SPATIAL CONFIGURATION OF MACROMOLECULAR CHAINS

Nobel Lecture, December 11, 1974 by

P AUL J. F L O R Y
Department of Chemistry Stanford University, Stanford, California The science of macromolecules has developed from primitive beginnings to a flourishing field of investigative activities within the comparatively brief span of some forty years. A wealth of knowledge has been acquired and new points of view have illumined various branches of the subject. These advances are the fruits of efforts of many dedicated investigators working in laboratories spread around the world. In a very real sense, I am before you on this occasion as their representative. In these circumstances, the presentation of a lecture of a scope commensurate with the supreme honor the Royal Swedish Academy of Sciences has bestowed in granting me the Nobel Prize for Chemistry is an insuperable task. Rather than attempt to cover the field comprehensively in keeping with the generous citation by the Royal Academy of Sciences, I have chosen to dwell on a single theme. This theme is central to the growth of ideas and concepts concerning macromolecules and their properties. Implemented by methods that have emerged in recent years, researches along lines I shall attempt to highlight in this lecture give promise of far-reaching advances in our understanding of macromolecular substances - materials that are invaluable to mankind. These polymeric substances are distinguished at the molecular level from other materials by the concatenation of atoms or groups to form chains, often of great length. That chemical structures of this design should occur is implicit in the multivalency manifested by certain atoms, notably carbon, silicon, oxygen, nitrogen, sulfur and phosphorus, and in the capacity of these atoms to enter into sequential combinations. The concept of a chain molecule consisting of atoms covalently linked is as old as modern chemistry. It dates from the origins of the graphic formula introduced by Couper in 1858 and advanced by Kekult, Loschmidt and others shortly thereafter. Nothing in chemical theory, either then apparent or later revealed, sets a limit on the number of atoms that may be thus joined together. The rules of chemical valency, even in their most primitive form, anticipate the occurrence of macromolecular structures. The importance of macromolecular substances, or polymers, is matched by their ubiquity. Examples too numerous to mention abound in biological systems. They comprise the structural materials of both plants and animals. Macromolecules elaborated through processes of evolution perform intricate regulatory and reproductive functions in living cells. Synthetic polymers in

P.J. Flory

157

great variety are familiar in articles of commerce. The prevailing structural motif is the linear chain of serially connected atoms, groups or structural units. Departures from strict linearity may sometimes occur through the agency of occasional branched units that impart a ramified pattern to the over-all structure. Linearity is predominant in most macromolecular substances, however. It is noteworthy that the chemical bonds in macromolecules differ in no discernible respect from those in monomeric compounds of low molecular weight. The same rules of valency apply; the lengths of the bonds, e.g., C-C, C-H, C-O, etc., are the same as the corresponding bonds in monomeric molecules within limits of experimental measurement. This seemingly trivial observation has two important implications: first, the chemistry of macromolecules is coextensive with that of low molecular substances; second, the chemical basis for the special properties of polymers that equip them for so many applications and functions, both in nature and in the artifacts of man, is not therefore to be sought in peculiarities of chemical bonding but rather in their macromolecular constitution, specifically, in the attributes of long molecular chains. Comprehension of the spatial relationships between the atoms of a molecule is a universal prerequisite for bridging the connection between the graphic formula and the properties of the substance so constituted. Structural chemistry has provided a wealth of information on bond lengths and bond angles. By means of this information the graphic formula, primarily a topological device, has been superseded by the structural formula and by the space model that affords a quantitative representation of the molecule in three dimensions. The stage was thus set for the consideration of rotations about chemical bonds, i.e., for conformational analysis of conventional organic compounds, especially cyclic ones. A proper account of bond rotations obviously is essential for a definitive analysis of the spatial geometry of a molecule whose structure permits such rotations. The configuration of a linear macromolecule in space involves circumstances of much greater complexity. A portion of such a molecule is shown schematically in Figure 1. Consecutive bonds comprising the chain skeleton are joined at angles fixed within narrow limits. Rotations q~ may occur about these skeletal bonds. Each such rotation is subject, however, to a potential determined by the character of the bond itself and by hindrances imposed by steric interactions between pendant atoms and groups. The number and variety of configurations (or conformations in the language of organic chemistry) that may be generated by execution of rotations about each of the skeletal bonds of a long chain, comprising thousands of bonds in a typical polymer, is prodigious beyond comprehension. When the macromolecule is free of constraints, e.g., when in dilutesolution, all of these configurations are accessible. Analysis of the manner in which such a molecule may arrange itself in space finds close analogies elsewhere in science, e.g., in the familiar problem of random walk, in diffusion, in the mathematical treatment of systems in one dimension, and in the behavior of real gases.

158

Chemistry 1974

Fig. 1. Representation of the skeletal bonds of a section of a chain molecule showing supplements of bond angles, and torsional rotations v for bonds i, i + 1, etc.

Inquiry into the spatial configuration of these long-chain molecules, fascinating in itself, derives compelling motivation from its close relevancy to the properties imparted by such molecules to the materials comprising them. Indeed, most of the properties that distinguish polymers from other substances are intimately related to the spatial configurations of their molecules, these configurations being available in profusion as noted. The phenomenon of rubber-like elasticity, the hydrodynamic and thermodynamic properties of polymer solutions, and various optical properties are but a few that reflect the spatial character of the random macromolecule. The subject is the nexus between chemical constitution and physical and chemical properties of polymeric substances, both biological and synthetic. The importance of gaining a grasp of the spatial character of polymeric chains became evident immediately upon the establishment, ca. 1930, of the hypothesis that they are covalently linked molecules rather than aggregates of smaller molecules, an achievement due in large measure to the compelling evidence adduced and forcefully presented by H. Staudinger, Nobelist for 1953. In 1932 K. H. Meyer1 adumbrated the theory of rubber-like elasticity by calling attention to the capacity of randomly coiled polymer chains to accommodate large deformations owing to the variety of configurations accessible to them. W. Kuhn 2 and E. Guth and H. Mark 3 made the first attempts at mathematical description of the spatial configurations of random chains. The complexities of bond geometry and of bond rotations, poorly understood at the time, were circumvented by taking refuge in the analogy to unrestricted random flights, the theory of which had been fully developed by Lord Rayleigh. The skeletal bonds of the molecular chain were thus likened to the steps in a random walk in three dimensions, the steps being uncorrelated one to another. Restrictions imposed by bond angles and hindrances to rotation were dismissed on the grounds that they should not affect the form of the results. For a random flight chain consisting of n bonds each of length l, the meansquare of the distance r between the ends of the chain is given by the familiar relation

The angle brackets denote the average taken over all configurations. Kuhn 4

P. J Flory

159

argued that the consequences of fixed bond angles and hindrances to rotation could be accommodated by letting several bonds of the chain molecule be represented by one longer equivalent bond, or step, of the random flight. This would require n to be diminished and l to be increased in Eq. 1. Equivalently, one may preserve the identification of n and l with the actual molecular quantities and replace Eq. (1) with

(2)
where C is a constant for polymers of a given homologous series, i.e., for polymers differing in length but composed of identical monomeric units. The proportionality between <r2> and chain length expressed in Eq. (2) may be shown to hold for any random chain of finite flexibility, regardless of the structure, provided that the chain is of sufficient length and that it is unperturbed by external forces or by effects due to excluded volume (cf. seq.). The result expressed in Eq. (2) is of the utmost importance. Closely associated with it is the assertion that the density distribution W(r) of values of the end-to-end vector r must be Gaussian for chains of sufficient length, irrespective of their chemical structure, provided only that the structure admits of some degree of flexibility. Hence, for large n the distribution of values of r is determined by the single parameter (r2) that defines the Gaussian distribution. Much of polymer theory has been propounded on the basis of the Kuhn equivalent random flight chain, with adjustment of n and l, or of C, as required to match experimental determination of (r 2) or of other configuration-dependent quantities. The validity of this model therefore invites critical examination. Its intrinsic artificiality is its foremost deficiency. Actual bond lengths, bond angles and rotational hindrances cannot be incorporated in this model. Hence, contact is broken at the outset with the features of chemical constitution that distinguish macromolecular chains of one kind from those of another. The model is therefore incapable of accounting for the vast differences in properties exhibited by the great variety of polymeric substances. The random flight chain is patently unsuited for the treatment of constitutive properties that are configuration-dependent, e.g., dipole moments, optical polarizabilities and dichroism. Inasmuch as the contribution to one of these properties from a structural unit of the chain is a vector or tensor, it cannot be referenced to an equivalent bond that is a mere line. Moreover, the equivalent bond cannot be embedded unambiguously in the real structure. Methods have recently been devised for treating macromolecular chains in a realistic manner. They take full account of the structural geometry of the given chain and, in excellent approximation, of the potentials affecting bond rotations as well. Before discussing these method, however, I must direct your attention to another aspect of the subject. I refer to the notorious effect of volume exclusion in a polymer chain. At the hazard of seeming trite, I should begin by pointing out that the chain molecule is forbidden to adopt a configuration in which two of its parts, or segments, occupy the same space. The fact is indisputable; its consequences are less obvious. It will be apparent, however, that volume exclusion vitiates

160

Chemistry 1974

Fig. 2. The effect of excluded volume. The configuration on the left represents the random coil in absence of volume exclusion, the chain being equivalent to a line in space. In the sketch on the right, the units of the chain occupy finite domains from which other units are excluded, with the result that the average size of the configuration is increased.

the analogy between the trajectory of a particle executing a random flight and the molecular chain, a material body. The particle may cross its own path at will, but self intersections of the polymer chain are forbidden. The effect of excluded volume must be dealt with regardless of the model chosen for representation of the chain. In practice, elimination of the effect of volume exclusion is a prerequisite to the analysis of experimental results, as I will explain in more detail later. The closely related problems of random flights with disallowance of self intersections and of volume exclusion within long-chain molecules have attracted the attention of many theorists. A variety of mathematical techniques have been applied to the treatment of these problems, and a profusion of theories have been put forward, some with a high order of sophistication. Extensive numerical computations of random walks on lattices of various sorts also have been carried out. Convergence of results obtained by the many investigators captivated by the subject over the past quarter century seems at last to be discernible. I shall confine myself to a brief sketch of an early, comparatively simple approach to the solution of this problem.5 The results it yields contrast with its simplicity. Returning to the analogy of the trajectory traced by a particle undergoing a sequence of finite displacements, we consider only those trajectories that are free of intersections as being acceptable for the chain molecule. Directions of successive steps may or may not be correlated, i.e., restrictions on bond angles and rotational hindrances may or may not be operative; this is immaterial with respect to the matter immediately at hand. Obviously, the set of eligible configurations will occupy a larger domain, on the average, than those having one or more self intersections. Hence, volume exclusion must cause <r2> to increase. The associated expansion of the spatial configuration is illustrated in Fig. 2. Other configuration-dependent quantities may be affected as well. This much is readily evident. Assessment of the magnitude of the perturbation of the configuration and its dependence on chain length require a more penetrating examination. The problem has two interrelated parts: (i) the mutual exclusion of the space occupied by segments comprising the chain tends to disperse them over a

P. J. Flory

161

larger volume, and (ii) the concomitant alteration of the chain configuration opposes expansion of the chain. Volume exclusion (i) is commonplace. It is prevalent in conventional dilute solutions and in real gases, molecules of which mutually exclude one another. In the polymer chain the same rules of exclusion apply, but treatment of the problem is complicated by its association with (ii). Pursuing the analogies to dilute solutions and gases, we adopt a smoothed density or mean field model. The segments of the chain, x in number, are considered to pervade a volume V, the connections between them being ignored insofar as part (i) is concerned. The segment need not be defined explicitly; it may be identified with a repeating unit or some other approximately isometric portion of the chain. In any case, x will be proportional to the number n of bonds; in general x n, however. For simplicity, we may consider the segment density Q to be uniform throughout the volume V; that is, e = x/V within V and Q = 0 outside of V. This volume should be proportional to ( r2 ) 3 / 2, where <r 2> is the mean-square separation of the ends of the chain averaged over those configurations not disallowed by excluded volume interactions. Accordingly, we let

where A is a numerical factor expected to be of the order of magnitude of unity. It is necessary to digress at this point for the purpose of drawing a distinction between <r2> for the chain perturbed by the effects of excluded volume and <r 2> 0 for the unperturbed chain in the absence of such effects. If denotes the factor by which a linear dimension of the configuration is altered, then (4) Equation (2), having been derived without regard for excluded volume interactions, should be replaced by

(2')
where C reaches a constant value with increase in n for any series of finitely flexible chains. The smoothed density within the domain of a linear macromolecule having a molecular weight of 100,000 or greater (i.e., n > 1000) is low, only on the order of one percent or less of the space being occupied by chain segments. For a random dispersion of the segments over the volume V, encounters in which segments overlap are rare in the sense that few of them are thus involved. However, the expectation that such a dispersion is entirely free of overlaps between any pair of segments is very small for a long chain. The attrition of configurations due to excluded volume is therefore severe. In light of the low segment density, it suffices to consider only binary encounters. Hence, if is the volume excluded by a segment, the probability that an arbitrary distribution of their centers within the volume V is free of conflicts between any pair of segments is

P. J. Flory

163

sums are executed over all configurations of the chain. The squared radius of gyration s 2, i.e., the mean-square of the distances of the segments from their center of gravity, is preferable to r2 as a parameter with which to characterize the spatial distribution. Treatments carried out with these refinements affirm the essential validity of the result expressed by Eq. (12) or (12). They show conclusively 7,8 that the form of the result should hold in the limit of large values of /3xri2, i.e., for large excluded volume and/or high chain length, and hence for a > > 1. In this limit, (as- a) /z = 1.67 according to H. Fujita and T. Norisuye. 8 For a < ~ 1.4, however, this ratio decreases, reaching a value of 1.276 at a = 1 .8,9 The general utility of the foregoing result derived from the most elementary considerations is thus substantiated by elaboration and refinement of the analysis, the quantitative inaccuracy of Eqs. (12) and (12) in the range 1.0 < a ,< 1.4 notwithstanding. The relationship between a and the parameter z prescribed by these equations, especially as refined by Fujita and Norisuye, 8 appears to be well supported by experiment.10,11 The principal conclusions to be drawn from the foregoing results are the following: the expansion of the configuration due to volume exclusion increases with chain length without limit for > 0; for very large values of /3x1/z relative to ((rz),,/x)32 it should increase as the l/10 power of the chain length. The sustained increase of the perturbation with chain length reflects the fact that interactions between segments that are remote in sequence along the chain are dominant in affecting the dimensions of the chain. It is on this account that the excluded volume effect is often referred to as a long-range intera c t i o n .9-12 The problem has been treated by a variety of other procedures. 9-12 Notable amongst these treatments is the self-consistent field theory of S. F. Edwards. 12 The asymptotic dependence of a on the one-tenth power of the chain length, and hence the dependence of <r2> on n 6 / 5 for large values of the parameter z, has been confirmed.12 The dilute solution is the milieu chosen for most physicochemical experiments conducted for the purpose of characterizing polymers. The effect of excluded volume is reflected in the properties of the polymer molecule thus determined, and must be taken into account if the measurements are to be properly interpreted, The viscosity of a dilute polymer solution is illustrative. Its usefulness for the characterization of polymers gained recognition largely through the work of Staudinger and his collaborators. Results are usually expressed as the intrinsic viscosity [q] defined as the ratio of the increase in the relative viscosity qrel by the polymeric solute to its concentration c in the limit of infinite dilution. That is,

the concentration c being expressed in weight per unit volume. The increment in viscosity due to a polymer molecule. is proportional to its hydrodynamic volume, which in turn should be proportional to <r 2> 3/2 for a typical polymer

164

Chemistry 1974

chain. Hence, ~~1-1 should be proportional to the product of <r2> 3/2 a n d the number density of solute molecules given by c/M where M is the molecular weight. It follows that

molecular weight, provided of course that the molecular weight, and hence the chain length, is sufficiently large. If the excluded volume effect could be ignored, the intrinsic viscosity should vary proportionally to M 1/2 . Since, however, increases with M, a stronger dependence on M generally is observed. Often the dependence of [r] on molecular weight can be represented in satisfactory approximation by the empirical relation

Measurement of light scattering as a function of angle, a method introduced by the late P. Debye, affords a convenient means for determining the meansquare radius of gyration. S ma ll-angle scattering of x-rays (and lately of neutrons) offers an alternative for securing the same information. From the radius of gyration one may obtain the parameter <r2> upon which attention is focused here. The results are affected, of course, by the perturbation due to excluded volume. Inasmuch as the perturbation is dependent on the solvent and temperature, the results directly obtained by these methods are not intrinsically characteristic of the macromolecule. Values obtained for <r 2> from the intrinsic viscosity by use of Eq. (13), or by other methods, must also be construed to be jointly dependent on the macromolecule and its environment. If the factor were known, the necessary correction could be introduced readily to obtain the more substantive quantities, such as <r 2> 0 and <s 2> 0 that characterize the macromolecule itself and are generally quite independent of the solvent. Evaluation of according to Eq. (11) and (12) would require the excluded volume . This parameter depends on the solvent in a manner

P. J. Flory

165

that eludes prediction. Fairly extensive experimental measurements are required for its estimation, or for otherwise making correction for the expansion . All these difficulties are circumvented if measurements on the polymer solution are conducted under conditions such that the effects of excluded volume are suppressed. The resistance of atoms to superposition cannot, of course, be set aside. But the consequences thereof can be neutralized. We have only to recall that the effects of excluded volume in a gas comprising real molecules of finite size are exactly compensated by intermolecular attractions at the Boyle temperature (up to moderately high gas densities). At this temperature the real gas masquerades as an ideal one. For the macromolecule in solution, realization of the analogous condition requires a relatively poor solvent in which the polymer segments prefer selfcontacts over contacts with the solvent. The incidence of self-contacts may then be adjusted by manipulating the temperature and/or the solvent composition until the required balance is established. Carrying the analogy to a real gas a step further, we require the excluded volume integral for the interaction between a pair of segments to vanish; that is, we require that =0. This is the necessary and sufficient condition. 5,6,13 As already noted, estimation of the value of is difficult; the prediction of conditions under which shall precisely vanish would be even more precarious. However, the Theta point, so-called, at which this condition is met is readily identified with high accuracy by any of several experimental procedures. An excluded volume of zero connotes a second virial coefficient of zero, and hence conformance of the osmotic pressure to the celebrated law of J. H. vant Hoff. The Theta point may be located directly from osmotic pressure determinations, from light scattering measured as a function of concentration, or from determination of the precipitation point as a function of molecular weight.6,13 The efficacy of this procedure, validated a number of years ago with the collaboration of T. G. Fox, W. R. Krigbaum, and others, 13,17,18 is illustrated in Figs. 3 and 4 by the lower plots of data representing intrinsic viscosities measured under ideal, or Theta conditions.6 The slopes of the lines drawn through the lower sets of points are exactly 1/2, as required by Eq. (13) when = 0 and hence = 1. The excellent agreement here illustrated has been abundantly confirmed for linear macromolecules of the widest variety, ranging from polyisobutylene and polyethylene to polyribonucleotides.19 At the Theta point the mean-square chain vector <r 2> 0 and the mean-square radius of gyration <s2> 0 invariably are found to be proportional to chain length. A highly effective strategy for characterization of macromolecules emerges from these findings. By conducting experiments at the Theta point, the disconcerting (albeit interesting!) effects of excluded volume on experimentally measured quantities may be eliminated. Parameters (e.g., <r 2> 0 and <s 2> 0) are thus obtained that are characteristic of the molecular chain. They are found to be virtually independent of the nature of the Theta solvent selected. Having eliminated the effects of long range interactions, one may turn

166

Chemistry 1974

Fig. 3. Intrinsic viscosities of polystyrene fractions plotted against their molecular weights on logarithmic scales in accordance with Eq. (14). The upper set of data was determined in benzene, a good solvent for this polymer. The lower set of data was determined in cyclohexane at the Theta point. The slopes of the lines are a = 0.75 and 0.50, respectively. From the results of Altares, Wyman and Allen.14

attention to the role of short range features: structural geometry, bond rotation potentials, and steric interactions between near-neighboring groups. It is here that the influences of chemical architecture are laid bare. If the marked differences in properties that distinguish the great variety of polymeric substances, both natural and synthetic, are to be rationally understood in fundamental, molecular terms, this must be the focus of future research. Rigorous theoretical methods have recently become available for dealing realistically with short-range features peculiar to a given structure. Most of the remainder of this lecture is devoted to a brief overview of these methods. Although the field is comparatively new and its exploration has only begun, space will not permit a digest of the results already obtained. The broad objective of the methods to which we now turn attention is to treat the structure and conformations accessible to the chain molecule in such a manner as will enable one to calculate configuration-dependent quantities and to average them over all conformations, or spatial configurations, of the unperturbed chain. The properties under consideration are constitutive; they

P. J. Flory I I I I I I I

167

- 1.0 e - 0.8
0.6

0.4

0.2

Fig. 4. Intrinsic viscosities of fractions of poly(methyl methacrylate) according to Chinai and Samuels15 plotted as in Fig. 3. The upper set of points was measured in methyl ethyl ketone, a good solvent. The lower set was determined in a mixture of methyl ethyl ketone and isopropanol at the Theta point. Slopes are a = 0.79 and 0.50, respectively.

represent sums of contributions from the individual units, or chemical groupings, comprising the chain. In addition to <r 2> 0 and <s2> 0, they include: mean-square dipole moments; the optical anisotropies underlying strain birefringence, depolarized light scattering and electric birefringence; dichroism; and the higher moments, both scalar and tensor, of the chainvector r. Classical statistical mechanics provides the basis for evaluating the configurational averages of these quantities. Since bond lengths and bond angles ordinarily may be regarded as fixed, the bond rotations are the variables over which averaging must be carried out. The procedure rests on the rotational isomeric state scheme, the foundations for which were set forth in large measure by M. V. Volkenstein 20 and his colleagues 21 in Leningrad in the late 1950s and early 1960s. It is best explained by examples. Consider rotation about an internal bond of an n-alkane chain. As is now well established, 22,23 the three staggered conformations, trans(t), gaucheplus(g + ) and its mirror image, gauche-minus(g-), are of lower energy than

168

Chemistry 1974

Fig. 5. Two of the staggered conformations for n-butane: trans on the left and gaucheminus on the right.

the eclipsed forms. The t and g- conformations of n-butane are shown in Fig. 5. The energies of the eclipsed conformations separating t from g+ and t from g- are about 3.5 kcal. mol-1 above the energy of the trans conformation. Hence, in good approximation, it is justified to consider each bond to occur in one of three rotational isomeric states centered near (but not necessarily precisely at) the energy minima associated with the three staggered conformations. 20-24 The gauche minima lie at an energy of about 500 cal. mol -1 above trans. Each of the former is therefore disfavored compared to the latter by a statistical weight factor we choose to call exp( -Eg/RT), where Eg is about 500 cal. mol -1 ; thus, ( 0.5 at T = 400 K. A complication arises from the fact that the potentials affecting bond rotations usually are neighbor dependent; i.e., the potential affecting pi depends on the rotations pi-1 and vi+r. Bond rotations cannot, therefore, be treated independently .20,21,24,25 The source of this interdependence in the case of an n-alkane chain is illustrated in Fig. 6 showing a pair of consecutive bonds in three of their nine conformations. In the conformations tt, tg + , g+ t, tg- and g-t, the two methylene groups pendant to this pair of bonds are well separated. For gauche rotations g+g+ and g-g- of the same hand (Fig. 6b), these groups are proximate but not appreciably overlapped. Semi-empirical calculations21,24,26, 27 show the intramolecular energy for these two equivalent conformations to be very nearly equal to the sum (cu. 1000 cal. mol-1) for two well-separated gauche bonds; i.e., the interdependence of the pair of rotations is negligible. In the remaining conformations, g + g - and g -g + , the steric overlap is severe (Fig. 6c). It may be alleviated somewhat by compromising rotations, but the excess energy associated therewith is nevertheless about 2.0 kcal. mol -1. Hence, a statistical weight factor exp (-2000/RT) is required for each such pair. 24,26,28 Inspection of models in detail shows that interactions dependent upon rotations about three, four of five consecutive bonds are disallowed by interferences of shorter range and hence may be ignored. 24 It suffices therefore to consider first neighbors only. The occurrence of interactions that depend on pairs of skeletal bonds is the rule in chain molecules. In some of them, notably in vinyl polymers, such interactions may affect most of the conformations. Hence, interdependence of rotations usually plays a major role in determining the spatial configuration

P. J Flory

169 H H

H H

of the chain. The rotational isometric state approximation, whereby the continuous variation of each is replaced by discrete states, provides the key to mathematical solution of the problem posed by rotational interdep e n d e n c e .2 0 , 2 1 , 2 4 , 2 5 It is necessary therefore to consider the bonds pairwise consecutively, and to formulate a set of statistical weights for bond i that take account of the state of bond i-l. These statistical weights are conveniently presented in the form of an array, or matrix, as follows:

where the rows are indexed in the order t, g + , g- to the state of bond i- 1, and the columns are indexed to the state of bond i in the same order. According to the analysis of the alkane chain conformations presented briefly above, U i takes the form 24,26,28

170

Chemistry 1974

(16)

A conformation of the chain is specified in the rotational isometric state approximation by stipulation of the states for all internal bonds 2 to n-l inclusive; e.g., by g + ttg -g -, etc. Owing to the three-fold symmetry of the terminal methyl groups of the alkane chain, rotations about the terminal bonds are inconsequential and hence are ignored. The statistical weight for the specified conformation of the chain is obtained by selecting the appropriate factor for each bond from the array (15) according to the state of this bond and of its predecessor, and taking the product of such factors for all bonds 2 to n - 1. In the example above this product is us+ us+t utt uts- us s , etc. It will be obvious that the first superscripted index in one of the factors u must repeat the second index of its predecessor since these indices refer to the same bond. The configuration partition function, representing the sum of all such factors, one for each conformation of the chain as represented by the scheme of rotational isomeric states, is

where the subscripts are serial indexes. Each ui must be assigned as specified above. The sum, which extends over all ordered combinations of rotational states, may be generated identically as the product of the arrays U i treated as matrices. That is, according to the rules of matrix multiplication

w h e r e U1 = row (1, 0, 0) and Un = column (1, 1, 1). Matrix multiplication generates products precisely of the character to which attention is directed at the close of the preceding paragraph. Serial multiplication of the statistical weight matrices generates this product for each and every conformation of the chain, and Eq. (18) with the operators U 1 and U n appended gives their sum. The foregoing procedure for evaluation of Z is a minor variant of the method 29 of H. A. Kramers and G. H. Wannier for treating a hypothetical onedimensional ferromagnet or lattice. A number of interesting characteristics of the chain molecule can be deduced from the partition function by application of familiar techniques of statistical mechanics. I shall resist the temptation to elaborate these beyond mentioning two properties of the molecule that may be derived directly from the partition function, namely, the incidences of the various rotational states and combinations thereof, and the equilibrium constants between isomeric structures of the chain in the presence of catalysts effectuating their inter-conversion. Vinyl polymers having the structure depicted in Fig. 7 with R R afford examples wherein the study of equilibria

between various diastereomeric forms arising from the local chirality of individual skeletal bonds has been especially fruitful. 30 Consider the evaluation of a configuration-dependent property for a given configuration, or conformation, of the chain. Since the configuration is seldom given, the problem as stated is artificial. Its solution, however, is a necessary precursor to the ultimate goal, which is to obtain the average of the property over all configurations. A property or characteristic of the chain that will serve for illustration is the end-to-end vector r. Suppose we wish to express this vector with reference to the first two bonds of the chain. For definiteness, let a Cartesian coordinate system be affixed to these two bonds with its X,-axis along the first bond and its Y -axis in the plane of bonds 1 and 2, as shown in Fig. 8.
1

in this reference frame.

In order to facilitate the task of transforming every bond vector to the reference frame affiliated with the first bond, it is helpful to define a reference frame for each skeletal bond of the chain. For example, one may place the axis Xi along bond i, the Yi-axis in the plane of bonds i - l and i, and choose the Zi-axis to complete a right-handed Cartesian system. Let Ti symbolize the transformation that, by premultiplication, converts the representation of a vector in reference frame i+l to its representation in the preceding reference frame i. Then bond i referred to the initial reference frame is given by

where li is presented in reference frame i. The required sum is just

(19)
This sum of products can be generated according to a simple algorithm. We first define a generator matrix A i as follows31,32

172

Chemistry 1974

Fig. 8. Specification of the coordinate axes affixed to each of the first two bonds of the chain: X1 Y 1 for bond 1 and X2 Y 2 for bond 2.

together with the two terminal matrices

(21) (22)
In these equations Ti is the matrix representation of the transformation specified above and 0 is the null matrix of order 1 x 3. The desired vector r is generated identically by taking the serial product of the As; i.e.,

as may easily be verified from the elementary rules of matrix multiplication. Each generator matrix A i depends on the length of bond i and, through T i, on both the angle f3i between bonds i and i+l and on the angle of rotation plr about bond i (see Fig. 1). In order to obtain the average of r over all configurations of the chain, it is necessary to evaluate the sum over all products of the kind given in Eq. (23) with each of them multiplied by the appropriate statistical weight for the specified configuration of the chain; see Eq. (17). That is,

P. J Flory

173

T h e n31

The matrix a i comprises the elements of Ui (see Eq. (15)) joined with the A matrix for the rotational state of bond i as prescribed by the column index. It will be apparent that serial multiplication of the a i according to Eq. (28) generates the statistical weight factor u2u3. . .u n-1 for every configuration of the chain in the same way that these factors are generated by serial multiplication of the statistical weight matrices U i in Eq. (18). Simultaneously, Eq. (28) generates the product of As (see Eq. (23)) that produces the vector r for each configuration thus weighted. The resulting products of statistical weights and of As are precisely the terms required by Eq. (24). The terminal factors in Eq. (28) yield their sum. With greater mathematical concision 31,32

If each bond vector li is expressed in its own reference frame i, then

(34)

174

Chemistry 1974

20
2

40

60

80

100

NUMBER OF BONDS, n
Fig. 9. Characteristic ratios <r > o/G plotted against the number of bonds n in the chain for polymethylene, and for isotactic and syndiotactic poly (methyl metacrylate) s. From the calculations of Abe, Jernigan and Flory26 and of Yoon.34

That is, (35) where G1 has the form of the first row, and G n that of final column of Eq. (34). Evaluation of <r2> 0 proceeds exactly as set forth above for <r>0.32,33 The foregoing method enjoys great versatility. The chain may be of any specified length and structure. If it comprises a variety of skeletal bonds and repeat units, the factors entering into the serial products have merely to be fashioned to introduce the characteristics of the bond represented by each of the successive factors. The mathematical methods are exact; the procedure is free of approximations beyond that involved in adoption of the rotational isometric state scheme. With judicious choice of rotational states, the error here involved is generally within the limits of accuracy of basic information on bond rotations, nonbonded interactions, etc. Other molecular properties that may be computed by straightforward 4 6 adaptation of these methods 24,32 include the higher scalar moments <r >0, <r >0, etc; the moment tensors formed from r; the radius ofgyration <s2> 0 = (n+ 1)-2 c c (r2jj); the optical polarizability and its invariants that govern the i j optical anistropy as manifested in depolarized light scattering, in strain bire-

P. J. Flory

175

fringence and in electric birefringence; x-ray scattering at small angles; and NMR chemical shifts. For illustration, characteristic ratios <r2>0/nl2 are plotted in Fig. 9 against the numbers n of bonds for n-alkanes and for isotactic and syndiotactic poly(methyl methacrylate), or PMMA. Isotactic PMMA is represented by the formula in Fig. 7 with R = COOCH, and R = CH, and with all dyads of the meso form, i.e., with R occurring consistently above (or below) the axis of the chain. In the syndiotactic stereoisomer, the substituents R and R alternate from one side to the other, all dyads being racemic. For the alkane and the isotactic PMMA chains the characteristic ratios increase monotonically with chain length, approaching asymptotic values for n 100 bonds. This behavior is typical. For syndiotactic PMMA, however, the characteristic ratio passes through a maximum at intermediate values of n, according to these computations by D. Y. Yoon.34 This behavior can be traced34 to the inequality of the skeletal bond angles in PMMA in conjunction with the preference for tt conformations in the syndiotactic chain.35 The maximum exhibited in Fig. 9 for this polymer is thus a direct consequence of its constitution. This peculiarity manifests itself in the small angle scattering of x-rays and neutrons by predominantly syndiotactic PMMA of high molecular weight. 36 Scattering intensities are enhanced at angles corresponding, roughly, to distances approximating < r2> 0 1 / 2 at the maximum in Fig. 9. This enhancement, heretofore considered anomalous, is in fact a direct consequence of the structure and configuration of syndiotactic PMMA. It is thus apparent that subtle features of the chemical architecture of polymeric chains are manifested in their molecular properties. Treatment in terms of the artificial models much in use at present may therefore be quite misleading. The analysis of the spatial configurations of macromolecular chains presented above is addressed primarily to an isolated molecule as it exists, for example, in a dilute solution. On theoretical grounds, the results obtained should be equally applicable to the molecules as they occur in an amorphous polymer, even in total absence of a diluent. This assertion follows unambiguously from the statistical thermodynamics of mixing of polymer chains, 5,6,37 including their mixtures with low molecular diluents. It has evoked much skepticism, however, and opinions to the contrary have been widespread. These opposing views stem primarily from qualitative arguments to the effect that difficulties inherent in the packing of long chains of consecutively connected segments to space-filling density can only be resolved either by alignment of the chains in bundle arrays, or by segregation of individual molecules in the form of compact globules. In either circumstance, the chain configuration would be altered drastically. Whereas dense packing of polymer chains may appear to be a distressing task, a thorough examination of the problem leads to the firm conclusion that macromolecular chains whose structures offer sufficient flexibility are capable of meeting the challenge without departure or deviation from their intrinsic proclivities. In brief, the number of configurations the chains may assume is

176

Chemistry 1974

sufficiently great to guarantee numerous combinations of arrangements in

which the condition of mutual exclusion of space is met throughout the system as a whole. Moreover, the task of packing chain molecules is not made easier by partial ordering of the chains or by segregating them.6,37 Any state of organization short of complete abandonment of disorder in favor of creation of a crystalline phase offers no advantage, in a statistical-thermodynamic sense. Theoretical arguments aside, experimental evidence is compelling in showing the chains to occur in random configurations in amorphous polymers, and further that these configurations correspond quantitatively with those of the 38 unperturbed state discussed above. The evidence comes from a variety of sources: from investigations on rubber elasticity, chemical cyclization equilibria, thermodynamics of solutions, and, most recently, from neutron scattering studies on protonated polymers in deuterated hosts (or vice versa). T h e investigations last mentioned go further. They confirm the prediction made twenty-five years ago that the excluded volume perturbation should be 5 annulled in the bulk amorphous state. The excluded volume effect is therefore an aberration of the dilute solution, which, unfortunately, is the medium preferred for physicochemical characterization of macromolecules. Knowledge gained through investigations, theoretical and experimental, on the spatial configuration and associated properties of random macromolecular chains acquires added significance and importance from its direct, quantitative applicability to the amorphous state. In a somewhat less quantitative sense, this knowledge applies to the intercrystalline regions of semicrystalline polymers as well. It is the special properties of polymeric materials in amorphous phases that render them uniquely suited to many of the functions they perform both in biological systems and in technological applications. These properties are intimately related to the nature of the spatial configurations of the constituent molecules. Investigation of the conformations and spatial configurations of macromolecular chains is motivated therefore by considerations that go much beyond
39

its appeal as a stimulating intellectual exercise. Acquisition of a thorough understanding of the subject must be regarded as indispensable to the comprehension of rational connections between chemical constitution and those properties that render polymers essential to living organisms and to the needs

of man. REFERENCES
1. 2. 3. 4. 5. 6. Meyer, K. H., von Susich, G., and Valk, E., Kolloid-Z, 59, 208 (1932). Kuhn, W., Kolloid-Z, 68, 2 (1934). Guth, E., and Mark, H., Monatsch., 65, 93 (1934). Kuhn, W., Kolloid-Z., 76, 258 (1936); 87, 3 (1939). Flory, P. J., J. Chem. Phys., 17, 303 (1949). Flory, P. J., Principles of Polymer Chemistry, Cornell University Press, Ithaca, N.Y., 1953. 7. Flory, P. J., and Fisk, S., J. Chem. Phys., 44, 2243 (1966). 8. Fujita, H., and Norisuye, T., J. Chem. Phys., 52, 115 (1971). 9. Fixman, M., J. Chem. Phys., 23, 1656 (1955).

P. J. Flory

177

10. Yamakawa, H., Modern Theory of Polymer Solutions, Harper and Row, New York, 1971. 11. Yamakawa, H., Pure and Appl. Chem., 31, 179 (1972). 12. Edwards, S. F., Proc. Phys. Soc., (London), 85, 613 (1965). 13. Fox, T. G., Jr., and Flory, P. J., J. Phys. and Coll. Chem., 53, 197 (1949). Flory, P. J., and Fox, T. G., Jr., J. Polymer Sci., 5, 745 (1950) ; J. Amer. Chem. Soc., 73, 1904 (1951). 14. Altares, T., Wyman, D. P. and Allen, V. R., J. Polymer Sci., A, 2, 4533 (1964). 15. Chinai, S. N., and Samuels, R. J., J. Polymer Sci., 19, 463 (1956). 16. Mandelkern, L., and Flory, P. J., J. Che m . Phys., 20, 212 (1952), Mandelkern, L., Krigbaum, W. R. and Flory, P. J., ibid., 20, 1392 (1952). 17. Fox, T. G., Jr., and Flory, P. J., J. Amer. Chem. Soc., 73, 1909, 1915 (1951). 18. Krigbaum, W. R., Mandelkern, L., and Flory, P. J., J. Polymer Sci., 9, 381 (1952). Krigbaum, W. R. and Flory, P. J., ibid, 11, 37 (1953), 19. Eisenberg, H., and Felsenfeld, G., J. Mol. Biol., 30, 17 (1967). Inners, L. D., and Felsenfeld, G., ibid., 50, 373 (1970). 20. Volkenstein, M. V., Configurational Statistics of Polymeric Chains, translated from the Russian ed., 1959, by S. N. and M. J. Timasheff, Interscience, New York, 1963. 21. Birshtein, T. M. and Ptitsyn, O. B., Conformations of Macromolecules, translated from the Russian ed., 1964, by S. N. and M. J. Timasheff, Interscience, New York, 1966. 22. Pitzer, K. S., Discussions Faraday Soc., 10, 66 (1951). 23. Mizushima, S., Structure of Molecules and Internal Rotation, Academic Press, New York, 1954. 24. Flory, P. J., Statistical Mechanics of Chain Molecules, Interscience Publishers, New York, 1969. 25. Gotlib, Yu. Ya., Zh. Fiz Tekhn, 29, 523 (1959). Birshtein, T. M., and Ptitsyn, itson, S., J. Chem. Phys., 30, 964 (1959). Nagai, O. B., ibid., 29, 1048 (1959). L K., ibid., 31, 1169 (1959), Hoeve, C. A. J., ibid., 32, 888 (1960). 26. Abe, A., Jernigan, R. L., and Flory, P. J., J. A mer. Chem. Soc., 88, 63 1 ( 1966). 27. Scott, R. A., and Scheraga, H. A., J. Chem. Phys., 44, 3054 (1966). 28. Hoeve, C. A. J., J. Chem. Phys., 35, 1266 (1961). 29. Kramers, H. A., and Wannier, G. H., Phys. Rev., 60, 252 (1941). 30. Williams, A. D., and Flory, P. J., J. Amer. Chem. Soc., 91, 3111, 3118 (1969). Flory, P. J., and Pickles, C. J., J. Chem. Soc., Faraday Trans. II, 69, 632 ( 1973). Suter, U. W., Pucci, S., and Pino, P., J. Amer. Chem. Soc., 97 1018 (1975). 31. Flory, P. J., Proc. Nat. Acad. Sci., 70, 1819 (1973). 32. Flory, P. J., Macromolecules, 7, 381 (1974). 33. Flory, P. J., and Abe, Y., J. Chem. Phys. 54, 1351 (1971). 34. Yoon, D. Y., unpublished results, Laboratory of Macromolecular Chemistry, Stanford University. 35. Sundararajan, P. R., and Flory, P. J., J. Amer. Chem. Soc., 96, 5025 (1974). 36. Kirste, R. G., and Kratky, O., Z. Physik, Chem. Neue Folge, 31, 363 (1962). Kirste, R. G., Makromol. Chem., 101, 91 (1967). Kirste, R. G., Kruse, W. A., and Ibel, K., Polymer, 16, 120 (1975). 37. Flory, P. J., Proc. Royal Soc., A, 234, 60 (1956). Flory, P. J., J. Polym. Sci., 49, 105 (1961). 38. Flory, P. J., Pure & Appl. Chem., Macromolecular Chem., 8, l-15 (1972). 39. Kirste, R. G., Kruse, W. A., and Schelten, J., Makromol. Chem., 162, 299 (1972). Benoit, H., Decker, D., Higgins, J. S., Picot, C., Cotton, J. P., Farnoux, B., Jannink, G., and Ober, R., Nature, Physical Sciences, 245, 13 (1973). Ballard, D. G. H., Wignall, G. D., and Schelten, J., Eur. Polymer J., 9, 965 (1973); ibid, 20, 861 (1974). Fischer, E. W., Leiser, G., and Ibel, K. Polymer Letters, 13, 39 (1975).