RECRYSTALLI ZATI ON

OF
METALLIC MATERIALS
Papers presentad at a
Seminar of the
Insti tu fr Metallkunde of the University
of Stuttgart
and the
Max-Planck-Institut fr Metallforschung,
Stuttgart
During Summer 1970
Edited by ...
FRANK HAESSNER Jl
CONTENTS
PREFACE 1
LIST OF CONTRIBUTORS 3
,P. HAESSNER
H. P. STWE
i
t R. W. CAHN
v S. HOFMANN and
F. HAESSNER
K. DETERT
J. GREWEN
SYSTEMATIC SURVEY AND FUNDAMENTAL
PROBLEMSOFRECRYSTALLIZATION 5
DRIVING FORCES FOR RECRYSTALLIZATION 21
NUCLEATION IN RECRYSTALLIZATION 43
MIGRATION OF HIGH ANCLE GRAIN BOUN-
DARIES 81
SECONDARY RECRYSTALLIZATION 109
ANNEALING TEXTURES 135
O.DIMITROV, EFFECTS OF TRACE IMPURITIES ON RE-
R. FROMAGEAU and CRYSTALLIZATION PHENOMENA 183
C. DIMITROV
U. KOSTER
H. MECKING and
F.KIRCH
I. L. DILLAMORE
1971 Alie Redite v orbehalten Dr. Riederer Verlg GmbH 7000 Stuttgart
Druck: Bbnecke-Clausfhal-Zcllerfeld
RECRYSTALLIZATION OF TWO-PHASE
ALLOYS 215
RECOVERY AND RECRYSTALLIZATION
DURING DEFORMATION 257
THE IMPORTANCE AND APPLICATION OF RE-
CRYSTALLIZATION IN THE PROCESSING OF
METALS AND THE MANUFACTURE OF SPECIAL
PRODUCTS 289
ISBN-Nr. 3-87675-007-5
PREFACE
The phenomenon of recrystallization has been utilized extensively in
metallurgical processing for some thousand years. Inspite of this con-
siderable practical importance, the systematic study of this phenomenon
only set in extensively about 50 years ago. Today the scientific and practi-
cal knowledge has reached a level which, in a number of cases, permits
the planned development of materials by suitable recrystallization treat -
ments. The conceptions and formulations originally developed for the
recrystallization of metis have in the meantime been applied successfully
to other crystalline materials as well, such as, for example, ceramics.
It therefore appeared desirable to treat the topic recrystallization in a
special seminar. Such a seminar was organized jointly by the Institu fr
Metallkunde of the University of Stuttgart and by the Max-Planck-Institut
fr Metallforschung, Stuttgart for advanced students, scientists and parti-
ctpants from industry. The lectures presented in the course of this
seminar appear in revised form in this book. They represent a survey of
our current knowledge and of the opinions and open questions in the most
significant subfields of the topic of recrystallization of metis . The in-
dividual chapters consist of contributions which, inspite of their self-con-
tained form, are matched to complement each other. Internal cross-re-
ferences , the use of uniform symbols and a detailed introduction by the
editor will help to tie the various contributions together. The treatment
of experimental technques, recrystallization of strongly disturbed thin
films and recrystallization of non-metallic crystalline materials has been
omitted for lack of space. It is not intended that this book should form a
complete text-book on recrystallization. Rather, it is hoped that the most
important aubjects of interest have been discussed in a way which should
enable the reader to gain a simple access to the current problems.
-
The printing of the book has been made possible through the courtesy of the
Dr. Riederer-Verlag whose staff I wish to thank for their help. Thanks are rlr .nfnrschune
Chaoter 1 F. HaeBner, Max-Planck-Institut fr Metamorst ing,
also extended to all those who have contributed to the success of the semina
Stutteart, Germany.
Stuttgart, March 1971
FRANK HAESSNER
Chapter 2.
1
Chapter 3.
Chapter 4.
Chapter 5.
Chapter 6.
Chapter 7.
Chapter 8.
Chapter 9.
Stuttgart, Germany.
H. P. Stwe, Institu fr Werkstoffkunde und Herstellungs-
verfahren der Technischen Universitat Braunschweig,
Braunschweig, Germany.
R. W. Cahn, Materials Science Divisin, School of Applied
Sciences, University of Sussex, Brighton, England.
S. Hofmann and F. Haefiner, Max-Planck-Institut fr Metall-
forschung, Stuttgart, Germany.
K. Detert, Institu fr metallische Werkstoffe im Bereich
Forschung und Entwicklung, AEG/Telefunken, Frankfurt,
Germany.
J. Grewen, Institu fr MetalUcunde und Metallphysik, Tech-
nische UniversitatClausthal, Clausthal-Zellerfeld, Germany.
O. Dimitrov, R. Fromageau and C. Dimitrov, Centre
d'EtudesdechimieMtallurgique, Vitry-sur-Seine, France.
U. Koster, Institut fr Werkstoffe, Ruhr-UniversitatBochum,
Bochum, Germany.
H. Mecking and F. Kirch, Institut fiir Allgemeine Metall-
kunde und Metallphysik, Rhein. -Westfl. Technische Hoch-
schule Aachen, Germany.
Chapter 1O. I. L. Dillamore, BISRA-The Corporate Laboratories British
Steel Corporation, Sheffield, England.
CHAPTER1
SYSTEMATIC SURVEY AND FUNDAMENTAL PROBLEMS
OF RECRYSTALLIZATION
BY
F. HAESSNER
The terminology used in describing the phenomena of recrystallization
has with the passing of time repeatedly been revised as a result of increas-
ing understanding. Henee, it is not surprising that the term "recrystalliz-
ation" frequently denotes a phenomenon that may differ in detail from case
to case. Consequently, in the following presentation that is intended as an
introduction to the individual chapters and as a type of guide for the book,
an attempt will first of all be made to classify the entirety of the re-
crystallization phenomena of metis in a scheme which facilitates a survey.
Subsequently, attention is turned to those categories of questions that al-
ways reappear in one way or another in connection with recrystallization,
namely, the "fundamental problems".
1. SYSTEMATIC SURVEY
1.1. Decrease in (Gibbs) Free Energy and Reaction of Lattice Defects
The essential processes occurring during recrystallization may be
discussed reasonably well by considering a plastically deformed
crystalline material. For the sake of simplicity, it is first assumed
that the material does not undergo a phase transition in the temperature
range considered. As a result of its increased content of physical
defects, such a material is in a thermodynamically unstable state.
Upon increasing the temperature the material can lower its free energy +'
by the reduction and rearrangement of the lattice defects. In particular
cases, this may occur in quite different ways. For a systematic
c las sif i catin, it has proved appropriate and useful to describe the
In a solid material the difference between the Gibbs free energy and
the Helmholtz free energy is normally neglegibly small. Therefore
the word free energy is used throughout this volume instead of Gibbs
free energy.
rnultitude o phenomena that occur below the melting point in terms of
five fundamental structural processes. These fundamental processes,
whereby a pur metal can approach a state of equilibrium, are
essentially the following:
1) Reactions of point defects and point defect agglomerates; in
particular the annihilation of these defects.
2) The annihilation of dislocations of opposed sign and the shrinking
of dislocation loops.
3) The rearrangement of dislocations to form energetically more
favourable configurations.
4) The absorption of point defects and dislocations by grain
boundaries migrating through the metal.
5) The reduction in the total grain boundary rea.
Generally a deformed polycrystalline metal does not attain the thermo-
dynamically more stable state by a single process. Instead, all of the
above-mentioned reactions proceed in turn and may overlap locally
and chronologically. If the state of increased free energy is brought
about in some other way than by plstic deformation, e. g. by quench-
ing of a single crystal from high temperatures or by particle bombard-
ment, then processes 1) and 2) only will essentially be observed upon
annealing. In any particular case, the behaviour will be governed by
the type, concentration and arrangement of the lattice defects intro-
duced into the metal by the treatment. These "structural details"
determine the decrease in the free energy in the course of each process,
i. e. they determine the driving forc (change in free energy per unit
volume) and the rate of reaction associated with the process under
consideration. In pur metis, the changes in the free energy are re-
latively small, i. e. not more than some few per cent of the latent
heat of fusin (cf. chapter 2). Regarding the rates of reaction, all
that can generally be said is that they obey exponential time-laws with
activation enthalpies which usually lie between those of lattice and
grain-boundary self-diffusion.
In a collective description, processes 1) and 2) are termed "recovery"
and processes 4) and 5) "recrystallization" (migration of large-angle
boundaries). If the intermedate process 3) contributes to the forma-
tion of large-angle boundaries, then to that extent it is also considered
as a part of recrystallization. In other words, recrystallization is
looked upon as the formation and migration of large angle boundaries ,
This definition is quite general and incorporales in particular many-
phase systems as well. On the other hand, the formation of a
crystalline stcucture from an amorphous phase would not be consider-
ed as recrystallization, since the definition of the term large-angle
boundary between these two phases would be meaningless.
It was assumed so far that the unstable state is formed prior to anneal-
ing. If, however, the unstable state is formed during annealing, e. g.
by plstic deformation at sufficiently high temperatures, then one is
dealing with the case of "dynamic recovery" or "dynamic recrystalliz-
ation" (see chapter 9).
The understanding of the processes occurring under such complicated
conditions is as yet so incomplete that a systematic classification in
the framework of the fundamental processes is not possible.
1. 2. Primary Recrystallization
Normally, in the course of the annealing of a sufficiently strongly
cold-worked metal, this being the most important case in practice, dis-
cretely located nuclei are formed which grow, thereby consuming the
'Occasionally, the formation of large-angle boundaries, i. e. the
"nucleation", is omitted intentionally in the definition of "recrystal-
lization" (cf. for example chapter 2 section 1.), since many re-
crystallization processes occur without nucleation.
deformed material. This process is called "primary recrystalliz-
ation". The nucleus comprises a lattice regin (of the order of ca.
1 A im) that is capable of growth. This regin contains a much lower
density of lattice defects, in particular dislocations, than the deform-
ed matrix. Such a nucleus is separated f r om its environment by a
newly formed large-angle boundary. In other words, nucleation and
pr imar y growth of grains during primary recrystallization correspond
to the fundamental processes 3) and 4) . The drivng forc for pr imar y
growth of grains results essentially f r om the difference between the
dislocation densities in the nucleus and the matrix (see chapter 2
section 2. 1. ).
Deviations from the normal case of primary recrystallization may be
observed after either very strong or very weak cold-working: After
strong deformation, the dislocations occasionally rearrange to such
an extent during annealing that small regions of low dislocation
density which are divided by large-angle boundaries are formed
everywhere . A new structure is thus formed in the absence of
large-angle boundary migration. Thereby, the texture of the material
is essentially retained. This process is called "recrystallization in
situ" . In a structural, atomistic sense, this behaviour corresponds
to process 3) , A fter weak def or mation, large-angle boundaries that
are present f r om the beginning of annealing may be displaced, leaving
a less defect regin behind. This process is called ' strain-induced
grain-boundary migration" and corresponds to process 4) (see chapter
3 section 3 d) ).
The r ear r angement of dislocaions leading to the formation of
small-anele boundaries (subgrain f or mation) is also called
polygor.i. 'ion.
The phenomenon of the f or mation of a new str uctur e in the ab-
sence of large-angle boundary migration can be studied parti-
cular ly well in some cold-worked mater ials that are capable of
pr ecipitatior . In analogy to the terminology of pr ecipitation
phenomena, this special type of recrystallization has been called
"continuous r ecr ystalli/ ation1' , as opposed to "discontirixtus rr-
cr ystallization", dur ing which "reaction fronts", i. ( . lar ge-
angle boundaries, pass br ou^h the def or med str uctur e f e . f.
chapter E section 2. 1. , 2.2. ).
The integral kinetics of primary recrystallization (in an isothermal
experiment) can be described by the recrystallized volume fraction
X (t ) (t : time) . In the normal case which is characterized by the
nucleation rate v = =- -~a-i- and the growth rate v = ~ M ~
n O K I - A (cj 2 d t
(n: number of observable grains, D: diameter of grain) X (t) may be
calculated under simplifyingassumptions, In the simplest case,
assuming v and v to be independent of time, growth to be isotropic
and nucleation to occur locally in a statistic manner throughout the
volume of the specimen, one obtains
X( t ) = 1 - exp (-v t 4 ) .
Under the same assumptions, the average grain size >
pleted primary recrystallization, i. e. X (t ) = 1 , is gi
., 1/4
The recrystallized fraction and all quantities derived therefrom are
suitable in a very limited sense only for a detailed analysis of primary
recrystallization . This is so because of the large number of re-
strictive assumptions that have to be made in their calculation. Thus,
for example, the determination of v and v from the integral
quantities X and D ylelds at best an estmate of the order of magni-
* o
tude.
1. 3. Grain Growth (Grain Coarsenlng)
If annealing is continuad after the completion of primary recrystal-
lization i. e. X( t ) = 1, then the average grain diameter D can
increase further. These processes are summarized by the term
" grain growth". In a structural, atomistic sense, they correspond
' This is also truc when general expressions are used for the re-
crystallized fraction, suchasfor example, the well-known Avrami-
relation X (t) =l -e xp( A t ) (A and K are constars).
1 O
to process 5). Phenomenologically, a distinction is made between
"normal or continuous grain growth" and "discontinuous grain growth" .
During normal grain growth, a relatively uniform increase in the
average grain diameter s observed, whereas in the case of discon-
tinuous grain growth, a small number of grains grow at the expense
of all the others (c.f. chapter 5, Fig. 5. 3. ). The latter case bears
a strong resemblance in external appearance to that of primary re-
crystallization, it is therefore frequently referred to as "secondary
recrystallization". The driving forc for grain growth stems from the
specific boundary energy of the grains, it is about three orders of
magnitude smaller than that for primary recrystallization (c. f. chap-
ter 2). Consequently, grain growth phenomena proceed much more
slowly. Furthermore, they can easily be influenced by external mani-
pulations.
The integral kinetics (isothermal experimental conditions) of normal
grain growth may be stated without difficulty, if a fixed distribution of
grain shapes and relative sizes is assumed and if furthermore the
quantity r- is assumed to be proporcional to the previously
mentioned driving forc ( c.f . chapters 2 section 2. 2. and 5 section 2.):
I? - D2 = k - t
(k: constant which depends also on temperatura).
This time-law has been f ound experimentally i nt he case of extremely
pur metis. Frequently, however, a smaller increase of mean grain
diameter with annealing time is observed. The main reason for this
probably is the obstruction of grain-boundary migration by impurity
The terminology used for these subphenomena vares f rom case to
case. The terms "grain growth" and "grain coarsening","discon-
tinuous grain growth", "exaggerated grain growth", "discontinuous
grain coarsening" and "secondary recrystallization" are used
synonymously. The employment of the conventional expression
"grain growth" to describe the behaviour defined by process 5) is
unfortunate, because it may give rise to confusin with the process
of pri mary growth of grains. The expression "grain coarsentng"is
a better characterization of the actual state of affairs.
1 1
atoms. This hindrance to normal grain growth can reach such an ex-
tent that the boundaries can move at a very small number of places
only. This is the case of discontinuous grain growth. The reason for
such a strong hindrance may be the existence of a very pronounced
texture, e. g. the cube texture in f, c. c. metis, or, in the case of
alloys, the precipitation of a second phase (see chapter 5). The time-
dependence of the secondary recrystallization process may be describ-
ed, in analogy to the case of primary recrystallization, by an Avrami-
relation.
Generally, normal grain growth ceases, when the mean grain diameter
has reached a valu of the order of the specimen dmensions, because
after that the grain boundaries start to become pnned at the surface
by thermal grooving (see chapters 2 section 3. 4. and 5 section 5.). If
this anchoring effect is small, however, then migration of grain boun-
daries at the surface can be observed in exceptional cases. The
corresponding driving forc results from the difference in the surface
energy of grains of different orientation (c. f. chapters 2 and 5). This
process is called "tertiary recrystallization".
1.4. The Inf luence of Foreign Atoms
In the presence of foreign atoms, the processes occuring during
annealing become very complicated in general as a result of the diversi-
ty of possible interactions between physical and Chemical defects. This
has twofold consequences. Firstly, the initially unstable state prior to
annealing is not a superposition of the influences of the two defect
types. An example for this is, in the case of f . c. c. metis, the
change of the deformed structure, when the stacking fault energy is
lowered by foreign atona. Secondly, foreign atoms modify the pre-
viously listed fundamental processes. Thereby one must distinguish
between systems whi ch, at the beglnning of annealing, contain the
foreign atoms either in solid solution, in supersaturation or in the form
of precipitates (see chapters 7 and 8). In the latter two cases, the fact
that during annealing precipitation and solution processes may occur
before, at the same ti me or after the structural basic processes has
to be taken into account as well (see chapter 8).
A partial classification of this variety of combinations is possible in
t erms of the modification of the driving forc due to forei gn atoms.
Such a procedure is relatively straightforward in the case of the funda-
mental process 4). The details of this modification depend, of course,
on the particular situation: If, for example, during grain boundary
migration precipitation occurs exclusvely at the boundaries, then an
additional driving forc of Chemical origin results (c.f. chapters 2
section 2. 5. and chapter 8 section 2. l.b). If, on the other hand, thermo-
dynamically stable precipitates already exist in front of the migrating
grain boundaries, then a friction forc resulting from the particles
will act on the boundaries. However, a more complicated situation
oc c urs if a metastable phase is precipitated out in front of the moving
boundaries. Thus the metastable particles can be transformed into the
stable equj li bri um phase by the mi grati ng boundari es. In thi s case, the
expression for the driving forc is supplemented by a transformation
t erm.
With the aid of the ti me-dependence of the i ndi vi dual dri vi ng forces the
kinetics of the processes can be classified as well (see chapter 8). The
fundament al processes 3) and 5) can also be discussed analogously. With
regard to the problem of nucleation, however, this treatment is of a
largely speculative nature at the present ti me, since even in the case
of pur metis little details are known about the process of nucleation.
1 3
2. FUNDAMENTAL PROBLEMS
2.1. Unstable Initial State
The physical and chemical structure of the initial state determines
the recrystallization process. For this reason the starting-point of
any analysis of recrystallization phenomena is the question concerning
the structure of the unstable initial state, in particular with respect
to the structural details relevant to the subphenomena. The answer
to this question is generally very unsati sfactory, the underlying reason
being that in most cases only mean vales characterizing the initial
state are known, such as the dislocation density, the texture, the stor-
ed energy and similar quantities. In addition to this, however, a
local description of the initial state is requi red in order to clarify the
relevant details. Three simple experimental observations may
illustrate this for the subphenomen of nucleation (c. f. chapter 3) :
a) Strained zinc single crystals in whi ch only one glide system has
been operated merely show a recovery upon annealing/' On the
other hand, single crystals def ormedby an equivalent amount in
mltiple slip exhibit recovery and recrystallization upon annealing.
b) In weakly rolled, polycrystalline alumi ni um, coloniens of nuclei,
originating from some intersection Unes of the grai n surfaces of
the initial state, grow into the deformed grains upon annealing.
c) In a deformed titan-molybdenum alloy the first recrystallization
nuclei are observed at the intersections of deformati on bands.
From such and related optical microscope observations it was already
concluded at an early stage that local misorientations in the defect <
lattice are a prerequisite for the formati on of large-angle boundaries.
It should be possible to gain furtherreaching information with the aid
of the technique of transmi ssi on electrn microscopy.
Analogous examples concerni ng the mi grati on of large-angle
boundari es can also be given.
14
The great expectations placed on this experimental technique have not, :
however, been fully realized. The essential reason is that, with the
electrn microscopes commonly used so f ar, only very thin foils could
be penetrated with the electrn beam. Recrystallization processes in !
such thin foils, however, differ f rom those in bulk material (the kinetics
are at least slower and probably also qualitatively different). In order
to obtain reliable results, a bulk specimen must be annealed from which
a thin foil is subsequently prepared. This means, however, that two
states of one and the same regin (in the bulk material) can never be
observed. Therefore, the hitherto employed investigation technique
provides at best statistical correlations.
In spite of these fundamental limitations, many studies of recrystalliz-f.
ation have been performed with the aid of the transmission electrn
microscopy technique in recent years on materials, ranging from
specifically deformed single crystals to complicated manyphase,
polycrystalline materials (see chapters 3,6 and chapter 8). Still, the
question concerning the relevant details in each individual case has
seldom been solved unambiguously. In particular, it remains to be
decided to what extent the experimental data, frequently interpreted in
a different sense by the individual authors, can be generalized. It is
to be expected that this unsatisfactory situation will soon be improved
by the application of the recently developed high voltage electrn
microscopes.
2.2. Formation of Large-Angle Boundaries {Nucleation)
The cardinal question of nucleation is: How can a perfect lattice
regin of another orientation that is capable of growth be formed in a
defect lattice ? In the attempt to answer this question, the fundamental
difficulty is encountered that the experimental data are always obtained
1 5
on individual objects which are transferred by some means into the
range of "perceptibility". In other words, measurement always
registersthe "perceptible nucleation" which, however, may be related
to a larger or to a smaller extent to growth processes. Experimental-
ly, a distinct separation between the so-called "processof nucleation"
and the growth of the nucleus is probably never possible. In the
evaluation of the information on nucleation that is contained in the ex-
perimental data, this aspect must be taken into account critically at
all times, especially when compatison is made with theoretical pre-
di ctions.
In the course of time four different conceptions of nucleation have
essentially been developed: the "classical model", the "umklapp model",
the model of "subgrain coarsening" and the "bulging mechanism". The
former two models rely on the conception that nucleation is a spontaneous,
self-contained process. In the case of a crystalline defect body in
which a large number of reactions can occur upon annealing, such an
approach appears to be a priori not too promising. The notion that
nucleation is a more or less continuous sequence of reactions of lattice
defects finally leading to a regin that is capable of growth in another
environment should be more appropriate to the state of the material.
This more realistic view is taken in the case of the latter two models
which are therefore in the focus of current interest. In the following
the essential features of the four models shall be briefly outlined (see
chapter 3 section 3.).
The classical model is based on the asaumption that the nucleus which
is capable of growth is formed as a result of thermal fluctuations in the
sense of the classical nucleation theory of phase transitons. The
critical nucleus size and the activation energy, as estimated on these
grounds for a homogeneous material, are, however, by far too large.
f"a solid, the obvious alternative is the assumption of heterogeneous
1 6
nucleation, e. g. n regions of local stress concentrations. This does
indeed lead to lower vales of the cr'^cal nucleus sizes and activation
energies. However, in order to obtain an agreement in the order of
magnitude with the experimental data, stress concentrations correspon
14 15 -2
ing to dislocation densities between about 10 and 10 cm are
required. Such high dislocation densities have not, however, been
observed so far. Even if such dislocation densities exist locally, the
nucleus cannot grow beyond this small regin of high dislocation densitj.
"
since outside this regin, it immediately becomes subcritical. In such
a materia!, a nucleus could well be formed but would not be capable of
growth. It is therefore fairly sure that nucleation cannot be explained
in terms of the classical model.
r
The umklapp model has been proposed to explain the formation of
nuclei in the cube orientation during the annealing of rolled f. c. c. metal
It is based on the assumption that, in the deformed state, lattice regin
of specific mutual orientation and with specifically oriented interfaces
occasionally lie next to one another. These two lattice regions are
supposed to be able to attain a new common orientation by an umklapp
process, the socalled "inverse Rowland transformation". In this case,
the lattice regin formed in this way is the nucleus which is capable of
growth. It has not been possible so far to obtain convincing experiment;
evidence in favour of such a mechanism.
1 7
places. The critical and controversial point is the transition from a
small- to a large-angle boundary. This is considered to be possible
in two ways, by "subgrain growth" and by "subgrain coalescence".
Evidence in favour of both mechanisms has been obtained by electrn
microscopy.
The bulging mechanism is basically a model of growth, since it is
assumed that the unstable state already comprises large-angle boun-
daries of macroscopic or microscopio dimensions and that these
boundaries must merely bow out. Experimental evidences in favour
of both kinds of bulging have been obtained repeatedly, e. g., in the
case of m aero-bulging, the previously mentioned phenomenon of
strain-induced boundary migration.
Considerable progress should be possible in the understanding of
nucleation by finding the answers to the following questions: Under
which circumstances are subgrain coarsening and bulging mechanisms
observed preferentially? Which are the corresponding relevant details?
What are the theoretical and the experimental orientation distributions
of the nuclei in both cases? The solution to the last question is also of
considerable importance for the understanding of the annealing textures,
since these represent the integral result of the anisotropy of nucleation
and grain-boundary migration (c. f. chapters 3, 6 and 1O).
The subgrain coarsening models are associated with heterogeneous
nucleation. In this case the assumption is that, in the material, pre-
formed regions of high dislocation density which are strongly mis-
oriented with respect to their environment exist. These regions are
supposed to be able to grow as a result of the slow and gradual recovery
and rearrangement of lattice defects, in particular dislocations. At
first, small-angle boundaries (subgrains) are thought to be formed,
and from these mobile large-angle boundaries fi nal l y evolve at some
. 3. Migration of Large-Angle Boundaries
The basic question of grain-boundary migration is: How does the
displacement occur and which are the associated transport phenomena?
He re, the present situation is more satisfactory than in the case of
nucleation. A relatively large number of consistent experimental re-
sulte on the rate of migration are available and, in addition, some
quantitative models exist (see chapters 4 and 7).
1 8
'
In the theoreticai mudis of grain-boundary migration, the boundary is
treated as a strongly disturbed layer of finite thickness. The displace-
ment is assumed to occur by uncorrelated j umps of individual atoms
from the lattices into this layer, transport of atoms through the layer
and the growth of individual atoms onto the lattices. Atoms that are
transferred from the lattices into the boundary must surmount an
activation energy barrier. The differences between the various theore-
tical approaches le in the extent to which details have been considered
in the calculation of the transfer of atoms from the lattices into the
boundary (or vice versa) and of the passage of atoms through the
boundary. The most important experimental findings can in principie
be understood in terms of the models (see chapter 4).
The most important unsolved questions concerning grain-boundary
migration in pur metis are the following: What are the exact re-
lationships between the rate and the driving forc and what is their
interpretation? Is the elementary procesa really the uncorrelated
jumping of individual atoms? If yes, which is the rate-determining
substep in the transfer of atoms from one lattice to the other? What
are the details of the orientation-dependence of the rate?
In the theoretical models that deal with the influence of solute impurity
atoms, it is irst assumed that the elementary process of material
transport from one lattice to the other is not affected by small amounts
of impurity atoms and is the same as in pur metis. The effect of
impurity atoms results from the existence of an interaction energy be-
tween the impurity atoms in the lattice and the boundary. Thereby the
impurity atoms exert a friction forc on the moving boundary. Compar-
ed with the situation in a pur metal, the behaviour becomes very com-
plex, because the friction forc itself is rate-dependent. Many experi-
ment al observations can be interpreted along these Unes (see chapter
7 section 2).
1 9
The questions of current interest in this rea are the following: To
what extent can the theory be verified by reliable experiments? What
is a "pur metal" in the theoretical sense? Do impurity atoms which
generally accumulate at the boundary affect the structure of the boun-
dary? If yes, is the assumption that the elementary process of material
transport is not changed by small contents of impurity atoms in the
boundary really valid in this case? What is the influence of impurity
atoms on the anisotropy of the rate of migration?
3. CONCLUDING REMARKS
The phenomena associated with recrystallization may be reduced to
a few fundamental processes. They may be classified clearly in terms
of the driving forcea responsible for these fundamental processes. This
approach facilitates the understanding of the individual processes on a
structural, atomistic basis.
The experimental and theoretical overall knowledge in this field has
in the meantime attained a standard permitting the detailed examination
of specific model conceptions on the one hand and the specifically planned
development of materials for practical purposes on the other hand (see
chapter 1O and also chapters 5, 6, 8). The significan! and to a large ex-
tent unsolved questions concern the interaction between structural and
Chemical defects in microscopic and atomic dimensions.
2 1
2 O
CHAPTER 2
DRIVING FORCES FOR RECRYSTALLIZATION
BY
H.- P. STWE
1. DEFINITIONS
,
Recrystallization can be defined as the migration of large angle grain
boundaries [1 ](c.f. chapter 1 section 1.1.). The boundaries will
move if thereby the Gibb' s free energy of the system is lowered.
Figure 2.1 shows schematically a cylindrcal body of cross-section F
and length L.
Fig. 2.1
Grain boundary between a growing crystal ( I )
and the rest of the specimen ( n ).
It consists of two parts. Part I is a crystal, e. g. a metallic grain.
Its boundary is represented by a vertical Une. Part II is the rest of
the system and may be a deformed single crystal, a polycrystal, or
2 2
a single crystal of a different orientation. Let g. and g_ be the free
energy ( c. f. chapter 1 section 1 . 1 . ) per mole of these two parte and
g__ > gT . The total free energy of the system is then
estimated as [2j
2 3
m = * (4)
i
tot
(U
and G. , wlll be lowered if the grain boundary moves to the right
(V: mole volume). It can be said that the grain boundary is pulled to
the right by a "driving forc" p, the size of which is
d G
tot
dx
-gj) y
(2)
This interpretation is generally accepted and shall be used in this
article. It should be noted, however, that it is not quite correct since
the boundary is not a mechanical body on which Eorces may act. The
situation is quite similar to the movement of dislocations which are
also said to be acted upon by external forces although they are not
mechanical bodies.
This paper will list a number of driving forces that can move grain
boundaries during an anneal. First "forward driving forces" will be
considered which permit small grains to grow into larger grains. Next,
"dragging forces" will be treated which handicap the forward movement
of boundaries. Finally, sonie effects will be discussed that can be unde
stood f rom a balance of several driving forces.
The velocity of a boundary moving through the lattice is proporcional to
the sum of all driving f orces, i. e.
v = m (3)
where b is the Burgers vector, V mole volume, R the gas constant,
T absolute temperature and D an appropriate diffusion coefficient. It
is equal or at least related to the coefficient of grain boundary self
diffusion [33 .
D depends very strongly (exponentially) on temperature. This is the
reason recrystallization is usually observed at elevated temperatures ,
By contrasi, the driving forces to be treated in this paper do not
appreclably depend on temperature (with the exception of p-).
FORWARD DRIVING FORCES
1. Stored Energy from Cold Work
The most importar* recrystallization phenomenon is primary re-
crystallization, i. e. the growth of new grains in a deformed matrix.
The driving forc for this procesa is the stored energy of cold work.
Most of the mechanical work necessary to deform a metal is converted
into heat. Only a small part is retained in the metal * - - not more
than a few percent at room temperature. This part is called "stored
energy". It is localized almost entirely in the stress fields around
the dislocations formed during cold work. The line energy of a dis-
2
location is about G- b , where G is the shear modulus and b the
Burgers vector. Denoting the dislocation density by N the driving
forc in a deformed matrix is given by
(5)
The factor m is called the "mobility" of the grain boundary. It can be
Recrystallization may also occur at room temperature if time is
measured on a geological scale (e. g., in ore deposits) [4 j.
2 4
This equation is not very accurate because the Une energy itself de-
pends somewhat on the dislocation density. Fortunately, this leads
only to a minor correction. More importantly, the stored energy also
depends on the details of dislocation arrangement as is illustrated in
Flg. 2.2.
a o
Fig. 2.2
Dislocation arrangement: (a) after deformation
(b) after polygonization
It shows two bodies having different arrangements of an equal number
of dislocations. In Fig. 2, 2a dislocations are piled up in slip planes,
symbolizing a cold worked metal. In Fig. 2. 2b dislocations are arrang-
ed in sub-boundaries, symbolizing a polygonized metal. The state in
Fig. 2. 2b is energetically more favourable than that in Fig. 2.2a; the
difference in energy has yielded the driving forc for polygonization.
Polygonized metal will therefore yield a lower driving forc for re-
crystallization than the cold worked metal - - even if dislocation den- 2. 3.
sities are equal. 4' Thus, eq. 5 should contain a factor taking into
account dislocation arrangement. Its valu lies somewhere between
0,1 and 1. Since this paper is mainly concerned with the orders of
magnitude of driving forces this factor has been omitted. Furthermore
experimentally it is difcult to determine the dislocation density to
better than a factor of 2.
2 5
1 2 2
Ufling plausible vales In eq. 5 (G =10 dyn/cm , b=r3-10 cm,
12 2
N 10 /cm ), the driving forc p., for primary recrystallization is
9 2
estimated as about 10 dyn/cm .
,2. Grain Boundary Energy
The grain boundary is associated with an interfacial specific energy
*. Therefore, a fine grained structure has a higher free energy than
a single crystal of equal size. This difference provides the driving
forc for normal grain growth and secondary recrystallization.
Imagine a polycrystal subdivided into cubical grains of edge D. Each
2
cube will then have six faces D which are shared with its neighbours.
The contribution of grain boundaries to the free energy of a polycrystal
is thus obtained as
6-D .yn 3^ . (6)
"2-
In addition, the dislocation density usually decreases during poly-
gonization.
'B
2D'
D
y
If #0 300 erg/cm , a single crystal growing into a matrix with
grains of 0,1 mm diameter will experience a driving forc of about
5 2
10 dyn/cm .
Surface Energy
If I and n in Fig. 2.1 are both annealed single crystals, they still
may differ in their orientation, and thus, in the specific energy of
their free surfaces. This too may yield a driving forc. If Av is the
difference in specific free surface energies, this driving forc is
equalto
PS = ~ ' (7> O W
f
where w is the thickness of the sheet and the factor 2 shows that
2
large grains in thin sheets have two free surfaces. A* 50 erg/cm
in a sheet of 0,1mm thickness may thus lead to a driving forc of abou
104 dyn/cm2.
In the annealing of thin, cold worked sheets the three driving forces
discussed so far may b ecome effective in the order of their size:
First, the stored energy of cold work, p., is released (primary re-
crystallization). Next, grain b oundary energy is released as far as
possible ( p_). This is called normal grain growth when the average
grain diameter increases steadily, and secondary recrystallization
when a few grains grow to excessive grain size in an otherwise fine
grained matrix (see chapter 5, Fig. 5. 3). Finally, grains of low
surface energy may grow at the expense of their competitors ( p., );
this has been called "tertiary recrystallization" C7] (c.f. chapter 5
section 6. and chapter 6 section 2. 3. 6.). Experimentally it is not
always easy to draw a clear Une between these three stages.
2.4. Magnetic Energy
Another driving forc can b e exerted b y a magnetic field on the
grain b oundaries of ferromagnetic materials. If for example, I and
II in Fig. 2.1 are two iron crystals, where I has a <100>-direction
and II a <111> -direction parallel to an external magnetic field, then
the specific magnetic energy of the two crystals di ffers b y an amount
of 143000 Gauss'Oerstedt; i. e,, the boundary is pulled to the right
b y a driving forc
2 7
crystals magnetically saturated. Larger fields will not increase the
driving forc. In smaller fields, however, the driving forc decreases
rapidly. It is essential that the material be ferromagnetic; the anneal
must therefore be carried out below the Curie point (769 C in iron).
Because p4 is rather small it is difficult to influencethe recrystalliz-
ation texture of steel sheets appreciably by magnetic fields.
5. Discontinuous Precipitation
1
The elementary process of discontinuous precipitation is again the
movement of a large angle grain boundary; so that this phenomenon,
too, can be understood as a form of recrystallization (see chapter 8).
In front of the moving grain boundary is a super saturated solid solution.
It leaves behind a two-phase structure which is near thermal equi-
librium. The driving forc comes from the gain in free energy effect-
ed by the decomposition of the solid solution. It is given by [8, 9]
^0,00114
cm cm
Jg =U[co (l-co ) -
+ l - c ) In (l In (9)
where c is the concentration of the alloy, T the temperature at
which c can be brought into solution, T- the annealing temperature
and c, the solubility corresponding to T- (Fig. 2. 3). For small
concent rations this simplifies to
~ u (C -
In In (9a)
TU, The constant U can be taken from the equation for the solubilitv
This valu holds when the external field is large enough to have both
2
0
o
c = e-U/RT
Inserting this in eq. 9a weobtain
> , ~ R (T- - T) c In c . 4
5 1 oo o V
i. DRAGGING FORCES
(3.1. Curvatureof the Grain Boundary
Consider a small spherical grain of radius r growing intoa deformed
matrix. Unlikegrain I in Fig. 2.1, the spherical grain must increase
its grain boundary rea during growth. This can beinterpreted as a
dragging forc that tries tomakethegrain shrink. Its magnitude is
Fg. 2.3
Section of a binary equilibrium diagram
with limited solid solubility.
For example4, 9 atomic percent silver can bebrought intosolid
solution in copper at 77 C and can bekept thereafter quenching. Sucl
an alloy will show discontinuous precipitation when anneafed at 300 C.
Using eq. 10and thesevales it follows that the growth front of the
"precipitation cells" is pulled forward by a driving forc of about
9 2
15 0 cal/Mol 6-10 dyn/cm . This is larger than all other driving
forces treated in this paper .
Themobility given by eq. 4 does not apply tothe process of dis-
continuous precipitation.
given by
d(4ifr
'B' 'B
d r
(11)
This result does not depend on details of geometry. It applies toany
general shape of thegrain boundary if r stands for the mean radius
of curvature.
Depending on the senseof curvature, pfi may not only bea dragging
forc but a driving forc as well.
By assuming various local curvatures thegrain boundary can experience
the samedriving forc at all points, even if the external driving forces
act upon it only at certain points or along certain Unes. Consider as
an examplethe driving forc p. It acts upon the boundary only in the
vicinity of thetwosurfaces of the sheet. Therefore, theboundary will
first bepulled forward near thesurface only. This causes theboundary
tobecurved as is shown in Fig. 2.4
Fig. 2.4
Curvatureof grain boundary during tertiary recrystallization.
3 O
Now, the grain boundary in the interior of the specimen also ex-
periences a driving forc; in the steady state it is equal to p_
3
so that all parts of the boundary move with the same velocity.
Measurng the curvature of the boundary thus leads to an estmate
f o r A / .
E
The situation is similar for the driving forc p, . Equation 6 is
!'
only an integral statement on the energy to be gained by transforming '
a fine grained structure into a single crystal ( or a coarse grained
structure). The individual grain boundary does not "know", however,
which way it should move in order to increase the average grain size
of material. It will rather move towards its center of curvature. .
r
Where this center of curvature lies is essentially determined by F
I
the number of faces a grain has. Since most grain boundaries have ' f
about the same interfacial energy *' they will (in equilibrium)
form angles of about 120 at the grain edges. This will make the
faces of a relatively small grain ( having few neighbours ) convex,
the faces of a relatively large grain ( having many neighbours )
concave. This is seen at one glance in the two dimensional example
of Fig. 2. 5.
pe/pon
Fig. 2. 5
Curvature of grain boundaries in n-sided,
two-dimensional grains (angle at corners 120 ).
Thus, small grains have the tendency to shrink (and finally disappear),
large ones have the tendency to grow. This is how the curvature of
grain boundaries increases the average grain size after primary re-
crystallization.
Twin boundaries and small angle grain boundaries have smaller
vales.
3.2. Second Phase Partiles
Figure 2. 6a show's a cubic particle intersected by a grain boundary.
The section occupied by the particle corresponds to a saving n grain
boundary energy. This is the reason position I is energetically
favourubh ; wdernal forces are necessary to pul the boundary from
this pusition.
E
Fig. 2.6.
Interaction between grain boundary and
(a) cubic inclusin, (b) spherical inclusin.
The sign of such forces depends on the exact position of the boundary.
A boundary in position K will be pulled to the right; a boundary in
position m to the left. Therefore, segregated particles cannot pul
the grain boundary over a large distance. If, however, another driv-
ing forc tries to pul the boundary through the matrix, the boundary
will have the tendency to stick to all second phase particles. In this
case, all inclusions exert dragging forces which can be summed up.
If f is the fraction of grain boundary rea occupied by inclusions this
means a saving of f Jn order to pul the boundary forward this
1
energy must be invested over a distance of ~

d.
i
where d. is the
i
diameter of the inclusions. The backward drag exerted by the in-
clusions is thus given by
9 . f .1.
'3
2 - f
P.,- (12)
3 3
This equation is exact only if the inclusions have the "cube-on-edge"
shape shown in Fig. 2. 6a. Other shapes will lead to a correction of
the order of 1; it is 2 in the case of spherical particles. For spheri-
cal particles, we obtain from Fig. 2.6b
d 2 d di2 2
where 2 is the number of particles per cm of grain boundary. Ob-
viously, the backward drag is largest for x = d./2. Inserting
2
Z (d./2) ir* f the mximum drag is given by
'B
(14)
Usually f is taken as the volume fraction of second phase particles.
This would be exact for a rigid grain boundary intersecting segregated
particles at random such as the surface of a metallographic specimen
does. Actually, a slightly higher valu should be used because the
grain boundary is flexible and likes to stick to segregates. A fine
dispersin of small particles will, however, lead only to a small
correction in eq. 14.
Setting f = 1 % and d. = 10 Aim a drag of about 10 dyn/cm is ob-
tained. A comparison with eq. 5 shows immediately that segregated
particles can have no influence on primary recrystallization: A much
higher driving forc will pul the grain boundary across all obstacles.
On the other hand, a comparison with eq. 6 will show that segregates
may well have a decisive influence on normal grain growth and second-
ary recrystallization.
i
3 4
3. 3. Solute Impurity Atoras
Equation 14 shows that a given amount of second phase inclusions
will exert greater drag on the grain boundaries the finer it is dispers-
ed in the matrix. One might be tempted to use eq. 14 for an estmate
of the influence of solute atoms on a migrating grain boundary usingth-
atomic diameter for d.; and the concentration c for f. A valu of
9 2
c-1% then yields a dragging forc pg~ 10 dyn/cm . This would
immediately explain why solute atoms (other than second phase
particles) can also influence primary recrystallization.
This simple consideration contains two errors, both of which are
caused by the mobility of the foreign atoms which - - unlike second
phase particles - - may freely move through the lattice.
a) Impurity atoms are not evenly distributed through the lattice but
segregated to the grain boundaries. Therefore f in eq. 14 is not
the average concentration of solute atoms but an elevated con-
centration at the grain boundary which, in equilibrium, should be
f c . exp UB/ K_T (15)
where UB is the interaction energy between grain boundary and im-
purity atom. It usually amounts to a few tenths of an eV [3] .
This may increase the drag exerted by impurity atoms by several
orders of magnitude. This explains why additions of only a few pprn
solute atoms may appreciably influence the recrystallization
kinetics of high purity metis.
It is feasible that UR might assume negative vales (conceivably
for such alloy additions that raise the melting point, such as Ni
in Cu). Then the concentration in the grain boundary would be lower
than in the matrix. Nevertheless, the resultant dragging forc
would be equally high because in this case solute foreign atoms
would be "pushed ahead" by the migrating grain boundary [5] .
3 5
b) The mobility of solute atoms enables them to travel wlth the moving
grain boundary. At very small grain boundary velocities the solute
atoms will therefore not exert any significan* drag. At very high
migration rates the grain boundary can "break away" from its im-
purity atoms and may again move with a velocity equal to that in a
pur metal. The dragging forc estimated from eqs. 14 and 15 will
therefore be effective only in an intermedate range of migration
rates when solute atoms can diffuse behind the migrating grain boun-
dary and exert their dragging forc. The details of this process are
quite complicated and have been discusses elsewhere [3, 5] .
It is difficult to verify the distribution of solute atoms around the grain
boundary experimentally. Therefore, this type of dragging forc has
frequently been neglected. The influence of solute atoms is then under-
stood as an influence on the grain boundary mobility which then does
not correspond to the simple eq. 4 but instead will be a complicated
function of grain boundary velocity and, thus, of driving forc [3, 5].
or is it then possible to describe its temperature dependence by one
simple activation energy. This is the reason why the "activation
energies for recrystallization11 reported in the literature show such
wide discrepancies, even for the same material (c. f. chapter 7).
1.4. Surface Grooves
Thermal etching will produce a groove on the metal surface where
it is intersected by a grain boundary. Such a groove is shown schema-
tically in Fig. 2. 7.
3 G
3 7
4. BALANCE BETWEEN SEVERAL DRIVING FORCES
Fig. 2. 7
Section through groove produced by
thermal etching on a metal surface.
The opening angle (180 - 2 (p) is determined by the ratio of grain
boundary energy tf and surface energy # as
4.1.
tg < p sin
(16)
If the boundary moves out of the groove it must expand; this produces
a dragging forc
Po -
'B tg < p
(17)
Setting tfg/a' g O , 5, a sheet of thickness 0, 1 mm will experience
a dragging forc of about 10 dyn/cm . If the surface energies of
grains I and n are different (as discussed in section 3), Fig. 2. 7
will be asymmetric. This (more complicated) situation has been dis-
cussed in detail by MULLINS [6 ] (c. f. chapter 5 section 5).
1. 2.
Stable Grain Size
From eq. 6 we obtain for the kinetics of normal grain growth
d D/ d t ~3 tfB /D (18)
1
2
so that D~r.
This is quite in accord with experiments showing the grain size to in-
crease with the square root of annealing time. This is true, however,
only up to a limiting grain size; from then on the grain size (in many
cases) remains constant. The existence of a limiting grain size shows
that the driving forc for grain growth has been lowered to the point
where it is equal to the dragging forc exerted by segregate partiles.
Cotnparing the forces p (eq. 6) and p? (eq. 14) we obtain
D = 4 f d.
and D.. =d. / .
lim. i
(19)
(20)
This equation permits the prediction of the limiting grain sizes achiev-
ed during the anneal of many technical alloys (see chapter 5 section 3
and chapter 8).
Specimen Thickness Effect
Sometimes normal grain growth will stop before the limiting grain
size predicted by eq. 20 has been reached. This occurs when the grain
size becomes comparable to the smallest dimensin of the specimen.
The phenomenon is called "specimen thickness effect"; the reason for
it is the anchoring of migrating boundaries by thermal grooving at the
3 8
specimen surface (c.f. chapter 5 section 5). The condition for this
efect is obtained by equating driving forces p2 and P9 as
B
w - g r , 2 w
(21
3 9
threshold strain below which no primary recrystaliization is observed.
Smaller strains may also lead to recrystaliization; since no nucleation
of new grains is possible one observes in this case the migration of
boundaries already present in the material. This "strain induced
grain boundary migration" is possible even when nucleation is not
because r is now of the order of the grain size (see chapter 3).
so that D
lim.
m 6 w
However, p0 will decrease even earlier. When the grain diameter I
t
comes larger than the thickness of the sheet, grains can grow in two
dimensions only. If the derivation of eq. 6 is modified accordingly
one obtains
D
Um.
21 4 W
in a sheet (and correspondingly D,. 2 w in a wire of thickness
Experimental vales for D.. are somewhat smaller.
4. 3. Critical Size of Nuclei
The critical size of a nucleus in recrystaliization must be such that
the driving forc permits the formation of new grain boundary rea.
The condition p, > pfi yields
N G > 2
so that a nucleus can grow only if
r >1 0D / Nc m.
(24
( 25
The recrystallization nuclei which so far have been observed seem to
have sizes of about 1 -2 ,um. Such nuclei can grow only if the dis-
10 2
location density exceeds 1 O /cm . This explains the existence of a
All driving forces except those for primary recrystallization and for
discontinuous precipitation are too small to permit spontaneous
nucleation. This is why normal grain growth and secondary re-
crystallization show no nucleation stage; in these processes grains
of macroscopic dimensions will serve as "nuclei" the growth of which
is permitted hy the condition that
> 2 =2 D/3.
( 2 6)
4.4. Influencing Directly the Recrystallization Texture of a Sheet
It would be most useful commercially if one were able to produce or
to prevent the formation of certain annealing textures at will (c. f.
chapters 6 and 10). Unfortunately, one is usually limited in finding
a favourable schedule of production (from the melt to the final pro-
duct) by trial and error.
Only two of the driving forces considered in this paper, p and p4,
are related to crystal orientation and therefore can be used to in-
fluence recrystallization textures directly . Both can become
effective only if they are larger than the dragging forc p_ caused by
second phase particles.
I ndirect influences, e. g. by way of the deformation texture or the
formation and dissolution of segregates are technically much more
important, but they are quite complex and beyond the scope of this
paper (c.f. chapters 6, 8, 10).
4 O
A comparison of eqs. 7 and 14 shows that tertiary recrystallization
will occur only if
w
\y suitable control of the furnace atmosphere it is sometimes possibleL
to favour the growth of grains with a desired orientation [7] . In such
cases g'p 0.1 j * so that the procedure can work only if
<^
f
A sheet of a few tenths of a mm in thickness containing particles of ^
a few Aim diameter may thus undergo tertiary recrystallization only
if their volume fraction is appreciably smaller than 0,1%.
This is why tertiary recrystallizationhas found practical application {
especially in the homogeneous Fe-Si alloys used for the production of ^
magnetic sheets with a favourable grain orientation (c.f. chapter 5).
A comparison of eqs. 8 and 14 shows that annealing a steel in a mag-
netic field will influence the recrystallization texture only if
REFERENCES
1. P. A. Beck:
2. J. E. Burke and
D. Turnbull:
3. K. Lcke and
H. -P.Stwe in:
4. H.Siemes:
5. J. W. Cahn:
6. W. W. Mullins:
7. J. L. Walter and
C. G. Dunn:
8. H. K. Hardy and
T. J. Heal in:
9. S. Horiuchi and
I. Gokyu in:
<
5 d.
cm
(29)
2
wheregLhas been assumed as about 500 erg/cm . Consequently also
in this case, second phase particles of about 2 ,um diameter may
not be present in a volume fraction exceeding 0,1%.
4 1
Phil. Mag, Supplement 3 (1954) 245.
Progr. in Metal Physics 3 (1952) 220.
"Recovery and Recrystallization
of Metals", L.Hitnmel Ed., John Wiley
and Sons, N. Y. 1963, 171.
N. Jahrb.f. Mineralogie 102 (1964) 1.
Acta Met. ^0 (1962) 789.
Acta Met. 6 (1958) 414.
Trans. AIME 215 (1959) 469.
"Progress in Metal Physics", B. Chalmers
andR.King, Eds., Pergamon Press,
London 5 (1954)146.
"Texturen in Forschung und Praxis",
J.Grewen und G. Wassermann, Ed.,
Springer, Berlin 1969, 312-
4 3
JHAPTER 3
/:C^->v
/> ' ''i-:
IUCLEATION IN RECRYSTALLIZATION
W. CAHN
NTRODUCTION
This chapter is concerned with the mechanisms by which new grains
come into existence when plastically deformed metis are heated, that
is to say, in primary recrystallization. The evidence from which one
can gain understanding of these mechanisms I ncludes direct micro-
graphic examination, by both light and electrons, and Information
about the individual and collective orientations of the new grains. A t
one time it was believed that theoretical arguments based on observad
kinetics of recrystallization, especially through attempts to fit parti-
cular equations to these kinetics, could yield indirect insght into
mcleation mechanisms, but this approach has been largely abandoned
now.
The nucleation problem in recrystallization has been very resistant
to generalization. This is largely due to the fact that it is remarkably
difficult to observe a nucleus under the microscope while it is stll so
small that it has only just begun steady growth - in formal thermo-
dynamic language, when it has only just grown beyond the stage of be-
ing an embryo. One of the characteristics of the nucleation process
in primary recrystallization is its extreme heterogeneity: nuclei
cluster in particular regions where the dislocation density is high.
'
This characteristic can be conveniently studied by optcal microscopy,
but the limited magnification available implies that nuclei can usually
Te seen only when they have already grown considerably. To overeme
his limitation - and to permit orientations of small grains to be de-
Srmined - a number of investigators have examined foils in the
4 4
electrn microscope. This approach has the drawback that only very
smallj volumes can be expiored, and the probability of finding an
interesting configuration is correspondingly small,
There have been many attempts to derive insight into the nature of ths
nucleation process by examining the detailed relationship between the
4 5
The mechanisms of nucleation
were first extensively discussed by
BURGERSflJ in his classic survey of the state of knowledge about re-
crystallization at the end of the thirtes. Following up one of BURGERS,
leads,CAHN [2] andBCK[3j independently proposed a model for the
nucleation process based on the newly discovered process of poly-
preferred orientation resulting from substantial deformation and the | T*~ ^*' ""t**"* Subse'uent *>* grains.
^ ! Next,BURKEandTUWBlL[4J comprehensively analyzed re-
annealing texture formed on subsequent heating. The drawback to this
strategy is, first, that one is seeking to interpret the nature of indi-
vidual events that collectively convert one statistical distribution into
another; there are too many disposable parameters for any unambiguo
conclusions to be drawn. Secondly, this approach is only feasible for
large deformations and yet small deformations may be more interest-
ing. Nevertheless, this type of work has been the direct cause of a
number of useful theoretical and experimental advances, and the
statistical methodology required has made great advances recently.
crystallization kinetics in terms of the constituent processes of
nucleation and grpwth, and sought in this way to throw light on the
nucleation process. In the same paper, they also applied classical
nucleation theory to recrystallization. Some details of this theory
were brought up to date in a book byBYRNE[5j . A further important
step forward was the detailed study byBAlIY andHIRSCH[6] of
strain-induced boundary migration, a form of nucleation originally
discovered byBECKandSPERRY[7j . The various mechanisms have
been previously reviewed by CAHN [8, 9] .
Plainly one could hope to learn more about the nucleation process if
one could determine the orientations of individual new grains in P^i^ACKGROUND EVIDENCF
ly recrystallized metal and relate these to the orientations of the in- [
dividual matrbc grains within which the new grains had been born. Thi: The following principies are generally valid, being based on extensive
strategy bypasses the problem of relating two statistical distributions bservational evidence, and any model of nucleation must be consistent
with them:
to each other, yet (provided one can examine an adequately large
population) allows one to relate the location and morphology of new
grains to their orientations, It may not even be necessary to catch tto
new grains at a very early stage of growth.
To achieve this aim one needs to combine the large field of view possi:
with the use of the light microscope with some method of determining
the orientations of small selected reas. Recent work in the author's
laboratory has been directed to this objective.
a) Nuclei form preferentially in a regin where residual damage due
to plstic deformation is greatest. More specifically, they form in
regions of large strain gradient. Thus it is not sufficient or indeed
necessary to subject metal to a high macroscopic strain: single
crystals of cadmium or zinc may be deformed by single glide to
strains of 100% or more without gaining the ability to recrystallize
on subsequent annealing. However, nuclei form freely even after
macroscopic strains of a few per cent only where dislocations
impinga on grain-boundaries, sizeable inclusions, or indeed on
4 6
other dislocations. Even free surfaces can, it appears, dam up
dislocations and enhance nucleation locally [10] .
This general observation has an important implication. Pile-ups
of dislocations shored up behind an obstacle necessarily cause a
local curvature of the lattice if - as is usual - they are all of the
same sign. Such curvatures are even more pronounced if two lots
of dislocations on intersecting glid "planes are dammed up by a
block of sessile dislocations. Figure 3.1 shows such a situation
and the contrast makes it obvious that the lattice is locally rotated
quite sharplyv On heating, such sites form recrystallization
nuclei, as can be seen in a different alloy in Fig. 3. 2.
Generally it can be asserted that local lattice misorientations are
a necessary and sufficient precondition for nucleation in a cold-
worked metal.
b) The larger the strain, the more nuclei are capable of being formed.
(The principie has to be expressed in this form, because increasing
strain may so accelerate growth of the nuclei that the matrix is
swallowed up before most of the potential nuclei have time to form) .
In terms of the preceding paragraph, this principie implies that in-
creasing strain leads to larger misorientations and more mis-
oriented sites. It is not generally realized how great a range of
orientations a grain in a polycrystal can acquire after quite a
modest strain. Thus Fig. 3. 3 shows a large grain in an aluimnium
foil which has been reduced 21% by compression, and then the
orientations of some 70 sites determined by the X-ray Kossel
technique (details will follow below). The results of this unpublish-
ed work by S. BELLIER and R. D. DOHERTY in the author' s laboraton
are shown in Fig. 3. 4, and it can be seen that macroscopic mis-
47
(001)
Fig. 3.1
Mo-35 at %Re alloy, deformed. Photographed in high-voltage
electrn microscope at 900 kV. (Dr. S. MAHAJAN, A. E. R. E. , Harwell).
Fig. 3.2
Ti-10 wt. % Mo alloy, reduced 3, 7% by rolling. annealed 5 min.
at 850C. ( Aft erT.H. SCHOFIELDand A.E. BACDN[32] ).
47a
Fig.3.3
: 99, 997% aluminium sample, deformed 21% by compression. Anodized
and photographed by polarized light. The spots indcate sites of Kossel
X-ray diffraction photographs. (S. BELUER and R. D. DOHERTY).
4 8
orientations of more than 20 have been created. These large-scale
misorientations also have an important role in nucleation, but it must
be remembered that large but highly localized misorientations of
the type shown in Fig. 3.1affect too small a lattice volume to be
detectable by means of X-ray diffraction photographs.
There is a special kind of deformation which is of particular im-
portance in connection with the formation of nuclei. This is the de^-
formation band. This term, once widely quoted, has somewhat
fallen out of use. It has been applied both to "artificial" boundaries
formed within the confines of a single grain as a consequence of
plstic deformation, or alternatively to bands (again created by
deformation) in which the orientation is sharply different f rom the
grain which envelops it on both sides. Either way, the essential
feature is a pseudo grain-boundary created by deformation in an
originally strain-free grain. Figure 3. 5 shows a well-developed
band (in the first sense of the word) in aluminium deformed 40%
by compression.BELLIERandDOHEFnYmeasured the misorientation
between points "1" and "2" and found it to be 35 . The lower grain
in the picture shows bands in the process of formation. Any such
strain-induced band must necessarily be the site of very high de-
formation gradierits, and correspondingly it has often been found
that nuclei form preferentially at deformation bands. Thus HU [11]
andWAL3RandKOCH[2l] have thoroughly documented such nucle-
ation in rolled sicon-iron crystals by both optical and electrn
microscopy. Figure 3. 6 shows a single new grain nucleated at a
"weak" deformation band in aluminium, deformed by 20%. At
twice this strain level, extensive nucleation tak^ s pluce at these
artificial grain boundaries.
Fig. 3.4
Stereographic plot of <lll>poles correaponding to the sites indicated
in Fig. 3. 3 (Misorientation between points 1 and 2 is 20 ).
Fig. 3.5
Deformation band in aluminium sample, deformed 40% in compression
). DOHERTY).
49a
F i e
Gttfi nucle ate d at Ae(0m*aR band i n alumi ni um de forme d 21%by
compre ssi on and anne ai cu 4i ... A . 330 C.(S.BELLIERand R.D.DOHERTY}.
c) A special interest attaches totwo-phase alloys, since many in-
dustrial materials are of this type. The considerable complications
in the behaviour of such alloys are deatt with elsewhere in this
volume (c.f. chapter 8). For present purposes, it is enough to
point out that coarse and fine disperions behave quite differently.
In summary, fine dispersions inhibit both nucleation and growth
severely, and therefore raise the recrystallization temperatura
for a given degree of deformation. Coarse dispersions aid nucle-
ation without seriously slowing the growth of nuclei, and therefore
the recrystallization temperature is lowered. Nucleation at oxide
inclusions in a Fe-0 alloy has been clearly demonstrated on micro-
graphsbylESLIE.MBCHAUKandAlJL [12] . The most detailed
study of the transition from hindrance to promotion of recrystalliz-
ation, as the dispersin is gradually made coarser, is due to MOULD
and COTTERILL [13] . They studied a series of dilute Al-Fe
alloys in the fully precipitated condition, deformed and then anneal-
ed below the re-solution temperature. For details of this work, the
reader must be referred to the original paper, but Fig. 3. 7 shows
some derived results, combined with some results for Al-Cu alloys
duetoDOHERTYandMARTIN[H] . This graph confirms the existence
of an optimum scale of dispersin, and also a second, smaller
critical inter-particle spacing which the authors explain in terms
of a particular nucleation model. The remarkably steep fall in
nucleation and growth rates below the critical 4 / um spacing ig
/
attributed to the "Zener drag" of the particles acting on sub-boun-
daries (and thereby inhibiting nucleation) or on large-angle boundaries
(and thereby inhibiting growth) (see chapter 2). Fine dispersions of
AlgFe in Al-Fe alloys can greatly inhibit recrystallization, and cer-
tainly nucleation is not enhanced. The behaviour of such alloys can
be very complicated if precipitation and recrystallization take place
concurrently [25] .
1
B-WWTlCLe SPACINC U
Fig. 3.7
Nucleation and growth rates in Al-Cu alloys (o) and Al-Fe alloys ().
In each case, 60% strain and anneals at 300C.(After P.RMOULDand
P. COTTHULL [13]).
Evidence concerning orien.tationa_ of nuclei will be deferred until
the various nucleation models have been reviewed.
3. MODELS OF NUCLEATION
Three principal models have been serious contenders in the various
attempts to explain nucleation. These may be t ermed classical nucle-
ation , subgrain-growth model and strain-induced boundary migration,
a) Classical Nucleation-BURKE andTURNBULLft ] first examined in
detail whether the classical Volmer-Becker theory of nucleation,
as ordinarily applied to the format ion of solid nuclei in supercooled
liquida, could be extended to the situation of primary recrystalliz-
ation. In this theory, a critical nucleus size is determined by the
condition that for the nucleus to be just stable, an infinitesimal in-
crease in size will result in zero change in free energy, because
increase in interfacial energy is just balanced by reduction in
volume internal energy. In the present case, this second t erm mus
represent the energy of dislocations destroyed or favourably re-
arranged as the nucleus grows.
The model implies a continuous agitation of the mat rix phase, so
that subcritical embyos are constantly being generated and relaps-
ing into the mat rix configuration. Occasionally, an embryo by
chance exceeds the critical size and will t hen grow.
This model implies a Hquid-like state at the core of the nucleus-to-
be, which is a crude approximation to the state of affairs where a
very high dislocation density exists locally - that is, just where
nuclei are observed to form. A low specific int erfacial energy im-
plies a smaller critical nucleus size, and low-angle boundaries havt
,53
a low interfacial energy: therefore nuclei not differing too much in
orientation from the matrix would be expected.
BURKEandTURNBULL conclude that this model is energetically feas-
ible if one allows for the fact that most of the measured stored
strain energy of only a few caloes per gram-atom is in fact con-
centrated in a few compact sites. BYRNE [5] attempts to quanti-
fy this concentration of energy, and concludes that an enormous
local strain of- 0,20 would be required to produce a reasonable
o
Cfitical nucleus size of = 6 A for his particular set of assumptions.
A size even slightly larger than this would be astronomically
improbable to crate by chance configurational fluctuatons. This
virtually amounts to saying that the material must be locally
amorphous.
This model has certain advantages; for instance, it can account for
the existence of substantial incubation periods. Since there are so
many disposable parameters, especially the degree of local con-
centration of strain energy, and since the critical size is far too
small to observe clearly even at the limit of resolution of the
electrn microscope, no experimental method of disproving this
model appears to exist, and one can only exclude it by applying
Occam' s Razor: the other models fit the facts without the necessi-
t yf or such extreme assumptions as to very large, very concentrated
local strains.
b) Subgrain-Growth Model - In its original form [2, 3] this model was
based on the idea that classical polygonization - the rearrangement
of uniformly distributed edge dislocations into arrays normal to
the Burgers vector - creates strain-free cells which can then grow
into the differently oriented surrounding matrix. CAHN [2]original-
ly justified this model in detail in t erms of the resulting kinetics -
54
particularly the resulting incubation period and subsequent mximum
in nucleation rate - but since then various different kinetics have
been recorded experimentally according to the material and circum-
stances, and arguments as to generally valid nucleation mechanisms
based on recrystallization kinetics are therefore of doubtful valu.
CAHN also based himself on observed relationships between de-
formation and annealing textures, but again subsequent work on
growth selectivity amongst nuclei has shown that this kind of argu-
ment cannot contribute usefuUy to the Identification of nucleation
mechanisms.
The model has now evolved into something whch can best be ex-
pressed by a sketch (Fig. 3. 8). The essence of the model is that
a small regin of high dislocation density, and therefore of high
strain gradient and by implication of substantial local misorient-
ation, turns into a small strain-free cell by a process of dislocation
climb and rearrangement. The structure becomes akin to a cell
structure created by creep, and it is known rom X-ray analysis
[15] that the boundaries of such cells have dislocation structures
similar to sub-boundaries formed by classical polygonization (of
macroscopically bent single crystals).
Once a dislocation-free subgrain has formed, it can grow into its
neighbourhood. Any subgrain population is characterized by two
distinct parameters - the distribution of sizes of the subgrains and
the distribution of the misorientations of subgrains relative to their
immediate neighbours. A subgrain can grow at the expense of its
immediate neighbours for two distinct reasons: (i) It is substantial-
ly larger than its neighbours, so that the triple-point angles where
its periphery abuts on other sub-boundaries depart from 120
equilibrium (Fig. 3. 9). The subgrain grows because of the attempt
to get these triple points back to equilibrium. This argument is
55
Fig. 3.8
Nucleation by subgrain growth (schematic). Subgrain boundaries
thickly populated by dislocations (dots) have a high misorientation
angle, and are the most likely to migrate.
Fig. 3.9
E ffect of relative size of a grain on the curvature of grain boundaries.
Arrows show direction of boundary migration, provided all boundary
energies are equal. (After R. L. COBLE and J. E. BURKE ).
5 6
exactly luce the accepted interpretaron of the coarsening of large
grains in secondary recrystallization, and indeed Fig. 3. 9 was
originally drawn in this connection. The argument only applies
strictly if all the sub-boundaries in Fig. 3.9 have the same energy,
1. e. all misorientations between nearest neighbours are of the
same type and magnitude. This however is not likely to be so,
and indeed (ii) highly disoriented subgrains tend to grow at the
expense of their neighbours even if they are jipi larger than the
average size, as ODTTRELL[16] was the first to emphasize. At
first sight this is very surprising. Classical nucleation theory
shows us that a small interfacial energy implies a small critical
nucleus and therefore readier nucleation. Now we can apply all
the mathematical arguments of classical nucleation theory to the
subgrain model; the only difference is that the nucleus is created
by dislocation rearrangement and not by random fluctuations of
atomic configuration in the nuclear regin. Therefore the observ-
ation that strongly misoriented subgrains form nuclei preferentially
amounts to asserting that nuclei perversely form under circum-
stances where their critical radius is unnecessarily large.
This is in fact a misconception. In classical nucleation theory, the
interfacial energy is constant and moreover an expanding nucleus
gains volume free energy which is proportional to the matrix volume
absorbed. In our present case the periphery of a growing subgrain
acquires an ever higher dislocation density and therefore a steadily
increasing specific energy. Instead of destroying all the dislocation
in the volume it absorbs, it sweeps most of them up (some will be
annihilated by reacting with other dislocations in the advancing sub-
5 7
m
\
Fig. 3.10
boundary). The main reason why a subgrain grows, then, must be Energy of an edge dislocation boundary as a function of
angle. (After W. T. READ)
jhat the swept-up dislocations will haye a smaller_tgtal energy when
captive in the sub-boundary which has just swept them yp_jhan they
hadejther as isolated dislocations or as constituents of the low-
sub-boundaries which have been consumed.
5 8
And indeed this inequality does apply to edge-dislocation boundaries
of progressively increasing misorientation angle. Read' s classical
theory of sub-boundary angle (clearly summarizedbyBYRNE [5J,
p. 110), shows that the energy of such a boundary increases more
slowly than the misorientation (Fig. 3.1O). Moreover, this theory
shows that the reduction in specific interf acial energy f or a given
increment A&o misorientation increases with the misorientation "&
Thus a moderately large-angle sub-boundary " digests" dislocations
more ef f iciently in terms of internal energy reduction than does a
small-angle boundary. 5n f act. tnis inequality is likely to be f ar
more influential in governing runaway growth of subgrains - i. e.
nucleation - than is the triple-point d isequilibriumin d icated in Fig. 3,
the atable triple-point conf iguration is in any case no longer a
simple 12Q-configuration once the three converging boundaries
have d if f eren t energies).
We have reached the f ollowing point: A highly misoriented subgrain,
whether large or small relative to its neighbours, will grow f reely
at the expense of its neighbours which are substantially misoriented
f r om the growing subgrain but only slightly f rom each other.
Eventually, however, the growing subgrain changes character.
The boundary angle approaches the valu $m(Fig. 3.10), at which
the individual dislocations on the boundary begin to lose their
identity, Somewhere n ear this point, the boundary must chan ge
f rom a sweeper-up of dislocations to a destryer of dislocations.
When this happens, of course, the internal energy disposed of per
un it volume of cold-worked matrix con sumed rises sharply, and
the growth of the nucleus may be expected to accelerate also. Here,
however, we en coun ter a parad ox. It is well established by prewar
workon recrystallization (rev ewed byBURGERS[l]) that large n ew
grains do not readily grow into matrix regions of orientations only
a f ew degrees d if f eren t f r om their own. This is shown, f or instance,
5 9
by the eurvival, in a large alumium grain f ormed by strain-anneal-
ing, of small island grains of the original cold-worked matrix; these
island grains all have orientations cise to (or in twin relation to)
that of the large grain in which they are embedded [17] . The
reason f or this reluctance to migrate have never been analyzed. '
We have theref ore the parad ox that small-angle sub-boundaries
consisting of identif iable edge dislocations migrate f reely, while
boundaries of rather larger angle, of more disordered structure,
do not unless the misorientation across them exceeds 10-15.
The use of the terms sub-boundary and boundary are intended to
distinguish between the two categories. We can now understand the
importance of long-range misorientations in governing nucleation
in cold-worked metis. The initial stages of nucleation, on the
subgrain model, involve highly localized misorientations and high-
ly mobile sub-boundaries growing f reely f or the reasons related
to Fig. 3.10. At some stage, not hitherto analyzed, the sub-
boundary -* boun d ary con versin happens, and then it helps con-
tinued growth of the nucleus if it can grow into matrix regions of
rapidly increasing misorientation.
It must be remembered that the migration process in d isord ered
large-angle boundaries (those above^ misorientation) must be
f undamentally dif f erent f rom the migration of sub-boundaries made
up of identif iable dislocations. The less d isord ered a large-an gle
boundary (i. e. the smaller its misorientation, subject to a lower
limit of " ^ ) the less easily can atoms rearran ge themselves as
migration proceeds. It must also be remembered that Fig. 3.10
only applies to the special case of edge-dislocation arrays, and
it is probable that a dislocation boun d ary consisting of several
types of dislocation would f in d it much harder to migrate even f or
small A vales. In f act, the notion of a tran sition f rom sub-
boundary behaviour to disordered boundary behaviour may only
be valid when the sub-boun d ary in question is a simple edge dis-
location boundary.
6 O
This is exactly what happens if new grains are nucleated at grain-
boundaries or deformation bands (i. e. artificial grain-boundaries).
Figure 3.11,for instance shows schematically what has been ob-
served to happen at the edge of a sharply misoriented deformation
band in cold-rolled silicon-iron. Quite large subgrains act as
"preformed nuclei", in the manner indicated and grow into neigh-
bouring subgrains of sharply different orientations. Soon these
nuclei grow large enough to abut as matrix regions so far mis-
oriented that they now have large-angle boundaries. Indeed, it
would be a simplifying convention to reserve the term "nucleus"
for subgrains which have become large enough so that their inter-
faces with the matrix have gone through_the_ sub-boundary *
boundary transition.
A number of investigators have observed the rapid growth of
highly misoriented subgrains into a cold-worked matrix. This
is most clearly seen in the electrn microscope. GRA NZER and
H AAEE [18J , both deformed and annealed aluminium foils in the
microscope and observed nucleations in the sense described in
this section. A particularly clear set of micrographs and associat-
ed electrn diffraction patterns was prepared by MICH ELS and
RICKETTS[19J , who worked with deformed aluminium and nickel
foils annealed in the microscope. Their study constitutes probably
the most convincing confirmation as yet of the subgrain-growth
mechanism of nucleation. MA RSDEN[20] has some micrographs
of individual subgrains growing at the expense of a number of
neighbouring subgrains in a rolled 2-phase A l-Mn-Si-Fe alloy,
to the point where the misorientation is about 15 and the nucleus
has become viable. This particular study was done with samples
deformed and annealed in the bulk and thinned afterwards, which
by-passes the objection that thin foils may behave anomalously
when annealed. The second-phase partiles hold up short segments
61
BAND DlRECTION -
A() bcundanei poro!!*! to th.j dirsti.O" nre 1,11
Fg. 3.11
Nucleation by subgrain growth on a deformation band in
rolled Fe-Si alloy. (BasedonJ. L. WA LTERandG. F. KOCH [21]).
6 2
of advancing sub-boundary; the resulting boundary cusps make it
possible to identify the direction in which the aub-boundary was
migrating.
63
c) Subgrain Coalescence - HU [21] has identified a variant of the
grain-growth model of nucleation, and termed it "subgrain coales-
cence". On the basis of careul observations on sicon-iron
crystals (similar to those by other investigators [21] on which
Fig. 3. 11 was based) he deduced that one subgrain can merge with
its neighbour not by sub-boundary motion but rather by a rotation
of one of the, subgrains till.its lattice is parallel to that of its neigh-
bour, More recently, HU [22] observed a similar process- in sub-
grains in heavily rolled copper. LI [23] has rationalized the pro-
cess by calculating energy balances: he showed for various con-
figurations that low-angle boundaries can have an energetic in-
centive to lower their misorientation angle and eventually dis-
appear to allow nearly higher- angle boundary to increase their
misorientations in consequence.
Presumably the later stages of growth of a nucleus produced by
this mechanism will be the same as for nuclei formed by subgrain
growth. It is not known what circumstances f avo ur subgrain
coalescence over subgrain growth. Both can oprate, it appears,
in the same metal; perhaps purity, degree and mode of deformation
or other variables exert a determining influence.
Figure 3, 12 reproduces two of a sequence of electrn micrographs
24 showing subgrain coalescence in 99. 99% aluminium def ormed
98% by rolling and annealed in situ in the microscope. The author
remarks that dislocations were seen to move f ro m right to left in
the subgrain- C, which eventually merges completely with D.
E
Fig. 3.12
Aluminium rolled 98%, thinned and heated in electrn microscope.
(a) As def ormed, (b) intermedate stage of annealing. At a later
stage the C/D boundary completely disappears. (After B. BAY [24 J ) .
6 4
d) Strain-induced Boundary Migration(SIBM) - This term was applied
byBECKandSERRY [7j to the phenomenon illustrated in Figs. 3.13
- 3.15. These micrographs, and also Figs. 3.3, 3.4, 3.5, 3.17,
3.18 and 3.19 form part of a detailed study of recrystallization of
aluminium by S. BELUERand R.D.DOHERTY [26]. In Fig. 3.13, the
original boundary between the white grain and the two darker grains
to ita right was ab. On annealing, a tongue (the uniform grey regin
in the centre ) grew out to the right. The orientation of this grey
regin (which may be termed the jmlge) is within a few degrees of
that of the white parent grain. This impliea that one of the sub-
grains in the white grain, immediately to the left of ab, has acted
as "nucleus" and spread into the differently oriented grains to the
right. ab is always a high-angle boundary where this mechanism
operates.
Sometimes (though this is unusual) the bulge has a uniform orient-
ation, as confirmed by examining an anodized surface n polarized
light, a technique sensitive to small variations in orientation.
Figure 3.13 is an instance of this. Figure 3.14 shows a more
common configuration: here the bulge comprises three coarse sub-
grains. Careful inspection will show that each one matches in tint
one of the minute subgrains in the "predator" grain with which it
comes in contact at the interface between the predator grain and
the bulge. This is consistent with the basic hypothesis that each
subgrain of the bulge is a continuation of a subgrain fnot necessari-
ly in the plae of section) of the parent grain. Figure 3.15 shows
a bulge with finer subdivisin than that of Fig. 3.14. It is not cer-
tain whether the seprate subgrains near the tip of the bulge, which
do not come in contact with the predator grain in the plae of
section, do so below or above the plae of section.
65
Fig. 3.13
Aluminium deformed 21% by
compression. annealed 11 min
at 395 C S. BELLIER and
R. D. DOHERTY).
Fig. 3.14
As Fig. 3.13.
(S. BELLIER and R. D. DOHERTY).
Fig. 3.15
As Fig. 3.13, but annealed
42 h. at 395C.
(S. BELLIER and R. D. DOHERTY).
6 6
Always the subgrains in the bulge are much coarser than those in
the predator grain. Figure 3.16, due to BECK [27], interprete the
phenomenon. He postlales that the predator grain has - at any
rate initially - a coarser aubgrain structure than the "victim" grain
which is to be consumed. Energy is then reduced when a small
volume of the victim grain is consumed, because the total rea o
sub-boundary destroyed in it is greater than the rea of sub-
boundary newly created in the bulge (particularly because the bulge
has an even coarser substructure than the predator grain had).
BAILEN and HRSCH [6] analyzed the energtica of this procesa, which
they also confirmed experimentally, in copper and silver - though
on a much finer geometrical scale than in Figa. 3. 13 - 3. 15. The
outcome of their calculations (summarized in re. 8) can be best
understood by reference to Fig. 3.17. I f I gA* is the stored
energy difference per unit volume (see chapter 2 sec. 1. ) between
the victim grain and the predator grain, a calculation on the same
lines as the classical Volmer-Becker calculation for a critical
nucleus radius shows that
must be positive for all vales of the angle & if the bulge is to grow
steadily. Properly speaking, in the light of micrographs such as
Figs. 3. 13 - 3. 15, ^g/y should be the difference in energy per
unit volume between the victim grain and the first part of the bulge
t ofor m, and this will normally somewhat exceed g/v as defined
above.
The inequality cited is equivalent to that defining the Volmer critical
nucleus, since here L must exceed a minimum valu if the bulge
is to grow. This means effectively that a particularly large sub-
grain abutting on to the grain boundary will cause that boundary to
67
(o) BEFOflE MIGRATION
(t > ) AFTEH S T ART OF
Fg. 3.16
Model for strain-induced boundary migration (After P. A. BECK).
GRAIN BOUNDARY
Fig. 3.17
N ucleation by strain-induced boundary migration (After BAI L EY and KI RSCH).
i;*
6 8
to bulge across the ent ire width of the subgrain. I n effect , t here-
fore, BAILEY and KI RS CH' a conditions reduces to the formulation
of a critical subgrain size in the grain-boundary regin of the pre-
dator grain. BAILEY and KIRSCH were able to roughly confirm their
criterion from their own work on heavily deformed silver, but no
attempt has been made to check it for the coarser scale of SIBM
in lightly deformed aluminium. SIBM has also been observed
during recrystallizatjon n deformed lead [31] .
BECK and SPEURY discovered SIBM in aluminium deformed by about
40%, and the micrographs published here apply to strains of ~ 20%.
The limited evideriee svalable suggests that SIBM declines in im-
portance at higher deformations.
The magnitude and inde'd the sign of 4g/_. can vary from one
part of a grain boundary to another, and it is presumably in such
situations that the configuraron of Fig. 3. 18 can arise, where
bulges grow in both directions along the same boundary.
4. ORIENTATION RELATIONSHIPS
As explained in the I ntroduction, one may hope to derive some in-
sight into the operating nucleation mechanisms in pri mary recryst alliz-
ation by measuring the orientation relationships between new grains
and matrix ( def ormed) grains in partially recrystallized specimens.
This has been the principal theme of work carried out by FERRAN,
BELLI ER and DOHERTY during the past few years in the author' g
laboratory. A detailed account of the first part of this work is in
process of publication [28] . It depends upon the following strategy:
The work is done with moderat ely def ormed al umi ni um of 99, 997%
purit y, which is pur enough to be part icularly susceptible to preerred
( i 9
Fig. 3.18
Al umi ni um deformed 50% by compression,
annealed 8 min at 330 C. (S. BELLIER and
R. D. DOHERTY).
nucleation al grain boundaries and especially grain edges (as first ob-
served in zone-refined aluminium by VAJDERMEER and CORDN [30]
in an importar paper) but not so pur that recrystallization can take
place at room temperatura. The aim then isto initiate new grains at
the periphery of fairly large matrix grains in sheets ~O, 06 mm thick,
and to determine the orientations of as many grains as possible. For
this purpose, the pseudo-Kossel method of X-ray diffraction was used,
as dse ribedin detall by FERRAN [29] . A different method - the micro-
Laue method - was used for considerably deformed grains which will
i ot give good Kossel patterns. Briefly, the sample is coated with a
thin layer of iron, and a microbeam of electrons focused on the desir-
ed spot. The resultant cone of FeK X-rays produces a transmission
diffraction pattern such as that shown in Fig. 3.19, and the orientation
of the grain responsible is readily deduced. It is possible to determine
the orientation of a regin a few hundredths of a millimetre in diameter.
The great advantage of the method is that a very large rea of sheet -
2
as much as 50 mm - can be examined in a single experiment, instead
of the few square microns in a thinned electron-microscope foil which
happen to be reduced to the appropriate thinness. The statistics of the
investigation are thorefore imcomparably more favourable, ard more-
over H'.s sheet is so thick that it can be regarded as bulk material.
Figure 3. 20 is an example of the type of regin that was examined in
detall, and indeetf the paper by FERRAN et al [28] represents an
examination and full statistical evaluation of the 63 grains nucleated
around the periphery of several large matrix grains. Of thesp 63 grains
5 were small and were recosrnizably the producs of SIBM, though not
necessarily very cise in orientation to the predator grain. The other
58 were larger and often had grown to both sides of the bcurdary on
which they liad been nucleated.
Fig. 3. 19
Transmission Kossel X-ray
diffraction oattern.
(S. BELLIER and G. FERRAN).
Fig. 3.20
Aluminium deformed 40% by
compression, annealed 38 min
at 330 C.
(S. BELLIER and R. D. DOHERTY).
7 2
Various forms of statistical analysis were applied to the new orientations
of the 63 new grains on the one hand and the orientation of the central
matrix grains and their immediate neighbours on the other hand. First-
ly, it turned out that the orientation correlation between the 63 new
grains and the central matrix grains, while not random was statistical-
ly weak. The interesting relationships only emerged when the cplonies
of new grains along the boundary between a single pair of matrix grains
(such as AB in Fig. 3. 20) were considered separately, and related to
the two contiguous matrix grains. It was found that the orientations
of an entire colony of new grains were always clustered (e.g. Fig. 3. 21)
and particularly oae set of (111) poles were always closely clustered
(usually within a small circle of radius r= 12 ), This small circle
sometimes contained a (111) pole of one of the two contiguous matrix
grains, but in any case the centre of the small circle was never more
than 20 from the nearest matrix (111) pole.
More recent workbyBELLIERandDOHEHry[26] has been extended to the
orientation of grains preferentially nucleated at deformation bands -
i. e. at "artificial grain boundaries". Figure 3, 22 shows such a
specimen:< the cluster of grains in the centre were nucleated at a
deformation band like that of Fig. 3. 5, but the band itself has been
swallowed up by the new grains. The plae of the band was nearly
parallel to the picture plae. Some 30 orientations of both new grains
and different regions of the matrix were obtained. The (111) poles of
both new grains and matrix points fell into two tight clusters on the
stereograms, each cluster containing some matrix and some new
grain poles. The centroids of the 2 families of orientations were
mutually rotated by about 20, Figure 3. 23 shows the outline of the
pole clusters, without details of individual poles.
; . 3. 21
1) poles of two deformed
rix grains either side of
oundary, and of a colony
5 recrystallized grains
tleated at this boundarv-
Ser G. FERRAN, R. D. er-
flERTY and R. W. CAHN
Fig. 3.23
(111) poles of matrix and re-
crystallized grains of Fig. 3. 22
in stereographic projection
(Aorientation of the <111>
poles of the new grains,
O + orientation of the<lll>
poles of the matrix).
Fig. 3.22
Al umi ni um deformed 40% by
compression, annealed 15 min
at 330C. A deformation band
is almost in the plae of the
mi'.-rograph. (S. BELIJER and
R. D. DOHERTY) .
7 4
The implications of these two sets of studies are clear, and can be
summarized as follows:
a) New grains nucleated at a single normal matrix grain boundary or
at a single matrix deformation band have closely clustered
orientations,
b) These orientations are often cise to the orientation of one of the
contiguous grains or one of the regions either side of the deform-
ation band, but may also be substantially rotated from the nearest
such matrix orientations.
c) This last fact is to be related to the observation (Fig. 3. 23) that
deformed grains have substantial mis orientations, sometimes
over quite small distances.
The general conclusin is that each nucleus has orientation of a matrix
subgrain adjacent to a genuine or artificial grain boundary, but that
this orientation may be substantially rotated relative to nearby regions
of the matrix, because of the large orientation gradients present,
especially in the neighbourhood of matrix boundaries.
This general conclusin is compatible with any of the nucleation
mechanisms described in the preceding section. However, for nuclei
formed at boundaries, subgrain growth and SIBM become indistinguish-
able; they are to all intents and purposes the same mechanism. We
cannot tell whether the individual subgrains reach the status of nuclei
by (a) simple subgrain growth, (b) subgrain coalescence, (c) random
nucleation. It is interesting, however, to note that the nucleation at
deformation bands provides such a very cise relationship between new
and od grains that it seems extremely improbable for it to have come
about by random nucleation.
7 5
As explained in the Introduction, it is very risky to draw conclusione
from relatonships between deformation textures and annealing textures.
Nevertheless, such analyses have become very sophisticated, especial-
ly in the handa of LCKEandhis clleagues. In their most recent work
[33] on textures of hexagonal zirconiunt and titanium (easier to analyze
than cubic metis) they reach the positive conclusin from their ana-
lysis that "the number of available nuclei is the smaller, the larger the
misorientation between the nucleus and the_matrix". The annealing
texture is formed primarily by selective competitive growth - i. e. com-
petition between the nuclei - but because of the non- randomness of
nuclei, the annealing texture is not ideal; instead, a "compromise
texture" is formed. These ideas are entirely cqnsistent with the new
,r
observations reported here.
SUMMARY
The principal experimental observations that can cast light on the
mechanism of nucleation in primary recrysfllization are reviewed.
Special emphasis is placed on the role of both localized and macroscopic
misorientations in the deformed grains, and in particular the role of
preferential nucleation at grain boundaries and deformation bands.
The possible nucleation mechanisms - classical nucleation, subgrain
growth, subgrain coalescence, and strain-inducedboundary migration -
are then considered. The progressively changing structure and speci-
fic energy of the boundary of a growing subgrain is discussed in some
detail, and it is proposed that a subgrain becomes an effective nucleus
when its boundary undergoes a transition in structure and properties.
Micrographic evidence based on recent work by BELLIER and DOHERTY
is presented, especially in connection with strain-induced boundary
migration.
7 G
In conclusin, hitherto unpublished evidence is presented concerning
the orientations of new grains nucleated at grain boundaries or de-
formation bands in aluminium, and the implications for the nucleation
mechanism probably involved are considered.
ACKNOWLEDGEMENTS
I am grateful to Drs. S. BELLIER, R. D. DOHERTY, G. L. FERRAN and
S. MAHAJAN for permitting me to use information and illustrations from
their work prior to publication, and for stimulating conversations.
REFERENCES
1. W. G. Burgers:
2. R.W. Cahn:
3 . P. A. Beck:
4. J. E.Burkeand
D. Turnbull:
5. J. G. Byrne:
6. J. E. Bailey and
P. B. Kirsch;
7. P.A.Beck and
P. R. Sperry:
8. R.W. Cahn in:
9. R.W. Cahn in:
10. A.Kimura,
R. Maddin and
H. Kimura:
11. H. Hui n :
12. W. C. Leslie,
J.T.Michalak and
F. W. Aul in:
13. P.R.Mouldand
P. Cotterill:
7 7
"Rekristallisation, Verformter Zustand und
Erholung", Akademische Verlagsgesellschaft,
Leipzig 1941.
See also: W. G. Burgers in: "The Art and
Science of Growing Crystals", J. J. Gilmann,
Ed., John Wiley and Sons, New York 1963,
p. 416.
Proc. Phys.Soc. (London) 63A (1950) 323.
J.Appl. Phys. ^0 (1948) 633.
Prog. Met.Phys. 3 (1952)220.
"Recovery, RecrystalliEation and Grain Growth",
Macmillan, London 1965, p. 71.
Proc. Roy. Soc. 267A (1962)11.
J.Appl. Phys. 21 (1950)150.
"Recrystallization, Grain Growth and Textures",
H. Margolin, Ed., ASM, Metals Park Ohio
1966, p. 99.
"Physical Metallurgy", R. W. Cahn, Ed., North-
Holland, Amsterdam 2nd edition, 1970, p.1129.
ActaMet. 12 (1964) 1167.
"Recovery and Recrystallization of Metals",
L. Himmel, Ed., Interscience, New York 1963,
p. 311.
"Iron and its Dilute Solid Solutions", C. W.
Spencer and F. E.Werner, Eds., Inter-
science, New York 1963, p. 119.
J. Mater. Sci. 2 (1967)241.
14. R.D.Doherty and
J. W. Martin:
15. R.W. Cahn,
I. J. Bear and
R. L. Bell:
16. A.H. Cottrell;
17. P. Lacombe and
A. Berghezan:
T. J.Tiedema,
W. May and
W.G.Burgers:
18. F.Granzerand
G. Haase:
19. L. C. Michels and
B.G.Ricketts:
20. P.K. Marsden:
21. J. L. Walter and
U. r. iv.ut.li:
22. H. Hu in:
23. J, C.M. Li:
24. B. Bay;
25. K. Holm and
E. Hornbogen:
26. S. Bellier and
R.D.Doherty:
27. P. A.Beck:
28. G.L. Ferran,
R. D. Dohert y and
R. W. Cahn:
29. G. L. Ferran:
7 8
J.Inst. Metals 91 (1962) 332.
J.Inst, Metals 82 (1954)481.
Prog.Met. Phys, 4 (1953)255.
Compt. Rend. Acad. Sci., (Pars) 228 (1949)
93;
Acta Cryst. 2 (1951) 151.
Z.Phys. 162 (1961) 504,517.
Trans. Met. Soc., AIME 239 (1967) 1841.
J. Mater. Sci. 6_ (1971) in the press.
Acta Met. ^(1963) 923.
"Textures in Research and Practice",
J. Grewen and G. Wassermann, Ed.,
Springer, Berlin 1969, p. 200.
J.Appl. Phys. 33 (1962) 2958.
J. Mater. Sci. 5 (1970) 617.
J. Mater. Sci. 5 (1970)655.
to be published
Advances in Physics 3 (1954) 245.
submitted to Acta Met.
Metallography 3 (1970) 441.
7 9
30. R. A. Vandermeer and Trans. AIME 2^5 (1959)577.
P. Cordn:
31. P. Tardy and
S. S. Iskander:
J. Mater. Sci. 4 (1969) 353.
32. T. H. Schofield and Acta Met. 9 (1961) 653.
A. E. Bacon:
33. K. LUcke and
R. Rixen:
Metallurgical Trans. 1 (1970) 259;
discussion, ibid, 1 (1970) 2342.
8 1
HAPTER 4
I GRATI ON OF HIGH ANGLE GRAIN BOUNDARIES
Y
HOFMANN AND F. HAESSNER
INTRODUCTION
One of the e l e me n t a r y processes of recrystallization is the move-
ment of high angle grain boundaries. A description of thls process
in general is very complicated (e. g. , with respect to topology). Its
main features, however, may be discussed quite reasonably in trras
of a migration rate. The discussion given here will be mainly confin-
ed to the grain boundary migration rate in pur metis; the influence
of dissolved for eign atoms and of second phase particles is discussed
elsewhere in this volume (see chapters 7 and 8).
The migration rate of a high angle boundary will be first derived with
the help of a simple model. Theoretical predictions obtained in this
way are then compared with the experimental results of various authors
and are finally discussed critically.
SIMPLE SINGLE PROCESS MODEL
Many experiments have confirmed that the grain boundary migration
rate increases exponentially with temper atur a. This clearly indicates
that the migration of a high angle boundary is a thermally activated
process. It seems, ther efor e, reasonable to derive an expression
for the boundary migration rate based on absolute reaction rate theory
[l] . Figure 4.1 gives an elementary conception of a grain boundary
in an energy-distance diagram.
8 2
Gibbs free energy
per gm-Atom
Grain Matrix
Distance
Fig. 4.1
Energy barrier for the j u m p o an atom across the boundary
according to the simple single process model.
The grain boundary separates two lattice regions of different orient-
ations. One of them is t ermed here "matrix" and has a Gibbs free
energy (per gram-at om) greater than the other by an mount lg. I
and u indcate possible atomic positions at the interface between the
two lattices under consideration. A movement of the boundary, for
example, in the direction of the matrix would occur when more
atoms j u m p from u to I in unit time (j ump frequ ency /"", ) than in the
opposite direction (Ju mP frequ ency r.). Let us assume that during
t his movement all p aramet ers determining the bou ndary remain con-
stant with the exception of its position on the x - axis. Then, if the
8 3
atomic distance (denoted by a ) is the same in both lattices,
ty can be expressed as:
v = - a - ( TJ - r 2 ) .
In case the atoms j u m p independently f r om each other and possess the
same vibrational frequency l ? in both lattices, then
r
;n , T
TV' 6XP
UI |_
(2)
(Here Q is a geometrical factor of the order of 1; R the universal gas
constant; T, the absolute temperature; 3g ,,, the Gibbs f r ee energy
of activation per gram-atom for the j u mp of one atom f r om n to I).
As/
Since for pur metis during recrystallization & RT<<C1 (see
chapter 2), the migration rate can be expressed as:
exp ( -
RT
(3)+)
a -
If B/_ (V: mole volu me) is taken t obe the driving forc which is act-
ing on the grain boundary (cf . chapter 2), then the expression
RT
( . _ ) v
1 RT ' V
(4)
has the character of a mobility, which represents a property ot the
boundary itself.
Formally the migration rate (eq. 3 ) can also be expressed t
Dm A
v = ~RT ~a~ '
(5)
where
D exp ( - - ^) f6 )
A relation of this form was first given by D. TmMULL [2]
84
representa an appropriate difusion coefficient characteristic o the
raigration mechaniam. It has been suggested that this may be the
diffusion coefficient for grain boundary self diffusion [2, 3.] . It
should, however, be kept in mind that grain boundary difuaion takes
place through atomic j uni ps along the boundary, whereas grain
boundary motion occurs due to atomic j umps across the boundary.
Thus, these two diffusion coefficients can have quite different
vales. The simple single-process model described hete gives a
linear dependence of the grain boundary migration rate on the
driving forc (see also section 4.1.). Since it has been assumed
that all the atoms at the grain boundary have the same j ump
probabilities and always find a free place on the adjoining grain
surface, the migration rate obtained here represents the mximum
possible valu (see also section 4. 3.}. The model does not provide
any evidence regarding the dependence of the migration rate on the
degree of misorientation of the two lattice regions. In order to
explain the anisotropic nature of the migration rate, which has been
repeatedly confirmed experimentally, the model has to be improved
taking into account structural details of the boundary.
8 5
3, IMPROVED SINGLE PROCESS MODEL
This model differs from the one described in section 2 essentially
in the following aspects:
1. The grain boundary is assumed to be a layer with a thickness of a
few atomic distances.
2. The atoms at the interface between each of the two lattices and the
boundary possess different j ump probabilities.
3. An atom can make a j ump only when there is a vacant site in its
neighbourhood.
Here the boundary is assumed to be an infinitely extended thin planar
regin lying perpendicular to the x - axis, and its thickness S is
assumed to remain constant with time. Let the elementary process
for the boundary migration be the transfer of single atoms from one
lattice (matrix) to the other (grain) across the boundary regin. This
process can be considered to becomposed of three partial fluxes jT ,
\
jR, jn (see Fig. 4. 2). In the steady state condtion it follows from
the equation of continuity that these fluxes are of equal magnitude and
proportional to the migration rate v:
II
i
B
(7)
As a generalization of Fig. 4.1, the Gibbs free energy of an atom may
vary as a function of its position as shown in Fig. 4. 2,
'in this regard we follow a theoretical treatment given by H.GLEITER
[*]-
For simplification, the atomic distance (denoted by a) has been
assumed to be the same in the grain, matrix and the boundary.
F
.
<
|
u
i
*
, _ . j . -
i '
8 6
Gibbs free enero/
per gm-Atom
Grain Matrix
IAMT
1 rfe
1 ! 9
i 1 , .
J * X Bl B fW U * *^
i -
I JB "1
D/s/once x
Fi g. 4. 2
Schemati c di agram showi ng Gi bbs free energy of an atom i n the
grai n bou ndary regi n accordi ng to the i mproved si ngle process
model.
Here i t i s vi su ali zed that the emi ssi on of an atom from one of the two
crystal latti ces i nto the bou ndary regi n i s consi dered as a si ngle
thermally acti vated process. The transfer reacti on throu gh the bou n-
dary i s assu med to take place by a self- di f fu si n mechani sm (di ffu si on
coeffi ci ent = DR). In Fi g. 4. 2, z_ and z,, roay be the nu mber of
8 7
o
atoms per cm at the su rfaces of grai n and matri x, respecti vely, whi ch
2
are capable of j u mpi ng. z_ . and z_ . , the nu mber of atoms per cm
DL D 1 1
at each si de of the bou ndary, and c_ . c,,, cn the nu mber, respecti vely,
2
of vacant si tes per cm i n grai n, matri xand bou ndary. Then based on
absolu te reacti on rate theory the parti al flu xes can be expressed as:
(5 A* Aa
n &T1 *) K ~ T tL
*I B BI I ^^ RT ' I I B *^ RT ' \^'
o a ^n B ^RT ; a "B BU - n expt RT ' f8fa)
DB z Bn- z Bi
B- a S ' _ (8c'
The di ffu si on coeffi ci ent for the transport of atoms tktoti gh the bou ndary
( D_ ) can be esti mated by
4 ^B
D r t Kl
whi ch i s analogou s to the expressi on for latti ce self- di ffu si on. Recall-
i ng that 4g RT, the mi grati on rate can be expressed, after combi n-
i ng eqs. 7, 8a - c and 9 and neglecti ng the di fferences i nthe i ndi vi du al
vi brati onal frequ enci es ( l > T =^TT =1 ^^ =V)> as
6 . m2 / A e_ i 1 z_ z, z,
a ' \ CR . exp - ( &n ) . u I I Jg
6
RT c t* ' f Wf4 1
Kl |Cn Cj Cj RTJ
T (1 0)
1 + a - c ( l + l )
8 B ( ci cn
Si nce a dri vi ng forc i s always necessary i n order to keep a grai n bou n-
dary steadi ly movi ng, v mu st be equ al to zero i f 4g =0, i . e.
I _ 1 . The mi grati on rate follows from thi s as
Cn ci
RT
. exp ( -
Bn
RT
v = (11
This relation simplifies for the following two limiting cases:
a) The rate controlling process may be the transport of atoms through
the boundary. This would mean that
) 1 which gives
(1 2 '
a
S
' CB
' a6,
f 1 + l
* P- - C Cj Cj.
..<>.
)

3 g
(a)
5- RT
exp
CI RT
b) The rate controlling process may be the emission of the atoms frorc
the matrix and their incorporation in the grain. This corresponda
to the condition:
o
(!
' B (CT
so that
exp (-
'n
RT
(13
4. DISCUSSION AND COMPARISON OF THE THEORY
WITH EXPERIMENTS
Even for a qualitative comparison of experimental observations w
the theory of grain boundary migration, as developed above, one hast
assume that the driving forc ( 3 g/V) and the structural parameters
of the boundary ( c_. c_ , c z _ ) in eq. 1 1 either remain constante
B 11 I 1
change slowly in a known way with time. In the few experimental case-
__ ^
For 8 = a and c_ = a" ( mximum valu) eq. 1 2 corresponda te
eq. 3 of the simple model in section 2 .
8 9
in which such an assumption is sufficiently well justified, the purity of
the metal is of critical importance and determines the applicability
of the model: A metal can be taken as "pur" in the sense of the
theory only, when the dragging forc on the moving boundary due to
the foreign atoms present is negligibly small compared to the driving
forc (c.f. chapter 7). In the case of foreign atoms in solid solution,
the magnitude of the dragging forc does not only depend upon the
concentration of the foreign atoms but also on the velocity of the
boundary. This means that the same metal, which can be regarded
as "pur" with respect to theory for a large driving forc and
corresponding high migration rate, may behave as "impure" in the
case of smaller driving forc and lower migration rate.
Information on grain boundary migration rate can be given either by
the observation of an individual boundary (for example in a bicrystal)
or of an assembly of many boundaries (for example in polycrystalline
material). The latter case can only provide average vales which
are difficult to interpret within the model. Investigations on a single
boundary, on the other hand, fulfill the assumptions of the theory to
a greater extent. As an example, the light micrograph in Fig. 4. 3
shows the movement of a grain boundary in a deformed single crystal
gold matrix.
Fig. 4. 3
Light micrograph showing successive positions
of a grain boundary during its migration in a
def ormed single-crystalline gold matrix.
9 1
The successive positions of the boundary which lies perpendicular to
the plae of the f igure obtained af ter various annealing times show a
parallel shift. In this case the condition that the boundary does not
change its orientation during the movement, as required by the model,
is very nearly fulfilled.
4.1. Relationship between Grain Boundary Migration Rate and Driving Forc
During normal grain growth of polycrystalline metis a test of the
relationship between v and 4g is possible through the time dependence
of the average grain diameter D: Since the driving forc in this case
should be inversely proportional to D , the relationship between v
and g is implicitly given by the time law D = D ( t) at constant
temperature. ( f or example, vOC/lg is valid for D OC \ f i - ~ [5] .
For the case of primary growth of grains in a def ormed matrix the
driving f orc depends on prior def ormation conditions and must
theref ore be determined by an independen! experiment ( f or example,
by means of a calorimetric measurement of the stored energy) [6, 7] .
Experiments of both types have often conf irmed a proportionality be-
tween driving f orc and migration rate as required by the theory
[7 - 10] .
Some experimental results, however, have shown a deviation f rom
jj
proportionality, which is of the f orm voc^g with k > 1. This
deviation has usually been explained in terms of the ef f ect of dissolv-
ed f oreign atoms [ll - 14 J , since they may cause a rate dependent
dragging f orc which, in the transition range of the Impurty-Drag-
Theory, leads to a non-linear relationship between v andg ( c.f .
chapter 7 section 2) . This explanation may be true f or particular
investigations; it however, appears to be no longer applicabJe when
9 2
in a certain temperature range a constant and low activation energy
(of the order of the activation energy for grain boundary dif f usion) is
obtained.
One way to explain the occurrence of a complex relationship between
the quantities v and Ig in a pur metal is in principie possible, if
sorae of the structural parameters in eq. 13 (e. g., c., and c.. )
are themselves dependent on velocity or driving f orc. An influence
of this type cannot be neglected in those cases in which the driving
forc is linked with the deformation structure of the matrix (e. g.,
migration of grain boundaries into a plastically def ormed matrix) or
with the curvature of the boundary (e. g., movement of the boundary
during grain coarsening). Thus, a definite driving f orc-migration
rate relation according to the theoretical predictions can only then
be expected when a fairly good separation of driving f orc and the
structural parameters of the grain boundary is experimentally
established. This condition can in principie be f ulf illed when, for
example, a grain boundary is made to move under the influence of
the difference in the magnetic energies of the two adjoining grains
( c. f . chapter 2 section 2. 4.). As MULLINS[15]has shown, this is
possible through annealing of a magnetically anisotropic metal in
high magnetic fields. However, such an experiment has not yet
been perf ormed with sufficient accuracy since other disturbing forces
like thermal grooving (c. F. chapters 2 section 3. 4. and 5 section 5)
have a strong inf luence due to the small magnitude of 4g.
4. 2. Temperature Dependence of Grain Boundary Movement
Regarding the inf luence of annealing temperature, all experiments
on grain boundary migration agree that this dependence can be ex-
pressed in the form of a Arrhenius relation :
v =
9 3
exp
Q
(14)
The valu of the activation energy Q is found to decrease as the
purity of the metal under investigation increases [7, 76] . For
highly pur metis Q is very nearly equal to half of the activation
energy for lattice self diffusion, i. e. of the order of the activation
energy for grain boundary self diffusion [16, 17] . Since at pre-
sent a calculation of the activation energy is not possible, an analy-
sis is restricted to a comparison of experimental vales with other
diffusion data [18 - 20] . However a reliable comparison of this
sort is very difficult to make since migration rate meagurements
generally have been perf ormed in small temperature intervals and
moreover on metis with dif f erent amounts of impurities.
4.3. Pre - Exponential Factor of Migration Rate
In contrast to the activation energy, the terms in eq. 11 constitu -
ing the pre-exponential factor give at least an upper limiting valu
which allows a quantitative comparison between theory and experiment.
With Ig = h_ - T as the pre-exponential f actor follows
f rom eq. 13 as:
v =
o
,6 u .2 Z T , sn ,
ci R
, a , 1 1 i
<5 BCT CT T
dg
" RT
(15)
A mximum valu for v is achieved when the boundary is only one
atomic distance thick ( 6= a), when each atom is capable of jumping
In case Q is found to be temperature dependent this behaviour
may be explained by an impuritv ef f ect according to the transition
regin of the I mpurity-Drag-Theory (c. f. chapter 7 section 3. 3.).
9 5
9 4
-2,
T i'l
and al ways fi nds a vacant posi ti on (z_ _ = c_ = a ) and when
the vi brati onal frequency i s atthe mxi mum (u = Debye l i mi ti ng
frequency). For the acti vati on entropy A s_ an upper l i mi t can be
taken, whi ch has a magni tude equal to that of the entropy of mel ti ng,
namel y a val u nearl y equal to 2 R i n the case of meti s. It
fol l ows from thi s:
_ (max) _ 9 , . ,, . ^g
vo . ~t D RT '
A compari son wi th experi mental data CT abl e 1) shows that for hi gh
dri vi ng forces v ;s> v ' max' hol ds. In the case of l ow dri vi ng forces
the experi mental val es of the pre-exponenti al factor are qui te cise
to the mxi mum val u whi ch i s theoreti cal l y possi bl e. Wi thi n the
frame of the theoreti cal model these resul ts are rather i ncomprehens
i bl e si nce the experi mental val es of v shoul d be al ways much
smal l er than v ^
o
In vi ew of the l ow acti vati on energy obtai ned i n the experi ments, i t
i s unl i kel y that thi s di screpancy i s due to the effect of di ssol ved
forei gn atoms. It rather appears to be a cri ti cal poi nt of the model
based on uncorrel ated si ngl e jumps. Pre-exponenti al factors hi gher
than that computed from eq. 16 may be expected if the el ementary
process for grai n boundary mi grati on i s based on a correl ated move-
ment of atoms as snown i n some recent studi es [29, 30].
4.4. Vacancy Concentrati on i n the Boundary
T he effect of "vacancy concentrati on" i n a boundary on i ts mobi l i ty
has been repeatedl y questi oned [31, 32] . Duri ng pri mary re-
crystal l i zati on of hi gh puri ty al umi ni um by i sothermal anneal i ng,
VArCEHMEIR [33] observed, for exampl e, a mi grati on rate whi ch de
creased strongl y wi th ti me. A transi ti on from an i ni ti al l y rapi d to
>
(16
:es
e
.tB
er
e-
ty
h
)
^^H
.
1 <~>
t.*
ht.
3
c
3
S
^3
T di
M *
c a
flj Q,
5
E? cti
? 3
fc. i-
1!
I
eS
U e
o
* .
c
H ^
C
o
r
n
p
:
B
o
u
n
d
s
K
7! 1
H
W 1 J
o Lia
3 MH . .
i ! si
S
Jc "S""" o
3 >
1 r-1
0> W
l s i
9"! " rt o
H > >
~ T i i
2 ""
t > * ?
E* bf> ,5
cj t rt
<W w
_ t J
CJ
u
t-
o
u .o
> Ifl
IH '
O ' '
T
y
p
e

o
f

S
a
m
p
l
e
a
n
d

E
x
p
e
r
i
m
e
n
t
"S
t-
i i
o
".
f
to
o
l-H
N
o
i r >
c-
o
^
^"
i
.

r
e
c
r
y
s
t
.
P
o
l
y
c
r
y
s
t
.

;
p
r
i
m
i i
M
1J
CM
O
to
es
m
O
co
1-1

co
o
^"
i
.

r
e
c
r
y
s
t
.
P
o
l
y
c
r
y
s
t
.

;
p
r
i
m
-
i / r1
CM
1 1
l-H
N
(O
*"
o
co

o
J,
"

.5
s
t
r
i
a
t
i
o
n
s
u
b
s
t
r
u
c
S
i
n
g
l
e

b
o
u
n
d
a
r
y
;
i <
<5
l-H
1 P
n
o
7
^-t
*~*

o
1-1
00
^"
o
CO
o
;
TU"
ts
^
o
cu
ti
E
n.
S
i
n
g
l
e
b
o
u
n
d
a
r
y
;
<
T
^T
i t
o
1-1
o
< n
i ft
o
i-H
m
m"
Oi
t-1
o
o

i l
.

r
e
c
r
y
s
t
.
P
o
l
y
c
r
y
s
t
.

;
p
r
i
m
<
i i
so
i i
n
o
1-4
T P
10
0
o
fH
to
...
I-l
o
co

o
SO
11
.

r
e
c
r
y
s
t
.
P
o
l
y
c
r
y
s
t
.

;
p
r
i
m
r
'oP
1 1
T-1
O
ei
"
c-
fO
-
o
1-1
-T
JS
D
L.
bn
c
P
o
l
y
c
r
y
s
t
.

;

g
r
a
i
s
-]
'T 1
M
eS
O
^H
1-^
.
to
"o
*7
T
(U
3
s
t
r
i
a
t
i
o
n
s
u
b
s
t
r
u
c
t
l
S
i
n
g
l
e

b
o
u
n
d
a
r
y
,
(
"
S
p
e
c
i
a
l
"
)

t-
i i
to
o
co
!-t
- -
^s
r
co
O
l-H
V
0>
3
s
t
r
i
a
t
i
o
n
s
u
b
s
t
r
u
c
t
l
S
i
n
g
l
e

b
o
u
n
d
a
r
y
;
(
"
r
a
n
d
o
m
"
)
A
co
CM
O
o"
V
o"

o
to
" o
r -t
T f
i l
i
n

g
r
o
w
t
h
t-
hc
13
a
f-
u
"o
OH
A
i i
co
s
o"
*
cT
-
o
to"
' o
1-1
i n
II
"~
L
n

g
r
o
w
t
h
bo
13
t
>,
~
u
1 1
o
OH
*.
rc~'
w
to
o"
i n
o"
i-t
to"

o
iH
i n
II
"
L
n

g
r
o
w
t
h
P
o
l
y
c
r
y
s
t
a
l
;

g
r
a
i
c.
w
9 6
a lower migration rate was also shown in the investigations of FELLER-
KNIEPMFJER and SCHWARTZKOPFF[32]jn the growth of nuclei in high
p urity copper. In a study of p rimary recry stallization of copper
NfiCHEERG, NEUflER and ltXE [34] were able to show that the
recrystallization temp erature was strongly reduced when the metal
was deformed at liquid nitrogen temperature instead of at room
temperature. All these observations can be explained qualitatively
through an increased grain boundary mobility at increased vacancy
concentration of the matrix or its decrease due to a reduction in
vacancy concentration during the recrystallization procesa.
Qualitatively the present model can give an explanation of these effects
in terms of the vacancy concentration C-, in the boundary (eq. 11).
t
The basic assumption here is that the excess vacancies in the deform-
ed matrix flow into the boundary during its movement and in this way
lead to an ulerease in c_ . However any attempt for a quantitative
13
interpretation encounters the following two diffcuKies:
a) Initially c_ must be relatively small so that it can be substantial-
D
ly increased through the additional vacancies introduced into the
grain boundary. This assumption, however, leads to extremely
low absolute vales of the boundary velocity (sec. 4. 3.).
b) For a constant migration rate. c_ must also be constant; i. e..
D
the number of vacancies flowing-in must be compensated through
a special annihilation mechanism [3lJ . Such a mechanism be-
comes very complicated and is rather unlikely when,as for example
in the case of primary grain growth [32] , the boundary surface
increase continuously for a constant migration rate.
The dificulty discussed under b) can be overeme if it is assumed that
c_ is unambiguously determined by the structure of the boundary and
changes only weakly through the influx of vacancies. The effect of
9 7
excess vacancies in the matrix is then only to influence the factors
related to the structure of the matrix ( c or z^ ; eq. 11). The rate
controlling process in that case is the emission of atoms from the
matrix [32] ( eq. 13). An indication that the migration rate is in-
fluenced only through a change in c_ without any change in c_ is
II H
given in an experiment in which the migration rate was measured in
deformed neutron-irradiated copper single crystals [35] . Figure 4. 4
shows the result: v increases with the neutrn dose, the activation
energy remains constant.
r [-c]
_ 2JO 200 190 180 !70 160150
0,01
(&)c=Q
( V <
2,0 2.05 2.1 2.J 5 2,2 2.25 2,3 2.35
Fig. 4.4
Temperature dependence of grain boundary migration rate in copper
for different neutrn irradiation doses.
(After F. HAESSNER and P. HOLZER [35] )
9 8
As the irradiaion of copper with fast neutrons gives rise to the
formation of dislocation loops of interstitial as well as of vacaney
type [36] , an increase in v is difficult to understand on the basis o
a vacaney influx model. The result in principie can be interpreted
in t erms of an increase in the number of lattice defects (and henee
of cn ) in the matrix with increasing neutrn dose, if the limiting
case of eq. 13 is assumed [37] .
4. 5 Orientation Dependence of the Migration Rate
The influence of orientation of a grain boundary on its migration
rate can be approached f rom two points of view: phenomenological-
geometric and structural-atomistic. Both these considerations
should be coordinated with one another in order to have a complete
model of a grain boundary. This is not possible at present. The
problem is complicated owing to the fact that five paramet ers alone
are needed to define the geometry of the system: The misorientatian
between the two adjoining grains can be given in t erms of an axis of
rotation (2 parameters) and an angle of rotation about this axis
(1 parameter), the position of the boundary itself can be expressed
by the position of its normal with respect to one of the two grains
(2 parameters). In terms of these coordinates a boundary is called
a pur tilt or a pur twist boundary, respectively, if the axis of
rotation lies in the boundary plae or perpendicular to it. In an
atomistic picture, an additional translation in the plae of the boun-
dary may occur, in specific cases, which however is not macros-
copically defined. In general, only the geometric parameters of the
boundary under investigation can be experimentally determined.
A relation between grain boundary migration rate and orientation car
be directly obtained through bi-crystal experiments (at a constant
driving forc) [21, 38, 39] . In addition, experiments of growth
9 9
selection [40] allow the evaluation of orientation relationships of the
grains of mximum growth rate with the deformed crystalline matrix
into which they are growing. By both types of experiments an aniso-
tropy o the migration rate has been confirmed. The highest migration
rates have been obtained for the following grain-matrix orientation
relationships:
i nf . c. c. metis: 30-40 rotation about <111>- axis [21,38,41]
inb.c.c. metis: 25-30 rotation about <110>- axis [40]
i nh. c. p. metis: 30 rotation about<0001> - axis [40, 42].
These relationships hold in the case of tilt boundaries; twist boundaries
generally have substantially lower migration rates. The general trend
of the velocity with orientation in case of a<lll> rotation axis is
shown in an cxperiment of LIEBMANN, LCKE and MASING[38](Fig.
4. 5), in which the migration rates of grains having different orient-
ations with a deformed aluminium matrix were determined.
A unique explanation of these results is difficult because at least in
the case of low driving forces there exists a strong influence of small
impurity contents: Thus AUST and RUTTER determined a relationship
between orientation dependence and the impurity effects through their
investigation on weakly doped leadbi-crystals[39]. Their results
are shown in Fig. 4. 6 (c. f. chapter 7) in which the migration rates
have been plotted as a function of the Sn-content for various types
of boundaries which are characterized by the orientation relationships
between the matrix and the grain. As shown in Fig. 4.6 there are
certain boundaries which move faster than others and whose migration
rate depends only weakly on the Sn-content (special boundaries). These
boundaries differ from others in the respect that they seprate ad-
joining grains with a misorientation of either 23 or 40 rotation
about a common<lll> axis or 28 rotation about a<100>- axis. The
1 00
g 2
?0 40 60
OIEMAWJ Dl f FT. RENCE Idgsl flBOUT COMVON <"!>
Fig. 4. 5
Orientation dependence
of grain bound ary
migration rate in al u-
minium. The mis-
orientation is given by
the rotation angle aboul
a c ommon<l l l > -axis
of the adjoining grains
(Af ter B. LI EBMANN,
K. L CKE , G.MAS I N G
[38])-
Fig. 4. 6
Variation of the migrat
velocity as a f unc tione
the concentration o tic
in lead at 300C, for
"rand om" and f or "spe:
grain boundaries. (Af l f :
K. T. AUST and J . W . R t
[2 5] ).
1 O 1
activation energy of "special boundaries" is independent of the Sn-
c onc entration whereas it increases strongly f or "random boundaries"
as shown by AUST and R UTTER . On the basis of the I mpurity-Drag-
Theory (c. f . chapter 7), this observation can be explained in terms
of the interaction of f oreign atoms which is smaller with the boundaries
having special orientations than with those having random orientations.
The anisotropy of the migration rate can be, in principie, interpreted
in terms of an orientation dependence of the driving f orc or of the
mobility. An anisotropy of the driving f orc is, however, out of
question at least in the case of highly def ormed material s because
of a complex dislocation structure [2 1] such that, in any case, an
orientation dependence of the mobility is to be expected. (For possible
interrelations between driving f orc and mobility see sec. 4.1.).
The orientation dependence of the grain boundary mobility can only be
explained in terms of a model which takes into account the dependence
of the grain boundary structure on the misorientation between the two
lattices and also accounts f or the specific orientation relationships in
case of "special boundaries". The f irst attempt in this direction c arne
f rom KR ONBER G and W ILSON[43]who gave a grain boundary model based
on l attice coincidences, whic h was later improved by Brandon et al.
[44J .BISHOP and CHALMER S [45] have proposed a model for the
c harac terization of special orientations which is entirel y based on
the concept of boundary c oinc idenc e, Their model seems to be the
best so far for a clear interpretation of experimental observations
on grain boundary motion. Figure 4. 7 shows the structure of a sym-
metric tilt boundary with a rotation of 2 8.1 about a<100>- axis bas-
ed on this model.
1

O

2
1
0
3

"
5
*

T
~
?

3
n

P
O
i

i

o

5
P

p
<
4

H
*

H
S
.

o

n

3
'
3

f
f
i

3
s
a
o

s
.
a
g
?
*
3

w

2
.

p

p

-
*

o
m

s
f

3

S
"

o

o

o

*
"

3

5
;

3
D

3
1
U
J

Q

W

Q
,

.

O
J
o

p

a

M
-

P

,-
H

"

-
r
-
&
0
3
M
"
j
w
g

<

o
4
fcf
1

n

B
J

A

Z
.

p

o

C

A

5
.
a

P

S
.

t

2

o
-

P

S
-
.
^
j

3

o

i

o

P

v
g
.
5
o
I
d
e
a
l

c
o
i
n
c
i
d
e
n
c
e

b
o
u
n
d
a
r
i
e
s

i
n

t
e
r
m
s

o
f

t
h
i
s

m
o
d
e
l

a
r
e

d
i
s
t
i
n
g
u
i
s
h
e
d
t
h
r
o
u
g
h

t
h
e
i
r

p
e
r
i
o
d
i
c
,

e
d
g
e

t
y
p
e

o
r
d
e
r
i
n
g

o
f

a
t
o
m
s

w
i
t
h

a

c
o
n
s
t
a
n
t
e
d
g
e
w
i
d
t
h
.

A
n
y

d
e
v
i
a
t
i
o
n

f
r
o
m

t
h
i
s

o
r
i
e
n
t
a
t
i
o
n

c
a
n

b
e

e
x
p
r
e
s
s
e
d
t
h
r
o
u
g
h

d
i
s
t
u
r
b
a
n
c
e
s

i
n

t
h
e

e
d
g
e

s
e
q
u
e
n
c
e
s

i
n

t
h
e

b
o
u
n
d
a
r
y
p
l
a

e
.
T
h
i
s

m
a
y

l
e
a
d

t
o

a
n

i
n
c
r
e
a
s
e

i
n

t
h
e

g
r
a
i
n

b
o
u
n
d
a
r
y

e
n
e
r
g
y

a
n
d

i
n
t
h
e

r
a
n
g
e

o
f

i
n
t
e
r
a
c
t
i
o
n

w
i
t
h

f
o
r
e
i
g
n

a
t
o
m
s

a
s

o
b
s
e
r
v
e
d

b
y

A
u
s
t

a
n
d
R
u
t
t
e
r
.

T
h
e

b
o
u
n
d
a
r
y
-
c
o
i
n
c
i
d
e
n
c
e

m
o
d
e
l

m
a
y

a
l
s
o

b
e

e
x
t
e
n
d
e
d

f
o
r
a
s
y
m
m
e
t
r
i
c

t
i
l
t

b
o
u
n
d
a
r
i
e
s
,

t
w
i
s
t

b
o
u
n
d
a
r
i
e
s

a
n
d

m
i
x
e
d

b
o
u
n
d
a
r
i
e
s
.
A

s
u
b
s
t
a
n
t
i
a
l

i
m
p
r
o
v
e
m
e
n
t

i
n

t
h
e

q
u
a
n
t
i
t
a
t
i
v
e

i
n
t
e
r
p
r
e
t
a
t
i
o
n

o
f

t
h
e
o
r
i
e
n
t
a
t
i
o
n

d
e
p
e
n
d
e
n
c
e

o
f

g
r
a
i
n

b
o
u
n
d
a
r
y

m
o
t
i
o
n

c
a
n

b
e

e
x
p
e
c
t
e
d

i
f
t
h
e

s
t
r
u
c
t
u
r
a
l

p
a
r
a
m
e
t
e
r
s

o
f

t
h
e

b
o
u
n
d
a
r
y

C
R
,

c
_

,

c

,

z
-

(
e
q
.

1
3
)
c
a
n

b
e

c
o
r
r
e
l
a
t
e
d

w
i
t
h

t
h
e

o
r
i
e
n
t
a
t
i
o
n
.

W
h
a
t

i
s

n
e
e
d
e
d
,

h
o
w
e
v
e
r
,

i
s
a

c
l
e
a
r

c
o
n
c
e
p
t
i
o
n

o
f

t
h
e

a
c
t
u
a
l

s
t
r
u
c
t
u
r
e

o
f

t
h
e

b
o
u
n
d
a
r
y
u
n
d
e
r
i
n
v
e
s
t
i
g
a
t
i
o
n

a

k
n
o
w
l
e
d
g
e

o
f

w
h
i
c
h

i
s

s
t
i
l
l

l
a
c
k
i
n
g
.
f

5

^

i
?

o

t

B
H

S
.

(

t
u

w
w
i

<
-

o
-

3

o
"

'

'
'

I
B

V
i

n

c

^

?

*
'"
l
'
'
'
l

S

S
w

*

a

"
*

I
'
l

'

S

^

w
1

i

i

-

g

J

i

1

*

3

0
3
1

'

1

T
O
-

3

-

W
1

'

M

R
-

;

p

O

r
t

-
T
.
8

;

*

H
-

3

c
:

o
.
f
i

o

A

o

a
!
3

"
>

e
W

-
i

5

S

v
<
r
e

"
s
*
1

*
l

O

P
*

^

*
J
1

!

5
'

2

f
f
l

*
*
^
1

H

S
"
f
i

s
r
*

O
.

h

3
-
3
i

'
'

t
u

'
!
1
'

1

1
^

V
P

'
.
:

a

.
-

-

2
.
i

"
:
-

"

^
5

o

.

X
^

^
^
-

.

*
*
*

^
!
^
k

^
S

&
$
$
$

)

^
-
'
^
m
^
S
'
s
^

^

^
^
T

?

^
^
^
^

"
^
|

^
^
l

f

!
0

n
"

&
$
&

Q
$
%

&
S
'

^
^

^
^
^
^
5

f
*
i
7
i

f

i
-
j

i

*
i
W
l
H

a

^
S
S
?
^

?
S
^
^
"

^
W
)

O
^
S
^
5
-

f
$
l
j
&
)

r
W
M
P

<
3

0

^
^

r
-
-
,

J

\
.
S

L
^
A

-
^
-

1

.
^
^

C
U

V

--

-
^
A

p
o
s
s
i
b
l
e

b
a
s
i
s

f
o
r

t
h
e

d
e
v
e
l
o
p
m
e
n
t

o
f

a

p
h
y
s
i
c
a
l

t
h
e
o
r
y

f
o
r

t
h
e
s
t
r
u
c
t
u
r
e

o
f

h
i
g
h

a
n
g
l
e

g
r
a
i
n

b
o
u
n
d
a
r
i
e
s

h
a
s

r
e
c
e
n
t
l
y

b
e
e
n

g
i
v
e
n
[
4
6

-

4
8
]

.

I
t

a
s
s
u
m
e
s

t
h
a
t

a
m
o
n
g

a
l
l

t
h
e

g
e
o
m
e
t
r
i
c
a
l
l
y

p
o
s
s
i
b
l
e
b
o
u
n
d
a
r
i
e
s

o
n
l
y

t
h
o
s
e

h
a
v
i
n
g

a

m

n
i
m
u
m

e
n
e
r
g
y

c
a
n

b
e
p
h
y
s
i
c
a
l
l
y
r
e
a
l
i
z
e
d
.

S
y
m
m
e
t
r
i
c

t
i
l
t

b
o
u
n
d
a
r
i
e
s

h
a
v
e
,

f
o
r

e
x
a
m
p
l
e
,

b
e
e
n

c
o
m
-
p
u
t
e
d

o
n

t
h
e

b
a
s
i
s

o
f

t
h
i
s

p
r
i
n
c
i
p
i
e

[
4
6
]

.

A
n

e
x
t
e
n
s
i

n

o
f

t
h
i
s
p
r
i
n
c
i
p
i
e

t
o

t
h
e

t
h
r
e
e
-
d
i
m
e
n
s
i
o
n
a
l

a
n
d

n
o
n
-
s
y
m
m
e
t
r
i
c

c
a
s
e
s

(
a
n
d
t
a
k
i
n
g

i
n
t
o

a
c
c
o
u
n
t

a
n

a
p
p
r
o
p
r
i
a
t
e

e
n
t
r
o
p
y

t
e
r
m
)

m
a
y

g
i
v
e

a

q
u
a
n
t
i
-
t
a
t
i
v
e

r
e
l
a
t
i
o
n

b
e
t
w
e
e
n

t
h
e

f
i
v
e

m
a
c
r
o
s
c
o
p
i
c

o
r
i
e
n
t
a
t
i
o
n

p
a
r
a
m
e
t
e
r
s
a
n
d

t
h
e

g
r
a
i
n

b
o
u
n
d
a
r
y

e
n
e
r
g
y
,

w
h
i
c
h

e
x
i
s
t
s

s
o

f
a
r

o
n
l
y

f
o
r

s
p
e
c
i
a
l
c
a
s
e
s

(
l
o
w

a
n
g
l
e
-
b
o
u
n
d
a
r
i
e
s
,

c
o
h
e
r
e
n
t

-

t
w
i
n

b
o
u
n
d
a
r
i
e
s
)
.

O
n
l
y
w
i
t
h

t
h
e

d
e
v
e
l
o
p
m
e
n
t

o
f

s
u
c
h

a
n

a
t
o
m
i
s
t
i
c

m
o
d
e
l

w
i
l
l

i
t

b
e

p
o
s
s
i
b
l
e
t
o

g
i
v
e

a

q
u
a
n
t
i
t
a
t
i
v
e

c
o
r
r
e
l
a
t
i
o
n

b
e
t
w
e
e
n

t
h
e

b
o
u
n
d
a
r
y

s
t
r
u
c
t
u
r
e
a
n
d

t
h
e

b
o
u
n
d
a
r
y

m
i
g
r
a
t
i
o
n

t
e
r
m
s

a
n
d

s
o

t
o

a
n
s
w
e
r

s
a
t
i
s
f
a
c
t
o
r
i
l
y
t
h
e

q
u
e
s
t
i
o
n
s

r
e
g
a
r
d
i
n
g

t
h
e

o
r
i
e
n
t
a
t
i
o
n

d
e
p
e
n
d
e
n
c
e

o
f

t
h
e

m
i
g
r
a
t
i
o
n
r
a
t
e
.
1 O 4
5. CONCLUSIONS
Al t hough t he mech ani sm of grain boundary mot ion in pur met is
is underst ood f ro m a phenomenological qual it at ive point of vi ew, a
quant i t at i ve explanation of the effect of the st ruct ural paramet ers of
grai n boundaries is impossible at present. This applies especi al l yfo
the mxi mum possible valu of the mobility and its dependence on
orient at ion. The main reason for t his is t hat t he st ruct ural details
of high angle boundari es are not yet cl ear. Precise experi ment s in
t he sense of t he t heory can be perfo rmed only when it is assured
t h at the driving f o rc is exactly known, is constant with respect to
t i me and space, and can be varied independen! of the position of the
boundary. During its migrat ion, the boundary shoul d shift only in
the direction of the normal to its surface. The high sensit ivit y of
the migrat ion rate wit h very small additions of foreign atoms pre-
sents a f u rt h er difficul t y. In regard to the last point, substantial
advancement is to be expected t hrough i mproved Chemical analytical
met hods and better preparat ion t echniques.
REFERENCES
1. S. Glasstone,
K. J. Laidler and
H. Eyring:
2. D.Turnbul l :
3. P.A. Beck,
P, R. Sperry and
H. Hu:
4. H.Gl eit er
5. P. A. Beck,
J. C. Kremer,
L. J. Demer and
M. L. Holzworth:
6. R. A. Vandermeer
P. Cordn in:
7. P. Cordn and
R. A. Vandermeer:
8. E. L. Holmes and
W. C. Winegard:
9. G. F. Bolling and
W. C. Winegard:
10. J. P.Drol et :
11. P. Cordn and
T. A. El-Bassjouni:
12. J. P.Drol et and
A.Gal ibois:
13. B. B. Rath and
H. Hu:
14. B. B. Rath and
H. Hu:
10 5
"The Theory of Rate Processes",
Me Graw-Hill, New York 1941.
Trans. AIME 191 (1951) 661.
J. Appl. Phys. 2\. (1950)420.
Acta Met. H (1969)853.
Trans. AIME 175 (1948) 372.
and
"Recovery and Recrystallization of Metals",
L. Hi mmel , Ed., Interscience, New York
1963, 211.
Trans. Met, Soc. AIME 224 (1962) 917.
Acta Met. 7 (1959)411.
Acta Met. 6 (1958) 283.
Dr. Se. Thesis, Laval Univ.Quebec (1969).
Trans. Met. Soc., AIME 233 (1965) 391.
Acta Met. U (1968) 1387.
Trans. Met. Soc., AIME 245^ (1969) 1577.
Trans. Met. Soc., AIME 245 (1969) 1243.
15. W. W. Mullins:
16. P. Cordn and
R. A. Vandermeer in:
17. ) G. F. Bolling and
W. C. Winegard:
18. E. L. Holmes and
W. C. Winegard:
19. K. Lcke and
H. P.Stwe in:
2ty G. F. Bolling:
21. B. B. Rath and
H. Hu:
22. K. T. Aust and.
J. W.Rutter:
23. C. Frois and
M. O. Dimitrov:
24. R. Fromageau:
25. K. T. Aust and
J.W. Rutter in:
26. K. Detert and
G. Dressler:
27. J. W. Rutter and
K. T. Aust:
28. J.P.Drolet and
A.Galibois-
29. H.Gleiter and
A. Lissowski:
30. F. HaeBner and
S. Hofmann:
1 0 6
Acta Met. ^(1956) 421.
"Recrystallization, Grain Growth
and Textures", H. Margolin, Ed. ,
A. S. M. , Metals Park Ohio 1966,205.
Acta Met. 6 (1958) 288.
Can.J.Phys. 37 (1959) 899.
"Recovery and Recrystallization o
Metals", L. Himmel, Ed., nter-
science, New York 1963, 271.
Acta Met. 16 (1968) 147.
Trans. Met.Soc., AIME 236 (1966) 113
Trans, Met.Soc., AIME 215(1959) 820
Ann. Chim. :l (1966) 113.
Mem.Sci.Rev. Met. 66(1969) 287.
"Ultra High Purity Metals", A. S. M.
Metals Park, Ohio 1962, 115.
Acta Met. \3 (1965) 845.
Trans. Met.Soc., AIME 218 (1960) 68!
Acta Met, 16 (1968) 1387.
Z. f. Metallkde. 62_ (1971) 237.
to be published.
1 0 7
31. W. in der Schmltten
P. Haasen and
F. Haefiner:
33. M. Feller-Kniepmeier and
K. Schwartzkopff:
33. R. A. Vandermeer:
34. H. D. Mengelberg,
H. Meixner and
K. LUcke:
35. F. Haefoer and
H.P.Holzer:
38. F. Hiusaerniann,
M. Rhle,
G. Roth and
G. P. Scheidler:
37. F. HaeBner,
S. Hofmann and
P. Holzer:
38. B. Liebmann,
K. Lcke and
G. Masing:
39. K. T. Aust and
J. W. Rutter:
40. G. Ibe and
K. Lcke in:
41. S.Kohara,
M. N. Parthasarathi and
P. A. Beck:
42. R. Klar and
K. LUcke:
43. M. L. Kronberg and
F. H. Wllson:
44. D. C. Brandon,
B. Ralph,
S. Ranganathan and
M. S. Wald:
Z. f. Metallkde. 51_ (1960) 101.
Acta Met. H (1969) 497.
Acta Met. 15 (1967) 447.
Acta Met. 13_ (1965) 855.
Scripta Met. 4 (1970) 161.
phya. stat. sol. 32 (1969) K 103.
3. Int. Reinststoff-Symposium,
Dresden 1970,64.
Z. f. Metallkde. 47 (1956) 57.
Trans. Met. Soc., AIME 215
(1959) 119.
"Recrystallization, Grain Growth
and Textures", H. Margolin, Ed.,
A.S. M., Metals Park, Ohio 1966,434.
Trans. Met. Soc., AIME 212 (1958) 875.
Z. f. Metallkde. 59 (1968) 194.
Trans. AIME 185 (1949) 501.
Acta Met. 12 (1964) 813.
1 O 8
45. G. H. Bi s hop and
B. Chalmers:
46. M. Weins,
B. Chalmers,
H.Gleiter and
M. Ashby:
47. M. Weins,
H. Gleiter and
B. Chalmers:
48. T. Johannesson and
A. Thblen:
Scripta Met. 2 (1968) 133.
Scripta Met. 3 (1969) 601
Scripta Met. 4 (1970) 235.
PhlLMag. 21 (1970) 1223.
1 09
HAPTER 5
ECONDARY RECRYSTALLIZATION
IY
IDETERT
,THE PHENOMENON
The phenomenon of secondary recrystallizatton is similar to that of
primary recrystallization. While primary recrystalllzation occurs
in a cold deformed matrix, secondary recrystallization takes place
in a recrystallized structure. As shown in Fig. 5.1 a few grains in
the recrystallized matrix grow vigorously during annealing until they
impinge on one another. Finally a coarse grained structure has re-
placed the fine grained matrix. The time dependence of the progress
of secondary recrystallization can be described as that of primary
recrystallization by the well known Avrami relation [ 1 ] :
X = 1 - exp (- A tK) (X fraction of secondary recry-
stallized grains,
t annealing time),
the constants A and K are to be determined by experiment.
ZTHEORETICAL ASPECTS
Secondary recrystallization also called discontinuous grain growth,
represents a special case of grain coarsening. The free energy g
(per mole) (c. f. chapter 1 section 1.1.) which is responsible for the
grain coarsening process originales from the presence of grain
boundaries with a grain boundary energy f-. The rea of grain
110
t
C O
t! 2
2 o
S .
LO-S
"S
~
t
3 i
o .3
u ^
1 1 1
boundaries present in the matrix gradually decreases during coarsenlng.
Henee the relation holds:
'B
(V mole volume, D mean grain diameter, /j- geometrc factor
between 1 and 3 ).
In the case of Cu, j'- = 500 erg/cm . With D = 10 A im: g/V =
6 3
1 10 erg/cm . This valu is very much smaller (by a factor
of 100) than the valu of free energy responsible for primary re-
crystallization in a deformed matrix.
The rate of grain growth v =
v = m - 4
dD
dt
can be expressed by the relation:
where m is a mobility term.
Therefore the time dependence of continuous grain coarsening is
proportional to VT
This relation holds for the mean grain diameter. The grain growth
behaviour of an individual grain is much more involved. It will be
governed by the individual processes which occur in the matrix dur-
ing annealing such as the straightening of a curved boundary, and
shifting of the position of a grain boundary junction into an equilibrium
position in respect to the stresses from the adjoining grain boundaries.
ULTHA M [ 2 ] has described the grain growth behaviour of individual
grains in more detall. Large grains with more than six sides in a
micrograph tend to grow at the expense of smaller grains (c. f. chap-
ter 2). In a matrix, a diatribution of grain sizes exists as shown in
1 1 2
Fig. 5. 2, and this can be described as a log normal distribution.
2 3 D Cmm]
Fig. 5. 2
Distribution of grain sizes in recrystallized Fe-Si3%.
As outlined in Fig. 5. 3. a, only the large grains continu in existence
during continuous grain coarsening. The majority of smaller grains
disappears in spite of the fact that the mean grain diameter grows as
the square root of the isothermal annealing time.
1 1 3
bu
fc
h
b e
B B
o
O
20
' ^ -4- CQ
rt 0>
E ^
1 1 4
Secondary recrystallization requires that the grain growth be strongly
impeded, with the exception of a few grains which act as nuclei for
the secondary recrystallization. Then the initial rate of grain grourth
is independent of time, and the motion of grain boundaries of those
growing grains is directed towards the center of curvature. During
the process of secondary recrystallization the distribution of grain
size becomes very inhomogeneous, as is shown in Fig. 5. 3.b.
3. THE INFLUENCE OF SMALL PARTICALS
In Fig. 5. 4 is shown that the coarsening of the Fe Si3% matrix
during annealing has come to an end once a certain grain size is
reached.
D
0,025
0,020
0,015
0 1 2 3 4 5 6 ^timefrrsJ
Fig. 5. 4 - Grain growth in FeSi3% at 800 C.
This behaviour is caused by the presence of very fine submicroscopic
partides as shown in Fig. 5. 5.
115
Fig. 5.5
Small particles present in Fe Si3% to prevent grain grourth
1 1 6
C. ZENEP [3] has outlined the idea how parcles impede grain growth.
A partile present in a grain boundary decreases the rea of grain
boundary by the cross section of the particle. The motion of a boun-
dary through a matrix with spheric partiles of a mean diameter d .
is exposed to a certain dragging energy which decreases the free
energy of grain boundary motion by A g. (c. f. chapter 2 section 3. 2):
( f volume fraction of particles
T2 geometric factor between
1 and 2).
The continuous grain growth comes to an end when Ag = U g. , henee
a limit grain size of D = D. . is defined:
D. = n .
d.
factor between 1 and 2).
In Fe Si3%, small particles of manganeso sulfide act to impede grain
coarsening. From the Fig. 5. 5 one measures d. = 0,1 A im, f = 1%
and one derives D,. = 10- 2 0 / um, which does agree very well
with the observed limit of grain size shown in Fig. 5. 4.
Therefore the presence of particles fulfillsthe one requirement that
secondary recrystallization may occur. P.A . BECK has shown for
the first time that secondary recrystallization in an A l-Mn alloy
is linked with the presence of fine particles of Mn A lfi [4] , However,
he only observed this phenomenon during annealing in a temperature
range where the fine particles start to dissolve.
In order to explain secondary recrystallization one must also explain
why exceptions exist for some individual grains while grain growth
in general is impeded in the matrix.
117
As was outlined by BURKEandTURNBULLfor a grain with diameter
D , the following relation must hold ih order to have a nucleus pro-
duced[5]: ^+ g. <g .
g- determines the free energy of a growing nucleus:
Therefore
dlferences in
I'* 12
Dlim
n .
D,
f \ i
- , or neglecting
D
D. > ^. D
Dlim
From this relation two conclusions can be drawn. First: Grains with
a larger diameter than average may act as nuclei. Second: D should
be smaller than A . (see chapter 10 section 3. 2.).
In the case of Fe Si3% with MnS particles, during annealing below
850 C continuous grain coarsening has come to an end as
D = D.. = O , 025 mm was reached. During continued anneal be-
tween 850 and 950 C the Mn S particles gradually begin to coarsen
and to dissolve, henee increasing D.. . Therefore secondary re-
crystallization is promoted. Rapidly heating the samples to tempera-
tures above 950 C produces a much faster coarsening and dissolv-
ing of MnS particles. Since D,. increases to substantially higher
vales than D , a larger number of grains fulfill the size require-
ment to act as nuclei for secondary recrystallization. The final
structure then becomes much less coarse. So the observation is ex-
1 1
plained that secondary recrystallization due to small particles is v ery
pronounced in its appearance only during annealing in that temperature
range where the particles v ery gradually start to coarsen and to dis-
solv e.
4. THE INFLUENCE OF TEXTURE
Another kind of impediment to grain coarsening can be obtained
when a v ery pronounced preferred crystal orientation is present in
the matrix [6j. C.G. D U N N has demonstrated and outlined how under
those circumstances secondary recrystallization in sheets of Fe S i3 %
can occur [7]. A preferred crystal orientaiion of the grains in the
matrix means, that only small angle boundaries exist between the
indiv idual grains of the preferred orientation. The energy of a small
angle boundary is much lower than that of a high angle boundary. A
boundary associated with a difference of crystal orientation in the
order of 5 - 1 0 has only O, 3 to O, 5 times of the energy v alu av er-
age of a high angle grain boundary for the same metal. Furthermore
the mobility of such a low angle boundary is generally small in com-
parison to the mobility of a high angle boundary.
Therefore the expression of v for rate of grain growth leads to
much smaller v ales when low angle boundaries are inv olv ed than
with ordinary grain boundaries.
One can also understand that exceptions from impeded grain growth
will exist, when still a few grains are present with crystal orient-
ations which di ffer f r om the general preferred orientation. For these
grains a tendency to shrink would persist if no size difference existed.
One of these grains with different orientation will not act as a nucleus
if the following relation be obeyed:
1 1 9

D.
'LAB LAB
Dl <
D
LAB
low anS le boundary
energy ).
This is to say that the grain of different crystal orientation must be
larger than av erage so that the diameter D1 2 - 3 times D.
Figure 5. 6 a and b show a process of secondary recrystallization
in Fe S i3 % , where continuous grain coarsening was prev ented by a
preferred (110) [001] orientation in the recrystallized matrix. At
the position of former grain boundaries before cold rolling, grains
with different orientation and larger size hav e originated during the
recrystallization anneal. During a continued anneal, secondary re-
crystallization was promoted from those grains.
5 . THE INFLUENCE OF SHEET THICKNESS
The influence of the shape of the material on grain coarsening dur-
ing annealing is well known. In a wire, grain coarsening stops when
the grain size coincides with the diameter of the wire. Then the
boundaries straighten across the wire and an equilibrium position
is reached.
One has explained the influence of sheet thickness on grain coarsening
in similar terms, since it was noted that grain coarsening stops when
the mean grain diameter has reached a size of twice the thickness of
the sheet [8J, Fig. 5. 7. W. MU LLIN S [9] has giv en a more thorough
^
-
J

'
"
t
f
i

i

n

-

o

i-
l
0
>
<
4
i

i

U
I
L
_
,

-
B
^
l
?

B

S
P
O

*
~
^
C

3

~
0
.
3
c

3
-

3
5
'

^
P

O
3

<

*
3

M
2

a
^
3
j
q

(
C
U

C
O
(
m
m
)
\
0
.
0
8
0
,
0
6
0
.
0
4
0
.
0
2
O
j
O
O
T
e
m
p
e
r
a
l
u
r
e
f
i
e
l
d
o
f

S
e
c
o
n
d
a
r
y
R
e
c
r
y
s
t
a
t
t
i

s
o
f
i

n
6
0
0

7
0
0

3
0
0

9
0
0

1
0
0
0

1
1
0
0

1
2
0
0

(

C
)
F
i
g
.

5
.

7
F
i
n
a
l

g
r
a
i
n

s
i
z
e

a
f
t
e
r

a
n
n
e

l
i
n
g

o
f

O
,
0
4

r
a
r
a

t
h
i
c
k

s
h
e
e
t
s

o
f
F
e

S
i
3
%
.
122
explanation. He has outlined the influence of shape in terms of therm
grooving where grain boundaries intersect the surface.
Thermal grooving alwavs occurs during high-temperature bright
anneal. In order to equalizethe stresses of the surface and of the
grain boundary along the Une of junction, a certain angle in the root
of the groove must be formed as demonstrated in Fig. 5. 8:
= sin
in general
' B
henee
surface energy),
= 10
The grain boundary of an individual grain which intersects the surface
with a curved Une will have as equilibriumshape a catenoid. The
curvature X can be described with two radii of curvature ortho-
1 1
gonal to each other so that X =
r2
= O (Fig. 5.9).
Since the grain boundary then forms an anglet f f with the surace
different from 90 the groove must be unsymmetrical in order to
stabilize the Une of junction (Fig. 5.10). The angle ^= 90 - f
is the limit that a stabilizing groove can be formed by an equilibrium
shape of the surface. That means the angle t ir of the intersecting
boundary must be yr - 90 - t p in order to be trapped by the groove.
In order to formsuch an angle a certain relation of radius of curva-
ture of the boundary to sheet thickness must hold:
- r.
2 sin ( w sheet thickness),
smce 10'
i
3 w
This means that grains with curved boundaries intersecting the sur-
tace are stabilizedwhen the radius of curvature of these boundaries
123
o
F
i
g
.

5
.
9
S
c
h
e
m
a
t
i
c

p
i
c
t
u
r
e

o
f

s
h
a
p
e

o
f

a

c
r
o
s
s
i
n
g

g
r
a
i
n

b
o
u
n
d
a
r
y

i
n

a

s
h
e
e
t
.
I

o
\
.
F
i
g
.

5
.
1
0
S
c
h
e
m
a
t
i
c

p
i
c
t
u
r
e

o
f

a

g
r
o
o
v
e

a
t

t
h
e

g
r
a
i
n

b
o
u
n
d
a
r
y

i
n
t
e
r
s
e
c
t
i
n
g

t
h
e

s
u
r
f
a
c
e
.
1 2 6
is greater than 3 - w . MULLINS then deduces f rom this fact that a
matrix with a mean grain diameter of D > 2 ty is stabilized by
the thermal grooves.
One has observed secondary recrystallization when coarsening of the
matrix was stopped by sheet thickness in Fe Si3% [10] . In order
to promote secondary recrystallization, a nucleation requirement
had to be fulfilled. This requirement is fulfilled when differences
of surface energy between individual grains exist. When a grain
has a difference of surface energy Afi so that its surface energy
s
is lower than the surrounding grains a relationship can be set up,
which determines its f ree energy of growth:
'1
# t
2
this grain will grow when D, _
' B
D.
2- Vw
This means D. can be even smaller than the average grain size
D = 2 w .
In Fig. 5.11, secondary recrystallization is present in a Fe Si3%
sheet with a thickness of O, 04 mm. The mean grain diameter before
secondary recrystallization has set in is shown in Fig. 5. 7 and
amounts to O, 09 mm. The grains of the secondary recrystallized
matrix had a crystal orientation so that cube planes of the lattice
were within 5 parallel to the surface. Different measurements of
surface energy have shown that one can observe this kind of secondary
recrystallization when Afi > 0,1 $ . That means that such
"o S
grains with a lower surface energy with a diameter of more than
O, 9 w will fulfill the requirement for nucleation.
127
, 0,5 mm ,
' !
ig. 5.11
in a
e a3% during annealing at 1100 C'
(a) 5 min, (b) 20 min, (c) 1 h.
sheet
1 2 8
6. TERT1ARY RECRYSTALLIZATION
The surface energy of a crystal does not only depend on the orient-
ation of its crystal plae at the surface. It also depends on adsorption.
Adsorption depends on the annealing atmosphere and on the temperature
of annealing. It has become possible to favor different crystal planes
of the same material for secondary recrystallization to occur simply
by changing the impurities in the atmosphere. One has even observed
that a sheet of Fe Si3% which started a secondary recrystallization
by favoring grains with cube plae (100) parallel to the surface, could
undergo a transition to a second kind of secondary recrystallization,
now favoring grains with (110) planes parallel to the surface, while
the purity of the annealing atmosphere was altered durng the anneal.
This type has been termed tertiary recrystallization [ll ].
7. EXAMPLES OF PRACTICAL USE
OF SECONDARY RECRYSTALLIZATION
In general, coarse grains should be avoided during the processing
of metallic materials, strength and ductility decrease wi th grain size;
f or mi ng processes are therefore more difficult to execute on coarse
grained material. Fur ther mor e, the smoothness of the metallic sur-
face is impaired by the deformation of a large-grained matr i x. The
conditions of an annealing treatment must therefore be chosen so as
to avoid secondary recrystallization.
In magnetic ferrous sheets there are a few cases where secondary
recrystallization is purposely applied lo produce a coarse-grained
structure with improverl magnetic properties. It is known that
coercive forc decreases with grain size [12]. In some cases se-
condary recrystallization is applied to produce a preferred crystal
orientation which creaies magnetic anisotropy in the sheets. Trans-
f or mer sheets of 0,30 - O, 35 mm thickness are at present being made
1 2 9
of grain-orientad Fe Si3%. A secondary recrystallization occurs dur-
ing the final annealing process in the temperature range of 900 - 950 C.
Small inclusions of MnS particles provide the inclusin inhibited normal
grain growth as outlined before. This" secondary recrystallization pro-
cess produces a (110) [001] texture with a directionally pr ef er r ed mag-
netization along the strip rolling direction. In this case the cold rolling
treatment provides a recrystallized matrix so that the longer grains
which will act as nuclei of secondary recrystallization have the desired
crystal orientation. To obtain the (100) [001] texture in sheets of
Fe Si3% which produces directionally preferred magnetization in both
the rolling direction and transverse to it, an annealing treatment is
required which causes a secondary recrystallization when sheet thick-
ness has stopped the normal grain coarsening process. A special cold
rolling treatment takes care that before secondary recrystallization
occurs only such grains with a (100) plae parallel to the surface are
present in the matrix which have a cube edge orientation parallel and
normal to the rolling direction.
In other cases, grain coarsening by secondary recrystallization destroys
the preferred crystal orientation which is desired for good magnetic
properties. An example is the annealing treatment of magnetic shees
Fe Ni 50%. Here a primary recrystallization produces the (100) [OOl]
texture whi ch gets destroyed when too high an annealing temperature
above 1100 C is applied causing secondary recrystallization.
A more detailed description of the development of annealing textures
is given in another paper of this volume (see chapter 6).
Here another example shall yet be mentioned, where secondary re-
crystallization is used to improve the mechanical properties of tungsten
filament wire. The tungsten material which is used for lamp filaments
is called doped tungsten because it contains small additions of alumi ni um,
silicon and potassium to produce a fibrelike structure of elongated
1 3 O
grains in the wires; see Fig. 5.12. This grain structure is necessan
to produce the required Ufe expectation at the incandescence temperah
of the filaments.
During annealing of the drawn filament wire, a so-called recrystalliz-
ation in situ occurs in the temperatura range between 1200 and 1500 C
This type of primary recrystallization resembles more a kind of re-
covery by subgrain coarsening. However, the dislocation density is:
moved and high angle boundaries are formed. The very fine particles
of the metallic additions appear to be responsible for this type of re-
crystallization behaviour [[13].
During the wire drawing process, the inclusions of metallic additions
such as mullite (Al-Si- ?O. ) which may also contain potassium or its
hydroxide will forra stringers which are difficult to recognize. Duriii
the annealing, strings of very fine spheric particles or voids show up
in the recrystallized matrix; see Fig. 5.13. There exists a differenct
of opinin whether the spheric voids in the tungsten matrix are filled
with glassy silicate or with potassium vapor. During a continuad
annealing above 2100 C a discontinuous grain growth occurs as seen
in Fig. 5.14, Yet the larger grains developing are elongated instead
of equiaxed. The anisotropy of the growth of these grains is due tott
rather anisotropic distributions of the strung-out voids.
There is some argument in the literature whether this process of gr
coarsening should be labeled primary or secondary recrystallization
while a regulary primary recrystallization is indeed missing. The
discontinuous grain growth is however controlled by grain boundary
energy. Therefore it is only justified to cali this process secondary
recrystallization, where normal grain growth is impeded by the fine
inclusions or voids. It has been demonstrated that at temperaturas
above 2100 C those voids start to coarsen thus fulfilling the require-
ments for this type of secondary recrystallization.
Fig. 5.12
Tungsten filament annealed at 3000C.
Fig. 5.13
Doped tungsten annealed 3, 5 h at 2200C afterWAUIR[l3].
132
,C;bn-m
Fig. 5.14
Large grains growing in doped tungsten at 2200C afterWALTEE(l3].
REFERENCES
1. M. Avrami:
2. P. Feltham:
3. C. Zener:
4. P. A. Beck,
M. L. Holzworth and
P.R. Sperry:
5. J. E. Burke and
D. Turnbull in:
6. P. A. Beck and
P.R. Sperry:
7. C. G.Dunn:
8. P. A. Beck,
J, C. Kremer,
J, L. Demer and
M. L. Holzworth:
9. W. Mullins:
10. F. ABmus,
K. Detert and
G.Ibe:
11. J. L. Walther and
C. G. Dunn:
12. A. Mager:
13. J. L. Walther:
1 3 3
J. Chem.Phys. 7 (1939) 1103, 8(1940)
212, 9 (1941) 174.
Acta Met. 5 (1957) 97,
Discussion Trans. AIME 175_ (1948) 15.
Trans. AIME 180 (1949)163.
"Progress in Metal Physics", B. Chal-
mers, Ed. , Pergamon Press, London
3 (1952) 220.
Metals Trans. Marchfl949^240.
Acta Met. 1 (1953) 163, 2 (1954) 173.
Trans. AIME 175 (1948) 372.
Acta Met. 6 (1958) 414.
Z. f. Metallk. 48^ (1957) 344.
Trans. AIME (1959) 465.
Ann. Phys. 6F U (1952) 16.
Trans Met. Soc. AIME 239 (1967) 272.
134
1 3 5
DHAPTER 6
ANNEALING TEXTURES
1 Y
IGREWEN
4
tINTRODUCTION AND BASIC PRINCIPLES
Crystal orientation is of great significance in the recrystallization
of metallic rnaterials. The two basic processes of primary recrystalliz-
ation, nucleation and growth of the nuclei by th migration of high-angle
grain boundaries, can both be orientation-dependent [1,2] . In secn-
dary recrystallization the orientation dependence of the grain boundary
mobility and of the surface energy must be taken into account. In addi-
tion secondary grains grow at the expense of stable matrix of small
grains; certain orientations cause matrix stabilization and thus actual-
ly initiate secondary recrystallization [3, 4] . Finally, the driving
forc for tertiary recrystallization arises from orientation-dependent
differences in the surface energy [4] .
Whenever one of the basic processes is orientation-dependent the re-
crystalized grains of a deformed and annealed single crystal can have
only certain quite definite orientations. This is also true in the case
of polycrystalline material in which the deformation has given rise to
a largely uniform orientation of all the grains, in other words a de^
formation texture [5] . A rise in temperature leading to softening
causes annealing textures to develop from these deformation textures.
All textures produced during the annealing of metallic materials are
termed annealing textures (not recrystallization textures) because re-
covery as well as recrystallization exerts a certain influence [6j.
1 3 7
136
1.1. Processes of Softening and Grain Growth
Recovery, which can occur before or alongside primary recrystallii-
ation, principally involves subgrain format i on and enlargement. I nt ht
following, recovery by means of subgrain format i on and growth ( form-
ation and migration or elimination of low-angle grain boundaries) will
be distinguished from recrystallization ( formation and migration of
high-angle grain boundaries) . So recrystallization will be defined
according toBEEK [6] . I t thus includes pri mary recrystallization in-
volving nucleation and growth until all the new grains are in mutual
" ' contact, and the grain growth processes subsequent to completad
primary recrystallization. These are either continuous ( normal
grain growth) or discontlnuous ( secondary recrystallization). Also
to be included are tertiary recrystallization, characterised according
* toDU^N andWALTER [4] as the surface energy controlled growth of
small grains at the expense of much larger grains, and all grain
boundary migrations for which the driving forc is Chemical energy.
Changes in texture associated with these will not, however, be describ-
ed in the following.
1. 2. Representation and Description of Textures
Textures are represented in stereographic projections as so-called
pole fi gures showing the intensities of a particular set of planes, e.g.
{lllV{lOoV or {llO\ The plae of projection is the plae of
the sheet in which lie the rolling di rect i on RD and the t ransverse
direction TD. Figure 6.1 shows the quantitative 4100J- and 011}
pole-figures of a rolled 99, 5% pur alumi ni um sheet [8] . I n a
quantitative pole fi gure the intensities are normali zedwi t h respect
to the random orientation. I t should be differentiated f rom the semi-
quant i t at i ve pole-figure in which although the intensities are measure
The format i on of a new st ruct ure in the absence of large-angle grai
boundary mi grat i on is called recrystallization in situ or, recently,
continuous recrystallization [ f j .
quantitatively they are not normalised.
Pole figures are two-dimensional projections of t hree-di mensi onal
space and do not therefore describe the actual texture completely
exactly. Since mathematical methods for the calculation of t hree
dimensional orientation distribution functions have only recently been
derived [9 - 11] there are not yet sufficient results available for
them to be taken into consideration in a discussion on annealing tex-
tures. I t is, however, already clear that our concept of the develop-
ment of textures will change considerably once sufficient calculations
are available.
Normal pole figures are not however a working basis for a theory of
annealing textures. Rather it is necessary to simplify the data even
furt her, that is;to find ideal orientations for the pole figures. I deal
orientations are single crystal orientations whi ch can be considered
representative of the texture; in the case of sheets t hey consist of
the ndices of the crystal plae ( hkl) parallel to the rolling plae and
the crystal direction [uvw] lying in the rolling direction. These ideal
orientations have to be doubled or quadrupled according to the planes
of symmetry of the sheet ( ND-RD and ND-TD, ND - - normal to the
sheet), so that there are either one, two or four crystallographically
equivalent components.
I n the case of f. c. c, metis and alloys whi ch up to a stacking fault
F F*
paramet er s/Gb ( s = stacking fault energy, G = shear modulus,
_ 3
b = Burgers vector) greater t han 5. 10 have the texture Hlustrated
in Fig. 6.1 ( copper type rolling texture), the mean ideal orientation
is frequent ly given as ( 123) [412] . This is plotted in the{lll}pole
fi gure in Fig. 6.1 ( one of four equivalent component s) [6,12],
Theoretically ( Ol ) [2l] + ( 112) [ll] is a better description [13].
1 38
0 , 5
{100}
(O
(1123 C">J
A (123) f <f ?J
Figs. 6.1 a and b
Quantitative pole figures of the rolling texture of a 99, 99%
aluminiutn foil of thickness 38 ,um and a rolling reductioh
of 95% (after GREWEN, SEGMLLER and WASSERMANN).
1 3 9
One each of the two components of these ideai orientations are plotted
in Fig. 6.1. The poles of the ideal orientations describe the texture
mxima quite well but the different intensities of the mxima and the
large scatter of the texture are both disregarded. The scatter can
be represented by a continuous series of single crystal orientations
[14] or by limited fibre textures [3,15l but the model is then too
cumbersome to describe the relationships between rolling and anneal-
ing textures more over to interpret them.
Such difficulties encountered in the study of polycrystalline material
have led to the increasing use of rolled and drawn single crystals.
Besides well-defined starting textures, such single crystals offer
advantages in respect of microscopic structure and the observation
of nucleation and crystal growth.
J. Different Types of Annealing Textures
There are three general possibilities for the formation of an anneal-
ing texture. In the first, inspite of softening, the deformation texture
is retained or only sharpened. This can be attributed to both subgrain
formation and enlargement and recrystallization (see section 2. 3. 1).
The second possibility consists of a change in the texture produced
either by primary recrystallization, normal grain growth, secondary
or tertiary recrystallization. The final possibility is that a random
orientation is set up on annealing but this occurs much lesa often than
is commonly supposed and is usually associated with a very coarse
grain which hinders the texture determination.
Only a few representative cases exemplifying the present state of
texture research will be discussed here; a summary of the extremely
numerous annealing textures possible in metis and their alloys may
be found in a number of comprehensive reviews [5 , 16 - 18] .
140
1. 4. Orientation Dependence of Grain Boundary Mobility
The orientation dependence of the grain boundary mobility is an
important experimental parameter for the interpretation of recrystalliz-
ation textures. Grain boundary migration and its dependence on cer-
tain f actors has already been described in detail in chapter 4. Accord-
i ngt ot h i s an orientation relationship giving mximum growth rate is
found in the case of f. c. c. materials characterised by a 30- 40
rotation about a common< 111> direction. In b. c. c. materials the
rotational orientation relationship is 25 <110> and in c.p. materials
30<0001>( c. f . chapter 4section 4. 5.).
If in a deformed single crystal nuclei are formed in very many, if
possible in all imaginable orientations frandom orientation), as a re-
sult for example of cutting off one end of the crystal and localized
heating, the nuclei growing most rapidly during annealing are those
having the specified orientation relationship with the matrix. In this
way growth selection occurs among the nuclei which is all the more
pronounced the more nuclei are present in random orientation [19].
The number of such nuclei increases with the degree of deformation
[20].
The orientation dependence of the grain boundary mobility can be up-
set by both dissolved alloying consituents and particles. AUS T and
HLHTER[2l] were the first to point out the influence of dissolved con-
stituents. They found - albeit for a very small driving forc of
3 4 3
10 to 10 erg/cm (cell striation structure in cast lead) - both a
decrease in the growth rate with increasing impurity f t i n to a max.
O, 006 wt. %) and an orientation dependence of the growth rate in a
certain alloy composition range only. In view of the much l arger
o
driving forc necessary for primary recrystallization (about 2. 10
3
erg/cm for copper [after 22] this result may have no significance
:
141
for primary recrystallization but grain coarsening processes may well
5 3
be impeded (driving forc about 2.10 erg/cm at a grain boundary
o
energy of 1000erg/cm and a grain radius of 0,1 mm [22]).
If particles are present these also hinder grain boundary migration,
which, according to 2NER [23] , comes to a complete standstill when
the particle drag equals the driving forc for grain boundary migration.
Again the greatest effect will be observed in grain growth processes
(c.f. chapter 2). If the particles are able to redissolve the orientation
dependence of the grain boundary mobility can also be upset because
it is replaced by a solution controlled selection principie [24]. The
grains orientated favourably with respect to the matrix are not
necessarily those unimpeded by particles.
The orientation dependence of grain boundary mobility also has an
efect on nucleation [l] in as far as a recrystallization nucleus must
not only be almost stress free and sufficiently large, as required by
all theories of nucleation, but may also possess neither the same
orientation as the deformed matrix or its twin orientation because
it would then be completely incapable of growth. How in a largely
uniformly orientated, deformed matrix nuclei arise with deviating
orientations, must be taken into account in the various theories of
nucleation (see section 2. 2. ).
5. S tructure of the Deformed State
Following on f rom the electrn microscope studles on iron-eilicon
single crystals by WALTER and KOCH [25] and by HU [26] it has be-
come more and more widely accepted that the structure of deformed
material has a decisive influence not only on primary recrystallization
but also on the texture thereby produced. According to the material,
*' In this connection it should be mentioned that the increase In ran-
dom orientation can be ascertainod only by quantitative pole figures.
143
1 4 2
degree of deformation, deformation t emperat ura and initial orientation
before rolling, widely varying dislocation configurations are possible
wh i ch influence nucleation and henee the primary recrystallization,
texture. In contrast to the od idea of a single nucleation mechanism,
current theories favour several mechanisms which can oprate de-
pending upon the dislocation structure in the deformed material (c.f.
chapter 3 section 3.).
An important sructural feat ure from the point of view of recrystalliz-
ation are macroscopic deformation bands such as those illustrated in !
Fig. 6. 2 [30] visible in the optical microscope, first observed raany
years ago [27] but only recently [25, 26, 28, 29] accorded mounting
attention. Macroscopic deformation bands are regions with the same
orientation having a band-like structure usually lying parallel to the
deformation direction.
The micrographs of drawn copper single crystals with increasing de-
grees of deformalion illustrated in Fig. 6. 3 [31] give an overall
pi ct ure of the structural features observable in the electrn mi cros-
cope. Investigations on polycrystalline material yielded a similar re-
sult [32] . It can be seen f rom Fig. 6. 3. a that at low degrees
of deformat i on a ch aracteristic cellular structure is f ormed.
The tendency to cell format i on as against randomly distributed con-
figurations of dislocations increases with the stacking fault parameter
E
s/Gb and is t h us in the case of the f. c. c. metis largest for
alumi ni um decreasing th rough nickel and copper to silver. In the case
of alumi ni um the cell boundaries can be very sh arply developed by the
deformation alone.
The cell size decreases with increasing deformation and, in the case
of the f. c. c. metis, reach es a limiting valu as a result of dynamic
recovery [33]. The limiting valu of O, 2 to O, 3 Aim [3l] found for
Figs. 6. 3 a to c
Electron micrograph s of drawn copper single
crystals, puri t y 99, 99%, drawing direction
= DD (after Ahlborn and Sauer):
(a) initial orientation approx. <321> , drawn 10%.
(b) initial orientation approx. <311> , drawn 50%.
(c) initial orientation approx. < 311> , drawn 90%.
Fig. 6. 2
Macroscopic deformation bands in the middle of
a rolled copper single crystal with an initial orient-
ation (111) [Ol] . Puri t y 99, 99%, rolling re-
duction 60%, plae of section ND-RD (after HEYE
and SATTLER).
144
the single crystals discussed here was in good agreement with that ob-
tained from polycrystalline material [33], In the case of b. c. c.
metis no constant final valu of the cell size is observed.
Figure 6. 3 b shows that at higher degrees of deformation attainment
of the limiting valu is followed by "ell elon^ation which proceeds with
the sharpening of the cell boundaries to-ui*. ..^.ation of so-<,alled
microbands (Fig. 6. 3 c). At a relativeiy late stage, that is, at a
strain of 70 to 90%, whole groups of these microbands form. In the
case of single crystals - as first shown for iron-silicon, r?c <f~ ! -
they represent transition regions between macroscopic defor
bands with different crystallogra^Viie opECTfattons^ Thus the graui
boundaries shown in Fifr," <Wg not tn' ai-& M
but transition regions buiit up o\o bauc-& \ C Va. C-26 T
shows microbands in iron-silicon.?*<+eTKYiev ccua.Uc,\ - *J T
direction.
The microbands differ ,y^ &.3^tu
separated by low-angle graiTi
continuously from one
differences in orientation can
single crystals in Fig. 6. 3 in
<100> has been produced by
55 . In the above mentioned iron^bi' ' oM
which has an initial orientation o [ Cu O ' 'l * O C?J
an end orientation (001) [230J f -f i > < * pp^C* x m.i
Figure 6.4 b illustrates the elet rpn / vucr o S f o p^ c
macroscopic deformation band as observed in iron-silicon single
crystals with an initial orientation of (001) [lOO]. It consists of a
largelyrandom dislocation structure, whereas in the deformation
bands in the drawn copper single crystals, cells and isolated bands
could be detected.

C<5"j
otteMf^-OM ^j
'-' o^'^
on- silicon
ion (001)
800 , cold
.-e =
oetween
"bands.
6. 5 a and b
Electron micrographs of a copper-chromium
alloy with 0,6 wt.%Cr; rnlling reduction 95%,
plae of section ND-RD - ! t ^i SAUER):
(a) microbands in the di' forn c-d condition.
(b) nucleation by bulging, anv -aledlOminat 500 c.
1 4 6
Macroscopic deformation bands connected by microbands are not
associated solely with the splitting of an initial orientation into several
end orientations. At high strain deformation bands also occur in cases
where the initial orientation is stable, e. g. ( Ol ) [21l] or (112) [lll]
in f ace centred cubic materials, but where oscillating slip takes place
in and out of the stable orientation [13] . If, however, the initial
orientation remains pompletely unchanged, as for example in the case
of (001) [llOH in lioji-silicon, the structure is highly uniform as in
Fig. 4 b [26] and no deformation bands are produced.
The structure of very heavily rolled polycrystalline material consists
wholly of microbands [34] . This can be established only by taking
a specimen in the plan N D - RD since superposition of microbands
creates a false impression in the rolling plae. Figure 6. 5 illustrates
such microbands in copper with O, 6% Cr [35] . The multiplicity of
orientations found in wires as a result of rotation about the f ibre axis
can be explained only in terms of a splitting up of the structure into
microbands.
It follows from these results, which do not pretend to give the complete
picture, t hat considerable diversity is possible in the st ruct ure of de-
formed material and henee that recrystallization can proceed f rom
widely differing starting st ruct ures. The effect of t his on nucleation
and on the orientation of the nuclei will be discussed in section 2. 2.
2. THEORY OF ANNEALING TEXTURES
2.1 . Historical Background
In the beginning of t ext ure research, the view was widely held that
nucleation was the controlling factor in texture formation. Above all
BIRGHE [36] and his school [37] andEEHUtCER [38] clai me d t h at
1 4 7
nuclei could be f orme d only in orientatiohs equivalent to the recrystalliz-
ation t ext ure. This is t h eref ore known as the thepry of oriented nucle-
ation .
I n t h e years between about 1950 and 1965, principally as a result of
studies on single crystals and quantitative measurement s of the orient-
ation dependence of grain boundary mobility, attentiori became centred
on t h e growt h stage of primary recrystallization as the t ext ure govern-
ing stage. This development is associated with the ame of BECK[6],
who with his co-workers evolved the it h eory of oriented growth [17] .
According to t h is theory, nuclei are f ormed in random orientations
and selective growth decides which of t h em give rise to the primary
recryst allizat ion texture. Again in the case of grain growth, the
texture is determined by the grains, which, as a result of favourable
orient at ion relationships, grow most rapidly at the expense of the
matrix.
A new turning point was reached with the advent of the electrn micros-
: cope which yielded important data concerning the nucleation stage. They
I showed that, as well as or even instead of growth selection as a con-
sequence of orientation dependent growth rate, the orientation of the
nuclei is of great significance. It must t herefore be decided in each in-
dividual case whet her and to what extent oriented nucleation and oriented
growt h oprate t oget her, This is however not possible f r om texture de-
t f rmi n at i on s alone. As can be seen f r om sections 1. 5 and 2. 2, electrn
microscope investigations are primarily necessary.
Orientation Effects in Nucleation
The principal mechanisms of nucleation [39 - 4l] are strain-induced
grain boundary migrat ion or macro- and micro-bulging, and processes

-
1 4 8
of subgrain growth or coaleacence by the migratlon or elimination
respectively of low-angle grain boundaries known collectively as sub-
grain enlargement. Since these have been discussed in detail in
chapter 3 only their influence on textures will be considerad here.
Strain-induced grain boundary migration, which is mainly discussed
for low to mdium degrees of deformation, was first detected by B E C K
andSE RY [42] and described as the growth of individual grains by the
migration of existing high-angle grain boundaries in lightly def ormed
material. The prerequisites for this mode of growth are a sufficient
difference in orientation between the growing grain and the one to be
consumed, and a higher dislocation density in the latter. The drving
f o rc for this process comes f ro m the diferent dislocation density in
the two grains.
ESECK [43]has described macroscopic bulglng interms of a model
which has been independently substantiated [44] by energetic consider-
ations. The "nuclei" for this high-angle migration are consideredto
be tongue-like subgrains of the growing grain which project themselves
into the matrix (chapter 3 Fig. 3.16). This is doubtlessly a case of
oriented nucleation.
The mechanism described involves high-angle grain boundaries which
are no longer found at high degrees of def ormation (see section 1. 5.).
In this case, instead of macroscopic bulging, nucleation occurs by sub-
grain enlargement. Subgrain growth, that is the enlargement of sub-
grains by the migration of low-angle grain boundaries, can look very
similar to macro-bulging, particularly in the presence of second phases
[45] as shown in Fig. 6. 5 b by a specimen of copper-chromium [35]
in which the bulge extends f ro m two particles at the subgrain boundary.
Finally Fig. 3.11 (chapter 3) shows a schematic representaron of
subgrain growth as f irst observed experimentally by WALTE RandK OC H
[25] in the transition regin (microbands) between macroscopic de-
1 4 9
f o rmatio n bands of a rolled iron-silicon single crystal. Here again
a low-angle grain boundary is not displaced along its whole length but
bows into the neighbouring subgrain.
Figure 6. 6 shows a nucleus f o rmed within the microband structure
of rolled polycrystalline copper [46] . The formation of this nucleus
is attributedto subgrain coalescence but the location of the nuclei with-
in the microbands is the same as in the above mentioned case of
nucleation by subgrain growth. If no microbands are present in the
structure, recrystallization should not occur and the deformation
energy should be dissipated only by subgrain formation and enlarge-
ment, as already found for the above-mentioned iron-silicon single
crystal with a stable (001) [lio] orientation [26]. This necessitates
extremely careful specimen preparation in order to avoid artificial
nucleation.
Whether the subgrains are enlarged by growth or by coalescence as
first discovered by HU[26, 46] is the subject of continuing dis-
cussion. This is important for the interpretation of annealing textures
in as far as subgrain growth is a mechanism of oriented nucleation
whereas coalescence can give rise to subgrain rotation. The conse-
quent change in orientation of the nucleus is not however arbitrary.
An orientation relationship to the matrix exists consisting, in the
case of polycrystalline copper, of a rotation about the<110> direction
of the f avo ured slip system [46] . The result of this is that although
random nucleation is no longer possible, a broad band of differently
oriented nuclei is formed as a consequence of the extended scatter of
the def o rmatio n texture [40, 46] .
It should be mentioned at this point that as early as 1931 W. G.B URGE RS
[47 ] had found an orientation relationship between deformed and re-
crystallized aluminium single crystals closely associated with the
150
Fig. 6.6
Electron micrograph of rolled and annealed
electrolytic cgpper (rolling reduction 99, 5%,
25 min at 100 C). Plae o picture ND-RD (after H U ).
1 5 1
deformation mechanism. This od model difers from the newer one
in that nucleation only in the most heavily deformed and therefore
uniformly oriented lattice regions, i. e. oriented nucleation, was
assumed.
Nucleation in the most heavily strained, uniform oriented parts of the
deformed matrix is also the principie of oriented nucleation by sub-
grain growth which was postulated for example for the transition
regions between macroscopic deformation bands in iron-silicon
single crystals with an initial orientation of (001) [lio] [25] . If,
however, more complicated changes in orientation and henee a more
complex structure are produced in single crystals by deformation,
or if the whole structure is divided into microbands as Is usually the
case in polycrystalline material, it is no longer possible to assign
the same orientation to the most heavily deformed regions of the
matrix that is the nucleation points. In this case the nuclei formed
by subgrain growth must also exhibit widely varying orientations.
The extent to which the orientation of the nuclei can be determined by
selected-area diffraction has been studied and discussed in detail for
the cube recrystallization texture [17, 46, 48, 49] . This texture,
which arises in f. c, c. materials from the deformation texture shown
in Fig. 6.1, can be seen in Fig. 6. 7. Up to now it has not been
possible to detect cube nuclei in deformed copper specimens taken
parallel to the rolling plae [48, 49] by selected-area diffraction.
Even thorough examnation throughout the thickness of 80 to 95%
deformed copper sheet in the plae N D - R D yielded a negative result
[46] . A consideration of the number of nuclei based on the final
grain size, which in the case of the cube texture is usually 10 um
(and more), shows that so few nuclei need to be present that even in
a survey of almost 1000 points there might not be a single one in the
cube orientation [17] .
152
Nevertheless nucleation in preformed cube subgrains must be excluded
on other grounds [46] . If a systematic search for the cube orientatior
o
in the regin of the microbands, which have a thickness of 500 to 1000 A
(95% deformation) in the plae ND-RD, yields a negative result it
means that possibly existing cube orientations must be limited to very
small regions. It is however incomprehensible on energetic grounds
that just these very small regions should be capable of growth at the
expense of much larger microbands. This finding supports the view
that the nuclei forming in heavily deformed, polycrystalline sheet have
not only the orientations of the subsequently developing texture but
other orientations also, and that nucleation must be succeeded by growtt
selection.
2. 3. Examples of the Interpretaron of A nnealing Textures
2. 3.1. Complete or Partial Retention of the Rolling Texture in f. c. c. Materials
(Recovery and Recrystallization): The best known example is that of
aluminium, which has the rolling texture shown in Fig. 6.1. A lthough
the cube texture (Fig. 6. 7 a) can develop from this texture it appears
considerably less often in aluminium than for example in copper and is
frequently accompanied by residual rolling texture, which can also be
seen in Fig. 6. 7 a. It is even more apparent in Fig. 6. 7 b which
shows a mixed texture typical of aluminium comprising almost equal
parts of retained rolling and cube texture.
The situation in the case of aluminium is particularly complicated in
that the retention of the rolling texture with simultaneous softening
can be attributed [57] as well to subgrain formation and enlargement,
that is to recovery [50- 53], as to primary recrystallizatkm [52,
54 - 56] . The two cases can be clearly differentiated by a specific
etch [50, 54] . This precludes the view that retention of the rolling
153
* TD
Figs. 6.7 a to d
A nnealing textures in aluminium sheet; quantitative{lll}p<?le figures:
(a) cube texture in 99,99% A l (continuous casting, cold rolling re-
duction 98, 8%, annealed 5min at 400C)(after GREWEN).
(b) mixed texture consisting of rolling and cube textures (0,15wt. %
Fe, O, 34 wt. % Si). Continuous casting in 8 mm, hot rolled start-
ing at 520 C. cold rolled to 0. 5mm, annealed 3 h at 400 C
(after GREWEN and v. HEI MENDA HL).
(c) rolling texture (O, 54 wt. % Fe, O, 035wt. % Si). Pretreated as in
b, only annealed 40 h at 250 C (after GREWEN and v. HEIMENDA HL).
(d) orientation relationship between two components of the rolling
tex ture ( 7 12 22) [845] near (123) [412j*and a component of
the rolling texture and the cube texture (001) [100] &; pole common
to both components of the rolling tex tureA fafter HU and BECK),
154
texture must always imply recovery and that only a change must mean
recrystallization. In fact only retention of the rolling texture accom-
panied by subgrain formation and enlargement precludes the possibility
of recrystallization. Recrystallization with the retention of a copper-
type rolling texture or a very similar orientation always occurs U the
conditions are unfavourable for the formation of the actual cube re-
crystallization texture. The interpretation of the recrystallization
rolling texture formed together with the cube texture is so closely
connected with this cube texture that the production of both textures
will be discussed together in the next section.
Figure 6, 7 c shows an example of a rolling texture retained by means
of recovery [51] . This {lll}pole figure was obtained from commer-
cially pur alurainium with O, 5wt. % iron and O, 035wt. % Silicon hot
rolled at 500C and then 95% cold worked. Atthough the hardness
reached a mnimum valu after a 40 hour anneal at 250 C, the texture
remained unchanged from that in the deformed state. The results of
the texture determination were corroborated by parallel studies with
the electrn microscope which revealed only continuous growth of
cells formed during deformation. The rolling texture was sharpened
by an anneal at 400C subsequent to the 40 hour anneal at 250 C. This
could be attributed to grain growth, which, according to the theory of
inevitably leads to texture sharpening.
Softening with simultaneous retention of the rolling texture is not al-
ways observed separately from sharpening as in the preceding example.
Recovery and recrystallization can occur in rapid succession or simul-
taneously, which can complcate the assessment quite severely. It
is therefore difficult to demnstrate experimentally the extent to which
recovery can be associated with a sharpening of the rolling texture.
Considerable evidence supports the view that sharpening can occur
by subgrain coalescence but not by subgrain growth [26] .
155
In the case of aluminium an important prerequisite for softening by
recovery is a high degree of cold working leading to a well developed
cell structure. There is some evidence that this is not the only con-
dition, for example, the observation of recrystallization in the pre-
sence of an excess of Silicon at a high degree of cold working. Only
in the presence of excess iron can recovery-softening occur with the
degrees of deformation employed in practice [50, 51, 53, 57, 59].
The solubility of iron in aluminium B low (O, 052 wt. % iron in
solution at the eutectic temperature of 655 and only O, 005wt. % at
500 ) but it can exist in supersaturated solution [55] . Since Al-Fe
v
is precipitated at a measurable rate only at temperatures above
300 C [53] whereas softening by recovery has been observed at
lower temperatures [51] , the iron must be effective as segregations
at the subgrain boundaries. Obviously it aggravates the existing
difficulty of nucleation in a cell structure and thereby inhibits re-
crystallization. If the iron is precipitated out by an excess of silicon
[55, 56] the inhibiting effect is lost.
A further example of the retention of the deformation texture on anneal-
ing is provided by aluminium-copper alloys with 5wt. % copper de-
formed 50 and 90% and annealed at 280^;[53, 60]. The retention of
the deformation texture was accompanied by subgrain enlargement
controlled by the growth of 0'particles and their transformation into
9 . If the supersaturation was less (alloys with 2 wt. % copper) re-
crystallization could occur because the subgrain boundaries were less
heavily blocked by particles.
The model for this process is based on the fact that the first step at
the beginning of the anneal is the development of a subgrain structure
[6l] . In many materials a dislocation network of this type can be
produced only at high degrees of deformation as established by studies
156
on a copper-zinc-nickel alloy (60 wt. % Cu, 28 wt. % Zn, 12 wt. % Ni)
and a copper-beryllium alloy with 1, 9 wt. % beryllium [62] . This
dislocation network must subsequently be stabilized by the precipitation
of a suf f iciently large number of particles. If too few particles are
present nucleation takes place by subgrain enlargement as observed
for aluminium 2% copper. Even if the dislocation network is stabilized,
however, discontinuous subgrain growth leading f lnally to nuclei can
occur as a result of the transf ormation of an unstable phase into a
stable phase or by particle coaraening by Osswald ripening, This
situation was observed in copper-nlckel-zinc, but was completely ab-
sent in copper-beryllium as a consequence of the much greater number
of precipitates. Between these two extremes, however, there is the
possibility of continuous subgrain enlargement as in the case of alu-
munium 5% copper (see also chapter 8).
In theearly days of texture research, retention of the deformation
texture on annealing was always interpreted as recrystallization with
oriented nucleation (c.f . [5]). The preceding examples show that
this is no longer the rule. But macroscopic bulging, being a mechanism
of oriented nucleation, represents an exception. It leads to partial re-
tention of the def ormation texture, since existing grains with a low dis-
location conten consume other grains with a dif f erent orientation and
higher dislocation conten. The prerequisite for this is an orientation-
dependent dislocation density. This can arise f rom either orientation-
dependent def ormation behaviour or an orientation-dependent rate of
recovery or recrystallization,
There are certain indications that in the case of single crystals the
tendency to recrystallization is related to the initial orientation prior
to def ormation [26, 63 - 66] . Dif f erences in recrystallization be-
haviour were also observed between cross-rolled and straight-rolled
1 57
material [67]. These results are however open to question, because
systematic studies have not yet been made on single crystals regard-
ing the relationship between the initial orientation bef ore def ormation,
final orientation, dislocation structure in the def ormed condition and
the stored energy of deformation.
The f ormation of the<100>annealing texture in extruded aluminium
wires with<lll> + <100> def ormation texture can be attributed to
bulging [68] because the most important prerequisite, a dif f erent
degree of work hardening in the two def ormation textures, was
established experimentally in both polycrystalline, verypure aluminium
(99,99%, extruded at -190C) [69] and in extruded single crystals
(A l 99, SZ-mounted in polycrystalline aluminium, extruded at 300 C)
[70] . There is considerable evidence that nucleation by bulging is
involved in the case of hot worked material which softens during or
shortly after deformation.
!.2. The Cube Texture in f. c. c. Metals (Primary Recrystallization and
Grain Growth): The cube texture is one of the most interesting and
important annealing textures because it is associated with a large
anisotropy of the mechanical and magnetic properties. It can thus be
both highly undesirable technically (earing c. f . [57] ) and desirable
(e. g. f erro-magnetic iron-nickel alloys). In Fig. 6. 7 a the cube
texture is illustrated in a 99, 99% pur aluminium sheet; Fig. 6. 7 b
shows the cube texture together with sharpened rolling texture in a
less pur material i(99 5% A l). A certain degree of purity is the
f irst requirement for the production of the cube texture. Particles
disturb its development on annealing and thus f avour the sharpened
rolling texture. The f ormation of the cube recrystallization texture
can also be hindered by dissolved alloying elements. The reason for
this, however, is usually a change in the rolling texture to the brass
1 5 8
type, as observed for example in copper-zinc or copper-tin alloys
(c.f. [5] ). The promotion of recovery and henee of the copper type
rolling texture by heavy cold working and iron in aluminium described
in the preceding section is also relevant in this context.
The basis for an interpretaron of the cube texture, the sharpened coppe
type rolling texture and the factors influencing the proportion of each is
provided by the theory of oriented growth [6, 17, 46] . Following on
the formation of a wide range of different orientations by the nucleation
process the cube texture and the sharpened rolling texture develop by
growth selection during the growth phase of primary recrystallization
and subsequent grain coarsening. An orientation relationship giving
mximum growth rate exists between the cube texture and the rolling
texture. Figure 6. 7 c shows the vil} Ples f the cube texture and
of two (of the four possible) components of the rolling texture described
by (7 12 22) [845] near (123) [412] [54] . Rolling and cube textures
each have a {lllj pole in common and can be transformed into one an-
other by a rotation through about 40 . On the other hand however one
{lll} pole each of the two components of the rolling texture also co-
incide (plae ND - RD ) so that a very similar orientation relationship
exists for these. This is the reason why orientations corresponding
to the deformation texture or very similar to it develop together with
the cube texture. Even in those cases where the cube orientation final-
ly becomes the exlusive texture, sharpened rolling texture is still pre-
sent at the end of primary recrystallization, and the more favoured
cube texture increases only by grain growth. This is true of all
materials exhibiting the cube texture: copper, iron-nickel and aluminiuc
In the case of aluminium, however, the further requirements such as
low grain size and only a small quantity of second phases are consider-
ably less often satisfied than for example in the case of copper.
1 5 9
The reason why the cube texture is better capable of growth than the
four components of the rolling texture, which also exhibit favourable
orientation relationships among themselves, has been the subject of
numerous scientific discussions. Since these have been reviewed in
detail by BBCKandHU [l?]it seems superfluous to renew the discussion
here. Only the most important argument in favour of the cube texture
will be mentioned. The cube texture is a so-called compromise texture
because it possesses an orientation relationship giving only approximate-
ly mximum growth rate but with all four components of the rolling
texture [6, 17]. This is not the case among the components of the
rolling texture themselves. In order to gain the growth advantage a
cube grain must if possible have contact with all the components of the
matrix during primary recrystallization or subsequent grain growth.
If a cube grain is surrounded by only one component of the rolling tex-
ture it is by no means certain which regin is capable of growth. The
growth advantage of the cube texture lies in the combination of the cube
nucleus with several grains in different matrix orientations. This re-
quires a fine grained starting material and no second phases to lie be-
tween the cube orientation and the rolling texture and prevent the con-
tact between them, According to the experimental results, which
corrobrate the theory, the fracton of cube texture increases with de-
creasing grain size of the starting material (c.f. [5] ) and,
example in copper-chromium, only very little cube texture,f
if chromium preciptales hinder the growth during grain jnj^rgement
[24, 35] .
\The (225) [733 ] Textura in Brass with 30 wt. % Zinc (P
crystallization): In the interpretation of recrystallization
possibility of the development of recrystallization texture different
rom that exactly derived from the principal orientations of the de-
formation texture, but having a quite favourable orientation
160
relationship as well to the principal orientations as to the scattering
must always be considerad. In other words the entire orientation
spread and not simply the ideal orientations must be taken into account
Like the cube texture, these textures, of which the (225) [ _ 7 3 4 J orienta-
tion, a primary recrystallization texture in brass with 3 0% zinc, is a
good example, count as compromise textures. If only the ideal
orientation of the deformation texture (011)[ 21J were taken into
account, a recrystallization texture (121) [ lOl] would be expected.
In its place however (225) [734] is found which possesses a quite
favourable orientation relationship with not only the two components
of the rolling texture but also with the orientation spread (Further to
compromise textures see also steel section 2. 3 . 4 . and chapter 10
section 3. 3 .).
2. 3 .4 . The (111) [lio] Texture in Steels (Primary Recrystallization or Grain
Growth): Whereas the textures discussed above are examples of
oriented growth there are some found in steels, in addition to those
similar to the above, for which oriented nucleation is indicated. In
the case of steel there is a technically very important texture , parti-
cularly suitable for deep drawing because the energy requirement is
considerably less than for other textures [71, 7 2] . This texture is
describedby the ideal orientation (111) [ lio] [73, 7 4 ] . The parti-
cular suitability for deep drawing due to the texture can be determined
from the r-value, which is the ratio of the change in width to the change
in thickness in a tensile test on fat specimens, for example at 20%
extensin (c.f. e.g. [71, 7 2, 7 5J ), An r-value of 1means that there
is no normal anisotropy. r-values greater or less than 1arise as a
result of normal anisotropy, which for r-values greater than 1 is
associated with an improvement in the deep drawing properties.
Planar anisotropy of the r-value, that is a variation in r as a function
of the angle between the longitudinal axis of the tensile specimen and
RD, always leads to earing. A combination of a high r-value and a
161
low planar anisotropy is therefore desirable (c.f. chapter 10sec. 2. 2.1.)
A complete absence of planar anisotropy occurs only in textures which
are rotationally symmetric about the normal to the sheet, that is in
fibre textures with the fibre axis normal to the rolling plae.
The (111) [ lio] texture does possess a certain surface anisotropy but -
it is possible by the observance of specific conditions in the production
of the texture to produce a mean r-value (r-value)of 2 [76-79] . K
has also become possible recently to reduce the surface anisotropy
considerably [80] . These improvements are the result of specific
research on the manufacture of advantageous textures, known as
texture tailoring [8l] . It will be shown in the followlng that there
exist several different ways of achieving a high mean r-value.
Figure 6. 8a shows the {lll}pole-figure of a rolled carbonyl iron
sheet. The ideal orientations for this texture can be derived by a
rotatlon about the [ llo] rolling direction and the most important are
(001), (112) and (111) [ lio]. (111) [21l] is also given [73, 74] .
A description of the texture as two<110>fibre textures has also proved
helpful [82] . The fibre axis of one fibre texture lies along the RD,
and of the other in the plae ND-RD. As mentioned earlier, however,
a model based on fibre textures is too complicated for the interpretation
of recrystallization textures so that discrete ideal orientations, or at
best the scatter regions around the [ IlOj rolling direction [83], are
always used.
The (111) [ lio] texture in Fig. 6. 8 b can be formed from the rolling
texture illustrated in Fig. 6. 8 a [77] . Frequently, however, depend-
ing on the degree of deformation and the annealing temperature other
textures arise in addition to or instead of (111) [ lio] including a
texture particularly unsuitable for deep drawing (001)[ llOj . Figure
6. 9 shows pole figures of steels with high and low r-values (1,8 or 1,1)
162
Figs. 6.8 aandb
(a) Quantitative 100 pole figure of the rolling texture
of carbonyl iron (99, 7% deformed) (after Liesner
and Wahl).
fb) Quantitative incompleto 100 pole figure of the
annealing texture of a rimming steel with a finely
dispersed second phase, 80% cola rolled. anneal-
ed 64 h at 780 C (after HUTCHISON, SMITH, WATSON
and DILLAMORE).
1 6 3
from which it can be seen that very slight changes in the texture can
cause considerable deterioration in the r-value [84] .
For this reason a more sensitive method of texture determination is
often chosen by which the intensities of the crystallographic planes ly-
ing in the plae of the sheet can be determined directly. This does not
however take into account the crystallographic direction parallel to the
RD. Figure 6.10 [84] shows the results of such measurements from
which it is clear that the texture differences between the material with
good and poor r-values involve differences in the fractions of (111) and
(100).
Since the (111) [llOj texture is associated with elongated fat grains
(pancake structure) it can be recognised in a micrograph. Globular
grains, like those associated with a texture having a poor r-value,
would be considerably more advantageous for deep drawing.
The oldest method of producing (111) [lio] starts from Al-killed
steel. An important prerequisite is the precipitation of AIN after de-
formation [76, 83, 84, 85, 86, 87] . This necessitates a slow rate of
heating ( 25 C/h; box annealing) or preliminary annealing at lower
temperatures (500C, 100C/h from 565 to 700C(according to [76])).
Rimming steel or Al-killed steel in which the AIN has been precipitated
before deformation or which has been heated up too rapidly does not
show the advantageous texture. It is also absent in Al-killed material
which has been deformed more than 70 to 80% by rolling and has henee
developed a different surface texture [88] . Certan vales for the hot
rolling temperature preceding the cold working are also required
(c.f. [78] ).
A very important observation is that in specimens with the (111) [lOJ
texture and the associated structure, AIN cannot be detected in the
*
.

n

2
-

o

T
O

T
3

W
H

1
3

3
.

<
D

3
5

n

t
u
s
-
f
f
c
z
.
s

<
*

i
i
K

o
.
n

_
u
(T
I

'"
"
'
O

1

5
"

.H

E
l
^

u

-
^

o

2
.

<
*

r
e
G

s
*
*
-

w
4
T
&

^
S
-
g

?

n

Z

-
f
t

=

o

S
o

n
-
)

3 T
3
0

=
T
2

3

^

O

-
3
s
r
1

a

o
-
r
e

r

n
^

o
-
^

r
n
n

o
1
"
i

i
^
3

m

9
.
0
u

o
S

S
N

T
r
o

o
O
.

i
'
a
*
a
a

i

S
-
i
>

i
>
-

O
s

-
*

g

?
3

n
I
I
a
|
2
^
0

^

[
U
C

3
-
1

g
..ro

fD
e
n
o u a c
r

5

5
|

F
r

&
5
'
^
o
<
r
t
r
a

r
e

-

p
r
a
3

2

c

-
a

S
"
^
i

(
t>

-
a
c
j
-

H

^

o
H

H

f
*
"

^
o
q

^
-

t
u
a

S

~
=
r

^

i
;

a

i
*
'
c
>
-
i.

h
-
,
3

^

E

m
p

c
o

3

w
^

s
^
o
^

^

p
m
8
?
J
-
l
C
o

i
O

L
O
*

1
N

i

H
l
G
H

I
N
*
}
i
6
h
r

A
T

T
E
M
P

O

L
O
*

A
1
N
A

H
l
f
.
H

4
1
N
*
]
i
6
h
r

A
T

T
E
M
P
0
5
U
l
1
0
0
0

1
0
2
5

1
0
5
0

1
0
7
5

H
O
O

1
2
0
0

i
3
0
0
T
E
M
P
E
R
S
T
U
R
E

f
9
0
0

iO
O
O

i0
2
5

(
0
5
0

1
0
7
5

1
1
0
0
T
E
M
P
E
R
A
T
U
R
E

*
F
1
2
0
0

(
3
0
0
F
i
g
.

6
.
1
0
I
n
f
l
u
e
n
c
e

o
f

t
h
e

a
n
n
e
a
l
i
n
g

t
e
m
p
e
r
a
t
u
r
e

o
n

t
h
e

r
e
l
a
t
i
v
e
i
n
t
e
n
s
i
t
i
e
s

o
f

t
h
e

(
0
0
1
)
,

(
1
1
2
)

a
n
d

(
1
1
1
)

p
l
a
n
e
s

i
n

t
h
e
r
o
l
l
i
n
g

p
l
a

e
.

M
a
t
e
r
i
a
l

a
s

i
n

F
i
g
.

6
.

9

,

r
a
t
e

o
f

h
e
a
t
-
i
n
g

4
,
4

C
/
h
.

L
o
w

A
1
N

a
s

i
n

F

E
.

6
.

9

a
,

h
i
g
h

A
1
N

a
s
i
n

F
i
g
.

6
.

9

b

(
a
f
t
e
r

M
I
C
H
A
L
A
K

a
n
d

S
C
H
O
O
N
E
)
.
1 6 6
newly recrystaUised state [76, 84, 87, 89] , and that U A1N is pre-
cipitated before deformation it does not hinder the growth of the re-
crystallized grains with the result that largely globular grains are
produced [84] . This indicates that the pancake structure and its
accompanying (111) [lio] texture are not, as previously assumed
[90, 9l] , causedby particles formed during tempering or preliminar?
annealing. MKHALfl and SODCNE [84] in particular were able to show
that subgrain enlargement is strongly impeded by clusters o aluminiura
and nitrogen but not by A1N particles [83, 84] . As indicated by the
later formed preciptales, the clusters lie preferentially at subgrain
boundaries. Whereas the first stage of recovery, the formation of
subgrains, proceeds similarly in material with and without nitrogen
precipitates, disturbances arise in the second stage of the nucleation
phase, subgrain growth, which lead to the formation of a considerably
smaller number of nuclei in the material without A1N precipitates [84].
Important in this context is an experiment on the tendency to recrystallii
ation of rolled single crystals with different initial orientations, in whirt
it was established that the (001) orientation recrystallized at a slower
rate than others [65] . This would indicate that t is the (100) grains
which are disturbed in the nucleation phase and that for this reason
(111) is formed predominantly during the anneal [84].
DILLAMORE, SMITH and WATSON [83] do not interpret the differencei
quite the same way but they do also believe that they orig nate in the
nucleation phase. Since, according to their view, (001)| _ lOj is the
only texture formed exclusively by oriented nucleation, the others be-
ing formed by both oriented nucleation and oriented growth, the
disturbing influence occuring during nucleation must act only on (001)
[lio] [86].
1 6 7
This statement is based oi\ consideration of a number of important
aspects associated with oriented growth. First of these is that nuclei
frequently form at grain boundaries and must therefore be capable from
the outset of growing into two different orientations [92, 93] . With
advancing growth the orientation relationships of the nuclei to the
orientation spread in the matrix become important, as already describ-
ed in the case of the compromise textures (see sections 2. 3. 2. and
2. 3. 3.).
Besides the orientation dependence of the grain boundary mobility it
must further be considered that the stored energy and henee also the
driving forc can be orientation dependent. This is indicated by the
already mentioned orientation dependence of the recrystallization
tendency in single crystals. Calculations for steel with deformation
textures derived from a rotation about the [llOj rolling direction re-
vealed the increase in stored energy from (001) to (110) shown in
Fig. 6.11.
Fig. 6.11
Orientation dependence of stored energy (after DILLAMORE, SMITH
and WATSON).
1 6 8
It must further be considered that the fraction of a recrystallization
texture produced by oriented growth is also determined by the quantity
of material in the deformation texture offering maximuin growth rate
possibilities for the nuclei. If only a small fraction of the deformed
material has a favourable orientation relationship with a particular
micleus orientation, this orientation will not be very strongly represent-
ed in the recrystallization texture. As shown by studies of iron 3%
silicon single crystals [94] a high growth rate is associated with a
27 rotation about a<110>direction. Applying these considerations
to steel textures it can be seen that nuclei in the orientation range i
Fig. 13 derived by a rotation about the [lio] rolling direction, lying
about 27 from both (001) and (110), have the greatest chance of
dominating by oriented growth. The driving forc is particularly
large (see Fig. 6.11) for nuclei with an orientation relationship to
(110), but to nuclei at 27 from (001) there is a relatively large fraction
of deformation texture availabie, in which (001) [lio] is very strongly
represented. (111) [Ho] also lies in the range of orientations of
type (hhl) [lio] derived in this way. (001) [lio] however cannot
be formed in this manner.
The theory of oriented growth is well founded in the case of steels in as
far as all the orientations of the recrystallized material are already
availabie inthe deformation texture. Moreover, energetic consider-
ations of the nucleation at grain boundaries by bulging or subgrain en-
largement within the grains render plausible oriented nucleation in
the same orientations as result from oriented growth. But (001) [l
can also be formed by oriented nucleation if no disturbances arise in
the growth stage of the subgrains.
It has recently also been possible to produce a texture with a high
r -valu of 2,1 and a low surface anisotropy in rimmed steel by com-
bining a strong < 111> fibre texture (fibre axis parallel to BN) with
169
Fig. 6.12
Quantitative incomplete^lOOJ-pole figure of rimming
steel with O, 063 wt. % C and O, 05 wt. % Mn. Final
hot rolling temperature 925 C, cold rolling reduction
75%, rate of heatinc in the final anneal 20 C/h,
annealed 20 h at 710 C in dry hydrogen (after HU and
GOODMAN).
1 . 5
O
05
OJ 04
wt %Mn
Fig. 6.13
r -valu of steels as a ' on of the manganeso con-
ten (after HU and C' N).
1 7 O
with (111) [IlO]f B O]. Fi gureB . 12 illustrates the -OOll pole f i gure
of such a steel. In addition to certain requi rement s, concerning the
hot rolling temperature [78] and the cold rolling reduction, a prere-
quisite f or the f ormation of this texture is a very low manganese con-
tent. The steel in Fig. 6. 12 contained O, 063 wt. % carbn and only
O, 05 wt. % manganese (vacuum melted). Figure 6. 13 shows the improve-
ment in the r- valu achieved by reducing the manganese conten f rom
O, 56 to O, 05 wt, %. Reduction of the manganese conten to vales rauch
less than are otherwise common f or steel ef f ects a pref erential re-
covery and an acceleration In the recrystallization of certain orient-
ations. Although he actual mechanism has not yet been elucidated,
data on the recovery and recrystallization rates suggest that oriented
(111) nucleation occurs. It is interesting to f i nd that oriented nucle-
ation is once again being considered important.
One technique f or producing a texture suitable f or deep drawing in
steels wi t h a normal manganese conten consists of removing a f inely
dispersed second phase during grain growth subsequent to pri mary re-
crystallization [77, 79, 95 - 97] . This is usually associated with an
increase in grain size, so that technically those manuf act uri ng processes
using pri mary recrystallization to produce (111) should be given pref er-
ence. From the scientif ic aspect, invesigations of the inf luence of
second phases such as carbides (cementite), nitrides or extremely
f inely dispersed austenite, are of such interest that an example will be
described in the f ollowing.
Hypoeutectic steel with O, 04 wt. % carbn is annealed at 920 C af ter
hot rolling and then quenched in oil to produce a f i ne dispersin of
pearlite [77] . Af ter cold rolling of about 70 to 85% the material is
annealed at 780C in an atmosphere of purif ied and dri ed hydrogen-
nitrogen. No only does the ransf ormaion into ausenite occur, whi ch
initially constitutes about 6% of the structure, but a slow decarburis-
ation also taf ees place by whi ch the austenite is removed.
1 7 1
In this material, primary recrystallization yields both (111) and
(001). If inhibition of the grain growth is prevented by eliminating the
austenite, pref erenlial growth of the (111) grains ensues and the (001)
grains are consumed. This is possible only because the (111) grains
whi nh f ormed earlier are larger than the (001) grains at the end of
pri mary recrystallization. Size is not the sol reason why the (111)
grains are more capable of growth than he matrix. They are also
better able to overeme any inhibiting ef f ects still remaining. The
basis f or this interpretation is he heory of grain growh according
toHElEHT[98] andthe T E N E R equation [23] dseribing the relation-
ship between driving f orc and particle drag (c. f . chapter 10 section
3. 2. ).
1 5 . Secondary Recrystallization with<10Q>and<lll)Orientation Relation-
ships in a Matrix with a Single Orientation: Secondary recrystalliz-
ation can occur only when pri mary recrystallization is complete and
normal grain growth is not possible or at a standstill (matrix stabiliz-
ation). Furthermore, nuclei are necessary with an orientation
dif f erent f rom he exture of the matrix and a size above a critical
valu. Since these relationships have been discussed in detail else-
where [4, 99] and also in chapter 5, only an example of secondary
recrystallization will be considered here. Texture governed secondary
recrystallization where particular orienaion relationships exist be-
tween nucleus and matrix is iteresting in wo respecs. No only does
it yield pronounced secondary recrystallization textures by oriented
growth but it is only possible at all because the strong, single-crysal-
like exure has a stabilizing ef f ect on he marix (matrix stabilization
by t ext ure inhibition). In texures with only one component the
orientation dif f erence is so small that normal grain growth is im-
possible. Hegions with a deviating orientation generally present in
every material nevertheless f inally lead to secondary recrystallization.
H
I
172
There are however also cases in which as a result of an extremely
sharp texture and the absence of nuclei secondary recrystallization
does not occur even at very high temperatures [6, 100, 101] .
Apart from deformed and annealed single crystals, the cube texture
in copper [102], iron-nickel [103] or 18/8 steel [104] is the best
known example o a matrix in which secondary recrystallization with
<111> or <100)rotational orientation relationships can occur [105] .
The reason why both <111> and <100>are possible axes for the
orientation relationship has been investigated in detail in 18/8 steel.
In this material a copper-type rolling texture can be produced by
rolling at 800C and annealing at 800 to 1000C from which the cube
texture subsequently forms [106] . Below 200C this steel has a hrass
type rolling texture. By means of a special roll-anneal programme
an extremely sharp cube texture could be produced illustrated in
Fig. 6.14 a. This transformed at 1000 (Fig. 6.14 b) into the (025)
[100] texture by a rotation about the [100] rolling direction and at
1200C into the texture shown in Fig. 6.14 c in which almost equal
fractions of (025) TlOOl and (123) 412 1 exist side by side. The
O
latter is related to the cube texture by a 40 rotation about <111>
(Fig. 6. 7 c ) and is identical with the principal orientation of the roll-
ing texture. In the case of other materials and textures it is also
common to observe the same texture in both the deformed and second-
ary recrystallized states.
Figure 6.15 shows a schematic representation of the different and
furthermore temperature-dependent growth rates for the <111> and
<100> orientation relationships, which explain these results. Where-
as at low temperatures the activation energy is lower and henee the
growth rate higher for the <100>than for the< lll>orientation re-
lationship the situation is probably exactly the reverse at high
173
Figs. 6.14 a to c
Quantitative incomplete {lll}pole figures of annealing textures
of 18-8 Ni-Cr steel. 90% rollin* reduction at 800C, sandwich
assembly (after GOODMAN and HU):
(a) 30minlOOO"C ((001) [100] , (122) [212] ).
(b) 96 h 1000 C ((001) Cl O O ] , A (025) [100] ).
(c) 2h 1200 C (A-(123) [412] , A (025) [100] ).
1 7 4
temperaturas. The mutual grain boundary between the growing grain
and the one it consumes differs according to the orientation relation-
ship between the grains (c.f. chapter 4 section 4. 5.) wlth the conse-
quence that at any given time both grain boundary types are inter-
acting with diferent impurities. Since the type and quantity of the
segregated impurities can vary according to the annealing temperature,
the growth rate is temperature-dependent.
2. 3. 6. Tertiary recrystallization: Ais already mentioned, tertiary recrystalliz-
ation is the growth of small grains in a favourable orientation at the
expense of a matrix consisting of much larger grains [4, 107]. The
driving forc for this process is the difference in surface energies.
A well known example of tertiary recrystallization is the transform-
ationof (001) [100] into (Ol) [lOO] in thin sheets (O, 3 and 0,15 mm)
of very pur iron-silicon (3 wt. % Si, O, 005 wt. % impurities). In
this material a (Ol) [lOO] texture with strong scatter about RD s
formed by primary recrystallization and subsequent grain growth from
a deformation texture with a strong (111) [211 ] orientation. The
material was annealed in vacuo (5.10'4 mm Hg). During a prolonged
anneal, for example for 5 hours at llOO^large grains in the cube
texture form from (Ol). These are however interspersed with sorae
( Ol ) grains smaller than those of the matrix. If the anneal is pro-
longed even further (e. g. 16 hours at 1200C)the (Ol) grains once
again become capable of growth at the expense of the (001) matrix
and (Ol) [I00]makes a renewed appearance.
Secondarv recrystalliz ation occurs in the cube texture because the sur-
face energy of the (001) grains is less than that of the (Ol) grains. Dur
ing the prolonged anneal it is possible that the situation becomes re-
versed and the surface energy of the (Ol) grains has the lower valu.
This could be the reason for the renewed growth of (Ol).
1 75
fc

t
I
5

3
m>
<;oo>
RECIPROCAL OF ABSOLUTE TEMPERATURE
Fig. 6.15
Schematic representation of the temperature
dependence of the growth rate for grains with
a <Tlll> and <O> orientation relationship
to the matrix.
176
REFERENCES
1. R.W. Cahn:
2. S. Hofmann and
F. HaeBner:
3. K. Detert:
4. C.G.Dunnand
J. L. Walter in:
5. G, Wassermann and
J. Grewen:
6. P. A. Beck:
7. U. Kbster:
8. J. Grewen.
A. Segmller and
G. Wassermann:
9. H. J. Bunge:
10. R.J.Roe:
11. G.Ibeand
K. Lcke:
12. H.Hu,
P. R. Sperry and
P. A. Beck:
13. W. Heye:
14. H. J, Bunge and
F.Haefiner:
15. J. Grewen and
G. Wassermann:
16. I. L.Dillamore and
W. T. Roberts:
17. P. A. Beck and
H. Hu in:
18. J. Grewen and
G. Wassermann:
This volume chapter 3.
This volume chapter 4 (see also Lit.
res.).
This volume chapter 5.
"Recrystallization, Grain Growth and
Textures", H. Margolin, Ed., A.S.M.,
Metals ParkOhio 1966, p. 461.
"Texturenmetallscher Werkstoffe".
Springer-Verlag, Berlin/Gottingen/
Heidelberg 1962, 2nd edition.
Phil. Mag. Suppl. 3 (1954) 245.
This volume chapter 8.
Aren. Eisenhttenw. 29 (1958) 114.
"Mathematische Methoden zur Textur-
analyse", Akademie-Verlag, Berln 191t
J. Appl. Phys. 36(1965) 2024.
probably Acta Met. and Met. Trans. 2
(1971).
Trans. AIME 194 (1952) 76.
Forschungsberichte aus dem Institu
fr Metallkunde und Metallphysik der
TU Clausthal, 3 and 4 (1969).
J. Appl. Phys. 39 (1968) 5503.
Acta Met. 3 (1955) 354 - G. Was ser ruar:
Z.f . Metallkde. j>4 (1963) 61.
Met.Rev. 10 (1965) 271.
"Recrystallization ..." (Re. 4),p. 393.
Ed., "Texturen in Forschung und Prai;'
Springer-Verlag, Berlin/Heidelberg/Nn
York 1969.
177
19. G.Ibeand
K. Lcke in:
20. B. B. Rath and
H.Hu:
21. K.T.Aust in:
22. K. Lcke in:
23. C. Zener, quoted in:
24. J. Grewen and
W. Leo:
25. J. L. Walter and
E.F.Koch:
26. H. Hu in:
27. C. S. Barrett and
T. B. Massalski:
28. W. Hueck and
G. Wassermann:
29. H. Ahlborn:
30.
31.
32.
33.
W. Heye and
P. Sattler:
H. Ahlborn and
D.Sauer:
D.Sauer:
P. R. Swann in:
34. H. Hu:
35. D. Sauer:
36. W.G.Burgers in:
37.
38. U.Dehlinger:
"Recrystallization ..." (Ref. 4), p. 434.
Trans. Met. Soc. AIME 236(1966) 1193;
245 (1969) 1243, 1577.
C.f, Ref. [18], p. 160.
"Metallphysik'1, VdEh, Ed., Verlag
Stahleisen, Dusseldorf 1967, p. 182.
C. S. Smith, Trans. AIME 175 (1948) 15.
Z. f. Metallkde: ^9 (1958) 770.
Acta Met. 10 (1962) 1059; U. (1963) 923;
(DiscussionTo 26).
"Recovery and Recrystallization of
Metals", L. Himmel, Ed., Interscience
Publ., New York 1964, p. 311.
C.f.: "Structure of Metals", Mc.Graw-
Hill, New York 1966, p.551, 3rd edition.
Z. f. Metallkde. 54 (1963) 32.
Z.f. Metallkde. 56_ (1965) 205 and 411;
57_ (1963) 32.
Z.f . Metallkde. shortly.
Z.f . Metallkde. 59^(1958)658.
unpublished results.
"Electron Microscopy and Strength of
Crystals", G. Thomas and J. Washburn,
Ed., Interscience Publ., New York
1963, p. 131.
Z.f. Metallkde. 6 (1969) 69.
Dissertation, TU Clausthal 1971.
Handbuch der Metallphysik I/2, G.
Masing, Ed., Akademische Verlagsge-
sellschaft, Leipzig 1941.
C.f. also [5], p.491.
Metallw. 7(1928)1172- Z. f. Metallkde.
52 (1961) 44.
39. R. W. Cahn in:
40. R. W. Cahn:
41. R.W. Cahnin:
42. P. A. Beck and
P. R. Sperry:
43. P. A. Beck in:
44. J. E. Bailey and
P. B. Hirsch:
45. R. D. Doherty and
J. W. Martin:
46. H. Hu in:
47. W. G. Burgers and
P. C. Louwerse:
W. G. Burgers and
T. J.Tiedema:
48. F.Haefiner,
U. Jakubowsky and
M. Wilkens:
49. H. Perlwitz,
K. Lcke and
W. Pitsch:
50. P. A. Beck in:
51. J. Grewen and
M. v. Heimendahl:
52. J. Grewen:
53. E. Hornbogen and
H. Kreye in:
54. P. A. Beck and
H. Hu:
55. W. Bunk:
56. W. Normann:
1 7 8
"Recrystallization ..." (Ref. 4), p. 99.
"Physical Metallurgy", North Holland PubL
Comp. , Amsterdam 1965, p. 925.
[18], p.150.
J. Appl. Phys. 21 (1950) 150.
"Metal Interfaces", A. S. M. , Metals
Park Ohio 1952, p. 208.
Proc. Roy. Soc. A 267 (1962) 11.
J. Inst. Metals 91 (1963) 332.
[18], p.200.
Z.Phys. 67(1931)605.
Proc. nederl. Akad. Wetensch. 53 (1950}
1525.
C.f. e.g. : Mat. Sci. Eng. 1 (1966) 30.
C. f. e.g.: Acta Met. 17 (1969) 1183.
D. Altenpohl, "Aluminium und AluminiuD
legierungen", Springer-Verlag, Berln/
G5ttingen/Heidelberg/New York 1965,
p.410.
Z. f. Metallkde. j > 9 (1968) 205.
Z. f. Metallkde. 59 (1968) 236.
[18], p. 274.
Trans. AIME 194 (1952) 83.
C. f. e.g.: Z. f. Metallkde. 56(1965)645:
57 (1966) 345.
C. f. e. g. : Z. f. Metallkde. 58(1967)151-
in[18], p. 361.
57. J. Grewen:
58.
59. I. Miki and
H. Warlimont:
60. H. Ahlborn,
E. Hornbogen and
U. Kij ster:
U. Kbster and
E. Hornbogen:
61. H. Kreye:
62. H. Kreye and
G. Y. Chin:
63. P. K. Koh and
C. G. Dunn:
64. H. Hu:
65. W. R. Hibbard and
W. R. Tully:
66.
67. H. Hu:
68. J. Grewen:
69. R. A. Vandermeer and
C. J. McHargue:
70. H. Ahlborn,
G. Wassermann and
S. Wiesner-Kaup:
71. W. T. Lankford,
S. C. Snyder and
J. A. Bauscher:
72. R. L. Whiteley in:
73.
74. J. Grewen in:
< w
1 7 9
Forschungsberichte aus dem Institu
fr Metallkunde und Metallphysik der
TU Clausthal 2 (1967).
Ref. 4, p. 507.
Z. f. Metallkde. 59 (1968) 254.
J. Mat. Sci. 4 (1969) 944.
Z.f. Metallkde. 59 (1968)792.
-
: J
Z. f. Metallkde. 61 (1970)108.
"Second Intern. Conf. on the Strength of
Metals and Alloys", Conf. Preprints,
1970.
Trans. AIME 203 (1955) 401; 206 (1956)
1017.
Trans. Met. Soc. AIME 215 (1959) 320.
Trans. Met. Soc. AIME 221 (1961) 336.
C.f. also[5], p. 466.
Trans. AIME 209 (1957) 1164.
Z. f. Metallkde. 57 (1966) 581.
Trans. Met. Soc. AIME 230 (1964) 667.
Z.f. Metallkde. 57(1966)22.
Trans. ASM 42 (1950) 1197.
W. A. Backofen et al Ed. , 9th Sagamore
Conf. , AMRA, "Fundamental of Deform-
ation Processing", Univ. Press, Syracuse
1964, p. 183.
C.f. also[5], p. 354.
Ref. 5, Stahl und Eisen 89 (1969) 1018
and 1071.
180
75. J. Grewen and
"Mechanische Umformtechnik"I O. K ienz l cj
181
G. Wassermann in:
i ]
76. P. N. Richar s:
77. W. B. Hutchison,
C. J. E. Smith,
T. W. Watson and
I. L. Dil l amore:
W. B. Hutchison,
T. W. Watson and
I. L. Dil l amore:
78. S. R. Goodman and ( ,.
79. S. R. Goodman and
80. H. Hu and
S. R. Goodman:
81. W. F. Hosford and
W. A. Backofen in:
W. F. Hosford in:
82. F. Haener and
H.Weik:
Hl III
83. I. L. Dil l amore,
C. J. E. Smith and
T.W.Watson:
84. J. T. Michal ak and
R. D. Schoone:
85. R.H.Goodenow:
86. I. L. Dil l amore and
S. F. H. Fl et cher in:
87. W. Jol l ey:
f 1 3
88. J. F. Hel d:
89. W. C. Lesl ie,
R. L. Rickett,
C. L. Dotson and
C. S. Wal ton:
90. R. L. Sol ter and
C. W. Beattie:
Ed. , Springer-Verl ag, Berl in/Heidel ber&
New York 1968, p. 93.
J. Austr. Inst. Metal s ^2 ( 1967) Feb, p.2.
in[l 8] ,p. 339 ff.
\
?V
J.Iron Steel Inst. 207 ( 1969) 1479.
Met. Trans. 1( 1970) 1629.
US Patent 3, 404, 047, Oct. 1968.
Met. Trans. \) 3057.
W. A. Backofen et al Publ . , 9th Saga-
more Conf. ( as under[72] , p. 259;
[18] , p. 414.
Arch. Eisenhttenw. 27 ( 1956) 153.
Met. Sci. J. 1 ( 1967)49.
Trans. Met. Soc. AIME 242 ( 1968) 1149.
Trans. Met. Soc. AIME 59 ( 1966) 804.
"Recrystal l iz ation. .." ( Re. 4),p. 448.
J.Iron Steel Inst. 305 ( 1967) 321.
Trans. Met. Soc. AIME 239 ( 1967) 573.
Trans. AIME 4j[ ( 1954) 1470.
Trans. AIME m ( 1951) 721.

91. H. Borchers,
Z. -Q. K im and
H. H. Hoff :
92. I. L. Dil l amore:
93. I. L. Dil l amore:
94. G. Ibe and
K . Lcke in:
95. H. Yoshida,
K . Sasako and
F. K anz aki:
96. K . Sasaki:
97. E. H. Mayer and
D. E. Wise:
98. M. Hil l ert:
99. J. L. Wal ter in:
100. P.A.Beck:
101. C.T.Wei,
M. N. Parthasarathi and
P A Rpr k- . J^ . J Ct - t V.
102. M. L. K ronberg and
F. H. Wil son:
103. G. W. Rathenau and
J. F. H. Custers:
104. S. R. Goodman and
H. Hu:
105.
106. H. Hu and
S. R. Goodman:
107. J. L. Wal ter and
C fl fl nnn* . VJ . LSUl l l l .
Arch. Eisenhttenw. 35 ( 1964) 57 ff.
"
Acta Met. 12 ( 1964) 1005.
Trans. Met. Soc. 233( 1965)702.
"Recrystal l iz ation. . . " ( Ref. 4) p. 434 ff.
IDDRG Col l oquium, London 1964.
US Patent 3, 239, 388, Marz 1966.
US Patent 3, 244, 565, April 1966.
Acta Met. 13 ( 1965) 227.
[18] , S. 227.
Acta Met. 1( 1953) 230.
J. Appl . Phys. 28 ( 1957) 874.
c. f. : Trans. AIME 185 ( 1949) 501.
^^^
Phil ips Res.Rep. 4 ( 1949) 241.
Trans. Met. Soc. AIME 236( 1966) 710.
c. f . al so[5] , S. 384.
Trans. Met. Soc. AIME 227 ( 1963) 1454;
230 ( 1964) 1413; 233 ( 1965) 103.
Trans. Met. Soc. AIME 215 ( 1959)465.
-
1 8 3
CHAPTER 7
EFFECTS OF TRACE IMPURITIES
ON RECRYSTALLIZATION PHENOMENA
BY
0. DIMITROV, R. FROMAGEAU AND C. DIMITROV
It has been known for a long time that impurities may have a strong
influence onthe recrystallization behavour of deformed metis, and
systematic investigations of this effect have been made, for instance
in copper [l]and in aluminium [2]. However, very significant ad-
vances in the understanding of the basic phenomena have taken place
more recently, since high purity metis have been made available.
It has then been possible to study the simpler cases of very pur
metis, doped with known amounts (eventually in the part per million
range) of one foreign element. In the present paper, we shall first
discuss the general features characterizing the influence of impurities
on recrystallization. We shall then give an outline of the theories
which have been proposed to account for the rate of migration of the
recrystallization interface. The results of a number of detailed ex-
perimental studies will then be described and compared with the
theoretical predictons. Finally, we shall mention a few consequences
resulting indirectly from the retarding effect of impurities. We shall
not intend to give here a general exhaustive review of the work done on
this subject, but we shall rather select a few characteristic examples
in order to illustrate the important points.
184
1. GENERAL EFFECTS OF IMPURITIES ON RECRYSTALLIZATION
1.1. The first and most conspicuous effect observed is that recrystallization
is_delayed by the presence of impurities. More precisely, this means
that, for comparable initial conditions, annealing at a given temperature
will require a longer time to achieve recrystallization, when impurities
are present, In some cases, however, the efect may be so large that
recrystaUization can no more be observed in any reasonable time at
the chosen temperature,
In order to characterize the effect of purity on recrystallization as a
whole, one can then give the temperature at which the phenomenon is
completed in some specified length of time. As recrystaUization can
be described on the basis of thermally activated processes, temperature
will enter exponentially in the rate formulations, and for widely differei
rates, it will be more convenient to give the temperature of isochrona
anneals than the length of isothermal ones. The temperature in itself
has no absolute signification but it is useful as a means of comparison,
if the various parameters are held constant (initial structure and
orientation, type and amount of deformation, etc,.}.
Table 1 gives the temperatures necessary to achieve complete re-
crystallization after large deformation by rolling, for commercially
pur and for high-purity zone-refined metis [3]. It can be seen,
and this is a general rule, that the purification of a metal leads to an
appreciable decrease of its recrystallization temperature. The effect
is very marked in the low-concentration range, In the case of high-
purity aluminium, recrystallization takes place below room temperature.
instead of above it. This, of course, has no fundamental significanceto
it has the consequence that pur aluminium cannot be retained at room
temperature in an appreciably cold worked state: it will spontaneously
soften on staying at that temperature. Also the study of the recrystallii-
ation of this material requires special, low-temperature, metallographi;
techniques.
185
Table 1
Recrystallization Temperatures of Metals of Different Purities.
Element Commercial pur metal Zone refined metal
(0 (C)
Aluminium
Copper
Iron
Nickel
Zirconium
200
180
480
600
450
- 50
80
300
300
170
,2. An important question related to the influence of impurities concerns the
state under which they are effective: Is the effect of foreign elements
most important when they are dissolved in the lattce of the host metal,
or when they are precipitated as a second phase? Experimental results
show that a given amount of impurity has a larger influence on primary
recrystallization when it is dissolved than when it is precipitated. For
instance, this is illustrated in the results of MKTARIQL[4] who studied
the recrystallization of zone refined aluminium doped with small amounts
of iron. For samples containing 25 p. p. m. iron, annealed at 600C be-
fore rolling, the vales of the nucleation rate v and growth rate v were:
o ,n2 -3 -1
v = 3.10cm s
n
and v = 7 . 106 cm . s"1 at 0C.
For samples heat-treated at 400UC before rolling, the corresponding
vales were
3 -3 -1 -4 -1
v = 1,9 . 10cm . s and v = 10 cm . s
The difference corresponds to the fact that iron is entirely in solid solu-
tion after annealing at 600C, whereas it is mostly precipitated after treat-
ing at 400C and has then a much smaller influence on recrystallization.
1 8 6
However, it must not be concluded f rom the preceding, that precipitated
impurities have no effect. It is known that grain growth is influenced by
the presence of preciptales, but even for primary recrystallization, small
volume fractions of preciptales can be effective if they are finely dispers-
ed in the cold worked matrix. This effect has been observed by F R O M A < " ^ A U
[5] during a study of the recrystallization of deformed aluminium single
crystals. Three series of samples were used, with the same {ll2|< lll>
orientation [6] .
- Sample SSC was prepared f rom a commercial high-purity aluminium
with an estimated impurity content of 5 p. p. m.
- Sample ZF C was obtained by zone-refining metal SSC and its estimat-
ed impurity content is O, 5 p. p. m.
- Sample ZF C is a metal prepared by electrolysis in organic mdium,
and further refined by zone melting. Its estimated impurity content is
about 0, 15 p. p. m.
The resistivity ratios
?294K' &lven in Fig. 7. 1, reflect the
increasing purity f r om SSC to ZF C.
A s for the recrystalllzation behaviour, the comparison of SSC and ZF C
shows the expected increase of recrystallization rate (here characterized
by a nucleation rate) with increasing purity. However crystal ZF C ex-
hibits a lower recrystalU zation rate, which seems to be in contradiction
with its higher overall purity. The reason for this discrepancy can be
found by examining the structure of the material in the electrn micros-
cope. Int he material obtained by organic- mdium electrolysis and zone
ref ining, black spots are f ound, which correspond to precipitates (F ig.
7. 2). These are absent in the metal prepared by zone refining f r om a
commercial high-purity material (Fig. 7. 3).
Thus, the presence of a small amount of a finely dispersad second phase
is rct detectad br th$ electrical resisTity. but has aa appreciable in-
f l uence on recrystallization,
This valu is the sum of the concentrations of determined foreign elemffl3
and of the sensitivlty limits of the undetected ones, for about forty analy
impurities,
1 8 7
logv
3 7
n
(vn:cms
NUCLEATION
OrientationC
(112)
3,5
T(SK)
Fig. 7. 1
Var i at i on of nucleation rate with t emperat ure, for three
cold-rolled al uminium crystals of different impurity content.
188
Fig. 7.2
Presence of precipitates in a recrystallized sample of
aluminium prepared by organic-medium electrolysis
and zone refining.
Fig. 7. 3
Absence of precipitates in a recrystallized sample of
aluminium prepared by industrial three-layer electro-
lysis and zone refining.
189
1.3. What parameter is most influenced by impurities: nucleation or growth?
The answer to thls question is difficult because a clear experimental
separation of the two processes is not easy to achieve. When new
crystals are formed, their growth rate can be defined simply and it
can be measured easily. However, the measurement of a nucleation
rate depende on the definition of the recrystallization nucleus. The
latter can be observed only after it has undergone some growth (c. f.
chapter 1 section 2. 2. and chapter 3). Thus, the measured valu
will depend not only on the nucleation process itself, but also on the
growth characteristics.
Experimental results suggest that the main effect of impurities is on
the growth process. In the case of aluminium, indications on the
effects of impurities may be derived from the electrn microscopic
observations of MESBGER [7], During the anneal of cold worked
aluminium, one first observes a rearrangement of dislocations inthe
cell structure, leading to the formation of well defined cell walls. At
higher temperatures, there is a general increase in size of the cells;
some of them grow much larger than the surrounding ones, and con-
tinu to grow as new recrystallized grains. The first process is
fundamentally the same as a nucleation, as it provides a well defined
interface in the metal, a necessary structure for subsequent growth.
The second process is based on the migration of such an interface and
corresponds to growth. Figure 7. 4 shows that impurities influence
both types of processes, but that growth is much more sensitive than
dislocation rearrangement. This is especially marked in the low con-
centration range.
In the following discussion we shall deal mainly with the influence of
impurities on growth. As we have seen, this seems to be the principal
effect and, further more, it has been much more studied, both theoretical-
ly and experimentally.
3 9 0
rotura
ore
20 0
200 K )
30010'
Fg. 7.4
liifluence o impurity conten on the structura]
c- hanges ubservod in aluminium by electrn
niicroscopy.
1 91
2 . THEORETICAL TREATMENTS OF THE INFLUENCE
OF IMPURITIES ON GRAIN BOUNDARY MIGRATION
The fundamental ideas underlylng most of the current theories have
been givenbyL C K E andE E TE K T [8] . Although their quantitative theory
is a simplified one, it gives nevertheless many of the general features
of the experimentally observed phenomena.
The model is based on the assumption that there is a forc of inter-
action between dissolved impurites and the interface separating a
new recrystallized grain from the cold worked matrix. Due to this
interaction, impurities concntrate near the interface and form an
atmosphere which must be dragged by the movingboundary. The
velocity of the boundary is thus decreased. At hgh concentrations,
the rate of migration is controlled by the diffusion of the impurity
atoms behind the boundary. When their concentration decreases, the
migration rate increases. If the rate exceeds a certain critical valu,
v
the foreign atoms can no longer follow the interface, which breaks
away from its impurity atmosphere. Thus, at low concentrations,
the effect of impurities should be negligible; the migration rate should
be concentration independent, and essentially equal to that of an ideal-
ly pur metal.
More complete and rigorous theoretical treatments of this probiem have
been gvenbyL JC K E andS T R E [9] and by C flH N [10] .
Fundamentally, one writes that the rate of migration v is proportional
to the net drivingforc on the boundary:
v = m (p - PJ ) (1)
where m is the mobility of the boundary. p is the drivingforc re-
sultingfrom the decrease of free energy of the crystal when the boun-
dary moves; p. is the restrainingforc exerted by the foreign atoms
on the boundary.
1 9 2
The i mpuri t y drag p. depends on impurity concentration and on b oun-
dary velocity. Its valu is calculated as a f unction of the energy o
interaction between impurity atom and grain boundary, U_ , and of
U
the dif f usion coefficient of the impurity in a direction perpendicular
to the grain boundary, D. Both U_ and D are f unctions of the
o
distance to the boundary, as shown schematically in Fig, 7. 5.
Fig. 7. 5
Schematic profiles of inter-
action energy U_( x) and
diffusion coefficient D (X)
as a function of the
distance to the grain boun-
dary plae.
D(x}
Finally, one obtains the approximate relation
a C v v
, _
1 + p v
where C = bulle impurity concentration, otand / ?= parameters de
pending on the exact shape of the U0 and D prof iles.
D
193
This relationship between velocity and driving forc is given in Fig. 7. 6
for 3dif f erent vales of the impurity concentration. For low concentra-
tions, the curve s continuous and there is only a small deviation f rom
the straight line corresponding to an ideally pur metal. For high con-
centrations, the curve is S-shaped and its central part is unstable; the
velocity will change discontinuously f rom one part o the curve to the
other, at some critical driving forc. This relationship is complex
and the calculations involve assumptions concerning Un and D.
u
For the purpose of comparison with experiment, it is usef ul to con-
sider two simpler limiting cases.
Fig. 7. 6
Grain b oundary velocity v as a
f unction of the driving f orc p
for different solute concentra-
tons C < C < C .
At low migration velocities, ( v<g; - ) , eq. 2 be comes
v
+ or C
(3.1
m
This shows that should be a linear f unction of C. Furthermore,
1 9 4
the apparent activation enerpy w j ] i be near to a mean valu of the
activation energy for impurity diffusion in the boundary regin. Accord-
ing to the shapes of the Un and D profiles thia can vary frona- a valu
j f ' - .
cise to the activation energy for bulle diffusion of the impurity, to
approximately the sum of grain boundary diffusion and the adsorption
energy of the impurity on the boundary.
At high velocities, (v - ) the following equation is obtained
v = m ' p -
e x C
T
(4)
It can be shown that the first term dominates so that the influence
of impurities is not large. The activation energy is mainly determined
by the intrisic mobility m, and should be of the order of the activation
energy for grain boundary self diffusion.
Between these two domains, there is a transition regin w here the
velocity will change continuously or discontinuously, according to the
specific conditions considered. The variation of activation energy in
the transition regin will depend on the driving forc. For low driving
forces, it will go continuously from the low valu to the high one. For
very high driving forces, there will be a discontinuity in the transition
regin. For the intermedate driving forces, the apparent activation
energy will go through a mximum between the vales corresponding
to the two limiting ranges.
Thus the theory is able to predict a number of consequences w hich can
be compared with experimental data. How ever, the a priori calculation
of the migration velocities is very difficult because of the assumptions
w hich have to be made about the vales of the different parameters.
The existence of such adj ustable parameters, together with the ex-
perimental uncertainties, strongly limits in many cases the signific-
ance of a direct comparison.
1 9 5
3. EXPERIMENTAL OBSERVATIONS
The theory outlined above describes the influence of impurities on
the motion of a grain boundary in a metal. It can be applied to pri mary
recrystallization, normal grain- growth in the recrystallized state,
secondary recrystallization, etc... For primary recrystallization,
the subject of this paper, the situation is somewhat simpler than in
the case, for instance, of grain- growth w here driving forc and
orientation relationships between adj acent grains change continuously
during the process.
For a comparison with theory, the ideal experiment should provide:
- a constant and known driving forc
- the presence of one impurity in known amount in an
otherwise highly pur metal
- known orientations of the new grain and matrix.
Such a set of requirements is seldom fulfilled. In the following para-
graphs we shall select some experiments approaching these conditions
and we shall consider how their results agree with the theoretical pre-
dictions.
3 .1 Influence of the Concentration
Let us consider the case of a cold worked high purity metal, studied
during anneals at a given temperature. It is possible to determine how
the concentration of a foreign element influences the growth rate of
the new crystals. The case of alumi ni um- based dilute alloys has been
studiedbyFROIS [U, 12] . The growth-rate of the fastest growing
new grains was measured by determining the variation of the maximal
diameter as a funct i on of annealing time.
196
y cm s
O Cu
-...Mg
132 C
Ironwlion rgion i.
Cu'
oo
C 10
ragion
Fig. 7. 7
Growth rate of new crystals at 132C in cold rolled
aluminium, in the presence of copper or magnesium
additions.
197
Figure 7. 7 shows typical results for additions of copper or of magnesium.
One can distinguish 3 domains of concentration:
fi fi
- a low concentration regin (below 2. 10 for copper, or 65.10" for
magnesium) where there is very little influence of the addition element.
fi P.
- a high concentration regin (above 50.10" for copper, or 390.10
for magnesium) where the growth rate is much lower and decreases
with increasing impurity conten.
- a transition regin, where both types of growth can be observed on
one given sample.
In the results given here, the transition between the two types of growth
4 6
is a discontinuous one: the velocity decreases by a factor of 10 to 10
when going from one branch of the curve to the other. However, for
the same experimental conditions other solutes, like silver [12, 13] ,
or iron [4] give a continuous transition between the impurity-independent
and the impurity-controlled rejiOTi.
Thus, the general behaviour is in good agreement with the theoretical
predictions: existence of two domains with different influences of the
foreign element, and possibility of various types of transitions be-
tween those two domains.
As for a quantitative comparison, we can first consider the high
velocity branch of the curve. Theoretically, this branch is limited
to velocities larger than one third the intrinsic velocity [10] and thus
no large impurity effect is expected. Due to the low precisin of the
experimental velocity determinations, it is not possible to make any
significant quantitative comparison in this range of concentrations.
For the low velocity branch of the curve, eq. 3 shows that there should
exist a linear relationship between inverse velocity and concentration.
198
This, in fact, is not observed and velocity vares more rapidly with
concentration than predicted theoretically. In the case of copper
additions the non-linearity could be partly due to recovery which is
more extensive in the concentrated alloys, and which resulta in a
decreased driving forc for the higher concentrations. For magnesium
additions, however, such a driving forc difference does not exist and
thus, within the frame of the theory, one is led to conclude that the
concentrated alloys are not entirely in the low-velocity limit, but are
still in the transition regin. Any quantitative comparison with
formula 3 should then be restricte" to the alloy with the highest con-
centration, measured at the lowest temperature. For the magnesium
alloy of concentration 1000 p.p. m. measured at 132C, the follow-
ing vales have been assumed:
Mximum interaction energy UR: 1200 cal/mole [14] ; constant
interaction forc over 1interatomic distance on each Bid of the grain
boundary; driving forc p equal to the stored energy in pur aluminium
7 3 r T
4, 5 . 10 ergs/cm [15] ; diffusion coefficient of Mg constant and
-17 2
equal to the bulk diffusion coefficient: 6, 9. 10 cm /s, extrapolated
from the measurements of BtKLE [16] .
- fl -1
This leads to a calculated growth-rate of 1, 95 . 10 cm. s , to be
-R -1
compared with the measured one: 2, 08 , 10 cm . s . The agree-
ment is excellent, although probably fortuitous, in view of the un-
certainties on the various parameters.
3. 2 Nat ure of the Foreign Element
The global effect of a given concentration of a foreign element is
very different according to the nature of this element. The measure-
ments of SJVWrand 3VHTH [l] in copper clearly show this phenomenon
in range of concentrations around 100 p.p. m.
1 99
The same is true in the range of fractional part per million. This is
seen for instance in the experiments of TCM^GEAU [17] on the re-
crystallization on aluminium single-crystals after cold-rolling. The
crystals, all of the same orientation, were taken along a zone refined
bar and had therefore different impurity contents. Fig. 7, 8 gives
the vales of a number of parameters for the different crystals. Let
L-S compare the central purest part of the bar, with samples situated
near each end.
The comparison of the central samples No. 2 and 3 with the sample
taken at the flnishing end of the bar (No. 5) shows a decrease in
growth rate by a factor of 10 and a decrease in nucleation rate by a
factor of 1000 . This corresponda to a difference in the total im-
purity conten of 6 parts per million, constituted mainly of iron,
copper and sicon. If we now compare the purest samples (No. 2
and 3) with the one taken near the starting end (No. 1), we find that
there is a decrease in growth rate by a factor of 2 and in nucleation
rate by a factor of 10 '. The difference in impurity content is here
only 0,1p. p. m. consisting mainly in Cr, H f, Mo and W. Thus, it
appears that this last group of impurities has a larger influence on
recrystallization. These elements have a common feature: their
partition coefficient between solid and liquid aluminium is greater
than 1. If we consider the grain boundary to have a liquid-like
structure, such elements will tend to avoid the boundary: the inter-
action will be repulsive ( Ur j > 0j H owever, theory indcales that the
sign of U_ is not a decisive factor in determining the importance of
D
the slowing effect, Thus one should admit that some other factor
(magnitude of interaction energy or of diffusion coefficient) is respons-
ible for the observed difference.
'The apparently larger effect on nucleation rate is not in contradiction
with the statement of section 1.3. It originales from the fact that the
technique of measurements gives the growth rate of the fastest grow-
ing grains. If some grains are less influenced by the presence of im-
purities (see section 3. 4.), the measured growth rate corresponds to
these grains and will be relatively less decreased. H owever for the
maj orit y of the recrystalllsed grains, the growth rate is smaller and
the number of new grains detected per unit time is decreased. This will
be reected as a strong lowering of the apparent nucleation rate.
2 0 0
(kcat.at-g) ' -
f S
9 t,nc. <
(tseconds) 3
STARTING
END
200 300 FINISHING
END
Variation of parameters characterizing the recrystallization
of 5 aluminium single crystals, as a function of their position
along a zone-refined bar.
H,
t.
inc
T
R
amb
= Ratio of electrical resistivity at 20 K and
294 K
Growth rate at - 25C
Nucleation rate at - 25 C
Activation energy for growth
Activation energy for nucleation
Incubation period at - 25 C
Temperature of beginning of recrystalliz-
ation for anneals of 20 hours.
2 O 1
If >. now consider specificaUy the influence on growth rate of different
ents, several characteristics can be defined from a phenomeno-
1 Point of view. Thus, from results of the tvpe given in Fig. 7 7
one can define:
- the critical concentration, below which no signific effect
of the foreign element is detected
- the type of transition (continuous or discontinuous)
- the growth rate in the low-velocity regin.
parameters differ widely with the nature of the foreign element
*ey can be characteristic only under specified particular con-
M. For instance, the critical concentration is temperature-
ependent. Also the type of transition could change for different
vales of the driving forc.
From a theoretical point o view, a foreign element should be characteriz-
d by the vales o the interaction energy and of the dfusion coefficient
y their variation in the regin of the grain boundary (tu. can de-
pend on the type of grain-boundarji
An interesting case inthis comection is tha, o, lead, wtere some soluteS
^own to diffuse very rapidly, tt dif,usion Coeficients several
rders of magnitude larger than the coefficient for self-diffusion. RUTTER
and AUST [18]have studied the migration of large ang.e grain boundaries
' and doped lead, under the driving forc of a striation substructure
produced by solidification. was shown that the effect of gold and sllver '
>ns was much larger than the effect of the aame concentration Of
Th,s means that, in their experimenta! conditions, the fast diffus-
Puritie8silver and gold decrease the migration rate more
vely than the slow-diffusing impurity tin. This is what theory
rts, if the samples are considered to be in the high velocity limit
202
A n even more extreme ^ituation is that of the in terstitial impurities
carbn and n itrogen dissolved in of-iron . The effect of these elemen ts
on the rate of growth of new crystals, durin g the recrystallization of
deformed iron , has been studied byA NTONIONE et al. [19j . They
show that the effect is very small in the ran ge of con cen tration s
examined (5 to 60weight p. p. m. ) Theoretically, the extremely high
diffusivity of such elemen ts leads to relatively high vales of
so that the experimen tal velocities always correspon d to the "low
velocity" ran ge ( v - ^- ) . Moreover the coefficien t (or ) before
the con cen tration in eq. 3 ia small; the effect of such impurities will
be negligible for any practical concentration. This is in agreemen t
with the experimen ts.
3. 3. In fluen ce of Temperatura
Gen erally, the n fluen ce of temperature on the velocity will be
described by an apparen t activation en ergy, derived from the slope of
an A rrhen ius plot, d log v/d (-=-). This does n ot n ecessarily have
a physical sign ifican ce, but it can be compared with the theoretical
prediction s.
A typical example is given by the studies of GRNWA LD an d H A ES S NERn
gold [20J . The migration velocity of a recrystallization boun dary was
measured in rolled gold sin gle crystals, con tain in g 20parts per
million iron as the main impurity. Figure 7. 9 shows the variation
of migration velocity as a fun ction of temperature, for a 30 - LlllJ
tilt boun dary. One can see that the curve ten ds to be a straight Un e
at high an d at low temperatures, with a S -shaped tran sition at in ter-
medate temperatures. This mean s that the apparen t activation
en ergy varies as a fun ction of temperature, startin g from a con stan t
valu of 30, 5 kcal/mole at low temperatures, in creasin g through a
203
mximum aroun d 310C, an d decreasin g to a low constant valu of
19,1 kcal/mole at high temperatures. S uch a behaviour correspon ds
to the theoretical prediction s, for the case of in termedate vales of
the drivin g forc.
The variation of apparen t activation en ergy with con cen tration has been
determin ed for the recrystallization of doped alumin ium [12] . Two
differen t types of behaviour are foun d, which both correspon d to
possible theoretical cases. For addition s of magn esium (Fig. 7.10)
there are two differen t sets of vales. The first on e at low con cen tr-
ations is con cen tration independent an d iden tical to that foun d in the
high-purity material. It should be n ear to the activation en ergy
correspon din g to the in trin sic mobility. A t high con cen tration s, much
higher vales of the activation en ergy are foun d, of the order of an
activation en ergy for bulk diffusion . There is n o con tin uous tran sition
between the two types of growth, and both vales can be obtained in
the ran ge of in termedate con cen tration s.
On the other han d, for silver addition s, the behaviour is differen t
(Fig. 7.11). In the low-con cen tration alloys, the activation en ergy
is the same as in zone-rened alumin ium but it increases progressive-
_C c
ly when the con cen tration in creases from 15.10 to 230.10 , The
studied con cen tration s were n ot sufficien tly high to decide if there was a
con tin uous tran sition between the region s of low an d high activation
en ergy, or if the apparen t activation en ergy goes through a mximum.
The latter case was assumed by CORDN an d VA NDERMEER [21] for
their experimen ts on copper addition s in alumin ium.
1 0 '
204
350 330 310 330 270 250 2)0'C
'O '
/,65 1.70 1.75 1.80 I, 5 ',90 1,95 2,0
Fig. 7. 9
Migration velocity of a 30 [fll] tilt boundary in
rolled gold, as a function of temperature.
Qg ""
Mg
pur^AI
Fig. 7. 10
Activation energy for growth in the recrystallization
of magnesium doped aluminium.
205
30
2C
Qg Kcal mol.'1
Ag
Atomic
concentro I ion
KT
1-3
Fig. 7 . 1 1
Activation energy for growth in the recrystallization
of silver-doped aluminium.
2 0 6
3. 4. Orientation Eects
The inf l uence of the orientation relationship between the two regions
separated by the moving grain-boundary ia clearly demonstrated in
the work of AUST and RU TTER on lead [22] . Their measurements of
growth rate of a perf ect crystal into a striated one are made in such
conditions that the orientation relationship between the two crystals,
and the driving f orc, are maintained constant throughout the experi-
ment. Figure 4, 6 (chapter 4) shows the infl uence of additions of tin
in lead, on the migration velocity measured at 300 C. It can be seen
that tin has a pronounced slowing effect for grain boundaries separat-
ing two crystals without any special orientation relationship. By con-
trast, the effect is much smaller for pairs of crystals related by
certain rotations around simple crystallographic axes. This effect
of orientation is exhibited in a restricted concentration range, f rom
5 to 40 p. p. m. of tin. Outside this range crystals with special
orventations do not appear any more, with the nucleation technique
which has been used.
This smal l er inf l uence of tin impu rity for special bou ndaries is not
expl icitl y taken into account in the theoretical calculations. A n ex-
pl anation shoul d impl y consideration of the val es of interaction
energy, and of di f f u si on coef f i ci ent for such boundaries, based on a
model of their structure on the atomic scale. Some remarks about
this point have been proposed by CA HN [10] but not detailed nter-
pretation has been given. The whol e domain of the interaction be-
tween orientation ef f ect s and i mpu ri t y ef f ects l argel y remai ns to be
expl ored, on a quantitative basis.
4. CONSEQUENCES OF THE EFFECT OF IMPURITIES
The important reduction in recrystal l ization rate induced by the
presence of i mpu ri t i es has consequences for rel ated phenomena. W e
207
shall mention. here two examples concerning the f ormation of textures,
and the occurence of recovery.
4.1. Texture Formation
If orientation relationships have an effect on the influence of impurities,
as described in section 3. 4. , this will have direct consequences on
texture formation.
Let us consider the case of a pur sample in which a large number of
potential nuclei is present. Its recrystallization will lead to crystals
with a certain spread in orientaions. In the presence of impurities,
the nuclei with special orientations will grow with velocities which
can be several orders of magnitude larger than the random ones.
Thus, they will represent the majority of the recrystallized grains.
In comparison with the pur sample, this will result in:
- a drastic decrease in the appearent nucleation rate, as only
the small number of nuclei having special orientations wil l
grow to measurable sizes.
- the formation of a sharp texture, as there will be a growth
selection of the favourably oriented grains.
These ef f ects will be observed only in the concentration range where
the special boundaries have a higher migration velocity.
Such a behaviour has been experimentally observed in al uminium doped
with copper [23] or magnesium [ll] and also with zone-ref ined
al u mi ni u m single crystal of varying total impurity conten [24] . This
is illustrated on the micrographs of Figs. 7. 12 and 13. Figure 7.12
shows the structure of a recrystallized sampl e of al u mi ni u m wi t h a
total impurty content of about 1 p. p. m. The sample has been polish-
ed and anodized, and the oxide f il m gives different shades in
2 U 8
polarized light. These v u r y with the structure of the oxide fi l m, re-
lated to the orientation of the underlying grain. From visual inspection
no promi nent orientation effect is detected. Figure 7.1'3 shows, in
the same conditions, the structure of a sample having a total impurity
conten of about 6 parts per million. The grains here exhibit very
similar shades, which indicates that there is a very marked preferred
orientation, The shape of the grains is different and reveis the
existence of preferential growth directions.
4. 2. Recovery Processes
In a metal where recrystallization is strongly delayed by the presence
of impurities, recovery can take place much more extensively. (We
consider here recovery in the restrictive sense of a decrease in
mechanical properties). This is especially true for large amounts
of cold work.
For instance, in the case of high-purity aluminium, recrystallization
after cold-rolling takes place below room temperature. It is preceded
by an important decrease of resistivity, attributed to the elimination
of point defects, but not by any significant decrease of mechanical
properties [25] . By contrast, in copper-doped samples, re-
crystallization is rejected to much higher temperatures. Table 2
shows t hat , during anneals at 87 C an important decrease in re-
sistivity and in mechanical properties is observed before recrystalliz-
ation t akes place.
i
I
209
Fig. 7.12
Microstructure of a recrystallized zone-refined aluminium
f t
Fig. 7. 13
Microstructure of a zone-refined al umi ni um sample fless
pur material). Observation in pol nni cd light after
anodization.
210
Table 2
Recovery of Electrical and Mechanical Properties at 87C,
in Aluminium Doped with 50 Atoro p.p. m. of Copper,
Annealing treatment
after rolling
90 s at 0C
(elimination of point
defects
same + 33 s at 87C
same + 470 s at 87C
Excess resistivity
remaining after
elimination of point
defects (%)
100
65
40
Fracture
stress
2
(Kg/mm )
28,5 t 0, 5
25, 2 - 1,7
22,7 -1,1
The metal recrystallized at 200 C has a fracture stress of
10,7 - 0,7 Kg/mm2
This effect can have technical consequences. In some applications
(deep-drawing for instance) one looks for low deformation stresses
but also one wants to avoid a recrystallized material because of
possible inhomogeneous deformation. Such a state can be achieved
by recovery. If recrystallization can be controlled by the pff ect of
addition elements, a range of temperatures is made availaMe, in
which an annealing treatment will provide the suitable structure.
5. CONCLUSIONS
From a general point of view, it can be said that the impurity-drag
mechanism gives a good interpretation of the effects of foreign elements
2 1 1
on recrystallization. The theoretical treatments derived from this
model seem to be well supported by experiment, although their flexibility
makes strict comparisons difficult.
Many interesting questions still wait for an answer, especially in re-
lation with orientation effects. For instance:
- What orientation effects exist in really pur materials?
- How are such effects related to grain-boundary structure?
- Is there any impurity effect on the intrinsic mobility (m in
eq. 1), which is usually assumed to beconstant?
- How is impurity drag influenced by orientation relationships,
through the effect on interaction energy or on diffusion-
coefficient?
Finally, it should be mentioned that two favourable circumstances
exist presently in this field.
- We have at hand a good theoretical tool, which makes possible
a better knowledge of the conditions, with respect to impurity-
drag, in which experiments are done. This allows more firm
conclusions to be drawn from such experiments.
- Techniques are available for obtaining materials of very high
degrees of purity, for introducing and for determining small
amounts of foreign elements. We can thus control the purity
parameter with much more confidence. Such a situation is
likely to promote rapid progress in the understanding of re-
crystallization phenomena.
212
REFERENCES
1. J. S. Smart Jr and
A. A. Smith Jr:
2. G. Masing,
K. Lcke and
P.Nolting:
3. O.Dimitrov:
4. F. Montariol:
5. R. Fromageau:
6. R. Fromageau and
Ph. Albert:
7. C. Messaser and
O.Dimitrov:
8./K. Lcke and
9. JK. Lcke and
H. P. Stwe in:
10. J.W. Cahn:
11. C. Frois and .
O. Dimitrov:
12. C. Frois and
O. Dimitrov:
13. C. Frois:
14. D. Me Lean:
15. E. Bonj our,
C. Frois and
O.Dimitrov:
16. H.Bckle:
17. R. Fromageau and
G. Revel:
18. J.W. Rutter and
K.T. Aust:
19. C. Antonione and
G. Della Gatta:
213
Trans. AIME 152, (1943) 103.
Z.f.Metallkde. 47_ (1956) 64.
Mem.Sci.Rev. Met. 62_, (1965)59.
Metaux et Corrosin W, (1963) 223.
Mem.Sci.Rev. Met. 66_ , ( 1969) 287.
Comptes Rendus Acad. Se. 260 , (1965)
895.
Comptes Rendus Acad. Se. 251, (1960)
88.
Acta Met. 5, (1957) 628.
"Recovery and Recrystallization o
Metals", L. Himmel, Ed., Interscieiw
Publishers, New York 1962, p. 171.
Acta Met. 10^, (1962) 789.
Comptes Rendus Acad. Se. 253, (196U
2532.
Ann. Chim. Paris !_ , {1966)113.
Comptes Rendus Acad. Se. 256 ,(1953
5137.
"Grain Boundaries in Metals" Clarer
Press, Oxford 1957, p. 124.
Comptes Rendus Acad. Se. 259, (1964)
4027.
Z.Elektrochem. ^3_ , (1943)238.
Compte sRe ndusA c ad. Sc . 266^, sen* '
(1968) 258,
Trans. Met. Soc., AIME 218, (196Q )
Colloque International sur le fer de' .-
haute puretfe Paris (1966); Mem.Sci. H
Met. 65, (special issue 15 j une 196:
p. 315 and 327.
20. W.Grnwald and
F. Haefiner:
21. P. Cordn and
R. A. Vandermeer:
22. K.T. Aust and
J. W. Rutter:
23. C. Frois and
O. Dimitrov:
24. R. Fromageau:
25. C. Frois
Acta Met. _ 18, (1970) 217.
Trans. Met. Soc., AIME, 224 , (1962)
1917.
Trans.Met. Soc., AIME 215, (1959) 119.
Comptes Rendus Acad. Se. 252, (1961)
1465.
Thesis p. 115, Paris (1968).
Acta Met. 14, (1966) 1325.
2 1 4
2 1 5
CHAPTER 8
RECRYSTALLIZATION OF TWO-PHASE ALLOYS
BY
U. KSTER
Many commercial alloys such as steels and aluminium alloys are
phase mixtures rather than homogeneous solld solutions. The re-
crystallization behaviour of such alloys fs known to be difficult to
predict [ 1 - 3] . It depends on whether second phase particles are
precipitated during recrystalization (Part 1.) or before plstic de-
formation (Part 2,), and on the nature, volume fraction and dispersin
of the second phase. Very little was known unti recently about the
recrystallization behaviour of coarse two-phase alloys such as or //3 -
brass in which the phases effectively form seprate grains, The pre-
sent discussion is therefore limited essentially to.the influence of
particles of a second phase on primary recrystallization, :i, e. t is
assumed that thte volume fraction of the second phase is only small
( ~10%)- The interpretation of the effects occring in coarse^two-
phase alloys in which the constituents are present i almost equal
proportions is discussed briefly in Part 3.
1, INFLUENCE OF PRECIPITARON
ON PRIMARY RECRYSTALLIZATION
If the defect state is produced in a supersaturated alloy by plstic
deformation, particles of the second phase form in the same temperatura
range over which recrystallization can also proceed (Fig. 8. la). Pre-
cipitation and recrystallization exert a mutual influence: the precipitat-
ing particles hinder both the rearrangement of dislocations to form re-
crystallization fronts and the migration of the latter, while the defects
216
present in the crystal lattice strongly accelerate the nucleati on of
parti cular phases, which themselves can affect dislocation rearrange-
ment and grai n boundary mi grati on. Only when recrystallization is
complete before the start of nucleation for precipitation are the re-
actions without influence on one another.
Inadditionto a survey byE. HORNBOGENandH.KREYE[3] , a large
number of publications on this subject involving various alloy systems
[4 - 12] have appeared in recent years.
1.1
Concentration c
Fig. 8,1 a
Schematic phase diagram: an alloy of concentration C Q
is homogenised at TH, quenched toTR and plastically
deformed i nthe supersaturated state. " " Precipitation
and recrystalli zati on processes can proceed on anneal-
ing at temperatures T < T,.
Order of Occurrence of Precipitation and Recrystallization
In order to gain an overall pi cture o the conditions under whi ch
the completi on of recrystallization precedes nucleation of preciptales,
2 1 7
or mutual influence of the precipitation and recrystallization processes
occurs, the temperature dependence of the start of both reactions will
be considered. For the start of recrystallization (assuming a constant
size for the recrystallization nuclei):
RT
where K_ is a factor containing the driving forc for the reaction,
an entropy term and geometric factors, N is the dislocation density
and Q0 (N) is the activation energy for the formation of a recrystalliz-
K
ation front. Q,,(N) has the order of magnitude of Q_ (activation-
t U
energy for self-diffusion}, decreases somewhat with i ncreasi ng dis-
location density and increases as a resutt of segregation of dissolved
atoms at dislocations.
For the start of precipitation (one stable phase only):
t,, = K - ea
P p
where K is a factor containing the driving forc, an entropy term
and geometric factors, c is concentrati on and Q (c, T) is the activ-
ation energy for nucleation of the second phase. O (c, T) depends
strongly on the supercooling below the equilibrium temperature T..
Only at low vales of T, that is considerable supercooling, does
Q (c,t) become small compared wi th Q - onl yt heni s tp dependent
solely on the activation energy for diffusion Q _.
Start of rtcrystalilzation
Logarithm of the onnealing time
Fig. 8. 1 b
Temperature dependence of the start of precipitation
and recrystallization.
Both curves are shown in Fig, 8.1 b. It can be seen that for an alloy
with a given dislocation density and concentration there is a temperature
T at which t = tp , below which there is an interaction between
recrystallization and precipitation in the total volume of the alloy. Thm
t emperat ure ranges in the recrystallization behaviour can be distinguish-
ed on the basis of this diagram:
(1) T > T- (T = equilibrium temperature)
Recrystallization is influenced only by segregation;
no precipitation
2 1 9
> ,T > T2 . : Recrystallization is influenced only by
. i segregation; precipitation proceeds in the
,: recrystallized structure after completion
. . of recrystallization
T < Tg Precipitation proceeds in the deformed
f structure; recrystallization s influenced
by particle. 3 .
i
The transition at T2 is naturally not sharp since a certain time is
necessary for the completion of recrystallization. In this transition
regin the behaviour characteristic of both ranges (2) and (3 ) occurs
simultaneously at different sites in the specimen. Below T_ ,
2
Qn (c,T) is Influenced by the lattice defects still present, in most
cases being considerably reduced. The precipitation curve changes
correspondingly compared with that for the defect-free crystal; pre-
cipitation is generlly accelerated as a result of which the start and
finish of recrystallization are further delayed.
1.2. Influence of Precipitation on Recrystallization
In the temperature range (3 ) the precipitating partlcles influence
both the rearrangement of dislocations to form recrystallization fronts
and the migration of the latter. Assuming that a suitable grain boundary
is available to act as recrystallization front or has formed, it will be
subject to a number of forces. In order to facil tate an understanding
of the recrystallization behaviour in this temperature range it is ad-
vantageous to set up an equation of forces.
1.2.1. Migration of Recrystallization Fronts (Discontinuous Recrystallization):
The driving forc for grain boundary migration can be expressed both
9
as the gain in free energy per unit volume (erg/cm ) traversed by the
220
o
grain boundary and as the forc (dyn/cm ) acting on unit rea of the
grain boundary, i. e. the pressure on the grain boundary (c.f. chapter 2).
The driving forc p together with the grain boundary mobility m de-
termine the growth rate v of the recrystallization front according to
v = m p . The grain boundary mobility depends strongly on
temperature, the orientations of the two grains and the nature of the
grain boundary. It is further strongly affected by the impurity atom
concentration in the grain boundary (c.f. chapters 4 and 7).
The following processes can lead to a reduction in the free energy in
supersaturated, deformed alloys by the migration of a grain boundary
(Fig. 8. 2a - d):
(a) driving forc p-, due to the elimination_pf dislocations or sub-
grain boundaries
If a grain boundary suitable as a recrystallization front migrates into
a deformed regin, it leaves behind it a regin with a much lower dis-
3
location density. The gain in energy per cm of volume traversed by
the grain boundary and henee the driving forc is given by:
2
= G. b ( No-
where G is the shear modulus, b is the Burgers vector and N, N.
are the dislocation densities in the deformed and recrystallized regions
respectively. The order of magnitude in the case of heavily deformed
o n
metis is 10 dyn/cm . The driving forc due to the elimination of
subgrain structure can be estimated correspondingly, e. g. for the
7 2
subgrain structure in Fig. 8.10b p 10 dyn/cm .
(b) driving forc p duet o discontinuous precipitation
In the case of partiles precipitating in the recrystallization front from
221
,v
Fig. 8. 2
Schematic sketch of discontinuous re-
crystallization (after E. HORNBOGEN[9] ):
(a) Discontinuous recrystallization. A
grain boundary serves as reaction
front and reduces the dislocation den-
sity from N to N-.
(b) Discontinuous precipitation. The super-
saturation of dissolved atoms C acts
analogously to the dislocations Ss a
"chemical"forc on the reaction front.
(c) Combined discontinuous reaction. Both
the dislocation density N and the con-
centration C are reducea to N- and C,
in the reaction front. Tlie prerequisite
for this is a supersaturated defect
solid solution.
(d) Dispersed particles exert a driving
forc in the reverse direction, If the
particles in equilibrium with the
st ruct ure of the grain boundary are
stable they mus be avoided: if they
are not stable they dissolve and re-
pr e< ' : ~ ! ' t e behind t he reaction front.
2 2 2
a supersaturated solid solution or in the case of metastable partiles
transforming into stable particles, a further driving forc ia set up.
Its magnitude can be calculated from the difference in free energies
of the supersaturated crystal and the crystal in which separation of
phases has occurred, J g
, and from the boundary surface energy
necessary for the particles
_ (V: mole volume) [13]. Thus:
with
where f is the volume fraction of the precipitates, A.S the interlamel-
lar spacing, ^R is the specific boundary surface energy and
= RT c log + ( 1 - c ) log 1 - - co
- cl
where c , c- are the concentrations before and after traversal by
the reaction front respectively (c.f. chapter 2 section 2. 5.). For an
Al-Zn alloy [14] at T ^500 K and a reduction in concentration due
to phase separation from c = 0. 05 to c- = O , 01, a driving f or cp
Q O O 1
of about 10 dyn/cm is obtained.
The gain in energy due to precipitation and henee the driving forc can
also be determined calorimetrically. It must be mentioned that only
in the extreme case of a reaction front with migration velocity v = O
does discontinuous precipitation bring complete phase separation
corresponding to thermodynamic equilibrium. As a rule a residual
supersaturation remains in the recrystallized matrix the extent of
2 2 3
which increases with the forward velocity of the recrystallization front.
If a grain boundary migrates under a forc p only, the process is
termed discontinuous precipitation. Assuming p and p to be
additive, a discontinuous reaction is possible in a deformed and super-
saturated crystal in which the dislocations are eliminated and atoms
precip tate simultaneously. The rate is determined by the effective
valu of PN + pc .
(c) retarding forc p due to precipitated particles^
Below T0 a high angle grain boundary acting as recrystallization front
t
is subject not only to an accelerating forc p = p,, + p but also to a
y* t-
retarding forc p due to precipitated stable particles. The same is
true in the case in which non-deformable particles were present be-
fore plstic deformation. Accordingto ZENER [5], for spherical
particles with diameter d., p is given by:
3f
pr =
'B
d.
where f is the volume fraction of the particles and # is the speci-
fic boundary surface energy. This equation is based on the assumption
that the interaction between grain boundary and particle is independent
of the nature of the particle. A refinement due to ASHBY [16] takes
into account whether the grain boundary energy within the particle is
greater or smaller than in the matrix. The above equation for the
forc is nevertheless a good approximation.
It follows from this expression, in agreement with experience, that for
a given volume fraction, smaller particles hinder grain boundary
migration more than large ones. This expression does however assume
a uniform distribut on of particles throughout the matrix. If the volume
224
fraction f is preferentlally precipitated at the grain boundaries, the
given expression is no longer valid and the restraining forc is larger.
The condition for migration of the recrystailization front (discontinuous
recrystallization) is
PN + Pc > Pr
as long as there is no further retarding forc due to segregation at the
grain boundary (which case will not be discussed any further here).
It must be stressed that it is irrelevant which type, or mainly which
type, of energy is responsible for the migration of high angle grain
boundaries. The only essential is that the sum of the forces exceeds
the critical valu for displacement. If p-, + p < p , migration o
the reaction front ceases.
1. 2. 2. Continuous Recrystallization: Apart from the condition that the sum o
the driving forces should exceed the retarding forc, the prerequisite
for the occurrence of discontinuous recrystallization is the existence
of a grain boundary capable of migration. For pur metis the activ-
ation energy for rearrangement (climbing) of dislocations produced
during deformation is Q _ _ ~ Q ,. (activation energy for self-diffusioel
K OL
This valu can be increased by more than one order of magnitude by
the segregation of impurity atoms (e. g. in Al-Fe alloys [12] ) or by
preferential precipitation of particles at dislocations (e. g. 0 ' in Al-Cu
alloys [4, 6] }, with the result that practically no rearrangement can
occur. If it is further assumed that no grain boundaries are present or
that existing grain boundaries are held by preferential precipitation, no
recrystallization by the formation and migration of a reaction front can
occur in the alloy. In such a case the annealing out of defects (dis-
locations, grain boundaries) is controlled by the formation or re- N
2 2 5
soluton of the precipitated particles. Precipitation of particles breaks
down the segregation at dislocations and henee facilitates their rearrange-
ment. Electron microscopio studies (e. g*- on Al-Cu alloys, see Fig. 8.10.)
indicate that the further rearrangement of dislocations is controlled by
the solution of small particles which pin dislocations or dislocation
groups, i. e. the extent of dislocation-free regions corresponds to the
distance between particles. The microscopic process is described quite
well by subgrain coalescence or Y-node motion (Fig. 8. 3.). As soon
as a particle located at a dislocation node dissolves, the activation ener-
gy at this point is reduced to Q qyy These dislocations can now migrate
much more easily than those still pinned by particles and a subgrain can
anneal out by Y-node motion. Another process involves a reduction of
the dislocation density of a subboundary which implies rotation of a sub-
grain. Both processes proceed as soon as new nodes have been freed
by particles dissolvng. The defect-free reas increase as do the angles
between them (as long as the dislocations do not annihilate one another).
Growth can proceed when the dislocation networks assume the character
of grain boundaries even though the condition for discontinuous recrystalliz-
ation is still unfulfilled. This process, without any migration of re-
erystallization fronts, finally leads to a completely recrystallized struc-
ture differing from the discontinuously recrystallized structure only in
the distribution of grain size and particles and in the texture [3 ].
1.3. Influence of Temperature, Supersaturation, Dislocation Density and
Annealing Time on the Recrystallization Behaviour: In the temperature
range (3), in which the precipitation and recrystallization processes
exert a mutual influence, discontinuous or continuous recrystallization
can occur. In an alloy with concentration c and dislocation density N
the volume fraction of particles precipitating at dislocations ii
from f = O at T with decreasing temperature because y x
creasing solubility; simultaneously the particle size diminjehss with
3
.

3

u
Q

"
"

T
O
3

"
"
o

E
1
O 3

e
r
a

.

1
>

3
&
&
.
W
3
r
o

r

'
H
.

3
*

w

*
"

g

P

2
g
r
t
j
p
f
O
B
f
f
q
S
,
-
)
,
-
.
3

t
-

O
*

1
^

(
B

^
i
^
^
n
.

^
-
s
i
i
w
t

g
:

5

g

^
g
.
a
i
i
a
l
i
l
"

C

'
c
r
i
m
4 u

g
a
f
t
t
u
r
o
M e
n
I

:

n
o
r
m
a
l

r
e
c
r
y
s
t
a
l
l
i
z
a
t
i
o
n
(
i
n
f
l
u
e
n
c
e
d

o
n
l
y

b
y

s
e
g
r
e
g
a
t
i
o
n
}
P

=
P
N
:

n
o
r
m
a
l

r
e
c
r
y
s
t
a
l
l
i
z
a
t

o
n
;

p
r
e
c
i
p
i
t

t
i
o
n

o
c
c
u
r
s
i
n

t
h
e

c
o
m
p
l
e
t
e
l
y

r
e
c
r
y
s
t
a
l
l
i
z
e
d

s
t
r
u
c
t
u
r
e
:
-
T
.
P

=
P
N
.
m

:

d
i
s
c
o
n
t
i
n
u
o
u
s

r
e
c
r
y
s
t
a
l
l
i
z
a
t
i
o
n

i
n
f
l
u
e
n
c
e
d

b
y
p
a
r
t
i
c
l
e

p
r
e
c
i
p
i
t

t
i
o
n
I
V

:

c
o
n
t
i
n
u
o
u
s

r
e
c
r
y
s
t
a
l
l
i
z
a
t
i
o
n

c
o
n
t
r
o
l
l
e
d

b
y

t
h
e
c
o
a
r
s
e
n
i
n
g

o
f

p
a
r
t
i
c
l
e
s
p

>

p
j
,

+

p

o
r

a
b
s
e
n
c
e

o
f

r
e
c
r
y
s
t
a
l
l
i
z
a
t
i
o
n
~

f
r
o
n
t
s
r
J
l
o
g
a
r
i
t
h
m
o
f

t
i
m
e
F
i
g
.

8
.
4

-

S
c
h
e
m
a
t
i
c

d
i
a
g
r
a
m

s
h
o
w
i
n
g

t
h
e

p
o
s
s
i
b
^
i
l
i
t
i
e
s

f
o
r

i
n
t
e
r
a
c
t
i
o
n
b
e
t
w
e
e
n

r
e
c
r
y
s
t
a
l
l
i
z
a
t
i
o
n

a
n
d

p
r
e
c
i
p
i
t

t
i
o
n

p
r
o
c
e
s
s
e
s
.
t
_

:

s
t
a
r
t

o
f

r
e
c
r
y
s
t
a
l
l
i
z
a
t
i
o
n
t
p

:

s
t
a
r
t

o
f

p
r
e
c
i
p
i
t

t
i
o
n

i
n

u
n
d
e
f
o
r
m
e
d

c
r
y
s
t
a
l

.
t

p

:

s
t
a
r
t

o
f

c
o
n
t
i
n
u
o
u
s

p
r
e
c
i
p
i
t
a
t
i
o
n

o
n

d
i
s
l
o
c
a
t
i
o
n
s
.
2 2
the result that the retarding forc p increases strongly with decreas-
ing temperatura. If discontinuous recrystallization is found to occur
just below T_ , the mobility of the reaction front and the resultant
driving forc p = pN + p - p will decrease with decreasing
temperature. A curve showing any given stage of the recrystallization
process becomes displaced to longer times. Below a certain temper-
ature Tg , p,, + p < p and the discontinuous process ceases. Dur-
ing this continuous process, the coarsening of the particles is rate-de-
termining (Fig. 8. 4. ). An increase in the concentration c at constant
dislocation density will cause a shift of the time t at which precipit-
ation sets in to smaller vales, if it is assumed that the change in the
concentration c has little effect on the beginning of recrystallization.
With this the vales of T., T and T~ are shifted to higher temper-
atures (Fig. 8. 5. ).
o*
-3
O
L_
O
QJ
(N=const)
Concentration c
Fig. 8. 5
I nfluence of concentration c on the recrystallization behaviour
of deformed, supersaturated solid solutions: Stage I normal
recrystallization, no precipitation; Stage n recrystallization
with subsequent precipitation; Stage HI discontinuous re-
crystallization with simultaneous precipitation; Stage IV con-
tinuous recrystallization.
2 2 9
An increase in the dislocation density N causes the time t to be dis-
placed to smaller vales, i. e. assuming that the start of precipitation
is not affected by the dislocations, the temperatures T? and T,, are
displaced to lower vales, and above a certain dislocation density re-
crystallization is always complete before the start of precipitation
(Fig. 8.6.).
I f, however, it can be assumed that an increasing number of dislocations
accelerates continuous precipitation, the behaviour becomes more com-
plicated. In the case of discontinuous precipitation, for example, an
increase in the dislocation density does not simply give rise to an
additional forc pN. These dislocations can in fact alter the super-
saturation ahead of the reaction front by making available low energy
nucleation sites, This displaces the incubation period for continuous
precipitation t_ (cont) to shorter times. It results that above a critical
dislocation density for a given time t. t (cont) < t.. Because of the
continuous precipitation which has already set in at the time t., the
chemical forc p is reduced due to the decrease in c. while the re-
*c
tarding forc p is si multan eously increased by the particles formed
at the dislocations. The result of superimposing these two effects is
that the resultant forc p = p,T + p - p on the reaction front at time
K N *c r
t- at a particular dislocation density N has the greatest possible valu.
It is in fact possible that above a given dislocation density at time t..
the condition for discontinuous recrystallization is no longer fulfilled
and instead of a combined discontinuous precipitation and recrystalliz-
ation reaction, continuous processes occur, i. e. individual precipit-
ation at dislocations and continuous recrystallization (Fig. 8. 7 a).
230
l n i
Fig. 8.6
Temperature dependence of incubation periods for precipt-
ation tp and recrystallization t_ as a function of the dis-
location density N , if it isassumed that the dislocations
have little effect on the beginning of precipitation:
Dislocation density N 1: no mutual influence of both pro-
cesses; dislocation density N ~ < N - : T< T- mutual in-
f l uen ce of precipitation and recrystallization. Schematic
diagram which is representative for Cu-Co alloys
( afterH.KRE 5 E , E .HORN BOGE N [ 7 j ).
231
Fig. 8.7
The forces acting at the reaction front as a function of
(a) dislocation density N and (b) aging time. Schematic
diagram which is representative for N i-Be alloys
(after H. KREYE, E. HORNBOGEN, F. HAESSNER [ 8J >
2 3 2
233
The conditions ahead of the reaction front are always time-dependent.
At constant temperatura the forces acting on the front are inluenced
with time by the following processes ahead of the reaction front. p..
decreases as a result of recovery; p decreases as a result of con-
tinuous precipitation ahead of the reaction front while p increases
for the same reason; p decreases as a result of coarsening of the
- p decreases, there- particles. The resultant forc p = p,, +
fore, with increasing annealing time until the condition for discontimi-
I ous recrystallization is no longer satisfied and the dlscontinuous reaction
* - comes to a standstill (Fig. 8. 7 b).
! ' i
1.3. Examples
With the help of the general principies discussed above, the re-
crystallization behaviour of deformed, supersaturated alloys can be
understood and classified qualitatively at least.
1. 3.1. Order of Occurence of Precipitation and Recrystallization in Al-Fe
[12] : Al-Fe alloys appear to be a simple subject for an attempt to
examine how far the schematic diagram in Fig. 8. 4 represents the
situation obtaining in real supersaturated and defect alloys. Useful in-
formation concerning the recrystallization and precipitation processes
can be obtained by combining the measurement of a suitable macroscopic
property e. g. electrical conductivity with light or electrn microscopy.
The first method is a very sensitive measure of the kinetics of anneal-
ing out even of extremely small amounts of dissolved atoms and of de-
fects. Since both effects lead to changes in the conductivity in the same
direction, this macroscopic method had to be supplemented by micros-
copic investigations.
Measurements of electrical resistivity during isothermal annealing
show (Fig. 8. 8) that at low temperatures the annealing out of dislocatiorj
and the precipitation of iron atoms occur simultaneously with the result
too
p/o
99
98
97
96-
95
tOO'
Pfo
99-
98-
97
96
l
PsX. 30OC
Rv
Adp/dlln*) \, V "
= 4 - X\-.
3 , .-.-" ' \
." " O .* .* .
< 2 . ,--' ' Vv
_) -1 X. ~>
UJ
aLn v
1 " E"
O-O1 O-t 1 1O tOO 1OOO f(h)
-^ 500C dp/d(lnj
Hs \e~X - ' ' f
E _2X. ,' --., -5 m
S \- A^
\ " 4
,' \-*
\LJ
,\\\
\t 1 5
' ' N. . ,^^
~" ^^ Ocr
i- , r- . ^'~' " - ' *.
OO1 O-1 t 1O tOO 1OOO (hj
Fig. 8. 8
Resistivity 9 of an Al-Fe alloy during isothermal ageing
at different temperatures. p ( ) and d 9 /dlnt()
plotted vs annealing time; 50% deformed alloys aged at
300 C (one-step reaction) and 500 C (two-step reacion).
1 2
R0) RC,: start and end of recrystallization; P_ , Pp : start
o r, o o
of precipitation before and after complete re-
crystallization. (after K. HOLM, E. HORNBOGEN [12J).
2 3 4
that there is only one mximum. At higher temperatures little or no
iron can preciptate during recrystallization. A second mximum
therefore occure after recrystallization is complete due to precipitation
starting by homogeneous nucleation in the recrystallized matrix. Under
favourable conditions recrystallization can nave finished completely be-
fore any precipitation can start.
In the temperature range in which recrystallization and precipitation
reactions proceed simultaneously, the former - as shown by optical
and electrn microscope investigations - can be continuous or discon-
tinuous. If the conditions for the formation and migration of recrystalliz-
fronts are fulfilled, the two types of reaction, continuous and dis-
continuous recrystallization, will always occur simultaneously. The
-X process of continuous recrystallization will reduce the drivng forc
< _
*\r discontinuous recrystallization because it decreases the dislocation
density. If the subgrains are able to grow relatively large before they
are consumed by the discontinuous process, a microstructure can re-
sult consisting of aggregates of grains formed by both discontinuous
and continuous recrystallization. Such microstructures are frequently
observed where the specimen has been aged at low temperatures.
1. 3. 2. Discontinuous Precipitation and Recrystallization in Ni-Be [8] : This
alloy is an example of the transference, with increasing deformation,
from combined discontinuous precipitation and recrystallization to
continuous processes, i. e. individual precipitation at dislocations and
continuous recrystallization.
In undeformed Ni-Be alloys discontinuous precipitation occurs in a
limited temperature-concentration range. The migrating grain boun-
daries acting as reaction front bring about precipitation of lamellae o
the equilibrium phase NiBe or a transformation of the already
2 3 5
precipitated metastable phase into the equilibrium phase. The reaction
ceases when particles of the stable phase can form ahead of the grain
boundary.
i
Deformation has the effect of accelerating discontinuous precipitation
by increasing the driving forc p.. and accelerating the formation of
the equilibrium phase ahead of the reaction front which requires a lower
nucleation energy at dislocations. At low degrees of deformation
( . 50%) the effect on discontinuous precipitation predomnales
and before the equilibrium phase can form at the dislocations thus halt-
ing the discontinuous process it has already swept through 70 to 100%
of the volume. At high degrees of deformation the high dislocation
density brings about a reduction in the Chemical forc p by virtue of
the associated high velocity of the reaction front with the result that
the velocity of the front does not rise quite as rapidly with further in-
crease in the dislocation density. On the other hand the high dislocation
density n addition to reducing the nucleation energy by increased
diffusion along dislocation channels also effects a further acceleration
of the front on the formation of the stable phase. This precipitates
out before the discontinuous reaction can begin. Discontinuous pre-
cipitation s thereby hndered and recrystallization proceeds continu-
ously.
1.3. Discontinuous and Continuous Recrystallization in Al-Cu Alloys [4, 6] :
The precipitation behaviour of this alloy is complicated but typical of
the frequently occurring case in which several metastable phases pre-
ciptate n addition to the stable phase; nucleation for these metastable
phases occurs preferentially at various defects, In the undeformed
solid solution the 0"-phase forms coherently in the perfect lattice,
the Q' -phase partially coherently at dislocations and the stable
Q -phase at grain boundaries. If no lattice defects are present a
237
I
2 3 6
direct transformation from the metastable to the stable phase occurs:
0" -- 0* -*- 0. In deformed alloys nucleation at lattice defects
predominates. All dislocations serve as nucleation points for 0'- .3
partiles so that precipitation of this phase begins as the dislocation
density increases. The recrystallization behaviour is further complica!-
ed in that the deformation structure in the specimen, i. e. the density,
configuration and type of dislocations produced by the deformation, is
not homogeneous. At low and mdium degrees of deformation, the
structure does in fact consist of a distribution of numerous dislocation $
spirals, dislocation rings and short, irregular dislocations, which be-j|
comes more uniform as the copper content increases, but at high
magnifications a band-like structure of high dislocation density zones
becomes apparent which become sharper as the degree of deformation "*
increases. These bands are obviously the boundaries of slightly mis-
oriented blocks. After a very short anneal these regions of very high
dislocation density can take on a structure similar to that of a high angle
grain boundary at which the 0-phase can form directly, Continuous O T
discontinuous recrystallization at these bands starts simultaneously
and extends thence into the intermediate regions (Fig. 8. 9).
(a) Continupusijrecrystallization; The dislocations produced by deform-
ation are very effectively immobilised by splitting whereupon 0' -
partiles form.
The density of the dislocations can change only by dissolution of the
0 ' -particles. At small initial dislocation densities isolated dislocatix.1
are found between the particles {Fig. 8.10.a), at high densities dis-
location networks (Fig. 8.10.b). Transformation of the largest 0'-
partcles into the stable 0-phase and re-solution of the others occur
simultaneously. An aluminium solid solution with low Cu-content is
in equilibrium with this phase. The dislocation networks are therefore
pinned not by segregations but by 0-particles. If a node is freed
Fig. 8.9
Influence of deformation structure on the recrystallization
behaviour of an Al-5% Cu alloy:
0% deformed, dislocation configuration.
(b) 50% deformed, 104 min at 28QC. 0 precipitation in
deformation bands together with 0 ' on individual dis-
locations.
(c) 90% defer,"'' 15 min at 240C. 0 precipitation and
subgrain forma, n in a deformation band.
238
by the dissolution of Its 0-particle the adjacent subgrain boundaries
can be annealed out by Y-node motlon or by subgrain rotation (Fig.
8.10 c).
(b) Discontinuous recrystallization: If the number of dislocations is
sufficiently large and the temperature high enough for the formation of
recrystallization fronts these can migrate under two different sets of
circumstances: .
(1) Ahead of the recrystallization front ' .ere are 0' --particles, dis-
locations or dislocation networks and rew 0 -particles at particular
sites (Fig. 8.11 a). The driving forc is then made up as follows:
ai _*. a Q
The 0'-*-0 transformation pro-
0-particles alr- *v p/esent liinder
P = PT .T ~ P
*
motes migraton of the front, ti
it (Fig. 8.12 a).
(2) Ahead of the recrystallization front there are subgrains or dislocation
networks and 0-particles (Fig. 8.11 b), i. e. pur recrystalliz.ition
239
occurs without superimposed discontinuous precipitation:
(Fig. 8.12b).
~ p "
It is possible to predict which conditions are necessar> . _ r which pro-
cess from the general principies discussed above. Due to differenef
in the deformation structure, regions with often quite differp'
crystallization structure are found in one and the same sppcimen. AH
increase in the dislocation density leads both to the formation re-
crystallization nuclei and to nuclei for the stable 0-phase. This has
the result that the discontinuous recrystallization reaction '1) can
occur only in a mdium dislocation density range and at not coo high
temperatures, i. e. under conditions at which several recrystallization
fronts can form but not too many 0-particles can form ahead of the
front. This reaction does not therefore occur at high degrees of de-
formation and supersaturation. Reaction (2) can be expected when the
Continuous recrystallization in Al-Cu alloys.
(a) Isolated dislocations are pinned by 9 * -particles.
Al-3% C u/ l , 5h 200C.
(b) Dislocation networks between 9 ' -particles.
AI-3% Cu/ 25% deformed/isochronal up to 370C.
(c) After the transformation O ' -* 6 subgrain
boundaries are pinned by 9 -particles; after dis-
solution of single particles the adjacent subgrains
can become more perfect by Y-node motion or by
subgrain rotation. Al-5% Cu/ 50% deformed/
104 minat280 C.
240
241
Fig. 8.11
Schematic sketch of discontinuous recrystallization.
( direction of motion)
(a) The recrystallization front sweeps out dislocations
and subboundaries and also transforms 0' to 9 .
(b) The recrystallization front sweeps out dislocations
and subboundaries.
Fig. 8.12
Discontinuous recrystallization in Al-Cu alloys.
(a) Al-2% Cu/ 90% reduction in rea/ isochronal up
to 300C;
(b) Al-5% Cu/ 50% reduction in rea/ isochronal up
to 400C.
242
condition pN > P is fulfilled at high dislocation densities. At constant
dislocation density and temperature and increasing supersaturation the
rearrangement of dislocations to recrystallization nuclei becomes in-
creasingly hinderedby O' and 0 particles and continuous recrystalliz-
ation is to be expected.
In the case of Isochronal heating the transition from the continuous to
the discontinuous reaction is shifted to higher temperatures. This is
tecause continuous recrystallization starts first and is replaced by the
discontinuous process only after particles that have formed at lower
temperatures have partially dissolved as a result of continued heating.
At very slow rates of heating, discontinuous recrystaization can be
avoided completely.
INFLUENCE OF DISPERSED PARTICLES
ON RECRYSTALLIZATION
If an alloy containing particles of a second phase is plastically deform-
ed there are two basic possibilities for the behaviour of the particles:
(a) The dislocations in the matrix can run through the particles, split
;hem and by the operation of many slip systems and after a high degree
of deformation lead virtually to the solution of the particles. The
situation is then the same as that discussed in section 1.
(b) The particles possess mechanical properties such that they cannot.
be plastically deformed.
Many high-temperature-strength alloys such as SAP (sintered al umi ni um
powder) or TD nickel (thoria dispersed nickel) have a microstructure
consisting of a matrix containing a high density of dislocations and a
finely divided dispersed second phase. As a rule a hardened material
of this type recrystalhzes at high temperatures and thus rapidly loses its
hardness. To maintain strength at high temperature, the dispersed phase
must retard recrystallization of the matrix (see Fig. 8.13). The recrystall1
ation behaviour of such materials is thus of great technical 'interest.
2 4 3
K>-
200
600
Anneoling Teirptr
800
1000 1200
Fig. 8.13
Room-temperature hardness after 1 hr anneal for
copper powder alloys internallv oxidized at 95C
id extruded. (after O. PRESTON, N. J. GRANT [17])
'I
244
Compared wtth the behaviour of a matrix without partiles, the dis-
persad particles can effect either an acceleration or a retardati
recrystallization. Some examples are given IB the following Table 1.
Largely independent of the alloy, an acceleration of recrystallization
ls observed for wide spacing and large diameter of the particles
( X > 1 /i, diameter > O , 3 m) and a retardation for clos<
spacing'andCmaU diameter ( X <1 /-, diameter < O , 3 ^m).
effect can be explained In general terms by considering that
particles will have the effect of increasing the dis^ation density when
the specimen is plasticaUy deformed thus increasing the driving or,
for recrystallization. O n the other hand the partiles hinder both the
rearrangement of dislocations and dislocation networks to mol
high-angle grain boundaries and the migration of these, leac
retardation of recrystallization.
2.1. Grain Boundary Pinning
According to the Zener model discussed in section 1, particles of
a secondphaseexert a retarding forc ?r on a mobile high-angle
grain boundary. For high degrees of deformation and coarse c
persions the effect of this retarding forc on the course of recryst
ation can be neglected. For a dispersin with f = O , 04, g
d. = 2/um and a boundary surface energy ^- 500 erg/c,
retarding forc is given by
JB
d.
c 2
= 3.10 dyn /cm
which can be neglected compared wit h a driving forc PN of approxi-
mately 107 to 108 dyn/cm2. At low degrees of deformation or whe
the partile diameter is of the order of 0,1 /.m at a volume frac,
f ^0 0 5, the retarding forc has a considerable effect on a migrat
24 5
c
.2
a
u
at
tu
IH
c
o
i
0)
"3
u
u
rt
a""
OJ V
u cu
S
Lri Z
5 .2
n. T J
CU
y ^
% *
a t
&
2
= &
>>
o
-^
rt
i i+ + + + + i
+ + ii , ,
B 1
^^ \<^O 05
M eo o o O o
_, ^ - O O i O
" O O co tn M
E
^
e e 7 o<|^
^ m C3 "^
A V ^^
i-5P -^ '
W i-H CN csj M
L_ f | | | j 1 _} ( _J
0 0 MO 0
tj W O O f rH
eo 5 u w <
+ + + + + +
3 ^H 3 bC D
o <o <: o
+
s
o
e-T
,
co
g
o
T -*
rjn
tL
u
(H
rJ
^
+
<
1
2 4 6
ing high-angle grain boundary. Since the mean partile diameter d. ,
the mean spacing between particles X and the volume fraction f are
connected by the equation
.
all the usual dispersions with such small partile diameter are associat-
ed with relatively cise partile spacing.
2.2. Influence of the Partile Spacing on Nucleation
Investigationsby R.D.DOHERTYandJ.W.MARTIN[18, 19] on cold
rolled dispersions of Al + AUCu alloys have shown that nucleation is
the rate determining stage of recrystallization {Fig. 8.14) and that
phase separation is far more important than partile size in determin-
ing recrystallization behaviour {the partile sizes in this case were
however in most cases greater than O, 25 Aim). Widely spaced
particles cause an acceleration of recrystallization as a result of a
higher work hardening rate in the two-phase alloy. During subgrain
growth the angle and mobility of the subgrain boundary increase until
high-angle mobile boundaries form. For easy nucleation in two-phase
alloys the subgrains must have reached this critical misorientation
before they impinge upon surrounding particles. If the subgrain
boundaries are pinned by particles before a grain boundary capable
of migration has been able to form by rearrangement of dislocations
or dislocation networks, the adjacent subgrains will be able to grow
to a comparable size thus reducing the driving forc for further growth.
Any further growth of the subgrains is then dependent on coarsening
of the 0 -phase (Al_Cu), i. e. we have the case described in Part I of
t
continuous recrystallization controlled by partile growth. Henee in
alloys with coarse dispersin the rate of nucleation is rapid and not
2 4 7
very sensitivo to changes in X, while below a critical valu of X
(O, 8 to 1, O / um) the rate of nucleation decreases very rapidly with
X .
Influence of Deformation Structure and Partile Size on Nucleation
A series of recent invcstigations [26 - 30]' has shown that although
the particles of a second phase indubitably hnder the rearrangement
of dislocation networks to grain boundaries capable of migration and
the migration of these also, the resistance to recrystallization is de-
termined essentially by the structure produced during deformation.
Investigations on TD nickel [21 - 24] have shown that this alloy attains
its full heat-resistance if it is repeatedly strained and "recovery-
annealed". Each yele rearranges the dislocation distribution in such
a way as to enhance resistance to recrystallization and thus improve
high temperature strength, In contrast, a 90% cold-worked TD nickel
specimen recrystallizes at about the same temperature as pur nickel.
The special strain-anneal cycling obviously produces a dislocation con-
figuration resistant to recrystallization.
J.L.mifl/ HALLet al. [26] have found that the presence of MgO partiles
in Ag {diameter - 50 A, X 500 A ) leads to a marked reduction in
both the random and long range variations in orientation in the deformed
structures. The structure of the deformed dispersion-hardened alloy
is characterised by an unusually high and relatively uniform dislocation
density, which appears in the early stages of deformation. After severe
deformation, however, crude cell structures were observed. This cell
sructure was noticeably smaller and less well-defined than in pur
sver. These differences are apparently related to the influence of
particles on the nucleation of dislocations on several different slp
2 4 8
I
f ** 3- O
rt- <
ET W
(D 1
Eg
S
^ p -
i
f l > o
MI
S
'
' n o <
re o a j
. OiQi -j
. i -j "o M.
><] M pj
ft re E
T n o
-3= "
:?a
&~-
& e a .
o csr
3N 2
"l|
rt- =
3- M . K
a >ro
3' ^49
Ene
re e 2.
M
^> O u
K ^ tS
t P
>-"
. H 3
S^f t
0re ^
w 3*
S-^,3
1*
3UlO
O"<
(
ff 3' & S2
^ &
O (D
I!
sro
s&
bu ^
^S
re i i o
interparticle spacing A m
.
en O)
O
Q **
O*
3' S
in * *
t
interparticle spacing inym
'O *!o ^
O
o
-i
Q
o
249
s ys tems a t low va les of s tra in, lea ding to a grea ter dis tribution of the
type a nd s ign of dis loca tions in ma teria ls conta ining dis pers ions . The
prevention of long ra nge motion of dis loca tions a lso prohibits the con-
grega tion of ma ny disloca tions of the s a me sign to form cellbounda ries
with la rge mis orienta tions . The la ck of a cells tructure with a s igni-
fica nt a mount of loca lcurva ture contributes to the difficulty of s ubgra in
forma tion. The s ma llinitia lmis orienta tion between the s ubgra ins a nd
their s urrounding res ults from the s ma llma gnitude of the mis orient-
a tion in the cells tructure of the s a mples conta ining dis pers ions . The
la ck of la ttice curva ture prohibits a ra pid increa s e in mis orienta tion
during growth of the s ubgra ins by a ccumula tion of the mis orienta tion
in the cons umed cells tructure. This, plus the low mobility of the low
a ngle bounda ries due to pa rtile res tra int of the dis loca tion rea rra nge-
ment proces s es neces s a ry for their migra tion produces a s itua tion in
which a number of s ubgra ins a re formed ea rly in the a nnea ling proces s
but do not grow to cons ume their environment. Annea ling proceeds by
the very slow growth of thes e s ubgra ins plus infrequent nuclea tion of
a dditiona ls ubgra ins .
Inves tiga tions byj . W.M ARTIN et a l. [27, 28] ha ve s hown tha t the pre-
s ence of Co or SiC> 2 precipita tes in copper crys ta ls (dia meter -800
a nd 1500 R res pectively, X ~ 0. 5 a nd 1, 7 ^um res pectiveiy) dur-
ing deforma tion tends to increa s e the homogeneity of dis tribution of
dis loca tions , thus res ulting in a reduction in the la ttice mis orienta tion.
Thus the cha nce of forming a s uita ble recrys ta lliza tion front on a nnea l-
ing is reduced even a bove the pa rtile s olidus Une for the Co-pa rticles .
Recrys ta lliza tion occurs only a fter hea vy deforma tion, a s s ocia ted with
a n increa s ing ra te of work ha rdening. Les s hea vily deformed s peci-
mens f orm s ta ble dis loca tion configura tions . Ana lys is of hes e net-
works ha s s hown tha t the s tructure ca n be cons idered to compris e
complex tilt bounda ries on pla nes a pproxima tely perpendicula r to the
prima ry s lip pla nes a nd s heets of networks on the prima ry s lip pla nes .
2 5
More recent investigations by T. CRQLL^ON and J. W .MA RTI Ni [29] on
the infl uence of particle size on deformation structure and recrystalliz-
ation behaviour have shown that misoriented subgrains associated with
the dispersed phase f orm on annealing only if the dislooation inter-
actions during straining lead to tha development of local lattice curva-
ture in the matrix at these regions. Two types of dislocation-particle
interactions were lound to have occurred, depending upon the size of
the particles. Dislocations had either f ormed as secondary glide loops,
involving local lattice curvature adjacent to each particle, or - when
the dispersed particle size was small (i. e. < 3000 A ) - existed as
prismatic dislocations, not involving local curvature of the matrix
lattice. I n that case no such recrystallization embryos were observed.
2.4. I nfluence of Particle Diameter and I nterparticle Spacing on Recrystalliz-
ation
The recrystallization behaviour may be considered as being influenced
by the following factors (see Fig. 8.15):
The dislocation density in the crystal is increased by non deformable
particles of a second phase by the amount of N OC-j
compared with a matrix without particles. The deformation structure
is f urthermore dependent on particle diameter: Large particles give
rise to local lattice curvature and henee to favoured nucleation sites
for recrystallization. Small particles ( < 3000 A )lead to an increas-
ed homogeneity of the dislocation distribution thus impeding nucleation.
Cise spacing of particles ( X 1 / un ) impedes the rearrange^
ment of dslocations or dislocation corigurations to grain boundaries
capable of migration.
A retarding f orc p
- r
boundary by the particles
f y
c^ar ^B
is exerted on a migrating grain
of a second phase.
2 5 1
On the basis of an increased dislocation density and the formation of
favoured nucleation sites, the recrystallization in coarse dispersions
(large particles, wide particle spacing) is found to be accelerated as
compared with a matrix without particles. Reduction in the particle
diameter (and henee also in the distance between particles) does in
fact lead to a further increase in the dislocation density but because
of the greater homogeneity of the dislocation distribution and the in-
creasing impedance to dislocation rearrangement and grain boundary
migration the rate of recrystallization is considerably reduced.
acceleration of
recrystallization
'single phase alldy
retardation of
recrystaltization
interparticle spacing
Fig. 8.15
Schematic diagram on the inf l uence of dispersed particles
on recrystal l ization.
252
The case must also be mentioned in which the partiles can be deformed
with the matrix but are not dissolved. In a case luce this recrystalliz-
ation is complicated even further. V.A.PHIIUPS [31] has found, for
example in Cu-Co alloys, that particles (diameter = 100 to 600 A)
deformed fairly homogeneously with the matrix becoming lamellae
after 95% cold-rolling. These particles greatly impeded softening
and recrystallization. The greate-t offect was found with particles
of ca. 300 A diameter, the size at which the particles lose their
coherence with the matrix.
3. RECRYSTALLIZATION BEHAVIOUR OF COARSE,
TWO-PHASE ALLOYS
Until recently very little was known about the recrystallization be-
haviour of coarse, two-phase alloys, suchas O f / / 3 -brass, in which
the phases effectively f orm seprate grains. R. W. K. Honeycombe and
W. BOAS (1948) [3] and L. M. CLAREBOROUGH (1950)[33] working
principally with or / / ? -brass have found an extremely complex re-
crystallization behaviour without being able to interpret their results
completely.
HONEYCOMBE and BOAS have studied the deformation and recrystalliz-
ation of a dplex brass containing 60% Cu and 40% Zn using micros-
copio and X-ray methods. In compression and tensile tests the or -
phase starts to def orm plastically at a lower stress than the / 7-phase,
and even after heavy deformations, it is def ormed more severely.
Microscopio examination shows that the ot-phase starts to recrystallize
at random whereas the /?-phase tends to nucleate at the boundaries
between the two phases. Which of the two phases starts to recrystallize
f irst on annealing depends on the annealing treatments prior to de-
formation. In wires quenched f rom 600C the of-phase starts to
253
recrystallize at about 100C lower than the fl-pba.se, whereas in slow-
ly cooled wires the /3-phase starts to recrystallize before the cr-phase.
The authors assume that this behaviour can be interpreted on the basis
o the order-disorder transformaron in /?-brass. L. M. CLAREBROUGH,
however, has shown that partile precipitation in /3-brass during
annealing must be the controlling factor for this behaviour, without
however understanding the influence of the particles on recrystallization.
The controlling factor in the recrystallization behaviour of coarse, two-
phase alloys is the deformation structure, that is, the type, density
and configuration of the dislocations in the grains of the two phases, under
certain circumstances the occurence of deformation-transformations
(e. g. in /3-brass). The deformation structure depends on the different
mechanical properties of the two phases. It is influenced by the total
deformation of the alloy and by the volume fractions and grain diameters
of the phases. It can f urther be influenced by precipitation occuring in
one or both of the phases prior to deformation. It can be assumed that
the recrystallization behaviour of coarse two-phase alloys could be
elucidated completely by a precise electrn microscopical analysis of
the deformation structure with the aid of the principies described in
Part I of the interaction between precipitation and recrystallization.
No electrn microscope studies of the recrystallization of such alloys
have however yet been reported.
Acknowledgement
I wish to thank Professor E. Hornbogen for numerous valuable dis-
cussions and suggestions.
2 5 4
REFERENCES
1. R.W. Cahn in:
2. K. Detert in:
3. E. Hornbogen and
H. Kreye in:
4. U. Kbster and
E. Hornbogen:
5. H. Kreye:
-Y6. H. Ahlborn,
E. Hornbogen and
U. Kbster:
7. H. Kreye and
E. Hornbogen:
8. H. Kreye,
E. Hornbogen and
F. H
9. E. Hornbogen:
10. H. Kreye,
E. Hornbogen and
F. HaeJner:
11. E. Hornbogen,
E. -M. Minuth and
E. Blank:
12. K. Holm and
E. Hornbogen:
13. J.W. Cahn:
14. S. Horiuchi and
I. Gokyu in:
15. G. Zener, cited in:
16. M. T. Ashbyand
J. Lewis:
"Recrystallization, Grain Growth and
Textures", H. Margolin, Ed., A. S. M,,
Metals Park, Ohio 1966, p. 109.
"Recrystallization, Grain Growth and
Textures", H. Margolin, Ed., A. S. M.,
Metals Park, Ohio 1966, p. 193.
"Texturen in Forschung und Praxis",
J.Grewen and G. Wassermann, Ed.,
Springer, Berln 1969, p. 274.
Z. f. Metallkde. 5f) (1968) 792.
"Rekristallisation und Ausscheidung",
Dissertation Stuttgart 1968.
J. Mat.Sci. 4 (1969) 944.
J.Mat.Sci. 5 (1970) 89.
phys.stat.sol. (a) 1_ (1970) 97.
Praktische Metallographie 9 (1970) 349.
Aren. Eisenhttenwes. 41 (1970) 439.
Arch, Eisenhttenwes. 4jl (1970) 883.
J.Mat.Sci. 5 (1970) 655.
ActaMet. 7 (1959) 18.
"Texturen in Forschung und Praxis",
J.Grewen and G. Wassermann, Ed.,
Springer, Berln 1969, p. 312.
C.S.Smith: Trans. AIME 175 (1948) 15.
Harvard Univ. Techn. Rep. 547 (1967) .
2 5 5
0. Preston and
N.J.Grant:
R. D. Doherty and
J.W. Martin:
R.D. Doherty and
J.W. Martin:
F. Haeiner,
E. Hornbogen and
M. Mukherjee:
J. E. White and
R.D. Carnahan:
ID. Webster:
LE. R. Kimmeland
M. C. Inman:
LD. Webster:
i P. R. Mould and
P. Cotterill:
LJ.L.Brimhall,
M. J. Klein and
R. A. Huggins:
' . F. J. Humphreys and
J.W. Martin:
I T. C. Rollason and
J. W. Martin:
I T. C. Rollason and
J.W. Martin:
f N.Ryum:
L V . A. Phillips:
I R. W. K. Honeycombe and
W.Boas:
LM. Clarebrough:
Trans. Met. Soc., AIME 221 (1961) 164.
J. Inst. Met. 91 (1962/63) 332.
Trans. Met. Soc., A. S. M. _ 57_ (1964) 874.
Z. f. Metallkde. 57(1966) 171.
Trans. Met. Soc., AIME 230 (1964) 1298.
Trans. Met. Soc., AIME 242 (1969) 390.
Trans. Met. Soc., A. S. M. 62_ (1969) 390.
Trans. A. S.M. 62_ (1969) 936.
J.Mat.Sci. 2 (1967) 241.
Acta Met. 1A (1966) 459.
Phil. Mag. 17 (1968) 365,
Acta Met. Ij (1970) 1267.
J.Mat.Sci. 5 (1970) 127.
Acta Met. 17 (1969) 269, 821, 831.
Trans. Met. Soc., AIME 236 (1965) 1302.
Austr.J. Sci. Res. A^ (1948) 70.
Austr.J. Sci. Res. A3_ (1950) 72.
256
257
CHAPTER 9
RECOVERY AND RECRYSTALLIZATION
DURING DEFORMATION
BY
K. MECKING AND F. KIRCH
1. INTRODUCTION
During plstic deformaion, dlslocation multiplication occurs by the
action of dislocation sources. In spite of the continuous operation of
these sources, however, the dislocation density does not reach the
mximum limiting valu which can be contained in the lattice. Instead,
the importance of the annihilation processes n competition with multi-
plication increases at high dislocation densities. Since they proceed
during deformation, these annihilation processes are called dynamic.
They are responsible for the attainment of a saturation valu of the
dislocation density at high degrees of deformation.
For dynamic recovery to occur, it is in general neceasary for dis-
locations to be able to leave their original slip-plane in order that
mutua! annihilation with dislocations of opposite sign may take place
or that they may take up an energetically more favourable position.
Screw dislocations achieve this by cross-slip and edge dislocations
by climb. Recrystallization transforms the defect lattice with a high
dislocation density produced during deformation into an undistorted
lattice by grain boundary migration.
258
All these processes are thermal l y activated. Which of the possible
processes are involved and their relative importance thus depends on
the temperature and strain rate as well as on the degree of deformation.
In practice it is simpler to consider the more easily measurable flow
stress tf than the dislocation density N. According to all the theoretic-
al models [l to 7] u' and N are related by an equation of the form
(1)
in which Q is a constant, the numrica! valu of which vares only
negligibly in the different models ( r\ 1), G is the shear modulus,
b the Burgers vector and N the dislocation density. (The flow stress
O 1 is defined as the tensile stress (or uniaxial compressive stress)
1 1 * f i 1
to cause plstic flow at a mdium rate ( 10 sec > f > 10 sec ).
In the case of single crystals, the shear stress T* in the most favour-
able shear system is usually given instead of the normal stress ff1 ;
as a guide, (f 3
Fig. 9.1 gives some idea of the influence of the deformation temperature.
It shows the stress TF at the end of elongation without reduction of
rea for a series of f. c. c. metis as a function of the deformation
temperature ( in units of the melting point temperature) . The two
branches of the curves indicate that different processes are in oper-
ation at T < T / and at T > T /. (T = melting point) .
m/z m/ m
Furthermore there are obviously differences between the various
materials. If the identical dislocation density were set up regardless
of material (see equation 1 ) the curves of f_ /, against T/T
Jb/u m
should all coincide. The reason for the discrepancies is that the
degree of difficulty encountered by climb and cross slip varies with
the metal. In the case of f. c. c. metis, differences in the stacking
fault energy are responsible. According to the stacking fault energy,
259
t> r
the dislocations split into partial dislocations of different width [9, 10] .
As the separation of the partial dislocations increases ( i. e, decreas-
ing stacking fault energy} the dislocations become more strongly bound
to their slip planes rendering cl imb and cross-slip processes more
difficult. Henee the curves in Fig. 9.1 lie in the order of relative
stacking fault energy E /Gb.
5
In the following, the processes operating in the various temperature
ranges will be discussed in more detail, mainly with reference to
f. c. c. metis since these have been studied particularly thoroughly.
2 . FORMAL DESCRIPTION OF WORK HARDENING BEHAV1OUR
Stress-strain behaviour where hardening and softening processes
are proceeding side by side can be described formal l y as follows
[11 to 14].
Movement of dislocations gives rise to macroscopic deformation and,
as a result of changes in the dislocation configuration and density, to
changes in the flow stress. In the time interval dt, the work harden-
ing processes ( suffix h) give rise to the strain and flow stress incre-
ments df, and da1, while the recovery processes ( suffix r) give
rise to df and -da1 . Thus the total changes df and do'are
given by
df = df.
r ; d^da-h -
(2)
Using the abbreviations doVdf = 6 = slope of the stress-strain
curve and do" , /df. = 6. , the two equations above can be combined
to give
o'r
9, (3)
260
T / T r
Fig. 9.1
Flow stress corrected for shear modulus T*F/G at the
end of elongation without rea reduction as function
of the homologous temperature for Al, Au, Cu and Ag
single crystals of easy glide orlentation. after [8] .
2 6 1
The fundamental mode of operation of the recovery processes can be
explained with the help of this equaton. Let us consider the boundary
cases:
1) The fraction of the strain caused by recovery processes is small
fr ** * " * ! / 6h '* the emPhass is on fr An example of this
is the process in whch edge dislocations of opposing sign climb to-
wards one another perpendicular to their slip plae and annihilate
one another e. g. as in the annealing out of dislocation dipoles. There
can be a large change in the flow stress without significan! associated
strain. Recrystallization is another such process.
2) The change in stress due to the recovery process is slight
( fr ^ ^r/ch ' but alrnost a11 the strain is accounted for by the
dislocation movement associated with the recovery. Such a case is
possibly observed in the cross-slip model, in which recovery takes
place by screw dislocations cross slipping to by-pass obstacles, mov-
ing over long distances and then suffering annihilation with dislocations
of opposing sign.
The actual cases will be mixed, with the emphasis laid more on the
one or the other term in eq. 3 according to the process.
3. CROSS SLIP
First to be considered is the recovery process operative at low
temperatures fT < T , ^. If a single crystal is deformed in this
temperature range, the well known work hardening curve is obtained
with the stages I, n and IH. Figure 9. 2 shows the curves for a silver
and a copper single crystal deformed at 295K. Stage I will not be
considered here. The course of the curve in stages n and ffl can be
explained as follows [3,17] . At small vales of strain, dislocations
2 6 2
pile up at obstacles. The stress fields of these pile ups are the cause
of the high work hardening rate in stage H. At the higher stresses in
stage mit becomes increasingly possible for screw dislocations to
leave the pile ups by cross slip and combine with dislocations of
opposite sign. The annihilation processes set off by cross slip are
increasingly favoured by increasing external stress and the work
hardening curve flattens off. At high stacking fautt energies because
of the narrow stacking fault width cross slip sets in at lower stresses
than at low stacking fault energies. This is the reason why the Cu-
curve lies below the Ag-curve. Acceptance of the cross slip mechanism
as controlling process was based originally on microscopic observations
o slip lines. The higher the degree of deformation the greater the ex-
tent of cross slip traces observed (Fig. 9. 3).
These microscopical results provided the inducement for applying the
cross slip theory developed by SCHOECK and SEEGER and WOLFF [18,19]
to the beginning of stage IH. It could beshown in numerous cases that
this theory was able to give a good description of the temperature
and rate dependence of the starting stress T"ra of stage m. MECKING
and LCKE [12, 20] were able to show that not only the start but the
whole of the work hardening curve in stage m can be understood on
the basis of the cross slip theory.
If in eq. 3 f , the part of the strain rate caused by recovery, and
& , the stress reduction caused by recovery are set proportional to
the frequency with which recovery events occur, we have
(1 - . ) or e*p ( -
yh
(4)
In this equation the temperature and rate dependence of the work harden-
ing curves yield an activation energy Q strongly dependent on the tempr
ature and degree of deformation or flow stress. This dependence is
2-33
Fig. 9. 2
Work hardening curves of a Ag and a Cu single crystal
with T/G plotted against the shear ff .
Fig. 9. 3
Slip lines ont he surface cf a Cu single crystal eformed
in Stage I. The slip bands runnin< r paraUel to the primary
slip plae are interconnected hv f i ne slip traces along the
cross slip plane.after [15J.
265
in aix reemenl with the cross slip theory. Figure 9. 4 shoivs the activ-
ation energies ubt ained in (his way for the work hardening curves in
Fig. 9. 2 as a function of strain.
Activation energies srongly dependent on temperature and applj ed
stress have also been observed in this temperature range in creep
tests [21 - 23] . It can be inferred froni this that the creep velocity
is also cont r' t l l ed by cross slip processes since strong stress or
l emperil ure dependence of the activation energy is typu'.al of these
processes.
4. DISLOCATION CLIMBJ=OLYGONIZATION
As illusrated in Fie. 9. 1 , different recovery mechanisnis oprate
at hi^ h deforruation temperatures T > T /2 from those in the low
l emperat ure range. Studies of dislocation structure show that dis-
location climb is occurring. Figure 9. 5 shows the dislocation sruc-
t ure of a coppsr single crystal strained at 875 K rendered visible by
t h etr! .^je. A polygonization structure of this type is
f vpi cal of hi/h t emperat ure deforni^tion [24 to 27j and is the result
The t endency to subgrain f'"-mation in f. c. c. metis depends strongly
un t hf stacking fault energy of t i i e metal, since widely spHt dis-
locations flow SFE) cainiot cl imb as easily as only slightly split or
pertect dislocations (high SFE). An example of this is illustrated
in Fitr 9. (!, which shows Laue phofographs of Al, Au, Cu and Ag
single crystals after almost identical strain at T/Tm - 0; 7. In the
case of AJ and Au, the two metis with high SFE, the Laue reflections
are split into many individual spots as a result of strong polygonization.
266
O C U tt 1Z
Fig. 9.4
Activation energies for dynamic recovery as function
of the degree of deformation for the work ' .dening
curves shown in Fig. 9. 2. The activa ,t energies
were calculated according to [12] and ["20],
Fig. 9. 5
Polygonization structure in a copper single crystal
strained at 875 K rendered visible by dislocation
etching, after [24].
2 6 7
The f orm of the reflections on the Ag and Cu photographs are however,
similar to those found at lower deformation temperares. The streaks
are the result of kink band formation which s favoured by these crystal
orientations [8, 28] . The following Fig. 9. 7 shows Laue photographs
of Al after various degrees of deformation also at T/T = 0,7. K
can be seen that in the case of Al deformation bands are^formed at
small strains but polygonization is favoured as the deformation in-
creases.
As an example o the stress-strain behaviour in this temperature range,
Fig^ 9. 8 shows the work hardening curve of an Al single crystal at
685K. The horizontal portion at high strains and the stress mximum
occurrW1-- certain temperature ranges are typical and were also ob-
served in hot-torsic- -xoeriments [29] In this temperature range
creep tests have always been us- r quantitative analyses [27] . On
the basis of the numerous available experimental results it is general-
ly assamed that dislocatioi, ^ govems the recovery processes in
high temperature deformation but the controlling mechanism has not
been elucidated.
The numerous theoretical models can be divided into two categories on
the basis of their fundamental assumptions:
*) The essential process is the climbing movement of those dislocation
ients with predominantly edge character. According to the model
special importance is attributed to the annihilation of dislocation groups
.3 0, 3 1] , dispersin of dislocation dipoles [32] , climbing over
obstadas togde [33] or the strain resulting from the climbing pro-
cesses themselves [34] .
b) Th e essential process is the movement of screw dislocations con-
ni ngj ogs [29, 3 5, 3 6]. These jogs have edge character and the
velocity of the screw dislocations depends on how rapidly the jogs climb.
268
Fig. 9. 6 iiB.^.
Laue points f r om Ag, Cu, /iu ai.^ .u.
af ter def or mation at the homologous tem"*1
T/T = O, 7. Back refiection p'' ' ^H
' m 7
.^^.1 \~ \
269
> f
-10"
E=25% e=55'/.
poiL. _ ^ f^ single crystal after various
strains at T/Tm = 03 7, after *[8],
0.3
'
^_
Al
, .* -
Y: VIO sec
665 K
1.0
Y
Work hardening curve of an Al single crystal.after
[8] . Def or mation temperature 685K.
270
Although the qualitative interpretation of
these modela is very different, quantitatively they lead to very simil ar
f and tf in eq, 3 by
relationships. Independer of whether f or is the controlling
parameter, the right-hand side of eq. 3 is proportional to the speed
with which the edge dislocations can climb. This is proportional to
the self diffusion coefficient since the climb velocity depends on the
rate at which vacancies diffuse towards or away from the dislocations.
Thus all models yield an equation of the form
D
3D
(5)
for the velocity of stationary creep ( 9 = O in eq. 3) which has re-
ceived most attention in the past and to which we shall limit ourselves
in the following. D is the self diffusion coefficient. In all the
oU
theories the function iff is strongly dependent on the applied stress
yr
yroro' with 3 < K < 5 and possibly also weakly dependent on
temperature. This dependence is differently predicted by the various
models providing the opportunity for deciding experimentally between
the different theoretical possibilities. Unfortunately the experimental-
ly determined relationships do not unequivocally favour a single n> odel
[38] since none of the theoretical expressions corresponda to the
experimental results over a wide range of stress o", The results
are best described by the empirical relationship [39]
' = B . [" sinh ( cu - a1) 1 K . exp ( - - (6)
where B, tu and K are constants. Figure 9. 9 shows that with the
aid of this equation the experimental results for Al from creep-torsien
and extrusin experiments, that is for the widest possible range of
temperaures and velocities, can be described with a constant valu
for Q. The valu obtained for Q is in the regin of the self diffusion
ncrgy U= 1, 63 eV/atom.
271
O
Q.

u l m
10 K
10
Itf
103
102
1(f
10
lo2
I 1 r
>" 3 2 0 - 6 1 6 " C j
i<m25Q-5SO-C
Torsi n I SS- BSO' C
20- 53- C
SLOPE i67
299-10'S.sJ'1
QO73k c o l / mo l
10
10
10
io7
1l ,
10"Z KT1 10 101 102
sinh((jj ffj
Fig. 9.9
f/B ' exp (-QAT) as function of sinh (cucT) for Al for
a wide variety of deform-Mon conditions,afer [46]
272
The dissatisfactory feature of this method is that in order to fit eq. 6
to the experimental results three empirical constants w , B and K are
required, the theoretical interpretation of which is difficult. An
essential result inthese considerations is that the essential recovery
process in this temperature range is diffusion controlled dislocation
climb. This is generally valid for f. c. c., b. c. c. and hexagonal
metis and for single crystals and polycrystalline material.
As already seen from Fig. 9. 6 the recovery processes in f. c. c. metis
become more difficult as the stacking fault energy decreases. This
can be seen particularly clearly from the vales of the rate of station-
ary creep, which is plotted in Fig. 9.10 as a function of the relative
stacking fautt energy :Eg/Gb.The creep velocity is corrected for
temperature: o'/G =10~4 for all the metis shown. It is seen that as
a function of E /Gb the creep velocity measured at comparable
temperaturas can vary by several powers of ten. This means that
the quantity yin eq. 5 depends strongly on the stacking fautt energy
(as well as on the stress).
5. DYNAMIC RECRYSTALLIZATION
For metis with a relativeiy high stacking fault energy, such as for
example aluminium [29, 40, 46] , recovery is the only softening
mechanism observed during high temperature deformation either by
creep or at technical deformation velocities, many powers of ten
higher. In the case of metis with low stacking fault energy such as
copper, recrystallization also comes under consideration as a soften-
ing mechanism [41 to 46] . Recrystallized grains in the deformed
matrix have been observed in polycrystalline material after high -
temperature deformation. Furthermore copper and nickel creep

10*
10-'
ic *
10
-u
10
- 1 5
273
O 100 200 300 400 500 600
Gb/ Es
Fig. 9. 10
Stationary creep rate (corrected for temperature and
shear modulus) as function of the relative stacking
fault energy E /Gb. From [27].
s
2 7 4
curves show characteristic strain rate cycles which could be connected
with recrystallization proceeding during deformation [45] .
In the case of polycrystalline material it is dificult to decide whether
the observed recrystallized grains arse during or after deformation
[47, 48] . In the following, therefore, the results of tensile tests on
single crystals ("8, 49] will be used in a discussion of this question.
It is easy to determine whether recrystallization has taken place in
single crystals. Furthermore it can be seen directly from the work
hardening curves whether the recrystallization occurred during de-
formation.
Metallographic and X-ray investigations conducted after high temper-
ature deformation showed that copper, silver and rapidly deformed
gold single crystals had become at least partially polycrystalline,
whereas aluminium and slowly deformed gold single crystals had re-
mained single crystals. This is illustrated in Fig. 9.11 by the sur-
faces of a copper , a silver and a gold crystal etched to reveal the
grain boundaries. Recrystallized regions are visible and it can also
be seen that recrystallization has occurred preferentially in certain
directions. These preferred directions arise because recrystalliz-
ation is favoured in kink bands. Recrystallization is frequently con-
fined to individual kink bands.
The question is whether recrystallization has occurred during or after
deformation. Figure 9.12 shows the work hardening curve of a copper
single crystal at a deformation temperature of 875 K. Initially the
curve has the normal shape until the stress makes a sudden steep drop.
Kecrystallization was always observed in specimens which were strain-
ed beyond the stress drop but never when the tensile test was broken
off shortly before the stress drop. The steep drop in the flow stress
and equally the further irregular course of the work hardening curve
are the result of recrystallization proceeding concurrently with de-
formation.
275
Fig. 9.11
276
0.9
Fig. 9.12
Work hardening curve of a Cu single crystal with con-
current recrystallization, after [8j , deformation
temperature 875 K.
2 7 7
This dynamic recrystallization follows different rules f r om the static
process. This can be seen f r om an experiment illustrated in Fig. 9.13.
It shows a section of a work hardening curve of a copper single crystal
which has undergone periodic deformation, that is the deformation
was interrupted by unloading for a given time (in this case five minutes)
after a given strain. During the unloaded period very slight recovery
occurred. On continuing with the deformation it can be seen that the
flow stress has a slightly lower valu than just before unloading. The
interesting thing in Fig. 9.13 is that after the last unloading the stress-
strain curve has the normal shape until suddenly the stress drops
steeply as a result of recrystallization. Clearly the recrystallization
is accelerated by the si multan eously occurring deformation because
there was considerably more time (5 min) avaable during the last un-
loaded period than was necessary for the recrystallization which
commenced immediately deformation was resumed. This recrystalliz-
ation covered macroscopic regions of the crystal in less than 1 sec
as can be estimated f r om the time necessary for the fall in stress.
Furthermore, the stress at which dynamic recrystallization starts is
sharply defined. In a series of experiments in which the tensile test
was broken off at stresses only a f ew percent below the recrystalliz-
ation point, no sign of recrystallization could be found by metallo-
graphic methods. This also follows fr om Fig. 9.13. The starting
stress for recrystallization is scarcely higher than that applied imme-
diately preceding the last unloading.
Dynamic recrystallization is always observed in Ag and Cu above
Tm/2 but never in Al. The reason is obvious. Critical conditons
must be f ulf illed for the onset of recrystallization and these are never
reached in AI because they are suppressed by the intensive recovery.
Au occupies an interesting intermediate position. Recrystallization
-3 -1
occurs at high strain rates >2. 10 sec but not at low ones.
278
0,8
0,7
Cu . lO^sec
975 'K
0. 5 Q5 0,55 0,6
Fig. 9. 13
Section of the work hardening curve of a step-wise de-
f ormed Cu single crystal. Af ter each def ormation step
of approx. 2% the specimen was unloaded f or about
5 min.
2 7 9
The processes by which an intensively active recovery suppresses re-
crystallization are not clear. It is possible that recovery hinders the
buld-up of f low stress necessary f or recrystallization, i. e. the
necessary stored energy (or driving f orc). It is however also con-
ceivable that dislocation structures favourable for recrystallization
cannot f orm during intensive recovery. For example, recovery
suppresses the f ormation of def ormation bands (see Figs. 9. 6 and 9. 7)
in which recrystallization can proceed particularly easily (see Fig. 9.11).
The extent to which the results obtained f rom single crystals can be
used to explain the behaviour of polycrystalline material will now be
considered. The relationship between f low stress, temperature and
strain rate valid in the hot working temperature range has been measur-
ed for a wide variety of polycrystalline materials and the results dls-
cussedinasurveybyJONASetal [46]. As shown by these authors,
an analysis of the experimental results is possible by means of the
empirical equation 6 already used f or recovery. Subsequent to this
analysis the metis can be divided into two groups on the basis of the
derived ctivation energy Q. In the f irst group Q is approximately
equalto the self -dif f usion energy Q^. In the second group vales
of Q are obtained considerably larger than Q^. The above authors
conclude that Q has a smaller valu when only dynamic recovery is
operative and a large valu when dynamic recrystallization occurs.
There is as yet no direct proof for this conclusin.
Since it is clear f rom the single crystal experiments discussed whether
recovery alone or both recovery and dynamic recrystallization determines
the work hardening curve at high strains, eq. 6 was also applied to single
crystal measurements [8, 49 ] . The technique will be explained brief -
281
2 8 0
For the application to single crystal experiments it is expedient to re-
place f by ^and o'by T * , to give
#= B [si sinh exp (-QAT)
(7)
in which the nuraerical vales o B and cj differ from those in eq. 6.
Rearranging to give Q we have
Q = kT
m
di n [sinh (U'f)]T\m/T
[sinh fcu.
(8)
With this equation, a valu of the activation energy Q can be determined
which can be used to describe the temperature and rate dependence of
the starting stress for recrystallization rR (mximum stress in Fig.
9. 12). The valu foro; in eqs. 7 and 8 is obtained from the rate
dependence of the flow stress [40]. For silver, copper and gold,
vales were found between 03 4 < w < O, 6 (mm /kg).
Knowing tu it is possible to plot sinh (w-T^) against In * at constant
temperature and against T m/T at constant # . Figure 9. 14 shows
In sinh (wO as a function of T /T for the recrystallizing metis
m
Ag, Cu and Au. In the temperature range investigated the relationship
is linear for all metis."1"' Figure 9. 15 shows sinh (oj-rR) as a
function of In #for silver, copper and gold with the temperature as
parameter. In the case of both silver and copper a host of paraUel
straight lines is obtained. At small g' the dependence in the case of
gold can also be described by a number of parallel straight lines
but at high the straight lines bend sharply. T he kink lies at the
strain rate above which gold recrystallizes.
T R for gold is not defined at low strain rates where no recrystalliz-
ation occurs. Instead of T the stress was taken at the highest
.
point of the work hardening curve, i. e. at the start of necking.
Fig. 9.14
Starting stress for recrystallization TL as function of
temperature for Ag, Cu and Au single crystals.after
[8] , plotted according to eq. 8.
2 2
3 >
Fig. 9.15
Starting stress for recrystallization TR for Ag, Cu and
Au single crystals as function of strain rate $ at
various temperatures, after [8], plotted according to
eq. 8; (see footnote p. 280 ).
2 83
The slope of the lines in Figs. 9.14 and 9. 15 gives the differential
quotients in eq. 8 and henee Q. For copper Q = 2, 6 eV
(QgD = 2, 1 eV [50]), for silver Q = 4, 8 eV (QgD = 1, 9 eV[5 / ]) and
or gold there are two vales corresponding to the steep branch in
Fig. 9jl 5(i . e. j f > 4 . 1 0 ~ 3 sec"1)for which Q >6, 5 eV and to the
fat branch ( % < 4. 10"3 sec"1) for which Q = 1, 5 eV (Q = 1, 9 eV5.?])
The resulta from gold, which recrystallizes only at high strain rates
-3 -1
( o >4. 10 sec ), are particularly informative. Activation energies
considerably greater than Qor, are obtained in this rate range whereas
su
at low strain rates at which no recrystallization occurs the activation
energies are below Q^. The Q-values for Ag and Cu, which re-
OLJ
crystallize at all rates investigated lie well above Q-^..
These single crystal results are important for two reasons. They con-
firm the assertion made byJONASetal- [46] for the deformation of
polycrystalline material that using eq. 6 high vales of Q are always
observed if recrystallization occurs. Furthermore the results show
that recrystallization in single crystals obeys quantitatively similar
rules to that in polycrystalline material.
The high vales obtained for the activation energy Q have only formal
significante, since no theoretical model exists as yet to explain the
relationship between the starting stress for recrystallization, strain
rate and temperature. The rules governing dynamic recrystallization
have been investigated scarcely at all experimentally while theoretical
studies are virtually non-existent.
I
2 8 4
6. CONCLUDING REMARKS
The recovery processes controlling the individual branches of the
curves in Fig. 9.1 can now be explained. Below T /2 cross slip is
the decisive mechanism in all four metis: above T /2 it is either
m
diffusion controlled climb or dynamic recrystallization. In aluminium
the course of the curve is determined solely by the rules governing
dislocation climb, whereas for Ag and Cu, in which the dislocations
cannot climb so well owing to the low stacking fault energy, dynamic
recrystallization becomes important. Gold occupies an intermedate
position.
Corresponding to the mechanisms operative in the different temperatura
ranges we speak of the temperature tanges of cold working and hot
working.
The cold working range includes the deformation temperatures at which
cross slip controls dynamic recovery, It is characteristic that only
at high stresses do the softening processes proceed sufficiently rapidly
to match work hardening (high work hardening). Only very little work
has been published on the stress-strain behaviour in this temperatura
range. The empirical knowledge and theoretical understanding of the
quantitative relationships are henee very meagre.
, he hot working range starts at the temperatures where diffusion con-
trolled climb or recrystallization become important. It is typical of
di ffusi on controlled climb that the flow stress reaches a relatively low
saturation valu after only a small strain. Good formability is associat-
ed with behaviour such as this. According to the work hardening curve
for Al shown in Fig. 9. 8 , for example, tensile extensin is possible
under constant stress without necking of the specimen, that is, Al
28 5
possesses properties similar to superplasticity. For these reasons
technical shaping processes associated with high degrees of deformation
are undertaken preferentially in this temperature range. The rules
governing dynamic recovery by dislocation climb processes have been
studied in detall and are widely known. Here also, however, some
important details remain unexplained. It s for example not understood
how the stacking fault energy is quantitatively involved in the recovery
rate of f. c. c. metis.
Dynamic recrystallization gives rise to simarly good deformation
properties. Only in recent years, however, have intensive studies of
dynamic recrystallization been made with the result that only general
experimental data are available while theoretical formulations are
almost completely lacking.
2 8 7
286
REFERENCES
1. G. I- Taylor:
2. N. E. Mott:
3. A. Seeger in:
4. Z. S. Basinksi:
5. P.B. Kirsch and
D.H. Warrington:
6. D. Kuhlmann Wilsdorf:
7. U. F. Kocks:
8. F.Kirch:
9. R. D. Heidenreich and
W. Shockley in:
10. N. Thompson:
11. E. Orowan:
12. H.Mecking and
K. Lcke:
13. J.Ishida and
D. Me Lean:
14. R.Gasca- Neri,
C. N. Ahlquist and
W. D. Nix:
15. S. Mader and
A.Seeger:
16. S. Mader in:
17. P.Haasen:
18. G. Schoeck and
A. Seeger in:
19. H.Wolf:
20. H. Mecking and
K. Lcke in:
21. O.D. Sherby,
J. L. Lytton and
J. E.Dorn:
Proc.Roy.Soc. A 145 (1934)362.
Trans. Met. Soc. AIME 218 (1960) 962.
"Handbuch der Physik", S. FNgEJ,Ed.
Springer-Verlag, Berln VII 2, (1958) p. 1.
Phil. Mag. 4 (1959)393.
PhlLMag. 6 (1961)735.
Trans. Met. Soc. AIME M_ 4 (1962) 1047.
PhiLMag. 13. (1966)541.
Thesis TH Aachen 1970.
Bristol Conf. Phys. Soc. London 1948.
Proc.Phys.Soc. 66 (1953)481.
j.Scot.lron Steel Inst. 54 (1946-47)45.
Acta Met. 17 (1969) 279.
J.lron Steel Inst. 205 (1967)88.
ActaMet.18 (1970)655.
Acta Met. 8 (1960) 513.
"Electron Microscopy and Strength of
Crystals", G.Thomas and J. Washburn,
Eds., Intersc. Publ. ,New York 1963, p. 18
Phil. Mag. 3 (1958) 384.
"Rep. Conf. Defects in Solids", Phys. Soc.
London, p. 340.
Z.Naturforschung 15a_ (1960)180.
"Sec. Intern. Conf. on The Strength of
Metals and Alloys", Vol.n p.470.
Acta Met. 5 (1957) 219.
22. J. L. Lytton,
L. A. Sheppard and
J. E. Dorn:
23. P. G. Me Cornick and
A. L. Ruoff in:
24. J. Hansen:
25. J. E. Dorn and
J. Mote:
26. C.R. Barrett:
27. A.K. Mukherje,
J. E. Bird and
J. E. Dorn in:
28. H. Mecking and
F.Kirch:
29. H.P. Stwe:
30. J.Weertman:
31. R.W. Christy:
32. R. Chang:
33. B.I lschner:
34. F. R. N. Nabarro:
35. N.F. Mott:
36. P. B. Kirsch and
D. H. Warrington:
37. C.R. Barrett and
W. D. Nix:
38. W. Blum and
B. Reppich:
39. F.Garafalo:
40. W. A.Wong and
J. J. Joas:
41. D. Hardwick,
C. M. Sellara and
W.J. Me G.Tegart:
Trans. AIME 212 (1958) 220.
"The Interaction between Dislocations
and Point Defects", B. L. Eyre, Ed.,
AERE- R 5944, 1968, p. 518.
Diplom Thesis, TH Aachen 1969-
"High- Temperature Structures and Materials",
Pergamon Press, New York 1963, p. 95.
Acta Met. 13 (1965) 1088.
"The Interaction between Dislocations
and Point Defects", B. L. Eyre, Ed.,
AERE- R 5944, 1968, p.422.
to be published.
Acta Met. 13_ (1965)1337.
J. Appl. Phys. 28_ (1957) 362.
J. Appl. Phys. 3 (1959) 760.
"The Physics and Chemistry of Ceramics",
Cordn and Breach, New York 1963, p. 275.
Z. Phys. 190 (1966) 258.
Phil. Mag. 16_ (1967) 231.
"Creep and Fracture of Metals at Elevated
Temperatures", London: H. M. Stationery
Office 1956, p.21.
Phil. Mag. 6 (1961) 735.
Acta Met. 13 (1965) 1247.
Acta Met. 17_ (1969) 959.
Trans. Met. Soc., AIME 227 (1963) 351.
Trans. Met. Soc. AIME 242 (1968) 2271.
J. Inst. Met._ 90_ (1961-62)21.
288
42. C. Rosart and
P. Blain:
43. C. M. Sellars and
W.J. Me G.Tegart:
44. G. J. Richardson,
C. M. Sellars and
W.J. Me G.Tegart:
45. M. J. Luton and
C. M. Sellars:
46. J. J. Joas,
C. M. Sellars and
W.J. Me G.Tegart:
47. H. P. Stwe in:
48. H. J. Me. Queen:
49. H. Mecking and
F. Kirch:
50. C. T. Thomizuka et al.
51. C. T. Thomizuka and
E.Sonder:
52. H. C. Gatos and
A. D. Kurtz:
Rev. Met. 55_ (1958) 573.
Mem. Sel. Rev. Met. 63_ (1966) 731.
Acta Met. 14 (1966)1225.
Acta Met. 17 (1969) 1033.
Met. Rev. (1969) 130.
"Deformation under Hot working Condions
Sp. Rep. Nr. 108, Iron and Steel Inst.
London 1968.
J.Metals^O_ (1968)31.
to be published.
Phys.Rev. 98(1955)1870.
Phys.Rev.J03 (1956) 1182.
J. Metals 6 (1954)616.
289
CHAPTER 10
THE IMPORTANCE AND APPLICATION
OF RECRYSTALLIZATION IN THE
PROCESSING OF METALS AND THE MANUFACTURE
OF SPECIAL PRODUCTS
BY
I. L. DILLAMORE
1. INTRODUCTION
Primary recrystallization and grain growth processes are of im-
portance in determining the grain structure in a material which has
been fabricated by mechanical working, and in discussing the im-
portance of recrystallization the intention is really to consider the
control of grain structures*nd thereby the properties of the material.
It is not possible to define a grain structure in simple terms; para-
meters of interest in influencing properties are: grain size, grain
size distribution, grain shape, grain orientation distribution - or
texture, and the degree of recrystallization. The phase composition
and distribution will not be considered as a primary variable here,
only materials which are essentially single phase will be dealt with,
although a small volume fraction of second phase is allowed within
the present terms of reference if its prime function is to influence
the grain structure of the majority phase.
2. THE INFLUENCE OF GRAIN STRUCTURE ON PROPERTIES
In this section the relationship between grain structure parameters
and the important physical and mechanical properties will be briefly
290
reviewed. More complete coverage of some of these relationships can
be found in the conference volume "Textures in Research and Practice"
[1 ] and in various review articles [2, 3] . Here the intention is
not to provide an exhaustive survey but roerely to highlight the com-
mercial importance of the control of grain structures.
2.1. Grain Size Effects
2.1.1. Mechanical Properties: The grain size dependence of yield stress
has been widely recognised since it was first demonstrated by
SYLVESTROWICZ and HALL [4] that the lower yield stress of iron
obeyed a relationship of the f orm
l.y.p.
= < f . + k D
i y
-V2
CD
where o is termed a friction stress, D is the grain diameter and
k is the constant of proportionality. As shown by WILSON and
RUSSELL [5] k is greatest for the annealed condition, and is an
order of magnitude lower in material which had initially been strained
through the yield point and subsequently aged (i. e. strain-aged). This
is illustrated by results tak en from the work of WILSON and RUSSELL
and shown in Fig. 10.1, where it may also be seen that the grain size
dependence of the flow stress, that is the stress at some fixed strain
beyond the Lders strain, is lower than that of the lower yield stress.
It is usual to express the grain size dependence of the flow stress by
k f - D
(2)
although recent work due to MORRISON [6] has indicated that, while
eq. 1 is well obeyed for the range of grain sizes from 1 yum to 300yu
eq. 2is a rather poor approximation for the full range of grain sizes.
291
200
IELD
T R ES S
M N / m 2
100
0 8 10
Fig. 10.1
Grain size dependence of the lower yield stress of low carbn
mild steel (1) annealed sample, (2) as (1) and subsequently
strained 5%, (3) as (2) and aged 20mn. 60 C, (4) as (2) and
aged 10, 000min. 60C (1 MN/m2 = 107dynes/cm2) (after
WILSON and RUSSELL [ 5 ]).
p
r
o
p
e
r
t
i
e
s

u
n
d
e
r

d
y
n
a
m
i
c
,

o
r

c
r
e
e
p
,

c
o
n
d
i
t
i
o
n
s

(
c
.
f
.

c
h
a
p
t
e
r

9
)
,

W
h
e
n
g
r
a
i
n

b
o
u
n
d
a
r
y

s
l
i
d
i
n
g

m
a
k
e
s

a
n

i
m
p
o
r
t
a
n
t

c
o
n
t
r
i
b
u
t
i
o
n

t
o

t
h
e

s
t
r
a
i
n
r
a
t
e

a

f
i
n
e

g
r
a
i
n
e
d

m
a
t
e
r
i
a
l

g
e
n
e
r
a
l
l
y

s
h
o
w
s

a

g
r
e
a
t
e
r

c
r
e
e
p

r
a
t
e

t
h
a
n
a

c
o
a
r
s
e

g
r
a
i
n
e
d

m
a
t
e
r
i
a
l

u
n
d
e
r

t
h
e

s
a
m
e

s
t
r
e
s
s
,

a
l
t
h
o
u
g
h

r
e
s
u
l
t
s

d
u
e
t
o

G
A
R
O
F
A
L
O

e
t

a
l

[
l

]

s
h
o
w

a

m
i
n
i
m
u
m

i
n

t
h
e

c
r
e
e
p

r
a
t
e

o
f

a
n
a
u
s
t
e
n
i
t
i
c

i
r
o
n

a
l
l
o
y

a
t

a

m
o
d
e
r
a
t
e
l
y

c
o
a
r
s
e

g
r
a
i
n

s
i
z
e

(
F
i
g
.

1
0
.
4
)
.

A
y
s
.
M
N
/
m
2
y
p
e

p
e
r
c
e
n
t
H
I
B
B
A
R
D

[
1
0
]
.
G
r
a
i
n

s
i
z
e

h
a
s

m
o
r
e

i
m
p
o
r
t
a
n
t

e
f
f
e
c
t
s

o
n

h
i
g
h

t
e
m
p
e
r
a
t
u
r
e

m
e
c
h
a
n
i
c
a
l
O

4
0
2
0
0 2 1 n
w
i
l
l

d
i
m
i
n
i
s
h

w
i
t
h

i
n
c
r
e
a
s
i
n
g

t
e
m
p
e
r
a
t
u
r
e
.

S
u
c
h

a
n

e
f
f
e
c
t

i
s

d
e
m
o
n
s
t
r
a
t
-
e
d

b
y

t
h
e

r
e
s
u
l
t
s

o
f

F
i
g
.

1
0
.

3

t
a
k
e
n

f
r
o
m

t
h
e

w
o
r
k

o
f

C
A
R
R
E
K
E
R

a
n
d
-

O A

a

t
h
a
n

c
u
b
i
c

m
e
t

i
s

[
7
]

.

C
l
i
m
b

a
n
d

c
r
o
s
s
-
s
l
i
p

e
f
f
e
c
t
i
v
e
l
y

i
n
c
r
e
a
s
e

t
h
e
n
u
m
b
e
r

o
f

a
v
a
i
l
a
b
l
e

d
e
f
o
r
m
a
t
i
o
n

m
o
d
e
s

a
n
d

s
i
n
c
e

t
h
e
s
e

p
r
o
c
e
s
s
e
s

b
e
-
c
o
m
e

i
n
c
r
e
a
s
i
n
g
l
y

i
m
p
o
r
t
a
n
t

a
s

t
h
e

t
e
m
p
e
r
a
t
u
r
e

o
f

d
e
f
o
r
m
a
t
i
o
n

i
s

r
a
i
s
-
e
d

i
t

i
s

t
o

b
e

e
x
p
e
c
t
e
d

t
h
e

g
r
a
i
n

s
i
z
e

d
e
p
e
n
d
e
n
c
e

o
f

t
h
e

s
t
a
t
i
c

f
l
o
w

s
t
r
e
s
s
2
3

j
J
m
7
3

j
j
m
I
4
2
j
j
m
J
&

U

I
>
'
d
e
p
e
n
d
e
n
c
e

o
f

f
l
o
w

s
t
r
e
s
s

i
s

r
e
l
a
t
e
d

t
o

t
h
e

n
u
m
b
e
r

o
f

a
v
a
i
l
a
b
l
e

d
e
f
o
r
m
-
a
t
i
o
n

m
o
d
e
s
,

t
h
e

s
m
a
l
l
e
r

t
h
e

n
u
m
b
e
r

t
h
e

g
r
e
a
t
e
r

t
h
e

g
r
a
i
n

s
i
z
e

d
e
p
e
n
d
-
e
n
c
e
.

T
h
u
s

h
e
x
a
g
o
n
a
l

m
e
t

i
s

s
h
o
w

a

g
r
e
a
t
e
r

g
r
a
i
n

s
i
z
e

d
e
p
e
n
d
e
n
c
e

[
9
]
-
J
&
^
*

.
p
^
o
-
0
B
g
-

0
-
^
.
Q
A
s

d
i
s
c
u
s
s
e
d

b
y

A
R
M
S
T
R
O
N
G

e
t

a
l

[
7
]

t
h
e

m
a
g
n
i
t
u
d
e

o
f

t
h
e

g
r
a
i
n

s
i
z
e

1
*
*
& .
O
-
O
-
a

w
i
t
h

i
n
c
r
e
a
s
i
n
g

s
t
r
a
i
n
.

T
h
e
r
e

i
s
,

h
o
w
e
v
e
r
,

n
o

c
l
e
a
r

e
v
i
d
e
n
c
e

o
f

a
n
y

j
c
u
b
i
c

m
e
t
a
l

s
h
o
w
i
n
g

a

s
t
r
o
n
g

g
r
a
i
n

s
i
z
e

d
e
p
e
n
d
e
n
c
e
o
f

t
h
e

s
t
r
a
i
n

t
o

i
n
s
t
a
b
i
l
i
t
y
.
j
-
<
<
^
>
"
*
*
*
*
"
-
<
r
-
*
f
t
a
t
u
r
e

i
s

n
o
t

s

m
p
l
y

d
e
t
e
r
m
i
n
e
d
.

I
n

f
.

c
.

c
.

m
e
t

i
s

t
h
e

g
r
a
i
n

s
i
z
e

d
e
p
e
n
d
-
e
n
c
e

o
f

f
l
o
w

s
t
r
e
s
s

i
s

p
r
o
g
r
e
s
s
i
v
e
l
y

r
e
d
u
c
e
d

o
n

s
t
r
a
i
n
i
n
g

w
h
i
l
e

i
n

b
.
c
.
c
.
m
e
t

i
s

t
h
e

g
r
a
i
n

s
i
z
e

d
e
p
e
n
d
e
n
c
e

f
i
r
s
t

i
n
c
r
e
a
s
e
s

[
8
]

a
n
d

t
h
e
n

d
e
c
r
e
a
s
e
s
O

Q
A
^

-
A
o
-
^
t
v
*
*
r

>
*
r
~
-
^
^
^
^
^
^
^
F
W
o
r
k

h
a
r
d
e
n
i
n
g

p
r
o
c
e
s
s
e
s
,

b
y

i
n
f
l
u
e
n
c
i
n
g

t
h
e

s
h
a
p
e

o
f

t
h
e

s
t
r
e
s
s
-
s
t
r
a
i
n
c
u
r
v
e

a
r
e

p
r
i
n
c
i
p
a
l
l
y

o
f

i
m
p
o
r
t
a
n
c
e

i
n

d
e
t
e
r
m
i
n
i
n
g

t
h
e

s
t
r
a
i
n

t
o

p
l

s
t
i
c
i
n
s
t
a
b
i
l
i
t
y
.

T
h
e

g
r
a
i
n

s
i
z
e

d
e
p
e
n
d
e
n
c
e

o
f

w
o
r
k
-
h
a
r
d
e
n
i
n
g

a
t

r
o
o
m

t
e
m
p
e
r
-
^

N
^
1

^
r

*
*
T
T
-
Q
^
-
>
^
^

r
-
a -
i
p
_
3 i
n
e
r
a
i r
e
B
-
r
e
i r
t
-
r
e
M
i
r
e
i r
l
-
=
r
(
D
T
O
"
1
P
_
5
'
e
n
N r
e
i
s

r
e
l
a
t
e
d

t
o

t
h
e

g
r
a
i
n

s
i
z
e

d
e
p
e
n
d
e
n
c
e
s
o
f

y
i
e
l
d

a
n
d

s
u
b
s
e
q
u
e
n
t

f
l
o
w
s
t
r
e
s
s
e
s

a
n
d

i
s

i
l
l
u
s
t
r
a
t
e
d

i
n

F
i
g
.

1
0
.
2
.

I
t

m
a
y

b
e

s
e
e
n

t
h
a
t

t
h
e

L

d
e
r
s
2
9
2
T
h
e

g
r
a
i
n

s
i
z
e

d
e
p
e
n
d
e
n
c
e

o
f

t
h
e

y
i
e
l
d

p
o
i
n
t

e
l
o
n
g
a
t
i
o
n

o
r

L

d
e
r
s

s
t
r
a
i
n
l
o
t
o
c
o
1
0

1
0
2

1
0
3

1
0
4
a
g
e
i
n
g

t
i
m
e

a
t

6
0

C
.
,

m
i
n
u
t
e
s
F
i
g
.

1
0
.
2
Y
i
e
l
d

s
t
r
e
s
s

i
n
c
r
e
m
e
n
t

(

A

y
.

s
.
)

a
n
d

y
i
e
l
d

p
o
i
n
t

e
l
o
n
g
a
t
i
o
n
(
y
p
e
)

a
s

a

f
u
n
c
t
i
o
n

o
f

a
g
e
i
n
g

t
i
m
e

a
t

6
0

C

f
o
r

t
h
r
e
e

g
r
a
i
n

s
i
z
e
s
i
n

l
o
w

c
a
r
b

n

m
i
l
d

s
t
e
e
l
,

t
h
e

s
a
m
e

m
a
t
e
r
i
a
l

a
s

i
n

F
i
g
.

1
0
.
1
.
2 9 4
Q 3
1t
SI
HH
N t
I
"p*2
is
4
o 8
ib. r*
Ss1
133 -i
>0
BS
LO
3 en
ro en
en
O
O
o
B?
w ^ O
W SI *^
ff *ta^
ro
295
However, it is generally truc that creep strength increases with grain
size. The creep ductility, on the other hand, decreases with increasing
grain size [12] (Fig. 1 0.5), and this also relates to grain boundary
sliding since creep cavitatlon is nucleated by triple point cracking at
the end of sliding boundaries [13] . The stress concentration at the
triple point increases with increasing free length of aliding boundary
so that cavitation occurs more readily in coarse grained materials [14].
Returning to low temperature properties, perhaps the most iraportant
effect of grain size from a technological viewpoint is on notch ductility,
or toughness, in steels.
In Fig. 1 0.6 both yield strength and impact transition temperature
(corresponding to an energy absorption of 55 Joules in a standard
-1/2
Charpy specimen) are plotted against D for three families of
steels all having a carbn content of - 0,1%. These resutts[/5]show that
all methods of increasing strength decrease toughness except refining
the grain size. Compared with the carbn manganese steels the as-
rolled C-Mn-Nb or C-Mn-V steels are strengthened by precipitation
hardening, while the quenched and tempered C-Mn-Nb or C-Mn-V
steels are further strengthened by dislocation hardening and additional
precipitation. The grain refinement achieved by control of the process
cycle allows the additionalstrengtheningto be achieved with no loss in
toughness.
1.1.2. Physical Properties: Most physical properties are, when compared to
mechanical properties, relatively insensitive to grain size. Electrical
conductivit y is affected [16] since grain boundaries scatter electrons,
thus reducing the mean free path, and, although the effect of grain size
is quite small it is worthwhile giving attention to this effect n power
transmission, so as to minimize power losses.
[
l
l
]
s
e
S
D
V
O
O

'
O
I
V
.

O
H
V
O
U
B

U
T

9
$
B
J

d
a
a
j
o

J
B
p
u
o
o
a
s
-

^
"
0
1
N
l
V
d
S
0
2
-
0

0
1
-
0

0
-
0

K
)
-
0

2
0
-
0
T
1
0
-
0 -

0
1
4
0
3
0
R

o
f

A
2
0
1
0
O
0
-
1

0
-
2

0
-
3

0
-
4
G
r
a
i
n

d
i
a
m
e
t
e
r

m
m
0
-
5
t
o
C
D
-
I
F
i
g
.

1
0
.
5
C
r
e
e
p

d
u
c
t
i
l
i
t
y
,

a
t

4
5
0

C

a
n
d

a

s
t
r
a
i
n

r
a
t
e

o
f

1
8
%
/
h
r
,

e
x
p
r
e
s
s
e
d
a
s

p
e
r
c
e
n
t

r
e
d
u
c
t
i
o
n

o
f

r
e
a
,

f
o
r

7
0
/
3
0

b
r
a
s
a

a
s

a

f
u
n
c
t
i
o
n

o
f

g
r
a
i
n
s
i
z
e

(
a
f
t
e
r

T
A
P
L
I
N

a
n
d

W
H
T
T
T
A
K
E
R

[
1
2
]

)
.
298
750
L Y P
MN/m2
500
250
; - Mn
Tc, C
55 Joules
- 80
- 160
o L C- Mn -
5
10
O'*
15
-V2
mm c
299
It is widely held [17] that the power losses in silicon-iron transformer
core material are influenced by grain size, and it is usually specified
that the material for this purpose should be coarse grained. However,
as will be apparent later, it is often difficult to control grain size and
grain orientation independently in thin sheet material and the apparent
effect of grain size on power losses may be a texture effect. The re-
sults of SWIFT et al [18] on iron sheet show no effect of grain size
(Fig. 10. 7): in the material used in this investigation grain size was
controlled by using the Of-J*phase transformation and by strain anneal-
ing thus avoiding textural effects. The general acceptance of a grain
size effect may also arise in this case from the fact that the desirable
texture (110) [001] is only obtained to the required perfection when
extensive grain growth has occurred.
I W E R
,oss
k g
Fig. 10. 6 - Semi schematic representaron of the yield strength
and 55 J oules transition temperaturo ranges indicated as a funct ion
the D " /2 (where D is the grain diameter). The steels, all of about
Oj 1% C are as follows: C-Mn-Nb/V Q. T. carbn-manganese steel
with secondary hardeningadditionsof niobium or vanadium auenched
directly after hot-rolling and subsequently tempered at 600 C;
C-Mn-Nb/V as rolled - steels of similar compositions but somewhat
lower Nb or V additions to the above, rolled to a low finishing temper-
ature and air-cooled; C-Mn as-rolled steels treated similarly to the
C-Mn-Nb/V as rolled steels (55 Joules r 5, 5 kg. m) (after KEWELL [l5j > -
O
_B _r n a x 1- 2 W b /m2
11
ffl
B
1- 0
!^. 0_6_W _b/m2
O 200 400 600 800
GR AI N- SI ZE
Fig. 10. 7 - Power loss at 50 cycles as a function of grain size in
decarburised iron at three levis of mximum inductance (B ).
(Wb 5 1 Weber) (resulte of SWIFT et al [18] ). max
300
2. 2. Grain Orientation Effects
2.2.1. Mechanical Properties: The most important atomic deformation
mechanism, slip, obeys a critical resolved shear stress lawso that
the strength of a crystal depends on its orentation relative to the
stress axes. The familiar Schmid law[19] defines the orientation
dependence for uniaxial stressing and this may be generalized for
three dimensional stress states.
The range of strengths which can be obtained in cubic metis by mere-
ly altering their orientation, or controlling the texture, is not large.
The mximum range obtained by, for instance, deforming the orienta-
tions shown in Fig. 10. 8 in plae strain compression is from
< f = 2 T io o" = 2 ^6 T . In polycrystals of cubic metis the
range of strengths obtained is generally much smaller, the results
shown in Fig. 10. 9 a for the variation of lower yield stress with angle
to the rolling direction [20] in a rimming steel sheet are fairly typic-
al; they show a - 10% variation about the mean valu.
A tensile specimen of an isotropic material contrais equally in all
directions normal to the axis of extensin during straining. A textured
metal being generally anisotropic shows different contractile strains
in different directions normal to the tensile axis. In sheet metal the
ratio of the contractile strains in the width and thickness directions of
a tensile specimen ( / ,} is called the r-value. The variation of
r-value with test direction is shown in Fig. 10. 9 a for the same
material from which the yield stress variation was determined. The
r-value has assumed some importance in relation to sheet formability
for two reasons; the average of the r-values {r) obtained f rom tests
in different directions correlates extremely well with the deep-
drawability of the material [22] , while the mean amplitude of varia-
tion in r-value -^- correlates with the propensity of the material
for ear formation [22] . The relationship between i- and the limiting
3 O 1
t
0) O
c _
OJ }
iw
u
^
f\
v
co
o
I-H
S f
k
u
1
E
t.
H1
.
u
T-l
I
XI
&
O
r-t
c
1-,
OT
1
c
E
u
< *s

o
-2

'
J

o
H
Q
&
01
5
0
0
L
O
W
E
R
Y
I
E
L
D
S
T
R
E
S
S
M
N
/
m
2
4
5
0

-
4
0
0

O
(
a
)

V
a
r
i
a
t
i
o
n

o
f

l
o
w
e
r

y
i
e
l
d

s
t
r
e
s
s

a
n
d

r
-
v
a
l
u
e

w
i
t
h

a
n
g
l
e

o
f
t
e
n
s
i
l
c

t
e
s
t
l
n
R

r
e
l
a
t
i
v
e

t
o

t
h
e

r
o
l
l
i
n
g

d
i
r
e
c
t
i
o
n

f
o
r

a

a
a
m
p
l
e
t
*
t

t
l
n
-
o
t

i
r
t

u
l
,

I
m
v
t
i
m

t
l
i

a
O
O

j
>
o
l

I
l
t
f
u
r

j
j
h
o
w
n
.
j
n

(
b
)

(
r
e
^
u
l
t
i
O U
T

3
304
drawing ration {L. D. R. E the ration o diameter o the largest blank
whi ch successfully completes the drawing operation to that of the punch)
is shown in Fig- 10.10 for a series of steels [23] and the correlata
of i y - wi t h 4h /h (the mean variation in cup height divided by the
mean cup height) is shown in Fig. 10.11 [22] .
The strain to fai lure in the tensile test is controlled in ductile metis
by the onset o plstic instability and this in t urn is controlled by the
work hardening behaviour of the material. The results shown in
Fig. 10.12 demnstrate that a preferred orientation can give wide
variations in work hardening and in the strain to f ai lur e [24] . The
case shown, which relates to cube-textured f. c. c. metis is, however,
rather extreme, multi-component textures generally show muc h smaller
dlfferenees between different test directions.
When f ai lur e is brittle in character the f ai lur e stress will generally
be texture sensitive, as exemplified by the result s of PRIESTNER and
LOUAT [25] and of WEBSTER [26] on grai n oriented silicon-iron. For
f ai lur e after general yield as well as at the yield point the var i at i on in
f r ac t ur e stress with orientation follows that of the yield stress.
The orientation dependence of flow stress diminishes with increasing
t emper at ur e for the same reason that grai n size dependence diminishes,
i. i > . c li mb and cross-slip effectively increase the number of available
d - t n r ma t i o n mechanisms. Following f r o n i t hi s t he orientation depend-
enre is generally slrongcr the f ewer the available deformat i on modes,
and in hexagonal met i s t ext ur e har deni ng may be seriously considered
as a viable means of strtmgthening the material [27] . Use of this
sourc e of st r engt h r equi r es a r at her sophi st i c at ed apprec i at i on of the
mat er i al properties relative to the service condltions of a c omponent
or st r uc t ur a l member and in consequence t here have been few applica-
tions of t ext ur e har deni ng.
S O 5
B.3-
2 - 4
L D. R.
2 - 3
O 1 - 5
f
2- 0
Fig. 10. 10
The dependence of deep drawability, as represented by the limit-
ing drawing ratio (L. D. R. E ratio of diameters of the largest blank
to successfully complete the drawing operation and the punch) , as a
func t i on of the average r-valu for the material (after ATKJNSON
and Mc LEAN [23] ).
306
3 0 7
.
V
V
\
V
\c
\N
\ \ \ \1

1 1
o
w -i
^^J ^J^^v
ui Ot
4
= -
\
\
o
IV
p
O
ir
^^^
p
^
m

LC *4 f
ci>
1
o o o o
OJ
II-

t*>
rt
cu
j=
a
3
O
V
g p
(-.
(
>
m
QJ
j:
+j
"8
a>
u
M
"t
a>
u
h
a>
O.
rt
n
rt
T)
W
e n
m
a>
t-i
l
S <
cu

bfi
'S
J^
h
rt
H
i*
rt
-H
g .

a
V
J3
-+J
a>
^
u
9
c
M
CU
fj
fc-l
^1
p
DD
O)
J
^
V
Li
0)
^:
&
m
*#
fi
M
+
O
O)
4=
+
O
J
n
*
o
O)
J2
+
o
II

u>
-

J
t - >
(8
O
t
V
1
"1
3
^
-H
"o
t - l
01
X
*->
o
*J
TJ
>
h
t i
V
< >H
U
h
C
O
*J
33
n
o
o.
o
O)
M
+
o
ti
u
t4
4)
4
4)
L-tl.
4F
"3
c
o
+J
u
c

(j
a
rt
a
< D
ti
o
"a
""S
f
j=
ci
t

.i
i_*j
P!
j
H
s
a
T3
d
2
J
E
t<

I
a
v
h
N
+
o
a
,
+
0
h
n
ib
o
o
> a
t
h
t *
i
T.S .
MN/ m2
180
160
/\\
60
40
20
% E l
O
30 60 90
&
E
't-
a
Fig. 10.12
Variation of nominal tensile strength and ductility with angle of
tensile test to the rolling direction in cube textured copper fafter
RICHARDS [24] ).
308
2 2 2 PhysicalProperties: Most physical properties are directional in
crystalline solids and are thus influenced by texture. Control of pre-
ferred orientation is, in principie, the only method of influem
the *>ung 's modulus of a material but there are not many example
o the conscious control of processing to develop the required
and associated elastic properties. A possible example has beer
suggested by CHIN and WONCIEWICZ [28] in which the (111
texture is developed to improve the modulua for phosphor bronze
spring material. It has been realized, however, that a good texture
in steel for sheet forming applications is accompanied by the best
modulas valu for the finished sheet structure [29, 30] and it is
generally true that the direction of mximum Young' s modulus .
incides with the direction of mximum uniaxial flow stress. The ex-
3 O 9
ceptions to this rule are the b. c. c. metis Mo, Nb, V and Cr which
1 C A T* - r")iof less than 1.
have a Zener coefficient (Az = - % U44 T^- C12
As may be seen from Fig. 10. 13 the two b.c. c. metis iron and
niobium, both havingtextures similar to that of Fig. 10. 9 b, a
the opposite variation of Young' s modulus with angle to the rol
direction [31] -
For iron the valu of AZ is 2,41 and for niobium it is 0,51 [32].
The most important example of the influence of texture on a physical
property is that on the magnetic susceptibility. The familiar results
of HONDA et al [33] showing the magnetization/field curves for the
field applied along <100> , <110> and <l l l > directions in iron
show that the ease of magnetization decreases in the given order
(Fig. 10. 14). In nickel the easy direction of magnetization is <
and in cobalt it is <0001> .
In silicon iron for transformer core applications the Goss texture
{110}<001> is deliberately sought in order to have the <001> easy
magnetization direction in the sheet plae and the process evolved
presents perhaps the most sophisticated application of recrystallizatio
in the development of properties.
23
20
YOUNG' S
MODULUS
!04MN/m2
1 5
10
1 1 _L II
O 30 60 90
DEGREES
Fig. 10.13
Variation of Young' s modulus with direction relative to the rolling
direction in iron and niobium sheet, both materials havlng a texture
similar to that shown in Fig. 10-9 b (after HODDINOTT and DAVI ES
[311),
310
OJ
E
o
O
.5 2
4 - " *
O)
c
O)
C
6 -
o
250
Fi el d- A/cm
500
Fig. 10. 14
Magnetizaron versus field curves with the field applied parallel
to<100> , <110> and<l l l > , showing the relative ease of
magnetization along these directions in iron (after HONDA et al
[33] ).
3 1 1
3. Grain Shape Effects
Elongated grain structures are fairly common in rolled, extruded
and drawn metis, as well as in as-cast materials with a columnar
structure, and it has often been supposed that the grain shape has some
inuence on mechanical properties. This possibility has been consider-
ed in the case of wrought aluminium [34] and for aluminium killed
steels [35] but there is no positive evidence that grain shape influences
the low temperatura mechanical properties. Creep rupture life is, how-
ever, affected by grain shape as shown by the results of Fig. 10.15.
These results, obtained by VERSNYDER and GU ARO [36] on columnar
grained allo-s of composition 75, 5% Ni, 21% Cr and 3, 5% Al show
that the rupture life when stressed along the columnar axis is as much
as an order of magnitude greater than when stressed perpendicular to
this direction. The nterpretation of this result, which will not be con-
sidered further since it does not relate to recrystallization, is that
cavities are nucleated and grow on grain boundaries transverse to the
stress direction, These transverse boundaries are larger in rea and
more numerous for tests perpendicular to the columnar axis.
Orientation Distribution Effects
In a textured metal the orientation population is indicated by a pole
figure but this gives no information about the spatial distribution of
the orientations present and it may arise that either grains of similar
or dissimilar orientation occur together with greater probability than
indicated by their representation in the material.
WILSON and CHAPMAN [9] recognized the possible contribution of
this effect on the grain size dependence of the flow stress. They inter-
preted some results obtained from round bar and fat bar extrusions
of magnesium in which the grain size dependence of flow stress parallel
c
u
*
>
f J
=
a r
t
3

1

3
t
h
e

e
x
t
r
u
s
i

n

d
i
r
e
c
t
i
o
n

w
a
s

g
r
e
a
t
e
r

f
o
r

t
h
e

r
o
u
n
d

b
a
r
0
-4
-*
s
i
m
i
l
a
r

i
n

t
h
e

r
e

a
x
i
s

d
i
s
t
r
i
b
u
t
i
o
n

a
l
o
n
g

t
h
e

e
x
t
r
u
s
i

n

d
i
r
e
c
t
i
o
n

w
a
s
A
J
f+
*
Q
J
U
i I
o

m
a
t
e
r
i
a
l
s
,

a
s

b
e
i
n
g

d
u
e

t
o

a

g
r
e
a
t
e
r

p
r
e
v
a
l
e
n
c
e

o
f

1
1
7
3 S c
o

t
o
u
n
d
a
r
i
e
s

i
n

t
h
e

f

a
t

e
x
t
r
u
s
i
o
n
s
.

T
h
e

i
m
p
l
i
c
a
t
i
o
n

h
e
r
e
0
J
2
C
U
"
& r
t
J
=

M
1
|
-
1
-
'
g
l
e

b
o
u
n
d
a
r
i
e
s

a
r
e

l
e
s
s

e
f
f
e
c
t
i
v
e

d
i
s
l
o
c
a
t
i
o
n

b
a
r
r
i
e
r
a

.
t
h
e

m
e
a
s
u
r
e
d
c
u
n
d
a
r
i
e
s

s
o

t
h
a
t

t
h
e

e
f
f
e
c
t
i
v
e

g
r
a
i
n

s
i
z
e

i
s

l
a
r
g
e
r

t
h
a
0
f
(_
,
a
i
n

s
i
z
e

i
f

g
r
a
i
n
s

o
f

s
i
m
i
l
a
r

o
r
i
e
n
t
a
t
i
o
n

o
c
c
u
r

t
o
g
e
t
h
e
L
.
b
e
b C
U
1
h C
O
,
e

i
n
t
e
r
p
r
e
t
a
t
i
o
n

o
f

W
I
L
S
O
N

a
n
d

C
H
A
P
M
A
N
'

s

r
e
s
u
l
t
s
j
=
H
e
f
f
e
c
t
s

w
h
i
c
h
e
c
u
l
a
t
i
v
e

b
u
t

t
h
e
r
e

a
r
e

o
t
h
e
r

e
x
a
m
p
l
e
s

o
f

c
o
r
r
e
l
a
t
i
o
n
c
,
0
1
c
o
m
m
e
r
c
i
a
l
a
*
e

b
e
t
t
e
r

d

c
u
m
e
n
t
e
d
,

a
n
d

o
n
e

i
n

p
a
r
t
i
c
u
l
a
r

i
s

o
f

s
o
m
i
t
-
r
t
n
o
m
e
n
o
n

o
f
c
u
0
,
t
u
J
3
-*
-
0
-
M c
u
c
o
T
!
O

c
u

a
J
3
_
0
f t
<
U
a
c
u
a w c
o
i H
c
u
0 r
t

Q
J
h
t
?
c
-

D
-
'
c
o
a
p
i
n
g
"

o
r

"
r
i
d
g
i
n
g
"

i
n

s
o
m
e

f
e
r
r
i
t
i
c

s
t
a
i
n
l
e
s
s

s
t
e
e
l
s
(

h

e
x
h
i
b
i
t
s

t
h
e
^
s
& c
u
T

1
t s r
t
"
S m

a
T
-
l
*
+
-t
C
U
o r
i
>
4
H
t
-
m Q
J
J
3
+
j
c
o
i
a
0
3
C
D
T
H
0 T
-
l
Q
)
t
-
&
E
l
^c
o
O s
-
r
1
M
H
I
d
g
i
n
g
"

p
h
e
n
o
m
e
n
o
n
,

a
n
d

i
e

t
a
k
e
n

f
r
o
m

t
h
e

w
o
r
k

o
f

C
h
r
i
d
g
e
s

s
e
p
a
r
a
t
-
c
e

s
u
r
f
a
c
e

o
f

a

d
e
f
o
r
m
e
d

t
e
n
s
i
l
e

s
p
e
c
i
m
e
n

i
s

r
a
i
s
e
d

l
i
j
a
H
B

o
f

s
i
m
i
l
a
r
a
'r
t
t
-
.
&
C
0
-
M t
u

.r
-
l
C
U
-
Q c
o
r
t
i
c
i
J
& t
u
O
T
W
S
t-
J

c
o

t
-
-
t-
>
D
h
J
=

h
o
m
o
g
e
n
e
o
u
s
-
t>
>

e
n
t
a
t
i
o
n

o
c
c
u
r
r
i
n
g

i
n

s
t
r
i
n
g
e
r
s
,

a
n
d

d
e
f
o
r
m
i
n
g

f
a
i
r
l
;
g 0
j
a
t
t
e
r
n
s

b
e
-
w
i
t
h
i
n

t
h
e

c
o
l
o
n
i
e
s

b
u
t

g
i
v
i
n
g

r
i
s
e

t
o

d
i
f
f
e
r
e
n
t

s
t
r
a
i
n
>

c
u
-
f
c
j
-
-
-.
!
f
l
Q
J
i
C
e
e
n

c
o
l
o
n
i
e
s
.

T
h
e

d
i
f
f
e
r
e
n
t

b
e
h
a
v
i
o
u
r

p
r
o
b
a
b
l
y

o
r
i
g
i
&
-f-*
|
n i
"
w
i
r
s
e

g
r
a
i
n
e
d

t
e
x
t
u
r
e
d

s
t
r
u
c
t
u
r
e

o
f

t
h
e

h
o
t
-
b
a
n
d

m
a
t
e
i
0 u
i
r
e
s
e
n
t

d
u
r
i
n
g
j
-

c
h
e
m
i
c
a
l

s
e
g
r
e
g
a
t
i
o
n

c
a
u
s
i
n
g

s
o
m
e
t
f
p
h
a
s
e

t
o

b
e
c
-p
-H
i
s

u
n
d
e
s
i
r
a
b
l
e
-*
-*
:

r
o
l
l
i
n
g

a
s

s
o
m
e

w
o
r
k
e
r
s

b
e
l
i
e
v
e
.

T
h
e

r
i
d
g
i
n
g

e
f
f
e
c
o
J
S
u 8 r
-
i
i
c
e

t
h
e

s
u
r
f
a
c
e

f
i
n
i
s
h

o
f

t
h
e

c
o
l
d

f
o
r
m
e
d

m
a
t
e
r
i
a
l

i
s

]
H
'c
o
0
>
n
't-
<
t
i
5 o
3 "
C <
u
o
s
s

v
a
r
i
a
t
i
o
n
s

i
n

t
e
x
t
u
r
e

f
r
o
m

p
o
i
n
t

t
o

p
o
i
n
t

i
n

a

m
a
t
f
.
0
s
u
r
f
a
c
e

f
r
i
c
t
i
o
n
:
h
e

m
a
t
e
r
i
a
l

i
s

r
o
l
l
e
d
,

d
r
a
w
n

o
r

e
x
t
r
u
d
e
d

w
i
t
h

a

h
i
g
h

c t

C
T
3
H W Z
s
h
o
w
n

b
y

D
I
L
L
A
M
O
R
E

a
n
d

R
O
B
E
R
T
S

[
3
9
]

a
n
d

R
E
G
c
o
<
&
-
H o
>

o 0
)
.
c
+
j
o
4
9

W
E

[
4
0
]

t
h
e

s
u
r
f
a
c
e

s
h
e
a
r

c
a
u
s
e
d

b
y

f
r
i
c
t
i
o
n

l
e
a
d
s
f
e
1
3
..-
i
h O
)

r
t
s Q
J
*
^
I
M O O
)
i
n
t

o
f

t
e
x
t
u
r
e
s

d
i
f
f
e
r
e
n
t

f
r
o
m

t
h
o
s
e

f
o
u
n
d

i
n

t
h
e

c
e
n
t
r
U
J
E
i
n

r
i
m
m
i
n
g

o Q
.I
P 1
-,
0
.
1
a
>

S U C
U
P

i
E
1

1
.
Q
J
-
c
:

i-
j
a -
i
-
'

o
1
c

n
A
'

i
T
-<
^
^
C
)
K
4
M K

-
i
u c
u

2
>
4
H
Q
J
n
i
M

3 o g o

4
J
1
r
1

s
,
c
u

a
q
c 0
-
-
-.
s 'E
<
M 0
)
u f
S
<
*
-.
g G 0
)
j
=
-
M
5
3
e
n
-
H Q
J
Q
J
1
o
1
0
0
0
S
T
R
E
S
S
M
N
/
m
2
1
0
0
1
0
E
q
u
i
a
x
e
d

s
a
m
p
l
e
s
1
0
1
0
0
1
0
0
0
1
0
0
0
0
R
U
P
T
U
R
E

L
I
F
E

A
T

8
0
0

C

(
h
r
s
)
F
i
g
.

1
0
.
1
5
R
u
p
t
u
r
e

l
i
f
e

a
t

8
0
0

C

a
s

a

f
u
n
c
t
i
o
n

o
f

t
h
e

a
p
p
l
i
e
d

s
t
r
e
s
s

f
o
r

a

c
a
s
t
N
i
/
C
r
/
A
l

a
l
l
o
y

w
i
t
h

c
o
l
u
m
n
a
r

g
r
a
i
n
s

s
t
r
e
s
s
e
d

a
l
o
n
g

a
n
d

t
r
a
n
s
v
e
r
s
e
t
o

t
h
e

c
o
l
u
m
n
a
r

a
x
i
s
.

S
q
u
a
r
e

s
y
m
b
o
l
s

i
n
d

c
a
t
e

r
e
s
u
l
t
s

f
r
o
m

a
s
i
m
i
l
a
r

m
a
t
e
r
i
a
l

b
u
t

w
i
t
h

e
q
u
i
a
x
e
d

g
r
a
i
n

s
t
r
u
c
t
u
r
e

{
a
f
t
e
r

V
E
R
S
N
Y
D
E
R
a
n
d

G
U
A
R
O

[
3
6
]

)
.
314
i
Fig. 10.16
Macrestructure of the surface of a 17% Cr-Fe alloy which has
developed ridges-or roping-on tensile strainlng (CHAO [38] ).
Fig. 10. 17
A tensile specimen of copper, and a cross-section of the same
sample showing that during tensile testint it has hecome bent
about the tensile axis The effect is due to a variation of
texture - and henee r-value-across the specimen thickness
(ROBERTS [42] ).
3 1 5
ROBERTS [42] observed an interesting phenomenon in a sample of
rolled and annealed copper as shown in Fig. 10.17. The tensile speci-
men has bent around the tensile axis during deformation. This is due
to different textures being formed on the opposite faces of the sample
because of differences in surface friction at these two locations during
rolling. On tensile straining the lateral contraction varies through the
thickness and the sample becomes bent.
2. 5. Work-Hardening Effects
In considering the control of recrystallization it must not be over-
looked that for a wide range of applications it is desirable to prevent
recrystallization. Aluminium and copper alloys are commonly market-
ed in the /4-hard, /2 hard, or fully hard condition, when use is
made of the cheap strengthening effect of work-hardening. In these
materials, and in lead and zinc alloys delibrate additions are made
to raise the recrystallization temperature.
As shown by GRANT [43] and by WILCOX and CLAU ER [44]the disloca-
tion structure resulting from the fabrication procedure and stabilized
by the thoria dispersin, makes a maj or contribution to the creep
strength of thoriated nickel. A procedure of repeated small deformations
and subsequent recovery anneals has been developed [45] to stabilize
the dislocation structure in thoriated nickel in.order to maximize the
high temperature strength of this material.
Partly recrystallized metis find little application but the recent ob-
servations of STOLOFF and DILLAMORE [46] and HENDERSON et al
[47] on a 49% Fe - 49% Co - 2% V ordered alloy and on chromi um
respectively suggest that for materials which are inherently brittle
a dplex structure of cold worked and fi ne grained recrystallized
316
material may be desirable. The resulta of STOLOFF and DILLAMORE
showing the variation of tensile strength and elongation to failure for
isochronal annealing In the range 500C to 800C are reproduced in
Fig. 10.18. The tensile strength variation shows the typical sigmoidal
recrystalUzation behaviour but the ductility exhibits a pronounced
mximum for the incompletely recrystallized material. As will be
discussed later tungsten lamp filaments probably benefit from this
effect, although this does not appear to have been appreciated.
3. THE CONTROL OF GRAIN STRUCTURE
In this section a number of examples of the delibrate control of
grain structure will be considered together with some interpretation
of the underlying mechanisms.
3.1. Inhibition of Recrystallization
Any alloy addition has the effect of raising the temperature for rapid
recrystallization. Solute elements are possibly effective in several
ways. They can affect the deformation structure or the recovery pro-
cesses which may lead to nucleation and they will undoubtedly contribute
to a solute drag retarding grain boundary migration [48, 49] . The
early results of SMART and SMTTH [50] on the effect of trace additions
on the temperature for half softening (T1/2) of spectrographically pur
copper shown in Table 1 illustrate the large effect which solute elements
can have on recrystallization (see chapter 7).
3 1 7
O
C\
w
O
O
OO
- i
iI Q M
r <S Q)
rt ...T3
o* 5 K -o
g o ^T *
C L. ' vifti?
o
o
vO
O
o
o
o
_
c ^a5
3 rt TJ^
^-S lo yj O "O CO

s g ri
-
S f i 3 B '
73 .2 -fi 4= -Mi
TJ ,. a) o S
O
O
O
O
o
C\
E
318
Table 1
Effect of Solute Elementa on the
Temperature for Half Softening
Composition (Wt%)
99,999% Cu
Cu + 0, 01 Ag
Cu + 0, 01 Cd
Cu + 0, 01 Sn
Cu + 0,01 Sb
Cu + 0, 01 Te
T r
Tl/2 L
140
205
305
315
320
370
These resulta make clear the reasons for adding small amounts of
tellurium to copper which is to be used in the cold worked condition.
Solute element additions are also made to the low melting metis zinc
and lead to confer reasonable room temperature strength. 1% copper
is added to zinc in material to be used in the cold worked condition and
this raises the softening temperature from 20C to 175C [51] .
Additions of up to 6% antimony are made in the case of lead but it is
not clear that the effect here is simply on recrystallization processes.
Elements which are held in supersaturated solid solution by cooling
sufficiently rapidly from the hot working (or casting) temperature can
be particularly effective in inhibiting recrystallization after cold work-
ing (c.f. chapter 8). A well known example of this effect is that of
aluminium killed steels, in which Al N is not precipitated if coiling is
carried out below about 550 C, but does start to preciptate at around
500 C after cold working. The recrystallization kinetics of an aluminium
3 1 9
killed steel at two temperaturas are compared in Fig. 10.19 with those
obtained from a rimming steel, both steels being annealed isothermally
after 60% deformation by rolling [52] .
As will be discussed later the influence of such precipitating phases on
recrystallization is important in relation to the control of preferred
orientation, but it is not generally applied in practice to affect the soften-
ing temperature.
Dispersed phases are used to stabilize the dislocation structure in a
variety of materials: S. A. P. (sintered aluminium powder), D. S. L.
(dispersin strengthened lead) and thoriated nickel represent perhaps
the most commonly reported materials of this kind. There now appears
to be little doubt that the high temperature strength of these materials
is due more to the effect of the particles in stabilizing the deformation
substructure than to any direct particle strengthening [43,44] (c.f.
chapter 8).
The mechanisms by which a dispersed phase stabilizes the deformed
structure has been discussed at length by a number of workers[45, 53, 54]
and HUMPHREYS and MARTIN [54] are undoubtedly correct in asserting
that the principie effect o the dispersed phase, if it is appropriately dis-
persed, is to homogenize the deformation substructures. Although a
dispersed phase undoubtedly retarda grain boundary migration it could
not truly stabilize the deformation structure if it did not prevent nuclea-
tion. A recrystallization nucleus must be a regin of microstructural
heterogeneity (c. f. chapter 3} and it follows that the dispersed phase
has to inhibit nucleation either by totally inhibiting grain boundary migra-
tion or by homogenizing the substructure.
The treatment applied in the case of thoriated nickel to stabilize the
substructure, namely repeated small deformations interspersed with
recovery anneals is one which will cause the me r estructure to become
3 2 0
(/)
*E
UJ
o
O
O
LO
B i1^1
.2 ^ L
5 S P
S i . S
in * * ^
S S
a"3 S
*"" 03 -H
IU 1-1 _.
E
c
g
c ~
OM
bfl n 3J
i-rt j 'r* _5
1*1 QT uPrt
321
progressively more homogeneous and a recent study [55] has shown
that this is indeed the case. In the work of B UTL E R [55] the recrystalliz-
ation behaviour of pur nickel and nickel 2, 5 vol % AlgOg feeometric
mean particle size 210 A - 10 A ) has been studied after both cold
rolling and warm rolling at 450C to reductions of 75%. The recrystalliz-
ation curves for annealing at 450C are reproduced in Fig. 10. 20 from
which it may be seen that recrystallization is always slower in the
nickel-alumina than in the nickel. After room temperature deformation
this is attributed to the particle drag effect on grain growth in the two
phase alloy. Warm deformation accelerates recrystallization in nickel
but gives a recrystallized grain size which is hardly different from
that after cold deformation; in the nickel-alumina warm deformation
retards recrystallization but not through any influence on growth rate.
The greatly increased grain size of the warm-rolled and recrystallized
material, as compared with the cold-rolled and recrystallized, indicates
an approximate three-fold reduction in the number of nuclei and the
kinetic differences suggest a reduction in the average nucleation rate
of about two orders of magnitude. This is entirely consistent with the
idea that the particles stabilize the substructure by making it more
homogeneous, since the effect of warm-working is similar to that of
the cold work-recovery treatments in homogenizingthe structure.
IRAN and L ATHAM [56] have reported good thermal stability of the
dislocation structure for plain carbn steels which have been warm-
worked in the pearlite transformation range after austenitizing (this
treatment has been called isoforming). The micrestructure developed,
shown in Fig. 10. 21, is of small subgrains of O, 5 / um diameter pinned
by coarsespheroidized Fe,,C particles of 0,2 - O, 3 / um diameter. The
particle-dislocation interactions are somewhat different in this case
than in the dispersin hardened nickel since in "isoforming" the dis-
locations are created and the Fe3C particles are formed simultaneous-
ly. The structural stability depends nonetheless on the uniformity of
the subgrain structure.
322
323
15
a
|u o
o
_
oo
I/)
O
0)
.c
O
o
o
m
^
3
t4
b > 3\H <U
O QO
X
= ou
, oo
1 oo
ta
41
U -
< U
o.c
>
i-H
rt
1-1
o
t-
tu ^
E"
a
o
o
N m
= ,in
e ^t p j
s w
ge J
g, s
Fig. 1 0.21
Ele ctron micrograp h s howingthe s ubgrain s tru ctu re
and coars e ce me ntite p artile s which re s ults f rom
is oforminga s teel of comp os ition O, 5% C, 1 % Cr
(afte r IRAN and LATHOM [56]}.
324
Partial stabilization of the deformed structure is in some cases un-
desirable, as for instance when it leads to sluggish recrystallization
in enamelling irons [57] and rimming steel, but in other cases it may
be desirable as is the case for tungsten lamp filament wire [58j . The
phenomena involved in both cases are similar and explanations of them
in the literature have concentrated on the segregation of impurities or
dopants to grain boundaries during recrystallization as the dominant
factor. This is certainly an important contribution leading to the in-
complete recrystallization but it is not the only factor. Textural effects
are involved and these will be dealt with in a later section.
In enamelling irons the segregating species are raanganese and sulphur,
which, when the oxygen content is high [59] , may be retained in solid
solution after hot-rolling, and subsequently preciptate, or segregate,
during recrystallization, with the effect either of preventing complete
recrystallization or of giving a dplex grain structure in the annealed
material. In either case the formability and surface finish of the pro-
duct is impaired.
The tungsten filament wire shows two stage annealing kinetics, as shown
by Fig. 10. 22, taken from the work of DAVIS [58]. The two stages have
been designated primary and secondary recrystallization respectively [58]
but this terminology does not coincide with the common usage. As shown
by Fig. 10. 23, a micrograph from a sample annealed to the end of stage
one, the structure at this point consists of a mixture of long recrystalliz-
ed grains and unrecrystallized reas. The recrystallized grains are
prevented from growing at the expense of the unrecrystallized regions
by impurities or dopants, such as K_O, ThO2 or MgO, which are de-
liberately introduced in very small quantities (see also chapter 5 section
7).
3 2 5
3000 -
TENSILE
STRENGTH
2000
MN/m2
1000 -
O
ELO N
%
20
10
O
O 600 1200 1800 2400
T C
Fig. 10.22
Tensile strength and % elongation as a funcion of
temperature for heavly drawn tungsten wire anneal-
ed for 10 minutes at temperature (after DAVIS [58] ).
3?6
i *dt-AJ "" 11 *-*.-"**-, t ,-*." '".u-nf ~~
tagf-,^--"- -,_.- ^ix'^Sg vr*.,
SPIli^'%s^~?^l
i_^r.^y^^?^""-~-;<^ii^F^^--i
*M-***talli**^l
u*T<w^ _
f**^-nc~^'^
f-ijF*^!-;/ j
--.g"1'-.''rffc*q!li|y? J>.->-^ ^-J"'^^! -1"< 1!
^^^_ _>"^(!l'"^"*v*, -^"^ 1_7^_X^tfl X^1^"***1"
-~.MI t- -^*-^ ^ ,-w^v %^-r'i^>-^tr "-.^ . - '/_ " ^i
Fig. 10. 23
Micr ostr uctur e of dr awn tuestenwir e similar to that
inFig. 10. 22 annealedto the er dof the fir sl stage,
showinp a mixedmicr ostr uctur of coldv /or kedand
r ecr ystallizedgr ains (after DAVI S [58] ).
3 2 7
DAVI S has shownthat the secondannealing stage is defer r edto higher
temper atur es the gr eater the r eductionimpar tedto the wir e befor e anneal-
ing. This obser v ationis consistent with the r esults of LEBER [60] onthe
annealing of dr awnmolybdenum wir e, for which it was found that r e-
cr ystallizationoccur r ed less r eadily the gr eater the defor mation. The
explanationof this effect lies inthe incr easing unifor mity of str uctur e
accompanying the shar pening of the single component<110>fibr e textur e.
The dplex str uctur e intungstenfilament wir e is useful ih that the
mater ial is least br ittle whenit is I n this condition, as indicatedby the
ductility v ar iationshowninFig. 10. 22.
2. Gr ainSize Contr ol
'
Contr ol of gr ainsize is usually effectedincor omer ciar pr actice by
applying the empir cal "Laws of Recr ystallization". For instance, if a
coar se gr ainedmater ial is r equir ed, either the pr ocess schedule is
ar r angedso that a low total r eductionpr ecedes the final anneal or a
long annealing tr eatment is giv enso as to allow extensv e gr aingr owth.
I neither case a single phase mater ial with low total impur ties would
giv e the lar gest gr ainsize inany giv enannealing tr eatment.
The long gr aingr owth anneal pr ocedur e is usedfor the dev elopment of
moder ately coar se gr ainedstr uctur es which ar e r equir edto minimize
the Lder s str ainandgiv e optimum for mablity inlow car bnsheet
steel. 78 hour cycles inbox fur naces with about 20 hour s at 690 C ar e
typical andthe r esultar gr ainsize is ar oundASTM 6 (gr aindiameter
- 45 uro).
Copper wir e usedfor elctr ica! power tr ansmissionis oftencoatedwith
enamel for insulationpur poses. The total pr ocess, as well as economic
consider ations r equir e that the stov ing oper ation, inwhich the enamel
is bakedonto the wir e, should ser v e also as the r ecr ystallizatfonanneal
328
for the drawn wire. The stoving operation is only a short time at 180 C
and to meet the requirement of good conductivity it is desirable to have
a coarse grain size, Clearly the elements tellurium4 antimony, Un
and cadmium Usted in Table 1 and others such as phosphorus and
arsenic which also retard recrystallization and grain growth are to be
avoided as impurities in this material, especially as they all impair con-
ductivity in other ways.
Secondary recrystallization leads to the development of coarse grain
structures in a rather different manner from normal grain growth, It
has frequently been stated (see e. g. the article by WALTER and DUNN
[61] ) that the essential requirement for secondary recrystallization to
develop is that primary grain growth should be restricted by one of
three effects, surface inhibition, texture inhibition or second phase
particle restraint. The last of theae is by far the most important and
will be the only one dealt with here.
An analysis of grain growth, due to HILLERT [62] shows clearly how
the matrix structure for secondary recrystallization is developed in the
presence of a second phase dispersin. For a single phase material
HIEKEsuggests that the driving forc causing a grain of diameter D
to grow (or shrink) will be proportional to
(3)
where D is the diameter of a grain, of near average size, which ex-
periences zero driving forc for growth, gf_ is the specific grain
boundary energy. This type of expression obviously recognises that
for grains of D < D shrinkage occurs and that grains having D > D
grow. In a two phase material the Zener drag forc, given by
3f ' " - a > where f is the volume fraction of second phase and d,
2
t i
d.
329
is the geometrical mean particle diameter, reduces the driving forc
for both growth and shrinkage, so that the driving forc for growth or
shrinkage of a grain of diameter D is now proportional to
1 + 3f
D 2d:
(4)
where the sign is chosen to minimize the term inside the brackets.
*if v /1 1 \y if the drag forc TJ^-^B is greater than / 1- 3 f _
2di \ D/ B
the grain of diameter D will nelther grow or shrink, so that grains of
diameters between
Di =
D
.
and
D2 -
D
(5)
- 3f
do not alter their size. Growth is restricted to grains larger than D-
which grow at the expense of those smaller than D... The result of this
is that as grain growth proceeds, the average grain diameter (taken to
be equal to D) increases, so D, increases and fewer and fewer grains
2
are able to grow. In the limit grain growth ceases completely when
D = 2d./3f, but the grain size distribution has been spread out, develop-
ing a tail at the large grain size end, as shown by the curves for all
grains in the distribution in Fig. 10.24. The resulta of Fig. 10. 24 are
taken from the work of HUTCHINSON [63] and relate to grain growth
in ferrite with an austenite dispersin.
2d
If growth has ceased at the point where D = - &- , the grain size dis-
tribution having meanwhile developed a tail as shown in Fig.10. 24,and the
volume fraction of second phase is then reduced progressively, so as
to always control the rate of grain growth, the grains that are allowed
A
L
L

G
R
A
I
N
S
%

G
R
A
I
N
S
O
F

S
I
Z
E
'
D
'
)

G
R
A
I
N
S

J
1
L

=
0
-
2
0
Z
t
o
t
a
l
O
O
1
0
2
0
3
0
G
R
A
I
N

S
I
Z
E

'
D
'
j
j
m
G
r
a
i
n

s
i
z
e

d
i
s
t
r
i
b
v
r
t
i
o
n
s

f
o
r

t
h
e

t
o
t
a
l

g
r
a
i
n

p
o
p
u
l
a
t
i
o
n

f
a
l
l

g
r
a
i
n
s
)
a
n
d

f
o
r

t
h
o
s
e

w
i
t
h

(
1
1
1
)

s
h
e
e
t

p
l
a

e

(
a
s

d
i
s
t
i
n
g
u
i
s
h
e
d

b
y

a

B
i
t
t
e
r
F
i
g
u
r
e

1
0
.

2
4

p
a
t
t
e
r
n

t
e
c
h
n
i
q
u
e
)

i
n

a

l
o
w

c
a
r
b

n

r
i
m
m
i
n
g

s
t
e
e
l

t
r
e
a
t
e
d

t
o

h
a
v
e
a

f
i
n
e

c
a
r
b
i
d
e

d
i
s
p
e
r
s
i

n

p
r
i
o
r

t
o

r
o
l
l
i
n
g

8
2
j

5
%

a
n
d

a
n
n
e
a
l
e
d
a
t

(
a
)

7
0
0

C

-

j
u
s
t

r
e
c
r
y
s
t
a
l
l
i
z
e
d

a
n
d

(
b
)

7
8
0

C

f
o
r

6
4

h
r
s
(
a
f
t
e
r

H
U
T
C
H
I
N
S
O
N

[
6
3
]
)
.
%

G
R
A
I
N
S

1
5
O
F

S
I
Z
E
'
D
'
6
4

H
R
S
=

0
-
4
6
A
L
L

G
R
A
I
N
S
O
3
0

6
0

9
0

1
2
0

1
5
0
G
R
A
I
N

S
I
Z
E

'
D
'
j
j
m
0
0
o
t
a
u
332
to grow are those at the large size extreme of the grain size distribution.
This reaults in the progressive extensin of the tail to the distribution
until the optical appearance is of a dplex grain structure. If now the
second phase is completely removed the largest grains grow rapidly at
the expense of the fine gained matrix and aecondary recrystallization
may be said to have occurred.
Obviously the removal of a second phase after recrystallization and some
grain growth is an important factor in the development of secondary re-
crystallization, as has been strikingly demonstrated by CALVET and
RENON [64] , and the sequence described here is of commercial signi-
ficance in the development of the Goss texture in silicon-iron [61],
deep drawing textures in rimming steels [65] and in the occurrence of
surface defects due to coarse grains on aluminium and brass extrusions.
The second phase may be removed either by being taken into solution
[61, 64] or by a surface reaction such as decarburisation, leading to
a change in composition [65].
A fine grain size finds application chiefly in structural steels, in which
a good combination of strength and toughness is sought. A detailed dis-
cussion of the means by which the fine grain sizes are obtained is not
appropriate here, since the phase transformation fr om austenite to
ferrite, or in normalized steels from ferrite-austenite-ferrite, con-
tributes appreciably to grain refinement. However there is in a given
steel a correlation between austenite grain size and the ferrite grain
size resulting from transformation so that recrystallization and grain
growth processes in austenite are of some importance.
Aluminium killed steels are preferred when the steel is to be used in
the normalized condition, although semi-killed or balanced steels con-
taining vanadium would be equally good. The underlying reason for the
cholee is that Al N particles (or VN partiles) prevent grain coarsening
in the Austenite during the normalizing treatment and thus contribute to
3 3 3
ferrite grain refinement. As may be appreciated from the discussion
of secondary recrystallization the use of particles in this way is not
without its dangers. Ideally the grain size distribution, resulting from
the ferrite to austenite transformation, and the dispersin parameters
should be such that the largest grain has a diameter, eq. 5
(6) D max D
2 '
GLADMAN [66] has considered this question in some detall, he arrives
at an expression similar to the one presented here with slight differences
in numerical constants.
Particle interactions with the recrystallization process in austenite are
of an essentially similar nature to those already discussed in the re-
crystallization of ferrite. In steels which are to be used n the hot rolled
condition good combinations of strength and toughness (see Fig. 10. 6)
can be obtained through the use of niobium containing steels [67] and
controlling the hot rolling sequence to retain the Nb (C, N) in solution
until the last stages of hot-rolling, these being carried out at temper-
atures below 1000C.
The effect of the Nb(C, N) precipitation is to prevent the recrystallization
of the austenite; the deformed austenite transforms to a fine grained
ferrite. It is commonly thought that precipitation on the dislocation
structure in the austenite prevens subgrain growth and thus the nuclea-
tion of recrystallization, but, as in the case of Al-killed sheet steels
and enamelling irons, it is more probable that precipitation on grain
boundaries is the dominant effect.
!.3. Texture Control
Positive steps to control preferred orientation are taken in severa!
instances: the avoidance of cube-texture in cold rolled and annealed
copper sheet, the avoidance of earring in aluminium sheet, the
334
3 3 5
development of good deep-drawabilityinlow carbn sheet steel and of
the Goss texture in silicon-iron sheet for transf ormer applications re-
present the most important examples (see also chapter 6).
In copper the cube texture developa after rolling reductions of 90 - 95%
and is essentially a single component texture which endows the material
with mechanical properties hardly better than those encountered in anneal-
ed single crystals. The obvious expedlent of avoiding large cold rolling
reductions in the processing of copper to sheet is one means of avoiding
this problem, another is to use phosphorus deoxidized copper which is
not prone to the development of the cube texture.
The reasons for the development of so strong a texture have exited a
great deal of interest for almost 50 years and there is still no general
agreement ont he mechanisms involved (see for instance the various
articles in Ref . 1). It would appear f rom the work of HU [68]and others
[69] that grains in the cube orientation are particularly f avoured in the
nucleation stage of recrystallization and establish an early size ad-
vantage which enables them to grow to dominate the structure during
grain growth following recrystallization. Nucleation of the cube texture
has been shown by HU [68] to be associated with regions of high lattice
curvature, or transition bands and such sites are f avoured for early
nucleation, even if subgrain rotation does not occur.
The mechanism by which phosphorus (and also arsenic, antimony and
germanium) interf eres with the development of the cube texture is not
known but may be associated with the fact that phosphorus bearing copper
alloys exhibit strain ageing phenomena, indicating that phosphorus
segregates to dislocations and may thus have a greater inf luence on
subgrain nucleation mechanisms than on grain boundary migration
mechanisms of nucleation.
The cube texture is less readily developed in aluminium than in copper;
whereas in copper the annealing texture is almost 100% cube orientation,
in commercially pur aluminium the cube texture co-exists with a
texture designated the "S-texture" by B UNK et al [70] . The texture
of 95% def ormed and annealed aluminium [71] is shown in Fig. 10. 25,
A proportion of the cube component is desirable in aluminium sheet for
deep drawing applications since the propensity f or earring can be re-
moved by obtaining the correct proportions of the "S" and cube com-
ponents. The "S-texture" gives ears at 45 and the cube texture at
O and 90 to the rolling direction. Clearly super-position of the two
ear positions gives essentially no earring.
It is possible by suitable control of chemistry and processing to obtain
textures which range f rom 100% "S" to 100% cube component. The
maj or contrbutions to understanding and perf ecting this control have
been made by Germn workers, notably SIEB EL, B UNK , NO RM A NN
and M A SING and work on the topic has been reviewedby GREWEN [72].
In general terms the principal impurities in commercial al uminium
are iron and silicon and if these are held in solution af ter hot rolling,
or af ter casting, precipitation occurs during the f inal anneal and this
interf eres with the development of the cube texture. In order to
develop the cube texture the A I-Fe-Si preciptales must be present be-
f ore the f inal cold rolling reduction, or the levis of iron and silicon
must be below the solubility limit at the annealing temperature.
In terms of the mechanisms involved the precipitation occurring con-
currently with recrystallization prevens the f irst f ormed nuclei, which
are of the cube orientation [73] , f rom capltalizing on ther advantage.
Precipitation on migrating high angle boundaries slows t hem down,
possibly to zero velocity, thus requiring the unrecrystallized volume
to continu to nucleate n order for recrystallization to proceed.
3 3 6
Fig. 10. 25
{lll}pole figure of commercial purity aluminium
showingthe "S-texture" () and cube texture (*}
components in a material rolled 95% and annealed
5 min. at 300C (after P. BECK and H. HU [71] ).
337
Later nuclei a-3 not cube oriented, indt I in aluminium rather few cube
nuclei exist and f the correct texture balance is to be achieved most of
these nuclei must b e allowed to prosper.
The relativo kinetics of recrystallization and of the precipitation of the
Al-Fe-Si phases make it unlikely that precipitation could occur in
aluminiurn prior to recrystallization but in other cases the heating
cycle can be adjusted to obtain the correct juxta-position of the two
processes. This is the case in aluminium kled low carbn sheet
steels but as a preliminary to reviewing the influence of AlN precipita-
tion the textures development in rimming steels after reductions in the
range 65% - 85% will be described.
The cold rolling texture common to all b. c. c. metis rolled to greater
than about 60% reduction is shown in Fig. 10. 26 a and the texture of
rimming steel rolled 84% and annealed to be just 100% recrystallized
is shown in Fig, 10. 26 b. Superimposed on both of these J200}pole
figures are the {200| poles of 40 grains chosen at random from among
the first 10% of recrystallized grains, as judged in the electrn
microscope. It is clear that the earliest recrystallized grains are
more typical of the annealing texture than the deformation texture,
although the orientations developed on annealing are well represe nted
in the deformation texture. In Fig. 10.27 the ptica! microstructure
of the material annealed to 67% recrystallization is shown, together
with the |200J-pole figure obtained from the same sample. It may be
seen from the pole figure that negligible changes have occurred in the
orientation distribution up to this point in the progress of recrystalliz-
ation. The remaining 33% of unrecrystallized material must clearly
be predominantly of orientations near to (100) [Ol] , as may be seen
by comparing Figs. 10. 26 a and b and Fig. 10.27, This is confirmed
by etch pit determinations of orientation.
3 3 8
339
8
13 o i
HJ 3 O
ST^ o
s SUS ^
g^ =
H
-HA
Cii
47
NV
-I -H Lj
33
tC
r-i iHM
tH t-( rt
OOi-l
f ti ^**^.^-"-^
~! 3 S - --1
0) rf -^ NO iH H
00
0*^0
Fig. 10.27
The micro-structure (a) and corresponding {200/ pole f igure
(b) of a sample of rimming steej annealed to 67% recrystalliz-
ation after a reduction of 80% by rolling.
* (100) COl l l
* (211) C011 ]
* (111) [112]
( a f *- - ""TCHINSON [63] ).
3 4 0
The electrn diffraction data of Fig. 10.26 indicates that the earliest
recrystallized grains, shown by Fig. 10.27 to f o rm in colonies, have
near (111) sheet plae orientations. The low stored energy and uni f orm
cell structure of the (100) [Ol] orientation render grains of this orienta-
tion incapable of providing subgrain nuclei, but, as has been discussed
by DILLAMORE et al [74] grain boundary bowing, as proposed by BECK
[75] may oprate early in the annealing sequence to form bulge nuclei
of the (100) [Ol] orientation. If the bulge mechanism does not oprate
at zero, or short times the (100) [Olljgrains resist recrystallization
until such time as the near (111) oriented grai ns have fully recrystallized
and competitive growth within the colonies has occurred to provide grains
of sufficiently large sze to grow into (100) [Ol] under the low driving
forc provided by the residual stored energy in this orientation. The
first grain to reach sufficiently large size to be able to grow into the
large unrecrystallized volume will establish for itself a clear size ad-
vantage in the 100% recrystallized structure, and this accounts for the
results of Fig. 10.24a, in which the (111) grains occur preferentially
at the large grain size end of the grain size distribution. As shown by
Fig. 10. 24 b the (111) grains are able, in the presence of a second phase
dispersin, to capitalize on this size advantage during subsequent growth.
The processing sequence used in the case of alumi ni um killed steels in-
volves final hot rolling temperatures above A 3 and coiling temperatures
around 550C [2lJ and, after cold reductions of the order of 60 - 80%,
a slow heating rate to the anneali ng temperatura of 690 C. This sequence
retains the aluminium nitride in solution until the f i nal anneal and its
precipitation just prior to recrystallization in addition to retarding re-
crystallization (Fig. 10.19) effects an i mprovernent in the texture
appropriate to deep drawing.
34 1
As may be seen f ro m the results of HELD [41] shown in Fig. 10. 28 the
texture required to give a high r valu and thus good deep drawability
(Fig. 10.10) has a high ratio of the volume fraction of materi al with
(111) sheet plae to that wi th (100) sheet plae. Precipitation of Al N
just pri or to recrystallization has the effect f increasing the volume fraction
of the (111) components and reducing the (100). It is perhaps worth-
whi le to bri ng forward evidence to demnstrate that Al N is indeed pre-
cipitated at the time when it is effective, since there has been some
dispute over the question of whether the Al N is precipitated or merely
segregated. Fig. 10. 29 is an electrn micrograph from a sample of
alumi ni um killed steel whi ch has been annealed for 1 hr at 500C follow-
ing a solution treatment. The Al N partiles are clearly indicated by
the strain contrast. Table 2 shows the texture ratio (lll)A for
the same alumi ni um killed steel following three processing
sequences, as indicated.
Table 2
'(100)
1
1 Process Sequence
Rolled 85% annealed 4 min 780C
Precipitation treatment 1 hr 500C
Rolled 85% annealed 4 min 780C
Roled 85% precipitation treated
, 1 hr 500 C annealed 4 min 780 C
I/i i i \ Average
1/100) ; Grai n Size ,um
2,1 5,8
4, 2 , 7, 5
5,5 9, 9
i . .. _ _
It can be seen that the precipitates present prior to cold rolling are
ef f ecti ve in raising the (111) : (100) ratio and increasing the grai n size
as compared with the materi al in whi ch precipitation did not precede
recrystallization, but that precipitation following deformation is, in
thi s i nstance, more effective in increasing both the (111) : (100) ratio
and the grai n size. As shown by Fig. 10. 30 both of the materials
342
OJ
O
; o
o 2
o
o
o
f\
O)
O C*
J= ct
U^4 O*
O -i-.
g ^
0JS
Sw

Og' rn
^ -S C
4- >"-1
+ rt
W
.s
Itl rt
rt
> .2
< u 5
>"o
5
343
fc
E
Fig. 10. 29
Electron nncrograph of aluminium killed steel solution
treated and aged 24hrs 500C, showing the presence of
sibly spherical, precipitates as evidenced by the
trata fields with lines of no contrast perpendicular to
; operating diffraction rector (after H U T CfflN S O N [63])
344
treated to cause precipitation before recrystallization have an elongated
grain structure, while the thlrd sample has not. The Al N particles
(Nb C, Ti( C,N) , WC and other precipitates act in a similar way although
the precipitation kinetics and corresponding process sequences differ)
are effective in two ways: one, as already inferred, is to pin grain
boundaries so that (100) [Ol] bulge nuclei do not form early in the
annealing sequence, and the other is to interfere with subgrain growth,
in a similar manner to the particle interaction with grain growth discuss-
ed previously, thereby resulting in more selective subgrain nucleation of
recrystallization. Both of these effects enhance the (111) components
at the expense of the (100).
The mechanisms involved in the development of the Goss texture in trans-
former steels have already been described in considering the effects of
a dispersed phase on grain growth processes. The only requirement
beyond that of having the appropriate dispersin (of MnS in this case)
is that the |llo}<001> grains occur at the large grain size end of the
distribution in much the same way as the (111) grains do in Fig. 10. 24.
Grain growth in the presence of the second phase and the subsequent re-
moval of the second phase by solution treatment, or, as in modern pro-
cessing [75] by surface reaction, then leads to secondary recrystalliz-
ation to the {lio} * C 001> texture. The only question that remains to be
answered is how the |110\<001> grains come to have a size advantage
in the initial recrystallized material, and although there is no direct ex-
perimental evidence on this point some circumstantial evidence is
available.
The process route requires a coarse grain size in the hot band material,
and this is achieved by using as high a finishing temperature as possible.
The hot band which is about 2 mm - 3 mm thick is cold rolled to about
O, 75 mm and annealed. The initial coarse grain size favours deformation
banding and grains of the initial {lto|<001>orientation are, as shown
345
-, Mf-
** *ZL^:-^""-_ C;
.I-^K""-'-*"""-^j" '\V^.L_.. fr_. '"^in-yUJ-j " ' ^
^m M i" ! v - **r^^'* "^-_" * JC Tj* ^-- ''1"^-^. . J^
-^Tv Ai' VrjS-*-^l29^t**-S
TV3=^ -*-V -v ---j-T^~t-!'i"~2^%'-- - *^i
.^^^Li::^^,, . -J-z^\ **--^r>*' -
f ^*-^**^~i _^r r-^Biii """pj" ^" ~ Cj^-' -**-*-* .^^' r
o,
J=
-t-1
i
a

o0

t-t-
.S 5
E 5
"e-
"O -t
Q o
i?
|a
a
00
" ^g
-S'o'"
'SS")
o b
e -s t, -a _
^
< l O
O.
t
3 4 6
by HU [77] and WALTER and KOCH [78] particularly prone to deforma-
tion, or transtion banding. Such transition bands are especially favour-
ed sites for nucleation so that the {lio}<001> component ig developed
to some extent in the O, 75 ram strip during the annealing treatment, A
second cold reduction, to about O, 3 mm thickness, followed by a further
annealing treatment further enhances the {llo}<001> component and
again the transition band nucleated grains forra early in the annealing
treatment and grow to domnate the large size end of the distribution
of grain sizes in the recrystallized material. The intermedate and
final recrystallization anneals at 850C are followed by a secondary re-
crystallization anneal at about 1200C in which the {lid}<001>grains
grow at the expense of all else.
3.4. The Influence of Preferred Orientation on Recrystallization
Textures can influence recrystallization kinetics, as has been inferred
earlier, and as recognised by LORIA et al [79] and by RICHARDS and
WAT SON [80] .
LORIA et al suggested, but brought forward no evidence to prove, that
the exponent n in the Avrami equation
X = 1 - exp (-AtK),
where X is the fraction recrystallized, t is time and A is a constant, is
influenced by microstructural variables, including texture, and by
Chemical variables. K is usually represented as indicating the
dimensionality of growth, but as pointed out by LORIA et al the optical
appearance and the exponent K rarely lead to the same conclusions in
this respect. Certainly the results shown in Fig. 10. 27 indcate that
spatially random nucleation, as assumed implicity in the Avrami analysis,
does not always occur and this is something that is influenced by texture.
347
The resistance to recrystallization shown by some regions in enamelling
irons also leads to microstructures similar to the one shown in Fig. 10. 27
and this may not be a mere coincidence. An arrest in the progresa of re-
crystallization occurs at the stage represented by Fig. 10. 27 and thus
the difference between this material and enamelling iron is only of
degree; the presence of MnO or MnS precipitates distributed uniform-
ly would obviously extend the arrest so it is not necessary to assum
manganese segregation as a major cause of this effect.
This interpretaron is reinforced by the results presented by BAIRD
and ARROWSMITH [81] and by LESLIE et al [57] . Bbth sets of workers
studied sluggish recrystallization in a low carbn iron but BAIRD and
ARROWSMITH used rolling and LESUE et al used drawing as the de-
formation process with the result that major differences in behaviour
were found, the drawn material being far more sluggish in recrystalliz-
ation. As pointed out by RICHARDS and WATSON this difference is not
unexpected. RICHARDS and WATSON drew on the results of DILLA-
MORE et al [74] in pointing out that the energy stored in a crystal dur-
ing deformation depends on its orientation relative to the axes of strain.
Differences in orientation are thus a cause of microstructural heterogene-
ity and thus probably of recrystallization nuclei (of the bulge type, as
discussed by DILLAMORE et al). A single component texture, such as
that formed in drawn wire of b. c. c. metis, will thus be deprived of one
source of nuclei and may be expected to be correspondingly slower to
recrystallize. This factor accounts for the results of LEBER [60],
contributes to the results of DAVIS [58] and accounts for the differences
between recrystallization after rolling and after drawing. It may also
be a factor in the stabilization of substructure in thoriated nickel, this
being more readily achieved in extruded than in rolled material.
348
4. CONCLUSIONS
An attempt has been made to outline some of the reasons both for
controlling grain structures and for the steps which are taken in practice
to achieve this control. It has not been the intention to give recipes for
improving producs, or to give an exhaustivo survey, but to provide
understanding of the control which is available.
Although there has been some recourse to theory and to speculation
a mainly phenomenological viewpoint has been adopted. It may be noted
in particular that, in discussing texture control, oriented nucleation and
oriented growth were not mentioned: such formal theories have been
consigned to history with the application of electrn microscopy to the
study of texture development.
The examples of texture control in cubic metis, which have been dis-
cussed, relatin gto deep drawing properties of steel, earrin g character-
istics in aluminium, and the development of the Goss texture in silicon-
iron are the only clear practical examples of the commercial application
of this form of control. It is possible that textures may in future be
influenced in order to develop specific elastic properties but there is
little prospect that texture control will achieve wider importance as a
primary tool in affecting strength per se or ductility. As a means of
control over secondary variables, as for instance avoiding the problem
of roping or of influencing recrystallization kinetics, textures will con-
tin u to develop in importance as one of the microstructural variables
which need to be considered in quality control.
Hexagonal metis, having fewer deformation modes than cubic metis
have mechanical properties which are more sensitive to texture, and
here more attention will be given in future to the range of control which
349
can be exercized. This is especially to be expected for titanium, as
this metal finds wider application.
The examples of grain size control which have been considered have
drawn attention particularly to the role of second phases, or grain
boundary segregatea, in retarding grain growth and stabilizing fine
grained structures. Other applications not dealt with, in which the
second phase dispersin acts in this way to give desirable properties,
relate to superplastic behaviour and to obtaining good creep ductility.
It should be remembered, however, that a dispersin can contribute
to the development of secondary recrystallization and coarse grained
structures so that grain refin emen t by a dispersin has to be used with
care.
Partice stimulation of nucleation has not been discussed: atthough the
rapid reepystallization usually obtained by stimulated nucleation would
help in shortening processing the optically observable particles
f > 1 yum) which are especially effective in this respect generally
have undesirable effects on properties.
The use of a second phase has been considered in relation to texture
development in steel and aluminium and the sluggish recrystallization
of tungsten wire and enamelling irons and the discussion of the processes
involved has been intended to show their common features. The
mechan isms discussed in detall for the examples concerning steel appear
to be applicable in most respects to those involving other metis. As
discussed by LESUE et al [82]steels and other metis have a great
deal in common.
3 5 0
REFERENCES
l.
2. D.V.Wilson:
3. W. Boas and
J. K. Mackenzie in:
4. W. Sylvestrowicz and
E. O. Hall:
5. D. V. Wilson and
B. Russell:
6. W. B. Morrison:
7. R, Armstrong,
I. Codd,
R. M. Douthwaite and
N. J. Petch:
8. J. T. Evans and
R. Rawlings:
9. D. V. Wilson and
J. A. Chapman:
10. R. P. Carreker and
W. R. Hibbard Jr.:
11. F.Garofalo,
W. Domis and
F. Gemmingen:
12. D. M. R. Taplin and
V. N. Whittaker:
13. J. A. Williams:
14. E. Smith and
J. T.Barnby:
15. J.Kewell:
16. M. Braunovic and
C.W. Haworth:
17. M. F. Littman:
18. J.Swift,
P. S. Dobson and
I. L. Dillamore:
"Textures in Research and Practico",
J. Grewen and G. Wassermann, Ed.,
Springer, Berln 1969.
Met. Revs. 14 (1969) 175.
"Progresa in Metal Physics", Vol. 2_,
B. Chalmers, Ed., Pergamon Press,
New York, 1952, p. 90.
Proc. Phys. Soc. B64 (1951)495.
Acta Met. 8_ (1960) 36.
Trans. ASM ^7 (1966)824.
Phil. Mag. 7 (1962) 45.
Met. Sci. J. 2 (1968) 221.
Phil. Mag. 8 (1963) 1543.
Acta. Met. 1 (1953) 654.
Trans. Met. Soc., AIME 230 (1964) 1460.
J.Inst. Met. 92 (1963/4)426.
Acta Met. ^5 (1967) 1119.
Met. Sci. J. 1 (1967) 1.
BISRA, Prvate Communication.
J. Appl. Phys. 40_ (1969) 3459.
J, Appl. Phys. ^8 (1967)1104.
Birmingham University - unpublished.
3 5 1
19. E.Schmid:
20. T.W. Watson:
21. R. L. Whiteley and
D. E. Wise in:
22. D.V.Wilson and
R. D. Butler:
23. M. Atkinson and
I. M. Me Lean:
24. T. U. Richards in:
25. R. Priestner and
N. Louat:
26. T.H. Webster:
27. W. F. Hosford and
W. A. Backofen:
28. G. Y. Chin and
B. Wonciewicz:
29. M. A. Adams:
30. I. L. Dillamore,
R. E. Smallman and
D.V.Wilson:
31. D. S. Hoddinott and
G. J.Davies:
32. L. K. France,
C. S. Hartley and
C.N.Reid:
33. K. Honda,
H. Masumoto and
S. Kaya:
34. H. Y. Hunsicker:
35. R. L.Rickett,
S. H. Kalin and
J. T. Mackenzie Jr.:
36. F.L.VerSnyder and
R. W. Guard:
Proc.Int. Gong. App. Mech.Delft (1924) 324.
Ph. D.Thesis, Birmingham University, 1966.
"Fat Rolled Products", Vol. m, AIME
Metallurgical Society Conference, Vol. 16 ,
Interscience, New York 1962, p. 47.
J.Inst. Met. Eto (1961/2)473.
Sheet Metal Ind. 42_ (1965) 290.
"X-ray Diffraction by Polycrystalline
Materials", H. S. Peiser, H. P. Roolesby
and A. J. C. Wilson, Eds., Inst. Phys.,
London. p. 462.
Acta. Met. U (1965) 563.
Acta Met. (1970) in press.
9th Sagamore Conference, Syracuse
University Press 1964, p. 259.
Trans. AIME 245 (1969) 1669.
J.Inst. Met. discussion 93 (1964/5)486.
9th Commonwealth Mining and Metallurgy
Conference, vol. 4 (1969) 249.
J.Inst. Met. 97 (1969) 155.
Met. Sci. J. 1 (1967) 65.
Sci. Rep. Tohoku Imp. Univ._17_ (1928)111.
Trans. TMS AIME 245 (1969) 29.
J. Metals J_ (1949) 242.
Trans. ASM 52 (1960)485.
3 5 2
37. H.Takechi,
H. Kato,
T. Sunami and
T. Nakayama:
38. Hung Chi Chao:
39. I. L. Dillamore and
W.T.Roberts:
40. P. Regenet and
H. P. Stwe:
41. J.F.Held:
42. W.T.Roberts:
43. N.J.Grant:
44. B. A. Wilcox and
A. H. Clauer:
45. M. C.Inman,
K. M. Zwilsky and
D. H. Boone:
46. N. S. Stoloff and
I. L. Dillamore:
47. F. Henderson,
F. P. Bullen and
H. L. Wain:
48. K. Lcke and
H. P. Stwe in:
49. J. W. Cahn:
50. J. S. Smart and
A. A. Smith:
51.
52. R. H.Goodenow:
53. D. Webster:
54. F. J. Humphreys and
J. W. Martin:
55. E. Butler:
56. J. J. Iran and
D. J. Lathauj:
Trans.Jap.Inst. Met. 8 (1968)233.
Trans. ASM G O (1967)37.
J.Inst. Met. 92 (1963/4).
Z. f. Metallkde. 54_ (1963)273.
Trans. TMS AIME ^39 (1967)573.
Sheet Met. Ind. 43 (1966) 237.
"The Strengthening of Metals", D. Peckner,
Ed., Reinhold Publishing Corp., New York
1964, p. 188.
Met. Sci. J. 3 (1969) 26.
Trans ASM 57 (1964) 701, ,
Bolton Landing Conference on Ordered
Alloys, Sept. 1969.
J.Inst. Met. 98 (1970) 65.
"Recovery and Recrystallization of MetalsJ'
L. Himmel, Ed. , Interscience, New York
1963, p. 171.
Acta. Met. 1 Q _ (1962) 789.
Trans. AI ME ^52 (1943)103.
ASM Metals Handbook 1961, Vol. 1.
Trans. ASM ^9 (1966)804.
Trans. Met. Soc.,AIME 242^ (1968)640.
Phil. Mag. 16 (1967) 927.
Ph.D. Thesis, Birmingham University 1970.
I. S.I.Spec. Rept. 114, p. 53.
353
57. W. C. Leslie,
J. T. Michalak,
A. S. Keh and
R.J.Sober:
58. G. L.Davis:
59. D. A. Witmer and
G. Krauss:
60. S. Leber:
61. C. G. Dunn and
J. L. Walter in:
62. M.Hillert:
63. W. B. Hutchinson:
I. L. Dillamore and
W. B. Hutchinson:
64. J. Calvet ,and
C, R^non:
65. W. B. Hutchinson,
T. W. Watson and
I. L. Dillamore:
66. T. Gladman:
67. W. E. Duckworth,
R. Phillips and
J. A. Chapman:
68. Hsun Hu:
69. S. Horiuchi,
T. Okubo and
I. Gokyxi:
70. W, Bunk,
K. Lcke and
G. Masing:
71. P. A.Beck and
H. Hu:
72. J. Grewen:
73. J. C. Blade:
74. I. L. DiUamore,
C. J. E. Smith and
T. W. Watson:
Trans ASM 5_8_ (1965) 672.
Metallurgia^8_ (1958) 177.
Trans. ASM 62^ (1969) 447.
Trans. ASM 53 (1960)697.
"Recrystallization,Grain Growth and
Textures", H. Margolin, Ed., ASM, Metals
ParkOhio 1966, p.461.
Acta Met. 13_ (1965) 227. "
Ph.D. Thesis, Birmingham University 1969.
International Conference on the Technology
of Iron and Steel, Tokyo, 1970.
-
Mem.Sci.Rev. Met.5_7_ (1960)345.
J. L S. I,_ 207_ (1969) 1479.
Proc.Roy. Soc. A294 (1966)298.
J.I.S.I. 203 (1965) 1108.
ref. [1], p. 200.
Trans. Jap. Inst. Met. _7_ (1966) 257.
Z. f. MetallKde. 45_ (1954)584.
Trans AIME, J. Metals 4 (1952)83.
Forschungsberichte aus dem Institu fiir
Metallkunde und Metallphysik Clausthal,
No. 2, (1967).
J. Anst. Inst. Met. 12 (1967) 55.
Met. Sc. J. J_ (1967)49.
75.
76.
77.
78.
79.
a t - r - .
i
80.
81 v !i
81.
pla ir i
P. A. Beck:
Br it ish Pa t ent , 1 128
Hsun Hu in:
J.L.Wa lt er a nd
E. F.Koch:
E. A. Lor ia ,
K. Det er t a nd
J.G. Mor r is:
C. Richa r ds a nd
T. W. Wa t son:
J. D. Ba ir da nd
- t M A - T- r v ujemi - h-
82. W. C. Leslie,
J. T.Micha la ka nd
F. W.Aul in:
354
Adv a nces in Physics 3 (1954) 245.
702 a ssigned t o Ar mco St eel Cor p.
"Recov er y a nd Recr yst a lliza t ion of Met a ls",
L. Himmel, Ed., Int er science, New Yor k
1963, p.311.
Act a . Met . IQ (1962) 1059.
Act a Met . 13 (1965) 929,
J. I. S. I. 207 (1969) 582.
J.I.S.I. 204 (1966) 240.

"Ir on a nd it s Dilut e Solid Solut ions",

C. W. Spencer a nd F. E. Wer ner , Eds.,
Int er science, New Yor k 1963, p. 119.