Naturwissenschaften (2006) 93: 374378 DOI 10.

1007/s00114-006-0118-x

ORIGINA L ARTI CLE

F. H. Kaatz

Measuring the order in ordered porous arrays: can bees outperform humans?

Received: 25 October 2005 / Accepted: 11 April 2006 / Published online: 13 May 2006 # Springer-Verlag 2006

Abstract A method that explains how to quantify the amount of order in ordered and highly ordered porous arrays is derived. Ordered arrays from bee honeycomb and several from the general field of nanoscience are compared. Accurate measures of the order in porous arrays are made using the discrete radial distribution function (RDF). Nanoporous anodized aluminum oxide (AAO), hexagonal arrays from functional materials, hexagonal arrays from nanosphere lithography, and square arrays defined by interference lithography (all taken from the literature) are compared to two-dimensional model systems. These arrays have a range of pore diameters from 60 to 180 nm. An order parameter, OP3, is defined to evaluate the total order in a given array such that an ideal network has the value of 1. When we compare RDFs of man-made arrays with that of our honeycomb (pore diameter 5.89 mm), a locally grown version made by Apismellifera without the use of foundation comb, we find OP3=0.399 for the honeycomb and OP3=0.572 for mans best hexagonal array. The nearest neighbor peaks range from 4.65 for the honeycomb to 5.77 for mans best hexagonal array, while the ideal hexagonal array has an average of 5.93 nearest neighbors. Ordered arrays are now becoming quite common in nanostructured science, while bee honeycombs were studied for millennia. This paper describes the first method to quantify the order found in these arrays with a simple yet elegant procedure that provides a precise measurement of the order in one array compared to other arrays.

Introduction
One frequently runs across the phrases ordered and highly ordered arrays used for two-dimensional images of arrays, especially from the general field of nanoscience. A method is described in this text of precisely quantifying the amount of order in these arrays. A discrete radial distribution function is used for each array. From this, an order parameter is derived which provides a number in the interval [0, 1] of the amount of order in the array. In this context, perfect order has the value 1 and disorder near 0. The radial distribution functions (RDF) also provide an estimate of the number of nearest neighbors from the height of the first peak. Bee honeycomb is compared with several recently manufactured arrays from nanotechnology. Honeybees are sometimes known as natures geometers. Bee honeycombs are one of the most regularly ordered structures found in nature (von Frisch 1974; Toth 1964). The honeycomb is where bees store honey. Zenodorus of Sicily (Betts 1921) proved in the second century BC that, of the three regular figures that will completely fill a plane surface (namely, the equilateral triangle, the square, and the regular hexagon), the hexagon has the greatest content for a given circumference. The classical honeycomb conjecture, which asserts that any partition of the plane into regions of equal area has perimeter at least that of the regular hexagonal honeycomb tiling, was recently proven (Hales 2001). These proofs have led to the general misconception that bees manufacture comb so as to enclose the greatest amount of space with a parsimonious use of wax. While a complete understanding of how bees actually make honeycomb is lacking, a model (Pirk et al. 2004) based on liquid equilibrium was recently put forth. In this model, the bees themselves provide the heat necessary to liquefy the wax and form the comb. They also use sensory organs in their necks as a geometric plumb line to help form comb material (von Frisch 1974). This is not an attempt to explain how bees actually make comb, but rather, to quantify how much order is in the structure of the comb through the use of a discrete radial distribution function.

F. H. Kaatz (*) Mesalands Community College, 911 South 10th Street, Tucumcari, NM 88401, USA e-mail: fhkaatz@yahoo.com F. H. Kaatz Owens Community College, P.O. Box 10,000, Toledo, OH 43699, USA

375

Hexagonal and other types of arrays are also of ever increasing interest in the rapidly expanding field of nanoscience. There are now many methods whereby man-made arrays are fabricated, including anodized aluminum oxide (Li et al. 1998a,b), arrays from functional materials (Sun et al. 2004), nanosphere lithographic (NSL) techniques (Haynes and Van Duyne 2001), and square arrays from methods based on lithography (Krishnan et al. 2005). The natural structure of these arrays is hexagonal unless forced otherwise. The hexagonal structure is sometimes a result of close-packing, i.e., a two-dimensional network created by the dense packing of cells. Order in arrays can result in physical properties not attainable from random or disordered arrays. Some examples include field emission from carbon nanotubes (Bocharov and Eletskii 2005; Hu and Huang 2003), magnetic arrays where the magnetism is affected by a demagnetization factor from neighboring magnetic nanowires (Sellmyer et al. 2001), the electromagnetic response of quantum dots (Mikhailov 1996), and semiconductor device characteristics using ordered dopant arrays (Shinada et al. 2005). Carbon nanotubes need to be spaced for field emission properties, so that the effects of screening are minimized. Arrays of nanotubes were fabricated through the use of anodized aluminum oxide (Li et al. 1999) and nanosphere lithographic techniques (Park et al. 2005). In the following section, a method is described for obtaining precise measurements of the order in highly ordered arrays based on using the RDF in two dimensions. The bee honeycomb measured has an order parameter OP3=0.399 and the best hexagonal array has an order parameter OP3=0.572, with everything else at lower levels of order. There is clearly room for improvement, but the general trend is that more recently defined arrays have more order. One might hope for higher ordered arrays from future progress.

2000) in two dimensions. In this approach, the RDF is obtained by: N ri ri X g2 ri n rj where ri bin size 2ri j1 (1) and n(rj) is the two-dimensional sequence of nearest neighbors n(rj)={1, 3, 2, 7, 3, } (see Table 1), and is the density of pores. In our experimentally determined RDFs, we used 19 sites of origin for the hexagonal arrays and 13 for arrays formed from nanosphere lithography and square arrays, or out to n(r3) from the center of the photo. While arrays from nanosphere lithography are technically hexagonal arrays, the nearest neighbor density is different (see Table 1). Typically, scanning electron microscopy (SEM) or atomic force microscopy (AFM) images from the literature are loaded from the PDF document into Adobe Photoshop for editing and contrast enhancement. The quality of contrast can be relative in the SEM or AFM image as originally taken, with some features not resolvable under standard conditions. Pore size and coordinates are determined using Image SXM (Barrett 2006), which uses a best-fitting ellipse to calculate areas and centers of particles. Excel macros (de Levie 2004) were used to generate the RDFs. The relationship between the pores and an RDF is explained in Fig. 1, which shows a hexagonal pore network. The largely drawn color circles represent the RDFs according to their spacing n(ri). The interstitial sites are the black circles around the pores. For nanosphere lithography, the interstitial sites are the open circles, or pores, so they have a different RDF. Computer-generated hexagonal and square arrays were used to calculate the RDFs shown in Fig. 2a,b. The number of nearest neighbors is given by the height of the first peak and is shown in Table 2 under N.N. The hexagonal array has an average of 5.93 and the square array has an average of 4.06. Table 2 also lists the height of the first peaks for the other arrays studied.

Materials and methods

The porous structure may be modeled by a discrete radial distribution (RDF) approach (Behnke and Sands

Table 1 Nearest-neighbor sequences for the hexagonal arrays, arrays from nanosphere lithography, and square arrays NN Hex n(ri) Hex NSL n(ri) NSL Square n(ri) Square 1 1 6 1 3 1 4 2 3 1.74 6 3 6 2 1.42 4 3 2 6 2 3 2 4 4 7 2.64 12 7 6 5 2.24 8 5 3 6 3 6 8 2.84 4 6 12 3.46 6 12 6 3 4 7 13 3.62 12 13 6 10 3.16 8 8 4 6 4 3 13 3.62 8 9

4 4

The density of pore variation with the sequences n(ri) are also listed

376

Fig. 1 Plot showing the origin of a hexagonal RDF with n(ri) up to i=5. The different colors represent n(ri) values. Pores are the solid black circles

Results
The honeycomb used for this study was purchased locally from an apiary and made without the use of a foundation comb. Therefore, it represents the capability of bees when using only a rectangular support frame. It has a pore diameter of 5.89 mm and a spacing of 6.66 mm. This was the only comb used in the study, so the results may not be representative for all honeycombs. There were no defects in the honeycomb structure, although dislocations and defects are known (Hepburn and Whiffler 1991) to exist in bee honeycomb. The experimentally determined RDF is shown in Fig. 3a. A disorder parameter is defined based on the data, as follows:
N X X G2 ri g2;d ri ri 0:06 g2;d ri i0:1 inri

Fig. 2 a RDF for a computer-generated hexagonal array. b RDF for a computer-generated square array

(2)

Here g2,d is the experimentally determined RDF. G2 is calculated from the RDF as N=4.1. The subtracted term
Table 2 Statistics for the arrays examined in this paper Average pore d Honeycomb AAO (Li et al. 1998a) AAO (Li et al. 1998b) Polystyrene (Sun et al. 2004) NSL (Hulteen et al. 1999) Square (Krishnan et al. 2005) Ideal hexagon Ideal square 5.89 mm 78.19 nm 57.81 nm 178.8 nm 110.1 nm 72.60 nm

represents the contribution in an ideal location and so is subtracted from the disorder term. The number 0.02 represents the bin size in the data. The number 0.06 is used since simulations show that it is difficult to generate an RDF of an ordered hexagonal array with G2=0 in a computer model unless one takes bins above and below n

Bond 6.66 mm 100.0 nm 94.30 nm 217.6 nm 286.8 nm 185.3 nm

 G2(r4:1 )

OP3 0.399 0.367 0.202 0.572 0.441 0.526 1.000 1.000

N.N. 4.65 3.97 4.95 5.77 1.91 1.87 5.93 4.06

1.106 1.223 1.438 0.878 0.369 0.529 0.000 0.000

A comparison of the various OP3 values can be made here

377

Fig. 3 a Experimentally determined RDF for the bee honeycomb. OP3=0.399 for the array. b RDF for an AAO array from Li et al. (1998a). Data were taken from their Fig. 1b. c RDF for an AAO array from Li et al. (1998b). Data were taken from their Fig. 12b. d RDF for an array made from the functional materials in Sun et al.

(2004). Data were taken from their Fig. 2f. e RDF for an array made from nanosphere lithography of Hulteen et al. 1999. Data were taken from their Fig. 1c. f RDF for a square array made by lithography from Krishnan et al. (2005). Data were taken from their Fig. 5a

(ri). This results in a rather sensitive order parameter that can be written as G2 ri OP3 1 N : (3) P g2;d ri ri
i0:1

Recall that OP3 has the value of 1 in an ideal network since there is no disorder contribution. This honeycomb has OP3=0.399. Table 2 lists the results for all the arrays studied in this paper. The Fig. 3 legend lists the source figures for the RDF data. The electrochemical anodization of Al creates anodized Al oxide (AAO) with arrays of pores. These nanopores can be used as templates to grow ordered arrays of various nanowire materials. The pore size and spacing is determined primarily by the selection of acid and applied voltages used (Li et al. 1998a). While defects can occur (Li et al. 1998b), highly ordered arrays with no obvious fault structures are used for the purposes of this study. The calculated RDF from an AAO array in Li et al. 1998a is

shown in Fig. 3b. It was formed in 0.3 M oxalic acid at 40 V. It has a pore diameter of 78.19 nm and a spacing of 100.0 nm. The OP3 is 0.367. A second AAO array is from an AFM image in Li et al. 1998b. The array was formed in 3% oxalic acid at 40 V at 15C. It has a pore diameter of 57.81 nm and a spacing of 94.30 nm. The RDF is shown in Fig. 3c and appears more disordered than Fig. 3b, but this may be due to the poor contrast in the AFM image leading to difficulty in assigning coordinates. The OP3 for this RDF is 0.202, confirming the looks of the RDF. Another hexagonal array can be made from functional materials (Sun et al. 2004). This involves a solutiondipping template strategy where polystyrene spheres form a hexagonal close packed array, which are then removed, and the precursor solution left behind is dried forming a porous array. A field emission SEM topographic image was used to provide high contrast. The array was made from a 0.8-M precursor solution dried at 80C and calcined at 400C. The pore diameter is 178.8 nm and the spacing is 217.6 nm. The RDF shown in Fig. 3d appears fairly ordered by the spacing of the peaks at n(ri). All the peaks in

378

the RDF line up with n(ri) and are well-separated. The OP3 value of 0.572 confirms the appearance of the RDF. The number of nearest neighbors is 5.77, the highest value in the arrays studied so far. This is the most ordered array measured in this study and is better than the bee honeycomb. Another form of hexagonal arrays can be made by nanosphere lithography (Haynes and Van Duyne 2001). In this method, a dense closed packed array of nanospheres is formed and the interstitial sites are used as the porous network. Nanosphere lithography can be done by singlelayer or double-layer techniques (Haynes and Van Duyne 2001). In the double-layer method, a standard hexagonal array is produced. In the single-layer method, since the interstitial sites are used, a different density of sites is obtained (see Table 1) with the first nearest neighbor threefold coordinated. Since the pores are used to define a pattern for the deposition of material, we used the array of silver nanoparticles (Hulteen et al. 1999) to calculate the RDF shown in Fig. 3e. The silver particles shown in their AFM image have a size of 110.1 nm and a spacing of 286.8 nm. The RDF looks good and has OP3=0.441. The average number of nearest neighbors is 1.91 and would be near 3.00 in an ideal array. Square arrays can be defined by lithographic techniques (Krishnan et al. 2005). They used a double-interference lithography technique on spun photoresist. Square arrays have both a different n(ri) and density sequence (see Table 1). Krishnan et al. 2005 was used to calculate the RDF shown in Fig. 3f. It has a pore size of 72.60 nm and a spacing of 185.3 nm (see Table 2). It appears fairly well ordered with an OP3=0.526, a little below that of the best hexagonal array. The average number of nearest neighbors is 1.87 as compared to the value 4.06 for the ideal array. In summary, the use of ordered arrays is becoming wellestablished in nanostructured science. While bee honeycombs have been studied for millennia, this paper describes the first method to quantify the order found in these arrays. It is worthwhile to mention that only one comb was used in this study and may not represent the best that honeybees can do. At the same time, bees are working with quite modest equipment, jaws, feet and sensory organs, and the fact that they can produce such well-ordered structures is amazing. The honeycomb is a macroscopic structure and is well-ordered in comparison with man-made arrays. In contrast, man uses all the technological advances possible, and produces some ordered structures as well. In the comparison of man with nature, the bees are not far behind.
Acknowledgements This research is indebted to Steve Barrett and his efforts with Image SXM for Mac OS X, which was used to determine the areas and coordinates for all the arrays. F.H. Kaatz would like to acknowledge the excellent library services at Owens Community College, the University of Toledo, and Sandia National Laboratories. The honeycomb used in this study was purchased from Sawyers Apiaries, Swanton, OH 43558, USA. F.H. Kaatz thanks P. Kaatz, T. Egami, and M.P. Siegal for a critical review of the manuscript.

References
Barrett S (2006) The web site of Image SXM. http://www.liv.ac. uk/sdb/ImageSXM/. December 2005 Behnke JF, Sands T (2000) Bimodal spatial distribution of pores in anodically oxidized aluminum thin films. J Appl Phys 88 (11):6875 Betts AD (1921) The structure of comb-I. Bee World 3:37 Bocharov GS, Eletskii AV (2005) Effect of screening on the emissivity of field electron emitters based on carbon nanotubes. Tech Phys 50(7):944 de Levie Robert (2004) Advanced excel for scientific data analysis. Oxford University Press, London, UK Hales TC (2001) The honeycomb conjecture. Discrete Comput Geom 25:1 Haynes CL, Van Duyne RP (2001) Nanosphere lithography: a versatile nanofabrication tool for studies of size-dependent nanoparticle optics. J Phys Chem B 105:5599 Hepburn HR, Whiffler LA (1991) Construction defects define pattern and method in comb building by honeybees. Apidologie 22:381 Hu Y, Huang CH (2003) Computer simulation of the field emission properties of multiwalled carbon nanotubes for flat panel displays. J Vac Sci Technol B 21(4):1648 Hulteen JC, Treichel DA, Smith MT, Duval ML, Jensen TR, Van Duyne RP (1999) Nanosphere lithography: size-tunable silver nanoparticle and surface cluster arrays. J Phys Chem B 103:3854 Krishnan R, Nguyen HQ, Thompson CV, Choi WK, Foo YL (2005) Wafer-level ordered arrays of aligned carbon nanotubes with controlled size and spacing on silicon. Nanotechnology 16:841 Li AP, Mller F, Birner A, Nielsch K, Gsele U (1998a) Hexagonal pore arrays with a 50420 nm interpore distance formed by self-organization in anodic alumina. J Appl Phys 84(11):6023 Li F, Zhang L, Metzger RM (1998b) On the growth of highly ordered pores in anodized aluminum oxide. Chem Mater 10:2470 Li J, Papadopoulos C, Xu JM, Moskovits M (1999) Highly-ordered carbon nanotube arrays for electronics applications. Appl Phys Lett 75:367 Mikhailov SA (1996) Radiative decay of collective excitations in an array of quantum dots. Phys Rev B 54(15):10335 Park KH et al (2005) Advanced nanosphere lithography for the areal-density variation of periodic arrays of vertically aligned carbon nanofibers. J Appl Phys 95:024311 Pirk CWW, Hepburn HR, Radloff SE, Tautz J (2004) Honeybee combs: construction through a liquid equilibrium process? Naturwissenschaften 91:350 Sellmyer DJ, Zheng M, Skomski R (2001) Magnetism of Fe, Co and Ni nanowires in self-assembled arrays. J Phys Condens Matter 13:R433 Shinada T, Okamoto S, Kobayashi T, Ohdomari I (2005) Enhancing semiconductor device performance using ordered dopant arrays. Nature 437:1128 Sun F, Cai W, Li Y, Cao B, Lei Y, Zhang L (2004) Morphologycontrolled growth of large area two-dimensional ordered pore arrays. Adv Funct Mater 14(3):283 Toth LF (1964) What the bees know and what they do not know. Bull Am Math Soc 70:469 von Frisch Karl (1974) Animal architecture. Harcourt Brace Jovanovich, New York