Crystallinity in polymers: an historical view Geoffrey Allen" We still have much to learn about the morphologies of natural and synthetic polymers crystallized in bulk. Nevertheless, conventional X-ray and thermodynamic studies of crystallinity in polymeric materials over the past 60 years have given us a clear picture of the concept of crystallinity and the molecular features which promote crystallinity in these substances. Apart from issues of chain connectivity, the science of conventional molecules proved adequate for the task! A few liquids of low molecular weight do not readily crystallize. Apparently these substances are composed of molecules that have some difficulty in arranging themselves into a crystalline array. The molecules usually have bulky substituent groups which offer steric resistance to folding into a compact form suitable for efficient packing in a crystal lattice. Such liquids, on cooling, become very viscous and ultimately form glasses at temperatures around — 100°C. In this solid state they show no sharp X-ray diffraction rings indicative of randomly orientated crystals. In fact, the diffuse X-ray diffraction pattern observed is very similar to that of the corresponding liquid, showing that the glassy solid has no long-range molecular order. Some of these liquids can be crystallized by holding a sample for many hours at a specific temperature, some 20°C above the glass formation temperature, until the substance suddenly begins to crystallize. On warming, the melting point of the crystals is well defined and located about 40—60°C above the temperature of glass formation. Given that some relatively simple small molecules find difficulty in forming crystals, what are we to expect of the behaviour of polymer molecules? A typical simple polymer molecule has some 10000 repeat units covalently linked into a long chain, for example: polyethylene- -(CH 2 -CH 2 HCH 2 -CH 2 HCH 2 -CH 2 HCH 2 -CH 2 )- polyethyleneoxide -(CH 2 -CH 2 -OHCH 2 -CH 2 -O)- and the chains may not be linear, as is often the case in polyethylene: for example —CH2— CH2— CH2— CH—CH2—CH2—CH2—CH2—CH—CH2—CH2— CH 2 CH2 CH 2 CH2 CH3 CH2 I I CH2 "Kobe Steel Ltd., Alton House, 174/177 High Holborn, London WC1V 7AA, UK. 1062-7987/93/020197-10 $10.00 © 1993 by John Wiley & Sons, Ltd. CH3 EUROPEAN REVIEW, Vol. 1, No. 2, 197-206 (1993) Geoffrey Allen 198 since molecular branches can be formed during the polymerization process. Even the linear chains of larger monomer units will have sizeable side groups: for example polystyrene CeHs C9H5 C6H5 I I I -CH 2 -CH-CH 2 -CH-CH 2 -CHCH3 CH3 CH3 I I I polypropylene -CH 2 -CH-CH 2 -CH-CH 2 -CHR R R I polypeptide I I -CH-C-NH-CH-C-NH-CH-C-NH- II O and R may represent several different side groups in the same polypeptide chain. Now, bear in mind that each molecule can curl into a randomly arranged coil, by virtue of rotations about the bonds in the main chain. The polymer melt is composed of these randomly coiled chains and each chain is wriggling rapidly through millions of conformations. Furthermore, the long chains are entangled. To fit into a crystal lattice, a chain must uncoil a well-defined section of say 10—20 monomer units and align that section with a similar one from another part of the same chain or of a neighbouring chain. If a sufficient number of segments can line up to form a stable nucleus then crystal growth can occur. A priori, it would seem to be an event of low probability even in polymers composed of linear long chains of the simplest chemical structure! In fact, crystallization offlexiblechain molecules of sufficient structural regularity is widely observed in naturally occurring polymer systems such as polypeptides, carbohydrates and natural rubber, as well as in those of synthetic origin. The first recorded observation was made by John Gough1 in a letter to the Manchester Literary and Philosophical Society in 1805: ' If a thong of Caoutchouc be stretched, in water warmer than itself, it retains its elasticity unimpaired. If the experiment be made in water cooler than itself, it loses part of its retractile power, being unable to recover its former figure; but let the thong be placed in hot water, while it remains extended for want of spring, and the heat will immediately make it contract briskly'. Gough did not realize he was witnessing the partial II O II O crystallization of the rubber strip followed by its melting in the warmer water. The systematic study of crystalline polymers parallels the development of polymer science from its origins in the early 1920s. Following the discovery that crystalline substances diffract X-rays and that the diffraction pattern can be resolved to give unit cells and hence crystal structures, Michael Polanyi (in his days as a physical chemist, before he became a professional philosopher) obtained diffraction patterns from a bundle of parallel ramie fibres. Polanyi2 concluded that the pattern of spots corresponded to crystallites of cellulose orientated with one axis parallel to the fibre axis. The empirical structure of cellulose was known to be C 6 H 10 O 5 ; Polanyi found that the unit cell was orthorhombic, measuring 7.9 x 8.4 x 10.2 A (1 A = 10~ lo m). He concluded that the cell contained either two cyclic disaccharides or two disaccharide units of polysaccharide chains. The concept of the existence of polymer chains was not at the time accepted by organic chemists and the second option was not well received. Polanyi3 wrote in his memoirs in 1962: 'Unfortunately I lacked the chemical sense for eliminating the first alternative . . . I was gleefully witnessing the chemists at cross purposes with conceptual reform . . . I should have been better occupied in definitely establishing the chain structure.' In the mid 1920s, Katz4 observed that an amorphous natural rubber strip when extended six-fold produces a sharp X-ray diffraction pattern ' characteristic of a body containing crystallites' in addition to the amorphous halo. Katz estimated Crystallinity in polymers that the same cell could only accommodate three isoprene units. In the same year, 1925, Staudinger and his colleagues5 published studies of 'oligomers and high polymers' (that is polymer chains so long that the molecular weight could not be determined by the methods then available) of formaldehyde, • • -O-CH 2 -O-CH 2 -O-CH 2 -O-CH 2 -- • • X-ray diffraction from the oligomers showed reflections characteristic of the length of the extended molecules but in addition there were reflections due to a subcell that enclosed sections of the chain. As the length of the oligomers increased, the subcell reflections became more pronounced and it was possible to retain by extrapolation the diffraction pattern of the high polymer. By 1928, X-ray studies by K. H. Meyer and Herman Mark6 demonstrated that the crystal structures of cellulose, silk fibroin, chitin and natural rubber were due to the packing of long-chain molecules. The concept of a long-chain molecule was fully accepted. The paper on silk fibroin described for the first time a polypeptide chain in the fully extended conformation—now known as the ^-structure. In the next decade Astbury7 found that there were two crystalline forms, a and /?. Later, from crystallographic measurements on aminoacids and peptides Pauling and his school argued that the /?-form was composed of helical molecules with 3.6 amino acids per turn and a pitch of 1.5 A. Pauling8 deduced this structure from molecular models as well as crystallographic data. Later still, this combination was to be crucial in the work of Crick and Watson9 on the structure of DNA. The X-ray diffraction method is now a cornerstone of molecular biology. Many common synthetic polymers such as polystyrene and polymethylmethacrylate (Perspex) did not crystallize, they were only available as glassy (amorphous) substances in the solid state. Furthermore, the various known forms of polyethylene crystallized only to a degree of about 40%, that is, they remained a mixture of crystalline and amorphous fractions. However, in the early 1950s Ziegler10 reported a new family of catalysts which produced highly crystalline polyethylene of high molecular weight. Natta11 reported similar catalysts which produced crystalline versions of polymers such as polystyrene, polymethylmethacrylate and polypropylene of high molecular weight. 199 These new materials were effectively the last links in establishing the molecular structure criteria for crystallizable polymers. The new forms of polyethylene contained linear chains or chains with very few branch points. Natta showed from crystallographic data that his polymers, which had pendant side groups, had structural regularity, i.e. stereoregularity. The so-called isotactic polymer chain of the R I polymer -CH-CH 2 - (e.g., polystyrene or polypropylene) has all the pendant groups on the same side of the plane of the carbon atoms as shown in Figure l(a). In the syndiotactic form, the pendant groups alternate strictly on either side of the chain as in Figure l(b). The atactic form has the pendant groups placed at random on either side of the chain as in Figure l(c). The Ziegler/Natta catalysts produce long runs of either iso- or syndiotactic chain sequences with much shorter sections of atactic sequences. The isotactic chains are regular and can fit into a crystal lattice; so can syndiotactic chains, although the resulting structure will be different. Polymers composed of long atactic sequences do not crystallize because they do not possess sufficient steric regularity to produce stable nuclei from which crystals can grow or allow the chain to fit into a crystal lattice. These forms of polymer exist only as glasses in the solid state. Thus, the principles governing the formation of R R R R R |al Isoiuclic H VH H VH H V V (f V H V H H (hi SMHIIOUI |c| ALICUC V R Figure 1. Schematic representation of isotactic, syndiotactic and atactic vinyl polymers. Geoffrey Allen 200 crystals in polymeric substances include the following: (i) The chemical repeat unit plays a role similar to discrete molecules in crystals of a normal organic compound. (ii) The molecular chains must have long sequences of stereoregularity. (iii) Linear polymer molecules facilitate crystallization. (iv) The unit cell contains only a small number of repeat units. (v) More than one chain may pass through a unit cell. On this basis, crystallization in all known synthetic or natural polymers can be rationalized in terms of the structure of the molecular chains. Apart from the fact that a polymer molecule passes through a large number of consecutive unit cells, the crystal lattice is analogous to that of a simple organic compound. Crystallization and melting Conceptual difficulties with the phenomenon of crystallization in polymeric materials arise from the fact that with the exception of polymethylene oxide -(CH 2 -O) n - and certain proteins, macroscopic single crystals are very rarely encountered. Polymers do not crystallize completely; the submicroscopic crystallites are embedded in the remaining amorphous matrix. X-ray diffraction again proved useful in establishing these facts. Pioneering studies in the early 1930s showed that when a polymer is cooled slowly from the melt (i.e., the rubbery state) sharp diffraction rings are observed, showing that the sample contains randomly orientated crystals. When the sample is held in the temperature range where the sharp rings increase in intensity, crystallization eventually ceases, as judged from the X-ray intensities, and there is a residual amorphous halo demonstrating that amorphous material remains. If cooling is continued to very low temperatures there is no change in the pattern, despite the fact that the sample will have changed from a rough leathery texture to a hard solid. On warming, the crystalline diffraction pattern disappears over a range of temperature, but there is a well-defined temperature at which the last traces disappear. This melting temperature limit is some 20°C above the temperature of the maximum rate of crystallization. Rapid cooling often results in no crystallization despite transformation of rubber to a hard brittle glassy material. Since there must be volume changes accompanying crystallization of the sample, dilatometric studies were made to parallel the X-ray observations. The pioneering results obtained by Bekkedahl12 in 1934 on natural rubber showed a volume-temperature curve obtained at slow rates of cooling with crystallization and a contraction in volume occurring near 0°C. Warming from liquid air temperature shows a second-order change in the volume—temperature relation at —72°C, followed by a first-order melting process in the range 6 to 16°C which is accompanied by an expansion of 2.7%. This agrees with the X-ray determination of the melting temperature of the crystallites. By contrast, when the sample is cooled rapidly to liquid air temperature it remains amorphous. On warming, only the second-order transition at — 72°C is observed as the sample changes from glass to rubber. Thus, the melting point of natural rubber is ~ 16°C and its glass-transition temperature — 72°C. A combination of X-ray and volumetric data shows that unstretched natural rubber, taken to its limit of crystallization, still contains some 70% of amorphous material. Further work by Wood and Bekkedahl13 revealed behaviour which is known now to be typical of the crystallization of many polymer systems. First, the rate of crystallization showed a maximum at about 0°C and the rate decreased on further cooling—behaviour which is characteristic of a system controlled by nucleation. Second, the temperature range of melting was found to be a function of the temperature at which crystallization occurred. As the temperature of crystallization is lowered, the temperature of melting of the crystals produced also falls but the temperature range over which melting is observed becomes greater, i.e., the melting is more diffuse. Under all conditions, melting only begins some 5-10°C above the crystallization temperature. To obtain the highest melting point, crystallization must be carried out above the temperature at which the optimum rate of crystallization occurs. Since this is a region where crystallization is slow, this effect has led to Crystallinity in polymers various melting points being reported for the same material. Taken together, these observations of melting point ranges and incomplete crystallization mean that when crystallization ceases, an equilibrium state is not reached between the amorphous phase and the crystalline phase. There will be a distribution of crystallites of different degrees of imperfection and also of different sizes. Because of surface energy considerations, the smaller crystals and more imperfect crystals will have lower melting points. Further, there may be entanglements and other constraints (for example, high viscosity) in the amorphous phase which prevents reorganization and recrystallization to produce an array of more stable crystallites. Polyethylene has provided further insights into the nature of crystallization in polymers. Since normal paraffins, CnH2n + 2, are well-defined chemicals, certainly up to C 94 H 190 , and within the series their melting behaviour is very similar to that of normal organic substances, it is possible to estimate by extrapolation the melting point of an infinitely long chain. n Tm°C 30 44 62.1 86.2 94 114.2 CO 140 The result is that polyethylene of high molecular weight should melt at about 140°C. This result was known at a time when polyethylene was manufactured from ethylene by a high-pressure process using a free radical catalyst. Unfortunately, the melting point of the product was only 115°C, well below the estimate for a very long linear chain. Furthermore, the polyethylene showed a limited degree of crystallizability. From X-ray and spectroscopic studies it was concluded that polyethylene molecules made by this process are highly branched. The whole issue was resolved when Ziegler's new catalyst, which synthesized chains with a very much lower degree of branching, yielded highly crystalline products, initially with melting temperatures of about 135°C and later in the region of 140°C. Now a strict comparison could be made of the crystallization and melting behaviour of a polymer 201 with its discrete short-chain homologues, in this case the n-paraffins. Figure 2(a) shows volumetemperature curves for C 44 H 90 and C 94 H 190 , and Figure 2(b), curves for an unfractionated linear polyethylene sample of number-average molecular weight 32000. The dilatometric results give evidence that it is possible to obtain melting curves in polymers that are comparable with those of single substances of low molecular weight. The temperature at which the last traces of crystallinity disappear is clearly defined for each sample. The fact that the polymer sample contains a distribution of chain lengths as well as some branched molecules and a distribution of crystal imperfections does broaden the melting range. Comparison of the results for the polymer sample which was cooled slowly from the melt, with the same sample carefully crystallized at 130°C for 40 days, shows that the latter sample attained a higher degree of crystallinity. Both samples gave Tm = 140°C. Thus, the differences in crystallization and melting between substances of low molecular weight and polymers is one of degree rather than kind.14 Fusion in polymers is also a first-order thermodynamic process. However, the main difference is that polymers crystallize only in part. This results in the observation of a glass transition, with second-order character, in the residual rubber phase at a temperature much lower than Tm. Fibre formation is one of the most important features of crystallizable polymers. The early X-ray workers on cellulose noted that the crystallites were orientated with their long axis parallel to the fibre axis. In fact, the drawing of the fibre to align the polymer molecules along the fibre axis and hence promote the growth of orientated crystallites is an essential feature of all synthetic fibre processes, nylon, terylene, polypropylene, rayon, etc. There are many natural examples, including cotton, wool and silk. However, although these systems attain quite high degrees of crystallinity and because of crystallite orientation the bulk properties are anisotropic, exactly the same concepts of crystallization apply. Crystallization and morphology Crystalline polymers have the capacity to present a complex hierarchy of morphologies extending from Geoffrey Allen 202 T(°C) 106 108 110 112 ^< ii6 118 11 <^- 14 •• 18 C 12 C H 44 90 94 H 190 J L 16 1 o <u 10 CO 8• J 6 (Q) 85 •o o • 14 • o 12 86 87 88 89 90 i I i i 1.3 1.25 o o d> E 1.2 1.15 o Specil o 1.1 1.05 (b) 0.95 -40-20 0 20 40 60 80 100 120 140160 180 T (°C) Figure 2. (a) Fusion of n-hydrocarbons. Plot of dilatometric scale reading against temperature for C 44 H 90 and C 94 H 190 . (b) Specific volume-temperature relations for an unfractionated linear polyethylene sample. Slowly cooled from melt o; isothermally crystallized at 130°C for 40 days, then cooled to room temperature •. Crystallinity in polymers 203 unit cell dimensions, to crystallites in fibres and isotropic substances of the order of 100-1000 A, on to polycrystalline aggregates extending up to a millimetre or so. In rare instances, macroscopic single crystals can be obtained. At the molecular level an individual chain may pass once or even twice through a unit cell. The observation of limited degrees of crystallinity attained in bulkcrystallized materials and estimates of crystallite size suggests that a long molecule may emerge from a crystallite, have a coiled segment embedded in an amorphous domain and then return to the same crystallite or enter a different one. Thus, for many years, the accepted morphological structure of an unorientated, crystallized polymer was the 'fringed micelle' model. 15 The long molecules meander through a two-phase structure in which the randomly arranged crystallites are bundles of orientated segments of polymer chain some 100-1000 A long and the amorphous domains are tangles of coiled chain. The model, largely due to Bunn, is shown in Figure 3. It incorporates the possibility that one molecule can traverse several crystallites and the intervening amorphous domains. Fibre formation is envisaged as a straightening out of the chain and a consequent growth of crystals orientated in the direction of draw. In 1945, Bunn 16 reported a larger scale of ordering, following a study of polyethylene in a polarizing optical microscope. These entities, called spherulites, are micrometers in size, consisting of approximately spherical arrays of equivalent radiating crystalline units. It was not obvious how (b) Figure 4- Polyethylene lamellae crystallized from solution: (a) lozenge-shaped; (b) truncated lozenge-shaped. (Courtesy of Dr. D. C. Bassett.) Figure 3. Schematic representation of the fringedmicelle model (after Bunn). the small crystallites of the fringed micelle model could give rise to these structures since the spherulites were orders of magnitude larger. Even more curious and more difficult to accommodate in the model was the observation deduced from the optical birefringence that the polyethylene chain axis was tangential rather than radial, i.e., perpendicular to the direction of the radiating fibrils. 204 Geoffrey Allen This difficulty remained until 1957 when studies were made of the new linear (i.e., unbranched) polyethylene crystallized as individual crystals from dilute solution. Examination in a transmission electron microscope showed the crystals to be lozenge-shaped, about 100 A in thickness and several micrometres in length. Keller17 demonstrated that within the crystal the molecular chain axis was roughly perpendicular to the plane of the lamella. Since the molecules were, on average, about 1 urn long, Keller deduced that they could be accommodated in a 100 A thick crystal only by the chain folding back on itself repeatedly at the crystal surfaces. These results triggered a major international research effort on individual lamellae and aggregates of lamellae grown from linear polymers in solution. Much work has been done on individual polyethylene lamellae to validate the remarkable finding by Keller. Electron diffraction has confirmed that the c-axis of the unit cell is close to perpendicular to the plane of the crystal and thus so is the chain axis. It is accepted that each polymer molecule folds back upon itself many times at the upper and lower crystal surfaces (the 'fold surfaces') rather like a Chinese cracker. However, there is still controversy over the exact nature of the fold. The Keller school argues that most of the folds are tight, having four CH2 groups per turn and molecular models show that this is quite possible. The Flory school, mainly devoted to thermodynamic studies, favours a model in which a molecular chain at a surface may leave the crystal and not necessarily return to an adjacent site via a tight fold—a switchboard model18 in which some folds contain a large number of carbon atoms to re-enter the crystal at a more distant site. Nevertheless in both models the polymer crystal can be considered to have two components with different properties—a well-ordered crystalline interior plus lamellar surfaces which incorporate the fold planes. We know now that many polymer chains adopt folded-chain conformations in lamellar structures. Keller himself19 has recently recorded the history of his discovery and the stormy debates which followed. The form of single-crystal lamellae obtained from a very dilute solution of polyethylene in xylene at lower temperatures is the simple lozenge shape with {110} growth faces. At higher temperatures these crystals develop {100} faces which give the lamellae the appearance of truncated lozenges. Both forms are shown in Figure 4. Because the monolayer crystal has the c-axis of the unit cell perpendicular to the plane of the crystal and the crystal is bounded by growth faces of low crystallographic index, the crystal habit is orthorhombic, reflecting the symmetry of the unit cell. In the same way, hexagonal monolayer crystals are produced by polyoxymethylene and isotactic polystyrene because they have hexagonal unit cells. Detailed studies show that all these ' single crystals' are actually multiple twins because the crystal habit referred to above exists in the lamellae in a number of crystallographic regions, each associated with a particular growth face. Single lamellae can only be obtained under carefully regulated conditions. More generally, crystallization from solution and sometimes from the melt produces multilayered lamellar aggregates. These multilayered aggregates result from screw dislocation imperfections in the crystal growth caused by chain folds which are staggered and which give rise to the proliferation of layers. Where such defects are limited by crystallization conditions and slow growth occurs in solutions of moderate concentration or where growth occurs in melts which are only slightly supercooled, well-ordered multilayered crystals can be obtained. More usually in polymers, screw dislocation defects and the splaying of lamellae generate sheaf-like structures and ultimately the spherically symmetrical spherulites which are the building blocks of the morphology of polymers crystallized in bulk. Spherulites and the arrangement of lamellae in spherulites continue to be studied extensively to elucidate the relation between their structure and the macroscopic properties of the material. Before proceeding with crystallization in the polymer melt we must note that spherulites are not found exclusively in polymeric substances; they are observed in many organic and inorganic materials where crystallization occurs from a viscous melt. However, the formation of spherulites is still not well understood. In polymer systems such understanding is made more difficult by the fact that the direct transmission electron microscopic observations made on individual lamellae grown from solution cannot be applied directly to lamellae grown from the melt. Specimens of melt-crystallized materials have to be prepared by Crystallinity in polymers 205 Direction of growth Figure 6. A model lamellar fibril in a polyethylene spherulite (after Takayanagi). The a, b, c axes depict the orientation of the unit cell along the spiralling fibril. Figure 5. Sheaf-like aggregates crystallized from the melt in a blend of linear and branched polyethylene at 125°C. (Courtesy Dr. D. C. Bassett.) crystallizing thin films or by sectioning bulkcrystallized samples. It is not certain that the forms produced by crystallization in a very thin film are representative of those produced in bulk. Neither is it certain that sectioning does not deform the structures. Lamellar structures in bulk often have to be deduced with the aid of two less direct methods. Staining of samples using chlorosulphonic acid or osmium tetroxide is one technique which was developed originally for biological specimens. The second technique involves a potassium permanganate/sulphuric acid etching reagent which selectively removes disordered material from the surface of a specimen so that the ordered structural features are revealed. Bassett and his colleagues20 have provided an insight into the origins of spherulites by crystallizing polymers under conditions where the nucleation density is very high and the volume available for growth constrained so that mature spherulites do not develop. At high crystallization temperatures, sheaf-like structures can be seen. Figure 5 shows the lamellar structure of a sheaf-like 'axialite' grown from a polyethylene melt. When these structures grow in an unconfined sample it is quite possible that the lamellae will branch through the mechanism of screw dislocation and then splay apart. Unfortunately it is difficult to quench polyethylene samples and freeze in structures developed through these stages of growth. However, in isotactic polystyrene it is observed that lamellae branch through screw dislocations and this generates spiral structures containing several layers. Successive layers then splay apart. Since screw dislocations develop on all growth faces this results in three-dimensional growth. If these mechanisms hold for the growth of the sheaf-like polyethylene structures, eventually all solid angles will be filled and a spherulite form produced. It is widely believed that axialite sheaves differ in degree only from spherulites. Spherulite structures observed in several different polymers have common features. By examining the edge of a spherulite it can be shown that the thicker lamellae develop first and, in doing so, enclose columns of melt which later crystallize isothermally near the centre of the spherulite or when the sample is quenched, near the tips of the growing lamellae. The forms of the overall structures observed are dictated by the early-formed, thicker or so-called dominant lamellae. The lamellae that form later are thinner and in-fill the already established dominant spherulite frame at interstitial sites. They are termed subsidiary lamellae. Using solvent extraction, it has been shown that in polyethylene samples the longest, most linear molecules crystallize first. Thus, they tend to be located in the dominant lamellae. Shorter molecules crystallize isothermally within the frame of dominant lamellae and shorter chains still, in regions which do not crystallize except on quenching. Chains which do not crystallize form a tenuous amorphous matrix. This molecular fractionation influences macroscopic properties. The regions in which material of lower molecular weight is concentrated provide relatively easy 206 Geoffrey Allen routes through which cracks propagate. This is one reason why spherulite size and structure are considered to be important in the control of bulk properties of crystalline polymers. In all structures, the observation made over 40 years ago by Bunn, that the polymer chain axes lie tangential and not radial to the spherulite structure, is explained by Keller's folded chain. The chains are folded within the spiralling lamellae with the chain axis perpendicular to the growing faces, as shown schematically in Figure 6. 14. L. Mandelkern (1989) In Comprehensive Polymer Science, edited by G. Allen and J. H. Bevington. (Pergamon Press) 2,, 363. 15. C.W. Bunn (1953) Fibres from Synthetic Polymers, edited by R. Hill. (Amsterdam: Elsevier), p. 240. 16. C.W. Bunn and T.C. Alcock (1945) Trans Faraday Soc 41, 317. 17. A. Keller (1957) Philos Mag [8] 2, 1171; (1958) Discuss Faraday Soc 25, 114. 18. P. J. Flory (1989) In Comprehensive Pofymer Science, edited by G. Allen and J. H. Bevington (Oxford: Pergamon Press). REFERENCES 1. J. Gough (1805) Mem Lit Phil Soc Manchester 2, 288. 2. M. Polanyi (1921) Naturwiss 9, 228. 3. M. Polanyi (1962) In 50 Years of X-ray Diffraction edited by P.P. Ewald. Vosthoek. 4. J.R. Katz (1925) Kolloid Z. 37, 19. 19. A. Keller (1991) In Sir Charles Frank: An Eightieth Birthday Tribute, edited by R.G. Chambers, J.E. Enderby, A. Keller, A.R. Lang and J.W. Steels (Bristol: Adam Hilger), p. 265. 20. A.S. Vaughan and D.C. Bassett (1989) In Comprehensive Polymer Science, edited by G. Allen and J.H. Bevington. (Oxford: Pergamon Press), 2, 415. 5. H. Staudinger and M. Luthy (1925) Helv Chim Acta 8,65. 6. K.H. Meyer and H. Mark (1928) Berichte B61, 593, 1932, 136, 1939. 7. W.T. Astbury (1938) Tram Faraday Soc 34, 378. 8. L. Pauling, R.B. Carey and H.R. Branson (1951) Proc Nat Acad Sci USA 37, 205. 9. F.H.C. Crick and J.D. Watson (1953) Nature 171, 737, 964. 10. K. Ziegler, E. Holzkamp, H. Brell and H. Martin (1955) Angew Chem 67, 541. 11. G. Natta, I.W. Bassi and P. Corradini (1955) Makromol Chem 18-19, 455. 12. N. Bekkedahl (1934) J Res Nat Bur Stand 13, 411. 13. L.A. Wood and N. Bekkedahl (1946) ] Appl Phys 17, 362. Author's biography: Geoffrey Allen is a Fellow of the Royal Society and a Foreign Member of the Engineering Academy of Japan. He was formerly Professor of Chemical Technology at the Imperial College of Science and Technology. He then became Chairman of the Science and Engineering Research Council and following that appointment was Research and Engineering Director of Unilever for nine years. He is now Executive Adviser to Kobe Steel Ltd. He has written numerous papers on the thermodynamic properties of polymers and on the application of structural techniques to plastics and rubbers.