interactive studies and visualization.
123
J Mater Sci (2007) 42:87958803 8797
NAMD has been developed by theoretical and computa- minimization using CHARMm force field [51]. Further,
tional biophysics group at the Beckham Institute for the optimized model is solvated with 1800 TIP3P
Advanced Science and Technology at the University of water molecules followed by energy minimization of the
Illinois at Urbana-Champaign. Minimization has been solvated N-Collagen. Furthermore, the temperature is
performed by conjugate gradient method on all models increased to 300 K from 0 K in steps of 100 K followed
before performing MD and SMD simulations. Isothermal by increasing pressure from 0 bar to 1.01 bar in steps of
and isobaric ensemble (NPT) is used for MD and SMD 0.25 bar. The resulting structure has been used to per-
simulations with periodic boundary conditions. Electro- form SMD simulations in absence of HAP. The same
static interactions between pairs are calculated by Particle geometrically optimized N-Collagen molecule has been
Mesh Ewald (PME) technique. The van der Waals cut off used with HAP for interaction studies. In order to ana-
distance of 9 A has been used in all simulations. Tem- lyze the role of water on the overall mechanical
perature of all models is controlled by Langevin dynamics response of N-Collagen, unsolvated N-Collagen is also
and pressure is increased to 1.01 bar in steps and is cal- used for SMD simulations in close proximity and in
culated by Nose-Hoover Langevin piston method [53, 54]. absence of HAP. Here, again the same route has been
Verlet algorithm is used to integrate the Newton equation taken to obtain unsolvated N-Collagen molecule for
of motion [53, 54]. MD simulations have been performed SMD simulations.
on all models followed by SMD simulations.
HAP crystal model
Model Construction And Simulations
HAP mineral has a hexagonal structure with space group
Collagen Molecule with N-Telopeptide P63/m [59]. The unit cell parameters are a = 9.424 A ,
b = 9.424 A, c = 6.879 A, a = 90, b = 90, c = 120. The
The collagen is a triple helix molecule with three poly- dimensions of the HAP used in the present study are
peptide chains with each polypeptide chain forming a left- a = 75.392 A , b = 75.392 A , c = 20.637 A , a = 90,
handed helix that is folded in a right-handed superhelix b = 90, c = 120 which correspond to 192 unit cells of
[55]. The structure of collagen molecule with telopeptide is HAP i.e., 8 unit cells along a-axis, 8 unit cells along b-axis
obtained from literature [49], where the (GPP)n model of and 3 unit cells along c-axis. The dimensions of periodic
collagen was obtained from the Protein Data Bank-ID boundary conditions are the same as dimensions of the
1 k6f.pdb [56]. Further, N-terminal telopeptide sequence HAP model used. CHARMm force field parameters for
used in the simulations is as follows [57, 58]: HAP have been obtained from our previous study of
interfaces between HAP and polyacrylic acid [60, 61]. The
a1-GlnLeuSerTyrGlyTyrAspGluLysSerThrGlyIleSer-
constructed HAP model is first geometrically optimized
ValPro-helix (GlyProMet-)
through energy minimization then its temperature is
a2-GlnPheAspAlaLysGlyGlyGlyPro-helix (GlyProMet-)
increased from 0 K to 300 K in steps of 100 K. Further,
The three stranded N-telopeptide are joined to triple
pressure is increased to 1.01 bar in steps of 0.25 bar.
helix molecule (GPP).
Finally, the model is again geometrically optimized
First, collagen molecule with N-terminal telopeptide through energy minimization. The optimized model has
(N-Collagen) is geometrically optimized through energy been used to create appropriate surfaces. It has been found
123
8798 J Mater Sci (2007) 42:87958803
from experiments that the c-axis of HAP aligns parallel to Steered molecular dynamics (SMD) simulations
collagen molecule, i.e. (0001) surface of HAP interacts
with telopeptide of collagen molecule (Fig. 2). The (0001) SMD simulation is a type of MD simulation technique in
surface of HAP has been created by extending the c-axis of which force is applied to selected atom/atoms. It provides
HAP model from 20.637 A to 320.637 A which creates the dynamics of binding and unbinding of analyzed mol-
300 A of vacuum between (0001) surface of HAP and its ecules with applied load [6474]. It also provides the
periodic image. This creates a pseudo 2D periodic mechanical response of molecules. SMD simulations can
boundary condition from 3D periodic boundary condition. be conducted at: constant force [64, 75] and constant
The (0001) surface of HAP is rendered non-dipolar by velocity [6467, 76]. We have used constant velocity SMD
removing half of the surface ions (calcium atoms) from (v-SMD) in which constant velocity is applied to selected
(0001) surface to its opposite surface. The construction of atom/atoms. Force is calculated in v-SMD as F = k(vtx),
non-dipolar surface from dipolar surface has been per- where k is the spring constant of the spring attached to the
formed in accordance with Tasker et al. [62]. The non- pulled atom/atoms, and v is the pulling velocity, t is the
dipolar surface has been constructed because it has been time and x is displacement. In order to analyze the response
observed previously that these surfaces are stable as com- of N-Collagen molecule under load, the center of mass of
pared to dipolar surfaces [63]. The HAP model with (0001) N-Collagen has been pulled at a velocity of 1 A /ps. The
1 2
surface is geometrically optimized through energy mini- force constant of the spring is 4.0 kcal mol A which
mization and then the temperature is increased from 0 K to corresponds to spatial (thermal)
p
fluctuation of the center of
300 K in steps of 100 K. Further, pressure is increased to mass of 0.77 A kB T=k [77] at 300 K. The ends of the
1.01 bar. Finally, the model is geometrically optimized in telopeptides chains are fixed (carbonyl carbon atoms of
order to obtain the structure of the model at the global pyroglutamic acid residues of all the chains). In order to
energy minimum. This optimized model has been used analyze the load-deformation behavior of N-Collagen, the
with solvated and unsolvated N-Collagen to construct center of mass of the N-Collagen is pulled in close prox-
HAP-Collagen model. imity of HAP and in absence of HAP. To understand the
role of water on the load-deformation behavior, another
Hydroxyapatite with N-Collagen molecule (HAP- type of v-SMD simulations have been performed in which
Collagen) center of mass of solvated and unsolvated N-Collagen is
pulled in close proximity of HAP separately and the layers
The optimized model of solvated and unsolvated N-Col- of HAP opposite to (0001) surface are fixed (Fig. 4).
lagen is brought in close proximity of minimized model of
HAP with (0001) surface separately (HAP-Collagen)
(Fig. 3). The model thus constructed is first geometrically Interaction energy calculation
optimized through energy minimization. Further, the same
procedure is used to reach a temperature of 300 K at a The interaction energy between the components has
pressure of 1.01 bar as mentioned in the previous sections. been evaluated by energy evaluation tool of NAMD,
Finally, such created HAP-Collagen model is used for MDEnergyTM [50]. Here, energy is calculated between
SMD simulations. defined atoms using a specified cut-off distance. This
123
J Mater Sci (2007) 42:87958803 8799
Fig. 4 Pulling of solvated N-Collagen in close proximity of HAP. Fig. 5 Load-Deformation plots of (a) Solvated and (b) Unsolvated
Here Layers opposite to (0001) surface of HAP were fixed (as shown collagen molecule in close proximity and in absence of HAP
in white color) (N-Collagen molecules are represented in VDW (a1
chains are in black color whereas a2 chain is in white color), Water is
represented in ball and stick, and HAP is represented in VDW) N-Collagen, and also from unbinding of water molecules
from N-Collagen. Finally, the region A5A6 is attributed to
the backbone chain of the N-Collagen molecule. Similarly,
technique uses the trajectory file obtained during simula- the same type of behavior is also obtained for N-Collagen
tions, the topology file of the structure, and the parameter in absence of HAP as shown in Fig. 5a. However, the
file. The interaction energy between the atoms has been region A2A3 which has been obtained for the N-Collagen
calculated for the entire 150 ps of v-SMD simulations. in presence of HAP is not observed for the N-Collagen in
absence of HAP. The reason for this is that at the start of
the region A2A3 (in the plot for N-Collagen in presence of
Results and discussion HAP), N-Collagen unbinds from the water molecules and
the water molecules interact with the charged surface of
Figure 5a shows the load-deformation response of solvated HAP, whereas in the absence of HAP, there is absence of
N-Collagen in close proximity (grey) and in absence interaction of water molecules with any surface.
(black) of HAP (Fig. 6). The load-deformation behavior of In order to understand the role of water on the
N-Collagen in close proximity of HAP shows five distinct mechanics of N-Collagen, unsolvated N-Collagen is also
regions and they are represented as A1A2, A2A3, A3A4, used and is shown in Fig. 5b, where pulling of unsolvated
A4A5, and A5A6. The slopes of these regions represent the N-Collagen in close proximity of HAP is represented by
stiffness of N-Collagen when HAP is present. The change grey color whereas the pulling of unsolvated N-Collagen in
in slope corresponds to unbinding and binding events absence of HAP is represented by black color. The
occurring at different time intervals. The region A1A2 response of N-Collagen in close proximity of HAP exhibits
results from the interaction of N-Collagen with water five regions: C1C2, C2C3, C3C4, C4C5, and C5C6
molecules and with HAP molecule. The change in slope whereas the N-Collagen in absence of HAP exhibits six
from region A1A2 to A2A3 results from unbinding of regions: D1D2, D2D3, D3D4, D4D5, D5D6, and D6D7. A
N-Collagen from water molecules that are interacting with similar type of trend has been observed for several proteins
N-Collagen and HAP surface. The regions A3A4 and when pulled at constant velocity [71, 78]. The regions
A4A5 result from breaking of hydrogen bonds between C5C6 and D6D7 result from backbone chain of
proline and glycine residues of helical regions of N-Collagen. On comparing the slopes of regions A1A2
123
8800 J Mater Sci (2007) 42:87958803
(44.93 kcal mol1A 2) and B1B2 (53.03 kcal mol1A 2), of observed that the solvated N-Collagen requires more
solvated N-Collagen with the slopes of the regions C1C2 energy (296.87 kcal/mol) to untie from the surface as
(53.24 kcal mol1A 2) and D1D2 (57.92 kcal mol1A 2) of compared to the unsolvated N-Collagen molecule
unsolvated N-Collagen, it has been observed that the slopes (210.34 kcal/mol). This comparison has been performed
of these regions are similar. This indicates that the presence for displacement of center of mass up to 50 A as beyond
of HAP and water does not affect stiffness of the backbone
50 A, the area does not change significantly. To understand
chain of N-Collagen molecule. However, the peak loads of the attachment of N-Collagen molecule with HAP, inter-
solvated and unsolvated N-Collagen molecules in close action energy between various components during v-SMD
proximity of HAP are 10600 pN and 9700 pN respectively, simulations has been calculated using MDEnergyTM as
at displacements of 34.7 A and 15.6 A respectively. These shown in Fig. 9. During v-SMD simulations, interaction
are observed as peak loads when the contribution of the energy between HAP-water is highest followed by inter-
backbone of N-Collagen is not significant in the load- action energy between collagen-water and then between
deformation response. Similarly, we have observed that the HAP-collagen. Although significant changes are apparent
peak loads of solvated and unsolvated N-Collagen mole- in the load-deformation response as shown in Fig. 7, the
cules in absence of water are 10000 pN and 6500 pN with corresponding changes are not significant in the Energy-
displacements 27.7 A and 15.6 A respectively. This indi- Time plot (Fig. 9). This indicates that the interaction
cates that water influences the load-deformation response between HAP and N-Collagen occurs through water i.e.,
of N-Collagen. In order to analyze the role of water on the the load is transferred to HAP from N-Collagen through the
interaction of N-Collagen with HAP, another type of water molecules present between HAP and N-Collagen.
v-SMD simulations have been performed as mentioned in
the SMD section. Here, solvated and unsolvated N-Colla-
gen molecules are pulled in close proximity of HAP while Conclusions
keeping the layers opposite to (0001) surface of HAP fixed
(Fig. 7). The load-deformation response thus obtained is Bone exhibits a complex structural hierarchy that spans
shown in Fig. 8a, b. The area under the load-deformation over nanometer to millimeter. At the nanostructural level,
plot represents the energy needed by a molecule to deform. bone is primarily composed of HAP and collagen. Here, we
The area under the plots of Fig. 8a, b represents the energy have evaluated the load-deformation behavior of N-Colla-
needed by solvated and unsolvated molecule to untie from gen molecules in close proximity and in absence of HAP
the HAP surface. On comparing the area of the using v-SMD simulations. It has been observed that the
load-deformation response of solvated and unsolvated load-deformation response of solvated N-collagen in close
N-Collagen in close proximity of HAP, it has been proximity of HAP has features which result from breaking
123
J Mater Sci (2007) 42:87958803 8801
of hydrogen bonds between N-Collagen and water, where N-Collagen. We have shown from the energy-time plots of
water is interacting with HAP significantly. These features various pairs that N-Collagen interacts with HAP through
are not present in the plot for N-Collagen in absence of water.
HAP due to lack of interaction between water and HAP. The present study is focused on the influence of mineral
The load-deformation response for solvated N-Collagen and water on the load deformation behavior of N-Collagen
differs from the load-deformation response of unsolvated in HAP-Collagen model system. This model system is the
N-Collagen. To understand the role of water on the load- initial step towards understanding the molecular and nano
deformation response of N-Collagen in close proximity of mechanical response of bone. However, in order to predict
HAP, solvated and unsolvated N-Collagen molecules are the molecular and nano mechanical response of bone
pulled separately in close proximity of HAP. It has been accurately, the present model needs to incorporate several
observed from the plots of load-deformation response that complex molecular and nano features of bone such as
the solvated N-Collagen requires more energy to untie intergrowth. Also, in order to evaluate the mechanical
from the HAP surface as compared to unsolvated response of nano and molecular structure of real bone
123
8802 J Mater Sci (2007) 42:87958803
References
123
J Mater Sci (2007) 42:87958803 8803
39. Sedlin ED, Hirsch C (1966) Acta Orthopaedica Scandinavica 59. Sudarsanan K, Young RA (1969) Acta Crystallographica B
37:29 25:1534
40. Smith JW, Walmsley R (1959) J Anatomy 93:503 60. Bhowmik R, Katti KS, Verma D, Katti DR (2007) Mater Sci Eng
41. Yamada H, Evans FG (1970) Strength of biological materials. C 27:352
Williams & Wilkins, Baltimore 61. Bhowmik R, Katti KS, Katti D (2006) Polymer 48:664
42. Nyman JS, Roy A, Shen X, Acuna RL, Tyler JH Wang X (2006) J 62. Tasker PW (1979) J Phys C Solid State Phys 12:4977
Biomech 39:931 63. Bertaut FCR (1958) Acad Sci Paris 246:3447
43. Currey JD (1965) J Exp Biol 43:279 64. Izrailev S, Stepaniants S, Balsera M, Oono Y, Schulten K (1997)
44. Fitton-Jackson S (1957) Proc R Soc London Ser B 146:270 Biophys J 72:1568
45. Nylen MU, Scott DB, Mosley VM (1960) In: Sognnaes RF (ed) 65. Isralewitz B, Izrailev S, Schulten K (1997) Biophys J 73:2972
Calcification in Biological Systems. Am Assoc Adv Sci p 129 66. Wriggers W, Schulten K (1999) Proteins Struct Funct Genet
46. Itoha S, Masanori K, Koyamac Y, Takakudac K, Shinomiyaa K, 35:262
Tanaka J (2002) Biomaterials 23:3919 67. Kosztin D, Izrailev S, Schulten K (1999) Biophys J 76:188
47. Kadler KE, Holmes DF, Trotter JA, Chapman JA (1996) Bio- 68. Molnar F, Ben-Nun M, MartiNez TJ, Schulten K (2000) J Mol
chem J 316:1 Struct (THEOCHEM) 506:169
48. Malone JP, Alvares K, Veis A (2005) Biochem 44:15269 69. Lu H, Isralewitz B, Krammer A, Vogel V, Schulten K (1998)
49. Malone JP, George A, Veis A (2004) Proteins: Struct Funct & Biophys J 75:662
Biol 54:206 70. Krammer A, Lu H, Isralewitz B, Schulten K, Vogel V (1999)
50. Kale L, Skeel R, Bhandarkar M, Brunner R, Gursoy A, Krawetz Proc Natl Acad Sci USA 96:1351
N, Phillips J, Shinozaki A, Varadarajan K, Schulten KJ (1999) 71. Lu H, Schulten K (1999) Proteins Struct Funct Genet 35:453
Comput Phys 151:283 http://www.ks.uiuc.edu/Research/namd/, 72. Lu H, Schulten K (1999) Chem Phys 247:141
as on 16 January 2007 73. Izrailev S, Stepaniants S, Isralewitz B, Kosztin D, Lu H, Molnar
51. Brooks BR, Bruccoleri RE, Olafson BD, States DJ, Swaminathan F, Wriggers W, Schulten K (1998) In: Computational molecular
S, Karplus MJ (1983) J Comput Chem 4:187 dynamics: challenges, methods, ideas. volume 4 of Lecture Notes
52. Humphrey W, Dalke A, Schulten KJ (1996) J Mol Graphics in Computational Science and Engineering, Springer-Verlag,
14:33, http://www.ks.uiuc.edu/Research/vmd/, as on 16 January Berlin p 39
2007 74. Schulten K, Schulten Z, Szabo A (1981) J Chem Phys 74:4426
53. Feller SE, Zhang Y, Pastor RW, Brooks BR (1995) J Chem Phys 75. Lu H, Krammer A, Isralewitz B, Vogel V, Schulten K (2000) In:
103:4613 Elastic filaments of the cell. Academic/Plenum Publishers, New
54. Martyna GJ, Tobias DJ, Klein ML (1994) J Chem Phys 101:4177 York, p 143
55. Beck K, Brodsky B (1998) J Struct Biol 122:17 76. Grubmuller H, Heymann B, Tavan P (1996) Science 271:997
56. Berisio R, Vitagliano L, Mazzarella L, Zagari A (2002) Protein 77. Balsera M, Stepaniants S, Oono Y, Schulten K (1997) Biophys J
Sci 11:262 73:1281
57. Jones EY, Miller A (1987) Biopolymers 26:463 78. Gao M, Wilmanns M, Schulten K (2002) Biophysical J 83:3435
58. Otter A, Kotovych G, Scott PG (1989) Biochemistry 28:8003
123
J Mater Sci (2007) 42:88048810
DOI 10.1007/s10853-007-1916-z
Received: 19 December 2006 / Accepted: 4 June 2007 / Published online: 10 July 2007
Springer Science+Business Media, LLC 2007
Abstract In this work, subchondral cortical bone mate- exhibits remarkable strength and resilience to fracture
rial is investigated from the joints of five horses, three of under most circumstances. Increasingly, the structure of
which presented with no clinical signs or radiographic bone on molecular length scales has been studied in order
signs of osteoarthritis and two of which suffered osteoar- to understand its material properties in more detail and so
thritis joint disease, as judged by clinical and radiographic to generate a design template for new synthetic materials
assessment and histological findings. The horse is a good mimicking its properties.
model for osteoarthritis in humans, so the aim of this study Bone is known to fail as a material under some
is to use nuclear magnetic resonance (NMR) for a detailed instances. Trauma generated fractures are common, as is
investigation of the bone structure in bone material affected repetitive overloading injury in athletes. The bone dis-
by osteoarthritis. In particular, we report on the assessment eases of osteoporosis and osteoarthritis (OA) have huge
of the mineral structure of these samples as viewed by economic and welfare impact in the human population. In
solid-state NMR. the case of osteoporosis, the material strength of the bone
is clearly compromised. However, the effect of osteoar-
thritis on the integrity of bone material is less well
understood. An osteoarthritis-associated increase in sub-
chondral bone volume and trabecular thickness has been
Introduction detected in subchondral cortical and cancellous bone from
a number of species [14]. An increase in bone mineral
Bone as a material is interesting from a number of points of density has been associated with osteoarthritis in human
view. From the biological perspective, it is the material joints [5] and it appears that osteoarthritis is less likely in
which gives higher animals their shape and form and osteoporotic individuals [57]. However, although in-
allows locomotion. From the materials science perspective, creased remodeling of subchondral bone in osteoarthritis
it is an organicinorganic composite material which has been indicated in a number of reports [2, 3, 810],
hypomineralization of subchondral bone has also been
noted [2, 11]. It is reported that the stiffness of sub-
S. Maltsev C. Jaeger (&)
Federal Institute for Materials Research and Testing, Division chondral cancellous bone increases in osteoarthritis [12,
I.3, Working Group NMR Spectroscopy, Richard Willstaetter 13], while the stiffness of subchondral cortical bone de-
Str. 11, D-12489 Berlin, Germany creases [1416]. It is clear from these various reports that
e-mail: christian.jaeger@bam.de
bone mineral structure plays a part in osteoarthritis,
M. J. Duer though less clear how. Accordingly, we have undertaken a
Department of Chemistry, University of Cambridge, Lensfield solid-state NMR study in order to characterize the bone
Road, Cambridge CB2 1EW, UK mineral in osteoarthritis, in order to improve our under-
standing of the molecular basis of the disease.
R. C. Murray
Centre for Equine Studies, Animal Health Trust, Lanwades Park, Solid-state NMR is an ideal method for studying com-
Kentford, Newmarket, Suffolk CB8 7UU, UK posite materials such as bone. An NMR spectrum for a
123
J Mater Sci (2007) 42:88048810 8805
Samples Histology
Osteochondral samples from the dorsal proximal aspect of Osteochondral samples underwent routine histological
the third tarsal bone were obtained from five horses fol- preparation, decalcification and paraffin-embedding.
lowing humane destruction for clinical reasons unrelated to Four lm thick sections were stained with Toluidine blue
this study. Three horses had no clinical, radiographic or and Haematoxylin and Eosin for evaluation of the articular
histological evidence of abnormality or pain (samples 27, cartilage and subchondral bone. Cartilage morphology was
82, 85, aged 3, 6 and 7 years, respectively). Two horses assessed using an Olympus DP12 microscope (Olympus
had clinical signs of pain localized to the tarsometatarsal UK Ltd., London, UK) and a polarized light. Each section
joint with radiographic and histological evidence of was assessed with respect to chondrocyte orientation and
osteoarthritis (samples 64 and 124, aged 10 and 17 years). morphology, staining patterns, articular surface and
Naturally-hydrated samples were frozen in liquid osteochondral junction defects; focal cartilage and sub-
nitrogen and ground in a Sartorius Dismembrator ball mill. chondral bone abnormalities; cartilage erosion; osteophyte
The samples were then immediately transferred via a fun- formation; abnormally increased area of dense bone and
nel into the NMR rotors (approximately 100 mg) and focal cancellous bone abnormalities. Presence of osteoar-
lightly pressed into place with a plastic rod. thritic change was based on previously described histo-
123
8806 J Mater Sci (2007) 42:88048810
logical scoring systems for cartilage and subchondral bone The spectra shown in Fig. 2 for the two different sam-
[20, 21]. ples are very similar in both 1H and 31P lineshapes and in
the relative intensities of the various signals to each other
for equivalent mixing times. Moreover, they are similar to
Results and discussion those reported previously by other workers examining the
mineral structure of bone from bovine and rat species [23],
Histology and bovine, rat and human [22]. Very similar spectra are
found for all the other samples in this work and results
Samples 27, 82 and 85 had no evidence of osteoarthritic from experiments which used HartmannHahn cross
change in either the cartilage or subchondral bone at any polarization rather than FSLG are similar also (a loss of the
site within the joint. Samples 64 and 124 had clear evi- water signal due to spin-diffusion processes can occur in
dence of osteoarthritic change within this joint cartilage HartmannHahn experiments). The latter is not surprising
structural disorganization with chondrone formation and as 1H1H distances in the mineral matrix are expected to be
subchondral bone fibrosis. Sample 64 had clear definition large so that at the MAS frequency (12.5 kHz), spin dif-
of the subchondral bone margins. Sample 124 had marked fusion between 1H spins is expected to be slow on the
thickening of the subchondral bone with generalized loss of timescale of the cross polarization transfer and under such
porosity, typical of osteoarthritic changes (Fig. 1). circumstances, the cross polarization dynamics are similar
for FSLG and HartmannHahn cross polarization.
NMR The spectra in Fig. 2 show at least three broad signals in
the 1H dimension: a sharper signal at approximately 0 ppm,
Fig. 2 shows the two-dimensional 1H31P HETCOR due to OH hydroxyl ions, a broad peak centered around
spectra for samples 27 and 64 as typical samples from the 5.5 ppm due to structural water in the mineral (or at least,
non-OA and OA groups. The spectra were recorded using water in close proximity to mineral phosphate, otherwise
FSLG (frequency switched Lee-Goldburg) cross polariza- this signal would not appear in the spectrum at all) and a
tion and 1H31P mixing times of 0.5 ms, 2 ms and 8 ms. very broad peak in the region 515 ppm from HPO2 4
These spectra show the spatial correlations between 1H hydrogen phosphate groups. The OH 1H signal is corre-
spins (from various groups) and 31P spins (primarily lated with a relatively narrow 31P resonance, indicating that
phosphate species contained in the mineral component) in the phosphate groups in close spatial proximity to the hy-
the sample, and so essentially contain information about droxyl groups are relatively well ordered, and thus belong
the spatial correlation of 1H containing groups in the to a relatively crystalline hydroxyapatite structure. The
mineral/closely bound to the mineral surface and mineral H2O and HPO2 4
1
H signals, on the other hand, are corre-
phosphate, there being little phosphate elsewhere in the lated with relatively broad 31P lines, indicating that these
1
bone matrix observable by NMR [22]. As most of the H and their nearby phosphate/hydrogen phosphate groups
protons (1H) of the protein matrix are not close to 31P, the are in relatively disordered environments.
protons in the protein matrix are not observed in this Two-dimensional 31P31P correlation spectra for the
experiment, making it an excellent method by which to samples under non-spinning conditions for mixing times of
examine the mineral component of a sample exclusively, 1 ms, 10 ms, 100 ms, 1 s and 10 s and room temperature
without resorting to deproteination of the sample, as dis- (not presented in this paper) were undertaken to establish
cussed previously by Ackermann et al. [22]. whether all mineral crystals are similar or whether there are
populations with different species of phosphate groups. At
short mixing times, the exchange pattern in the two-
dimensional spectrum is simply a diagonal lineshape as
there is essentially no spin diffusion between the 31P spins
on this timescale. As the mixing time increases, there are
increasing amounts of exchange intensity correlating dif-
ferent 31P signals, until at mixing times of 1 s and above,
the exchange pattern is essentially circular, indicating that
all 31P signals are correlated with each other. Moreover, the
buildup of exchange intensity with mixing time to this
position is smooth. In other words, all the 31P sites repre-
sented in these spectra (and therefore, also the HETCOR
spectra in Fig. 2) exist in every crystallite in the sample,
Fig. 1 Stained microscopy of the osteoarthritic sample 124 i.e. the different 1H sites and phosphate groups inferred
123
J Mater Sci (2007) 42:88048810 8807
7 6 5 4 3 2 1 0 -1 -2 -3 7 6 5 4 3 2 1 0 -1 -2 -3
-2 -2 -2 -2
0 0 0 0
2 2 2 2
4 4 4
( H), ppm
( H), ppm
4
6 6 6 6
8 8 8 8
10 10 10 10
1
1
12 12 12 12
14 14 14 14
16 16 16 16
18 18 18 18
20 20 20 20
7 6 5 4 3 2 1 0 -1 -2 -3 7 6 5 4 3 2 1 0 -1 -2 -3
31
( P), ppm 31
( P), ppm
7 6 5 4 3 2 1 0 -1 -2 -3
-2 -2
0 0
2 2
4 4
( H), ppm
6 6
8 8
10 10
12 12
1
14 14
16 16
18 18
20 20
7 6 5 4 3 2 1 0 -1 -2 -3
31
( P), ppm
Fig. 2 2D 1H31P FSLG HETCOR for sample 27 and sample 64 at mixing times of 1 ms and 1.5 ms respectively and sample 27 at mixing times
of 8 ms
from the HETCOR spectra in Fig. 2 exist in every crys- The rate constants were determined by fitting the
tallite; there are not different mineral phases in different experimental data to Eq. (1) [24].
crystallites for instance.
We thus ascribe the more disordered region containing St a1 expt=sHP expt=sd 1
H2O/phosphate and HPO2 4 groups to surface regions of the
crystallites and the more ordered OH/PO3 4 groups to a more where S(t) is the 31P signal intensity for a mixing time of t,
crystalline core in the crystallites as it was shown recently for sHP is the rate constant for the 1H31P cross polarization
nanocrystalline hydroxyapatite in detail [22]. This is con- process and sd is a decay constant describing the decay of
sistent with the idea that the surface regions of any crystal are the 31P signal through various relaxation processes affect-
expected to be more disordered due to the lack of coordi- ing both the 1H and 31P spins.
nation around surface ions, and/ or coordination of surface The sHP rate constants determined for the 31P signal
ions by heterogeneous molecules from the protein matrix in intensity in the three regions indicated in Fig. 3 are given
the case of mineral crystals in bone. The H2O/phosphate in Tables 1 and 2 for HETCOR spectra using Hartman
region may arise in part or whole from matrix water bound to Hahn and FSLG cross polarization, respectively.
the surface of the mineral crystallites. It should be noted that sHP for the OH/PO3 4 region is
HETCOR spectra in Fig. 2 were recorded in fact for difficult to determine accurately due to the near linear
between four and ten mixing times for each sample. This profile of the signal buildup for the 31P intensity in this
allows us to determine the approximate rate constants for spectral region over the range of mixing times used.
the (presumed exponential) build up of 31P magnetization The data in Tables 1 and 2 shows that the sHP rate
by cross polarization from 1H, which in turn depends on the constants for each 31P spectral region are very similar
1
H31P distances involved, as well as other interactions within the limits of error for all samples studied. This, plus
affecting both the 1H and 31P spins and any local molecular the similarity of the two-dimensional HETCOR spectra
dynamics (for instance, local rotations or librations of themselves leads us to conclude that the mineral compo-
water molecules). sitions and structures are very similar for all samples and
123
8808 J Mater Sci (2007) 42:88048810
12 12
HPO42- One possible explanation for this effect is that at the longer
14 14
16 16
-
18 18 1msec LGCP OH
-
20 20 4msec LGCP OH
-
7 6 5 4 3 2 1 0 -1 -2 -3 1.0 8msec LGCP OH
31
( P), ppm
0.8
Fig. 3 OH, water and HPO2
4
1
H spectral assignment on the 1H31P
correlation spectra 0.6
0.4
Table 1 The rate constants sHP for the 31P intensity in the three
spectral regions shown in Fig. 2 of 1H31P HETCOR spectra recorded 0.2
using HartmanHahn cross polarization
Sample sHP (OH/PO3 sHP (H2O/PO3 sHP (HPO2 0.0
4 4 4
region)/ms region)/ms region) /ms 15 10 5 0 -5 -10
31
( P), ppm
27 2.9 0.5 0.6 0.1 0.4 0.2
64 10 4 0.6 0.1 0.6 0.1
1msec LGCP H2O
124 2.6 0.8 0.7 0.1 0.6 0.2 4msec LGCP H2O
1.0 8msec LGCP H2O
Error estimates are from curve fitting and do not take direct account
of experimental errors
0.8
0.6
Table 2 The rate constants sHP for the 31P intensity in the three
spectral regions shown in Fig. 2 of 1H31P HETCOR spectra recorded 0.4
using FSLG cross polarization
0.2
Sample sHP (OH/PO3
4 sHP (H2O/PO3
4 sHP (HPO2
4
region) /ms region)/ms region)/ms
0.0
64 1.7 0.3 0.5 0.1 0.4 0.1 15 10 5 0 -5 -10
31
124 21 0.3 0.1 0.2 0.2 ( P), ppm
Error estimates are from curve fitting and do not take direct account of 1msec LGCP HPO4
2-
thus that joint disease has not changed the mineral structure 0.8
in any way that we can measure here.
It is interesting to examine the 31P lineshapes extracted 0.6
from the two-dimensional HETCOR spectra in Fig. 2.
0.4
These are plotted for sample 27 in Fig. 4 for different
1
H31P FSLG mixing times for the three spectral regions 0.2
indicated in Fig. 3. Other samples give near identical plots,
indicating that the phosphate environment and distribution 0.0
is very similar in all samples. The 31P lineshape for the 15 10 5 0 -5 -10
31
OH/PO3 4 spectral region is invariant with mixing time in
( P), ppm
both the FSLG and HartmanHahn mixing time experi-
Fig. 4 The 31P lineshapes extracted from the two-dimensional FSLG
ments. This is the behaviour expected for a single com- HETCOR spectra for the three spectral regions defined in Fig. 2 for
ponent (crystalline) system in which there is one type of sample 27 for FSLG mixing times of 18 ms
123
J Mater Sci (2007) 42:88048810 8809
mixing times, there is spin diffusion from H2O 1H to OH Table 3 The 1H and 31P T1 relaxation time constants for the bone
groups in the ordered hydroxyapatite regions somewhere in samples used in this work. The measurement method is described in
the Experimental section
the experiment prior to cross polarization to 31P, so that
magnetization which initially started on H2O ends up on Sample 27 82 85 64 124
the ordered phosphate groups (and 31P spins thereof) in the 1
H T1/s 1.1 0.3 1.1 0.2 1.0 0.2 1.1 0.2 1.2 0.3
hydroxyapatite core of the crystallites. However, FSLG 31
P T1/s 244 4 207 3 213 3 228 3 206 6
mixing does not permit spin diffusion between 1H spins
during the mixing period and the longest t1 time used in the
HETCOR experiment is 6 ms which from previous work experimental data. Thus, from the values in Table 3, we
[25] is too shorter a time for there to be significant spin believe that the T1 values for both mineral 1H and 31P are
diffusion between 1H spins (significant spin diffusion very similar between all samples once all sources of error
between OH and H2O only occurs on a timescale of are taken into account. Two methods were used for
200 ms [25]). Thus, the existence of the narrowed 31P line determining the T1 values: (1) direct T1 measurements,
does not arise from spin diffusion between 1H, but must after first selecting the mineral 1H spectroscopically (which
instead arise from a second phosphate site spatially corre- is done here using a rotor-synchronized echo; the protein
1
lated with H2O which has different cross polarization H signals have relatively short transverse relaxation times
dynamics to the sites which account for the main cross and so decay during the echo period, leaving only the
polarization events at shorter mixing times. The fact that mineral 1H signal at the end of the echo experiments), and
this second site does not become apparent until longer (2) indirect measurement using cross polarization from 1H
mixing times suggests that either (a) its cross polarization to 31P (which automatically selects the mineral 1H, as only
transfer rate constant sHP is significantly longer, or (b) the these are close enough to 31P spins to cross polarize). In
decay constant governing the decay of the 31P cross first case, we obtain relaxation information of every 1H
polarization signal is significantly less than, those for the signal independently. In the second case, we obtain the
sites which account for the majority of polarization transfer information about 1H relaxation through the 31P nuclei. In
at short mixing times. Alternatively, both (a) and (b) may general, the T1 relaxation times measured in these two
occur to some extent. In case (a), the conclusion would be ways can be different. The difference tells us about the
that the H2O from which the cross polarization is occurring internal structure of proton lines: if, for example, T1
for this second H2O/phosphate site is relatively distant relaxation measured via cross polarization to 31P is faster,
from the phosphate to which it cross polarizes and in case then there are at least two components forming the 1H line
(b), that the rotating-frame spin-lattice relaxation times for and this should be taken account of when analyzing the
the 31P and/or 1H involved in the cross polarization process experimental relaxation data. However, T1 values mea-
for the second H2O/phosphate site are significantly smaller sured by methods (1) and (2) were very similar, strongly
than for other H2O/phosphate groups with the faster cross suggesting that a one exponent analysis is sufficient.
polarization rates. Moreover, the phosphate groups Furthermore, in our work, using different cross polari-
involved in this second site are (from the narrowed 31P zation contact times (100 ls and 9 ms) yields very similar
lineshape) more ordered than for the rapidly cross polar- T1 behaviour, despite the fact, that for a 100 ls contact
izing H2O/phosphate sites. A likely explanation is that the time, there is almost no signal from hydroxyapatite OH 1H
mineral bone crystals consists of crystalline hydroxyapatite present in the spectrum and at 9 ms, almost no HPO2 4 or
as core covered with an amorphous layer as found for H2O signal contributes to the spectra. This further suggests
nanocrystalline hydroxyapatite [23]. It must then be con- that one exponent fitting of the experimental T1 relaxation
cluded that the water molecules are close to this interface curves may be used and, within error limits, this gave
which in turn would explain the higher degree of order of acceptable fits in every case.
the phosphate anions due to the near hydroxyapatite
structure.
For completeness in this study, we have recorded the Conclusions
mineral 1H and 31P T1 relaxation time constants for each
sample also. These are shown in Table 3. It should be We have investigated the molecular-level structure of the
noted that a single T1 time constant describes the relaxation mineral in bone taken from the subchondral regions of
of the entire 1H spectrum within experimental error. The equine joints where there was clinical and histological
estimates of error are estimates of the error from fitting the evidence of OA joint disease. Our study used solid-state
experimental data points to an exponential curve. They do NMR as the primary method of structure investigation and
not take into account noise in the experimental data nor the we employed a number of different experiments including
effects of any baseline errors (for instance) in the two-dimensional 1H-31P HETCOR spectra and relaxation
123
8810 J Mater Sci (2007) 42:88048810
time measurements. Our results do not find any differences 9. Murray RC, Vedi S, Birch HL, Lakhani KH, Goodship AE (2001)
between the mineral structure and composition of the dis- J Orthop Res 1035:19
10. Shimizu M, Tsuji H, Matsui H, Katoh Y, Sano A (1993) Clin
eased bone samples as compared with bone which from Orthop 229:293
clinical findings and histology is not diseased. This sug- 11. Mansell JP, Bailey AJ (1998) J Clin Invest 1596:101
gests that the osteoarthritis may have more influence on the 12. Finlay JB, Bourne RB, Kraemer WJ, Moroz TK, Rorabeck CH
bone organic matrix or its interaction with the mineral, than (1989) Clin Orthop 193:247
13. Li B, Aspden RM (1997) J Bone Miner Res 641:12
alteration in the mineral structure itself. These results will 14. Ding M, Dalstra M, Line F, Hvid I (1998) Acta Orthop Scand
no doubt add to the debate about how precisely OA affects 358:69
underlying bone structure and strength. 15. Ewers BJ, Newberry WN, Garcia JJ, Haut RC (1998) Trans
Orthop Res Soc 459:23
16. Li B, Aspden RM (1997) Ann Rheum Dis 247:56
17. Maudsley AA, Ernst RR (1977) Chem Phys Lett 50:368
References 18. Lee M, Goldburg WI (1965) Phys Rev A1261:140
19. Torchia DA (1984) J Magn Reson 613:30
1. Dequeker J, Mokassa L, Aerssens J, Boonene S (1997) Microsc 20. Kim DY, Taylor HW, Moore RM, Paulsen DB, Cho DY (2003)
Res Tech 358:37 Vet J 166:52
2. Grynpas MD, Alpert B, Katz I, Lieberman I, Pritzker KPH (1991) 21. Yagi R, McBurney D, Laverty D, Weiner S, Horton WE Jr (2005)
Calcif Tissue Int 20:49 J Orthop Res 23:1128
3. Panula HE, Nieminen J, Parkkinen JJ, Arnala I, Kroger H, Alhava 22. Cho G, Wu Y, Ackermann JL (2003) Science 613:300
E (1998) Acta Orthop Scand 627:69 23. Wilson EE, Awomusi A, Morris MD, Kohn DH, Tecklenburg
4. Radin EL, Martin BM, Burr DB, Caterson B, Boyd RD, Goodwin MMJ, Beck LW (2006) Biophys J 3722:90
J (1984) J Orthop Res 221:2 24. Wilson MA (1987) NMR techniques and applications in geo-
5. Burr DB (1998) Curr Opin Rheumatol 256:10 chemistry and soil chemistry. Pergamon Press, London
6. Burr DB, Schaffler MB (1997) Microsc Res Tech 37:343 25. Jager C, Welzel T, Meyer-Zaika W, Epple M (2006) Magn Reson
7. Dequeker J (1985) Clin Rheum Dis 271:11 Chem Chem 573:44
8. Matsui H, Shimizu M, Tsuhi H (1997) Microsc Res Tech 333:37
123
J Mater Sci (2007) 42:88118823
DOI 10.1007/s10853-007-1917-y
Received: 29 December 2006 / Accepted: 4 June 2007 / Published online: 14 July 2007
Springer Science+Business Media, LLC 2007
Abstract A new remodeling theory accounting for algorithm is applied to simulate the uniaxial loading of a
mechanically driven collagen fiber reorientation in car- cylindrical tendon and the complex multiaxial loading
diovascular tissues is proposed. The constitutive equations situation in a model artery. After investigating different
for the living tissues are motivated by phenomenologically material and spatial stress and strain measures as potential
based microstructural considerations on the collagen fiber driving forces, we conclude that the Cauchy stress, i.e., the
level. Homogenization from this molecular microscale to true stress acting on the deformed configuration, seems to
the macroscale of the cardiovascular tissue is performed be a reasonable candidate to drive the remodeling process.
via the concept of chain network models. In contrast to
purely invariant-based macroscopic approaches, the pres-
ent approach is thus governed by a limited set of physically
motivated material parameters. Its particular feature is the
underlying orthotropic unit cell which inherently incorpo- Introduction and motivation
rates transverse isotropy and standard isotropy as special
cases. To account for mechanically induced remodeling, Living tissues are able to adapt to physiological and path-
the unit cell dimensions are postulated to change gradually ophysiological stimuli in order to keep adequate perfusion
in response to mechanical loading. From an algorithmic according to the metabolic demand of the tissue. For
point of view, rather than updating vector-valued micro- example, changes in mechanical stimuli lead to altered
structural directions, as in previously suggested models, we cellular and extracellular activities, and typical observed
update the scalar-valued dimensions of this orthotropic unit biological responses are related to growth, remodeling,
cell with respect to the positive eigenvalues of a tensorial adaptation, and repair, i.e., mechanobiology (see, e.g.,
driving force. This update is straightforward, experiences Humphrey [33], Huang et al. [31], Ingber [34], Klein-Nu-
no singularities and leads to a stable and robust remodeling lend et al. [35], Wang and Thampatty [55], Holzapfel and
algorithm. Embedded in a finite element framework, the Ogden [28, 29], Mofrad and Kamm [47] or Lehoux et al.
[40]). Changes in the material (and structural) properties of,
for example, the artery wall through alterations in its
E. Kuhl (&)
Department of Mechanical Engineering, Stanford University, internal microstructure constitute an active process that
Stanford, CA 94305, USA occurs in response to changes of mechanical parameters, a
e-mail: ekuhl@stanford.edu process called arterial remodeling. It is the endothelium
cell that sense mechanical and humoral parameters, trans-
G. A. Holzapfel
Department of Solid Mechanics, School of Engineering duce signals to the underlying smooth muscle cells and to
Sciences, Royal Institute of Technology, Stockholm 100 44, the surrounding tissue, and relay mechanical and bio-
Sweden chemical changes into biomolecular events. Therefore, the
endothelium at the interface of the blood plays a crucial role
G. A. Holzapfel
Institute for Biomechanics, Center for Biomedical Engineering, in the initiation of arterial remodeling. In particular, in
Graz University of Technology, Graz 8010, Austria cardiovascular tissues, remodeling processes are important
123
8812 J Mater Sci (2007) 42:88118823
in the context of arterial development, atherosclerosis, and capture, predict and explain basic trends observed in col-
healing in response to arterial injury. A great deal of inter- lagen fiber remodeling and its impact on the structural
disciplinary research effort is devoted to the (mechanical) response at the tissue or organ level. There seems to be a
signaling pathways because they may enable the identification general agreement that the interplay between matrix stress,
of therapeutic targets and the development of new pharma- fibroblast alignment and stress in the actin network is
cological strategies. Moreover, understanding the interplay responsible for collagen fibril reorientation as reported by
between the architecture of the internal microstructure and the Stopak and Harris [50] and described in detail by Gariki-
mechanical loading is of fundamental importance to engineer, pati et al. [20]. However, we do not aim at explaining the
e.g., blood vessel substitutes, see Nerem and Seliktar [48]. origin of remodeling which is governed by many highly
The function and integrity of organs are maintained by complex interactive phenomena on the cellular level that
the tension in collagen fibers, which contribute signifi- involve altered gene expression in response to altered
cantly to the stability and strength of organs. Collagen loading (i.e., gene transcription, translation, protein syn-
fibers are typically considered as the main load bearing thesis, packing, and activation) which eventually results in
constituent of the extracellular matrix. Accordingly, altered rates of turnover of cells and matrix. Nor do we aim
changes in the material (and structural) properties can at following the classical continuum mechanics approach
primarily be attributed to variations in collagen content, and develop a purely invariant based macroscopic theory
type and thickness and, of course, in the orientation within governed by a number of abstract material parameters. Our
the tissue. Leung et al. [41, 42] were amongst the first to goal is to apply suitable homogenization techniques to
verify experimentally that mechanical forces relate to derive a sound phenomenologically and micromechani-
pressure and flow direct medial cell biosynthesis and cally based formulation with a limited number of param-
modulate structural adaptations to hemodynamic changes. eters that have a clear physical interpretation.
Based on in vitro studies of smooth muscle cells, they To this end, we begin our investigations on the
reported that aortic medial cells attached to elastic mem- microstructural or rather molecular level focusing on the
branes and subjected to cyclic stretching consistently syn- mechanical description of the individual collagen fibers.
thesized collagen of types I and III much more rapidly than The characteristic feature of typical collagen molecules
did cells growing on stationary membranes. In the present is their long, stiff, triple-stranded helical structure in
manuscript we use the word remodeling exclusively with which three collagen polypeptide chains are wound
respect to collagen fiber reorientation, while the type and around one another in rope-like superhelical structures
thickness of collagen as well as its content and its con- which are stabilized by numerous hydrogen bonds. The
centration are assumed to be constant. In addition, we do mechanical properties of these helical structures are un-
not address adaptation in the form of volumetric growth like those of any other natural or synthetic polymers.
which is addressed in detail elsewhere in the literature, see, Collagen helices display a remarkable stiffness which
e.g., Rodriguez et al. [49], Lubarda and Hoger [43] or Kuhl may be characterized appropriately through the so-called
et al. [38]. There is, however, strong evidence that growth wormlike chain model. The wormlike chain, or rather
and remodeling can indeed be viewed as separate indi- Kratky and Porod model [36], imagines the polymer as a
vidual processes. Stopak and Harris [50] studied the ori- rod that bends smoothly under thermal fluctuations.
entation of collagen fibrils due to the forces exerted on Traditionally applied to model the DNA double helix,
them by fibroblast in gels. Fiber reorientation was found to see Bustamante et al. [7, 8] and Marko and Siggia [44],
take place in response to changes in the mechanical loading the Kradky and Porod model was recently adopted to
although no significant growth, resorption, and production simulate the behavior of the collagenous triple helix by
of new fibers was reported. Motivated by these findings, Bischoff et al. [3, 4], Garikipati et al. [19, 20] and Kuhl
Garikipati et al. [20] provided a theoretical framework that et al. [37, 39].
focuses exclusively on collagen fiber remodeling and After the collagen fibrils have formed in the extra-
supported their theory by a set of remodeling experiments. cellular space, they are greatly strengthened by the for-
For a more sophisticated theoretical approach that captures mation of covalent crosslinks between lysine residues of
the interaction of the individual phenomena of growth and the constituent collagen molecules. If cross-linking is
remodeling, the reader is referred to Menzel [45]. inhibited, the tensile strength of the fibrils is drastically
To gain further insight into the complex biomechanical reduced, the collagenous tissue becomes fragile and the
phenomena related to tissue remodeling we aim at formu- structure tends to tear, see Alberts et al. [1]. To incor-
lating and implementing a novel constitutive framework porate these characteristic cross-linking network effects,
for collagen fiber remodeling with particular emphasis on different isotropic chain network models have been pro-
the arterial wall, in which type I collagen is the major posed in the past, see, e.g., Flory [17], Treloar [53] and
constituent. As such, the central focus of this study is to Arruda and Boyce [2, 5, 6]. In order to account for the
123
J Mater Sci (2007) 42:88118823 8813
anisotropic nature of cardiovascular tissues on the mes- principal stresses (or strains). Hence, as the stress (or
oscopic extracellular matrix level, we generalize the strain) state varies with the radial position, the orientations
cubic isotropic unit cell of the Arruda and Boyce model of the two collagen fiber families also vary across the
to obtain the orthotropic eight-chain model suggested thickness of the arterial wall, as reported by, e.g., Taber
recently by Bischoff et al. [3, 4]. and Humphrey [52] and Holzapfel et al. [30]. This varia-
Finally, it remains to incorporate the living nature of the tion across the wall thickness was successfully obtained
tissue and its ability to adapt its collagenous microstructure from the discrete collagen fiber reorientation models by
to the mechanical loading environment. Naturally, fiber Driessen et al. [13] and Hariton et al. [25]. In [13] it was
directions will evolve in vivo to optimize the load bearing assumed that the collagen fibers align along preferred
capacity while keeping the required compliance. Tradi- directions, situated in between the principal stretch direc-
tionally, remodeling theories in arteries can be classified tions, while in [25] the remodeling process was assumed to
into stress driven and strain driven approaches. The former be stress driven. Within the present manuscript, we com-
are typically based on the assumption that the cardiovas- bine these basic assumptions with the fundamental concept
cular tissue remodels its geometry to restore circumferen- of chain network models to obtain a three-dimensional
tial wall stress due to pressurization and wall shear stress remodeling theory which is general enough to predict
due to blood flow to normal levels, see, e.g., Taber and remodeling in complex multiaxial loading situations. In
Humphrey [52], Gleason and Humphrey [22] or Hariton contrast to existing theories, which strongly rely on com-
et al. [25]. Alternatively, motivated by successful predic- plex rotational updates, this new approach can be algo-
tions in hard tissue mechanics, the authors of the latter type rithmically realized in terms of remarkably simple and
of models suggest that strain rather than stress is the rele- straightforward updates of scalar-valued spatial dimen-
vant driving force for the remodeling process, see Kuhl sions.
et al. [37], Himpel et al. [26] or Driessen et al. [13]. Either The manuscript is organized as follows: the governing
of the two theories is able to identify characteristic equations of the micromechanically motivated remodeling
microstructural directions which are allowed to reorient theory are derived in Sect. Governing equations. Starting
with respect to the eigendirections of a mechanically rel- from the molecular level, we derive the constitutive
evant second-order tensor. In terms of algorithmic proce- equations for anisotropic soft biological tissues based on
dures, this vector reorientation typically leads to complex the concept of orthotropic chain network models. Section
rotational updates which usually involve singularities due Computational examples then focuses on two particular
to the trigonometric nature of the underlying update model problems, a cylindrical tendon subject to uniaxial
equation, see, e.g., Menzel [45, 46]. Although very elegant tension, and a tube-like artery subject to uniaxial stretch in
from a mathematical point of view and maybe well-suited combination with a distending pressure. Section Discus-
for microstructures with one predominant orientation, these sion closes with some final remarks.
reorientation models seem rather cumbersome in the con-
text of arterial walls where multiple fiber families need to
be accounted for. Governing equations
When aiming to develop reliable constitutive theories
for remodeling in cardiovascular tissues it is crucial to have In what follows, we summarize our set of constitutive
detailed insight in the structural arrangement of the colla- equations for anisotropic cardiovascular tissues. To this
gen fiber distribution. By using the birefringent properties end, we apply the following hypotheses:
of collagen, Finlay et al. [16] elaborated tangential sections
Hypothesis I: Large arteries seek to restore wall stress
of cerebral arterial walls to examine the integrated struc-
to within a range of homeostatic values.
tural order of the individual layers. Alternative techniques
Hypothesis II: Collagen fibers as the main load bearing
providing information about the collagen fiber distribution
constituent of the extracellular matrix adapt their
in arterial walls were discussed recently by Elbischger
orientation and align with respect to the principal stress
et al. [14, 15]. Along these lines, continuously distributed
directions in order to minimize wall stress.
collagen fiber orientations were incorporated in the more
Hypothesis III: Collagen fiber remodeling can be
recent models by Driessen et al. [11, 12], Freed et al. [18]
modeled and simulated phenomenologically to improve
and Gasser et al. [21].
the understanding of fiber orientation and provide
Experimental findings suggest that at biological equi-
further insight in the structural arrangement of the
librium two predominant fiber orientations can be identi-
individual arterial layers.
fied in each layer of the arterial wall. Typically, these two
discrete families of collagen fibers are found to be located It should be mentioned, however, that although the pres-
somewhere in between the directions of the two maximal ent model takes into account microstructural information,
123
8814 J Mater Sci (2007) 42:88118823
it is still based on a rather phenomenological approach in the constant and h is the absolute temperature. In the case of
sense that it does not explain the mechanisms how the indi- living tissues, we suggest h = 310 K, i.e., h = 37 C.
vidual cells actually sense changes in loading and commu- The two parameters governing the chain behavior are
nicate this information. Many different receptors on the the contour length L and the persistence length A, as
surface of endothelial cells and vascular smooth muscle cells illustrated in Fig. 1.
are able to detect subtle changes in the mechanical envi- The force required to pull the ends of the chain away
ronment. They initiate various different mechanotransduc- from each other by a distance r thus follows straightfor-
tion cascades according to the nature of the mechanical wardly by taking the derivative of the free energy wchn with
stimulus perceived. The cytoskeleton and other structural respect to the end-to-end length.
components play an important role in mechanotransduction " #
as they are able to transmit and modulate tension between chn dwchn 1 r 1
f kh 4 1 2
focal adhesion sites, integrins and the extracellular matrix. dr 4A L 1 r=L2
Moreover, changes in the mechanical environment may also
initiate changes in the ionic composition of the cells, medi- Note that due to the particular nature of the free energy
ated by ion channels, stimulate various membrane receptors chn
w , the end-to-end length r of a wormlike chain cannot
and induce complex biochemical responses, see, e.g., Huang extend beyond its contour length L as 0 < r < L.
et al. [31], Mofrad and Kamm [47] or Lehoux et al. [40] for
excellent overviews. Since we do not aim at simulating these Remark 1 [Parameters on the collagen fiber level] The
molecular mechanisms of mechanotransduction, all these wormlike chain model is essentially a two-parameter
phenomena are modeled phenomenologically through a set model governed by the contour length L and the persistence
of continuum based remodeling equations which we describe length A. Figure 1 illustrates the physical meaning of the
in the sequel. persistence length. It shows the force-displacement curves
We begin on the microstructural level with the con- of a single collagen fiber indicating the increase in initial
stitutive description of the individual collagen fibers. On stiffness with increasing persistence length for, say,
the mesolevel, we then elaborate a representative volume A = 0.1, A = 0.4, and A = 0.8. Note that throughout the
element representing the extracellular matrix. On the entire manuscript, all lengths of the model have been
macroscopic level, we finally characterize the overall rendered non-dimensional by dividing them by the link
tissue behavior through a micromechanically motivated length of the chain, as proposed by Garikipati et al. [19].
constitutive model which is able to account for micro-
structural adaptation in response to changes in the Remark 2 [Specific data for the persistence length]
mechanical loading. F-actin (a filamentous protein responsible for the con-
traction and relaxation of muscle) has a persistence
On the collagen fiber level length A of approximately 16 lm. For nanotubes A is in
and Siggia [44], and Bustamante et al. [7, 8], and recently 8
applied to collagen fibers by Bischoff et al. [3, 4], Gari- 7
f(r/L) times 1/(k)
kipati et al. [19, 20] and Kuhl et al. [37, 39]. In the sta-
6
tistical mechanics of long chain molecules such as collagen
fibrils, the key kinematic variable that characterizes the 5
conformation of the chain is the end-to-end length r. 4
According to the wormlike chain model, the free energy
3
wchn of a single collagen fiber can be expressed in terms of
the end-to-end length in the following form. 2
2 1
chn L r 1 r
w wchn
0 kh 2 2 1 0
0 0.2 0.4 0.6 0.8 1
4A L 1 r=L L
r/L
Herein, wchn
0 is the value of the chain energy in the Fig. 1 Collagen fiber level s Single chain force vs. chain stretch for
unperturbed state, k = 1.381 1023 J/K is the Boltzmann varying persistence lengths A
123
J Mater Sci (2007) 42:88118823 8815
the millimeter range. Note, however, that A for DNA The overall chain energy Wchn follows by summing up
in vivo has a value of ~50 nm (Hagerman [24]), and A the contributions wchn of eight individual chains weightened
for synthetic polymers is typically only a few nanome- by the overall chain number density nchn, i.e., the number of
ters. Hence, the persistence lengths A L of a typical chains per unit volume. According to the original eight-
DNA molecule and a synthetic molecule are considerably chain model by Arruda and Boyce, each of these chains
smaller than their contour lengths L. Recent studies connect the corners of a regular cuboid of dimensions 2 l1,
performed by means of optical tweezers seem to indicate 2 l2, and 2 l3 with its center, compare Fig. 2, left. The end-
that under physiological conditions collagen I molecules to-end length r0 in the undeformedpconfiguration, thus,
p
have a persistence length of ~14.5 nm which would be follows straightforwardly as r0 l2I l21 l22 l23 :
less than 5% of their contour length of ~309 nm, see Sun The unit cell is postulated to deform in the principal stretch
et al. [51]. Accordingly, collagen would be much more space. Accordingly, the end-to-end length r in the deformed
flexible than previously assumed, yet even more flexible configuration can be expressed in terms of the deformation
than DNA. Although we suggest to stick to the wormlike gradient F or rather in terms of the right Cauchy-Green
chain approach in the sequel, this is not a general limi- tensor C,
tation of the overall constitutive model as such. Due to q q q
the modular structure of the overall framework, the free r l2I n
I n I l2I n0I C n0I l2I IIC 5
energy function for the individual fibers (1) can easily be
modified, adapted and integrated straightforwardly in the with explicit summation over all three direction I = 1,2,3.
macroscopic model. Here, we have introduced the non-standard invariants
IIC n I n0I C n0I with the understanding that IIC
I n
On the extracellular matrix level represents the stretch in the n0I direction squared. Thereby,
n0I are the unit normal vectors of the unit cell axes in the
From a mechanical point of view, the extracellular matrix undeformed reference configuration, see figure 3. After the
is modeled as a surrounding substrate in which the indi- deformation, they map onto the vectors n I F n0I which
vidual collagen fibers are embedded. A representative are obviously no longer of unit length.
volume element of the extracellular matrix thus consists of Finally, the repulsive contribution W rep lnIC IC IC
1 2 3
a substrate of elastin, proteoglycans and cell, characterized is constructed to compensate for the chain stresses in the
through the isotropic free energy wiso, and an anisotropic reference configuration caused by non-vanishing initial
contribution due to the individual chains wchn. Moreover,
we introduce a repulsive chain contribution wrep to char-
acterize the tissues behavior of the initial configuration
such that the total free energy may be written in the fol-
lowing form.
n03 =nIII l2 +
II
For the isotropic term Wiso, we apply a standard
F, u l3 +
III
neo-Hookean model expressed in terms of the first
invariant I1C C : I of the right Cauchy-Green tensor
Fig. 3 Remodeling based on changes of cell dimensions s
C Ft F;where F rX u denotes the deformation Instantaneous alignment of cell axes n0I with eigenvectors nrI and
gradient and J detF > 0 is its determinant. Moreover, k gradual adaptation of cell dimensions lIwith respect to eigenvalues
and l are the standard Lame constants. kr
I
123
8816 J Mater Sci (2007) 42:88118823
end-to-end lengths r0. Classically, the free energy W assume the chain number density to be constant, however, its
introduced in (3) defines the Cauchy stress r in the fol- evolution in time due to changes in collagen content and
lowing form, thickness could be incorporated straightforwardly. The
degree of anisotropy is governed by the unit cell dimensions
1 dW t l1, l2,p
and
r F riso rchn rrep 6 l3 which implicitly define the end-to-end length
J dF r0 l2I in the initial configuration.
whereby the individual stress contributions can be
expressed as follows.
On the tissue level
1
riso k lnJ lI lb In cardiovascular tissues, the collagen fibers are not arbi-
J " #
chn trarily distributed in space but follow rather a particular
n 1 1 1
rchn kh chn
r 7
pattern. To account for remodeling in the form of micro-
4AJ L 4r1 r=L2 4r structural rearrangement, we allow the fiber direction to
" #
nchn 1 1 1 rotate in response to the current mechanical stress envi-
rep
r kh rep
r ronment. We assume the Cauchy stress tensor
4AJ L 4r0 1 r0 =L2 4r0
r krI nrI nrI 8
Herein, the Finger deformation tensor b F Ft repre-
sents a characteristic spatial strain measure. The tensorial to be the driving force of the remodeling process. Here, we
basis of the chain stress follows straightforwardly as have introduced its eigenvalue decomposition with krI and
chn l2I n
r I :Note that the repulsive energy Wrep has
I n nI being the corresponding eigenvalues and eigenvectors,
been constructed such that the corresponding tensorial respectively. For notational simplicity, we have implicitly
rep l2I =IIC n
basis of the repulsive stress r I n
I ensures a assumed the summation over all I = 1,2,3 components. Our
stress free reference configuration as rrep rchn jrr0 for remodeling theory is based on two fundamental hypothe-
which IIC 1 for all I = 1,2,3. ses:
Remarks 3 [Special cases of transverse isotropy and Hypothesis I: The characteristic directions n0I of the
isotropy] The orthotropic eight-chain model, as introduced microstructure align instantaneously with respect to the
by Bischoff et al. [3, 4], can be understood as a kinematic eigenvectors nrI :
generalization of the original isotropic eight-chain model, Hypothesis II: The unit cell dimensions lI adapt
as indicated in Fig. 2. gradually with respect to the positive eigenvalues kr
I :
123
J Mater Sci (2007) 42:88118823 8817
r
kI l0I dt n01 x n01
dt lI j expjtr0 10
jjkr
I jj r0
which is obviously more cumbersome due to numerical
where j denotes a relaxation parameter and l0I ; I = 1,2,3, singularities introduced through trigonometric functions.
are the dimensions of the cuboid in the undeformed In contrast to the rather simple explicit equation for
configuration. Alternatively, the above equation could be the adaptation of the unit cell dimensions (11), the
integrated in time to render an explicit update equation for fiber direction can be expressed through the exponential
all cell dimensions lI. update
1
kr l0I n0k1
3
expDt e x n0k x n0k1 nrmax
lI I
1 expjtr0 l0I 11 1 1
j 1
jjkr
I jj r 0
3
where x and e denote the time discrete rotation vector and
Recall that the reorientation
p process itself leaves the
the third-order permutation tensor, respectively. Recall that
initial chain length r0 l2I unaltered.
in addition to potential numerical difficulties, this previous
reorientation approach is restricted to transversely isotropic
Remarks 5 [Remodeling based on fiber rotation] Previous
biological tissues and its extension to reorienting multiple
remodeling approaches in the literature focused on the
directions seems to be a rather complex task.
reorientation of a single microstructural direction, see, e.g.,
Menzel [46], Kuhl et al. [37] and Himpel et al. [26].
Remarks 6 [Dissipation] It is a well-accepted fact that a
Applications thus focused primarily on transversely
purely mechanical theory is thermodynamically inadmis-
isotropic biological tissues with one distinguished fiber
sible for remodeling processes that stiffen the material, see,
orientation such as muscles, tendons or ligaments for which
e.g., the discussions in Menzel [46], Kuhl et al. [37], and
l2 = l3 l1. Similar to the previous approach, reorientation
Himpel et al. [26] or Garikipati et al. [20]. The fact that the
was assumed to be driven by either strain or stress, based on
dissipation would be positive for stiffening materials in a
an eigenvalue decomposition, as given in (8). In contrast to
purely mechanical theory indicates that other thermody-
the new model suggested in this contribution, the following
namic phenomena, e.g., of chemo-mechanical nature,
assumptions were postulated for the remodeling process:
should indeed be taken into account. Alternatively, mixture
The characteristic direction n01 of the microstructure theories could be considered in which energy and entropy
aligns gradually with respect to the eigenvector nrmax are exchanged amongst the individual constituents of the
related to the maximum positive eigenvalue kr
max : tissue, see, e.g., Gleason and Humphrey [22, 23].
The unit cell dimensions lI remain unaffected by the Accordingly, to date, there is no general agreement of how
process of remodeling. evolution laws for reorientation of microstructural direc-
tions should be formulated. In the context of linear elas-
Consequently, the reorientation-based analogue of Eq.
ticity, it has been shown that the free energy attains an
(9) could be expressed as follows, compare Fig. 4.
extremum if strain and stress share the same principal
n01 ! nrmax and lI const. directions, see, e.g., Cowin [10] or Vianello [54]. In non-
linear elasticity, however, it is not even clear to date
Rather than updating the scalar-valued unit cell lengths whether stress or strain is the relevant driving force for the
lI, this reorientation approach requires an update of the remodeling process. The previous n01 ! nrmax non-linear
vector-valued fiber direction n0I reorientation model addressed in Remark 5 is based on the
general paradigm that nature always tries to find the
extremum, see Menzel [46], Kuhl et al. [37] and Himpel
et al. [26]. However, the recent reorientation model for
arteries by Hariton et al. [25] just postulated a general
stress driven remodeling process and so does the model
presented herein. For detailed discussions about the impact
of the dissipation inequality in remodeling, the reader is
referred to the excellent discussion by Garikipati et al.
[20].
Fig. 4 Remodeling based on changes of cell orientation s Gradual
alignment of the cell axes n01 with eigenvector nrmax of related Remarks 7 [Parameters on the tissue level] An additional
maximum principal eigenvalue kr
max and constant cell dimensions
lI = const convincing benefit of the incorporation of remodeling is
123
8818 J Mater Sci (2007) 42:88118823
40
123
J Mater Sci (2007) 42:88118823 8819
capture these general trends is 10 units long and has an combination with an internal pressure representing the blood
initial cross-section of 10 units2. It is discretized with 10 flow. It has long been recognized that the layered structure of
elements in height and 48 elements per cross-section, thus cardiovascular tissues displays a geometric complexity way
with 480 tri-linear finite elements in total. The load is beyond the highly organized parallel bundles of collagen
increased incrementally with DF = 25 from 0 < t 12 until fibers that constitute the mechanical backbone of tendons.
a final load level of F = 300 is reached. From 12 < t 150, Motivated by earlier theoretical studies of Driessen et al.
we apply another 138 remodeling steps until biological [13] and Hariton et al. [25], and supported by the experi-
equilibrium occurs. mental findings of Finlay et al. [16], we analyze the
Figure 6 illustrates the remodeling history of the cylin- remodeling history in a tube-like artery. At this point, we
drical tendon. At first, when the load is applied, the tendon restrict ourselves to a rather qualitative analysis aiming at
is stretched in the axial direction up to about 175% of its including physiologically realistic data at later stages of this
original length, compare with Fig. 6, left. However, as time project. The initial dimensions are given through a length of
evolves, the collagen fibers tend to align gradually with the 8 units and an inner and outer radius of 1 and 3 units,
loading axis. Accordingly, the structure stiffens signifi- respectively. Along the height we apply 12 elements, we use
cantly and the overall stretch decreases to about 140%. At 8 elements across the thickness and 16 elements in circum-
the final and yet stiffest state, all fibers are aligned with the ferential direction. Here, we apply standard tri-linear finite
direction of the mechanical load. elements. Note, however, that shell elements would be more
During the remodeling process, the unit cell lengths appropriate when it comes to quantitative studies. The load is
p evolved from initially random values 0 lI
have obviously applied incrementally in 25 load steps from 0 < t25 with
r0 with l2I r0 to l1 = r0 and l2 = l3 = 0 representing Dp = 0.192 p0 and Dl = 0.032 until a final state of p = 4.8 p0
the special case of transverse isotropy. Selected snapshots and Dl = 0.8, i.e., an axial stretch of 10%, is reached. Again,
of the remodeling history are depicted in Fig. 6, right. For the load is then held constant for another 125 time steps for
each element, the end-to-end vectors of each chain have the time period 25 < t150 to allow for remodeling towards
been projected on the tendons surface. The individual a final state of biological equilibrium. Figure 7 shows five
colors represent the collagen fiber angle measured against selected snapshots of the remodeling process. In between the
the loading axis. Red colors indicate a full alignment with an first and the second snapshot, the internal pressure and the
angle zero, while blue colors indicate that the collagen fibers prescribed axial stretch are increased incrementally.
are oriented orthogonal to the loading axis. In particular, in Accordingly, the tube blows up and stretches along its axial
the second snapshot of the series at t = 12 corresponding to direction occur. Due to the nearly incompressible behavior of
the final loading state, the spatial inhomogeneity is nicely soft biological tissues, the tube thickness decreases drasti-
visible. The red elements with aligned collagen fibers display cally in response to loading. As time evolves, the collagen
very stiff response and deform only marginally while the soft fibers tend to reorient so that they are finally located between
blue elements undergo significant stretches. As expected, the two directions of maximum principal stress. For this
this initial inhomogeneity tends to vanish gradually in the particular case of loading, these are obviously the longitu-
course of remodeling. dinal and the circumferential directions.
Starting from an initial random orientation
p at random
Tube-like artery subject to uniaxial stretch and internal unit cell lengths 0 lI r0 with l2I r0 in the first
pressure snapshot of the series, the unit cell lengths evolve pro-
gressively driven by the positive eigenvalues of the Cauchy
We turn now to the more challenging example of a multi- stress kr
I : As a natural consequence, the chain directions
axial loading state induced by a uniaxial stretching in remodel gradually and the tube stiffens with respect to the
123
8820 J Mater Sci (2007) 42:88118823
applied loading. Accordingly, the outer radius, which had The predicted inhomogeneity in the radial direction with
increased during the first stages of loading, decreases an almost circumferential fiber orientation at the luminal
remarkably in response to remodeling, compare time t = 24 side of the artery and an increase of the inclination towards
with t = 150. While the chains had originally been oriented the outer side is closely related to the phenomenon of
randomly in space at t = 1, their component in radial prestress, typically encountered in arterial specimens. For
direction vanishes gradually as l3 goes to zero in response the considered model problem, the degree of inhomoge-
to the compressive stresses kr3 0 in the radial direction at neity, obviously, strongly depends on the stretch-to-pres-
t = 150. Thus, the suggested remodeling algorithm proves sure ratio. Motivated by a recent study by Gleason and
able to produce not only transversely isotropic but also Humphrey [23] who analyzed the elastin, collagen and
orthotropic microstructures in a natural way. smooth muscle cell turnover and remodeling in response to
Recall that the final collagen fiber angles developed different loading scenarios, we systematically elaborate
naturally in response to the positive eigenvalues krI of the combinations of transmural pressure and axial stress. To
Cauchy stress r which had been postulated to be the illustrate the influence of the mechanical loading situation
driving force of the remodeling process. Due to the inho- on the remodeling process, we display Fig. 9 to show the
mogeneous stress state across the radial direction, different final biological equilibrium stages for different stretches at
fiber angles arise in the individual layers. different pressure levels. For the sake of visibility, we have
Figure 8 depicts the final collagen fiber orientations at virtually peeled off the individual layers after the calcula-
three different radial locations representing the intima, the tion of the entire tube had been performed. Experimentally,
media and the adventitia. The results suggest that the the definition of these specific layers could be made
transmural pitch of the fiber orientation increases from the apparent by the use of stains differentiating smooth mus-
inner to the outer wall. The analytically predicted double- cles cells from collagen fibers. For pure stretching, as
helix architecture of the collagen fibers agrees very well depicted in Fig. 9 (left), the fibers of all layers are aligned
with experimental observations by Finlay et al. [16] and with the loading axis, i.e., l1 = r0 and l2 = l3 = 0 with
Holzapfel et al. [27, 30]. In addition, our results nicely n01 pointing in the axial direction. For pure pressure,
agree with the more recent numerical studies by Driessen as depicted in Fig. 9 (right), all fibers are oriented in
et al. [13] and Hariton et al. [25]. the circumferential direction, i.e., again, l1 = r0 and
l2 = l3 = 0, however, now with n01 pointing in the circum-
ferential direction. These computational results agree well
with the experimental findings of human brain arteries by
Finlay et al. [16] who reported a significant realignment of
collagen with increasing distending pressure for all arterial
layers and a recruitment towards the circumferential
direction. Intermediate stages of combined stretch and
pressure loading, as shown in the three images between the
left and the right image of Fig. 9, reveal the usually
observed helical fiber arrangement. Based on these results,
it is obvious to suggest that the collagenous architecture
strongly depends on the mechanical environment.
In the biomechanics community there is an ongoing
debate over which quantity drives the remodeling process.
Fig. 8 Tube-like artery subject to axial stretch and internal pressure
s Final stage of the remodeling process with collagen fiber To this end, we analyze the influence of different driving
orientations projected on the individual layers forces of tensorial nature. For purpose of comparison with
123
J Mater Sci (2007) 42:88118823 8821
the (spatial) Cauchy stress r used in the previous examples, Finally, we elaborate the difference of a stress and
we now elaborate the use of the (material) second Piola strain-based remodeling criterion, as used by, e.g., Driessen
Kirchhoff stress S JF1 r Ft ; as the representative et al. [13], Menzel [46] and Humphrey [32], Hariton et al.
material stress measure. Figure 10 compares now the out- [25], respectively. To this end, we compare the outcome of
come of a remodeling process driven by spatial and two different remodeling processes based on either the
material stresses. The diagram illustrates the fiber orien- Cauchy stress r as the characteristic spatial stress measure,
tation angle over the radius. Variations in fiber orientation see Fig. 11 (left), or on the Finger deformation tensor b as
seem to be slightly more pronounced when using the sec- the characteristic spatial strain measure, see Fig. 11 (right).
ond PiolaKirchhoff stress as the driving force. Neverthe- Both the curves as well as the plots of the structure dem-
less, for this particular problem, it appears that differences onstrate that strain-based remodeling is much less sensitive
are rather marginal. We suggest that the Cauchy stress, to the radial variation. The difference between an almost
which is the true stress experienced by the deformed circumferential arrangement at the luminal side and a
structure, is the more reasonable choice. rather helical structure towards the outermost layer is less
10
0
2 2.2 2.4 2.6 2.8 3 3.2 3.4
radius
123
8822 J Mater Sci (2007) 42:88118823
pronounced for the strain driven case. This finding is, of these more complex inhomogeneous (living) structures, the
course, rooted in the non-linear nature of the force- remodeling algorithm was embedded in a non-linear finite
displacement behavior of the individual chains, recall element program, the equations were linearized consis-
Fig. 1. While the strain varies slowly close to the locking tently and solved with an incremental iterative Newton
stretch, corresponding stresses may vary significantly. Raphson solution strategy. Starting with initially random
Accordingly, the stress-based remodeling criterion predicts fiber orientation, the algorithm generated collagenous
a wider variation in collagen fiber angles than the strain architectures which qualitatively resembled experimental
driven remodeling algorithm. Although for hard tissues, observations. In all computational examples, the suggested
such as bone, several studies including the one by Cowin algorithm convinced through its remarkable stability and
[9] suggest that strain is the critical mechanical factor in robustness. What remains, however, is the realization of
growth and adaptation, we postulate that stress is a more quantitative validations of the suggested theory and its
reasonable choice as the driving force in the context of soft algorithmic realization. In addition to the collagen fiber
tissues such as arteries, tendons or ligaments. This obser- orientation, which could eventually be determined through
vation is along with Taber and Humphrey [52] who suggest techniques such as microscopy, histology, magnetic reso-
that stress and not strain correlates well with growth in nance diffusion tensor imaging, the chain number density
arteries. and the turnover rate remain to be classified. The latter is
assumed to be of particular relevance in the context of
improving cardiovascular surgery and predicting patient-
Discussion specific remodeling in response to medical treatment.
Finally we would like to point out once again, that
A continuum theory of remodeling for soft biological tis- although the reorientation approach outlined in the present
sues has been proposed with a particular focus on cardio- manuscript is motivated by micromechanical consider-
vascular tissue. In this context, remodeling is attributed to ations, it is still rather phenomenological to most extend. In
the collagen fiber reorientation as a natural consequence of contrast to purely invariant based formulations, it intro-
changes in the mechanical loading environment. Hence, in duces material parameters such as the contour length or the
our approach the optimal configuration of collagen fibers persistence length which have a clear physical interpreta-
depends solely on the external loading acting upon the tion. The computational mechanics community would thus
living tissue. Based on a homogenization strategy from the certainly classify the model as bottom-up or rather micro-
molecular microscale via the mesoscale of the extracellular mechanically based and non-phenomenological. In the cell
matrix to the macroscale of the overall tissue, we derived a biology community, however, the same model would most
statistical mechanics-based chain network model governed probably be considered rather phenomenological or top-
by a limited set of physically motivated material parame- down since it does not explicitly address cell level
ters. In addition to the two (classical) Lame constants, five phenomena like mechanotransduction or the biochemical
additional parameters sufficed to characterize the highly origin of the remodeling process as such. Nevertheless, to
non-linear, exponentially stiffening of the collagen mor- date it seems unmanageable to simulate large tissue
phology: the contour length, the persistence length and the structures or organs and yet at the same time to account for
initial end-to-end length of the collagen chains, the chain every single biochemical phenomenon individually. To the
number density accounting for the degree of anisotropy and most extend, these complex mechanisms are not even fully
the turnover rate. The concept of chain network models understood at this point.
was applied to transmit information from the molecular Successful constitutive models require an understanding
level to the extracellular matrix level. In contrast to exist- of the functional interactions between the key components
ing remodeling theories which are based on complex of cells up to the organs, and how these interactions change
rotational updates of characteristic microstructural direc- from a physiological to a pathological state. Such infor-
tions, our theory essentially captures remodeling in the mation resides neither in the individual genome nor in the
form of changes of the dimensions of this representative protein. This information is contained in the interactions of
element. By means of simple model problems, we were proteins with cellular, organ and system structures. The
able to show that the theory indeed succeeded in charac- identification of these interactive relationships is clearly
terizing mechanically introduced remodeling of collagen within the focus of intense current research. Continuum
fibers. models like the one presented herein will certainly benefit
Finally, we elaborated two biomechanically relevant from gradually incorporating more and more information
boundary-value problems, a cylindrical tendon subject to from the protein, subcellular and cellular level in the future
uniaxial tension and a tube-like artery loaded by an axial to define precisely how mechanical forces translate into
stretch in combination with different internal pressures. For chemical signals that initiate the process of remodeling.
123
J Mater Sci (2007) 42:88118823 8823
References 27. Holzapfel GA, Gasser TC, Stadler M (2002) European J Mech A
Solids 21:441
1. Alberts B, Johnson A, Lewis J, Raff M, Roberts K, Walter P 28. Holzapfel GA, Ogden RW (2003) Biomechanics of soft tissue in
(2002) Molecular biology of the cell, 4th edn. Garland Science cardiovascular systems CISM courses and lectures no 441.
Taylor Francis Group, New York Springer Verlag, WienNew York
2. Arruda EM, Boyce MC (1993) J Mech Phys Solids 41:389 29. Holzapfel GA, Ogden RW (2006) mechanics of biological tissue.
3. Bischoff JE, Arruda EM, Grosh K (2002) J Appl Mech 69:570 Springer, BerlinHeidelbergNew York
4. Bischoff JE, Arruda EM, Grosh K (2002) J Appl Mech 69:198 30. Holzapfel GA, Stadler M, Schulze-Bauer CAJ (2002) Ann Bio-
5. Boyce MC (1996) Rubber Chem Technol 69:781 med Eng 30:753
6. Boyce MC, Arruda EM (2000) Rubber Chem Technol 73:504 31. Huang H, Kamm R, Lee RT (2003) Am J Physiol Cell Physiol
7. Bustamante C, Bryant Z, Smith SB (2003) Nature 421:423 287:C1
8. Bustamante C, Smith S, Marko JF, Siggia ED (1994) Science 32. Humphrey JD (2001) J Biomech Eng 123:638
265:1599 33. Humphrey JD (2002) Cardiovasular solid mechanics. Springer
9. Cowin SC (1984) Calc Tissue Int 36:S99 Verlag, BerlinHeidelbergNew York
10. Cowin SC (1995) J Elast 34:45 34. Ingber DE (2003) Ann Intern Med 35:564
11. Driessen NJB, Bouten CVC, Baaijens FPT (2005) J Biomech Eng 35. Klein-Nulend J, Bacabac RG, Mullender MG (2005) Pathol Biol
127:494 (Paris) 53:576
12. Driessen NJB, Peters GWM, Huyghe JM, Bouten CVC, Baaijens 36. Kratky O, Porod G (1949) Recl Trav Chim 68:1106
FPT (2003) J Biomech 36:1151 37. Kuhl E, Garikipati K, Arruda EM, Grosh K (2005) J Mech Phys
13. Driessen NJB, Wilson W, Bouten CVC, Baaijens FPT (2004) J Solids 53:1552
Theor Biol 36:53 38. Kuhl E, Maas R, Himpel G, Menzel A (2006) Biomechanics and
14. Elbischger PJ, Bischof H, Holzapfel GA, Regitnig P (2005) In: modeling in mechanobiology, pp available online, DOI 10.1007/
Suri JS, Yuan C, Wilson DL, Laxminarayan S (eds) Plaque s10237-006-0062-x
imaging: pixel to molecular level, vol 113. Studies in Health 39. Kuhl E, Menzel A, Garikipati K (2006) Phil Mag 86:3241
Technology and Informatics, IOS Press, p 97 40. Lehoux S, Castier Y, Tedgui A (2006) J Intern Med 259:381
15. Elbischger PJ, Bischof H, Regitnig P, Holzapfel GA (2004) 41. Leung DYM, Glagov S, Mathews MB (1976) Science 191:475
Pattern Anal Appl 7:269 42. Leung DYM, Glagov S, Mathews MB (1977) Circ Res 41:316
16. Finlay HM, Mc Cullough L, Canham PB (1995) J Vasc Res 43. Lubarda VA, Hoger A (2002) Int J Solids Struct 39:4627
32:301 44. Marko JF, Siggia ED (1995) Macromolecules 28:8759
17. Flory PJ (1969) Statistical mechanics of chain molecules. John 45. Menzel A (2006) In: Holzapfel GA, Ogden RW (eds) IUTAM
Wiley Sons, ChichesterNew York Symposium mechanics of biological tissue. Springer Verlag, p 91
18. Freed AD, Einstein DR, Vesely I (2005) Biomech Model 46. Menzel A (2005) Biomech Model Mechanobiol 3:147
Mechanobiol 4:100 47. Mofrad MRK, Kamm R (eds) (2006) Cytoskeletal mechanics
19. Garikipati K, Arruda EM, Grosh K, Narayanan H, Calve S (2004) models and measurements. Cambridge University Press
J Mech Phys Solids 52:1595 48. Nerem RM, Seliktar D (2001) Annu Rev Biomed Eng 3:225
20. Garikipati K, Olberding JE, Narayanan H, Arruda EM, Grosh K, 49. Rodriguez EK, Hoger A, Mc Culloch AD (1994) J Biomech
Calve S (2006) J Mech Phys Solids 54:1493 27:455
21. Gasser TC, Ogden RW, Holzapfel GA (2006) J R Soc Interface 50. Stopak D, Harris AK (1982) Dev Biol 90:383
3:15 51. Sun YL, Luo ZP, Fertala A, An KN (2002) Biochem Biophys Res
22. Gleason RL, Humphrey JD (2004) J Vasc Res 41:352 Commun 295:382
23. Gleason RL, Humphrey JD (2005) Math Med Biol 22:347 52. Taber LA, Humphrey JD (2001) J Biomech Eng 123:528
24. Hagerman PJ (1988) Annu Rev Biophys Biophys Chem 17:265 53. Treloar LRG (1975) The physics of rubber elasticity. Clarendon
25. Hariton I, de Botton G, Gasser TC, Holzapfel GA (2006) Bio- Press, Oxford
mechanics and modeling in mechanobiology, pp available online, 54. Vianello M (1996) J Elast 42:283
DOI 10.1007/s10237-006-0049-7 55. Wang JH, Thampatty BP (2006) Biomech Model Mechanobio 5:1
26. Himpel G, Menzel A, Kuhl E, Steinmann P (2007) Int J Num
Meth Eng, in press
123
J Mater Sci (2007) 42:88248837
DOI 10.1007/s10853-007-1859-4
Received: 20 February 2007 / Accepted: 17 May 2007 / Published online: 7 August 2007
Springer Science+Business Media, LLC 2007
Abstract Interfaces are often believed to play a role in access to effective stresses representing the stress hetero-
the mechanical behavior of mineralized biological and geneities in the microstructures, which induce brittle fail-
biomimetic materials. This motivates the micromechanical ure. Thereby, Coulomb-type brittle failure of the crystalline
description of the elasticity and brittle failure of interfaces interfaces implies DruckerPrager-type (brittle, elastic
between crystals in a (dense) polycrystal, which serves as limit-type) failure properties at the scale of the polycrystal.
the skeleton of a porous material defined one observation At the even higher scale of the porous material, high
scale above. Equilibrium and compatibility conditions, interfacial rigidities or low interfacial friction angles may
together with a suitable matrix-inclusion problem with a result in closed elastic domains, indicating material failure
compliant interface, yield the homogenized elastic prop- even under hydrostatic pressure. This micromechanics
erties of the polycrystal, and of the porous material with model can satisfactorily reproduce the experimental
polycrystalline solid phase. Incompressibility of single strength data of different (brittle) hydroxyapatite bioma-
crystals guarantees finite shear stiffness of the polycrystal, terials, across largely variable porosities. Thereby, the
even for vanishing interface stiffness, while increasing the brittle failure criteria can be well approximated by micro-
latter generally leads to an increase of polycrystal shear mechanically derived criteria referring to ductile solid
stiffness. Corresponding elastic energy expressions give matrices, both criteria being even identical if the solid
matrix is incompressible.
123
J Mater Sci (2007) 42:88248837 8825
123
8826 J Mater Sci (2007) 42:88248837
123
J Mater Sci (2007) 42:88248837 8827
Z !
Table 1 continued 1 XZ s
Epoly exdV n ndS
Rpoly Second-order macroscopic stress tensor (related to RVE Vpoly Vpoly ij I ij
Vpoly of polycrystal with weak interfaces) Z X fi Z
1 X s
n ndS
s
n ndS
Rpoly;m Macroscopic mean stress (related to RVE Vpoly of 2
polycrystal with weak interfaces) Vpoly i oVi i
V i oVi
Rpoly;d Macroscopic equivalent deviatoric stress (related to
RVE Vpoly of polycrystal with weak interfaces) with location vector x; normal n onto the spherical surface
RPORO Second-order macroscopic stress tensor (related to RVE of the crystals,
VPORO of porous material the solid phase of which is a
polycrystal with weak interfaces)
n ni nj =2 nj n=2 ni n=2 3
RPORO;m Macroscopic mean stress (related to RVE VPORO of
porous material the solid phase of which is a as the mean displacement at the interface I ij ; Vi and
polycrystal with weak interfaces)
fi Vi =Vpoly as the volume and the volume fraction of the
RPORO;d Macroscopic equivalent deviatoric stress (related to RVE
VPORO of porous material the solid phase of which is a ith crystal, and Vi as its surface with area Ai. For crystals
polycrystal with weak interfaces) of the same shape and size (with volume VC and surface
r Second-order stress tensor field with in single crystals VC), and indiscernible average mean displacements at
filling RVE Vpoly of polycrystal with compliant their surfaces, (2) can be transformed to
interfaces
rin Stress field inside the inclusion surrounded by compliant Z
1 s
n ndS
interface and infinite matrix (related to generalized Epoly 4
matrix-inclusion problem) VC oVC
rex Stress field throughout the matrix surrounding inclusion
coated by compliant interface (related to generalized
The corresponding macroscopic stresses Rpoly are equal
matrix-inclusion problem) to the spatial average of the (equilibrated) local stresses
r Scattering factor in two-membered evolution strategy rx inside the RVE Vpoly,
/ Longitudinal coordinate of spherical coordinate system Z
u Volume fraction of pores within the RVE VPORO 1
Rpoly hrxi rxdV
v lC =kC Dimensionless quantity related to compressibility of Vpoly Vpoly
single crystals X fi Z
W Macroscopic energy density rxdV
i
Vi V i
1 Second-order identity tensor
X fi Z
First-order tensor contraction x rx nxdS 5
: Second-order tensor contraction i
Vi oVi
Dyadic product of tensors
For spherical crystals with radius a, surface VC with
area AC 4pa2 , and volume VC 4=3pa3 , (5) can be
Micromechanics of polycrystal with weak interfaces further transformed,
X fi Z
Micromechanical representation Rpoly aer x rx er xdS
4 3
i 3 pa oVC
with x as the position vector within the RVE. The with radial unit vector er being identical to the normal n.
geometrical compatibility of (1) with the local Since the microscopic stresses are equilibrated (div r 0),
microscopic strains ex in the crystals and the (5) and (6) imply [9, p. 118, 28], that the macroscopic
displacement discontinuities n nj ni at the stresses act as traction forces Rpoly n both at the boundary
interfaces I ij between the crystals i and j implies [14] of the RVE, oVpoly ; and those of single crystals, oVC ;
123
8828 J Mater Sci (2007) 42:88248837
on oVpoly and oVC : rx nx Rpoly nx 7 homogenization. This was recently shown quantitatively
for polycrystals consisting of perfectly disordered needles,
The relation between Rpoly and Epoly depends on the being either isotropic or anisotropic [18].
constitutive behavior of the single crystals and of the
interfaces between them. Homogenized elasticity of polycrystal with compliant
interfaces
Constitutive behavior of interfaces and single crystals
As long as the interfaces behave elastically, the relation
The interfaces are the weakest locations of the material, the between Rpoly and Epoly reads as
load bearing capacities of which are bounded according to
a Coulomb-type law, Rpoly Cpoly : Epoly 11
Tx K nx r\a : r CC : e
with 9 ra : T K 0 n
K Kn n n Kt 1 n n; Kn ! 1 with n n=2; K 0 2K 12
r>a : r Cpoly : e
K is the second-order interface stiffness tensor with infinite
r!1 : n ! E0 x
normal component Kn (no mutual interpenetration of
crystals), and positive tangential component Kt (allowing
For determination of kpoly, a purely spherical deforma-
for relative tangential movements of crystal surfaces). Also
tion, E0 E0 1 is imposed at r ! 1. Spherical symmetry
the bulk crystal phase inside the RVE Vpoly behaves linear
of both the loading and the geometry of the considered
elastically,
solid implies vanishing tangential displacement disconti-
nuities at the inclusion interface, nt 0. Since Kn ! 1,
8x 2 Vi : rx CC : ex 10 also nn 0 (no mutual interpenetration of crystals), and
the matrix-inclusion problem with compliant interfaces
with CC 3kC J 2lC K as the isotropic elastic stiffness
of the bulk material phase comprising all single crystals;
with bulk modulus kC and shear modulus lC. J 1=31 1
and K I J are the volumetric and the deviatoric part of
the fourth-order identity tensor I; with components
Iijkl 1=2dik djl dil dkj ; the components of the second-
order unit tensor 1, dij (Kronecker delta), read as dij = 1 for
i = j and dij = 0 for i 6 j:
The assumption of crystal isotropy deserves to be
commented, since single crystals are generally anisotropic,
including approximately transversely isotropic hydroxay-
apatite [34]. However, hydroxyapatite anisotropy is not
very pronounced [34], and in addition, the disorder Fig. 3 Matrix-inclusion problem with compliant interface (general-
ized Eshelby problem): a spherical inclusion with interface is
of crystals (and of their principal material directions)
embedded in an infinite matrix subjected to uniform strain E0 at
probably renders isotropic phase proporties as suitable infinity. The elastic properties of the matrix are those of the
approximation for the purpose of polycrystal property homogenized material
123
J Mater Sci (2007) 42:88248837 8829
reduces to the classical Eshelby-type inclusion problem nin r a . Their use for estimating the traction forces
with a perfect, rigid interface [17]. Then, consideration of at the interfaces within the polycrystalline RVE Vpoly yields
only one bulk phase (the crystals) implies that the overall the corresponding macroscopic stress Rpoly according to
bulk modulus kpoly is identical to the crystal bulk modulus (6) as
k C,
Z
1
kpoly kC 13 Rpoly an r nr adS 18
VC oVC
For determination of lpoly, a purely deviatoric defor- The solution for the displacements at r = a+ turns out to
mation, E0 E0 e1 e1 e3 e3 , is imposed (see Fig. 3 be, according to (12)2 and (3), nr a ni n=2; a
for the Cartesian base frame e1 ; e2 ; e3 ). The mathematical suitable estimate for the mean displacement n at the crystal
form of the displacement field in the exterior region, r > a interface I ij . Use of this quantity in (4) yields the corre-
(the homogenized material), nex , is established in the line sponding macroscopic strains Epoly in the form
of [27], and reads in spherical coordinates (see Fig. 3 for
Z
Eulerian angles / and h) as 1 s
Epoly n a ndS 19
VC oVC ex
nex;r Bex 5 4mpoly Cex
Aex r 3 4 cos2 / sin2 h cos2 h
E0 r 1 2mpoly r 2 Shear components Rpoly;12 and Epoly;12 of macroscopic
nex;h 1 Bex Cex stresses (18) and strains (19), together with (14)(17) and
Aex r 2 4 2 2 sin 2h1 cos2 /
E0 2 r r (50)(54), give access to lpoly , via lpoly R12 =2E12 ,
nex;/ 1 Bex Cex yielding (after elimination of E0) the following expression,
Aex r 2 4 2 2 sin h sin 2/
E0 2 r r 2 !1 31
14 lC 5j lC 6kC 17lC
1 34 5 20
where mpoly is the Poissons ratio of the polycrystal with lpoly 2 8lpoly 57kC 4lC
weak interfaces,
with the dimensionless quantity j Kt0 a=lC . j ! 1
3kpoly 2lpoly relates to a rigid interface. The higher the rigidity j of
mpoly 15 the interface, the higher the overall polycrystal shear
6kpoly 2lpoly
modulus (Fig. 4), for different (dimensionless)
compressibilities v lC =kC of the single crystals.
The boundary condition in (12)4 directly implies Aex = 1,
Thereby, crystal incompressibility (v ! 0) guarantees
while the constants Bex and Cex will follow from interface
finite overall shear stiffness even for an interface with
conditions.
vanishing stiffness (j 0), while a polycrystal built up of
Inside the inclusion (r < a, the solid crystal phase), the
crystals with zero bulk modulus (v ! 1) and connected
displacement field nin reads as
through zero-stiffness interfaces (j = 0) does not exhibit
nin;r
Ain rBin r 3 cos2 /sin2 hcos2 h
E0
nin;h 1 11lC 15kC Bin r 3
Ain r sin2h1cos2 / 16 1
E0 2 33kC 2lC
nin;/ 1 11lC 15kC Bin r 3 0.8
Ain r sinhsin2/
E0 2 33kC 2lC
0.6
The four remaining constants Bex ; Cex ; Ain and Bin are
determined by enforcing equilibrium of forces at the 0.4
interface r = a:
0.2
T rin n rex n K 0 n 17
123
8830 J Mater Sci (2007) 42:88248837
any shear stiffness (Fig. 4), but still the bulk stiffness of However, use of the average tangential traction force hTt i
the single crystals according to (13). In case of an in failure criterion (8) is problematic since force peaks ini-
incompressible solid (kC ! 1; v lC =kC ! 0), it tializing failure may be cancelled out in the averaging process.
p
follows from (13) that kpoly ! 1, and (20) reduces to As a remedy, we use the second-order moment hTt2 i (also
called quadratic average) as a characteristic or effective value
Tt x, in the line of [13, 14, 35]. The relation between
for
lpoly 2 lpoly p
485 j 114 9j 57j 0 21 hTt2 i and Rpoly is established through energy consider-
lC lC
ations: The energy stored in the RVE Vpoly can be expressed
through the global macroscopic energy density W as
Upscaled failure properties of polycrystal with weak 1
interfaces Vpoly W Vpoly Rpoly : Epoly
2
1
In order to determine the effective failure properties Vpoly Epoly : Cpoly : Epoly
2
resulting from local failure characteristics (8) and from the 1 2 2
interactions between interfaces and bulk single crystals, we Vpoly kpoly Epoly;v 2lpoly Epoly;d 24
2
are left with relating the local interface forces Tx 2 I to
the macroscopic stresses Rpoly , see (5). The tangential and with macroscopic volumetric strain Epoly;v tr Epoly
normal traction forces, Tt and Tn, occuring in the interface and equivalent deviatoric strain Epoly;d
p
failure criterion (8), are non-homogeneously distributed 1=2 Epoly;d : Epoly;d ; Epoly;d Epoly 1=3Epoly;v 1.
across the interfaces. Failure will occur where relatively In order to express W from a microstructural viewpoint,
high tangential traction forces encounter a relative low we consider the local constitutive behavior of the interface
resistance due to relatively low normal traction forces. (Eq. 9) and of the bulk phase (Eq. 10). The corresponding
Instead of trying to model the actual force fields across the macroscopic elastic energy stored in the RVE reads as
interfaces, we estimate the effect of the actual force dis- Z Z
tribution through so-called effective traction forces, as it is 1 1
Vpoly W r : edV T ndS
commonly done for stress, strain, or force fields in the 2 Vpoly 2 I
Z Z
context of continuum micromechanics [14, 47]. In this line, 1 1
e : CC : edV n K ndS 25
we represent the failure-inducing interplay between mod- 2 Vpoly 2 I
erate normal traction forces and tangential traction force
R
peaks by means of two different effective measures for the In order to extract hTt2 i A1C I Tt2 dS from (25), varia-
normal and the tangential traction forces, respectively: (i) tions of W with varying Kt (holding merely Epoly fixed) are
first-order moments of normal forces, and (ii) second-order studied,
moments of tangential forces.
Z Z
The first-order moment of the normal traction forces, oW oe on
Vpoly : rdV TdS
hTn i, is related to the macroscopic mean stress Rpoly;m oKt Vpoly oKt I oKt
through Z Z
1 oe
n 1 n n ndS : rdV
1 2 I Vpoly oK t
Rpoly;m trRpoly Z Z
3 Z o 1
n ndI : rdV nt 2 dS
1 3 Vpoly oK t 2 I
tr nx rx nxdS
3 AC oVC 26
Z Z
1 1
rrr xdS Tn xdS where T r n was considered and
AC oVC AC oVC R where dI Ris the Dirac
distribution of support I ; V dI f dV I f dS. For
hTn i 22 transformation of (26), we extend Hills lemma [28] to
(22) establishes a first link between the macroscopic the case of displacement discontinuities at the interfaces
stress Rpoly and the interface tractions Tx: We use this [14]. Considering (5) and the format (2) for the
average (or first-order moment) of normal traction forces as macroscopic strains Epoly , (26) can be transformed to
to estimate the average interface resistance Ttcr in (8), Z
oW o
according to Vpoly e n ndI : rdV
oKt Vpoly oKt
Z Z
1 oEpoly 1
Ttcr ah hTn i 23 nt 2 dS : Rpoly n 2 dS 27
2 I oKt 2 I t
123
J Mater Sci (2007) 42:88248837 8831
Z 0.5
oW 1 ID E
Vpoly nt 2 dS nt 2 28
oKt 2 I 2 0.4
0.3
Identification of (28) with the derivation of the mac-
roscopic expression for the energy density (24) with 0.2
respect to Kt yields
0.1
I D E ok
poly 2 olpoly 2
nt 2 Epoly;v 4 E 29 0
10 20 30 40 50
Vpoly oKt oKt poly;d
Fig. 5 Concentration factor BTt relating macroscopic deviatoric
When considering hTt2 i Kt2 n2t according to (9), stress on polycrystal to effective tangential traction in intercrystalline
okpoly =oKt 0 according to (13), and Rpoly;d interfaces, as function of dimensionless quantity j Kt0 a=lC
2lpoly Epoly;d , (29) reduces to (interfacial rigidity), for different crystal compressibilities
v lC =kC , Eq. 32
!
I o 1 increases the peaks of tangential traction p force,
i.e., the
hTt2 i Kt2 R2poly;d 30 proportionality factor BTt , again bounded by 2=5 (Fig. 5),
Vpoly oKt lpoly
r
where Rpoly;d is the equivalent deviatoric stress of the 2
lim BTt v 34
macroscopic second-order stress tensor Rpoly , j!1 5
r Use of the micro traction-macro stress relationships (22)
1
Rpoly;d Rpoly;d : Rpoly;d and (32) in the local interface criterion (8) yields a mac-
2
roscopic polycrystal-specific brittle-failure criterion in the
with Rpoly;d Rpoly Rpoly;m 1; 31
form
1
and Rpoly;m trRpoly
3 BTt Rpoly;d ah Rpoly;m 35
Combination of (30) with I =Vpoly 3=2a and with (35) expresses that Coulomb-type brittle failure (8) in the
j Kt0 a=lC yields interfaces between spherical crystals inside the RVE results
q in DruckerPrager-type (brittle) failure properties at the
hTt2 i BTt Rpoly;d scale of the polycrystal.
v
!
u
l u 1 o lC Micromechanics of porous material with polycrystalline
with BTt v C ; j t j2 32 skeleton
kC 3 oj lpoly
On the other hand, for any constant crystal compress- with the Eshelby tensor S for spherical inclusions reading
ibility v, stiffening the interface (enlarging j Kt0 a=lC ) also as [17]
123
8832 J Mater Sci (2007) 42:88248837
" 2 #
3kpoly 6kpoly 2lpoly 3u a
S J K 37 R2PORO;m
3kpoly 4lpoly 53kpoly 4lpoly 4 B Tt
" #
so that 2u23 50mpoly 35m2poly
1 R2PORO;d
7 5mpoly 2
4kpoly lpoly 1 u 2
kPORO 38 a
3kpoly u 4lpoly 2 h1 uRPORO;m
B Tt
2
1 u9kpoly 8lpoly a
lPORO lpoly 39 h2 1 u2 44
9kpoly 1 23 u 8lpoly 1 32 u B Tt
We consider brittle failure of the overall porous medium with mpoly mpoly kpoly ; lpoly according to (15),
if the polycrystal failure criterion (35) is reached in highly lpoly lpoly kC ; lC ; j according to (20), and
stressed regions of the polycrystalline matrix. The corre- BTt BTt v lkCC ; j according to (32).
sponding (micro-) heterogeneity within the solid matrix The elastic stress domain of the porous medium the
has recently been shown [13] to be reasonably considerable matrix of which is a polycrystal with brittle interfaces in-
through so-called (homogeneous) effective (micro-) creases with decreasing crystal compressibility v (Fig. 6).
stresses, such as the square root of the spatial average over For the incompressible limit case, v ! 0, (44) reduces to
" 2 #
the solid material phase, of the squares of equivalent
3u a 2
deviatoric (micro-) stresses, R2PORO;m 1 u R2PORO;d
4 B Tt 3
2
s
Z a
q 2 h1 uRPORO;m
1 1 B Tt
hr2d iS rd x : rd xdV 40
VS V S 2 2
a
h2 1 u2 45
1 B Tt
with rd x rx trrx1 41
3 For a crystal compressibility of hydroxyapatite,
v 0:54 (see also Section Application to hydroxyapatite
The effective deviatoric stress (40), used to approximate biomaterials), the elastic domain increases with decreas-
Rpoly;d in (35), is accessible through energy considerations ing interfacial rigidity (Fig. 7) and with increasing friction
similar to those of (24) to (30), and result to be ([9, p. 132, angle a (Fig. 8). High interfacial rigidities j or low friction
13]) angles a result in closed elastic domains, indicating pos-
sible failure of the porous material even under hydrostatic
"
o 1 stress states R 1Rm , while low interfacial rigidities j or
R2poly;d hr2d iS R2 high friction angles a are related to open elastic domains,
olpoly kPORO PORO;m
# related to infinite resistance of the porous material, as long
o 1 l2poly
R2PORO;d 42
olpoly lPORO 1 u
0.5
Z
1 1 0.2
Rpoly;m hrm iS trrxdV
VS V S 3
RPORO;m 0.1
43
1u
0
2 1.5 1 0.5 0 0.5
Use of Eqs. 43 and 42, together with (38)(41), (13), and
(20), in (35) yields a failure criterion at the scale of the Fig. 6 Elastic limits of a porous material the matrix of which is a
polycrystal with brittle interfaces, for different crystal compressibil-
porous material with polycrystalline interfaces in the solid ities v lC =kC (Eq. 44): u 0:5; a 0:3; j ! 1. Uniaxial load
phase, path indicated (thin solid line)
123
J Mater Sci (2007) 42:88248837 8833
Porous hydroxyapatite (HA) biomaterials are widely used In the hierarchical organization of synthetic hydroxyapatite
for replacement of hard tissue defects, because of their ceramics, we identify two different scales which will be
chemical composition, microstructure and Youngs mod- considered in the framework of a two-step homogenization
ulus being similar to the bone mineral, called carbonated or scheme. The first homogenization step refers to an obser-
calcium-deficient hydroxyapatite (CDHA) [26, 36, 46]. If vation scale of several hundreds of microns where
porous scaffolds are used as bone replacement material in hydroxyapatite crystals are separated by boundaries or
highly loaded anatomical locations, reliability of their interfaces (Fig. 9a). The latter will be shown to be a po-
mechanical properties is particularly important for the tential nucleus for failure of the material. The corre-
performance of the implants. Therefore, the prediction of sponding homogenized material is called hydroxyapatite
strength of HA biomaterials from their microstructure and polycrystal with interfaces. At the microstructural scale
porosity is of particular interest. To the knowledge of the with a characteristic length of some millimeters (Fig. 9b),
authors, corresponding micromechanical approaches are pores are embedded in a matrix which is made up of the
extremely rare or inexistent, so that we check in this material which was homogenized in the first upscaling step.
123
8834 J Mater Sci (2007) 42:88248837
123
J Mater Sci (2007) 42:88248837 8835
3u 2
a02 R2PORO;m 1 uR2PORO;d
700 4 3
Peelen et al. 1978
600 Akao et al. 1981 2a02 h1 uRPORO;m a02 h2 1 u2 49
Martin and Brown 1995
micromechanical model
500 with a0 a=BTt j and h as only two parameters being left
for an optimization procedure to match the experimental
400
data of Fig. 10 and Table 2. This procedure delivers a
300 cohesion hopt 16:51 MPa (close to the values obtained for
the brittle case in Section Biomaterial-independent prop-
200
erties of interfaces between hydroxyapatite crystals,
100 a; h; jback-analysis) and ratio a0opt 1:61 which is
quasi-identical to the one obtained for the brittle case
0 0.2 0.4 0.6 0.8 1 (Section Biomaterial-independent properties of interfaces
between hydroxyapatite crystals, a; h; jback-analysis),
Fig. 10 Uniaxial compressive strength fc of porous hydroxyapatite
biomaterial as function of porosity u: model prediction accordingpto
implying an aj-relationship quasi-identical to that of
Eq. 44 or Eq. 49, evaluated with RPORO;m fc =3; RPORO;d fc = 3, Fig. 11. This means that the failure of porous hydroxyapa-
compared to experimental data (Table 2). Correlation coefficient tite biomaterials can be equally well represented by a brittle
r2 = 0.97 elastic-limit-type micromechanics model and a ductile one
related to limit analysis. In this context, it is very interesting
to note that the ductile criterion (49) is even identical to the
1
elastic domain for incompressible solid matrices, Eq. 45.
Accordingly, one might argue that the nature of the
0.8
heterogeneity of the stresses in the solid matrix (considered
herein by quadratic averages) is far more important for the
0.6 overall failure of the material than the precise mode of
local interface failure (brittle or ductile). However, as re-
0.4 gards hydroxyapatite biomaterials, experiments [7, 38, 43]
strongly support brittle failure: A comprehensive mechan-
0.2 ical formulation for its possible origin, namely breaking of
weak interfaces between hydroxyapatite crystals, was the
0 20 40 60 80 100 main focus of the present paper.
Fig. 11 Friction angleinterface rigidity relationship a(j) suitable for
representation of strength of hydroxyapatite biomaterials (Fig. 10)
123
8836 J Mater Sci (2007) 42:88248837
Cex 5a3 48l2poly aKt kC 240l2poly kC lC 10. Dormieux L, Barthelemy J-F, Maghous S (2006) Comptes Ren-
dus Mecanique 334:111, In French
136l2poly aKt lC 176l2poly l2C 8lpoly l3C 11. Dormieux L, Kondo D, Ulm F-J (2006) Microporomechanics.
Wiley
9lpoly aKt kC lC 114lpoly l2C kC 132lpoly aKt l2C 12. Dormieux L, Maghous S (2000) CR Acad Sci IIb Math Mech
328(8):593, In French
57l2C aKt kC 4l3C aKt lpoly =N 51
13. Dormieux L, Molinari A, Kondo D (2002) J Mech Phys Solids
50:2203
Ain 5544l2poly aKt lC 192l2poly aKt kC 320l2poly l2C 14. Dormieux L, Sanahuja J, Maalej Y (2007) Comptes Rendus
Mecanique 335:25, In French
1536l2poly kC lC 16lpoly aKt l2C 228lpoly aKt kC lC 15. Dormieux L, Sanahuja J, Maghous S (2006) Comptes Rendus
408akpoly lpoly Kt lC 144akpoly lpoly Kt kC Mecanique 334:19
16. Dormieux L, Ulm F-J (eds) (2005) CISM courses and lectures,
240kpoly lpoly l2C 1152kpoly lpoly kC lC vol 480. Applied micromechanics of porous media. Springer,
Wien
12aKt l2C kpoly 171aKt kC kpoly lC lpoly =N 52 17. Eshelby JD (1957) Proc Roy Soc Lond, Ser A 241:376
18. Fritsch A, Dormieux L, Hellmich Ch (2006) Comptes Rendus
Bin 240l2poly lC 8lpoly lC 6kpoly lC Mecanique 334(3):151
19. Fritsch A, Hellmich Ch (2007) J Theor Biol 244(4):597
12lpoly kC 9kpoly kC =a2 N 53 20. de Gennes P-G, Okumura K (2000) CR Acad Sci IV 1(2):257
21. Hashin Z (1991) J Appl Mech 58:444
N 1408l3poly l2C 192l2poly l3C 24l3C akpoly Kt 22. Hellmich Ch, Barthelemy J-F, Dormieux L (2004) Eur J Mech A-
Solids 23:783
342l2C akpoly Kt kC 1920l3poly kC lC 23. Hellmich Ch, Mang HA (2005) J Mater Civil Eng (ASCE)
17(3):246
1088l3poly lC aKt 384l3poly aKt kC 24. Hellmich Ch, Ulm F-J (2002) J Eng Mech (ASCE) 128(8):898
1664al2poly Kt l2C 1584l2poly kpoly l2C 25. Hellmich Ch, Ulm F-J, Dormieux L (2004) Biomech Model
Mechanobiol 2:219
1032al2poly Kt kC lC 2160l2poly kpoly kC lC 26. Hench LL, Jones JR (eds) (2005) Biomaterials, artificial organs
and tissue engineering. Woodhead Publishing, Cambridge, UK
2736l2poly l2C kC 432l2poly akpoly Kt kC 27. Herve E, Zaoui A (1993) Int J Eng Sci 31(1):1
1224l2poly akpoly Kt lC 801aKt kC kpoly lpoly lC 28. Hill R (1963) J Mech Phys Solids 11:357
29. Hofstetter K, Hellmich Ch, Eberhardsteiner J (2005) Eur J Mech
852aKt l2C kpoly lpoly 1710lpoly l2C kpoly kC A Solids 24(6):1030
30. Hofstetter K, Hellmich Ch, Eberhardsteiner J (2006) Comput
120lpoly l3C kpoly 684lpoly l2C aKt kC Assisted Mech Eng Sci 13(4):523
48lpoly l3C aKt 54 31. Hunt BR, Eipsman RL, Rosenberg JM (2001) A guide to
MATLAB for beginners and experienced users, 1st edn. Cam-
bridge University Press, Cambridge, UK
They define the displacement fields (14) and (16), which 32. Katti DR, Katti KS (2001) J Mater Sci 36(6):1411
give access to strains e Os n, stresses R (via (12)1, and 33. Katti DR, Katti KS, Sopp JM, Sarikaya M (2001) Comput Theor
(12)3 respectively), mean interface displacements n and Polym Sci 11(5):397404
34. Katz JK, Ukraincik K (1971) J Biomech 4:221
interface tractions t (via (12)2). 35. Kreher W (1990) J Mech Phys Solids 38(1):115
36. LeGeros RZ (2002) Clin Orthop Relat Res 395:81
37. Lemarchand E, Ulm F-J, Dormieux L (2002) J Eng Mech (ASCE)
References 128(8):876
38. Martin RI, Brown PW (1995) J Mater Sci: Mater Med 6:138
1. Akao M, Aoki H, Kato K (1981) J Mater Sci 16:809 39. Okumura K (2002) Eur Phys JE Soft Matter 7(4):303
2. Barthelat F, Tang H, Zavattieri PD, Li C-M, Espinosa HD (2007) 40. Okumura K (2003) Europhys Lett 63(5):701
J Mech Phys Solids 55(2):306 41. Okumura K, de Gennes P-G (2001) The Eur Phys JE Soft
3. Barthelemy J-F, Dormieux L (2003) Comptes Rendus Mecanique Matter 4(1):121
331:271 42. Peelen JGJ, Rejda BV, de Groot K (1978) Ceramurgia Int 4(2):71
4. Barthelemy J-F, Dormieux L (2004) Int J Numer Anal Methods 43. Pramanik S, Agarwal AK, Rai KN, Garg A (2007) Ceram Int
Geomech 28(78):565 33(3):419
5. Bernard O, Ulm FJ, Lemarchand E (2003) Cement Concr Res 44. Sanahuja J, Dormieux L (2005) Comptes Rendus Mecanique
33(9):1293 333:818, In French
6. Bernaud D, Deude V, Dormieux L, Maghous S, Schmitt DP 45. Schwefel H (1977) Numerische Optimierung von Computer-
(2002) Int J Numer Anal Methods Geomech 26:845 Modellen mittels der Evolutionsstrategie. Birkhauser Basel,
7. Chu T-MG, Orton DG, Hollister SJ, Feinberg SE, Halloran JW Switzerland, In German
(2002) Biomaterials 23:1283 46. Suchanek W, Yoshimura M (1998) J Mater Res 13(1):94
8. de With G, van Dijk HJA, Hattu N, Prijs K (1981) J Mater Sci 47. Suquet P (ed) (1997) Continuum micromechanics. Springer,
16:1592 Wien
9. Dormieux L (2005) In: Dormieux and Ulm [16], p 109 48. Tai K, Ulm F-J, Ortiz Ch (2006) Nano Lett 6(11):2520
123
J Mater Sci (2007) 42:88248837 8837
49. Ulm F-J, Constantinides G, Heukamp FH (2004) Materials and 51. Zaoui A (1997) In: Suquet [47], p 291
structures 37(1):43 52. Zaoui A (2002) J Eng Mech (ASCE) 128(8):808
50. Ulm F-J, Delafargue A, Constantinides G (2005) In: Dormieux 53. Zhong Z, Meguid SA (1997) J Elasticity 46(2):91
and Ulm [16], p 207
123
J Mater Sci (2007) 42:88388843
DOI 10.1007/s10853-007-1846-9
Received: 29 January 2007 / Accepted: 14 May 2007 / Published online: 17 July 2007
Springer Science+Business Media, LLC 2007
Abstract The present paper investigates the force- confinement free energy and the longitudinal extension
extension behavior of a stiff polymer under stretching were proposed as functions of the confinement cross-sec-
inside a small tube. We develop a theory and perform tion. This line of work was significantly elaborated for
Brownian dynamic simulations based on a recently devel- polymer models such as the Gaussian and wormlike chains
oped generalized bead-rod model (GBR) to show that the [2]. For a stiff polymer chain1 tightly2 confined inside a
force-extension relation of such a strongly confined poly- tube or nematic liquid, an interesting scaling law for the
mer chain can be described by that of an unconfined confinement free energy of the chain was derived by Odijk
polymer subject to an effective force which is derived et al. [1016], with results later confirmed by numerical
based on Odijks theory of a confined polymer chain. simulations [1719].
Over the last decade, direct experimental investigations
have become possible for static and dynamic properties of
single polymer molecules subject to external forces, such
as the measurement of force-extension relation of DNAs in
Introduction electrolytes [2022]. Bustamante et al. [21] have shown
that the force-extension curve of a 97004 bps DNA mol-
Physical properties of macromolecules and polymers in ecule in 10 mM Na+ can be described by a wormlike chain
complex environments are usually influenced by external model provided that the contour length L and persistence
conditions such as geometrical confinements and applied length p are treated as adjustable parameters to fit experi-
forces [1, 2]. Polymers in geometrical confinements that mental data. Marko and Siggia [23] interpolated a formula
are smaller than their unconfined molecular sizes are of for the extension of polymers which can almost exactly fit
great significance in fields from polymeric liquid crystals the force-extension curve of a long DNA chain. For short
[3] to biological structures such as nucleosomes [4] and polymer chains, Kierfeld et al. [24] developed a discrete
viruses [5, 6]. In concentrated polymeric solutions and harmonic chain model which has been numerically verified
melts, a single polymer can be subjected to large resis- [19] and shown to be valid even for a charged short chain if
tances to motion from surrounding chains perpendicular to an effective persistent length is introduced [25].
its contour while moving along its contour with relative
ease. This situation is similar to a polymer confined in a
tube-like region [2]. Early discussions on the behaviors of a 1
In an unconfined solution, a stiff polymer usually means that its
single polymer chain confined in a tube can be found in de contour length L is smaller than its persistence length p. On the other
Gennes work [1, 79], where simple scaling laws for the hand, a polymer confined in a small tube can be regarded as stiff as
long as the persistence length p of a polymer is larger than the tube
radius R, irrespective of the ratio L/p. We do not distinguish these two
J. Wang (&) H. Gao situations in the current study.
2
Division of Engineering, Brown University, 182 Hope Street, Here, a tightly or strongly confined polymer implies that the
Providence, RI 02912, USA typical length scale of the confinement is smaller than the polymers
e-mail: Jizeng_Wang@brown.edu persistence length.
123
J Mater Sci (2007) 42:88388843 8839
Figure 1 shows the model system of a continuous wormlike Burkhard [26] has shown that the confinement free energy
chain confined inside a cylindrical tube with radius R. A set of a polymer with respect to the harmonic potential in
of Cartesian coordinates (x, y, z) are placed at the center of Eq. 7 has the same general form as that based on Eq. 6,
the tube so that the z-axis is along the tube axis. The chain with difference only in a dimensionless prefactor which
is stretched by a tensile force f f k where k is the unit depends on the particular potential. Following this
vector along the z-axis. The position vector along the arc approach, we will establish the basic form of the solution
length s of the chain is based on the harmonic potential in Eq. 7 and then deter-
mine a prefactor based on Odijks theory of a confined
r r? s zsk 1 polymer chain.
In Eq. 5, the potential energy with respect to the
where r? x; y is perpendicular to the z-axis, and external force can be rewritten as
r0 0: The derivatives
Z L
or? dz f
u? ; ujj k 2 f rL r0 fz fL u2? ds 8
os ds 2 0
define the tangential vector. In the case of strong where we have used Eq. 3 and z(0) = 0. Inserting Eqs. 4, 7
confinement, the undulation of the chain due to thermal and 8 into Eq. 5 yields
fluctuation will be small so that ku? k\\1: The
Z L
inextensibility condition of the wormlike chain 1 ou? 2
kuk kor=osk 1 and Eq. 2 lead to H pkB T ds
2 0 os
Z Z Z s 2
f L 2 N L
dz 1 u? ds u? ndn ds 9
1 u2? Ou4? : 3 2 0 2 0 0
ds 2
Therefore, we have where we have dropped a constant term and used relation
Z s
o2 r o2 r ? d 2 z o u? Ou2? k ou?
4
2 2k : r? s u? ndn: 10
os2 os ds os os 0
The Hamiltonian of the confined wormlike chain under Following earlier studies [3, 23, 29], we introduce
stretching can be expressed as the summation of bending Fourier transform
123
8840 J Mater Sci (2007) 42:88388843
Z Z L
~? x eixs u? sds 11 1
u zL L u2? ds; 20
2 0
123
J Mater Sci (2007) 42:88388843 8841
normalized tensile force, as well as p/R, p/L and b/p. Only with current position rnj under cylindrical confinement to
when L>>p and b<<p Eq. 24 can be reduced to Eq. 23. illustrate the algorithm. If the bead is located close enough
to the reflecting wall, e.g.,
q
Brownian dynamics simulations Snj R x2nj y2nj
p 27
5DDt
Here we briefly describe Brownian dynamics simulations
that were performed to verify the simple force-extension the stochastic movement caused by the reflecting wall is
relations proposed in the previous section. The simulations given by
were based on the GBR model of a wormlike chain under xnj
strong confinements [19]. dxwall
nj q
In the GBR model, a semiflexible polymer or filament is x2nj y2nj
described as N identical virtual beads of radius a connected
Snj p Snj p
by N1 inextensible rods of length b with the unit f1 p DDt f2 p DDU 28
DDt DDt
tangential vectors uj (juj j 1, j 1; 2; . . . ; N 1). The
contour length of the chain is L = (N1)b. The N virtual ynj
dywall
nj q
beads with coordinates rj xj ; yj ; zj (j 1; 2; . . . ; N) are x2nj y2nj
introduced for modeling hydrodynamic interactions 29
between different chain segments. The Brownian dynamics Snj p Snj p
f1 p DDt f2 p DDU
of a discrete wormlike chain involves the collective motion DDt DDt
of N identical beads in solution. After the hard rod con-
straint is implemented via the linear constraint solver dzwall
nj 0 30
(LINCS) [30], the new position vector rn1 of the N beads
where j 1; 2; . . . ; N, Snj is the current relative radial
is determined from
position of the jthpbeadand functions f1, f2 are defined in
[19]. The value 5DDt used above was suggested by
rn1 I Tn Bn rn
Peters et al. [31] in their treatment of a sticky wall. In the
Dt
Dn Fn nn Tn d 25 present case, the wall of the confining tube is assumed to
kB T influence the motion of a bead only if the distance between
where rn (3N vector) is the current position of the beads, the bead and the wall is smaller than this value. We
Fn is the collective vector of internal (inter-beads) and therefore set dxwall wall wall
nj dynj dznj 0 if the current
external forces, nn is the random force generated at each distance
p between the jth bead and the wall exceeds
p
time step from a Gaussian distribution with zero mean and 5DDt, i.e., Snj > 5DDt. For convenience, we can
variance equal to express the above stochastic displacements given by
Eqs. 2830 as a 3N vector v, which consists of
\nn nn0 > 2Dn Dtdnn0 : 26 vwall wall wall wall
nj dxnj ; dynj ; dznj , j 1; 2; . . . ; N. Including v
in the current position vector rn in Eq. 25, we obtain
Here, Dt is the time step and dnn0 is the Kronecker delta
symbol, I Tn Bn is a projection matrix which sets the rn1 I Tn Bn rn vwall
n
constraints and Dn is the translational diffusion matrix Dt
determined through hydrodynamic interactions between Dn Fn n Tn d 31
kB T
beads.
Special considerations are needed for numerical simu-
lations of spatially confined wormlike chains. Peters et al. Results and discussions
[31] introduced an efficient algorithm for the Brownian
dynamics of a particle near reflecting wall. In their method, For a confined wormlike chain subjected to a constant
the errors of discretization are kept on the order of O(Dt) to tensile force f in the z-direction on both ends of the chain,
handle the boundary conditions, while a naive treatment of the force vector Fn in Eq. 31 can be written as
identifying reflection processes by checking pboundary
crossings usually yields errors on the order of O Dt [31]. Fn Fbn Ft 32
Their approach has been adapted to polymer confinements
in the GBR model [19]. For the collective motion of many where Ft is the tensile force and Fbn is the effective force
beads in the bead-rod wormlike chain, we take the jth bead on the beads due to the bending rigidity of the chain.
123
8842 J Mater Sci (2007) 42:88388843
-1
0.12 p=p0, R=10nm 10
~t1/2
p=2p0, R=20nm
p=2p0, R=10nm
0.1
p=p0, R=5nm
CWLC, p/R=5.3, c=2.5
0.08 DWLC, p/R=5.3, c=2.5
~t
CWLC, p/R=10.6, c=2.7
1-<z>
1-<z>
DWLC, p/R=10.6, c=2.7
-2
0.06 10
p=p0, R=10nm, fp/kBT=8
p=1.5p0, R=15nm, fp/kBT=8
0.04
p=2p0, R=20 nm, fp/kBT=8
fp/kBT=31, R=Infinity
0.02
f=0, R=8nm
-3
0 -2 0 1 2 3 4
10 -2 -1 0 1
10 10 10 10 10 10 10 10 10 10
fp/k T t ( s)
B
Fig. 2 Comparisons of Brownian dynamics simulation results for the Fig. 3 The dynamic evolution of the relative extension of a wormlike
relative extension of nanotube confined wormlike chains with chain under stretching by a force of different magnitudes inside
corresponding theoretical predictions based on the concept of an nanotubes of different radii, where the effective force is fixed as
effective force in Eq. 22. CWLC stands for the continuous 31kB T=p and p0 53:248 nm corresponds to the persistence length of
wormlike chain model and DWLC stands for the discrete wormlike a DNA chain
chain model. The theoretical predictions are shown as continuous
curves. In plotting the curves, we have taken the basic measure of
Dt 300 ps, a = 1.3nm and T = 293 K. The effective
persistence length to be p0 53:248 nm , corresponding to that of a
DNA chain force is fixed at fe p=kB T 31 for each trajectory, which
can be realized by setting p=R 5:3 and fp=kB T 8, or by
setting fp=kB T 31 for an unconfined chain, or by con-
Based on the GBR model, Brownian dynamics simula- fining a free chain in a tube. The results indicate that
tions have been performed for wormlike chains in equilibrium is reached within a few microseconds. It can be
nanotubes of different radius. In all simulations, the chains observed from Fig. 3 that, under the same normalized
are initially set in a straight configuration. Tube confine- effective force fe p=kB T 31, the crossover times between
ments and constant tensile forces are then applied during ballistic ( t) and diffusive ( t1=2 ) behaviors are different
the chains relaxation. We record the normalized end-to- for tubes with different radii. These results confirm that the
end distance <z> of a chain along z-axis at each time force-extension behavior of a wormlike chain confined in a
increment. Figure 2 compares the simulation results with nanotube can be well described by the discrete wormlike
corresponding theoretical predictions based on the effec- chain model of Eq. 24 incorporating the concept of effec-
tive force concept in Eq. 22. The chains were simulated for tive force defined in Eq. 22.
a total time of 10 ls with parameters b/p 0.19,
a 1:3 nm, T = 293 K, Dt 300 ps (in cases of
fp=kB T\100) or Dt 30 ps (in cases of fp=kB T > 100). Conclusion
Each data point in Fig. 2 is obtained by averaging the
recorded values of <z> for 200 trajectories with different Theories of confined polymers suggest that the effect of
random seeds. The abbreviation DWLC stands for the geometrical confinements on the force-extension relation
discrete wormlike chain model in Eq. 24, and CWLC of a polymer can be represented by an effective force
stands for the continuous wormlike chain model in Eq. 23. applied to an unconfined polymer chain. Here we have
The simulations were conducted for chains within different theoretically investigated this problem and derived an
tubes under parameters p/R 5.3, p=R 10:6 and analytical expression of this effective force for a polymer
R = 5 nm, 10 nm, 20 nm. It can be seen from Fig. 2 that chain confined in a cylindrical tube based on Odijks theory
the simulation results are in good agreement with those on confined polymers. We have also performed Brownian
predicted by the discrete wormlike chain model when the dynamics simulations, based on a recently developed
effective force expression is applied. Figure 3 illustrates generalized bead-rod model, to investigate the force-
the dynamic evolution, averaged over 200 different tra- extension behavior of a wormlike chain confined in a
jectories, of the relative extension of wormlike chains nanotube, with results in excellent agreement with the
confined in nanotubes of different radius over a total sim- discrete wormlike chain model of Kierfeld et al. [24]
ulation time of 10 ls under parameters b/p = 0.19, incorporating the effective force concept defined in Eq. 22.
123
J Mater Sci (2007) 42:88388843 8843
References 18. Bicout DJ, Burkhard TW (2001) J Phys A: Math Gen 34:5745
19. Wang J, Gao H (2005) J Chem Phys 123:084906
1. De Gennes PG (1979) Scaling concepts in polymer physics. 20. Smith S, Finzi L, Bustamante D (1992) Science 258:1122
Cornell University Press, Ithaca, NY 21. Bustamante C, Marko JF, Siggia ED, Smith S (1994) Science
2. Doi M, Edwards SF (1986) The theory of polymer dynamics. 265:1599
Oxford University Press, New York 22. Strick T, Allemand J, Bensimon D, Bensimon A, Croquette V
3. Odijk T (1986) Macromolecules 19:2313 (1996) Science 271:1835
4. Shiessel H (2003) J Phys Condens Matter 15:R699 23. Marko JF, Siggia ED (1995) Macromolecules 28:8759
5. Earshaw WC, Harrison SC (1977) Nature 268:598 24. Kierfeld J, Niampoly O, Sa-Yakanit V, Lipowsky R (2004) Eur
6. Purohit PK, Kondev J, Philips R (2003) Proc Natl Acad Sci USA Phys J E 14:17
100:3173 25. Wang J, Fan X, Gao H (2006) Mol Cell Biomech 3(1):13
7. De Gennes PG (1976) Macromolecules 9:587 26. Burkhardt TW (1995) J Phys A: Math Gen 28:L629
8. De Gennes PG (1976) Macromolecules 9:594 27. Burkhardt TW (1997) J Phys A: Math Gen 30:L167
9. De Gennes PG (1971) J Chem Phys 55:572 28. Jo K, Dhinhra DM, Odijk T, De Pablo JJ, Graham MD, Runn-
10. Odijk T (1983) Macromolecules 16:1340 heim R, Forrest D, Schwartz DC (2007) Proc Natl Acad Sci USA
11. Odijk T (1993) Macromolecules 26:6897 104:2673
12. Khokhlov AR, Semenov AN (1981) Physica A 108:546 29. Landau LD, Lifshitz EM (1958) Statistical physics. Addison-
13. Khokhlov AR, Semenov AN (1982) Physica A 112:605 Wesley, Reading, MA
14. Helfrich W, Harbich W (1985) Chem Scr 25:32 30. Hess B, Bekker H, Berendsen HJC, Fraaije JGEM (1997)
15. Helfrich W (1978) Z Naturforsch A 33:305 J Comput Chem 18:1463
16. Helfrich W, Servuss RM (1984) Nuovo Cimento D 3:137 31. Peters EAJF, Barenbrug TMAOM (2002) Phys Rev E 66:056701
17. Dijkstra M, Frenkel D, Lekkerkerker HNW (1993) Physica A
193:374
123
J Mater Sci (2007) 42:88448852
DOI 10.1007/s10853-007-1820-6
Received: 9 February 2007 / Accepted: 3 May 2007 / Published online: 14 July 2007
Springer Science+Business Media, LLC 2007
Abstract Theoretical analysis of Poly-(L)-Lactic Acid PLLA nanofibers. The theoretical results correlate well
(PLLA) nanofibers is a necessary step towards designing with reported experimental data.
novel biomedical applications. This paper aims to analyze
the mechanical properties of PLLA nanofibers so that
optimal scaffolds in tissue engineering applications can be
developed. We carry out analysis of PLLA nanofibers to
estimate the mechanical properties from basic building Introduction
blocks to the nanofibrous structures. A single PLLA
nanofiber is made up of ShishKebab like fibrils inter- Polymeric nanofibers are attractive materials for a wide
twined together and can contain both amorphous and range of applications in the bio-medical, textile and other
crystalline phases. The elastic modulus of the Lactic acid emerging technologies. This is primarily due to their large
monomeric formation in the crystalline phase is derived surface area to volume ratio and the unique features at the
using second-derivative of the strain energy using molec- nanometer scale [1]. Structures of fibrous polymers are
ular dynamics simulation. The mechanical property of the generally very flexible, and their conformation is easily
ShishKebab fibril is derived by homogenization. The fiber deformed against mechanical extension or induced motion
modulus is then obtained using the Northolt and van der between its atoms. In any industrial application, the suit-
Houts continuous chain theory. One of the significant ability of a material and/or structure relies significantly on
contributions in this paper is the use of modified continu- their physical properties, especially their mechanical and
ous chain theory, where a combined multiscale approach is electrical properties. Whilst the mechanical design ensures
used in the estimation of the mechanical properties of dimensional stability and structural integrity, the electrical
design aims to fulfill the functionality of the products.
In recent years, polymeric nanofibers have been devel-
V. U. Unnikrishnan G. U. Unnikrishnan oped for a variety of applications such as tissue engineering,
J. N. Reddy molecular filters, sensors and protective clothing [26]. For
Advanced Computational Mechanics Laboratory,
example, polymeric nanofibers can be used to form nano-
Department of Mechanical Engineering, Texas A&M University,
College Station, TX 77843-3123, USA fibrous scaffolds for tissue engineering application [7].
These polymeric scaffolds allow cells to proliferate and
J. N. Reddy (&) grow into tissues with defined sizes and shapes for trans-
Engineering Science Programme,
plantation purposes [8, 9]. An understanding of the struc-
National University of Singapore, Singapore, Singapore
e-mail: jnreddy@tamu.edu tureproperty relationship is essential for the engineering
applications of polymeric nanofibers since they are affected
C. T. Lim (&) by the mechanical properties arising from the internal
Division of Bioengineering & Department of Mechanical
molecular structures. Tremendous savings in cost can be
Engineering, National University of Singapore,
9 Engineering Drive 1, Singapore 117576, Singapore achieved if preliminary experimental designs can be eval-
e-mail: ctlim@nus.edu.sg uated theoretically to eliminate inferior designs and reduce
123
J Mater Sci (2007) 42:88448852 8845
the number of experiments. The proposed theoretical work lactic acid using MD simulation for the estimation of the
in the analysis of nanofiber is primarily aimed at providing a mechanical properties in the atomistic scale. Homogeniza-
computational framework for the estimation of the tion and description of the ShishKebab model is discussed
mechanical properties and to provide a strong connection in section (ShishKebab modelelectrospun nanofibers).
between experimental observations and theoretical analysis. The multiscale transfer of quantities of interest from the
Fibers prepared from polymer solution or melt by con- atomistic scale to the mesoscale by micromechanical
ventional methods (melt, dry and wet spinning) have methods is discussed in sections (Micromechanical analy-
diameters in the range of 5500 mm [10]. Recently, there sis) and (Continuum volume averaging: micromechanical
has been increased interest in the fabrication of nanofibers method). In section (Continuous chain model of polymeric
(with diameters in the range from tens to hundreds of fibers), the formulation of the continuum chain model
nanometers) using electrospinning [1, 10, 11] for mechan- to scaleup the material properties is discussed. Section
ical characterization studies. Using molecular dynamics (Results and discussion) combines the results from various
(MD) simulation, crystalline lactic acid monomer units are methods and finally the paper concludes with a summary in
equilibrated and thermostatted to the experimental condi- section (Conclusion).
tions by a series of NVE ensemble (Microcanonical
ensemble) and NVT ensemble (Canonical ensemble) anal-
ysis and subjected to isothermal strain conditions to obtain Atomistic simulation
the mechanical properties [12, 13]. To develop an optimal
scaffold for tissue engineering application, it is required to The knowledge of structure and molecular motion in
manipulate the mechanical characteristics of the nanofi- polymers is essential to understand the properties of prac-
brous scaffolds. There has been numerous experimental tical interest. Theoretical simulation of the physical pro-
studies on the design of optimal scaffolds [8]. However, cesses forms the first step in this work. The estimation of
very few theoretical studies exist in predicting the the mechanical properties of the PLLA fibers needs to be
mechanical properties and behavior of nanofibers under carried out. There are various methods of estimating the
external mechanical loads using multiscale simulation. This physical properties of atomistic structures, molecular
paper aims to analyze the mechanical properties of PLLA dynamics (MD) simulation being one of them and is used
nanofibers via an atomistic-mesoscale stimulation method. here. Molecular dynamics has been a very popular tool for
All molecular dynamics simulations were performed using the determination of mechanical, thermal and other prop-
Cerius2 (version 4.6, Accelrys, Inc.) simulation package. erties of interest in atomistic structures [2125]. The
Analysis of the orientation process during uniaxial starting point of a MD simulation is the non-relativistic
drawing of a polymer has long been investigated in many quantum mechanical time dependent Schrodinger equation.
theoretical and experimental studies [7, 14, 15]. Based on The thermodynamic state characterized by the fixed num-
the deformation of cellulose fibers, analytical models were ber of atoms, volume and temperature called the canonical
developed for rodlets connected by crosslinks. These ensemble [25] forms the basis of the MD simulation here.
models were modified with the various additions like cross- The simulated system and the heat bath couple to form a
linking with forces applied to the ends of the chains as well composite system. The conservation of the energy still
as changes in material properties. This research lead to two holds in the composite system but the total energy of the
different formulations for the analysis of polymeric chains: simulated system fluctuates. The motion of the particles in
the rubber elasticity theory based on complex constitutive the system is governed by the Hamiltonian, which is a
relations and the orientation based mechanism for the function of the position and momentum of the particles [13,
analysis of semi-crystalline polymers leading to the aggre- 25]. The Hamiltonian representing the total energy of an
gate model [16]. The fibrils in a nanofibrous material are isolated system is given as the sum of the potential and
found to intertwine to form polymeric nanofibers. The fiber kinetic energy terms and thermodynamic terms, as given by
modulus is obtained using the Northolt and van der Houts P
continuous chain theory [14, 1720]. This is an enhance- Hr N ; pN 2m
1
p2i U r N
ment over conventional homogenization techniques, i 1
E pN ; r N Ek pN U r N
because the effect of shear deformation of the fibrils is not
taken into consideration. The continuum chain formulation where Ur N is the potential energy from intermolecular
P 2
used in this paper gives relationships between the macro- interactions as a function of the spatial ordinate r N ; 2m
1
pi
i
scopic elastic constants and the orientation parameters is the kinetic energy, which represents the momentum pi of
based on the spatial distribution of the nanofibrils. the particle i with mass mi and pi is a function of the
The paper is organized as follows. Section (Atomistic absolute temperature. The time derivative of the
simulation) describes the atomistic simulation of crystalline Hamiltonian gives
123
8846 J Mater Sci (2007) 42:88448852
dH 1 X X oU oE
pi p_ i r_i 0 2 fi t Dt 8
dt m i i
ori ori tDt
and the spatial derivative of the Hamiltonian gives the The elastic constants can be obtained directly from the
equation of motion as variation of the potential for nanofibers [13, 23, 26]. The
total potential energy due to the strain which is the elastic
dH oU strain energy can be expanded as a Taylor series for small
3
dri ori displacements, with the initial position being represented
by the equillibrium position. The elastic moduli tensor can
The reliability of a MD simulation depends mainly on be written as
the type of potential functions used. The total potential of
2 3
the computational unit cell is given by the sum of valence
!
bond energies and the nonbonding interactions 1 X6 2
6 d U 1 dU a b c d 1 dU a c 7 7
Cabcd 6 a a a a d a a 7
X Xh i 2NXa j6i 4 drij2 rij drij ij ij ij ij
bd
rij drij ij ij 5
U tot VijB VijNB 4 |{z}
j j>i Aac 0
9
where VijB is the potential energy due to bonding and VijNB is
the potential energy due to nonbonding interactions. The Here Cabcd is the elastic moduli; U Urij is the potential
force of attraction and repulsion (Fab) experienced by each energy as a function of the interatomic distance rij ; Aac is the
molecule is obtained from the gradient of the potential field internal stress tensor, which at equilibrium is equal to zero;
Wa is the average volume of an atom; N is the number of
oU tot atoms; dab is the Dirac-Delta function; a; b; c; d are the
Fab 5
orab spatial dimensions and aaij is the undeformed value of rij in
the a -direction. The deformation distance is given by
This force is used in calculating the updated position of uaij rij aaij and it is related to the strain by ubij
the atoms and is carried out by the Velocity-Verlet time aaij eab where eab are the components of the homogenous
integration scheme [13]. In this analysis, the time step is infinitesimal strain tensor associated with atoms i and j.
chosen in such a way that the material reaches a metastable The knowledge of structure and molecular motion in
state at a given ambient temperature. This is normally in polymers is essential to the understanding of mechanical
the range of femtoseconds (1015 s), and the local variation and thermal properties. The crystallization behavior of
of the velocity and the kinetic energy about a small PLLA shows that it is a semicrystalline polymer that
increment in time is very small. It has been shown that crystallizes from melt and from solution to form fibers [27].
without a sudden change in the ambient conditions, the Studies on crystal structure of lactide copolymers by var-
molecules vibrate about the mean energy position, which ious studies have shown that the unit cell of PLLA is a
ensures that there is no change in the inherent temperature pseudo-orthorhombic structure a 10:6 A; b 6:1 A;
of the material. At the beginning of each time step of the and c 28:8 A, which is used here [2831]. X-ray dif-
simulation, updated velocities vi (t) are calculated for each fraction experiments and Nuclear Magnetic Resonance
particle by (NMR) analysis for the estimation of the fibrous and crystal
structure of PLLA has shown that the crystalline structure
Dt Dt fi t of PLLA differs slightly [31] from that used by Hoogsteen
vi t vi t Dt 6
2 2 mi et al. [28] and De Santis et al. [29]. However, these
where Dt is the time step in the MD simulation, fi t is the differences are not high enough to cause a change in the
total force acting on particle i at time t and mi is the mass of properties of the PLLA structure. MD analysis of crystal-
the particle. The coordinates of the particles ri t are line PLLA is carried out with the crystal structure and the
updated from the velocities entire computational model is equilibrated to the experi-
mental conditions (see Fig. 1). The minimum energy con-
Dt dition is the starting point for the thermostating analysis,
ri t Dt ri t vi t Dt 7 where the crystal structure is analyzed under the influence
2
of thermal energies. Isothermal strain conditions were
From the updated particle position, the interatomic for- applied to the thermally equilibrated unit cell and the
ces are computed from the first derivatives of the potential elastic constants were obtained using second derivative
energy field E with respect to the atomic coordinates ri elastic constant analysis [13, 32].
123
J Mater Sci (2007) 42:88448852 8847
Micromechanical analysis For a simple EC, the average stresses due to the atomic
ensemble is equal to the average stress due to volume
Direct application of the micromechanical methods for the averaging, establishing the relationship between the mate-
nanofiber raises several questions. Volume averaging of the rial constants derived from the MD simulation and volume
123
8848 J Mater Sci (2007) 42:88448852
123
J Mater Sci (2007) 42:88448852 8849
where k; n ff ; g; . . . ; K 1g, and Sk is the Eshelby of the chains towards the fiber axis. The deformation of the
Tensor for the dispersed inclusions. The effective modulus fiber is taken as the average deformation of a polymer
of the composite (i.e., matrix with inclusions), C, is found chain in the direction of the fiber axis as is shown in
from Fig. 5. Detailed description of the continuous chain model
" # can be found in Northolt [27], and Northolt and van der
X
K1
Hout [14]. We propose to modify the elastic and shear
C Cm I vk Adil 16
k modulus Ec ; Gc of the chain used in this model by the
k1
homogenized elastic and shear modulus Ec S; vS ; vk ;
Gc S; vs ; vk , which are functions of (1) the Eshelby tensor
(S) for the circular inclusions, and (2) the volume fraction
Continuous chain model of polymeric fibers of the Shish and Kebab vS ; vk or an equivalent homoge-
nized structure. Thus, the effective fiber modulus is
Subsequent to the homogenization of the ShishKebab obtained by
model, it is found that a fibril intertwines around other 2
fibrils to form the nanofiber. Tan and Lim [11] have 1 1 sin h E
17
reported that a fibril might terminate by connecting another Efiber Ec S; vs ; vk 2Gc S; vs ; vk
fibril or it may branch into two others. This type of com- where Efiber is the fiber modulus, hsin2 hiE is the strain
plex intertwining cannot be modeled by simple homoge- orientation parameter and is given by the following
nization techniques and therefore a detailed analytical equation. [14, 17]:
procedures need to be considered. The deformation char-
acteristics of an oriented crystalline polymeric fiber have to p=2
R
take into consideration, apart from the mechanical prop- qhcosh sin3 h dh
erties of the material, the molecular arrangement in the 0
sin2 h E 18
nanoscale, and at larger length scales [20, 42]. The model p=2
R
that we use in this paper is based on the analysis of qhcosh sinh dh
0
extension of oriented crystalline fibers called the contin-
uum chain model. However, we extend this theoretical The modified fibril strain can subsequently be written as
formulation by incorporating the effect of the smaller-scale the sum of the elastic strain and the strain due to elastic
material properties by adequate homogenization tech- rotation or shear of the fibrils as given by
niques. It has also been experimentally shown that a
2
polymeric fiber experiences shear deformation when sub- r sin h E r
123
8850 J Mater Sci (2007) 42:88448852
strategy used here is novel as it considers the inhomoge- average Poissons ratio of 0.4 was obtained using MD
neity of the nanofiber and the orientation of the fibrils simulations. The resultant effective modulus of the nano-
[7, 11, 15, 36]. This modeling procedure is carried out by fibrous structure obtained using MoriTanaka method was
using mathematically well-established multiscale modeling 5.77 GPa, by considering about 0.2 volume fraction of the
simulation techniques coupling the atomistic scale to Shish in Kebab in the nanofiber. This obtained effective
macroscopic scales. To the best knowledge of the authors, modulus closely matches with many of the effective
such a methodology of extracting the material properties properties obtained [28, 36, 50]. The effective modulus of
from a completely computational point of view (indepen- the ShishKebab configuration is now used in the modified
dent of experimental data) for nanofibers has not been continuum chain model to obtain the variation of the
attempted. As homogenization methods considering only effective properties of the nanofiber with varying orienta-
the individual aspects of material modeling at different tion parameter values.
scales are available in literature, a multiscale computa- The elastic stiffness calculated using the MoriTanaka,
tional framework is being proposed in this work. The 3D open cell and honeycomb structure models is given in
material properties are extracted from the molecular to the Fig. 6. It can be seen that the predicted stiffness values has
macro level, and finally validated with independent a maximum of 5.9 GPa and decreases with an increase in
experimental results. The uniqueness is in the multiscale porosity. The amount of porosity is also indicative of the
approach proposed here. diameter of the fiber, as the diameter increases the porosity
High strength PLLA fibers of the order of 16 GPa with of the fiber also increases as seen from experimental
high crystallinity and porosity has been produced by dry observations [11]. It can be seen from Fig. 6 that with an
spinning [38]. Leenslag and Pennings have reported a tensile increase in diameter the elastic stiffness would decrease.
modulus of 14 GPa for solution-spun PLA fibers [39]. This method, however, fails to provide an accurate esti-
Numerous studies by researchers have produced high mation of the elastic modulus of nanofibrous materials,
modulus PLLA fibers for various uses, having elastic when compared to experiments (see, for example, Inai
modulus ranging from 1 to 20 GPa. For example, Tan and et al. [1]). The average elastic property obtained by
Lim [36] reported elastic modulus values of 110 GPa, homogenization of the ShishKebab model is used in the
Hoogsteen et al. [28] 16 GPa, Yuan et al. [47] 15 GPa, modified continuum chain model. In this continuum chain
Broz et al. [48] 3.0 GPa, and Cicero and Dorgan [49] model, the homogenized elastic property is predicted by
reported 1.53.0 GPa for different draw ratios. Inai et al. [1] Eq. 17 using the strain orientation parameter derived from
reported the elastic modulus in the range of 2.9 0.4 GPa the birefringence data (Eq. 18). This analytically predicted
for semi-crystalline electrospun polymeric PLLA fibers. elastic modulus of PLLA fibers (as shown in Fig. 7) closely
Most of the above reported values were attained by the matches with the experimental values of Mezghani [54].
estimation of elastic stiffness in tension. Flexural modulus The stress-strain curves, using the continuum chain model
in the range of 69 GPa was obtained by Lim et al. [50]. and the experimentally determined values for different
The Youngs modulus obtained from MD analysis of draw velocities by Inai et al. [1] is compared in Fig. 8.
crystalline lactic acid should conform to the experimental
value of the modulus of crystalline PLLA [51]. The
experimentally obtained elastic modulus for a ~90% crys-
talline PLLA made by a hot drawn (melt spinning) process 8
Mori Tanaka
is 9.2 GPa [51] and a Poissons ratio of 0.44 has been
3D-Open Cell
Elastic Stiff ness ( G Pa)
123
J Mater Sci (2007) 42:88448852 8851
25
References
123
8852 J Mater Sci (2007) 42:88448852
26. Blonski S, Brostow W, Kubyt J (1994) Phys Rev B 49:6494 43. Baltussen JJM, Northolt MG (2001) Polymer 42:3835
27. Kalb B, Pennings AJ (1980) Polymer 21:607 44. Baltussen JJM, Northolt MG, Hout RVD (1997) J Rheol 41:549
28. Hoogsteen W, Postema AR, Pennings AJ, Ten Brinke G (1990) 45. Ruotsalainen T, Turku J, Heikkila P, Ruokolainen J, Nykanen A,
Macromolecules 23:634 Laitinen T, Torkkeli M, Serimaa R, Ten Brinke G, Harlin A,
29. De-Santis P, Kovacs AJ (1968) Biopolymers 6:299 Ikkala O (2005) Adv Mater 17:1048
30. Yamane H, Sasai K, Takano M (2004) J Rheol 48:599 46. Demir MM, Yilgor I, Yilgor E, Erman B (2002) Polymer 43:3303
31. Fujita M, Doi Y (2003) Biomacromolecules 4:1301 47. Yuan XY, Mak AFT, Yao KD (2003) Polym Degrad Stability
32. Namilae S, Chandra N, Shet C (2004) Chem Phys Lett 387:247 79(1):45
33. Somania RH, Yanga L, Zhub L, Hsiao BS (2005) Polymer 48. Broz ME, Vander Hart DL, Washburn NR (2003) Biomaterials
46:8587 24:4181
34. Welch P, Muthukumar M (2001) Phys Rev Lett 87:218302 49. Cicero JA, Dorgan JR (2001) J Polym Environ 9:1
35. Zussman E, Rittel D, Yarin AL (2003) Appl Phys Lett 82:3958 50. Lim JY, Kim SH, Lim S, Kim YH (2002) Macromol Chem Phys
36. Tan EPS, Lim CT (2004) Appl Phys Lett 84:3400 202:2447
37. Thelen S, Barthelat F, Brinson LC (2004) J Biomed Mat Res 51. Fambri L, Pergoretti A, Fenner R, Incardona D, Migliaresi C
69A:601 (1997) Polymer 38:79
38. Leenslag JW, Pennings AJ (1987) Polym Commun 28:92 52. Balac I, Milovancevic M, Tang C-Y, Uskokovic PS, Uskokovic
39. Leenslag JW, Pennings AJ (1987) Polymer 28:1695 DP (2004) Mater Lett 58:2437
40. Fisher FT, Bradshaw RD, Brinson LC (2003) Compo Sci Technol 53. Rappe AK, Casewit CJ, Colwell KS, Goddard WA, Skiff WM
63(11):1689 (1992) J Am Chem Soc 114:10024
41. Mura T (1997) In: Micromechanics of defects in solids. Kluwer 54. Mezghani K, Spruiell JE (1998) J Polym Sci: Part B: Polym Phys
Academic 36:1005
42. Ward IM (1967) Br J Appl Phys 18:1165
123
J Mater Sci (2007) 42:88538863
DOI 10.1007/s10853-007-1812-6
Received: 15 December 2006 / Accepted: 30 April 2007 / Published online: 12 August 2007
Springer Science+Business Media, LLC 2007
Abstract In this paper a generalized anisotropic hyper- two- and three-dimensional networks in skin, see e.g., [2].
elastic constitutive model for fiber-reinforced materials is Collagen fibers are usually crimped or undulated in a nat-
proposed. Collagen fiber alignment in biological tissues is ural state. With increasing strain they line-up, straighten,
taken into account by means of structural tensors, where and finally become the main load bearing elements. As a
orthotropic and transversely isotropic material symmetries consequence, the stress response of soft tissue samples is in
appear as special cases. The model is capable to describe general characterized by highly non-linear behavior with
the anisotropic stress response of soft tissues at large J-shaped or nearly exponential stress-strain curves. Hence,
strains and is applied for example to different types of from a biomechanical perspective, both anisotropy and
arteries. The proposed strain energy function is polyconvex non-linear behavior at large elastic strains should be
and coercive. This guarantees the existence of a global regarded by a constitutive model in order to describe soft
minimizer of the total elastic energy, which is important in tissues appropriately. In the elastic domain, hyperelastic
the context of a boundary value problem. models meet these requirements and have therefore
extensively been utilized in soft tissue mechanics.
The diversity among the mechanical characteristics of
soft tissues motivated a great number of constitutive for-
Introduction mulations for different tissue types. For example, the
reader is referred to [3] for a survey on strain energy
In soft biological tissues the distribution, arrangement and functions for planar biological tissues in connection with
interaction of the constituents lead to a diversity of biaxial testing techniques. Modeling approaches for vari-
mechanical characteristics and thus high specificity and ous components of the cardiovascular system the reader
functionality. In particular, the extracellular matrix plays may find in [1] and references therein. In particular, arterial
an important role. Its main constituents are proteins, gly- wall mechanics as well as several hyperelastic models for
cosaminoglycans as well as bound and unbound water, see arterial tissue are studied in [4]. Several strain energy
e.g., [1]. Among those proteins, different collagen types are functions for myocardium have recently been compared by
crucial for the mechanical properties of soft tissues. Indeed, Schmid et al. [5]. Although the major part of these models
some collagen types form fibers or networks and thus is based on phenomenological approaches, a number of
provide reinforcing structures. By ordered arrangement, micromechanically based models have been suggested. For
this finally leads to anisotropy. The alignment of these example, chain network models (see e.g., [6]), taking into
structures is manifold and reaches from parallel fiber account single collagen fibrils by means of statistical
bundles in tendons to helical arrangement in arteries and mechanics, have been applied to model orthotropic and
transversely isotropic soft tissues [7, 8]. Other authors
derived constitutive relations based on sinusoidal, zig-zag
A. E. Ehret (&) M. Itskov or circular helix representations for the crimped collagen
Department of Continuum Mechanics, RWTH Aachen
University, Eilfschornsteinstr. 18, 52062 Aachen, Germany
fibers, see e.g., [911]. The material parameters appearing
e-mail: ehret@km.rwth-aachen.de in microstructurally based models generally have a
123
8854 J Mater Sci (2007) 42:88538863
physical meaning. However, estimation of these parameters [19] investigated polyconvexity of numerous functions of
might be a difficult task in view of the variance among isotropic and anisotropic strain invariants. On this basis,
individual tissue samples. In general, phenomenological they proposed a variety of polyconvex free energy terms
models are able to adequately describe the macroscopic for a transversely isotropic material as well as an extension
behavior of soft tissues observed in experiments and have for orthotropic materials. Furthermore, for these material
therefore wide applicability. Although the related material symmetries, sufficient conditions for the polyconvexity and
parameters do not usually provide a clear physical inter- coercivity of the strain energy function were elaborated by
pretation, they can be obtained by fitting the model to Steigmann [21]. Itskov and Aksel [22] presented a class of
experimental data. Notwithstanding, even within a phe- transversely isotropic and orthotropic polyconvex and
nomenological approach it is preferable to include as much coercive strain energy functions. An advantage of the latter
information about the structure as possible. Thus, motivated formulation is that the functions fulfill the condition of an
by the histological structure of blood vessels, Holzapfel energy and stress-free undeformed configuration a priori.
et al. [4] considered each layer of an arterial wall as a fiber- Their hyperelastic model is based on a power series rep-
reinforced composite of an isotropic matrix and two sym- resentation with an arbitrary number of terms. A modified
metrically arranged fiber helices. This led to a multi-layer version of the model, based on an exponential function
constitutive model with an elastic potential composed of an representation was successfully utilized to describe the
isotropic part and an anisotropic part associated with the two mechanical behavior of soft collagenous tissues by Itskov
fiber contributions. In a recent paper, Gasser et al. [12] have et al. [23]. Balzani et al. [24] followed the approach by
presented an approach to replace the mean fiber directions in Holzapfel [4] and presented a number of polyconvex strain
the latter model by a collagen fiber dispersion as experi- energies for soft biological tissues consisting of an isotro-
mentally observed by means of polarized light microscopy. pic part and the superposition of several transversely iso-
Besides these biomechanical aspects, the issue of tropic contributions. We remark that some of the material
material stability should be taken into account if the tissue parameters in these models should be restricted further to
is assumed to maintain its integrity throughout the elastic prevent the initial stiffness associated with the fiber part
deformations. Possible criteria for material stability have from being infinite. In the models [4, 24], the transversely
recently been discussed [13]. One of those is the strong isotropic parts are switched off if the associated preferred
ellipticity condition, which also implicates positive defi- directions are under compression. This guarantees
niteness of the acoustic tensor so that the speed of dis- convexity of the strain energy function, however, it leads
placement waves is real for any direction of propagation. to purely isotropic behavior when all fibers are in a
Contrariwise, the loss of ellipticity has recently been used compressive state.
to describe several failure mechanisms in fiber-reinforced In this paper we present a polyconvex anisotropic
materials like, e.g., fiber kinking, splitting and de-bonding hyperelastic constitutive model for materials consisting of
[14, 15]. Investigation of ellipticity for several constitutive an isotropic matrix and reinforced by an arbitrary number of
models (see e.g., [16]) showed, for example, that the fiber families. Each fiber family is explicitly taken into
so-called Fung elastic model [2, 17] widely and success- account by a structural tensor and associated with a weight
fully applied to soft biological tissues is in general not factor. The model is formulated in terms of so-called gen-
elliptic and that enforcement of the strong ellipticity con- eralized structural tensors. It is shown that for a material
dition imposes severe restrictions on the material constants. reinforced by a single family of fibers, these tensors coin-
For polyconvex strain energy functions [18] on the other cide with a recently proposed structure tensor that includes
hand, ellipticity is guaranteed (see e.g., [19]). Another collagen fiber dispersion [12]. The model is developed in a
advantage of polyconvex functions appears in the context generalized form offering maximum flexibility for use with
of boundary value problems. The question whether there is different types of engineering materials and biological
a solution of the boundary value problem is bound to the tissues. Earlier models for orthotropic and transversely
existence of a global minimizer of the total elastic energy isotropic materials are included as special cases.
of a body. According to Ball [18, 20] this minimizer exists The paper is organized as follows: We begin with the
if the strain energy function is polyconvex and satisfies a basic mathematical notations and definitions as well as a
certain growth condition referred to as coercivity. Hence, short discussion of hyperelasticity. Then, the generalized
polyconvexity provides an excellent starting point to for- structural tensors and a functional basis for the anisotropic
mulate strain energy functions that guarantee both ellip- strain energy function are introduced. We proceed with the
ticity and existence of the global minimizer. derivation of the generalized polyconvex and coercive
Instead of checking various anisotropic hyperelastic strain energy function. Finally, numerical examples are
models for polyconvexity, one can formulate polyconvex presented, showing special cases of the model in applica-
strain energy functions from scratch. Schroder and Neff tion to arterial tissue.
123
J Mater Sci (2007) 42:88538863 8855
Preliminaries Hyperelasticity
123
8856 J Mater Sci (2007) 42:88538863
G Q 2 Orth : QLi QT Li ; i 0; 1; . . .; n : 11 For orthotropic and transversely isotropic materials,
another set of generalized structural tensors has recently
For a hyperelastic material, the condition of material been presented [22]. Orthotropy is characterized by sym-
symmetry is written in terms of the strain energy function metry with respect to three mutually orthogonal planes by
(5) and the symmetry group (11) by reflections from which the material properties remain
unchanged. The axes normal to these planes are referred to
WC QCQT WC C; 8Q 2 G: 12 as the principal material directions. Introducing unit vec-
According to Rychlewskis theorem [29], this condition tors li ; i 1; 2; 3; in these directions allows to form the
is satisfied if and only if the strain energy can be structural tensors (see e.g., [22])
represented as an isotropic tensor function of arguments
b 1 l1 l1 ; L
L b 2 l2 l2 ;
containing the structural tensors. In view of (11) one can 16
thus write (see e.g., [22]) b3 I L
L b1 Lb 2 l3 l3 ;
WCL QCQT ; QLi QT WCL C; Li ; where a superposed hat serves to distinguish them from
13 those associated with fiber directions (14). Transverse
i 0; 1; . . .; n; 8Q 2 Orth:
isotropy represents a material symmetry with respect to
At first, we consider a general fiber-reinforced material only one preferred direction. Rotations about this axis and
consisting of an isotropic matrix and an arbitrary number n reflections from planes parallel or orthogonal to it preserve
of fiber families as schematically shown in Fig. 1. Let the the material properties. Specifying this principal material
alignment direction of each fiber family be given by a unit direction by a unit vector l1 ; two structural tensors can be
vector mi, i = 1,2,...,n. Then we define n + 1 structural defined by (see e.g., [22])
tensors Li, i = 0,1,...,n by
b 1 l1 l1 ;
L b 2 1 I L1 :
L 17
1 2
Li mi mi ; L0 I; i 1; 2; . . .; n; 14
3 In a similar manner as (15), linear combinations of the
where I denotes the identity tensor of second order. The tensors (16) or (17) lead to the generalized structural
tensor L0 is assumed to be associated with the isotropic tensors (cf. [22, 23])
matrix. In the next step, we form linear combinations of the
tensors (14) and thus define so-called generalized structural X
m
r b
X
m
r
~r
L wi L i; wi 1; r 1; 2; . . .; 18
tensors by i1 i1
X
n
r
X
n
r
where m = 3 for orthotropy, m = 2 for transverse isotropy
~r
L vi Li ; vi 1; r 1; 2; . . .; 15 and w(r)
i 0 denote scalar weight factors. In fiber-rein-
i0 i0
forced materials orthotropy and transverse isotropy arise as
where v(r)
i 0, i = 0,1,...,n denote scalar weight factors. special cases. Indeed, certain arrangement of two or more
Note that both the structural tensors (14) and the general- fiber families may lead to orthotropic material symmetry
ized structural tensors (15) are characterized by the prop- and one or more fiber families aligned in one single
~ r 1; i 0; 1; . . .; n; r 1; 2; . . ..
erty trLi trL direction result in transverse isotropy (see e.g., [30]). In
this case, the generalized structural tensors defined by (15)
and (18) should coincide so that consequently the weight
factors w(r)
i , i = 1,2,,m, associated with principal material
directions (16, 17) are related to the factors v(r) i ,
i = 0,1,,n, weighting the fiber directions (14) (see
Appendix).
In a recent work [12], the dispersion of collagen fibers
was included into a constitutive model for arterial tissue.
Therein, it is assumed that the fiber distribution is char-
acterized by rotational symmetry about a mean preferred
direction given by the unit vector a0 . As a measure for the
fiber distribution, a scalar parameter j is introduced such
that a so-called structure tensor H takes the form [12]
123
J Mater Sci (2007) 42:88538863 8857
Functional basis 1 F W
WF F W 2 adjF W
3 detF; 24
The right CauchyGreen tensor C and the structural tensors where W i ; i 1; 2; 3, are convex functions of their argu-
(14) form a finite system of tensors. According to Hilberts ment, respectively.
theorem, one can find a finite set of isotropic invariants Vk, We recall the following basic property of convex
k = 1,2,...,t of these tensors called the functional basis, in functions: Let uA : Inv ! R be convex and p : R ! R
terms of which all other isotropic invariants can be be convex and monotone increasing, then p(u(B)) is con-
expressed. Thus in view of (13), one can write vex. Indeed, for some B1, B2 [ Inv and k [ [0, 1], we have
^ V k ;
W W k 1; 2; . . .; t: 20 pukB1 1 kB2 pkuB1 1 kuB2
25
Let us first consider the case where all fiber families are
either parallel or mutually orthogonal. Then, taking into due to the convexity of u and monotonicity of p. Since the
account the symmetry of C and Li, i = 1,2,...,n, and latter is also convex we can further write
applying the classical invariant theory (see e.g., [25, 31]), a
functional basis can be given by pkuB1 1 kuB2 kpuB1 1 kpuB2
26
V 1 trC; V 2 trC2 ; V 3 trC3 ;
which implies convexity of p(u(B)).
V 3i trCLi ; V 3ni tr C2 Li ; i 1; 2; . . .; n;
Now let us consider the convexity properties of the in-
21 variants Vk, k = 1,2,...,t, (21), bearing in mind the
see also [30]. For a general fiber-reinforced material with requirements for polyconvex functions (23). Indeed it can
arbitrary fiber alignment, the functional basis (21) is be shown that the invariants V3+n+i, i = 1,2,...,n, are not
completed by the invariants convex with respect to F, while the remaining ones in (21)
satisfy this condition [19]. However, on the basis of the
tr CLi Lj ; tr Li Lj ; tr Li Lj Lk ; i\j\k 1; 2; . . .; n: CayleyHamilton theorem, the invariants (21) can be
expressed uniquely in terms of other invariants [22]
22
In regard to the constitutive modeling, however, it is Ii trCLi ; Ji trcofCLi ; IIIC detC;
27
useful to restrict the arguments of the strain energy i 0; 1; . . .; n;
function. The invariants (22)2,3 are constants and will be
neglected for this reason. Further, we omit the dependence which are convex with respect to F, adjF and detF,
of the strain energy on the coupled terms (22)1 which respectively [19], and thus provide a suitable basis to
reflect deformations of the fiber families mp and mq formulate polyconvex functions. On account of the positive
relative to each other. Accordingly, in the following, we semi-definiteness of the structural tensors (14), we notice
reduce the list of arguments of the strain energy to the the following important property
invariants (21) even in the case of an arbitrary fiber
Ii [ 0; J i [ 0; i 0; 1; . . .; n: 28
orientation and set t = 3 + 2n.
The convexity properties remain unaffected by forming
linear combinations of the invariants (27)1,2 with non-
Polyconvex strain energy functions negative weight factors (see e.g., [32]). Thus, in view of
(15) one can define generalized invariants based on the
Polyconvexity generalized structural tensors as
X
n
r X
n
r
A strain energy function W W F F : Inv ! R is said to I~r ~ r ; J~r
vi Ii tr CL ~r ;
vi Ji tr cofCL
be polyconvex [18] if there exists a convex function i0 i0
W WF; adjF; detF : Inv; Inv; R ! R such that r 1; 2; . . .; 29
123
8858 J Mater Sci (2007) 42:88538863
which are likewise convex with respect to F and adjF, On the other hand, the generalized representation (31)
respectively. The definitions (27) and (29) finally allow to enables to account for various material characteristics in a
represent the strain energy function in the form straightforward manner by an appropriate choice of any
convex and monotone increasing functions fr, gr and a
~ I~r ; J~r ; IIIC ;
W W r 1; 2; . . .: 30 convex function hr satisfying (34). Clearly, the earlier
models can be obtained as special cases. Indeed, for fiber
Utilizing the additive representation (24), a set of
orientations with orthotropic or transversely isotropic
polyconvex strain energy functions can hence be
symmetry, the power functions
constructed by
h i 1 ~ar 1 ~br
1X s 1=2 fr I~r Ir 1 ; gr J~r Jr 1 ;
W lr fr I~r gr J~r hr IIIC ; 31 ar br
4 r1
1=2 1 cr
hr IIIC IIIC 1 35
where lr 0 are material parameters with the dimension cr
of stress. According to the above mentioned statement recover the strain energy proposed in [22] and showing
about convexity, the functions fr and gr are convex and good agreement with experiments on calendered rubber,
monotone increasing functions of their arguments, while hr where ar 1, br 1, cr > 0 denote material constants.
are convex with respect to III1/2 C = detF. The strain energy Setting
function is given in terms of a series with an arbitrary
h i h i
number of terms s, which offers wide flexibility in appli-
fr I~r a1r ear Ir 1 1 ; gr J~r b1 ebr Jr 1 1 ;
~ ~
cation to experimental data. Nevertheless, truncating the c r
1=2
series after the first term, so that s = 1, is often sufficient to hr IIIC c1 IIIC r 1 ;
r
describe experimental results adequately as, e.g., shown in 36
the numerical examples presented hereinafter.
In the next step, we consider the condition of the energy where ar 0, br 0, cr > 0, yields the exponential
and stress free undeformed configuration. To this end, we model successfully utilized to describe pericardial tissue
set C = I and require and rabbit skin in [23].
1X s
WjCI l fr 1 gr 1 hr 1 0: 32 Coercivity
4 r1 r
Applying (6), the condition of a stress free undeformed For the existence of a global minimizer of the total elastic
state reads as energy of a body, polyconvexity of the strain energy
function is not sufficient. The strain energy function
123
J Mater Sci (2007) 42:88538863 8859
have in consideration of (34) C lr fr00 I~r L
~r L
~r
r1
fr I~r fr 1 fr0 1 I~r 1 I~r 1 ; g00r J~r III2C C1 L
~ r C1 C1 L
~ r C1
gr J~r gr 1 g0r 1 J~r 1 J~r 1 ; r 1; 2; . . .; s: g0r J~r g00r J~r J~r IIIC C1 C1 L ~ r C1
39 C1 L~ r C1 C1
Consequently, in view of (29), the conditions (38)1,2 are 1 1
1
g0r J~r J~r g00r J~r J~r2 h0r IIIC 2 IIIC 2
satisfied if 4
n 1 1
P
s P r h00r IIIC 2 IIIC C1 C1
1
4 lr vi Ii 1 c0 trCp d1 ; 4
r1 i0 40 1 1
1 S
P
s P n
r g0r J~r J~r h0r IIIC 2 IIIC 2 C1 C1
1
4 lr vi Ji 1 c0 trcofCq d2 : 2
r1 i0 1 o
gr Jr IIIC C C1 L
0 ~ ~ r C1 C1 L~ r C1 C1 S :
Bearing in mind (14)2 and (27)1,2 and choosing
p = q = 1, one can rearrange the latter inequalities as 44
n In vivo, biological soft tissues contain a large amount of
1
P
s
r 1
Ps P r
12 lr v0 c0
trC 4 lr vi Ii 1 d1 0; water and thus show a very slight compressibility. In the
s r1 r1
i1 constitutive modeling they are often considered as
1
P r 1
P
s Pn
r
12 l v
r 0 c 0 trcofC 4 lr v i J i 1 d2 0: incompressible. The incompressibility constraint IIIC = 1
r1 r1 i1
can be taken into account setting the constraint function f
41 in (9) to
Taking into account the positive definiteness of Ii and Ji 1
123
8860 J Mater Sci (2007) 42:88538863
123
J Mater Sci (2007) 42:88538863 8861
experiment axial
experiment circumferential
80 model
60
Cauchy Stress [kPa]
Adventitia
60 =20.265 kPa
=86.340 40
=77.521
40 1=0.082
=69.552 20 experiment 0.5:1
experiment 1:1
nd
20 experiment 1:0.5
model
0
0 0 0.2 0.4 0.6 0.8 1
1 1.05 1.1 1.15 1.2 1.25 Green-Lagrange Strain E
Stretch
60
rubber materials (see e.g., [38]). For this reason, we choose experiment 1:0.5
model
the functions fr and gr in a logarithmic form motivated by 0
the Gent model for rubber elasticity [39]. Setting s = 1 in 0 0.2 0.4 0.6 0.8 1
Green-Lagrange Strain Ez
(36), this leads to
Fig. 3 Comparison of the polyconvex model (53) with experimental
l I~ 1 K~ 1
W a ln 1 b ln 1 ; 53 data [37]. Material constants: l = 81.048 kPa, a = 2.060, b = 0.408,
4 a b v1 = 0.498, u = 47.59
123
8862 J Mater Sci (2007) 42:88538863
0 1 1 1
prediction axial 3
0 1
r b 2
r b
0 0.2 0.4 0.6 0.8 1 1 2v1 L 1 1 v1 L 2:
Green-Lagrange Strain E, Ez
3 3
Fig. 4 Experimental data [37] and prediction by the polyconvex In a fiber reinforced material, orthotropy may be the
model (53) result of different fiber configurations. We first consider the
case where fibers are aligned in three mutually orthogonal
fiber directions coinciding with the principal material
These are, however, considered negligible in comparison to
directions, so that li mi ; i 1; 2; 3. Then, in view of
the variation in data among individual tissue samples.
(15,16,18) one obtains
3 h
X i
Conclusions ~r 1
L
r r r r b
1 v1 v2 v3 3vi L i:
3 i1
In the present paper we have presented a generalized A material with two orthogonal fiber families is
polyconvex and coercive strain energy function for fiber- likewise orthotropic, where the principal directions are
reinforced materials. The model is able to take into account given by the two fiber directions, so that l1 m1 and
the collagen fiber structure of soft biological tissues. The l2 m2 , and the direction normal to the plane in which the
strain energy function is given in a generalized form fibers lie. The generalized structural tensors (18) are thus
offering wide flexibility in application to different types of given by
tissue. For example, the model agreed appropriately with
the mechanical behavior of arterial tissue in various h i h i
~ r 1 1 2vr vr L
L b 1 1 1 vr 2vr Lb2
experiments. The agreement was achieved with a small 3 1 2
3 1 2
123
J Mater Sci (2007) 42:88538863 8863
123
J Mater Sci (2007) 42:88648872
DOI 10.1007/s10853-007-1784-6
Received: 15 December 2006 / Accepted: 17 April 2007 / Published online: 17 July 2007
Springer Science+Business Media, LLC 2007
Abstract Microtubules (MTs) are fundamental structural with data obtained by other previous experimental and
elements in the cytoskeleton of eukaryotic cells. Their theoretical studies. We report here a Young modulus of
unique mechanical properties depend on the properties of 1.66 GPa for the MT under axial tension. In perspective
the tubulin dimer, its interactions with surrounding dimers our approach provides a simple tool for the analysis of MT
and the geometric organization within the MT. While the mechanical behaviour under different conditions.
geometry has already been well described in experimental
works, the mechanical characteristics of the dimer as well
as of the individual monomers have up to date not been
described. These may therefore provide new, additional
insight to the microtubule tensile properties. In this paper Introduction
we construct a mesoscale model of MT with a bottom-up
approach. First, we evaluate the elastic constants of each of Microtubules (MTs) are long, hollow cylinders made of
the two monomers together with the interaction force be- protofilaments which bind together laterally along the long
tween them by means of molecular dynamics (MD) sim- axis of the cylinder and form the microtubule wall [1]. The
ulations carried out in an explicit water environment. Using elemental base of the protofilaments is the tubulin hetero-
the MD results, we model a 1 lm long MT as a cylinder dimer which consists of an a- and a b-tubulin monomer.
constituted by interacting elastic elements and examine its Both are globular molecules with a relatively high degree
properties via finite element method (FEM). The obtained of similarity from the primary to the tertiary structure level.
results show an elastic constant value for a-tubulin of Their orientation in the dimer is almost the same, and
11 N/m, while for the b-tubulin the elastic constant was stacked together on top of each other as they are in the
measured to be 15.6 N/m. Concerning interactions between protofilament they seem very alike [2]. However, the
neighbouring monomers, the elastic constants along the structures do also have individual features that separate
protofilament (45 N/m for the intra-dimer interface and them. One important difference is that in the MT the
18 N/m for the inter-dimer interface) are more rigid than a-monomer binds a GTP molecule while the b-monomer
elastic constants calculated for lateral interfaces (11 and binds a GDP molecule.
15 N/m). The mesoscale model provides mechanical Different MT configurations exist depending on the
properties of the whole MT, thus allowing the comparison number of protofilaments included in the MT wall. The
number of protofilaments in a MT observed in-vivo and
in-vitro varies widely, from 8 to 19 [3], making the MT
M. A. Deriu (&) F. M. Montevecchi
Department of Mechanical Engineering, Politecnico di Torino, structure highly polymorphic. Most cellular microtubules
Corso Duca degli Abruzzi, 24, Torino 10129, Italy have 13 protofilaments.
e-mail: marco.deriu@polito.it In microtubules the protofilaments bind together later-
ally with a slight shift generating a spiral with a pitch of 2,
S. Enemark M. Soncini A. Redaelli
Department of Bioengineering, Politecnico di Milano, 3 or 4 monomers length [3]. For MTs with 3 start-helices,
Milan, Italy the neighbouring monomers form spirals of all-a- and
123
J Mater Sci (2007) 42:88648872 8865
all-b-tubulins except at the seam of the microtubule where desired resolution at the single monomer level needed for a
a-tubulins bind laterally to b-tubulins and vice versa. For thorough understanding of the MTs mechanical behaviour.
MTs with 2 or 4 start-helices, there is no seam and the On the theoretical level several attempts have been
spirals consist of all-a- and all-b-tubulins. made to produce a bottom-up approach corresponding to
Microtubules are essential structural elements present in the one here presented. Such attempts include finite ele-
all eukaryotic cells. They form motor protein tracks ment models, but so far all these models have been based
which are used for directed transport within the cell and on assumptions on how the interactions vary with distance
they are part of the spindle apparatus used to move and [911]. In particular these works use reported experimental
segregate the chromosomes during cell division. They also values of Young modulus and shear modulus to define the
play a purely mechanical role for the cell maintaining its interactions between the mechanical properties of the
shape. building blocks of the MT.
Despite the many experimental and computational In this work we use molecular dynamics (MD) to obtain
efforts the mechanical properties of the microtubule are still mechanical characteristics of the different types of inter-
debated. The experimental methods included optical twee- actions present in the MT both between and within
zers [4, 5], hydrodynamics flow [6], atomic force micros- monomers. The characteristics are used to create a simple
copy (AFM) [7], and thermal fluctuation and viscosity MT model with finite elements methods (FEM) from which
measurements [8]. The reported values of Youngs modulus we are able to evaluate the overall MTs mechanical
vary between 0.1 GPa [7] and 2.5 GPa [5]. Gittes et al. [8] properties. In particular, we demonstrate that it is possible
estimated a length independent MT flexural rigidity of to make a full description of the microtubule using MD
2.151023 Nm2 based on thermal fluctuations in shape. combined with the available experimental data on the
From this value they calculated the Youngs modulus for atomistic structure.
MTs to be 1.2 GPa using the assumption that the MT is both
homogeneous and isotropic in structure. The length of the
MTs in the experiments ranges between 25 lm and 65 lm. Methods
MTs of this length are typically found in cilia.
It should be remarked that both the isotropy assumption Molecular models and equilibration of the structures
and the length independence are now generally questioned. for MD simulations
In a more recent experiment by Kis et al. [7] the elastic
deformation of MTs bound to a surface with holes of dif- With the aim to characterize the tubulin monomers and all
ferent sizes was directly measured with AFM. The AFM tip the monomer-to-monomer interactions involved in the MT
was used to apply a force on the MT and the deflection was structure, we used the atomic structure of ab-tubulin,
measured and used to estimate simultaneously both 1TUB.pdb [2], available from the RSCB Protein Data Bank
Youngs and shear moduli. The results show a significant [12]. Due to a large variation among monomer isotypes
difference between the Youngs modulus (Y) of 100 MPa present in the crystallographic assay a smaller part of the
and the shear modulus (G) of 1.4 MPa, evidencing that the C-terminal is missing in both monomers. Other pdb
MT is highly anisotropic. structures exist [13] but were not taking into consideration
These results indicate that the MT flexural rigidity is not since a larger number of amino acids are missing despite
length independent and consequently, it does not seem to the fact that some of them have a better resolution.
be an appropriate parameter to describe the MT mechanical The chosen structure was modified before use. A missing
behaviour at least not for sufficiently short MTs (few lm Mg2+ ion present in other structures was added based on data
long MTs). For short microtubules the sliding between contained in the refined structure 1JFF.pdb [13]. Further-
adjacent protofilaments is relevant during MT bending. On more, a TaxolTM molecule was removed. TaxolTM is used as
the contrary very long MTs (2565 lm) are more rigid and a stabilizing agent for crystal formation of the microtubules
the Youngs modulus dominates the mechanical behaviour but is not present under physiological conditions.
since only a slight sliding occurs between adjacent proto- Gromacs 3.3.1 software [14] with the GROMOS96 43a1
filaments in bending [9]. Thus, the value reported by Gittes force field was used to perform all the simulations using
et al. [8] might result in a correct evaluation of the Youngs cut-offs of 1 nm for non-bond interactions (van der Waals
modulus of the MT, since neglecting the shearing for and electrostatic). The time step was set to 2 fs for all the
sufficiently long MT merely introduces small errors. MD simulations.
However, even at their best, these measurements are not The structure was first energy minimized in vacuum
able to identify the origin of the shear and Youngs mod- before it was centred and orientated along the x-axis in a
ulus values on the basis of the different forces between and rectangular box of size 1898 nm with periodic boundary
within the monomers. Therefore, they cannot provide the conditions. The rest of the box was filled with Single Point
123
8866 J Mater Sci (2007) 42:88648872
Charge (SPC) [15] water molecules to explicitly model For each system, the interaction interfaces were identi-
water in the system. To balance the strong negative charge fied and then created on the basis of the contact surfaces
of the dimer and neutralize the total system charge 40 Na+ described in literature [17]. After the set up of the relative
ions were added to the solution. position of the two monomers, the water molecules and
The entire system was then energy minimized again and ions were added and each system was heated and equili-
after that heated to 300 K by coupling it to an external heat brated for 200 ps adding a position restraint in the form of
bath for 35 ps [16]. Finally the system was left to equili- a harmonic potential (Vpr) to each Ca-atom:
brate for 800 ps in order to stabilize the structure in terms
1 2
of temperature and energy oscillations. Vpr rCa;i = kpr rCa;i t rCa;i 0 1
The equilibrated structure was the basis for all further 2
dynamics done for the mechanical characterization.
where kpr is the elastic constant of the harmonic potential
and rCa,i is the position of the Ca atom of the i-th residue.
Mechanical characterization of interactions between
For each molecular system several configurations were
monomers
generated with different distances (d) between the centres
of mass (CM) of the two monomers. A pulling MD pro-
The mechanical characterization was done with MD simu-
cedure was used to generate different initial configurations
lations in which the interaction energy between two
moving the monomers (Dd = dd0) along the line con-
monomers was measured at different distances. On the basis
necting the two CMs (Fig. 2) closer and apart with respect
of the previously mentioned equilibrated structure four
to the initial position (d0).
different molecular systems were generated, each consisting
For each distance the interaction energy term, i.e., the
of two monomers: the intra-dimer (ab-interaction), inter-
sum of the Coulomb and van der Waals terms, was ob-
dimer (ba-interaction) and lateral (aa- and bb-interaction)
tained by running a MD simulation for at least 100 ps
complex (Fig. 1).
while the protein backbone structure was maintained
restraining the monomer Ca-atoms. This prevents marked
structural changes in the monomer shape, which are con-
sidered in the single monomer testing.
During the MD simulation, the monomer-to-monomer
interaction energy and CMs distance were sampled every
0.2 ps. The last 50 ps are considered to determine the mean
value of interaction energy and CMs distance for each
configuration.
123
J Mater Sci (2007) 42:88648872 8867
The mean value of the interaction potential energy (Vint) MD simulations moving S and S with a velocity equal to
between the two monomers as a function of the mean value 5104 nm/ps.
of the CMs distance (d) was fitted with a 3rd order poly- The force F(t) applied to the molecule is given by:
nomial function. Consequently, the force (Fint) can be
obtained as first derivative with the opposite sign of the F t ks jxS t xP tj jxS0 t xP0 tj; 2
interaction potential energy with respect to the distance.
The elastic constant (kint) is calculated as second derivative where xS(t), xS(t), xP(t) and xP(t) are the CM positions of S,
of the interaction potential energy. S, P and P, respectively. The spring forces at the begin-
ning of the simulation are equal to zero since the free ends
of each spring and the CM of its corresponding pull group
Mechanical characterization of single monomers have the same position.
The deformation e(t) in elongation and compression
The mechanical properties of a- and b-monomers were tests was calculated as:
tested using an AFM-like approach in which springs were
applied to deform the monomer (Fig. 3). DLt Lt L0
et ; 3
First, water and ions were removed from the equili- L0 L0
brated structure and two independent systems consisting of
where L(t) is the monomer length as function of time
a- or b-tubulin were extracted. Single monomer systems
calculated as the distance between the CMs of groups P and P
were then solvated by adding SPC water molecules and
neutralized with Na+ ions. Then the system was energy
minimized, heated to 300 K and equilibrated for 200 ps Lt jxP t xP0 tj 4
while all the Ca-atoms of the monomer were restrained (see
Eq. 1). and L0 is the initial monomer length, i.e., for t = 0.
Two pull groups (P and P) were defined, selecting In all tests structure deformations of 10% were carried
Ca-atoms of the residues involved in the longitudinal out.
interactions between adjacent monomers along the pro- The elastic constant (km) of the monomer then is given
tofilament. A spring was connected to the CM of each by the Hooks law:
group and characterized by an elastic constant (ks) equal to
F t
10 nN/nm. The spring elastic constant value was chosen on km m a; b 5
DLt
the basis of a preliminary ks sensitivity analysis. The value
of ks = 10 nN/nm represents a good compromise between which corresponds to the slope of the F-Dl curve at the
low ks values that require very time consuming pulling time t.
simulations and high ks values which allow faster simula-
tions but introduce large oscillations in the force values.
During MD simulation, the free end of each spring (S and MT mesoscale model
S) was moved with a constant velocity (vs) along the x-axis
corresponding to the MT longitudinal direction (Fig. 3). The mesoscale model of an entire microtubule was built
Compression and elongation tests were carried out for 2 ns using a finite element (FE) approach. a-tubulin and
b-tubulin monomers and all the interactions between
monomers in the microtubule lattice were represented as
springs. The mechanical properties were set on the basis of
the results obtained with MD simulations. In this way the
microtubule was simulated as a network of springs.
In order to simplify the microtubule model, thus
reducing the number of elements along the protofilament,
two different springs k~ab and k~ba were defined representing
the mechanical behaviour of ab interface (ba interface)
together with the elastic properties of half of the related
monomers (Fig. 4a, b):
Fig. 3 Scheme of mechanical test for the monomer. The CMs of pull 1
1 1 1
groups (P, P) are each connected to a spring. Constant rate k~ab 6
displacements are imposed to the free end of the springs (S, S) 2ka 2kb kab
123
8868 J Mater Sci (2007) 42:88648872
123
J Mater Sci (2007) 42:88648872 8869
1000 Table 1 Minimum interaction energies (Vmin) together with the dis-
tance between the CM of the two monomers, maximum forces of
t (ps) interaction (Fmax) between the two monomers occurring when they
0 d=5.78 nm are dmax apart, and elastic constants (kint) for the interaction evaluated
20 40 60 80 100
at the interaction energy minimum distance dmin
Vint (kJ/mol)
-1000 ab ba aa bb
d= 4 . 2 n m
- 1 0 00
over time is given by the difference between the position
- 2 0 00
of the free end of the spring and the corresponding pull
group.
- 3 0 00 The monomer contraction and elongation (Dl(t); Eq. 3),
and the force (F(t); Eq. 2), imposed by the spring were
- 4 0 00
calculated based on the filtered data from the simulation
-5 00 0 output. The curves showing F(t) versus Dl(t) for a-tubulin
and b-tubulin in both pulling and compression tests are
-6 00 0 reported in Fig. 9. Positive Dl(t) values imply an elongation
of the monomer while negative Dl(t) values correspond to a
Fig. 6 Interaction energy as function of distance between monomers.
Crosses: ab. Triangles: ba. Squares: aa. Diamonds: bb. The curves in
compression.
grey tone scale from dark to light show the fitted lines in the same A straight tendency line is superimposed on the
order F(t)Dl(t) curves for each monomer. Based on Eq. 5, the
123
8870 J Mater Sci (2007) 42:88648872
S 14
0.3
12
0.2
k (pN/nm x 10 3 )
P
10
0.1
8
x(nm)
0
100 200 300 400 500 600 6
-0.1 t (ps)
4
-0.2 P 2
-0.3
S -1.0 -0 .5 0 0.5 1.0 1.5
l(mn)
Fig. 8 Relative motion (x(t)) of the two pull groups (P, light grey; P,
dark grey) and the one free spring end as a function of time during the Fig. 10 Mechanical properties of the spring elements along the
a-tubulin elongation test (black) protofilaments calculated on the basis of MD data (k~ab (diamonds) for
the inter- and k~ba (squares) for the intra-dimer interactions)
Discussion
slope of the tendency lines provides the elastic constant
values ka = 11 N/m for the a-monomer model and The aim of the present paper is to increase the under-
kb = 15.6 N/m for the b-monomer model. standing of the MTs mechanical behaviour starting from
The b-tubulin model seems to be slightly more rigid available knowledge about the dimers molecular structure
than the a-tubulin model, despite the monomer similarity. and the geometry of the MT lattice. A new bottom-up
However, the value of km is affected by the total elongation hierarchical approach was developed. The basic building
considered for the analysis. In particular, considering a blocks and their interaction interfaces were identified con-
strain of the monomer up to 5% in elongation and com- sidering the structure of the MT as reported by electron
pression the elastic constant value for both a- and b-tubulin crystallography analyses [2, 13]. The mechanical properties
is in the range 1516 N/m. of the building blocks and their interactions were calculated
123
J Mater Sci (2007) 42:88648872 8871
through MD simulations. The results together with the data Limitations in the method used to obtain the interactions
on the permitted MT lattice structure obtained from litera- between monomers include the positioning and interaction
ture [3] were used as input for a FE model. Using advanced path used. Different permitted MT configurations like 10:2
FE structural code, a spring elements model was used to or 13:3 should ideally be constructed with different posi-
investigate the overall MT behaviour under axial tension. tions and interaction paths of the monomers. Simplifying,
The results obtained are directly comparable to results from we study only the interactions along two perpendicular
MT experiments [48]. In contrast, no direct experimental paths: along the protofilament and laterally. In correspon-
measurements on the dimer are available for comparison. dence, the original structure, 1TUB.pdb, taken from the
Previously, experimental works have estimated the dimers RCSB Protein Data Bank was obtained from protofilaments
mechanical properties starting from the MT structure [7, 8]. in the 2-dimensional planar zinc-sheet configuration. A full
Apart from the difficulties in separating the dimers investigation in 3 dimensions could be interesting for a
mechanical properties from the lateral and inter-dimer later model refinement.
interactions, these experiments cannot provide detailed In order to measure the interaction energy a novel
knowledge about the individual MT building blocks and approach was used rather than a constrained potential of
their specific interactions. The present work in contrast mean force (PMF). Preliminary docking tests demonstrated
takes the investigation of the MT to the molecular level. that using a distance constraint method or a harmonic
Also our model has some limitations. The single umbrella potential acting on the CMs of the monomers has
monomer AFM-like measurements are done at a particular limitations in our case due to the globular shape and size of
velocity. However, globular protein structures are known to the investigated molecules. Indeed when arranging the two
have viscoelastic properties, consequently the strain rate monomers very close to each other (e.g., with a harmonic
imposed can influence the obtained results. Strain rates umbrella potential acting on each monomers CM) their
imposed in typical AFM pulling/compression experiments structures, squeezed by the high repulsion forces, widely
on proteins are in the order of 109 nm/ps [18, 19], while deformed during the simulation. In this way the real dis-
we impose a velocity of 5104 nm/ps. tance between the outer surfaces of the two docked
However a preliminary analysis aimed at evaluating monomers cannot be controlled.
how the velocity influences the obtained elastic constant Our approach, which restraints each Ca-atom (resulting in a
values (data not published) resulted in no substantial dif- backbone restraint) allowed the structure of each monomer to
ferences between force values (and thus in elastic constant maintain its shape throughout the whole simulation while
values) calculated at a strain rate of 103 nm/ps with lateral groups of each residues on the surface moved and
respect to force values calculated at 5104 nm/ps. rearranged their positions. The spring constants describing the
The atoms chosen for the pull groups and the use of the interactions between the monomers correspond with knowl-
position restraints introduce some arbitrary choices that edge about the interactions as described in literature [2, 17]
require further discussion. Different parts of the mono- and with experimental evidence concerning the dynamic
mers surface structure have different mechanical proper- instability of the MT structure [21]. The general opinion is that
ties, and this is also valid for different parts of the pull during the MTs disassembly, protofilaments break the lateral
group which contains loops and regions weaker than contacts between the tubulin dimers and peel off from MTs
others. Without taking this into account and without a structure. Subsequently, free protofilament segments then
proper pull group definition a pulling approach would only break at the inter-dimer interfaces splitting into ab-tubulin
report the mechanical behaviour of the weakest region of dimers. On the contrary, the ab-tubulin dimer is known as a
the surface. In our work we try to estimate an average very stable structure and can be dissociated only by means of
behaviour of the whole monomer by maintaining the severe chemical treatments [1]. In accordance, our
interface surface fixed. The position restraints used serve results show that lateral (Fmax = 6.3 nN for aa-interaction
precisely this purpose. and Fmax = 4.1 nN for bb-interaction) and inter-dimer
Our approach differs from experimental AFM mea- (Fmax = 6.0 nN) interactions are much weaker than the
surements [18] and simple AFM simulations with the interaction between the monomers within the dimer
spring attached to a single atom [19, 20]. In these studies (Fmax = 11.9 nN). In addition, we note that no substantial
force values are typically of order 100 pN, while we report difference is found between the values measured for lateral
values of about 1 nN. Preliminary tests demonstrated that and inter-dimer interactions. Not surprising the energy needed
the choice of the Ca-atoms of residues used to set the pull to pull the two monomers within the dimer apart (ab-inter-
groups (P and P) influences the output force values. In action) is much greater than the energy needed to pull any of
particular, moving the entire interface instead of a single the other dimers apart (Table 1).
atom (as in typical AFM simulations) results in higher It is important to observe that the force-distance relations
calculated force values. reported in Fig. 7 were calculated based on interaction
123
8872 J Mater Sci (2007) 42:88648872
potential energy curves fitted with a 3rd order polynomial 3. Chretien D, Wade RH (1991) Biol Cell 71:161
function. With this approximation, the force-distance 4. Felgner H, Frank R, Schliwa M (1996) J Cell Sci 109:509
5. Kurachi M, Hoshi M, Tashiro H (1997) Cell Mot Cyt 38:201
curves are valid around the minimum value and are not 6. Venier P, Maggs AC, Pantaloni D (1994) J Biol Chem 269:13353
applicable far from this minimum value. In particular, it is 7. Kis A, Kasas S, Babi B, Kulik AJ, Benot W, Briggs GAD,
clear that there is no repulsion force when the monomers are Schonenberger C, Catsicas S, Forro L (2002) Phys Rev Lett
far apart, and the force should be go zero for large distances. 89:248101
8. Gittes F, Mickey B, Nettleton J, Howard J (1993) J Cell Biol
The mesoscale model here developed is based on a 120:923
simple representation of the MT structure. Indeed, the 9. Kasas S, Kis A, Riederer BM, Forro L, Dietler G, Catsicas S
monomers deformations and interactions are accounted (2004) Chem Phys Chem 5:252
for by means of mono-dimensional elements. The meso- 10. Janosi IM, Chretien D, Flyvbjerg H (2002) Biophys J 83:1317
11. VanBuren V, Cassimeris L, Odde DJ (2005) Biophys J 89:2911
scale model of the MT does not include thermal fluctua- 12. Berman HM, Westbrook J, Feng Z, Gilliland G, Bhat TN,
tions. However considering the persistence length of the Weissig H, Shindyalov IN, Bourne PE (2000) Nucl Acid Res
MT about 16 mm [1] the bending effect of thermal fluc- 28:235
tuations on 1 lm long MT can be neglected. 13. Lowe J, Li H, Downing KH, Nogales E (2001) J Mol Biol
313:1045
The FE analysis of a 1 lm long MT under axial tension 14. Lindahl E, Hess B, van der Spoel D (2001) J Mol Mod 7:306
allowed calculating mechanical characteristics of an entire 15. Berendsen HJC, Postma JPM, van Gunstersen WF, Hermans J
MT. The Young modulus obtained using our bottom-up (1981) In: Interaction model for water in relation to protein
approach is 1.66 GPa in agreement with data reported in hydration. Reidel Publishing Company Dordrecht, Dordrecht, p
331
literature. Up to date, all the experimental works in the past 16. Berendsen HJC, Postma JPM, DiNola A, Haak JR (1984) J Chem
produced a wide range of values for mechanical charac- Phys 81:3684
teristics of the microtubule (0.12.5 GPa) [48], and the 17. Nogales E, Whittaker M, Miligan RA, Downing KH (1999) Cell
most of these studies reported a Young modulus of 96:79
18. Lenne PF, Raae AJ, Altmann SM, Saraste M, Horber JKH (2000)
12 GPa. FEBS Lett 476:124
19. Altmann SM, Grunberg RG, Lenne PF, Ylanne J, Raae AJ,
Acknowledgments This research has been supported by the EST Herbert K, Saraste M, Nilges M, Horber JKH (2002) Struct
Marie Curie programme contract number MEST-CT-2004-504465 10:1085
and by the Active Biomics STREP project contract number NMP4- 20. Soncini M, Vesentini S, Ruffoni D, Orsi M, Deriu MA, Redaelli
CT-2004-516989. A (2006) Biomech Model Mechanobiol Nov 18;[Epub ahead of
print]
21. Erickson HP, Stoffler D (1996) J Cell Biol 135:5
References
123
J Mater Sci (2007) 42:88738884
DOI 10.1007/s10853-007-1792-6
Received: 12 December 2006 / Accepted: 20 April 2007 / Published online: 7 August 2007
Springer Science+Business Media, LLC 2007
Abstract Surface remodeling of biological tissues kinetics is higher. This reinforces the need for proper ref-
through tissue growth or dissolution is deemed critical to erence frame definition for experimental strain measure-
their proper functioning, and is influenced by the defor- ments.
mation of the tissues during physiological activities. The
present work attempts to develop a constitutive framework
for deformation modulated surface remodeling of biologi-
cal tissues. The framework is developed assuming finite
deformation of the tissue, and the effect of deformation on Introduction
the driving force for surface remodeling is determined from
thermodynamic principles. The microscopic trends are Many biological tissues continuously undergo remodeling
upscaled to yield the remodeling-induced change in a (growth or resorption) during their natural life spans. For
macroscopic porous tissue. By way of application, the example, about 25% of trabecular bone and 3% of cortical
effect of deformation on the remodeling kinetics is deter- bone in a mature adult is remodeled annually. The tissue
mined for an incompressible elastic tissue. Depending on remodeling phenomena has important consequences, and is
the ratio of the specific elastic stiffness and the specific deemed critical for a number of physiological events
Gibbs energy variation induced by the cell, the effect of including the healing of bone fracture [13], wound healing
deformation on the remodeling kinetics can be significant. in the skin [4, 5], hypertrophy of the heart (i.e., size in-
It is found that both tensile and compressive deformation crease in muscles of an overloaded heart, [6]) and changes
aid tissue dissolution (and dissuade growth). However, the of pulmonary blood vessels in hypertension [7]. The tissue
magnitude of the effect is found to be different under remodeling mechanism and kinetics are affected by the
tensile and compressive loadings, and critically depends on tissue deformation during the remodeling process, e.g., the
the reference frame used for the strain measurements. For contractile forces generated during skin wound contraction
Lagrangian strain measures (e.g., stretch, engineering is associated with scar tissue formation [8], and the
strain), the increase in the dissolution kinetics per unit importance of in-vivo stress in bone healing is generally
strain is higher under compressive loadings. On the other accepted, though the exact nature of the effect is uncertain
hand, for Eulerian strain measures (e.g., logarithmic or true [1, 9]. Thus modeling of deformation-coupled remodeling
strain), the effect of tensile loading on the dissolution processes in biological tissues is important.
The interaction of the remodeling characteristics and the
stress/deformation in biological tissues have been numeri-
P. Ganguly cally examined by several investigators [1013]. Skalak
Schlumberger Doll Research, Cambridge, MA 02141, USA [10] showed that residual stress may be developed due to
non-uniform (and incompatible) growth in biological tis-
F.-J. Ulm (&)
sues. Hoger and co-workers [1113] have modeled defor-
Massachusetts Institute of Technology, 77 Massachusetts Ave.,
Cambridge, MA 02139, USA mation coupled growth by introducing a multiplicative split
e-mail: ulm@mit.edu of the total deformation gradient (i.e., due to both
123
8874 J Mater Sci (2007) 42:88738884
deformation and growth) into its elastic deformation and these materials can be conveniently divided into two length
growth parts, and have analyzed problems related to the scales: (a) the microscale (see Fig. 1b), which is typically
hypertrophy of the heart and growth in bones under stress. at the length scale of individual struts or pores for the
The analysis is similar to the multiplicative split of the porous structures, and the relevant scale for biochemical
elasto-plastic deformation gradient into the elastic and phenomena related to remodeling processes, and (b) the
plastic components in conventional finite deformation macroscale, which is at the scale length of the entire porous
elasto-plastic constitutive analysis (e.g., [14, 15]), and tissue, and is the relevant length scale for determining the
estimate the growth phenomenon in terms of its effects on a effect of remodeling on tissue characteristics. Typically the
given set of material points (i.e., the analysis is based on a length scales corresponding to the micro and the macro
material reference frame). On the other hand, the remod- scale in porous biomaterials differ by one or two orders of
eling phenomenon involves growth or resorption of the magnitude. For example, in porous trabecular bone, the
biological tissue, and thus, implicitly involves a change in microscale corresponding to the size of individual trabec-
the number of material points (increase for growth, ula and individual pores are of the sizes 100500 lm [16,
decrease for resorption), and a continuously variable set of 17], while the porous structure, corresponding to the scale
material points. Hoger and co-workers have resolved this of the trabecular bone samples generally used in experi-
anomaly by assuming growth to be volumetric and dis- ments, is 10100 mm in size [18]. Similarly, in problems
tributed, i.e., no volume element is destroyed or created of growth in skin tissues, the length of individual struts and
during the growth process, rather the existing volume the pore diameter is approximately 100 lm [5], while the
elements are extended or compressed as new material scaffolds used in experiments are 1100 mm in size [5]. It
points are added or some material points disappear. Such is noteworthy that in both cases, the material addition (or
an approach is effective for volumetric remodeling pro- removal) occurs at the lower scale length. The size of the
cesses in biological tissues. However, in surface remodel- osteoclast cells responsible for resorption in bone is 50 lm
ing, the appearance of new material points (or the [19], while the fibroblast cells involved in skin growth are
disappearance of old ones) determine the characteristic of 20 lm in size [5]. In order to predict the effect of
the newly formed surface, and the growth process is spa- remodeling (which occurs at the microscale) on the mac-
tially discontinuous. Such situations arise in a number of roscopic tissue characteristics, consistent upscaling
biomaterial remodeling scenarios, including trabecular schemes need to be developed. Such upscaling may be
bone growth/resorption under osteoblast/osteoclast cell inconvenient when a multiplicative split of the overall
action at the bone surface and wound healing in skin tissues deformation gradient is assumed, since the estimation of
(which involve the adhesion of the fibroblast cells at the the macroscopic deformation and growth components of
surface of the skin matrix). The applicability of the mul- the deformation gradient from the microscopic measure-
tiplicative split in such situations is questionable. ments is not straightforward.
Another characteristic of the trabecular bone/skin The focus of the present work is the analysis of the
remodeling problem deserves mention at this point. Both remodeling phenomenon (at different length scales) in
bone and the skin are porous tissues, composed of struts porous biological tissues. The problem arising from the
and pores (see Fig. 1a). The remodeling phenomenon in variability of the material points in such systems is
Fig. 1 (a) The porous tissue at the macroscale, consisting of solid area of the solid strut is As, n is the normal to the solid surface, the
struts and fluid saturated pores. The biological cells are shown solid-cell interface area is f and u is the velocity. A negative u:n at the
attached to the solid. (b) Schematic representation of the microscale, solid-cell interface signifies dissolution of the solid and a positive u:n
where the multinucleated biological cells may attach at the surface of signifies growth
the solid strut, and initiate the remodeling phenomena. The surface
123
J Mater Sci (2007) 42:88738884 8875
circumvented by developing the theory in terms of a spatial gradient of the ud vector (i.e., gradud oud =ox). This
coordinate system. The model is developed both w.r.t. the defines the particle velocity at any spatial point inside the
undeformed reference frame (i.e., at time t = t0), and the solid (approaching the solid-cell interface), and un ud
deformed (or current) reference frame.The present model (see Fig. 2a). On the other hand, the solid-cell interface
follows the analytical lines earlier employed by Silva and velocity is defined by the rate of mass addition to (or
Ulm [2] for bone resorption, Lemarchand et al. [20] for removal from) the solid as well as the solid strain rate. Let
mechanically induced dissolution processes, and Ulm et al. the velocity component due to the mass addition/removal
[21] for calcium leaching. However, the present work treats be uc , defined such that uc n > 0 (n is the outward
the biochemical phenomena at finite strains (whereas normal in the current configuration to the solid-cell inter-
infinitesimal strain assumptions were used in the above face, see Fig. 2) for addition, and uc n\0 for mass
studies) in order to develop a model for biological tissue removal. The velocity of the solid-cell interface un is
remodeling, which may involve small deformations for defined by the sum of the two components uc and ud ,
hard tissues (e.g., bone) and relatively large deformations un ud uc . This results in a discontinuity in the
for soft tissues (e.g., skin). velocity field at the solid-cell interface, and the jump in
velocity u is defined by
u un un uc 1
Remodeling analysis at the microscale
In case of density, the discontinuity occurs across the
Problem definition
solid-cell interface. The solid density at the solid side is
defined by solid mass per unit volume qs t, while that at
The representative volume element (r.e.v) at this scale
the cell side is the fluid mass per unit volume, qf t.
contains a solid domain Vs and a cell attached along the
cell-solid interface f (see Fig. 1b). The mass transfer takes qt qn ; t qn ; t qf t qs t 2
place between the solid surface and the cellular fluid
around it, by deposition or dissolution of solid mass from In the present work, the change in the solid density
or into the cell fluid. The cell fluid is considered to be a across the interface is also significant. The solid density
mixture (at pressure p) of a solvent and a solute (partial jumps from qs t on the solid side to 0 in the fluid.
pressure pi), generated by the biochemical activity of the
cell. The problem is defined in terms of the spatial coor- qs t qs n ; t qs n ; t qs t 3
dinates, in order to circumvent the difficulty of defining a
suitable material coordinate system for a problem with Similarly, the jump in the free energy mass-density (free
continuous mass (and material points) addition or removal energy per unit mass) across the interface is defined by the
at the cell-solid interface. Since concurrent deformation difference in the free energy mass-density in the cell fluid
and growth processes are assumed, the attachment or (wi) and the solid (ws).
removal of the material points occur in the deformed solid
configuration. Thus, the remodeling problem is initially w wn wn wi ws 4
defined in terms of quantities measured in the deformed
configuration (i.e., current reference frame) x, and subse- The discontinuities in the density and the free
quently suitable transformation equations are used to energy mass-density are schematically shown in
express these quantities in the original (i.e., undeformed) Figs. 2b and c.
reference frame, x.
Figure 2 shows the evolution of several field variables Preliminaries
across the solid-cell interface (x = n). The solid deforms
under the applied loading (external stress and fluid pres- The focus of the present work is the analysis of deforma-
sure), and simultaneously undergoes mass addition (or re- tion coupled remodeling in porous biological tissues, and
moval) at the solid-cell interface, which in turn changes the the analysis is conducted in both the initial (i.e., unde-
volume of the solid. In the current configuration, the mass formed) and the current (i.e., deformed) reference frames.
addition (or removal) at the interface may be represented in The analysis thus involves repeated transformation of
terms of velocity jump at the solid-cell interface. The quantities between the two reference frames. The equations
particle velocity in the solid bulk (ud ) is defined
by the
for volume, density and surface area transformations for a
microscopic strain rate in the solid, d 12 gradud deforming material (without any growth/resorption) are
grad t ud , where grad t ud is the transpose of the given below.
123
8876 J Mater Sci (2007) 42:88738884
u = ud + uc =s =s
n
n n
solid ( Vs ) cell ( Vf ) sosliodli(dVs ) cell c(eVlfl) sosliodli(dVs ) ceclle(lV
l f)
=f =f
u = ud
- + - + - +
Fig. 2 Schematic representation of the evolution of (a) velocity, (b) represented by n. The area of the solid-cell interface is f. The
density and (c) free energy per unit mass across the solid-cell primary purpose of this figure is to show the jump in the field
interface (x = n) in the current configuration. The vertical lines quantities across the solid-cell interface. The field quantities have
corresponding to x = n and x = n+ are infinitesimally distant from the been shown constant in each of the domains (Vs and Vf) for
interface, and lie in the solid domain (Vs) and the cellular domain (Vf), convenience. No such assumptions have been employed in the
respectively. The outward normal to the solid-cell interface is analysis
If vs t0 is the original (undeformed) infinitesimal vol- fields that exhibit discontinuities within the given domain.
ume of a solid, then the volume vs t after deformation is For example, if the field x exhibits a discontinuity at R
given by within the domain V of perimeter A, then
Z Z Z
vs t jvs t0 5
x nA dA divxdV x nR dR 9
A V R
where, j is the determinant of the deformation gradient f ,
defined in terms of the undeformed (X) and deformed (X) where, nA and nR are outward normals at the boundary A
reference frames (f Gradx ox=oX). For a solid with and the surface of discontinuity R, respectively, and x is
constant mass, the density is inversely proportional to the the jump in the field x across the surface of discontinuity.
volume, and the densities in the two reference frames are
related by Mass and volume conservation
123
J Mater Sci (2007) 42:88738884 8877
(see Eq. (1) and Fig. 2). Thus using the generalized Z Z
dVs t
divergence theorem (Eq. 9) on the solid domain, the rate of divud dvs t uc ndac t 17
dt Vs t ft
change of the solid mass can be expressed by
Z The first term reflects the change in volume due to the
dMs t dqs t d deformation undergone by the solid, while the second term
qs tdivu dvs t
dt Vs t dt signifies the corresponding volume increase (or decrease)
Z due to biological growth (or dissolution) at the solid surface.
qs tu ndac t 12 The two activities thus can be clearly demarcated in the
ft
deformed configuration. This demarcation is less clear in
where ft is the solid-cell interface area in the deformed the undeformed configuration, as evident from the volume
configuration (the infinitesimal solid-cell interface area is change rate derived applying Eqs. (7, 8, 16) to Eq. (17).
given by dac t) and n is the outward normal (outward Z Z
from the solid) at the interface. The discontinuity of the dVs t dj
dvs t0 u c Ndac t0
je 18
velocity field u is defined by uc (see Eq. 1). The above dt Vs t0 dt ft0
equation contains three terms. The first two terms
dqs t=dt and qs tdivud integrated over the solid In contrast to Eq. (17), the second term in Eq. (18),
volume vs t describe the mass conservation in a reflecting undeformed measures, contains both the bio-
deforming solid. The last term defines the mass loss or chemical (e u c N) and the mechanical (j detf ) contri-
addition at the solid-cell interface. Considering mass butions. The coupling arises due to the fact that Eq. (18)
conservation in the solid bulk (i.e., neglecting diffusional measures the rate of change of current volume (i.e., volume
mass exchange between the growing/dissolving solid after coupled deformation and mass change) in the unde-
surface and the solid bulk), i.e., formed reference frame, and must account for the change
in density due to deformation. It may be convenient to
Z
dqs t introduce the (Lagrangian) mass rate per unit (undeformed)
qs tdivud dvs t 0 13 surface involved in the deposition or dissolution process,
Vs t dt
m u c N. Use of m
c qs t0 e c in Eqs. (15, 18) yield the
the mass gain (or loss) is only due to the surface Lagrangian expressions in the form
contribution. Z
dMs t
Z c dac t0
m 19
dMs t dt ft0
qs tuc ndac t 14
dt f t
Z Z
dVs t dj c
m
Since the rate of mass change is invariant with the dvs t0 j dac t0 20
change in configuration, the rate in the deformed and dt Vs t0 dt ft0 qs t0
undeformed configurations can be equated
Z Z
dMs t
qs tuc ndac t u c Ndac t0
qs t0 e Thermodynamic formulation
dt f t ft0
123
8878 J Mater Sci (2007) 42:88738884
Z Z
In the deformed configuration, the external work rate is jp pi
the sum of two terms. The first term is the work rate P ext f : pdvs t0 mc dac t0
Vs t0 ft0 qs t0 qi
developed by the traction field t r n (with r the Cauchy
25
stress tensor and n the outward normal to the solid surface)
and the total solid velocity u (composed of velocities due to
where f is the rate of change of the deformation gradient,
deformation and chemical growth/dissolution) along the
and p jr t f 1 is the Boussinesq (or first Piola-Kirch-
solid surface in the deformed configuration As(t)
hoff) stress tensor.
Z Z The change of the Helmholtz free energy of the system
Pext u u r ndas t divud rdvs t under consideration is due to the change of the free energy,
As t Vs t
Z between t and t + dt, in the solid bulk and along the surface
uc tdac t 22 of discontinuity. Expressed in terms of the specific free
ft energies, this change in the deformed reference frame is
where the generalized divergence theorem Eq. (9), and Eq. given by
(1) have been used. The last term in Eq. (22) represents the Z Z
dW dws
work rate pdV, involving the fluid pressure p on the boundary qs t dvs t wi ws qs tuc ndac t
dt Vs t dt ft
of the considered domain and the dissolved solid volume dV,
which is obtained by letting t pn in Eq. (22) 26
Z Z where wi and ws are the free energy per unit mass of the
Pext u r : ddvs t puc ndac t 23 solute ions (in the cell fluid) and the solid, respectively.
Vs t f t While the first term in Eq. (26) is classical, the second
term expresses the spontaneous change at the solid-cell
where d is the (symmetric) Eulerian strain rate tensor, interface, induced by w wi ws which can be
obtained from developing divud r gradud : r, while interpreted as the discontinuity in Eq. (4) in free
considering divr 0 and r rt in Vs t. The second energy per unit mass between the solute in the cell
contribution to P ext of the considered system arises from fluid (wi ) and the solid phase (ws). The Lagrangian
the introduction (or extraction) of the solute mass dMi into counterpart of Eq. (26) is readily obtained using Eqs. (5,
the cell-fluid, which, given mass conservation of the sys- 6, 15, 19)
tem under consideration, equals with opposite sign the
change in solid mass (given by Eq. 15), i.e., Z Z
dW dw
dMs =dt dMi =dt qi dVi , where qi is the mass density qs t0 s dvs t0 mc dac t0
wi ws
dt Vs t0 dt ft0
of the reactant ions in the cell-fluid (assumed for purpose of
clarity constant in dV). The resulting work contribution is 27
the work rate Piext pi dVi developed by the partial pressure
of the Rsolute pi along the rate of change of solute volume Finally, using Eqs. (24, 26) in Eq. (21) yields the
dVi ft qs t=qi uc ndft. The total external work Eulerian expression of the dissipation rate
rate in the deformed reference frame thus reads Z
dws
Z D r : d qs t dvs t
Vs t dt
P ext Pext u Piext r : ddvs t Z
p
Z
Vs t
gi ws q tuc ndac t 0 28
p pi f t qs t s
q tuc ndac t 24
ft qs t qi s where gi wi pi =qi is the Gibbs potential or free mass
enthalpy (per mass unit) of the solute (ions in the solution
In the undeformed configuration, the external work of the cell fluid). Analogously, Eqs. (25) and (27) yield the
terms can be evaluated through suitable transformation of Lagrangian counterpart
Eq. (24), using Eqs. (6, 8, 15, 19)2
Z
dws
D f : p qs t0 dvs t0
Vs t0 dt
Z " #
2
Using grad: Grad: f 1 and p jr t f 1 , divud r jp
h i h i gi ws c dac t0 0
m 29
1
j Gradu f
d 1
: p t f 1j f : p : Given that dvs t jdvs t0 , f t 0 qst0
R R
d
Vs t divu rdvs t Vs t0 f : p dvs t0 .
123
J Mater Sci (2007) 42:88738884 8879
The first integral (the volume integral) in these two p
expressions of the Clausius-Duhem inequality represents A Mgi gs M wel
s 33
qs t
the intrinsic dissipation rate in the solid bulk. It is of the
standard format classically employed in continuum From Eqs. (3033) it is apparent that a positive affinity
mechanics. Assuming elastic deformation in the solid bulk, encourages mass addition or surface growth
and thus zero dissipation, the terms simply signify that the (uc uc n 0), while a negative driving force encourages
rate of change of elastic energy in the bulk is governed by mass removal or dissolution at the solid-cell surface
the stressstrain work. More generally, in the absence of (uc 0). This is readily seen for an undeformed solid
bio-chemical processes at the solid-cell surface (i.e., phase, for which uft uft0 , and for which, following
uc n 0 , mc 0), it represents the amount of exter- chemical thermodynamics, the first term Mgi gs in Eq.
nally supplied energy which is not stored in the solid (33) is identified as the Gibbs energy variation per mole;
microstructure, but dissipated into heat form. Given its that is the difference between the chemical potential of the
intrinsic nature, associated with solid deformation, it may dissolved ions in the cell fluid, which is controlled by the
be assumed to be non-negative irrespective of the phe- cell, and which is therefore conveniently referred to as
nomena at the solid-cell surface. Consequently, the dissi- biologically generated potential lBGP [2], and the chemical
pation associated with the resorption process which is potential of the same substance bound in the solid phase ls
captured by the surface integrals in Eqs. (28) and (29), (which in turn is related to the solubility product of the
must be non-negative as well, i.e., solid). In this case, the chemical affinity coincides with the
" # Gibbs energy variation (see e.g., [24]), i.e.,
p A0 lBGP ls . Thus, lBGP ls > 0 is expected to
uft gi ws qs tuc n 0 30
qst encourage growth at the solid surface, as evident from the
non-negativity of the dissipation. Similarly, for a deform-
able medium, cellular fluid pressure and/or elastic defor-
" #
jp mation in the solid will encourage mass dissolution at the
uft0 gi ws uc N 0
qs t0 e 31 solid surface. Adopting the stated link between the Gibbs
qst0
potential gi and the biological generated potential lBGP for
constant solute pressure and temperature (for which
In contrast to the volumetric dissipation in the solid
gi = const), the biochemical affinity becomes
bulk, the dissipation rateh ufti is a surface dissipation rate
density (of dimension uft MT 3 ). Following stan- M el
dard thermodynamic arguments, expressions Eqs. (30) and A lBGP ls v p 34
qs t v
(31) may be used to formally identify the term
el
gi ws p=qs t as the driving force of the influx (in the where vel
v qs tws is the elastic energy per unit current
case of mass deposition) or the outflux (in the case of volume, while M=qs t is the current molar volume.
el
dissolution) of solid mass per unit surface occupied by Denoting by e v el
v qs t0 ws and M=qs t0 the elastic
cells, which is expressed by qs tuc n in the deformed energy per unit volume and the molar volume, respectively,
configuration, and m u c N in the undeformed
c t0 qs t0 e in the undeformed configuration, the driving force can be
configuration. It is convenient to distinguish in ws the en- expressed in Lagrangian variables.
ergy associated with elastic deformation of the solid from
M el
those related to the chemical composition of the solid, by A lBGP ls e
v jp 35
considering ws wel el qs t0 v
s gs , where ws is the elastic free
energy and gs is the chemical potential per unit solid mass.
Finally, following standard thermodynamics, the iden-
Furthermore, the dissipation expressions may be rewritten
tification of affinity A as the driving force of the bio-
in terms of the molar flux J qs t=Muc n in Eq. (30)
chemical growth/dissolution at the solid surface, (which
and Je qs t0 =Me u c N in Eq. (31), where M is the
qualitatively defines the effect of the different components,
molar mass, and M=qs t and M=qs t0 the molar vol-
e.g., pressure, deformation and biochemical potential dif-
umes of the solid in the deformed and reference configu-
ference, on the growth/dissolution tendency at the solid
rations
surface) leads to the definition of the kinetics of the growth/
uft A J 0; uft0 A Je 0 32 dissolution process by a constitutive equation relating the
driving force (A) to the molar flux in the deformed con-
where A is the so-called chemical affinity (see [22, 23]) figuration J J A, or in the undeformed configuration
here of the cell-mediated biochemical reaction. Je JeA. The simplest form of such a growth/dissolution
123
8880 J Mater Sci (2007) 42:88738884
law that satisfies the non-negativity of the local dissipation biological tissue. At the macroscale, the primary objective
in Eq. (32) is a linear form which captures, similar to a is to determine the effects of the remodeling processes
discrete Ficks Law, the diffusion of the ionic species (which occurs at the microscale) on variables that are rel-
between the solid and the cellular fluid. evant at the scale of the porous material. At this scale, the
r.e.v is composed of the solid domain (Vs t and the fluid
M e M e domain (Vf t; which occupies in the current configuration
A k J k J 36
qs t qs t0 the volume Vf t V t Vs t; V t being the volume of
the porous tissue. The ratio of the current fluid volume over
Or equivalently, the total initial volume of the r.e.v V(t0) is referred to as the
Lagrangian porosity, and its rate of change is defined as
A kuc keuec 37
d/ 1 dVt dVs t
where uc uc n and uec uec N. The diffusion 39
dt Vt0 dt dt
coefficient k (of dimension k LMT 1 mole1 ) is
primarily governed by the diffusion mechanism in the
Using Vt JVt0 (where J is the determinant of the
deformed configuration, and is independent of the factors
macroscopic deformation gradient F) and defining
governing the driving force A, i.e., pressure, deformation
dVs t=dt by the upscaled version of Eq. (18), Eq. (39) can
and biochemical potential difference. In return, the
be rewritten as
diffusion coefficient ke in the reference configuration
takes into account the change of orientation of the solid Z Z !
d/ 1 dJ dj c
surface due to deformation. The two expressions for the Vt0 dvs t0 u Ndac t0
j~
diffusion coefficients k and ke can be related by using Eqs. dt Vt0 dt Vs t0 dt ft0
(16, 37) 40
where Vs t0 is now the total solid volume in the porous
ke material, and ft0 is the total cell-solid interaction
n f N 1 38
k surface (for the entire porous tissue being considered),
both expressed in the undeformed configuration.
Finally, it should be noted that surface curvature effects
Expressing
dj
the volume
R average of the rate
of change
have been neglected in our derivation. The surface curva- dj
of j by dt , i.e., Vs t0 dt dvs t0 Vs t0 djdts Vs t0 , Eq. (40)
ture affects both the dissolution kinetics, as well as the
can be reduced to
solid stress state, because of the associated stress discon-
htinuity
h ii at the solid-fluid interface (term of the form Z
r n cjn, where j is the mean curvature of the d/ dJ dj 1
cs t0 uc Ndac t0
j~
solid-cell surface3, and c is the surface energy per unit area, dt dt dt Vs t0 Vt0 ft0
of dimension MT2). This stress discontinuity would enter 41
through Eq. (22) the energy derivation, and would ulti-
mately lead to an effect of this curvature term on the dis- or equivalently, using the upscaled version of Eq. (20),
solution kinetics in a similar way as the fluid pressure. Z
Indeed, it suffices to replace p in Eqs. (34, 35) by an d/ dJ dj 1 c
m
cs t0 j dac t0
effective pressure p0 p cj. It is no surprise, then, to dt dt dt Vs t0 Vt0 ft0 qs t0
confirm that a high positive (i.e., outward) curvature of a 42
solid surface leads to a higher dissolution rate, as it has
been recognized in skeletal tissue mechanics [25]. where cs t0 Vs t0 =Vt0 is the solid volume fraction in
the undeformed porous material, and m u c N is
c qs t0 e
the mass rate per unit (undeformed) surface involved in the
Application to porous media cell-mediated deposition or dissolution process. The first
and second terms in Eqs. (41) and (42) represent the con-
Porosity change tribution of pure mechanical deformation (the entire porous
material and the solid phase, respectively) to the change in
At the microscopic scale, the formulation analyzed defor- porosity, and are the classical terms employed in the
mation coupled remodeling of a single strut of the porous Biot-Coussy theory of porous media [22]. The third term
3
signifies the contribution of the biochemical growth/dis-
2j 1=R1 1=R2 , with R1 and R2 the radius of curvature at the
solution process, amplified by deformation.
major and minor axis of the solid-cell surface.
123
J Mater Sci (2007) 42:88738884 8881
123
8882 J Mater Sci (2007) 42:88738884
el !
d/ vv
de M el
_
Dm Vt0 F : P p cs t0 52 A A0 e
v 56
dt dt Vs t0 qs t0 v
123
J Mater Sci (2007) 42:88738884 8883
1000 1000
(a) (b)
100 100
= -100 = -100
10 10
= -10 = -10
= -1 = -1
1 1
0 .2 5 0 .5 0 .7 5 1 1 .2 5 1 .5 1 .7 5 2 2 .2 5 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8
Str e tch in load ing dir e ctio n T rue strain
Fig. 3 Representation of the normalized dissolution velocity varia- deformation measure is defined by the true strain in the loading
tion for different soft biological tissues under uniaxial loading, in (a) direction (natural log of stretch), signifying compression for negative
Lagrangian and (b) Eulerian strain measures. The Lagrangian true strain and tension for positive strain. The tissue properties are
measure of tissue deformation is defined by the stretch in the loading characterized by E, which signifies a ratio of the elastic deformation
direction (stretch = (final length)/(initial length)). This signifies energy and the biochemical energy for the tissue remodeling (see text)
compression for stretch <1 and tension for stretch >1. The Eulerian
measure is employed, such as the true (logarithmic) strain an increase in the chemical potential of the dissolving ions
which is related to the (Lagrangian) stretch by ln lnk1 . (compared to their potential in the solid) was observed to
The results
displayed
in Fig. 3b indicate, for the same encourage tissue growth. The microscopic equations were
value of ln , a higher dissolution rate enhancement in upscaled to yield the tissue characteristics (e.g., change in
tension than in compression. This underscores the necessity Lagrangian porosity) at the macroscale. A numerical
to specify (both theoretically and experimentally) the ref- example on the effect of deformation on the dissolution
erence frame (Lagrangian or Eulerian) used for measuring rate for an incompressible soft tissue showed that the rate
field quantities. increased for both tensile and compressive deformations.
The change of the dissolution rate (per unit strain) de-
pended on the strain measure. For Lagrangian measures
Conclusion
(e.g., stretch), the rate increase per unit strain was higher
under compressive loading than under tensile loading. On
The present work attempts to develop an integrated anal-
the other hand, the rate increase per unit strain was higher
ysis for deformation coupled surface remodeling in porous
under tensile deformation when an Eulerian measure (e.g.,
biomaterials. In such materials (examples include trabec-
true strain) was used.
ular bone and skin), the biochemical phenomenon related
to remodeling (growth/dissolution) occurs at a microscopic Acknowledgments The authors wish to thank the Natural Sciences
scale and estimation of its effect on the tissue level and Engineering Research Council (NSERC) of Canada and the Es-
(macroscopic) properties is important. In the present work, ther and Harold E. Edgerton Chair at MIT for financial assistance.
The authors are thankful to Profs. Lorna Gibson (MIT), Olivier
the remodeling and deformation (at the microscale) in the
Coussy (Institute de Navier, France), and Luc Dormieux (Institute de
biological tissues is modeled through development of Navier, France), and Dr. Anirban Sain (McGill University, Canada)
suitable equations for mass, volume and the relevant for helpful discussions.
thermodynamic quantities for the growing solid, which are
then upscaled to yield the macroscopic effects. The equa-
tions are developed assuming a spatial reference frame, to Appendix: Hill Lemma
circumvent the problems of defining a suitable material
reference frame for growth/dissolution processes, charac- This demonstration of the Hill Lemma is inspired by the
terized by a continuous change in the number of material presentation of Zaoui [29]. We denote by p and f_ the
(and mass) points. The analysis examined the effects of microscopic Boussinesq tensor which (in the absence of
different field quantities on the growth/dissolution potential body forces) satisfies Divp 0 in V(t0), and the micro-
of the solid-cell interface. A positive pressure of the cel- scopic deformation gradient rate tensor f_ Gradu. If,
lular fluid and deformation of the solid was observed to either f_ satisfies a uniform deformation boundary condi-
encourage solid dissolution (and tissue resorption), while tion, or p a uniform traction boundary condition, then
123
8884 J Mater Sci (2007) 42:88738884
D E D E D E
on oV t0 : u F_ X 60 References
where F_ represents the deformation rate of the r.e.v at the 1. Cowin SC (1983) Ann Biomed Eng 11:263
macroscopic scale, while u is defined at the microscopic 2. Silva EC, Ulm F-J (2002) In: Karihaloo BL (ed) Proc. IUTAM
scale. The work rate provided to the r.e.v by the surface Symp. on Analytical and Computational Fracture Mechanics of
Non-Homogeneous Materials, Kluwer Acad. Pub., London,
traction t p n in the undeformed configuration reads (in p 355
components) 3. Ballarini R, Kayacan R, Ulm F-J, Belytschko T, Heuer AH
(2005) Int J Fracture 135(14):187
Z Z 4. Grinnell F (1994) J Cell Biol 124:401
P ui ti dat0 F_ ij Xj pik Nk dat0 61 5. Freyman TM, Yannas IV, Pek Y-S, Yokoo R, Gibson LJ (2001)
oV t0 oV t0 Exp Cell Res 269:140
6. Lund D, Tomanek RJ (1978) Am J Anat 152:141
Application of the divergence theorem yields, for any 7. Sobin SS, Tremer HM, Hardy JD, Chiodi HP (1983) J Appl
Physiol 55:1445
stress field pik that satisfies pik;k 0 (that is Divp 0) 8. Yannas IV, Lee E, Orgill DP, Skrabut EM, Murphy GF (1989)
yields Proc Natl Acad Sci USA 86:933
Z Z 9. Pauwels F (1980) Biomechanics of the locomotor apparatus.
Springer-Verlag, Berlin
Xj pik Nk dat0 Xj pik ;k dvt0 10. Skalak R (1981) In: Carlson DE, Shield RT (eds) Proc. IUTAM
oV t0 V t0
Z Symp. on Finite Elasticity, Martinus Nijhoff, The Hague, p 348
11. Rodriguez EK, Hoger A, McCulloch AD (1994) J Biomech
pik djk Xj pik;k dvt0
V t0 27:455
Z 12. Chen Y-C, Hoger A (2000) J Elast 59:175
pij dvt0 62 13. Lubarda VA, Hoger A (2002) Int J Solids Struct 39:4627
V t0 14. Lee EH (1969) J Appl Mech 36:1
15. Lubarda VA (1991) Int J Solids Struct 27:885
16. Gibson LJ (1985) J Biomech 18:317
Thus, inputing pij V t0 Pij in Eq. (61)
17. Hellmich C, Ulm F-J (2002) J Biomech 35:1199
18. Zysset PK, Curnier A (1996) J Biomech 29:1549
P V t0 F_ ij Pij V t0 F_ : P 63 19. Blair HC (1998) BioEssays 20:837
20. Lemarchand E, Dormieux L, Ulm F-J (2003) In: Huyghe JM (ed)
(2) Consider the r.e.v subjected to a uniform traction Proc. IUTAM Symp. on the Mechanics of Physicochemical and
Electromechanical Interactions in Porous Media, Kerkrade
boundary condition (Netherlands)
21. Ulm F-J, Lemarchand E, Heukamp FH (2003) Eng Fract Mech
on oV t0 : t p N P N 64 70:871
22. Coussy O (1995) Mechanics of porous continua. J. Wiley & Sons,
where P is the macroscopic Boussinesq tensor, while t is Chichester UK
the microscopic stress vector (defined on the undeformed 23. Coussy O, Ulm F-J (1996) Arch Appl Mech 66(8):523
configuration). The work rate is developed in the form: 24. Atkins PW (1994) Physical chemistry. WH Freeman and Com-
pany, New York, p 216
Z Z Z 25. Martin RB, Burr DB, Sharkey NA (1998) Skeletal tissue
mechanics. Springer-Verlag, New York
P ui ti dA ui pij Nj dA ui Nj dA Pij 65 26. Dormieux L, Djimedo K, Ulm F-J (2006) Microporomechanics.
oV oV oV
J. Wiley & Sons, Chichester UK
27. Attard MM (2003) Int J Solids and Struct 40:4563
Application of the divergence theorem yields 28. Rho JY, Roy ME, Tsui TY, Pharr GM (1999) J Biomed Mater
Z Z Z Res 45:48
29. Zaoui A (1997) Materiaux heterogene et composites (lecture
ui Nj dA ui djk Nk dA ui djk ;k dV t0 notes, Edition 1997). Departement de Mecanique, Ecole Poly-
oV oV V t0
Z Z technique, France
ui;k djk dV t0 f_ij dV t0 66
V t0 V t0
123
J Mater Sci (2007) 42:88858893
DOI 10.1007/s10853-007-1759-7
Received: 18 December 2006 / Accepted: 10 April 2007 / Published online: 17 July 2007
Springer Science+Business Media, LLC 2007
Abstract Wet adhesion is widely adopted in biological scientific interest, since it can give insights into the
adhesion systems in nature. Wet adhesion is studied in this workings of nature for shaping the structures in evolu-
paper with the focus on the effect of different contact tionary processes. Also, we can discover the detailed
shapes (flat, concave, convex, and ring-like) on the adhe- chemical and physical properties of the materials which
sion force. The evolution of the liquid bridge between a have evolved, and can learn about their use as structural
fiber tip and substrate during the detaching process shows elements and their biological role and function for the
two transition points. The first transition from the radius- guidelines of man made adhesion systems.
controlled to the contact-angle controlled process is critical Previous works [13] have studied the hierarchical
to influence the strength and robustness of adhesion. We structure of the animals feet and the scaling law for dry
show that a concave shape is more effective than a flat one, adhesion. The adhesive system with dry adhesion is based
while a convex shape has no advantage. A ring-like contact on van der Waals forces or electrostatic force between the
shape has advantages in a hydrophobic environment and on finely structured feet and the substrate. A number of
a rough surface. adhesion measures [47] on the attachment devices such
as: flies, spiders, geckos, Tettigonia viridissima, etc, pro-
vide the evidence that van der Waals force plays dominant
role in dry adhesion, and theoretical analysis [1, 8, 9]
proved that biological systems indeed can use van der
Introduction Waals force to achieve strong adhesion with finely struc-
ture design of the attachment systems.
Insects and geckos have evolved a variety of well-defined In contrast, wet adhesion is based on capillary for-
shapes and structures for adhesion. Although often intricate cesthrough a liquid bridge between the fiber and the
and fragile, they can nonetheless deal with extreme surface of the substrate [2]. Capillary force is a long range
mechanical loads with high efficiency. Some insects live force in comparison with the van der Waals force. The
attached to a substrate with dry adhesion; others use wet capillary force between a hemisphere and a plane had been
adhesion. These abilities are based on a variety of inge- studied several decades ago [10]. The determination of li-
nious structural solutions. Understanding these is of great quid-bridge profile is the key to solve a wet adhesion
problem [11]. The stability, breakage, elastic properties of
liquid bridge [12] at nanometer scale are studied by both
Y. Su B. Ji (&) K. Hwang
theoretical and experimental approaches. Experiments
Department of Engineering Mechanics, Tsinghua University,
Beijing 100084, China showed that humidity contributes significantly to gecko
e-mail: bhji@mail.tsinghua.edu.cn adhesion implying the importance of capillary force for
adhesion at larger scale [13, 14]. A fiber-substrate model
Y. Huang
has been recently studied by Qian and Gao [15]. They find
Department of Mechanical Science and Engineering, University
of Illinois, Urbana, IL 61801, USA that the size of the fiber plays an important role in bio-
e-mail: huang9@uiuc.edu logical wet adhesion, and the strength of wet adhesion is
123
8886 J Mater Sci (2007) 42:88858893
highly enhanced by radius reduction of the fiber. A scaling model for the liquid bridge is shown in Fig. 2. We make
law of the adhesion strength was obtained for the self- the following assumptions in order to simplify the analysis:
similar fiber-substrate systems.
(a) No phase change such as liquid to gas during the
In previous studies, the tip of the fiber in the adhesion
detaching process such that the volume of liquid is
system was often assumed to be flat or hemisphere-like
conserved;
[16]. Experimental studies show that nature provides a
(b) Since the fiber and substrate are much stiffer than the
variety of contact shapes, e.g., horseshoe, suction cup, and
liquid and therefore have essentially no deformation
torus, etc., for good adhesion [1621]. What is the effect of
during the process, both the fiber and the substrate are
the contact shape on the adhesion force? For dry adhesion,
assumed to be rigid;
it has been found that the contact shape exerts a strong
(c) The detaching process is quasi-static such that the
influence on adhesion strength, e.g., the best shape is the
inertia and viscoelastic effects are negligible;
flat tip for perfectly smooth substrate [16]. Gao and Yao
(d) Since the typical size (diameter) of the fiber is at the
[19] also discussed how to design an optimum contact
micrometer scale, the effect of gravity is negligible in
shape for dry adhesion at nanoscale. In this paper, we study
comparison with the surface tension at this scale.
the effects of contact shape on the adhesion force for the
wet adhesion system. This study is orientated to help For this problem, the YoungLaplace equation connects
improve the design of micro- and nano- manipulation the pressure difference inside and outside the meniscus to
system based on capillary force by the biomimicking the local liquid profile by
approach. For example, recent experimental studies [21,
22] suggested that a concave tip can generate a much larger 1 1
DP c 1
capillary force than a flat one. However, the mechanisms R1 R2
have not been well understood. where 1/R1 and 1/R2 denote the two local principal
The organization of the paper is as follows: In Section curvatures of the liquid profile, respectively, DP is the
The model, we introduce the models for wet adhesion pressure difference which is constant within the meniscus
with four different contact shapes of the fiber tip, mainly due to the negligible gravity, and c is interface energy
focusing on how to calculate the profile of the liquid bridge between vapor and liquid. c is connected with the contact
and the adhesion force. These four contact shapes are flat, angle h via the Youngs equation,
concave, convex, and ring-like, which represent the flat
punch, suction cup, hemisphere, and torus structure cSV cSL
evolved in nature [18], respectively. In Section Results, cosh 2
c
these four contact shapes are studied, and the detaching
force for each contact shape is calculated and compared. where cSV and cSL are interface energy of solidvapor and
Finally, discussion and conclusions are made in Section solidliquid, respectively.
Discussion and conclusions. The liquid profile should be obtained first in order to
calculate the adhesive force. An ordinary differential
The model equation can be derived from YoungLaplace equation [10,
15, 23],
The wet adhesion of the hairs of biological attachment
system is modeled as a fiber contacting with the substrate du DP sinu dx dz
; where cosu; sinu 3
mediated by a liquid bridge. To study the effect of contact ds c x ds ds
shape on the adhesion strength, we introduce four different
geometries of the fiber tip, i.e. flat, concave, convex, and where x and z are the coordinates of the axisymmetric li-
ring-like, to model the flat punch, suction cup, hemisphere, quid bridge, u is the angle between the local tangent of
and torus structure evolved in nature [18], respectively, liquid surface and the horizontal axis, and s is the arc
shown in Fig. 1. The fiber is assumed axisymmetric. The length of the liquid profile as shown in Fig. 2.
Fig. 1 The adhesion systems
with different contact shapes.
(a) flat; (b) concave; (c) convex;
(d) ring-like
123
J Mater Sci (2007) 42:88858893 8887
123
8888 J Mater Sci (2007) 42:88858893
123
J Mater Sci (2007) 42:88858893 8889
Fig. 6 Schematic illustration of the fiber-liquid-substrate system with convex shape with an equivalent contact angle heq > h1 and an
three different shapes. (a) the concave shape with an effective contact effective liquid volume Veq V DV ; (c) the ring-like contact
angle heq < h1 and an effective liquid volume Veq V DV ; (b) the shape with two contact lines
123
8890 J Mater Sci (2007) 42:88858893
123
J Mater Sci (2007) 42:88858893 8891
Appendix
123
8892 J Mater Sci (2007) 42:88858893
Table 1 continued
Meridional curvature Average Profile Profile radius Remark
curvature radius
du
p
b \0 H 6 0 X>0 sinu sin2 uHX up HX up 2sinuup
dS Xu H
8 p
> 2
< sinu sin uHX up HXup 2sinuup du > 0
du p
H dS
c \0 or H 6 0 X>0 Xu
dS > 2
: sinu sin uHXup HX up 2sinuup du \0
du
dS
0 or ( H dS
2 X
du
\0 Inflexion Xc Hup HX up 2sinuup > 0
dS du
dS
0 sin2 uc HXup HX up 2sinuup M0\M 1
du
d dS
0 (global) H 6 0 X>0 X H1 u p2
du
e dS
0 (global) H0 X>0 sin u = 0 (u = 0 oru = p)
du X up sinuup
f dS
6 0 or du
dS
0 (local) H 0 X>0 Xu sinu
123
J Mater Sci (2007) 42:88858893 8893
Table 1 continued
Meridional curvature Average Profile Profile radius Remark
curvature radius
g du
dS
6 0 H 6 0 X0 Xu 2sin
H
u
Zu 2cos
H
u
(local)
123
J Mater Sci (2007) 42:88948903
DOI 10.1007/s10853-007-1901-6
Received: 9 January 2007 / Accepted: 2 April 2007 / Published online: 17 July 2007
Springer Science+Business Media, LLC 2007
Abstract When the DNA double helix is subjected to previously published models for stretched DNA. The cal-
external forces it can stretch elastically to elongations culated electronic states are then used to parametrize an
reaching 100% of its natural length. These distortions, effective tight-binding model that can describe electron
imposed at the mesoscopic or macroscopic scales, have a hopping in the presence of environmental effects, such as
dramatic effect on electronic properties at the atomic scale the presence of stray water molecules on the backbone or
and on electrical transport along DNA. Accordingly, a structural features of the substrate. These effects introduce
multiscale approach is necessary to capture the electronic disorder in the model hamiltonian which leads to electron
behavior of the stretched DNA helix. To construct such a localization. The localization length is smaller by several
model, we begin with accurate density-functional-theory orders of magnitude in stretched DNA relative to that in the
calculations for electronic states in DNA bases and base unstretched structure.
pairs in various relative configurations encountered in the
equilibrium and stretched forms. These results are com-
plemented by semi-empirical quantum mechanical calcu-
lations for the states of a small size [18 base pair Introduction
poly(CG)poly(CG)] dry, neutral DNA sequence, using
Soon after Watson and Cricks discovery of the DNA
double-helix structure [1], Eley and Spivey [2] introduced
R. L. Barnett A. Turner M. Fyta E. Kaxiras the notion of efficient charge transport along the stacked p
Department of Physics, Harvard University, Cambridge, orbitals of the bases. The mechanism of charge transport
MA 02138, USA has been the subject of numerous studies in the intervening
years, with renewed interest fuelled recently by both bio-
Present Address:
R. L. Barnett logical and technological considerations. Over a decade
Department of Physics, California Institute of Technology, ago, Barton and co-workers observed distance-independent
Pasadena, CA 91125, USA charge transfer between DNA-intercalated transition-metal
complexes [3] and argued that it would be relevant for
P. Maragakis
Department of Chemistry and Chemical Biology, biology and biotechnology. More recent electron transport
Harvard University, Cambridge, MA 02138, USA experiments on DNA have yielded widely varying results,
showing alternatively insulating behavior [48], semicon-
Present Address:
ducting behavior [9], Ohmic conductivity [1013], and
P. Maragakis
D.E. Shaw Group, 120 West Forty-Fifth St., New York, proximity induced superconductivity [14]. The large
NY 10036, USA number of relevant variables endemic to such experiments,
like the DNA-electrode contact, and the rich variety of
E. Kaxiras (&)
structures that DNA can assume, are the causes of vari-
School of Engineering and Applied Sciences,
Harvard University, Cambridge, MA 02138, USA ability in the experimental measurements (for a recent
e-mail: kaxiras@physics.harvard.edu review of transport theory and experiments see Ref. [15]).
123
J Mater Sci (2007) 42:88948903 8895
Specifically, there is a large diversity of the DNA forms structures obtained in the pioneering study of Lebrun and
in terms of its composition, length, and structure. Experi- Lavery [29] as the representative structure for stretching
ments done long ago, suggested that DNA substantially effects. This study modeled the adiabatic elongation of
longer than its natural length (also referred to as over- selected DNA molecules in two modes of stretching, cor-
stretched DNA) can undergo a transition to an elongated responding to pulling on opposite 33 ends or 55 ends
structure up to twice the length of relaxed DNA [16]. This of the molecule: In the 33 stretching mode, the DNA
was also confirmed by recent single molecule stretching helix is unwound leading to a ribbon-like structure, while
experiments [1719], which showed that the molecule can in the 55 stretching mode the DNA helix contracts.
be reversibly stretched up to 90% of its natural length. Such We begin with a set of detailed calculations for the
important deformations of the double helix may occur in electronic structure of DNA bases (A,T,C,G) and repre-
biological environments. Stretching of DNA is also related sentative base pairs (AT-AT, CG-CG, AT-CG, CG-GC) in
to cellular processes, such as transcription and replication. various relative configurations, as they are likely to appear
For example, proteins often induce important local distor- in the stretched forms, These calculations are based on
tions in the double helix while they diffuse along the mol- density-functional theory [30, 31] and serve to set the stage
ecule in search of their target sequences. The electronic and for more extensive calculations which employ successive
transport properties of DNA are directly influenced by its levels of approximations necessary to handle the compu-
different conformations as well as by environmental factors, tational demands. Specifically, we extract the salient fea-
such as counterions, impurities or temperature. A full tures of electronic structure of the individual DNA bases
account of these effects based on a realistic, atomic scale and base pairs from the ab initio calculations; these are
description of the structure and the electronic properties compared to an efficient and realistic semi-empirical model
challenges the capabilities of theoretical models. [32], in order to establish the validity of the latter approach.
Theoretical efforts to understand the electronic behavior At this intermediate scale, we consider an 18 base pair
and transport in DNA can be divided into two general poly(CG)poly(CG) DNA sequence which has been stret-
categories: ched by 30%, 60% and 90% relative to the natural length of
(i) Model calculations that use effective hamiltonians the unstretched B form. The atomic structure of these
and master equations to describe the dynamics of electrons forms has been established by Lebrun and Lavery [29],
and holes in DNA (see, for instance, Refs. [2023]). Recent using empirical interatomic potentials. We next use the
results [24] have led to considerable insights concerning information from this approximate description to build an
the sequence-independent delocalization of electronic effective hamiltonian for the electronic behavior at much
states in DNA. The main limitation of such approaches lies larger scales. This allows us to describe electron localiza-
in the difficulty of determining accurate values for the tion, due to the combined effects of stretching and envi-
parameters in the effective hamiltonians. ronmental factors, over mesoscopic to macroscopic length
(ii) Ab initio calculations that can provide an accurate scales. The essence of the approach and the different scales
and detailed description of the electronic features [2527]. involved are shown schematically in Fig. 1. We emphasize
These approaches are typically limited to a small number that we address here issues related only to dry and neutral
of atoms due to computational costs, and cannot readily DNA structures, where the negatively charged groups on
handle the full complexity of DNA molecules in various the backbone are passivated by protons, conditions that are
conformations. In particular, stretching of DNA can induce relevant to the experiments we consider for comparison to
a very significant deviation from the B form which is stable our theoretical results; water molecules or counterions
under normal conditions in aqueous solution. Such struc- (such as Na+) are not considered in our calculations.
tural distortions are bound to have a profound effect on the
electronic behavior. A realistic description of these effects
makes it necessary to handle both the atomic scale features Theoretical methods
and the overall state of the macromolecule.
In the present work, we address the problem of DNA Ab initio calculations
stretching effects on the electronic states and the electron
localization by providing a bridge between the two As our first step toward establishing the electronic behavior
extremes of the length scale; a similar methodology was of dry, neutral DNA, we study the nature of electronic
recently used to study hole transfer in DNA [28]. Theo- states in individual bases and in base pairs. For these cal-
retically, there are different ways of pulling the opposite culations we used three different implementations of den-
ends of the DNA strands, leading to different stretched sity-functional theory [30]: a method that uses atomic-like
DNA forms, which are determined largely by base pair orbitals as the basis [34], one that uses plane waves [35]
reorientations. Here, we use the poly(CG)poly(CG) and a third that uses a real-space grid [36]. In all three
123
8896 J Mater Sci (2007) 42:88948903
nm
abinitio semiempirical effective ab initio methods used for the DNA bases and base pairs.
density electronic tightbinding
functional theory structure hamiltonian Accordingly, for the electronic structure calculations of
100
these structures we use an efficient semi-empirical quan-
tum-mechanical approach which employs a minimal basis
set [32]. The consistency of this approach is then verified
against the ab initio calculations. Within the semi-
empirical scheme, the electronic eigenfunctions are
10
expressed as
X
jwn i cn
m jum i 1
1500 bp
18 bp m
provide a thorough check on the consistency of various which minimizes the sum of the variances
computational schemes to reproduce the electronic features
of interest. The results are in excellent agreement across X
f ~ i j^z2 jw
hw ~ i i hw
~ i j^zjw
~ i i2 3
the three approaches. Since in these calculations there are
i
no adjustable parameters, we refer to them in the following
as ab initio results. under the constraint hw ~ j i dij where z is the position
~ i jw
along the helical axis. Similar and more general
Construction of semi-empirical model methodologies have been developed in the past for
obtaining maximally localized states from extended ones
The stretched forms contain a large number of atoms, [42, 43]. Due to the invariance of the trace, the first term in
typically beyond what can be efficiently treated with the Eq. 3 is independent of the unitary transformation and the
123
J Mater Sci (2007) 42:88948903 8897
where Slm hul jum i is the overlap matrix between the two 1X n
z z gx dx Ek 12
atomic orbitals and zlm l 2 m is the average z-value for the N k;n
atoms located at sites given by the labels l and m. Once the
localized states are constructed, the hopping parameters can be readily obtained. These quantities are essential in
can be computed as describing electron localization along the DNA double
helix under different conditions.
X
~ i jHjw
tij hw ~ j i ~ i jwn ihwn jw
en hw ~ j i 8
n
123
8898 J Mater Sci (2007) 42:88948903
1 U2
PU p exp 2 : 14
c 2p 2c
Fig. 3 The DNA base pairs AT (top) and CG (bottom), with the
Once the disorder hamiltonian is constructed with a specific atoms labeled. The purines (A, G) are on the right, the pyrimidines (T,
C) on the left. Atom labeling follows standard notation convention
set of random on-site energies, by direct diagonalization we [46]. All rotations were performed with respect to the helical axis
find the eigenstates jWi i of H Hdis (we use capital denoted by the black circle (see text)
symbols to denote the new wavefunctions from the
hamiltonian that includes the disorder term) and then
calculate the localization length defined as in Fig. 4. Specifically, the HOMO state of the AT pair is
h i1=2 exactly the same as that of the HOMO state of the isolated
Li hWi j^ njWi i2
n2 jWi i hWi j^ 15 A, and the LUMO state of AT the same as that of the
isolated T. Similarly, the HOMO state of CG is identified
where n^ is the position operator along the DNA helix with that of the isolated G and the LUMO state with that of
defined as:
X
n^ ncyn cn 16 T
n
123
J Mater Sci (2007) 42:88948903 8899
the isolated C. Thus, the purines (A or G) give rise to the These calculations were performed with the SIESTA code
HOMO state, while the pyrimidines (T or C) are respon- [34] and the relaxed configurations were used as input to
sible for the LUMO states of each pair. It is clear from the calculate the electronic structure with the other two
same figure, that essentially all atomic pz orbitals which methodologies [35, 36]. In Fig. 5 we show complete results
belong to a purine or pyrimidine contribute to the respec- from the SIESTA calculations and selected results from
tive HOMO or LUMO p state of the base pair. This is in one of the other two approaches.
agreement with calculations on the optical absorption The results of Fig. 5 show clearly that only in the region
spectra of DNA bases and base pairs [47]. A closer where the backbone distance becomes significantly smaller
inspection of Fig. 4 shows that the molecular frontier states than the equilibrium value, interaction between the two
of both AT and CG can be identified as similar contribu- bases shifts the eigenvalues of the electronic states appre-
tions (up to sign changes) from specific groups of carbon ciably, but even then the shifts are relatively small for the
and nitrogen atoms. Specifically, in the purines (A and G) frontier states. It is also noteworthy that the band gap of the
three distinct groups of atoms are mainly involved in AT pair is significantly larger (~3 eV) than that of the CG
forming the HOMO orbital and include atoms (C8N7), pair (~2 eV) and that the frontier states of CG lie within the
(C2N3) and (N1C6C5C4N9), respectively. In the band gap of the AT pair. This observation is important
pyrimidines (T and C) the main groups involved in forming because it indicates that in an arbitrary sequence of base
the LUMO orbital are two, (C4C5N1) and (N3N7C6). pairs, the frontier states will be associated with those of the
In both base pairs the atoms that are less involved in the CG pairs. This statement is verified by calculations of
frontier molecular states are the carbon atoms that form a electronic states in the AT-AT, CG-CG and AT-CG base
double bond with an oxygen atom, such as C2 of A and C pair combinations, to which we turn next.
and the four-fold bonded C7 atom of A. For more detailed comparisons, we collect in Table 1
The frontier states are very little affected when the two the eigenvalues of the frontier states for the DNA pairs and
components of the base-pair are separated along the the pair combinations, at different equilibrium configura-
direction in which they are hydrogen-bonded. To demon- tions in the three relevant variables, the backbone distance,
strate this, we show in Fig. 5 the change in the eigenvalues the axial distance and the rotation angle. Some results on
of the frontier states in AT and CG as a function of the the CG-GC base pair combination are also shown, to allow
distance between the two atoms that are bonded to the two for comparison to the poly(C)poly(G) sequence.
backbones (we call this the backbone distance). For both When two base pairs are stacked on top of each other,
base pairs the nitrogen atoms labeled N1 and N9, are the there are two degrees of freedom for motion of one relative
ones attached to the backbone (see Fig. 3). In order to to the other: a separation along the helical axis, which we
obtain realistic structures, for each value of the backbone will call axial distance, and a relative rotation around the
distance we hold the atoms of each base that are bonded to helical axis. We take the helical axis to be that which
the backbone fixed and allow all other atoms to relax fully. corresponds to stacking of successive base pairs in the B
form of the DNA double helix. According to the notation of
Fig. 3, the helical axis for both base-pairs is normal to the
line connecting atoms C4 and C6 and is closer (about one
2.5 third of their distance) to the purine atom C6. For each
configuration we fix the atoms that are bonded to the
TLUMO CLUMO backbone at a given relative position and allow all other
(eV)
AHOMO
axial distance and the rotation angle. As above, the
eigenvalues show little dependence on these two variables,
except for rather small values of the axial distance which
2.5
7 8 9 10 11 7 8 9 10 11 correspond to unphysically small separation between the
backbone distance () backbone distance ()
o o
123
8900 J Mater Sci (2007) 42:88948903
method based on a minimal atomic orbital basis [32] is in of wet DNA and in the presence of counterions, as shown
excellent agreement with the value obtained from the in Ref. [48], for a Z-DNA helix. The band gaps between all
SIESTA calculation (2.0 eV and 2.1 eV, respectively). The three ab initio methods are identical within the accuracy of
band gap is expected to be significantly smaller in the case these methods. The nature of electronic wavefunctions
obtained by the different methods is also in good qualita-
2.5
tive agreement. Accordingly, in the rest of this paper we
focus our attention to electron localization in the dry,
(eV)
TLUMO
CLUMO
0 CGCG
GHOMO
In Fig. 7, we show the unstretched and the three stretched
2.5 forms of the poly(CG)poly(CG) sequences at 30%, 60%,
2.5
90% elongation, along with the features of the frontier
(eV)
ATCG
CLUMO
states. For visualization purposes, we represent the calcu-
0 GHOMO
lated wavefunction magnitude of the frontier states by blue
2.5 (HOMO) and red (LUMO) spheres, centered at the sites
2.9 3.4 3.9 4.4 4.9 5.4 06 0 120 180 240 300 360
axial distance ()
o
angle (deg.)
where the atomic orbitals are located. The radius of the
sphere centered on a particular atom is proportional to the
Fig. 6 Eigenvalues of states in the AT-AT, CG-CG and AT-CG base magnitude of the dominant coefficient |c(n)m|2 at this site
pair combinations as a function of the distance along the helical axis (see Eq. 1), which is essentially proportional to the local
(at zero angle of rotation) and the rotation angle around the helical
axis (at the equilibrium axial distance). Lines are results from
electronic density. It is evident from this figure that the
SIESTA calculations, points are results from VASP calculations (see nature of the orbitals themselves, represented by the radii
text). In each case three states are included above and below the band of the colored spheres, does not change much in the dif-
gap. The value of the distance or the rotation angle that correspond to ferent stretched DNA forms, but the overlap between
equilibrium configurations are indicated by vertical dashed lines
(there are five almost equivalent local minima in rotation). As in
orbitals at neighboring bases is affected greatly by the
Fig. 5, frontier orbitals are identified as the corresponding orbital of amount of stretching. For the poly(CG)poly(CG)
one base only sequence shown, the HOMO orbitals are always associated
123
J Mater Sci (2007) 42:88948903 8901
involve only the G sites; those for the LUMO state involve
only the C sites. As a consistency check, we have also
calculated matrix elements for farther neighbors and found
33 55
those to be much smaller in magnitude. We have calculated
the values of t1,t2,t3 by repeating the same procedure as
(a) above for the stretched forms of the poly(CG)poly(CG)
DNA sequence. We note that if t2 = t3 = 0 electrons will
not be able to migrate along the DNA molecule even if t1 is
quite large, because at least one of the other two hops is
(b) necessary for migration (see Fig. 2). From this simple
picture, it is evident that the conductivity will be deter-
5
C G mined by which matrix element dominates. Quantitatively,
3 the bottleneck hopping matrix element is given by
(c)
3
G C 5 t maxminjt1 j; jt2 j; jt3 j: 17
Fig. 7 The DNA structures for the unstretched (top) and the different In Fig. 8 we show the value of the bottleneck hop-
amounts of stretching in the 33 and the 55 modes with features of ping matrix element calculated as a function of stretching.
the frontier orbitals described by the blue (HOMO) and red (LUMO)
spheres (see text for details). For both modes the amount of stretching
This indicates that hopping conductivity will dramatically
is (a) 30%, (b) 60%, and (c) 90% relative to the unstretched structure, decrease by several orders of magnitude upon stretching
which is the B-DNA form. The 35 orientations of the poly(CG) the molecule and that the hopping will decrease more from
poly(CG) sequence are shown in the left panel at 90% stretching, stretching in the 33 mode than in the 55 mode. This is
where the structure is easier to visualize
due to the conformational changes induced by the different
stretching modes, described earlier.
with the G sites for all the stretching modes, while the
LUMO orbitals are related to the C sites. However, as the Localization length
DNA becomes more elongated, the orbitals overlap even
less and become localized for high stretching modes. The The significant dropping of the hopping matrix elements
elongation to the overstretched form is achieved by upon stretching as described in the previous section is
changing the dihedral angle configuration of the DNA indicative of electron localization with a weak amount of
backbone, which leaves the local part of the orbitals disorder. To investigate this possibility in detail, we focus
essentially intact. Note how the orbitals rotate and spread on effects of stretching in the 33 mode. The evolution of
out as the structure is being ovestretched, following the the density of HOMO states upon stretching is shown in
rotation of bases.
We now turn to a discussion of the results for the hop-
101
ping matrix elements of Eq. 9. Our discussion here is rel-
evant to what happens when the occupation of a frontier
(2)
state is changed from complete filling (for the HOMO) or (3)
101 (2)
complete depletion (for the LUMO), that is, the physics of (2)
(2)
123
8902 J Mater Sci (2007) 42:88948903
Fig. 9; similar behavior is observed for the LUMO states. produced by the dipole potential terms, for instance, due to
The dramatic narrowing of the DOS width (equivalent to the presence of a stray water molecule situated on the
reduced dispersion in a band-structure picture) is strongly substrate roughly 15 A away from the DNA bases. We find
suggestive of electron localization [49], in this case that changing the value of c by an order of magnitude
induced by stretching. This localization length is controlled (either smaller or larger) does not affect the qualitative
by the hopping elements t, since e is the same at each site. picture presented here. Note that the localization length is
For a more quantitative description, we show in Fig. 9 not a strict function of the energy, as it depends on the
the localization length Li for each eigenstate for a 1500 disorder near where a given state happens to be localized.
base-pair DNA strand under different amounts of stretch- As the molecule is stretched, the localization length
ing. The value of L(i) for each state is obtained from Eq. 15, dramatically decreases until, for 60% stretching, the
with disorder strength c = 0.3 meV, which determines the eigenstates are completely localized on single base pairs.
width of the gaussian given in Eq. 14. This disorder The charge localization length as a function of DNA
strength is much smaller than the band width of the stretching has been recently studied in the experiment of
unstretched DNA, but becomes comparable to the band Heim et al. [33]. This study focuses on k-DNA which has
width as the molecule is stretched. The magnitude of such an irregular sequence of base pairs, and can be compared to
variations in on-site energies is consistent with those our theoretical results for poly(CG)poly(CG) recalling
that the frontier states even for a random sequence are
associated with those of the CG base-pairs. In the experi-
1000 ment, ropes of k -DNA on a substrate are overstretched by
a receding meniscus technique. The DNA ropes in this
experimental setup are slightly positively charged, corre-
sponding to a depletion of a few electrons per 1000 base
100 pairs. We suggest that this situation is approximated by the
structures of dry and neutral DNA that we considered
L (base pairs)
0.6
0.4 Summary
123
J Mater Sci (2007) 42:88948903 8903
all combinations of bases and base pairs studied here, the 14. Kasumov A, Kociak M, Gueron S, Reulet B, Volkov V, Klinov
nature of these states was not affected by separation of the D, Bouchiat H (2001) Science 291:280
15. Endres RG, Cox DL, Singh RRP (2004) Rev Mod Phys 76:195
bases or base pairs along different directions or rotation 16. Wilkins MHF, Gosling RG, Seeds WE (1951) Nature (London)
along the helical axis. 167:759
Turning to the next length scale and the semi-empirical 17. Smith SB, Cui YJ, Bustamante C (1996) Science 271:795
calculations, we have calculated the bottleneck matrix 18. Cluzel P, Lebrun A, Heller C, Lavery R, Viovy JL, Chatenay D,
Caron F (1996) Science 271:792
elements for electron hopping along the DNA molecule, as 19. Strick RT, Allemand JF, Bensimon D, Croquette V (2000) Annu
a function of stretching. These show a significant decrease Rev Biophys Biomolec Struct 29:523
with elongation of DNA, which is stronger for stretching in 20. Yamada Y (2004) Int J Mod Phys B 18:1697
the 33 mode than in the 55 mode. We were able to 21. Iguchi K (2004) Int J Mod Phys B 18:1845
22. Yi JY (2003) Phys Rev B 68:193103
show quantitatively that stretching of DNA dramatically 23. Chang CM, Neto AHC, Bishop AR (2004) Chem Phys 303:189
narrows the DOS width of frontier states. A small amount 24. Caetano RA, Schulz PA (2005) Phys Rev Lett 95:126601
of disorder produced by environmental factors will natu- 25. Barnett RN, Cleveland CL, Joy A, Landmand U, Schuster GB
rally lead to localization of the electrons along the DNA. (2001) Science 294:567
26. Artacho E, Machado M, Sanchez-Portal S, Ordejon P, Soler JM
Our estimate for the degree of localization, based on a (2003) Molec Phys 101:1587
reasonable (and quite small) amount of disorder in the 27. Alexandre SS, Artacho E, Soler JM (2003) Phys Rev Lett
on-site energies for the electron states, is in very good 91:108105
agreement with recent experimental observations. This 28. Senthilkumar K et al (2005) J Am Chem Soc 127:14894
29. Lebrun A, Lavery R (1996) Nucl Ac Res 24:2260
provides direct validation for the consistency and com- 30. Hohenberg P, Kohn W (1964) Phys Rev 136:B864; Kohn W,
pleteness of the multiscale method presented here. Sham LJ (1965) Phys Rev 140:A1133
31. Perdew JP, Zunger A (1981) Phys Rev B 23:5048
Acknowledgments The authors are grateful to Richard Lavery for 32. Elstner M, Porezag D, Jungnickel G, Elsner J, Haugk M, Frau-
providing the overstretched structures. MF acknowledges support by enheim T, Suhai S, Seifert G (1998) Phys Rev B 58:7260
Harvards Nanoscale Science and Engineering Center, funded by the 33. Heim T, Melin T, Deresmes D, Vuillaume D (2004) Appl Phys
National Science Foundation, Award Number PHY-0117795. Lett 85:2637
34. Soler JM, Artacho E, Gale JD, Garca A, Junguera J, Ordejon P,
Sanchez-Portal D (2002) J Phys Condens Matter 14:2745
35. Kresse G, Furthmuller (1996) J Phys Rev B 54:11169
References 36. Waghmare UV, Kim H, Park IJ, Modine N, Maragakis P, Kaxiras
E (2001) Computer Phys Comm 137:341
1. Watson JD, Crick FHC (1953) Nature 171:737 37. Perdew JP, Wang Y (1992) Phys Rev B 45:13244; Perdew JP,
2. Eley DD, Spivey DI (1961) Trans Faraday Soc 58:411 Burke K, Ernzerhof M (1996) Phys Rev Lett 77:3865
3. Murphy CJ, Arkin MR, Jenkins Y, Ghatlia ND, Bossmann SH, 38. Trouiller N, Martins JL (1991) Phys Rev B 43:8861
Turro NJ, Barton JK (1993) Science 262:1025 39. Vanderbilt D (1990) Phys Rev B 41:7892
4. Braun E, Eichen Y, Sivan U, Ben-Yoseph G (1998) Nature 391: 40. Hamann DR, Schluter M, Chiang C (1979) Phys Rev Lett
775 43:1494
5. De pablo P, Moreno-Herrero F, Colchero J, Herrero J, Herrero P, 41. Maragakis P, Barnett RL, Kaxiras E, Elstner M, Frauenheim T
Baro A, Ordejon P, Soler J, Artacho E (2000) Phys Rev Lett (2002) Phys Rev B 66:241104(R)
85:4992 42. Marzari N, Vanderbilt D (1997) Phys Rev B 56:12847
6. Storm AJ, Van Noort J, Devries S, Dekker C (2001) Appl Phys 43. Sgiarovello C, Peressi M, Resta R (2001) Phys Rev B 64:115202
Lett 79:3881 44. Lee Y-S, Buongiorno Nardelli M, Marzari N (2005) Phys Rev
7. Zhang Y, Austin RH, Kraeft J, Cox EC, Ong NP (2002) Phys Rev Lett 95:076804
Lett 89:198102 45. Thouless DJ (1972) J Phys C 5:77
8. Baba Y, Sekiguchi T, Shimoyama I, Hirao N, Nath KG (2006) 46. Sinden RR (1994) In: DNA Structure and Function, Academic
Phys Rev B 74:205433 Press, London
9. Porath D, Bezryadin A, Devries S, Dekker C (2000) Nature 47. Varsano D, Felice RI, Marques MAL, Rubio A (2006) J Phys
403:635 Chem B 110:7129
10. Fink H-W, Schonenberger C (1999) Nature 398:407 48. Gervasio FL, Carloni P, Parrinello M (2002) Phys Rev Lett
11. Cai L, Tabata H, Kawai T (2000) Appl Phys Lett 77:3105 89:108102
12. Tran P, Alavi B, Gruner G (2000) Phys Rev Lett 85:1564 49. Anderson PW (1958) Phys Rev 109:1492
13. Cohen H, Nogues C, Naaman R, Porath D (2005) Proc Natl Acad
Sci 102:11589
123
J Mater Sci (2007) 42:89048910
DOI 10.1007/s10853-007-1720-9
Received: 16 December 2006 / Accepted: 7 March 2007 / Published online: 17 July 2007
Springer Science+Business Media, LLC 2007
Abstract The adhesion of a living cell to an extracellular For cell-to-cell or cell-to-extracellular matrix adhesion to
matrix surface is effected through the bonding of receptor occur, binding proteins (integrins or ligands) within the cell
molecules in the cell membrane to compatible ligand membrane must aggregate into compact clusters in order to
molecules on the surface. In a series of experiments on bond specifically with molecules outside the cell. These
adhesion of cells to a substrate surface with a controlled adhesion patches can continue to grow by recruiting more
density of ligand binding sites, Arnold et al. (ChemPhys- membrane bound adhesion molecules from the vicinity
Chem 5:383, 2004) showed that tight cell adhesions could [35]. This occurs naturally because the bonding of a
form only if the areal density of binding sites on the sub- molecule in the membrane lowers the entropy of the binder
strate was higher than some critical value. Furthermore, distribution locally, giving rise to an osmotic pressure
this critical value was consistent across the four cell types gradient which, in turn, induces binder transport into the
examined in the experiments. For ligand density below the region. As is shown in [6], cell adhesion is influenced by
critical level, on the other hand, virtually no adhesions numerous factors including binding affinity between inte-
formed. In this article, we examine the competition grins and ligands, the concentration of binding molecules,
between thermal undulations of the cell membrane and its the properties of the bonding surfaces, and so on.
adhesion to the substrate. In particular, we show that Arnold and coworkers [7] have reported results of
thermal undulations destabilize membrane bonding to the experiments which revealed remarkable aspects of adhe-
substrate unless the bond spacing is below a certain level. sion at a size scale between the dimensions of the indi-
By following this line of reasoning in the context of clas- vidual bonds (a few nanometers) and the dimensions of a
sical statistical mechanics, we obtain an estimate of the tight adhesion patch (a few microns). They deposited gold
critical value of spacing which is in reasonable agreement nanodots periodically on a polyethylene glycol based
with the observations. substrate, chosen because it had no bonding affinity for cell
membranes. One binding molecule, an RGD ligand, was
then attached to each dot; a dot size of 8 nm or lessa
dimension below the diameter of a ligandwas chosen in
order to limit the number of molecules per dot to just one.
Introduction By controlling the spacing of the ligands in this way, it was
possible to observe the influence of density of ligands on
One of the most extensively studied topics in modern bio- cell adhesion. No feature of the system other than the
physics is cell adhesion, a phenomenon central to cell spacing was modified from experiment to experiment. It
motility [1], endocytosis [2], and other biological processes. was reported in [7] that a critical spacing between dots
exists for a class of cell types. For dots spaced more closely
together than this critical spacing, tight cell adhesion pat-
Y. Lin L. B. Freund (&)
ches formed. On the other hand, for spacings larger than
Division of Engineering, Brown University, 184 Hope Street,
Providence, RI 02912, USA this critical value, almost no bonds between the cell and the
e-mail: freund@brown.edu substrate were formed.
123
J Mater Sci (2007) 42:89048910 8905
123
8906 J Mater Sci (2007) 42:89048910
K=2
1
Ue Ce h00 x2 dx: 5
b
2 K=2
5
Notice that the physical dimensions of Ce are
force length2. Assuming that the binding potential is
the same for all potential binding sites, the total binding
10 energy is
2 1 0 1 2
q
Fig. 2 Shape of the energy well as described in (1) (solid line), and X
Nb
Of course, one can proceed by using (1) to describe the Following substitution of the expressions for Ue and Ub
effect of adhesion, as was done in Lin et al. [14]. However, into (7), the partition function becomes
in the current study we mainly focus on the conditions
Z 1 Z 1 Z 1
where stable adhesions are formed. In this case, the T
123
J Mater Sci (2007) 42:89048910 8907
Y
N odd
1=2 0.16
N b cb p
Z e : 11
n1;3;5... Qnn
0.14
/
Z 1 Z 1 Z 1 0.1
1
hgi ... ga1 ; a3 ; . . . ; aNodd
Z 1 1 1
0.08
eUe Ub =kT da1 da3 . . . daNodd : 12
v
!2 +
interaction potential defined in (1). It is clear that these two
u*
u X
Nodd potentials yield nearly identical results, as expected. From
r dt an simple analysis, it can be shown that the difference between
n1;3;5...
the predictions obtained by using these two potentials is a
by means of (12), and the normalized result is power series in the dimensionless factor ecb . If cb is rela-
tively large, which is the case for real cell adhesions where
X
Nodd the binding energy per bond is about Cb = 10 kT, then this
r2 Pnn
13 difference becomes very small.
d2 n1;3;5... 2Q nn The numerical results also show that only the first few
deformation modes, corresponding to relatively large
where the matrix P is defined as P AT PA, and P is a
wavelengths, contribute significantly to the value of r/d.
matrix with each component equal to 1. The ratio r/d
This outcome can be understood by realizing that the bend-
compares the standard deviation of fluctuations in the
ing energy associated with any mode, identified by wave-
membrane deflection at the binding point to the width of
number n, is proportional to n4, as indicated by (5), so the
the energy well. For adhesion to be stable, this quantity
magnitude of this mode an necessarily decreases rapidly with
should be less than 1, and perhaps substantially less. If the
increasing n because the energy cost becomes increasingly
criterion for stable adhesion is given in the form
high. This suggests that the discrete bonding assumption can
be relaxed and that the behavior can be captured by means of
r rcr 14 a model based on continuous bonding. In other words, at the
scale of long wavelengths the spacing D between binding
then the smaller the value of rcr, the higher the confidence
points becomes relatively small so that, effectively, adhesion
one has that the adhesion will remain stable. For example, if
is possible continuously along the membrane.
the distribution of membrane deflection at the binding point
Therefore, we re-examine the question on the basis of an
is assumed to be Gaussian, then choosing rcr/d = 1 gives us
assumption that the adhesion is continuous along the
about 68% confidence level that the adhesion will remain
membrane and that the density of interaction potential for
stable, whereas the confidence level increases to about 95%
adhesion is still given by (2). Now, ub represents the
if one chooses rcr/d = 1/2. The dependence of the ratio r/d
adhesion energy per unit length. In this case Cb has the
on other parameters, such as binding energy Cb or spacing
physical units of force and W has the dimensions force/
between two sites D, for example, can be studied numerically
length2. The total elastic energy is still given by (5) and the
on the basis of (13), as described by Lin et al. [14]. With the
total binding energy now takes the form
parameters chosen as k = 100, Nb = 41, ce = 104 and
cb = 6, Fig. 3 shows the sensitivity of r/d to the number of Z K=2
modes included in the calculation. In the figure, the solid line Ub ub dx: 15
K=2
represents the result of using the harmonic potential (2) and
the diamond symbols indicate results obtained by using the The partition function Z, defined in (7), becomes
123
8908 J Mater Sci (2007) 42:89048910
1=2
0.4
Nodd
X 1
r2 2ce n4
cb k 3 : 17 0.3
d2 n1;3;5... k
0.2
0 5 10 15 20 25 30
Typically, the size of a cluster of molecular bonds that Number of binding sites
makes up a focal adhesion is in the micrometer range
whereas the width of the energy well is in the nanometer Fig. 4 Dependence of the critical adhesion energy density from the
discrete bonding model, normalized by the corresponding value
range, at most. Consequently, the value of k is at least on implied by the continuous bonding model, on the number of binding
the order of 100 and usually larger. Taking the fact that k is points Nb along the membrane
large and letting Nodd ! 1, the summation of the infinite
series appearing in (17) is convergent and the result is
123
J Mater Sci (2007) 42:89048910 8909
where
/ - critical spacing
120
p4 Ce K Cb d2 Xd4
ce ; k ; cb 24 80
4 kT d kT 2kT
are dimensionless variables similar to those defined for 40
the one-dimensional case. For sufficiently large values of
Nodd, the value of the double sum in (23) can be 0
0 0.2 0.4 0.6 0.8 1
approximated accurately by a double integral. The double / - standard deviation
series is convergent so there is no disadvantage in
allowing the total number of modes Nodd to become Fig. 6 Dependence of the normalized spacing between membrane
binding sites on the normalized standard deviation in membrane
large. Accounting for the fact that the increments in both fluctuation at a particular binding site within the adhesion energy
m and n in the series is 2, the integral approximation to well, illustrated for membrane bending stiffness Ce = 20 kT and
the double series is chemical potential Gb = 10 kT
123
8910 J Mater Sci (2007) 42:89048910
To compare the model results to experimental results, it adhesion. The competition between thermal fluctuations
is necessary to choose a value for d, the half width of the and formation of adhesive bonds has been examined within
energy well representing interaction between bonding ele- the framework of classical statistical mechanics. Mathe-
ments. Suppose that we adopt the value of d 1 nm, a matical models for adhesion have been examined for two
large value relative to atomic dimensions, in order to reflect cases, one in which adhesion sites are widely spaced and
the compliance of the adhesion molecules. From a statis- therefore discrete and a second in which adhesion site
tical point of view, the value of r/d should be less than 1, spacing is small enough so that adhesion can be viewed as
and perhaps substantially less, in order for the behavior to continuous over the membrane surface. The equivalence
correspond to a stable adhesion. If we assume that the between these two cases has been demonstrated under
critical values of spacing Dcr lies within the range for circumstances for which the size of the membrane being
which the value of r/d varies from 1/2 to 1 then, on the considered is much larger than the spacing between bind-
basis of (26), the implied range of critical spacing is from ing sites.
40 nm to 160 nm. This range includes spacings compara- It has been found that background thermal undulations
ble to the critical spacing range 5873 nm observed by set an upper limit on bond site spacing for stable
Arnold et al. [7]. It is noted that the spacing in the exper- adhesions to form, in a statistical sense. A contact region
iments referred to a regular hexagonal array of binding with adhesions sites space more widely than this upper
sites. From simple geometrical considerations and the bound will likely not adhere; if bonds are formed, their
assumption that bond energy per unit area is the same for state is unstable and they will be overcome by thermal
either square or hexagonal arrays of binding sites, the fluctuations. An estimate of the critical spacing based on
critical spacing for a hexagonal array falls within the range reasonable values of system parameters is in good
43172 nm. agreement with experimental observations that have been
Finally, we note that the assumption that the outer reported on the actual spacing for a number of cell
boundaries of the square membrane are constrained against types.
transverse deflection is somewhat arbitrary. Instead, if it is
assumed that the edges are subjected to periodic boundary Acknowledgments The research support of the National Science
Foundation through the Brown University MRSEC Program (DMR-
conditions, rather than conditions of vanishing deflection,
0079964) is gratefully acknowledged.
then the final result corresponding to (26) is modified
slightly, taking the form
r References
r2 p2 2
: 27
d2 8 ce cb 1. Harris AK, Wild P, Stopak D (1980) Science 208:177
2. Alberts B, Bray D, Lewis J, Raff M, Roberts K, Watson JD
In this case, the critical spacing is implied to be in the (1989) Molecular biology of the cell, 2nd edn. Garland Publish-
ing, New York and London
range 2080 nm for a square array of binding sites, and in 3. Boulbitch A, Guttenberg Z, Sackmann E (2001) Biophys J
the range 2186 nm for a hexagonal array. Again, these 81:2743
estimates correspond to the full range 1=2\r=d\1 for the 4. Freund LB, Lin Y (2004) J Mech Phys Solids 52:2455
standard deviation in fluctuation of in transverse deflection 5. Shenoy VB, Freund LB (2005) Proc Natl Acad Sci USA
102:3213
for the representative bond site of the membrane. 6. Hersel U, Dahmen C, Kessler H (2003) Biomaterials 24:4385
7. Arnold M, Cavalcanti-Adam EA, Glass R, Blummel J, Eck W,
Kantlehner M, Kessler H, Spatz JP (2004) ChemPhysChem 5:383
Conclusions 8. Boal D (2002) Mechanics of the cell. Cambridge University
Press, Cambridge
9. de Gennes PG, Taupin C (1982) J Phys Chem 86:2294
In this article the stability of adhesive contact between a 10. Gov N, Zilman AG, Safran S (2003) Phys Rev Lett 90:228101
compliant cell membrane and a substrate has been con- 11. Lin LC-L, Brown FL (2004) Biophys J 86:764
sidered. The binding sites have been assumed to be peri- 12. Gov N, Safran SA (2004) Phys Rev E 69:011101
13. Merath R-J, Seifert U (2006) Phys Rev E 73:010401
odically distributed over the substrate surface. The 14. Lin Y, Inamdar M, Freund LB (2007) J Mech Phys Solids 55: in
membrane undergoes continual thermal fluctuations due to press
its immersion in a large heat bath, and these fluctuations 15. Goennenwein S, Tanaka M, Hu B, Moroder L, Sackmann E
are necessarily suppressed to some degree in the process of (2003) Biophys J 85:646
123
J Mater Sci (2007) 42:89118918
DOI 10.1007/s10853-007-1698-3
Received: 7 December 2006 / Accepted: 15 March 2007 / Published online: 7 August 2007
Springer Science+Business Media, LLC 2007
Abstract This article reviews current work on the of maintaining their integrity by continual repair of damage
strength and toughness of bone, its mechanisms of fracture and by adapting to changes in their stress environment.
and its ability to repair and adapt its structure. These These two aspects of boneits resistance to fracture and
properties are affected at all size levels, from the nano- its functional adaptationhave evolved hand-in-hand,
structure of collagen molecules and mineral crystals, giving us a material, which is supremely well adapted to its
through the microstructure of osteons and trabeculae, up to role.
the macroscopic shape and density variations that occur at Two themes will run through this article. The first is that
the level of a whole bone. the subject is a hierarchical one which must be viewed on
different scales: it is convenient to divide these into three,
which I will call the macro, micro, and nano scales, though
inevitably these divisions are somewhat arbitrary and
overlapping. The second theme is that, despite some unique
Introduction characteristics, bone is a structural material, which can be
investigated in the same way, using the same tools and
The strength and fracture of bone has interested scientists ideas, which we use to investigate other structural materi-
for a very long time. Galileo Galilei, in his pioneering work als. Thus the expertise of materials scientists, who under-
on mechanics, Dialogues Concerning the Two New Sci- stand the relationships between structure and mechanical
ences [1], discusses the shapes of bones in relation to their properties, is vital to our research activities in this field.
strength as an example of a problem in scaling. He worked
in the University of Padova, and it can be no coincidence
that it was there, around the same time, that the first The macro scale
anatomy theatre was built, allowing students to observe the
dissection of cadavers. Interdisciplinary research in this In the world of man-made materials it is normal to think of
field has a long history. the material and the structure (i.e. the macroscopic
In this article I will discuss the current state of our structure of the object in question) as two different things,
understanding of fracture in bone, and point out lines of the first being broadly the province of the materials sci-
developing research. It is impossible, or at least unwise, to entist whilst the second is the concern of the engineer. In
discuss the fracture of this material without also discussing natural materials, however, there can be no such easy
its repair, because our bones are living structures, capable distinction: in structures such as a human bone or tendon or
the stem of a plant, the properties of the material vary from
place to place within the structure in a smooth, continuous
D. Taylor (&) manner which we are only beginning to imitate in, for
Department of Mechanical Engineering, Trinity Centre
instance, functionally-graded materials. For example, in the
for Bioengineering, Trinity College Dublin, College Green,
Dublin 2, Ireland head of the femur (Fig. 1) the material varies from the solid
e-mail: dtaylor@tcd.ie outer shell of cortical bone to the inner network of
123
8912 J Mater Sci (2007) 42:89118918
trabecular bone, in which the size and orientation of indi- under conditions of disuse or disease, forming osteoporotic
vidual trabeculae is driven by local stress. It is difficult to bone; further increases in porosity give rise to a spongy,
know where the material stops and the structure begins, so open structure known as cancellous or trabecular bone,
we must start our discussion by considering an entire bone, which is found near joints where it provides a low-weight
for example a typical long bone such as the femur. At this solution to the problem of transferring compressive stresses
scale three factors exert an influence on the mechanical from the joints down into the bone tube, as we saw above
performance of the structure, these are: shape, size and (Fig. 1).
density. The prediction of bone shape (in practice largely con-
fined to predicting the local thickness of the cortex) and the
Shape distribution of density (or, in some cases the local structure
of the trabecular bone), has been extensively studied using
It is often said that the shapes of our bones have been FE simulations. These features are not only determined by
optimised for their purpose through long ages of evolution. evolutionary pressures, but can also change within the
In fact optimised is not quite the right word, because lifetime of an individual, as a result of changes in the stress
evolution does not act to optimise a structure but rather to environment as caused by, for example, the adoption of a
make it just good enough to give the organsm a competitive more active lifestyle or the implantation of an artificial
advantage, a better chance of survival. Alexander [3] and joint [7]. The most successful simulations are those, which
Martin [4] considered the various factors involved: cru- take account of the development of damage over time, its
cially, the aim here is not to avoid fracture at all costs, in continual repair and the functional adaptation, which will
fact a certain probability of fracture (which, for the indi- occur if the rates of damage and repair are not equal [8].
vidual, may mean death) is essential in order to reduce This work will not be discussed in detail here, as it takes
weight, ensuring a gracile support structure, one which is us out of the realm of materials science, being largely
light and efficient. Nevertheless, the optimisation problem concerned with continuum-mechanics techniques such as
is an interesting one for the student, and gives us some damage mechanics, using control variables such as strain-
useful insights. Consider the problem of designing a energy density. Some workers have achieved considerable
bonesimplified to a regular cylinder of outer radius r and predictive success (see for example [9]), though their work
thickness t: what will be the optimum value of the shape is hampered by the limited amount of experimental data
factor r/t, assuming that the aim is to minimise weight for a available and the large degree of scatter inevitable in such
given strength? Long bones are subjected to a mixture of data.
loading types: tension, compression, bending and torsion. It
is a simple matter to show that, for the cases of tension,
compression and torsion, the factor r/t has no effectit
cannot be optimised. On the other hand, under applied
bending the optimum value will be infinity: clearly no
sensible solution arises. Pauwels [5] pointed out that there
is an extra weight term, because the bone tube is full of
marrow, and this leads to a finite optimum value for r/t.
Currey and Alexander [6] developed this idea still further,
considering different modes of failure (yielding, impact,
etc); their results suggested an optimum value of approxi-
mately 2 which, despite considerable variation, is a typical
value for long bones.
Modern approaches to this problem tend to use com-
puter simulationsusually finite element (FE) analysisto
study the development of the complex shapes of real bones.
These will be discussed in the next section since they are
also capable of simulating density variations.
Density
123
J Mater Sci (2007) 42:89118918 8913
123
8914 J Mater Sci (2007) 42:89118918
Fig. 2 Cracks in bone show increasing toughness with crack length, increasing toughness is correlated to the presence of unbroken
as the graph here demonstrates for a crack growing from a notch. The ligaments which span the crack faces [16]
photographs, taken at early and late stages in the test, show how the
Fig. 3 A transverse section through a human femur, showing (on the left) two osteons with a crack running between them. At higher
magnification (on the right) the crack is seen to be white, indicating that it has been filled in by subsequent mineralization [24]
crack growth behaviour becomes completely dominated by stitial bone between osteons (Fig. 3). Much work has gone
these features and this is especially apparent under cyclic into the detection and measurement of these cracks, using
loading [27; 28]. Small fatigue cracks initiate in the inter- penetrant dyes [29]. Unlike metallic materials (but very
123
J Mater Sci (2007) 42:89118918 8915
Repair
123
8916 J Mater Sci (2007) 42:89118918
Trabecular Bone
123
J Mater Sci (2007) 42:89118918 8917
References
Fig. 9 Cross section of a rat ulna after extensive fatigue loading [45]:
microcracks can be seen (labelled lCr) along with a region which has 1. Galileo G (1638) Dialogues concerning the two new sciences.
become stained due to the presence of diffuse damage (labelled Dfdx) Elzevir, Leyden, The Netherlands
123
8918 J Mater Sci (2007) 42:89118918
2. Gray H (1959) Anatomy of the human body. Lee and Febiger, 35. Muir P, McCarthy J, Radtke CL, Markel MD, Santschi EM,
Philadelphia, USA Scollay MC, Kalscheur VL (2006) Bone 38:342
3. Alexander RM (1984) J Theo Biol 109:621 36. Frost HM (1969) Calci Tissue Res 3:211
4. Martin RB (2003) J Theo Biol 220:271 37. Burr DB (2002) Bone 30:2
5. Pauwels F(1980) Biomechanics of the Locomotor Apparatus. 38. Burr DB, Martin B (1993) J Biomech 26:613
Berlin, Springer Verlag 39. Lee TC, Staines A, Taylor D (2002) J Anatomy 201:437
6. Currey JD, Alexander RM (1985) J Zool Lon A 206:453 40. Colopy SA, Benz-Dean J, Barrett JG, Sample SJ, Lu Y, Danova
7. Huiskes R (1988) In: Fitzgerald R (ed) Non-cemented total hip NA, Kalscheur VL, Vanderby J, Markel MD, Muir P (2004) Bone
arthroplasty, Raven Press, New York, pp 283302 35:881
8. Prendergast PJ, Taylor D (1994) J Biomech 27:1067 41. Burger EH, Klein-Nulend J, Smit TH (2003) J Biomech 36:1453
9. Huiskes R, Rulmerman R, Van Lenthe GH, Janssen JD (2000) 42. Klein-Nulend J, Bacabac RG, Mullender MG (2005) Pathologie
Nature 405:704 Biologie 53:576580
10. Taylor D (2000) J Theo Biol 206:299 43. Hazenberg JG, Freeley M, Foran E, Lee TC, Taylor D (2006)
11. Taylor D, Casolari E, Bignardi C (2004) J Orthopaed Res 22:487 J Biomech (In Press)
12. Lakes RS, Nakamura S, Behiri JC, Bonfield W (1990) J Biomech 44. Verborgt O, Gibson GJ, Schaffler MB (2000) J Bone Miner Res
23:967 15:60
13. Seltzer KL, Stover SM, Taylor KT, Willits NH (1996) Veter Surg 45. Bentolila V, Boyce TM, Fyhrie DP, Drumb R, Skerry TM,
25:371 Schaffler MB (1998) Bone 23:275
14. Nalla RK, Kruzic JJ, Kinney JH, Ritchie RO (2004) Bone 46. Martin B (1995) J Orthopaed Res 13:309
35:1240 47. Hazelwood SJ, Bruce Martin R, Rashid MM, Rodrigo JJ (2001)
15. Malik CL, Stover SM, Martin RB, Gibeling JC (2003) J Biomech J Biomech 34:299
36:191 48. Taylor D, Lee TC (2003) J Anatomy 203:203
16. Nalla RK, Kruzic JJ, Kinney JH, Ritchie RO (2005) Biomaterials 49. Nyman JS, Reyes M, Wang X (2005) Micron 36:566
26:217 50. Ascenzi A, Baschieri P, Benvenuti A (1990) J Biomech 23:763
17. Bazant ZP (2004) Mater Struct 37:1 51. Vashishth D, Koontz J, Qiu SJ, Lundin-Cannon D, Yeni YN,
18. Taylor D (2004) Eng Frac Mech 71:2407 Schaffler MB, Fyhrie DP (2000) Bone 26:147
19. Nalla RK, Kinney JH, Ritchie RO (2003) Nat Mater 2:164 52. Nyman JS, Roy A, Shen X, Acuna RL, Tyler JH, Wang X (2006)
20. Vashishth D, Tanner KE, Bonfield W (2003) J Biomech 36:121 J Biomech 39:931
21. Yeni YN, Fyhrie DP (2003) J Biomech 36:1343 53. Wang X, Xiaoe LI, Shen X, Agrawal CM (2003) Ann Biomed
22. Natali AN, Meroi EA (1989) J Biomed Eng 11:266 Eng 31:1365
23. Tanabe Y (1999) In: Takahashi HE (ed) Mechanical Loading of 54. Zioupos P, Currey JD, Hamer AJ (1999) J Biomed Mater Res
Bones and Joints. Springer-Verlag, Tokyo, Japan 45:108
24. Boyde A (2003) J Anatomy 203:173 55. Eyre DR, Dickson IR, Van Ness K (1988) Biochem J 252:495
25. Hazenberg JG, Taylor D, Lee TC (2006) Biomaterials 27:2114 56. Nalla RK, Kruzic JJ, Kinney JH, Balooch M, Ager III JW,
26. Nalla RK, Kruzic JJ, Kinney JH, Ritchie RO (2005) Biomaterials Ritchie RO (2006) Mater Sci Eng C 26:1251
26:2183 57. Knott L, Whitehead CC, Fleming RH, Bailey AJ (1995) Biochem
27. OBrien FJ, Taylor D, Lee TC (2005) J Orthopaed Res 23:475 J 310:1045
28. Akkus O, Rimnac CM (2001) J Biomech 34:757 58. Akkus O, Rimnac CM (2001) J Orthopaed Res 19:927
29. Frost HM (1960) Henry Ford Hosp Med Bull 8:25 59. Akkus O, Adar F, Schaffler MB (2004) Bone 34:443
30. OBrien F, Taylor D, Lee TC (2002) J Biomech 35:523 60. Currey JD (1984) Biol Sci 304:509
31. OBrien F, Taylor D, Lee TC (2000) J Anatomy 197:413 61. Wasserman N, Yerramshetty J, Akkus O (2005) Eur J Morphol
32. Qiu S, Rao DS, Fyhrie DP, Palnitkar S, Parfitt AM (2005) Bone 42:43
37:10 62. Akkus O, Yeni YN, Wasserman N (2004) Crit Rev Biomed Eng
33. Frank JD, Ryan M, Kalscheur VL, Ruaux-Mason CP, Hozak RR, 32:379
Muir P (2002) Bone 30:201 63. Vashishth D (2005) Crit Rev Euk Gene Expr 15:343357
34. Fazzalari NL, Forwood MR, Smith K, Manthey BA, Herreen P
(1998) Bone 22:381
123
J Mater Sci (2007) 42:89198933
DOI 10.1007/s10853-007-1693-8
Received: 12 December 2006 / Accepted: 13 March 2007 / Published online: 17 July 2007
Springer Science+Business Media, LLC 2007
Abstract Understanding the relations between the into structure-function relations. We also outline the
mechanical responses of whole entities, their materials potential for future studies.
properties and their structures, is a challenge. This chal-
lenge is greatly enhanced when the material itself is
complex, and when the entity it forms has a convoluted
shape. It is for these reasons that it is still beyond the state-
of-the-art to predict and fully understand the mechanical Introduction
functions of whole biological entities such as bones and
teeth. Recent advances in optical metrology open up new Understanding the relations between the mechanical
opportunities as they enable the precise and accurate responses of whole entities, their materials properties and
mapping of the manner in which the entire surface of a their structures, is a challenge. This challenge is greatly
whole stiff mineralized tissue deforms. Furthermore these enhanced when the material itself is complex, and when the
data can be obtained non-destructively and without contact entity it forms has a convoluted shape. This situation is
with the sample. Data of this kind create the exciting almost always the case in biology and was well recognized
possibility of relating the complex distribution of by DArcy Thompson [1] in his classic publication entitled
mechanical properties of loaded biological materials such On Growth and Form. He wrote in 1917 Matter as such
as bone and teeth and their microstructures to deformations produces nothing, changes nothing, does nothing; and
and strains. Such studies could improve our understanding however convenient it may afterwards be to abbreviate our
of normal physiological processes such as skeletal aging, as nomenclature and our descriptions, we must carefully
well as disease processes such as osteoporosis. They also realize in the outset that the spermatozooan, the nucleus,
provide opportunities for engineers designing bio-inspired the chromosomes or the germ-plasma can never act as
materials to study the principles, advantages, and charac- matter alone, but only as seats of energy and as centers of
teristics of the behavior of hierarchical and multifunctional force (p. 20, volume I). Despite the years that have
materials. elapsed since 1917, and the enormous technological ad-
In this manuscript we review optical metrology meth- vances made, it is still beyond the state-of-the-art to predict
ods, highlight studies of whole body function for bones and and fully understand the mechanical functions of whole
teeth, and in particular those studies that provide insights biological entities such as bones and teeth.
Various attempts have been made to achieve this goal.
They were based mainly on measuring deformations of
R. Shahar (&)
Koret School of Veterinary Medicine, The Hebrew University of whole bones or teeth at single points, or developing con-
Jerusalem, P.O. Box 12, Rehovot 76100, Israel stitutive theoretical models that can predict the behavior of
e-mail: shahar@agri.huji.ac.il the whole bone or tooth based on the mechanical properties
of its individual constituent materials. However both
S. Weiner
Department of Structural Biology, Weizmann Institute of approaches were, for the most part, only partially suc-
Science, Rehovot 76100, Israel cessful. Recent advances in optical metrology open up new
123
8920 J Mater Sci (2007) 42:89198933
Bones and the bulk of teeth are composed of the same basic
materials [4]. Only the thin hard outer layer of teeth
(enamel) is quite different [5]. The materials that bones are
composed of are also collectively called bone, whereas in
teeth (aside from the enamel) they are collectively called
dentin. They both have the same constituents which
include the mineral, carbonated hydroxyapatite (also known
as dahllite), the framework protein, type I collagen, many
other so-called non-collagenous proteins and water [6]. Fig. 1 Hierarchical structure of bone (from Weiner and Wagner [4],
Enamel also contains carbonated hydroxyapatite, but the with permission)
crystals are orders of magnitude larger than those in bone
and dentin [7]. Furthermore, in most vertebrate teeth (except
for those of many fish), collagen is absent in enamel. these fibrils. They are 3080 nm wide and 23 nm thick
The materials of bones and teeth have hierarchical [10]. At the next hierarchical level the mineralized fibrils
structures, in that they change at different length scales (see are organized in a variety of patterns, the most common of
Fig. 1). They are also graded materials, as their composi- which is the lamella (level 4 bottom left in Figs. 1 and 2).
tion, structure, and mechanical properties may vary con- Each lamella is 23 microns thick and has a complex
tinuously or in discrete steps, from one location to another rotated plywood structure [11].
[8]. The combination of these two attributes results in a very At the next hierarchical level, lamellae are arranged in
complex material type that certainly cannot be described in one of several possible ways, depending on location and
terms of one value for a particular materials property. The species; in mature human bone the most common
following is a brief description of the hierarchical structures arrangement is concentric layers (called secondary ost-
of bone (the material), dentin, and enamel. eons). These form cylinders 150250 lm in diameter and
contain a central hollow tube 80 lm in diameter, which
The hierarchical structure of bone contain blood vessels and nerves (level 5 Fig. 1) [12]. At
the next level bone can be mostly solid (cortical) or
Bone is a term used to denote both the organ and the sponge-like (cancellous) (level 6 Fig. 1). This entire com-
material of which it is composed. The material bone is a plex arrangement, with different but characteristic 3-D
bio-composite with a complex and hierarchical structure. geometric morphologies, ultimately forms the organ bone.
The various hierarchies of bone span many different
length scales. The basic building blocks of the material The structures of tooth enamel and dentin
bone are an organic matrix, most of which is type-I col-
lagen, and a mineral phase (level 1 Fig. 1) [4]. At a higher Teeth consist of a crown (the part above the gum line) and
length scale bone is composed of mineralized collagen one or more roots (the part below the gum line). The crown
fibrils 80100 nm in diameter (level 2 Fig. 1), arranged in has an exterior layer of enamel and the root has an exterior
a regular, staggered array of collagen molecules (level 3 layer of cementum. Both parts have a continuous interior
Fig. 1) [9]. The crystals are located within and around layer of dentin, which surrounds the pulp cavity (see Fig. 3).
123
J Mater Sci (2007) 42:89198933 8921
Dentin
123
8922 J Mater Sci (2007) 42:89198933
123
J Mater Sci (2007) 42:89198933 8923
Optical metrology methods method, the structure under investigation is coated with a
photoelastic material, or a replica made of photoelastic
The main advantages of all optical metrology techniques material is produced whose geometry is identical to that of
are the full-field nature of the measurements they provide, the structure investigated. When such a sample is illumi-
and that they are non-destructive [24]. In essence a mea- nated by polarized light after loading, its birefringence
surement is obtained from each pixel in the field of view. causes the light to become refracted in the two orthogonal
The number of pixels is dependant upon optical resolution, principle stress directions. The difference in the refractive
and is usually around 105. In this regard, results obtained indices leads to a relative phase retardation between the
experimentally by optical methods are similar to results two component waves. The two waves are then brought
that would be obtained by a very large (and practically together in a polariscope, and interference takes place. As a
impossible) array of strain gauges. result fringe patterns form, which depend on the relative
Optical methods provide full-field surface measurements retardation. These fringe patterns allow the determination
of either stress (photoelasticity) or displacements and of the state of stress at the surface of the object tested.
strains (various interferometry methods and digital image Despite its appeal as a method able to map surface stress
correlation (DIC)). Full field measurements, be they distributions, photoelasticity has several inherent weak-
deformation, strain or stress, have several advantages: they nesses. When a sample is coated with a photoelastic
can identify local strain or stress peaks and gradients, which material, potential errors include a mismatch of Poissons
would be missed by measurements made with few strain ratio between the coating material and the investigated
gauges. This is particularly significant in the case of inho- structure, incorrect light incidence, uneven coating thick-
mogeneous, anisotropic, graded biological materials. They ness and the potential reinforcing effect of the photoelastic
are also compatible with the results of finite element anal- material [25]. When a replica is used, only the geometry is
ysis, and therefore facilitate model validation. Furthermore, simulated, and the complex materials properties are not
the full-field data set makes it possible to vary finite element taken into account. Furthermore, replicas or coated models
models in terms of material properties until the predicted cannot interact in a biologically relevant way with physi-
numerical results agree with measured results, providing ologic fluids, thus in essence a dry sample is tested. This
deeper insight into local material properties. further reduces the validity of the experimental results.
All optical metrology methods are based in one way or
another on interference. Interference is a phenomenon that Teeth
occurs when two or more waves overlap each other in
space. The superposition principle states that the resulting One of the first applications of 3-D photoelasticity to study
field is the sum of the original fields. The resultant intensity the behavior of whole teeth was by Johnson et al. [26].
however is not merely the algebraic sum of the intensities They measured the stress fields occurring in epoxy resin
but also depends on the phase difference between the dif- replicas of human molar teeth with different types of cavity
ferent waves. Thus interference may be constructive or preparations. They showed that when cavities are prepared
destructive. with rounded angles, stress concentration is minimized
Here we review optical metrology methods, highlight compared to cavities prepared with sharp angles.
studies of whole body function for bones and teeth, and in The nature of the stress distribution from the tooth root to
particular those studies that provide insights into structure- its supporting alveolar bone was investigated by Asundi and
function relations. We also outline the potential for future Kishen using a combination of in vivo strain gauges and
studies. in vitro photoelasticity experiments [2731]. They used
epoxy resin to create a model of a normal tooth and its sup-
Photoelasticity porting alveolar bone, which included a thin layer of silicon
rubber between the bone and the tooth to simulate the peri-
Principle of the method odontal ligament. Using digital photoelasticity, they showed
the stress distribution pattern in the tooth and supporting
Photoelasticity was one of the first optical metrology bone replica (see Fig. 5). The replicas were loaded at several
techniques used for the study of whole bones and teeth. The angles relative to the long axis of the tooth. As the loading
method relies on the property of birefringence that certain force direction formed a larger angle with the long axis of the
so-called photoelastic materials exhibit when subjected to tooth, stresses in the alveolar bone increased. They also
stress. Birefringent materials have two different refractive showed that stresses decreased from the top (cervical) to the
indices when light passes through them. Furthermore, the base (apical) region of the root, and that most of the bite
magnitude of the refractive indices at each point is directly forces along the tooth axis caused loading of the alveolar
related to the state of stress at that point. When using this bone in the cervical region, and to a lesser extent in the apical
123
8924 J Mater Sci (2007) 42:89198933
123
J Mater Sci (2007) 42:89198933 8925
non-contact methods such as HI and ESPI overcome these attached to the specimen surface by a thin layer of epoxy.
deficiencies (see below). The dentin was divided into an outer region (close to the
enamel) and an inner region. They showed that when hy-
Teeth drated, both regions underwent the same lateral strains
(along a line perpendicular to the long axis of the tooth).
A key insight into the design strategy of teeth was obtained However, when dehydrated there was a significant differ-
using Moire fringes on slices of human teeth [18]. The ence in lateral strain between the two regions (higher in the
study showed that under compression much of the strain is outer region). They found that dehydration of dentin in-
taken up in a zone some 200 microns thick that separates duced residual compressive strains in the outer region of
the stiff outer enamel layer from the crown dentin. This dentin, as was observed by Wood et al. [37] for uncon-
zone was known from hardness measurements to be much strained dentin. This showed that free water has a signifi-
softer than the enamel and the underlying dentin. It was cant impact on the stress-strain response of dentin in a
proposed that this soft zone is an important working part direction parallel to the tubules and perpendicular to them.
of the teeth in that it acts as a cushion or a gasket during The presence of water increased the toughness of dentin,
mastication. while loss of water increased its stiffness. The authors point
Huo [36] extended this study by using a finite element out that the dentinal collagen fibrils are made of microfi-
model with the same geometry as the tooth slices. Four brils separated by spaces filled with water molecules.
finite element models, using different types of material Dehydration of the dentin leads to loss of these spaces and
properties for dentin were tested. They ranged from allows the collagen polypeptide chains to contact each
homogeneous and isotropic properties to an inhomoge- other, and form molecular associations not formed in the
neous and anisotropic model. The latter model yielded hydrated state. These associations stabilize the structure
numerical results that most resembled the experimental and increase its stiffness.
data produced by Wang and Weiner [18]. Clearly the Moire interferometry has also been used to determine
model can now be used to better understand the structure the deformation gradients of enamel and dentin [39]. The
mechanical relations of teeth by systematically varying the purpose of this study was to look for a biomechanical basis
input parameters and the geometry. This is a very prom- for a significant problem affecting the teeth of humans,
ising use of a finite element model to extend the data and namely non-carious cervical lesions, by studying the strain
improve our understanding of the functional attributes of distributions in enamel and dentin in human teeth loaded in
this soft zone, or for that matter other design features of a compression. They observed that while the enamel under-
tooth or bone. went marked strain gradients in the direction perpendicular
A follow-up of the Wang and Weiner experiment was to the long axis of the tooth, the crown dentin experienced
performed by Wood et al. [37] using MI. This experiment marked strain gradients along the tooth axis (see Fig. 6).
was also designed to investigate the material properties of As load increased, the strains in the enamel increased in the
the soft zone, but using a completely different approach. area above the base of the crown (cemento-enamel junc-
Slices of the tooth crown were cut perpendicular to the long tion), while strains in the dentin increased below the
axis of the tooth. In this way the enamel formed a con- cemento-enamel junction. These unique in-plane strain
tinuous ring around the dentin. In other slices the enamel patterns led the authors to hypothesize that loading caused
was not continuous. The changes in shape of the slices by biting forces contributes to strain concentration at the
were then monitored as a function of changing humidity. junction between the crown and the root of the tooth. This
The results clearly showed that if the enamel constrained in turn could cause the formation of cavity-like lesions in
the dentin then very little deformation occurred during this area which are attributed to mechanics rather than
drying. If however the enamel was discontinuous, then bacterial activity.
much of the strain was localized in the soft zone as the This example illustrates the ability of optical metrology
dentin dried. This too therefore demonstrated the unique methods in investigations that aim to gain a fundamental
materials properties of this zone. In addition, the experi- understanding of the design strategies of mineralized
ment showed that a continuous enamel ring prevents the organs like teeth and bone. In particular the mechanical
dentin from contracting in this zone during drying. This behavior that results from the complex interplay of the
presumably implies a very strong interface between the organs external morphology and internal structure, com-
dentin and the enamel. bined with the spatial distribution of material properties
The role of water in the mechanical behavior of dentin can be measured. Such data helps explain the normal
was also investigated by Kishen and Asundi [38]. Several behavior of teeth, and the occurrence of common lesions
human incisors were ground to prepare parallel-sided such as non-carious defects in the enamel at the crown-root
longitudinal sections, and a high-frequency grating was junction.
123
8926 J Mater Sci (2007) 42:89198933
123
J Mater Sci (2007) 42:89198933 8927
information on the effect of the bones geometry on the strains. This method is based on causing interference between
strain fields. rays returned from two neighboring surface points. When two
Double-exposure holography was also used to determine such points lie on the same Cartesian axis and are very close to
bone deformation (swelling and shrinking) due to thermal each other, the relative phase change occurring before and after
stress [42]. Thermal stresses were used simply because deformation can be related to the partial derivative of the
they were simple to apply. The experimental results were deformation along that axis. One of the advantages of this setup
compared to measurements made with a thermo-graphic is its reduced sensitivity to vibrations [47].
measuring system, and a good correlation was found The potential of ESPI to measure deformations of loa-
between the temperature and displacement distributions. It ded biomineralized tissues such as teeth and bones
was also shown that the deformation fields differed increased significantly after it was demonstrated that such
depending on which point was used for the application of measurements can be made with the sample under water
the heat source. [19, 48, 49]. This is essential for the investigation of bio-
Osten and coworkers [43, 44] reviewed the potential of logical specimens. This technique allows not only the
HI in biomechanics. They described a system, which uses investigation of deformations and strains of entire organs,
two laser sources and two fast CCD cameras. The latter but also the determination of the elastic properties of
concurrently record the out-of-plane and in-plane defor- millimeter-sized samples with good precision and accuracy
mation components. The deformation fields of notched in teeth [48] and bone [49]. Furthermore, by improving the
samples of antler bone were followed dynamically at a rate signal-to-noise ratio it will be possible to determine local
of 30 hologram pairs per second. The deformation maps strain variations within these small samples and correlate
clearly demonstrated the presence of high displacement them with microstructures.
gradients (equivalent to strains) at the tip of the notch (See Electronic speckle pattern interferometry has several
Fig. 7). Further refinement of this system, using ultra-fast limitations. It is extremely sensitive to ambient vibrations,
and high-resolution cameras and use of three lasers and and requires an isolated environment, which usually
three fast cameras will allow experiments to follow fracture includes an acoustically insulated system mounted on top
phenomena in hitherto unattainable detail and obtain all 3 of a floating optical table. Applied loads must be small to
displacement components, thus significantly extending the avoid large displacements which could lead to decorrela-
potential application of HI to the study of biologic samples. tion. Therefore when larger loads are of interest, they must
be reached by a series of small and consecutive incremental
Electronic speckle pattern interferometry loads, with displacement fields determined for each incre-
ment, and then summed.
Principle of the method Another technical difficulty is the fact that fluctuations
in speckle intensities may occur during the time between
Another optical method which allows whole-field measure- obtaining the reference image and loaded image, especially
ment of displacements of optically rough surfaces is electronic when wet biologic samples are tested, resulting in decor-
speckle pattern interferometry. It is based on the observation relation. This makes data analysis very challenging. Kirk-
that when a rough surface is illuminated by a coherent source of patrick and Brooks [50], in an attempt to overcome the
light, the reflected light beams interfere locally with each other, difficulties associated with speckle fluctuations, used a
creating a grainy image termed a speckle pattern. The speckle modification of classical laser speckle data analysis, which
pattern changes when the specimen is loaded. The differences was based on a sequential sampling algorithm which is
between patterns of images obtained before and after applying insensitive to slow speckle decorrelation (relative to the
load are used to detect shifts in the phases of speckles [45, 46]. sampling rate). In a preliminary proof-of-concept type of
These phase differences correspond to surface displacements, experiment, they showed that when a sample interval of
and their magnitudes can be determined using one of several 0.02 s was used, reliable results could be obtained. They
available phase-shifting algorithms. Phase unwrapping allows applied this technique to obtain the Youngs moduli of
the determination of the entire displacement field. One of the samples of porcine cortical bone. Another approach was
main advantages of this method is its extreme sensitivity, based upon applying very small incremental loads between
which allows detection of displacements as small as tens of successive images (reference and loaded state), thus cre-
nanometers. (The limit of detection is ~k/30, were k is the ating displacements which are much smaller than the pixel
wavelength of the laser light; typically several hundreds of dimensions. In this technique each experiment consists of
nanometers). There are several additional ways to utilize many individual steps, and avoids decorrelation. Using this
speckle-based methods. One such method, called shearogra- approach, Youngs modulus (human dentin and equine
phy, deserves special mention, because it allows direct mea- cortical bone), and Poissons ratio (equine cortical bone)
surements of displacement derivatives, which can be related to were measured [48, 49].
123
8928 J Mater Sci (2007) 42:89198933
Fig. 7 Deformation of a notched piece of antler bone immersed in the phase maps and (b), (d), and (f) are the corresponding
water, measured by digital HI, at three different times (0.5, 1, and deformations in the y-direction
1.5 s after the beginning of the loading process). (a), (c), and (e) are
Last, obtaining the three components of displacement bone (mandible) resulting from two different loading loca-
requires changing the laser configuration during measure- tions. However in this preliminary study only the general
ments, and this process requires time. This makes ESPI less low-resolution pattern was evaluated and quantitative dis-
suitable for dynamic measurements. A novel approach taken placement data were not obtained. It should also be noted that
to shorten the acquisition time of in-plane and out-of-plane this set of experiments was performed on dry specimens, and
deformations was reported by Rodriguez et al. [51]. They therefore does not represent the real deformations this bone
developed an ESPI system which allowed easy, precise and will undergo under physiologic load.
rapid alteration of the direction of the laser beams illumi-
nating the surface of the object, by use of optical fibers Teeth
attached to a rotating platform to split and guide the light
beam. To demonstrate the abilities of their system, they Zaslansky et al. [19] used ESPI to measure the deformation
measured the surface displacements of a human lower jaw distributions of whole human premolar teeth (see Fig. 8).
123
J Mater Sci (2007) 42:89198933 8929
Experiments were performed in a custom-designed femora when load was applied either along the long axis of
micromechanical compression device. The entire setup was the femur by compressing the femoral head, or from medial
located within a sealed stainless-steel chamber, and the to lateral surfaces at the knee region [52, 53]. It has shown
teeth were submerged in water during the experiment (see that the two different modes of loading resulted in very
setup details in [48]). different patterns of strain distribution on the surface of the
Compressive load was applied incrementally either to a entire femur. In these studies the femora were tested dry,
localized area at the tip of the main cusp (simulating hard and therefore the results may not represent their true
food mastication) or over most of the upper region of the physiologic behavior. Studies such as this further under-
main cusp (simulating soft food mastication). The same standing of the response of whole bone to load. They also
protocol was also applied to exact replicas of the tested provide insight into the remodeling process that occurs
teeth. It was hypothesized that any difference in the when a change in loading pattern is applied to bones.
deformation patterns between the real tooth and its replica Another potential use of ESPI was recently reported by
will be the result of the tooth structure. In order to Mohr et al. [54]. They assessed the process of fracture
interpret the spatial distribution of the whole-field dis- healing by harvesting mid-sagittal sections of fracture
placements a novel approach was developed, named rel- callus (a layer of cartilage and bone produced by the body
ative displacement maps (RDM). These provide a to bridge the fracture area) from sheep 8 weeks after
reference-independent map of the 3-D surface displace- fractures were created in the middle of their sheen bone
ments (see [19] for details). (tibia). The fractures were stabilized with an external metal
The teeth and their replicas deformed in roughly the device (external fixator frames). The slices were subjected
same manner. This indicates that the overall deformation to axial compression, and the strain distribution in the
pattern of the premolar is mostly controlled by shape, and entire callus was measured by ESPI. Slices with complete
to a much lesser extent by its complex internal structure. bony bridging were shown to have higher stiffness than
An interesting difference was seen between the defor- those consisting mostly of connective tissue bridging.
mations occurring due to soft food and hard food masti- Highest strains in all sections were seen in the fracture gap.
cation. When subjected to a more concentrated load Continuation of this study could significantly further
(simulation of hard food) the tooth undergoes substantial understanding of the fracture healing process and provide
bending. Another significant finding was the much more answers to such questions as the mechanical consequences
uniform deformation of natural teeth compared to their of the sequence of events occurring during the fracture
replicas. This was attributed to the properties of the healing process, and the most suitable conditions required
enamel cap that behaved mainly as a rigid body, with for successful healing.
only moderate deformation at higher loads. One particu-
larly interesting observation was that the location of Digital image correlation
minimal deformation of the natural teeth, but not the
replicas, was in the region of contact between neighboring Principles of the method
teeth. This is presumably a built-in design feature to
minimize abrasion between teeth. Digital image correlation is an optical metrology tech-
nique which is based on entirely different principles
Bone from those of interferometry-based techniques described
thus far. Digital image correlation is performed by
Electronic speckle pattern interferometry has been used to matching high-resolution structural patterns or color-
study the three-dimensional strains on the surface of mice sprayed random image patterns of the surface of objects
123
8930 J Mater Sci (2007) 42:89198933
123
J Mater Sci (2007) 42:89198933 8931
obtain images at rates of up to 10 KHz, and readily available methodology limitations. Using optical metrology it is now
powerful computing power which can easily handle vast possible to obtain a quantitatively precise and accurate
quantities of data in real time, will allow researchers in the answer to this question. Bones can be scanned to determine
field of whole-bone and whole tooth function to use inter- with high resolution their structure (in terms of voids,
ferometry techniques and DIC to answer questions which mineral content, sizes, and orientation of cancellous struts,
seemed beyond our reach only a short time ago. etc.), and then loaded while an optical measurement
The most basic and intriguing question is that of the method is used to measure surface displacements. Since the
deformation (and resulting strain) distributions of whole methods are non-destructive, the load can be removed and
organs under various types of load. Until recently it was the bone treated to remove a predetermined amount of
considered impossible to elucidate the strains and stresses cancellous bone from a selected region (quantified by a
occurring in bones and teeth of complicated shape and post-removal microCT scan). The loading experiment can
structure. As has been shown in the preceding sections, now be repeated, and surface displacements of the modified
while optical metrology techniques show great promise in bone measured. Direct comparison can be made between
this respect, attempts to use these methodologies to the pre- and post-removal displacement and strain fields,
investigate this question are only beginning. showing quantitatively the effect of the removal of can-
Bones come in an impressively large assortment of cellous bone regions. Preliminary results of such an
shapes and sizes. Within each bone the different hierarchies experiment demonstrate an obvious difference in strain
are themselves distributed in a very varied way. There are distributions between the two states in a rat femur.
areas which are more mineralized than others, areas with One of the most outstanding attributes of all optical
secondary osteonal bone while other areas have primarily metrology methods is their compatibility with the results
parallel lamellae, areas with thick cortex and others with obtained by finite element analysis. The finite element
very thin cortex, areas with cancellous bone or without method (FEM) is a commonly used engineering numerical
cancellous bone, and so on. It is reasonable to assume that method which is used to determine the stresses, strains and
most of these variations can be rationalized in mechanical deformations of structures of complex geometry and
terms, but until now such rationalizations could only be material properties when placed under load. Parametric
based on speculation and very little experimental support. analysis of the contribution of each component can be
Testing whole bones with optical metrology methods can easily achieved by this technique. FEM is frequently used
allow a much more detailed exploration of the mechanical to model various biomineralized tissues, such as bones and
significance of the shape and structure of bones. teeth [36, 6165]. In such models, the geometry, even
The organ bone must be both stiff (to resist deformation when complicated, can be simulated very accurately by
under the action of muscles) and strong (to bear load advanced CAD tools. However the need to assign correct
without breaking). A further requirement is the basic need material properties to the elements of these models poses a
to minimize weight so as to decrease the energy cost of much greater challenge since biologic materials are
locomotion. The solution nature found to these conflicting anisotropic, inhomogeneous and graded in nature. Thus it
demands is a structural motif common to almost all bones is possible to create a model of a bone or tooth, and then
in all speciesan outer cortical shell of varying thickness, experimentally test the same bone under conditions iden-
which surrounds cancellous bone in some regions. Cortical tical to those simulated in the numerical analysis, mea-
bone is almost solid, with very few voids in it, while suring displacements with optical metrology. It is then
cancellous bone (also termed spongy or trabecular bone) possible to compare the analysis results to the experimental
consists of slender (50150 lm in diameter) and relatively results. The model can be modified and fine-tuned up to the
long (12 mm) struts surrounded by large spaces. It has point when the predicted and experimental results are
been shown that cortical and cancellous bones are made of close, indicating by reverse engineering the correct spatial
the same material, possibly with minor differences in the distribution of the material properties within the bone or
level of mineralization. However due to the different void tooth. Several simple examples of this approach were
contents, a volume of cortical bone is much stiffer provided above ([25, 36]), but the potential is far from
(10100 times) than an equal volume of cancellous bone. tapped.
Therefore it is commonly believed that most of the stiffness Optical whole-field metrology methods can also be
of bones is due to their cortical shells. The cortico-can- applied to good advantage in studying questions arising in
cellous structural arrangement raises the question: what is clinical settings. For example, dentists are not certain of the
the mechanical function of the cancellous bone compo- effect carries, or various restoration materials and methods
nent? used to treat them, have on the mechanical behavior of
Only a small number of studies tried to address this whole teeth. To answer such questions, we are currently
question experimentally, and results were limited due to testing intact premolar teeth under load with ESPI. We first
123
8932 J Mater Sci (2007) 42:89198933
123
J Mater Sci (2007) 42:89198933 8933
43. Alexeenko W, Pedrini G, Zaslansky P, Kuzmina E, Osten W, tern interferometry. 5th world congress of biomechanics, Munich,
Weiner S (2004) Digital holographic interferometry for the Germany
investigation of the elastic properties of bone. 12th international 55. Schmidt T, Tyson J, Galanulis K (2003) Exp Tech 27:22
conference on experimental mechanics, Bari, Italy 56. Zhang D, Arola DD (2004) J Biomed Opt 9:691
44. Pedrini G, Alexeenko IV, Zaslansky P, Tiziani HJ, Osten W 57. Nicolella DP, Nicholls AE, Lankford J, Davy DT (2001) J Bio-
(2005) Proc SPIEThe Int Soc Opt Eng 5776:325 mech 34:135
45. Jones R, Wykes C (1989) Holographic and speckle interferom- 58. Nicolella DP, Lankford J (2002) J Musculoskelet Neuronal
etry. Cambridge University Press, Cambridge Interact 2:261
46. Vest CM (1979) Holographic interferometry. Wiley, New York 59. Nicolella DP, Bonewald LF, Moravits DE, Lankford J (2005) Eur
47. Schmidt T, Tyson J, Galanulis K (2003) Exp Tech 27:47 J Morphol 42:23
48. Zaslansky P, Currey JD, Friesem AA, Weiner S (2005) J Biomed 60. Nicolella DP, Moravits DE, Gale AM, Bonewald LF, Lankford J
Opt 10:024020 (2006) J Biomech 39:1735
49. Shahar R, Zaslansky P, Barak M, Friesem AA, Currey JD, 61. Koca OL, Eskitascioglu G, Usumez A (2005) J Prosthet Dent
Weiner S (2006) J Biomech 40:252 93:38
50. Kirkpatrick SJ, Brooks BW (1998) J Biomed Mater Res 39:373 62. Kuijs RH, Fennis WMM, Kreulen CM, Barink M, Verdonschot N
51. Rodriguez D, Moreno V, Gallas M, Abeleira MT, Suarez D (2003) J Dent Res 82:967
(2004) Med Eng Phys 26:371 63. Simon U, Augat P, Ignatius A, Claes L (2003) J Biomech
52. Samala PR, Su M, Liu S, Jiang HH, Yokota H, Yang LX (2005) 36:1079
Strain measurement of a mouse bone by 3D-electronic speckle 64. Jaecques SVN, Van Oosterwyck H, Muraru L, Van Cleynenb-
pattern interferometry (3d_ESPI). SPIE, Bellingham, WA reugel T, De Smet E, Wevers M, Naert I, Van der Sloten J (2004)
53. Su M, Samala PR, Jiang HH, Liu S, Yang L (2005) J Hologr Biomaterials 25:1683
Speckle 2:34 65. Taylor WR, Roland E, Ploeg H, Hertig D, Klabunde R, Warner
54. Mohr M, Simon U, Claes L, Nbottlang M (2006) Full-field strain MD, Hobatho MC, Rakotomanana L, Clift SE (2002) J Biomech
acquisition on ovine fracture callus with electronic speckle pat- 35:767
123
J Mater Sci (2007) 42:89348942
DOI 10.1007/s10853-007-1648-0
Received: 16 December 2006 / Accepted: 1 March 2007 / Published online: 7 August 2007
Springer Science+Business Media, LLC 2007
Abstract Marine mussels produce an impressive adhe- must account for performance within the turbulent and
sive material for affixing themselves to rocks in the tur- harsh marine environment. We are curious to see what
bulent marine environment. This glue is generated by links exist between small-scale chemical bonds and the
application of proteins to the surface followed by extensive resulting macroscopic materials.
cross-linking to yield the final matrix. Prior studies have Perhaps the two most well studied marine biological
shown that simple oxidation or reactivity brought about by materials are barnacle cement [16] and mussel adhesive
metal ions may be key to this protein cross-linking process. [710]. In both cases, the material is produced by the
Here we have explored protein cross-linking reactivity in animal depositing a mixture of proteins onto a surface.
which combinations of metals and oxidants may display Cross-linking of the proteins then yields the final, cured
synergistic effects with respect to adhesive curing. matrix. The exact details of the cross-linking process and
Extracted adhesive proteins were mixed with a series of surface adhesion remain to be determined. In the case of
metals, oxidants, and combinations thereof. In some cases, barnacles, the proteins contain cysteine thiols that may
synergistic curing was observed. For example, we found become oxidized to disulfides for cross-linking [16].
that iron(II) ions with hydrogen peroxide brought about a Mussel adhesive proteins contain the unusual amino acid
greater degree of protein cross-linking than the sum of the 3,4-dihydroxyphenylalanine (DOPA) [710]. These DOPA
individual components. These studies were performed as residues are essential for cross-linking and can comprise
part of our efforts to provide perspectives on the connec- high levels (~1022%) of the total protein amino acid
tions between biology, chemistry, and functional materials. content. Also present in mussel adhesive at remarkably
high levels are transition metals such as iron and zinc
[1116].
The majority of studies related to cross-linking of
mussel adhesive proteins have focused upon oxidation
Introduction reactions. Oxidants such as NaIO4 [1719] and the oxi-
dizing tyrosinase enzyme [1821] readily react with
The oceans abound with a fascinating collection of unique DOPA-containing proteins, peptides, and synthetic poly-
materials, ranging from barnacle, tube worm, and oyster mers to yield cross-linked products. A recent atomic force
cements to the adhesives of sea stars, mussels, and corals. microscopy study indicates that DOPA oxidation to the
We cannot help but marvel at how these organisms design quinone may be responsible for curing the bulk material
and generate such functional materials for affixing them- [22]. Surface adhesive bonding, by contrast, may depend
selves to underwater surfaces. This material development upon the reduced (i.e., not oxidized) catechol sidechain
of DOPA interacting with the animals chosen substrate
[22].
L. M. Hight J. J. Wilker (&)
Our laboratory [2328] and others [17, 18, 2932] have
Department of Chemistry, Purdue University, West Lafayette,
IN 47907-2084, USA been exploring potential roles of metals in the formation of
e-mail: wilker@purdue.edu mussel adhesive. Metal ions such as Fe3+ have the ability to
123
J Mater Sci (2007) 42:89348942 8935
both become chelated by the DOPA sidechain [25] as well simple oxidants (e.g., H2O2, Na2S2O8) cross-linked the
as bring about DOPA oxidation [7, 33, 34]. From spec- protein, but only to a limited degree [27]. The most
troscopic and reactivity studies, we found that the iron in effective reagents were oxidizing metal ions such as Fe3+,
mussel adhesive may be bound by three DOPA-containing Cr6+ (in Cr2 O2
7 ), and Mn
7+
(in MnO 4 ) [27]. By contrast,
+ 2+ 2+
protein strands [26]. Furthermore, these Fe(DOPA)3 cen- simple metal ions (e.g., Na , Co , Ni ) did not afford any
ters can react with oxygen to generate protein-based radi- curing. For oxidizing reagents, no correlation was found
cals [24, 26]. Such oxidized protein may go on to afford between curing and reduction potentials. In general, con-
proteinprotein coupling for curing or protein-surface centration dependences were found, with higher concen-
bonding for direct adhesion. trations of a given reagent generally bringing about more
We wondered how such chemical and bonding pro- curing. Of the reagents examined that may typically be
cesses may contribute to the actual generation and perfor- available to mussels, Fe3+ provided the most pronounced
mance of mussel adhesive. In particular, we were curious cross-linking [28]. Given the high concentrations of iron in
to see how well metals or oxidants might be able to bring mussel adhesive plaques [1116], we concluded that iron
about cross-linking of mussel adhesive protein. Adhesive may be the reagent responsible for bringing together DOPA
precursor proteins were extracted from mussels prior to proteins to form this material [26, 28].
cross-linking [27, 28]. After precipitation from aqueous In general, metal ions may react with oxidants and
solutions by addition of an organic solvent and centrifu- reductants to generate reactive species. Perhaps the most
gation, a slightly viscous protein hydrogel was obtained. well known cases are from metalloenzymes. For example,
We reacted this gel with a collection of potential cross- in the cytochrome P450cam system, a metal ion (iron in a
linking agents [27, 28]. The various resulting matrices were heme) reacts with an oxidant (O2) and a reductant (elec-
examined for materials properties employing a direct trons) to generate a reactive intermediate capable of oxi-
measure of curing [27, 28]. Using an Instron materials dizing an unactivated hydrocarbon (camphor) to a product
testing system, a rod was moved into the gel at constant alcohol [35]. Small molecule systems are also well studied
velocity, all the while recording force (c.f., Fig. 1). This in which a metal complex (e.g., Fe EDTA) can react with
measurement of compressibility and shear properties an oxidant (e.g., H2O2) to produce reactive oxygen species
allows rapid analysis of many samples and provides (e.g., hydroxyl radical) [36, 37]. Thus metals and oxidants
quantitative comparisons (c.f., Table 1) [27, 28]. We have can react in synergistic ways to bring about chemistry
recently also reported on the rheological characteristics of unavailable to either component alone. Perhaps both metals
these hydrogels both before and after cross-linking [23]. and oxidants play a role in the curing of mussel adhesive.
We reacted the protein hydrogel with a collection of We found oxidized protein, in the form of an organic
metals and oxidants [27, 28]. In general, we found that radical, within fresh adhesive plaques harvested from
mussels [26]. Earlier studies showed the presence of oxi-
dase activity in the plaques [37]. Thus oxidation reactions
are clearly part of the protein curing process. However, the
active oxidant generated by mussels for this cross-linking
has not yet been identified. The usual suspects list of
common biological oxidants include oxidizing enzymes
(e.g., tyrosinase) and the most common small molecule
oxidizers found in biology such as dioxygen (O2), hydro-
gen peroxide (H2O2), superoxide (O 2 ), and hydroxyl rad-
icals (OH ) [38, 39]. Generally speaking, most protein
cross-linking processes require the presence of oxidants
[33]. With regard to mussel proteins, nearly all cross-
linking studies have focused on oxidation, with periodate
(IO4 ) [1719] and the tyrosinase enzyme [1821] com-
prising the majority of work.
Another interesting issue related to oxidation within this
biological material is that of metal ion transport. Prior to
internalization by organisms, the majority of environmental
Fig. 1 Photograph of the experimental setup used for direct iron is in the +3 oxidation state [4042]. In open ocean
measurements of curing mussel adhesive extract with various
reagents. A rod was run into the protein hydrogel at constant velocity
water, for example, nearly all iron is particulate and typi-
while force was recorded. The sample tube in the photograph is raised cally in various forms of Fe3+ [4042]. Mussels collect this
for visibility insoluble iron in their bivalve filtration system [1214].
123
8936 J Mater Sci (2007) 42:89348942
The process by which these animals transport this accu- strainer with 3.5 mm holes. Using a ceramic container with
mulated iron to their adhesive system has not been de- a diameter slightly smaller than the strainer, the pellets
scribed in detail. Generally speaking, divalent Fe2+ is more were forced through the strainer for homogenization, while
soluble and easier to transport than trivalent Fe3+ [43]. chilled with ice. Deionized water was added to the strained
Thus most organisms accumulate Fe3+ and reduce the ion pellets by the following steps: strained pellets were trans-
down to Fe2+ for actual use [43]. Given our results indi- ferred to a tared 100 mL beaker, mass was recorded and
cating that Fe3+, rather than Fe2+, appears to be required for multiplied by 1/3. This resulting value was the amount of
mussel protein cross-linking [27, 28], we can envision a deionized water added in grams to the beaker of strained
scenario in which mussels accumulate Fe3+ in their filters, pellets such that the final amount of water in the mixture
reduce the metal to Fe2+ for transport to the byssal adhesive was 25% w/w. The strained pellets and water were thor-
system, then reoxidize to Fe3+ for initiating the curing. If oughly mixed with a spatula to obtain a homogenous
such processes are taking place, an oxidant is required for mixture. This mixture was then transferred in 1011 g
the final Fe2 ! Fe3 transformation. Here we report on increments to a 10 mL plastic syringe with a 2 mm open-
the ability of oxidants to influence the curing of DOPA ing used to deposit the homogenous material into plastic
proteins by metal ions. At this time, we can only speculate 2 mL microcentrifuge tubes (9 mm in diameter, 35 mm
as to which oxidants might be pertinent to mussel adhesive. deep). During the transfer process, the mixture remaining
Thus we have focused on two: hydrogen peroxide (H2O2) in the beaker was kept on ice and covered with Parafilm
and sodium periodate (NaIO4). Hydrogen peroxide is a until placed in the syringe. The tube was capped and tapped
thermodynamically strong (E = 1.78 V H2O2/H2O) [44] on the bench 67 times to eliminate air bubbles and then
but kinetically slow oxidant found in many biological placed on ice until all of the mixture was dispensed. The
systems. When examining the oxidation of biological sample tubes were covered with aluminum foil and stored
molecules, H2O2 makes for an obvious choice. Our prior in a 4 C cold box until used, within 7 days of extraction.
studies showed that H2O2 brings about only modest curing
of mussel adhesive proteins [27, 28]. We chose NaIO4 for a Preparation of cross-linking reagent solutions
contrast to H2O2. Previously we found that this strong
oxidant (E = 1.59 V IO4 /IO3 ) [45] cured mussel For all studies discussed below, the final concentration of
adhesive proteins appreciably, to a similar degree as Fe3+ cross-linkers were 45 mM. This concentration was chosen
[27]. This reagent has also been popular with in vitro to be consistent with our prior studies [27, 28] as well as
studies of mussel adhesive protein cross-linking [1719]. close to that of the natural abundance in mussel adhesive.
As will be seen below, synergistic curing effects of metals Total iron in mussel adhesive plaques is approximately one
and oxidants were found. These results may provide in- part per thousand [11, 12] or about 20 mM when converted
sights on how mussels use chemical reactivity to generate to solution concentrations. To obtain a final cross-linker
their adhesive material. concentration of 45 mM in the gel, 1 M solutions of the
desired reagents were prepared fresh daily. In previous
work, 0.5 M solutions were used [27, 28], but experiments
Experimental discussed below require a 1 M solution to compensate for
the dilution that will occur upon addition of two reagents
Protein extractions (metal and oxidant) instead of only one. To prepare solu-
tions, the calculated amount of reagent required for a 1 M
Mussel adhesive protein was extracted from the animals solution in 5 mL volume was weighed out and placed in a
according to a literature method [46]. This extraction 10 mL volumetric flask. After placing all compounds in
method relies upon acid precipitation of most biomolecules individual flasks, 5 mL of deionized water was added and
while leaving the DOPA-containing proteins in solution the flask vortexed to aid dissolution. For solutions that did
[46]. Subsequent removal of the proteins from solution, not completely dissolve with mixing alone, mild heat was
using an organic solvent precipitant, yields a gel containing applied by hotplate for 710 min. The heated solutions
two of the protein variants, with a relative composition of included NaIO4, ZnCl2, Mn(OOCCH3)3, CuCl, and TiF3.
~80% Mefp-1 and ~20% Mefp-2 [46]. After precipitation All solutions were transferred to 15 45 mm screw thread
of these DOPA proteins from solution with acetone, cen- glass vials and capped to prevent any evaporation. For
trifugation yielded hydrogel pellets [46]. The pellets were reagents where solubility did not allow a 1 M concentra-
stored in a 4 C cold box under a blanket of deionized tion, saturated solutions were used instead. The saturated
water until used, within 7 days of extraction. Roughly 85 solutions included NaIO4, Mn(OOCCH3)3, CuCl, and TiF3.
pellets yielded enough sample to complete one set of All water used in this study was purified to at least 18 MW/
penetration tests. The pellets were placed in a plastic cm using a Barnstead Nanopure Infinity system. In our
123
J Mater Sci (2007) 42:89348942 8937
force (mN)
160
NaIO
Cross-linking reactions 4
120 2+
Cu +
80 H2O2
Into each sample tube was dispensed 1.001.05 g of the 2+
40 Cu
homogenized adhesive protein extract. Then 50 lL of a HO
1 M non-metal oxidant solution was added followed by a 0 water 2 2
50 lL solution of a 1 M metal salt. The reagents were 0 5 10 15 20 25
added to the sample tube by 100 lL glass syringes. For rod extension (mm)
controls, 100 lL of deionized water was added to the tube. Fig. 2 Extension versus force data for a rod penetrating into a protein
When adding reagent combinations, the non-metal oxidant hydrogel extracted from marine mussels. The protein gel was reacted
was consistently added first. In the case of combining a with water, Cu2+, H2O2, NaIO4, Cu2+ + H2O2, and Cu2+ + NaIO4.
metal and water, the water was added first. Each sample Higher force indicates greater cross-linking of the protein. Each trace
is averaged from multiple runs. Error bars show one standard
tube was mixed thoroughly with a microspatula for 56 s. deviation for a given data point
The spatula was then scraped on the inside rim of the tube
to return any mixture that stuck back into the sample tube.
Each tube was capped and tapped on the bench 67 times 600
to eliminate air bubbles. The sample was then allowed to 2+
react at room temperature for one hour. 500 Fe +
NaIO4
400
force (mN)
diameter rod (blank drill bit) was run into the gel mixtures 100 HO
2 2
2+
at constant velocity, all the while measuring force on the 0 Fe
rod. Previously we used a 3.5 mm rod, thus data in this 0 5 10 15 20 25
current study yields higher penetration force values [27, rod extension (mm)
28]. The theory behind this mechanical test predicts that
Fig. 3 Penetration plots for the mussel adhesive protein gel reacted
linear relationships will be found between penetration with Fe2+, H2O2, NaIO4, and combinations thereof. Each trace is
depth and force [48, 49]. Some deviations from this averaged from multiple runs and error bars are shown with one
expected linearity will be seen below and are a result of standard deviation
inhomogeneity in the samples resulting from cross-linking.
The starting protein hydrogel is homogeneous, as are many
of the cross-linked products. Extreme levels of cross-link- 700
ing, however, can yield hardened solids suspended in 600 3+
solutions. Thus the materials no longer behave ideally. We Fe +
500 NaIO4
discussed this phenomenon in our recent paper in which we
force (mN)
123
8938 J Mater Sci (2007) 42:89348942
Table 1 Penetration forces in mN observed for an extract of mussel adhesive proteins mixed with various combinations of reagents
Reagent Water n H2O2 n NaIO4 n
Water 30 12 32 34 20 10 144 57 10
NaCl 31 16 5 19 2 4 113 54 5
CaCl2 22 9 5 38 1 4 126 38 5
ZnCl2 27 12 5 42 12 5 180 42 5
AlCl3 22 15 5 29 9 5 146 51 5
Ga(NO3)3 23 10 6 35 10 5 322 89 5
CoCl2 28 16 5 24 8 5 111 4 4
TiF3 24 9 5 27 9 5 211 38 5
CuCl 30 11 5 32 6 5 221 66 5
CuCl2 29 9 5 65 22 5 215 61 5
VCl3 24 10 6 77 13 4 392 79 4
Na3VO4 1201 638 5 741 264 5 1201 619 5
Na2Cr2O7 1563 664 5 630 550 5 1859 983 6
MnCl2 26 13 6 41 12 5 626 192 5
Mn(OOCCH3)3 25 9 5 23 7 5 147 42 5
KMnO4 887 319 6 490 251 5 1646 665 6
FeCl2 28 7 5 222 25 4 478 83 4
FeCl3 134 47 5 213 63 5 560 137 5
Each value is an average of n runs and one standard deviation is shown
Reported maximum loads were taken at a final penetration between the water control and H2O2. Thus Figs. 3 and 4
depth of 20 mm. For a given experimental run on a given will not show water traces in the interest of clarity. Below
day, every potential cross-linking reagent or combination of we provide the effects of cross-linking with each of these
Table 1 was examined. The final force values shown for a oxidants in combination with various metal ions. All curing
given entry in Table 1, below, were averaged from five data are shown in Table 1. We will present the data starting
different runs each performed on five different days. Out- with low levels of curing and progress to higher degrees of
lying data points were removed if they failed a Q test at 90% cross-linking.
confidence level. Reported errors denote one standard
deviation. The number of samples used for each data point Non-redox active metal ions
is provided in Table 1. When comparing the averaged data
in the Discussion section, standard t-tests were per- For the Na+ ion, from NaCl, no difference was observed for
formed on the raw data, thereby yielding probabilities that a protein cross-linking (31 16 mN) versus the water con-
given pair of samples showed the same curing. trol (30 12 mN), as is seen in Table 1. Addition of H2O2
provided no increase (19 2 mN) and NaIO4 was higher
(113 54 mN) but no more so than NaIO4 alone
Results (144 57 mN). Similar results were found for CaCl2 and
ZnCl2. Aluminum(III) is often used, along with Ga3+, to
Curing data mimic the coordination chemistry of Fe3+ but without the
available redox chemistry. Aluminum(III) alone
In order to provide a context for the data obtained, we start (22 15 mN) or with H2O2 (29 9 mN) showed no cur-
the discussion with controls. Water added to the adhesive ing nor did Al3+ with NaIO4 (146 51 mN) beyond that of
extract showed a penetration force of 30 12 mN. the individual components. Gallium(III), however,
Hydrogen peroxide, alone, did not exhibit curing, with a appeared different. Alone Ga3+ (23 10 mN) or with
force of 34 20 mN. On its own, cross-linking was found H2O2 (35 10 mN) exhibited no notable effect. When
for NaIO4 at 144 57 mN. Figure 2 shows a penetration combined with NaIO4, hardening from Ga3+
test with water, H2O2, and NaIO4 each added to separate (322 89 mN) was near double that of the individual
extracts. As can be seen, there is no appreciable difference components summed together (144 + 23 = 167 mN).
123
J Mater Sci (2007) 42:89348942 8939
Redox active metal ions We examined the cross-linking afforded by three man-
ganese ions, Mn2+, Mn3+, and Mn7+ (Table 1). Starting
Cobalt has redox activity available, with the Co2+ and Co3+ with Mn2+ (26 13 mN) and H2O2 combined
ions most common. Results found for Co2+ were similar to (41 12 mN) no detectable effect was noted. Pronounced
that of Na+, with Co2+ alone (28 16 mN), mixed with synergistic curing, however, was seen with both Mn2+ and
H2O2 (24 8 mN), or NaIO4 (111 4 mN) no greater NaIO4 together (626 192 mN). By contrast, Mn3+
than any of the controls. Titanium also has multiple oxi- (25 9 mN) did not cure any more effectively when H2O2
dation states available (e.g., 2+, 3+, 4+). For this study we (23 7 mN) or NaIO4 (147 42 mN) were added, rela-
chose a Ti3+ salt based upon prior experience showing that tive to the oxidant-only solutions. As shown in Table 1,
TiF3 is easiest to work with in water, minimizing imme- Mn7+ brought about dramatic curing (887 319 mN). No
diate formation of TiO2 powder [52]. Titanium(III) alone enhancement was found with H2O2 (490 251 mN) but a
(24 9 mN) or with H2O2 (27 9 mN) did not cure the doubling was seen with Mn7+ and NaIO4 (1646
adhesive extract to any significant degree. With both Ti3+ 665 mN).
and NaIO4 the observed curing was high (211 38 mN) Spectroscopic studies from our laboratory have impli-
although still close to the sum of the individual components cated Fe3+ in the curing of mussel adhesive [26]. Thus we
(24 + 144 = 168 mN). were particularly interested to explore possible influences
Given that the copper-containing tyrosinase enzymes of oxidation upon iron-induced cross-linking of adhesive
can bring about oxidation of substrates such as tyrosine, proteins. Although Fe2+ did not yield any curing alone
DOPA, phenols, catechol [53], we were curious to see (28 7 mN), in combination with H2O2 much greater
about potential cross-linking from copper in conjunction penetration forces were required (222 25 mN) as can be
with oxidants. Inorganic copper can also catalyze the oxi- seen in Fig. 3. Further enhancement of curing was found
dation of small molecules such as catechol [54]. The Cu+ with Fe2+ and NaIO4 (478 83 mN). Alone, Fe3+ cured
salt CuCl provided data similar to the TiF3 case-little the protein significantly (134 47 mN) and this effect
curing alone (30 11 mN) or with H2O2 (32 6 mN). increased to a small degree when combined with H2O2
Periodate addition may have provided extra curing (213 63 mN). Both Fe3+ and IO 4 , together, showed
(221 66 mN) relative to each reagent on their own synergistic cross-linking (560 137 mN) greater than that
(30 + 144 = 174 mN), although the difference was not with Fe3+ and H2O2, presented in Fig. 4. Thus oxidants do
great. Curing from the Cu2+ ion alone (29 9 mN) was enhance the curing ability of iron. Also worth noting is
insignificant and when combined with NaIO4 (215 that, of all the metal ions examined here, only Fe2+ dis-
61 mN) was high, but not conspicuously so. For Cu2+ and played a dramatic enhancement of curing with H2O2
H2O2 together (65 22 mN), the observed penetration (Fig. 3; Table 1).
force was no greater than the sum of the components
(29 + 34 = 63), but was higher than any of the reagents Visual observations of curing
mentioned so far with added H2O2. These data can be seen
in both Fig. 2 and Table 1. The starting gel had a light tan color, a mostly thin and
Vanadium(III), similarly, did not cure alone liquidy consistency, with no clumping. To make a food
(24 10 mN) or with H2O2 (77 13 mN) much beyond analogy, the material began somewhat like apple sauce. We
the sum of each separate reagent (24 + 34 = 58 mN). A have reported on the rheological properties of this protein
mild synergistic effect was observed with V3+ and NaIO4 extract, both before and after cross-linking [23]. In general,
at 392 79 mN. On its own, Na3VO4 cured the protein if the numbers of Table 1 indicate little or no curing (e.g.,
extract substantially (1201 638 mN), consistent with ~80 mN and less), no changes in color or consistency were
prior observations of V5+ being a potent cross-linker [27]. found except for added color in the following cases:
Periodate did not contribute to curing (1201 619) nor CuCl2 + H2O2 (light brown), CoCl2 + H2O2 (slight pink),
did H2O2 (741 264). We also reported previously that VCl3 + H2O2 (grey), and Mn3+ + NaIO4 (rose). Hydrogen
Cr6+, in the dichromate form Cr2 O2 7 , is one of the peroxide brought about no observable changes on its own.
strongest curing agents we have found to date For NaIO4, alone, the protein gel took on a pinkish-brown
(1563 664 mN) [27]. Neither H2O2 (630 550 mN) color with no conspicuous change in consistency. Like-
nor NaIO4 (1859 983 mN) enhanced the Cr6+ reactivity wise, the metal ions and NaIO4 with curing under
significantly. Indeed, like what was found with V5+, the ~220 mN and lower, along with Ga3 IO 4 , were pink-
metal with H2O2 actually showed less curing than the ish-brown. Reactions with color changes, but without ma-
sum of the two parts. However, such decreases may be a jor viscosity increases were FeCl3 (black), FeCl3 + H2O2
result of limitations in the methods employed here (vide (dark brown), FeCl2 + H2O2 (dark brown), and TiF3 +
infra). NaIO4 (reddish brown). The latter two samples separated
123
8940 J Mater Sci (2007) 42:89348942
somewhat, with a darker color solid residing below a earlier starting material, at this time we do not have a
lighter colored solution above. satisfactory explanation for previously observed Mn3+
The powerful oxidant Na2Cr2O7, alone or with H2O2, cross-linking. Also earlier we noted that Ca2+ (21 1 mN)
generated a yellow-tan extremely viscous gel that stuck to cured slightly versus the water control (14 3). Here Ca2+
the spatulas when handled. Addition of NaIO4 formed an (22 9 mN) was not notably different from the water
orange-tan solid with a yellow solution above. For KMnO4, control (30 12 mN). In both cases the differences were
alone or with H2O2, a colorless solution was found to small. According to a Student t-test, there is a 21.5%
reside above a black gel containing tan flecks. With probability of the Ca2+ and water control results being the
KMnO4 and NaIO4, the clear solution was present above a same.
purple, then red solid. In all cases, the increase in viscosity
was immediate and extreme. With Na3VO4, alone, also a Potential curing synergy between metals and oxidants
strong oxidant and cross-linker, a brown and then dark
green, very viscous and sticky gel was noted with a clear The V3+ ion with IO 4 (392 79 mN) cured significantly
solution above. Similar results were found with Na3- more than either component alone (24 10 and
VO4 + NaIO4 and VCl3 + NaIO4. A combination Na3VO4 144 57 mN), thereby indicating the presence of syner-
and H2O2 showed a pink to yellow to green to brown to gistic curing. Only a 1.5% probability exists that V3+ with
yellowbrown transformation within 20 s, also producing a IO
4 cured to the same degree as IO4 alone, according to
viscous gel with a light yellow solution above. Combined a Student t-test. Hydrogen peroxide did not provide an
with NaIO4, MnCl2 showed a substantial increase in analogous effect with V3+. By contrast, the V5+ ion did
viscosity, a red-brown gel, and a colorless solution above. not enhance curing when H2O2 or NaIO4 were added.
Both FeCl2 and FeCl3 mixed with NaIO4 yielded viscous However, V5+ curing was very pronounced
yellowbrownred gels. (1201 638 mN). As noted above, the extreme cross-
linking yielded separation between the solution and solid,
thereby making measurements of the homogeneous gel
Discussion difficult. Thus a limit of our method may be found here.
Were enhanced cross-linking to occur by addition of an
Comparisons with prior data oxidant, such reactivity may not be observable. This
separation of solid and solution also became problematic
The general trends of curing reported here agree with our when we measured the rheological properties of protein
earlier data [27, 28]. For example, H2O2 cross-linking is gels cross-linked at the extreme limits [23]. The current
minimal and that of IO 4 and Fe
3+
are appreciable. Some lack of observed curing enhancement with oxidant
specific changes, however, could be noted in that reagents addition to V5+ does not necessarily mean such an
such as H2O2 showed higher force values here enhancement does not exist.
(34 20 mN) than earlier (20 4 mN). These differences Curing observed from VCl3 + NaIO4 may be a result of
can be attributed to incremental improvements we made to partial V3 ! V5 oxidation brought about by the perio-
our test procedure. The most conspicuous change was use date. The resulting V5+ could then cross-link in a manner
of a 5 mm penetration rod here versus a 3.5 mm rod pre- analogous to that shown for Na3VO4 alone. Prior work
viously. Thus the force data tend to now be higher. Other showed no reactivity from V4+. Oxidation of V3+ to V5+ is
changes in the procedure worth noting include extracting thermodynamically uphill by ~1.3 V, based upon the sum
the protein from mussel feet collected in a different season of reduction potentials for the VO2+/V3+ and
and a slightly different method of mixing reagents into the HV2 O 5 =HV2 O7
3
couples [44, 55]. Reduction of IO 4
gel extract. Previously we massed the pellets, strained them occurs at a comparable potential, at 1.59 V for IO
4 /IO3
3 5
to homogeneity, and added water for normalization of [45], thereby indicating such V ! V oxidations are
volumes based upon the pre-strained mass. Here we feasible. The degree of cross-linking from V3+ and NaIO4
avoided variations from loss in the strainer by using the (392 79 mN) was less than that of V5+ alone
post-strained mass for all calculations [27]. (1201 638 mN, same as V3+ and NaIO4 at 4.5%) indi-
The only significant observed variation from our pre- cating that if such metal-based oxidation took place, the
vious work is that of Mn3+ alone. Here curing was not reaction did not go to completion. Alternatively, a more
observed for this ion, but previously it was. Earlier we complex cross-linking mechanism, possibly incorporating
noted a similar darkening and thickening of the sample for enzyme catalysis principles, could be at play. We have
both Mn(OOCCH3)3 and KMnO4. As will be discussed tested the ability of oxidizing enzymes to cure this adhesive
below, Mn2+ + NaIO4 curing may be a result of initial protein abstract [27]. The lack of observed hardening with
oxidation to Mn7+. Beyond possible Mn7+ impurities in the enzymes may have been a result of the enzymes having
123
J Mater Sci (2007) 42:89348942 8941
poor access to substrate when suspended in the viscous (213 63 mN) is nearly identical to that of Fe2+ + H2O2
gelatinous extract [27]. (222 25 mN), with a 33.6% chance of the values being
For the manganese ions, Mn7+ exhibited curing at the identical. Also similar, at 93.4% probability, are Fe3+ with
high extreme, characteristic of oxidizing metal ions. No NaIO4 (560 137 mN) and Fe2+ with NaIO4
additional curing was noted with H2O2 or NaIO4 although (478 83 mN), shown in Fig. 4. However, the high degree
that from Mn7+, alone, may be approaching the limit of of curing found with Fe3+ and NaIO4 cannot be easily
what is observable. Manganese(III) did not cure beyond the attributed to simple oxidation of the metal center. Thus a
appropriate control reactions. By contrast, Mn2+ provided synergistic effect of protein cross-linking appears to be
some interesting results. Alone (26 13 mN) or with H2O2 present.
(41 12 mN) Mn2+ induced no cross-linking. An obvious Although Fe3+, alone, may be a more effective curing
synergistic effect was detected when NaIO4 was added agent than Fe2+, the aqueous solubility of Fe2+ is generally
(626 192 mN), with only a 0.7% probability of this greater. Iron in marine environments is readily available at
hardening being the same as NaIO4 alone. With no curing ~10 parts per billion [56], predominantly existing as Fe3+
for Mn3+ or Mn4+ ions, oxidation of Mn2+ to Mn3+ or Mn4+ and predominantly as insoluble, particulate species
cannot be involved. Oxidation of Mn2+ all the way up to [4042]. Mussels are known to collect iron from sea water
Mn7+, at 1.51 V, is thermodynamically possible [44] with with their bivalve filtration system, accumulating high
a strong 1.59 V (for IO
4 /IO3 ) oxidant such as NaIO4 [45]. concentrations in their filter [12, 13]. How this particulate
Thus the curing seen for Mn2+ + NaIO4 may simply be a iron transfers to the adhesion system is not known. If the
result of partial oxidation to Mn7+. However, a lack of animal were to internalize the metal for transport, reduction
added curing for Mn3+ with NaIO4 implies that this ion was to more soluble Fe2+ may be required [57]. Our results here
not brought to Mn7+. A more complex interplay between show that extrusion of only this Fe2+ along with the
metal, oxidant, and protein may well be at work here. adhesive proteins would not bring about curing and adhe-
Perhaps the most unexpected observation found is that sive production. If mussels apply oxidants along with the
of enhanced cross-linking for Ga3+ (23 10 mN alone) Fe2+ and protein, oxidation to Fe3+ could then result,
with added IO 4 (322 89 mN). This apparent enhanced thereby initiating the adhesive curing process. Such a
curing of Ga3+ and IO 4 has a 10.2% probability of being mechanism is only speculation at this stage. However, we
the same as only IO 4 . The Ga
3+
ion is inert to redox [26] and others [37] have found oxidative activity in
chemistry except under extreme conditions. Typically, adhesive plaques.
when such an ion binds to oxidizable ligands such as thi-
olates, electron density is removed from the ligand and
placed onto the metal ion, thereby making the ligand more Conclusions
difficult to oxidize. Thus we might expect no major
enhancement of Ga3+-induced curing with an added oxi- Mussels are among the rich collection of marine species
dant. A contrasting case may occur when a redox active that use clever and unique chemistry for the construction of
metal ion such as Fe3+ binds to a redox active ligand. Metal macroscopic structures. The results shown above are pre-
reduction and ligand oxidation can take place. If such sented as part of our efforts to describe the links between
oxidation of the ligand takes place within the framework biology, chemistry, and materials. Combinations of metals
of a DOPA protein, for example, cross-linking may then and oxidants do appear to exhibit synergy with respect to
begin. At this time we have no suitable explanation for the adhesive curing. We find that the effects of some metal-
Ga3+ + NaIO4 result. induced cross-linking is enhanced with added oxidants to
In terms of potential synergistic curing when both a degrees greater than the sum of the individual components.
metal and oxidant are present, Fe2+ provides a nice Metal concentrations in mussel adhesive are quite high and,
example. Figure 3 shows that the combination of FeCl2 and perhaps, can generate all the cross-linking needed by the
H2O2 at 222 25 mN at 20 mm extension is clearly animal. Metal-induced cross-linking by metals such as iron
greater than the simple sum of Fe2+ (28 7 mN) and H2O2 requires the oxidized ion (i.e., Fe3+ and not Fe2+). However
(34 20 mN). The likelihood of overlap between these the reduced forms are often easier to transport in biological
values is only 0.1%. Similarly Fe2+ and NaIO4 at contexts [57]. Perhaps oxidants enhance the curing ability
478 83 mN is enhanced significantly over only Fe2+ and of metal ions by generating sufficient concentrations of
NaIO4 summed (28 + 144 = 172 mN), with only a 0.6% the particular ion most needed for materials generation.
probability of the values being coincident. These oxidants Oxidants could be in the form of small molecules such as
may start by bringing about an Fe2+ Fe3+ reaction. H2O2 or even enzymes. These results raise the possibility
We know that Fe3+ is an effective cross-linker (Fig. 4; that mussels may use a combination of metals and oxidants
Table 1). Indeed, the cross-linking of Fe3+ + H2O2 to generate their impressive adhesive material.
123
8942 J Mater Sci (2007) 42:89348942
Acknowledgements LMH was a summer researcher in the Purdue 27. Monahan J, Wilker JJ (2004) Langmuir 20:3724
University Chemical Biology Research Experience for Undergradu- 28. Monahan J, Wilker JJ (2003) Chem Commun 1672
ates program, sponsored by the National Science Foundation. This 29. Taylor SW, Luther GW III, Waite JH (1994) Inorg Chem 33:5819
work was also supported by the Lord Corporation and the Office of 30. Taylor SW, Chase DB, Emptage MH, Nelson MJ, Waite JH
Naval Research. (1996) Inorg Chem 35:7572
31. Frank BP, Belfort G (2001) Langmuir 17:1905
32. Frank BP, Belfort G (2002) Biotechnol Prog 18:580
33. Waite JH (1992) Biol Bull 183:178
References
34. Rzepecki LM, Waite JH (1991) In: Scheuer PJ (ed) Bioorganic
marine chemistry. Springer-Verlag
1. Naldrett MJ, Kaplan DL (1997) Marine Biol 127:629 35. Denisov IG, Makris TM, Sligar SG, Schlichting I (2005) Chem
2. Naldrett MJ (1993) J Mar Biol Assoc UK 73:689 Rev 105:2253
3. Dougherty WJ (1990) J Crustacean Biol 10:469 36. Gozzo F (2001) J Mol Catal A 171:1
4. Kamino K, Inoue K, Maruyama T, Takamatsu N, Harayama S, 37. Waite JH (1985) J Mar Biol Assoc UK 65:359
Shizuri Y (2000) J Biol Chem 275:27360 38. Mishra B, Hassan PA, Priyadarsini KI, Mohan H (2005) J Phys
5. Kamino K, Odo S, Maruyama Y (1996) Biol Bull 190:403 Chem B 109:12718
6. Yamamoto H, Nagai A (1992) Marine Chem 37:131 39. Halliwell B (1990) Free Radic Res Commun 9:1
7. Deming TJ (1999) Curr Opin Chem Biol 3:100 40. Sunda WG (2001) In: Turner DR, Hunter KA (ed) The biogeo-
8. Waite JH, Andersen NH, Jewhurst S, Sun C (2005) J Adhes chemistry of iron in seawater. Wiley
81:297 41. Ussher SJ, Achterberg EP, Worsfold PJ (2004) Environ Chem
9. Waite JH (2002) Integr Comp Biol 42:1172 1:67
10. Waite JH, Lichtenegger HC, Stucky GD, Hansma P (2004) 42. Achterberg EP, Holland TW, Bowie AR, Fauzi R, Mantoura C,
Biochemistry 43:7653 Worsfold PJ (2001) Anal Chim Acta 442:1
11. Coombs TL, Keller PJ (1981) Aquat Toxicol 1:291 43. Richardson DR (2005) Curr Med Chem 12:2711
12. George SG, Pirie BJS, Coombs TL (1976) J Exp Mar Biol 23:71 44. Lide DR (1992) CRC handbook of chemistry and physics, 73rd
13. Pentreath RJ (1973) J Mar Biol Assoc UK 53:127 edn. CRC Press, Boca Raton
14. Hobden DJ (1967) J Mar Biol Assoc UK 47:597 45. Bratsch SG (1989) J Phys Chem Ref Data 18:1
15. Szefer P, Geldon J, Ahmed-Ali A, Osuna FP, Ruiz-Fernandes 46. Waite JH (1995) Methods Enzymol 258:1
AC, Galvan SRG (1998) Environ Int 24:359 47. Bourne MC (1966) J Food Sci 31:282
16. Koide M, Lee DS, Goldberg ED (1982) Estuar Coastal Shelf Sci 48. Bourne MC (1967) J Food Sci 32:601
15:679 49. Bourne MC (1975) J Texture Stud 5:459
17. Fant C, Elwing H, Hook F (2002) Biomacromolecules 3:732 50. Bourne MC (1982) In: Food texture and viscosity: concept and
18. Fant C, Sott K, Elwing H, Hook F (2000) Biofouling 16:119 measurement. New York
19. Suci PA, Geesey GG (2000) J Colloid Interface Sci 230:340 51. deMan JM (1969) J Texture Stud 1:114
20. Hansen DC, Corcoran SG, Waite JH (1998) Langmuir 14:1139 52. Sever MJ, Wilker JJ (2004) Dalton Trans 1061
21. Burzio LA, Burzio VA, Pardo L, Burzio LO (2000) Comp Bio- 53. Decker H, Tuczek F (2000) Trends Biochem Sci 25:392
chem Physiol B 126:383 54. Balla J, Kiss T, Jameson RF (1992) Inorg Chem 31:58
22. Lee H, Scherer NF, Messersmith PB (2006) Proc Natl Acad Sci 55. Bard AJ, Parsons R, Jordan J (1985) In: Standard potentials in
USA 103:12999 aqueous solution. New York
23. Loizou E, Weisser JT, Dundigalla A, Porcar L, Schmidt G, 56. Donat JR, Bruland KW (1995) In: Salbu B, Steinnes E (ed) Trace
Wilker JJ (2006) Macromol Biosci 6:711 elements in natural waters. CRC Press
24. Weisser JT, Nilges MJ, Sever MJ, Wilker JJ (2006) Inorg Chem 57. Crichton RR, Wilmet S, Legssyer R, Ward RJ (2002) J Inorg
45:7736 Biochem 91:9
25. Sever MJ, Wilker JJ (2006) Dalton Trans 813
26. Sever MJ, Weisser JT, Monahan J, Srinivasan S, Wilker JJ (2004)
Angew Chem Int Ed 43:448
123
J Mater Sci (2007) 42:89438956
DOI 10.1007/s10853-007-1649-z
Received: 21 December 2006 / Accepted: 1 March 2007 / Published online: 7 August 2007
Springer Science+Business Media, LLC 2007
Background
123
8944 J Mater Sci (2007) 42:89438956
120
[8] showed that the preCOLs organize themselves into
hexagonal (6 + 1) bundles that are banana-shaped
Nominal Stress [MPa]
100
Distal ~ 1 cm Proximal
(Fig. 3b). The presence of two bumps along the preCOLs
d ~ 0.1 mm
80 has also been observed. AFM images of a stretched byssal
60 thread suggest that the macroscopic strain is accommo-
40
~ 2 cm Distal dated by both the elimination of the tilt seen in the
Proximal unstrained thread and the unraveling of the flanking
20
Plaque globular domains. The initial angular or bent structure
0
0 0.5 1 1.5 2 2.5
may act to lock the neighboring fibres and prevent
Eng.Strain intermolecular shear/sliding within the bundles.
123
J Mater Sci (2007) 42:89438956 8945
Distal region
80
70
Nominal Stress [MPa]
60
1000 mm/min
50
40
30
20 10 mm/min
10
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Eng. Strain
Fig. 4 Nominal stress-engineering strain behavior in uniaxial tension Gosline [5]). The true stresstrue strain behavior of polyurethane
of the distal region of Mytilus californianus at extension rates of PU1 at quasi-static and high strain rate regimes is shown on the right
10 mm/min and 1,000/mm (left) (data replotted from Carrington and (data replotted from Yi et al. [9])
123
8946 J Mater Sci (2007) 42:89438956
Distal region
1.2
(a) Max. strain= 0.08 (b) Max. strain= 0.16 Polyurethane
Force [N]
0.4
0
1.2
(c) Max. strain=0.35 (d) Max. strain= 0.65
Force [N]
0.8
0.4
True Strain
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0 0.1 0. 2 0. 3 0. 4 0. 5 0. 6
Eng. Strain Eng. Strain
Fig. 5 Force-engineering strain behavior of distal portions of byssal (extension rate = 10 mm/min). The true stresstrue strain behavior of
threads of Mytilus californianus at cyclic uniaxial tests (data replotted polyurethane PU1 at cyclic uniaxial compression tests at engineering
from Carrington and Gosline [5]). Four different strain levels are strain rate _ 0:063 /s is shown on the right (data replotted from Yi
considered: (a) 0.08 (extension rate = 5 mm/min), (b) 0.16 (extension et al. [9])
rate = 5 mm/min), (c) 0.35 (extension rate = 10 mm/min), (d) 0.65
In addition, after unloading, the distal region increases material microstructure (see The representative vol-
in stiffness with time and gradually returns to its initial ume element (RVE) section).
structure and mechanical behavior [5] (see Fig. 6). the constitutive model for the stressstrain behavior of
each phase. For this study, the distal portion of the
Micromechanical model byssal thread will be regarded as a two-phase compos-
ite: the constitutive model for the tissue will be
To understand the mechanisms that govern the behavior of presented in Material constitutive behavior section,
the distal portion of byssal threads, a micromechanical whereas the fluid filling the cavities will be simply
model, representative of the underlying structure, is modeled as a hydrostatic incompressible fluid.
developed. The model framework permits us to derive the the constitutive model of the boundaries between
macroscopic mechanical response from the material different phases. Pressure equilibrium is enforced
microstructure. The development of a micromechanical across the fluid/tissue interface.
model requires: the computation of the macroscopic mechanical behav-
ior of the material based on the response of the RVE
the geometric definition of a representative volume (see RVE loading and macroscopic response
element (RVE) which captures the main features of the section).
123
J Mater Sci (2007) 42:89438956 8947
(where xj is the position vector and n the outward normal), higher order invariants is generally weak, so that the strain
allowing us to work in terms of tractions on the boundary. energy can be simplified as
i
Material constitutive behavior W Wiso I1 ; I3 Wani I4 : 9
The byssal thread of marine mussels is a fiber-reinforced where the isotropic term Wiso reflects the contribution of
composite material, in which a microfibrillar matrix is the microfibrillar matrix and the anisotropic term Wani
reinforced by stronger fibers. Following pioneering work of captures the stiffening effect of the preferentially aligned
Spencer [14], the mechanical response of hyper-elastic fibrous network.
123
8948 J Mater Sci (2007) 42:89438956
Force [pN]
mechanically nearly-incompressible Neo-Hookean mate- 30
rial. To include compressibility a bulk strain energy term is
A
added to the strain energy [22, 23], so that the total strain
20
energy of the material is given by
l K 10
Wiso I1 3 l log J J 12 ; 10 A B C D
2 2
0
where l and K denote the shear and bulk modulus, 0 20 40 60 80 100 120 140 160 180 200
respectively. The first Piola-Kirchhoff stress tensor Siso is Extension [nm]
found by differentiation with respect to C
Fig. 8 Typical saw-tooth profile obtained from single molecule
force-extension tests of protein molecules. (a) The molecule is stretch
oW until the held force is such that unfolding becomes highly probable.
S 2F ; 11 (b) Unfolding of a domain abruptly reduces the holding force, since
oC
additional contour length is released. (cd) The molecule is stretched
yielding again until a force is reached where the next domain unfolds. (From
[24])
123
J Mater Sci (2007) 42:89438956 8949
To capture the initial orthotropy of the network with a Fiber strain energy: The three-dimensional stressstrain
preferred fiber orientation, an eight-chain orthotropic unit behavior of the fibrous network can be determined
cell is employed, as proposed by Bischoff et al. [15, 16]. employing the orthotropic eight-chain network together
The eight-chain network considers an idealized structure of with a representation of the axial force-stretch behavior
8 fibers located along the diagonals of a prismatic cell (with of a bent protein fiber containing folded domains. In
dimensions a, b and c along the material axes singled out determining the axial force-stretch behavior of such fi-
by the unit vectors ^ea , ^eb and ^ec , respectively) (Fig. 10). bers, we need to derive an expression for the force
However, in this study the material axes will be aligned displacement relation when axially stretching the fiber by
with the coordinate axes singled out by the unit vectors ^e1 , displacing the two end points and thus increasing the
^e2 and ^e3 and we will assume a = c, so that the initial cell end-to-end distance from r0 to r. The fiber stretching is
can be completely defined by the undeformed end-to-end thus given by kf = r/r0 and may be multiplicatively
distance r0 and the average fiber orientation h0, decomposed into unbending and axial stretching com-
p ponents following a KronerLee decomposition [28, 29]
ac 2r0 sin h0 ; b 2r0 cos h0 : 14 (Fig. 11),
Therefore, the initial direction of each fiber in the cell is kf ksf kuf ; 18
described by the unit vectors
where kuf and ksf denote the stretch accommodated by
1 5
v0 v0 v01^e1 v02^e2 v01^e3 ; unbending and stretching, respectively. The force-stretch
2 6 behavior will consist of an initially compliant region where
v0 v0 v01^e1 v02^e2 v01^e3 ;
3 7
15 the axial stretch is primarily accommodated by the
v0 v0 v01^e1 v02^e2 v01^e3 ; unbending of the fiber; as the fiber straightens, the force-
4 8
v0 v0 v01^e1 v02^e2 v01^e3 ; stretch behavior stiffens, since the fiber will now be
directly axially stretched.
where
Idealizing the bent fiber to consist of two long, straight,
stiff segments of length L1 and L2 joined by a bend
1 as shown in Fig. 12, an expression for the fiber strain
v01 p sin h0 ; v02 cos h0 : 16
2 energy contribution due to unbending is obtained (see
The long, continuous nature of the fibers permits the Appendix A)
assumption of affine deformation, giving the fiber stretch
EI
k(i)
f of the i-th fiber to be wuf a a0 2 19
2q0 a0
where EI is the bending stiffness of the fiber, q0 the initial
bend radius and a0 = a10 + a20 and a = a1 + a2 the initial
banana and current bend angle, respectively, given by
b
u
f
x2 unbending
s
f
stretching
x1 initial configuration
a f= f f
s u
x3
current configuration
a
Fig. 11 Schematic of decomposition of kf into unbending and
Fig. 10 Eight-chain, three-dimensional orthotropic unit cell stretching parts
123
8950 J Mater Sci (2007) 42:89438956
Ar
2q3
4L22 kuf 2 r02 kuf 2 r02 L22 L21
2
_ f c xu
4 5; Lu au Lmax Lt exp ; 25
a1 arcsin 2ku r0 L1
kB T
f
20
where
kuf 2 r02 L22 L21
a2 arccos 2kuf r0 L2 :
_ DGu
Models treating the chain as an entropic wormlike chain au L0u exp 26
kB T
(WLC) and incorporating additional stretching terms
appropriate for enthalpic contributions have been is a lumped parameter. In Eqs. 25 and 26 L_0u denoted the
developed for polymers [30, 31]. However, here, the pre-exponential factor, DGu the energy barrier to unfolding,
increase of the fiber contour length during unfolding is xu the width of the activation barrier, Lt the current fibril
taken to be gradual and slow, so that its thermal contour length, Lmax the maximum fibril contour length, T
fluctuations may be neglected. Thus, the fiber strain the absolute temperature and kB the Boltzmann constant.
energy contribution due to axial stretching is simply Therefore, the fiber contour length is updated as
taken as
LtDt Lt L_u Dt: 27
EA 2 s
wsf ks r k 12 ; 21 The limited stressstrain data available at different rates
2L 0 f
makes it difficult to more fully assess the approximation of
where EA denotes the fiber axial stiffness and L the fiber
an Eyring model; more data is needed to either confirm this
contour length. The axial force versus extension
approach or to point to other process models.
relationship is obtained via f ow
or , yielding Note that refolding is an extremely low probability
during stretching and thus is neglected. This will be dis-
2EI kuf r0 a a0 cussed later.
fu r
h i2 ; 22
a0 q0 Fibrous network stressstrain behavior: The strain en-
4L22 kuf 2 r02 kuf 2 r02 L22 L21
ergy density of the fibrous network is determined using
both the expression for the fibril strain energy and the or-
and
thotropic eight-chain network structure.
When the cell is stretched, its strain energy is simply
EA given by the sum of the strain energy of the eight con-
fs r0 ksf 1; 23
L stituent fibers
and fiber equilibrium requires that
X
4
i
Wcell 2 wkf ; 28
f f u f s: 24 i1
14
Therefore, employing Eqs. 18 and 2224, kuf and ksf can be (the fact that the stretch of fibers v0 is the same as in
58
solved. fibers v0 has been used), so that the strain energy
Unfolding criterion: The unfolding is a thermally acti- density is given by
vated process whereby stress on the folded domains lowers
the energy barrier of unfolding and hence increases its Wcell m X 4
i
probability. It is introduced to the fibrous network by Wani wkf ; 29
abc 4 i1
specifying the evolution of the contour length of its con-
stituents. Here, following Qi et al. [32], the rate depen- where m denotes the fiber density (number of fibers per unit
dence of unfolding is captured by an Eyring model [33, 34] reference volume). At this point, the first Piola-Kirchhoff
123
J Mater Sci (2007) 42:89438956 8951
q q
m X4
1 EA 2 si r0 H02 L210 H02 L220 : 39
rani r k 1 vi vi : 34
4J i1 ki L 0 f
f
Unfolding
123
8952 J Mater Sci (2007) 42:89438956
parameters and also describes additional tests that are the banana shaped protein fiber; as the fiber straightens and
needed to better facilitate the identification of these it is stretched directly, the response becomes much stiffer.
material parameters. Finally, it is important to note that the model also captures
the rate dependence of the unfolding process.
Constitutive model Figure 14 shows the material behavior for the case of
simple shear loading, where
In this section, the response of the constitutive model for
the distal tissue described in Material constitutive X X0 tan c x02 ^e1 ; 41
behavior section is investigated.
where c denotes the shear angle. In this case four fibers of
Uniaxial tension the eight-fiber network are observed to extend (denoted
with Fiber A in Fig. 14) and four are observed to com-
Figure 13 demonstrates the ability of the constitutive press (denoted with Fiber B in Fig. 14), where the four
model to capture the characteristic features of material compressed fibers begin to elongate at the largest strains
response to large uniaxial tensile deformations: initial lin- since they have also rotated. When a fiber is compressed,
ear response, yielding, stretch-induced softening and the unfolding process is not activated and therefore the rate
strain hardening. The evolution of the contour length is dependent aspect of the response of that fiber is not acti-
also reported in the plot, showing the link between the vated. All fibers are observed to rotate towards the shear
material response and the unfolding process. The transition direction further demonstrating how macroscopic defor-
from the initially stiff to a very compliant response is due mation is accommodated by both rotation and extension/
to the start of the unfolding process, whereas strain hard- compression of the fibers.
ening begins when the unfolding process terminates. Fig- In the case of simple shear, it is not just the azimuthal
ure 13 also shows an initially (at strain levels lower than orientation (i.e., the azimuth angle with respect to the fiber
0.01) compliant response generated by the unbending of axis given by the angle h0 in Fig. 15left) of the fiber that
90 70 1.33/s
Nominal Stress [Mpa]
0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Eng. Strain Eng. Strain
380
Fiber A
Nominal Shear Stress [Mpa]
123
J Mater Sci (2007) 42:89438956 8953
fc [N]
and 1/s in the case of a 16-chain 0.3 G
x3
network (left). On the right, the 25
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
determines its stretch. Therefore, in addition to an eight- Figure 16 shows the simulation (for both the constitu-
chain arrangement of fibers with initial azimuthal orienta- tive and the micromechanical model) and experimental
tion of 25, a case considering a cone of 16 fibers oriented results for uniaxial tension at a strain rate of 0.0133/s and
at the azimuthal angle of 25 with the same overall fiber 1.33/s. The simulations results agree very well with the
density is also subjected to simple shear. Figure 15 shows experimental data and correctly capture the rate dependent
the material behavior for this case of 16-fiber cone oriented behavior of the stressstrain behavior. Using the micro-
at h0 = 25 and the fibers equally spaced around the cone graph of Fig. 2 as a guide, the cavity is taken to be ellip-
(see detail in Fig. 15). In the uniaxial tension case the soidal with major and minor semi-axes a and b equal to 425
material response obtained employing the 16-chain net- and 75 nm, respectively (see detail in Fig. 16), whereas the
work is exactly the same as the one obtained employing the dimensions of the RVE are taken as 575 and 150 nm, so
eight-chain network. In contrast, when simple shear load- that the volume fraction of cavities f is equal to 0.12. In the
ing is considered, the responses given by the two different left part of Fig. 16 the cavities are taken to be water filled,
networks is slightly different, since the constituents fibers whereas the right part of Fig. 16 shows the material
are loaded differently (Fig. 15right). behavior comparing the cases when taking the cavities to
be empty versus water-filled.
For the case of uniaxial tension, the water-filled and
Micromechanical model empty cavities representations yield very similar stress
strain behavior for the thread. This nearly identical behavior
A two-dimensional axisymmetric finite element model of is a result of the cavity not providing any shear stiffness
the RVE of the distal segment of the thread has been (whether water filled or empty), for this case of uniaxial
constructed in the commercial code ABAQUS with a mesh tension, noting that uniaxial tension primarily samples the
constructed of 4-node, bilinear, hybrid, axisymmetric ele- shear stiffness of a material for materials possessing a low
ments (CAX4H). The fluid-filled cavity has been modeled shear modulus relative to the bulk modulus. In contrast, in
employing hydrostatic fluid elements (FAX2) and the fluid the case of a hydrostatic loading situation, the two cases
has been assumed to be incompressible with a density of would give very different results, since we would be sam-
q = 1,000 kg/m3. pling the volumetric stiffness of the thread.
60 1.33/s 60
tension at _ 0:0133 /s and
_ 1:33 /s. Experimental and
Nominal Stress [Mpa]
50 50 1.33/s
Aa
123
8954 J Mater Sci (2007) 42:89438956
In Figs. 17 and 18, the contours of the axial stress and Conclusions
axial strain, respectively, are plotted in the RVE for both
_ 0:0133 /s and _ 1:33 /s at strain level equal to 0.2, A constitutive model for the hysteretic large strain behavior
0.4 and 0.6. of the distal portion of mussel byssus threads has been
Simulations on cyclic loading-unloading tests were presented, providing new insight into its rate dependent
conducted at a strain rate _ 0:0133 /s and _ 1:33 /s and behavior. The constitutive model results demonstrate that
Fig. 19 shows cyclic loading to different maximum strains. the yielding and the following plateau region in the
The unloading-reloading behavior shown in Fig. 19 pro- stressstrain curve correspond to simultaneous fiber
vides the key that the yield is not a classic plasticity type stretching and module unfolding taking place in the dif-
of yield, but it is a structural breakdown that changes the ferent constituents fibers. The rate effects on the stress
hyperelastic behavior due to unfolding. strain behavior due to the rate-dependence of unfolding are
60
Max. strain = 0.35 Max. strain = 0.65
40
20
(c) (d)
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Eng. Strain Eng. Strain
123
J Mater Sci (2007) 42:89438956 8955
also captured by the model. However the model does not 2q3
4L22 k2 r02 k2 r02 L22 L21 2
capture the difference between the unloading and reloading a1 arcsin4 5;
path seen in the cyclic tests. During unloading the unfolded 2kr0 L1
A:4
domains partially refolds, giving a reloading path some- 2 2
k r0 L22 L21
what stiffer than the unloading path (depending on the hold a2 arccos :
time between unloading and reloading). This important 2kr0 L2
aspect of material behavior shall be incorporated into fu- The strain energy of the fiber due to the unbending of the
ture extensions of the model. The model shall also be ex- bend is given by
tended to the proximal region and adhesive plaque, so that
it will be possible to use model-based FEM simulations to 1
systematically investigate the behavior of multiple threads wu EIDj2 q0 a0 ; A:5
2
connected to the animal and their load sharing mecha-
nisms, which will ultimately provide insights into the where EI is the bending stiffness and Dj denotes the
adhesive mechanisms of this biologically engineered sys- change in curvature of the bend
tem.
Finally, we would like to remark the fact that the pro- 1 1 a a0
Dj : A:6
posed constitutive model for mussel byssus threads is a q q0 q0 a0
starting point to understand the role of modular protein in
Substitution of Eq. A.6 into Eq. A.5 yields
many biological materials, including muscles [35, 36], red
blood cell membranes [37, 38] (M Arslan submitted),
EI
spider silk [39] and organic matrix in nacre [27, 40]. wu a a0 2 ; A:7
2q0 a0
Acknowledgments This research was supported by the Dupont
so that the fiber force-stretch behavior is given by
MIT Alliance.
123
8956 J Mater Sci (2007) 42:89438956
123
J Mater Sci (2007) 42:89578965
DOI 10.1007/s10853-007-1653-3
Received: 13 December 2006 / Accepted: 1 March 2007 / Published online: 19 July 2007
Springer Science+Business Media, LLC 2007
Abstract Tendons are multi-level fibre-reinforced com- component to tendon response, which may indicate a
posites, designed to transmit muscle forces to the skeleton. helical organisation to the tendon matrix.
During physiological loading, tendons experience tensile
loads, which are transmitted through the structure to the
cells, where they may initiate mechanotransduction path-
ways. The current study examines the structural reorgani-
sation and resulting local strain fields within the tendon Introduction
matrix under tensile load. It uses confocal microscopy to
photobleached a grid onto the collagen and image its Tendons are fibrous connective tissues, specialised in the
deformation under the application of incremental tensile function of uniaxial, tensile force transmission [1]. Struc-
strain. Six parameters are used to quantify fibril and fibre turally, tendons are natural, multi-level fibre-reinforced
movement and examine the mechanisms of extension composites, built from the basic collagen molecule into
employed by fascicles. units of increasing diameter [2, 3]. The fibres are pre-
Results demonstrated an inhomogeneous strain response dominantly collagen type I, comprising between 70% and
throughout the matrix and large variability between sam- 90% of the tendon dry mass, depending on the location and
ples. Local strains in the loading axis were significantly functional role of the tendon within the body and the
smaller than the applied values. However, large compres- species of test subject [4, 5]. The collagen units are gen-
sive strains, perpendicular to the loading axis, were re- erally aligned in the direction of loading and are the
corded. The average Poissons ratio (0.8) suggested cells principle load-bearing components of tendon, conferring
may experience significant compression during loading. much of its mechanical strength [6]. At the molecular level,
Deflection of the grid lines, indicating sliding between individual tropocollagen molecules, approximately 280 nm
collagen fibres, and rotation of the grid were also recorded. in length, are packed at repeating 64 nm intervals to form
These data highlight the non-homogenous strain environ- microfibrils, which aggregate into fibrils [7, 8]. Structural
ment of fascicles and provide further evidence for fibre integrity of the fibril is ensured by inter-molecular and
sliding under tensile load. They also suggested a rotary inter-microfibrillar cross-links, which aid strain transfer
between the discontinuous collagen molecules [9, 10].
The fibrils aggregate within the proteoglycan rich non-
V. W. T. Cheng H. R. C. Screen (&) collagenous matrix to build fibres and fascicles, the higher
Medical Engineering Division, Department of Engineering,
order structural levels of the tendon. However, the detailed
Queen Mary, University of London, Mile End Road,
London E1 4NS, UK organisation of these levels and the mechanisms by which
e-mail: H.R.C.Screen@qmul.ac.uk they facilitate the transfer of tensile load are less clearly
characterised. A number of studies have attempted to
V. W. T. Cheng H. R. C. Screen
determine the length and continuity of the collagen fibrils,
IRC in Biomedical Materials, Department of Engineering,
Queen Mary, University of London, Mile End Road, although their large aspect ratio and imaging limitations
London E1 4NS, UK have precluded a conclusive answer. Scanning electron
123
8958 J Mater Sci (2007) 42:89578965
microscopy images, showing very few fibril ends in mature controlling the binding of fluorophores to particular mol-
tissue, have led some authors to conclude that the fibrils are ecules within a sample, specific structures can be visual-
continuous throughout the length of the tendon [11]. ised. Photobleaching techniques, which further utilise this
However, analysis of the composition and arrangement of principle, have gained considerable popularity in recent
the non-collagenous matrix between fibrils, and an exam- years, with the development of techniques such as fluo-
ination local mechanical properties have led other authors rescent recovery after photobleaching (FRAP) and fluo-
to hypothesise a model with discontinuous fibrils embed- rescent loss in photobleaching (FLIP) for the analysis of
ded within a functional non-collagenous matrix that di- cell dynamics [24]. In typical experiments a fluorescent
rectly facilitates strain transfer [1214]. labelled protein is incorporated into a sample and small
Electron microscopy studies have located the proteo- target regions of the dye then photobleached by repeated
glycan decorin, bridging adjacent collagen fibrils within exposure to high intensity light from a laser beam. At high
tendon [15]. This has led to the development of a shape intensities, the fluorophores are over excited causing them
molecule hypothesis, in which decorin is responsible for to permanently lose their ability to fluoresce. If the sample
controlling shear resistance and maintaining structural is then excited by a lower intensity laser, the undamaged
integrity between adjacent discontinuous fibrils [16]. De- fluorophores in the non-photobleached areas, will still
corin has a horseshoe shaped core protein that binds around fluoresce clearly demarcating the two regions. The mobility
collagen fibrils and a single, orthogonally aligned dermatan of cellular components can be examined through moni-
sulphate glycosaminoglycan (GAG) chain which bridges toring the movement of the remaining fluorescing fluoro-
adjacent fibrils [13, 17]. It is hypothesised that the phores [25]. Although most commonly utilised to image
recruitment of a large number of weak van der Waals cellular structures, it is possible to use photobleached lines
interactions between dermatan sulphate chains can provide as markers of strain within tissue samples, by staining the
sufficient shear resistance to control strain transfer between complete tissue matrix with a fluorescent dye and then
fibrils [16, 18]. selectively photobleaching regions to providing well con-
X-ray diffraction studies also support the discontinuous trolled areas of the sample that will not fluoresce when the
fibril hypothesis, reporting that only 40% of the total strain fluorophores are excited. The current study utilises this
applied to a tendon fascicle can be accounted for by fibril technique to examine strain transfer through tendon fasci-
extension, indicating the remaining 60% must occur cles, testing the hypothesis that sliding will occur at both
through relative movement between the higher structural the fibril and fibre levels of the tendon hierarchy.
levels [19, 20]. In agreement, confocal microscopy studies
implicate fibre sliding as a dominant extension mechanism
for tendon and other hierarchical collagenous materials [14, Materials and methods
21]. Using the tenocyte cell nuclei as markers of fibre
motion, relative sliding of adjacent fibres has been visual- For all experiments, fascicles were taken from the proximal
ised under the application of tensile load. In addition, fur- end of the tails of male Wistar rats aged between 4 and
ther studies have demonstrated that altering the composition 8 months. Fascicles, approximately 300 lm in diameter,
of the non-collagenous proteoglycan matrix influences the were teased from the tails within five hours of sacrifice, cut
relationships between fibril and fibre sliding within tendon to lengths of around 45 mm, and maintained in a moistened
fascicles, supporting a possible role for proteoglycans in state, in paper towel soaked with phosphate buffered saline
regulating strain transfer mechanisms [22]. These studies (PBS; Sigma, Poole, UK).
provide important data relating to cellular deformation and
mechanotransduction; however the relationships between Straining system
the cells and the collagen matrix will influence the observed
response. In order to examine the matrix response directly, a Testing utilised a custom designed tensile straining rig
recent study has utilised a fluorescent photobleaching (Fig. 1), able to load fascicles while in situ on the stage of a
technique to image the movement of the collagen fibrils confocal microscope (TCS SP2, Leica Microsystems
themselves and examine the strain response of the tissue GmbH, Wetzlar, Germany). A series of preliminary tests
[23]. The authors photobleached lines through the collagen were carried out to validate the system for the testing of
matrix prior to the application of tensile strain. From images tendon fascicles and to determine the loading environment.
of the lines under strain, they have suggested that fibril The rig and the grips are made from stainless steel, to
sliding dominates extension, shielding the tenocytes from minimise corrosion and facilitate ease of cleaning and
excessive shear strains. sterilisation. The grips are shaped as triangular wedges
Fluorescent microscopy exploits the ability to visualise (Fig. 1b) in order to hold the fascicle adjacent to the glass
fluorophores by exciting them with laser light. By coverslip against the base of the rig, enabling the objective
123
J Mater Sci (2007) 42:89578965 8959
Microstrain analysis
123
8960 J Mater Sci (2007) 42:89578965
series of ten images were taken at 15 lm increments and Tukey post-hoc tests for analysing multiple compari-
through the thickness of the sample, at the 8% applied sons of mean were applied. Statistical significance was set
strain increment. at p < 0.05.
Image analysis
Results
Images were processed using the analysis software Image J
(1.34s, National Institute of Health, USA) following a Straining system
similar procedure to that previously described by Bruehl-
mann and co-workers [23]. Two identical copies of each The grip displacement at each strain increment, measured
image were filtered using Gaussian blur filters with radii of from the 4 magnification brightfield images, closely fol-
10 and 20 pixels. The images were subtracted before lowed that applied to the stepper motors, with a variability
thresholding and skeletonising, to produce a single pixel between predicted and measured displacement of 5 lm
trace of the photobleached grid following the midline of the (3.3%). The grips successfully held the individual fascicles
original conformation (Fig. 2a). Any dirt or noise not related and no slippage occurred in any of the samples. Twelve of
to the photobleached region was removed, and the pixel the samples (80%) failed in the central region at a distance
co-ordinates for the grid converted to data points for analysis from each grip, and failure occurred through a slow
in Excel (Excel 2003, Microsoft Corp., Redmond, WA). unravelling of the collagen fibres.
Six parameters, shown in Fig. 2b, were qualitatively Examination of the fascicles under 10 magnification
analysed to describe the in situ micromechanical environ- brightfield light microscopy demonstrated that the samples
ment. were aligned in the grips, and that the collagen fibres were
fully aligned in the direction of loading (Fig. 3). At the
x strain (100Dx/x) Percentage strain in x direction (with
gross level that fascicle was loaded uniformally and main-
axis of loading)
tained the same orientation throughout testing. The crimped
y strain (100Dy/y) Percentage strain in y direction
fibres were seen to straighten and extend and no rotation
(perpendicular to loading axis)
was visible. Sample failure was visualised as a pull out of
y angle (hy) Angle of perpendicular lines relative to the y
adjacent groups of fibres. Commencing at a strain increment
direction
between 9% and 13%, visible relaxation of the matrix be-
x angle (hx) Angle of parallel lines relative to the x
came apparent, followed by sliding between groups of fibres
direction
at random locations across the sample width. The same
Deflection (d) Step deflection along the length of the y line
sliding behaviour was visible along the length of the sam-
Displacement (D) Displacement between ends of y line
ple, and failure was finally seen to occur through complete
along the x axis
pull out of adjacent fibre groups, occurring at the 14%
increment in the majority (70%) of samples.
Statistical analysis
Microstrain results
Each parameter was analysed for the effect of increasing
strain application using StatsDirect (2005, StatsDirect Ltd., A series of images from an example fascicle at 0%, 4% and
Cheshire). Analysis of Variance (ANOVA) statistical tests 8% strain are shown in Fig. 4. Distortion of the photo-
x + x
Applied Load
123
J Mater Sci (2007) 42:89578965 8961
(a) 1
0.8
0.4
0.2
0
0 2 4 6 8
Applied Strain (%)
Fig. 3 Typical image of an unstrained fascicle under brightfield
imaging, as taken at the start of every test. Image shows the alignment (b) 0
of the collagen fibres at the start of loading 0 2 4 6 8
-2
123
8962 J Mater Sci (2007) 42:89578965
16
Discussion
Mean Total x displacement (um)
14
The confocal rig used for the current study provides a
12
mechanism of incrementally loading tendon fascicles in a
10
controlled manner, whilst visualising their in situ micro-
8
mechancial environment. Using the rat tail tendon model,
6
direct staining and photobleaching of the collagen matrix
4
was employed to provide a clear mechanism of monitoring
2
collagen behaviour directly. Validation experiments con-
0
0 2 4 6 8 firmed a controlled loading environment, in which the
Applied Strain (%) fascicles were securely gripped and subjected to uniform
loading at the gross level, enabling the sub-structural
Fig. 7 Analysis of the total displacement (D) enabled across the
response to be examined.
length of the grid. Data depicts mean and standard error (n = 20)
The use of the two viable fluorescent dyes, 5-DTAF and
Acridine Orange, enabled both the cells and the photo-
The gridlines perpendicular to the loading axis demon- bleached grid to be viewed simultaneously. The cells
strated a rotation, relative to their original orientation (hy). provided an important marker to ensure all images were
Rotational direction varied between samples, hence taken at the same focal depth, and the ability to image the
absolute values are reported, in order to assess the extent of cells and grid at the same wavelength ensured a minimal
the reorientation. The reorientation hy demonstrated a sta- imaging period at each increment. The suitability of both
tistically significant trend of increasing with applied strain stains has been demonstrated in previous publications, with
to reach a mean absolute value of 6.2 (Fig. 6b). 5-DTAF successfully staining collagen fibres in vitro and
The angular reorientation of the parallel grid lines (hx) in vivo [23, 26, 27], and Acridine Orange staining tenocyte
followed the direction of hy for each individual fascicle but nuclei [14, 28]. However, the two dyes are rarely used
was consistently statistically smaller. The mean absolute together, as they share similar absorption/emission max-
angle followed a statistically significant trend of increasing ima. Results demonstrated competitive staining between
123
J Mater Sci (2007) 42:89578965 8963
Table 1 Summary of mean data and range of values recorded for each deformation parameter at the 8% applied load increment
Values at 8% applied strain x strain (%) y strain (%) Deflection, d (lm) Angle, hy () Angle hx () Displacement, D (lm)
Sample number 20 11 20 11 20 20
Minimum value 0.17 0.48 0.73 1.05 0.84 1.7
Maximum 2.49 11.16 16.05 15.3 3.35 35.6
Mean value 0.83 6.84 6.22 6.2 1.9 12.3
Standard error 0.16 2.8 0.83 1.3 0.37 2.7
(a)
y studies by the authors have demonstrated a swelling of
x fascicles in response to incubation in different solutions,
and a subsequent change in the local strain parameters [22].
z
It is not possible to carry out these analyses without an
incubation step, and the fascicles in the current study have
(b) y strain (%) x strain (%) been treated in a similar fashion to the control (unswollen)
0
-12 -10 -8 -6 -4 -2 0 2 4
fascicles in the previous work. The differences recorded
-1 between fascicles in the earlier work were attributed to the
-2
longer twenty-four hour incubation period, however, it
z i n c r e me n t ( i n c r e a s i n g d e pt h )
123
8964 J Mater Sci (2007) 42:89578965
indicative of a reduction in fibre diameter as the fascicle would not have been able to differentiate between these
was subjected to tensile loading (Fig. 5b). The mean two parameters, hence the previous data may well have
Poissons ratio, calculated to be 0.8, suggests that there been ascribed to rotation as well as fibre sliding and may
may be significant compressive forces generated within the provide further explanation of why cells are seen to fall in
matrix under physiological loading conditions, that could and out of focus.
be transmitted to the cells, influencing the mechanotrans- In the current study, rotation was measured indepen-
duction pathways. One possible explanation for the large dently, through an analysis of angular reorientation of grid
Poissons ratio is the exudation of fluid from the extra- lines, both perpendicular and parallel to the loading axis.
cellular matrix during stretching, as previously reported Reorientation of perpendicular lines, reached a mean angle
[31, 32]. Electron microscopy has demonstrated inter- of 6.2 and coupled with reorientation and distortion of the
fibrillar space within the structure, which may well parallel grid lines, indicates there may be a rotary element
facilitate this response [22]. A visible improvement in the to the sample strain response, supporting the hypothesis of
definition of the collagen fibrils with increasing strain was a helical tendon structure. Helical crimp patterns have
also noted, and images demonstrated fibrils falling in and previously been reported in tendon [35, 36]. Using inter-
out of the focal plane. Both of these factors may be asso- ference and polarised light microscopy, de Campos Vidal
ciated with fibre alignment to the axis of loading and result [36] demonstrated wide variations in geometric features
in a reduction of interfibrillar space. and birefringence in the crimp structure of bovine and rat
The compressive strains in the fascicle matrix are larger tail tendon sections, which was attributed to a spiralling
than those previously reported for compressive nucleus fibril arrangement and a helical supra-organisation. In
strains under tensile load [28]. However, previous cell addition, Kannus [37] described how fascicles, or tertiary
mechanics studies have demonstrated that the nucleus of tendon bundles, frequently showed a spiral formation along
connective tissue cells is generally stiffer than the sur- the course of the tendon. The chirality of the collagen
rounding cytoplasm [33], hence strains may be represen- molecule has led a number of authors to hypothesise a
tative of those experienced by the cell body. Compression superstructure showing elements of handedness, as a
of the tendon ultrastructure under tensile load has also been helical organisation would improve stability of higher
described by Wang and co-workers [34] using laser Raman order structures through reciprocity in winding patterns
spectroscopy to examine the response of rat tail tendon [3840]. This has caused some controversy, with other
fascicles to uniaxial tension. The authors reported a rise in authors suggesting that the reduced stiffness associated
wavelength shift after the toe region of the loading curve, with a helical structure would be too inefficient for the
indicating lateral compression of the collagen fibre, pos- tendon to effectively transfer force [41]. Although such a
sibly associated with the reorganization of the chemical structure would imply a less efficient means of transferring
groups (particularly carbonyls) aligned orthogonally to the muscle action to the skeleton, a helical organisation would
axis of applied strain. be better suited to resist flexion, lateral compression or
Deflection (d) was employed as a measure of relative multidirectional deformation, all of which have been
sliding between collagen structures. Most samples dem- reported to occur within tendon [4].
onstrated a single deflection across the width of the 100 lm This study has concentrated on extension mechanisms
grid, indicative of sliding behaviour at the fibre level. enabled by fibril and fibre movement within a single focal
However, the limited capability to image cells meant it was plane. However, as a three dimensional structure, it would
not possible to confirm that this point of deflection was at a be anticipated that similar extension mechanisms should be
fibre edge. Deflections of up to 16 lm were seen in sam- seen in the z-plane. The z-series of images taken at 8%
ples, with a mean value of 6.2 lm at 8% applied strain. applied strain confirmed that strain dispersal through the
This accounts for less than 1% of the applied displacement, matrix incorporated relative sliding between fibres in the
however the grid covered approximately one third of the z-plane also. In agreement with data relating to individual
fascicle width, so further deflection sites may be present xy planes (Fig. 5), greater variation was recorded in the y
across the width. These values were smaller than those strains. However, it was notable that the extent of the
previously reported for sliding behaviour between rows of variability equated to approximately a quarter of the mean
tenocytes [14, 22]. However, it is notable that the sliding strain value for that parameter.
values previously reported by Screen and co-workers clo- Sliding has been implicated in a number of microme-
sely match the maximum displacement measurement D chanical studies as a mechanism for strain transfer and the
taken in the current study. The displacement D incorpo- subsequent extension of tendon fascicles [14, 20, 21]. The
rated both rotation and displacement of the grid and current study provides further evidence of sliding behav-
reached a mean maximum value of 12.2 lm at 8% applied iour within the tendon ultrastructure, but also indicates
strain. In previous studies, the analysis of nuclei position that rotation occurs in undamaged fascicular samples. It
123
J Mater Sci (2007) 42:89578965 8965
indicated that tendon extension could be consistent with a 19. Sasaki N, Odajima S (1996) J. Biomech 29(5):655
helical or spiral geometry in the fascicular structure, 20. Puxkandl R, Zizak I, Paris O, Keckes J, Tesch W, Bernstorff S,
Purslow P, Fratzl P (2002) Philos Trans R Soc Lond B Biol Sci
whereby extension is facilitated by a straightening of the 357:191
coil in addition to fibre sliding. 21. Bruehlmann SB, Matyas JR, Duncan NA (2004) Spine 29:2612
22. Screen HR, Shelton JC, Chhaya VH, Kayser MV, Bader DL, Lee
Acknowledgement Many thanks to Dr. Martin Knight, for his DA (2005) Ann Biomed Eng 33(8):1090
expert advice and assistance with the confocal microscopy. 23. Bruehlmann SB, Kelly EJ, Duncan NA (2005) Trans Orthop Res
Soc 30:389
24. Goodwin JS, Kenworthy AK (2005) Methods 37:154
25. Koster M, Frahm T, Hauser H (2005) Curr Opin Biotech 16:28
References 26. Woo HM, Kim MS, Kweon OK, Kim DY, Nam TC, Kim JH
(2001) Br J Ophthalmol 85:345
1. Woo SLY (1982) Biorheology 19:385 27. Davison PF, Galbavy EJ (1985) Invest Ophthalmol Vis Sci
2. Harris B (1980) Symp Soc Exp Biol 34:37 26:1202
3. Hiltner A, Cassidy JJ, Baer E (1985) Ann Rev Mater Sci 15:455 28. Arnoczky SP, Lavagnino M, Whallon JH, Hoonjan A (2002) J
4. Benjamin M, Ralphs JR (1997) Histol Histopathol 12:1135 Orthop Res 20:29
5. Ker RF (2002) CBPA 133:987 29. Petran M, Boyde A, Hadravsky M (1990) In: Confocal micros-
6. Elliott DM, Robinson PS, Gimbel JA, Sarver JJ, Abboud JA, copy. Academic Press, London, vol 9, p 262
Iozzo RV, Soslowsky LJ (2003) Ann Biomed Eng 31:599 30. Hansen KA, Weiss JA, Barton JK (2002) J Biomech Eng 124:72
7. Wess TJ, Hammersley AP, Wess L, Miller A (1998) J Struct Biol 31. Lanir Y, Salant EL, Foux A (1988) Biorheology 25:591
122:92 32. Hannafin JA, Arnoczky SP (1994) J Orthop Res 12:350
8. Buehler MJ (2006) PNAS 103(33):12285 33. Knight MM, van de Breevaart Bravenboor J, Lee DA, van Osch
9. Avery NC, Bailey AJ (2005) Scan J Med Sci Sports 15:231 GJVM, Weinans H, Bader DL (2002) Biochim Biophys Acta
10. Wess TJ, Cairns DE (2005) J Synchrotron Rad 12:751 1570:1
11. Provenzano PP, Vanderby R Jr (2006) Matrix Biol 25:271 34. Wang YN, Galiotis C, Bader DL (2000) J Biomech 33:483
12. Derwin KA, Soslowsky LJ, Kimura JH, Plaas AH (2001) J 35. Yahia LH, Drouin G (1989) J Orthop Res 7:2243
Orthop Res 19:269 36. de Campos Vidal B (2003) Micron 34:423
13. Redaelli A, Vesentini S, Soncini M, Vena P, Mantero S, 37. Kannus P (2000) Scand J Med Sci Sports 10:312
Montevecchi FM (2003) J Biomech 36:1555 38. Ottani V, Martini D, Franchi M, Ruggeri A, Raspanti M (2002)
14. Screen HR, Lee DA, Bader DL, Shelton JC (2004) J Eng Med Micron 33:587
218:109 39. Wess TJ, Hammersley AP, Wess L, Miller A (1998) J Mol Biol
15. Scott JE, Orford R (1981) Biochem J 197:573 275:255
16. Scott JE (2003) J Physiol 55:2335 40. de Campos Vidal B (2006) Matrix Biol 25:132
17. Weber IT, Harrison RW, Iozzo RV (1996) J Biol Chem 41. Raspanti M, Manelli A, Franchi M, Ruggeri A (2005) Matrix Biol
271:31767 24:503
18. Vesentini S, Redaelli A, Montevecchi FM (2005) J Biomech
38:433
123
J Mater Sci (2007) 42:89668973
DOI 10.1007/s10853-007-1586-x
Received: 7 December 2006 / Accepted: 6 February 2007 / Published online: 19 July 2007
Springer Science+Business Media, LLC 2007
Abstract The investigation of the atomistic mechanisms from reviewing related work we outline future directions
of crystal nucleation constitutes a major challenge to both and discuss the perspectives of simulation studies for the
experiment and theory. Understanding the underlying investigation of biomineralization processes in general.
principles of composite materials formation represents an
even harder task. For the investigation of the mechanisms
of crystal nucleation a profound knowledge of the ion
solvent and the ionion interactions in solution is required. Introduction
Studying biocomposites like fluorapatitecollagen materi-
als, we must furthermore account for the biomolecules and The nucleation and growth of crystals is a phenomenon of
their effect on the growth process. Molecular simulation fundamental interest. This does not only apply to physics,
approaches directly offer atomistic resolution and hence chemistry and materials science, but also to a specific field
appear particularly suited for detailed mechanistic analy- of bio-sciences and medicine: biominerals reflect a fasci-
ses. However, the computational effort is typically nating blend of organic soft matter and inorganic nano-
immense and for a long time the investigation of crystal crystals. The resulting composites exhibit properties of
nucleation from atomistic simulations was considered as both hard and soft matter. Indeed, biominerals combine the
impossible. We therefore developed special simulation toughness of an inorganic material and the flexibility of
strategies, which allowed to considerably extent the limi- biological tissues. A profound understanding of these
tations of computational studies in this field. In combina- properties requires investigating the interplay of the
tion with advanced experimental investigations this organic molecules with the ions of the inorganic compo-
provided new insights into the nucleation of biomimetic nent. This task is far from trivial. Exploring the atomistic
apatitegelatin composites and the mechanisms of hierar- structure of a composite is complicated by non-periodic
chical growth at the micro- and mesoscopic scale. Along local constellations of the organic molecules. Crystallo-
this line, molecular simulation studies reflect a powerful graphic methods may hence only provide insights in the
tool to achieve a profound understanding of the complex inorganic, crystalline component of biominerals. Apart
growth processes of apatite/collagen composites. Apart from identifying the atomistic structure, understanding the
mechanisms leading to these structures reflects an even
D. Zahn (&) O. Hochrein A. Kawska harder challenge. It requires an atomistic in-situ investi-
R. Kniep gation of ion aggregation, mineralization of the biomole-
Max-Planck-Institut fur Chemische Physik fester Stoffe,
cules and self-organization of the composite material.
Nothnitzer Str. 40, 01187 Dresden, Germany
e-mail: zahn@cpfs.mpg.de Molecular simulation approaches may easily achieve the
microscopic resolution needed for such detailed mecha-
J. Brickmann nistic analyses. In particular molecular dynamics (MD)
Eduard-Zintl-Institut fur Anorganische und Physikalische
simulations appear to be a promising tool for studying
Chemie, Technische Universitat Darmstadt, Petersenstrae 20,
64287 Darmstadt, Germany biomineral formation. However, the computational effort is
e-mail: brickmann@molcad.de immense and special MD techniques are mandatory to
123
J Mater Sci (2007) 42:89668973 8967
make simulations reasonable. The present review cannot are based on quantum mechanical calculations, exclu-
account for all of such approaches. We instead focus on sively. Lacking reliable empirical force-fields, ab-initio
particularly promising methods, which have proven suc- approaches were chosen to investigate bulk crystal struc-
cessful in a series of apatite/gelatin aggregation studies. tures and the local arrangement of defects in apatites.
Therein, the computational investigations are intended to Many of the variety of apatite species may be written as
complement experimental work and cross-links are of vital an isomorphous Ca10(PO4)6X2 series with X = OH, F,
importance. Indeed, we feel that combined efforts from Cl, or Br. Therein the ions X are embedded in channels
both experiment and theory are needed to tackle the chal- formed by staggered triangles of calcium ions. While suf-
lenging complexity of biomineral formation [1, 2]. ficiently small ionic species X = F are located in the
Apatite/collagen composites belong to the most abun- center of the Ca-triangles, larger ions were found off-center
dant biominerals in both humans and animal life forms. The along the [001] direction [3]. In hydroxyapatite, the most
importance of this material as the predominant component relevant apatite for human hard materials, the OH ions
of bones and teeth motivated a large number of both may be located either above or below the Ca-triangles.
experimental and theoretical studies [1]. Despite these From both crystallographic structure refinements and
efforts, our understanding of biomineral formation is still ab-initio structure relaxation studies, the distance between
limited. This particularly applies to phenomena taking place the hydroxide oxygen atom and the center of the Ca-tri-
at the atomistic scale. Indeed, only little is known about the angle was found as about 0.2 A [4, 5]. In the stable con-
interplay of the apatite ions and the collagen proteins. The figurations, the positively charged hydrogen atom of the
latter issue reflects a key aspect of the composite as it hydroxide ion is pointing away from the nearest Ca-trian-
accounts for crucial characteristics of the biomineral such as gle. Transformation of the OH ordering therefore requires
composition, morphogenesis and mechanic properties. not only crossing the embedding Ca-triangle by dislocation
of the oxygen atom by 0.4 A , but also an orientation
inversion of the complex anion.
Atomistic simulation approaches Depending on the preparation conditions pure
hydroxyapatite exhibits either monoclinic (space group
Atomistic simulations approaches may be divided in two P21/b; a = 9.4214(8) A , b = 2a, c = 6.8814(7) A ,
categories depending on the use of quantum or classical b = 120.00(8), Z = 4) [6] (Fig. 1, top) or hexagonal
mechanics. The latter concept relies on empirical potential (space group P63/m; a = b = 9.432 A , c = 6.881 A
, Z = 2)
energy terms, which may account for hydrogen-bonding [5] symmetry (Fig. 1, bottom). The latter modification
and non-bonding interactions with reasonable accuracy. corresponds to the high temperature phase, which can be
However, the appropriate theoretical treatment of bond obtained from the monoclinic modification via an
breaking and formation usually requires quantum order disorder transition affecting the orientation of
approaches. In contrast to classical force fields, quantum the hydroxide ions, only. By means of optical birefringence
chemical calculations are considerably more demanding experiments [7], X-ray and difference scanning calorimetry
with respect to computational resources. The desire to measurements this transformation was observed at tem-
combine the advantages of both classes of methods has peratures above about 200 C [8]. In principle, two dif-
motivated the formulation of mixed quantum/classical ferent models may account for the (average) hydroxide ion
schemes. These approaches are based on reducing the orientation (occupation of 0.5 for each crystallographic
computationally very demanding quantum calculations on position) in the high temperature phase: the mirror plane
the most relevant degrees of freedom, while describing the perpendicular to the c-axis of the space group P63/m may
remaining modes at much lower costs by classical be associated to (i) a disordered column model in which
mechanics. Implemented in an appropriate way, quantum/ the orientation of the hydroxide ions is inverted at random
classical schemes may preserve much of the accuracy sites within the channels or (ii) an ordered column
obtained from quantum chemistry. On the other hand, the model in which all OH ions in a single row point in the
benefits from transferring the efficiency of classical force same direction, while the orientation of each specific col-
fields to large parts of the simulation model may easily umn with respect to the collective of all rows is randomly
save orders of magnitudes of the computational costs. distributed [5].
Both models are indistinguishable by X-ray or neutron
Apatite simulation studies scattering experiments [9] and so far the only evidence
which picture accounts for the hexagonal modification of
The importance of apatite has motivated a manifold of hydroxyapatite was obtained from theory [4, 10, 11]. De
atomistic simulation studies. Early theoretical works on Leeuw performed a series of quantum mechanical structure
bulk apatite crystals date back up to about 10 years ago and optimization studies based on various predefined OH
123
8968 J Mater Sci (2007) 42:89668973
arrangements and energy minimization to local minima [4]. less than 10 kJ mol1 at 0, 300, and 573 K, respectively
This approach is typically used for low-temperature [4, 10] (S. Hauptmann, Private communication).
structure determinations from ab-initio calculations. Apatites grown in biologic systems typically contain a
Indeed, the energy minimization corresponds to zero manifold of defects (carbonates, water, etc.) [9]. Atomistic
Kelvin. Simulation setups of non-zero temperature require simulations offer very detailed insights into the arrange-
Monte-Carlo or molecular dynamics approaches which are ment of defects and local structure changes arising from
computationally much more demanding. This motivated impurity incorporation. Most of the defect studies related to
the development of empirical potential energy functions for apatites were dedicated to hydroxyapatite. A central
modeling apatites at favorable costs [12, 13]. On the motivation for this is given by the fact that the inorganic
basis of such force-fields we performed molecular component of bones is represented by non-stoichiometric
dynamics simulations of the temperature-induced mono- (Ca-deficient) hydroxyapatite [9] whose impurities are
clinic hexagonal transition [10]. The results of the believed to be relevant for biological functions [14].
latter study and the structure optimization calculations of The general perspectives of atomistic simulations to
de Leeuw are quite controversial. While de Leeuw non-stoichiometric apatite studies may be very nicely
predicted concerted inversions of ordered columns [4], illustrated by the work of Astala and Stott [15]. Focusing
independent flipping events of single hydroxide ions were on carbonate incorporation in hydroxyapatite, these authors
observed from the molecular dynamics simulations [10]. investigated both hydroxide and phosphate substitution. To
Quantum chemical approaches allow very precise avoid net charges, replacing OH or PO43 by CO32 ions
treatment of the atomic interactions. At first sight one requires further changes in the system and Schottky defects
might therefore tend to favor the predictions made on the and protons (associated to hydroxide or phosphate ions)
basis of ab-initio calculations. However, in quantum cal- accompany carbonate incorporation. Astala and Stott pre-
culations limited computational resources typically imply pared a series of model systems corresponding to a variety
considerable model simplifications whose effect may out- of possible defect arrangements [15]. Each of the simula-
balance the benefits compared to empirical force-fields. tion models was then optimized by relaxation to a local
Indeed, in the present case modeling a temperature-induced minimum in the potential energy landscape. From this an
phase transition at zero Kelvin reflects such an oversim- energetic scoring was obtained which may be related to the
plification. Exploring inversion events of single hydroxide occurrence of specific defect arrangements.
ions on the basis of free energy calculations at room tem- Apart from carbonates, the most abundant impurities in
perature, Hauptmann observed a barrier of 87 kJ mol1 biologic hydroxyapatite are H+ (associated to phosphate or
(S. Hauptmann, Private communication). At higher tem- hydroxide ions), fluoride and chloride ions [16]. The
perature, the lattice expands and the edges of the Ca-triangles incorporation of halide ions is related to hydroxide sub-
are more susceptible to elongations. This reduces the stitution and directly preserves the total charge of the
hydroxide-flipping barrier to 50 kJ mol1 at 200 C, i.e., crystal. From a dentists point of view, the most important
the temperature at which the monoclinic hexagonal of these defect species is represented by fluoride ions,
transition takes place [10]. Moreover, the changes in which are known to protect human tooth enamel from
potential energy related to a single OH inversion event in caries. OH F ion replacement in bulk hydroxyapatite
monoclinic hydroxyapatite reduces from 42, 23 kJ mol1 to was investigated in a series of simulation studies employing
123
J Mater Sci (2007) 42:89668973 8969
similar strategies as described above [4, 11]. On this basis, relaxation times. Decorrelation is typically achieved by
fluoride defects were found to promote the mono- setting the barostat relaxation time about ten times larger
clinic hexagonal transition. Indeed, the preferred than the thermostat relaxation time.
fluoride incorporation corresponds to OHFHOOH To check the effect of the isothermalisobaric algorithm
constellations. This fully compensates the energetic costs on the apatite/water interface reported in [17] we rerun 5 ns
for inverting the orientation of a hydroxide ion within the sketches of our model system using 0.5 ps/0.5 ps, 0.5 ps/
aligned OHOHOHOH columns of monoclinic 5 ps, and 0.5 ps/125 ps as thermostat and barostat relaxa-
hydroxyapatite [11]. tion times, respectively. The larger the thermostat/barostat
decoupling, the more reasonably small cell fluctuations
were observed. This particularly applies to the cell edge
Ion/water and apatite/water simulation studies parallel to the c-axis of the apatite slab, which varies from
0.5 to +10% for the 0.5 ps/0.5 ps setup. Applying 0.5 ps/
Interface studies 5 ps already reduces the corresponding fluctuations to less
than 1% (0.1% for the 0.5 ps/125 ps setup). Increased
In biologic systems, apatites are formed from aqueous flexibility, orientation inversion, and release of interfacial
solutions and remain in contact with water during their use hydroxide ions are strongly promoted by simulation cell
as bone or dental materials. The latter issue, i.e., the deformation. In fact, we expect the anomalously large
structure of apatite/water interfaces may be investigated compression/expansion fluctuations of the apatite/water
from relatively straightforward molecular simulation interface study of de Leeuw to account for the observed
approaches. In 2003, we presented a molecular dynamics hydroxide dissociation.
simulation study of the hydration of (001) surfaces of
hydroxyapatite [17]. The investigated surface types were Aggregation studies
obtained from different (001) cuttings from the ideal
crystal and should be considered as limiting cases locally For the investigation of apatite nucleation a profound
covering the features of slightly rough crystal faces. Unless knowledge of both the ionwater and the ionion interac-
acidic conditions are applied, the solubility of hydroxyap- tions in solution is required. While there are well-estab-
atite in aqueous solution is extremely low. Indeed, our lished empirical force-fields for studying calcium,
0.5 ns molecular dynamics runs reflect a stable apatite/ phosphate, hydroxide and fluoride ions in aqueous solution,
water interface for which a series of structural and proton transfer events cannot be assessed without quantum
dynamical properties could be sampled. Soon after this mechanics so far. Apatite formation is known to dramati-
work a similar molecular dynamics simulation was pre- cally depend on the pH [1]. A full account of the ionion/
sented by de Leeuw from which hydroxide ion dissociation ionsolvent interactions hence must include the consider-
was observed [18]. While no comparison of the two studies ation of proton transfer reactions. At ambient conditions, it
was given in [18], we feel the striking differences must be is reasonable to assume that these processes only occur in
discussed before continuing the use of either simulation solution and at the apatite/water interfaces. Proton transfer
model. reactions may hence be expected during ion aggregation
The two pictures may be easily discriminated from and crystal growth, but not inside the bulk material.
experiment by adding dry apatite and pure water. Even The crystallization of calciumphosphates from aqueous
pulverized apatite samples with large surface to volume solution is a process of considerable complexity and the
fractions do not alter the pH of the system. This is in clear degree of protonation of the phosphate ion can play an
contradiction to the observation of hydroxide ion release in important role. Depending on the pH, phosphoric acid
the simulations of de Leeuw [18] giving rise to the exhibits several stages of deprotonation (pKa1 = 2.16,
assumption that the underlying empirical interaction model pKa2 = 7.21, pKa3 = 12.32; at 25 C [19]). At physiologic
might be inappropriate. However, comparing the de Leeuw conditions the fraction of completely deprotonated phos-
force field and the Hauptmann force field [12] used in our phate ions may hence be expected to be extremely small.
simulations [17] we found both models to be very similar. Nevertheless, crystallization may yield calciumphosphates
Instead, it seems the different setups of the molecular (including hydroxyapatite) consisting of PO43 ions even
dynamics simulations caused the lack of concordance. De under acid conditions. In such cases, the depronation of the
Leeuw applied identical relaxation times of 0.5 ps to the (di)hydrogenphosphate ion has to occur during crystal
isothermalisobaric algorithm. This choice is rather unu- growth. We investigated the initial steps of calcium and
sual as in most constant pressureconstant temperature hydrogenphosphate ion aggregation in aqueous solution by
molecular dynamics simulations artificial couplings of the means of quantum/classical molecular mechanics simula-
respective variables are avoided by assigning different tions [20]. By combination of studying calcium ion
123
8970 J Mater Sci (2007) 42:89668973
aggregation to (hydrogen)phosphate ions and investigating reaction coordinate. Adding further ions, we already have
hydrogenphosphate deprotonation, we identified the to consider at least two degrees of freedom. Apart from the
[Ca2+PO43Ca2+] ion triple as the smallest stable distance coordinate also the orientation of the aggregate
aggregate, which may be expected to contain an entirely with respect to the newly attached ion must be considered.
deprotonated phosphate ion. This example demonstrates For solute/solvent combinations of very low solubility
the deprotonation of the hydrogenphosphate ions to be such as calciumphosphate/water or CaF2/water systems we
promoted by its local environment, i.e., by positively found very stable ion complexes which did not dissolve
charged calcium ions: even during high-temperature annealing runs [20, 21].
Aggregates larger than ion pairs or triples exhibit several
adsorption sites. While the energy score of one ion docking
process might be more favorable than another, a variety of
adsorption sites may lead to stable aggregates. In such
systems, crystal growth does not necessarily follow the
minimum energy path, but is also controlled by the local
availability of ions which is directly connected to ion dif-
fusion in the dilute solution.
As computational costs force us to avoid direct molec-
ular dynamics simulations of slow diffusion processes we
developed a simulation scheme which instead mimics ion
During the initial steps of calciumphosphate nucleation distribution in solution originating from diffusion [21].
in aqueous solution the acidicity of HPO42 increases This is accomplished by a Monte-Carlo step in which an
dramatically. By association of one Ca2+ ion, the anionic ion is placed at a random position in the vicinity of the
acid HPO42 is converted into a [Ca2+HPO42]0 complex, aggregate. The nearest adsorption site is then guessed from
which may be interpreted in terms of a neutral acid. While steepest descent energy optimizations. Finally, full relax-
more likely to be deprotonated, the acidicity of ation of the total simulation system provides a new
[Ca2+HPO42]0 is still rather weak. Proton transfer aggregate to which further ions may be attached. During
becomes likely only after aggregation of a second calcium this iterative procedure the ions are added one-by-one to
ion leading to a [Ca2+HPO42Ca2+]2+ complex which the aggregate being formed. The initial aggregate consists
may be considered as a cationic acid. of a single ion and each growth step represents the asso-
For a realistic molecular dynamics study of crystalliza- ciation of a further ion. Despite its appealing simplicity, we
tion processes from solution simulation models corre- wish to point out that this ion-by-ion attachment procedure
sponding to slightly super-saturated salt solutions are relies on a series of approximations which should be
required. In the case of low salt solubility, this implies carefully checked for each specific study of crystal
immense numbers of solvent molecules. The major obsta- formation from solution.
cle to straightforward molecular dynamics simulation of The example of CaF2 aggregation from aqueous solu-
crystal formation from strongly diluted solutions is related tions nicely demonstrates the perspectives of our aggregate
to the mainly diffusion controlled character of such sys- growth model. Starting from single ions, we can track
tems. In the extreme limit the ion concentration is so low crystal nucleation from the very beginning. The early stage
that most of the solutes does not interact with each other of CaF2 formation is of particular interest because the
and behave as if being in an infinite dilution. In this picture evolution of small complexes to aggregates counting
the ions diffuse freely, until some of them happen to get at 50100 ions reflects a disorder order transition. Indeed,
close distance to each other. Attractive ionion interactions our simulations offer insights into the self-organization
may then lead to the formation of aggregates. process leading to the CaF2 crystal structure. We identified
The formation of ion pairs and triples as described above the formation of regular motifs to nucleate in the inner core
was investigated by ignoring ion diffusion and focusing on of the aggregate (Fig. 2). Adding further ions to the aggre-
the association of nearby ions. The related free energy gate surface implies structural rearrangements which cause
landscape is scanned as a function of a few degrees of the growth of the ordered domain and eventually leads to
freedom and the minimum energy path is used as a model crystallite formation.
reaction coordinate. In complex systems like ion solutions The limitations of the aggregate growth model may be
the real reaction coordinate depends on all degrees of nicely demonstrated by the example of NaCl aggregation
freedom, including the solvent molecules. However, for from aqueous solution, i.e., a solute/solvent combination of
studying the association of an ion pair, in many cases the high solubility. Transition path sampling molecular
ionion distance reflects a very suitable model of the dynamics studies of this process revealed a competition of
123
J Mater Sci (2007) 42:89668973 8971
Apatite/collagen/water systems
123
8972 J Mater Sci (2007) 42:89668973
123
J Mater Sci (2007) 42:89668973 8973
far from a generally applicable mesoscopic model of apa- 13. de Leeuw NH (2004) Phys Chem Chem Phys 6:1860
tite/collagen materials and complexity reduction by coarse 14. Posner PA (1969) Physiol Rev 49:760
15. Astala R, Stott MJ (2005) Chem Mater 17:4125
graining of biominerals remains a major task for the future. 16. Dorozhkin SV, Epple M (2002) Angew Chem 114:3260
17. Zahn D, Hochrein O (2003) Phys Chem Chem Phys 5:4004
Acknowledgment Financial support was provided by the Deutsche 18. de Leeuw NH (2004) J Phys Chem B 108:1809
Forschungsgemeinschaft. 19. Lide DR (19971998) CRC handbook of chemistry and physics,
78th edn. CRC Press, Boca Raton, New York
20. Zahn D (2004) Z Anorg Allgem Chem 630:1507
References 21. Kawska A, Brickmann J, Kniep R, Hochrein O, Zahn D (2006)
J Chem Phys 124:24513
22. Zahn D (2004) Phys Rev Lett 92:40801
1. Kniep R, Simon P (2007) Top Curr Chem 270:73 23. Bhowmik R, Katti SK, Katti D (2007) Polymer 48:664
2. Harding JH, Duffy DM (2005) J Mater Chem 16:1105 24. Buehler M (2006) J Mater Res 21:1947
3. Rulis P, Ouyang L, Ching WY (2004) Phys Rev B 70:155104 25. Buehler M (2006) Proc Nat acad Sci 103:12285
4. de Leeuw NH (2002) Phys Chem Chem Phys 4:3865 26. Schepers T, Brickmann J, Hochrein O, Zahn D (2007) Z Anorg
5. Kay MI, Young RA, Posner RA (1964) Nature 204:1050 Allgem Chem 633:411
6. Elliot JC, Mackie PE, Young RA (1973) Science 180:1055 27. Tlatlik H, Simon P, Kawska A, Zahn D, Kniep R (2006) Angew
7. van Rees HB, Menngeot M, Kostiner E (1973) Mater Res Bull Chem 118:1939
8:1307 28. Thompson JB, Kindt JH, Drake B, Hansma HG, Morse DE,
8. Suda H, Yashima M. Kakihana M, Yoshimura M (1995) J Phys Hansma PK (2001) Nature 414:773
Chem 99:6752 29. Currey J (2001) Nature 414:699
9. Elliot JC (1994) Structure and chemistry of the apatites and other 30. Fantner GE, Hassenkam T, Kindt JH, Weaver JC, Birkedal H,
calcium orthophosphates in studies in Inorganic Chemistry 18. Pechenik L, Cutroni JA, Cidade GAG, Stucky GD, Morse DE,
Elsevier, Amsterdam Hansma PK (2005) Nat Mater 4:612
10. Hochrein O, Kniep R, Zahn D (2005) Chem Mater 7:1978 31. Simon P, Zahn D, Lichte H, Kniep R (2006) Angew Chem Int Ed
11. Zahn D, Hochrein O (2006) Z Anorg Allgem Chem 632:79 45:1911
12. Hauptmann S, Duffner H, Brickmann J, Kast SM, Berry RS
(2003) Phys Chem Chem Phys 5:635
123
J Mater Sci (2007) 42:89748985
DOI 10.1007/s10853-007-1642-6
Received: 5 December 2006 / Accepted: 26 February 2007 / Published online: 17 July 2007
Springer Science+Business Media, LLC 2007
123
J Mater Sci (2007) 42:89748985 8975
b-turn structures, and a series of these motifs would form and a 3BspE I restriction sites allowing the use of a previ-
spring-like spirals, or nanosprings, conferring elasticity to ously described multimerization or doubling strategy based
the fiber [1113]. Structural studies of the recombinant on compatible but non-regenerable restriction sites [20]. The
polypeptide 1 showed that its sequence, [(GPGGS two synthetic genes constructed by successive doubling of
GPGGY)2 GPGGK]11, is composed of type II b-turns [13]. each different basic repeat unit were called (A1S8)20 and
The central (PG) of the (GPGGX) consensus motif, also (Y1S8)20. These sequences were cloned in a pBluescriptII
present in several elastomeric proteins such as elastin SK+ plasmid and used to transform electrocompetent XL1-
[1416], and glutenin [17], has been found to be favorable Blue cells (Stratagene). The recombinant plasmids contain-
for the formation of type II b-turns [18, 19]. Thus, the ing the final (A1S8)20 and (Y1S8)20 silk-like sequences
mechanism of silk elasticity is thought to resemble the one (2,300 bp each) were partially sequenced on both strands
described for these b-turn forming polypeptides and seems using vector specific primers.
to be entropically driven [3, 13]. Therefore, the extremely
high number of (GPGGX)n motifs in the Flag protein is Cloning of the silk-like sequences in the pET19b
believed to be responsible for the extraordinary elasticity (KanR) expression vector
measured for the flagelliform silk (200%) [1, 2]. This
argument is reinforced by the fact that a lower amount of The pET19b (KanR) expression vector was previously
this motif in the dragline protein MaSp2, and its absence in obtained by replacing the Bpu1102/AlwNI fragment con-
the minor ampullate silk proteins Misp1 and Misp2, seem to taining the ampicillin resistance gene in pET19b (AmpR)
result in much lowered elasticity of these two silk fibers [6]. (Novagen) by the Bpu1102/AlwNI fragment containing the
We have engineered and produced recombinant repeti- kanamycin resistance gene (KanR) from pET26b ()
tive silk-like proteins with basic repeats containing variants (Novagen) (Dr M. B. Hinman, unpublished data).
of the elastic (GPGGX) motifs found in the Flag protein, The plasmid DNAs of the bacterial pBluescript clones
either (GPGGA)4 (=A1 motif) or [(GPGGY) (GPGGS)]2 containing the confirmed (A1S8)20 and (Y1S8)20 silk-like
(=Y1 motif) combined with a strength motif [Linker-(A)8] sequences, as well as the pET19b (KanR) expression vec-
(=S8 motif) found in the dragline silk MaSp2 protein. We tor, were isolated using an alkaline lysis protocol [21].
have produced and purified the two proteins (A1S820 and The expression vector and each recombinant silk plas-
Y1S820), confirmed their primary structures by amino acid mid were subjected to a Bam HI restriction enzyme
analyses, characterized their secondary structures using digestion (New England Biolabs Inc.). The totality of the
circular dichroism, and generated two types of fibers from restriction enzyme digestion sample was subjected to
both proteins either from aqueous solutions or organic electrophoresis in a 0.8% agarose gel. The silk inserts and
solvents. The synthetic fibers were mechanically tested to the pET19b (KanR) vector were purified by electroelution
give us an evaluation of the performance resulting from the following standard protocols [22]. The DNA fragments
combination of either of the Flag protein elastic motifs, and were then recovered by addition of ammonium acetate to a
the dragline silk protein strength motif. final 1.5 M concentration, followed by the addition of two
volumes of cold ethanol. The samples were placed at
80 C for 45 min and the purified DNA fragments were
Materials and methods recovered by centrifugation at 18,000g for 25 min at room
temperature. The supernatant was discarded and each DNA
Gene construction and cloning in pBluescriptII SK+ pellet was dried and resuspended in 30 lL of TE buffer
in E. coli (10 mM TrisHCl/1 mM EDTA, pH 8).
Each purified Bam HI silk fragment was ligated to the
Three sets of complementary synthetic oligonucleotides purified Bam HI pET19b (KanR) expression vector using
(roughly 60 bp each, Electronic Supplementary data or T4 DNA ligase following the recommendations of the
ESM, Fig. 1) were designed (Dr M. B. Hinman): (1) flagel- Manufacturer (Promega). Each ligation reaction was used
liform putative elastic encoding motifs A1 and Y1 (coding to transform electrocompetent E. coli XL1-Blue cells
respectively for (GPGGA)4 and (GPGGSGPGGY)2), (2) the (Stratagene). The plasmid DNA from the recombinant
dragline silk putative strength encoding motif S8 (coding for clones were extracted by alkaline lysis and characterized
a linker-polyalanine: (GGPSGPGS(A)8)). These three types by restriction enzyme digestion with Bam HI (New Eng-
of synthetic oligonucleotides (Y1, A1 and S8) were synthe- land Biolabs Inc.). These plasmids were also subjected to a
sized and assembled as double stranded cassettes each restriction enzyme digestion with Xma I (New England
cloned in pBluescriptII SK+ (Stratagene) at the Hind III/ Biolabs Inc.) to verify the inserts orientation. The two
Sma I sites (Midland Certified Reagents Inc, Texas). These kinds of recombinant pET clones displaying the (A1S8)20
double stranded sequences were engineered with a 5Xma I, or (Y1S8)20) silk insert in the right orientation were
123
8976 J Mater Sci (2007) 42:89748985
E. coli cell lysis The heat-treated protein extracts and the purified protein
fractions were analyzed by SDS-PAGE. For all SDS-PAGE
The cell pellets were resuspended in 1 lysis buffer analyses, 4% stacking and 10% separating polyacrylamide
(50 mM TrisHCl pH 8, 10 mM MgCl2, 10 mM NaCl) at gels containing 0.1% SDS were made in TrisHCl buffers
3 mL of buffer per every gram of cell. Lysozyme (Sigma) (Mini PROTEAN3 Cell protocol for SDS-PAGE buffer
was added to each sample to a final concentration of system, Biorad). The composition of the sample buffer and
0.2 mg/mL and the samples were incubated on ice for the 5 electrode buffer used are the ones recommended by
30 min swirling periodically. At this stage, 1 mM PMSF the manufacturer (Biorad). Typically, 10 lL of protein
(phenylmethylsulphonylfluoride; Sigma) was added to the sample were subjected to SDS-PAGE analysis. A Kalei-
lysates to prevent protein degradation. Then 1.5 g of doscopeTM Prestained Standards (Biorad, 10 lL) or a
deoxycholic acid (MP Biomedicals LLC) were added per Precision Plus Protein Dual Color Standard (Biorad, 8 lL)
gram of cells and the lysates were incubated 20 min at were used as a molecular weight marker. All electrophor-
37 C. At this point, 0.02 mg of DNAse I (Sigma) was eses were performed using the Mini PROTEAN3 Cell
added to the cell lysates per gram of cell and the samples (Biorad) at a constant voltage of 80 V.
were incubated at room temperature for 30 min on a
platform shaker. The lysates were then subjected to cen- Staining of polyacrylamide gels
trifugation at 3,300g for 15 min to pellet the cellular
debris. After a heat-treatment at 80 C for 10 min, the After SDS-PAGE analysis, the gels were stained with
recovered lysates, or supernatants, were again subjected to Coomassie Brillant Blue (R-250) dye according to the
centrifugation at 3,300g for 15 min to pellet the denatured method published [23]. The stained gels were placed in a
proteins. The cleared heat-treated lysates were stored at 10% glycerol solution for 1 h before being dried between
80 C until use for protein purification. two sheets of Ultra Clear Cellophane (Research Products
123
J Mater Sci (2007) 42:89748985 8977
International Corp., Mount Prospect, Illinois) using a pure soluble Y1S820 dilutions used for CD were 0.75 mg/mL
plexiglass frame. in TrisHCl pH 8 and 0.54 mg/mL in 0.1 PBS. The CD
spectra were recorded on a JASCO-810 spectropolarimeter
Western blot analyses using the Spectra Manager for Windows 95/NT software
(Version 1.18.00). A total of 8 spectra accumulations were
The proteins samples separated by SDS-PAGE were recorded from 185 nm to 260 nm with a resolution of
transferred to a PVDF/ImmobilonTM P transfer membrane 0.1 nm and a path length of 0.01 cm. Melt and anneal
(Millipore) by electroblotting using the Mini Trans-Blot experiments were conducted successively from 0 C to
Electrophoretic Transfer Cell (Biorad). Blots were set up as 85 C and 85 C to 0 C at 5 C intervals. The data
specified by the manufacturer (Biorad). All transfers were obtained were submitted to the following online site for
performed overnight at room temperature under a constant analyses: www.cryst.bbk.ac.uk/cdweb/html/home.htwl [24].
current of 25 mA. After fixation of the proteins, the We used the SELCON 3, CONTIN and CDSSTR methods
membranes were subjected to Western blot analyses using for protein secondary structure predictions (see [24] for full
the HisTag antibody (Novagen) directed against the references).
(histidine)10 tag ((His)10) present in the amino terminus of
the fusion proteins. The secondary antibody used was the Production of synthetic fibers
Goat Anti-Mouse IgG HRP conjugate (H + L) (Novagen).
All Western blot analyses were performed according to the Two methods of fiber production were used for both types
protocols described by the manufacturers. Chemilumines- of synthetic proteins (Fig. 1). The first one relied on the
cent detections were performed using the ECLTM Western natural ability of these purified synthetic fibers to sponta-
Blotting Detection Reagents (Amersham Biosciences). The neously form fibers in aqueous solutions, and the second
membranes were then wrapped in plastic wrap and placed relied on the wet spinning/extrusion of a silk dope made in
in a light proof cassette against a high performance an organic solvent.
chemiluminescence film (HyperfilmTM ECL, Amersham
Biosciences) to visualize the proteins of interest. Hand pulled fibers
Amino acid analyses The Y1S820 pure protein spontaneously formed an oily
looking film at the surface of the elution or strip solutions
The purified protein samples were dialyzed against 5 mM collected in a glass dish, thus naturally separating from the
ammonium bicarbonate and lyophilized. The samples were solution. We used forceps to pull the edge of the Y1S820
treated using the Waters AccQTagTMMethod for hydro- film that lifted into single fibers (Fig. 1a). While A1S820
lysate amino acid analysis (Millipore) and then subjected to did not spontaneously form a film at the surface of the
HPLC using a Hitachi LC 6500. The amino acid compo- elution or strip fractions collected, we noticed a film-like
sitions were expressed in mole % (ESM, Table 1). structure at the bottom of the dish. Thus shaking vigorously
123
8978 J Mater Sci (2007) 42:89748985
the dish to mix the contents resulted in the surfacing of Results and discussion
preformed A1S820 films that could then be pulled into
fibers using forceps (Fig. 1a). Gene expression and protein production
123
J Mater Sci (2007) 42:89748985 8979
(a)
MG(H)10SSGHIDDDDKHMLEDPP-[A1S820 or Y1S820 silk repeat]20-EISGSGC
Silk repeats:
A1S820: [(GGAGPGGAGPGGAGPGGAGP)1 (GGPSGPGSAAAAAAAAGP)]20 57.64 kDa
(b) (c)
1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9
250
150
100
76
50
37
25
Fig. 2 Primary structures and Western Blot analyses of the Y1S820 secondary antibody. In (b) and (c): Lane 1 is a Molecular weight
and A1S820 recombinant proteins. (a) Sequences and molecular marker; Lanes 25 show the A1S820 purification; Lanes 69 show the
weights of the spider silk-like proteins produced in E. coli. The Y1S820 purification; Lanes 2 and 6 show the unbound protein
molecular weights of the fusion proteins are indicated next to each fractions; Lanes 3 and 7 show the 40 mM imidazole washes; Lanes 4
silk repeat. (b) SDS-PAGE analysis of the purification of A1S820 and and 8 show the 50 mM imidazole washes; Lanes 5 and 9 show the
Y1S820 from heat-treated extracts (stained with Coomassie Brillant eluted/strip fractions. In both (b) and (c), the blue and green arrows
Blue). (c) Western blot analysis of (b) using the HisTag antibody as point at the purified A1S820 and Y1S820 proteins respectively
primary antibody and the Goat Anti-Mouse IgG HRP conjugate as
Protein purification and characterization bottom of the forming fiber thus allowing the film at the
periphery of the droplet to fuse into a single fiber (Fig. 3,
The A1S820 and Y1S820 proteins were easily purified with d and e). Moreover, we determined that the aggregation of
classic metal affinity chromatography and were recovered these proteins was concentration dependent as dilution of
using as low as 60 mM imidazole in the elution buffer. The the original heat-treated protein extract loaded on the col-
SDS-PAGE and Western Blot analyses of the purification umn resulted in no such fiber formation. We believe that
of Y1S820 and A1S820 are shown in Fig. 2 (b and c). The fiber formation is also purity dependant as these fractions
two pure full-size proteins were visible when stained with were extremely pure. It is important to stress the fact that
Coomassie blue (Fig. 2b). Note that on the Western blot these fibers were able to form in an aqueous solution close
analyses, we can see both full-size proteins, as well as to physiological pH rather than in harsh solvents. More
truncation products only for Y1S820 (Fig. 2c). Truncated importantly, these aqueous solutions might help preserve
proteins are likely the direct result of translation from a the original secondary structures of the recombinant silk
prematurely terminated transcript and are commonly proteins that might be critical for the resulting mechanical
observed for native [3, 10] or recombinant spider silk properties of the fiber.
proteins [20, 2730]. After expression, heat-treatment and After further observation, we noticed that the eluted
affinity chromatography purification, we were able to Y1S820 protein could self-assemble to form a very uniform
recover of 710 mg/L of pure A1S820 or Y1S820 proteins. film with an oily appearance that floated at the surface of
The results of the amino acid analyses of the purified the purified protein fraction. A major difference in behavior
A1S820 and Y1S820 proteins confirmed their identities and between the two pure proteins was that A1S820 did not
their high purity (ESM, Table 1). form an apparent film at the surface of the pure protein
While purifying Y1S820 and A1S820, we observed the fraction like the one observed for Y1S820 but would rather
formation of fibers at the bottom of the affinity chroma- form a viscous layer at the bottom of the dish covered by a
tography columns as the elution or strip fractions were more fluid layer. However, by vigorously shaking the dish
being collected (Fig. 3a and b). Samples of these in situ containing the pure A1S820 fraction, we were able bring
formed fibers were collected on a slide and observed under the viscous layer to the surface as a somewhat broken up
a light microscope (Fig. 3, c through e). It looked as though film. We used forceps to pull each type of films made by
the proteins were forming a film at the periphery of an these two proteins into fibers (Fig. 1a) and mounted the
extremely viscous droplet. These fibers were drawn out of pulled fibers on testing cards for mechanical testing. The
the column by the weight of the droplet that dangled at the film formed by Y1S820 could be reeled in and be wound
123
8980 J Mater Sci (2007) 42:89748985
around a tube very easily without breaking. The fibers from 65 C to 85 C show an average of 6.33% sheets,
pulled from the A1S820 small films were shorter, however, 9.10% turns, 8.30% helices, and 77.68% random coils
limited by the size of the broken film pieces. while the ones obtained for the same temperature bracket
with CDSSTR show 21.20% sheets, 17.80% turns, 8.60%
Structural analyses and hypotheses correlating helices, and 53.00% random coils. From the CD spectra of
secondary structure and self assembly the A1S820 melt presented in Fig. 4, we can confirm that
there is indeed an increase in sheets and turns (see the
Circular dichroism was used to determine the secondary increase at 200 nm and the decrease at 220 nm for the
structures of the A1S820 and Y1S820 silk-like proteins in sheets, and the increase at 210 nm for the turns). However,
aqueous solutions. The CD spectra of both melting (from there is still a high content of random coils (lower maxi-
0 C to 85 C) and successive annealing (from 85 C to mum before 200 nm) thus we feel that the results given by
0 C) for both proteins in the 5 mM TrisHCl pH 8 are SELCON 3 and CONTIN seem most probable for the
shown in Fig. 4. We chose to use a TrisHCl buffer pH 8 A1S820 protein. The same trends were observed for this
for CD analyses since we observed protein self-assembly in protein in PBS (ESM, Fig. 2). The successive annealing of
the elution and strip buffers containing TrisHCl pH 8. the A1S820 protein from 85 C to 0 C clearly shows that
Even though these two proteins only differ by two amino the heat induced changes in secondary structure observed
acids (in the 5th and 10th positions in the [GPGGX]4 during the melting are almost totally reversible (Fig. 4, top
repeat), there were substantial differences in the secondary right).
structures observed for both proteins. For Y1S820, the CD analyses show that at 0 C in Tris
For A1S820, the CD analyses show that at 0 C, the HCl, the protein is mostly unordered (about 55% in both
protein is mostly unordered (about 76.30% and 64.26% of buffers) but it contains less random coil than was observed
random coils observed respectively in TrisHCl pH 8 and for A1S820 (Fig. 4, bottom left). According to the melt
PBS). Random coils are noticeable in the melt profile of spectra and the results of the SELCON 3, CONTIN, and
A1S820 in TrisHCl (Fig. 4, top left) by the presence of CDSSTR analyses however, there are substantial changes
a lower maximum near 200 nm. As the temperature in the secondary structures when the temperature increases
increases, the SELCON 3, CONTIN and CDSSTR analyses to 85 C. For Y1S820, in the melt profiles shown in Fig. 4,
of the CD spectra for the A1S820 protein in TrisHCl show the maximum peak observed at 200 nm accompanied by a
a slight increase in helices (to a maximum of 10%). minimum at 220 nm are indicative of a noticeable increase
Although there seems to be a slight increase in sheets and in sheets (up to 31.32%), and the maximum peak observed
turns, the percentages of these two species at higher tem- at 210 nm (broadened by the sheet peak occurring at
peratures indicated by the SELCON 3 and CONTIN 200 nm) is indicative of an increase in turns (about 20%)
methods are substantially lower than indicated by the that is visible as soon as the temperature reaches 15 C.
CDSSTR method. Indeed, for both SELCON3 and CON- The same trend was observed for this protein in PBS (ESM,
TIN, the results of the analyses for A1S820 in TrisHCl Fig. 2). The successive anneal experiment for Y1S820
123
J Mater Sci (2007) 42:89748985 8981
10
A1S820- Melt 0 A1S8 20- Anneal
5
0 -5
-5 -10
CD (m deg)
CD (m deg)
-10
-15
-15
-20 -20
-25 -25
-30
-30
-35
-40 -35
190 200 210 220 230 240 250 260 190 200 210 220 230 240 250 260
Wavelength (nm) Wavelength (nm)
CD (m deg)
-5
0
-10
-5
-15
-10
-20
-15
190 200 210 220 230 240 250 260 190 200 210 220 230 240 250 260
Wavelength (nm) Wavelength (nm)
Fig. 4 Circular dichroism spectra for A1S820 and Y1S820. Melt the scale from one graph to another is not the same for the CD [mdeg]
(from 0 C to 85 C) and successive anneal (from 85 C to 0 C) units (vertical axes)
spectra obtained for both proteins in 5mM TrisHCl pH 8. Note that
(Fig. 4, bottom left) shows that once the secondary struc- subjected to a heat-treatment prior to purification reinforces
tures have formed during the melt, there is no reversion to this argument (Fig. 3). Moreover, changing the hydrophilic
the initial structure of the protein. 5th (Y) and 10th (S) residues and replacing them by
According to this data, for Y1S820, higher temperatures hydrophobic residues such as alanine (A) in A1S820
promote irreversible sheet and turn formation through deprives the protein of the ability to retain large amounts of
hydrogen bonding that remain stable. Such irreversible stable turn and sheet structures even when tested at a higher
temperature-induced b-sheet formation has also been concentration than Y1S820. This may be the reason shear
observed in melt experiments of major and minor ampul- forces are needed to force film formation from a liquid
late silk gland contents [31]. These structures for A1S820, crystalline phase from the pure A1S820 fractions.
though less pronounced, are reversible. Moreover, the fact Regarding the native Flag protein in the flagelliform
that the amount of turns increase and remain stable for gland, by having both [GPGGY GPGGS] alternating with
Y1S820 may indicate that the [(GPGGY)(GPGGS)]2 motif [GPGGA GPGGV/A], the protein may have parts of its
is able to adopt the b-turns structures proposed earlier for elastic segments in a stable b-turn conformation, possibly
this motif [6] and may be stabilized by intra-molecular under salt and pH control as suggested for other silk glands
hydrogen bonds involving tyrosine (Y) and serine (S). We [32, 33], while the rest of the elastic components are
hypothesize that by achieving a better organization or unfolded, or more disordered, thus allowing the molecules
folding of this latter motif, the protein might be able to to remain in a liquid crystalline phase avoiding premature
better self organize, thus allowing the poly-alanine seg- aggregation in the gland. Although in the Flag protein there
ments to interact with one another and lock into b-sheets by is no poly-alanine segments [6, 9, 10], the GGX repeats are
a nucleation process. This nucleation process may initiate known to promote self-assembly in proteins such as lam-
the self-assembly of the Y1S820 molecules into a supra- prin [34] and might play the same role as the poly-alanine
molecular structure. The fact that we do get spontaneous segments, or poly-(glycine/alanine) in other silks, in initi-
self-assembly of the pure Y1S820 proteins that were ating the self-assembly of the Flag molecules. A film
123
8982 J Mater Sci (2007) 42:89748985
similar to the ones seen in the A1S820 and Y1S820 pure proline seem to promote self aggregation in an amyloido-
fractions may self-assemble in the spiders silk glands. genic fashion [35], thus reinforcing the argument stated
Then, shear forces resulting from a combination of the above about the role of GGX sequences in the self-
spiders pull on the fiber and the extremely small size of the assembly of the native Flag protein.
spinning duct the proteins travel through might be enough
to induce the proper folding of the [GPGGA GPGGV] Mechanical performances of the synthetic pulled and
motifs. By locking all its [GPGGX]n motifs into the proper spun fibers
secondary structures, the rest of the molecules would then
be able to self organize and assemble into a fiber. Moreover Using an extruder, we were able to spin fibers out of both
in our case, by pulling the structure together, and having the A1S820 and Y1S820 proteins that were resolubilized in
properly folded motifs, the (GPS)2 linker sequence directly 100% HFIP. However, we determined that adding 15%
preceding the poly-alanine segments might be able to form water to the dope prior to spinning dramatically improved
stable intermolecular hydrogen bonds that would stabilize the mechanical properties as well as the appearance of the
the supramolecular structure. In the native Flag protein, we fibers. This can be explained by the fact that although HFIP
can hypothesize that this stabilizing role of the supramo- might promote intramolecular hydrogen bonding [36, 37],
lecular structure may be fulfilled by the spacer region, water was necessary to force the hydrophobic poly-alanine
positioned next to the (GGX)12 sequence, containing both segments away from the water phase and into proper sheet
highly hydrophobic and highly hydrophilic amino acids. structures, thus initiating self-assembly. The stress/strain
Recent molecular dynamic simulation studies on the curves for A1S820 and Y1S820, both pulled (P) and spun
(GVPGV)7 elastin pentapeptide suggest that such a (S) are shown in Fig. 5 while fiber specifics (number of
sequence has a higher propensity for PPII (poly-proline II) fibers used and diameter) and mechanical performance data
structures since they cannot stabilize any turns through (maximum stress, maximum extension, Youngs modulus,
hydrogen bonding [35]. It also retains a high backbone and toughness) are shown in Table 1. Before tensile test,
hydration level that would constrain these peptides in dis- pictures of each type of fibers used in this study were taken,
ordered conformations that are unable to exclude water and the pictures of the fibers displaying the best extension,
when aggregated, thus providing elastomeric properties to best maximum stress, and average maximum stress are
the matrix formed [35]. Such a model might be applicable shown in Fig. 6. According to our data, the pulled fibers,
to the more hydrophobic elastic motifs found in flagelli- which had a much smaller diameter on average than the
form silk as well as the elastic motif of the A1S820 protein. spun ones, outperformed the spun fibers in average maxi-
Moreover in elastins, glycine-rich repeats deprived of mum extension, maximum stress and toughness (Table 1
123
J Mater Sci (2007) 42:89748985 8983
A1S820 P 15 12.20 4.99 1,706.8 791.87 28.64 8.41 18.99 12.88 3.41 2.61
A1S820 S 19 32.15 16.24 759.68 540.27 28.58 17.18 3.72 1.24 0.464 0.30
Y1S820 P 31 15.79 6.05 1,081.49 1,000 49.64 19.35 34.06 25.30 10.6 10.2
Y1S820 S 18 28.4 11.32 933.62 727.14 10.21 7.32 1.59 1.03 0.089 0.11
Average values measured for all types and kinds of synthetic fibers. The standard deviation of each value is indicated ( STD). P = pulled;
S = spun
and Fig. 5). The appearance of these pulled fibers was also [6], and it is interesting to note that the Y1S820 pulled
strikingly different than the spun ones (Fig. 6). Not only fibers exhibit the same elasticity as the native dragline silk
were their surfaces smoother, but they were also more even (34% vs. 35%), thus confirming the role of the MaSp 2
and had greater sheen than any of the spun fibers. This can protein in the overall elasticity of the dragline. The Y1S820
be explained by the fact that the action of pulling the pulled fibers, however, cannot withstand as much stress as
preassembled film into a fiber was probably close to post the native dragline silk (0.050 GPa vs. 4 GPa), or flagel-
spin drawing that would improve the overall organization liform silk (0.050 GPa vs. 0.5 GPa). Consequently, their
of the fiber, hence its performance. toughness (10.6 MJ/m3, Table 1) is lower than the ones
Within the pulled group, the Y1S820 fibers were the reported for both dragline silk (160 MJ/m3) and flagelli-
toughest (average toughness = 10.6 MJ/m3). They were form silk (150 MJ/m3) [3, 4]. We can speculate that this
also twice as extensible (average maximum exten- difference in stress threshold is the result of two major
sion = 34.06%), and could withstand almost twice as much factors. The first one is that the synthetic proteins are much
maximum stress (average maximum stress = 49.64 MPa) smaller than the native silk proteins forming the fiber thus
as the A1S820 fibers which as a result displayed a higher limiting the number of intermolecular chain interactions
stiffness (average Youngs modulus = 1.706 GPa). This is necessary to stabilize the overall fiber structure and
understandable if we consider the secondary structures allowing weaker spots by lack of molecular overlap.
present in these Y1S820 proteins in an aqueous buffer at pH Indeed, the sizes of both dragline silk proteins are between
8 (more b-sheets and more turns than for A1S820), and the 300 kDa and 350 kDa [38], and although the size of Flag
fact that the elastic motifs in Y1S820 can be stabilized by has not been determined, the fact that the Flag mRNA is
hydrogen bonds due to the presence of Y and S in the bigger (15 kb) [6] than both MaSp mRNAs (11 kb and
elastic motif. 12 kb) [39, 5] suggests that this protein is at least bigger
Of the two synthetic constructs studied here, the primary than 350 kDa. Additionally, it is possible that a difference
structure of the Y1S820 protein consensus repeat more in sizes of the crystals formed and their lack of orientation
closely resembles the native N. clavipes dragline MaSp2 affect the strength of the synthetic fiber. The second may
consensus repeat ([(GPGQQ GPGGY)2 GPSGPS (A)9]n) be attributed to the fact that the native dragline contains a
123
8984 J Mater Sci (2007) 42:89748985
second protein, MaSp1 [39], rich in GGX repeats and protein batches to make the dopes, the age of the dope,
containing both (GA) and (A)n crystalline forming motifs the diameters of the fibers for instance, could account for
that can impart additional strength to the fiber. the distinct mechanical behaviors observed within a given
The comparison between the structures and perfor- fiber group. This variability is important to notice as it
mances of both the Y1S820 pulled fibers and the dragline defines the upper and lower limits for the mechanical
silk might give some insight as to the necessity of two performance of any fiber. Refining these limits, that is
proteins in the native dragline silk (MaSp 1 and MaSp2). lowering the variability in the mechanical performance of
The amount of (GPGXX)4 motifs adjacent to crystalline any synthetic fibers, should be a goal when making such
forming motifs in the MaSp 2 consensus repeat, as well as synthetic fibers and may be achieved by mastering the
their spacing, seem to be optimal for retaining reasonable spinning conditions. Hence, we want to stress the impor-
elasticity (35%) while maintaining high strength. There- tance of repetitions when testing synthetic materials to
fore, a second protein is necessary for additional strength know the full spectra of their performance.
possibly because adding extra crystalline forming motifs to
the MaSp2 consensus repeat might decrease the overall
elasticity of fiber. Moreover, water is a known plasticizer Conclusion
of silks. It has been shown that native dragline silk
extruded in water is stronger thus tougher than those We demonstrated a temperature dependent b-sheet induc-
extruded in air [40]. In the case of flagelliform silks, the tion in aqueous solution irreversible for Y1S820 and
presence of a natural aqueous glue coating is critical in reversible for A1S820, possibly explaining the difference in
promoting extreme elasticity (200%) [1]. It is possible that spontaneous film/fiber formation observed for the two
the mechanical performances of these pulled fibers may proteins in aqueous solutions.
be modified, even improved, when wet although no such The mechanical data obtained from both types of syn-
mechanical data is available to support this statement. thetic fibers made from the recombinant silk protein ana-
Regarding the performances of the spun fibers (Fig. 5 logs (A1S820 and Y1S820) clearly demonstrates that the
and Table 1), we found, not surprisingly, that the A1S820 [GPGGX]n motif is indeed responsible for the elasticity
fibers on average outperformed the Y1S820 fibers for displayed by both native dragline and flagelliform silks.
maximum stress (28.58 MPa vs. 10.21 MPa), maximum Moreover, the nature of the [GPGGX]n motif, that is the
extension (3.72% vs. 1.59%), and toughness (0.464 MJ/m3 hydrophobicity or hydrophilicity of the 5th and 10th amino
vs. 0.089 MJ/m3). In HFIP, we think that the elastic motifs acid residues, has a large impact in the level of elasticity
of the Y1S820 protein may not be able to properly form probably due to stability issues (presence or absence of
stabilized b-turns. Indeed, in this solvent, having bulky internal hydrogen bonding in the motif) as well as in the
residues present in the 5th and 10th positions (Y and S) of self organization process of the proteins into films or fibers.
the elastic motif may impair its folding, while the A1S820
protein has the advantage of small residues (A) in these two Acknowledgements We would like to thank Dr Michael B Hinman
(Department of Molecular Biology, University of Wyoming) for
positions in the elastic motif. This may better accommo-
providing the original A1, Y1 and S8 pBluescript clones, and for
date the proper folding of the molecule providing better performing the amino acid analyses of the recombinant proteins. We
overall organization allowing the poly-alanine to be locked also want to thank Professor Michael S Ellison (School of Materials
together into sheets and promote self assembly. However, Science and Engineering, Clemson University) for his valuable advice
on the mechanical testing methods and analyses. This research was
the extension of the spun A1S820 fibers is poor compared to
supported by grants from NIH, NSF and AFoSR.
the pulled A1S820 fibers reinforcing the idea that additional
shear forces, such as post-spin draw, might be necessary for
this protein to achieve better mechanical properties. References
From our results, it appears that the Y1S820 protein may
be more suitable to generate fibers with reasonable 1. Gosline J, Denny M, DeMont M (1984) Nature 309:551
mechanical properties when using aqueous solutions while 2. Stauffer S, Coguill S, Lewis R (1994) J Arachnol 22:5
it is not a good candidate to generate fibers from organic 3. Gosline J, Guerette P, Ortlepp C, Savage K (1999) J Exp Biol
202:3295
solvents. 4. Denny M (1976) J Exp Biol 65:483
During these studies, we noticed a high variability in the 5. Hinman M, Lewis R (1992) J Biol Chem 267:19320
performance of the pulled fibers as well as the spun ones 6. Hayashi C, Shipley N, Lewis R (1999) Int J Biol Macromol
within each type of fiber (Fig. 5). Such variability, al- 24:271
7. Simmons A, Michal C, Jelinski L (1996) Science 271:84
though not as substantial, exists in native silk fibers [41, 8. Termonia Y (1994) Macromolecules 27:7378
42] as well as synthetic polymeric fibers. In our experi- 9. Hayashi C, Lewis R (1998) Mol Biol 275:773
ments, no discernible factors such as the use of different 10. Hayashi C, Lewis R (2000) Science 287:1477
123
J Mater Sci (2007) 42:89748985 8985
11. Becker N, Oroudjev E, Mutz S, Cleveland J, Hansma P, Hayashi 27. Prince J, McGrath K, DiGirolamo C, Kaplan D (1995) Bio-
C, Makarov D, Hansma H (2003) Nat Mater 2:278 chemistry 34:10879
12. Hayashi C, Lewis R (2001) BioEssays 23:750 28. Fahnestock S, Irvin S (1997) Appl Microbiol Biotechnol 47:23
13. Zhou Y, Wu S, Conticello V (2001) Biomacromolecules 2:111 29. Fahnestock S, Bedzyk L (1997) Appl Microbiol Biotechnol 47:33
14. Chang D, Ventachalam C, Prasad K, Urry D (1989) J Biomol 30. Arcidiacono S, Mello C, Kaplan D, Cheley S, Bayley H (1998)
Struct Dynamics 6(i5):851 Appl Microbiol Biotechnol 49:31
15. Urry D, Shaw R, Prasad K (1985) Biochem Biophys Res Com- 31. Dicko C, Knight D, Kenney J, Vollrath F (2004) Biomacromol-
mun 130(i1):50 ecules 5:2105
16. Urry D, Hugel T, Seitz M, Gaub H, Sheiba L, Dea J, Xu J, Parker 32. Vollrath F, Knight D (2001) Nature 410:541
T (2002) Phylos Trans R Soc London, Ser B 357(i1418):169 33. Knight D, Vollrath F (2001) Naturwissenschaften 88:179
17. Tatham A, Hayes L, Shewry P, Urry D (2001) Biochim Biophys 34. Robson P, Wright G, Sitarz E, Maiti A, Rawa M, Youson J,
Acta 1548:187 Keeley F (1993) J Biol Chem 268(2):1440
18. Hutchinson E, Thornton J (1994) Protein Sci 3:2207 35. Rauscher S, Baud S, Miao M, Keeley F, Pome`s R (2006)
19. Wilmot C, Thornton J (1988) J Mol Biol 5:221 Structure 14:1667
20. Lewis R, Hinman M, Kothakota S, Fournier M (1996) Protein 36. Creighton T (1993) Proteins: structures and molecular properties,
Expr Purif 7:400 Second Edn. Freeman, New York
21. Birnboim H, Doly J (1979) Nucleic Acids Res 7:1513 37. Haq S, Khan R (2005) Int J Biol Biomacromolecules 36:47
22. Sambrook J, Fritsch E, Maniatis T (1989) Molecular cloning. A 38. Sponner A, Schlott B, Vollrath F, Unger E, Grosse F, Weisshart
laboratory manual. Second Edn. Cold Spring Harbor Laboratory K (2005) Biochemistry 44:4727
23. Wong C, Sridhara S, Bardwell J, Jakob U (2000) BioTechniques 39. Xu M, Lewis R (1990) Proc Natl Acad Sci USA 87:7120
28:426 40. Chen X, Shao Z, Vollrath F (2006) Soft Matter 2:448
24. Whithmore L, Wallace B (2004) Nucleic Acids Res, Web Server 41. Blackledge T, Hayashi C (2006) J Exp Biol 209:2452
issue 24:W668 42. Blackledge T, Hayashi C (2006) J Exp Biol 209:3131
25. Scheller J, Guhrs K, Grosse F, Conrad U (2001) Nature 19:573
26. Teule F, Jung S, Wood J, Marcotte W, Ellison M, Abbott A
(2002) In: Brebbia C, Sucharov L, Pascolo P (eds) Design and
nature, Udine, Italy. WIT press, p 379
123
J Mater Sci (2007) 42:89868994
DOI 10.1007/s10853-007-1831-3
Received: 30 March 2007 / Accepted: 8 May 2007 / Published online: 17 July 2007
Springer Science+Business Media, LLC 2007
123
J Mater Sci (2007) 42:89868994 8987
associated with a chemical reaction that occurs between rinsed with room temperature water and allowed to dry on
strained crack tip bonds and an environmental species such filter paper (Grade 595 General-Purpose Filter Paper,
as H2O that leads to bond rupture and crack advance [9]. In Schleicher & Schuell MicroScience GmbH, Dassel, Ger-
other materials, inherent creep or relaxation processes at many) then removed and stored in ambient conditions of
the crack tip may lead to time-dependent material separa- 1823 0.8 C and 3555 2% relative humidity (RH)
tion and crack advance [10, 11]. The phenomena is often (Dickson TM325 Temperature & Humidity Data Logger,
referred to as subcritical cracking or delamination since it Addison, Illinois). Comparative tests were performed on
occurs at G < Gc, that is, at applied loads lower than those single donor tissue specimens to reduce variability within
required for fracture in the absence of an environmental test sequences. During separation, orientation of the outer
species or inherent relaxation process. Some materials SC surface was recorded.
exhibit no detectable subcritical cracking. In the case of SC To vary SC hydration, isolated tissue was exposed to
no studies have reported on time-dependent intercellular environments with different RH values at ambient tem-
delamination. perature (1823 C). A portion of the SC was placed in an
In the present study we demonstrate that SC is suscep- enclosure held at 40% RH while additional SC was placed
tible to subcritical delamination for applied G < Gc. in another enclosure along with an open container of water
Intercellular delamination growth rates, v, were character- to create a 100% RH environment to increase SC moisture
ized in selected moist air environments. We employed content. The tissue was allowed to equilibrate for at least
methods recently applied to measure subcritical delami- 24 h prior to fabrication of test specimens.
nation behavior in thin-film and layered structures for For comparison, SC, 60 60 mm2, was chloroform-
nanoscience and biomedical applications [9, 1216]. methanol treated (CMT) to delipidize the structure with a
Delamination growth rates were shown to be a strong 120 min 30 mL chloroform:methanol (2:1 by volume) soak
function of the applied G resulting in a characteristic vG followed by two 30 min 30 mL water rinses. Additional
curve. Conditioning of the SC tissue by changing hydration tissue was treated in selected surfactant solutions to examine
content and chemical treatments using surfactants were their effects on SC mechanical integrity. These specimens of
demonstrated to affect both critical and subcritical delam- SC 30 50 mm2 were immersed in 100 mL of 10% wt/wt
ination behavior. Increasing SC hydration is shown to H2O solutions of alkyl amidosulfosuccinate (AAS), alkyl
accelerate delamination growth rates consistent with polyglucoside (APG), and SDS for 18 h, then rinsed with
behavior anticipated from critical delamination testing. distilled water and dried on filter paper. During treatment of
Chemical exposure including 10% wt/wt sodium dodecyl SC, the solution pH values were pH 5.2, 8.18.7, and 7 0.1
sulfate (SDS) and chloroform-methanol (2:1 v:v) soaks (Accumet Research AR25 Dual Channel pH/Ion Meter,
that remove lipid suppressed growth rates compared to Accumet Research, Fisher Scientific, Fair Lawn, NJ) for
untreated controls. These treatments have been shown each treatment, respectively. These surfactant solutions have
previously to affect lipid content and fluidity [4, 17]. been shown to cause varying levels of lipid and protein
Measured subcritical delamination behavior as a function damage as indicated in a comparative manner in Table 1.
of treatment is interpreted in terms of changes to SC Note that our previous studies have revealed that donor
microstructure and constituents. age and body location has an effect on measured delami-
nation behavior [1, 2]. In the present study we compare
results from single donors to negate these variations.
Materials and methods
Delamination energy measurements
Tissue preparation
The fracture mechanics technique developed to measure
Human cadaver SC tissue was isolated from two female the delamination energy of SC tissue has been described
Caucasian donors, 57 and 101 years of age, from the thigh. elsewhere [1]. Briefly, specimens were prepared by
Epidermal tissue was separated from dermis by immersion adhering SC tissue between two elastic substrates of
of donor tissue cleared of adipose tissue in a 35 C water polycarbonate (Hyzod GP, Sheffield Plastics Inc., Shef-
bath for 10 min followed by a 1 min soak at 60 C and field, Massachusetts) with cyanoacrylate adhesive (Instant
subsequent mechanical separation from the dermis using a Krazy Glue Gel, Elmers Products Inc., Columbus, Ohio)
flat-tipped spatula. Stratum corneum was isolated from to form a double-cantilever beam (DCB) fracture
underlying epidermis by soaking in a trypsin enzymatic mechanics specimen as shown in Fig. 1. The substrate
digest solution (0.1% wt/wt in 0.05 M, pH 7.9 Tris buffer) dimensions of 40 10 2.88 mm3 were selected to ensure
at 35 C for 120 min. During separation, the orientation of elastic deformation and the valid application of linear
the outer SC surface was recorded. The isolated SC was elastic fracture mechanics.
123
8988 J Mater Sci (2007) 42:89868994
Table 1 Table of showing SC tissue treatments, their relative protein and lipid damaging properties, and critical and subcritical delamination
failure parameters
Treatment Protein damage Lipid damage Gc (J/m2) Slope, m (da/dt/J/m2) GTH (J/m2)
D 2 a 0:64h3 1=3
C 1 D E0 bh3
P 3 E0 I a 0:64h 3
P 8
where C is the DCB specimen compliance, P is the load,
D/2 is the corresponding displacement of each beam from During the load-relaxation component of the subcritical
its original position at the loading point, E = E/(1V2) is test, the displacement, D, is held fixed with time and
the plane strain Youngs modulus for the polycarbonate, m knowledge of the initial debond length, ai, and peak load,
is Poissons ratio, I = bh3/12 is the area moment of inertia, Pi, from the initial loading of the specimen, allows calcu-
b is the polycarbonate substrate width, and h is the height lation of the delamination length as a function of time:
123
J Mater Sci (2007) 42:89868994 8989
h Pi 1=3 (a) 2.4
at ai 1 0:64 0:64h 4 Pi o
25 C 40% RH
ai Pt 2.0
Load, P (N)
taking the time derivative of the above expression:
1.2
!
1=3
da ai h Pi dP 0.8
1 0:64 4=3
5
dt 3 ai Pt dt
0.4
25 C 40% RH
of applied G as shown in Fig. 3. The corresponding
-5
critical delamination energy, Gc, is included for compar- 10
ison. Delamination growth rates were characterized over
-6
nearly five orders of magnitude, and were clearly sensi- 10
tive to the applied loads. Intermediate growth rates in the
-7
range of 108105 m/s exhibited an exponential depen- 10
dence on G. Alternatively, a power-law relationship of the -8
form da/dt = CGm can be employed to characterize the 10
intermediate delamination growth rates, where C and the -9 G TH
10
exponent m are constants dependent on tissue condition
and testing environment. Threshold behavior for applied -10
10
loads approaching a threshold applied strain energy 1 2 3 4
2
release rate, GTH, was clearly apparent at low growth Applied Strain Energy Release Rate, G (J/m )
rates approaching 109 m/s. Delamination is presumed
Fig. 3 Typical delamination growth rate (da/dt) versus applied strain
dormant at applied G less than GTH. Not all materials
energy release rate (G) for SC conditioned at 1823 C, 40% RH and
exhibit a clear threshold load for crack or debond growth tested in a 25 C, 45% RH environment showing threshold behavior
as reported in a recent study of debonding of a polymer/ (GTH) and critical delamination energy values (Gc)
123
8990 J Mater Sci (2007) 42:89868994
100% RH
-5
10 (92% RH) higher humidity testing environment was approximately
2.6 104 s.
-6
10 increasing Tissue conditioned at 100% RH and tested in the drier
hydration 45% RH environment exhibited behavior essentially the
-7
10 40% RH same as that for specimens treated and tested at the higher
(45% RH)
RH. However after 4 103 s, the tissue was affected by the
-8
10 drier testing environment and exhibited some contraction
that was evidenced by the measured load beginning to
-9
10 precond. RH increase with time. This behavior invalidated the load
(chamber RH) relaxation analyses and the test was terminated at the last
valid growth rate in the vicinity of 108 m/s.
-10
10
1 2 3 4 5 6
2
Applied Strain Energy Release Rate, G (J/m )
Effects of tissue surfactant treatments
Fig. 4 Delamination growth rate (da/dt) versus applied strain energy
release rate (G) for SC specimens conditioned in environments of Delamination was performed on CMT, AAS, APG, and
different RH at ambient temperatures 1823 C and tested in
SDS treated tissue to examine the effects of these solutions
environments with RH similar to their preconditioning values. Labels
show the preconditioning RH and environmental test chamber RH in on SC delamination growth rate behavior compared to that
parentheses of untreated controls. These specimens were conditioned
123
J Mater Sci (2007) 42:89868994 8991
and tested at 45% RH at ambient temperatures 1823 C bonds and an environmental species such as H2O which
and 25 C, respectively. These data are presented in leads to bond rupture and crack advance [9, 1416]. This
Fig. 6 with parameters listed in Table 1. The shapes of type of environmentally assisted cracking generally has
the delamination growth rate curves were similar to the characteristic features of the resulting vG curve. Inter-
untreated SC although the slope, and particularly the mediate growth rates in the range of 108105 m/s are
position of the vG curves along the G-axis, exhibited a strongly sensitive to applied G, the activity of the envi-
marked sensitivity to the tissue treatment. The curves all ronmental species and temperature, indicative of the
exhibited threshold behavior at low delamination growth environmentally assisted reaction kinetics associated with
rates. The harsh CMT and SDS surfactant treated tissue the crack tip bond rupture process. Further decreases in G
exhibited delamination growth rate curves at significantly often result in threshold behavior below approximately
higher G values compared to the untreated control, and the 109 m/s where crack growth rates decrease below
tissue treated in the milder surfactants, AAS and APG, detectable limits. At higher growth rates, a transport-lim-
exhibited curves at lower applied G values indicative of ited region largely independent of G is characterized by a
decreased resistance to delamination. Critical delamination plateau in the vG curve [9, 1416]. We did not observe a
energy values, Gc, exhibited similar trends with treatment transport limited region in the present study, but such
to the time-dependent delamination curves. behavior is often observed at higher crack growth rates,
which we did not characterize. With still further increases
in applied loads where G approaches Gc fracture occurs at
Discussion high crack growth rates and is largely independent of
environmental test conditions.
Susceptibility to subcritical delamination Time dependent fracture may also be related to inherent
creep or relaxation processes that occur in a process zone
The data presented in the present study demonstrates that surrounding the crack tip that lead to material separation
SC tissue is susceptible to time-dependent intercellular and crack advance. Such behavior has been reported for
delamination growth under load in moist environments. metals, inorganic, and organic materials [10, 11]. In the
Not all materials or interfaces exhibit such time dependent case of polymers, the time dependence of crack growth
delamination or cracking at applied loads less than the results from viscoelastic deformation in the crack tip pro-
fracture energy Gc. cess zone which is dependent on both testing temperature
The observed behavior may be associated with a and crack growth rate [11]. While we were not able to
chemical reaction that occurs between strained crack tip unambiguously establish the precise time dependent deb-
onding mechanism in the present study, the intercellular
separation process that occurs during delamination of SC
10
-4
GC GC GC
involving separation of intercellular lipids and corneodes-
GC GC
mosomes suggests that the latter process involving an
Delam. Growth Rate, da/dt (m/s)
-5 APG AAS Ctrl SDS CMT inherent time dependent molecular separation process is
10
more likely. We note also that the SC does not contain
-6
atomic bonds like silanol bonds in inorganic glasses that
10
are particularly sensitive to moisture adsorption and asso-
-7
ciated bond cleavage. Organic molecules containing carbon
10
backbones are generally insensitive to cleavage by H2O or
-8
OH molecules in moist environments and therefore typi-
10 cally insensitive to environmentally assisted cracking.
-9
10 Hydration effects
o
25 C 45% RH
-10
10 The reduction in critical delamination energies, Gc, with
0 1 2 3 4 5 6 7 8
2 SC hydration (Figs. 4, 5) is consistent with prior examin-
Applied Strain Energy Release Rate, G (J/m )
ations of critical delamination energy values [1, 2]. The
Fig. 6 Effect of chemical treatments on critical and subcritical effects of hydration on SC components are numerous
delamination behavior. Critical and subcritical delamination energy including corneocytes swelling and the presence of water
measurement results for various pH and surfactant treatments (SDS:
sodium dodecyl sulfate (10% wt/wt); CMT: chloroform-methanol
pooling between cells [21, 22], disruption of intercellular
(2:1 v:v) treatment; APG: alkyl polyglucoside (10% wt/wt); AAS: lipid lamellae [23, 24], and even the degradation of cor-
alkyl amidosulfosuccinate (10% wt/wt)) neodesmosomes [7, 21]. We discussed the decreased
123
8992 J Mater Sci (2007) 42:89868994
123
J Mater Sci (2007) 42:89868994 8993
where C a constant that includes the concentration of layer leading to overall increases in delamination energies
reactant molecules, the interfacial bond length and areal compared to untreated controls [2]. The strong surfactant
bond density, aH2 O is the moisture activity, D0 is the pre- quality of SDS likely plays a similar role in extracting
exponential diffusion coefficient of water in the polymer, intercellular lipids leading to increases in critical delami-
Ed is the activation energy for diffusion, a is a constant nation energies.
involving the activation volume associated with the diffu- Extending our knowledge of surfactant behavior on
sion process, R the universal gas constant, and T the tem- critical delamination energies to subcritical behavior, we
perature in Kelvin. The constant a/RT in the exponential find delamination growth rate curves shifted to higher
was found to be approximately 1 kJ1/2 m/mol. While is it G-values with strong surfactant treatments (Fig. 6) similar
premature to attempt a similar detailed fit of the above to trends observed between critical delamination energy
model to the present study where the structure of the values. The similar shapes of the vG curves between un-
intercellular delamination path including details of the lipid treated and treated specimens indicate that the mechanisms
and corneodesmosome bond length and areal density are driving delamination growth are similar but that higher
not known, it is nevertheless instructive to note that the applied G values are required to facilitate activation of the
observed plateau in growth rates is likely related to a very kinetic processes leading to subcritical delamination
similar mechanism involving moisture diffusion through growth. The similar behavior to reaction-controlled curves
the SC layer. If that is the case, then it is also possible that for treated delamination growth velocities as a function
the moisture diffusion will exhibit the same stress depen- of G suggests that similar mechanisms to those seen
dence observed in other organic polymer materials. We in untreated SC may be occurring. The differences in
should then expect to observe a similar weak dependence subcritical curve positions along the G axis may be
of growth rates in the plateau region on the applied G as attributed to modifications of the intercellular material
predicted by Eq. 6. Using a simplified form of Eq. 6 we through which the debonding occurs. For the CMT and
find that a best fit to the data yields a value of SDS treated SC, depletion of intercellular lipid bonds with
a/RT = 1.5 kJ1/2 m/mol which is in close agreement with strongly corneocyte envelope lipid interactions provides an
the previous value of a for weakly bound polymer inter- obvious explanation of the increased G-values; however,
faces, and further that the model captures the plateau the accelerated delamination rates that occur with exposure
growth rate dependence on G very well. In fact, the to milder surfactants (AAS and APG) cannot be explained
successful application of the stress-dependent transport in the same manner.
model suggests that like other organic polymer materials, Examining the milder surfactant treatments, their critical
water diffusion in SC may well be stress dependent. delamination energies fall toward lower values than those
of the untreated controls. These results suggest that these
milder surfactants do not affect the SC in the same manner
Chloroform-methanol and surfactant effects as the harsher surfactants. The results are surprising
because any lipid extraction would be expected to cause
When exposed to chemical treatments, SC exhibits sig- increases in SC delamination energies while exposure to
nificant changes in delamination behavior as seen in Fig. 6 these solutions led to decreases. One competing mecha-
and quantified in Table 1 for both critical and subcritical nism that may explain the incongruous behavior is that of
measurements. The treatments used to condition SC in SC hydration content. While the described tests control
these subcritical tests have been shown to damage protein external factors such as environmental RH, the actual water
and lipids structures to varying degrees as compared in content of the SC may vary with treatment as water-hold-
Table 1. Chloroform-methanol extraction has been shown ing capacity is modulated if hygroscopic natural moistur-
previously to remove the majority of intercellular lipids izing factor (NMF) content is changed in the system.
from SC tissue facilitating the direct apposition of unex- Alternatively, the treatments may increase water holding
tracted lipids covalently bound to the cornified envelopes capacity of the skin as is the case with glycerol which has
of adjacent corneocytes [4, 2529]. The increased critical been shown to absorb into SC and whose large water-
delamination energies of delipidized SC have been attrib- holding capacity is well known [32, 33]. This effect of
uted to the interdigitation of opposing corneocyte envelope hydration may explain the decreases in delamination
lipids [28, 30]. Only with extraction of the lamellae and energies with AAS and APG treatment but require further
envelope lipids is dissociation of SC tissue into individual examination to provide evidence of these effects. The
cells observed, supporting this explanation [31]. Through subcritical data, however, faithfully track with the critical
thickness multiple delamination tests on CMT SC has values and exhibit similar behavior to that of untreated and
indicated that the extraction is pervasive throughout the SC harsher surfactant-treated SC.
123
8994 J Mater Sci (2007) 42:89868994
123
J Mater Sci (2007) 42:89959004
DOI 10.1007/s10853-007-1741-4
Received: 26 February 2007 / Accepted: 2 April 2007 / Published online: 17 July 2007
Springer Science+Business Media, LLC 2007
123
8996 J Mater Sci (2007) 42:89959004
To begin with, viral capsids are a particularly impressive capsid diameters vary roughly from about 20 nm to a few
result of two-dimensional protein assembly, which serve to hundred nm. Constituent proteins are roughly similar in size
contain and protect viral genome molecules (either DNA or for most viruses, such that capsid size tends to increase
RNA). These protein shells assemble from multiple copies directly with T number. For most viruses, capsid size is
of only a few capsid proteins (in some cases only one) held uniformly and precisely determined and is consistently
together by non-covalent forces into very regular structures reproduced under physiological conditions. One notable
that are, in most cases, either cylindrical or spherical in exception is the human immunodeficiency virus, HIV-1,
shape. In rare cases, capsid self-assembly is achievable the capsid of which is polymorphic, marked by variation
in vitro from purified protein components [6]. Because of in size and shape [13]. Interestingly, similarities in the
their regularity, the structures of many capsids have been multiscale structural ordering extend to sub-protein,
determined to very high resolution (better than 3 A ) molecular spatial scales as well. For example, as revealed in
through combined X-ray crystallography and cryo-electron structures determined by X-ray crystallography and cryo-
microscopy (cryo-EM) studies. Figure 1 shows represen- EM, the coat proteins of many plant, insect, and animal
tations of two viruses that have been extensively studied viruses having icosahedral structure are formed around the
both structurally and mechanically: the plant virus, cowpea same core structural motif, that of the so called jelly roll
chlorotic mottle virus (CCMV), and the bacteriophage b barrel [14].
/29. The structure of CCMV is known at a resolution of Despite their precise ordering, significant configura-
about 3 A [7], and so the atomic coordinates are known tional changes have been observed at various stages
with a high degree of certainty, as shown in Fig. 1a. The throughout the viral life cycle for some capsids. For
low-resolution representation of the CCMV capsid is instance, many viruses first assemble as procapsids
shown in Fig. 1b, enabling the icosahedral nature of the which then mature into infectious capsids. Maturation can
capsid to be clearly seen. The structure of the /29 capsid is involve structural transitions where proteins are rearranged
not yet known at atomic resolution, thus the only data (e.g., HIV-1 [15]), existing covalent bonds are broken (e.g.,
available is a density map created from low-resolution flock house virus (FHV) [16]), or new covalent bonds are
cryo-EM data. The density map as determined by Morais formed (e.g., HK97 [17]). Structural transitions in capsids
et al. [8] is shown in Fig. 1c. can also be triggered by environmental changes, as is the
The structure of spherical capsids is ordered according to case for the pH controlled swelling of CCMV [7, 18] and
the symmetry of an icosahedron, with capsid proteins the pH triggered release of a pentamer by tymovirus [19].
grouped locally into fivefold and sixfold capsomers. Five- Similarly, calcium-mediated deformation of simian virus
fold capsomers (pentamers) adopt the role of the 12 icosa- 40 (SV40) is thought to play a role in its infection process
hedral vertex positions, whereas the sixfold capsomers [20]. The mechanics of these structural transitions is not
(hexamers) fill in the 20 icosahedral face regions. The well understood nor is it generally known how these con-
pentamerhexamer count can vary according to the formational changes can be controlled by mechanical force
T-number classification proposed by Caspar and Klug stimulus. A notable exception is the maturation of FHV
[12] in the 1960s, which groups the total number of 60T which is influenced by pressure on the capsid [16].
capsid proteins into 12 pentamers and 10(T1) hexamers. Recently, technological advances in single-molecule
The T-number index can only adopt certain integer values, experimentation have enabled direct measurement of the
1, 3, 4, 7, 9, 13, . Among the different spherical viruses, strength and elasticity of viral capsids. Measurement of the
Fig. 1 Representations of the CCMV (a) and (b) and /29 (c) viral atomic resolution, thus only low-resolution representations such as
capsids. The structure of CCMV has been determined to atomic density maps are available, and shown in (c). Renderings of CCMV
resolution, and therefore the atomic positions are known with a high and /29 were generated with the molecular imaging software,
degree of precision, as shown in the atomic representation (a). The Chimera [9]. The CCMV atomic coordinates were downloaded from
low-resolution representation of CCMV in (b) shows the icosahedral the RCSB Protein Data Bank [10], and the /29 density map was
structure. The capsid structure of /29 has not beed determined to downloaded from EMBL-EBI [11]
123
J Mater Sci (2007) 42:89959004 8997
forces required to package DNA into bacteriophage /29 exhibit some degree of structural flexibility, as they may
using optical tweezers has shown that its capsid is strong undergo conformational changes in which their size can
enough to sustain effective pressures estimated on the expand or contract, and may self-assemble from pure
order of 60 atm [2124]. Traditional experimental methods components. Effective utilization of these properties re-
for determining elastic properties of materials involve quires a full understanding of the underlying mechanisms.
stretching and compressing them. Atomic force micros- Complementary theories are needed to bring about the
copy (AFM) allows for this to be done with single mole- understanding of these mechanisms, and mechanical
cules and their assemblies. AFM has been used to study the models provide an excellent foundation for determining the
mechanical properties of the capsids of several viruses: properties that control these behaviors.
/29 [25], CCMV [26, 27], parvovirus minute virus [28], To date, the use of bottom-up mechanical models such
and murine leukemia virus [29]. In fact the technique is as molecular dynamics (MD) and normal mode analysis
useful for a general class of protein assemblies and has (NMA) have been the most widely used methods to study
also been used recently to study buckling of microtubules the unforced vibrational dynamics of viral capsids. These
[30]. In these experiments, capsids are indented with the methods tend to involve an all-atom description of capsid
AFM cantilever tip, producing a force-deflection curve. structure, although coarse-graining has gained in use. In
Figure 2 shows force-indentation results from experiments fact, viral capsids, which are large macromolecules with
performed on CCMV at two different pH levels [26, 27]. many thousands of atoms, are nearly impossible to study
At pH 5 a jump in force is observed at a compression of with these methods without some form of coarse-graining.
about 30% and accompanied by a permanent decrease in Because these methods are meant to study the unforced
stiffness (Fig. 2b) indicating an irreversible mechanical vibrational dynamics of a system, they give an indirect
failure of the capsid. In contrast, the pH 6 force response look at the mechanics of viral capsids. One of the important
remains linear without significant hysteresis throughout the results to come out of the coarse-grained studies of viral
complete indentation range, even to the point where the capsids is that the overall structure, and not the atomic
capsid is compressed such that the top and bottom inner detail, is the main factor in controlling the unforced
surfaces of the capsid shell are pressed into contact. Thus, mechanics. With the development of new direct experi-
at pH 6 CCMV is almost perfectly elastic, and effectively mental methods of probing viral capsids, such as the AFM
indestructible under nanoindentation. experiments described above, mechanical models are nee-
The collection of experiments on viral capsids reveal a ded to describe the forced mechanical behavior. In the last
number of mechanical properties that contribute to their few years, top-down approaches such as continuum mod-
allure for materials applications. Namely, their surprising eling have been developed and used as a theoretical com-
strength and flexibility as found by AFM experiments, even plement to describe the response of viral capsids under
under large deformations, sets them apart from most tra- forced conditions, with the results suggesting that the
ditional materials. Additionally, the ability to control and overall shape of the capsid also plays a dominant role in
alter the material properties of some viral capsids as a governing the forced mechanical behavior. Ultimately, it is
function of adjustable environmental conditions, such as likely that novel multiscale methods will be required to
pH, makes them versatile. And yet, viral capsids are not capture both the global structural response of the capsid in
weak; as noted above, some viral capsids are able to sustain addition to the detailed atomic response that will govern
extremely high pressures [2124]. Finally, viral capsids can complex behaviors such as failure.
Fig. 2 Indentation force plotted versus substrate displacement for and capsid in series). Loading is described by the red curves,
CCMV capsids at pH 6 (left) and pH 5 (right) [26, 27]. For a unloading by the blue curves. The large drop in force and presence of
corresponding force, the indentation is given by the distance between significant hysteresis in the pH 5 case is indicative of capsid failure.
the black curve (cantilever alone) and the colored curves (cantilever (Figure adapted from [27].)
123
8998 J Mater Sci (2007) 42:89959004
Modeling viral capsids from the bottom up environment of the entire capsid is modeled, while the
smaller number of atoms in a single subunit is utilized. The
Experimental methods such as X-ray crystallography and use of the rotational symmetry boundary conditions
cryo-EM used to determine the atomic structures of viral showed a significant reduction in the edge effects as atoms
capsids show that they are not entirely static. At the atomic near the boundaries remained close to the crystal coordi-
scale, thermal motions are detectable and the equilibrium nates. However, the total simulation time only varied from
structure describes the average position of all atoms over 15 ps [35] to 60 ps [36].
time. These structural methods are also able to detail changes Until very recently, MD simulations performed on viral
in the atomic positions in response to the binding of external capsids were limited to the asymmetric unit (the collection
molecules, such as antiviral complexes, and the structural of subunits that comprise one icosahedral unit, and in the
rearrangements seen during conformational changes. In or- case of T = 1 capsids, a single subunit) alone, as detailed
der to understand both the equilibrium fluctuations and above. The first molecular dynamics simulation on an en-
structural changes, mechanical models of viral capsids were tire virus was completed in 2006 [40]. The simulation was
first developed to study their unforced vibrational dynamic performed on the T = 1 satellite tobacco mosaic virus
behavior almost twenty years ago. Methods such as molec- (STMV). The T = 1 capsid structure is the smallest among
ular dynamics (MD) and normal mode analysis (NMA) spherical viruses. The simulation was performed with both
represent a bottom-up approach to modeling viral capsids; the capsid proteins and a highly simplified model of the
the models were initially used with atomic level detail, and nucleic acid, which was built from identical RNA segments
more recently, coarse-grained models have been introduced. to match the icoshedrally averaged electron density of
RNA, in a completely solvated environment, with
~1,060,000 total atoms, ~900,000 of which were water.
Molecular dynamics The time scale of the simulation was 10 ns. The aim of the
simulation was to investigate the stability of the viral
Beyond studying the equilibrium fluctuations of viral caps- capsid with and without the RNA present, and it was found
ids, MD has been used to model several specific types of that the capsid without the RNA was unstable after 10 ns.
problems: assembly [31, 32], stability [33, 34], and antiviral The capsid was, however, extremely stable with the RNA
activity [3538], to name a few. MD is useful for the mod- present in the simulation. Unlike some viruses that have
eling of very short time scale dynamics (on the order of pico been observed to self-assemble in vitro into empty capsids
and nanoseconds), and captures both anharmonic and har- (no nucleic acid packaged), empty STMV has not been
monic motions. The main challenge in the application of MD observed [40].
to viral capsids is the number of atoms in the system. There is In order to perform MD on viral capsids, an all-atom
no general rule for estimating the number of atoms per pro- structure needs to be available for the system of interest.
tein subunit, and thus the number of atoms for different These are not available for all viral capsids. All-atom
viruses with the same T-number may widely vary. However, descriptions necessarily limit the size of the system that can
among several representative T = 3 viruses, the total number be studied, and at the time scales sampled by MD, large
of atoms varies from about 150,000 to 250,000. Conse- deformations of a capsid, such as conformational changes,
quently, most MD simulations performed on viral capsids are cannot be modeled. Even the studies that use the rotational
limited to single subunits, with symmetry conditions symmetry boundary condition method to model only a
enforced so as to capture the icosahedral environment. small fraction of the capsid are limited in the types of
Viral capsids have been studied using MD for almost motions that are output; any type of asymmetric or global
20 years, with the first simulation, performed in 1988, motion will not be captured. Bottom-up descriptions such
consisting of a single human T = 1 rhinovirus subunit in as MD are advancing very slowly in terms of their capa-
vacuo, with an antiviral compound bound to the subunit in bility. The use of MD to model externally applied forces,
order to determine the dynamic effects of the binding [39]. such as would be present in experimental methods such as
This simulation was performed on ~8000 atoms, and the AFM, is probably not feasible. Experimental methods such
results showed large edge effects as no boundary condi- as AFM incur large deformations and occur on a time scale
tions were applied, and thus the icosahedral environment of of milliseconds to seconds. These deformation and time
the entire capsid was not modeled. In 1991, a method was scales are simply inaccessible by all-atom methods, unless
developed to model the complete capsid environement by some form of coarse-graining is applied. Aware of these
taking advantage of the rotational symmetry boundary limitations, a new coarse-grained MD method has been
conditions present in icosahedral viruses, and MD was recently developed to study the dynamic stability of larger
again applied to the human rhinovirus subunit in vacuo viral capsids, extending the timescale to the microsecond
[35], and fully solvated [36]. In this way, the physical range [65].
123
J Mater Sci (2007) 42:89959004 8999
Normal mode analysis such as viral capsids to be studied. In addition, the intital
energy minimization required when using the complex in-
Normal mode analysis (NMA) was developed as a more teratomic potential is also avoided. The EN model is useful
computationally efficient alternative to MD. Initially, NMA for the lowest modes of vibration as the high frequency
simulations were all-atom and involved the same interatomic modes cannot be modeled well with simplified potentials.
potential field as MD, but by using a quadratic approximation In two studies, Tama and Brooks [50, 51] used coarse-
for the potential energy, the problem is reduced to a normal grained NMA with the EN model to study the conforma-
mode analysis on a structure and only outputs harmonic tional changes observed in icosahedral virus capsids. In
motions. For an excellent reference on both the theory and 2002, they reported that a coarse-grained NMA of the
applications of NMA, see Cui and Bahar [41]. The potential CCMV capsid, which undergoes a pH induced swelling
energy, or dynamical, matrix is formulated from the second with ~10% size increase in the radial direction, showed that
derivatives of the potential function. In this way, the time just a few of the lowest symmetric normal modes con-
propagation of MD is eliminated, and the computationally tributed to the swelling displacement [50]. The goal was to
expensive step is diagonalizing the dynamical matrix, after model several pathway intermediate structures, to get a
which the output eigenvalues (frequencies) and eigenvectors better idea of the structural changes during swelling. The
(normal modes) are easily extracted. However, for larger CCMV capsid was coarse-grained using the Ca atoms, the
systems, the diagonalization becomes a major bottleneck as simplified Hookean potential was used, and the rotations-
the size of the dynamical matrix grows proportionally to the translations of blocks (RTB) method [56, 57] was used to
size of the system being studied. Thus, the use of all-atom simplify the problem by assuming each individual subunit
NMA was limited for many years to small proteins due to was rigid. The normal modes are computed as a linear
computational limitations. combination of the rotations and translations of the rigid
It has been shown that the lowest modes tend to repre- blocks. In 2005, the same methods were applied to study
sent coordinated global motions of the structure being conformational changes of other viruses [51]. It was shown
studied, and are therefore most relevant for large motions that for viruses with known conformational changes
such as conformational changes [42]. Generally, NMA is (CCMV, HK97, and NxV) one symmetric mode captures
used to find normal modes that capture conformational the bulk of the motion. The next few lowest symmetric
changes of proteins [43], as an aid in the refinement of low- modes are needed to more accurately describe the known
resolution cryo-EM density data [4446], and to describe conformational pathway. They argue that this shows the
the general unforced dynamics of macromolecules [47, 48]. method is robust, and can be predictively applied to other
As applied to viral capsid structures, NMA is used to study viruses for which both conformational states are not
maturation dynamics and conformational changes such as known, or for viruses for which it is not known if con-
capsid swelling [4951]. The time scales associated with formational changes exist. Also in 2005, Rader et al. [49]
the NMA frequencies are longer than that of MD, but are applied coarse-grained NMA to study the conformational
still limited to the microsecond range. changes that the HK97 virus undergoes during maturation.
An early all-atom NMA study was performed on the rod It was shown that the first eleven modes capture over 98%
shaped tobacco mosaic virus (TMV), in which one full unit of the observed maturation pathway. The low-frequency
that is necessary for symmetryseventeen subunits that modes also showed regions of relative flexibility, at sub-
make up on turnwas modeled for the analysis [52]. The strate recognition sites, and relative stiffness, at anchors or
standard CHARMM force field was used [53]. In 2005, van hinge sites. This study proposed a maturation pathway
Vlijmen and Karplus [54] reported the first atomic-level ultimately leading to a cross-linked chain-mail complex.
NMA calculations on a full icosahedral virus capsid, also If EN models are used in a normal mode analysis, at any
using a full interatomic potential. However, all-atom NMA level of coarse-graining, the need for the initial minimi-
calculations using standard interatomic potentials remain zation of the potential energy is eliminated, but the results
quite rare due to the computational demands, the devel- are less physical. As the leading constant in the Hookean
opment of simplified potentials, and the use of coarse- potential is the only tunable parameter in the model, the
grained models. value is found by fitting the normal mode results to
In 1996, Tirion [55] introduced a simplified pairwise experimental vibrational data. There is then a loss of some
Hookean potential to replace the complex interatomic predictive power. Similarly, using NMA to predict the
potential, and the method became known as the elastic pathways of large conformational changes is only possible
network (EN) model. The potential involves a single if the initial and final states are known, as it is not obvious
parameter that is equal for all interactions, thus the need for which of the lowest modes capture the relevant pathway.
all atoms is avoided. This allows for NMA to be applied to The only exception to this might be viral capsid swelling,
coarse-grained models, and consequently for larger systems which appears to involve only symmetric motion, and thus
123
9000 J Mater Sci (2007) 42:89959004
comprised of only the symmetric modes. However, van theory of thin elastic shells to study the shapes and shape
Vlijmen and Karplus [54] argue that symmetric modes, transitions of the icosahedral viruses. These models treat
which Tama and Brooks [50, 51] use in their coarse- the total energy of the system as a sum of both bending and
grained NMA analyses of viral capsids, are not necessarily stretching contributions. Building off of previous work on
the true pathway of a conformational change, and non- the stability of planar elastic sheets [58], Lidmar, Mirny
symmetric modes may need to be studied to fully under- and Nelson (LMN) [59] hypothesized that the pentamers of
stand the pathway. an icosahedral capsid act as fivefold disclinations, which
It is fair to say that NMA is currently the best method are necessary to form a closed shell from an otherwise
available to study unforced vibrational macromolecular hexameric lattice. They showed that the degree of faceting
dynamics. Currently, all-atom NMA simulations are very is dependent on the relative stiffness of bending and
complex, so coarse-grained models are the preferred stretching and the size of the shell. In particular, the energy
method for large systems such as viral capsids. Not only minimizing shape of the capsid shell transitions from
has NMA established coarse-graining as a useful step in roughly spherical to icosahedrally faceted with an increase
studying large systems, but as a necessary one as well. in value of a dimensionless parameter, the Foppl-
Coarse-graining allows for systems without full atomic vonKarman (FVK) number c = YR2/j, where Y is the two-
descriptions to be studied, in addition to highlighting the dimensional Youngs modulus, j is the bending rigidity,
fact that the atomic details are not a main factor in deter- and R is the average radius of the capsid. There is a critical
mining the global motions. One of the main conclusions value of the FVK number at which a transition occurs that
that has been drawn from the numerous coarse-grained can be described as buckling. For sub-critical FVK num-
NMA studies is that the overall shape of a structure gov- bers bending stiffness dominates over stretching and the
erns its unforced vibrational dynamics [42]. This may have capsid shape is roughly spherical. Above this transition,
far-reaching implications for modeling viral capsid systems the structure has a lower energy if the surface is bent near
under forced conditions. In fact, it is reasonable to assume the disclination, thereby relaxing the otherwise dominant
that shape governs forced mechanics, and based on recent stretching energy. This effect is argued to be a cause for the
computational studies of viral capsids where only the shape visible faceting. As the subunits of different viruses have
is preserved, this appears to be precisely the case. similar molecular structure, they should be expected to
have similar mechanical properties. Thus, the spherical
or faceted shape of a virus is predicted to scale with its size.
Modeling viral capsids from the top down Following the LMN study, Nguyen et al. [60, 61] built on
the thin elastic shell theory to include spontaneous curva-
Recent single molecule experiments such as atomic force ture in an attempt to discover non-spherical equilibrium
microscopy have given rise to a number of new top-down capsid shapes, such as the conical capsid shape often
two- and three-dimensional continuum elasticity models seen in HIV. The two-dimensional icosahedral models of
involving the mechanical response of viral capsids. As Zandi and Reguera [62] showed that in addition to the size
discussed above, a major result of previous coarse-grained dependent shape character of viral capsids, there is a size
NMA studies was that the most important factor deter- dependent stress character; the degree of stress concentra-
mining a structures unforced vibrational response was its tion at the pentamer sites grows with increasing T-number.
shape, not atomic detail. Similarly, here it is argued that the Icosahedral models of viruses have very recently been
same is true for forced mechanical response, as the results extended to model the mechanical response of capsids
of several recent continuum studies of viral capsid under applied loads. Simulations of AFM experiments have
mechanics seem to support. been performed to study the effect of the shape transitions
implied by the FVK number. Vliegenthart and Gompper
[63] simulated indentation of T-number lattice triangula-
Two-dimensional continuum modeling tions, which for sufficiently large T-numbers are consistent
with the Foppl-vonKarman continuum elastic description.
From the number of viral capsid structures determined Using a molecular dynamics approach to solve for equi-
experimentally by X-ray crystallography and cryo-EM, it librium shapes, they showed that the buckling instabilities
has been observed that icosahedral capsids, which come in of fivefold disclinations can be excited during indentation,
a variety of sizes, have what appears to be a size dependent leading to discontinuous drops in indentation force. Fur-
shape character. Smaller viruses tend to appear more thermore Vliegenthart and Gommper argue based on their
spherical and smooth, whereas larger viruses tend to show results that size effects may be important for smaller
faceting and visible ridges. Recent computational models capsids, causing their force-indentation character to differ
of icosahedral capsids have been created on the basis of the from that predicted by continuum mechanics. Using a
123
J Mater Sci (2007) 42:89959004 9001
thin-shell continuum finite element approach, Klug et al. In 2004, the first AFM study on a viral capsid along
[27] recently modeled AFM experiments on the empty with continuum model elasticity studies was reported by
CCMV viral capsid modeled as an icosahedron with Ivanovska et al. [25]. The virus studied was the bacterio-
varying values of the FVK number. The shells with FVK phage /29, which has icosahedral endcaps and a cylindrical
numbers below the buckling threshold (more round) show center section made from a ring of subunits arranged into
linear behavior consistent with pH 6 experiments, while hexameric groups. It was shown that the empty capsids had a
the shells with FVK numbers above the buckling threshold linear response up to displacements of ~30% of the capsid
(more faceted) show sharp discontinuities in the force due height, and was completely reversible unless the capsid was
to inversion of buckled fivefold disclinations, reminiscent displaced to the point of failure, at which drops in the contact
of the failure of capsids seen experimentally at pH 5. In force were seen. In order to extract material parameters for
particular it is argued that the pH indued swelling transition the capsid, a three-dimensional continuum elastic finite
of CCMV detailed earlier has the effect of softening the element model was created, on which the AFM experiment
mechanical response of the capsid by lowering the effec- was simulated. The Youngs modulus was varied until the
tive Youngs modulus of the capsid, and thus lowering the slopes of the experimental and simulated contact force
FVK number. This is proposed as a mechanism responsible curves matched. Similar studies were performed on the plant
for the presense and absense of failure in AFM indentation virus CCMV and subsequently modeled using three-
experiments at pH 5 and 6 respectively. dimensional continuum elastic finite elements by Michel
The results of the LMN study support the notion estab- et al. [26], and Gibbons and Klug [64]. It was shown by
lished by NMA studies [42] of a strong link between shape Gibbons and Klug in a series of parametric studies that the
and mechanical properties for macromolecular assemblies. observed linear behavior may be understood as a combina-
Indeed, whereas NMA studies have suggested that shape tion of several geometric effects; mainly that the thickness
governs mechanics, the LMN model represents an example of the capsid and size of the AFM tip (which is on the order
of the reverse, i.e., how mechanics can govern shape. The of the capsid size itself) largely determine the mechanical
results of Klug et al., and Vliegenthart and Gompper go response to applied loads [64]. Interestingly, the response
further by examining the link between shape and mechan- was not sensitive to the constitutive law chosen, as long as
ical response of icosahedral capsids under the influence of large strain measures were used. (The use of linearized small
externally applied mechanical forces. The one common strains was shown to be inadequate for such analyses.) From
feature of these two-dimensional elastic shell models is the these studies, it appears that the geometry of the capsid and
assumption that because of their geometry the coat proteins AFM tip, rather than local material response, have the most
most naturally fit together in a planar hexagonal lattice, impact on the resulting force curves.
which represents a stress-free reference configuration. Thus Three-dimensional continuum models were also used to
the presence of fivefold disclinations necessarily introduces model immature and mature murine leukemia virus (MLV)
pre-stresses in the closed capsid shell. This assumption re- capsids [29]. AFM studies on the immature and mature
sides at a midpoint within the multiscale hierarchy of capsid capsid showed quantitative and qualitative differences in
structure, at a length scale above that of primary and sec- the contact force response curves; the immature capsid is
ondary molecular structure, but below that of the global 20 nm thick, and showed a response similar to Hertzian
capsid assembly. It may perhaps be surprising then, that contact behavior, while the mature virus, which is only
other recent continuum models working purely at the global 4 nm thick, showed a linear contact force response. The
structural scale also seem to do well in explaining features nominal external radius in both cases is the same, at
of capsid mechanics observed in experiments. ~50 nm. These results agree very well with the contact
force curves of the AFM experiments.
Similar AFM studies were performed on the minute virus
Three-dimensional continuum modeling of mice (MVM) with and without the DNA genome enclosed
[28]. The studies showed that the capsid without the genome
As a key theoretical complement to the recent experimental showed force responses that were equal for all three sym-
advances in single molecule experiments such as atomic metry faces, but showed an anisotropic increase in stiffness
force microscopy (AFM), three-dimensional continuum when the DNA was enclosed. A corresponding continuum
modeling has emerged as a simple but powerful tool in model was created in which the capsid was modeled as an
understanding viral capsid mechanics. Because the problem icosahedral shell with the average thickness of the capsid,
has complex boundary conditions and geometry, analytical with the DNA modeled as circular disks that add thickness to
solutions are not feasible without generalizations that would the capsid at different points on a given face of the icosa-
result in a loss of precision, and so finite element methods hedron. They were able to show the anisotropic increase in
have been used to simulate the AFM experiments. stiffness with the capsid and DNA model.
123
9002 J Mater Sci (2007) 42:89959004
Continuum finite element models for both empty and full factor in determining the mechanical response to external
capsids have served as a sufficient means of capturing some forces, while the models are not highly dependent on the
of the most important mechanical behaviors observed constitutive modeling details. These results are comple-
experimentally. Although icosahedral viruses can have very mentary to the results of the molecular mechanical models,
complex topographies, these features do not appear to play a which have shown that the details of the atomic structure
main role in the observed elastic response of the capsids; and interatomic potentials are not essential to capture the
rather, the details will more likely play an important role in unforced vibrational behavior. In particular it is important
failure mechanisms, which may depend more heavily on the to recognize the consistency between the insensitivity to
specific regions that are weak, and are likely to be directly constitutive law for continuum models and the insensitivity
related to the strength of individual capsomercapsomer to interatomic potentials for the molecular models. This
interactions. The studies cited above have shown that con- provides some justification for the modeling assumption of
tinuum models that take into account the overall shape of constitutive homogeneity. It may seem incredible that
viral capsids do quite well at reproducing experimentally continuum elasticity theory would be meaningful for a
observed contact force curves. This serves as an inspiration nanometer-scale molecular structure such as a viral capsid.
for further continuum and coarse-grained approaches, Yet, the results of recent modeling studies establish that
specifically to investigate mechanisms of failure. this is precisely the case.
Yet, caution should be exercised in celebrating the
sucesses of continuum theory. The established continuum
Conclusions models of capsid mechanics are not without limitations, the
largest of which is the exclusion of dynamic behavior. In
Molecular dynamics and normal mode analysis have particular, thermal forces and fluctuations are not consid-
revealed important mechanical properties of viral capsids ered, and although these details may not be important to the
by studying the unforced vibrational dynamics. One of the quasi-static global mechanical response of the viral capsids,
main conclusions of molecular modeling is that the overall time sensitive behaviors are not captured. Dynamic behav-
shape of viral capsids, and not the atomic detail, governs iors are extremely important when studying assembly and
the unforced mechanical response. In particular, coarse- disassembly, and viscoelastic/rheological behavior. Fur-
grained NMA models are able to capture the dynamic thermore, in the quest for deeper understanding of the
behavior of viral capsids, such as conformational changes, material properties of capsid shells, some questions that
without the fine scale detail used in MD simulations. simple continuum elasticity models have difficulty answer-
Despite the success of these bottom-up methods, there are ing will need to be addressed. For instance, elasticity isby
limitations to both. Computational expense is a large definitionincapable of explaining the irreversible damage
problem that has only recently been addressed with coarse- observed in the experiments on CCMV [26] and /29 [29];
grained methods, the time scales on which both MD and what are the mechanisms that lead to irreversible failure at
NMA are valid are still quite small, and perhaps most one pH, and nigh-invulnerability at another? More generally,
importantly, the methods are limited to unforced dynamic how are the microscopic conformational changes of
problems. The latter limitation excludes the study of a individual proteins manifested in the macroscopic struc-
range of problems for which experimental data is now tural mechanics of a large protein aggregate? Perhaps con-
available, such as the newly developed AFM experiments tinuum methods can be stretched and extended via multiscale
that were detailed above. Additionally, the trend towards modeling to provide answers to these questions.
coarse-grained models has moved the models closer to As informative as bottom-up methods such as MD and
continuum descriptions, and while the ability to describe NMA are on short time scales and at atomic levels, top-
movements at atomistic detail is lost, the bulk dynamic down continuum models hold the most promise for
behavior is retained, and this suggests that continuum understanding viral capsid behavior under the variety of
models may provide insights into the forced mechanical applications they now serve in. Yet to overcome the limi-
behavior of viral capsids without a loss of physicality. tations of the current continuum methods, incorporating
It has been demonstrated that continuum models provide atomic level detail when needed may be necessary. Thus,
a wealth of information about the response of viral capsids the development of multiscale methods will most likely be
to applied loads. By modeling the global shape of viral needed to understand complex dynamic behavior of viral
capsids, without atomic level structural information, both capsids, such as conformational changes. NMA has been
thin and thick shell mechanics seem to consistently predict used to describe the pathway of such deformations, but
what is observed experimentally. Results of several con- perhaps continuum and multiscale methods will deepen the
tinuum finite element studies have demonstrated that the understanding about the molecular mechanisms behind
shell-like geometry of the capsid is the most important these behaviors. For example, continuum models have
123
J Mater Sci (2007) 42:89959004 9003
been used to study the conformational changes of the 22. Kindt J, Tzlil S, Ben-Shaul A, Gelbart WM (2001) PNAS
CCMV capsid, and have shown that relevant modes may be 98(24):13671
23. Tzlil S, Kindt J, Gelbart WM, Ben-Shaul A (2003) Biophys J
exictable when forces are applied, and this exictation will 84:1616
manifest as a change in material properties [27]. The exact 24. Purohit PK, Kondev J, Phillips R (2003) PNAS 100(6):3173
molecular mechanisms responsible are as of yet unknown. 25. Ivanovska IL, de Pablo PJ, Ibarra B, Sgalari G, MacKintosh FC,
Additionally, multiscale models will be needed to model Carrascosa JL, Schmidt CF, Wuite GJL (2004) Proc Natl Acad
Sci USA 101:6700
the forced mechanical response of viral capsids beyond the 26. Michel J-P, Ivanovska IL, Gibbons MM, Klug WS, Knobler CM,
elastic regime, to study the failure of viral capsids, as Schmidt CF, Wuite GJL (2006) Proc Natl Acad Sci USA
failure will be initiated at the atomic level due to the 103(16):6184
breaking of bonds. 27. Klug WS, Bruinsma RF, Michel J-P, Knobler CM, Ivanovska IL,
Schmidt CF, Wuite GJL (2006) Phys Rev Lett 97:228101
28. Carrasco C, Carreira A, Schaap IAT, Serena PA, Gomez-Herrero
Note: Just before this article went to print, an article J, Mateu MG, de Pablo PJ (2006). Proc Natl Acad Sci USA
appeared detailing AFM experiments on the bacteriophage 103(37):13706
k, studying the effects of packed genome on the strength of 29. Kol N, Gladnikoff M, Barlam D, Shneck RZ, Rein A, Rousso I
(2006) Biophys J 91(2):767
the capsid [66]. We would also like to report that this same 30. Schaap IAT, Carrasco C, de Pablo PJ, MacKintosh FC, Schmidt
group is preparing an article on the subject of viral capsid CF (2006) Biophys J 91(4):1521
failure. 31. Reddy VS, Giesing HA, Morton RT, Kumar A, Post CB, Brooks
CL, Johnson JE (1998) Biophys J 74(1):546
32. Zlotnick A, Johnson JM, Wingfield PW, Stahl SJ, Endres D
References (1999) Biochemistry 38:14644
33. Warwicker J (1992) J Mol Biol 223:247
1. Lee SW, Mao C, Flynn CE, Belcher AM (2002) Science 296:892 34. van Vlijmen HWT, Curry S, Schaefer M, Karplus M (1998) J Mol
2. Blum AS, Soto CD, Wilson JD, Cole M, Kim, Gnade B, Chatterji Biol 275:295
A, Ochoa WF, Lin T, Johnson JE, Ratna BR (2004) Nano Lett 35. Cagin T, Holder M, Pettitt BM (1991) J Comput Chem 12(5):627
4(5):867 36. Yoneda S, Kitazawa M, Umeyama H (1996) J Comput Chem
3. Strable E, Johnson JE, Finn G (2004) Nano Lett 4(8):1385 17(2):191
4. Douglas T, Young M (1998) Nature 393(6681):152 37. Phelps DK, Post CB (1995) J Mol Biol 254:544
5. Douglas T, Strable E, Willits D, Aitouchen A, Libera M, Young 38. Phelps DK, Rossky PJ, Post CB (1998) J Mol Biol 276:331
M (2002) Adv Mater 14:415 39. Lau WF, Pettitt BM, Lybrand TP (1988) Mol Simul 1:385
6. Bancroft JB, Hiebert E (1967) Virology 32:354 40. Freddolino PL, Arkhipov AS, Larson SB, McPherson A, Schulten
7. Speir JA, Munshi S, Wand GJ, Baker TS, Johnson JE (1995) K (2006) Structure 14:437
Structure 3(1):63 41. Cui Q, Bahar I (eds) (2005) Normal mode analysis: theory and
8. Morais MC, Choi KH, Koti JS, Chipman PR, Anderson DL, applications to biological and chemical systems. Chapman &
Rossmann MG (2005) Mol Cell 18:149 Hall/CRC, Boca Raton
9. Pettersen EF, Goddard TD, Huang CC, Couch GS, Greenblatt 42. Tama F, Brooks CL III (2006) Annu Rev Biophys Biomol Struct
DM, Meng EC, Ferrin TE (2004). J Comput Chem 25(13):1605 35:115
10. Berman HM, Henrick K, Nakamura H (2003) Nat Struct Biol 43. Tama F, Sanejouand Y-H (2001) Protein Eng 14:1
10(12):980, http://www.pdb.org/ 44. Ming D, Kong Y, Lambert MA, Huang Z, Ma J (2002) PNAS
11. Brooksbank C, Camon E, Harris MA, Magrane M, Martin MJ, 99(13):8620
Mulder N, ODonovan C, Parkinson H, Tuli MA, Apweiler R, 45. Tama F, Wriggers W, Brooks CL III (2002) J Mol Biol
Birney E, Brazma A, Henrick K, Lopez R, Stoesser G, Stoehr P, 321(2):297
Cameron G (2003) Nucl Acids Res 31:43, http://www.ebi.ac.uk. 46. Delarue M, Dumas P (2004) PNAS 101:6957
12. Caspar DL, Klug A (1962) Cold Spring Harb Symp Quant Biol 27: 1 47. Go N, Noguti T, Nishikawa T (1983) Proc Natl Acad Sci USA
13. Ganser-Pornillos BK, von Schwedler UK, Stray KM, Aiken C, 80:3696
Sundquist WI (2004) J Virol 78:2545 48. Brooks BR, Karplus M (1983) Proc Natl Acad Sci USA 80:6571
14. Branden C, Tooze J (1999) Introduction to protein structure. 49. Rader AJ, Vlad DH, Bahar I (2005) Structure 13:413
Garland Publications, New York 50. Tama F, Brooks CL III (2002) J Mol Biol 318:733
15. Wilk T, Gross I, Gowen BE, Rutten T, de Haas F, Welker R, 51. Tama F, Brooks CL III (2005) J Mol Biol 345(2):299
Krausslich H-G, Boulanger P, Fuller SD (2001) J Virol 75:759 52. Simonson T, Perahia D (1992) Biophys J 61:410
16. Schwarcz WD, Barroso SP, Gomes AM, Johnson JE, Schnee- 53. Brooks BR, Bruccoleri RE, Olafson BD, States DJ, Swaminathan
mann A, Oliveira AC, Silva JL (2004) Cell Mol Biol 50:419 S, Karplus M (1983) J Comput Chem 4(2):187
17. Wikoff WR, Liljas L, Duda RL, Tsuruta H, Hendrix RW, John- 54. van Vlijmen HWT, Karplus M (2005) J Mol Biol 350:528
son JE (2000) Science 289:2129 55. Tirion MM (1996) Phys Rev Lett 77(9):1905
18. Liu H, Qu C, Johnson JE, Case DA (2003) J Struct Biol 142:356 56. Durand P, Trinquier G, Sanejouand Y-H (1994) Biopolymers
19. Michels B, Leimkhler M, Lechner MD, Adrian M, Lorber B, 34:759
Witz J (1999) Eur J Biochem 264:965 57. Tama F, Gadea FX, Marques O, Sanejouand Y-H (2000) Pro-
20. Li PP, Naknanishi A, Tran MA, Ishizu K-I, Kawano M, Phillips teins: Struct, Funct, Genet 41:1
M, Handa H, Liddington RC, Kasamatsu H (2003) J Virol 58. Seung HS, Nelson DR (1988) Phys Rev A 38(2):1005
77:7527 59. Lidmar J, Mirny L, Nelson DR (2003) Phys Rev E 68:051910
21. Smith DE, Tans SJ, Smith SB, Grimes S, Anderson DL, Busta- 60. Nguyen TT, Bruinsma RF, Gelbart WM (2005) Phys Rev E
mante C (2001) Nature 413(6857):748 72:051923
123
9004 J Mater Sci (2007) 42:89959004
61. Nguyen TT, Bruinsma RF, Gelbart WM (2006) Phys Rev Lett 64. Gibbons MM, Klug WS (2007) Phys Rev E 75:031901.
96:078102 65. Arkhipov A, Freddolino PL, Schulten K (2006) Structure 14:1767
62. Zandi R, Reguera D (2005) Phys Rev E 72(2):021917 66. Ivanovska I, Wuite G, Jonsson B, Evilevitch A (2007) Proc Natl
63. Vliegenthart GA, Gompper G (2006) Biophys J 91(3):834 Acad Sci USA 104:9603
123
J Mater Sci (2007) 42:90059014
DOI 10.1007/s10853-007-1936-8
Received: 31 January 2007 / Accepted: 12 June 2007 / Published online: 7 August 2007
Springer Science+Business Media, LLC 2007
Abstract An optimised structure is one which uses the strength parallel to the tube axis differ from those in the
smallest quantity of the best material to perform its circumferential direction; if this difference is chosen
function, with adequate safety factor or margin for error. properly, the orthotropic tube is both stiffer and stronger in
Structural optimisation occurs not only in mechanical bending than the equivalent isotropic one. Orthotropic
engineering, but also in nature: plants with hollow stems or tubes are exploited both by engineers (composites, highly-
stalks gain a height advantage, and are thus more efficient, drawn metals) and by nature (stalks, stems, bamboo culms)
by approaching the optimum shape. Here we consider the and are almost invariably structured so that the stiff, strong
optimisation of orthotropic tubes, typifying, in a mechan- direction lies parallel to the axis of the tube. The question
ical sense, stalk and stem. The stiffness and strength of we address here is this: how can the tube shape (that is, the
orthotropic tubes of initially circular section are reviewed, ratio of wall-thickness to tube radius) and the anisotropy
and diagrams are proposed which allow the optimum ratio (ratio of axial to radial modulus and strength) be
section shape to be selected. optimised to maximise the performance?
To answer this requires a study of the potential failure
modes of the tube. When a tube is bent, its section tends to
Introduction ovalise, loosing stiffness. Bent far enough, it fails in mode
1: ovalising and kinking with catastrophic stiffness
In creating structures to carry loads, hollow, thin-walled losslike a plastic drinking straw, bent until it collapses.
sections can be more efficient than thick-walled or solid But bending also creates tensile and compressive stresses
ones. They use less material and are therefore lighter (more in the tube wall; if either of these exceeds the uniaxial
economical) while resisting the same bending or tor- strength of the tube wall, the tube fails in mode 2: tensile
sional load; and this is true whether the design is based on yield or fracture, or compressive collapselike a stick of
stiffness or on strength. When the mode of loading is celery, bent until it snaps. Finally, ovalisation has another,
bending and the direction of loading is unknown, circular subtler, consequence: it creates circumferential stresses in
tubes are better than other shapes. And if nature is to act as the tube wall which, if they exceed the circumferential
a guide, circular tubes which are orthotropic are better than strength (almost always the lower one), cause mode 3
those that are isotropic using the same amount of material. failure: longitudinal splittinglike a stick of celery,
An orthotropic tube is one in which the modulus and pinched between the fingers.
Observations and analyses of the bending response of
thin-walled orthotropic tubes appear in two quite separate
U. G. K. Wegst (&)
Department of Materials Science and Engineering, Drexel bodies of literature. That relating to plant stems and stalks
University, 3141 Chestnut Street, Philadelphia, PA 19104, USA is largely experimental [114]. That focusing on light-
e-mail: uwegst@coe.drexel.edu weight engineering structures, and particularly on poly-
mer-composite tubes are predominantly analytical, and
M. F. Ashby
Department of Engineering, Cambridge University, generally treat only one aspect of what is a multi-faceted
Trumpington Street, Cambridge CB2 1PZ, UK problem [1526]. Only one paper [27] attempts, as we do
123
9006 J Mater Sci (2007) 42:90059014
here, a comprehensive survey of competing failure modes mode, but we shall ignore this and identify the elastic
and explores how well plantsparticularly bambooare collapse moment with the Brazier moment.
structured to combat them. We now review the stiffness To find it, we follow the method and notation of [29].
and strength of thin-walled orthotropic tubes in more Ovalisation is measured by f, the cross-sectional shape-
detail, allowing for ovalisation during bending and for change parameter (Fig. 1):
four distinct failure modes. Much of the analysis closely
ra
parallels that of [27] and will therefore be kept brief. The 1 5
r
recasting of all of the results in terms of section area and
shape, and the novel construction and optimisation, both where r and a are the radius of the original circular shape
of shape and of anisotropy this allows, are new. and the minor axis of the ellipse respectively.
Then the total strain energy per unit length, U, of an
Bending of thin-walled orthotropic tubes isotropic tube which has been deformed into an arc of
curvature C, replacing Eq. 4, becomes
In designing a thin-walled tube to carry bending moments,
1 3 2 3 5 2 3 pt3 2
two sets of consideration arise. The first set relates to U pr tEC 1 1 1 1E 6
2 2 8 8 r
stiffness, the second to the collapse moment (or strength).
The first term of the right hand side of this equation
Stiffness, ovalisation and the Brazier moment for describes the strain energy due to longitudinal stretching,
orthotropic tubes the second that due to ovalisation (circumferential bending).
For an orthotropic material this expression is modified
The curvature, C, of an elastically isotropic beam is related by replacing E with the longitudinal modulus, Ejj , in
to the bending moment, M, which it carries by the longitudinal stretching term and with the transverse
modulus, E? , in the ovalisation term.
M IEC 1
where E is the Youngs modulus of the material of which 1 3 2 3 5 2 3 pt3 2
U pr tEjj C 1 1 1 1 E? 7
the beam is made and I is its second moment of area. We 2 2 8 8 r
focus on a beam which is a thin-walled tube of initial Omitting the f2 term enclosed in square brackets (justified
radius, r, and wall-thickness, t, with cross-section A 2prt. in a moment), we seek the ovalisation, f, for a given
If the bending is slight, the cross-section remains circular, curvature, C, which minimises the total energy, U. Setting
when dU=df 0 leads to
I pr 3 t 2
r 4 Ejj 4
1 2 C 2 c2 8
Then the moment is related to the curvature by t E? 3
where the dimensionless curvature, c, is defined by
M pr 3 tEC 3
and the elastic strain energy per unit length of the tube is 3 r 4 Ejj
c2 C2 9
4 t 2 E?
1
U pr 3 tEC 2 4 Substituting Eqs. 8 and 9 (inverted) for f and C2 in Eq. 7,
2 gives the strain energy, U, as a function of c alone
However, if the bending is substantial, the tensile and
compressive stresses in the tube walls due to its longitu- 2p t3 2
U E? c c 4 10
dinal curvature cause the cross-section to ovalise as in 3 r
Fig. 1, and this has several consequences. The ovalisation From this we calculate the bending moment, M Cc dU
dc , as
reduces the second moment of area, I; further bending
causes the ovalisation to increase and the stiffness to 2p 1=2
decrease, until a maximum bending moment, the Brazier M p rt2 Ejj E? c 2c3 11
3
moment, is reached and catastrophic failure follows [28].
In practice tubes fail at bending moments which are a little Equation 11 is the generalisation of the simple Eq. 3 for
less than this because local defects trigger a local buckling orthotropic tubes, including the effect of ovalisation, and
123
J Mater Sci (2007) 42:90059014 9007
123
9008 J Mater Sci (2007) 42:90059014
M2 pr 2 trcjj 20
123
J Mater Sci (2007) 42:90059014 9009
123
9010 J Mater Sci (2007) 42:90059014
Optimisation of shape
/A /B 35
Fig. 4 (a) A failure map of loading coefficient M/A3/2 plotted against
the shape factor / for the three failure curves M1 (fracture/yield), M2 giving (by Eqs. 32 and 33) the condition for the transition
(buckling) and M3 (splitting). The intersections at I, II and III indicate
the transitions from one failure mechanism to another. (b) A graph to no-splitting as:
showing bending moment, M, plotted against dimensionless curva-
ture, c. (c) A failure map showing the three failure domains fracture/ 1=2
yield (1), buckling (2) and splitting (3)
rt? E?
0:38 36
rcjj Ejj
Domain (2), splitting determines the failure moment (since
the lowermost curve is that of M3), but this is not so. The When this condition is met, the shape which gives the
peak moment in Domain (3) is set by the Brazier moment, not largest value of M/A3/2 is that corresponding to /B.
123
J Mater Sci (2007) 42:90059014 9011
In the practice of engineering design, gradual or be- anisotropy is possible by adjusting the lay-up. It is then
nign failure modes, such as yielding, are viewed with less reasonable to ask: what is the best value for the
alarm than those which are sudden or catastrophic, anisotropy? To answer this, consider an idealised
among which are buckling and splitting. Responding to example: that of a tube made of fibres of modulus Ef
this, the designer applies a greater safety factor to cata- and strength rf, in a matrix of modulus Em and strength
strophic modes than to those which are benign, and there is rm. Let a fraction c of the total volume fraction of fibres,
some evidence (see below) that nature may do the same. f, lie parallel to the tube axis; the remaining fraction
The practical consequence of this is to shift the optimum (1 c) is oriented circumferentially. Then, to an adequate
shape a little to the left on Fig. 4c, that is, to slightly approximation for illustrative purposes, the moduli of the
smaller values of /. material are:
Optimisation of anisotropy
Ejj 1 f Em c f Ef and
39
The last section identified two expressions for the optimum E? 1 f Em 1 c f Ef
shape, /, for orthotropic tubes, expressed in terms of
material properties. By substituting these back in the fail- Defining E as the value of E for which c = 1/2 (when the
ure equations, the failure moment of a tube of optimum material is quasi-isotropic)
shape can be written as a function of material properties
alone. If splitting is possible, (that is, Eq. 36 is not satis- 1
E 1 f Em f Ef 40
fied), the ideal shape is that corresponding to /A. Substi- 2
tuting this into Eq. 30 gives the optimum moment that can
be achieved with a given set of material properties: we find
Mopt 1=2 1 Ef
0:18 rt? Ejj Ejj E 1 c f and
A 3=2 2 E
! 41
rt? Ejj 37 1 Ef
with /opt 0:81 E? E 1 c f
r2cjj 2 E
with similar expressions for the strengths in terms of the
If, on the other hand, splitting is suppressed (that is, Eq. 36
strength r at u = 1/2:
is satisfied), the ideal shape is that corresponding to /B.
Substituting this into Eq. 30 gives
1 rf
rjj r
1 c f and
Mopt 1=2 1=2 2 r
0:11 rcjj E jj E ? 42
A3=2 1 rf
1=2 38 r? r 1 c f
Ejj E? 2 r
with /opt 0:31
rcjj
We take, as an example, the case in which the optimum
Many orthotropic materials are fibre-reinforced shape is limited by /B, when Eq. 38 applies. Substituting
composites, and for these a degree of control of the for Ejj ; E? , and rcjj gives:
123
9012 J Mater Sci (2007) 42:90059014
Mopt 1=2
E
0:11r
A3=2 ( 1=2
1
1 c b
2
1=4 )
1 1
1 c a 1 c a
2 2
43
E r
with a f Ef and b f rf .
The limits for a are determined for the extreme cases
that Em Ef , when
Ef f Ef f Ef
af 1
1 2 44
E 1 f Em 2 f Ef 2 f Ef
Fig. 5 The graph shows the expression in curly brackets ({}) of
and that Em Ef =2 , when Eq. 43 plotted against c for a number of different values for a and b.
Note that a and b will vary independently but in parallel. In the
Ef f Ef extreme case that the material is truly homogeneous (a = b = 0), the
af
2f 45 failure moment is, correctly, independent of the tubes shape (black
E 1=2 1 f Ef f Ef line)
123
J Mater Sci (2007) 42:90059014 9013
1 p
Fig. 6 (a) and (b) Failure maps plotting the loading coefficient C A2
M/A3/2 against shape factor / for typical green (a) and dry (b) R 21
bamboo culms. The green, blue and red sets of lines indicate low, Substituting this expression for C in Eq. 8 and inserting the
medium and high property values as listed in Table 1. The boxes
circumscribe the region in which bamboo typically lies according result for c in Eq. 13b gives the bending moment, M, which
to its shape and the stiffness criterion. The numbers (1), (2) and (3) bends the stem into a quarter-circle
label the three failure domains fracture/yield, splitting and
p
3p2 r 2 2 Ejj
buckling, respectively. (c) A failure map comparing green (green
lines) and dry (red lines) bamboo culms of medium material p r 3=2 1=2
M p A / Ejj 1 / A3
properties 4 2 l 8 l E?
123
9014 J Mater Sci (2007) 42:90059014
The second expression in the square bracket is very small 19. Harursampath D, Hodges DH (1999) Int J Non-Linear Mech
due to the high slenderness ratios of bamboo (/ = l/r = 34:1003
20. Kedward KT (1978) In: Proceedings of the 2nd international
5501000, [43]) and may therefore be neglected. The conference on composite materials (ICCM/2). Met Soc of AIME
moment which bends a stem to a quarter-circle may there- p 353
fore be rewritten as 21. Libai A, Bert CW (1994a) Int J Sol Struct 31:1003
22. Libai A, Bert CW (1994b) Int J Sol Struct 31:1019
p 23. NASA (1968) Space vehicles design criteria (structures)buck-
p r 3=2 1=2
M4 p A / Ejj A4 ling of thin-walled circular cylinders. Technical Report NASA
4 2 l SP-8007, NASA
24. Tatting BF, Gurdal Z, Vasiliev VV (1996) AIAA J 34:1934
If the plant is to function as a flexibility strategist, it must 25. Tatting BF, Gurdal Z, Vasiliev VV (1997) Int J Solids Struct
be able to do this without failing by any of the other three 34:1419
26. Tsai SW, Wu EM (1971) J Comp Mater 5:58
mechanisms analysed above. 27. Schulgasser K, Witztum A (1992) J Theor Biol 155:497
28. Brazier LG (1927) Proc Roy Soc Lond A 116:104
29. Calladine CR (1983) Theory of shell structures. Cambridge
References University Press, Cambridge
30. Ghavami K (1990) In: Ramanuja Rao I, Gnanaharan R, Sastry C
(eds) Proceedings of the international bamboo workshop held in
1. Amada S, Munekata T, Nagase Y, Ichikawa Y, Kirigai A, Yang
Cochin, India, 1418 November. The Kerala Forest Research
ZF (1996) J Comp Mat 30:800
Institute, India and International Development Research Centre,
2. Arce-Villalobos OA (1993) Fundamentals of the design of
bamboo structures. PhD Thesis, Eindhoven University of Tech- Canada, p 235
31. Li S, Fu S, Zeng Q, Zhao X, Zhou B (1994) Biomimetics 2:15
nology, Netherlands
32. Oda J (1980) Trans Japan Soc Mech Eng Ser A 46:997
3. Niklas KJ (1989) Am J Bot 76:521
33. Ueda K (1980) Res Bull Coll Exp For Hokkaido University
4. Niklas KJ (1992) Plant biomechanics: an engineering approach to
plant form and function. University of Chicago Press, Chicago 37:817
34. Akselrad EL (1965) Izv Akad Nauk USSR, Otdelenie Tek-
5. Niklas KJ (1995) Ann Bot 75:133
nicheskikh Nauk Mech No 4:123
6. Niklas KJ (1997a) Ann Bot 80:275
35. Seide P, Weingarten V (1961) J Appl Mech 28:112
7. Niklas KJ (1997b) Ann Bot 80:437
36. Suo Z (1990a) J Appl Mech 57:627
8. Niklas KJ (1998) Ann Bot 81:11
37. Suo Z (1990b) Proc R Soc Lond A 427:331
9. Spatz HC, Speck T, Vogellehner D (1990) Bot Acta 103:123
38. Wegst UGK, Ashby MF (2004) Phil Mag 84:2167
10. Spatz HC, Boomgaarden C, Speck T (1993) Bot Acta 106:254
11. Spatz HC, Speck T (1994) Biomimetics 2:149 39. Karam GN, Gibson JL (1995a) Int J Sol Struct 32:1259
40. Karam GN, Gibson JL (1995b) Int J Sol Struct 32:1285
12. Spatz HC, Kohler L, Speck T (1998) Am J Bot 85:305
41. Young WC (1989) Roarks formulas for stress and strain, 6th edn.
13. Speck T, Speck O, Emanns A, Spatz HC (1998) Bot Acta
McGraw-Hill, London
111:366
14. Schulgasser K, Witztum A (1997) Ann Bot 80:35 42. Mosbrugger V (1990) The tree habit in land plants. Lecture notes
in Earth Sciences: 28. Springer, Berlin
15. Cecchini LS, Weaver PM (2002) AIAA J 40:2136
43. Liese W (1985) Bamboos: biology, silvics, properties, utilization
16. Corona E, Rodrigues A (1995) Comp Eng 5:163
(Schriftenreihe der GTZ, Nr. 180, Deutsche Gesellschaft fur
17. Fabian O (1977) Int J Sol Struct 13:1257
18. Gerard G (1968) Minimum weight design of compressive struc- Technische Zusammenarbeit (GTZ) GmbH, Dag-Hammerskjold-
Weg 1+2, D-65760 Eschborn, Germany)
tures. New York University Press/Interscience, New York
123
J Mater Sci (2007) 42:90159020
DOI 10.1007/s10853-007-1834-0
Received: 18 September 2006 / Accepted: 8 May 2007 / Published online: 20 July 2007
Springer Science+Business Media, LLC 2007
Abstract A novel method to prepare iron oxide nano- treatment of solid tumors [14]. For these applications
particles by precipitation in bicontinuous microemulsions the magnetic particles should be coated with such
is reported. Precipitation reactions were carried out in materials as natural or synthetic polymers to allow for
microemulsions stabilized with a mixture of dodecyltrim- the attachment of drugs and other molecules capable of
ethylammonium bromide/didodecyldimethylammonium specific recognition of, and binding to, malignant cells.
bromide (3/2, w/w). Nanoparticles were characterized by Although many synthesis methods are available for
X-ray diffraction, transmission electronic microscopy preparing superparamagnetic iron oxide particles [3],
(TEM) and vibrating sample magnetometry (VSM), and many papers have demonstrated that precipitation in re-
consisted of magnetite or a mixture of magnetite and verse microemulsions is the most adequate method for
maghemite. The particles have an average diameter of preparing magnetic particles with diameters smaller than
8 nm with a relatively narrow particle size distribution and 10 nm and narrow particle size distribution in a direct
show possible superparamagnetic behavior. Noticeably way [514]. In the other known methods the obtained
higher yields of precipitate are observed for this new particles must be subjected to a size-selection process in
approach in comparison with those typicall obtained using order to narrow the particle size distribution [15, 16].
reverse microemulsions. Within the reports on precipitation in reverse micro-
emulsions are noticeable the works of: Gobe et al. [5],
Bandow et al. [6], Dresco et al. [10], Santra et al. [12]
and, more recently, Lee et al. [14], who prepared mag-
netic particles with mean diameters of 3, 3.6, 7.3, 1 and
Introduction 3 nm, respectively, and with relatively narrow particle
size distribution. In spite of all this work, which deals
Nanoscale superparamagnetic iron oxide particles, such mainly with the synthesis and characterization of the
as magnetite (Fe3O4) and maghemite (c-Fe2O3), offer particles, little attention has been paid to the mechanistic
very attractive possibilities in biomedical applications of such systems [810, 12]. However, it is now recog-
because of their magnetic properties and size specifically nized that the nucleation and growth of the nanoparticles
in chemotherapy, radiotherapy and hyperthermia for in these systems occur inside of the swollen micelles (or
droplets) of the reverse microemulsions, which usually
have diameters smaller than 10 nm. In this way, swollen
J. Esquivel I. A. Facundo M. E. Trevino
micelles function as nanoreactors for the preparation of
R. G. Lopez (&)
Centro de Investigacion en Qumica Aplicada, Boul. Ing. the particles [17].
Enrique Reyna No. 140, Saltillo, Coahuila 25253, Mexico Despite the size control and distribution of the magnetic
e-mail: glopez@ciqa.mx particles, the precipitation method using the reverse
microemulsions has the drawback of a relatively low
J. Esquivel
Facultad de Ciencias Qumicas, Universidad Autonoma de yield. Calculations based on the reported recipes for pre-
Coahuila, Saltillo, Coahuila 25240, Mexico paring magnetic nanoparticles using this technique, show
123
9016 J Mater Sci (2007) 42:90159020
yields from 0.1 g [7] to 0.4 g [5] of product per 100 g of Phase diagram determinations
total mixture. Surprisingly, and as far as we know, the
precipitation in bicontinuous microemulsions has not been Microemulsion regions at 80 C were determined by titra-
considered as an option that could increase this yield. In tion of solutions of the DTAB/DDAB mixture (3/2, w/w) in
contrast to reverse microemulsions, which consist of MMA with 0.75 M aqueous solution of a FeCl36H2O/
aqueous droplets dispersed in an oleic continuous phase, FeCl24H2O mixture (3/2, mol/mol). Phase boundary was
bicontinuous microemulsions are formed from intercon- detected visually at each one of the constant (DTAB/
nected aqueous channels with diameters usually smaller DDAB)/MMA lines studied. Then, samples with composi-
than 10 nm immersed in an oleic continuous phase [18]. tions slightly below and above that of the visually deter-
This structure allows bicontinuous microemulsions to mined phase boundary were prepared by weighting each
accept contents of the aqueous phase usually between 50 component and allowing to reach equilibrium in a water
and 100% higher than those accepted by reverse micro- bath at 80 C to determine more precisely the phase
emulsions [18, 19]. We speculate that the reason why boundary.
bicontinuous microemulsions have not been considered
earlier as media for the precipitation of magnetic particles Synthesis of magnetic particles
is because of their microstructure. On the first hand one
would not expect to obtain spherical or equiaxial particles Precipitation reactions were carried out at 80 C in a
from the cylindrical channels of a bicontinuous micro- 100 mL jacketed glass reactor equipped with a reflux
emulsion in contrast to what can be obtained or expected condenser and inlets for argon, mechanical agitation and
from the spherical droplets of reverse microemulsions. feed of the microemulsion and ammonium hydroxide
Nowadays, it is accepted that the mechanism through aqueous solution. To help stabilize the system during the
which particles are obtained in reverse microemulsions reaction, the reactor was immersed in an ultrasonic bath.
includes nucleation in those droplets that contain a mini- Two precipitation reactions in bicontinuous microemul-
mum number of precipitate molecules and further growth sions were carried out along with one precipitation reac-
through recruiting precursor ions from those droplets tion in a reverse microemulsion with the same surfactant
containing no particles inside either by diffusion or droplet content (33 wt.%) for comparison. When the reaction
collision [17]. Although particle growth in bicontinuous media was the reverse microemulsion a stirrer speed of
microemulsions cannot proceed by droplet collision, there 300 rpm was used; mechanical agitation was not provided
is no obvious reason why nucleation inside the channels when the reactions were carried out in the bicontinuous
and further growth of nuclei through recruiting precursors microemulsions. Reverse microemulsion was prepared
ions from surroundings could not occur. with 57 wt.% MMA and 10 wt.% aqueous solution of a
In this paper, we report a new method that uses the mixture of ferric and ferrous chlorides in a molar ratio of
channels of bicontinuous microemulsions as media to 3/2 with an overall concentration of 0.75 M. The bicon-
prepare nanoscale magnetic particles. The results of the tinuous microemulsions were composed by 30 and
characterization of these particles by X-ray diffraction, 40 wt.% of the same aqueous solution of the mixture of
transmission electron microscopy and vibrating sample ferric and ferrous chlorides and 37 and 27 wt.% MMA,
magnetometry were compared with those of the particles respectively. In all cases, an excess of ammonium
prepared in a reverse microemulsion. hydroxide (twice the stoichiometric requirements) was
used. The typical procedure for the precipitation reaction
started with charging the previously deoxygenated mi-
Experimental croemulsion in the reactor and then raising the tempera-
ture to 80 C. To initiate the reaction aqueous solution of
Materials ammonium hydroxide was added to the reactor. The
reacting system was continuously purged with argon
All reagents were high-purity grades from Aldrich used during the entire process. The precipitation reaction was
without further purification: ferric chloride (FeCl34H2O, allowed to proceed for 30 min whereupon a mixture of
99%), ferrous chloride (FeCl26H2O, 98%), methyl meth- acetone and water was added to cause the final precipitate
acrylate (MMA, 99%), aqueous ammonia (NH4OH, recovered by magnetic separation. Noticeably the super-
57.6 wt.%), dodecyltrimethylammonium bromide (DTAB, natant contained no visible particles in suspension. After
99%) and didodecyldimethylammonium bromide (DDAB, several washes with acetone, the precipitate was dried
98%). Water was triple-distilled deionized grade. yielding a black powder.
123
J Mater Sci (2007) 42:90159020 9017
Characterization
123
9018 J Mater Sci (2007) 42:90159020
particles. It should be noted that we found it difficult to (8 nm) with relatively narrow particle size distributions.
compare our results with those reported in other studies on The results demonstrate that nucleation and growth of
precipitation of magnetic particles in reverse microemul- nanoparticles can be carried out in the channels of bicon-
sions. The reason is that the authors do not report the yields tinuous microemulsions. This observation opens the pos-
obtained in their studies but only a few of them provide the sibility that the channel diameters may play a role in
sufficient data for calculating the theoretical yields. Using determining the final particle size.
the data reported was possible to estimate the yields in X-ray diffraction patterns of the samples are shown in
those studies carried out by Gobe et al. [5], Lee et al. [7] Fig. 4. Peak broadening observed is consistent with the
and Selim [20], as 0.4, 0.1 and 0.2 g of magnetic particles small particle size [12, 21] with patterns matching those of
per 100 g of total mixture, respectively. In comparison, the
yields obtained in our study resulted in similar values for
the reverse microemulsion; however, those obtained from
precipitation in bicontinuous microemulsion were higher.
Low and high magnification TEM images are shown in
Fig. 3 along with histograms. Table 1 shows the number-
average diameter (Dn) and the polydispersity index
(Dw/Dn) calculated from the TEM photomicrographs.
Data in Table 1 indicate that the precipitation reaction
carried out in the reverse microemulsion produces small
diameter particles (5.6 nm) with a relatively narrow parti-
cle size distribution. For particles prepared in the bicon-
tinuous microemulsions, small diameters are also observed
123
J Mater Sci (2007) 42:90159020 9019
Conclusions
123
9020 J Mater Sci (2007) 42:90159020
13. Liu ZL, Wang X, Yao KL, Du GH, Lu QH, Ding ZH, Tao J, Ning 19. Salager JL, Anton RE (1999) Handbook of microemulsion sci-
Q, Luo XP, Tian DY, Xi D (2004) J Mater Sci 39:2633 ence and technology, chap. 8. Marcel Dekker, Inc., New York
14. Lee Y, Lee J, Bae CJ, Park JG, Noh HJ, Park JH, Hyeon T (2005) 20. Selim S (1997) U.S. Patent 5695901
Adv Funct Mater 15:503 21. Lopez Perez JA, Lopez Quintela MA, Mira J, Rivas J, Charles
15. Lopez-Lopez MT, Duran JDG, Delgado AV, Gonzalez-Caballero SW (1997) J Phys Chem B 101:8045
F (2005) J Colloid Interf Sci 291:144 22. Lopez-Perez JA, Lopez-Quintela MA, Mira J, Rivas J (1997)
16. Mikhaylova M, Kim DK, Bobrysheva N, Osmolowsky M, IEEE Trans Magn 33:4359
Semenov V, Tsakalakos T, Muhammed M (2004) Langmuir 23. Kodama RH (1999) J Magn Magn Mater 200:359
20:2472 24. Cornell RM, Schwertmann U (1996) The iron oxides. structure,
17. Osseo-Asare K (1999) Handbook of microemulsion science and properties, reactions, occurrence and uses. VCH, Weinheim,
technology, chap. 18. Marcel Dekker, Inc., New York Germany
18. Ezrahi S, Aserin A, Garti N (1999) Handbook of microemulsion 25. Lee Y, Lee J, Bae CJ, Park JG, Noh HJ, Park JH, Hyeon T (2005)
science and technology, chap. 7. Marcel Dekker, Inc., New York Adv Funct Mater 15:2036
123
J Mater Sci (2007) 42:90219029
DOI 10.1007/s10853-007-1881-6
Received: 7 March 2006 / Accepted: 25 May 2007 / Published online: 20 July 2007
Springer Science+Business Media, LLC 2007
Abstract Wrought material as well as physically simu- austenite microstructure [16]. In particular, in the high
lated welding heat affected zone (HAZ) samples of an UNS temperature heat affected zone (HAZ), a coarse micro-
S32304 duplex stainless steel were subjected to electro- structure is formed by a high fraction of ferrite, allotri-
chemical corrosion tests and electron microscopy charac- morphic and Widmanstatten austenite, and extensive
terization. An impaired corrosion resistance of the HAZ chromium nitride precipitation within ferritic grains [16].
microstructures compared to the wrought material micro- When this as-welded microstructure is reheated during
structure was observed. Calphad-based numerical simula- multipass welding or during post-weld heat treatment, new
tion of phase transformations and solute redistribution austenite, called secondary austenite (c2), is formed. This
taking place during welding provided an explanation of the c2 may grow from preexisting austenite particles (so called
observed corrosion behavior. The poor corrosion resistance primary austenite, c1) and is called intergranular c2.
of the HAZ microstructures studied was mainly attributed Another form of c2 can be formed by the nucleation and
to a decrease in corrosion resistance of ferritic grains after growth of new intragranular c2 particles, mostly within the
welding, which exhibited lower chromium content than nitride colonies. Nitride distribution will also be modified,
ferritic grains in the wrought material. with some intragranular nitrides dissolving and/or new
nitride precipitation occurring mainly along ferrite/
austenite interfaces [27].
It has been frequently reported that the DSS HAZ
microstructure exhibits impaired corrosion resistance when
Introduction
compared to the wrought material [14]. On the other hand,
there is no consensus on the origin of this reduction in
The welding of duplex stainless steels (DSSs) strongly
corrosion resistance or on the role of c2 precipitation.
affects their fine-grained and duplex-balanced ferrite plus
The detrimental effect of intragranular c2 precipitation
on the localized corrosion resistance of DSSs welded joints
C. M. Garzon A. J. Ramirez (&) has been reported, and this reduced corrosion resistance has
Brazilian Synchrotron Light Laboratory, Caixa Postal
been associated with the lower N, Cr, and Mo content of c2
6192 - CEP 13083-970 Campinas, SP, Brazil
e-mail: ramirez@lnls.br when compared to c1 [2, 3, 8, 9]. It has also been suggested
that the poor corrosion resistance of the HAZ microstruc-
Present Address: tures is mainly associated with chromium depletion around
C. M. Garzon
chromium nitrides within ferritic grains [1].
Physics Department, National University of Colombia,
Avenida Cra. 30 No 45-03, Ciudad Universitaria, Edificio 404, The objective of this investigation was to study the role of
Bogota, Colombia alloy element distribution among ferrite, c1, and intergran-
ular c2 on corrosion resistance of UNS S32304 DSS HAZ
C. A. Serna S. D. Brandi
microstructures. Thus, both electron microscopy character-
Metallurgical and Materials Engineering Department,
University of Sao Paulo, Av. Prof. Mello Moraes 2463, ization and electrochemical corrosion tests were performed
CEP 05508-900 Sao Paulo, SP, Brazil on simulated HAZ microstructures. The as-welded HAZ and
123
9022 J Mater Sci (2007) 42:90219029
reheated-HAZ microstructures were obtained in a Gleeble The pitting corrosion resistance was evaluated by cyclic
thermo-mechanical simulator. In addition, the analysis of the potentiodynamic polarization experiments carried out at
experimental results was supported by thermodynamic and 25 1 C in a 3.5 wt-% NaCl naturally aerated solution
kinetic simulation of the phase transformations taking place using a potential scan rate of 1 mVs1.
during the thermal cycles. The degree of sensitization (DOS) was evaluated
through double loop electrochemical potentiokinetic
reactivation (DL-EPR) tests. The DOS was measured as
Experimental the Ir/Ia ratio, where Ir is the maximum value of the
anodic current in the reactivation peak and Ia is the
A commercially available UNS S32304 duplex stainless maximum value of the anodic current in the activation
steel was subjected to physical simulation to obtain peak [10]. The DL-EPR tests were performed in a 0.5 M
samples with both as-welded HAZ microstructures H2SO4 + 0.01 M KSCN naturally aerated solution at
(ferritization treatments) and reheated (multipass) HAZ 30 1 C, using a potential scan rate of 1 mVs1. The
microstructures (reheating treatments). The chemical scanned potential range was from the corrosion potential
composition of the DSS is shown in Table 1. The ferriti- up to 500 mV-SCE.
zation treatments consisted of an isothermal anneal at The cyclic polarization and DL-EPR tests were
1,350 C for 5 s followed by rapid cooling; the average performed using a Princeton Applied Research Potentio-
heating rate was 300 Cs1 and the cooling rate at 800 C statGalvanostat 273A. Saturated calomel and platinum
was 46 Cs1. The reheating treatments were applied to electrodes were used as reference and counter electrodes,
prior ferritized samples, and consisted of isothermal heat respectively. Prior to the corrosion tests, the samples were
treatments at 900 or 1,200 C for 10 s. The heating and mechanically ground using abrasive paper to 600 mesh
cooling conditions for the reheating and ferritization final finish. Five tests were performed for each studied
treatments were similar. The physical simulations were condition.
performed in a thermo-mechanical simulator Gleeble
1500, using prismatic samples of 6 6 90 mm. Detailed
information about these treatments was presented Numerical simulation
elsewhere [6, 7].
Microstructural characterization was performed using The thermodynamics and kinetics assessment of both
optical microscopy (OM), scanning electron microscopy austenite growth and chromium nitride dissolution and/or
(SEM) and transmission electron microscopy (TEM). precipitation in the HAZ microstructures was performed
Detailed microstructural characterization of the simulated using the Calphad-based Thermocalc and Dictra soft-
samples is presented elsewhere [6, 7]. Metallographic ware [11] along with the Fe-data and Mob2 databases. A
preparation of samples for OM and SEM consisted of detailed analysis of this numerical modeling is presented
abrasive paper grinding followed by diamond paste pol- elsewhere [12] and only the most important features are
ishing and a final stage of 0.05 lm colloidal silica pol- presented in this work.
ishing. The specimens were electrolytically etched using a The diffusion-controlled growth of a c (austenite) region
40%-vol HNO3 aqueous solution at 25 C, and applying into a a + M2N (ferrite) region (where M2N is a nitride
1.1 V during ~70 s. The SEM analyses were performed with two metallic (M) atoms per each nitrogen (N) atom)
using a JEOL JSM 5900 LV microscope as well as a was simulated. A closed one-dimensional multi-component
CAMBRIDGE Steroscan-440 microscope coupled with a system and a planar sharp c/a interface were assumed.
wavelength-dispersive X-ray spectrometer (XWDS). TEM The atomic diffusion in both ferritic matrix and
samples were prepared by jet polishing with a 70%-vol austenitic grains was computed by solving the multicom-
ethyl alcohol, 20%-vol glycerin and 10%-vol perchloric ponent non-steady state atomic diffusion equation 1 [13].
acid (HClO4) solution at 20 C and 30 V. The TEM The interface velocity was calculated using the mobile
analyses were performed using a JEOL JEM 3010 URP interface model [11], which is implemented in Dictra
microscope coupled with an energy dispersive X-ray software. The establishment of the local equilibrium (LE)
spectrometer (XEDS) system. condition at the migrating c/a interface was assumed.
Table 1 Chemical composition of the UNS S32304 duplex stainless steel (wt-%)
Fe Cr Ni Mn Si Mo Cu C N
123
J Mater Sci (2007) 42:90219029 9023
X n
0 ol Table 2 Chemical analysis (XWDS) and numerical calculations
JK Lki i 1 (NC) of the main elements in the ferritic and austenitic grains of a
i1
oz ferritized sample (wt-%)
0 Phase Fe Cr Ni C N
In Eq. 1, JK is the flux of the atomic species k; Lki is a
proportionality factor that depends on the mobility of the a XWDS Bal. 24.0 4.6 N.M N.M
atomic species; li is the chemical potential for the species NC Bal. 22.7 4.8 0.009 0.06
in the system; and z is the spatial coordinate. c1 XWDS Bal. 23.3 4.9 N.M 0.48
Nitrides can precipitate or dissolve within the ferritic NC Bal. 22.4 4.8 0.039 0.31
matrix as function of temperature and local alloying con-
N.M: Non measurable value (content was lower than the detection
tent. This aspect was taken into consideration by using the limit)
diffusion in dispersed systems model [11], which is also
implemented in Dictra software.
To compute the growth of austenitic grains, the so-
must be noted that large experimental errors are expected
called primary austenite (c1), during ferritization treat-
with XWDS measurements of light element such as
ments, it was assumed the prior existence of 1 nm thick
nitrogen.
gamma nucleus that begins to grow during the cooling
Table 3 presents the Fe, Cr, and Ni weight fraction of
stage at 1,200, 1,100, 1,000, or 900 C. The chemical
nitrides and the ferritic matrix after the ferritization treat-
composition of this gamma nucleus was set at the equi-
ments as measured by TEM-XEDS and as determined via
librium value corresponding to the temperature at which
computational modeling. Because the chemical composi-
this nucleus begins to grow. However, to compute the
tion of the ferrite measured via XWDS-SEM was used as a
growth of secondary austenite (c2) during reheating treat-
standard for the XEDS measurement in the TEM, both
ments, it was assumed that c2 grows over the previously
compositions are forced to be identical. It is not possible to
precipitated c1 grains. The average c1 width was deter-
measure the C and N contents using XEDS in the TEM.
mined by metallographic measurements and its average
Therefore, only the composition of the primary solute
chemical composition taken from the previously described
elements (metal) is reported. The XEDS analyses of the
precipitation simulations. The size and chemical compo-
nitrides confirmed their expected high Cr content. The
sition of the remnant ferritic regions were established
experimental and calculated values presented in this table
through mass balance computation by calculating the
show that the average chromium depletion induced in the
amount of a + M2N necessary to reach the steel nominal
matrix during the ferritization treatment and as a conse-
composition.
quence of nitride precipitation is low; the average chro-
mium content of the as-welded a + M2N (ferritic) regions
is ~22.7 wt-% and the computed average chromium
Results metallic fraction of the ferritic matrix is 22.3 wt-%.
However, according to literature [11], narrow chromium
Microstructure and chemical composition depleted zones are established around chromium nitride
precipitates, which can be strongly detrimental to localized
The previously reported ferrite grain size and c1 fraction on corrosion resistance.
ferritized samples were 240 lm and 16 2%-vol, respec- Figure 1 shows the calculated chromium, nickel and
tively [6]. In addition, the weighted average width of c1 nitrogen composition profiles established in the overall
grains was ~7 lm. microstructure after reheating treatments. As presented in
After numerical simulation of c1 precipitation, it was Fig. 1a, intergranular c2 formed during the 1,200 C iso-
predicted that c1 grains grow with negligible substitutional thermal anneal exhibited a heterogeneous chemical com-
solute partitioning with the matrix. Furthermore, it was position profile, which has been experimentally verified by
observed that the average chemical composition of c1 did the authors using XEDS microanalysis in the TEM [12].
not change appreciably as the precipitation temperature This c2 is composed of an outer shell (next to the c/a
was varied between 1,200 and 900 C. Therefore, the interface) enriched in Cr and N, and a core region (next to
average chemical compositions of c1 and remnant ferritic c1) depleted in these alloying elements. These unusual
regions were obtained from these calculations. The calcu- solute profiles are established due to the intergranular c2
lated chemical compositions as well as SEM-XWDS growth during the high temperature isothermal anneal. The
measurements are presented in Table 2. One can see that high diffusivity of the substitutional alloying elements at
the calculated values are close to the measured ones. The high temperatures makes this partitioning possible. How-
only important difference is the nitrogen content of c1. It ever, during the cooling stage the limited diffusivity of the
123
9024 J Mater Sci (2007) 42:90219029
Table 3 TEM measured chemical analysis (XEDS) and numerical In contrast to the solute partitioning observed as a result
calculations (NC) of the metallic fraction of the main elements in the of the isothermal anneal at 1,200 C, the c2 formed at
ferritic matrix and nitrides in a ferritized sample (wt-%)
900 C exhibits a nearly homogeneous chemical compo-
Phase Fe Cr Ni sition, as presented in Fig. 1b. The c2 formed at the lower
temperature inherits the substitutional alloying concentra-
a XEDS 70.4 24.0 4.6
tion of the ferritic matrix due to low diffusivity of substi-
NC 72.85 22.3 4.84
tutional elements.
M2N XEDS 8.8 89.9 0.18
SEM and TEM analyses of nitride precipitation were in
NC 2.4 97.6 0.05
good agreement with numerical simulations. A consider-
able amount of chromium nitride precipitates were ob-
served in ferritized as well as 900 C reheated samples,
whereas virtually no nitrides were observed in the 1,200 C
reheated samples. Figure 2 shows TEM micrographs of the
general view of intragranular nitrides observed in a sample
reheated to 900 C. The intragranular nitrides varied in
width between approximately 40 and 120 nm. Figure 2a
shows a general view of a ferritic region containing Cr2N
nitrides (marked with arrows in the figure) and also some
dislocations. In Fig. 2b, a selected area diffraction (SAD)
pattern from the ferritic matrix, at the axis zone [011] a is
shown. Figure 2c presents an enlarged detail of one of the
nitrides in Fig. 2a; wider darkbright contrast fringes
inside nitrides can be seen, which correspond to Moire
patterns. A detailed crystallographic study of this nitride
precipitation is presented in a previous work [6], where
these nitrides were identified, by nano diffraction analysis,
as trigonal Cr2N.
Electrochemical experiments
123
J Mater Sci (2007) 42:90219029 9025
Fig. 3 Current density vs. Potential during pitting (a) and DL-EPR
(b) electrochemical corrosion tests
123
9026 J Mater Sci (2007) 42:90219029
corrosion test; in these regions (Fig. 4a) a low density, 1,200 C reheated samples. Corrosion attack during DL-
between approximately 50 and 200 pitmm2, of small pits, EPR tests was mainly observed in two locations: (i) the
between 1 and 3 lm in diameter, was observed. However, middle region of ferritic grains where intense chromium
a few regions, with a number of regions per unit area nitride precipitation is expected and (ii) the near region of
ranging between approximately 0.15 and 0.35 regionmm some c/a interfaces, where chromium nitride precipitation
2
,were observed that were severely corroded (Fig 4b). is also expected, as previously discussed. The superior
These regions, which varied in size from approximately localized corrosion resistance of both c1 and c2 grains was
200450 lm, consisted of a network of connected and observed.
unconnected pits of diverse sizes. Qualitatively, it was
observed that the extent of damage in these regions fol- The effect of alloying redistribution on the computed
lowed the same order as the pitting potential: wrought PRE index profiles
material < 1,200 C reheated < 900 C reheated < ferri-
tized. Figure 7 shows the calculated pitting resistance equivalent
Although it is not unambiguously clear, large side plates index (PRE = wt-% Cr + 3.3 wt-% Mo + 16wt-%N) pro-
(possibly austenite) are evident at the sample surface of the files, established in the overall microstructure after
corroded samples, Fig. 4b, that may possibly act as barriers reheating treatments. Aiming to simplify the kinetic com-
to additional pitting damage. Aiming to further clarify the puting simulations, which computing time strongly in-
actual localization of pit-corrode regions among ferrite or creases as a consequence of increasing the number of
austenite regions in microstructure, a sample reheated at alloying elements included in the system, the molybdenum
1,200 C was electrolytic-etched after being submitted to content was not used in the PRE computations. This sim-
pitting corrosion test. Figure 5 shows the appearance of plification should exert a minor effect on the overall results
that sample reheated at 1,200 C after the pitting corrosion because, (i) the Mo-content of the alloy is low (0.3 wt-%)
test followed by electrolytic etching. Although the elec- and (ii) prior thermodynamic simulations (not presented)
trolytic etching significantly modifies the prior appearance shown atomic partitioning of this element in microstructure
of the surface subjected to pitting test, it can be seen that contributes to PRE differences in a minor extent when
pit-damaged regions were inside prior ferritic grains. As compared to the effect of Cr- and N-content. In addition,
can be seen in this figure, the coarse intergranular allotri- only the elements in solid solution within the ferrite and
omorphic austenite and intragranular austenite regions austenite were taken into consideration to compute the PRE
acted as barriers for the growth of pitting damage. The index profiles, i.e. the elements present in the form of M2N
detail presented on the top right of Fig. 5 shows the barrier where not taken into account.
effect of these austenite particles. The most significant aspect regarding the PRE index
Figure 6 shows the appearance of the sample surface profiles (Fig. 7) is the low PRE index (22.3) of ferritic
after DL-EPR tests. Corrosion attack of the sensitized areas regions in the as-welded high temperature HAZ micro-
in the ferritized sample as well as the one reheated at structures (ferritized samples) compared to the wrought
900 C occurred, but no such attach was observed in the material. In the wrought material the PRE index of a and c
are 25.3 and 23.1, respectively. On the other hand, the PRE
index of reheated high temperature HAZ microstructures
varies between 21.8 and 22.6 for a and 23.5 and 32.7 for c.
This reduction in PRE index suffered by ferritic regions, as
a consequence of the welding thermal cycles, is mainly due
to the almost absence of chromium partitioning between
the austenitic (c1 + c2) and ferritic grains in the HAZ mi-
crostructures. This behavior is quite different to the ob-
served in the wrought material, where the carefully
selected combination of chemical composition and solu-
bilization heat treatment results on an almost balanced PRE
index between the two phases. Care should be taken be-
cause the reported PRE reduction within the ferritic regions
(a + M2N) has not taken into account the localized Cr
Fig. 5 SEM micrographs showing the appearance of the sample depletions around M2N precipitates, which will have a
surface after pitting electrochemical corrosion test pursued by strong influence on the localized corrosion resistance of
electrolytic etching. Sample reheated to 1,200 C such microstructures.
123
J Mater Sci (2007) 42:90219029 9027
123
9028 J Mater Sci (2007) 42:90219029
which grows as c2 within these changing nitride colonies supersaturated ferrite during the fast cooling following the
during reheating treatments [7]. Therefore, these similari- ferritization treatment, dissolved during the reheating treat-
ties strongly suggest that the results here presented for ments and the nitrogen is incorporated by the growing aus-
intergranular c2 growth can be a reasonable approximation tenite. The long diffusion paths during reheating at high
to intragranular c2 growth and of course to its effects on the temperatures allow for the nitrogen to be mostly incorporated
corrosion resistance of UNS S32304 DSS. Further by preexisting austenite particles that are growing, resulting
research, including numerical simulation is necessary to in nitride-free ferrite, as verified by the presented calculation.
clarify these similarities. However, the simulation of On the other hand, the shorter diffusion paths experienced at
intragranular c2 nucleation and growth requires more lower temperatures favor a modest growth of preexisting
elaborated models and one dimension approximations as austenite particles and the formation of acicular-type intra-
the ones used here to model intergranular c2 growth may granular c2 colonies within the preexisting nitrides colonies
not be considered satisfactory. that were far apart from austenite particles [7].
The calculated chemical profiles presented in Fig. 1 do The superior localized corrosion resistance of both c1
not fully agree with previously reported results on the and c2 grains was observed, even for the 1,200 C reheat-
chemical composition of intragranular c2 on super-duplex ing treatments in which relatively low PRE index value in
stainless steels with additions of Cu and W [8, 9]. On the the middle of c2 grains was predicted (Fig. 7). This result
other hand, the predicted and measured [12] composition agrees with reported research on DSS UNS S31803 [14].
profiles support corrosion resistance results for DSS UNS Besides localized corrosion occurrence in sensitized areas,
S13803 [14], which exhibits a similar chemical composi- DL-EPR test leads to generalized corrosion [15]. The
tion to the studied alloy UNS S32304. However, in addi- extent of localized corrosion at phase interfaces was dif-
tion to the differences in chemical composition of the ferent in the diverse HAZ microstructures (Fig. 6). In the
Superduplex alloys with Cu and W additions, the calcula- ferritized samples as well as the 900 C reheated samples
tions presented here are based on more elaborate kinetics the corroded regions adjacent to the phase interfaces were
models, which are expected to better represent the system. narrow, contrary to the 1,200 C reheated samples in which
However, the simplifications here used include the reduc- this corrosion damage occurred along a broader region
tion of the system to one dimension and the isothermal around to the interface. This is probably due to occurrence
treatments. Thus, important differences with actual welds of galvanic effects between a and c grains, which have
may be expected, especially for Superduplex alloys. Nev- different Cr- and Ni-content in the near interface region
ertheless, this new contribution helps to better understand (Fig. 1). The largest difference in Cr- and Ni-content in the
the chemical partitioning among ferrite and austenite in near interface region between a and c was in the 1,200 C
duplex stainless steels and particularly its effects on cor- reheated samples (see Fig. 1), which were the samples
rosion resistance. displaying the broadest corrosion at phase interfaces. One
The PRE index of ferritic regions among the different can also see in Fig. 6 that some intragranular austenitic
HAZ microstructures, namely as-welded HAZ (ferritized grains can be detached from the ferritic matrix as a con-
samples) and reheated-HAZ (1,200 and 900 C reheated sequence of phase interface corrosion, which is due to the
samples), suffered only minor variations. However, the above pointed out galvanic effect, as well as the large
nitrogen content within the ferrite, and thus the fraction of surface/volume ratio and small size of these c particles.
chromium nitrides precipitated in this phase, underwent large Thus, this small intragranular austenite particles detach-
changes. These changes have an important effect on the ment from the sample will worsen the corrosion damage in
corrosion resistance of such microstructures. The predicted the material. This can be the mechanism behind the pre-
supersaturation of ferritic regions with nitrogen in the as- viously reported damaging effect of intragranular c2 on the
welded HAZ microstructure, which results in abundant in- corrosion resistance of duplex stainless steels. However,
tragranular M2N precipitation, was reduced after reheating more research in different duplex alloys will be necessary
treatments. The 900 C reheating treatment induced a mod- to generalize this mechanism.
erate nitrogen content reduction in the ferritic regions near to Summarizing, the numerical simulation and experi-
the growing austenite, as shown in Fig. 1b. However, the mental results revealed that the moderately low PRE index
ferritic region immediately adjacent to the a/c2 interface established in ferritic regions of welded DSS UNS S32304
underwent an important nitrogen enrichment, which explains is the main cause for lower localized corrosion resistance
the previously reported nitride (M2N) precipitation at the a/c of HAZ microstructures when compared with the wrought
interface [6]. On the other hand, the 1,200 C reheating material. In addition, it can be also inferred that chromium
treatment induced an almost complete exhaustion of nitrogen nitride precipitation within ferritic regions further impairs
from ferritic regions. In both cases, a large amount or almost the corrosion resistance of HAZ microstructures due to
all the M2N nitrides that formed within nitrogen sensitization occurrence.
123
J Mater Sci (2007) 42:90219029 9029
Thus, selective corrosion inside gamma grains, c1 and nitrogen from ferritic regions, which occurred as a
c2, is not the main cause for the poor corrosion resistance of consequence of c2 growth.
DSS UNS S32304 HAZ microstructures. However, the The observed corrosion resistance improvement was
fraction of chromium nitrides within the ferritic regions more pronounced after isothermal reheating to
indirectly depends on the total fraction of austenite 1,200 C, compared to the 900 C reheating treatment,
(c1 + c2), as austenite grains exhaust the nitrogen from the which appears to be directly related to the removal of
ferritic regions. In addition, at high reheating temperatures, nitrogen within the ferrite grains.
the nitrides modify the substitutional alloy content profiles In contrast with what has been reported in literature for
in the region near to the c/a interfaces. superduplex stainless steels [2, 3], a direct detrimental
Precipitation of austenite displaying superior PRE index effect of intergranular c2 precipitation on corrosion
in the region adjacent to the c/a interfaces when compared resistance was not observed. Selective corrosion inside
with ferrite can deteriorate the generalized corrosion gamma grains, c1 and c2, was not the main cause for the
resistance of HAZ microstructures due to galvanic effects. impaired corrosion resistance of DSS UNS S32304
This effect should be particularly important in regions HAZ microstructures. However, the observed ferritic
where profuse precipitation of small intragranular austenite matrix dissolution around intragranular c2 suggest that
particles occurs. In this case ferrite around these c particles small gamma particles detachment from the material is
can be easily damaged due to both the large surface/volume the probable cause for a high corrosion rate.
ratio of c particles and their small size.
Acknowledgments The authors would like to acknowledge to
LNLS and FAPESP, for funding the actual and predecessor projects.
Conclusions
The results and analyses from the calculations and exper- References
imental measurements on heat-treated DSS UNS S32304
1. Ogawa TO, Koseki T (1989) Weld J 68:181
have led to the following conclusions:
2. Atamert S, King JE (1991) Z Metallkd 82:230
The electrochemical pitting and DL-EPR tests showed 3. Nilsson JO, Karlsson L, Andersson JO (1995) Mat Sci Techn
11:276
an impaired corrosion resistance of HAZ microstruc- 4. Gooch TG (1996) Weld J 75:135
tures with respect to the wrought material. This 5. Gregory A, Nilsson J-O (2002) Met Mat Trans A 33A:1009
reduction of corrosion resistance after welding was 6. Ramirez AJ, Lippold JC, Brandi SD (2003) Met Mat Trans A
mainly attributed to a decrease in the corrosion 34A:1575
7. Ramirez AJ, Ph.D. Thesis, University of Sao Paulo, Sao Paulo,
resistance of the HAZ ferritic grains. This was attrib- Brazil, 2001, p. 241. Full text available at http://www.te-
uted to (i) the mean chromium content of these ferritic ses.usp.br/teses/disponiveis/3/3133/tde02012002175418/
grains is the same chromium content of the alloy (i.e. 8. Hertman S, Huhtala T, Karlsson L, Nilsson J-O, Jargelius-
ferritic grains in the HAZ exhibited lower chromium Pettersson R, Wilson A (1997) Mater Sci Tech 13:604
9. Nilsson J-O, Huhtala T, Jonsson P, Karlsson L, Wilson A (1996)
content than ferritic grains in the wrought material) and Met Mat Trans A 27A:2196
(ii) the presence of chromium depleted regions around 10. Majidi AP, Streicher MA (1984) Corrosion 40:393
nitrides, which were often observed. 11. Andersson JO, Helander T, Hoglund LH, Shi PF, Sundman B
As-welded HAZ microstructures displayed lower local- (2002) CALPHAD 26:273
12. Garzon CM, Ramirez AJ (2006) Acta Mater 54:3321
ized corrosion resistance and higher degree of sensiti- 13. Andersson JO, A gren J (1992) J Appl Phys 72:1350
zation than reheated HAZ microstructures. The extent 14. Muthupandi V, Bala Srinivasan P, Seshadri SK, Sundaresan S
of corrosion resistance improvement after reheating (2004) Sci Technol Weld Joi 9:47
treatment was mainly caused by the removal of 15. Menendez H, Devine TM (1990) Corrosion 46:410
123
J Mater Sci (2007) 42:90309036
DOI 10.1007/s10853-007-1860-y
Received: 21 February 2007 / Accepted: 17 May 2007 / Published online: 20 July 2007
Springer Science+Business Media, LLC 2007
123
J Mater Sci (2007) 42:90309036 9031
complex ac susceptibility shows two peaks, Tp and Tg, at up to 0.6 T and transport critical current density in order to
high ac fields [12, 13]. The first peak (at high field) appears investigate the effect of the partial substitution of Ca by Sm
at a temperature Tg slightly bellow Tc due to intragranular on superconducting properties of Bi-2223 samples. We
supercurrents, and a second loss peak (at low-field) appears also reported XRD measurements to extract the lattice
at a temperature Tp lower than Tg due to sample-circulating parameters and the relative portion of Bi-2223 and that of
intergranular supercurrents. Bi-2212 phases and EDXRF measurements to estimate
Doping with different elements at various amounts and stoichiometric ratios of the considered samples.
adjusting preparation condition affect phase formation and
physical properties in Bi-based system. Many studies of
doping into superconductor oxide ceramics have been Experimental details
made in order to improve their superconducting, magnetic
and mechanical properties [1421]. The objectives of The Bi1.6Pb0.4Sr2(Ca2xSmx)Cu3Oy samples with
doping studies are to optimize the hole concentration, to 0 < x < 0.005 Sm were prepared by using the standard
introduce pinning centers and to enhance the formation of solid-state reaction method [16]. Rectangular bars were cut
Bi-2223 phase. Li, Ti, W and Mo have been doped into from the sintered samples for electrical resistivity, critical
BiPbSrCaCuO bulk samples, and it was reported that Li current and ac susceptibility measurements. The typical
doping led to an increase in critical current density (Jc), sample size was 2.9 2.1 12.1 mm3.
while Ti doping led to a decrease, and W and Mo doping XRD data were taken using a Rigaku D/Max-IIIC dif-
did not contribute a substantial change in Jc [23, 24]. fractometer with CuKa radiation in the range 2h = 460
Substitution of Eu, Dy and Tm in the Bi-2212 system has with a scan speed of 3/min and a step increment of 0.02
caused a transition from superconductor to insulator at room temperature. Phase purity and the lattice parame-
[2426]. The partial replacements of Ca by Y, Gd, and Sm ters were obtained from these XRD patterns. The accuracy
in the Bi-2212 and Bi-2223 systems decreased carrier in determining the lattice parameter c was 0.001 A . The
concentration due to structural modulations and changes in mean values of lattice parameter c for Bi-2223 phase are
the valency of Cu [2528]. On the other hand, it was determined from (002), (0010), (0012), (115) and (0014)
observed that the substitution of divalent Ca by trivalent peaks of the XRD measurements for pure and Sm substi-
rare-earth elements in the Bi-2212 system causes a sharp tuted Bi1.6Pb0.4Sr2(Ca2xSmx)Cu3Oy samples. The relative
change in transition from a superconductor to an insulator proportions of the Bi-2223 phase were determined from
and a decrease in the lattice parameter c because of a (002), (008), (0010), (115), (0012), (200), (1111) and
drastic change in the carrier concentration [22, 27]. A (0212) peaks and the relative proportions of Bi-2212 phase
comprehensive study of Sm addition to Bi-2212 and were determined from (115), (220), (315), (317), (1115)
Bi-2223 was reported in the literature [1619, 29]. In our and (1112) peaks. The relative proportions of the Bi-2223
previous study, we investigated the effect of the partial phase were determined from (002), (008), (0010), (115),
substitution of Ca by Sm in the Bi-2223 superconducting (0012), (200), (1111) and (0212) peaks and the relative
samples on its structural and physical and magnetic prop- proportions of Bi-2212 phase were determined from (115),
erties [1619]. It was obtained that superconducting (the (220), (315), (317), (1115) and (1112) peaks, using the
values of Tc and Jc) and mechanical properties degrade and following well-known expressions [30, 31];
the lattice parameter c and the Bi-2223 phase decreased P
with increasing Sm content. It was also reported that for IH hkl
f2223 P P 1
x = 1.0 and 1.5 the compounds showed semiconducting IH hkl ILhkl
behavior. We investigated effect of the partial substitution P
of Ca by Sm in Bi(Pb)SrCaCuO system by using the low ILhkl
f2212 P P : 2
field ac magnetic susceptibility method [18]. The investi- IH hkl ILhkl
gations consisted of ac susceptibility, the temperature
dependence of intergranular critical current density, Here IH(hkl) and IL(hkl) are the intensities of the (hkl)
Jcinter Tp , and scanning electron microscopy (SEM) mea- diffraction lines for Bi-2223 and Bi-2212 phases, respec-
surements. It was observed that the grain size of the pure tively (Fig. 2).
sample is considerably larger than the Sm-doped samples, The transport critical current (Ic) (77 K, self field,
the overall susceptibility curves of the samples are shifted 1 lV cm1) and dc resistivity (5 mA dc current) were
to lower temperature and the value of Jcinter Tp decreases measured by a conventional four-probe method. The con-
by increasing the Sm concentration. tacts were made with silver paint. The outside pin-points
In this work, we report measurements of electrical feed the current and the intermediate ones (l = 7 mm)
resistivity as a function of temperature under magnetic field allow the voltage drop measurement. A Keithley 220
123
9032 J Mater Sci (2007) 42:90309036
123
J Mater Sci (2007) 42:90309036 9033
(0012)H
(0010) H
(a) pure sample and Sm
substituted samples (b) x =
(0014)H
0.0005, (c) 0.001, and (d) 0.005.
(200)H
(002)H
(119)H
(1111)H
(115)H
The peaks indexed (hkl)L and
(0212)H
(0024)H
(220)H
(220)L
(1119)H
(hkl)H represent the Bi-2212
(115)L
(008)H
(317)L
(315)L
(004)H
and Bi-2223 phases,
respectively
b) x=0.0005
(0012)H
(0010)H
(0014)H
(200)H
(002)H
(1111)H
(119)H
I nt e ns i t y (a . u. )
(115)H
(0024)H
(0212)H
(1119)H
(220)H
(008)H
(2012)L
(220)L
(315)L
(004)H
(317)L
(115)L
c) x=0.001
(0010) H
(0012)H
(0014)H
(002)H
(200)H
(119)H
(1111)H
(115)H
(0024)H
(220)H
(0212) H
(1119)H
(220)L
(2012)L
(008)H
(315)L
(317)L
(004)H
(115)L
d) x=0.005
(0012)H
(0010)H
(0014)H
(002)H
(119)H
(200)H
(115)H
(1111) H
(0212) H
(0024)H
(220)L
(1119)H
(220)H
(008)L
(2012)L
(115)L
(1115)L
(008)H
(004)H
5 10 15 20 25 30 35 40 45 50 55 60
2 ( Degree)
of the Bi-2212 peaks increase with increasing substitution lattice parameter c is due to incorporation of Sm ions into
of Sm for Ca. The determined lattice parameters from the the interstitial sites in the unit cell rather than occupation of
(00l) peaks of the XRD data are given in Table 1. It is the Ca sites. The relative volume fractions of the Bi-2223
observed that the lattice parameter c decreases significantly and Bi-2212 phases were determined from the peak
with increasing Sm content as listed in Table 1. Similar intensities of the same particular reflections using Eqs. (1)
result has been reported by Zandbergen et al. [35]. As and (2). As listed in Table 1, the volume fraction of
pointed out by Zandbergen et al., the behavior of lattice Bi-2223 phase decreased and that of Bi-2212 phase
parameter can be explained by the increase of the oxygen increased with increasing the Sm content.
content in the unit cell by the replacement of Ca2+ by Sm3+ The stoichiometry of Sm substituted annealed
in the structure. It was speculated that the excess of oxygen Bi1.6Pb0.4Sr2Ca2xSmxCu3Oy samples were investigated by
goes into the bismuth oxide layers causing a decrease in EDXRF technique. Figure 3 shows the EDXRF spectrum
lattice parameter c. It is also believed that the decrease of of Sm substituted Bi1.6Pb0.4Sr2Ca2xSmxCu3Oy samples
with x = 0.0005, 0.001 and 0.005. Stoichiometric ratios of
the analysed samples are given in Table 2 [19]. In this
analysis, Bi atomic ratio is given as 1.6. As seen from the
Table 1 Some characteristics of superconducting samples
table, Sm amount in the sample up to x = 0.001 is not
Samples Lattice Volume Tconset Tp (K) (for enough to examine by means of the EDXRF technique.
parameter fraction (%) (K) 111 A/m)
)
c (A
The electrical resistivity as a function of temperature
Bi-2223 Bi-2212 was measured in an external magnetic field up to 0.6 T for
Pure 37.236 94 06 105 97
the Bi1.6Pb0.4Sr2(Ca2xSmx)Cu3Oy samples with
0 < x < 0.005 Sm. Figure 4 shows temperature depen-
x = 0.0005 37.206 92 08 102 88
dence of resistivity for Bi1.6Pb0.4Sr2(Ca2xSmx)Cu3Oy with
x = 0.001 37.197 83 17 99 86
x = 0, 0.0005, 0.0010 and 0.0050 at zero fields. The
x = 0.005 37.190 79 21 97 85
resistivity versus temperature plots show that the resistivity
123
9034 J Mater Sci (2007) 42:90309036
decreases linearly with temperature in the normal state. As measurements [16] except the values of transition tem-
can be seen from the figure the transition temperature, perature and room temperature resistivity. In the present
Tcoffset , decreases from 90 K to 75 K with increasing the Sm work, dc resistivity measurements on the same samples
concentration at zero magnetic field. Room temperature were made approximately 2 years later than the previous
resistivity increases while critical transition temperature work. It is observed that the values of Tc decrease while the
decreases with increasing x from zero to 0.005 Sm. The values of room temperature resistivity increase (10 times)
behavior of the resistivity is the same as our previous in the present study. The transition curves from normal to
123
J Mater Sci (2007) 42:90309036 9035
Table 2 Calculated
Samples Bi Sr Ca Sm Cu
stoichiometry of Sm substituted
Bi1.6Pb0.4Sr2Ca2xSmxCu3Oy Pure 1.600 0.950 0.320 0.003 1.790
samples from EDXRF results
[19] x = 0.0005 1.600 0.860 0.240 0.000 1.730
x = 0.0010 1.600 0.850 0.240 0.000 1.760
x = 0.0050 1.600 0.820 0.2300 0.005 1.680
x = 0.0100 1.600 0.800 0.220 0.007 1.580
x = 0.1000 1.600 0.820 0.200 0.063 1.650
123
9036 J Mater Sci (2007) 42:90309036
123
J Mater Sci (2007) 42:90379044
DOI 10.1007/s10853-007-1819-z
Received: 2 April 2007 / Accepted: 2 May 2007 / Published online: 20 July 2007
Springer Science+Business Media, LLC 2007
Abstract The oxidation of the bulk amorphous alloy temperature and a remarkable resistance to crystallization.
Zr65Cu17.5Ni10Al7.5 in air in its amorphous and the super- The bulk metallic glasses require cooling rates of about
cooled liquid states was studied in the temperature range 1100 K/s or less unlike conventional melt-spun metallic
573663 K using X-ray photoelectron spectroscopy (XPS). glasses which require very high cooling rates of about
The oxide film mainly consisted of the oxides of Zr (as 106 K/s for processing [1]. The interest in bulk amorphous
ZrO2) and Al (as Al2O3). No Cu or Ni was found in the oxide glasses has considerably grown in recent times due to their
film formed on the amorphous state of the alloy while sig- ability to offer the scope for carrying out research inves-
nificant Cu (as CuO) was present in the oxide film formed on tigations both in the amorphous and the supercooled liquid
the alloy in its supercooled liquid state. The role of the states of a glassy alloy. Novel multicomponent Zr-based
various alloying elements during oxidation at high temper- alloys like ZrTiCuNiBe and ZrCuNiAl form an
atures in air is discussed in the paper. The XPS data from important class of bulk metallic glasses with some inter-
oxide film support the previously suggested mechanism for esting applications [2].
oxidation of this alloy, i.e. the rate controlling process In contrast to several studies done on the oxidation
during oxidation of the alloy at low temperatures (in the behaviour of conventional melt-spun amorphous alloys
amorphous state) is the back-diffusion of Ni and Cu, while [37 references therein], not many studies are available on
the oxidation at high temperatures (in the supercooled liquid oxidation of bulk metallic glasses [816]. In one of the
state) is dominated by the inward diffusion of oxygen. early investigations on high temperature oxidation of the
bulk amorphous alloy Zr60Al15Ni25 in air it was suggested
Introduction that the growth of the oxide is controlled by Ni back-dif-
fusion in the alloy [11]. Contrary to these observations
Bulk amorphous alloys show a fairly wide temperature Triwikantoro et al. [12] investigated the oxidation of sev-
range between glass transition and the crystallization eral Zr-based amorphous alloys (ZrCuNiAl) in air and
reported the presence of Cu and Ni in the oxide scale. In
another investigation by the present authors on oxidation of
the bulk amorphous alloy Zr65Cu17.5Ni10Al7.5 [14], a
A. Dhawan S. K. Sharma (&)
Department of Physics, Malaviya National Institute of mechanism for its oxidation in air was proposed suggesting
Technology, Jaipur 302 017, India that the back diffusion of Ni, and possibly Cu also, is the
e-mail: sksh@datainfosys.net rate limiting process during the oxidation of the alloy in the
amorphous or the glassy state, while oxidation process in
V. Zaporojtchenko F. Faupel
Faculty of Engineering, University of Kiel, Kaiserstr. 2, 24143 the supercooled liquid state is more likely dominated by the
Kiel, Germany diffusion of oxygen through the oxide layer [14]. In order
to verify this suggested mechanism, it is necessary to
Present Address:
obtain information about the presence of various alloy
A. Dhawan
Department of Physics, Jagannath Gupta Institute of Engineering constituents, especially Ni and Cu, in the oxide film using
& Technology, Jaipur 302 022, India some depth profiling technique. This was a motivating
123
9038 J Mater Sci (2007) 42:90379044
factor for carrying out the present investigations. The Results and discussion
technique of X-ray photoelectron spectroscopy (XPS) was
employed in conjunction with Ar+ ion sputtering for The oxide films formed on the alloy surface as a result of
obtaining the desired information from the oxide film high temperature (in the temperature range 573663 K)
formed on amorphous and supercooled liquid states of the oxidation in air were characterized by XPS. The main
bulk alloy Zr65Cu17.5Ni10Al7.5 during its oxidation in air in features of XPS peaks from specimens annealed at 573 and
the temperature range 573663 K. 603 K (corresponding to the oxidation of specimens in the
amorphous state) were quite similar. Similar observations
were made from XPS peaks obtained from specimens
Experimental annealed at 633 and 663 K (corresponding to the oxidation
of specimens in the supercooled liquid state). Therefore, the
Specimens of size (10 mm 15 mm) were cut from the figures at two typical temperatures 573 and 633 K which
as-cast amorphous ribbon (10 mm wide 30 lm thick) of correspond to oxidation of specimens in the amorphous and
the bulk amorphous alloys Zr65Cu17.5Ni10Al7.5. The the supercooled liquid states of the alloy, respectively, are
samples were first cleaned ultrasonically in acetone and shown here and discussed in the following.
ethanol and later dried under a jet of pressurised air. Figure 1 depicts the Zr 3d spectra recorded from the
These samples were put in an open-ended quartz tube oxide films formed on specimens of amorphous
which was inserted in a tubular furnace and annealings Zr65Cu17.5Ni10Al7.5 at temperatures 573 and 633 K for the
were performed at 573 K for 20 h, 603 K for 8 h, 633 K as-received specimens and after sputtering with 5 keV Ar+
for 3 h and 663 K for 1 h in dry air environment inside ions. The XPS peaks for Zr 3d are reported to occur at
the open-ended quartz tube. As the oxidation was carried 178.7 and 182.2 eV for Zr0 (in metallic form) and Zr4+ (in
out ex-situ in a furnace it was not possible to remove the oxide form), respectively [19]. Therefore, the Zr 3d XPS
native oxide layer from the specimens by carrying out ion peaks in Fig. 1a and b occurring at 181.9 eV and 182.4 eV
beam sputter cleaning or scribing in a vacuum chamber. for the specimen oxidized at 573 K from the as-received
The study of the native oxide formed on this alloy has and the sputtered specimen surface, respectively and also
earlier been reported in another investigation [9]. The the corresponding XPS peaks (c) and (d) in the same figure
choice of timetemperature combination for annealing occurring at 182.0 and 182.5 eV for oxidation at 633 K, are
was made on the basis of the results obtained in suggestive of the presence of Zr4+ species on the top sur-
our previous investigation [14] and the known TTT face [9, 1921]. A small shift of about 0.5 eV in Zr 3d peak
diagram for this alloy [17, 18] so that the alloy towards higher binding energy side after sputtering , i.e.
Zr65Cu17.5Ni10Al7.5 remained after oxidation treatment after removal of the adsorbed hydrocarbon impurity layer
either in the amorphous (573 and 603 K) or in the after argon ion sputtering, may be attributed to chemical
supercooled liquid state (633 and 663 K) and did not
crystallize during annealing. Then the annealed samples
were taken out of the furnace and the shiny side (free
surface during melt- spinning of the alloy) was analysed Zr 3d peaks at 573 K
by X-ray photoelectron spectroscopy (XPS). The XPS d a as-received
b spt. 5 min
measurements were done using an electron spectrometer c Zr 3d peaks at 633 K
(VG MK II) equipped with a non-monochromatic Mg Ka c
d
as-received
spt. 4 min
source energy (hm = 1253.6 eV) in a base pressure better b
Intensity [a.u.]
123
J Mater Sci (2007) 42:90379044 9039
Intensity [a.u.]
recorded from the specimen surface (Fig. 2a, b) oxidized at
573 K indicate hardly any presence of Cu. The Cu 2p peak c
in Fig. 2c occurs at 933.6 eV from the as-received amor-
phous specimen Zr65Cu17.5Ni10Al7.5 oxidized at 633 K d
indicating the presence of Cu2+, possibly as CuO [19].
After sputtering for 4 min with Ar+ ions, the Cu 2p3/2 peak
(Fig. 2d) shifts to the lower binding energy side, i.e. 870 865 860 855 850 845
932.5 eV indicating the reduction of oxidized copper Binding Energy [eV]
(Cu2+) to lower valent states (Cu+, Cu0) [22, 23]. It is
noteworthy here that the possibility of preferential sput- Fig. 3 XPS peaks for Ni 2p from specimens of amorphous
tering and the reduction of the oxides to lower valent states Zr65Cu17.5Ni10Al7.5 oxidised at 573 K: (a) as-received, (b) after
5 min sputtering; and oxidised at 633 K: (c) as-received, (d) after
during argon ion sputtering of the oxidised alloy surface is 4 min sputtering
not ruled out [22, 23]. However, the qualitative nature of
conclusions derived from observations made in the present
study is not changed by these effects. It is further inter- peaks in Fig. 4 occur about at 119.0 eV in these spectra
esting to note from Fig. 3 that hardly any Ni is present as and thus point to the presence of Al3+, possibly as Al2O3 on
can be inferred from the extremely poor signals in the the surface along with ZrO2. However, it is interesting to
observed Ni 2p peaks from specimens oxidized at 573 and note from this figure that the peak of Al is seen after oxi-
633 K. It is worth mentioning here that features for Cu and dation at 573 K (Fig. 4a, b) while no significant Al peak is
Ni peaks at 603 and 663 K (figures not shown here) were observed after oxidation at 633 K (Fig. 4c, d). A broad
similar to those at 573 and 633 K, respectively. Figure 4 peak seen at about 122.0 eV in Fig. 4d stems from Cu 3s
shows Al 2s XPS peaks recorded from the specimen [19]. These observations can be understood in conjunction
oxidized at 573 and 633 K. The Al 2s peak (Fig. 4) was with Cu 2p peaks shown in Fig 2c and d which indicate the
recorded instead of the more intense Al 2p peak due to the presence of significant amount of Cu in the oxide film
overlap of the latter with the Cu 3p peak [19]. The Al 2s formed at 633 K and thus the relative concentration of Al
peaks for oxidic Al3+ and metallic Al0 are known to occur in the oxide layer becomes too low to give a significant
at 119.0 and 117.0 eV, respectively [9, 19]. The Al 2s
Al 2s peaks at 573 K
a as-received
b spt. 5 min
d Cu 2p peaks at 573 K
a as-received Al 2s peaks at 633 K
b spt 5 min c as-received
b d spt. 4 min
Cu 2p peaks at 633 K
c as-received
d spt. 4 min a
Intensity [a.u.]
In t e n s i t y [ a . u . ]
c
d
b
c
a
950 945 940 935 930 925 920 124 122 120 118 116 114 112
Binding Energy [eV] Binding Energy [eV]
Fig. 2 XPS peaks for Cu 2p from specimens of amorphous Fig. 4 XPS peaks for Al 2s from specimens of amorphous
Zr65Cu17.5Ni10Al7.5 oxidised at 573 K: (a) as-received, (b) after Zr65Cu17.5Ni10Al7.5 oxidised at 573 K: (a) as-received, (b) after
5 min sputtering; and oxidised at 633 K: (c) as-received, (d) after 5 min sputtering; and oxidised at 633 K: (c) as-received, (d) after
4 min sputtering 4 min sputtering
123
9040 J Mater Sci (2007) 42:90379044
XPS peak of Al from this sample (Fig. 4c, d). It is further temperatures (in the temperature range 573663 K) can be
interesting to note here that no significant peak due to Cu is obtained from the available values for heat of oxide
seen in the oxide film on the sample oxidised at 573 K formation. The heats of formation for ZrO2, Al2O3, NiO
(Fig. 2a, b) while Al is present in the oxide film of this and CuO are 1101.1, 1117.6, 497.7 and 314.8 KJ/mol
sample (Fig. 4a, b). Further, the peak features of Cu and Al O2, respectively [26]. This suggests that oxides of Zr and
peaks after oxidation at 603 and 663 K (figures not shown) Al in the alloy Zr65Cu17.5Ni10Al7.5 are likely to be formed
were similar to those observed at 573 and 633 K, respec- first because of their strong affinity for oxygen, followed by
tively. Thus from an analysis of Al (Fig. 4) and Cu (Fig. 2) Ni and Cu. However, the back-diffusion or segregation of
peaks it may be concluded that due to strong segregation of alloy constituents during oxidation controls the oxidation
Cu at higher oxidation temperatures (633 and 663 K) the kinetics and thus the formation of oxidation products in the
relative amount of Al in the oxide film formed at these oxide film.
temperatures becomes too small to yield significant XPS In the following discussion the role of various alloying
peaks due to Al. (Fig. 4c, d). This is also evident from elements during oxidation is analysed. It is interesting to
Al/Zr depth profile shown in Fig. 7 later. note here (Figs. 2 and 3) that hardly any Cu or Ni is seen in
The formation of various oxides (ZrO2, CuO, Al2O3) is the oxide film after oxidation of the alloy at 573 and 603 K
supported by the peak positions of the O 1s peak along with (which correspond to the oxidation of the alloy in the
those of the respective alloy constituents (Figs. 1, 2 and 4). amorphous state). However, significant amount of Cu (as
Figure 5 depicts the O 1s XPS peaks from oxidized spec- CuO) is observed in the oxide film after oxidation at higher
imens. The O 1s peaks from the as-received surfaces temperatures 633 and 663 K (which correspond to the
appear at 530.0 eV (Fig. 5a, c) for oxidation at 573 and oxidation of the alloy in the supercooled liquid state). On
633 K, respectively and are shifted to higher binding the other hand Al peaks (Fig. 4) show contrasting behav-
energy side after sputtering (530.4 eV and 530.6 eV) iour with Cu peaks. In order to clearly highlight these
(Fig. 2b, d). The O 1s peak for oxide formation (O2 2 ) has observations depth profiles for the concentration ratio of
been reported to occur at 530 0.4 eV [9, 19, 20, 24, 25]. these elements with respect to Zr in the alloy were obtained
The observed asymmetry at high binding energy side in the and are shown in Figs. 6 and 7. An interesting feature of
O 1s curves in Fig. 5a and c before sputtering most likely Fig. 6 pertains to the absence of both Cu and Ni in the
corresponds to the presence of hydroxyl ions on the surface oxide film formed at 573 and 603 K which correspond to
[9, 19, 20, 25]. This feature disappears after sputtering as is the oxidation of the alloy sample in the amorphous state. In
evident from Fig. 5b and d. Therefore, the O 1s spectra contrast, appreciable amount of Cu is seen in the oxide film
shown in Fig. 5 correspond to the formation of oxides of formed at 633 and 663 K which correspond to the forma-
various alloy constituents (ZrO2, Al2O3, CuO) on the top tion of oxide film in the supercooled liquid state of the
surface at these high temperatures [9, 19, 20, 24, 25]. alloy. It is interesting to mention here that depth profiles for
An understanding about the formation of various oxides on Al shown in Fig. 7 are in sharp contrast with those for Cu
the alloy surface during oxidation in air at various (Fig. 6). Al is present on the surface after oxidation at all
temperatures, but hardly any Al is seen deeper in the oxide
d
O1s peaks at 573 K
a as-received
1.2
b spt 5 min
c O1s peaks at 633 K
Cu/Zr
573 K
c as-received 1.0 603 K
d spt 4 min
b 633 K
663 K
Intensity [a.u]
Concentration Ratio
Ni/Zr
0.8
573 K
a 603 K
633 K
0.6 663 K
0.4
0.2
123
J Mater Sci (2007) 42:90379044 9041
p
Al/Zr
An estimate shows that the diffusion length 4Dt of Ni at
0.12 573 K 573 K is about ~34 nm (DNi at 573 K = 3.9 1021 m2/s
603 K
0.10 633 K and t = 20 h, being the time corresponding to the oxidation
Concentration Ratio
663 K
of the alloy in its amorphous state). A similar calculation at
0.08
603 K yields the value of diffusion length as 56 nm (DNi at
0.06 603 K = 2.7 1020 m2/s and t = 8 h). These estimates are
indicative of the fact that due to back diffusion of Ni during
0.04
oxidation of the alloy specimen in the amorphous state Ni
0.02 has moved by about 3456 nm deeper into the alloy
beyond the oxide-alloy interface. Though no data for dif-
0.00
fusion of Cu in this alloy have been reported in the liter-
0 4 8 12 16 20 24 28 32 ature, diffusion of both Cu and Ni have been investigated
Sputter Depth [nm]
in another conventional type amorphous alloy Zr50Cu50
Fig. 7 Plots of atomic concentration ratio [Al/Zr] versus sputter [29]. The reported data for diffusion of Ni and Cu in
depth for the alloy Zr65Cu17.5Ni10Al7.5 oxidised at different oxidation amorphous Zr50Ni50 [29] point to the fact that diffusion of
temperatures Ni is faster than that of Cu by about an order of magnitude
at these temperatures. This is supported by the observed
film formed at higher temperatures, i.e. at 633 and 663 K size effects on diffusion in amorphous alloys [28, 30]
(Fig. 7). It is very likely that due to strong segregation of suggesting that smaller atoms diffuse faster than bigger
Cu at these temperatures (Fig. 6) the relative concentarion atoms in amorphous alloys. It thus appears from the above
of Al in the oxide film containing Zr and Cu becomes too analysis that both Cu and Ni have moved beyond the oxide-
low to yield a significant XPS peak of Al (Figs. 4 and 7). In alloy interface by several tens of nanometers after oxida-
fact, Al seen on the surface just at the beginning of the tion at 573 and 603 K. Figures 6 and 7 show depth profiles
oxidation (see Fig. 7) pertains to the initial amount of Al up to about 2329 nm only and the oxide-alloy interface is
on the surface, which undergoes oxidation, but its amount not yet reached as evident from the plots. Due to practical
is drastically reduced as a result of strong segregation of Cu difficulties encountered in experimentation, especially in
in the oxide film formed at higher temperatures (Fig. 6). increasing the sputtering rate, it was not possible to carry
This is supported by the fact that no segregation of Cu is out the sputtering of the entire oxide film and thus to obtain
observed during oxidation at 573 and 603 K (Fig. 6) and an estimate of the oxide film thickness. Sun et al. reported
thus significant amount of Al is seen in the oxide film at oxide film thickness of about 60 nm formed on amorphous
these temperatures (Fig. 7). In view of these observations it Zr60Al15Ni25 after its oxidation in air at 603 K for 3 h and
is suggested that alloy constituents, especially Cu and Ni at 623 K for about 5 h [11]. It is thus likely that the oxide
seem to play an important role during oxidation of the alloy film thickness on the specimens of Zr65Cu17.5Ni10Al7.5 in
Zr65Cu17.5Ni10Al7.5. The role of Cu and Ni is explained by the present study might be of this order, i. e. about several
the controlling oxidation mechanisms and is discussed in tens of nanometers and the depth profiles in Fig. 6 origi-
the following paragraphs. nated from a part of the oxide film. However, the above
It was suggested [14] in our previous investigation on analysis and interpretation of these depth profiles do pro-
oxidation of this alloy that the oxidation in the amorphous vide valuable information about possible oxidation mech-
state is controlled by the back-diffusion of Ni and Cu while anisms for this alloy.
the inward diffusion of oxygen becomes dominant during It is, thus, suggested from the above discussion that the
oxidation of the supercooled liquid state of the alloy. The depletion of Ni and Cu in the oxide film, as seen in Fig. 6,
fact that no Ni and Cu are seen during oxidation of the possibly arises due to their back diffusion into the alloy
alloy in its amorphous state clearly supports this mecha- during oxidation of the alloy and the absence of these
nism for oxidation of the alloy [14] that the back diffusion elements (Ni and Cu) in depth profiles after oxidation of
of Ni and Cu seems to be the rate controlling process the alloy specimen in the amorphous state can be under-
during oxidation of the alloy in its amorphous state. In stood. As the alloy reaches the supercooled liquid state
order to analyse the depth profile in Fig. 6, it would be during oxidation (at temperatures 633 and 663 K) strong
interesting to get an estimate of the diffusion lengths of segregation of Cu is noticed in the oxide film (Fig. 6)
these elements using the known diffusivity data [2729] in which point to a change in the mechanism of oxidation of
this and other similar types of amorphous alloys. the alloy in its supercooled liquid state, i.e. the oxidation
Diffusion rates of Ni in amorphous Zr65Cu17.5Ni10Al7.5 mechanism is dominated by the inward diffusion of oxygen
are available in the literature [27, 28] while no diffusion than the back diffusion of Ni and Cu into the alloy [11, 14].
data for Cu diffusion in this alloy have so far been reported. During oxidation of the alloy the rate controlling process is
123
9042 J Mater Sci (2007) 42:90379044
the back diffusion of Cu and Ni so long as the alloy alloys in the literature [1113]. Sun et al. [11] during
remains in its amorphous state (which can be ascertained oxidation of amorphous Zr60Ni25Al15 observed the absence
from the known TTT diagram of the alloy [17, 18] and it of Ni in the oxide film formed on amorphous Zr65Ni25Al15
is most likely that Cu, being a slower diffuser than Ni, is after oxidation in air at temperatures below the glass
present closer to the oxide-alloy interface than Ni and thus transition temperature of the alloy. It was proposed [11]
gets easily oxidized at 633 and 663 K when the alloy that Ni was rejected from the oxide film and its back dif-
attains the supercooled liquid state after initial oxidation. fusion into the alloy was the rate controlling process. The
This is evident from the plots shown in Figs. 2 and 6. In the observations made in the present study also point to the fact
supercooled liquid state the oxide film formation is domi- that a similar mechanism appears to hold for oxidation of
nated by the inward diffusion of oxygen anions and out- the amorphous Zr65Cu17.5Ni10Al7.5 in the amorphous state
ward diffusion of cations of zirconium and copper. The as suggested by Sun et al. [11] for the oxidation of the
diffusion rates of smaller atom (Cu here) are, in fact, about amorphous Zr60Ni25Al15. It is noteworthy here that the
two orders of magnitude larger at temperatures of interest oxidation of the alloy in amorphous and the supercooled
in the supercooled liquid state (633 and 663 K) than that in liquid states was found to obey parabolic rate law with
the amorphous state (573 and 603 K) [2730]. This facil- different activation energies in the respective states [14].
itates the outward diffusion of Cu and its consequent oxi- The observations of significant amount of Cu in the oxide
dation by inwardly moving oxygen atoms. Moreover, the films formed after oxidation in the supercooled liquid state
larger amount of Cu (17.5 at.%) in the alloy than Ni (10 clearly points to a different oxidation process. It is sug-
at.%) and the proximity of the former to the oxide-alloy gested that during oxidation in the supercooled liquid state
interface after the initial oxidation till the alloy remained in the inward diffusion of oxygen becomes a dominant
the amorphous state are the favourable factors leading to and the rate controlling process for the oxidation of the
the oxidation of Cu and the corresponding suppression of alloy in the supercooled liquid state. Triwikantoro et al.
the oxidation of Ni in the supercooled liquid state of the [12] and Koester and Trwikantoro [13] have suggested a
alloy specimen. This explains the absence of Ni in depth similar mechanism for the oxidation of amorphous
profiles for oxidation of the alloy in the supercooled liquid Zr69.5Cu12Ni11Al7.5 in air at 633 K for 4 h. It was men-
state (Fig. 6). Some traces of Cu and Ni are seen (Figs. 2 tioned in their investigation [13] that the alloy had a glass
and 3) on the as-received surface at 573 and 603 K (cor- transition temperature of 635 K and the formation of some
responding to the amorphous state) which possibly arise quasicrystals was noticed in the oxidised alloy specimen. It
due to oxidation of some initial amount of Cu and Ni which is very likely that the alloy oxidation did not correspond to
was available just at the beginning of the oxidation process the oxidation in the fully amorphous state and perhaps due
and these elements get subsequently depleted in the oxide to this reason no rejection of Ni and Cu was observed in
layer and are possibly enriched near the oxide-alloy inter- their investigation and based on the results of their study it
face as the dominant rate controlling process for oxidation was suggested that the inward diffusion of oxygen was the
of the alloy in its amorphous state is their back-diffusion dominant process for the growth of the oxide film on
into the alloy. As a result no further presence of Cu or Ni is Zr69.5Cu12Ni11Al7.5. Therefore, the oxidation studies by
seen in the depth profiles of these elements in the oxide Sun et al. Zr60Ni25Al15 [11] and by Triwikantoro et al. [13]
film formed on the alloy surface in its amorphous state (i.e. on Zr69.5Cu12Ni11Al7.5 are available evidences in the lit-
at 573 and 603 K). erature showing different oxidation mechanisms for high
Therefore, the observed trends of the depth profiles in temperature air oxidation of a ZrCuNiAl amorphous
Fig. 6 point to the role of Cu and Ni during the oxidation of alloy depending on whether the alloy is fully amorphous
the amorphous Zr65Cu17.5Ni10Al7.5 in air and support our or not. The results of our oxidation investigations carried
previously proposed mechanism for oxidation of this alloy out in air in the temperature range 573663 K on
[14]. The change in diffusion mechanism at the glass Zr65Cu17.5Ni10Al7.5 also point to different type of oxidation
transition temperature may likely be associated with the processes for the alloy being in the amorphous or the
differences in the diffusion behaviour of smaller atoms (Cu supercooled liquid state during oxidation treatment.
and Ni) in the amorphous and the supercooled liquid states In order to get some insight into the nature of oxide
of the alloy [27, 28]. It is suggested from the above analysis films formed on specimens oxidised in the amorphous and
that during oxidation in the amorphous state both Zr and Al the supercooled liquid states, surface morphology of the
having the most negative values of the heat of formation oxide film was examined by SEM. Figures 8 and 9 repre-
get preferentially oxidised with the rejection of Ni and Cu sent surface morphologies obtained from specimens oxi-
from the oxide film. Oxidation mechanisms suggested for dised at two typical temperatures 603 K (corresponding to
this alloy are indeed consistent with the previously pro- the amorphous state) and 663 K (corresponding to the
posed mechanisms for air oxidation of similar type of supercooled liquid state). Figure 8 shows a relatively
123
J Mater Sci (2007) 42:90379044 9043
[17, 18]. The high ratio between the volume of oxide and
metal leads to a compressive stress in the oxide film
resulting in a partial cracking of the oxide film. This is
supported by the fact that the Pilling-Bedworth ratio of
CuO, Cu2O and ZrO2 have values 1.73, 1.67 and 1.57,
respectively [31]. These observations support our propo-
sition that the rate kinetics for oxidation at higher tem-
peratures (in supercooled liquid state) is dominated by
inward diffusion of oxygen anions and outward diffusion of
metal cations, especially of Zr and Cu [14].
Therefore, the observations reported in this study
support the view that different oxidation mechanisms are
operative for oxidation of the alloy Zr65Cu17.5Ni10Al7.5 at
low temperatures (corresponding to the amorphous state
Fig. 8 Scanning electron micrograph of the oxide film formed on the
amorphous Zr65Cu17.5Ni10Al7.5 after oxidation at 603 K for 3 h (alloy
of the alloy) and at high temperatures (corresponding to
specimen in the amorphous state) the supercooled liquid state of the alloy). It may, how-
ever, be argued that the change in mechanism could
purely be a kinetic effect not being related to the change
granular homogeneous distribution of oxidation products
of structural state of the alloy. Though this possibility
on the sample oxidised in its amorphous state at 603 K. On
seems to be rather unlikely as the proposition for change
the other hand sample oxidised at 663 K (in its supercooled
in the oxidation mechanism was made on the basis of
liquid state) showed a different morphology with some
results obtained by study of oxidation kinetics of the alloy
noticeable cracks in the oxide film. Such cracks have ear-
using the TGA [14] and the XPS analysis of the oxide
lier been reported in the oxide films formed during oxi-
films (present study), it may not be totally ruled out on
dation of ZrNi amorphous alloys in air [7]. In fact, the
the basis of the observations made in the present study.
presence of cracks in the oxide film formed at higher
Further careful investigations to study the segregation/
temperatures points to the presence of metallic species in
diffusion behaviour of Cu and Ni in the formation of
oxidised forms. Cracks may have developed by expansion
oxide films at higher temperatures with much smaller
due to oxidation of metallic components into oxides,
oxidation times (e.g. permissible oxidation time is only
especially by the oxidation of Cu in the alloy under Zr-rich
20 s at 633 K from the TTT diagram for the alloy
top layer. In fact, during the beginning of oxidation an
Zr65Cu17.5Ni10Al7.5 [17, 18] so that the sample remains
initial Zr-rich oxide layer is already formed on the alloy
fully in the amorphous state in contrast to the oxidation
specimen before it is transformed into its supercooled
time of 3 h used in the present study, which corresponds
liquid state after short annealing time which can be
to the supercooled liquid state of the alloy) are required in
ascertained from the known TTT diagram of the alloy
order to unambiguously prove this point.
Conclusions
123
9044 J Mater Sci (2007) 42:90379044
(2) No Ni and Cu were detected in the oxide film formed 7. Asami K, Kimura HM, Hashimoto K, Masumoto T (1995) Mat
at 573 and 603 K (corresponding to the amorphous Trans JIM 36:988
8. Kiene M, Strunskus T, Hasse G, Faupel F (1999) Mater Res Soc
state of the specimen) while appreciable amount of Symp Proc. Materials Research Society, Pittsburgh, PA, USA,
Cu (as CuO) was found in the oxide film formed at 554:167
633 and 663 K (corresponding to the supercooled 9. Sharma SK, Strunskus T, Ladebusch H, Faupel F (2001) Mater
liquid state). In the later case no Ni was detected due Sci Eng A 304306:747
10. Dhawan A, Raetzke K, Faupel F, Sharma SK (2001) Bull Mater
to strong segregation of Cu (Cu being present in much Sci 24:281
higher concentration in the alloy than Ni) and the 11. Sun X, Schneider S, Geyer U, Johnson WL, Nicolet M-A (1996)
proximity of Cu to the oxidealloy interface. J Mater Res 11:2738
(3) The above results are consistent with the previously 12. Triwikantoro, Toma D, Meuris M, Koster U (1999) J Non-Cryst
Solids 250252:719
proposed mechanism for the air oxidation of this alloy 13. Koester U, Triwikantoro D (2001) Mater Sci Forum 360362:29
suggesting that the rate controlling process during the 14. Dhawan A, Raetzke K, Faupel F, Sharma SK, Phys Stat Sol (a)
oxidation of the alloy at low temperatures (in the 199 No. 3 (2003) 431438
amorphous state) is the back diffusion of Ni and Cu 15. Tam CY, Shek CH (2005) J Mater Res 20:1396
16. Kai W, Hsieh HH, Nieh TG, Kawamura Y (2002) Intermetallics
while the oxidation at higher temperatures (in the 10:1265
supercooled liquid state) is dominated by the inward 17. Busch R (Private Communication)
diffusion of oxygen and the outward diffusion of 18. Knorr K (1999) In: Selbstdiffusion in Metallischen Massivg-
zirconium and copper. laesern. Ph. D Thesis, University of Muenster, Muenster,
Germany
19. Wanger CD, Riggs WM, Davies LE, Moulder JF, Muilenberg GE
Acknowledgements The financial support for this work from the (1979) Handbook of X-ray photoelectron spectroscopy. Perkin-
Board of research in Nuclear Sciences (BRNS), Department of Elmer, Eden Prairie, MN
Atomic Energy (DAE), Government of India under the DAE/BRNS 20. Walz B, Oelhafen P, Guentherodt H-J, Baiker A (1989) Appl Surf
research Project (Grants No. 99/37/25/BRNS) is gratefully acknowl- Sci 37:337
edged. Thanks are due to Dr. K Raetzke for a critical reading of the 21. Satoh H, Nakane H, Adachi H (1996) Appl Surf Sci 94/95:247
manuscript and for many useful suggestions. The help provided by 22. Betz G, Wehner GK (1983) Sputtering by particle bombardment
Ms Jugrita Zekonyte and Mr. Stefan Rehders during XPS measure- II, vol 52. In: Behriech R (ed) Springer Verlag, Berlin, p 11
ment is gratefully acknowledged. Thanks are due to Dr. M. Sunda- 23. Malherbe JB, Hofmann S, Sanz JM (1986) Appl Surf Sci 27:355
raraman, Dr. J. G. Shah and Ms. M. S. Tawade from BARC, Mumbai 24. Brundle CR, Chuang TJ, Wandelt K (1977) Surf Sci 68:459
for their help in recording SEM micrographs from oxidised alloy 25. Wandelt K (1982) Surf Sci Rep 2:75
specimens. 26. CRC handbook of chemistry and physics (199596) 76th edn.
Lide DR (ed) CRC Press, Boca Raton, pp 3447
27. Knorr K, Macht M-P, Mehrer H (1999) Bulk metallic glasses. In:
Johnson WL, Liu CT, Inoue A (ed) MRS symposia proceedings
References No. 554. Materials Research Society, Pittsburgh, p 269
28. Faupel F, Frank W, Macht M-P, Mehrer H, Naundorf V, Raetzke
1. Zhang T, Inoue A, Masumoto T (1991) Mater Trans JIM 32:1005 K, Schober HR, Sharma SK, Teichler H (2003) Revi Modern
2. Johnson WL (1996) Curr Opin Solid State Mater Sci 1:383 Phys 75:237
3. Peker A, Johnson WL (1993) Appl Phys Lett 63:2342 29. Hahn H, Averback RS, Shyu H-M (1988) J Less-Common Met
4. Hashimoto K (1983) Amorphous metallic alloys. In: Luborsky FE 140:345
(ed) Butterworths, London, p 471 30. Sharma SK, Banerjee S, Kuldeep, Jain AK (1989) J Mater Res
5. Naka M, Hashimoto K, Masumoto T (1976) Corrosion 32:146 4:603
6. Dey GK, Savalia RT, Sharma SK, Kulkarni SK (1989) Corros Sci 31. Bradford SA (1990) In: ASM handbook, vol 13. ASM Interna-
29:823 tional, Materials Park, OH, p 64
123
J Mater Sci (2007) 42:90459056
DOI 10.1007/s10853-007-1824-2
Received: 25 February 2007 / Accepted: 4 May 2007 / Published online: 20 July 2007
Springer Science+Business Media, LLC 2007
123
9046 J Mater Sci (2007) 42:90459056
extenders, 1,4-butanediol, are replaced with glycerol that is Table 1 Composition of linear and cross-linked SMPU
used as a connecting spot with other polyurethane chains. Sample Reactant, mol ratio wt.% of
The target polyurethane has a network-structure compared code glycerol
to the randomly cross-linked one, and a substantial MDI PTMG BD Glycerol PEG-200
improvement in mechanical and shape memory properties L-0 5.0 2.0 3.0
is achieved. In this investigation, PEG-200 is selected as the C1-5 5.0 2.0 2.775 0.15 5
spacer because of its flexibility and short chain length, and C1-10 5.0 2.0 2.55 0.3 10
the mechanical and shape memory properties are compared C1-15 5.0 2.0 2.325 0.45 15
with linear or randomly cross-linked ones. C2-5 5.0 2.0 2.85 0.15 0.15 5
C2-10 5.0 2.0 2.7 0.3 0.3 10
C2-15 5.0 2.0 2.55 0.45 0.45 15
Experimental
C2-20 5.0 2.0 2.4 0.6 0.6 20
C2-25 5.0 2.0 2.25 0.75 0.75 25
Materials
123
J Mater Sci (2007) 42:90459056 9047
(b)
O O O O
CNH CH2 NHCO CH2CH2CH2CH2O CNH CH2 NHCOCH2CHCH2O
n
O
PU chain
(c)
O O O O
CNH CH2 NHCO CH2CH2CH2CH2O CNH CH2 NHCOCH2CHCH2O
n
O
MDI + PEG-200
O O O O O
CNH CH2 NHCO CH2CH2CH2CH2O CNH CH2 NHCOCH2CHCH2O
n
400 m1 resolution, 25 scans, and a scan speed of 2 shrink at 20 C below the Tm for 1800 s. The shape
103 m/s. X-ray diffraction spectra were scanned under the recovery rate was calculated by Eq. (4). L2 was measured
conditions of 2h = 540 and 0.016/second by a wide- after the stretched specimen had been incubated at 20 C
angle X-ray diffractormeter (WAXD, Rigaku Rint 2000, above the Tm for 600 s and cooled at 20 C below the Tm.
CuKa, 40,000 V, and 0.03 A). A differential scanning The shape memory test was repeated three times.
calorimeter (DSC-2010, TA instrument) was used for the
heating and cooling scan at a rate of 0.16 C/s between Shape retention rate L1 Lo 100=Lo % 3
100 C and 250 C. The tensile mechanical properties
Shape recovery rate 2Lo L2 100=Lo % 4
were measured by a Universal Testing Machine (LR50K,
Lloyd instrument, UK) according to ASTM D638 using
0.025 m gauge length, 1.6 105 m/s crosshead speed, and
a 2500 N load cell. The dynamic mechanical properties
Results and discussion
were measured by a dynamic mechanical analyzer (DMA-
2980, TA instrument). The storage modulus and loss tan-
Synthesis
gent (tan d) were scanned between 80 C and 65 C at a
heating rate of 0.05 C/s, and 1 Hz. The cross-link density
Previously, the cross-linking method was dependent on the
was calculated according to the FloryRehner equation for
three-way linker that connected PU chains in a random
a specimen (0.025 0.025 0.001 m) swelled in toluene,
manner. Connection of the nearby PU chains by a flexible
a process for which further explanation will appear in the
cross-linker and control of the cross-link density may
results section.
improve the mechanical and shape memory properties due
Shape memory test to the resultant network structure. In this investigation,
some of BD, a chain-extender, is replaced with glycerol
The UTM was also used for the cyclic shape memory test. and the extra hydroxyl group of glycerol is used as the
Lo was the original length of the specimen. The shape cross-linking point. The extra hydroxyl group is activated
retention rate was calculated by Eq. (3). L1 was measured by MDI, and PEG-200 reacts with the MDI moiety on each
after the specimen had been drawn 100% at 20 C above chain. Because a PU chain is connected side by side at the
the Tm of the soft segment for 300 s, and then allowed to various sites along the PU chain, the overall strength can be
123
9048 J Mater Sci (2007) 42:90459056
IR analysis
O O
128
130
126
129
124
128 122
0 100 200 300 400 500 600 700 0 1000 2000 3000 4000
Time (s) Time (s)
Fig. 3 (a) Time course of the phthalic anhydride hydrolysis. (b) Titration of the periodically sampled reaction mixture after glycerol addition
123
J Mater Sci (2007) 42:90459056 9049
Table 2 Cross-link density of SMPU by swelling experiment The ratio, ebonded/efree, is close to 1.0 [18]. The degree of
a 3 b phase separation (DPS) and the degree of phase mixing
Sample code q (kg/m ) Q vc2 3 d
10 n
(DPM) are calculated from the following equations using
L-0 932 the R value of Eq. (7) [18]. The combined result is
C1-5 999 2.620 0.276 0.547 summarized in Table 3.
C1-10 895 2.516 0.284 0.581
C1-15 975 2.364 0.297 0.632 Cbonded R
DPS 8
C2-5 893 2.072 0.326 0.784 Cbonded Cfree R 1
C2-10 934 1.832 0.353 0.954
C2-15 968 1.637 0.379 1.139 DPM 1 DPS 9
C2-20 987 1.380 0.420 1.477
As the glycerol content increases, R decreases from 1.15
C2-25 965 1.176 0.460 1.866
(L-0) to 1.02 (C1-15) and 0.97 (C2-25). In the C1 series, R
a
Density; bDegree of swelling; cVolume fraction of polymer in decreases from 1.08 (C1-5) to 1.02 (C1-15). In the C2
swollen mass; dCross-link density
series, R decreases from 1.13 (C2-5) to 0.97 (C2-25).
Hydrogen bonding becomes weaker as the glycerol content
The peak around 1730 cm1 increases slightly with the increases. The change in phase separation is more obvious if
increase of glycerol content for both C1 and C2 series, the DPS is compared. The DPS decreases from 53.4% (L-0)
while the peak around 1700 cm1 decreases (Fig. 4). Based to 50.5% (C1-15) and 49.4% (C2-25). When the DPS at the
on IR results, the shift from bonded carbonyl to free car- same glycerol content is compared, no difference is found:
bonyl goes on as the glycerol content increases for both C1 52.1% (C1-5) vs. 52.9% (C2-5), and 50.5% (C1-15) vs.
and C2 series. Therefore, it seems that glycerol cross- 50.9% (C2-15). Now, it is clear from the IR experiment that
linking reduces the hydrogen bonding, and disrupts hard phase separation for both C1 and C2 series is decreased by
segment interaction. glycerol cross-linking and the difference in phase separa-
The peak shift can be used in quantifying the phase tion between them at the same glycerol content is minimal.
separation of hard and soft segments. Phase separation can
be also calculated from the IR spectra by comparing the XRD analysis
peak intensity ratio of carbonyl groups at 1703 cm1 and
1733 cm1 [17]. The parameter indicating phase separa- The X-ray deflection peak of PU, irrespective of glycerol
tion, hydrogen bonding index (R), is calculated by Eq. (7). content, is observed at 2Q = 19.5 [7]. Linear PU (L-0)
Cbonded ebonded A1703 shows a very high peak intensity compared to cross-linked
R 7 PUs, and the peak intensity decreases with any increase in
Cfree efree A1733
glycerol content (Fig. 5). In line with the IR data, hydrogen
In Eq. (7), A is absorbance, C is concentration, ebonded is bonding between hard segments is unaffected by cross-
the extinction coefficient of the peak at 1703 cm1, and linking in L-0, and the ordered hard segment alignment
efree is the extinction coefficient of the peak at 1733 cm1. increases the peak intensity. However, glycerol cross-linking
(a)
(b)
Transmittance (%)
Transmittance (%)
C2-5
L-0 C2-10
C1-5 C2-15
C1-10 C2-20
C1-15 C2-25
1800 1750 1700 1650 1600 1800 1750 1700 1650 1600
Wave number (cm-1) Wave number (cm-1)
Fig. 4 Infrared spectra of (a) L-0 and C1 series, and (b) C2 series
123
9050 J Mater Sci (2007) 42:90459056
Table 3 Comparison of thermal properties of SMPU Table 4 Degree of phase separation (DPS) and degree of phase
a mixing (DPM) of SMPU
Sample code R (A1703/A1733) DPS
a
Sample code Tg (C) Tm (C) DHm (J/kg) DHc (J/kg)
L-0 1.15 53.4
C1-5 1.08 52.1 L-0 52.47 10.00 5050 3550
C1-10 1.04 51.1 C1-5 50.72 11.35 1190 520
C1-15 1.02 50.5 C1-10 50.68 12.35 1180 110
C2-5 1.13 52.9 C1-15 49.96 11.57 1060 500
C2-10 1.07 51.6 C2-5 46.69 18.08 4040 420
C2-15 1.04 50.9 C2-10 42.99 20.35 3110 370
C2-20 1.01 50.4 C2-15 41.99 18.65 860 330
C2-25 0.97 49.4 C2-20 40.04 18.85 840 420
a C2-25 38.70 20.21 280 180
A1703: absorbance at 1703 cm1, and A1733: absorbance at 1733 cm1
a
DHc: heat of crystallization
disrupts hydrogen bonding and hard segment alignment, both C1 and C2 series. C2 series also exhibit higher Tm at
and reduces the peak intensity in proportion to the glycerol the same glycerol content: 11.35 C (C1-5) vs. 18.08 C
content. In the extreme case of C2-25, the peak intensity is (C2-5) and 11.57 C (C1-15) vs. 18.6 C (C2-15). Similar
too reduced to be considered as a peak. In line with the to the IR and XRD results, the network structure of C2
DPS data, it is also very difficult to differentiate C1 and C2 series is more influential in raising both Tg and Tm than the
series simply by XRD spectra. randomly cross-linked C1 series. It is very interesting that
Tm did not increase after cross-linking although the glyc-
Thermal analysis erol content is raised for both C1 and C2 series. It seems
that the cross-linking density of hard segment does not
The soft segment Tg is detected by the more sensitive DMA affect the soft segment melting because soft segment gets
analysis, because it is very hard to detect Tg by DSC due to enough energy for melting at Tm, a relatively high tem-
the very low temperature range and weak signal. However, perature compared to Tg. In DSC heating scans, L-0 shows
it is not impossible to detect Tm of the soft segment by DSC. a large endothermic peak around Tm, and the endothermic
The combined thermal analysis data are summarized in peak shrinks as the glycerol content increases for both C1
Table 4. Tg is very low for L-0 (52.47 C) and increases and C2 series (Fig. 6). The peak area can be quantitatively
with cross-linking for both C1 and C2 series. C2 series compared by heat of melting (D Hm) in Table 4. In
shows a higher Tg for a given glycerol content: 50.72 C accordance with the peak shrinkage with cross-linking in
(C1-5) vs. 46.69 C (C2-5) and 49.96 C (C1-15) vs. Fig. 6, DHm decreases significantly from 5050 J/kg to
41.99 C (C2-15). Tg increases up to 38.70 C in 1060 J/kg (C1-15) and 280 J/kg (C2-25). The restricted
the case of C2-25. Tm of L-0 (10.00 C) is quite high soft segment melting by glycerol cross-linking is respon-
compared to Tg and also increases with cross-linking for sible for the decrease of DHm. The heat of crystallization
7000
(a) (b) 7000
6000
L-0 6000 C2-5
C1 -5
C2-10
5000 C1 -10 5000
C1 -15 C2-15
C2-20
Intensity
Intensity
3000 3000
2000 2000
1000 1000
0 0
10 15 20 25 30 35 10 15 20 25 30 35
2 (deg ree) 2 (degree)
Fig. 5 X-ray diffraction spectra of (a) L-0 and C1 series, and (b) C2 series
123
J Mater Sci (2007) 42:90459056 9051
(a) (b)
Endothermic
E n d othermic
L-0 C2-5
C2-10
C1-5
C2-15
C1-10
C2-20
C1-15
C2-25
Fig. 6 DSC heating scans of (a) L-0 and C1 series, and (b) C2 series
C1 series C1 series
maximum stress, (b) strain at 25
C2 series 12 C2 series
break, (c) tensile modulus, and 20
(d) yield stress 9
15
6
10
3
5
0 0
0 10 20 30 0 5 10 15 20 25 30
Glycerol contents (wt%) Glycerol contents (wt%)
(b) 1500 (d) 5
Strain at break (% )
1200 4
900 3
600 2
(DHc) obtained from the cooling scan also supports that increase of glycerol content. Yield stress of C2 series is
cross-linking restricts soft segment movement (Table 4). about 1.5 times higher than that of C1 series at 5 wt%
glycerol, and decreases at higher wt% of glycerol. It is
Tensile mechanical properties remarkable that changing the cross-linking method makes
a big difference in maximum stress while maintaining a
Tensile mechanical properties averaged from five tests are similar modulus, and the strain at break goes up to 1200%
compared in Fig. 7. Maximum stress of cross-linked PU even after cross-linking. The tensile mechanical behavior
(C2-10) steeply increases up to 800% compared to L-0, is also compared in a stressstrain curve (Fig. 8). C1 and
although further increase of cross-linked content reduces C2 series show similar slopes up to 600% strain, but a
the maximum stress. Maximum stress of the C2 series is sudden increase in slope is observed for C2 series, espe-
300% higher than that of C1 series at 10 wt% cross- cially C2-5, C2-10, and C2-15. Because the behavior of
linking. The strain at break is similar for both C1 and C2 C2 series is similar to that of natural rubber, the four
series, and increases with the increase of cross-linking polymers, L-0, C1-10, C2-10, and natural rubber (NR),
content. Tensile modulus is maximized at 5 wt% glycerol are plotted in Fig. 8(c). Notably, C2-10 is different from
for both C1 and C2 series, and decreases with further other PUs and is close to natural rubber in overall shape
123
9052 J Mater Sci (2007) 42:90459056
(a) 25
L-0
(b) 25
C2-5
C1-5 C2-10
C1-10 C2-15
20 C1-15
20
C2-20
C2-25
Stress (MPa)
Stress (MPa)
15 15
10 10
5 5
0 0
0 200 400 600 800 1000 1200 1400 0 200 400 600 800 1000 1200 1400
Strain (%) Strain (%)
(c) 25
L-0
C1-10
20 C2-10
NR
Stress (MPa)
15
10
0
0 200 400 600 800 1000 1200 1400
Strain (% )
Fig. 8 Stressstrain curve of (a) C1 series, (b) C2 series, and (c) SMPU and natural rubber (NR)
except the higher stress and strain. The steep rise of stress crystallization. But C1 series, connected in a random
in natural rubber originates from the strain-induced crys- manner, is unable to form the aligned structure and shows
tallization, and the same reasoning can be applied to the no sign of strain-induced crystallization. Previous exper-
explanation of C2 series behavior. C2 series polymer iments on SMPU have demonstrated that cross-linking of
aligns along the chain length because each polymer chain hard segment can increase maximum stress but strain is
is connected by a PEG-200 spacer at its side and becomes sacrificed instead. But the adoption of a flexible spacer
stronger when stretched due to the strain-induced and cross-linking at its side in C2 series elevates the
(b) 2500
(a) 2000 L-0 C2 -5
C1-5 C2 -1 0
2000
C1-10 C2 -1 5
1500
C2 -2 0
C1-15
E' (MPa)
1500 C2 -2 5
E' (MPa)
1000
1000
500 500
0
0 -80 -60 -40 -20 0 20
-80 -60 -40 -20 0 20
Temperature (C)
Temperature (C)
Fig. 9 Storage modulus vs. temperature profile of (a) L-0 and C1 series, and (b) C2 series
123
J Mater Sci (2007) 42:90459056 9053
T an
0.3 0.3
0.2 0.2
0.1 0.1
0 0
-80 -60 -40 -20 0 20 40 60 -80 -30 20 70
Temperature (C) Temperature (C)
Fig. 10 Loss tangents vs. temperature profile of (a) L-0 and C1 series, and (b) C2 series
Strees (MPa)
2
2
1
1
0 0
0 20 40 60 80 100 0 20 40 60 80 100
Strain (%) Strain (%)
(b) 4 (d) 4
Cycle 1 Cycle 1
Cycle 2 Cycle 2
3 Cycle 3 3 Cycle 3
Strees (MPa)
Strees (MPa)
Cycle 4 Cycle 4
2 2
1 1
0 0
0 20 40 60 80 100 0 20 40 60 80 100
Strain (%) Strain (%)
123
9054 J Mater Sci (2007) 42:90459056
Stress (MPa)
Stress (MPa)
2 2
1 1
0 0
0 20 40 60 80 100 0 20 40 60 80 100
Strain (%) Strain (%)
Stress (MPa)
Stress (MPa)
2 2
1 1
0 0
0 20 40 60 80 100 0 20 40 60 80 100
Strain (%) Strain (%)
(e) 3 Cycle 1
Cycle 2
Cycle 3
Stress (MPa)
Cycle 4
2
0
0 20 40 60 80 100
Strain (%)
such type of shape memory polymer is called Tm-type- analytical and mechanical results, the shape memory test
shape memory polymer to differentiate it from Tg-type one. also demonstrates that the alignment of polymer chains and
This SMPU belongs to Tm-type, because Tg is too low and the cross-linking at the side with a flexible spacer play a
it is hard to control the temperature around Tg [2, 12, 13]. decisive role in improving the overall properties compared
As compared in Table 5, shape recovery process is over in to the randomly cross-linked C1 series and linear L-0.
900 s at Tm + 20 C. Shape recovery rate increases from Therefore, the finding of C2 series is very encouraging and
83.7% for L-0 to 89.3% for C1-5 and 96.8% for C2-10. C2 argues the importance of polymer chain alignment and
series exhibits a higher shape recovery rate than C1 series cross-linking method. More variations of SMPU based on
at the same glycerol content: 89.3% for C1-5 vs. 94.8% for C2 series can be produced in near future.
C2-5, and 85.1% for C1-15 vs. 94.0% for C2-15 (Table 5).
The shape recovery rate of C2-10, the best one of C2 series, Table 5 Comparison of shape recovery speed of SMPU
does not fall below 93% even after cycle 4. Although many
Sample code Shape recovery rate (%)
shape memory polymers, either polyester or polyurethane,
have been tested by us, shape recovery rate is usually less 300 s 600 s Maximum
than 95% and decreases to below 90% after cycle 3. L-0 80.2 83.1 83.7
Because the shape recovering force of C2 series is so C1-5 80.0 82.7 89.3
strong, shape retention rate decreases compared to C1 C1-10 81.0 85.8 88.7
series (Fig. 13). Shape recovery pictures of L-0, C1-10, and C1-15 80.4 82.3 85.1
C2-10 are shown in Fig. 14. After each specimen is rolled, C2-5 88.8 90.5 94.8
the time required for unfolding is compared. C1-10 reduces
C2-10 90.7 93.7 96.8
the unfolding time from 200 second of L-0 to 121 second,
C2-15 91.2 92.5 94.0
and C2-10 achieves the unfolding in just 66 second.
C2-20 88.0 90.8 90.8
Therefore, C2-10 recovers the original shape in one third of
C2-25 89.9 91.3 91.5
the time necessary for L-0. Along with all the above
123
J Mater Sci (2007) 42:90459056 9055
Shape recovery (% )
90 90
85 85
80 80
75 75
C1 series C1 series
70 70 C2 series
C2 series
65 65
0 5 10 15 20 25 30 0 5 10 15 20 25 30
Glycerol content (wt%) Glycerol content (wt%)
Fig. 13 (a) Shape retention and (b) shape recovery vs. glycerol content profile of SMPU
Fig. 14 Photographic comparison of shape recovery speed of (a) L-0, (b) C1-10, and C2-10
123
9056 J Mater Sci (2007) 42:90459056
Acknowledgments The authors of this paper would like to thank 7. Takahashi T, Hayashi N, Hayashi S (1996) J Appl Polym Sci
the Korea Science and Engineering Foundation (KOSEF) for spon- 60:1061
soring this research through the SRC/ERC Program of MOST/KOSEF 8. Lee BS, Chun BC, Chung YC, Sul KI, Cho JW (2001) Macro-
(R11-2005-065). molecules 34:6431
9. Ota S (1981) Radia Phys Chem 18:81
10. Chen LW, Lin JR (1998) J Appl Polym Sci 69:1563
11. Chen LW, Lin JR (1998) J Appl Polym Sci 69:1575
References
12. Hu J, Yang Z, Yeung L, Ji F, Liu Y (2005) Polym Int 54:854
13. Yang Z, Hu J, Liu Y, Yeung L (2006) Mater Chem Phys 98:368
1. Yang JH, Chun BC, Chung YC, Cho JH (2003) Polymer 44:3251 14. Carey MA, Wellons SL, Elder DK (1984) J Cellular Plastic 20:42
2. Li F, Chen Y, Zhu W, Zhang X, Xu M (1998) Polymer 39:6929 15. Petrovic ZS, Javni I, Divjakovic V (1998) J Polym Sci B: Polym
3. Liang C, Rogers CA, Malafeew E (1997) J Intell Mater Syst Phys 36:221
Struct 8:380 16. Hur MK, Kwak JM, Hur T (1996) Polymer (Korea) 20:392
4. Chun BC, Cha SH, Chung YC, Cho JW (2002) J Appl Polym Sci 17. Coleman MM, Lee KH, Skrovanek DJ, Painter PC (1986) Mac-
83:27 romolecules 19:2149
5. Kim BK, Lee SY, Xu M (1996) Polymer 37:5781 18. Seymour RW, Estes GM, Cooper SL (1970) Macromolecules
6. Tobushi H, Hara H, Yamada E, Hayashi S (1996) Smart Mater 3:579
Struct 5:483
123
J Mater Sci (2007) 42:90579062
DOI 10.1007/s10853-007-1833-1
Received: 16 January 2007 / Accepted: 8 May 2007 / Published online: 20 July 2007
Springer Science+Business Media, LLC 2007
Abstract Microwave radiation simplifies synthesis controlling size, and in some cases shape, it would be
methods by reducing reaction times, requiring fewer desirable to quickly and uniformly synthesize nanomate-
materials, and also controlling reaction processes. We have rials. One approach to obtaining this control is the use of
successfully synthesized nanoparticles of iron oxide and microwave systems in synthesis. Besides the obvious
zinc oxide coated on zeolite A using microwaves. The advantage of reducing reaction times from days to minutes,
radiation assisted in displacing either ferrous or zinc ions the nucleation processes are better controlled under
from the pre-loaded zeolite network and increasing microwave conditions since the reaction is driven by the
reaction speed with solution at the interface. Products were quick heat and pressure produced by the microwave system
characterized by TEM, XRD, VSM, ICP-AES, and [13]. The resulting physical characteristics of the nano-
fluorescence. We demonstrate the ability of using cation- particles are controllable and more reproducible [4]. For
exchanged zeolites as microreactors to bias reactions onto example, microwave methods now exist to synthesize
the zeolite surface. Efficient structure-directed surface nanoparticles with high degrees of monodispersity [5]. An
reactions are a potential route to making unique supported added advantage is that the amounts of reactants and sol-
nanomaterials for applications such as sensors, environ- vents used in the synthesis are reduced, providing a
mental remediation, and chemical catalysis. greener synthesis route with fewer cleanups required.
Of particular importance to this report is that microwave
synthesis of submicron-sized magnetite powders has been
shown [4]. They reacted both ferrous and ferric salts
together and the ferrous salt alone with sodium hydroxide
Introduction in teflon digestion vessels in a high-powered (1,200 W)
microwave. By using only two chemicals in the reaction,
Many of the unique phenomena that nanomaterials exhibit the conversion has few byproducts, resulting in a product
are highly dependent not only on the chemical constituents, with little impurity. The complement of microwave power
but also on the size and shape of the nanomaterial. Given facilitates rapid synthesis and relatively consistent heating
that controlling the rate of nucleation is the key to within the vessel.
Solution-based nanoparticles are not always the optimal
path of synthesis since the functionality of the particles can
J. A. Heuser (&) W. U. Spendel A. N. Pisarenko be realized only after removal of the solvent and placement
G. E. Pacey
of the particles in another media. For instance, some
Miami University Center for Nanotechnology, Department
of Chemistry and Biochemistry, Miami University, Oxford, nanoparticle metal oxides are important in the development
OH 45056, USA of industrial catalysts [6]. It would be ideal to synthesize
e-mail: heuserja@muohio.edu the desired nanoparticles within the proper medium,
thereby eliminating the transfer step. One widely-used
C. Yu M. J. Pechan
Department of Physics, Miami University, Oxford, OH 45056, catalyst is zeolite A, a white powder of aluminosilicate
USA networks. Oliveria et al. used this zeolite not as a catalyst,
123
9058 J Mater Sci (2007) 42:90579062
but rather as an ion-extracting agent in water [7]. The tometry (zinc synthesis only), and inductively coupled
group used solution-based synthesis to produce a magnetic plasma (ICP) emission spectroscopy. The specifications are
zeolite by incorporating magnetite (Fe3O4) about the zeo- as follow. TEM pictures were obtained using a Zeiss 10C
lite particles. The zeolite framework remained open to operated at 100 KeV with the images being captured on
extract cations from the solution and the attached magnetite Kodak 4489 film and scanned in on an Agfa Duoscan at
allowed powder withdrawal by using a simple magnet. 1000PPI. XRD patterns were obtained using a Scintag X1
As an extension of prior work, we are reporting a powder X-ray diffractometer. VSM measurements were
microwave synthesis method to yield submicron-sized made on a home-assembled apparatus utilizing the Faraday
magnetite and zinc oxide particles coated on zeolite A. Law to detect induced voltage in the pick-up coils by the
Microwave radiation served to both rapidly displace ions vibration of the measured magnetic sample in the magnetic
from the zeolite matrix and position them at the zeolite/ field. The frequency used for the vibrator is about 48 Hz
solution interface. After preloading a zeolite with selected and the system was calibrated using a standard Pd sample
cations, they were displaced from the framework to the with known moment. This VSM system has a moment of
solution interface where the appropriate chemical reactions sensitivity of about 10 micro-emu. Fluorescence data was
took place. The significance is our method of employing obtained on a Horiba Jobin Yvon Fluorolog-3 FL3-22
microwaves to a chemical system yielding nanoparticles spectrophotometer. ICP emission data was obtained on a
coated directly on the support matrix with little solution Varian Liberty 150 spectrometer.
reaction competition. Minimal preparation and reaction Glycerin (60% v/v) was added to a ferrous ion/zeolite/
time enhance the success of this method. hydroxide solution to increase the solution viscosity and
affect magnetite particle nucleation and growth. It was
hypothesized that different solvent properties would affect
Experimental procedure formation of the iron oxide particles. Microwave condi-
tions and ion concentrations remained unchanged from the
Chemicals were used without additional purification unless previous experiments.
otherwise noted. Ferrous sulfate, zinc nitrate, and sodium Since zeolite is often used as a cation-exchange med-
hydroxide were purchased from Aldrich. Valfor 100 ium, the exchange capability was tested using calcium ions.
zeolite was donated by The PQ Corporation. After the products were obtained, filtered, and rinsed, sat-
Zeolite was first preloaded with the desired cation by urated calcium solutions were added. They were allowed to
stirring 100 mL of 1 M ferrous sulfate or zinc nitrate equilibrate for at least 24 h. The filtered solids were dried
solution with 8 g of zeolite. The suspension was allowed to under ambient conditions and digested with hydrochloric
equilibrate for at least 24 h to ion-exchange either iron or acid. The resulting solution was analyzed using ICP
zinc, respectively. After sufficient mixing the slurry was emission spectroscopy.
filtered and thoroughly washed with deionized water, then
diluted to 30 mL. Once the zeolite was loaded and washed
it was ready for reaction in the microwave vessel. Results and discussion
Sodium hydroxide solution was used for both syntheses
as the other reactant. A 4.8 mL aliquot of 0.25 M NaOH The results obtained clearly demonstrate the hypothesized
(reaction concentration of 0.2 M) was combined with interface-directed nanoparticle syntheses. Both iron oxide
1.2 mL of loaded zeolite suspension in a glass reaction and zinc oxide nanoparticles were synthesized about the
vessel and capped with a pressure-sensitive lid. Identical zeolite surface after exposure to microwave radiation.
samples were also left to sit overnight without being Samples not exposed to microwave radiation took hours to
exposed to radiation to observe unassisted reactions. react and the particles were of micrometer size. The reac-
Solution reactions without zeolite present were synthesized tion did not appear to go to completion after 24 h. How-
by combining 1.2 mL of either 0.2 M ferrous sulfate or ever, under microwave radiation conditions, small, more
zinc nitrate solutions with 4.8 mL of 0.25 M NaOH. The desirable nanoparticles were synthesized rapidly.
reaction tubes were exposed to microwave irradiation in a A representative TEM picture of the iron oxide/zeolite
CEM Discover unit using the following parameters: product is given by Fig. 1. Glycerin was added to the same
Power300 W; Ramp time10 s; Hold time10 min; reaction mixture as the typical ferrous ion/zeolite mix. A
Max temperature150 C; Max pressure200 PSI; TEM image (not shown due to rapid decomposition under the
Stirring and powerMaxOn. electron beam) of the glycerin samples shows a significantly
Products were investigated with transmission electron narrower size distribution relative to the water-only syn-
microscopy (TEM), X-ray diffraction (XRD), vibrating theses. Particles about the zeolite were between one to three
sample magnetometry (VSM), fluorescence spectropho- nanometers and showed a high degree of uniformity.
123
J Mater Sci (2007) 42:90579062 9059
123
9060 J Mater Sci (2007) 42:90579062
123
J Mater Sci (2007) 42:90579062 9061
on the zeolite. Had the solution-synthesized zinc oxide bound to the zeolite. Despite this, the numbers confirm the
reaction dominated we would not have seen the predomi- majority of the metal oxide reacts at the zeolite/solution
nant spectral fine structure between 480 nm and 550 nm; interface.
the broad fluorescence between 510 nm and 620 nm would
have been more prominent.
The exchange capacities of the following solids were Conclusions
investigated: magnetite on zeolite, zinc oxide on zeolite,
microwaved zeolite, and zeolite. Results from the acid We have demonstrated a zeolite interface-directed metal
digestion and subsequent ICP analysis indicate that no oxide nanoparticle synthesis method potentially useful in
exchange capacity was lost when compared to the reagent making catalysts, sensors, and a new approach to making
zeolite sample, given in Table 1. An interesting point is nanostructured interfaces. These experiments show the
that the microwaved zeolite control had a somewhat higher ability of using zeolite ion-exchange as a means of deliv-
degree of exchange capacity, possibly attributed to the ering cations in reactions where the zeolite serves as a
radiation increasing cavity activity compared with the microreactor. Microwave radiation increases the ion ex-
untreated commercial product. The magnetite also shows a change rate in a controlled manner to displace ions from
higher level of exchange, but this value may be enhanced the zeolite network to form magnetite or zinc oxide by
by calcium adsorption on the magnetite. Magnetic zeolite reactions of simple ferrous or zinc salts with hydroxide.
has previously been shown to be a more efficient cation Cation delivery from the zeolite adds the ability to control
extractor than zeolite alone [7]. nanoparticle growth and nucleation at the zeolite interface.
Also listed in Table 1 are the theoretical maximum Nucleation control at the interface and solvent effects
percentages of product expected based on the ICP values provide a pathway to synthesize new nanostructured
obtained and the 5.5 milliequivalent per gram exchange interfaces.
capacity of zeolite A [8]. The data we obtained are slightly Future work in our laboratory is directed towards
lower than the maximum values mainly due to inefficient establishing microwave solution reaction conditions to
exchange and incomplete cation reaction at the zeolite control nanoparticle shape and size. Nanoparticle micro-
interface. Some metal oxide product formed may not have wave synthesis provides a significant opportunity for time-
efficient (less than 10 min) and green synthesis of unique
Table 1 Calcium exchange data and oxide percent values, both interfaces with potential applications in catalysis, chemical
theoretical and experimental analysis, electronic devices, and sensors.
meq/g Ca2+ % Oxide Theoret. %
123
9062 J Mater Sci (2007) 42:90579062
3. He R, Qian X-F, Yin J, Xi H-A, Bian L-J, Zhu Z-K (2003) Colloid 6. Glaspell G, Fuoco L, El-Shall MS (2005) J Phys Chem B
Surf A 220:151 109:17350
4. Khollam YB, Dhage SR, Potdar HS, Deshpande SB, Bakare PP, 7. Oliveira LCA, Petkowicz DI, Smaniotto A, Pergher SBC (2004)
Kulkarni SD, Date SK (2002) Mater Lett 56:571 Water Res 38:3699
5. Gerbec JA, Magana D, Washington A, Strouse GF (2005) J Am 8. Breck DW (1974) Zeolite molecular sieves, structure, chemistry,
Chem Soc 127:15791 and use. John Wiley & Sons
123
J Mater Sci (2007) 42:90639069
DOI 10.1007/s10853-007-1692-9
Received: 24 November 2006 / Accepted: 13 March 2007 / Published online: 20 July 2007
Springer Science+Business Media, LLC 2007
Abstract The effect of Cr on the microstructure, tensile regard, alloying chemistry may be used to achieve
properties and creep resistance of as-cast Ti48AlxCr balanced mechanical properties.
(x = 0, 2, 4 at.%) alloys were studied. The dependence of The alloying elements can be classified to two broad
the tensile properties and creep resistance of as-cast TiAl groups. The first group of elements is the solid solution
on the solid solution strengthening and formation of b strengthening elements or beta stabilizers such as Cr, Mn,
phase due to addition of Cr was investigated. V, Nb, Fe, Mo, W, etc., and the second group of elements
consists of B, Si and C which provide precipitation or
particle strengthening [6]. Among the various alloying
elements, Cr is unique in several ways [7]. Firstly, TiAlCr
Introduction
alloys containing the ternary Laves phase Ti(Cr,Al)2 not
only have outstanding oxidation resistance [8], but more
TiAl-based alloys are being considered as potential light-
importantly, have very similar thermal expansion coeffi-
weight materials for number of high temperature applica-
cients to those of c-alloys. Secondly, theoretical calcula-
tions [1, 2]. These alloys consist of high volume fraction of
tions based on the total energy LMTO method have shown
c phase together with small volume fraction of a2 phase.
that interaction between Cr and Al does not stabilize x
One of the major challenges encountered in the develop-
phase formation in the b phase [9, 10]. This makes it
ment and potential application of two-phase TiAl is how to
possible to ductilise the c-alloys by compositing the c
optimize the mechanical properties. With respect to opti-
phase with a ductile b phase of A2 or B2 structure [7].
mization of the mechanical properties of TiAl, alloying and
This investigation is to study the effect of Cr addition on
microstructure control must be considered [3]. Engineering
the microstructure, tensile properties and creep resistance
TiAl based alloys should exhibit balanced room tempera-
of as-cast TiAl alloy. The microstructure and mechanical
ture ductility, fracture toughness, high temperature
properties of as-cast gamma titanium aluminide containing
strength, creep and oxidation resistance [4]. So far, much
beta phase, Ti48Al4Cr are described and a comparison is
effort world-wide has been devoted to studying the effects
made with typical two-phase, first generation aluminides
of thermomechanical processing and alloying in TiAl. The
exemplified by binary Ti48Al and solid solution
general consensus is, however, that where termomechani-
strengthen, Ti48Al2Cr. The dependence of the tensile
cal processing results in the improvement of a specific
properties and creep resistance of as-cast TiAl on the solid
property, it is often at the expense of another [5]. In this
solution strengthening and formation of b phase due to
addition of Cr was investigated.
E. Hamzah (&) M. Kanniah
Faculty of Mechanical Engineering, Universiti Teknologi
Malaysia, Skudai, Johor 81310, Malaysia
e-mail: esah@fkm.utm.my Materials and experimental procedures
M. Harun
Industrial Technology Division, Malaysian Institute for Nuclear The materials investigated in this study have the nominal
Technology Research, Bangi, Kajang, Selangor 43000, Malaysia composition of Ti48 at.%Al, Ti48 at.%Al2 at.%Cr and
123
9064 J Mater Sci (2007) 42:90639069
Ti48 at.%Al4 at.%Cr. The alloys were produced by IRC peritectic reactions, L + b a and L + a c, is
University of Birmingham, United Kingdom by plasma hardly to be completed due to limited diffusion caused by
melting casting into 2 kg buttons. All the alloys will be the formation of a solid envelope of the peritectic phase,
referred in atomic percentage hereafter. The samples for avoiding the physical contact between the reactants [11].
metallography analysis were cut using Manho EDM wire As far b phase is concerned, even though it is found to exist
cut to 10 mm 10 mm 10 mm cube. The samples were together with a(a2) and c phases in b stabilizer added TiAl,
ground and polished according to standard procedures. The attention has been paid mainly to the role of b phase in the
polished surface was etched using Krolls reagent (94 ml mechanical properties but not to the formation mechanism
H2O + 4 ml HNO3 + 2 ml HF). The microstructures were of the microstructure as well as the phase equilibria among
examined using Phillips XL40 scanning electron micro- b, a and c phases. Therefore, the microstructural formation
scope. The grain size and volume fraction is measured in the alloys under development or study has in most case
using the intercept and point counting methods respec- been interpreted based on the binary phase diagram, since
tively. The lattice parameters of the alloys were measured the amount of the additive elements is not large [12].
by Siemens D 5000 X-ray diffractometer. However, little attempt has been made to establish the
Flat samples with dimension, 3 mm 2 mm with phase relationship among the b, a and c at elevated tem-
15 mm gauge length were prepared for the tensile and peratures [7, 13]. Figure 1a, b and c shows the micro-
creep test. Prior to testing, the test pieces were ground structure of Ti48Al, Ti48Al2Cr, and Ti48Al4Cr
along the gauge length in the longitudinal direction to respectively.
prevent premature crack initiation at surface defects during Figure 1 clearly shows that Ti48Al and Ti48Al2Cr
tension or creep. Instron tensile testing machine was used exhibits a nearly lamellae microstructure which consists
to perform the tensile testing at room temperature. Con- mainly the lamellae grain with a grain size of 500800 lm
stant load tensile creep test was performed at 600800 C and 25% of fine gamma grains at the grain boundaries.
with initial stresses of 150180 MPa. The constant load Even though the lamellae grain size of Ti48Al and
tensile creep test was conducted using Mayes TC 20 High Ti48Al2Cr is in a similar range, the volume fraction of
Temperature Creep Testing Machine, consisting three zone c phase in Ti48Al2Cr is significantly higher than in
temperature furnace and a loading lever arm ratio of 10:1 Ti48Al. The microstructure of Ti48Al4Cr consists
in air. The test temperatures were maintained constant with three phase which are the lamellae (300600 lm), fine
a precision of 1 C and monitored for 2 h before the test. gamma (510%) and beta (25%) phase. The beta grains
The temperature was monitored during the creep test by are observed to be irregular in shape together with fine c
using a thermocouple in contact with the gauge section of grains along grain boundaries. Therefore, it is very hard to
the sample. The sample was allowed to soak at test tem- distinguish between c and b grains of Ti48Al4Cr at low
perature for 1 h before the load was applied. The test was magnifications. SEM image in Fig. 1c clearly reveals the
interrupted after 20 h and the samples were left to cool in presence of b phase (bright contrast) along lamellae grain
air to room temperature. boundaries. This type microstructure appears to be com-
monly observed in b containing TiAl alloy, e.g., Ti46Al
(1.52)Cr2Mo0.25Si0.3B [14, 15], Ti46.5Al2Cr
Experimental results 2Nb0.8Mo0.2W0.2Si [16] and Ti-4510Nb [17].
Table 1 provides the lamellae spacing and grain size of the
Microstructural characterisation alloys. Generally, these features are in the similar range.
The composition of all the phases, such as a2 + c
Ti48Al and Ti48Al2Cr exhibits nearly lamellae lamellae, c grains and b phase, were measured using EDX
microstructure consisting high volume fraction of c phase method. The solution of Cr in c phase increases with
(tetragonal, L10 structure) and a small amount of a2 phase addition of Cr (04 at.%) in Ti48Al. The solution of Cr in
(hexagonal, D019). Ti48Al4Cr too exhibits nearly the b phase of Ti48Al4Cr is 11.15 at.%, indicating that
lamellae microstructure but consist of lamellae, fine c and Cr is a strong b stabilizer. This results shows that Cr sta-
b phase. Addition of Cr up to 4 at.%, introduces b phase bilizes b and slightly c phase. Table 2 shows the compo-
(bcc structure). The lamellae structure consists of alter- sitions of all the phases of the alloys studied.
nating laths of the c-TiAl and a2-Ti3Al phases. Such Table 3 shows the variation of lattice parameter a and c
lamellae structure results from the solid-state phase trans- of the c phase and increment of Cr solution in c phase with
formation of the primary disordered a dendrites. The single addition of Cr in Ti48Al. The variation of lattice param-
fine c regions surrounding the lamellae grains result from eter in the c phase is complex. The variation of lattice
the transformation of the aluminum rich interdendritic melt parameter and increment of Cr in c phase shows the solid
due to incompleteness of the peritectic reactions. The two solution strengthening of Cr in c phase.
123
J Mater Sci (2007) 42:90639069 9065
123
9066 J Mater Sci (2007) 42:90639069
Ductility (%)
400
1.5
300
1
200
0.5
100
0 0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Amount of Chromium (%) Amount of Chromium (%)
Creep resistance
123
J Mater Sci (2007) 42:90639069 9067
0 0
0 5 10 15 20 25 0 5 10 15 20 25
Time (hours) Time (hours)
2.5 10
(c) (d) Ti-48Al-4Cr
Ti-48Al 9 ruptured
2 interrupted 8
Creep strain (%)
1 Ti-48Al-2Cr
interrupted
0 0
0 5 10 15 20 25 0 5 10 15 20 25
Time (hours) Time (hours)
these results that the effect of beta phase on creep resis- Typical creep curve at 800 C with initial stress of
tance depends on the temperature. 150 MPa are shown in Fig. 4d. Ti48Al and Ti48Al2Cr
Figure 4c shows the creep curves at 800 C with initial exhibits steady state creep behaviour whereas Ti48Al4Cr
stress of 180 MPa. Ti48Al and Ti48Al2Cr exhibits exhibits minimum creep behaviour. After the initial
0.53% and 0.6% instantaneous creep strain respectively instantaneous strain, Ti48Al and Ti48Al2Cr undergoes
when loaded with initial stress of 180 MPa. Both alloys a period of primary creep where the strain rate (de/dt)
showed steady state behaviour. However, Ti48Al4Cr decreases with time to an apparent steady state value that
failed upon loading. It is quite interesting to note that persists for a long period of time until interruption. On the
Ti48Al4Cr which exhibits low primary creep strains and other hand, Ti48Al4Cr which exhibits low instantaneous
steady state creep rate compared to Ti48Al at 600 and creep strain compared to Ti48Al and Ti48Al2Cr,
700 C failed upon loading at 800 C whereas Ti48Al and undergoes a short primary creep and reaches a minimum
Ti48Al2Cr exhibits steady state value until interruption. value before the onset of tertiary creep. This type of creep
Therefore a creep test is attempted with initial stress of behaviour is also referred as inverse creep where the
150 MPa to study the behaviour of Ti48Al4Cr at 800 C creep curve can be divided into two regimes, before and
and possible cause for the sudden failure during loading of after the minimum creep rate, instead of the common three.
180 MPa. Ti48Al4Cr sample failed after 16.5 h. Ti48Al and
123
9068 J Mater Sci (2007) 42:90639069
15
and none on the other.
Microstructural features such as lamellae spacing and
10 grain size has significant effect on the mechanical proper-
ties of TiAl. These features are usually controlled by heat
5 treatment. However, controlled alloying chemistry, e.g., in
this present study; systematic addition of Cr (04 at.% Cr)
provides a path to study the minor effect of alloying
0
0.8 1 chemistry on microstructural features. The result of
Creep strain (%) microstructural characterization shows that the lamellae
spacing and grain size is in a similar range for all the alloys
Fig. 5 Time needed to reach 0.8 and 1% of Creep Strain for the
Binary and Ternary Alloys investigated. Therefore, it is concluded that addition of Cr
in the present study does not have significant effect on
microstructural features. However addition of 4 at.% Cr
Ti48Al2Cr exhibits 0.53% and 0.47% of instantaneous introduces b phase. Therefore, the variation in mechanical
creep strain respectively, whereas Ti48Al4Cr exhibits properties between the alloys studied is solely on the solid
0.32%. The instantaneous creep strain of Ti48Al and solution strengthening and b phase.
Ti48Al2Cr is in a similar range. Note also that the creep
strain of Ti48Al2Cr is just slightly less than of Ti48Al. Effect of chromium addition on tensile and creep
On the other hand, Ti48Al4Cr exhibits greater creep properties
strain before rupture. The period of succeeding accelerating
creep rates in the tertiary regime was found to be large and The increase of the fracture strength at room temperature
corresponds to roughly 8590% of the total creep strain to with the Cr content is chiefly attributed to the solid-solution
fracture. The time in the tertiary creep regime was about strengthening of Cr in c phase. However, presence of b
8085% of the total time to rupture. The secondary creep phase in TiAl does not improve, but deteriorates the frac-
rate of Ti48Al and Ti48Al2Cr is almost identical, ture strength and ductility at room temperature. It is
however, Ti48Al4Cr exhibit a greater secondary creep because the b phase, with a high content of Al and Cr, is
rate. Although, it is common to assume minimum creep brittle at room temperature. Tensile tests show that the
rate is equal to steady state creep rate for creep analysis or Ti48Al4Cr has intermediate ductility between Ti48Al
comparison, however, it does not really accurate. To fur- and Ti48Al2Cr. Although the fracture strength and
ther compare the alloys, the time necessary to reach 0.8 and ductility of Ti48Al4Cr is higher than of Ti48Al, how-
1% creep strains is recorded. Figure 5 shows that Ti48 ever it is doubted whether it is due to beta phase or not. The
Al2Cr took the longest time to reach 0.8 and 1% creep cleavage like fracture surface of Ti48Al4Cr after room
strains followed by Ti48Al and Ti48Al4Cr respec- temperature tension indicates the brittle characteristic of
tively. This indicates that solid solution strengthening by beta phase. Therefore, b phase is undoubtedly detrimental
chromium in fine c phase, slightly improves the creep at room temperature due to its extreme brittleness.
strength. On the other hand, Ti48Al4Cr had reached 0.8 Although it had been reported that solid solution
and 1% creep strains in a short period which indicates low strengthening by addition of Cr increases the creep life of
creep strength. Beta phase deteriorates the creep resistance TiAl at high stresses [6], however it has no significant effect
at high temperature. to the applications. This is due the creep strain exhibited by
solid solution strengthening TiAlCr (Ti48Al2Cr) is
only slightly lesser than binary Ti48Al at the intermediate
Discussion to high temperature. Aerospace and automotive applica-
tions not only concerns creep life but creep strain too.
Role of chromium addition on strengthening Steady state creep rate and the time necessary to reach 1%
mechanisms and microstructural modification creep strain of Ti48Al2Cr are not significantly improved.
On the other hand, b phase becomes very ductile at high
The variation of lattice parameters of c phase and incre- temperature due to its open bcc structure [16]. It is evident
ment of Cr in c phase with addition of Cr in Ti48Al shows from the high creep strain exhibited by Ti48Al4Cr before
123
J Mater Sci (2007) 42:90639069 9069
rupture during creep test at 800 C with initial stress of 4. All the alloys showed a mixed mode fracture surface
150 MPa as shown in Fig. 4d. High creep strain would after tensile test. Generally, transgranular brittle frac-
exceed the tolerance limit of a component and thus prevents ture is more evident due to large cleavage facet cor-
the applicability of the alloy. Therefore, b phase is detri- responding to c and b phase. Interlamellae separation
mental at both room and high temperature due to its extreme observed are due to crack along the a2/c interface.
brittle-ductile character. However, this study shows that b Slight dimple-like appearance of fracture surface of
phase might be useful at intermediate temperatures. This is Ti48Al2Cr reflects higher ductility compared to
due to the excellent creep resistance of Ti48Al4Cr Ti48Al and Ti48Al4Cr.
compared to Ti48Al and Ti48Al2Cr at temperatures 5. Solid solution strengthening by chromium in c phase
from 600700 C. The brittle-ductile character of b phase is has no significant effect on creep resistance with
suggested to be moderate and therefore have good signifi- respect to applications.
cant effect on mechanical properties at the intermediate 6. b phase deteriorates the creep resistance at high tem-
temperatures. From the present study, it is concluded that perature. However, presence of b phase has good
solid solution strengthening by Cr in c phase does not have effect on creep resistance at intermediate temperatures,
significant effect on creep strength of TiAl and introducing 600700 C.
b phase with addition of Cr is found to be detrimental for 7. Beta phase is very sensitive to temperature and detri-
creep strength with respect to the applications. mental for mechanical properties at both room and
Modifications to the alloy chemistry have been high temperature.
attempted in order to maintain a balance between the room
temperature ductility and high temperature creep resis- Acknowledgements The authors would like to thank the Malaysian
tance. It is clear that as with all alloys, the high and low Ministry of Science, Technology and Innovation (MOSTI) for the
temperature properties cannot be generally be optimized research fund under the IRPA program (Project No. 09-02-06-0002
simultaneously and the aim of the present program is to EA-002).
assess the optimum alloying routes which allow an
acceptable compromise between these properties. It is of
References
course widely accepted limiting factor of the applications
of these alloys is their cost and development of cost- 1. Yamaguchi M, Inui H (1993) In: Darolia R, Liu CT, Martin PL,
effective processing is not been addressed in this paper. Miracle DB, Wagner R (eds) Structural intermetallics. TMS,
Warrendale, PA, USA, p 127
2. Kim YW, Dimiduk DM (1991) J Metals 43:40
3. Crofts PD, Bowen P, Jones IP (1996) Scripta Mater 35:1391
Conclusions 4. Marketz WT, Fischer FD, Clemens H (2003) Plasticity 19:281
5. Cheng TT, Wills MR, Jones IP (1999) Intermetallics 7:89
1. As-cast Ti48Al and Ti48Al2Cr exhibited a nearly 6. Sun FS, Cao CX, Kim SE, Lee YT, Yan MG (2001) Metall Mater
Trans A 32:1573
lamellae microstructure consisting of a high volume 7. Shao G, Tsakiropoulos P (1999) Intermetallics 7:579
fraction of c phase and a small amount of a2 phase. 8. Brady MP, Smialek JL, Smith J, Humphrey DL (1997) Acta
Ti48Al4Cr too exhibits a nearly lamellae micro- Mater 45:2357
structure but the lamellae grain boundaries is occupied 9. Shao G, Tsakiropoulos P, Nguyen-Manh D, Pettifor DG (1997)
Phil Mag Lett 76:207
by fine c and body centered cubic (bcc) structured b 10. Shao G, Passa E, Tsakiropoulos P (1997) In: Nathal MV, Darolia
phase. R, Liu CT, Martin PL, Miracle DB, Wagner R, Yamaguchi M
2. Addition of Cr stabilizes c (TiAl) and b phase. The (eds) Structural Materials. TMS, Warrendale, PA, USA. p 825
solubility of Cr in c and b increases with the addition 11. Duarte A, Viana F, Henrique M, Santos CM (1999) Mater Res 2:1
12. Takeyama M, Ohmura Y, Kikuchi M, Matsuo T (1998) Inter-
of Cr from 2 at.% to 4 at.%. metallics 6:643
3. Ti48Al2Cr exhibited the highest fracture 13. Jewett TJ, Ahrens B, Dahms M (1997) Mater Sci Eng A 225:29
strength and ductility followed by Ti48Al4Cr and 14. Schillinger W, Clemens H, Dehm G, Bartels A (2002) Inter-
Ti48Al. The solid solution strengthening by Cr in metallics 10:459
15. Zhang D, Arzt E, Clemens H (1999) Intermetallics 7:1081
single c phase is responsible for this improvement. 16. Wang JG, Nieh TG (2000) Intermetallics 8:737
However, further addition of Cr stabilizes the b phase 17. Zhang WJ, Chen GL, Appel F, Nieh TG, Deevi SC (2001) Mater
(Ti48Al4Cr) which decreases the tensile properties. Sci Eng A 315:250
123
J Mater Sci (2007) 42:90709074
DOI 10.1007/s10853-007-1640-8
Received: 9 February 2006 / Accepted: 26 February 2007 / Published online: 18 July 2007
Springer Science+Business Media, LLC 2007
123
J Mater Sci (2007) 42:90709074 9071
An aerated 1 M H2SO4 aqueous solution, prepared from corrosion products formed at the electrode surface al-
ultra pure reagent, was used as corrosive electrolyte. The lowed a pseudo-passivation of the material.
halide salt was C8H17Ph3P+,I and it was prepared in the When the inhibitor is added, we observed an increase of
laboratory by adding triphenylphosphine to heptane iodine. the cathodic current density especially for region I. Then,
The electrochemical instrumentation consisted of a the current density tends to a plateau. Thus, the cathodic
Taccussel potentiostat-galvanostat PGP 201. Voltamaster 1 reaction is limited by oxygen diffusion. Finally, the current
software was employed for instrumentation control and density increases corresponding to H+ reduction.
data treatment. One can conclude that the organic halide salt denotes an
inhibition activity for the cathodic reaction taking place at
the surface of nickel in 1 M H2SO4.
Results and discussion For the anodic domains (regions II, III and IV), one can
notes that the corrosion potential tends to more anodic
Figure 1 shows anodic and cathodic polarization curves for value. Indeed, only the second activation peak is observed.
nickel electrode immersed for 1 h at open circuit potential Furthermore, the steady state current decreased for the
in 1 M sulphuric acid aqueous media, in the absence and entire over potential interval considered.
presence of 103 M of the iodine alkyl triphenyl phos- Hence, one can conclude that organic iodine salt acts as
phonium. The scan rate has been fixed at 8.33 mV/s and mixed inhibitor. The observed inhibitive action could be
the temperature at 298 K. due to an adsorption phenomenon at the surface of nickel
In absence of the halide salt and independently on the making a barrier for charge and mass transfer between the
temperature, one notes the presence of four regions on metal and its environment.
the polarization curves. The region I represent the To characterize the nickel interfacial behaviour, we
cathodic domain characterized by a charge transfer carried out the voltammetric investigation at various tem-
process. It can be assigned to the strong concentration of peratures ranging from 308 K to 328 K. The polarization
acid (1 M) and the mobility of the H+ ion that conceals curves obtained are reported in Fig. 2.
the contribution of the oxygen in the global kinetics of We have measured the corrosion potential (Ecorr) and the
the cathodic reaction. In the region II the electrochemical corrosion current densities (jcorr) from the points of inter-
process at the NickelH2SO4 interface is controlled by section of the extrapolated anodic and cathodic Tafel
activation kinetic. The extrapolation of this part to the regions. Then, the Inhibition Efficiency percentage (IE%)
potential of corrosion permits us calculating the material and the degree of surface coverage (h) were calculated as
corrosion rate. The region III is characterized by the follow:
presence of two oxidation peaks (1a) and (2a) corre-
sponding to the electrochemical generation of nickel j0 j1
IE% 100 1
oxides and/or sulfates. Then, the current falls (region IV). j0
A surface steady state appeared indicating that the
j0 j1
h 2
j0
jss
j/mA/cm
123
9072 J Mater Sci (2007) 42:90709074
298K 308 K
nickel in 1 M H2SO4 at different 2
10 2
phosphonium iodine 10
Log j /mA/cm2
concentrations for various 10
1
1
10
temperatures raging from 298 K
2
j/mA/cm
0
10
to 328 K 0
10
-1
10
-1
-2 10
10
-2
10
-3 10
-4 -3
10 10
-1000 -500 0 500 1000 -1000 -500 0 500 1000
E/mV/SCE E/mV/SCE
3 3
10 10
318 K 328 K
2 2
10 10
1 1
10 10
2
2
j/mA/cm
j/mA/cm
0 0
10 10
-1 -1
10 10
-2 -2
10 10
-3 -3
10 10
-1000 -500 0 500 1000 -1000 -500 0 500 1000
E/mV/SCE E/mV/SCE
-6 -5 -4 -3
0 M (inh); 10 M (inh); 10 M (inh); 10 M (inh); 10 M (inh);
123
J Mater Sci (2007) 42:90709074 9073
2
308K R = 0.998
DEa 318K
Logjcorr constant 3 328K
2:303RT 1,0x10
-3
-6
Y = 1.07X + 310
2
where: R =1
C /
-6
Y = 1.04X + 610
DEa: the activation energy of the corrosion reaction; -4
2
R = 0.999
5,0x10
T: the absolute temperature;
R: the universal gas constant. -6
Y = 1.02X + 910
2
Figure 3 shows the Arrhenius plots for the corrosion rate R = 0.999
of nickel in 1 M H2SO4 with and without the presence of 0,
-4 -4 -4 -3
00
103 M of phosphonium iodine slat. 2,50x10 5,00x10 7,50x10 1,00x10
-1
The results give rise to satisfactory straight lines from the C (mol.L )
slopes of which one can calculate the activation energy. DEa
Fig. 4 Langmuir adsorption isotherms on nickel in 1 M H2SO4
values were found to be 19.3 kJ mol1 for the uninhibited solution with and without the presence of 103 M of the phosphonium
solution and 71.1 kJ mol1 in the presence of 103 M of iodine slat
phosphonium salt. The value of DEa for the inhibited solution
is higher than that for uninhibited one, indicating the greater considered that the surface coverage (h) of the inhibitor on
tendency of such compound to react at the surface of nickel, the nickel specimen is related to the concentration (C) of
in 1 M H2SO4 electrolyte [57]. the inhibitor in the bulk of the solution according to
From the above mentioned results it is concluded that Langmuir adsorption isotherm:
the inhibition of nickel corrosion in 1 M H2SO4 solution
kC
occurs by adsorption of the additive used [8, 9]. h 4
The most commonly used isotherms for studying the 1 kC
adsorption mechanism of an inhibitor on a metal electrode
where k is the equilibrium constant for the adsorption
surface are listed in the literature [1019].
process.
In order to choose the appropriate isotherm model to
Rearranging Eq. 4:
describe the adsorption mechanism of phosphonium iodine
at nickel surface, we plotted C/h versus concentration C
1
(Fig. 4). C=h C 5
k
A linear relationship is obtained, independently on
the temperature considered. To explain such result, we One can conclude that the absorption of phosphonium
iodine salt on nickel surface in 1 M sulphuric acid medium
occurs according to the Langmuir adsorption isotherm.
2,0
From the intercepts of the straight lines on the C/h axis,
k values could be calculated (Table 2).
1,5 Furthermore, the equilibrium constant (k) for the
adsorption process of the inhibitor could also be related to
the free energy of adsorption DadsG by the relationship:
1,0 Without inhibitor
log jcorr
1 Dads G
With inhibitor k exp 6
55:5 RT
0,5
The calculated values are gathered in Table 2. Then, we
plotted DadsG against temperature (Fig. 5).
0,0
Independently on the temperature considered, negative
values are obtained for DadsG indicating the spontaneous
3,0 3,1 3,2 3,3 3,4
3 -1 -1
interaction between phosphonium iodine and nickel sur-
10 T (K ) face.
Fig. 3 Arrhenius plots for nickel in 1 M H2SO4 with and without the However, the free energy decreases from 28.8 kJ mol1
presence of 103 M of the phosphonium iodine slat at 298 K to 33.6 kJ mol1 at 318 K. A linear relationship
123
9074 J Mater Sci (2007) 42:90709074
Table 2 Values of k and DadsG for the adsorption of the phospho- could be attributed to the competition adsorption/desorp-
nium iodine salt on the nickel surface in the temperature domain 298 tion of the additive on nickel surface for T 318 K.
328 K
T k (kJ mol1) log k DadsG = RT log k
(kJ mol1) Conclusion
298 111 11.62 28.8
308 167 12.02 30.8 The results of these experiments revealed that:
318 333 12.72 33.6 (1) phosphonium iodine decreases the corrosion rate of
328 33.3 10.41 28.4 nickel in 1 M H2SO4 electrolyte;
(2) at constant temperature, IE% increases with increas-
ing inhibitor concentration;
(3) at the same concentration, IE% decreases with
-28 increasing temperature;
(3) inhibition occurs by adsorption according to the
-29 Y = -0.24X + 43.42 Langmuir isotherm;
2
R = 0.999
(4) inhibitor desorption increases for T 318 K.
-30
adsG (kJ.mol )
-1
-33 References
-34 1. Morad MS (2000) Corros Sci 42:1307
2. Khaled KF (2004) Appl Surf Sci 230:307
295 300 305 310 315 320 325 330 3. Niass SO, Touhami ME, Hajjaji N, Srhiri A, Takenouti H (2001)
() J Appl Electrochem 31:85
4. Said F, Souissi N, Dermaj A, Hajjaji N, Triki E, Srhiri A (2005)
Fig. 5 Adsorption standard free energy (DadsG) versus temperature Mater Corrosion 56(9):619
5. Saleh JM, Al Haidari YK (1989) Bull Chem Soc Jpn 62:1237
could be used to describe DadsG variation versus T in this 6. Abd El Nabey BA, Kamis E, Ramadan MS, El Gindy A (1996)
interval. Then, other thermodynamic functions can be Corrosion 52:671
calculated through the following equation: 7. Abd El Aal EE, Zakria W, Diab A, Abd El Haleem SM (1999)
J Chem Technol Biotechnol 74:1061
8. Dinnappa RK, Mayanna SM (1982) Corrosion 38:525
Dads G Dads H TDads S 7 9. Rudresh HB, Mayanna SM (1977) Surf Technol 6:139
10. Langmuir I (1918) J Am Chem Soc 40:1361
where DadsH and DadsS the enthalpy and the entropy of 11. Frumkin AN (1925) Z Phys Chem 116:466
adsorption. They were equal to +43.4 kJ mol1 and 12. Hill TL (1952) J Chem Phys 20:141
13. de Boer JH (1953) In: The dynamical character of adsorption.
+240 J K1 mol1, respectively.
Oxford University Press, Oxford
In the temperature domain considered, both thermody- 14. Parsons R (1964) J Electroanal Chem 8:93
namic functions values are positive reflecting for phos- 15. Damaskin BB, Petrii OA, Batrakov VV (1971) In: Adsorption of
phonium iodine adsorption on nickel surface: organic compounds on electrodes. Plenum Press, New York,
p 86, 94 and 247
(1) an endothermic behaviour; 16. Kastening B, Holleck L (1965) Talanta 12:1259
(2) an increase of the molecular disorder. 17. Bockris JO_M, Swinkels DAJ (1964) J Electrochem Soc 111:736
18. El-Awady AA, Abd-El-Nabey BA, Aziz SG (1992) J Electro-
Then, the free energy exhibited a singular behaviour. In chem Soc 139:2149
fact, it increases reaching 28.4 kJ mol1. Such result 19. Khamis E, Hosny A, EL-Hadary S. (1995) Affinidad 95:456
123
J Mater Sci (2007) 42:90759082
DOI 10.1007/s10853-007-1894-1
Received: 22 January 2007 / Accepted: 29 May 2007 / Published online: 27 July 2007
Springer Science+Business Media, LLC 2007
Abstract An induced optical anisotropy is observed as a causes the interacted medium to behave as an optically
result of interaction of a high-power CW Ar+ laser beam, uniaxial one, with the optical axis along the direction of the
with silver-ion-exchanged glasses. We have shown that the beam propagation. It is possible to induce similar effects by
absorption of the beam by the thin layer of Ag+ produces a unpolarized light [15]. Interaction of Ag-containing thin
temperature gradient resulting in a radial stress on the layers with polarized laser beam could make the material
surface of the sample. The induced anisotropy makes the anisotropic, leading to a dichroism [1618], which is out of
sample behave as a thin uniaxial optical medium with axis the scope of the present article.
along the direction of the beam propagation. For the In laser materials such as Nd+: YAG rod, the tempera-
polarized light, the induced anisotropy restricts the appli- ture gradient created by the flash lamp generates a
cation of micro-lenses, which are made by this method. mechanical stress in the material, which leads to a thermal
The average refraction index of the interaction area is strain resulting in the appearance of a birefringence [68,
measured. 14]. Irradiating crystals such as Tb3Ga5O12 or semicon-
ductor-doped glasses by a laser beam also results in gen-
eration of stress birefringence (Dn) [4, 5, 10, 12, 13, 19],
Introduction due to the thermal gradient induced by the Gaussian laser
beam. In above-mentioned cases if one places the inter-
It is well-known that placing an anisotropic absorbing acting medium between two crossed polarizers and illu-
crystal between two crossed polarizers and illuminating it minate it with a white light, a characteristic conoscopic
with a white light, results in the appearance of a charac- pattern could be observed. This pattern means that a dis-
teristic image with isogyres and isochromates, called the tinguished direction is induced (z-direction, along the
conoscopic pattern [1]. propagation of the laser beam), which coincides with the
Induced anisotropy as a thermal effect of the interaction dielectric principal axis ez i.e. ex ey 6 ez . Thus, the
of light (laser beam, flash lamp,...) with crystals and some medium becomes optically uniaxial similar to the case of a
other amorphous materials, leading to formation of cono- natural crystal [1]. The magnitude of Dn increases with the
scopic patterns, has already been observed and studied by radius, by quadratic law [5]. Therefore, a linearly polarized
many researchers [214]. In all above-mentioned cases, light passing through the sample will experience a phase
especially the laser materials, the light induced anisotropy difference between the two components along nr and n/
(the radial and tangential photoelastic changes in the index
of refraction, respectively). Thus, except for points located
A. Nahal (&)
Department of Physics, University of Tehran, Karegar Shomali along the directions of the polarizer and the analyzer (x and
Ave., Tehran 14399-55961, Iran y) for which the polarization remains linear, at each point
e-mail: nahal@khayam.ut.ac.ir of the transverse section, a polarization change yields a
spatial elliptical polarization pattern.
F. Moslehirad
Qazvin Branch, Azad University, Daneshgah Ave., We have observed the same phenomenon during the
P.O. Box 34185-1416, Qazvin, Iran interaction of a CW Ar+ laser beam (Pmax = 8 W), working
123
9076 J Mater Sci (2007) 42:90759082
in the multiline regime with Ag+-doped glasses. Absorption preliminary explanation of the obtained results. The con-
of the laser beam energy by the thin layer of the embedded clusion of report is given in the last section.
silver ions in the glass matrix increases the temperature of the
incident point of the laser beam on the sample. The Gaussian
intensity distribution of the beam results in establishing a Experiment
temperature gradient across the incident point leading to the
appearance of a stress birefringence in the sample. We have Sample preparation
found that if one put the irradiated sample between two
crossed polarizers, using white light, then a characteristic The samples are prepared using the well-known ion-
conoscopic pattern would be observed. That is, the irradiated exchange method [26]. For this, slides of commercial soda-
ion-exchanged glass behaves as uniaxial optical medium, lime glasses (20 30 0.8 mm) were placed in mixed
with axis along the beam propagation. This effect has not molten salts of NaNO3/AgNO3 (96/4 wt.%) in a tempera-
been studied for silver-ion-exchanged glasses neither ture of 400 C for 4 h, in the atmospheric pressure. In the
reported for thin layer of silver ions in a glass matrix. ion-exchange process some of the sodium ions of the glass
Doping a glass matrix with metal clusters produces a matrix would be exchanged by some of the silver ions of
nonlinear optical medium, which is promising for opto- the molten mixed salt, due to the diffusion process. The
electronics applications like optical switches and slab temperature and duration time of the ion-exchange process
waveguides. One of the most investigated doped glasses is determines the concentration, diffusion depth in the matrix,
the Na+Ag+ ion-exchanged one. Studying the interaction and even the size of the formed silver ionic clusters in the
of a high-power laser beam with Ag+-doped glass develops glass matrix. The composition of the glass also affects the
our knowledge about the limits of the above-mentioned above-mentioned parameters. As a result, a thin layer of
optical material, and also the temporary and permanent ion-exchanged glass would be formed in the glass
physical changes, which happen during and after the substrate. In our experiments the average thickness of the
interaction. Here we report one of the effects (the induced ion-exchanged layer was about 180 lm, measured by a
optical anisotropy), which we have observed. There are microscope equipped with a micrometer. Then the prepared
other effects, as the results of the interaction of a high- samples were irradiated by a CW high-power Ar+ laser
power laser beam with Ag+-doped glass, which we have beam (Pmax = 8 W), working in multiline regime.
reported elsewhere in detail [17, 20, 21]. The ion-exchanged glass has a brownish color with a
Our experimental results show clearly that the micro- continuous absorption spectra in visible region with a peak
lenses produced by irradiating the ion-exchanged glass in the UV region at about wavelength ka = 310 nm
using a high-power laser beam [22, 23] are not suitable for (Fig. 1), which is related to very small ionic silver clusters
use with polarized white light, due to the induced optical [27, 28]. The color of the incident point of the laser on the
anisotropy in it. This important point should be considered sample becomes yellowish after the interaction, with an
during the production of above-mentioned micro-lenses. orange colored periphery, which is made out of silver
Local heating of doped glasses or polymers by a laser beam neutral clusters.
could produce a micro-lense. The formation of the micro-
lens occurs, due to the local melting of subsurface layer of Interaction with the laser beam
the glass. The glass melting causes by an increase of the
light absorption with temperature. Thus, when laser radi- The setup shown in Fig. 2 was used for studying the
ation heats the glass the light penetration depth decreases, thermal effect of the laser beam on the optical properties of
and in spite of the thermal diffusion the process of heat the ion-exchanged glass in real time.
deposition becomes increasingly localized at the surface of The sample was placed in front of the Ar+ laser with a
the glass. This thermal runaway causes the host glass to little tilt to prevent back-reflection of the beam from the
melt locally under laser radiation. Because of the lower sample towards the laser cavity. An expanded linearly
density of the molten glass compared with that of the solid polarized HeNe laser beam and a polaroid plate, placed
glass, the material wells out and solidifies forming a micro- after the sample, were used as a probe (Pprobe = 1 mW) to
lense [22, 23]. Micro-lenses have many applications such study the induced-optical anisotropy and any optical path
as microlense array for telescope application [24], micro- changes during the interaction, in real time. The transmit-
lense-based projection displays [25] and etc. ted probe beam was used to study the induced anisotropy,
The organization of the paper is as follows. In and the reflected beams from both sides of the sample were
Sect. Experiment, the experimental setup would be ex- used to study the optical path changes (Fig.2).
plained in detail the results of the experiments are reported Our experiments show that by increasing the power of
in this section. The Sect. Discussion includes a the Ar+ laser beam, the temperature of the incident point of
123
J Mater Sci (2007) 42:90759082 9077
123
9078 J Mater Sci (2007) 42:90759082
123
J Mater Sci (2007) 42:90759082 9079
123
9080 J Mater Sci (2007) 42:90759082
p
Let us have a brief review for propagation of light in an I I1 I2 2 I1 I2 cos d 10
optical anisotropic medium, which results in forming the
!2 !2
conoscopic patterns. When a light beam under an incident where I1 OF , I2 OG . Using Eqs. 8, 9, and 10 the
angle h1 enters into an anisotropic medium with thickness intensity distribution right after the analyzer would be
d, the beam is broken up into two beams with different equal to:
refractive indices n, n, where k0 nk0 , k00 nk0 ; with k the
wavelength in the vacuum. Hence, they propagate in the I d
cos2 v sin2 /sin2/ vsin2 11
medium with different velocities, which results in retar- I 2
dation between them. That is, the exiting beams from the 2
!
medium according to the Fig. 6a, have a phase difference where I OE .
~ A
For the case when P? ~ (v p we have:
equal to (see the Appendix A): 2
0 s1 I d
q 2
2pd @no sin h 1
sin2 2/sin2 12
d n2e sin2 h1 no 1 A 7 I 2
k ne no
~A
and for the case when Pjj ~ (v = 0) we have:
123
J Mater Sci (2007) 42:90759082 9081
Appendix A
2p 00 2p 00 2p
d AB B C 0 AB0 a
k00 k k
with AB00 cosdh00 , AB0 cosd h0 , B00 C dsinh1
0 0
tan h tan h and using the Snells law with n1 = 1 for
air, we have:
2pd 00
d n cos h00 n0 cos h0 b
k
Fig. 8 The calculated isochromates for our samples using the relation If we take the AB beam as the ordinary beam, then:
of Eq. 11. The blue isochromate (bold circle) has smaller radius in
comparison to the red one (the dashed circle). Their values are n0 : no ; h0 : ho
normalized to one, for simplicity. This calculation is in good
s
accordance to the experimental results (Fig. 4b)
0 0 sin h1 2
sin h1 no sin ho ) n cos h no 1
isochromates for both blue and red lights (Fig. 8) using the no
parameters of our samples. It can be seen that the iso- c
chromates of blue light have smaller radius relative to the
isochromates of red light for the same order of interference. For the extraordinary beam we can write:
Our experimental results (Fig. 4b) show the same result for
the irradiated Ag+-doped glass by the Ar+ laser beam. As it h00 : he ) n00 sin he sin h1 d
can be seen from Fig. 4b the blue fringes are closer to the
center compared to the red ones. If the optical axis is normal to the surface, then the
relation between n and ne can be obtained from the Fres-
nels formula:
Conclusion
no ne
n00 q e
In this article, we have reported the observation of an 2
n2o sin he n2e cos2 he
induced optical anisotropy in Ag+-doped glasses as a result
123
9082 J Mater Sci (2007) 42:90759082
Therefore, using Eqs. d, e and Fig. 6a we obtain: 12. Chen X, Calemczuk R, Salce B, Lavorel B, Akir C, Rajanoha L
(1999) Solid State Commun 110:431
13. Chen X, Lavorel B, Dreier T, Genetier N, Misserey H, Michaut X
n2o n2e n2o sin2 h1
cot2 he f (1998) Opt Commun 153:301
n2e sin2 h1 14. Koechner W, Rice DK (1970) IEEE J Quant Electron QE-
6(9):557
After some simple trigonometric calculation and using 15. Tikhomirov VK, Elliott SR (1994) Phys Rev B 49:17476
16. Kolobov AV, Lyubin VM, Tikhomirov VK (1992) Philos Mag
Eqs. d, e, and f the relation for n00 cos h00 could be written Lett 65:67
as: 17. Nahal A, Khalesifard HRM, Mostafavi-Amjad J (2004) Appl
q Phys B 79:513
no 18. Nahal A, Miloslavsky VK, Ageev LA (1998) Opt Commun
n00 cos h00 n2e sin2 h1 g 154:234
ne
19. Chen X, Lavorel B, Boquillon JP, Saint-Loup R, Jannin M (1998)
Solid-State Electron 42:1765
Substituting Eqs. c and g in b we end up with: 20. Nahal A, Mostafavi-Amjad J, Ghods A, Khajehpour MRH, Re-
0 s1 ihani SNS, Kolahchi MR (2006) J Appl Phys 100:053503
q 2 21. Nahal A, Khalesifard HRM (2007) Opt Mater 29:987
2pd @no sin h 1 A
d n2e sin2 h1 no 1 h 22. Kaganovski Yu, Antonov I, Bass F, Rosenbluh M (2001) J Appl
k ne no Phys 89:8273
23. Antonov I, Bass F, Kaganovski Yu, Rosenbluh M (2003) J Appl
Phys 93:2343
24. http://www.aqua.epfl.ch./projects.html
25. http://www.electronicimaging.org/program/04/
References 26. Najafi SI (1992) Introduction to glass integrated optics. Artech
House, Boston
1. Born M, Wolf E (1964) Principle of optics. Pergamon, Oxford 27. Nahal A, Rezae N (2005) In: Proc. 12th Conf. on Optics and
2. Chen X, Berger H (1999) J Phys: Condens Matter 11:7377 Photonics, Shiraz Univ., Iran, pp 537539 (In Persian)
3. Froehly L, Verrier I, Froehly C, Burn G, Veillas C (1999) Opt 28. Kreibig U, Vollmer M (1995) Optical properties of metal clusters.
Commun 167:27 Springer, Berlin
4. Petrov DV, Gomes ASL, De Araujo CB (1994) Phys Rev B 29. Horowitz M, Daisy R, Werner O, Fischer B (1992) Opt Lett
50:9092 17(7):475
5. Chen X, Gonzalez S (1998) Appl Phys B 67:611 30. Durbin SD, Arakelian SM, Shen YR (1981) Opt Lett 6(9):411
6. Foster JD, Osternik LM (1970) J Appl Phys 41(9):3656 31. http://www.tulane.edu/~sanelson/geo1211/interference_of_light.
7. Sims SD, Stein A, Roth C (1967) Appl Opt 6(3):579 htm
8. Eichler HJ, Haase A, Menzel R, Siemoneit A (1993) J Phys D: 32. Pedrotti FL, Pedrotti LS (1993) Introduction to optics. Prentice
Appl Phys 26:1884 Hall, New Jersey
9. Ayras PH, Friberg AT, Kaivola MAJ, Salomaa MM (1999) Appl 33. Takeda S, Yamamoto K, Matsumoto K (2000) Non-Crystalline
Opt 38(25):5399 Solids 265:133
10. Chen X, Chaux R (1999) Opt Commun 171:119 34. Martin M, Videau JJ, Canioni L, Adamietz F, Sarger L, Le Flem
11. Sims SD, Stein A, Roth C (1966) Appl Opt 5(4):621 G (2000) Appl Opt 39:435
123
J Mater Sci (2007) 42:90839091
DOI 10.1007/s10853-006-0957-z
Received: 26 May 2005 / Accepted: 8 September 2006 / Published online: 20 July 2007
Springer Science+Business Media, LLC 2007
123
9084 J Mater Sci (2007) 42:90839091
There are many of papers on studying the crystallization N2 inletoutlet and a cooler, 117 mL TEOS, 190 mL
kinetics of PA6 nanocomposites [1113], but no publica- ethanol and 13 g GPTES were added. After vigorous stir-
tion has been found concerning the crystallization kinetics ring with a magnetic at room temperature for 15 min,
of the PA6/silica hybrid nanocomposites made by func- 36 mL 0.05 M hydrochloric acid was added. The stirring
tionalized silica nanoparticles in situ generated and cap- was carried out for about 30 min under ambient condition
rolactam polymerized in one process. This study and a transparent solution was obtained; (2) 3,000 g e-
investigated the crystallization behavior of these PA6/ caprolactam, the solution obtained by step (1) and 1 g
functionalized silica hybrid nanocomposites under non- antioxidant were added into a reactor equipped with a
isothermal conditions. The Avrami method and the Mo mechanical stirrer. The mixture was heated to and held at
method were used for the analysis. XRD was also used to 383 K for 12 h under vacuum; (3) 80 g distilled water as
clarify the crystalline structure of PA6FS hybrid nano- the initiator was added into the reactor, the mixture was
composites obtained under different cooling conditions. heated to and held at 453493 K for 1 h and then heated at
538 K for 5 h at 0.5~0.8 MPa in a N2 atmosphere, then at
0~0.06 MPa for about 1 h. The material was extruded to
Experimental produce cylindrical extrudates by N2, followed by im-
mersed immediately in a cold-water (about 293 K) and
Materials pelletized with an adjustable rotating knife, located after
water bath, into 5 mm length. Then these granules were
e-caprolactam was purchased from Yueyang Chemical washed with water at 357 K for 10 h. The chemical reac-
Plant (China) without further purification. 3-glycidoxy- tions of the synthesis of PA6FS hybrid nanocomposites are
propyltrimethoxysilane (GPTES) was purchased from shown in Scheme 1. Pure PA6 and PA6S (prepared by
Chemical Reagent Co. of wuhan University (China) and reported [9]) hybrid nanocomposites were also prepared
was purified by distillation before use. Tetraethoxysilane under the same condition.
(TEOS) and ethanol (both chemical reagent grade), were
ordered from Xilong Chemical Fractory, Guangdong Differential scanning calorimetry measurement
(China).
The non-isothermal analyses were carried out using a TA
Preparation of PA6FS hybrid nanocomposites DSC-Q10 differential scanning calorimeter thermal ana-
lyzer. The samples were heated to 523 K at a rate of
The PA6FS hybrid nanocomposites used in this study was 20 K/min under nitrogen atmosphere, then kept for 5 min
prepared in three steps as follows: (1) Into a 500 mL three- at this temperature to eliminate the heat history before
necked round-bottomed flask, which was equipped with a cooling at a specified cooling rate. Constant cooling rate
Scheme 1
O Si
ethanol,H2O O Si O
Si(OR)4 + O
CO(CH2)3Si(OR)3 CO(CH2)3 Si O Si Si
H+ O O O
O O Si
Si O
( Silica CH CH2 )
O
O Silica CHCH2NH
O
OH
Silica HO C(CH2)5NH C(CH2)5NH2
CH CH2 Silica CHCOCO
O OH
H
Silica C CH2
OH N
O C
123
J Mater Sci (2007) 42:90839091 9085
2.5, 5, 10, 15, 20 and 40 K/min were applied. The ther- isothermal crystallization system. Just like isothermal
mograms corresponding to the heating and cooling cycles analysis, non-isothermal crystallization can also be ana-
were recorded and analyzed to estimate the non-isothermal lyzed by the Avrami equation, but, considering the non-
crystallization kinetics and crystallinity degree. isothermal characterization, Jeziorny [15] presented the
final form of the parameter characterizing the kinetics of
Transmission electron microscope (TEM) observation non-isothermal crystallization as follows:
Thin plates were cut by ultramicrotomy from the systems log Zc log Zt =/ 3
and observed under a TECNAI G2 20 Transmission Elec-
tron Microscope (TEM) at a high voltage of 120 kV. where / is the cooling rate.
The non-isothermal crystallization can also be analyzed
X-ray diffraction (XRD) using the Ozawa method [16]. The Ozawa equation is as
follows
To eliminate the influence of the specimen thickness, about
1 Xc expkT=/m 4
100 lm films of PA6 and PA6FS were prepared as reported
[8]: The granules of PA6 and PA6FS were put between two where k(T) is the crystallization rate constant, Xc is the
glass slides, and were heated in oil-bath up to 523 K. After relative crystallinity, / is the cooling rate, and m is the
the granules entirely melted, a load was applied on the Ozawa exponent depending on the crystal growth and
surfaces of the slides to press the melted materials into thin nucleation mechanism.
films. These films were kept on 523 K for 5 min to elim- Mo and Liu [17, 18] have proposed a new kinetic
inate the heat history, then cooled under three kinds of equation of non-isothermal crystallization by combing the
conditions: (1) The films were cooled down in the oil-bath Avrami and Ozawa equations:
from 523 K to 293 K naturally; (2) The films were cooled
down in the air at 293 K; (3) The films were quenched in log / log FT a log t 5
the water at 293 K.
XRD patterns were recorded by a D/max 2550 X-ray where a refers to the ratio of the Avrami exponent n to the
diffractometer. The CuKa radiation source was operated at Ozawa exponent m(a = n/m); the parameter
40 kV and 300 mA. Patterns were recorded by monitoring FT KT=Zt 1=m represents the value of cooling rate,
those diffractions from 5 to 40. The scan speed was which has to be chosen at unit crystallization time when the
4/min. measured system amounts to a certain degree of crystal-
linity. According to Eq. 5, the plot of ln/ versus ln t at a
given crystallinity will be a straight line. Parameters a and
F(T) can be obtained from the slope and the intercept of the
Non-isothermal crystallization kinetics model line.
123
9086 J Mater Sci (2007) 42:90839091
16
14
40K/min
12 20K/min
10K/min
10 5K/min
2.5K/min
8
exo
6
0
380 400 420 440 460 480 500 520
Fig. 1 TEM graphs of PA6S (a) and PA6FS (b) Temperature (K)
22
14
20
a. 40 K/min
b. 20 K/min 12
18 40K/min
c. 10 K/min
16 20K/min
d. 5 K/min 10
10K/min
14 e. 2.5 K/min
5K/min
12 8 2.5K/min
exo
exo
10
6
8
6 4
4
2
2
0 0
380 400 420 440 460 480 500 360 380 400 420 440 460 480 500
Temperature (K) Temperature(K)
Fig. 2 DSC thermograms of non-isothermal crystallization for pure Fig. 4 DSC thermograms of non-isothermal crystallization for
PA6 at different cooling rates PA6FS nanocomposites at different cooling rates
123
J Mater Sci (2007) 42:90839091 9087
Table 1 Values of DH, Ti, Tp and tp at various cooling rates for pure PA6, PA6S and PA6FS
R (K/min) DH (J/g) Ti (K) Tp (K) tp (min)
Pure PA6
2.5 60.62 472.7 455.5 6.9
5 59.76 469.5 449.6 4.0
10 57.15 465.8 443.1 2.3
20 54.40 461.5 436.0 1.3
40 53.7 455.7 425.6 0.8
PA6S
2.5 48.36 469.5 457.0 5
5 45.80 465.1 452.0 2.7
10 44.29 460.1 445.8 1.4
20 42.14 455.3 437.9 0.9
40 40.50 448.4 427.0 0.6
PA6FS
2.5 50.71 473.1 467.0 2.4
5 50.85 468.9 462.2 1.3
10 51.32 465.1 456.9 0.8
20 52.10 459.7 450.4 0.5
40 53.39 454.0 442.8 0.3
cooling rate. The rate of crystallization of PA6FS is the approximately the same shape, indicating that only the
highest one at all cooling rates, and the rate of PA6S is retardation effect of cooling rate on the crystallization is
located between in the rates of PA6FS and pure PA6, observed in these curves.
which means the addition of silica nanoparticles in situ
generated and functionalized by GPTES accelerates crys- Non-isothermal crystallization kinetics analyses
tallization greatly.
Figures 6, 7 and 8 present the relative crystallinity de- Figures 9, 10 and 11 present plots of log[ln(1 Xc(t))] as
gree as a function of temperature for these three materials a function of logt for pure PA6, PA6S and PA6FS,
crystallized at various cooling rates. The higher the cooling respectively. The parameters n, Zt, Zc and t1/2 showed in
rate, the lower the temperature range at which the crys- Table 2 were obtained from Figs. 911. n values of pure
tallization occurs, therefore, the transformation is con- PA6 vary from 2.56 to 3.11, and those of PA6S range from
trolled by nucleation. In addition, all curves have 2.72 to 3.56, which means the addition of silica influences
3.5
100
pure PA6
3.0 PA6S
PA6FS
80
Relative crystallinit y(%)
2.5
e
60
2.0
1/tp (min )
d
-1
a. 2.5K/min c
1.5 40
b. 5K/min b
c. 10K/min
1.0 a
20 d. 20K/min
e. 40K/min
0.5
0
0.0
0 5 10 15 20 25 30 35 40 45 400 420 440 460 480
Cooling rate (K/min) Temperature (K)
Fig. 5 Plots of 1/tp as a function of cooling rate for pure PA6, PA6S Fig. 6 Plot of relative crystallinity as a function of temperature for
and PA6FS nanocomposites pure PA6 crystallized non-isothermally at various cooling rates
123
9088 J Mater Sci (2007) 42:90839091
100 1.0
0.5
Relative crystallinity (%)
80
0.0
e
log(-ln(1-Xc))
60
-0.5
d
40
c -1.0
a.2.5K/min 2.5 K/min
b.5K/min b 5 K/min
20 c.10K/min -1.5 10 K/min
d.20K/min a 20 K/min
e.40K/min -2.0 40 K/min
0
e
samples Zc increases and t1/2 decreases as the cooling rate
60 increases, which means an increase of the rate of crystal-
d
lization. At the same cooling rate, the higher Zc of PA6FS
40
a.2.5K/min c than that of pure PA6 and PA6S indicates that the silica
b.5K/min
c.10K/min modified by GPTES prompted crystallization effectively.
b
20 d.20K/min
a
Figures 12, 13 and 14 present plots of log/ as a function
e.40K/min
of logt for pure PA6, PA6S and PA6FS, respectively. The
0 slope a and intercept log F(T) showed in Table 3 were
obtained from the fitted straight line. It can be seen from
400 420 440 460 480
Table 3 that the values of F(T) systematically increase with
Temperature (K)
an increase in the relative degree of crystallinity. At a given
Fig. 8 Plot of relative crystallinity as a function of temperature for degree of crystallinity, the higher the F(T) value, the higher
PA6FS crystallized non-isothermally at various cooling rates cooling rate is needed within unit crystallization time,
indicating the difficulty of polymer crystallization. By
comparing the values of F(T) of different samples, we have
1.0
1.0
0.5
0.5
0.0
0.0
log(-ln(1-Xc))
log(-ln(1-Xc))
-0.5
-0.5
-1.0
-3.0 -2.5
-1.0 -0.5 0.0 0.5 1.0 -0.5 0.0 0.5 1.0
log t(min) log t (min)
Fig. 9 Plot of log[ln(1 Xc)] as a function of logt pure PA6 Fig. 11 Plot of log[ln(1 Xc)] as a function of logt for PA6FS
123
J Mater Sci (2007) 42:90839091 9089
Table 2 Values of n, Zt, Zc and t1/2 at various cooling rates for pure PA6, PA6S and PA6FS nanocomposites
/ (K/min) n Zt Zc (minn)a t1/2 (s)
Pure PA6
2.5 2.56 0.01 0.13 392
5 2.87 0.02 0.44 230
10 2.99 0.07 0.76 134
20 3.02 0.30 0.94 79
40 3.11 1.33 1.01 47
PA6S
2.5 2.72 0.01 0.13 291
5 2.98 0.04 0.51 154
10 3.21 0.14 0.82 85
20 3.33 1.41 1.02 53
40 3.56 6.83 1.05 31
PA6FS
2.5 2.95 0.07 0.35 129
5 3.23 0.17 0.70 84
10 3.68 1.54 1.04 52
20 3.91 7.52 1.11 31
40 3.86 85.22 1.12 21
a
Is calculated from Eq. 3
found that the value of PA6FS is lower than those of pure the influence of silica on the crystalline fraction. It was
PA6 and PA6S, meaning that the crystallization rate of previously reported [20] that the a form exhibited two
PA6FS is faster than those of pure PA6 and PA6S. This is reflection peaks at 2h 20 [a1 (200), d = 0.44 nm] and
in accordance with the result obtained from the Avrami 24 [a2 (002), d = 0.37 nm], where only one reflection
approach. peak at 2h 21 [c (200), d = 0.42 nm] was observed for
the c form. The a1 peak is sensitive to the distance between
Crystalline form hydrogen-bonded chains inside the sheet-like structures
whereas the a2 peak is sensitive to the separation distance
PA6 is known to crystallize into various phases: the a between the sheets [21].
stable crystal and among others the c-phase [8, 19]. In the Figures 15 and 16 show the XRD patterns of pure PA6
a-phase the hydrogen bonds are formed between antipar- and PA6FS hybrid nanocomposites under different cooling
allel chains whereas in the c-phase they are formed conditions, respectively. The results indicate that in the
between parallel chains. XRD was used to gain insight into
log (K/min)
0.8 0.8
0.6 0.6
0.4 0.4
-0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6
log t (min) log t (min)
Fig. 12 Plot of log/ as a function of logt for pure PA6 Fig. 13 Plot of log/ as a function of logt for PA6S nanocomposites
123
9090 J Mater Sci (2007) 42:90839091
1(200)
1.4
Xc=0.3
2(002)
Xc=0.4
1.2 Xc=0.5
Xc=0.6
Intensity(*1000 cps)
Xc=0.7
1.0
log (K/min)
0.8
1
0.6
2
0.4 3 2(200)
Fig. 14 Plot of log/ as a function of logt for PA6FS nanocomposites Fig. 15 XRD patterns of pure PA6 under various cooling conditions:
(1) oil-bath cooling, (2) air cooling and (3) water cooling; the curves
Table 3 Values of F(T) and a at various relative crystallinity for were vertically offset for clarity
pure PA6, PA6S and PA6FS
2(002)
Xc F(T) a
1(200)
Pure PA6
0.3 9.68 1.30
Intensity(*1000 cps)
PA6FS 2(Deg)
0.3 8.44 1.19
Fig. 16 XRD patterns of PA6FS nanocomposites under various
0.4 9.23 1.21 cooling conditions: (1) oil-bath cooling, (2) air cooling and (3) water
0.5 10.34 1.23 cooling; the curves were vertically offset for clarity
0.6 10.96 1.24
0.7 11.01 1.26 reduced [8]. Vaia et al. [22] suggested that the addition of
silicate clay layers forces the amide groups of PA6 out of
the plane formed by the chains. This results in conforma-
slow and the medium cooling rates (cooling in oil-bath and tional changes of the chains, which limits the formation of
in air), there exists only the a form, while cooled rapidly H-bonded sheets and the c-phase is favored. We think that
(quenched in water), the a and c form co-existed in pure the same reason result in the c-phase is favored in PA6FS
PA6 (Fig. 15). The a and c form coexist in PA6/modified hybrid nanocomposites.
silica hybrid nanocomposites films under air cooling rate.
But the c form becomes dominant in higher cooling rate
(Fig. 16). Under cooling in oil-bath, in pure PA6, the a1 Conclusion
peak is higher than the a2 peak (Fig. 15, curve 1). How-
ever, in PA6/modified silica hybrid nanocomposites film, A systematic study of the non-isothermal crystallization
the a1 peak is lower than the a2 peak (Fig. 16, curve 1). kinetics of pure PA6, PA6S and PA6FS has been per-
This means that the extent of hydrogen-bonded NH asso- formed by the DSC technique. The crystallization kinetics
ciated with the a-phase in PA6FS hybrid nanocomposites is of each sample was investigate according to two different
123
J Mater Sci (2007) 42:90839091 9091
kinetic models, namely, the Avrami and the Mo. Both 3. Yang F, Ou YC, Yu ZZ (1998) J Appl Polym Sci 69:355
models can describe the experimental data very well. 4. Balazs AC, Singh C, Zhulina E (1998) Macromolecules 31:8370
5. Hajji P, David L, Gerard JF, Paul C (1999) J Polym Sci B37:3172
The DSC results imply that the nano-sized SiO2 in situ 6. Reynaud E, Jouen T, Gauthier C, Vigier G, Varlet J (2001)
generated and functionlized by GPTES in PA6FS hybrid Polymer 42:8759
nanocomposites act as nucleation agents and accelerates 7. Li Y, Yu J, Guo ZX (2002) J Appl Polym Sci 84:827
the crystallization during the cooling process from the 8. Wu QJ, Liu XH, Lars AB (2002) Polymer 43:2445
9. Zhang P, Wang XY, Wei SS (2004) A direct method for synthesize
melt. But the introduction of silica nanoparticles does not PA6/inorganic nanocomposites. China Patent: ZL021398369.4
have great effect on the onset temperature of crystalliza- 10. Zhao CX, Zhang P, He JP, Lu SR, Wang XY (2007) Poly Mater
tion. The addition of silica influences the mechanism of Sci Eng (Chinese) 23(1):218
nucleation and the growth of polyamide crystallites. 11. Liu XH, Wu QJ (2002) Eur Poly J 38:1383
12. Fornes TD, Paul DR (2003) Polymer 44:3945
XRD results indicated that the addition of functionalized 13. Tol RT, Mathot VBF, Groeninckx G (2005) Ploymer 46:2955
nano-silica particles favored the formation of the c crys- 14. Avrami M (1940) J Chem Phys 8:212
talline form. 15. Celina M, George GA (1995) Degrad Stab 48:297
16. Ozawa T (1971) Polymer 12:150
Acknowledgements The financial support from the National Nat- 17. Liu JP, Mo ZS (1991) China Polym Bull (in Chinese) 4:199
ural Science Foundation of China (No. 10372087), Scientific Re- 18. Liu TX, Mo ZS, Wang SG et al (1997) Polym Eng Sci 37:568
search Fund of Hunan Provincial Education Department (No. 19. Men YF, Rieger J (2004) Euro Poly J 40:2629
05B004) and the College Students Innovation Foundation of 20. Campoy I, Gomez MA, Marco C (1998) Polymer 39:6279
Xiangtan University is greatly acknowledged. 21. Murthy NS, Curran SA, Aharoni SM, Minor H (1991) Macro-
molecules 24:3215
22. Lincoln DM, Vaia RA, Wang ZG, Hsiao BS (2001) Polymer
42:1621
References
123
J Mater Sci (2007) 42:90929096
DOI 10.1007/s10853-006-1260-8
Received: 26 May 2006 / Accepted: 28 August 2006 / Published online: 4 August 2007
Springer Science+Business Media, LLC 2007
Abstract The significance of severe plastic deformation emerged as arguably the most promising one. While a large
by equal channel angular pressing (ECAP) as a promising amount of experimental data have been collected over the
technique for extreme grain refinement down to sub- past decade, often supported by finite element calculations
micrometer and sometimes nanometer scale is generally based on simplified constitutive models, microstructure
recognised. We present a modelling frame for describing based modelling has been relatively scarce. The present
ECAP based on microstructure evolution. Following a authors, together with colleagues, have been working in the
particular scenario of grain refinement, in which a dislo- field of ECAP modelling for some five years. This article
cation cell structure is considered as a precursor of the outlines the general frame of their modelling approach and
developing grain structure, the variation of the cell size with gives a summary of results, along with an outlook for
the number of ECAP passes is traced. Finite element sim- outstanding problems.
ulations based on the model compare favourably with the Among the possible scenarios of grain structure evolu-
experimental data. Further features of the model such as a tion under ECAP processing, we select a gradual
provision for modelling the variation of the misorientation transformation of the dislocation cell structure to a new,
angle distribution and texture evolution are also discussed. refined grain structure. This involves accumulation of
misorientation between neighbouring dislocation cells with
strain turning low angle boundaries to large angle ones
[35]. Prangnell et al. [6] proposed a picture in which a
Introduction decrease of lateral dimensions of elongated grains termi-
nates in their subdivision into smaller grains once the grain
Among the severe plastic deformation techniques used as a thickness drops down to the dislocation cell size. Basically,
means of extreme grain refinement in bulk metallic mate- in their model, too, it is the dislocation cell size that
rials [1, 2], equal channel angular pressing (ECAP) has determines the final grain size in the material undergoing
large strain deformation by ECAP.
The ECAP model providing a description of the dislo-
Y. Estrin (&) cation cell size evolution goes back to the strain hardening
ARC Centre of Excellence for Design in Light Alloys,
Department of Materials Engineering, Monash University,
model for large strains [7]. A brief outline of the model, as
Clayton, Vic. 3800, Australia well as a description of some recent developments, are
e-mail: yuri.estrin@eng.monash.edu.au given in the subsequent sections. Experimental data vali-
dating numerical simulation results are also presented there.
Y. Estrin
CSIRO Division of Manufacturing and Materials Technology,
Clayton, Vic. 3168, Australia
Outline of the microstructure evolution model
H. S. Kim
Department of Metallurgical Engineering,
Chungnam National University, Daejeon 305-764, Korea
The model of large strain deformation [7] underlying our
e-mail: hskim@cnu.ac.kr approach applies to dislocation cell-forming materials. The
123
J Mater Sci (2007) 42:90929096 9093
initial stage of the process when a primordial cell struc- statistical dislocations, become sharper. This effect
ture is formed is not considered. The cells of average size d outstrips the concurrent increase of the total grain
are assumed to have a cubic shape. The volume fraction of boundary area, so that the volume fraction of the wall
the cell walls is given by material turns out to be a decreasing function of strain,
which can be represented [2, 3] by
d 3 d w3
f ; 1
d3 f f1 fo f1 expc=~c: 6
where w denotes the wall thickness. The evolution of the This function was chosen to fit available experimental
dislocation densities in the cell walls and the cell interiors, data for Cu and describes the variation with strain c of the
denoted respectively qw and qc, is described by the set of volume fraction f from an initial value fo to an asymptotic
coupled differential equations value f?, which is smaller than fo. The parameter ~c
p p 1=n represents the inverse of the rate of this variation.
dqw 6b 1 f 2=3 3b 1 f qw c_ Strain hardening is considered by relating the equivalent
k0 qw ;
dc bdf fb c_ 0 resolved shear stresses src and srw in the cell interiors and the
2 cell walls, respectively, to the equivalent resolved plastic
shear rates in these phases (both assumed equal to c_ ) and
p 1=n the respective dislocation densities:
dqc 1 qw 6 c_
a p b k 0 qc :
dc 3 b bd1 f 1=3 c_ 0 1=m
p c_
3 src aGb qc ; 7
c_ 0
In the above equations it was assumed that the shear 1=m
strain c in the cell walls and the cell interiors is the same. p c_
srw aGb qw : 8
The quantity c_ o denotes a reference shear rate and b the c_ 0
magnitude of the Burgers vector. The quantities a* and b*
Here G is the shear modulus, 1/m is the strain rate
are numerical constants. The physical origin of the various
sensitivity parameter and a is a numerical constant. The
terms in the evolution equations representing possible
strain hardening behaviour of the composite is defined by
dislocation reactions involved (such as loss of cell interior
a scalar quantity, sr, that is obtained by applying the rule of
dislocations to become entrapped in the walls given by the
mixtures:
first term of Eq. 1 and the second term of Eq. 2) has been
explained elsewhere [35, 7]. The exponent n in the last, sr f srw 1 f src : 9
dynamic recovery, terms in both equations can be taken to
be inversely proportional to the absolute temperature, given The model outlined was shown to describe strain
the fact that ECAP processing is typically conducted below hardening behaviour of dislocation-cell forming metallic
half the melting temperature, while the coefficient ko can materials, including late stages of hardening, very well [7].
be considered constant. It is particularly suited for simulating ECAP processing, in
Whereas Eqs. 2 and 3 describe the detail of the evolution which very large strains are involved.
of the dislocation density, that of the other microstruc- In the scenario of grain refinement adopted, the smallest
tural variables, d and f, involves certain assumptions. dislocation cell size achievable for a given material pre-
Following Ref. [7] we assume that the average grain size d determines the eventual grain size. With the highest dis-
scales with the inverse square root of the total dislocation location density not exceeding 2.5 1015 m2, cf. [8], and
density K of the order of 10, the smallest grain size in copper
ECAP can produce would be, as predicted by Eq. 5, of the
q f qw 1 f qc ; 4 order of 250 nma value found experimentally [4].
p Equation 5 can roughly be re-written as
d K= q; 5
d G
where K is a constant. With the variation of the average KMa ; 10
grain size thus defined, a description of the variation of f b r
requires the knowledge of the evolution of the wall where r is the flow stress and M is the Taylor factor. From
thickness w. The latter quantity undergoes a gradual this relation it is easily seen that for the grain size to
decrease with straining, as the cell walls, initially fuzzy become smaller than a 100-fold of the lattice parameter,
due to the abundance of geometrically unnecessary say, the stress should become higher than one tenth of the
123
9094 J Mater Sci (2007) 42:90929096
Fig. 3 Simulated (right) and experimental (left) (111) pole figures for
Fig. 1 TEM micrograph of the grain structure in copper after 8 the uniform part of a copper billet deformed by 2 and 4 ECAP passes
ECAP passes (Route Bc [9]) (Route A [11])
123
J Mater Sci (2007) 42:90929096 9095
7
growth of qgw was written in the form Calculation 7
6
6
6b c_ c 1 f 2=3
q_ gw n : 13 5 5
bdf
4 4
It should be noted that this evolution equation does not
3
include any recovery terms, as geometrically necessary 3
2
dislocations are assumed to be unrecoverable. Adjustments 2 1
of the evolution equations for the statistical dislocation
density in the walls and the total dislocation density in cell 1
interiors are straightforward [12]. A very simple equation,
0
0 1 2 3 4 5 6 7 8
dh v 1
; 14 Accumulated Equivalent Strain
dc 2 h
Fig. 4 Evolution of the average misorientation angle with strain (or
that follows from Eqs. 12 and 13 can easily be solved. Here the number of ECAP passes in a 90 ECAP die) for copper:
the notation calculation results for b 6b vis-a`-vis experiment [13]
123
9096 J Mater Sci (2007) 42:90929096
Of great interest for modelling the effects of ECAP on structure and the effect of this evolution on texture devel-
the material microstructure is the misorientation angle opment. It is the belief of the authors that these aspects of
distribution. A provision for a probabilistic description modelling of ECAP deformation should be among the main
involving distribution functions can be made in terms of a targets of research in this area in a near future.
Fokker-Planck equation [12, 14]. A particular form of the
Fokker-Planck equation, derived using the Langevin Acknowledgements The authors have greatly benefited from col-
laboration and innumerable discussions with Y. Brechet, A. Molinari,
approach [12], reads L. Toth, R. Hellmig and many other colleagues to whom they wish to
express their sincere gratitude. Financial support from DFG (grant ES
oPh; c kT o2 Ph; c o 1 74/122) and Korea Research Foundation Grant (KRF-2005-202-
FPh; c ; 16 D00205) is gratefully acknowledged.
oc g oh2 oh g
where Ph; c represents the misorientation angle
distribution function for a given value of the shear strain References
c, while the quantities g and Fare given by
1. Valiev RZ, Islamgaliev RK, Alexandrov IV (2000) Prog Mater
v Sci 45:103
g 2 17 2. Valiev RZ, Estrin Y, Horita Z, Langdon TG, Zehetbauer MJ, Zhu
2h
YT (2006) JOM 58:33
and 3. Baik SC, Estrin Y, Kim HS, Hellmig RJ (2003) Mater Sci Eng A
351:86
1 dv 4. Baik SC, Hellmig RJ, Estrin Y, Kim HS (2003) Z Metallkd
F : 18 94:754
2h dc 5. Kim HS, Estrin Y (2005) Mater Sci Eng A 410411:285
6. Prangnell PB, Bowen JR, Apps PJ (2004) Mater Sci Eng A 375
A discussion of this probabilistic approach and an 377:178
analysis of the ECAP history on the misorientation angle 7. Toth LS, Molinari A, Estrin Y (2002) J Eng Mater Technol
distribution will be given elsewhere. 124:71
8. Gubicza J, Balogh L, Hellmig RJ, Estrin Y, Ungar T (2005)
Mater Sci Eng A 400401:334
9. Hadzima B, Janecek M, Hellmig RJ, Kutnyakova Y, Estrin Y
Conclusion (2006) Mater Sci Forum 503504:883
10. Gendelman OV, Shapiro M, Estrin Y, Hellmig RJ, Lekhtmakher S
In a snapshot presentation of our research on the evolu- (2006) Mater Sci Eng A 438:88
11. Baik SC, Estrin Y, Hellmig RJ, Jeong H-T, Brokmeier HG, Kim
tion of the microstructure and texture during severe plastic HS (2003) Z Metallkd 94:1189
deformation by equal channel angular pressing, we 12. Estrin Y, Toth LS, Brechet Y, Kim HS (2006) Mater Sci Forum
attempted to show the recent successes of modelling and 503504:675
also indicate some outstanding problems. These involve 13. Hellmig RJ, Kim HS, Estrin Y (2005) Trans Indian Inst Met
58:1107
particularly the understanding and modelling of the evo- 14. Pantleon W (2005) Mater Sci Eng A 400401:118
lution of misorientation angles in the dislocation cell
123
J Mater Sci (2007) 42:90979111
DOI 10.1007/s10853-006-1261-7
Received: 2 June 2006 / Accepted: 9 November 2006 / Published online: 4 August 2007
Springer Science+Business Media, LLC 2007
Abstract Results on the microstructure and the tensile post-ECAE compression and a saturation in the strain rate
properties of equal channel angular extruded (ECAE) sensitivity after N = 4 resulting in maximum values of
copper processed for one to 16 passes are presented and *0.02. Our results suggest that rather than a change in
compared with the available literature data. With increas- deformation mechanism, the increase in ductility with
ing number of passes (N), the microstructure changes from increasing N is associated with an increase in the mean free
a strongly elongated shear band structure after N = 1 and 2, path of dislocationswith the grain boundaries remaining
towards a more equiaxed subgrain and grain structure. This actively involved as the transmitter of plastic strain and
is accompanied by a decrease in the cell wall or subgrain- their interaction with dislocations being the rate controlling
boundary widths and an increase in recovered or even deformation mechanism.
recrystallised grain structures with low dislocation densi-
ties. Electron backscatter diffraction measurements have
indicated that for lower N, the location of R3 boundaries is
restricted to shear bands, while at greater N, R3 boundaries Introduction
were found to be more widely distributed. Texture mea-
surements indicate close similarity with simple shear Copper (Cu) represents an ideal model material to study the
texture components and a spread of the orientation com- process of deformation and recrystallisation due to its
ponents with greater N. Upon comparing the tensile intrinsic softness and formability, low cost, fcc crystal
behaviour of as-ECAE Cu with the surveyed literature, structure and medium stacking fault energy. In early
broad agreement on the strength of the material is studies on large plastic deformation, Cu was often chosen
achieved. However, a strong variation in the percentage for deformation experiments which were conducted using
elongation to failure is also noted. Strain hardening and conventional methods such as rolling, torsion, extrusion or
deformation kinetic analysis via strain rate jump tests compression, all of which are capable of imparting large
indicate an evolution from stage III to V hardening during strains to the workpiece ([13] and the references therein).
These and other studies conducted between 1950 and 1980
represent the basis of our understanding of the materials
F. H. Dalla Torre (&)
Laboratory of Metal Physics and Technology, Department of
properties and its associated microstructure resulting from
Materials, ETH Zurich, Wolfgang-Pauli-Str. 10, Zurich 8093, severe plastic deformation (SPD), a term coined by Russian
Switzerland researchers in the 1980s (see [4] among others). Compared
e-mail: Florian.DallaTorre@mat.ethz.ch to classical deformation processes, the big advantage of
A. A. Gazder E. V. Pereloma C. H. J. Davies
SPD techniques (represented in particular by equal channel
Department of Materials Engineering, Monash University, angular extrusion (ECAE) [5, 6] and high pressure torsion
Melbourne, VIC 3800, Australia (HPT) [79]) is the lack of shape-change post-deformation
and the consequent possibility to impart extremely large
A. A. Gazder E. V. Pereloma C. H. J. Davies
Victorian Centre for Advanced Materials Manufacturing,
strains. Thanks to Gleiters research on nanocrystalline
Belmont, VIC, Australia materials [10, 11], SPD has received enormous interest
123
9098 J Mater Sci (2007) 42:90979111
123
J Mater Sci (2007) 42:90979111 9099
many cases, regions of similar orientation have similar Microstructure of ECAE processed Cu
colours; however, this is not always the case due to the
discontinuities in Euler space [19]. In this section, a survey on ECAE Cu is presented with the
Texture analysis was performed on 20 20 3 mm3 idea of imparting a general description of the microstruc-
ED-samples using a GBC-MMA texture goniometer ture and to highlight the typical features found by several
working at 40 kV and 25 mA and equipped with a CuKa research groups. For comparison of subgrain and grain size
anode and a polycapillary beam enhancer, resulting in a measurements, the available literature data has been
collimated beam of 10 10 mm2. Raw data in the form of compiled in Table 1.
incomplete pole figures (PFs) (v = 080) from the (111), In the early 1990s, studies on the microstructure of SPD
(200), and (220) crystallographic planes were used to cal- metals produced via HPT or ECAE were conducted by the
culate the orientation distribution functions (ODFs) f(g) Russian group led by Prof. R.Z. Valiev. The research was
using the spherical harmonic method without imposing any mainly focussed on characterising the end-physical prop-
sample symmetry conditions. g = (/1, /, /2) refers to erties of SPD materials after a minimal (preferably, nano-
grain orientation denoted by the three Euler angles in sized), homogeneous grain structure was reached. The
Bunges notation [21]. All end-texture PFs and ODFs were differences between these ultra-fine grained (UFG) met-
subsequently rotated to TD-view and will be discussed in als and their coarse-grained counterparts [2224] in terms
the xyz laboratory system (Fig. 1). Correspondingly, the of their elastic moduli, diffusion coefficients, internal
TD, ED and ND-axes are denoted along the /1, / and /2 friction, Curie and Debye temperatures and superplasticity
directions, respectively. Further details regarding the ideal was highlighted and attributed to their small grain sizes and
texture components developed during ECAE are included the development of non-equilibrium grain boundaries [23,
in Table 2. 2527].
In order to produce tension and compression specimens, One of the first mentioned studies on ECAE Cu pro-
the billets were first sectioned into quarters. Three such cessed at RT up to N = 4 passes using route BC led to an
sections were used to produce the circular tensile speci- equiaxed grain structure and an average grain size of
mens with a diameter of 4 mm and a gage length of 210 nm [27]. The authors observed that smaller grains
25 mm. Correspondingly, the center of the fourth section (<100 nm) with a volume fraction of *15% were absent of
was used to produce cylindrical compression test samples lattice dislocations, while larger grains (400500 nm) with
of 4 mm diameter and 8 mm length. The tensile tests were a volume fraction of *48% indicate the presence of a cell
performed at RT on a screw-driven Instron operating at a structure with subgrain boundaries and the presence of
cross head speed of 2.5 mm min1, resulting in a starting lattice dislocations having a mean density inside grains of
strain rate of e_ 1:6 103 s1 : The strain was measured 5 1014 m2 [27]. About 37% of the volume is reported to
with a clip-gage mounted directly onto the tensile speci- be constituting of intermediate grains of 100200 nm in
mens, yielding a Youngs modulus of about 130 GPa, in size with a chaotic dislocation structure [27].
close agreement with the theoretical value of Cu. It should Similarly, Ferasse et al. [14] also investigated the effect
be noted that while the gage length is situated well within of the number of passes on pure Cu and concluded that
the uniformly sheared stable billet length for N 8, the after N = 4, grain refinement reaches a minimal size of
gage lengths of specimens machined after N = 12 and 16 200300 nm, concurrent with a stabilisation of strength
could contain non-uniformly sheared volumes. For the and hardness. Their results on the grain size measured by
compression tests, a cross head speed of 0.5 mm min1 TEM were later confirmed by Mishin and Gottstein [28]
was used on the same screw-driven Instron resulting in an who, apart from dislocation density estimates of
initial strain rate of e_ 1 103 s1 : Cyclic strain rate 5 1014 m2 (in agreement with the value given by [25,
jump tests in compression were performed on a hydraulic 27]), reported the occurrence of twins formed during
Instron to evaluate the influence on strain rate sensitivity of ECAE and a major fraction of random high-angle grain
N. The strain rate jumps were performed between 1 104 boundaries (HAGBs). However, the latter observation has
and 1 102 s1. In both type of compression tests, the been proven incorrect in a subsequent work by Mishin
strain was measured by the cross-head displacement only, et al. [29, 30] which concentrated on detailed misorienta-
which due to the machine compliance leads to a signifi- tion measurements of the ECAE substructure deformed up
cantly lower apparent Youngs modulus in the uncorrected to N = 8, at U = 90 via route BC. Misorientation mea-
stressstrain curves shown in section Mechanical prop- surements on 219 misorientations showed that 63% are
erties of ECAE-Cu. Beyond 30% strain a steady increase low-angle grain boundaries (LAGBs) and that the intra-
in the stress with strain has been associated with the granular volumes are never surrounded solely by HAGBs.
gradual increase in frictional forces. Hence only data below Statistical measurements revealed a HAGB spacing per-
30% strain were considered as valid for the analysis. pendicular to the dominant band structure between 0.2 lm
123
9100 J Mater Sci (2007) 42:90979111
Table 1 Grain and subgrain size measurements of ECAE-Cu for varying number of passes and processing routes taken from various references
and own data
Processing parameters Method Avg. grain Avg. subgrain / % volume Ref.
a
size (nm) cell size (nm) HAGB
ECAE die angle Passes (N) Route Sample plane
Inner angle (U) Outer angle (W)
and 2 lm. In agreement with the comprehensive investi- [3644]. However, in most of these studies the grain size
gation of Agnew [31], Mishin et al. [29, 30] concluded has been defined by the shape of the boundaries rather than
that: (i) the reported grains size of 0.20.3 lm measured on by an accurate statistical misorientation measurement.
ECAE Cu included boundaries of high- and low- angle and, These results indicate an average size between 200 nm and
(ii) ECAE deformed metals should rather be characterised 400 nm with a typical log-normal distribution function
in terms of deformation structures rather than be classified spread from 100 nm to 600 nm (see Table 1). In order to
as a UFG material. These authors [29, 31] further differentiate between the high and low misorientation
recognised that ECAE Cu exhibits elongated subgrain angles of the boundaries, dark-field (DF) images were
structures, which are subdivided transversely by low-angle proposed for grain size measurements. However, these
boundaries similar to what has been observed as incident studies also yielded similar grain sizes to previous calcu-
dislocation boundaries in cold-rolling deformed structures lations based on BF images [38].
[32]. An important point highlighted in [33] concerns the The microstructural evolution of Cu processed via route
technique of measuring misorientations by means of TEM C for N = 18 passes has been studied by Baik et al. [36]
selected area diffraction (SAD) patterns. Measurements in order to deduce the mean cell size as a function of N.
based on SAD patterns as performed in [13, 34, 35] do not They observed a minimum value of *200 nm peak cell
account for misorientations adjacent to a grain, leading to sizes already after the first pass, but indicated a decrease in
an overestimation of the volume of HAGBs in ECAE the spread of the cell size distribution with increasing N.
metals [33]. This was explained on the basis of the spatial On the other hand, using a U = 90 die Huang et al. [41]
limitation in conventional TEM studies with large dia- studied the influence of the strain path via routes BC and C
phragms (in the micrometer-range) yielded diffraction for up to N = 8 by comparing the microstructural differ-
patterns from multiple boundaries (which in bright-field ences with respect to sample orientations x and z (see
(BF) mode were shown to be spaced only few hundreds of Fig. 1). They showed that while the x-plane of the micro-
nanometers apart). structure of route BC and C samples exhibited a nearly
Several other TEM studies have quantitatively repro- equiaxed cell/subgrain structure, the z-plane indicated a
duced microstructures similar to those characterised above higher fraction of elongated cells/subgrains for both routes
123
J Mater Sci (2007) 42:90979111 9101
by approximately the same amount. In agreement with has been reproduced more quantitatively by EBSD mis-
previous results [1315], they observed a smaller grain and orientation measurements (Fig. 4a).
cell/subgrain size and a higher fraction of HAGBs for route After N = 4 and 8, the microstructure is inhomogeneous
BC compared to route C. Analysis of 100 boundaries with respect to the various grain shapes. In some areas
indicated that LAGBs always showed a diffused appear- grains and subgrains are still elongated, while other areas
ance due to the dense dislocation forests at the boundaries. are representative of equiaxed grain and substructures
Alternatively, HAGBs comprised sharp contrast and (Fig. 3c, d). With greater N, the volume fraction of equi-
mainly dislocation-free adjacent areas [41]. Their obser- axed microstructural features increases and the distribution
vation and the statistical TEM misorientation of the diffraction spots is more uniform. This change
measurements of Mishin et al. [29] now serve as a guide in diffraction infers an increase in the misorientation
for qualitatively measuring the misorientation from the (Fig. 4bd), and subsequently recovery within the grains
boundary morphology in order to achieve higher statistics. and their grain boundaries (Fig. 4e, f). An enhancement of
recovered grain structures showing low dislocation densities
in the interior and sharp grain boundaries with increasing N
Microstructural evolution of as-ECAE Cu is related to the higher effectiveness of dislocation annihi-
lation due to an increase in the boundary misorientations.
Upon adopting the methodology of Huang et al. [41], our This in turn, is manifested by a decrease in the average
microstructure investigations [18, 45] on ECAE Cu for up boundary width from 52 nm to 34 nm and a decrease in
to N = 16 via route BC have yielded similar results. the dislocation density (6.9 1013 m2) at N = 16 after
Figure 2 shows that cell boundaries of low-angle configu- reaching a maximum (3.2 1014 m2) for N = 4 (see
ration (i.e. LAGBs) are identified depending on the Ref. [18] for details). Furthermore, in agreement with [41]
orientation of the adjacent cells either by clearly distin- for route C,1 some larger (up to 500 nm) grains character-
guishable high densities of dislocations and dislocation- ised by sharp and straight boundaries were observed for
debris or by fuzzy diffraction contrasts (), while the samples with N 8 and have been said to be precursors to
sharper boundaries (c) (often also confined to grains where recrystallised grains [18]. Also, stacking fault tetrahedra and
thickness fringes are visible), are considered to be HAGBs dislocation loops observed in larger grains containing only
[41]. As seen in the coupled TEM images and SAD pat- few lattice dislocations, lend further support to the occur-
terns after N = 1 and 2 (Fig. 3a, b), the microstructure is rence of recovery processes (Fig. 4c, d in Ref. [18]).
dominated by lamellar boundaries (LBs) oriented in the To improve the statistics on the variation of grain size
h110i direction (along the trace of the {111} plane), due to recovery and/or recrystallisation structures occur-
resulting in a microstructure similar to rolling [32]. Dis- ring during ECAE we performed EBSD measurements
location cell boundaries are evident and as indicated by the over a 40 30 lm area. Figure 5 shows a section taken
SAD pattern mainly LAGBs are present. This observation from this larger EBSD map of the N = 16 specimen. The
black lines represent HAGBs, while thicker red lines cor-
respond to R3 boundaries. The few larger grains (up to
1 lm in size) with a high abundance of R3 boundaries are
surrounded by smaller grains *170 nm in size. Over the
entire map of 40 30 lm about 12% of the grains can be
classed as recrystallised grains with a mean size of 1.0 lm,
based upon the qualifying assumptions that: (i) the
recrystallised grain size is six times larger than the mean
subgrain size, (ii) a recrystallised grain is bounded by at
least 10% HAGB and, (iii) that the image quality remains
constant.2 A similar observation of such grains of 1 lm
size with a volume fraction of 1.5% have recently been
reported for a Cu specimen subjected to N = 8 passes,
route BC [46]. From the current standpoint, the processes
that lead to the formation of such large grains requires
1
They measured a volume fraction of 1.5% represented by dynam-
Fig. 2 TEM bright field (BF) image showing cell wall boundaries ically recrystallised grains.
2
and/or subgrain boundaries () and high angle grain boundaries (c) With increasing fraction of recrystallised grains, the image quality
in a specimen deformed up to N = 16 is expected to improve.
123
9102 J Mater Sci (2007) 42:90979111
V o lu m e f r a c ti o n [ % ]
20 20
10 10
0 0
0-4
4-8
8-12
12-16
16-20
20-28
28-32
32-36
36-40
40-44
44-48
48-52
52-56
56-60
60-64
0-4
4-8
8-12
12-16
16-20
20-28
28-32
32-36
36-40
40-44
44-48
48-52
52-56
56-60
60-64
40 40
V o l u m e f r a c ti o n [ % ]
V o l u m e f r a c ti o n [ % ]
30 30
20 20
10 10
0 0
0-4
4-8
8-12
12-16
16-20
20-28
28-32
32-36
36-40
40-44
44-48
48-52
52-56
56-60
60-64
0-4
4-8
8-12
12-16
16-20
20-28
28-32
32-36
36-40
40-44
44-48
48-52
52-56
56-60
60-64
123
J Mater Sci (2007) 42:90979111 9103
123
9104 J Mater Sci (2007) 42:90979111
boundaries [65]. In their model, Meyers and Murr [65] might suggest that these serve as nuclei for the growth of
explain the driving force for the nucleation of twins by the annealing twins. However, this caveat requires the defor-
reduction in dislocation density, which is again in agree- mation and annealing twins to be crystallographically
ment with our observation of lower dislocation densities in identical as they consist of stacking faults on consecutive
larger grains as those shown in Fig. 7. Here, a clearer {111} planes [66].
identification of the nucleation mode (i.e. annealing or
deformation) is not possible. In Fig. 7c, two stacking faults Texture evolution in ECAE-Cu
are shown crossing an entire grain. Considering the
observations made by TEM and EBSD, it is important to Figure 8 shows the (111) experimental pole figure of the
emphasise the difference in appearance between these 0-pass billet prior to ECAE. Similar to a previous study
twins and those associated with the large grain shown in [67], this initial texture is weak and consists mainly of a
Fig. 5. It is however interesting to note that R3 boundaries h111i h001i duplex fiber with the fiber axis parallel to
also occur within the fine grained matrix (Fig. 5), which the billet longitudinal axis.
123
J Mater Sci (2007) 42:90979111 9105
Table 2 Ideal orientations and fibers for fcc materials after a single pass 90 ECAE
Notation Euler angles () for /2 = 0 and 45 sections Equivalent TD-rotation Fiber family
by h = 45
/1 / /2
123
9106 J Mater Sci (2007) 42:90979111
{111} A*2
*
A 1 A
Max = 3.02
A
(d) ND (e) ND
ED ED
densities between N = 4 compared to N = 12 and 16 which the texture scan is performed [67, 75] also mean that
(Fig. 10d, e) is that greater strain not only produces a net direct comparisons of our texture results with those found
spread of orientation components but also results in a in literature must be made with care.
weakening of all ideal shear-type orientations in agreement
with previous work [67, 7274]. Interestingly and common
to N = 4, 8, 12 and 16 pass conditions, the re-appearance Mechanical properties of ECAE-Cu
of weak texture components of the initial N = 0 texture
reflects glide-reversal effects that return part of the During ECAE processing, accurate stress and strain levels
substructure into an equiaxed configuration. cannot be quantified due the operation of friction forces at
It must be noted that the weakening trend of f(g) is in the billet-die interface. Consequently, variations in material
contradiction with other investigations [75, 76] that pos- properties as a function of the employed strain-path chan-
tulate an increase in texture strengthening with greater ges between successive passes cannot be properly
imparted strain. Mainly, these disparities allude to the evaluated. From strain reversal experiments, it is well
heterogeneity of the as-deformed billet due to varying known that a change in the strain path produces a change in
friction conditions at the billet-die interface [69, 73, 77] the internal stressgenerally described as the Bauschinger
and also the degree of difficulty in locating the exact center effect. In cases, where the reloading stress exceeds the pre-
of the billet stable length (from where the ED-plane sample deformation stress level, a dissolution of the previous cell
is cut) for equivalent texture scans with increasing N. In structures and a untangling of formerly strong dislocation
addition, differences such as: (i) the starting texture of the arrangements within the walls is accompanied by a stag-
material at N = 0 [74] and, (ii) the area of the sample over nation of the strain hardeningbefore a steady state is
123
J Mater Sci (2007) 42:90979111 9107
UTS [MPa]
40
350
important to account for such effects. 35
300 30
250 25
Tensile testing 20
YS,
200
Uniform,
15
150
Figure 11 shows the true strain true stress curves in (a) 10
100
tension and, (b) compression derived from engineering 5
50 0
stressstrain curves using standard equations. Alterna-
tively, Fig. 12 shows a quantitative summary of tensile test 0 2 4 6 8 10 12 14 16
300 N = 16
[79].
Slightly longer post-necking elongation values, but still
much larger than the uniform strain, have been recently
200
reported for specimens deformed for N = 16 and 20 via
route BC (etot * 15%) and for N = 2 to 4 via route C
100 (etot * 1719%) [37, 8082]. Yet, significantly larger total
elongation has been reported for higher numbers of passes
0 using route BC [83]. Recently, a total elongation value of
0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14
etot * 35% (which exceeds that of the annealed N = 0
True strain sample) was presented for Cu samples subjected to routes
(b) 500 C and E for N = 8 and 16 passes processed in a U = 120
angle die [84]. Since the tensile sample dimensions in these
400 results are not indicative of any consistent size effect, other
reasons have to be considered to explain the measured
True stress [MPa]
300
differences in total and/or uniform elongation. Note that
the UTS for the tensile strains given in [37, 8084] are all
within the same range as measured here, indicating that the
200 N=0
N=1 enhanced strain is not due to a larger grain size.
N=2
N=4
Wang and Ma [40] suggested that an increase in the
100 N=8 strain rate sensitivity (SRS) would be one way to improve
N = 12
N = 16 the tensile ductility as superplastically deformed materials
0 have typical SRS values that range between 0.3 and 0.8.
0.00 0.05 0.10 0.15 0.20 0.25
Earlier, such an explanation was also given by Valiev et al.
True strain [83] for explaining the high tensile strength of their ECAE
Fig. 11 Stressstrain curves for 016 passes in tension (a) and Cu samples deformed at RT [83]. However, our data
compression (b) extracted from strain rate jump tests in compression shown
123
9108 J Mater Sci (2007) 42:90979111
Table 3 Compressive yield strength and strain rate sensitivity (m) of Cu specimens subjected to N = 116 passes ECAE (route BC)
Mechanical property Number of passes (N)
0 1 4 8 12 16
in Table 3 (also plotted in [85]), indicate that the SRS evolution of the deformation microstructure with respect to
value saturates at *0.02 and is in agreement with the N [18]. With increasing N (up to N = 16), a decrease in the
values given in [79] where smaller strain rate steps in- wall thickness and dislocation density within the subgrains
between the jumps had been used. It should be noted that occurs. All the while, the cell or subgrain size remains
such low values of SRS are not sufficient to delay defor- roughly constant. These findings are in agreement with a
mation instability (and consequent necking) according to corresponding decrease in the microstrain measured by
Harts criterion [86]. As pointed out by Wang and Ma, only XRD (see Ref. [18] for more details). This would strongly
if the strain rate is sufficiently low (*106 s1) and suggest that the mean free path of dislocations increases in
enough time is given for the operation of diffusional samples processed beyond N 4. It can be therefore
mechanismswill an increase in the SRS occur and pro- suggested that an increased volume fraction of HAGBs
duce an observable increase in the uniform elongation [40]. (which themselves are known to be effective sinks for
It has also been suggested that increasing the percentage incoming lattice dislocations), and a lower dislocation
volume fraction of HAGBs proves beneficial to the uniform density within subgrains (and hence, a greater mean free
elongation [83, 87]. A recent study on a Cu0.36Cr alloy path of dislocations) help reduce sources of plastic insta-
deformed by route A up to N = 16, exhibits uniform and bility such as stress concentrations and thereby makes
total elongations of *10% and *23% respectively, and a possible an increase in percent elongation. Apropos to the
UTS of 410 MPa. On the other hand, a sample deformed agreement between microstructural analysis of the ECAE
up to N = 16 via an alternated route C and A type pro- material performed before mechanical testing and conse-
cessing schedule, showed uniform and total elongations of quent mechanical behaviour during testing, it can be
*4% and *20%, respectively and a UTS of 435 MPa. assumed that microstructural changes occurring during the
EBSD studies on these samples indicate the prevalence of a few percentage of strain reached during tensile testing are
greater fraction of HAGBs (fHAGB *6775%) for the of minor importance for the interpretation of the stress
sample with greater uniform strain compared to the sample strain curves.
(fHAGB *5263%) exhibiting less uniform strain [88]. It should also be noted that while the above mechanisms
Qualitatively, this is in agreement with our observations could explain the mechanical behaviour of SPD-UFG
wherein a lower fraction of HAGB and a lower uniform materials to a very large extent, another source of incon-
elongation was measured for samples at N 4 compared sistency in the available literature data could originate from
to those deformed for N > 4. It had been suggested [83, 87] a higher material-specific sensitivity to external flaws, a
that a microstructure comprising mainly HAGBs would phenomena which is also known to affect nc metals [89].
promote deformation mechanisms other than those medi- Thus, further experimental results are still needed to
ated by dislocations. But taking the low magnitude of resolve these differences and/or ambiguities.
measured SRS values into consideration (see Table 3),
mechanisms which are solely dependent on diffusion pro-
cess are of minor importance. However, diffusion Strain hardening stages III, IV and V measured in
phenomena are indirectly connected to the major operative compression
mechanismwhich is characterised by dislocation glide
through the crystals and their accommodation and evolu- Unlike tensile tests, uniaxial compression tests do not
tion within the cell wall structure. The possibility of texture suffer from necking instability and can be used to describe
effects (whereby an increase in ductility via dislocation slip the strain hardening behaviour. In the following passages,
along more preferential orientations can be induced) is also the hardening behaviour of the ECAE deformed samples is
negated as the present texture results show a generally described in the classical context of strain hardening theory
route-specific and reproducible trend of preferred orienta- reviewed in [13]. In doing so, the effects of strain path
tions with increasing N. changes from simple shear (or, ideal ECAE) to pure shear
In order to further clarify this question, we have also (compression) are neglected. Nevertheless, this description
carried out detailed TEM and XRD analysis on the showed usefulness in describing the hardening behaviour
123
J Mater Sci (2007) 42:90979111 9109
(a) 1000 Table 4 Calculated shear stress values for the different hardening
N=1 stages of Cu specimens subjected to N = 116 passes ECAE, via route
N=2 BC processing
800 N=4
N=8 Shear stress Number of passes (N)
N = 12
600 N = 16 1 2 4 8 12 16
[MPa]
0
100 110 120 130 140 150 160 170
III S During stage III, a net increase in the dislocation density
[MPa] III S
is still prevalent, which when contrasted with stage II is
(b) 120 less dominant and bears a dislocation annihilation com-
N=1
100 N=2 ponent [13]. As compared to stage II, stage III is also
N=4 commonly described by a higher strain rate and tempera-
N = 16
80 ture sensitivity which increases with increasing stress and
is associated with processes of dynamic recovery [13, 91].
[M P a]
123
9110 J Mater Sci (2007) 42:90979111
Fig. 4 in Ref. [85]). Similar microstructural observations in show a slight increase in uniform elongation for greater N.
strain reversal experiments are also associated with the Despite the large amount of pre-strain imparted via ECAE,
Bauschinger effect. According to these observations, it is the strain hardening measured during compression tests
currently not clear how far the stage V hardening observed indicates a recovery of stage III, which is followed by stage
in Fig. 13 is influenced by the change in the strain path. IV on further deformation. For samples processed to higher
Conversely, as a change in the stress causes a corre- N, a hardening behaviour associated with stage V is also
sponding change in the SRS, the apparent activation observed. Strain rate sensitivity values of the order of 0.02
volume Vap kT D ln c_ Ds is also transformed for a given and activation volume estimates, which yield a length scale
change in the hardening stage (see Fig. 7 in [85]). For of the same order as deduced from dislocation spacings in
N > 4, activation volumes between 48 and 110 b3, where the cell wall boundaries suggest that the deformation
b is the Burgers vector, have been measured with lowest kinetics are dictated by the interaction of lattice disloca-
values reached at very low strain rates, which are achieved tions with the boundaries.
in relaxation tests (for further details, see [85]). These
values are also in agreement with those presented recently Acknowledgements Mrs. C. Gu is kindly acknowledged for assis-
tance in texture measurements. This work was supported by the Swiss
by Wei et al. [92]. Separating the effective activation National Science Foundation, Grant # 200021-105647 (FHDT), Mo-
volume into its constituents Veff b x l; (with x being nash International Post-Graduate Research Scholarship (AAG) and by
roughly equal b), a length scale (l) of *1530 nm results. the Australian Research Council Discovery Project, DP0557255
This length scale is in the same range as deduced from our (CHJD, EVP).
TEM analysis of the dislocation density within cell walls
(qwall *1.9 1015 m2 [18]) if it is assumed that
p References
lTEM 1= qwall : This gives credence to the assumption that
the rate controlling mechanism in ECAE Cu at higher N 1. Sevillano JG, Van Houtte P, Aernoudt E (1981) Prog Mater Sci
( 8) is the interaction of lattice dislocations with cell wall 25:69
or subgrain boundaries as also stated by [92, 93]. 2. Nes E (1998) Prog Mater Sci 41:129
3. Kocks UF, Mecking H (2003) Prog Mater Sci 48:171
4. Valiev RZ, Islamgaliev RK, Alexandrov IV (2000) Prog Mater
Sci 45:103
Summary 5. Segal VM, Reznikov VI, Drobyshevskiy AE and Kopylov VI
(1981) Russian Metall (Engl. Transl.) 1:115
This work summarises our and current understanding of the 6. Segal VM (1995) Mater Sci Eng A 197:157
microstructural evolution and the mechanical behaviour of 7. Zhorin VA, Shashkin DP, Yenikoponyan NS (1984) DAN, SSSR
278:144
ECAE pre-deformed Cu via route BC. While good agree- 8. Kuznetsov RI, Bykov VI, Chernyshov VP, Pilyugin VP, Yefre-
ment on the sub-grain/grain size and shape as a function of mov NA, Posheyev VV, Sverdlovsk (1985) IFM UNTS RAN,
number of passes (N) results from a comparison with Preprint 4/85 (in Russian)
available literature values, a less consistent picture can be 9. Bridgman PW (1952) Studies in large plastic flow and fracture.
McGraw-Hill, New York and London, p 174 (Reprinted by 870
deduced from misorientation measurements. Our results Harvard University Press, Cambridge, Mass., USA, 1964)
performed by electron backscatter diffraction (EBSD) 10. Birringer R, Gleiter H, Klein H-P, Marquardt P (1984) Phys Lett
indicate an increase in the amount of HAGBs with A 102:365
increasing N, which saturates at about 50%. In agreement 11. Gleiter H (1989) Prog Mater Sci 33:223
12. Lowe TC (2006) Mater Sci Eng 503504:355
with literature an increase in the amount of R3 boundaries, 13. Iwahashi Y, Horita Z, Nemoto M, Langdon TG (1998) Acta
stacking faults, recovered or even partially recrystallised Mater 46:3317
grains and decrease in boundary width with increasing N is 14. Ferrasse S, Segal VM, Hartwig KT, Goforth RE (1997) Metall
observed. Consistent with previous texture work a spread Mater Trans A 28:1047
15. Furukawa M, Iwahashi Y, Horita Z, Nemoto M, Langdon TG
in the texture components away from the ideal simple shear (1998) Mater Sci Eng A 257:328
components is observed with greater N. 16. Lapovok RYe (2005) J Mater Sci 40:341 DOI:10.1007/
Upon comparing with available literature on the s10853-005-6088-0
mechanical behaviour of ECAE Cu good agreement has 17. Bowen JR, Gholinia A, Roberts SM, Prangnell PB (2000) Mater
Sci Eng A 287:87
been found in terms of maximum yield and tensile stress 18. Dalla Torre FH, Lapovok R, Sandlin J, Thompson PF, Davies
reached. In this study a softening after N = 8, in agreement CHJ, Pereloma EV (2004) Acta Mater 52:4819
with microstructural evolution is observed. With respect to 19. Humphreys FJ (2001) J Mater Sci 36:3833 DOI:10.1023/
the maximum tensile strain reached before failure in ECAE A:1017973432592
20. Bowen JR, Mishin OV, Prangnell PB, Juul Jensen D (2002)
Cu larger disparities were found, which might be associ- Scripta Mater 47:289
ated with testing conditions and external flaws. Here, total 21. Bunge H-J (1982) Texture analysis in materials science: mathe-
elongations do not exceed 10% strain for all passes, but do matical methods, 1st edn. Butterworth & Co., Berlin
123
J Mater Sci (2007) 42:90979111 9111
22. Valiev RZ, Krasilnikov NA, Tzenev NK (1991) Mater Sci Eng A 55. Christian JW, Mahajan S (1995) Prog Mater Sci 39:1
137:35 56. El-Danaf E, Kalidindi SR, Doherty RD (1999) Metall Mater
23. Valiev RZ, Mulyukov RR, Ovchinnikov VV (1990) Phil Mag Trans A 30:1223
Lett 62:253 57. Randle V (2004) Acta Mater 52:4067
24. Valiev RZ, Kozlov EV, Mulyukov RR (1993) Mater Sci Eng A 58. Sutton AP, Balluffi RW (1996) Interfaces in crystalline materials.
168:141 Oxford Science Publications, Clarendon Press, p 305
25. Valiev RZ, Mulyukov RR, Ovchinnikov VV, Shabashov VA 59. Venables JA (1961) Philos Mag 6:379
(1991) Scripta Metall Mater 25:2717 60. Yamakov V, Wolf D, Phillpot SR, Mukherjee AK, Gleiter H
26. Valiev RZ, Korznikova GF, Mulyukov KY, Mishra RS, Muk- (2002) Nat Mater 1:45
herjee AK (1997) Phil Mag B 75:803 61. Chen MW, Ma E, Hemker KJ, Sheng HW, Wang YM, Cheng
27. Valiev RZ, Kozlov EV, Ivanov YuF, Lian J, Nazarov AA, Bau- XM (2003) Science 300:1275
delet B (1994) Acta Metall Mater 42:2467 62. Liao XZ, Zhou F, Lavernia EJ, Srinivasan SG, Baskes MI, He
28. Mishin OV, Gottstein G (1998) Philos Mag A 78:373 DW, Zhu YT (2003) Appl Phys Lett 83:632
29. Mishin OV, Juul Jensen D, Hansen N (2000) In: Hansen N, 63. Liao XZ, Zhao YH, Srinivasan SG, Zhu YT, Valiev RZ, Gun-
Huang X, Juul Jensen D (eds) Proc. 21nd Ris international symp derov DV (2004) Appl Phys Lett 84:592
mater. sci.: recrystallisation-fundamental aspects and relations to 64. Rosner H, Markmann J, Weissmuller J (2004) Phil Mag Lett
deformation microstructure. National Laboratory, Roskilde, 84:321
Denmark, p 445 65. Meyers MA, Murr LE (1978) Acta Metall 26:951
30. Mishin OV, Juul Jensen D, Hansen N (2003) Mater Sci Eng A 66. Mahajan S, Pande CS, Imam MA, Rath BB (1997) Acta Mater
342:320 45:2633
31. Agnew SR (1998) Ph.D. thesis, Northwestern University 67. Li S, Beyerlein IJ, Alexander DJ, Vogel SC (2005) Scripta Mater
32. Hughes DA, Hansen N (2000) Acta Mater 48:2985 52:1099
33. Mishin OV, Huang X, Brown JR, Juul Jensen D (2001) In: 68. Li S, Beyerlein IJ, Bourke MAM (2005) Mater Sci Eng A 394:66
Hansen N, Huang X, Juul Jensen D (eds) Proc. 22nd Ris 69. Li S, Beyerlein IJ, Necker CT, Alexander DJ, Bourke MA (2004)
internat. symp. mater. sci.: recrystallisation-fundamental aspects Acta Mater 52:4859
and relations to deformation microstructure. National Laboratory, 70. Agnew SR, Weertman JR (1998) Mater Sci Eng A 242:174
Roskilde, Denmark, p 335 71. Huang WH, Chang L, Kao PW, Chang CP (2001) Mater Sci Eng
34. Iwahashi Y, Horita Z, Nemoto M, Langdon TG (1997) Acta A 307:113
Mater 45:4733 72. Gholinia A, Bate P, Prangnell PB (2002) Acta Mater 50:2121
35. Komura S, Horita Z, Nemoto M, Langdon TG (1999) J Mater Res 73. Gazder AA, Dalla Torre FH, Gu CF, Davies CHJ, Pereloma EV
14:4044 (2006) Mater Sci Eng A 415:126
36. Baik SC, Hellmig RJ, Estrin Y, Kim HS (2003) Z Metallkd 94:759 74. Ferrasse S, Segal VM, Kalidindi SR, Alford F (2004) Mater Sci
37. Vinogradov A, Hashimoto S, Patlan V, Kitagawa K (2001) Mater Eng A 368:28
Sci Eng A 319321:862 75. Gubicza J, Balogh L, Hellmig RJ, Estrin Y, Ungar T (2005)
38. Mingler B, Karnthaler HP, Zehetbauer M, Valiev RZ (2001) Mater Sci Eng A 400401:334
Mater Sci Eng A 319321:242 76. Cao WQ, Godfrey A, Liu W, Liu Q (2003) Mater Sci Eng A
39. Hoppel HW, Zhou ZM, Mughrabi H, Valiev RZ (2002) Philos 360:420
Mag A 82:1781 77. Li S, Bourke MAM, Beyerlein IJ, Alexander DJ, Clausen B
40. Wang YM, Ma E (2004) Acta Mater 52:1699 (2004) Mater Sci Eng A 382:217
41. Huang WH, Yu CY, Kao PW, Chang CP (2004) Mater Sci Eng A 78. Shih MH, Yu CY, Kao PW, Chang CP (2001) Scripta Mater
366:221 45:793
42. Wu SD, Wang ZG, Jiang CB, Li GY, Alexandrov IV, Valiev RZ 79. Wang YM, Ma E (2003) Appl Phys Lett 83:3165
(2004) Mater Sci Eng A 387389:560 80. Haouaoui M, Karaman I, Maier HJ, Hartwig KT (2004) Metall
43. Han S, Lim C, Kim C, Kim S (2005) Metall Mater Trans A Mater Trans A 35:2935
36:467 81. Wang JT, Du ZZ, Kang F, Chen G (2006) Mater Sci For 503
44. Vinogradov A, Suzuki T, Hashimoto S, Kitagawa K, Kuznetsov 504:663
A, Dobatkin S (2006) Mater Sci For 503504:971 82. Krishanaiah A, Chakkingal U, Venugopal P (2006) Mater Sci For
45. Dalla Torre FH, Gazder AA, Gu CF, Davies CHJ, Pereloma EV, 503504:733
Met Mater Trans A (submitted) 83. Valiev RZ, Alexandrov IV, Zhu YT, Lowe TC (2002) J Mater
46. Etter AL, Solas D, Baudin T, Penelle R (2005) Mater Sci For Res 17:5
495497:845 84. Maier HJ, Gabor P, Gupta N, Karaman I, Haouaoui M (2006) Int
47. Baik SC, Estrin Y, Kim HS, Hellmig RJ (2003) Mater Sci Eng A J Fatigue 28:243
351:86 85. Dalla Torre FH, Pereloma EV, Davies CHJ (2006) Acta Mater
48. Chang CP, Sun PL, Kao PW (2000) Acta Mater 48:3377 54:1135
49. Zehetbauer MJ, Steiner G, Schafler E, Korznikov A, Korznikova 86. Hart EW (1967) Acta Metall 15:351
E (2006) Mater Sci For 503504:57 87. Valiev RZ (2003) Adv Eng Mater 5:296
50. Huang CX, Wang K, Wu SD, Zhang ZF, Li GY, Li SX (2006) 88. Vinogradov A, Ishida T, Kitagawa K, Kopylov VI (2005) Acta
Acta Mater 54:655 Mater 53:2181
51. Liao XZ, Zhao YH, Srinivasan SG, Zhu YT, Valiev RZ, Gun- 89. Dalla Torre FH, Van Swygenhoven H, Victoria M (2002) Acta
derov DV (2004) Appl Phys Lett 84:592 Mater 50:3957
52. Liao XZ, Zhao YH, Zhu YT, Valiev RZ, Gunderov DV (2004) 90. Hollang L, Thiele E, Holste C, Brunner D (2006) Mater Sci Eng
J Appl Phys 96:636 A 424:138
53. Meyers MA, Vohringer O, Lubarda VA (2001) Acta Mater 91. Zehetbauer M, Seumer V (1993) Acta Metall Mater 41:577
49:4025 92. Wei Q, Cheng S, Ramesh KT, Ma E (2004) Mater Sci Eng 381:71
54. Meyers MA, Andrade RU, Chokshi HA (1995) Metall Mater 93. Wang YM, Ma E (2004) Appl Phys Lett 85:2750
Trans A 26:2881
123
J Mater Sci (2007) 42:91129115
DOI 10.1007/s10853-007-2065-0
LETTER
Received: 21 May 2007 / Accepted: 26 July 2007 / Published online: 13 August 2007
Springer Science+Business Media, LLC 2007
Austempering is an important thermal processing opera- Poestenkill, NY). A diametrical dilatometer was mounted
tion, where strong and tough bainitic steel is produced in a on the specimen to measure the diameter change during the
single heat treatment [1, 2]. During the austempering pro- thermal processing. The austempering experiments were
cess, the steel is first austenitized and then cooled rapidly performed in two cycles, where the first cycle provides the
just above the martensite start temperature until bainite same initial microstructure prior to each experiment. After
nucleates and grows, usually until the transformation stops completing the first cycle, the cylindrical specimens were
and then it is cooled to room temperature. Due to the heated to the austenitizing temperature (850 C), held for
sluggish solid-state transformation kinetics, industrial 5 min and then cooled to different austempering tempera-
austempering necessitates isothermal holds of 224 h, tures, where the bainite transformation was monitored for
depending on the size and composition of steel. In contrast the desired period of time, followed by cooling to room
to conventional isothermal processing, cyclic thermal temperature. The cooling rate was sufficiently fast to avoid
processing has been shown to accelerate the kinetics of any transformation occurring before reaching the austem-
several phase transformations [35], with a significant pering temperature (Fig. 1). The isothermal experiments
beneficial impact on productivity and energy consumption were carried out at austempering temperatures of 260 and
of these energy intensive operations. This was attributed to 300 C, whereas the cyclic experiments were carried out
the non-isothermal effects resulting from cyclic treatment. between 260 C and 300 C at two different heating/
The non-isothermal effect on phase transformations has cooling rates of 1 and 5 C/min. The percentage of bainitic
also been utilized to enhance the productivity of a modern transformation as a function of time was computed from
batch annealing operation [6]. In the present work, the the dilation curve [2] in conjunction with the microstruc-
effect of cyclic processing on austempering kinetics has tural examination by optical microscopy and SEM.
been compared with conventional isothermal processing The normalized dilatation curves for isothermal aus-
for 1080 steel. Furthermore, the effect of cyclic frequency tempering at 260, 300 C, and cyclic austempering
on the austempering kinetics has been studied. between 260 C and 300 C for 1 C/min and 5 C/min are
Austempering kinetics experiments were performed on shown in Fig. 2. In this figure, only the portion of the curve
6 mm diameter cylindrical samples of a 1080 steel using corresponding to austempering is shown and the dilatation
a GleebleTM 3500 thermo-mechanical simulator (DSI curve has been offset to start at zero. The plateau of the
dilatation curve with respect to the time represents the end
of the transformation to bainite. It must be noted that the
V. Sista P. Nash (&)
end of the bainite transformation does not necessarily
Thermal Processing Technology Center, IIT, 10 W 32nd St.,
Chicago, IL 60616, USA correspond to 100% volume fraction of bainite [1]. Previ-
e-mail: philip.nash@iit.edu ously, using the same type of Gleeble experiments, it was
shown that the end of the transformation corresponds
S. S. Sahay
closely to the austenite carbon content corresponding to a
Division of Tata Consultancy Services Ltd., Tata Research
Development and Design Centre, 54B, Hadapsar Industrial T0 equal to the austempering temperature [2, 7]. Further-
Estate, Pune 411 013, India more the transformation dilatation at the end of the
123
J Mater Sci (2007) 42:91129115 9113
1000
o
850 C, 5 min
Temperature,oC
750
500
o
Austempering: 260-300 C
250
o
Ms ~231.2 C
0
Time
123
9114 J Mater Sci (2007) 42:91129115
100
50
0
260 C 300 C 1 C/min 5 C/min
Austempering Condition
123
J Mater Sci (2007) 42:91129115 9115
123
J Mater Sci (2007) 42:91169120
DOI 10.1007/s10853-007-2063-2
LETTER
Received: 27 January 2007 / Accepted: 26 July 2007 / Published online: 12 August 2007
Springer Science+Business Media, LLC 2007
Structural components under contact stresses are designed (substrate) with a thin layer of epoxy adhesive facilitated
so as to have high damage tolerance. Unexpected degra- in situ observation of cracks during contact loading. We
dation of initial strength under the components actual have investigated the role of substrates of different elastic
environmental conditions leads to economic loss, and even modulus with respect to crack initiation by in situ obser-
potential loss of life, through unanticipated permanent vations of cone cracking from the surface and radial
deformation and fractures. Therefore, material design to cracking from damage at the interface between two layers.
suppress contact stress is of practical importance in engi- In the present study, we investigate the role of porosity
neering applications, such as for the components utilized in in the substrate on the cracking of a hard-coating layer. To
machinery, and bio-, thermal or energy applications [14]. this end, we prepared porous substrates with different
Layered composite design has been investigated with porosity and the same epoxy and glass plate. We extended
consideration of damage tolerance in previous studies the system to a hard substrate with a high elastic modulus.
[5, 6]. Hard outer layers provide surface resistance against Two starting powders were used to prepare the porous
damage from contact loading and soft underlayers con- substrate, Al2O3 (AKP-50, AES-11, ALCOA, Japan), and
tribute to stress redistribution and delayed fracture of the SiO2 (SE-8, SO-R31, Tokuyama, Japan). In order to vary
coating layer. An adhesive is sometimes used to bond a the porosity in the porous substrate, we used different
hard layer with a soft material in making dental-layered powder sizes and also varied the sintering temperature. In
composites. The coating thickness and the elastic modulus Al2O3, the starting powders were 1.22 lm, 5.5 lm, and
mismatch are crucial parameters in geometrical and 9.8 lm in size. We varied the sintering temperatures from
material designing to suppress Hertzian contact stresses 1,000 C to 1,400 C at an increasing interval of 100 C.
[7]. The design of such layered composites is one of the Two types of SiO2 powders were used, (SE-8, and
important issues, because a softer support causes the SO-R31), 10 lm and 1 lm, and the ratio of staring powders
coating to flex under the constrained contacts. was varied, 10 lm: 1 lm = 6:4, 7:3, 8:2, and 9:1, respec-
Recently, we have investigated the cracking of brittle tively. The powders were sintered at 1,300 C or 1,350 C.
coatings adhesively bonded to substrates with different Therefore, we could vary the porosity and the material. The
modulus [8]. A trilayer system consisting of an upper soda- porosity was controlled in the range of 0.4242.6% and
lime glass layer (coating) bonded to a thick glass layer 8.225.4% for Al2O3 and SiO2 substrates, respectively,
through the control of sintering temperature and/or starting
powder size.
K. S. Lee (&) S. K. Kim C. Kim T. W. Kim
The layered composite systems were fabricated from
School of Mechanical and Automotive Engineering, Kookmin
University, Seoul 136-702, Korea cover glass with dimensions of 4 10 0.16 mm bonded
e-mail: keeslee@kookmin.ac.kr to a 4 4 36 mm substrate with a 10 lm thick layer
of epoxy adhesive. The surface was polished to 1-lm
D. K. Kim
diamond finish before bonding, and both the top and
Department of Materials Science and Engineering,
Korea Advanced Institute of Science and Technology, bottom surfaces of the glass layers were abraded with a
Yuseong, Daejeon 305-701, Korea slurry of SiC grit (#600), in order to induce uniform flaw
123
J Mater Sci (2007) 42:91169120 9117
distribution. The upper glass and the lower substrate were experimentally determined constant [16]. Solid curves are
bonded with epoxy adhesive (Harcos Chemicals, Belles- fits from Eq. 1 and the data fall on a theoretical curve.
ville, NJ, USA) under light pressure. In Fig. 2, it is noteworthy that the elastic modulus of
Spherical (Hertzian) indentation was used to induce Al2O3 is higher than that of the glass-coating layer, and that
contact stress on the layered composite systems, as shown of SiO2 is lower than the coating layer. The epoxy adhesive
in Fig. 1 [914]. Indentations were made with a universal has a low modulus, as compared to the Al2O3 and SiO2.
static testing machine (Microload test system, Unitec., Conversely, Al2O3 ceramics with higher porosity present a
Korea, or Instron 5567, Instron, USA) with a tungsten small elastic modulus mismatch after the layered com-
carbide (WC, J & L Industrial Supply Co., Livonia, MI, posite is constructed, and SiO2 ceramics with higher
USA) ball of radius r = 3.18 mm at P = 150 N, at a con- porosity have a large elastic modulus mismatch. A soft
stant crosshead speed of 0.09 mm/min. We have applied support by epoxy adhesive between the coating and the
the spherical WC indenter (r = 3.18 mm) up to 150 N substrate layer allows the coating to flex beneath the con-
contact load on the layered composites, where a glass tact. This leads to changes in the fracture patterns in the
coating with both surfaces abraded to ensure uniform flaw coating layer, from the mode of surface cracking to sub-
density was utilized. surface radial cracking, as found in a previous study [8].
The elastic modulus of each layered material is given in The previous study shows that the initiation of radial
Fig. 2. The graphs in Fig. 2 show the change in the elastic cracking is related with elastic mismatch, the following
modulus of the substrate monolith according to the porosity equation is applied:
in the substrate. The elastic modulus of each layer monolith
is measured by acoustic impulse excitation apparatus
(Tektronics, 5800PR, USA) and plotted as a function of
porosity. The elastic modulus of the coating material
(a) 400
monolith and epoxy adhesive are also measured using the
same method and included in the graphs. The elastic
modulus of porous ceramics such as Al2O3 and SiO2
Elastic modulus, E(GPa)
300
decreases as the porosity increases, according to the Rice
equation [1517].
200
E Eo expbP 1
where Eo is the elastic modulus of a material with no
porosity (0% of porosity), P is porosity, and b an 100 glass
adhesive
0
0 10 20 30 40 50
Porosity(%)
(b) 80
glass
Elastic modulus, E(GPa)
60
40
Ring Crack
Dense Glass
20
0
Porous Ceramic 0 10 20 30
Porosity(%)
Fig. 1 Schematic diagram of Hertzian indentation test on densely
coating/porous-substrate layered composite bonded with thin Fig. 2 Plot of elastic modulus of each layer material as a function of
adhesive porosity for (a) Al2O3 and (b) SiO2
123
9118 J Mater Sci (2007) 42:91169120
123
J Mater Sci (2007) 42:91169120 9119
cracking than for SiO2 substrate). The results in Fig. 3a Although cracking appears in the coating layer of the glass/
indicate that the contact cracking crucially depends on the Al2O3 layered composite with small porosity, it is notable
porosity, as well as the elastic modulus mismatch. that the damage is not critical in the substrate layer of
Therefore, the rigidity of substrate for the layered Fig. 4a and b relative to Fig. 4c. On the other hand, the
composite bonded with the substrate controls the initiation result suggests that if yielding of the substrate occurs
and propagation of contact cracks in the coating layer. during contact with the higher porosity substrate as shown
Strong and stiffer substrates protect coating layers bonded in Fig. 4c, extensive cracking in the coating layer is pro-
with very-soft adhesive layers (d = 10 lm) from cracking duced. An extensive yield in the substrate of the layered
caused by flexure stress (a high modulus of substrate composite is observed with more extensive cracking in the
delays initiation of subsurface radial cracking according to coating layer, as indicated in Fig. 4c. The extensive
Eq. 2). This is because the controlling material quantity in cracking in the coating layer can be appeared if the
Eq. 2 is strength, rF, relating with flexure. Plots in Fig. 3a extensive yielding occurs in the substrate even though
illustrate the beneficial role of well-controlled pores in the the elastic modulus mismatch is diminished. Therefore, the
ceramic substrate with respect to contact fracture of the increase of size of cracks as the amount of the porosity
coating layer. For small porosity in the substrate (stage I), in the Al2O3 substrate increases to a maximum of 42.6% in
the pores may suppress cracking by damage re-distribu- stage II of Fig. 3a is closely due to yield of subsurface in
tion. On the other hand, radial cracking by flexural stress the substrate. The determining factor is whether yield
due to elastic modulus mismatch, especially apparent in occurs in the porous substrate during the contact stress. The
the layered composite with SiO2 substrate as shown in result indicates that the substrate should not yield before
Fig. 3b, causes severe damage to the coating layer. Contact the coating layer fractures.
cracks including radial cracking are apparent in the porous In designing the coating/substrate-layered composite,
SiO2 support at any range of porosity, because the hard/ bonding with adhesive requires a higher substrate modulus
soft-layered composite has a high elastic modulus mis- and hardness with a matching of the substrate, therefore
match, which strongly influences cracking in the coating higher damage tolerance. The smaller volumes of pores
layer. under the critical porosity range can contribute to delayed
SEM micrographs of section views showing subsurface cracking if they are not yielded or deformed before coating
damages in the brittle substrate monolith as well as crack fracture. More detailed and systematic investigations in this
profiles produced in the coating layers are shown in Fig. 4. regard should be conducted in future study.
Fig. 4 Section views of SEM micrographs of the glass/Al2O3 layered composite from Hertzian indentation at load P = 150 N. The porosity of
the substrate layer is (a) 13.3%, (b) 34.3%, and (c) 41.2%
123
9120 J Mater Sci (2007) 42:91169120
Acknowledgements This work was partly supported by Korea 7. Lee KS, Wuttiphan S, Hu XZ, Lee SK, Lawn BR (1998) J Am
Research Foundation Grant funded by the Korean Government Ceram Soc 81:571
(MOEHRD) (KRF-2004-003-D00002) and partly by a grant from 8. Lee KS, Lee YW, Blackburn DH, Chai H, Lawn BR (2000)
Electric Power Industry R&D Project funded by the Ministry of J Mater Res 15:1653
Commerce, Industry and Energy. 9. Hertz H (1896) Hertzs miscellaneous papers, Ch. 5, 6. Mac-
millan, London
10. Lawn BR (1993) Fracture of brittle solids, 2 edn. Cambridge
Univ. Press, Cambridge
References
11. Pajares A, Wei L, Lawn BR, Padture NP, Berndt CC (1996)
Mater Sci Eng A208:158
1. Lawn BR (1998) J Am Ceram Soc 81:1977 12. Lee KS, Wuttiphan S, Hu XZ, Lee SK, Lawn BR (1998) J Am
2. Lee KS, Seo DW, Yu JH, Woo SK (2003) J Kor Ceram Soc Ceram Soc 81:571
40:241 13. Lee KS, Lee SK, Lawn BR, Kim DK (1998) J Am Ceram Soc
3. Jung HH, Kim DJ, Han JS, Lee DY (2004) J Kor Ceram Soc 81:2394
41:508 14. Rhee YW, Kim HW, Deng Y, Lawn BR (2001) J Am Ceram Soc
4. Beele W, Marinjiniseen G, van Lieshout A (1999) Surf Coat Tech 84:1066
120121:61 15. Rice RW (1993) J Mater Sci 28:2187
5. Lee KS, Lee SK, Lawn BR, Kim DK (1998) J Am Ceram Soc 16. Rice RW (1989) Mater Sci Eng A112:215
81:2394 17. Ha JH, Kim JH, Kim DK (2004) J Jap Ceram Soc 112:1084
6. Lawn BR, Lee KS, Chal H, Pajares A, Kim DK, Wuttiphan S,
Peterson IM, Hu X (2000) Adv Eng Mater 11:745
123
J Mater Sci (2007) 42:91219124
DOI 10.1007/s10853-007-2056-1
LETTER
Received: 18 March 2007 / Accepted: 25 July 2007 / Published online: 11 August 2007
Springer Science+Business Media, LLC 2007
Ordered alloys and intermetallic compounds are playing an mm 2 mm specimens after 24 h immersion in 20%
increasingly important role in advanced corrosion applica- boiling hydrochloric acid. Transmission electron micros-
tions, e.g. [1]. As design schemes increase in sophistication, copy and diffraction as well as x-ray diffraction were used
the search continues for alloys, which offer higher strength to characterize the microstructure.
and can handle a larger variety of complex aggressive Figure 1 shows the effect of up to 1,000 h of exposure at
media in comparison with the conventional alloys. How- 700 C on the room temperature tensile properties and
ever, the issue of low ductility remains to be an important corrosion rate. It is observed that after about 1 h of expo-
one and, therefore, efforts are focused on improving duc- sure, the 0.2% yield strength was nearly doubled relative to
tility and/or innovations in manufacturing technology. the annealed condition; however, about 70% of the initial
Among the most important alloys in this class are the those tensile ductility in 50.8 mm gage length was still main-
based upon the NiMo and NiMoCr [2, 3, 4]. A common tained. Corresponding to these changes there was no
feature of these alloys is that they undergo a series of long- significant change in corrosion rate. With continued ther-
range ordering reactions upon exposure at elevated tem- mal exposure, the yield strength remained nearly
peratures [5, 6]. This involves a slight atoms rearrangement unchanged after up to 1,000 h. In contrast, there was a
on {420}fcc planes resulting in closely-related superalat- significant loss of tensile ductility as well as a substantial
tices: D1a (Ni4Mo), Pt2Mo-type: {(Ni2Cr, and Ni2(Cr,Mo)}, increase in corrosion rate. Such changes in properties were
and DO22 (basis for the Ni3Mo structure) [7]. correlated with long-range ordering to Ni4Mo as demon-
It was the objective of this study to determine the effect strated below.
of long-range ordering in a NiMo alloy of commercial Figure 2 shows x-ray diffraction patterns illustrating the
grade on its mechanical properties and corrosion resistance. effect of exposure time at 700 C on the average structure.
The alloy was processed into sheets about 2 mm in thick- In the annealed condition, only characteristic face-centered
ness. Its chemical composition is listed in Table 1. cubic (fcc) reflections are observed indicative of a disor-
Specimens were annealed at 1,065 C for 20 min and then dered solidsolution. After thermal exposure at 700 C,
water quenched. Inorder to induce long-range ordering, characteristic superlattice reflections of Ni4Mo were
annealed specimens were exposed up to 1,000 h at 700 C observed as expected (tetragonal D1a superlattice:
and then air-cooled. Mechanical properties were deter- a = 0.5720 nm, c = 0.3564 nm). However, as can be seen
mined from room temperature tensile tests (50.8 mm gage the relative intensities of the superlattice reflections
length). Aqueous corrosion rates were calculated from remained nearly unchanged as the exposure time was
weight loss measurements made on 25.4 mm 25.4 increased from 1 h to 1,000 h. This suggested that ordering
was completed after about 1 h of exposure at 700 C.
An example illustrating the evolution of the micro-
H. M. Tawancy (&) F. K. Alyousef structure of Ni4Mo within the alloy matrix is given in
Center for Engineering Research, Research Institute,
Fig. 3. Arrays of discrete particles were observed during
King Fahd University of Petroleum and Minerals,
Dhahran 31261, Saudi Arabia the early stages of exposure (Fig. 3a). This was followed
e-mail: tawancy@kfupm.edu.sa by the formation and growth of a mosaic assembly of
123
9122 J Mater Sci (2007) 42:91219124
Table 1 Chemical composition (weight %) demonstrated earlier (Fig. 1). Corresponding deformation
Ni Mo Cr Fe Mn Si Co Al C
substructures are shown in Fig. 4. As expected, in the
annealed condition, deformation occurred by slip as shown
Balance 26.92 0.64 0.93 0.24 <0.02 <0.1 0.19 0.002 in Fig. 4a. Slip lines could be clearly distinguished in the
deformation substructure with evidence of cross-slip sug-
twin-related variants along {100}fcc planes (Fig. 3b, c). gesting medium stacking fault energy. However, after
Corresponding to these microstructural changes, the 0.2% long-range ordering to Ni4Mo, the predominant deforma-
yield strength was nearly doubled relative to the annealed tion mode became twinning on the {111}fcc planes as
condition, however, about 70% of the initial tensile duc- shown in the example of Fig. 4b. This observation indi-
tility in 50.8 mm gage length was still maintained as cated that was a favorable deformation mode in the ordered
80
Tensile Elonagation
in 50. 8 mm ( % )
60
40
20
955
0. 2% Y iel d Str e ngth (MP a )
820
685
550
415
280
0 1/4 1/2 1 2 4 8 24 100 1000
Exposure Time (Hours)
123
J Mater Sci (2007) 42:91219124 9123
state consistent with the crystallographic features of the fcc the fracture mode was transgranular (dimple-type rupture).
(disordered) ? D1a superlattice (Ni4Mo). It can readily be Also, after 24 h of exposure the fracture mode remained to
shown that all {111} <110> slip systems are suppressed in be transgranular, however, the dimples became rather
the ordered state, however, eight of the {111} <112> shallower indicating less strain prior to fracture. Although
twinning systems remain to be energetically favorable [8]. after 1,000 h of exposure, the fracture mode became
Corresponding to these changes in tensile properties, there interngranular, dimples were observed at separated grain
was only a slight increase in corrosion rate, which could be facets (Fig. 5c) indicating highly localized deformation
interpreted in terms of Mo-depleted zones in the vicinity of alongside grain boundaries. This behavior is usually asso-
Ni4Mo domains, as reported in an earlier study [2]. ciated with alloy depletion in the vicinity of grain
After extended exposure corresponding to almost com- boundaries as demonstrated in Fig. 6. The respective val-
plete loss of tensile ductility (1,000 h at 700 C), only ues of Mo concentrations were obtained from
{111}fcc stacking faults were observed in the deformation concentration profiles derived by in-situ fracture and
substructure as shown in Fig. 4c. Noting that a stacking sputtering of aged specimens in an Auger spectrometer.
fault is a thin layer of a twin, this observation indicated that The value of 4.7 atomic % corresponds to a freshly exposed
fracture occurred prior to the development of deformation grain boundary, which is to be compared with the bulk
twins. Figure 5 illustrates characteristic morphologies of value of 19.1 atomic %. It is observed that ordering in the
tensile fracture surfaces as functions of exposure time at vicinity of grain boundaries occurred by a discontinuous
700 C. As expected, in the annealed condition (Fig. 5a), mechanism resulting in alternating lamellae of Ni4Mo and
123
9124 J Mater Sci (2007) 42:91219124
123
J Mater Sci (2007) 42:91259129
DOI 10.1007/s10853-007-2010-2
LETTER
Received: 28 June 2007 / Accepted: 20 July 2007 / Published online: 4 August 2007
Springer Science+Business Media, LLC 2007
The mechanical properties of particle reinforced compos- microstructure-based models simulating the tensile behav-
ites are largely dependent on the reinforcement particle ior of heterogeneous materials correlated well with
distribution and volume fraction [1, 2]. Crack growth in experimental results. Most of the 3D microstructure-based
particle reinforced metal matrix composites (MMCs), such modeling efforts have focused on predicting the elastic
as SiC particle reinforced aluminum (Al/SiCp), is very modulus and simulating the elastic-plastic behavior of the
much dependent on the reinforcement microstructure [1, 3] composite in tension. The onset and evolution of damage
as well as the matrix characteristics [4, 5]. A robust by particle fracture has been studied by a few researchers
numerical model to simulate crack growth must incorpo- [1014], although the reinforcement particles are modeled
rate the true particle geometry, orientation, size as simple spheres. Crack growth in 3D has not been
distribution, and spatial distribution [6]. Ayyar and Chawla modeled in particle reinforced metal matrix composites. In
[6] modeled the crack growth behavior in an Al/SiCp this paper, crack growth in a SiC particle reinforced Al
composite, using 2D microstructures obtained from optical composite was simulated, using the actual 3D microstruc-
microscopy. The results of these models showed that the ture of the composite.
incorporation of the actual microstructural attributes A serial sectioning process was used to capture the
significantly affected the simulated crack growth response. complex geometry of the SiC particles from a series of
Of course, 2D models must be conducted under sim- micrographs of a SiC particle reinforced 2080 Al matrix
plified stress states, such as plane stress or plane strain composite [7, 8]. The SiC particles had a volume fraction
conditions. Thus, 2D simulations do not provide as com- of approximately 20%, average particle size of about 8 lm,
plete a picture of the crack growth processes as 3D and an average aspect ratio of 2. The serial sectioning
simulations. In reality, the stress state in 3D is quite process allows one to capture the realistic microstructure of
complex. This is particularly true of composite materials, the particles, including the geometry, orientation, and dis-
where the geometry of the SiC particle in the third tribution of the particles. A typical flow chart of the serial
dimension and the matrix around the particle play an sectioning and 3D reconstruction process is shown in
important role. Because the geometry of the SiC particle is Fig. 1. Details of this process can be found in reference [7].
very complex, a 2D representation of the microstructure The sample was cut and mounted for polishing and a
cannot adequately capture the behavior of the composite. representative region of the microstructure was selected.
Chawla et al. [79] have shown that predictions from 3D The term representative is somewhat subjective as it is
dependent on the size of the SiC particles, spatial distri-
bution, etc. As the number of particles increases, however,
A. Ayyar
Department of Mechanical and Aerospace Engineering, Arizona the computational demands also increase. For the Al/SiCp
State University, Tempe, AZ 85287-8706, USA composites modeled in this paper, a volume of about 32
particles was included in the model (which is about half of
N. Chawla (&)
the total particles reconstructed after serial sectioning),
School of Materials and Department of Mechanical Engineering,
Arizona State University, Tempe, AZ 85287-8706, USA because of reasonable limits on computational efficiency.
e-mail: nchawla@asu.edu Fiducial marks were made by Vickers indentation. These
123
9126 J Mater Sci (2007) 42:91259129
marks were used to measure the material thickness loss the calculations when a certain parameter in the element
during the polishing process. The thickness removed is a exceeds a predefined value. The mesh was refined, since
function of the microstructural feature size. In this case, the crack profile is highly dependent on the mesh size and a
since the particles were approximately 68 lm in size, a fine mesh is required to obtain accurate results. Although
removal rate of about 1 lm/section was selected. Using the Al is highly ductile, for the sake of simplicity, and to
cyclic process of polishing and imaging, it was possible to illustrate the feasibility of crack growth modeling using
generate a series of 2D micrographs. These 2D micro- actual 3D microstructures, the Rankine criterion was used
graphs were segmented to differentiate between the for element elimination (failure occurs when the maximum
particles and the matrix. The 3D morphology of the par- principal stress in the element is exceeded). The element
ticles was reconstructed from the series of 2D segmented elimination process, using the Rankine criterion, was
images using commercially available software (SURF- simulated in the Dynamic-Explicit module in ABAQUS.
driver, Kailua, Hawaii). This module requires that a mass be assigned to all the
The 3D reconstructed microstructure was then meshed materials. Therefore, it is extremely important that the load
using HyperMesh1 (Altair, Los Angeles, CA), Fig. 2. A be applied very slowly to the system to avoid any inertia
2D surface mesh was first created on the SiC particle and effects. A velocity of 0.01 unit/s was applied to the nodes
the Al matrix surfaces. This 2D mesh consisted of 3-node on the upper surface, as shown in Fig. 2d, until the crack
triangular elements. The 2D mesh was then used to create a had propagated through the matrix material. The value of
3D mesh consisting of 4-node tetrahedral elements (C3D4). the velocity was obtained from an iterative procedure to
The tetrahedral elements conformed well to the irregular quantify any inertia effects. The elements in the Al matrix
shape of the SiC particles. A variable mesh density were eliminated when the stress reached the yield stress for
approach was used to reduce the overall size of the model the Al alloy matrix (approximately 320 MPa). Fracture of
without sacrificing the accuracy of the model. The mesh in the SiC particles was not incorporated in this model,
the SiC particles and the Al matrix near the particles were although the authors have used it in modeling 2D micro-
finer compared to the mesh on the surfaces of the Al structures [15]. All three components, SiC particles, Al
matrix, Fig. 2c. An embedded cell, with average com- matrix, and the average composite were modeled as
posite properties, was used to minimize the effects of free purely elastic. The elastic properties of SiC and Al alloy
edges. There were approximately 45,000 nodes and are shown in Table 1. The properties for the Average
240,400 tetrahedral elements (C3D4) in the model. The composite were computed using a rule-of-mixtures
discretized model was then imported into a commercial approach. The density was 2.81 g/cm3, the Poissons ratio
finite element analysis software (ABAQUS, Pawtucket, was 0.31 and the Youngs modulus was approximately
RI). Boundary conditions were applied to the model as 120 GPa. Particles were assumed to be bonded perfectly to
shown in Fig. 2d. The element elimination method was the matrix. The simulation time for this model, on a DELL
used to simulate crack growth in the particle reinforced Precision workstation with single processor (3.0 GHz
composite. In this method an element is eliminated from speed) and 2 GB of RAM, was approximately 6 hours.
Material Indentation of
Polishing
preparation fiducial marks
Imaging
123
J Mater Sci (2007) 42:91259129 9127
Plane 1 (ABCD)
Plane 2 (EFGH)
boundary conditions applied to
the model
127 m
55 m
X
D H
20 m E
A
Al Matrix Y Z
Average composite
G
(b) (c) Vx=0.01 unit/s
SiC
C
F
D E
Y Z
Uy,z=0
Al Matrix
A Ux=0
Table 1 Material properties (Elastic) used in the model increases, as has been observed experimentally [1]. When
Material Density Youngs Poissons particle fracture is modeled, the crack goes through the SiC
(g/cc) modulus (GPa) ratio particles, and the crack path is quite linear [15]. Crack
growth in a homogeneous model with average composite
Aluminum alloy 2080-T6 2.75 74 0.33
properties, i.e., where the microstructure of the SiC parti-
SiC particles 3.2 410 0.19
cles was not explicitly considered was also modeled. The
distance between plane 1 and 2, in the homogeneous
Figure 3 shows the simulated 3D crack profile for the model, is the same as that in the microstructure-based
model of Al/SiCp composite. The model clearly shows the model. As expected, here the crack profile was planar and
complex crack growth phenomenon in particle reinforced not very different between the outer and inner surfaces.
composites. The crack profile on the outer surface was very When one incorporates the explicit microstructure, it is
different from the crack profile on the inner surface quite clear that the crack profile through the SiC particles is
because of the nature of the local microstructure. The crack captured. The tortuous nature of crack growth, as well as
initiates in the Al matrix along plane 1 (ABCD) and ter- the deflection of the crack due to particles can be seen.
minates in the matrix at plane 2 (EFGH). On the inner Crack deflection, of course, reduces the crack-tip driving
surface it was observed that the crack did not reach plane 2 force of the material. The applied load is plotted versus the
(EFGH) as it was arrested by a SiC particle. Note that the nodal displacements for both models (microstructure-based
crack branches around one of the SiC particles, Fig. 3b, model and homogeneous model) in Fig. 4. The crack ini-
and is forced to grow around it, Fig. 3c, as the load tiates at plane 1 (ABCD) and ends at plane 2 (EFGH).
123
9128 J Mater Sci (2007) 42:91259129
Fig. 3 3D microstructure-based
(a) (a)
simulation of crack growth
showing crack arrest, crack
deflection, and tortuous 3D
crack surface
Plane 1 Plane 1
d = 0.053 d = 0.053
(b) (b)
Plane 1 Plane 1
d = 0.066 d = 0.066
(c) (c)
Plane 1 Plane 1
d = 0.069 d = 0.069
Fig. 4 Load versus Crack initiates at Plane 1 and terminates at Plane 2 Displacement of this node
displacement plots. In the 100
microstructure-based model Plane 1
90
more energy is required to
80
propagate the crack through the
Applied load (GN)
70
same distance (from Plane 1 to Plane 1 Plane 2
60
Plane 2)
50 Plane 2
40
30
Plane 2
Plane 1
20
10
0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 Homogeneous model
Displacement of node (units)
Microstructure-based
model
Clearly, more energy was required for the crack to prop- approach. Three-dimensional crack growth is a complex
agate in the microstructure-based model. This is a result of phenomenon, which involved crack deflection and crack
crack deflection and crack arrest by the particles in the arrest. A complex and tortuous 3D crack profile was sim-
microstructure-based model. ulated (assuming no particle fracture), which cannot be
In summary, a 3D microstructure-based model was adequately represented using conventional 2D representa-
developed to study crack growth in Al/SiCp composites. tions. This crack deflection due to reinforcement particles
The complex geometry of the SiC particles, in all three makes the crack path tortuous and improves the crack
dimensions, was captured using a serial sectioning growth resistance of the composite.
123
J Mater Sci (2007) 42:91259129 9129
Acknowledgements The authors acknowledge useful discussions 5. Sugimura Y, Suresh S (1992) Metall Trans A 23:2231
with Prof. K. K. Chawla, University of Alabama at Birmingham, and 6. Ayyar A, Chawla N (2006) Compos Sci Technol 66:1980
Prof. Y.-L. She, University of New Mexico. 7. Chawla N, Ganesh VV, Wunsch B (2004) Scr Mater 51:161
8. Chawla N, Sidhu RS, Ganesh VV (2006) Acta Mater 54:1541
9. Chawla N, Sidhu RS (2007) J Mater Sci-Mater Electron 18:175
10. Han W, Eckschlager A, Bohm HJ (2001) Compos Sci Technol
References
61:1581
11. Bohm HJ, Eckschlager A, Han W (2002) Comput Mater Sci
1. Chawla N, Chawla KK (2006) Metal matrix composites. 25:42
Springer, New York 12. Eckschlager A, Han W, Bohm HJ (2002) Comput Mater sci
2. Divecha AP, Fishman SG, Karmakar SD (1981) J Metals 9:12 25:8591
3. Shang JK, Yu WK, Ritchie RO (1988) Mater Sci Eng A102:181 13. Segurado J, Gonzalez C, Llorca J (2003) Acta Mater 51:2355
4. Bonnen JJ, You CP, Allison JE, Jones JW (1990) In: Proceedings 14. Segurado J, Llorca J (2006) Mech Mater 38:873
of the international conference on fatigue. Pergamon Press, New 15. Ayyar A, Chawla N (2007) Acta Mater (in press)
York, p 887
123