bniowp The Theory of Polymer Dynamics M. DOI Department of Applied Physics, Nagoya University and $8. F. EDWARDS Cavendish Laboratory, University of Cambridge CLARENDON PRESS - OXFORDOxford University Press, Walton Street, Oxford 0x2 6DP Oxford New York Toronto Delhi Bornbay Caicutta Madras Karachi Kuala Lumpur Singapore Hong Kong Tokyo Nairobi Dar es Salaam Cape Town Melbourne Auckland Madrid and associated companies in Berlin Ibadan Oxford is a trade mark of Oxford University Press Published in the United States by Oxford University Press Inc. New York © M. Doi and S. F. Edwards, 1986 First published 1986 First published in paperback (with corrections) 1988 Reprinted 1989, 1992, 1994 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press. Within the UK, exceptions are allowed in respect of any {fair dealing for ihe purpose of research or private study, or criticism or review, as permitted under the Copyright, Designs and Patents Act, 1988, or in the case of reprographic reproduction in accordance with the terms of licences issued by the Copyright Licensing Agency. Enquiries concerning reproduction outside those terms and in other countries should be sent to the Rights Department, Oxford University Press, at the address above. This book is sold subject to the condition that it shall not, by way of trade or otherwise, be lent, resold, hired out, or otherwise circulated without the publisher's consent in any form of binding or cover other than that in which it is published and without a similar condition including this condition being imposed on the subsequent purchaser. British Library Cataloguing in Publication Data Doi, M. The theory of polymer dynamics 1. Polymers and polymerization I. Title _II. Edwards, S. F. 1. Polymers and polymerization I, Edwards, S. F. _ Il. Tide QD281.P6E27 1986 547.7 _85-29854 ISBN 0 19 852033 6 (Pbk) Printed in Northern Ireland by The Universities Press (Belfast) LidPREFACE Many people have helped us write this book. We had very helpful comments during the writing of the book from R. C. Ball, F. Boue, J. des Cloizeaux, W. Griffin and J. Klein and the manuscript was read by G. Berry, R. Hayakawa, M. Kurata, Y. Oono, D. Pearson, and E. Samulski who all made useful improvements to it. We had meticulous help with the proofs from Fiona Miller and Ed Samulski. M. D. would like to give special thanks to Dr. Osaki who gave continual help and encouragement throughout and to the Japan Society for the Promotion of Science and the Royal Society for making his stay in Cambridge with S. F. E. in 1984/5 possible when the book was written. It is a pleasure for S. F. E. to record, with gratitude, that it is now twenty years since he was introduced to this field by Professor Geoffrey Gee and Sir Geoffrey Allen; it has proved of abiding interest. Tokyo and M.D. Cambridge S.F.E. March 1986CONTENTS 1 INTRODUCTION 2 STATIC PROPERTIES OF POLYMERS 2.1 2.2 2.3 2.4 2.5 2.6 The random flight model 2.1.1 The freely jointed model 2.1.2 General random flight models 2.1.3 Distribution of the end-to-end vector The Gaussian chain Chain conformation under an external field 2.3.1 The Green function 2.3.2. Example—chain confined in a box Scattering function Excluded volume effect 2.5.1 Introduction 2.5.2. Model of the excluded volume chain 2.5.3 Theoretical approaches Scaling Appendix 2.1 Gaussian distribution functions Appendix 2.II Differential equation for G(R, R'; N) Appendix 2.III Perturbation calculation for the excluded volume effect References 3 BROWNIAN MOTION 3.1 3.2 _ 33 3.4 3.5 3.6 Introduction The Smoluchowski equation 3.2.1 Diffusion of particles 3.2.2 Diffusion in phase space 3.2.3 Irreversibility of the Smoluchowski equation The Langevin equation Time correlation function and response function 3.4.1 Time correlation function 3.4.2 Microscopic expression for the time correlation function 3.4.3 Fluctuation dissipation theorem Brownian motion in a harmonic potential 3.5.1 Smoluchowski equation 3.5.2 Langevin equation Interacting Brownian particles3.7 3.8 CONTENTS Microscopic basis of viscoelasticity 3.7.1 Introduction 3.7.2 Constitutive equation 3.7.3. The Smoluchowski equation for a system in macroscopic flow 3.7.4 Expression for the stress tensor 3.7.5 Principle of virtual work Systems with rigid constraints 3.8.1 Introduction 3.8.2 The method of generalized coordinates 3.8.3 The method of Lagrangian multipliers 3.8.4 Elastic stress and viscous stress 3.8.5 Variational formulation Appendix 3.1 Eigenfunctions of the Smoluchowski equation Appendix 3.11 Relationship between the Langevin equation and the Smoluchowski equation Appendix 3.1II The Oseen tensor References 4 DYNAMICS OF FLEXIBLE POLYMERS IN DILUTE SOLUTION 41 4.2 4.3 4.4 4.5 4.6 47 The Rouse model 4.1.1 Dynamics of a polymer with localized interaction 4.1.2 Normal coordinates The Zimm model 4.2.1 Zimm model in © conditions 4.2.2 Zimm model in good solvent Dynamical scaling Dynamic light scattering Viscoelasticity 4.5.1 Introduction 4.5.2 Microscopic expression for the stress tensor 4.5.3 Calculation of the intrinsic viscosity 4.5.4 Intrinsic moduli Variational bounds for the transport coefficients 4.6.1 Introduction 4.6.2 Bounds for the intrinsic viscosity 4.6.3 Bounds for the diffusion constant Birefringence 4.7.1 Birefringence of polymer solutions 4.7.2 Molecular expression for birefringence 4.7.3 Flow birefringence 69 70 71 1S 76 76 79 85 89 91 91 91 94 97 100 103 105 108 108 110 112 114 116 116 116 119 121 121 127CONTENTS Appendix 4.1 The Verdier—Stockmayer model Appendix 4.11 Derivation of the normal modes Appendix 4.III Dynamic structure factor Appendix 4.IV Polarizability tensor of a Gaussian chain References 5 MANY CHAIN SYSTEMS 5.1 Semidilute and concentrated solutions 5.2 Gaussian approximation for concentration fluctuations 5.2.1 Collective coordinates 5.2.2. Pair correlation function 5.2.3 Osmotic pressure 5.2.4 Size of a single chain 5.3 Scaling theory—statics 5.4 Topological interaction in polymer dynamics 5.4.1 Entanglement effect 5.4.2 Rigorous approach 5.4.3. The tube model 5.5 Dynamics of concentration fluctuations 5.5.1 Kinetic equation 5.5.2 Dynamic light scattering 5.5.3 Form birefringence 5.6 Scaling theory—dynamics 5.7 Effective medium theory 5.7.1 Failure of the scalar field description 5.7.2. Hydrodynamic screening 5.7.3 Effective medium theory 5.7.4 Example Appendix 5,1 Transformation to collective coordinates Appendix 5.II Osmotic pressure in concentrated solution Appendix 5.III Perturbation calculation of (R?) References 6 DYNAMICS OF A POLYMER IN A FIXED NETWORK 6.1 Tube model 6.1.1 Tube model in crosslinked systems 6.1.2 Tube model in uncrosslinked systems 6.2 Reptation 129 131 132 135 137 140 140 143 143 147 148 149 152 156 156 158 160 161 161 166 169 172 172 173 174 176 180 182 183 184 188 188 188 189 1916.3 6.4 CONTENTS 6.2.1 Primitive chain 6.2.2 Simple application Reptation dynamics 6.3.1 Stochastic equation for reptation dynamics 6.3.2 Segmental motion 6.3.3 Correlation function of the tangent vectors 6.3.4 Dynamic structure factor 6.3.5 General time correlation function Contour length fluctuation 6.4.1 Statistical distribution of the contour length 6.4.2 Dynamics of the contour length fluctuation 6.4.3 Effect of the contour length fluctuation on reptation 6.4.4 Other small-scale fluctuations and their effects on the segmental motion 6.4.5 Branched polymers Appendix 6.1 Entropy of a polymer in a tube References 7 MOLECULAR THEORY FOR THE VISCOELASTICITY OF POLYMERIC LIQUIDS 71 72 73 74 VS 76 17 Tube model in concentrate¢| solutions and melts Microscopic expression for the stress tensor 7.2.1 Stress in polymeric liquids 7.2.2. Stress optical law Linear viscoelasticity 7.3.1 Background of phenomenological theory 7.3.2 Calculation by Rouse model 7.3.3 Calculation by reptation model 7.3.4 Comparison with experiments 7.3.5 Tube diameter in melts 7.3.6 Semidilute and concentrated solutions Other relaxation modes 7.4.1 Discrepancy between the theory and experiments 7.4.2. Contour length fluctuation and tube reorganization Stress relaxation after large step strain 7.5.1 Experimental setup 7.5.2 Calculation by Rouse model 7.5.3 Calculation by reptation model 7.5.4 Comparison with experimental results 7.5.5 Discussion Nonlinear viscoelasticity 7.6.1 Phenomena of nonlinear viscoelasticity 7.6.2 Deformation gradient tensor 7.6.3 Constitutive equation derived from Rouse model Approximate constitutive equation for reptation model 191 193 218 218 22278 19 CONTENTS 7.1.1 Deformation of the primitive chain 7.1.2 Independent alignment approximation 7.7.3 Constitutive equation 7.7.4 Comparison with experiments 7.7.5 Discussion Stress relaxation after double step strain Rigorous constitutive equation for reptation model 7.10 Further applications 7.10.1 Branched polymers 7.10.2 Molecular weight distribution 7.10.3 Future problems References 8 DILUTE SOLUTIONS OF RIGID RODLIKE POLYMERS 8.1 8.2 8.3 8.4 8.5 8.6 8.7 8.8 Rodlike polymers Rotational diffusion 8.2.1 Rotational Brownian motion 8.2.2 Hydrodynamics of rotational motion 8.2.3 Smoluchowski equation for rotational motion Translational diffusion 8.3.1 Hydrodynamics of translational motion 8.3.2 Smoluchowski equation including both translational and rotational diffusion Brownian motion in the equilibrium state 8.4.1 Vector correlation function (u(t) + u(0)) 8.4.2. Translational diffusion 8.4.3 Dynamic light scattering Orientation by an electric field 8.5.1 The effect of an electric field 8.5.2 Dielectric relaxation 8.5.3 Electric birefringence Linear viscoelasticity 8.6.1 Expression for the stress tensor 8.6.2 Calculation for weak velocity gradient Nonlinear viscoelasticity 8.7.1 Decoupling approximation 8.7.2 Elongational flow 8.7.3 Shear flow Effect of flexibility Appendix 8.1 Derivation of the Smoluchowski equation by the Kirkwood theory References 261 262 270 274 278 278 282 289 BW WWW Www RRR YVYwRNYYY PReAA ASS SSS SSES BESS 288 s288seses BBRLSs 318 322xii CONTENTS 9 SEMIDILUTE SOLUTIONS OF RIGID RODLIKE POLYMERS 9.1 Semidilute and concentrated solutions of rodlike polymers 9.2 Entanglement effect in rodlike polymers 9.2.1 Tube model 9.2.2 Translational diffusion 9.2.3 Rotational diffusion 9.2.4 Estimation of the tube radius and the rotational diffusion constant 9.3 Brownian motion in equilibrium 9.3.1 Time correlation functions 9.3.2 Dynamic light scattering 9.4 Orientation by external fields 9.4.1 Linear regime 9.4.2 Nonlinear regime—tube dilation 9.4.3 Experimental study of the rotational diffusion constant 9.5 Viscoelasticity 9.5.1 Expression for the stress tensor 9.5.2 Linear viscoelasticity 9.5.3 Nonlinear viscoelasticity 9.6 Short time-scale motion 9.6.1 Chopstick model 9.6.2 Local equilibrium approximation 9.6.3 Example Appendix 9.1 Tube dilation by orientational ordering Appendix 9.II Effective potential for the tube References 10 CONCENTRATED SOLUTIONS OF RIGID RODLIKE POLYMERS . 10.1 Introduction 10.2 The phase transition of rigid rods 10.2.1 Free energy for a given orientational distribution function 10.2.2 Equilibrium distribution 10.3 The kinetic equation 10.3.1 Dynamical mean field theory 10.3.2 Relaxation of the order parameter 10.4 Pretransitional phenomena 10.4.1 Introduction 10.4.2 Magnetic birefringence 324 324 326 326 327 327 328 330 330 332 333 333 334 335 336 336 337 339 340 342 343 345 346 348 350 350 351 351 354 358 358 362 362 363CONTENTS 10.4.3 Viscoelasticity 10.5 Linear viscosity in the nematic phase 10.5.1 Introduction 10.5.2 Perturbation scheme 10.5.3. Approximate calculation 10.5.4 Example—shear flow 10.5.5 The Leslie coefficients 10.6 Future problems 10.6.1 Nonlinear viscoelasticity in nematics 10.6.2 Spatial inhomogeneity and domain structure 10.6.3 Thermotropic liquid crystals References SUBJECT INDEX AUTHOR INDEX 374 376 376 378 379 381 3861 INTRODUCTION Macromolecules play a central role in chemical technology and indeed in biology. Their role and the richness of their properties mean that a whole series of monographs would be required to constitute a comprehensive treatise. This book concentrates on one aspect, the dynamics of polymers in the liquid state. That is, the dynamics of polymer solutions and melts, where in the last decade it has become possible to offer theories which explain the salient features of these systems. Among various dynamical properties of polymeric liquids, an impor- tant and conspicuous property is their mechanics. As one knows in everyday life, polymeric liquids (chewing gum, dough, or egg white, for example) show quite distinct flow behaviours from the usual liquids like water: a polymeric liquid is usually quite viscous and has visible elasticity. For example if one stretches chewing gum and releases it quickly, then it will shrink like rubber, yet chewing gum is a liquid and can fill a container of any shape. This property, called the viscoelasticity, is just one of the many distinctive properties of polymeric liquids. The purpose of this book is to understand such properties from the molecular point of view. Given the complexity of the polymer molecules, the theories are astonishingly simple, and before embarking on the main text it is worth explaining why it is reasonable to attack the problem of interacting macromolecules with a confidence which would not be justified for say liquid benzene, let alone water. Polymer molecules are formed when the condition required to add one chemical unit (monomeric unit) to a system is almost independent of its size. If one has a small hydrocarbon, the synthesis of the next homologue can be a distinct process from the one after that, i.e., to go from (A) to (A) — (A) involves a rather different pathway than going from (A) — (A) to (A) — (A) — (A). But if one starts with (A) — (A) — «+ -(A), (An, Say), the energy required for A,—>A,4, is almost identical to that of An+i1—Aj,+2, 80 that the process continues (polymerization reaction) and in principle can go to indefinitely long chains. Mixtures of species will often polymerize together: (A) - (A) - (8) - (4) - (4) - (B) - (8) - @) and special agents can branch polymers -~(4)- (A)-(4)-() (A)-() O<(4)- (4-4)2 INTRODUCTION @ / @ (b) ) cl) Fig. 1.1. (a) Flexible polymer, (6) rodiike polymer, (c), (d) their concentrated solutions. An enormous number of variants are possible, are formed, and can indeed be designed. There are two simple cases and these are the two studied in this book: (a) highly flexible polymers, where, on a sufficiently large scale, the polymer appears as a random walk, or a spaghetti-like shape (Fig. 1.12), and (b) a rigid rod (Fig. 1.1b). These are, of course, extreme cases and cases intermediate to (a) and (b) can be found. Now consider a closely packed assembly of these molecules (see Fig. 1.1¢ and d). At a sufficiently high temperature, they are in a high state of thermal agitation and form a viscoelastic liquid. (The reader will have a good intuitive feeling for the macroscopic behaviour of a rubbery liquid or a syrup.) Obviously the systems are quite complicated, but there are reasons why we can say that such systems are easier to understand than normal liquids. One reason is the multiplicity of the interactions: one polymer is simultaneously interacting with many others, perhaps hundreds. Each one of these interactions has only a small effect so that a sound startingINTRODUCTION 3 point is to add up their effects independently. This is in contrast with a normal liquid where each molecule has only a few, say eight or ten, neighbours and it is quite invalid to believe that these neighbours can move independently of one another. For polymeric liquids, any polymer molecule experiences an average of its surrounding and the problem reduces to deducing the mean properties from the behaviour of individual molecules. This task is made (comparatively) easy because of the essential feature of polymers, i.e., the molecule itself is very large and the macroscopic behaviour is dominated by this large scale property of the molecule. Let us give a simple example of this. Suppose a single molecule sits in solution and contributes to the viscosity of a surrounding fluid. Let us compare two polymers of the same size and conformation, but one is composed of spherical segments and the other is composed of triangular segments (see Fig. 1.2a and b). If the segments are separated from each other as in Fig. 1.2c and d, the shape of the segments matters; the viscosities of (c) and (d) are different. However, if the segments are connected to form a large polymer, the C) ) 0 fg A 000 Vv v 0° 10° 2922000000056, p=0,30 gm/ml 2220000000 e00ce00%006 eo 10 1 L Lt 1 107 10" 10° 10° 10? 10° x Shear rate (sec) Fig. 1.3. Logarithmic plot of non-Newtonian viscosity against shear rate for solutions of polystyrene, M,=4.11%10° in n-butyl benzene with various concentrations as shown. Reproduced from Graessley, Hazleton and Lindeman, Trans. Soc. Rheol. 11, 267 (1967).INTRODUCTION 5 + at 4 1 L 1 0.01 O41 1.0 10.0 100.0 Kt, Fig. 1.4. Reduced plot of the non-Newtonian viscosity shown in Fig. 1.3, where No is the viscosity at zero shear rate, and the time constant 1, is chosen empirically for each solution. Reproduced from W. W. Graessley Adv. Polym. Sci. 16, 1 (1974). temperature, and chemical species. The ability of such superimposition indicates the existence of an inherent simplicity hidden behind the apparent complexity of the polymer systems. It is this feature that we would like to discuss in this book. Now, in concentrated solutions such as shown in Fig. 1.1¢ and d, the key concept which allows us to express the relationship of the behaviour of a single molecule to the average behaviour of its neighbours is that of a surrounding tube. This is clearly specified for rodlike polymers. In order to draw a picture and without loss of generality, let us consider the freedom of a rod to move in a plane containing itself. Other molecules are represented in this picture by dots. Then in a high density of rods, the Picture is as shown in Fig. 1.5. Suppose other rods are fixed for the moment. The rod is free to move along itself, but rotations and displacement perpendicular to its length Permit it to move only in the shaded region; i.e., our rod is confined to a tube as far as rotations and lateral motions are concerned. Of course the tube is made of other rods which are diffusing, but for a high enough density this is not consequential. Thus what we must understand is how a d moves in a fixed, or a very slowly diffusing tube made of its Neighbours. In the case of rodlike polymers it is easy to calculate the size of the tube, but it is not so easy for flexible polymers where the picture of Fig.6 INTRODUCTION Fig. 1.5. Tube for a rod. 1.2 has to be replaced by something like Fig. 1.6. It is clear that in this case the polymer can go for extensive ‘excursions’ (Fig. 1.7) so that the tube is a fuzzy concept. It is also difficult to calculate the effective radius and the step length of the tube. However, by just calling it a, and assuming that a is a certain function of concentration, one can predict various dynamical properties in the molten state. At this stage, however, the reader has to be warned that the ‘fine structure’ of the behaviour of real polymers is by no means resolved. The contents of this book will produce answers to many questions, and these answers will have qualitatively the right forms. But they may not be correct in detail. For example, the viscosity of a monodisperse melt of flexible polymers is shown to be proportional to M? whereas experimen- tally it is known to be M**. Possible explanations of the discrepancy are discussed, but the problem is not entirely resolved. Thus the authors believe they are surveying real progress, but by no means the whole Fig. 1.6. Tube for a flexible polymer.INTRODUCTION 7 Fig. 1.7. Excursion of a flexible polymer. story, and have tried in this book to make it clear where approximations are being made and to give a reasoned assessment of their validity. This book is oriented to one aspect of polymer theory, but in order to reach it the authors have felt it necessary to give brief accounts of the basis of the subject in discussions of single chains and of polymer solutions. These topics too have made major advances in the last twenty years and deserve books of their own, but can only be given a minimal treatment in this book.2 STATIC PROPERTIES OF POLYMERS 2.1 The random flight model Flexible polymers can take up an enormous number of configurations by the rotation of chemical bonds. The shape of the polymers can therefore only be usefully described statistically. In this chapter we shall study the statistical properties of a single polymer in the equilibrium state. 2.1.1 The freely jointed model To study the statistical properties of flexible polymers, let us start from a very simple model: a chain consisting of N links, each of length by and able to point in any direction independently of each other (see Fig. 2.1). Such a model is called the freely jointed chain. The conformation of the freely jointed chain is represented by the set of (N +1) position vectors {R,,} =(Ro...Rwy) of the joints, or alterna- tively by the set of bond vectors {7,} =(n...1v), where mR, —Ry-1, 2 =1,2,...,M. (2.1) Since the bond vectors 7, are independent of each other, the distribution function for the polymer conformation is written as N W({r,}) = Ut vn) (2.2) where p(r) denotes the random distribution of a vector of constant length bg: 1 > v= nb 5((r| — bo). (2.3) This distribution is normalized to | dry(r) =1. (2.4) To characterize the size of a polymer, we consider the end-to-end vector R of the chain, N R=Ry-Ro= > m (2.5) n=l Since (7,) =0, (R) is zero, but (R?) has a finite value, which can beTHE RANDOM FLIGHT MODEL 9 tw n Fig. 2.1. Freely jointed chain, used as a characteristic length of the chain. Let R be defined by R= (R?)'? = (Ry — Ro))”. (2.6) From eqn (2.5) (R?) is given by y= Sista) = 3 (A)4+2.D (arm) =NOE ON) nme because for n#m, (I, * tm) =4tm)*(Im) =0. Thus R is given by R=VN bo. (2.8) 21.2 General random flight models Though the freely jointed chain is a very simple model, the result (R’)«N holds for more general models. Consider for example the model shown in Fig. 2.2, called the freely rotating chain, in which the n-th bond is connected to the (n — 1)-th bond with a fixed angle @ and can rotate freely around the (n — 1)-th bond. - For such a model, (r,-7,) does not vanish for n +m. However, (r+ Fa) decreases rapidly as |z—m| increases, and the relationship (R*) « N again holds for large N. To see this we calculate (7, * tm). If the average of r, is taken with the rest of the chain (i.e., fm, Mm+is +--+ > M1) fixed (n > m being assumed), we obtain (see Fig. 2.2b) (Fn) tpotmetstestized = COS OF, 1. (2.9) Multiplying both sides of this equation by 7, and taking the average over10 STATIC PROPERTIES OF POLYMERS (b) Fig. 2.2. (a) Freely rotating chain. (b) The average of r, with r,_, fixed gives Tr—1 00S 8. Tm» Tm+is ++ +> Tnr—1» We have (Iq Im) = COS O{Fy-1* Im) (2.10) This recursion equation, with the initial condition (r7,) = 53, is solved by (i Tm) = bcos 0)" (2.11) which decreases exponentially with |n — m|. Thus for large N, (R?) is given by (R’) = s > (in Tn) = > > (iy Task) => 3 (aera). 12) n=l k=—0 From eqn (2.11) 140086 SS ecteak) = bi(1+2 5, costa) = BG. 3) pole Hence (R?) = = Nj 88 (2.14) cos 8” which again shows (R?) «N.THE RANDOM FLIGHT MODEL 11 In general if the distribution function of r, is written in the following form HCCI) = TT We teats inva «+ Fatma (2.15) (R’) is written, for large N, as (R?) = Nb?. (2.16) The constant b is called the effective bond length. The ratio C= b7/b5 (2.17) represents the stiffnesst of the polymer and can be calculated from the local structure of the chains. The results are summarized in refs 1-3. 2.1.3 Distribution of the end-to-end vector Next we consider the statistical distribution of the end-to-end vector of the random flight model. Let @(R, N) be the probability distribution function that the end-to-end vector of the chain consisting of N links is R. Given the conformational distribution for Y({r,}), ®(R, N) is calculated by N O(R, N)= far, J dr... [dry 5(R—- Sm) ¥((ra}), 18) nal which is rewritten using the “my a(n)= om | dke*", (2.19) as ei OR, W)= Gass Jak an far. : - fry . N xexp(ik-(R— Bm) Cn}: @20) For the freely jointed chain, eqns (2.2) and (2.20) give 1 , ra (RN) = Gs fake fdr... dry TT exr(—ik- 1) ¥(0) 1 \N “ay | dke**| | dr exp(—ik - nwt] . (2.21) 38 the aitrature, the stiffness is often represented by the Kuhn statistical segment length bx = (R?)/Raax Where Rugs is the maximum length of the end-to-end vector.12 STATIC PROPERTIES OF POLYMERS The integral over r is evaluated by introducing polar coordinates (r, 0, 3 the reference axis of 6 being taken along the vector k), giving for exp(—ik - r)p(r) - alo [oe [eosin @ exp(—ikr cos 6) 5(r — b) sin kb kb kb where k =|k| and bg is replaced by b (since by=b for a freely jointed chain). From eqns (2.21) and (2.22) (2.22) @(R, N)= oF i acs)” If N is large, ((sin kb)/kb)” becomes very small unless kb is small. For kb &1, ((sin kb)/kb)% can be approximated as (1282) = (1 2) mene - kb 6) OP This approximation holds also for kb =1 since both sides of eqn (2.24) are nearly zero in such a case. Thus ®(R, N) is calculated as (2.23) Nk?’ (2.24) OR, N= a 1; [ake exp( -™ Ne . (2.25) The integral over k is a standard Gaussian integral. (Some useful Gaussian integrals are summarized in Appendix 2.I.) If k, and Ra (@ =x, y, z) denote the components of the vectors k and R, then using | eqn (2.1.2) i ®(R, N)=(2x)? TI J l i dk, exp(ikyRx —Ni26*16)| | “09° ($3) ea -abo') | = (3/2nNb2)"” exp( - 3%) . (2.26) Thus the distribution function of the end-to-end vector is Gaussian. The distribution (2.26) has the unrealistic feature that |R| can be larger than the maximum extended length Nb of the chain. A more realisticTHE RANDOM FLIGHT MODEL 13 Fig. 2.3. A chain divided into N submolecules. distribution function which does not have this feature is available in the literature.'?* In this book, however, such highly extended states of polymers are not considered and egn (2.26) is sufficient for our purpose. Although the above derivation is for the freely jointed chain, the result actually holds more generally. In general it can be shown that provided the conformational distribution is described by eqn (2.15), the distribu- tion of the end-to-end vector R of a long chain (N>>1) is given by eqn (2.26). This is a result of the central limit theorem in statistics.* To prove this, suppose that the chain is divided into N submolecules, each consisting of 4 links (see Fig. 2.3). Clearly N=NAA. (2.27) Let 7, be the end-to-end vector of the n-th submolecule. The end-to-end vector of the chain is written as Ww R=> i, (2.28) n=l Now if N is very large, both N and A can be taken large enough so that N>1, and A>>1. If 4>>1, the vectors #, become independent of each Other, and the distribution of {7,} can be written as N wm = TT OG). (2.29) From eqns (2.28) and (2.29), the distribution of R is derived without ‘owing the actual form of (7). Again using the identities (2.21) and (2.29), we have ®(R, N)= oy f dke®®A(ky* (2.30)14 STATIC PROPERTIES OF POLYMERS where A(k) = I dr exp(—ik + r)0(0). (2.31) Since #(r) depends only on |r|, the integral over the direction of r can be carried out as 2 o« 2x A(k) = [ore [ao sin @ [as exp(—ikr cos 0)p(r) oO 0 0 - J drdnr? “o a b= (= r) (2.32) where . (. dom [ard G0), Le (2.33) 0 Since we are interested in the small & region, we may evaluate A(k) by expanding (sin kr)/kr with respect to k A(k) = (1-2? +...) g=1- HRP) 5. (2.34) For A4>>1, (r*), is given by Ab?, whence A(k) =1- 34D. (2.35) Thus ®(R, N) is evaluated as 1 F < ® -<— ik. 19 ,2,.2\0 (R, N) p fate (1 — 267k?) = aay [exe exp(—3ANb7k?) -(came) ov(-aae) G0 Since AN = N, eqn (2.36) agrees with eqn (2.26). 2.2 The Gaussian chain We have seen that, in the statistical distribution of the end-to-end vector, the local structure of the chain appears only through the effective bond length b. This is generally true: the local structure affects only the effective bond length but does not otherwise appear in the problem. Therefore, if we are interested in the global properties of polymers, we can start from the simplest model available.THE GAUSSIAN CHAIN 15 We consider a chain whose bond length has the Gaussian distribution pe 3 y(r) = [=] exp( ~ a) (2.37) (Pr) =b, (2.38) The conformational distribution function of such a chain is given by Ny 3 732 32 won=T Lea] ee al n=l so that Such a chain is called the Gaussian chain. The Gaussian chain does not describe correctly the local structure of the polymer, but does correctly describe the property on large length-scale. The advantage of the Gaussian chain as a model is that it is mathematically much easier to handle than any other of the models considered in Section 2.1. The Gaussian chain is often represented by a mechanical model (see Fig. 2.4): (N + 1) ‘beads’ are considered to be connected by a harmonic spring whose potential rey is given by UMAR) = 335 3 keT s (Ry ~ Raa) (2.40) At equilibrium, the Boltzmann distribution for such a model is exactly the same as eqn (2.39). An important property ‘of the Gaussian chain is that the distribution of the vector R, —R,, between any two units n and m is Gaussian, being given by _ 3 pa 3(R, — Rn)? POR y — Rm mn) = Fea —all Plain moe] 40) Ry R, Fig. 2.4. Gaussian chain.16 STATIC PROPERTIES OF POLYMERS This follows from the properties of the Gaussian integral (see Appendix 2.1). Especially for any n and m (Rn — Rm)?) = | — m| b?. (2.42) The suffix n of the Gaussian chain is often regarded as a continuous variable. In such cases R,, — R,-1 is replaced by OR,,/dn and eqn (2.39) is written as N WIR, ] = const exp| - i i dn( 38)" . (2.43) This distribution is known as the Wiener distribution. Mathematically, there is some subtlety in going from discrete n to continuous n, but for the present purpose it is sufficient to understand that eqn (2.43) is a formal rewriting of eqn (2.39). In this book, we use the discrete n and continuous 7 interchangeably. The transformation tules from the discrete variables to continuous variables are summarized in Table 2.1. Table 2.1 Discrete ‘Continuous kK « > > | dn nak 1 R,—R,-1 > OR,,/On Rai t Rea 2R, > R,/an* Sum > d(n-m) Kronecker delta Dirac delta function 2 . 8 aR, ar, | TIdR, > for. +,§ These symbols refer to the functional derivative and the functional integral respectively. In the usual notation these are written as 6/6R(n) and [ 6R(n) in order to stress that R is a function of the continuous variable n. However, as the discrete representation and continuous representation are used here interchangeably, we use the same symbol R,, in both representations.CHAIN CONFORMATION UNDER AN EXTERNAL FIELD 7 2.3. Chain conformation under an external field 23.1 The Green function If there is an external field U,(r) acting on each segment, the equilibrium distribution of the Gaussian chain is modified by the Boltzmann factor oxo{ ~z2 [emu]. 2.44) and the conformational distribution function becomes WER] exp] 32 ss [a oy it ty fomvsn]. (2.45) To discuss the statistical properties of such a system it is convenient to consider the ‘Green function’ defined “ Ryo fr [om wal The = where Ry=R and Ry=R’ in the functional integral means that the integral is taken for all the conformations which start from R’ and end at R. (Notice that the denominator of eqn (2.46) is independent of R and R') For U,=0, G(R, R'; N) reduces to the Gaussian distribution function bay. (20NB?\ 32 3(R - RY’) GR-R sN)=( ; ) exp( - OME): 47) In the general case of U,+0, G(R, R';N) represents the statistical Weight (or the partition function) of the chain which starts from R’ and ends at R in N steps. The partition function for all possible conformations is given by G(R, R';N)= Ze I dR dR'G(R, R'sN). (2.48) From the definition of G(R, R’; N), the following identity holds: G(R, R'; N) = | dR’G(R, R’; N—n)G(R", R’;n), (for 0m being assumed), then (A(Rn, Rm)) = [ein dk, AR, ARG Ry, Rs N—n)G(Ry, Rim) i X G Rms Roi m)A(Rns Rn) x| [ARy dROG(Ry, RosN)]—@.51) Though the Green function G(R, R’; N) has a physical meaning only na N>O0, it is convenient to define G(R, R'; N) for N <0 in such a way at G(R, R';N)=0 for N> L,, Ly, L..» 20 STATIC PROPERTIES OF POLYMERS The partition function is then given by Z= [aR aR’G(R, R';N) = Z,2,2, 2.59) with L Z= f dk, f dRigx(Re» Rz; N) 8, ¥ a? NB? p? HL, = -60) mee on - 6 Zl - 2.60) The free energy of the system is calculated by A=—kgT InZ, (2.61) so that the pressure acting on the wall normal to the x axis is 1 3A = > 2.62, Pe L,L, aL, (2.62) Let us consider the two limiting cases: (i) In the case when the polymer is much smaller than the box (i.e., VNb«L,, Ly, Lz), Z, is approximated by 8 1 Zeal, alk, 2.63 a pbeaP? 2-63) whence __keT _keT kT P= LLL LL, =F and similarly P,=P,= vv? (2.64) which is simply the equation of state of an ideal gas. (ii) In the case when the polymer is much larger than the box (i.e., VNb>L,, L,, L,), Z, is dominated by the first term in the sum 8 w?Nb? 2,-3.1, exp| -722 |, (2.68) which yields ___1 0A =I SIND) NY bal Re L,L,8L, V 1 312) 312 Vv (2.66) Therefore, in this case P, is not equal to P, and P, (unless L, = L, = L,), i.e., the force acting on the wall is not isotropic. Later it will be shown that this anisotropy in the pressure due to the anisotropic confinement is responsible for a variety of mechanical properties peculiar to polymers. The Green function method has been applied in various problemsSCATTERING FUNCTION 21 related to polymers in nonhomogeneous situations such as adsorption of polymers, polymers near the phase boundary, polymers in a porous media, etc. Many such examples can be found in the literature.*’° 2.4 Scattering function The size of a polymer can be measured by various scattering experiments (light scattering, small angle X-ray scattering and neutron scattering). The principles and the practical details of such scattering experiments are given in refs 11-13. Suppose that the polymer is modelled by a series of units at each R, which have a scattering amplitude a,. Then the scattering intensity at the scattering vector k =k, —k; (k; and ky being the wave vectors of the incident and the scattered beam) is written as N Dane explik + (R, —Rn)] (2.67) nm=1 where N, is the number of scattering units in the system and the sum is taken over all the scattering units in the system. In particular, for a homogeneous polymer one has M la? 2. exp[ik - (R, — Rn)]. (2.68) This is proportional to MN, since the sum )™_, exp[ik-(R,—R,,)] Temains finite for No. We shall use the structure factor g(k) defined by No atk) =, ©, (explik (Ry ~ Ry) (2.6) which is independent of the system size (or Ny). lt the polymer solution is sufficiently dilute, the interference among different polymers can be neglected, so that eqn (2.69) is written as idk . ak) =x > (explik « (R, — Rn) (2.70) nm Where N is, as before, the number of segments constituting the polymer. Since the direction of R,,— R,, is random, eqn (2.70) can be rewritten as Mn eqn (2.32), sin(\k| IR, = Ral) ; 71) 1 8-5 (ae oR22 STATIC PROPERTIES OF POLYMERS The characteristic size of the polymer is obtained from g(k) in the small k region. Expanding eqn (2.71) for small k, we have 1 a(k)=5 D (1- 3K, — Rn) +.) =N(1-4PR2+...) (2.72) where R, is defined by Ree =n > & (Bn Rn) (2.73) which represents the mean square length between all the pairs of the segments in the chain. The quantity R2 is called the mean square radius of gyration since it can be rewritten as 1 N Rea yy (Rn Roy) (2.74) n=l where Rg is the position of the centre of mass of the chain: = is R,. (2.75) Nya The derivation of eqn €2.74) is straightforward: substitution of eqn (2.75) into eqn (2.74) gives R}= > (RI- 2R,, ‘Rg + RZ) = (Ri oR. DR 5 Rn -R) = (jERI~ jp Re Re) = 52D (R—Ro?)- (2.76) In general, the size of the polymer is more appropriately represented by R, than R (where R? = (R’)) because R, is well defined for branched polymers while R is not. For a linear Gaussian polymer R, is easily calculated. From eqns (2.42)SCATTERING FUNCTION 23 and (2.73) 5 iMz iMz ln m6 an fom n— m| b? = fen sen (n— mb? = ANB? 277) oO 0 Thus R, is given by R/V6 for the linear polymer. Let us now calculate the structure factor for the Gaussian chain. Since the distribution of R, —R,, is Gaussian, (exp[ik + (R,—R,,)]) is calcu- lated using eqn (2.1.20) in the Appendix as (explik (Ry Ryd) =(exp[_S_ikRan ~ Rre)]) =exp| - BD, FMRne ~ Raat?) | . (2.78) Since ((Rma — Rna)*) = |m — n| b?/3, we get (explik (R,-Rp)}) =exp[ == nmi]. 2.79) Hence the structure factor is given by gtk) = 1 Tan [om exp -Ein—mi]=afceR) 2.0) where o 98 fis)=3(e*-142) 81) which is called the Debye function. The asymptotic form of g(k) is _ {N(1—JPR2/3) for |k|Ry«1 8) = ONner? for |k|R,>1. (2.82) convenience of calculation, the Debye function is often approximated N 8(k) “TF RR : (2.83) The error of this equation is less than 15% for the entire region of k.a4 STATIC PROPERTIES OF POLYMERS 2.5 Excluded volume effect 2.5.1 Introduction In the models of polymers considered in the previous sections, the interaction among the polymer segments is limited to within a few neighbours along the chain. In reality, however, segments distant along the chain do interact if they come close to each other in space. An obvious interaction is the steric effect: since the segment has finite volume, other segments cannot come into its own region. This interaction swells the polymer; the coil size of a chain with such an interaction is larger than that of the ideal chain which has no such interaction. Even when there are attractive forces, as long as the repulsive force dominates, the polymer will swell. This effect is called the excluded volume effect. The excluded volume effect represents the effect of the interaction between segments which are far apart along the chain (see Fig. 2.7). Such an interaction is often called the ‘long range interaction’ in contrast to the ‘short range interaction’ which represents the interaction among a few neighbouring segments and is included in p(n,..., m+n.) in eqn (2.15). (Note that the terms ‘long’ and ‘short’ represent the distance along the chain, not the spatial separation.) The excluded volume effect was first discussed by Kuhn,’* and the modem development was initiated by Flory.""* It had been recognized by these pioneers that the long range interaction changes the statistical property of the chain entirely. For example, (R?) is no longer propor- tional to N but to a higher power of N (R?) « N?”, (2.84) The exponent v is about 3/5, so that the excluded volume effect is very important for long chains. Once the long range interaction is introduced, exact calculation: becomes impossible. A great deal of work has been done on this: Fig. 2.7. Excluded volume interaction.EXCLUDED VOLUME EFFECT 25 problem, and a detailed description is given in various literature.*7-17-'8 Here we shall outline only a few typical approaches. 2.5.2 Model of the excluded volume chain In real polymers, the nature of the long range interaction is quite complicated: the interaction will include steric effects, van der Waals attraction, and also may involve other specific interactions mediated by solvent molecules. However, as far as the property of large length scale is concerned, the detail of the interaction will not matter because the excluded volume effect is controlled by the interaction between distant parts of the chain. Thus the interaction between the polymer segments n and m can be expressed by a short range function kp TO(R, — Rn) (2.85) which can usually be approximated even further to a delta function’ ukgT 6(R, — Rn) (2.86) where v is the excluded volume and has the dimension of volume.t The total interaction energy is thus written as U; =foksT [dn fom 6(R, ~R,)- (2.87) oO 0 Using the local concentration of the segments cr) = > 6(r-R,) = [an 6(r-R,), (2.88) ” ° eqn (2.87) may be rewritten U,= | driukpTe(ry’. (2.89) This expression indicates that eqn (2.87) is the first term in the virial expansion of the free energy with respect to the local concentration c(r). Therefore the excluded volume parameter v can be regarded as the virial coefficient between the segments. In principle the virial expansion can be continued to include higher + The reader may be worried by the use of the delta function which diverges when R,, = R,,- It turns out, however, that all physically measurable quantities involve integrals of this potential and are well behaved. (The absolute entropy of the chain does diverge, but the entropy difference, which matters experimentally, is well behaved also.) This situation is familiar in quantum field theories and is discussed in great detail in the textbooks. The reader can be assured that this point causes no real difficulties.26 STATIC PROPERTIES OF POLYMERS order terms such as U= | dr[duk,Te(r)? + dwkpTe(r) +. .J- (2.90) However, the higher order terms may be neglected since the segment density inside the polymer coil is small: the segment density is estimated as é =e N'-*=N-*“S (when v = 3/5) (2.91) which becomes very small for large N. Therefore the essential features of the excluded volume effect can be studied using the potential given by eqn (2.87). For a given combination of polymer and solvent, v varies with temperature and can be zero at a certain temperature, called the © or Flory temperature. At the © temperature, the chain becomes nearly ideal.¢ An appropriate expression for the temperature dependence of v may be obtained as follows. Suppose that the interaction between the segments is expressed by a potential energy u(r) which depends only on their separation r. Then the second virial coefficient is evaluated by the standard formula for an imperfect gas” v= far[1-exp(-72)] . (2.92) Usually u(r) consists of a hard core potential u,,.a(7) and a weak attractive potential u,..(r) (see Fig. 2.8). In such a case, v is estimated as B = far{ ero — te) (1-4) 4-2 @ . fools OP kgT )e: er) A-p 9%) where A and B are constants independent of temperature. For this model, the © temperature is defined as B/A, and eqn (2.93) may be written as'® ° v=v(1-$). (2.94) Now if the interaction (eqn (2.87)) is taken into account, the distribution function of R,, becomes WIR, x ero| 33 [an 28) - ju | an] a(R, -R,,)| . (2.95) 0° oO oO + Even at the © temperature the chain is not ideal since there is a three-body collision term.”?! However the effect of the three-body collision is quite weak and gives only a ic correction to (R?).EXCLUDED VOLUME EFFECT 27 u(r) Yard) Yard) IL iL LH v r 7 @ ) © Fig. 2.8. A sketch of a potential. The potential can be decomposed into a hard core potential uy.,a(r) and a weak attractive potential u,,,(r). This model includes only two parameters; b, which represents the short range interaction, and v, which represents the long range interaction. The basic assumption of this two-parameter model is that there is a sharp distinction between the short range interaction and the long range interaction. Though the validity of this assumption can be questioned,” it is generally believed to be a correct starting point for the analysis of the excluded volume problem. 2.5.3. Theoretical approaches We shall now discuss the statistical properties of the chain of eqn (2.95). We shall limit the discussion to the case of v >0, i.e., the case of repulsive interation. The cases of v = 0 and v <0 are discussed in refs 24 and 25. A simple theory. The original idea of Flory’® for calculating the size of a polymer is to consider the balance of two effects: a repulsive excluded volume interaction which tends to swell the polymer, and the elastic energy arising from the chain connectivity which tends to shrink the polymer. This idea can be put into a particularly simple form of theory.”° Consider the free energy of a chain whose end-to-end vector is fixed at R. This is given by A(R) =—kgT In ®(R, N) + terms independent of R (2.96) because the equilibrium distribution function is proportional to exp(—A/ kgT). If v=0, the free energy A(R) is obtained from eqn (2.26) as 2. A(R) = kpT- x + terms independent of R. (2.97) To estimate the effect of the excluded volume interaction, we disregard the connectivity of the chain, and calculate the interaction energy of a28 STATIC PROPERTIES OF POLYMERS ‘segment gas’ confined in a volume R? (R = |R\). Since the concentration of the segment gas is ¢~ N/R, the interaction energy is estimated as vk T@R°. (Here the numerical factor is disregarded at this level of approximation.) Thus 3R? + w) 2b?” RS) The average size of the polymer can be estimated from the value of R which minimizes A. From 0A/3R =0, we have R ~vno( This gives the exponent v=3/5 already quoted, which is close to the experimental value.” Though the exponent v is close to the experimental value, the prediction of this theory for other quantities turns out to be inadequate: for example, the expression for the free energy (eqn (2.98)) is not consistent with the distribution function of the end-to-end vector ob- tained by computer simulation.” Also it suffers from an unreasonable behaviour of the entropy. Because of the very simple structure of the theory, the theory is generally regarded as a prototype of a mean field theory. It must, however, be realized that it is not a real mean field theory in that no mean field has been calculated. The mean field theory will be discussed later on. AR)=kT (35 (2.98) yrs oN, (2.99) Perturbation calculation. If the excluded volume v is small, the distribu- tion function (2.95) can be expanded with respect to v.*°3! Hence (R?) can be calculated as a power series of v. The calculation of the first term is given in Appendix 2.III. Though straightforward, such a calculation becomes quite tedious. The latest result® i (R?) = Nb*(1 + 4z — 2.0752? + 6.29723 — 25.057z4 +116.13525 — 594.7172+...) (2.100) where the expansion parameter is defined by = (3)" ne . (2.101) Note that z is proportional to VN. Thus the perturbation expansion becomes useless for large N. This situation is entirely different from the virial expansion of the imperfect gas, where the expansion parameter is the average concentration 2, which is proportional to N*~*” and is veryEXCLUDED VOLUME EFFECT 29 small in the case of polymer problems. The reason for this difference is that for polymers, the collision between any pair of segments affects the end-to-end vector seriously (the effect being of order 1), whilst in imperfect gases the effect is only of order of 1/N. The perturbation calculation in the polymer problem is justified only when none of the polymer segments is likely to collide with the other segments. Thus what must be small is the average number of collisions taking place in the whole polymer coil: N vVN Nue~ Nuys Be? (2.102) which is z. It has been shown that the series (2.100) is asymptotic, and suffers from an explosive increase in its coefficients which increase roughly like n". However, by an appropriate resummation technique, useful infor- mation can be drawn about the asymptotic behaviour of large z.°>°7 In particular, the exponent v is estimated as*® v =0.588 + 0.001. (2.103) Uniform expansion model. The difficulty in the perturbation calculation can be improved by a simple scheme of calculation.” In this scheme, it is assumed that the expansion of the chain is represented by the expansion of the bond length, i.e., that the distribution function of R, is well approximated by N , 3 aR,,\? was ero| ~ 5 | an( SH) | e108 where b' is the expanded bond length to be determined. Let (...)' be the average for this distribution function. The average (R?) for the distribution function (2.95) is written as (2) = (eSB (—BIR AD (Ry = Roy") (exp(-BIR,D)" (2.105) where N aR, 2 N N B[R,] =3(6-?- bi) [ dn(S) +4u [ dn fam 8(R,-R,,). (2.106) o 0 0 Now if eqn (2.104) is a good approximation to eqn (2.95), B[R,] can be regarded as small. Hence eqn (2.105) is evaluated as _ _ 27 (Rn?) = {G= BURa) (Rw —Ro)")" se a (2.107)30 STATIC PROPERTIES OF POLYMERS The average is evaluated straightforwardly (see Appendix 2.111) as (R?) = (Ry — Ro)’)' — (BIRn|(Rw — Ro)”)' + (Rw — Ro)”)'{BERn})’ 6” 4b? = 12 Ce] a = Nb’?+Nb Ge 1-353 z) (2.108) where z is a parameter defined by eqn (2.101). Now if eqn (2.104) is the best approximation of eqn (2.95), the first-order correction (the under- lined part) must vanish. This condition gives b? 4b° Bl paz 0 (2.109) Let us define a swelling coefficient «: 2 (RY) =r (2.110) Equation (2.108) gives a = b'/b, so that eqn (2.109) is written as a — a? = 42, (2.111) Equations (2.110) and (2.111) determine (R?). For small z, this theory agrees with the result of the first-order perturbation expansion (eqn (2.100)). On the other hand, for large z the theory gives R«N*° in agreement with eqn (2.99). Thus the theory gives an interpolation between the two cases. A warning must be given about the assumption involved in this theory. Although eqn (2.104) gives a good approximation for (R7), it gives erroneous predictions for other quantities; for example eqn (2.104) wrongly predicts that the distribution of R is Gaussian. In general the optimum choice of b’ depends on the quantity under consideration: to calculate a certain quantity A[R,], we expand (A[R,,]) as (A[R,]) = (ALR, 1)’ — (BER, 14[Rn])’ + (BERn])' (ATR a1)! (2.112) and choose the parameter b’ so that the underlined term vanishes. This prescription works well. For example, the distribution function of the end-to-end vector is calculated as*” 2 exp| -i ) | for R=R, “el -@°3G)"] oe eeEXCLUDED VOLUME EFFECT 31 This function has been found by Domb et al.” in numerical simulation of the problem on a lattice, and has also been discussed theoretically by Fisher.” Mean field theory. The simple theory given in the first part of this section can be improved by mean field calculation.*’ Let us consider an ensemble of chains whose ends are fixed in space, one at the origin and the other at point R. Let c(r) be the average density of the segment: atr)= { dn(ar—R,)) . (2.114) oO Then the segment at r feels the mean field potential ¢(r)vkg7T, whence the statistical distribution of the polymer becomes WyelR,] exp| - 3, | an( 2) -v [ancer| | (2.115) 0 0 For the distribution function, the Green function G(R, 0, 7) is calculated by (see eqn (2.53)) a 8 # [2-744 ve(R) |GR, 0, n) = 8(R) 5(N). 2.116) Given G(R, 0,7), the segment density ¢(r) is calculated by &(r)= farce, r, N—n)G(r, 0,n). (2.117) 1 G(R, 0, N) Equations (2.116) and (2.117) give a closed equation for G(R, 0, N). The detailed calculation within this theory is involved, but it predicts many interesting features of the excluded volume chain. For example, the structure factor g(k) at high k region is shown to be g(k)=k-™ for kR,>1 (2.118) which can also be obtained from the uniform expansion model. Fluctua- tions have been added to this model by Kosmas and Freed,” but the essential features of the result turned out to be the same. Renormalization group theory. The mean field theory neglects the fact that the collisions between the segments in the polymer are strongly correlated: a collision of any pair of segments is likely to induce other collisions since the segments are connected. To take into account such32 STATIC PROPERTIES OF POLYMERS correlation, the renormalization group technique, developed in the study of the critical phenomena, has been applied to the polymer problem. This method, originally invented by Wilson,” was first applied to polymer problems by de Gennes“ and des Cloizeaux.** A particularly clear result is given by Wilson and Fisher® based on the observation that in a space of higher dimensionality, the effect of correlations becomes weak and the simple perturbation expansion becomes applicable. For example, in the excluded volume problem if the polymer is embedded in d-dimensional Euclidean space, the expansion parameter z is N’v/(VNb)¢ «NO, which becomes small for large N if d > 4. (The dimension 4 is called the critical dimension.) It is thus possible to develop an expansion scheme tegarding e = 4~d as an expansion parameter. Such a scheme gives v as va\itge+ e’...). (2.119) For d =3 this gives v = 0.592, which compares well with the more exact result?©** y = 0.588 + 0.001. In any case, the deviation in the indices from 0.6 is quite small. The actual calculation method of the renormalization group theory i is quite complex and has many variations. Details can be found in references 18, 45, 47, and 48. 2.6 Scaling Though the renormalization group method is highly sophisticated, certain conclusions derived from the theory are easy to understand and quite powerful in understanding the nature of the excluded volume chain.’” The basis of the renormalization group theory is to study how physical quantities change when the basic units of the physics are changed. Before explaining this idea for the excluded volume chain, let us first consider the Gaussian chain. As was explained in Section 2.2, the statistical property of the Gaussian chain does not depend on the local structure of the chain. Therefore instead of the original Gaussian chain consisting of N segments of bond length b, we can start from a new Gaussian chain which consists of N’=N/A segments with the bond length VAb (see Fig. 2.9). The transformation from the old chain to the new chain is the change in the parameter NON/A, b->bvA. (2.120) If one knows how a physical quantity changes under this transformation, one can draw conclusions about the dependence of the physical quantity on the parameters N and b.SCALING 33 Fig. 2.9. (a) Original chain and (b) new chain, in which A =2. As an example, consider the size of the Gaussian chain. Various length can be defined to characterize it. For example: (i) Root of the mean square of the end-to-end distance, R = (R?)'? (ii) Root mean square of radius of gyration, Ry = (In (R, ~Ro)/N)? (iii) Mean end-to-end distance (|R]). (iv) The average of the longest distance connecting two beads in the chain, (max,,m|R, —Rml)- The first three quantities are easily calculated, but the last is not. However, without doing any calculation, one can show that all these quantities are proportional to VN b. This can be shown by the following argument: the average size of the polymer, however it is defined, has the dimension of length and must be written as average size = F(N)b. (2.121) The size of the polymer must be invariant under the transformation (2.120), ive., F(N)b = F(N/A)VAB (2.122) which is satisfied only when F(N)b = numerical constant x VN b. (2.123) Thus the distinction among the various lengths is only in the numerical constant. The scaling argument is developed quite generally in statistical mechanics" and indeed historically was the source of the renor- malization group theory. Suppose therefore that a similar property exists4 STATIC PROPERTIES OF POLYMERS for the excluded volume chain: i.e. the excluded volume chain is characterized by N and 5, and its size is independent under the following transformation: N>WN/A, b> ba” (2.124) where v is the exponent in R«N”. By the same argument as above, we can show that the average size of the excluded volume chain must have the following form: average size = numerical constant N’b. (2.125) Equation (2.125) indicates that there is only one length-scale to charac- terize the macroscopic size of the polymer. In particular, the difference between R and R, is only in the numerical factor R, = numerical factor x R = N°. (2.126) According to the renormalization group calculation,*'* the numerical factor is about 0.406 which is close to 1/V6 = 0.40825. In general, under the transformation (2.124), the physical quantity A changes as A> AA. (2.127) The parameter x depends on the nature of A and can be inferred by physical argument. From this property much information can be obtained on the nature of the polymer chain. As an example, consider the structure factor g(k) of an excluded volume chain. From the dimensional analysis, g(k) must be written as 8(k) = F(kb, N). (2.128) Under the transformation, g(k) changes from g(k) to g(k)/A since g(k) is proportional to the number of scattering units N. Thus the function F must satisfy F(kbA”, N/A) = FF kb, N). (2.129) For this to hold for arbitrary A, g(k) must have the following form:+ (k) = NF(kbN"). (2.130) Since N’b « R,, this equation may be rewritten 8(k) = NF(kR,). (2.131) This equation determines the general functional form of the scattering + Here a common symbol F is used to denote various functions. The functional form of F in eqn (2.131) is not the same as in eqn (2.130).GAUSSIAN DISTRIBUTION FUNCTIONS 35 function. Particularly useful information is obtained from eqn (2.131) for the high k region. If KR, >>1, g(k) should be independent of N. This happens only when g(k) is written as g(k) = const N(KN’)"¥” ) AnnX%n%m20 for all x,, (2.1.10) and C is a normalization constant given by C= (det[Ann])’7(22)-*”. (2.1.11) Using the coordinate transformation x, =x, —B,, eqn (2.1.9) is trans- formed to Wx, X25... 5 XN) = Cexp| +3 > Armen | . (2.1.12) Hence we shall consider only this form. For the distribution (2.1.12), the generalization of the formula (2.1.5) is (cxp[E Eete])=er0[3 D(A Yanfsén] —@L13) where (A~"), is the inverse of the matrix Of Ajm, D(A) amAme = One (2.1.14)GAUSSIAN DISTRIBUTION FUNCTIONS 37 Proof: Eqn (2.1.3) gives fon exp[-> Arm nm + 2, bet | =V2n/Ay exp| max.,( - 3 3 Anm¥ntm +5, Est)| . (2.1.15) am n The exponent on the right-hand side of eqn (2.1.15) is a quadratic function of x2. Thus the integral over x2 is repeated using eqn (2.1.3). Repeating this process for x3, x4,..., we get le) = const exp) maty,.a4(-3 5s Anmtntn +3, ta) |- (2.116) The maximum is obtained at x, = Din (A ")am&m» and 1 1 Ae, 24 5 Ds Anmtndin + Enka) =5 3 (AV annbe (2117) nm m am The constant in eqn (2.1.16) must be 1 since eqn (2.1.16) must hold for &,=& = --- =&,=0. Thus eqn (2.1.13) is proved. Since the moment (x,x,,) is calculated as a atm) =| soa 7 2.1.18 Gon =[seragz(emlE ea), eH it follows from eqns (2.1.13) and (2.1.18), that (%pXm) = (AW ame (2.1.19) Equation (2.1.13) is thus rewritten as (exp(& 0%)) = exp[ 3 DEkaintn)|- —— @120) By straightforward calculation, it can be shown from eqn (2.1.20) that a3 I inNmXeX) = See ae ae XP! 5 i, Contnte) = SE oe BE, 0, PLZ BIC) || = (XnXm) (Hi) + (Karke) me) + nx) Xe)» (2.1.21) In general we have the following formula (Wick’s theorem): (XnyZngs © nap) = ae (XmsXina) (Xms%mg) + + + Amap-%may) (2.1.22)38 STATIC PROPERTIES OF POLYMERS where (7, mz,.-.,M,) stands for the permutation of (1, 12,..., Ny) and the summation is taken over all possible pairings. An important property of the Gaussian distribution is that if the distribution of x, is Gaussian, any linear combination of x,, X= a,%n, (2.1.23) obeys the Gaussian distribution, i.e., W(X) = [20 (X?)] 2 exp| - a5] (2.1.24) where (X?) = >) and (XnXm) (2.1.25) The proof is straightforward. By definition WO = {fi dx, (x4, Xap. Xn) a(x -> ant) . After the integral over x, which is carried out easily because of the delta function, the integrand becomes a Gaussian function of x2, ¥3,..., xy and X. Successive integrations over x2, x3,...,Xj are done using eqn (2.1.2) and give a Gaussian function of X, which is eqn (2.1.24). 21.3 Gaussian distribution for complex variables For the complex variables z;,2:,...,2y, we can also define the Gaussian distribution function Wz, 2, .- + Zw) = Cexp| -> Aanz22n| (2.1.26) am where the matrix A,,, is assumed Hermitian with positive eigenvalues, Axn=Amny and >) AnnZ*Zm 20. (2.1.27) am The integral over z, is defined to be made over the entire surface of the complex plane; i.e., if x and y denote the real part and the imaginary part of z respectively, then +e te foz--= fae dy.... (2.1.28) The average of a function F(z, z,..., Zw) is defined as +2 (Fl2u, 2a +.) = [Te | Tere. day ve )Bl2tr Zar).GAUSSIAN DISTRIBUTION FUNCTIONS 39 The generalization of eqn (2.1.3) to complex variables is, for a positive A and a complex &, fee exp[—Azz* + z§* + 2*&] =“ explmax,(—Azz* +2&* +2*&)] (2.1.29) where max,(...) denotes the maximum value when z is varied over the entire complex plane. Equation (2.1.29) can be proved by direct calculation. Using eqn (2.1.29), one can prove: @ (exe, D Cues + 822,)]) =exp[ S280(202%)] 130) = (i) (222m) = (AM Dans (2.1.31) 2.1.4 Functional integral If the subscript m of x, is regarded as a continuous variable, the set of points (x), x2, ..., 4) represents a continuous function, and the integral over the set (%:,%2,...,Xn) reduces to the integration over all the functions, and is called the ‘functional integral’. It is denoted by the symbol 6x,,; i.e., I I o, continuous limit I Om The formula in the functional integral is obtained by taking the continuous limit of the discrete variables. For example, as a generaliza- tion of eqn (2.1.16), we have N N N [or en| =| dn | date + fanter| 0% 0 N N N sexp| max -5 | dn | dA nmXnkm + [ons.x,)] (2.1.32) Oo 0 oO or (ex dnt ,sn]) = ew|5 [an] amc ntstn] (2.1.33) where ° 8 [ama )onAry = 3(0 =). 2.1.340 STATIC PROPERTIES OF POLYMERS Appendix 2.11 Differential equation for G(R, R’; N) For U,=0, G(R, R';N) becomes the probability distribution of the end-to-end vector and is given by GAR ~R';N) = (ey ren (2 ow) (2.11.1) where 1, N>O0 ew)={7 N<0. (2.11.2) Using the Fourier transform (see eqn (2.25)) Go(R ~ RIN) = [SB exo(-ik (Rk = RV exp( -wE Jew), one can easily check @.01.3) a ten) = | exo(—ik-(R—R)| 2 Su GAR R'sN) = [A serp(-ik-(R—R’)| ~™ NOR x exp( - = Joc) +6009] b? a . , = 25 GAR-R'N)+9(R-RY SN). 2.11.4) To derive an equation for G(R, R'; N) for U.#0, we use eqn (2.49) for small AN: G(R, R';N+AN)= | dR"G(R, R"; AN)G(R", R'; N). (2.11.5) For n between N and N + AN, R, is not far apart from R. Thus if U,(r) is a smooth function of r, the energy is approximated as N+A4N | dnU,(R,) ~ ANU,(R). (2.11.6) N Then G(R, R”; AN) is easily obtained from eqn (2.46), G(R, R"; AN) = exo( = ANULR))GoR —R’"; AN). (2.11.7) B From eqns (2.11.5) and (2.11.7), G(R, R';N+ AN) = exp( - pANuR)) B x farecar —R’"; AN)G(R", R';N). (2.11.8)PERTURBATION CALCULATION 4 For small AN, Go(R — R"; AN) has a sharp peak at R = R’. Therefore the integral in eqn (2.11.8) can be evaluated by expanding G(R", R’; N) with respect to r= R — R": I= | dR"G)(R — R"; AN)G(R", R’; N) = force AN)G(R — 1, R';N) = farccces an(1 - w+ Ar, G(R, R';N). (2.11.9) noe ) aR, OR, ORs Since 2 | drGo(r; AN)rx = 0, | rGo(r; AN)rats = ane bap, (2.11.10) the integral becomes & = 1 2 1 I= (1 +2ANB GR, R';N). (2.11.11) Thus to the order of AN, eqn (2.11.8) is written as 3 nnya(p—-— (1+an5)o(R, R's) = (1-7 ANUAR)) 2 & x (i +3ANb aOR, R'N). (2.11.12) Comparing the terms of order AN, we have a be 1 , (e-em? ET U)CR, R'sN)=0. (2.11.13) This equation hoids for N > 0. To account for the singularity at N =0, we use eqn (2.11.7) for small N, i.e., ai keT which gives 5(R — R') 6(N) on the right-hand side of eqn (2.53). G(R, R';N) = exp( NUAR))Go(R -R'5N) (2.11.14) Appendix 2.11 Perturbation calculation for the excluded volume effect Let (...)o denote the average for the distribution function of the ideal chain, then for the excluded volume chain (R?) = (exp(—Ui)(Rw ~ Ro)”)o {exp(=Ui))o” (2.111)42 STATIC PROPERTIES OF POLYMERS where N m Uj=v | dm | dn 6(R, — Rn) (2.11.2) oO 0 (here k,T is put equal to unity), Expanding eqn (2.III.1) to the first order in U,, we have — (C= Ui)(Rw ~ Ro)” R= Ue = ((Rw — Ro)?) od + (Ui )o) — (Ui (Rw — Ro)”)o = No-(1 +f f anon -Rn))o) ~v fam fan(5(R,—R,)(Ry—Ro)*)o 2-113) First we calculate T= (8(R, — Rn)(Rw — Ro))o» (2.1114) which is written, using the Green function G,(R, R', N) in free space ‘see eqns (2.50) and (2.51)), as T= far AR,, dR, Go(Ry — Rms N - m) X Gy(Rys — Raj mt — n)G(Ry — Ron) 5(Ry — Ry)(Ryy— Ro}? = | ak dR,G9(Ry,— Ry; N—m)Gy(0; m—n) X G(R, — Ro; n)(Ry — Ro)? (2.1115) To do the integral, we write Ry—Ro as (Rv—R,)—(R,— Ro) and ategrate first over Ry and then over R,. The result is I= G)(0; m —n)(N —m + n)b?. (2.111.6) “hus N m (U,(Ry — Ro?)o= v | dm | dn(N —m +n)b?G,(0;m—n). (2.1117) imilarly ° 8 (U1) =v | dm { anG(0sm —n). QL)REFERENCES 43 The integrals in eqns (2.III.7) and (2.III.8) diverge, but the divergent terms cancel with each other in (R?). In fact, from eqns (2.III.3), (2.11.7), and (2.11.8), we have (R?) = No*(1 +u [am] anc m—n)(m— n)) oO oO = No*(1 +0 [an [olan on) - n)), (2.1119) which converges and gives the first term in eqn (2.100). Next we derive eqn (2.108). The integral we now have to evaluate is q= (2) as - R,)) (2.11.10) To calculate this we replace 9R,,/dn by R, — R,-1, and rewrite Ry — Ro as Ry — Ro= (Ry — Ry) + (Ry —Rn-1) + (Ry-1—Ro). (2.1.11) Since there is no correlation among Ry — R,, Ry — Rn-1, and R,-, — Ro, eqn (2.II1.10) becomes T= (Ry —Rn-a)?)'((Rw — Ray)! + (Rn — Rn-1)*)! + (Ry — Rn-1)")'(Rn-1-Ro)’)'. (2-112) Since ((R,, — R,)*)' = (m —n)b” and (Rj, —R,)*)' = (5/3)(m — n)°b4, eqn (2.1II.12) becomes J=Nb'* +36", (2.11.13) which gives eqn (2.108). References 1. Flory, P. J., Statistical Mechanics of Chain Molecules. Interscience, New York (1969). 2. Volkenstein, M. V., Configurational Statistics of Polymeric Chains, Interscience, New York (1963). 3. Birshtein, T. M., and Ptitsyn, O. B., Conformations of Macromolecules. Interscience, New York, (1966). 4. Yamakawa, H., Modern Theory of Polymer Solutions. Harper & Row, New York (1971). 5. Feller, W., An Introduction to Probability Theory and its Applications (3rd edn). Wiley, New York, (1968).29. }. Teramoto, E., Busseiron Kenkyu 39, 1 (1951). 31. . Yamakawa, H. and Tanaka, G., J. Chem. Phys. 47, 3991 (1967). 33. . Edwards, S. F., J. Phys. A 8, 1171 (1975). See also Oono, Y., J. Phys. Soc. STATIC PROPERTIES OF POLYMERS . Edwards, S. F., The configurations and dynamics of polymer chains, in Molecular Fluids (eds. R. Balian and G. Weill). Gordon & Breach, London (1976). . Freed, K. F., Adv. Chem. Phys. 22, 1 (1972). Edwards, S. F., and Freed, K. F., J. Phys. A 2, 145 (1969). Helfand, G., in Polymer Compatibility and Incompatibility (ed. K. Solc), p. 143. Harwood Academic Publishers (1982). ). Napper, D. H., Polymeric Stabilization of Colloidal Dispersions. Academic Press, London (1983). . Chu, B., Laser Light Scattering. Academic Press, New York (1974). . Berne, B. J., and Pecora, R., Dynamic Light Scattering. Wiley, New York (1976). . Marshall, W. and Lovesey, S. W., Theory of Thermal Neutron Scattering. Oxford Univ. Press (1971). . Kuhn, W., Kolloid Z. 68, 2 (1934). . Flory, P. J., J. Chem. Phys. 17, 303 (1949). . Flory, P. J., Principles of Polymer Chemistry. Cornell Univ. Press, Ithaca, (1953). . de Gennes, P.'G., Scaling Concepts in Polymer Physics. Cornell Univ. Press, Ithaca (1979). . Oono, Y., Adv. Chem Phys. 61, 301 (1985). . Stockmayer, W. H., Macromol. Chem, 35, 54 (1960). The potential of the delta function was first used by Zimm, B. H., Stockmayer, W. H., and Fixman, M., J. Chem. Phys. 21, 1716 (1953). . Stephen, M. J., Phys. Lett. 53A, 363 (1975). . de Gennes, P. G., J. Physique 36, L55 (1975); 39, L299 (1978). See for example, Reichl, L. E., A Modern Course in Statistical Physics, Chap. 11, Univ. of Texas Press, Austin (1980). See Ref. 4, Section 39. A recent discussion on this problem can be found in Tanaka, G., Macromolecules 13, 1513 (1980) and ref. 18. . Lifshitz, I. M., Grosberg, A. Y., and Khokhlov, A. R., Rev. Mod. Phys. 50, 683 (1978). . Williams, C., Brochard, F., and Frisch, H. L., Ann. Rev. Phys. Chem. 32, 433 (1981). . Fisher, M. E., J. Phys. Soc. Jpn 26 (Suppl.), 44 (1969). . See Cotton, J. P., J. Physique Lett. L231 (1980), which summarizes experimental results obtained by Miyaki, Y., Einaga, Y., and Fujita, H., Macromolecules 11, 1180 (1978); Fukuda, M., Fukutomi, M., and Hashi- moto, T., J. Polym. Sci. 12, 87 (1974); Yamamoto, A., Fujii, M., Tanaka, G., and Yamakawa, H., Polymer J. 2, 799 (1971). . Wall, F. T., Windwer, S., and Gans, P. J., Monte Carlo methods applied to configurations of flexible polymer molecules, J. Chem. Phys. 38, 2220; 2228 Son. C., Gillis, J., and Wilmers, G., Proc. Phys. Soc. 85, 625 (1965). Fixman, M., J. Chem. Phys. 23, 1656 (1955). Muthukumar, M., and Nickel, B. G., J. Chem. Phys., 80, 5839 (1984). Jpn. 39, 25 (1975); and 41, 787 (1976).REFERENCES 45 35. Domb, C., Adv. Chem. Phys. 15, 229 (1969). 36. McKenzie, D. S., Phys. Rep. 27C, 35 (1976). 37, Brézin, E., Le Guillou, J. C., and Zinn-Justin, J., Phys. Rev. D15, 1544, 1558 (1977). 38. Le Guillou, J. C., and Zinn-Justin, J., Phys. Rev. Lett. 39, 95 (1977). 39, Edwards, S. F., and Singh, P., J. Chem. Soc. Faraday Trans. II 75, 1001 (1979). 40. Fisher, M. E., J. Chem. Phys. 44, 616 (1966). 41. Edwards, S. F., Proc. Phys. Soc. 85, 613 (1965). 42. Kosmas, M. K., and Freed, K. F., J. Chem. Phys. 68, 4878 (1978). 43. Wilson, K. G., Phys, Rev. B4, 3184 (1971); Wilson, K. G., and Kogut, J. B., Phys. Rep. C12, 75 (1974). 44, de Gennes, P. G., Phys. Lett. 38A, 339 (1972). 45. des Cloizeaux, J., Phys. Rev. A10, 1665 (1974); J. Physique 42, 635 (1981). 46. Wilson, K. G., and Fisher, M. E., Phys. Rev. Lett. 28, 240 (1972). 47. Burch, D. J., and ‘Moore, M. A., J. Phys. A9, 435 (1976). 48. Witten, T. A. Jr. and Schiifer, L., J. Chem. Phys. 74, 2582 (1981). 49. Widom, B., J. Chem. Phys. 43, 3898 (1965). 50. Kadanoff, L., Physics A2, 263 (1966). 51. Witten, T., Jr., J. Chem. Phys. 76, 3300 (1982). 52. Ohta, T., Oono, Y., and Freed, K. F., Phys. Rev. A25, 2801 (1982).3 ’ BROWNIAN MOTION 3.1 - Introduction Polymers in solutions incessantly change both their shape and position randomly by thermal agitation. This Brownian motion dominates various time-dependent phenomena in polymer solutions such as viscoelasticity, diffusion, birefringence, and dynamic light scattering, which are to be discussed in subsequent chapters. In this chapter, we study the basic theory of Brownian motion. Since the general aspects of the theory of Brownian motion have already been discussed in many articles,’> we shall limit the discussion to topics which will be useful in the application to polymer solutions and suspensions. In principle, Brownian motion can be discussed starting from the dynamical equation of motion of the Brownian particle and the fluid molecules. However, this microscopic approach is not useful for calculat- ing the various dynamical quantities we are interested in. Here we take a phenomenological approach, regarding Brownian motion as a kind of stochastic process, and construct a phenomenological equation describing Brownian motion based on known macroscopic laws. This approach, originated by Einstein,‘ is limited by several conditions, such as that the time-scale and the length-scale under consideration are much longer than those characteristic of solvent molecules, and that a linear relation holds between fluxes and forces. For polymer solutions and suspensions, these conditions are normally satisfied without any problems, and the theory we shall describe in this chapter can be regarded as a general base for describing their dynamics. The phenomenological equation for Brownian motion has two seem- ingly different, but essentially the same, forms—the Smoluchowski equation and the Langevin equation. The Smoluchowski equation is derived from the generalization of the diffusion equation and has a clear relevance to the thermodynamics of irreversible processes. The Langevin equation, on the other hand, has no direct relationship to thermo- dynamics, but is capable of describing wider classes of stochastic processes.” We shall first study the Smoluchowski equation, and then consider its equivalence to the Langevin equation. 3.2 The Smoluchowski equation 3.2.1 Diffusion of particles The phenomena in which the effect of Brownian motion appears most clearly is diffusion: small particles placed at a certain point will spread outTHE SMOLUCHOWSKI EQUATION 47 in time. For the sake of simplicity we will consider one-dimensional diffusion. Let c(x, f) be the concentration at x and t. The process of diffusion is phenomenologically described by Fick’s law, which says that if the concentration is not uniform, there is a flux j(x,t) which is proportional to the spatial gradient of the concentration, i.e., i@,)=-D& @G.1) where the constant D is called the diffusion constant. The microscopic origin of the flux (eqn (3.1)) is the random motion of the particles: if the concentration is not uniform, the number of particles which happen to flow from the higher concentration region to the lower concentration region is larger than the number of particles flowing in the other direction (see Fig. 3.1). This imbalance in the number of flowing particles gives (to the first order in the concentration gradient) eqn (3.1) for the flux. Note that the flux comes entirely from fluctuations in the o(x) x Dix) x Cr Or ~o +O “0 Oo eg Oo Oo © +9 oe =O Oly <9 O- Fig. 3.1. Microscopic explanation for Fick’s law. Suppose that the particle concentration c(x) is not uniform. If the particles move randomly as shown by the arrows, there is a net flux of particles flowing from the higher concentration Tegion to the lower concentration region. Here the diffusion constant of the particle, which determines the average length of the arrows, is assumed to be constant.48 BROWNIAN MOTION velocity: the average velocity of individual particles is zero as their motion is independent of each other. Equation (3.1), together with the continuity equation oc oj Bo ae (3.2) gives the well-known diffusion equation for constant D, ac ce 3 Oa (3.3) If there is an external potential U(x), Fick’s law must be modified. The potential U(x) exerts a force 8U F=-— 3.4 Bn GB.4) on the particle, and gives a non-vanishing average velocity v, which in the usual condition of weak force, is linear in F so that 13U =-s- 3.5 v= ee (3.5) The constant € is calied the friction constant and its inverse 1/€ is called the mobility. If the particle is sufficiently large, € can be obtained from hydrodynamics. For example, if the particle is a sphere of radius a, and the viscosity of the solvent is 7,, then the hydrodynamic calculation indicates® € = 6a7n,a. (3.6) If the particle is not spherical, the formula for the velocity is not simple (see Section 3.8 and Chapter 8), but the linear relationship between the force and the velocity always holds provided that the force is weak. The average velocity of the particle gives an additional flux cv, so that the total flux will be dc caU j=-D—-=—. . ] ox € ax 67 An important relation is obtained from eqn (3.7). In the equilibrium state, the concentration c(x, t) is given by the Boltzmann distribution Ceq(x) * exp(—U(x)/kaT) (3.8) for which the flux must vanish: a 1 oU ~D ay Cea Boca ay (3.9)THE SMOLUCHOWSKI EQUATION 49 From eqns (3.8) and (3.9), it follows that in equilibrium D= ket (3.10) c° This relation is called the Einstein relation. The Einstein relation states that the quantity D which characterizes the thermal motion is related to the quantity € which specifies the response to the external force. We shall show later that the Einstein relation is a special case of a more general theorem, called the fluctuation dissipation theorem, which states that the characteristics of the spontaneous thermal flucutation are related to the characteristics of the response of the system to an external field. Using eqn (3.10), eqn (3.7) can be written as dc 3U jm ber <2) ax Hence the diffusion equation becomes ac_ al a | aU a axe e(kor + cS): ox This equation is called the Smoluchowski equation. The above argument for deriving the Smoluchowski equation can be summarized in a more formal may. Equation (3.11) can be rewritten as (3.12) 1 jn-7ee (ksT Inc + U), (3.13) which has a thermodynamic significance. The quantity U(x) + kgT Inc is the chemical potential of noninteracting particles of concentration c. Thus eqn (3.13) states that the flux is proportional to the spatial gradient of the chemical potential. This is a natural generalization of Fick’s law because when the external field is nonzero, what must be constant in the equilibrium state is not the concentration, but the chemical potential. Now if we define the flux ey by vu, =j/c, eqn (3.13) gives Yast Zor Inc+U) (3.14) which is quite similar to eqn 9, The only difference is that the potential U(x) is now replaced by the chemical potential U(x) + kgT Inc(x, ¢). If this replacement is made, the Smoluchowski equation is derived from the usual continuity cavation Oc _ on 2 cus). 3.15)50 BROWNIAN MOTION Thus the Smoluchowski equation can be formally derived from the macroscopic equations (eqns (3.14) and (3.15)), where the fluctuations have been dealt with by invoking the thermodynamic variable—the chemical potential U(x) + ksT In c(x, t}—instead of the potential U(x). 3.2.2 Diffusion in phase space So far we have been considering the diffusion of concentration. The same argument will hold for the probability distribution function W(x, t) that a particular particle is found at point x at time ¢ since the distinction between c(x, t) and W(x, t) is, for non-interacting particles, only in the fact that W is normalized. Thus the evolution equation for the probability W(x, 2) is written as a a1 “at “ax t which will also be termed the Smoluchowski equation. (In some literature this equation is called the Fokker—Planck equation, or the generalized diffusion equation, but the original Fokker—Planck equation was for diffusion in velocity space, so we do not employ the term here.) Having seen the basic principle, it is easy to derive the Smoluchowski equation for a system which has many degrees of freedom. Let Xi, %2,...,Xn~ ={x} be the set of dynamical variables describing the state of Brownian particles. To construct the Smoluchowski equation, we have to know first the relation between the average velocity v, and the force F, = —9U/éx,. Such a relation is generally written as Un =D Lam {2})Fins (3.17) The coefficients L,,, are called the mobility matrix, and may be obtained using hydrodynamics. It can be proved that L,,, is a symmetric positive definite matrix: oe =v) (er = (3.16) Lym =Lmn and >) F,FnLiym 20 for all F,. (3.18) nm Given the mobility matrix, the Smoluchowski equation is obtained from the continuity equation ow a e774 3x, me) (3.19) with the flux velocity being given by med Lam5o— (ka In +U). (3.20)THE SMOLUCHOWSKI EQUATION 51 Hence ow aU. Balle SS) G2 which is the basic equation for the dynamics of polymer solutions and suspensions. The Smoluchowski equation may be regarded as a phenomenological tool for describing the fluctuation of physical quantities, and can be applied to more general situations: for example the equation can be used to describe the fluctuation of thermodynamic variables (such as con- centration). In such cases, the potential U(x) must be regarded as the free energy which determines the equilibrium distribution of those variables, and the relation (3.17) must be replaced by phenomenological kinetic equations. We shall see such applications in Chapter 5. 3.2.3 Irreversibility of the Smoluchowski equation An important property of the Smoluchowski equation is that if U({x}) is independent of time and if there is no flux at the boundary, the distribution function W always approaches W equilibrium (Yq); W,4= exp(—U((x})/keT) / | d{x}exp(-U({x}/keT). (3.22) To prove this we consider a functional® Af} = | d{z}'(k,T In + U) (3.23) for the general solution of eqn (3.21). The time derivative of is calculated as $ a= face {= (keT In¥ + U)+ ket]. (3.24) Using eqn (3.21) and the integral by parts (the integral at the boundary being zero by the assumption), we have Gea ~ | AYE Lan [= -(kaT n+ U) || Gear in ¥ + 0) - | be) PE Lan| ba In (Y/Y) [+7 5m Cr.) | (3.25) which is negative according to eqn (3.18) unless W is identical to Wg. As sf will eventually reach the minimum value, W also reaches V., after a sufficiently long time.52 BROWNIAN MOTION At equilibrium, [WY] becomes the free energy defined in the equilibrium statistical mechanics: [Wea] = —kpT in( | d{x}exp(— UiksT)). (3.26) When the system is not in equilibrium, #[] is larger than é[W,,] and the difference [W] — s[W,] represents how far the system is away from the equilibrium state. We shall call sf the dynamical free energy. This quantity plays an important role in the subsequent discussions. The approach to the equilibrium state is also shown directly by the eigenfunction expansion of the distribution function. This is discussed in Appendix 3.I. More detailed discussion on the Smoluchowski equation is given in ref. 7. 3.3 The Langevin equation An alternative description of Brownian motion is to study the equation of motion of the Brownian particle writing the random force f(t) explicitly in the Langevin form: dx 6G--Z +10. @.27 Physically, the random force f(t) represents the sum of the forces due to - the incessant collision of the fluid molecules with the Brownian particle. As we cannot know the precise time-dependence of such a force, we regard it as a stochastic variable, and assume a plausible distribution W[f(t)] for it. In this scheme, the average of a physical quantity A(x(t)) is calculated, in principle, by the following procedure. First eqn (3.27) is solved for any given f(t), and A(x(¢)) is expressed by f(t). The average is then taken with respect to f(t) for the given distribution function [f(t)}. Though various distributions can be conceived for f(t) depending on physical modelling, here we shall consider only the process which is equivalent to that described by the Smoluchowski equation. It is shown in Appendix 3.II that if the distribution of f(¢) is assumed to be Gaussian characterized by the moment (f@) =0, (FOF) = 2keT 6(e- £'), (3.28) i.e., if the functional probability distribution of f(t) is Wyler — Fe [ ance?) @.29) then the distribution of x(¢) determined by eqn (3.27) satisfies the Smoluchowski equation. Here we check this for a simple special case.THE LANGEVIN EQUATION 53 Consider the Brownian motion of a free particle (U = 0) for which the Langevin equation reads dx sF=f0. (3.30) If the particle was at x’ at time t= 0, its position at time ¢ is given by are x(th=x +5 fare ). (3.31) Equation (3.31) indicates that x(t)—x’ is a linear combination of Gaussian random variables f(t). Therefore, according to the theorem in Appendix 2.1, the distribution of x(f) must be Gaussian. Hence the probability distribution of x(¢) is written as W(x, 1) = (20By? exp(- &-4y) (3.32) where A=(x@)), B=((x(t)-A)’). (3.33) From eqns (3.28) and (3.31), these moments are easily calculated: , 1f a ” _y! A=x +g fargen =x and ' : Ll seen) (1 a= (gf arnen)(5 [arnen)) Pa 1 . iw , -2 | ar J arncfceyge)) (3.34) kal ff an gry om _2koT =e Ja [a dt = t (3.35) or by the Einstein relation (3.10) B=2Dt. Thus = 1? exp ( = 2'Y Wx, 1) = (41) exp( = ). (3.36) This is the solution of the Smoluchowski equation ow e One (3.37)54 BROWNIAN MOTION ctx) D(x) oO e+ *—o =O <0 =o Qo e 7 & er es * =O a oe Qo Or Fig. 3.2. Explanation for the term 9D/6x in eqn (3.38). Consider the equilibrium state for the case of U=0 and 9D/dx >0, The random force f(r) in eqn (3.38) causes displacement of individual particles as shown by the arrows: the particles on the right are more mobile than those on the left. This creates a net flux of particles toward the negative direction. To compensate for this, the term 0D/dx is needed. The Langevin equation (3.27) is equivalent to the Smoluchowski equation only when the diffusion constant D=k,T/€ is independent of x. If D depends on x, an additional term must be added to eqn (3.27)°+ pent, +f + i (3.38) } In some literature, the added term is 9D/dx. This difference comes from the difference in the definition of the integral over the random force which is not properly defined in the conventional Stieltjes integration. Here we have followed the definition given in ref. 2. Such complication can be avoided if one uses the Langevin equation written in a difference form in a finite time interval Ar:°1° x(t+ At) =x(t)— ie Att VD (x()g(0) +3 2 where g(f) is a random variable sashes (g))=0, — (g()g(t’)) =2At 6. See ref. 7 for more detailed discussion.TIME CORRELATION FUNCTION AND RESPONSE FUNCTION 55 The derivation is given in Appendix 3.II. Physically, the term 0D/dx is needed to compensate for the flux caused by the random force which is dependent on particle position (see Fig. 3.2). The Langevin equation corresponding to the Smoluchowski equation in multidimensional phase space (eqn (3.21)) is given by” de _ su 1 2 an = Lam ae, + fa() + teeT beg am (3.39) The distribution of the random force is Gaussian, characterized by the moment a(t) = 0, (3.40) n()fnlt")) =2L uma T 5(t -t'), (3.41) where (L~"),m denotes the inverse matrix of Lym DY Lam (L mk = Snts (3.42) The Langevin equation (3.39) represents the same motion as the Smoluchowski equation (3.21). However, each of the equations has advantages and disadvantages in solving our problems; we shall therefore use both equations interchangeably. 3.4 Time correlation function and response function 3.4.1 Time correlation function An important quantity characterizing the Brownian motion is the time correlation function, which is operationally defined in the following way. Suppose we measure a physical quantity A of a system of Brownian particles for many samples in the equilibrium state. Let A(t) be the measured values of A at time ¢. Usually A(t) looks like a noise pattern as shown in Fig. 3.3a. The time correlation function C,,(t) is defined as the average of the product A(t)A(0) over many measurements: Caa(t) = (A(QA(0)). (3.43) Typical behaviour of C,,4(t) is shown in Fig. 3.3b: at t=0, C44(0) is positive and is equal to the mean square of A in the equilibrium state, (A?). As time passes, C44(t) usually decreases with time since the value of A(t) becomes uncorrelated to that at t=0. After a sufficiently long time, the correlation between A(t) and A(0) vanishes completeiy, and Caa(t) becomes equal to (A(t)) (A(0)) = (A). The characteristic time with which C,,(¢) approaches the asymptotic value is called the correlation time.56 BROWNIAN MOTION A(t) Cult) () Fig. 3.3. (a) Example of measured values of a certain physical quantity A as a function of time. (b) A typical behaviour of the time correlation function Cia(t) = (A()A(0)). The correlation time is denoted by rt... The time correlation function (3.43) is termed the auto correlation function as it expresses the correlation of the same physical quantities at different times. Time correlation functions are also defined for different physical quantities, Caa(t) = (A()B(O)) (3.44) which is called the cross correlation function. 3.4.2 Microscopic expression for the time correlation function Given the Smoluchowski equation, time correlation functions can be calculated. For the sake of simplicity we use x to denote the whole set of coordinates x,, x2,..., x, appearing in the Smoluchowski equation. Let G(x, x’; t) be the probability that the system which was in the state x’ at time t=0 is in the state x at time ¢. Clearly such probability is obtainedTIME CORRELATION FUNCTION AND RESPONSE FUNCTION 57 from the Smoluchowski equation 8 oe, xu) = 5 Let 224 Y 5Ou 0-5 See bom (Ko a ) (3.48) and the initial condition Ge, x1;1=0) = 8(e- x’) =T] 8(&, — 24). (46) Given G(x, x’; t), the time correlation function is evaluated by (A()B(0)) = | dx | de'AG)B(X'}G(e, x's) Wale) G47) where A(x) and B(x) denotes the value of the physical quantity A and B when the system is in the state x.+ The meaning of eqn (3.47) is clear. If the system is in the state x‘ at time t=0 and in the state x at time t, the measured value of A and B are A(x) and B(x’), respectively. The probability that this happens at equilibrium is G(x, x’; t)W.q(x’). Averaging A(x)B(x’) for this probability, we get eqn (3.47). Though eqn (3.47) gives a general method for calculating the time correlation function, it is not easy to carry out this procedure since G(x, x'; t) is difficult to obtain. Usually, more convenient methods are available, which will be demonstrated in subsequent sections. However, the initial slope of the time correlation function can be calculated directly from eqn (3.47). The time derivative of eqn (3.47) is calculated as §,(4(0B00)) = fax ax’A 2 Bee Wale) a aG au = faxfaramd = Lin(kaTS= + GS) X B(x'PWeg(x’). At t=0, G(x, x’; t) becomes 5(x — x’), hence GAOBO)| _ = [axa GD ge Lam] base BOM a(t) +B Mae) 2], 6.48) + Note that here two functions, A(t) which represents the measured value of the physical quantity at time t, and A(x) which represents the dependence of the physical quantity on the dynamical variables x, are distinguished by their arguments. To avoid confusion, it may be preferable to denote the latter as A(x), then A(t) can be written as A(x()). However, since this leads to over cumbersome nomenclature, we do not use it here.58 BROWNIAN MOTION Using integration by parts and eqn (3.22), we have (430) = —ksT | ax Weal FOL om 220) oB =), (3.49) ~ ~keTD Ft x, The average in the final expression is for the equilibrium distribution function W.,(x). The initial decay rate P defined by 1 =-5(A(B))| (AB) (A)(B)) 50) is thus given by kaT dA OB 1 = RABY (aes (51) 3.4.3 Fluctuation dissipation theorem Consider a time-dependent external field A(t) (magnetic field, electric field, or velocity gradient field) applied to a system in equilibrium. In general, the field perturbs the system, and changes the average values of physical quantities from those in the equilibrium state. If the field is weak, the change in any physical quantity is a linear functional of the field, and is written as (A(0))a — (A)o= | dt'u(t—t')h(t'), (3.52) where (A(t)), denotes the value of A at time t when the field is applied and (A)o the equilibrium value of A in the absence of the field. The function y(t) is called the response function. In many cases, the effect of the field on the system is expressed by a potential such as Vex (%, t) = —h(t) B(x). (3.53) The quantity B(x) is said to be conjugate to the field h(t). In such cases, the response function is related to the time correlation function by ui) =-- 4 x50. (3.54) keT Tai This theorem is called the fluctuation dissipation theorem.” To prove eqn (3.54), we consider the situation that a constant field / is applied for a long time until the system reaches equilibrium, and that theTIME CORRELATION FUNCTION AND RESPONSE FUNCTION 59 Ait t (o) Fig. 3.4. Relaxation function. When a constant external field is switched off at t=0, the average of a physical quantity A relaxes from the equilibrium value in the presence of the field, (A),, to that in the absence of the field (A)o. The time-dependence for t>0 is described by the relaxation function a(t) as (A)o + a(th. field is then switched off at t = 0 (see Fig. 3.4). In such a case, the average value of A will change as shown in Fig. 3.45. (A(t) a = @(a)h + (Ao (3.55) The function a(f), called the relaxation function, is expressed by the response function y(t) as 0 “ a(t)= | di'ule— 1’) = fadea(e. (3.56) Now (A(Z)), can be calculated if the distribution function P(x, f) is known: (AO). = faacwe, 1). (3.57) Since there is no external field for t>0, W(x, ¢) is related to V(x, t=0) by the Green function G(x, x’; t) in the absence of the field W(x, 1) = J dx'G(x, x"; )W(x, 0). (3.58)1) BROWNIAN MOTION Since W is at equilibrium in the presence of the field at time ¢ = 0: exp[—(U(x) — AB(x))/ksT] W(x, 0) = (3.59) | dx exp[—(U(x) — hB(x))/kaT] which can be expanded with respect to h as We, 0) = PCO k aT FBG ka) _ dx exp(—U(x)/ka nla +h(B)olkpT) =¥.,(1+ ee (od) (3.60) From eqns (3.57), (3.58), and (3.60), it follows that h "\—(B (A())a= | dx | dx’A(x)G(, x'; Yea(x')(1 +HBE)— Pre), (3.61) Using the stationary property of the equilibrium state, W(x) = | dr'G(x, x':)B(e"), (3.62) and the definition of the time correlation function, we get h (A())n = (Ado+ pr} {s'aae, x'5 1)(B(x’) ~ (B)o)Peq(x") = (Ayo 7a L(ACBO))a~ (A)o(B)al: 6.63) From eqns (3.63) and (3.55), it follows that 1 a(t) = 5,7) — (A)o(B)o) (3.64) which leads to eqn (3.54) after differentiation with respect to 1. Note that in the above proof the explicit form of the time evolution equation for W is not used. Therefore the proof applies to a pure dynamical system which is described by the Liouville equation. The fluctuation dissipation theorem holds quite generally in physical systems near equilibrium. In the case of A = B, the fluctuation dissipation theorem can be stated in a more convenient form. Let us define the growth function A(¢) as the response to the sudden application of a step field (see Fig. 3.5): (A())a— (A)o= BOA (3.65)TIME CORRELATION FUNCTION AND RESPONSE FUNCTION 61 a(t) @ t () Fig. 3.5. Growth function. When a stepwise external field is applied (a), the average of a physical quantity A changes as shown in (b). The time evolution is described by the growth function A(t) as (A)o+ B(t)h. From eqn (3.52) we have Bi) = | dt'u(?’). (3.66) 0 Using eqn (3.54), BU) == 7(Caal0)~ Caal) = 35 (AW?) + (A(0}?) ~2(4AO))) 1 2. =3,7 (AO — A(0))?). (3.67) A simple application of eqn (3.67) is the Einstein relation. Let us consider that A denotes the x component of the position vector of the Brownian particle. The field conjugate to x is the external force Fix. When Fx, is applied, the particle begins to move with the constant62. BROWNIAN MOTION velocity F./¢. Thus t BO= (3.68) Hence eqn (3.67) becomes Ey TeO-xOP). (3.69) Since, by definition of D, ((x(#) — x(0))?) is 2D#, eqn (3.69) leads to the Einstein relation D “Er (3.70) 3.5 Brownian motion in a harmonic potential In this section we study a simple system, the one-dimensional Brownian motion of a particle in a harmonic potential: U =4kx?. (3.71) Although this is a very simple system, it is a prototype of the problems which we shall discuss later on. 3.5.1 Smoluchowski equation Let us first calculate the time correlation function {x(¢)x(0)) using the Smoluchowski equation. In principle this is obtained from eqn (3.47) {x()x(0)) = | dx | dx'xx'G(x, x's: (') G72) We(2)= (2 faa - 5) (3.73) and the Green function G(x,x’;t) is to be determined from the Smoluchowski equation where ay. aG SO x5 Pg; (ko So + kx) (3.74) under the initial condition G(x, x'; 0) = (x — x’). (3.75)BROWNIAN MOTION IN A HARMONIC POTENTIAL 63 However, we can show that (x(r)x(0)) can be calculated without complete knowledge of G(x, x’; t). Using eqn (3.74), the time derivative of eqn (3.72) is calculated as 5 (xx) = faxfarax ‘lis 2 (wre SS + bxG) |Weale'). 3.7) The right-hand side is rewritten by integral by parts as a - 6 9609 E28 x) 2 9] ae Ox) = [arfer GW eal Hf € ax ax ex!) — € ox er") = ~§ faxfaveecee, x’; t)Peg(x’) 1 =—7 (x(Ox(0)) (3.77) where t= Elk. (3.78) The initial condition for the differential equation is obtained from (x(0)2) = | dxxW,,(x) =tt . (3.79) Hence (x(A)x(0)) = ‘af exp(-t/t). (3.80) 3.5.2 Langevin equation The same result as eqn (3.80) is obtained from the Langevin equation. For the potential (eqn (3.71)), the Langevin equation is written as dx bane ttO) (3.81) with . F(t)) =0, CF (OF(t)) = 2SkaT S(t —-#’). (3.82) To calculate (x(t)x(0)), we first express x(t) in terms of f(t’): 0-3 | de’ exp(—(t—#")/a)f(t'). (3.83)64 BROWNIAN MOTION Hence t 0 (x(x(0)) = 35 | des fas expl—(e— 0, ean) “2 | ae f drsempl-( 11 ~ 2)/e]2koT 5(4,~ 1) =“ J dt, exp(~(¢ = 24)/2) = 827 exp(-t/t) (3.84) which agrees with eqn (3.80). Finally we derive an explicit form of the Green function G(x, x’; 1). Again we use the same argument as used in deriving eqn (3.36). Since x(t) is a linear combination of f(t), the distribution of x is Gaussian, which is generally written as _ 2. Gs x's) =aBOY? en - SFO") (3.85) where AQ) = (x()), BO = (@M- AMY). (3.86) To calculate these quantities, we solve the Langevin equation under the initial condition x(0) = x’: x(t) =x! exp(-t/1) +7 J dt’ exp(~(t—r'y/n)f(r’). (3.87) 0 Using eqn (3.82), A(t) and B(t) are calculated from eqns (3.86) and 3.87): A(t) =x’ exp(-t/r) (3.88) and BO =z fu fo ta enp(—@— 1 + 2-H )/e)UfCAdfCt)) oO “2 fae , exp[-2(0—14)/2] - 26k pT 0 ='s “ty — exp(—2t/r)]. (3.89)INTERACTING BROWNIAN PARTICLES 65 Hence the Green function is given by kT -V2 G(x, x30 = [= (- exp(-21/1))| k[x — x' exp(-t/1) | - eo 3,90 * exo] 2pT{l —exp(—2t/t)] (3.90) Consider the two limiting cases: (i) When fis small, t<< 1, G(x, x’; t) is the same as free diffusion: G(x, x!) = (4nDt)-!? exp| - sa) (3.91) Gi) When ¢ is large, t>>1, G(x,x';1) becomes the Boltzmann distribution, GQ, x';)= (k/2kpT)” exp| - (3.92) tb 2kyT T in accordance with the second law of thermodynamics. Using Mehler’s formula’! for the Hermite polynomial: § HH go - peng 2-E=™F] say pal 2p ! with H,(&) defined by H,(&) = (-1P exp) ae exp(—8?), (3.94) eqn (3.90) can be written in the eigenfunction expansion form: G(x, x= x exp(—A,t) P(x) Pp(x’) Weq(x) (3.95) p= where Ap = PIT; — Yp(x)=(2?p!)"7Hw/y), with y= (2kpT/k)"”. (3.96) 3.6 Interacting Brownian particles Having seen the basic tools for describing Brownian motion, we now consider more realistic situations. Suppose that a collection of spherical Brownian particles, all having equal size, are suspended in a fluid and interacting with each other. Such systems are often found in colloidal suspensions.” As we shall discuss in the next section, the study of this66 BROWNIAN MOTION system is the basis for the general theory of polymer solutions and suspensions. To obtain the Smoluchowski equation for such a system, we first calculate the mobility matrix. Let Ri, R2,...,Rv={R} be the posi- tions of the spheres and F,, &,..., Fy be the forces acting on them. We assume that there are no external torques acting on the particles. Then the velocities of the particles are written ast Va =>) Ham * Fn (3.97) which defines the mobility matrix H,,,. (Note that V, and F, are vectors and therefore each component of the mobility matrix H,,,, is a tensor.) In a very dilute suspension, the velocity of a particle is determined only by the force acting on it, and the mobility matrix becomes Fyn = (3.98) where € = 627,a is the friction constant of the particle, and / is the unit tensor (4g = Sag). In general, however, the velocities of particles depend on the forces acting on their surrounding particles, because the force acting on a certain particle causes the fluid motion around it and affects the velocity of the other particles (see Fig.3.6). This interaction, mediated by the motion of the solvent fluid, is called the hydrodynamic interaction. As a result of the hydrodynamic interaction, the off-diagonal components of the mobility matrix become nonzero. , To calculate the particle velocities V, (n=1,2,...N) we have to know the fluid velocity v(r) created by the external forces acting on the Particles. In the usual condition of Brownian motion, the relevant hydrodynamic equation of motion is that of the low Reynolds number t This form of the mobility matrix is correct only for spherical particles. In general, particles of finite shape have both translational and the rotational degrees of freedom, and the general form of the mobility matrix is written as Va = MGR Bg + HS +News 0, =D HGP «Fy + HQ? Nyy phere {m is the angular velocity of the particle m, and N,, is the external torque acting on + The coefficients H77, HO, etc, depend on both the spatial arrangement of the Patticles and their orientation. However, in the case of spherical particles, HZ?) is independent of the orientation, so that if N,,=0, the rotational motion need not be considered explicitly.INTERACTING BROWNIAN PARTICLES 67 Fig. 3.6. The hydrodynamic interaction. The force acting on the particle m creates a velocity field and causes the motion of other particles. hydrodynamics,°""> which assumes: {i) The fluid is incompressible ate =O (3.99) (In this book Greek indices a, B, u,v... are used to indicate the components of vectors and tensors, and the summation convention is used over the repeated indices.) (ii) The inertia force of the fluid is negligibly small, so that if o2,(r) and g,(r) denote the stress tensor and the external force acting on unit volume of the fluid, respectively, a 31 8 = 8a). (3.100) The stress in the fluid is written as =r,( 2 3s) Cap = nal are arp) * Pap (3.101) (P being the pressure). From eqns (3.99)-(3.101) it follows that # a Nea Ma +o, = 8a: (3.102) Equations (3.99) and (3.102) are called the Stokes approximation, and are the basis of the hydrodynamic interactions in suspensions and polymer solutions.68 BROWNIAN MOTION Let us now calculate the flow field created by the external forces acting on the particles. First we regard the particles as points, and write g(r) as 8) = DF, d(r-R,), (3.103) then eqn (3.102) reads n.Vv + VP =—)>, F,6(r—R,). (3.104) Equations (3.99) and (3.104) are easily solved by the use of the Fourier transform (see Appendix 3-III). The result is v@)=DACr-R,)-F, (3.105) with . He) =a I+) (3.106) where # denotes the unit vector eat to r. The tensor H(r) is called the Oseen tensor. Since the particles move with the same velocity as the fluid, their velocities are given by Vi, = (Rn) = H(Ry — Rn) * En (3.107) Thus Ham = H(Ry — Rm)- (3.108) Unfortunately, eqn (3.108) is not appropriate since H,;=H(0) is infinite. This failure comes from the approximation that the particles are regarded as points. If we start from a collection of particles with finite size, this difficulty does not arise. Unfortunately, for a collection of particles with finite sizes the solution of Stokes equation is obtained only in the form of a perturbation expansion,'*"° which is not easy to handle. A simple approximation commonly adopted in the theory of polymer solutions is to use #/€ for H,,, i.e., A, = Hymn =H(R,—Rn), nfm. (3.109) I eB With this definition of the mobility matrix, the general form of the Smoluchowski equation is written as o> x Ham (ur + *3R,”) G10) This is the basic equation describing the interacting Brownian particles.MICROSCOPIC BASIS OF VISCOELASTICITY 69 3.7 Microscopic basis of viscoelasticity 3.7.1 Introduction Colloidal suspensions and polymer solutions have interesting mechanical properties. In general these materials have both viscosity and elasticity and hence are called viscoelastic. Colloidal suspensions show curious nonlinear hysteresis effects called thixotropy, rheopexy, and dilatancy. These unusual flow behaviours are the central problems of rheology.'7”° A fundamental question in rheology is how those phenomena can be understood from the microscopic characteristic of the materials, i.e., their structure and the type of interaction. In later chapters, we shall discuss this in detail for polymeric liquids. In this section, we shall give a general base for developing microscopic theory for the mechanical properties of suspensions and polymer solutions. The theory presented here is based on the classical work of Kirkwood,”° who summarized the earlier works of Burgers,?4 Kuhn,” and Kramers* and established how to take Brownian motion into account, how to include the effect of the macroscopic flow, and how to calculate the stress tensor.“ Though the original theory of Kirkwood was for dilute polymer solutions, the theory equally applies to concentrated polymer solutions and suspensions. The basis of the Kirkwood theory is to assume that the microscopic dynamics is described by a system of spherical Brownian particles of equal size interacting via the potential U({R}) and the hydrodynamic interaction. Each Brownian particle is called a bead. At first sight this model may appear to be quite specific, but it actually represents a quite wide class of systems. For example: (i) Polymer solutions: A polymer in solution can be modeled as a collection of beads connected sequentially (see Fig. 3.72). The connec- @ Fig. 3.7. Systems described by interacting beads. (a) Polymer (the freely jointed model). (6) Solid particle of an arbitrary shape. In this model, it is convenient to assume that the inside of the particle is filled with the same fluid as that outside, as shown in (c).70 BROWNIAN MOTION tivity of the chain and the excluded volume interaction are included in the potential U({R,,}). (ii) Suspensions of solid particles of arbitrary shape. A solid particle in suspension can be modelled as a collection of beads fixed on the surface the solid particle (see Fig. 3.7b). In this model, the bead m represents a discretized surface element of the particle, and when the size of the beads is taken to be infinitesimal, the formulation becomes equivalent to that obtained by hydrodynamical considerations.**+ To apply the theory to such general systems, we have to consider a system with rigid constraints. However, in this section we shall first consider the case in which there are no rigid constraints, i.e., the force 9U/AR,,, is finite and well behaved. 3.7.2 Constitutive equation The mechanical property of a homogeneous material is expressed by the constitutive equation which relates the stress tensor 0, to the velocity gradient tensor K,,, where 00,(r, t) 3.111 on @.111) Kap (r,t) = and w(r, t) is the macroscopic velocity field. Generally, both 0, and Kas depend on position and time. However, in considering the constitutive equation, the positional dependence of kag can be eliminated because the stress at a certain fluid element depends only on the previous values of the velocity gradient evaluated at that fluid element (principle of locality). This is a consequence of the length-scales of the particles being much smaller than the macroscopic length: although the stress depends, strictly speaking, not only on Kj but also on its spatial derivative 9x,,/dr,, the effect is of the order of (size of the particle)/ (macroscopic length), which is usually negligibly small. Therefore to study the constitutive equation, we may assume without loss of generality that the macroscopic velocity gradient is constant throughout the system, and that the velocity field is given by Da(P, t) = Kap (t)rp- (3.112) In this case, the stress tensor da is independent of position. Such flow is called homogeneous. + In this modelling of solid particles, it is convenient to assume that the frictional elements are placed only on the surface and that the interior of the particle is filled with fluid of the same viscosity as that of the outside fluid (see Fig. 3.7c). Provided that the shell undergoes only translation and rotation, the interior fluid behaves like a solid (the strain rate inside the fluid is zero). In such a case the stress is zero inside the particle and there is a discontinuity in the stress at the surface of the particle, i.e., if F is the surface force density and m the unit vector normal to the surface then o,gip = F, at the surface.MICROSCOPIC BASIS OF VISCOELASTICITY 7 The stress tensor 0,s can be expressed as the sum of an isotropic tensor 66,5 and an anisotropic tensor {3 whose trace is zero: Oop = Bap + OG (3.113) with F=30qe, 98R= 0. (3.114) In an incompressible fluid for which x = Kyg =0 (3.115) the isotropic part is determined by the external conditions and is irrelevant in the discussion of the constitutive equation. In this book we shall consider only such fluids, and neglect the difference in the isotropic part of the stress tensor. Thus two stress tensors of and off are regarded as equal if their difference, o4 — 0, is an isotropic tensor. 3.7.3. The Smoluchowski equation for a system in macroscopic flow Now our aim is to calculate the stress for given history of macroscopic velocity gradient. First we consider the form of the Smoluchowski equation for a given macroscopic velocity field 0(r, t)=«(t)+r. To do this we have to know the microscopic velocity field u(r, 1). This is obtained from the conditions: (i) v(r, ¢) is a solution of eqn (3.104), and (ii) the average of u(r, t) is the macroscopic field, i.e., &(r, t) = (v(r, 0)) (3.116) where (...) means the configurational average of the beads, (. -)= faceywceR)so. . (3.117) Though at first sight it may seem difficult to find the velocity field ®(r, t), the answer is simple: v(r, )=K() r+ > Ar —R,)* Fr (3.118) For this flow, the first condition is obviously satisfied. That the second condition is satisfied is seen by the symmetry argument: in the homoge- neous flow, the average a(n) (= H(r-R,,)« F,) (3.119)72 BROWNIAN MOTION must be a constant vector which can depend only on the tensor K.t Since one cannot construct a vector from a single tensor «, v'(r, ft) must be zero.$ Given the microscopic velocity field u(r, ¢), the velocity of the beads is immediately obtained: Vn =U (Rin t) = (1) * Rin +) Han Fr (3.120) so that the Smoluchowski equation becomes a a ow aU a Sto Daeg Me (maT + VSR) — Dag MO Roll (3.121) This equation describes the change in the distribution function of a system under macroscopic velocity gradient. 3.7.4 Expression for the stress tensor Next we consider how to calculate the stress tensor, say the component Ouz(@ =x, y,Z). To this end we consider a region of volume V in the Fig. 3.8. Microscopic definition of the stress tensor. The stress component 0,:(@ =X, y,Z) is the @ component of the force (per area) that the material above the plane (denoted by the dashed line) exerts on the material below the plane. The force § consists of two parts S® and S®. §® is the force acting through the solvent at the plane, and §” is the sum of the forces that the upper beads exert on the lower beads via a potential. + Strictly speaking a more careful analysis is needed to justify this argument since the sum in eqn (3.119) is ill converged, and depends on the shape of the outer boundary. However, as in the case of dieletrics, the contribution from the outer boundary can be included in the imposed field x(¢), so that in practice the contribution from the outer boundary can be neglected. (See the discussion in ref. 26.) +If there is an external force field (such as gravitational force) 0’(r, t) does not vanish. In such a case, the problem must be treated by a multi-fluid model,” which is not considered in this book.MICROSCOPIC BASIS OF VISCOELASTICITY B homogenous flow, and divide it by a hypothetical plane which is perpendicular to the z-axis (see Fig. 3.8). By the definition of the stress tensor, ,, is given by the force S, which the upper part exerts on the lower part through the plane: Oz = (Sz )/A (3.122) where A is the area of the plane. Now S, consists of two parts, the force S® which acts through the solvent fluid and the force 5) which acts directly between the beads. The former is written as Bue, dv. SO(n) = Jol. (e+ =) +P bx. |6(r.—h), 3.123) where h is the position of the plane defined in Fig. 3.8, and the integral is carried out in the region V. The force S® is given by the force F,,, which the bead 7 exerts on the bead m:t SE(h) =X Franz O(h — Rmz)O(Rnz — ht), (3.124) where 1 for x>0, O@)= {0 for x<0. 6.125) The two © functions in eqn (3.124) restrict the bead n to the upper part and the bead m to the lower part. In the homogeneous flow, the average (S,(h)) is independent of h, so that eqn (3.122) can be written as L L 1 1 : ue = Zr | an(ss(h)) =Zr faw(SPA) +20) G.126 The integral for S® is x= [an [ar na (ve) +5 (ued) + (P) bax (02m) or, ev 2 Toe = J aol (se oe) +3 (eed) +P} G27 re’ + Note that here the size of the beads is taken to be infinitely small. Since rigid bodies of finite size are expressed by the model shown in Fig. 3.76, this does not impose any limitation on the applicability of the present formulation.14 BROWNIAN MOTION By eqn (3.116), xP = [ arn Couns) += + (mts)) + (P) Bue] =V[ns(Kaz + Kea) + (P) Sac]: (3.128) The integral for S® is rewritten as xp=( f JhSPUH)) = © (Fn i GhO(h ~ Rp JOCRye — h)) = (3 Fane (Rre ~ Re) (Rae ~ Rr) (3.129) ixchanging m and n, and using Newton’s third law, F,,, = —Fim, we have XP = (15 [Fano (Bas ~ Rs) ~ Rr) 4 Fama(Rye ~ Rr) ®(Rne ~ Ros) = (8D Fane Rr ~ Rne)[(Rne ~ Rne) + (Re ~ Rr))) = (45 Enna (Rmz — 2): (3.130) -et F,, be the sum of the nonhydrodynamic forces acting on the bead m: En =>) Fan: (3.131) Then eqn (3.130) is rewritten as XD =(-E FanoRne +3, FaneRae) = (> FneRne): (3.132) rom eqns (3.126), (3.128), and (3.132), we finally obtain the stress ensor 1 Oap = s(Kap + Kpa) + (P) Sap — ye (EnaRmp )- (3.133) Che first term in eqn (3.133) represents the stress in the absence of theMICROSCOPIC BASIS OF VISCOELASTICITY oP) beads, and the last term, 1 of}= “ya (Fang) (3.134) ™ denotes the extra stress due to the beads. It is important to notice that £,, is given by o FE, = ~ Gp, hoT ne + U) (3.135) and includes the thermodynamic force dk,T In ¥/OR,,. The necessity of such a term can be understood by studying a special situation. Consider for example that the suspension is in equilibrium under the gravitational field U=-—gz. If the term kgT 91n W/AR,, is not included, the stress tensor is not isotropic, which contradicts the condition that the stress must be isotropic in a liquid in equilibrium. Equations (3.121) and (3.133) together may be regarded as a constitu- tive equation: for a given macroscopic velocity gradient Ky,g(t), the distribution function W is obtained from eqn (3.121) and the stress is calculated using eqn (3.133). 3.7.5. Principle of virtual work The stress formula (3.134) can be put in the form equivalent to the principle of virtual work.° Consider a virtual deformation which displaces the point r, on the material to r, + 5&,grg in a very short time 6¢. In the limit of 60, the velocity gradient Kyg = d€,/5t becomes very large, so that the time evolution of W in the time interval dr is dominated by Kapst ow a Bae Role: (3.136) Hence the change in W by the hypothetical deformation is 8 Rn Let xf be the dynamical free energy of the system: éw= e ét=-> + (de-R,,P). (3.137) A= [eceweer InW+U) It can be shown that the stress off} due to the Brownian particles is +This statement is not true if there are rigid constraints which produce infinite forces. See the discussion in the next section.16 BROWNIAN MOTION related to the change in the dynamical free energy by 5A = OSE qgV. (3.138) The proof is straightforward. The change in the dynamical free energy is 64 = fateresTow In + kgTSW + USW). (3.139) Substituting eqn (3.137) and using the integration by parts, we have oat = [ary ~ se (BeapRag YM kaT in V+ kyT + U) = [ory BeapRypW 5 e— (kaT In W + v) = —6£0p.>) (Rmp Fina) = be p08}V. (3.140) Equation (3.138) represents the principle of virtual work. 3.8 Systems with rigid constraints 3.8.1 Introduction We now consider the case where the beads are subject to rigid constraints. This is necessary to deal with the problems of suspensions of a rigid body, or polymers with rigid constraints (such as the rodlike polymer, or the freely jointed model), but the reader who is interested only in flexible polymers can omit this section. Here we shall consider only the holonomic constraints} which can be expressed as equations connecting the coordinates of the beads: GCUR})=0, p=1,2,...,N. (3.141) Examples of systems which have such constraints are: (i) the freely jointed model (Fig. 3.72), in which the beads are successively connected at constant distance b, so that (R,—R,1)?- 6? =0, 1 =1,2,...,N. (3.142) (ii) the rigid body model (Fig. 3.7b), in which the mutual distance between the beads is fixed. When the constraints are introduced, the force F,, is no longer a known function of {R} and must be determined by the equation of motion. This tthe topological constraints that the chains cannot cross each other do not belong to this lass,SYSTEMS WITH RIGID CONSTRAINTS 7 situation is familiar in classical mechanics, where the forces are determined using the condition that the solution of the equation of motion must satisfy the constraints. The same rule applies to the present system except that in our problem, the equation of motion is not Newton’s equation, but the hydrodynamic relation Vn = Rn tS Hn Ere (3.143) From a practical viewpoint, there are two ways of doing this. One is to introduce generalized coordinates which are independent of each other, and specify the configuration of the beads uniquely.” This method is suitable when the positions of the beads are expressed explicitly as a function of such coordinates. For example, rigid body problems are conveniently handled by this method. In this example, the generalized coordinates will stand for the three components of the position vector of the centre of mass, and the three Euler angles specifying the orientation of the rigid body. However it is impractical to apply this method to the freely jointed model. Another method is to use the Lagrangian multipliers for the con- straints. This method is complementary to the first, and indeed has been successfully used for the freely jointed model (and semiflexible polymer models”), Here we shall describe the formal part of the methods,” leaving the detail of calculation to the literature.°~ 3.8.2 The method of generalized coordinates Let {Q}=Q1, Q2,..., On, be the set of generalized coordinates. The position R,,, is expressed as a function of {Q} as Rr=Rn({Q}), m=1,2,...,N. (3.144) If the velocity of the generalized coordinate is V,, the velocity of the particle is given as MN Oi Mn 25 a. In this section we shall employ the summation convention, and write eqn (3.145) as (3.145) aR 90. To obtain F,,, we use the principle of virtual work.” Let us consider Vn = Va. (3.146)2B BROWNIAN MOTION the work necessary to change Q, by 5Q,, which is 8(U + kpT In W) = ls: (U+kgT in woe. (3.147) Alternatively, the work can also be calculated using the force F,, and the displacement 5R,,, caused by the change in Q,, i.e. 6(U + keT nV) = —F,,- OR, (3.148) where ORm OR, = 30. 6Q,. (3.149) From eqns (3.147)—(3.149) we have OR, Fn* 20, 30, (kaT In ¥ + U). (3.150) Equations (3.143), (3.145), and (3.150) determine V, and F,,. To obtain F,, and V, explicitly, we first solve eqn (3.143) for F,: FB, = (H™)am * Vn — K+ Rin) aR, =(H-? mn . = (Ham * ( 30, * R,,) (3.151) where (H™)am is the inverse Of Hyn! (Ham * Hy = nile (3.152) ~s this into eqn (3.150), we have =), 2 ayy 50. CA am [se -K-R, m |= 39, Ut keT nV), (3.153) Putting aR, aR, Bop = (Ham 3.154) ae = 39, A) 30; (3.154) define a FP =—30(U+keT In), (3.155) and VY = hase Yam 0 Rims (3.156)SYSTEMS WITH RIGID CONSTRAINTS 79 we can rewrite eqn (3.153) as (A) aa(Vo — VY) = FO (3.157) which can be solved using /,,, the inverse of (h~'),», giving Va = VS) + hag FP = hy 2 (U + he IB) 4 V. (3.158) ‘ab a 0, Hence =(H" (= Rm grip) 4 Rm yon Rn): . F= (Ham * 30, feFb sey K- (3.159) In the generalized coordinate space, the conservation equation is written as*° aw ot Vg a=1 9Q, where g is the determinant of the matrix g,, defined by Wev.w] (3.160) ORin Rm 8 = 30." 30, 20, (3.161) Thus the diffusion equation is obtained ast = 1 [ ( ow | aU )- | = += Ory |. 3.162 ~yga0, Vel (tI 59, +39, 8) YEW] 180 This equation was first given by Kirkwood.” 3.8.3 The method of Lagrangian multipliers The constraints can be treated by the alternative method of Lagrangian multipliers.” Here {R} are regarded as independent variables, and the constraining forces are explicitly added to the right-hand side of eqn (3.135): =o 3G Fog = ~ 5p (koT NW + U) + dy EE (3.163) +The factor Vg appears naturally as a result of the coordinate transformation, but it is unnecessary from the viewpoint of the general theory of Brownian motion, Indeed, in terms of B= WV and U=U- ‘ In Vg, eqn (3.162) can be written in the form of eqn (3.21) ov = we, Wy Vy a 30. 3 [ha( kyT.~ + ¥) ~vs e], 3Q, 3Q, which includes no Vg factor. Notice that the theory here can be presented without using the Riemannian geometry.80 BROWNIAN MOTION The unknown parameters A, are chosen such that the velocity V,, determined by eqn (3.143) satisfies. aC, 2G yy = aR Yam 0 (3.164) Using eqns (3.143), (3.163), and (3.164), ac, 3 -. aC, =(h-). —4. -_— —~(h-'). —2-K- Ay = (i Nye ge" Hon” Sq knT IE + 0) — pg pe KR (3.165) where (f~'),, is the inverse of the matrix AG yy fina = 3g. Ham * Se (3.166) The Smoluchowski equation is obtained if eqn (3.163) is substituted into the continuity equation: wa at oR, oP) a a a aC, a =. . — —4. — Th w+ ar, fm “(aR ar, ° ap, Me ae? in +U) a a, py aC, ) 4 _ . . _ ~—2 Ke . 1 5+ (Wg Hn SEE Np Set HR G.187) 3.8.4 Elastic stress and viscous stress If the distribution function is obtained from eqns (3.162) or (3.167), the Stress can be calculated from eqn (3.134) in which F,, is now given by either eqn (3.159) or (3.163). In either expression, the force F,, consists of two terms, one independent of « and the other proportional to K: Ena = FE) + EmapyKpy- (3.168) For example in the generalized coordinate representation, eqns (3.159), (3.155), and (3.156) give aR, (E) 1 . (E) P= (Hmm 56" hab aR a =—-(H") .—,,—~ (Amn 20, he 39, U t ket In) (3.169)SYSTEMS WITH RIGID CONSTRAINTS 81 and aR, 3Q. OR, OR; =(H-).. (2 1mm (50550, The force F® represents the force due to the potential, and &,,:#« represents the effect of the constraining force. Hence the excess stress due to the Brownian particles is generally written as B= (Hn *( Vo Ry) (Hy KR) «-R,). (3.170) of} = 0} + off (3.171) where 1 N OP = — 7D (FndRnp)s (3.172) m=1 N ofp = -4 2, (EnnauvRnp ) Kuv(t)- (3.173) We shall call o® the elastic stress and o™ the viscous stress. The viscous stress is proportional to the current velocity gradient «(t) and can be written as OB (2) = 1 arKuv(t), (3.174) while the elastic stress does not include «(t) explicitly. Phenomenologically, the viscous stress is the stress which vanishes instantaneously when the flow is stopped. On the other hand the elastic stress does not vanish until the system is in equilibrium. The elastic stress is dominant in concentrated polymer solutions, while viscous stress often dominates in the suspensions of larger particles for which the Brownian motion is not effective. Whichever stress dominates, the rheological properties can be quite complex since both of and %},, are functions of the configuration of the beads and therefore depend on the previous values of the velocity gradient. Note that the viscous stress a only appears in the system with rigid constraints. The actual formula for the stress tensor is complicated. However, a neat expression is obtained in the form of the principle of virtual work. By a straightforward calculation, it can be shown® + It must be mentioned that the distinction between the elastic stress and the viscous stress is a matter of time-scale. The elastic stress which has very short relaxation times, cannot be distinguished from the viscous stress. A discussion from a phenomenological viewpoint is given by J. D. Goddard, J. Non-Newtonian Fluid Mech. 14, 41 (1984).82. BROWNIAN MOTION (i) The viscous stress off or the coefficient ny, are related to the energy dissipation function W(LV}) = > (Van — Rn) (HY mn (Va R,) (3-175) . (Mini W) = nGuvKopKuvV (3.176) Note that Mini W does not vanish since V,, must satisfy eqn (3.145) (or eqn (3.164)) for the system with constraints. (ii) The elastic stress is related to the change in the dynamical free energy # caused by the instantaneous deformation d&g = Kapdt: =o8 where 6s = 0®5EapV, (3.177) 6A = [acryceeTow In + kg TOW + USW). (3.178) Here 5¥ is given by 3 3 =~ SR + (Vnde¥) (3.179) where Y,, is the velocity which minimizes eqn (3.175). 3.85 Variational formulation The theory described above can be formulated in the form of a variational principle® which is similar to the Lagrangian formulation of classical mechanics. The advantage of this formulation is that it is independent of the coordinate system, and allows a great flexibility in choosing coordinates. Let us regard V,, as a function of {R}, and consider the functional defined by K=iwis (3.180) with We [dERYYD Vn KR) (Han (Ve KR) (3.181) mn and a= J A{RY(kaTY nV + kg TH + WU), (3.182) where W is defined by =-y2.( w= 25 oR. (VirP). (3.183)EIGENFUNCTIONS OF SMOLUCHOWSKI EQUATIONS 83 It is then shown that: (i) The minimization of K for all variations of V,, subject to the constraints (3.145) (or (3.164)) determines the time evolution of W to be a . 3p PURE 1) = B({R}). (3.184) (ii) The stress is given by oO Mini K Oop = ra + 1y(Kep + Kpa) + Pdap (3.185) ap Some applications of this principle are given in ref. 6. In closing this section it is worthwhile to stress again that no condition has been imposed on the concentration of the particles. Therefore the theory will apply to concentrated suspensions as well as dilute suspensions. Appendix 3.1 Eigenfunctions of the Smoluchowski equations In this Appendix we discuss the eigenfunction expansion of the Smol- uchowski equation (3.21). For the sake of simplicity, we represent the set of variables {x} =(x,,...,Xy) by x, and write the Smoluchowski equation as ow a —TQ@)¥(, 4) G.1.1) where I is a linear differential operator: a ow aU rey¥=-S Lemke T + v2"). The conjugate of the operator I(x) is denoted by I'*(x): for any W(x) and @(x) (3.1.2) | ax ¥(x)(P(x)@(x)) = | ax(F*(x)P(x))®(x). (3.13) From eqns (3.1.2) and (3.1.3), P*(x) is obtained as a au, a + — — — —_ — Mea)= 3 (ber = 5) nm oy G14) Let W,(x) be the right-hand eigenfunctions and y,(x) be the left-hand eigenfunctions: T@)¥, (x) =2,B,(x) (3.1.5)84 BROWNIAN MOTION and T* (x) pp(x) = A, pp(x). (3.1.6) The eigenfunctions are chosen to be orthonormal so that J ALY, (x) Yg() = Spy. G17) It is easy to prove by direct substitution that the right-hand and left-hand eigenfunctions are related by W(x) = Veg) ¥p(x) (3.1.8) Hence eqn (3.1.7) can be written as [aves = (Vp a )ea= Spa (3.1.9) where {...)_q denotes the equilibrium average. The equilibrium distribution function V., is an eigenfunction with eigenvalue 0, which will be denoted by the suffix p=0, so that w= 1. All the other eigenvalues are positive. To show this we multiply both sides of eqn (3.1.5) by y,(x) and integrate over x: A, [ dxaqy3= | dv, TPeavp) (@.1.10) Using eqn (3.1.2) and the integral by parts, the right-hand side is rewritten as ths =kgT i a> a (1.11) which is positive due to eqn (3.18). Therefore all 4, are positive except Ag=0. Now the distribution function W(x, 7) can be expanded by the eigenfunctions as wx, 1) =D (Vo) Fos) (3.1.12) where by eqn (3.1.9) a,(t) = { dey, (x) P(x, t). (3.1.13) Since Wo = 1, and (x, £) satisfies the normalization condition, a= 1. 3.1.14)RELATIONSHIP BETWEEN EQUATIONS 85 From eqns (3.1.1), (3.1.5), and (3.1.12), a,(£) satisfies £ ay(l) = ipa) @.113) which gives a,(t) = a,(O)exp(—A,/). (3.1.16) Thus W(x, t) =D a,(O)exp(—A,t) Pp (x) Peq(x) (3.1.17) = Wos(x) + 2 o(OexP(— Apt) Vp (#) Poa). (3.1.18) After a long time, the underlined terms become very small, and W(x, t) becomes the equilibrium distribution function. As a special case of formula (3.1.18), the Green function G(x, x’; t) which satisfies the initial condition (3.46) is obtained as G(x, 2°51) = exp(—Apl)Yp(x)Yp(x'eg(x)- -G.1.19) Appendix 3.11 Relationship between the Langevin equation and the Smoluchowski equation Here, we shall show that the probability distribution of the solution of the Langevin equation (3.27) satisfies the Smoluchowski equation (3.16). First we consider the case when € is independent of x. We write the Langevin equation (3.27) as 2 = V(x) + 08(t) (3.11.1) where v@)=-12U@) 5 kp 12 C ox’ (* =D”, (3.11.2) and g(t) is a random variable satisfying (8) =0, (g@g(t’)) = 26(¢-2'). (3.11.3) Suppose that the particle was at x at time ¢. The displacement & of the Particle in a very short time At is easily obtained from eqn (3.II.1) since86 BROWNIAN MOTION V(x) can be regarded as constant in a short time interval: t+Ar £=V@)Al+o | dtyg(t). @.1L4) t Since & is a linear combination of the Gaussian random variable g(t), its probability distribution (&, At; x) is also Gaussian, the moments of which are obtained from eqn (3.II.3) and (3.11.4): (&) =V(x)At, (3.11.5) (E-(£)P) =? | dt, | dta(g(t,Jg(t,)) =20°At=2DAt. (3.11.6) t Hence (3.11.7) (&, At; x) = (4aDAt)"'” exp| -E—Y@)40"), 4DAt If the probability that the particle is at x at time t is W(x, ¢), then the probability that it is at x at time ¢ + At is given by Wx, + At) = fos[axroce ~ x! — B)O(E, Atsx'WOr’, 1) _E=V(e- 8)ary = | d&(4aD At)!” exp| 4DAt |ve -&0). (3.11.8) Since the integrand has a sharp peak at & = 0, the integral is evaluated by expanding V(x — §) and W(x — &, 1) with respect to &: +) ((1 +Zar)g - var} Wr, t+ A= [os@xpan~? exp] — aDAt a & x (1-82 +32 )we, 0, Neglecting the terms of order A?’ and higher, we get dv aw ry w sean= (1-2 ) -avee oY 6. (x. ) ce Atyw — Atv ox +DAt a (3.11.9)RELATIONSHIP BETWEEN EQUATIONS 87 Collecting terms of order At, we get aw ow oo, 1a /au oo 2 venyes 9) +D 53 =Doa+t is (> w) (3.11.10) This agrees with eqn (3.16) Next we consider the case when € depends on x. The Langevin equation (3.38) gives f= V(x) + a(x)g(t) + a(x) “, (3.11.11) where o(x)? =kgT/€(x). Let X(t) be defined by xO 1 X(t) = | dx’ 3.11.12 © ou) @.11.12) The Langevin equation for X is then obtained from eqn (3.11.11) == V(X) +8() (3.11.13) with ; V do Vwo=ste (3.11.14) Equation (3.11.13) is the Langevin equation studied earlier. Thus, the probability distribution function for X satisfies we ae .. o-alS- ). (3.11.15) From eqn (3.II.12), it follows that W(X, 1) = a(x) V(x, 1) (3.11.16) and 9 4F ax Can" (3.11.17) From eqns (3.II.15)—(3.II.17), we can show that W satisfies ow a 2 ow a1 ot 5” -v¥) a5 oe ee (eT y+ w2) (3.11.18) which is the Smoluchowski equation.88 BROWNIAN MOTION Appendix 3.III The Oseen tensor Defining the Fourier transform as uaz {arvinet”. . (3.01.1) we can rewrite eqns (3.102) and (3.99) as —n,k’v, — ikP, = —g, k+v,=0, (3. 111.2) which gives 1 2. alow A i @.1IL.3) where & indicates a unit vector in the direction of k. Hence vi= farcete —r’)-a(r') (3. 111.4) where 1 1 a . HO= ap i dk el kBexp(—ik-7). IIL) Since the tensor H(r) depends on the vector r only, it can be written in terms of the scalars A and B and the unit vector 7 parallel to r, as Hap(t) = Adap + BraFp. (3.11.6) The scalars A and B are determined from the two equations Hag =3A+B, Haphafp = A+B, (3.11.7) ie., 3A+B=— [dk -ik 3.18 “Qn nie exp(—ik - r) (3.11.8) and . -_! 1-k- A+B =o [ak Gem em-ik-9. GIL) The integrals are easily evaluated by introducing the coordinates t= k «7 and & =|k| |r| to give _ 2 fag, f ian 1 fyesing 34+ B= | om I ar exp(—i8) == J a 1 = 2an,r (3.11.10)REFERENCES 89 and o 1 A+B “rl — Pexp(—i&t) sin§ 1 “ign faz(. +m) — 4nn,r G.OL11) From eqns (3.III.10) and (3.III.11), A and B are obtained as 1 A=B=e (3.1IT.12) Hence 1 na H(r) Sane (1+ #9). (3.11.13) References 1. Wax, N., Noise and Stochastic Processes. Dover Publishing Co., New York (1954) represents a collection of classic papers. . Lax, M., Rev. Mod. Phys. 32, 25 (1960); 38, 541 (1966). . Kubo, R., Rep. Prog. Phys. 29, 255 (1966). . Einstein, A., Ann. Physik. 17, 549 (1905); 19, 371 (1906). See also Investigation on the Theory of the Brownian Movement. E. P. Dutton and Copy Inc, New York (1926). 5. Batchelor, G. K., An Introduction to Fluid Dynamics, Chap. 4. Cambridge Univ. Press (1970). 6. Doi, M., J. Chem. Phys. 79, 5080 (1983). 7. van Kampen, N. G. Stochastic Processes in Physics and Chemistry. North- Holland, Amsterdam (1981). 8. Risken, H., The Fokker—Planck Equation. Springer, Berlin (1984). 9. Zwanzig, R., Adv. Chem. Phys. 15, 325 (1969). 10. Fixman, M., J. Chem. Phys. 69, 1527 (1978). 11. Rainville, E. D., Special Functions. Macmillan, New York (1960). 12. Recent researches on the dynamics of interacting colloidal particles can be seen in Faraday Discuss. Chem. Soc. 76 (1983). 13. Happel, J., and Brenner, H., Low Reynolds Number Hydrodynamics. Prentice Hall, Englewood Cliffs, N. J. (1965). 14, Batchelor, G. K., J. Fluid. Mech. 74, 1 (1976). 15. Felderhof, B. U., Physica 89A, 373 (1977); J. Phys. All, 929 (1978). 16. Mazur, P., and van Saarloos, W., Physica 110A, 147 (1982). 17. For the theology of suspensions see Maron, S. H., and Krieger, I. M., in Rheology, Vol 3. (ed. F. R. Eirich) Academic Press, New York (1960); and Bauer, W. H., and Collins, E. A., in Rheology, Vol 4. (ed. F. R. Eirich) Academic Press; New York (1967). rN recent review on thixotropy is given by Mewis, J., J. Non-Newtonian Fluid Mech. 6 1 (1979). Pun90 18. 19. 20. 21. 22. SRR 31. 32. 33. BROWNIAN MOTION For the rheology of granular materials, see Shahinpoor, M., (ed.) Advances in the Mechanics and the Flow of Granular Materials, vol. I and II. Trans. Tech. Publications, Clausthal-Zellerfeld, Germany (1983). For the rheology of polymeric liquids, see Bird, R. B., Armstrong, R. C., and Hassager, O., Dynamics of Polymeric Liquids, vol. 1. John Wiley, New York (1977). Kirkwood, J. G., Rec. Trav. Chim. 68, 649 (1949); see also Kirkwood, J. G., Macromolecules, in Documents on Modern Physics. Gordon & Breach, New York (1967). Burgers, J. M., Second Report on Viscosity and Plasticity, Chap. 3. Amsterdam Academy of Sciences, North-Holland, Nordermann (1938), Kuhn, W., and Kuhn, H., Helv. Chim. Acta 37, 97 (1944); 38, 1533 (1945); 39, 71 (1946). . Kramers, H. A., J. Chem. Phys. 14, 415 (1946). 24. A collection of classic papers is given in Hermans, J. J., (ed.) Polymer Solutions Properties II, Hydrodynamics and Light Scattering. Dowden Hutchinson & Ross Inc., Stroudsburg, Pa. (1978). Also a history of the development of the theory of suspensions is described in Frisch, H. L., and Simha, R., The viscosity of colloidal suspensions and macromolecular solutions, in Rheology Vol 1 (ed. F. R. Hirich). Academic Press, New York (1956). . Batchelor, G. K., J. Fluid Mech. 41, 545 (1970). . Peterson, J. M., and Fixman, M., J. Chem. Phys. 39, 2516 (1963). . See for example Meyer, R. E., (ed.) Theory of Dispersed Multiphase Flow. Academic Press, New York, (1983). . See for example, Goldstein, H., Classical Mechanics. Addison-Wesley, Reading, Mass (1959). . Erpenbeck, J. J., and Kirkwood, J. G., J. Chem. Phys. 29, 909 (1958); 38, 1023 (1963). . Fixman, M., and Kovac, J., J. Chem. Phys., 61, 4939 (1974); ibid. 61, 4950 (1974); ibid. 63, 935 (1975). Hassager, O., J. Chem. Phys. 60, 2111, 4001 (1974). Nakajima, H., and Wada, Y., Biopolymers 16, 875 (1977); 17, 2291 (1978). Doi, M., Nakajima, H., and Wada, Y., Colloid Polym. Sci. 253, 905 (1975); 254, 559 (1976). 4 . Iwata, K., J. Chem. Phys. 71, 931 (1979). 35. ian review is given by Yamakawa, H., Ann. Rev. Phys. Chem. 35, 23 1984), . See for example, ref. 19 Appendix A.4 DYNAMICS OF FLEXIBLE POLYMERS IN DILUTE SOLUTION 4.1 The Rouse model 4.1.1. Dynamics of a polymer with localized interaction Having seen the general background of Brownian motion, we shall now discuss the dynamics of a polymer in solution. As we have seen in Chapter 2 the static properties of a polymer can be represented by a set of beads connected along a chain. It is natural to model the dynamics of the polymer by the Brownian motion of such beads. Such a model was first proposed by Rouse’ and has been the basis of the dynamics of dilute polymer solutions. Let (Ri, Ro,..., Rw) ={R,} be the positions of the beads (see Fig. 4.1a). The equation of motion of the beads is described by either the Smoluchowski equation “- eqn (3.21) or ce ow Sr ap, Hen [bo Se or the ee equation ee eqn (3.39)) Sul) =D Hm (— Set tal)) +4haT D sp" Howe 4.2) In the Rouse model, the excluded volume interaction and the hydrodynamic interaction are disregarded and the mobility tensor and the interaction potential are written as +2U wy] (4.1) 7m Hon “+ Sam (4.3) and k N U=5 2 Rn Rea)? (44) with 3kaT k= Be (4.5) In this model the Langevin equation (4.2) becomes a linear equation for R,,. For internal beads (n =2, 3,..., N—1), aR, Ge ~KORn~ Rass ~ Rua) the (4.6)92 DYNAMICS OF FLEXIBLE POLYMERS Ry fa) Fig. 4.1. (a) Rouse model and (6) local jump model. For the end beads (n = 1 and N) dR dR, STR Ra) th, SE —k(Ry— Rv th (4.7) The distribution of the random force f, is Gaussian, characterized by the moments given by eqns (3.40) and (3.41): (h(t) = 0, (fra( Mme (t')) = 26k TS nmdap5(t — t'). (4.8) As in the case of the Gaussian chain, the suffix n in the Rouse model can be regarded as a continuous variable. In the continuous limit, eqn (4.6) is rewritten as (see the transformation rule given in Table 2.1, Section 2.2) aR, _, FR, SGpak arth (4.9) To rewrite eqn (4.7) in the continuous limit, we note that eqn (4.7) is included in the general equation (4.6) if the hypothetical beads Ry and Ry, are defined as Ro=R,, Ry =Rv (4.10) which become, in the continuous limit, OR, =0. 9R,, = an Inno” On |n=n - (4.11) Also, the moments of the random forces are now given as (ha(t)) = 0, (free(t)fnp (t')) = 26keTS(n — m)5ap5(t—t'). (4.12)THE ROUSE MODEL 93 Equations (4.9), (4.11), and (4.12) define the continuous Rouse model. The results of the discrete model and the continuous model agree with each other for properties on a long time-scale, but do not for short times. The discrepancy, however, has no serious physical significance since the description of the polymer by discretized beads is an artefact, and the results which depend on the discrete nature of the beads have no validity. It should be emphasized that the essence of the Rouse model is in the universal nature of the modelling of the dynamics of a connected object. The central assumption in the Rouse model is that the dynamics is governed by the interactions localized along the chain. In fact, if one assumes a linear Langevin equation for R,, with localized interaction, one ends up with the Rouse model in the long time-scale behaviour. To see this, consider the general form of the linearized Langevin equation dR, _ dt where A,,, is a constant matrix representing the interaction among the beads, and g,, is a random force. Since the system is homogeneous, Ajm depends only on n — m, so that eqn (4.13) can be written as aR = DAmRnim + Bn (4.14) dt “i with A,, =A,,.+m- In the long time-scale motion, R,, varies slowly with n, which allows R,,4m to be expanded with respect to m, giving (4.13) 3 e D AmB om =D Am(Ry + mS Ry +4" Rt. ) = agR, +0, = R, +u25R, +. (4.15) where a= > An a= > mAnz, a Dd mA, (4.16) The assumption of local interaction guarantees that these sums converge to finite values. The coefficient aj must vanish since the equation must be invariant under a spatial translation (R, > R, +r), and a; vanishes since A,, is an even function of m (because the polymer cannot distinguish head and tail). Therefore, the asymptotic behaviour of eqn (4.14) is given by a e BRn= 425 Rn + Bat), (4.17) which is equivalent to eqn (4.9).94 DYNAMICS OF FLEXIBLE POLYMERS The Rouse model displays the general features of any model which assumes local interactions. One can conceive of other dynamical models.”* For example, one can start from the freely jointed chain, and simulate its dynamics by allowing the local jump process depicted in Fig. 4,16. This model can be shown to give the same results as the Rouse model for slow modes (see Appendix 4.1). It is generally believed® that the Rouse model represents the long time-scale behaviour of the ‘local jump’ model in the same way as the Gaussian chain represents the large length-scale properties of a polymer which has only short range interaction.t 4.1.2 Normal coordinates Let us now study the consequence of the Rouse model. Equation (4.9) Tepresents a Brownian motion of coupled oscillators. A standard way of treating such a system is to find the normal coordinates, each capable of independent motion. It is shown in Appendix 4.II, that in terms of the coordinates X, defined by N i Px) i = x, no cos aT )Ra(t) with p=0,1,2,..., (4.18) eqn (4.9) can be rewritten as a bo ap Xe = —heXe +h (4.19) where fo =NE and 6,=2NE for p=1,2,... (4.20) ky = 27?kp?/N = Sake ry for p=0,1,2,... (4.21) Nb and the f,’s are the random forces which satisfy (foa) =0, fra Ofap(t')) = 28 pp SapbokaTO(t—t'). (4.22) Since the random forces are independent of each other, the motions of the X,’s are also independent of each other. Thus the motion of the polymer is decomposed into independent modes. The time correlation functions of the normal coordinates can be calculated immediately from eqn (4.19) by using the results of Section + However, this statement is a conjecture. Usually the dynamics of the local jump model becomes a nonlinear equation for R,, so that the proof given above does not apply to the general local jump model. In fact a counter example was given by Hilhorst and Deutch.”THE ROUSE MODEL 95 3.5. For p>0 kT (Xpa(t)Xop(0)) = Spq5o8 xe exp(—t/t,) (4.23) where ° 4 Tp) = T/p* (4.24) any tN*b? (4.25) ne i 3xkyT On the other hand for p =0 (ol) ~ Kewl) = Xeg(O)) = Bap AEE t= Bag ETH, (4.26) The inverse transform of eqn (4.18) is R,=Xy+2 > x cos (Pa) (4.27) Let us now consider the physical significance of the normal coordin- ates. The coordinate Xp represents the position of the centre of mass a + dnR,, =Xo. (4.28) Thus the mean square displacement of Rg is calculated from eqn (4.26) as ((Re()—Re@)) = X ((Xoalt) — Xoa(0))) = ott re (4.29) OKY,Z The self diffusion constant of the centre of mass is defined by Do = lim = ((Re(®) ~ Ro(0)?). (4.30) From eqns (4.29) and (4.30) Do = kat (4.31) 'G NE 7 Pe The normal coordinate X, with p > 0 represents the internal conforma- tion of the polymer. Consider for example the end-to-end vector P(t)=Ry(t) — Ro(t) (4.32)96 DYNAMICS OF FLEXIBLE POLYMERS which is expressed by X, as P)=-4 DD X(t). (4.33) p:odd integer The time correlation function (P(t) - P(0)) is calculated from eqns (4.23) and (4.33) as (PO) PO) =16 E (%0-X (0) (434 piodd =16 Ye exp(—t/2,) prota Kp Feel). 4.38) Equation (4.35) indicates that the motion of the end-to-end vector is mainly governed by the first mode X,. In general, X, represents the local motion of the chain which includes N/p segments and corresponds to the motion with the length-scale of the order (Nb?/p)'”. The rotational relaxation time t, of a polymer can be defined by the longest relaxation time of the correlation function (P(t) - P(0)):+ (P(t): P(O)) «exp(-t/t,) for t®t,. (4.36) From eqn (4.35) we see that EN NAEP (4.37) T= Since N is proportional to the molecular weight M, eqns (4.31) and (4.37) indicate that Dg and +, depend on the molecular weight M as De=M", t,~ M?, (4.38) This prediction is not consistent with experimental results, which, in © conditions, are summarized as Dg&M~?, 5, M32, (4.39) This failure comes from the neglect of the hydrodynamic interaction, which will be discussed in the next section. Because of this failure the Rouse model is now regarded as inappropriate as a model in dilute solution. However, the model is conceptually quite important, and also it has turned out that it is a useful model in the dynamics of polymers in melts, which will be discussed in later chapters. t It has been shown*? that the most probable shape of the Gaussian chain is not spherical but an ellipsoid, the long axis being, on average, parallel to the end-to-end vegtor. Thus it is possible to talk about ‘rotational motion’ even for fiexible polymers.THE ZIMM MODEL 7 4.2 The Zimm model 4.2.1 Zimm model in @ conditions To describe the dynamics of polymers in dilute solution, we have to take into account the hydrodynamic interaction, which is expressed by the mobility matrix calculated in Chapter 3, An = UE 1 rn Han = Bam, Mreol [RimPam +4] for nm (4.40) where fam = R, — Ry and Fyn is the unit vector in the direction of Fy». For the tensor eqn (4.40), it can be shown that 8 Be * Han = 0. (4.41) Thus the Langevin equation (4.2) becomes a oU SRo= 2 Hm (Ze ttn(d)- (4.42) In particular, for the © condition, eqns (4.4) and (4.42) give (in the continuous limit) 3 e S RoE Ham (5a Rn + fl) (4.43) This model was first presented by Zimm.'° Since H,, is a nonlinear function of R,—R,», eqn (4.43) is quite difficult to handle. To simplify the analysis, Zimm’° introduced the preaveraging approximation, which replaces H,,,, by its average, Han (Ham) = [4 (Re} Han YE} 8 (4.44) If we are considering problems near equilibrium, which is the case in the subsequent part of this chapter, we may use the equilibrium distribution function Y.,({R,}) in the average of eqn (4.44), Hn > (Hh) ea {Ry} Man BoC (Ra). (4.45) Since the distribution of #,,, is independent Of |fiml, (Hhm)eq is written as 1 1 (Honea ™ Sm, (ial) Fin + Pe (4.46)98 DYNAMICS OF FLEXIBLE POLYMERS Using Gta) ea™ 5 (447) we have (Hm dea= (ea (4.48) “62, \|Rn — Rm , In the © condition, the distribution of R, —R,, is Gaussian with the variance |n — m| b”; hence { 3 a 3r? i (Ham dea = [ersme(s, |n —m| 7) exo( ~2|n—m| z) 620, I "Gh = mp yb = te mh (4.49) Thus in the preaveraging approximation, eqn (4.43) becomes a linear equation for R,,, 2 = R,(0) = = h(n - m)(k 25 Ral) +fn(0)) : (4.50) At first sight this approximation may appear quite crude. However, it has been shown that the results of this approximation are not very different from those of more sophisticated calculations which will be described later. Note that since h(n—m) decreases quite slowly (h(n —m) «(|n— m|)~12), the moment a, of the interaction matrix A,,, in eqn (4.16) does not converge. Thus in the Zimm model the interaction among the segments is not localized. This gives the qualitative difference between the Rouse model and the Zimm model. To analyse eqn (4.50) we rewrite it in terms of the Rouse normal coordinates X, defined by eqn (4.18): 2X) = 5 he(—heXe +h) 4.51) where k, is defined by eqn (4.21) and hiya = ju fom fam cos( PX" )cos(\i(n —m). (4.52)THE ZIMM MODEL 9 From eqns (4.49) and (4.52) h,, is calculated as bie | f dmh(m)cos( 2 )ens( 2% +) (4.53) = ea (P22) Fated 2) — cos(P2)sin( en) [annem Sr) (4.54) The underlined integrals converge quickly to the following values if g is large: am) VN J dmah (meos( 2 N = G'g)™,b (4.55) and _ (gam _ J dmh (mm)sin( 25" ) 0. If we replace the integrals by these asymptotic values, we obtain VN ‘pan qaun’ Moa = aoa | cos N )eos( N ) VN 1 = Gap) ?nb BIN 5p (4.56) This equation shows that h,, is nearly diagonal. Thus if we neglect the off diagonal component of h,,, we have an equation which has the same structure as that of the Rouse model: bp 2x0) =-kX, +h) (4.57) where bp = (Ipp)~1 = (1223)! (N2p)”? (p=1,2,...) (4.58) and 2eyT k= TET» (p=0,1,2,...) (4.59) Equation (4.58) is not correct for p =0, but fo is immediately calculated100 DYNAMICS OF FLEXIBLE POLYMERS from eqn (4.52) as So (han)! = [a [anf daicin — ml) | =4(6n°)""q,bVN. (4.60) Given €, and k,, the diffusion constant and the relaxation times are obtained as kaT 8k_T kpT = = = 19 4.61 Dome 3G) nbVN aR @s) and ; Tp = Cp/k, = 1p? (4.62) “ t = = TYNE) _ 4 45 128 (4.63) "NV Ba)keT kyT" ° where R = VNb has been used. Equations (4.61) and (4.63) predict the molecular weight dependence of Dg and t,, De«N/A, b-bd” (4.85) any physical quantities, which may be static or dynamic, are changed as A>WA. (4.86) The exponent v is the same as that which appeared in the static scaling (v is 1/2 in © solvent and about 3/5 in a good solvent). Though experimental confirmation of dynamical scaling is not so satisfactory as static scaling,j and there is a delicate theoretical problem + The situation is similar to that in critical phenomena, where although accepted, dynamical scaling has often been found to be more delicate and less universal than static scaling. In the polymer problems, the scaling prediction is largely supported by various experiments, but minute discrepancies have been found, which, at this stage, it is difficult to tell whether they are due to the experiments being done outside the scaling regime, or they indicate the failure of the scaling (see the discussion in ref. 21, Chapter 6). The renormalization group calculation by Oono™ indicates that the dynamical scaling is not rigorous, though the error is usually quite small.104 DYNAMICS OF FLEXIBLE POLYMERS related to the topological interaction which will be discussed later, the dynamical scaling is generally believed to hold for polymers of very large molecular weight. To show the usefulness of the scaling argument, let us consider the diffusion constant Dg. As indicated by the Zimm model, the parameters appearing in the problem are N, b, kgT and n,. From the dimensional analysis, Dg is written as kaT Do= TI): (4.87) Since Dg is invariant under the scaling transformation, = NIA . eal iN) = th ek 1 ): (4.88) For this to be true for arity A, f(N) must have the following form f(N) = numerical constant * N~’. (4.89) Hence keT ,_, _ keT Dg = constant N7Y =, . 1c = constant * nb aR, (4.90) By similar argument, we can show that the rotational relaxation time of such a chain is _nN°b)> _ 1,R3 ” kaT k,T° Equations (4.90) and (4.91) agree with eqn (4.83). For the Rouse model, which has no hydrodynamic interaction, a similar scaling property exists. When the 4 segments are grouped, the parameters in the Rouse model change as (4.91) N—> NA, b> bi’, o> fA. (4.92) Under this transformation, a physical quantity A changes as A> MA. (4.93) From this it can be shown that Dg and t, depend on N as Dg&N™, 1, ¢N1*2¥, (4.94) These results have been checked by computer simulation.”* 4.4 Dynamic light scattering The Brownian motion of, polymers can be experimentally studied by dynamic light scattering.” By measuring the time correlation of theDYNAMIC LIGHT SCATTERING 105 intensity of the scattered light, one can extract the dynamical structure factor tk =< 5 (explik (R,() R01). (4.95) The behaviour of g(k, t) has two limiting regimes. The regime kR, «1. In this regime, only the overall translational motion of the polymer can be seen because g(k, t) is written as 1 . alk, =, D Cexplik - (Rolt) - Rc) +ik- (R, (0) — Ra() — ik - (Rn(0) — Ra(0))))- The underlined terms may be put at zero since they are of the order of kR,, which is much less than unity. Thus a(k, t) = N(exp[ik - (Ro(t) — Ro(0))]) - (4.96) If is large, the distribution of Rg(t) — Rg(0) becomes Gaussiant with the variance 2Dgt, hence Pr = ik - — 3/2 — ok, )=N f dr exp(ik «r)(4nDgt) exp( iba) =N exp(—Dgk’t). (4.97) Thus the decay of g(k, t) for a long time region is written as g(k, t) >1. In this regime, we can see the internal motion of the chain. Calculation of g(k, t) is involved even if the linearization approximation is used,“ (see Appendix 4.III), but the characteristic features of g(k, t) can be obtained from the scaling argument. By the same line of argument as in Section 4.3, we can show thatDYNAMIC LIGHT SCATTERING 107 g(k, t) is written as g(k, t) = NF(KR,, tDg/R?). (4.104) If kR, >>1, g(k, t) should be independent of N since the local motion is independent of the total chain length. Since R, and Dg depend on N as R, «NY and Dg «N~ "(vp = 1 for the Rouse model and v for the Zimm model), g(k, t) must have the following functional form, (Kk, t) = N(KR,)~ YF (tDcRz7(kR,)") (4.105) where ra24i?, (4.106) Thus if g(k, t)/g(k, 0) is plotted against tT, where Ty = DgRz7(kR,y (4.107) one curve is obtained for all values of k. The dynamical structure factor (4.105) is characterized by the decay rate I’, which is given by T, = Dgk*R? for the Rouse model (vp = 1, v = 1/2) = Dgk*R, for the Zimm model (vp = v). (4.108) Note that in the Zimm model the relation T,«k? holds both in © solvents and in good solvents. In the region KR, >> 1, the decay of g(k, t) is not a single exponential (see Appendix 4.III). The decay curve, however, is conveniently charac- terized by the initial decay rate I. To evaluate eqn (4.100) for KR, > 1, it is convenient to rewrite H,, by use of the Fourier transform of the Oseen tensor (see eqn (3.III.5)) Hons J an >1, the integral is dominated from the large q region. Hence using the asymptotic behaviour of g(k) «k~” (see eqn (2.132)), we have keT [ k “eee +e) atk ] (0) — Ket rm 4n7n, Jaa (3) tog |S il- 1 v(3v —1) ksT |, ~ dav — Dav a "(55 ) nm & (4.112) which becomes kT TP = 0.0625 — ke? for v=0.5 (4.113) = 0.07e 22 3 for v=0.6. (4.114) Note that eqns (4.113) and (4.114) are rigorous and include no adjustable parameters. Experimentally the k*® dependence of I? is well confirmed.?-?*© However, the experimental value is about 25% smaller than the theoretical value.*° The reason for this discrepancy is thought to have the same origin as that for Dg.*° The characteristic behaviour of the dynamical structure factor has also been confirmed by neutron scattering.“”-? 4.5 Viscoelasticity 4.5.1 Introduction The dynamics of polymers in solution can be studied by measuring their viscoelastic properties. Shear flows, for which the velocity components are given by u,(r, t) = K(t)r,, uy =u, =0, (4.115) are commonly used for studying these properties. If the shear rate x(f) is small enough, the shear stress depends linearly on x(t) and can be written as y(t) = / dr'G(t—1')k(t’) (4.116) where G(t) is called the shear relaxation modulus. For dilute solutions, in which the effect of the polymer is small, it is convenient to write eqn (4.116) as Oey(t) = ngx(t) + | dr GOU—1)K("’). (4.117)VISCOELASTICITY 109 The first term represents the property of the pure solvent and the second term represents the effect of the polymers. Two special cases are important. (i) Steady shear flow: k(t) = K = constant. (4.118) In this case, the shear stress is constant, which defines the steady state viscosity 1 = sy: (4.119) From eqns (4.117) and (4.118), it follows that n=net | diG(2), (4.120) 0 The increase in the viscosity due to the presence of polymers is usually expressed by the intrinsic viscosity, or viscosity number, defined by [n]= tim a (4.121) where p is the weight of the polymer in the unit volume of solution. Using c (the number of segments per volume), N (the number of segments per polymer), M (the molecular weight) and N, (the Avogadro number), p is written as Pray. (4.122) {ii) Oscillatory flow: K(t) = Ky CoS(wt) = Ky Re(e”) (4.123) (where Re stands for the real part.) The response for this flow defines the complex modulus G*(w): Ony(t) = Ky Re( ate) e). (4.124) Since eqn (4.117) gives ' Ouy(t) = Ko Re(eln, + | dG (e— 1") = ko Re{e™(n, + [orcmeye)] . (4.125)110 DYNAMICS OF FLEXIBLE POLYMERS G*(q@) is written as G*(w) =ion, + iw | dte"G(t) ° =G"(w) +iG"(a). (4.126) The real part G’(@) and the imaginary part G"(@) are called the storage modulus and the loss modulus. Experimental results are often expressed by the dimensionless intrinsic moduli defined by: [6'(w)]n= lim = ae FO (@) = lim = =e a too sin(ot)G°%¢) (4.1272) [G"(@)]e = lim —— Ht (Cw) - on.) = ine — |i cos(ct)G(t) (4.127b) where R = Nykez is the gas constant. 4.5.2 Microscopic expression for the stress tensor Let us now study the viscoelastic properties using molecular models. As was discussed in Chapter 3, the macroscopic stress of the polymer solutions is written as (see eqn (3.133)) Gap = Ne(Kaplt) + Kpa(t)) + 0%) + Pdap (4.128) where og= -< 3 (EucRap)- (4.129) Here the factor c/N accounts for the number of polymers in the unit volume. Since F, is written as n Tin¥+U), (4.130) eqn (4.129) is rewritten as a B= 5D [ARV sR hal In W + UR = Hd | [atRdtoT Re Rue + (Se Re) | (4.131) The underlined term gives the isotropic stress kg7T6,g by integration by Parts, and can be dropped in the incompressible fluid (see Section 3.7.2).VISCOELASTICITY 111 Thus c aU p=. 5 (ae Ras) (4.132) Under © conditions, U is given by eqn (4.4), hence ke T ofp SETS (— (ye + Runs 2R oR) af 3kpT “yo 2% (Bre —Ry)a(Rn+i ~Rn)p) (4.133) or in the continuous limit 3k, TT ORpa OR, ai ol ee) BT on an) (4.134) In a good solvent, one has to add the excluded volume potential eqn (4.65) to (4.132). However, the stress arising from this potential can be neglected because (5 Stine) Bear 3 (gh sm) ORraw bot S (se 6(R, — Ry) (Rap — Ros) uv 4 tar S (ge — a(R, - omy) . (4.138) —Rn)(Rop — Rng) The first term is zero, and the second term can be omitted because it is isotropic. Therefore eqn (4.134) holds even for the chain with the excluded volume effect. (This of course does not mean that the excluded volume interaction plays no part in the viscoelastic properties. The excluded volume does affect the viscoelastic properties through the distribution function Y over which the average in eqn (4.134) is taken.) Equation (4.134) can be rewritten using normal coordinates (4.18) as oe € kal ea) ( Alt) | sl? sin( Pom) sin( =") a OS pa (t)) - (4.136)112 DYNAMICS OF FLEXIBLE POLYMERS In © conditions, this can be written using eqn (4.21) c OB = 7) Ds ko Koa Xpp)- (4.137) In a good solvent, eqn (4.137) is not rigorous since k, is now given by eqn (4.72). However, use of eqn (4.137) may be justified in the linearization approximation, which is to assume that the potential U is given by = U= deal Day 7 aa i 2 kyX?, (4.138) Equation (4.137) can be derived from this potential by the principle of virtual work described in Section 3.7.5. Thus we shall use eqn (4.137) for both © and good solvents. 4.5.3 Calculation of the intrinsic viscosity We shall now calculate the stress using the linearization approximation. As was shown in Section 3.7.3, under the velocity field u(r, t) = K(t)-7, each bead has an additional velocity «(r)-R, (see eqn (3.120)). This gives the velocity «(r)-X, in normal coordinates. Thus the Langevin equation for X, now becomes k, 2x,0= ~ EO +E HO +KO-¥0. (4.139) To calculate (X,.Xpp) it is convenient to rewrite this into an equivalent Smoluchowski equation.+ win a /, aw a Ee (eer ax, + w)— Dag KO-wY. (140 If we multiply both sides of eqn (4.140) by X,,X,s and integrate over all the normal coordinates, we get, after integration by parts a 17a a a 9p (AeaXop) = (5 z lax : kT x (XpaXpp) — keXq* ox, Xre%e)| a +S KO ae (XeXop)) + Equation (4.141) can also be derived from the Langevin equation if we use the relation, (Xpofe ()) = keTSap which follows from eqn (4.22), and the short time solution of eqn (4.139), X,(Q=X,(¢- a) + ar( — 22x40 At) +Kx(t- At) -X,(¢- an) + futVISCOELASTICITY 113 1 = 5 PkaT bap — Bol XpaXos)] + Kan Xow os) + Kou (XpuXpe) (4. 141) This equation can be solved for arbitrary k,g(t) (see Section 7.6.3), but here we consider a special case of the shear flow given by eqn (4.115). To calculate the shear stress 9,,(r), we have to solve 5 (Kok) = 278 (MoaXy) +R) (4142) For small x, (X;,) may be replaced by the equilibrium value kT /k,. Hence eqn (4.142) becomes kyT a k, = =—2—2 3p AoXoy) = —2 (Xo X,y) + KS a? (4.143) In the steady state, eqn (4.143) gives (XpxXpy) = Fi 5fakeTK. (4.144) The shear stress is calculated from eqns (4.137) and (4.144) as =-£ 5 2 ne kg XpeXoy) = “we kaTk (4.145) Thus the intrinsic viscosity is given by _ keaT c & =a 4.146 Inl= pn,k "2pm, N~ k, (4.146) or, by eqn (4.122), NakoT S by 14 (= Win, 2k, (4.147) The sum is evaluated for various models: (i) The Rouse model: eqns (4.20) and (4.21) give Na NES 1 _ Ng N° n?_ Ny NOE (= ain, 6x? Zp ay 6x? 6 Mn, 36° (4148) (ii) The Zimm model for © solvent: eqns (4.58) and (4. 59) give Na VN b In =s4 ioe =P" sa tho, 425(VN by. (4.149)114 DYNAMICS OF FLEXIBLE POLYMERS (iii) The Zimm model for good solvent: eqns (4.71) and (4.79) give [n] a) p= A ND? (4.150) If we write the molecular weight dependence of [7] as [n]= Ms (4.151) we have 1 Rouse model (© solvent), Vq = 4 0.5 Zimm model (© solvent), (4.152) 3v—1=0.8 Zimm model (good solvent). Experimental results show that in the © solvent v, is 0.5, in agreement with the Zimm model. The agreement is actually more quantitative. Eqn (4.149) is written in the form [nl=$e(voR,) (4.153) where the constant ®,, called the Flory—Fox parameter, is ®, =0.425N, = 2.56 x 107. (4.154) The experimental value of ®, is about 2.5 x 10%. Perhaps this good agreement is fortuitous since a more accurate analysis’? of the pre- averaged equation (4.50) gives ®,=2.84x 10”. Various theoretical results” give ®, values ranging from 2.2 x 10” to 2.87 x 10°. In any case this level of agreement again indicates the success of the Zimm model in the dilute solution theory. In good solvent, the experimental value of v,, is slightly smaller than 0.8. This is perhaps because, as in the case of Dg, the molecular weight is not sufficiently high for the asymptotic behaviour to be observed. In such regions the Flory-Fox parameter decreases with increasing molecular weight and increasing excluded volume parameter. Detailed calculations for ®, are described in the book by Yamakawa.** 4.5.4 Intrinsic moduli Next we consider the case that the shear rate is not constant. Equation (4.143) is solved for general x(t), ake? at | de! exp(—(t-#')/2,)x("") (4.155) where Tp = 6,/2kp = tp" (4.156)VISCOELASTICITY us 2 Rouse model, = 9 3/2 Zimm model (© condition), (4.157) 3v_Zimm model (good solvent). (Note that t, is different from the rotational relaxation time 1, by a factor 1/2.) From eqns (4.137) and (4.155), G®(t) is calculated as G0) =KkeT > exp(-t/t,) = eRT Dexp(-t/t,). (4.158) P P Hence [G'(w)la= j ate sin(cn) 5 exp(—t,) = 3S ee e) Fy 4159) [6° oyle | ae cox(or) expt!) = Sap Miso 0 Pp The expressions are simplified in two cases. (i) wt, <1: In this case, [G’(@)]z and [G"(@)] are approximated as [G’'()]x = (ot)? » pe, (4.161) x [G"(o)]x = or, See (4.162) Hence [G’(@)]z and [G"(@)]z are proportional to w? and w respectively. (ii) @t,>>1: In this case the sum over p can be replaced by an integral, so that (wxjp* xe [6'(oyle= { op a= (oe 1 fact x 0 7 2p sin(x/2p) = (wt) (4.163) [G'@)le= (or,)""* if ar =(@t,) _— : Thus 2p cos(7/2u) (a) 4 =2 (Rouse model) (4.164) [G'(@)le = [G"(@) le =35 (or) =1.11(w1,). (4.165)116 DYNAMICS OF FLEXIBLE POLYMERS (b) » =3/2 (Zimm model in © solvent) [G’(@Je=1.21(@01)?: — [G"(@) x =2.09(w41)*. (4.166) (c) # =9/5 (Zimm model in good solvent) [G'(@) Jn = 1.1411)": [G"(@) Jn = 1.38(7,)™. (4.167) For @ conditions, eqn (4.166) is well confirmed.**** For good solvent, experimental data have been interpreted in terms of the so called ‘draining parameter’®°, but the data might be interpreted by the above theory. Indeed eqns (4.166) and (4.167) indicate that the asymptotic slope of [G’(@)]z and [G"(w)]z decreases as the solubility of the polymer increases, which is consistent with the experimental results. 4.6 Variational bounds for the transport coefficients 4.6.1 Introduction When the hydrodynamic interaction is introduced, rigorous analysis of the Smoluchowski equation or the Langevin equation becomes impos- sible. In the previous sections, we used the preaveraging approximation to avoid this difficulty. Another way of handling the hydrodynamic interaction is to use a variational principle.“*> This method gives rigorous bounds for the transport coefficients such as [j] and Dg. The method has an advantage that the resulting formulae are easily applied to general polymers such as stiff polymers, branched polymers, or colloidal suspensions. A cautionary remark has to be made. The variational principle in this section is based on the positive definiteness of the mobility matrix, i.e., for any vector {F,}: D Fi Ham Fn 20 (4.168) This condition is guaranteed for the correct mobility matrix. However, the mobility matrix given by eqn (4.40) is an approximate one, and does not satisfy the inequality (4.168) in a certain configuration in which the beads are too close to each other. An improved formula which guarantees the inequality is proposed by Rotne and Prager.*” However, this correction is irrelevant for the asymptotic behaviour of N >>1, which is determined by the hydrodynamic interaction between beads far apart from each other. Thus we shall use eqn (4.40) for H,,,- 4.6.2. Bounds for the intrinsic viscosity Lower bound. According to the formalism described in Section 3.7, the intrinsic viscosity [7] can be calculated in the following way (the shearVARIATIONAL BOUNDS 117 flow (eqn 4.115) being considered): (i) Firstly, the steady-state distribution function Y is obtained by solving the Smoluchowski equation in shear flow a ow aU a Diggs Hm [ura SR ¥| —Dapz Kho ¥=0. (4.169) (ii) Secondly, [n] is calculated from eqns (4.121) and (4.129) = op = I= OF = ~ ade (Rohan) pn. Na 2 wate w[o] (4.177)118 DYNAMICS OF FLEXIBLE POLYMERS where aD | ab W[®]=ks r(- kal DSP Ho Set aR Re 0) (4.178) Equation (4.177) gives a lower bound for the intrina viscosity. To prove eqn (4.177), we note that for the true solution ®* which satisfies eqn (4.173) the following equality holds for arbitrary ®: (tot D Sp Hoe Se) = (3 ae Ray) (4.179) This can be proved by integration by parts. Using eqn (4.179) for the special case of & = &*, we can show that ao* ; W[o*] = kot (E ree Ray) =[n]n?MINy. (4.180) mm ORmx 4 Furthermore, from eqns (4.178), (4.180), and (4.168), it follows that od* ad W[o*]-W[®]=kp (kT 3 5 : (Rae) aD* ad “Hon *(: oR. ~5x)) 2 (4.181) Equation (4.177) follows from eqns (4.180) and (4.181). A simple choice of the trial function is O=1+E>D RR, (4.182) where R,=R,-Re (4.183) and & is a variational parameter to be determined. Substituting eqn (4.182) into eqn (4.178), we have 2 Ww[%] = ~ 7 MoT YE N'Rn + 4xksTENR;, (4.184) where , Rus = Hi D ((RoRiyHrnys)oq + (RinsRiHrny Dog) (4.185) mn and we have used the relations (RiRayHmnys eq = (RimyRncHnnzy Yea (RineRrxHmnyy Yea = (RinyRnyHmnx ea: (4.186) The best estimate for [7] is obtained by maximizing eqn (4.184) withVARIATIONAL BOUNDS 119 respect to &, which gives Na Ri (> ser (4.187) The right-hand side includes only the average over the equilibrium distribution function. It can be shown that eqn (4.187) gives a correct viscosity for two limiting cases, i.e., spherical particles and rigid rodlike particles. For flexible polymers in the © condition, however, the bound is rather weak: in terms of the Flory-Fox parameter, one obtains (after some tedious calculation)*® ® > 2ig(627°)'7N, = 1.90 x 107. (4.188) Upper bound. A formula for the upper bounds for [n] was given by Fixman*®® In <4 5 ((V,— Ry) (Ham “(Vin —KRrn) eq, (4.189) MK im where V, is a function of {R,,} satisfying 3 —. =0. 4.190 2 aR, (Va Peq) = 0 ¢ ) Use of this variational principle is not easy since it requires the evaluation of (H™)nm. Nevertheless, the calculation can be done numerically," and the results indicate that the error of the preaveraging approximation is less than 30% for [7]. 4.6.3 Bounds for the diffusion constant Upper bound. Next we consider the translational diffusion constant. To calculate Dg, we consider that a weak constant field 1 Uext = —FRez = -ym FRyz (4.191) is applied to the solution in uniform concentration. The field will cause a uniform motion of the centres of mass of each polymer with a constant velocity (V) which is proportional to F. According to the fluctuation dissipation theorem, Dg is obtained from ( V) by (see eqn (3.67)) — VezdkaT -O To calculate (Vg,) we have to solve the Smoluchowski equation for De (4.192)120 DYNAMICS OF FLEXIBLE POLYMERS the steady state: 7k F ) Dag “Hy Wos( ko ar, N/~° where F = Fe, (e, being a unit vector in the direction of the z axis). Since we are considering the time-scale in which polymer concentration is homogeneous, only the translational invariant solution is needed; i.e., ®({R,}) = O({R, +4}). (4.194) (Mathematically this condition is needed again to ensure that the surface integral vanishes.) Now, since the velocity of the n-th bead is given as (4.193) Ve= 3 Ham Go (ke In W+ U4 Use) =-D An («or Be ‘), (4.195) (Vc) is calculated from 1 1 ab oF (Vo) = (OV. )e4 79D (Hee (- bel SH), (4.196) where terms of higher order in F have been neglected. The term 0®/OR,, denotes the driving force due to the deformation of the molecule. If we neglect this term, we have F (Vez) =i (Hmzz dea = ma > : RR R a). - 4.197) This gives the Kirkwood formula De = nD = S RR mR). (4.198) To see the effect of the deformation term 3@/3R,,, we have to solve eqn (4.193). The effect can be calculated by a variational principle. It can be shown that for any ® that satisfies eqn (4.194) keT De =< Wel®] (4.199) where Wl] = 3B ((ka e-~ FZ) - Hon (ko SP ~ 5), (4200) Equation (4.199) indicates that the Kirkwood formula (4.198) cor- responds to the choice of © = 1 and is actually an upper bound for Dg.BIREFRINGENCE 121 Lower bound. A lower bound for the diffusion constant is given in a similar form to eqn (4.189).*° Let V, be the function which satisfies eqn (4.190), then De = DP ~"EF(E (Wa VP): (Han (Vn VO) (4.201) where “m “ VO =D) Him + €:- (4.202) Given the upper and lower bounds, one can estimate the error of the preaveraging approximation. Fixman® showed that the Kirkwood for- mula gives quite accurate estimation for D;: for flexible polymers in © condition, the error is of the order of a few per cent. 4.7 Birefringence 4.7.1 Birefringence of polymer solutions Polymer solutions are isotropic at equilibrium. If there is a velocity gradient, the statistical distribution of the polymer is deformed from the isotropic state, and the optical property of the solution becomes anisotropic. This phenomena is called flow birefringence (or the Maxwell effect).°'? Other external fields such as electric or magnetic fields also cause birefringence, which is called electric birefringence (or Kerr effect) and magnetic birefringence (Cotton—-Mouton effect), respectively. The birefringence of a material is expressed by the anisotropic dielectric tensor @,, at optical frequency, or the refractive index tensor fgg Which is the square root of é,, i.e., Reutap = bap: (4.203) It is convenient to decompose é, and fgg into an isotropic tensor and a purely anisotropic tensor whose trace is zero: Bap = €5ap + Exp, Figg = NO ap + Nop (4.204) where Exo =O and Nyg =0. (4.205) In polymeric systems, the anisotropic part €4, is usually much smaller than the isotropic part ed, (\€ag|/€ < 107°). When this is so, eqn (4.203) gives _1 Nap = 3p fab (4.206) Let us now consider the flow birefringence. If the velocity gradient122 DYNAMICS OF FLEXIBLE POLYMERS tensor Kye is small, the birefringence is linear to Kgg and written ast t naplt) = | Ault CV (Kap t+ Kpalt)). 4.207) In a steady state Nop = Au (Kap + Kpa)- (4.208) The constant Ay is called the Maxwell constant. In dilute polymer solutions, experimental results are often expressed by the ‘intrinsic Maxwell constant’ defined by [7] = tim Be. (4.209) s 4.7.2. Molecular expression for birefringence The birefringence of polymer solutions has two origins. Firstly, since the polymer segments have anisotropic polarizability, the orientation of the bond vector of the main chain causes birefringence, called the intrinsic birefringence. Secondly, since the dielectric constant of the polymer coil region is different from that of the solvent, the anisotropy of the shape of the polymer coil creates an anisotropic internal field, which also contributes to the birefringence. This is called the form birefringence. Intrinsic birefringence. The intrinsic birefringence of polymeric materials was first calculated by Kuhn.® He considered a polymer segment whose. end-to-end vector is fixed at r, and calculated its average polarizability g(r) assuming that (i) The segment is made up of n, bonds connected by universal joints. Each bond has anisotropic polarizability, a, along itself and o, perpendicular to itself (see Fig. 4.2a). (ii) The total polarizability is the sum of the polarizability of individual bonds. Kuhn’s result is written as“ 2 Ns Gap(r) =F (ay + 201) dap Ir| +3 rary ~ Hap?) 14H (34 ly +...]- 4210) If the polymer segment is not extremely extended, |r| is of the order of + In electric (or magnetic) birefringence, the birefringence is not linear to the field, so that the response function is not written in the simple form (see Chapter 8).BIREFRINGENCE 123 c a fe — : , 8 A A (@) Fig. 4.2, (a) Kuhn’s model, (b) Additivity assumption: if alt) and ap (te) | are the polarizability of the part AB and BC respectively, the total polarizability is Gap (t) = (aap(ti) + Wap (ts) ntner Vn,b, so that the underlined terms can be neglected. Thus the anisotropic part of the polarizability tensor Gap (t) = Gap lt) — 3G (1) Sap (4.211) is given by @ap(r) = A(rats — 45 apt”) (4.212) where 3(ay- a.) Aa5 a a . (4.213) An important result of this model is that in the Gaussian region A is independent of r and proportional to 1/n,. Actually this conclusion holds for more general models. In Appendix 4.IV, it is shown that the polarizability tensor is written in the form of eqn (4.212) with a constant A (proportional to n>") provided the following conditions are satisfied. (i) The statistics of the chain is Gaussian.+ (ii) The polarizability a.,(r) is additive; i.e., when the segment is divided into two parts, 1 and 2, the total polarizability is given by (see Fig. 4.2b) Gap (r) = (ap (Ki) + ap (%) n+ner (4.214) where @,(r1) and a,¢(r,) are the polarizability of each part and the average is taken for the equilibrium state with r, + 7, being fixed at r. In The effect of the excluded volume interaction on A is not known, but if the size of the segment is taken to be small, the effect will be only to change the constant A. +The additivity is not satisfied if the dipole-dij lipole interaction within the segment is taken into account. It has been shown by Copic™ that this effect is included in the form birefringence.124 DYNAMICS OF FLEXIBLE POLYMERS this general case, A is written as AY (eq where (r’).q is the equilibrium average of r° in free state and Ay is a constant independent of the size of the segment: Ay depends on the detailed chemical structure and can be positive or negative. A micro- scopic derivation of Ay is described in refs 64 and 65, and the experimental values are summarized in ref. 61. Now, eqn (4.212) can be used for each Rouse segment bounded by two beads, say n and n — 1, then rand (r*),, can be replaced by R, — R,-1= OR,,/dn and b?, respectively. The polarizability of this Rouse segment is then written as A= (4.215) _ AY (ORna ORap_ 2m) Snap = 52 ( an On 4a an) )* (4.216) To relate the polarizability of individual segments to macroscopic birefringence, Kuhn® used the Clausius-Mossotti equation, according to which the (isotropic) refractive index n is related to the (isotropic) polarizability of individual molecules as: n?-1 40 aD 32 a, (4.217) where the summation is taken for all the molecules in unit volume. According to eqn (4.217), if a; undergoes a slight change da;, n changes to én= = (n? +2)? D da; (4.218) i Using this relation for each component of the anisotropic polarizability tensor @,2g, which is much smaller than the isotropic part, one gets i) 2" N= 5, +2 Danas (4.219) all segments in unit volume where the superscript (i) stands for the contribution of the intrinsic birefringence. Equations (4.216) and (4.219) give 2 *)]) 220) x ; 1 [ ORpw OR, OR, a= = | ne Shs _ nop (|e zl on an 4.o( an where =2% 72 2 PNa K= on (n? + 2) Ay. (4.221) The factor pN,/M accounts for the number of polymers in a unit volume.. | y @) () (©) 2 Fig. 4.3. (a) Polymer, (b) spheroidal model, and (c) fuzzy spheroidal model. Form birefringence. The form birefringence arises from the difference in the isotropic part of the polarizability between polymer segments and solvent molecules. A simple way to understand this is to regard the polymer coil as an ellipsoid which has a different dielectric constant é+ de from that of the outside (see Fig. 4.3). For simplicity let us consider the coil as a prolate spheroid, and denote the direction of the jong axis by a unit vector u. The polarizability of the spheroid is larger when u is parallel to the electric field than when it is perpendicular to the field, so that if the distribution of u is not isotropic, the dielectric constant is not isotropic. The average dielectric constant of a material which includes a small number of such spheroids is calculated by electrostatics (see ref. 68, for example) as _ (Sup — tat), (tate) ] fap = £805 + ee +5e(1—N,2*e+den,| 722) where @ is the volume fraction of the spheroid and Nj is a numerical factor called the ‘demagnetization factor’ which takes values between 0 and 1/3 depending on the aspect ratio of the spheroid (Nz = 1/3 for a sphere and 0 for a rod). For smali 6¢, the anisotropic part €9} = é., — (81n)5ap/3 is calculated from eqn (4.222) to be ei= ¢ OD 404 — Nay ua) — 80g) (4.223) Note that the underlined coefficient is always positive independent of the sign of de. Since the prolate orients toward the flow direction, the form birefringence always gives a positive contribution to the Maxwell constant.126 DYNAMICS OF FLEXIBLE POLYMERS Though the spheroid model is convenient for displaying the charac- teristic aspects of the form birefringence, it is not suitable for quantitative calculation. A more general approacht is to consider that the dielectric constant varies with position around the mean value & (see Fig. 4.3c) such that e(r) =2 + de(r). (4.224) If e(r) fluctuates, the dipole moment density P = (e — 1)E/4z fluctuates by 5P,(r) = 6e(r)E.(r)/4n. (4.225) This causes a fluctuation of the electric field: E,(r) = E, + 6E,(r), (4.226) 6E,{r) = [or'Gastr —1r')6Ps(r’) (4.227) where Gap(r) = a (FFs — 48a) (4.228) denotes the field created by a dipole moment. Thus the average electric displacement D is written as Dz = (€(r)Ea(r)) = ((E + 5€)(Es + 5Ea(r))) = 8, + (6e(r)6E,(r)) = 66, + x | dr! Gup(t — 7°)(5e(#)5e(r')Ep(r")). (4.229) Since the fluctuation of 5e(r) is very small, E,(r’) on the right-hand side can be replaced by £5. Hence D.= [exp + x | dr’ Gag (r - r')(de(r)e("')) |B. (4.230) The second term becomes anisotropic if the correlation (5e(r)de(r')) is not isotropic. The form birefringence is thus given by (note Gua =0) 1 ef = an [erGap(r)(de(7)6e(0)). (4.231) It can be shown that eqn (4.231) reduces to eqn (4.223) for dilute spheroid. + The formulation presented here was adapted from the work of Onuki and Kawasaki® for the binary solution near the critical point. Here the static approximation is used for the electromagnetic field. This treatment is similar to that given in refs 66 and 70.BIREFRINGENCE 127 Equation (4.231) is rewritten using eqn (4.206) and e(r)=n(r), in terms of the refractive index: nQ =x J drGup(r)(6n(r)dn(0)). (4.232) In the polymer solutions, dn(r) can be assumed, as in the theory of light scattering,” to be proportional to the local segment density dc(r), én(r) = ($)oce (4.233) Hence, eqn (4.232) is rewritten as nQat (= ny’ | drGup(r)Se(r)6c(0)) (4.234) 3e op : . If we use the Fourier transform of G(r), Gap(r) = a lows (kakp - 45ap)exp(ik +r), (4.235) and express (ectry8et) by the structure factor g(k), dk : (8c(r)8c(0)) =e | Bayt exPC-ik-N(), (4.236) we finally have Bat (3: 2) oes y (fe aks — 45p)8(k). (4.237) The form birefringence is thus directly related to the anisotropy in the structure factor. 4.7.3 Flow birefringence Having obtained the basic formula, we can now study the flow birefrin- gence of dilute polymer solutions. The contribution from the intrinsic birefringence is easily calculated. Comparing eqn (4.220) with eqn (4.134), we note that the intrinsic birefringence is proportional to the stress: n®, = Co} (4.238) where __2n (Wn? +2) c= 2keT on (4.239) is a constant called the stress optical coefficient, which depends only on128 DYNAMICS OF FLEXIBLE POLYMERS the local structure of the polymer. Hence the contribution to [n} from the intrinsic birefringence is easily calculated from eqn (4.146) as [n°] =2C{n}. (4.240) However, calculation of the form birefringence is tedious (see refs 66 and 70, and also Section 5.5). Here we will give a simple approximate treatment. Under weak shear flow eqn (4.115), the deformation of the structure factor g(k) is proportional to the dimensionless shear rate x/T,, and will be written as 8(k) = Beq(k) + 5g(k) with dg(k)~(K/Tg)geq(k) (4.241) where g.(k) is the structure factor at equilibrium and I, is the characteristic decay rate of the dynamical structure factor (see Section 4.4). From eqns (4.237) and (4.241), the form birefringence is estimated as wl (OMY [ak Beak) ng «<(%) lem? lh (4.242) Since I, and g.,(k) are given by 1 = z F(KR,), — Beq(k) = NF(KR,), (4.243) eqn (4.242) is evaluated as nQa~«e < (3) tes (zy. neXt ac} R3 Oc} kgT a? b (Sn): Kn, 5p) MRT (4.244) This result is also derived from eqn (4.223) if we use (be 5 (SE) > uty) =a, J=N,=1. (4.245) ee The contribution to [n] from the form birefringence is 2 1/on\? M 0} =n = [n®] ox” ky “(2 o) RE RT (4.246) where k; is a certain numerical constant. This formula was first derived by Janeschitz-Kriegl,©*7! who estimated k, from experimental results to be 0.34. Equations (4.240) and (4.246) indicate that [n®] and [n] increaseTHE VERDIER-STOCKMAYER MODEL 129 with molecular weight M as [n®] «M1, [n®] x M. (4.247) Hence for large molecular weight, the form birefringence dominates. From eqns (4.240) and (4.246), it follows that [a] _ fn) + [2 (8n/apy? M- [a] [n] nRT [n] This relationship has been confirmed experimentally.” The form birefringence [n] is always positive while the intrinsic birefringence [n] can be positive or negative depending on the sign of Ay. Since the observed birefringence is the sum of the two, the analysis of the flow birefringence is not easy in dilute solutions. Indeed, in the case of Ay<0O, the measured birefringence shows anomalous behaviour.*” This complication can be avoided if the effect of the form birefringence is eliminated by choosing the solvent such that dn/dp = 0. The experimental results on flow birefringence of dilute polymer solu- tions are reviewed in refs 61 and 62. =2C +h (4.248) Appendix 4.1 The Verdier-Stockmayer model” In this model the polymer is made up of N beads connected by N—1 bonds, each having constant length b (see Fig. 4.1b). In a small time interval At, each bead makes the following jump with probability w Az. (i) For the internal beads (i.e., beads 2,3,...,N—1) R,> (Ravi + Ra-1— Rn). (4.11) (ii) For the end beads (n = 1 or N) R\ > R,- vp (4.1.2) Ry Ry-1 + ¥y (4.1.3) where v) and vy are randomly chosen vectors of length b. To analyse this model, it is convenient to look at the bond vector v, = Rai —R, rather than R,,. The transition rule for v, is U,41 This happens with probability w Az, v,—> 4 U,-1 This happens with probability wAz, (4.1.4) v, This happens with probability 1 — 2wAz. For n =1 or N—1, v or vy,, mean the random vector. Now let us calculate the bond correlation function Cam(t) = (v(t) + Un (O)). (4.1.5)130 DYNAMICS OF FLEXIBLE POLYMERS This is obtained as follows. If the jump v,—v,4, occurs in the time interval Ar, then C,n(t + At) becomes equal to C,41,n(t). This happens with probability wAt. The jump v,—v,_, occurs also with probability wAt. The probability that no jump occurs is 1 — 2wAr; hence Cam(t + At) = Casim(Q)wAt + C,—1,m(t)wAt + Cum(t)(1—2wAt) (4.1.6) or for At—>0 S Com(l)= W(Cosisnl)+ Carin) 2Con(l)) LT) Com(t) and Cym(t) are zero since vo and vy are chosen without any correlation to the other bond vectors, i.e., Com(t)=0, — Cxm(t) = 0. (4.18) The initial condition is Gum(0) = B75 jn (4.19) By using the coordinate defined by v2 H,()= > (2) sin( 2 om) Can(?) (4.1.10) the simultaneous difference equation (4.1.7) is solved giving Cum(t) = 0? S sin( Pt Pe sin Pa x 7 exp(— -a,t) (41.11) where , Ap = 4w sin?(2*) . (4.112) On the other hand, for the Rouse model, the quantity corresponding to Cim{(t) can be obtained by identifying v, with IR, /dn: 8) = (0. al) aa (4.1.13) From eqns (4.23) and (4.27) C&W) =4 > a (2 pat sin(2 a) sin PO Pa )exp(— t/t) p= =2 vo z sin( 2 Pe sin( PO pam )exp( — ed pt). (4.1.14)DERIVATION OF THE NORMAL MODES 131 This agrees with eqn (4.1.11) if one notes that for small p ww A= ye p. Thus w can be identified with 3k,7T/&b?. (4.1.15) Appendix 4.II Derivation of the normal modes To find the normal coordinates, we consider the linear transformation of R,(0) N X,()= { dng,,R,(t) (4.11.1) 0 and choose ¢,, so that the equation of motion for X, has the following form, & 3x, =-kX, +h. (4.11.2) From eqns (4.9) and (4.11.1) N N ox, 3 & | OR, ) Se —R, = : 11.3 5 Se=t J Snr 5 Ry = 7 [ant.(t Seth): (413) Using integration by parts, we can rewrite this as rhs = 9h aR,” [ate RJ | an[e Sten, + boa. (4.114) The first term vanishes due to eqn (4.11). Hence eqn (4.11.2) is rewritten as ~2 [x 2b e, [ + fan[eSten, + nd | N = | dn(—kpbonR,) +f. (4.115) For eqn (4.11.5) to hold, we must have a het sb (4.1.6)132 DYNAMICS OF FLEXIBLE POLYMERS with 2b pn _ = = on =0 at n=Oandn=N (4.11.7) and t N = z [art Su. (4.11.8) 0 Equations (4.11.6) and (4.1.7) are the eigenfunction equations for ¢,,, which are well known and have the solution $m = Hoos 222) (p=0,1,2,...) (4.11.9) and =e (PZ) i, =k (ER). (4.11.10) Now €, can be chosen arbitrarily, so we choose €, such that f, satisfies the same formula as the oscillator (see eqn (3.82)); i.e., (fox(O)fox(0)) = 2,ks T(t). (4.11.11) The left-hand side can be calculated from eqns (4.11.8) and (4.11.9) as {nf 0)) = FERS be anf am cs (PRP cos ET) funn (0)) . (4.11.12) — bp. (pan qan - ies J dn cos( 2 eos (Fr 2th. T5456(0) 2 146 = yay ONG 2k TB 8(0) (4.11.13) where eqn (4.12) has been used. Comparing eqns (4.11.13) and (4.11.11) we have eqn (4.20) and thereby eqn (4.21) from eqn (4.11.10). Appendix 4.11 Dynamic structure factor First we consider the Rouse model, for which g(k, t) can be calculated rigorously.“ According to the theorem in Appendix 2.1, a linear combination of the Gaussian random variables obeys the Gaussian distribution. Since R,,(¢)—R,(0) is a linear function of f,(t), which is Gaussian, the distribution of R,,(¢) — R,(0) is also Gaussian. Hence, eqnDYNAMIC STRUCTURE FACTOR 133 (2.1.20) gives (expfik - [Rm(t) — R.(0)]]) = . T, expl — 4k2((Rma(t) — Rne(0))?)] = exp| -# -% - b(t) (4.1111) where Pmt) = ((Rm(t) — R,(0))*). (4.11.2) Using eqn (4.27), Pmn(t) is written as matt) = {[ 1) ~ XO] +2 (cool P")ASCO — c08(2= om), “0)[ ). (4.1113) Since the correlation between different modes vanishes, eqn (4.III.3) can be rewritten as ma) = (0) XA OIF) +4 5, [cost PR) (H(07) + 00s"( 2") (x, (07) ~ 2.005( 2) cos( 2X") (x,(0) x,(0))]. (4.11.4) From eqn (4.23) and S lel) -eof)f-Zi-m. mts we have 2 «© Pnnlt) = 6Dagt + [n — m| 624 AE J cos(P2)cos( 2X) pap? N N xX (1—exp(—p7t/tr)). (4.1116) where Tt, is the Rouse relaxation time given by eqn (4.25). Thus alk, 0 =5 > exp| Dot 2 |n — m| k?b? 2Nb*K? S 1 Poo pam 2 |: 3a? 2 P* Cyr N Or a exp(—ip?/tp)] |. (4-11.7) The form of the g(k, t) is simplified in the two limiting cases: (i) Small angle regime: If k?Nb?<1, the second and third terms in eqn (4.I1I.7) may be neglected since their magnitude is less than k7Nb?.134 DYNAMICS OF FLEXIBLE POLYMERS Hence g(k, t) becomes (k, t)=N exp(—-k?Dgt) (4.1118) which agrees with eqn (4.97). (ii) Large angle regime: If k°Nb*>>1, we may limit consideration to the time region t< The integrand has a sharp peak at n=m, so the double integral is evaluated as N N oo dn {dm =N | d(n-m). (4.11.11) [a fan=n J The final form is written as 122 | gk, t) =< | du exp[—u —(Ty)7ACU(T)-))-— (4.00.12) | « x where kaT Tee kp? (4.11.13) and A(u) =2 | x 28H) (1—exp(-x2)). (4,111.14) 0POLARIZABILITY TENSOR OF A GAUSSIAN CHAIN 135 For I, >1, this expression is further simplified to a(k, #)= atk, 0) { du expl—u — Cu)!*A(0)] =a(k, Ojexp| - a ro] . (4.11115) For the Zimm model, the structure factor can be obtained in the same way if the linearization approximation“ is used. The result for the © condition is a(k, ) =8(k, 0) f du exp[—u — (Tt) *h(u(Tyt)™)] (4.11.16) o with _keT Te Sn, Re (4.11.17) h(u) =2 | 2) —exp(-27/V2)). (4.11.18) oO For [,t>>1, g(k, t) = g(k, O)exp(—1.35(T,£)”"). (4.11.19) Appendix 4.1V_ Polarizability tensor of a Gaussian chain Let aag(r, N) be the average polarizability of a polymer chain consisting of N units with its end-to-end vector being fixed at r. By symmetry, g(r, N) is written as aap(r, N)= AC, N)atp — 366") (4.1V.1) where A(r’, N) is a scalar. Equation (4.IV.1) can be expanded with respect to 77: Gap(t, N) = (ao(N) + a,(N)P? + ...)(ratp —45ag"). (4.1V.2) We shall show that ao(N) is proportional to 1/N, and a,(N) = a,(N) = ... =0 provided that (i) the statistics of the chain are Gaussian and that (ii) the polarizability is additive. The second condition is written as follows. Suppose that the chain is divided into two parts each consisting of N, and N, units, then eran ts Ni + Ns) = [ar drs (r., Ma) + dope, NJIO(N, ) (4.1V.3)136 DYNAMICS OF FLEXIBLE POLYMERS where ®(r,,™) is the equilibrium distribution of r, and 7, under the constraint that r, + 7 is fixed at 7. For the Gaussian chain, this is given by O(n, m) = O( +m — PVH, Ni) (H, No)/(r, N+ Nz) (4-IV.4) where n 2 3 3 Wr, N)= (=) exp( -s) . (4.1V.5) If we multiply eqn (4.IV.3) by ep (r, N, + Np) and integrate over r, we have Gap (ky Ni + Nz) P(k, Ny + No) = B(k, N2)&ap(k, Ni) Pk, Ni) + i(k, Ni)dop(k, N2P(k, N2) (4.1V.6) where . W(k, N) = exp(—Nb7k?/6) (4.1V.7) and & p(k, N) is a differential operator: = F a0, # Sap(k, N)= ~(a0- asst. Mae 3g ne az) (4.1V.8) Using eqn (4.IV.7), eqn (4.IV.6) is written as 2 2 exp(= RN, + Na) Gap (k N+ N)exp( - Fem + ™)) 2 2 = exp(= IN, ) ap ky Nvexo( ae eN,) 6 6 Ba). e, + exp( k N:) ap Naexp( — ze ) . (41V.9) For this to be true for arbitrary N, and N,, it must follow that 2 2 exp(= en) dap(k, Nyexp( - ew) =NByp(k) (4.1V.10) where B,(k) is a certain function of k. Substitution of eqn (4.IV.8) and a straightforward calculation gives 4 4 = N(kakp - 3baph?)| aol) + (we - 30N)a\(N) tee | = NByg(k). (4.1V.11) Comparison of the coefficients of the power of N indicates that this equality is satisfied only when aN) =co/N, a(N)=a,(N)=...=0. —(4.1V.12)REFERENCES 137 Choosing the constant cy as Ay/b, we have A ean 05 N) = pp (Fala — 48 op?) Ay = Pn (ratg — 45ap?”) (4.1V.13) which is eqn (4.215). References 1. Rouse, P. E., J. Chem. Phys. 21, 1272 (1953). 2. Verdier, P. H., and Stockmayer, W. H., J. Chem. Phys. 36, 227 (1962); and ona FPS pon ee Verdier, P. H., J. Chem. Phys. 52, 5512 (1970). . Orwoll, R. A., and Stockmayer, W. H., Adv. Chem. Phys. 15, 305 (1969). . Iwata, K., J. Chem. Phys. 54, 12 (1971). Edwards, S. F., and Goodyear, A. G., J. Phys. A5, 965, 1188 (1972). See general review, Stockmayer, W. H., Dynamics of chain molecules, in Molecular Fluids (eds R. Balian and G. Weill). Gordon & Breach, New York (1976). . Hilhorst, H. J., and Deutch, J. M., J. Chem. Phys. 63, 5153 (1975). . Kuhn, W., Kolloid Z. 68, 2 (1934). . Sole, K., and Stockmayer, W. H., J. Chem. Phys. 54, 2756 (1971); see also Doi, M., and Nakajima, H., Chem. Phys. 6, 124 (1974). . Zimm, B. H., J. Chem. Phys. 24, 269 (1956). . Zimm, B. H., Roe, G. M., Epstein, L. F., J. Chem. Phys. 24, 279 (1962). Hearst, H. E., J. Chem. Phys. 37, 2547 (1962). . Yamakawa, H., Modern Theory of Polymer Solutions, Chapter 6. Harper & Row, New York (1971). |. Edwards, S. F., and Freed, K. F., J. Chem. Phys. 61, 1189 (1974). |. Edwards, S. F., and Muthukumar, M., Macromolecules 17, 586 (1984). . Peterlin, A., J. Chem. Phys. 23, 2464 (1955). . Kurata, M., Yamakawa, H., and Teramoto, E., J. Chem. Phys. 28, 785 (1958); Chikahisa, Y., and Tanake, T., J. Polym. Sci. C 30, 105 (1970). . Matsushita, Y., Noda, I., Nagasawa, M., Lodge, T. P., Amis, E. J., and Han, C. C., Macromolecules 17, 1785 (1984). . Tanford, C., Physical Chemistry of Macromolecules. Wiley, New York 1961. . Oono, Y., and Kohmoto, M., J. Chem. Phys. 78, 520 (1983). }. de Gennes, P. G., Macromolecules 9, 587, 594 (1976). . de Gennes, P. G., Scaling Concepts in Polymer Physics, Chapter 6. Cornell University Press, Ithaca (1979). . Hohenberg, P. C., and Halperin, B. I., Rev. Mod. Phys. 49, 435 (1977). . Oono, Y., J. Chem. Phys. 79, 4629 (1983). |. Ceperley, D., Kalos, M. H., and Lebowitz, J. L., Phys. Rev. Lett. 41, 313 (1978); Macromolecules 14, 1472 (1981). . Berne, B. J., and Pecora, R., Dynamic Light Scattering. Wiley, New York (1976).138 BIR 41, & DYNAMICS OF FLEXIBLE POLYMERS . Akcasu, Z., and Gurol, H., J. Polym. Sci. Phys. ed. 14, 1 (1976). . Kirkwood, J. G., J. Polym. Sci. 12, 1 (1954). . King, T. A., Knox, A., Lee, W. I., and McAdam, J. D. G., Polymer 14, 151 (1973). . Adam, M., and Delsanti, M., J. Physique 37, 1045 (1976); Macromolecules 10, 1229 (1977). . Nose, T., and Chu, B., Macromolecules 12, 590 (1979). . Han, C. C., and Akcasu, A. Z., Macromolecules 14, 1080 (1981). . Schmidt, M., and Burchard, W., Macromolecules 14, 210 (1981). . Stockmayer, W. H., and Hammouda, B., Pure & Appl. Chem. 56, 1373 (1984). . Guttman, C. M., McCrackin, F. L., and Han, C. C., Macromolecules 15, 1205 (1982). . Yamakawa, H., J. Chem. Phys. 53, 436 (1970). . Fixman, M., and Mansfield, M. L., Macromolecules 17, 522 (1984). . Jasnow, D., and Moore, M. A., J. Phys. (Paris) 38, LA67 (1978); Al-Noaimi, G. F., Martinez-Mekler, G. C., and Wilson, C. A., J. Phys. (Paris) 39, L373 (1978). . Lee, A., Baldwin, P. R., and Oono, Y., Phys. Rev. A 30, 968 (1984). . Tsunashima, Y., Nemoto, N., and Kurata, M., Macromolecules 16, 584, 1184 (1983); Nemoto, N., Makita, Y., Tsunashima, Y., and Kurata, M., Macromolecules 17, 425 (1984). des Cloizeaux, J., J. Physique Let. 39, L151 (1978). Stockmayer, W. H., and Albrecht, A. C., J. Polymer Sci. 32, 215 (1958). . Kurata, M., and Yamakawa, H., J. Chem. Phys. 29, 311 (1958). . Weill, G., and des Cloizeaux, J., J. Physique 40, 99 (1979). . de Gennes, P. G., Physics 3, 37 (1967); Dubois-Violette, E., and de Gennes, P. G., Physics 3, 181 (1967). . Akcasu, A. Z., Benmouna, M., and Han, C. C., Polymer 21, 866 (1980). . Schaefer, D. W., and Han, C. C., in Dynamic Light Scattering (ed. R. Pecora). Plenum Press, New York (1985). . Allen, G., Ghosh, R., Higgins, J. S., Cotton, J. P., Farnoux, B., Jannink, G., and Weill, G., Chem. Phys. Lett. 38, 577 (1976). . Richter, D., Hayter, J. B., Mezei, F., and Ewen, B., Phys. Rev. Lett. 41, 1484 (1978). Nicholson, L. K., Higgins, J. S., and Hayter, J. B., Macromolecules 14, 836 (1981). . Ferry, J. D., Viscoelastic Properties of Polymers. (3rd edn). Wiley, New York (1980). . Johnson, R. M., Schrag, J. L., and Ferry, J. D., Polymer J. 742 (1970). . Tanaka, H., Sakanishi, A., Kaneko, M., and Furuichi, J., J. Polymer Sci. C15, 317 (1966); Sakanishi, A., J. Chem. Phys. 48, 3850 (1968). . Osaki, K., Adv. Polym. Sci. 12, 1 (1973). . Prager, S., J. Chem. Phys. 75, 72 (1971). . Fixman, M., J. Chem. Phys. 78, 1588 (1983). Wilemski, G., and Tanaka, G., Macromolecules 14, 1531 (1981). . De Wames, R. E., Holland, W. F., and Shen, M. C., J. Chem. Phys. 40, 2782 (1967); see also Zwanzig, R., Kiefer, J., and Weiss, G. H., Proc. Natl. Acad. Sci. US 60, 381 (1968).57. 58. 59. . Fixman, M., J. Chem. Phys. 78, 1594 (1983). 61. 62. REFERENCES 139 Rotne, J., and Prager, S., J. Chem. Phys. 50, 4831 (1969). Doi, M., unpublished work. Zimm, B. H., Macromolecules 13, 592 (1980). Tsvetkoy, V. N., in Newer Methods of Polymer Characterization (ed. B. Ke), Polymer Rev. 6. Interscience, New York (1964). Janeschitz-Kriegl, H., Flow birefringence of elastico viscous polymer solu- tions. Adv. Polym. Sci. 6, 170 (1969). . Kuhn, W., Kolloid Z 68, 2 (1934); Kuhn, W., and Grin, F., Kolloid Z. 101, 248 (1942); Kuhn, W., J. Polym. Sci. 1, 360 (1946). . Volkenstein, M. V., Configurational Statistics of Polymeric Chains. Interscience, New York (1963). . Flory, P., Statistical Mechanics of Chain Molecules. Interscience, New York (1969). . Copic, M., J. Chem. Phys. 26, 1382 (1957). . Peterlin, A., and Stuart, H. A., Z. Physik 112, 1 (1939); see also Peterlin, A., in Rheology, (ed. F. R. Eirich) Vol 1, p. 615. Academic Press, New York (1956). . Landau, L. D., and Lifshitz, E. M., Electrodynamics of Continuous Media. Pergamon Press, Oxford (1960). . Onuki, A., and Kawasaki, K., Physica 111A, 607 (1982); Onuki, A., and Doi, M., J. Chem. Phys. (1986) to be published. . Koyama, R., J. Phys. Soc. Jpn 16, 1366 (1961); 19, 1709 (1964). Daudi, S., J. Physique 38 1301 (1977). . Janeschitz-Kriegl, H., Macromol. Chem. 40, 140, (1959). . Frisman, E. V., and Tsvetkov, V. N., J. Polym. Sci. 30, 297 (1958).~ MANY CHAIN SYSTEMS 5.1 Semidilute and concentrated solutions The physical properties of a polymer solution depend on solvent, temperature, and concentration. The solvents for polymers are broadly classified into two categories, good and poor solvents. Good solvents have strong attractive energy with polymers and dissolve polymers over a wide range of temperature. In such a solvent, the net interaction between polymer segments is repulsive (since they tend to contact with solvent molecules rather than themselves), and the excluded volume parameter v is positive and large. Poor solvents, on the other hand, are less keen to accommodate polymers, and precipitate polymers when the temperature is changed or the polymer concentration is increased. The excluded volume v in such a solvent can be positive, zero, or negative, depending on the temperature. For the purpose of the discussion, polymer solutions in good solvents can be divided into three regions: dilute, semidilute, and concentrated (see Fig. 5.1). A dilute solution is defined as one of sufficiently low concentration that the polymers are separated from each other; each polymer on average occupying a spherical region of radius R,. (Fig. 5.1a) In this solution, the polymer-polymer interaction has only a small effect, and any physical property is expressed as a power series with respect to the polymer concentration p (weight of polymer in unit volume). Consider for example the osmotic pressure II and the viscosity 7. In the dilute limit, these are written as l= Gar. (5.12) n=n,(1 + e[n]) (5.15) where R is the gas constant, M the molecular weight, T the temperature and 7,, the solvent viscosity. Equation (5.1@) is van’t Hoff’s law and eqn (5.1b) is the definition of the intrinsic viscosity [7]. The interaction among the polymers gives terms of order p”, so that eqns (5.1a) and (5.16) become m= Rr(E+ Asp? + . ) (5.22) n=n(1+ pln] + kxe'[nF +. ..)- (5.2b)SEMIDILUTE AND CONCENTRATED SOLUTIONS 141 e(x) ex) e(x) é é é x x x @ () Fig. 5.1. Three concentration regimes in good solvent: (a) dilute, (b) semidilute, and (c) concentrated. c(x) denotes the concentration profile along the dot-dashed lines. (o) The parameters A, and k;, are called the second virial coefficient and the Huggins coefficient, respectively. As the concentration increases, the polymer coils come closer and start to overlap each other. Since the number of polymers in a unit volume is pNg/M the concentration p* at which the overlap starts is estimated as p*Na yr Re= 1. (5.3) Note that the concentration p* can be quite low. As R, is proportional to M”, p* depends on the molecular weight as p*txM'3=M-"5 (for v = 3/5). (5.4) Thus for large molecular weight, p* becomes quite small; e.g., for polystyrene of M~10°, p* is about 0.005g/ml (about 0.5% in weight). Hence we can easily get a solution in which the molecules are strongly overlapped, but still occupy a small volume fraction. Such a solution is called semidilute. A semidilute solution is characterized by the large and strongly correlated fluctuations in the segment density such as we have in dilute solutions. Although the fluctuations decrease with increased polymer concentration, the semidilute solution still retains the same character as critical phenomena in statistical mechanics, and presents the same kind of problem as we discussed in Sections 2.5 and 2.6.! If the concentration becomes sufficiently large, the fluctuations become small and can be treated by a simple mean field theory. Such a solution is called concentrated. As we shall show later, the cross-over concentration142 MANY CHAIN SYSTEMS from semidilute to concentrated is estimated by pre = OMAN) (5.5) where M, is the molecular mass of the segment. The high concentration limit is the melt in which there are no solvent molecules. We may thus classify the polymer solutions in good solvent into three Tegimes: (i) dilute; p , 7p2 Ran ~ Ron-1)° (5.10) an is the energy of the chain connectivity and U,[RanV/keT =4>) v5(Ran — Rom) (5.11) a,b am144 MANY CHAIN SYSTEMS is the excluded volume interaction, which includes both the intramolecu- lar interaction a = b and the intermolecular interaction a+b. Since eqn (5.11) uses the pseudo potential, this model needs some alterations at high concentration where the detailed chemical structure matters, but here we will proceed using the simplest form. Although the starting point is clear, theoretical development is not easy since calculation of any average from eqn (5.9) involves a highly nonseparable integration which is common to many body problems. However, under certain conditions we can progress by using collective coordinates.” Instead of describing the problem in terms of Ru,, we focus our attention on the local segment density c(r) defined by cr) = ¥ (7 — Ran) (5.12) and consider the distribution function V{c(r)] for c(r). This method is effective if the physical quantiy under consideration is expressed by c(r) For the mathematical development, it is convenient to use the Fourier transform of ¢(r): = < | dr exp(ik + r)e(r) = z Sesplik - Ror] (5.13) c= See eme(-ik- = Ges | dkcye*"* (5.14) where V is the volume of the system. The component co is equal to the average concentration c, and does not fluctuate at all, while c, with k #0 denotes the fluctuation of the local concentration. A technical remark has to be made here. In the representation of {c,}, not all c, are independent of each other since c, and c_, are related as Cy~=CE (5.15) where cj is the complex conjugate of c,. We shall choose those c, with positive k, values as the independent components, and use the abbreviated symbol k >0 to denote the set of the independent components. When the component c, with negative k, appears in the subsequent calculation, it should be remembered that it is cy. Thus for example T= > ccs (5.16) k>0GAUSSIAN APPROXIMATION 145 is the same as the sum => > 5 act (6.17) K,>0 ky=— kyo which is expressed by the sum over the entire k space as > CyC_4 = a> 40-4 — 308= ay cyck — 4c5. (5.18) k>o k Now the distribution function of c, is given by W(4})% { TT 8Ran exp(—(UoLRon] + UilRon)V/kaT) } en xT (ce -ZDewlik- R..)) (5.19) By use of the formula o(r-r')= > exp(ik -(r—r')), (6.20) U, [Ran] is rewritten as Ui(RonlkaT = 5555, 2, explik* Ran — Rim] k ab vw 4 . 1 . =2 (Fz exp(ik RFE, exp(-ik Rin). (6.21) If this is substituted into eqn (5.19), the underlined terms can be replaced by c, and c_, because of the delta functions. Equation (5.19) is thus Tewritten as W({cx}) & exp[-Fed cxc-a| x | TT 6Ran exp(—UolRanl/keT) TT 5(c - 3 Sexplik . Ra») 6.20) ° k>0 an Let Up({c,}) and U,({c,}) be defined by Up({cx})/keT = -inf T] 6... exp(—UclRen]/ksT) 1 . x TT 6(cx—pEexrGkRen)) (5.23) U({cx})/kaT = x Dy CeCe (5.24) then W({cx}) © exp[—(Uol {cg}) + Ui({ce}))/ka TI- (5.25)146 MANY CHAIN SYSTEMS Calculation of Uo({c,}) is difficult and it is only possible to give the explicit form as a series. A systematic method? is described in Appendix 5.1. (Another method is described in ref. 11). Here we use the Gaussian approximation (or random phase approximation), which is to say that the distribution of c, is Gaussian, i.e., the free energy is written as Up({cx})/kaT = > AokCeC—« + higher order terms inc, (5.26) ik where Ao, is a constant to be determined. This approximation is justified in the concentrated solutions where the density fluctuations are small, while it is not adequate in the semidilute solutions. To determine Ay, we use the fact that for noninteracting polymers (v=0), (cyc_,) can be calculated exactly as a sum of the correlation functions of independent Gaussian chains, (cal—4)o= a0lk) (5.27) where (...)o denotes the average for the state of v = 0, and go(k) is the scattering function of ideal polymers (see eqn (2.80)), __2N [ (-mee NBR | | sdb) = Gapqap [&%P(——g—) +1} (5.28) On the other hand, for v =0, the distribution of c, becomes Pe((c4)) #exp|~E Aacac-+]=exp|-25) Awcact | (5:29) which gives (see eqn (2.1.31)) (cxe-4)0= lexi do= 5 (5.30) Comparing eqn (5.30) with eqn (5.27) we obtain Ao, as Vv Ax = Zegolky” (5.31) From eqns (5.24), (5.26), and (5.31) the total energy U({c4}) = Uo({en}) + Ui({cx}) (5.32) is obtained as Vv 1 UleakeT = 53 (Say t)ae-« (5.33) For large k, the asymptotic form of go(k) gives + 4,-KH,, 5.34) cBo(k) We (5.34)GAUSSIAN APPROXIMATION 147 This expression is also valid for small k because 1/cgo(0)=1/cN is negligibly small compared to v in the concentrated solution. Thus for the entire & values we have 242 2 Ue kaT =ZE (St vreu= ELEC +E Mac» (6.5) where & =(b7/12cv)"” (5.36) is called the correlation length. The distribution function for c, is Vu _ W({c4}) & exp| -5 Dae +E "exc (5.37) 2 12c¢ We shall now examine some consequences of this expression. 5.2.2 Pair correlation function From eqn (5.37), (cyc-x) is given by 1 12¢ (cac—z) “TPE (5.38) whence the scattering function per segment is _V _ 2 8(k) = c (cac—«) ~ pe +E) e) (5.39) The Fourier transform of g(k) gives the pair correlation function:* ~2=c{ p(k) (e(r)e(0)) -e=ef GE saKe __ famkedk 12 sin(kr) )@ay BE+E?) kr = 5 exr(-r/8), (5.40) Equation (5.40) indicates the physical significance of &:& represents the correlation length of the concentration fluctuation. Experimentally, the correlation length can be determined from the behaviour of g(k) in the small k region: 0) SD = 1+ 1Elag for k—0. (5.41)148 MANY CHAIN SYSTEMS The definition of this apparent correlation length can be extended to dilute regimes, in which case §,p) gives R,/V3 (see eqn (2.72). The behaviour of &,p, is entirely different in dilute solutions and concentrated solutions. In dilute solutions, §,)~R, increases with the molecular weight and the excluded volume, while in concentrated solutions §,5, = & is independent of the molecular weight and decreases as a function of concentration and excluded volume (see eqn (5.36)). The reason can be easily understood from Fig. 5.1: once polymers overlap each other, the excluded volume interaction tends to make the concentration homogeneous. The simple theory given above is valid only at rather high concentra- tion or at small excluded volume, i.e., near the © condition. At both these limits there are additional difficulties. At very high concentration the precise form of the potential matters. Also, near the © conditions the precise details of the interaction, in the sense of a cluster expansion passed to the two-body term, can also matter. In both of these limits it is possible to make an appropriate improvement,®’ and the results have been found to be in good agreement with experiments.'”" 5.2.3 Osmotic pressure The free energy A of the system is obtained from the integration of eqn (5.9). It is convenient to take the system of non-interacting polymers (uv =0) as a reference state. The difference in the free energy A — Ao between the two states is [ Hoa ex(—(WolRan] + Ui (Ranke) exp(—(A — Ag)/kgP) = R$ SHH | TAR. exP(-UolRanl/ko7) (5.42) which can be rewritten as exp(—(A — Ao)/kpT) | Haed(ce- 2 exptk- )) {tor k>0 Van on ao X exp(~(UolRaq] + UsfRen) Vo) | a de45 (ce - z E exp(ik- R.,)) T1 ORe, exp(—(UolRan]/koT) J Hy dee exp(—(WolCee}) + U({ea)))/ko7) (5.43) | By, dex exp(—Uol ex) ka T)GAUSSIAN APPROXIMATION 149 In the Gaussian approximation, the integrals on the right-hand side can be performed rigorously (see Appendix 5.II). Given the free energy, the osmotic pressure is calculated by oA c a I= - $6 = eT (Z) ~5y(A-A0). (5.44 The result is’ V3 (cv)? I= ket (5 + 4uct— ver). (5.45) The first two terms are easily understood: the first term represents van’t Hoff’s law (c/N)ksT (c/N being the number of polymers in unit volume), and the second term is the excluded volume interaction between the segments. The last term represents the correction to the second term due to the chain connectivity: i.e., the effect that the intramolecular excluded volume interaction does not contribute to the osmotic pressure. From eqn (5.45) it will be seen that as c—>0, the osmotic compres- sibility (@11/9c) becomes negative, which is of course incorrect. It follows that the theory presented above is only valid if 3/2. er ie, czc* = (5.46) vc? 2 This gives eqn (5.5). The situation c (2 k>0 sy tH exp(-ik- R,,) io Cese-a) +chu>; exp(ik- R.))| cconp[ D & (cee-1)o*expl—ik-(R, ~Ry)]| -en[5D > (cxc-4)v? exp[—ik + (R, ~ I} (5.50) Hence WLR] exp| 5520 Ry — Rana)? —4D OR, -Rn) | 51) where B(r) = vd(7) - 2 (cxc—4)v?e*"*. (5.52)GAUSSIAN APPROXIMATION 151 Using eqns (5.36) and (5.38), we get -2 Hr) = oo) -7> pepomik -n| = of Gopnik . vf z= = »(5@) - one). (5.53) The potential 0(r) denotes an effective potential between the segments of the test polymer. The effective potential consists of a strong repulsive part (v6(r)) of very short interaction range, and a weak attractive part (—v exp(—r/&)/42&"r) of interaction range &. In the length-scale larger than &, these two parts cancel precisely: indeed from eqn (5.53) it can be shown that Jaro =0. (5.54) Thus there is no excluded volume interaction among the segments whose mean separation is larger than &. This effect is called the screening of the excluded volume interaction. As a consequence of the screening, the distribution of the conforma- tion of the test polymer becomes Gaussian. Indeed given the effective excluded volume potential 0(r), it is easy to calculate the mean size of the polymer by use of the method described in Section 2.5.'° For example, the straightforward perturbation calculation gives (see Appen- dix 5.111) 2 NI/A, c-cls, = bb > BA”. (5.57) If we know how a certain physical quantity changes under this transfor- mation, we can discuss the dependence of this quantity on those parameters. As an example, let us consider the apparent correlation length &,), defined by eqn (5.41). We can take the same line of argument as given in Section 2.6. (i) From dimensional analysis &,, is written as Sepp = OF(N, cb). (5.58) (ii) Since &,,, is invariant under the transformation (5.57) we have bAYF(NIA, cb°2*-) = F(N, cb). (5.59) For this to be satisfied &,,, must be written ast Eapp = N°DF(cb°N?*"?). (5.60) This is rewritten by using the radius of gyration in infinite dilution tHere a single symbol F is used to denote various functions for the sake of notational simplicity.SCALING THEORY—STATICS 153 RY = N° and the segment density c* at the overlap concentration ct = N/(RO) = N*-"/b3, (5.61) as Expp = ROF(c/c*) (dilute and semidilute). (5.62) This equation is valid in both the dilute and semidilute regimes. In the semidilute regime, the functional form can be further specified. (iii) In the overlapped state of polymers, the correlation length & (=&,pp in this region) should be a function of the segment density c only, and be independent of N. (To see this, imagine that the polymers in the highly entangled state shown in Fig. 5.1b are cut into halves. The behaviour of the concentration fluctuation would not be changed by this operation.) For eqn (5.60) to be independent of N, it must be written as £=N°6(cb°N2"!)* (5.63) with : Vv v+(3v—-1)x=0, ie, x= Wor (5.64) Thus E= Ro(S) ee VG") «e944 (semidilute). (5.65) Next we consider the structure factor g(k). Since g(k) is dimensionless, and changes as g(k)—>g(k)/A under the scaling transformation, it is written as g(k) = F(kb, N, cb*) (5.66) with F(kbA", NAW, cb°A3"") = A-'F (kb, N, cb). (5.67) Thus g(k) can be written as 8(k) = cb°N**F(KN’, cb°N**-"), (5.68) or using eqn (5.60) 8(k) = CE 2p pF (kEapp, c/c*) (dilute and semidilute). (5.69) In the semidilute regime, g(k) must be independent of N, which gives &(k)=c&°F(kE) (semidilute). (5.70) _ Similarly it can be shown that the osmotic pressure is given by ky = Se” F(ele*) (dilute and semidilute). (5.71)154 MANY CHAIN SYSTEMS In the semidilute regime, II will again be independent of N, so that eqn (5.71) predicts ckgT N To compare these results with experiments, it is convenient to express the concentration by p (the weight of polymer in unit volume). Equation (5.71) is then rewritten as T= (cle*"Gr-) « 0% (semidilute). (5.72) = PRE Epip*). (5.73) M In the dilute region, this must be expanded with respect to p 2- n= [1+06+0(4) | (5.74) where a2, a; are some numerical factors. Comparing eqn (5.74) with eqn (5.1), we have a 7 — Aa= ype” Poe MU, (5.75) In the semidilute region, eqn (5.72) gives «pm, (5.76) WBv-1) n=l (4) M \p* These results are confirmed by the experiments of Noda et al.” (see Fig. 5.3). The radius of gyration of a single chain is also derived by the scaling argument. By dimensional analysis and the scaling argument, one can show that the radius of gyration must be written as R, = ROF(cic*). (5.77) In the semidilute region, the statistical distribution of the chain becomes Gaussian and R, must be proportional to N’. This requirement gives Ry = RO (cic? )8~ 220-39) ce N%e-18, (5.78) This prediction on the shrinkage of the polymer chains in semidilute solution was confirmed by neutron scattering.”* Though the scaling argument gives only a qualitative feature of the parameter dependence, explicit functional form can be calculated by the renormalization group method.’°?3* For example Ohta et al.”* gives by €SCALING THEORY—STATICS 155 2.0) o log (ITMipRT) -1.6 0 1.6 log (p/p") Fig. 5.3. Reduced osmotic pressure of poly(a methylstyrenes) in toluene at 25°C is plotted against reduced concentration p/p*. Polymers of four molecular weights ranging from 7.1 x 10* to 1.2 x 10° are shown. The slope of the full line is 1.32, Reproduced from ref. 22. expansion M am Loy _ 2m. +3) wt (3p) +g [9x 2+ * onl; (e (1-X™)In(i + »)] (5.79) where X is a parameter proportional to c/c* = p/p*, or is more explicitly expressed by the second virial coefficient A, X= AnpM. (5.80) This prediction is in good agreement with the experimental result of Wiltzius et al.” Explicit calculation for Eapp was done by Nakanishi and Ohta.156 MANY CHAIN SYSTEMS 5.4 Topological interaction in polymer dynamics 5.4.1 Entanglement effect Having discussed static properties, we now consider dynamical problems in many chain systems. Here we have to consider another very important type of interaction which arises from the very nature of the polymer: polymers are one-dimensionally connected objects and cannot cross each other. A good way of looking at this is to imagine that the polymers have no thickness, and no attractive force, like a mathematical curve in space. Clearly the excluded volume of such a chain is zero. However, even such chains can interact strongly due to the topological constraints that chains cannot cross each other. Consider two loops initially placed separately in space as in Fig. 5.4a. If one wants to calculate the partition functions of such systems, one has to exclude the configurations such as shown in Fig. 5.46 and c because such configurations are not accessible due to the topological constraints. The net effect of this is that the effective interaction between the loops is repulsive; the second virial coefficient A; is positive even if v = 0.78? The topological interaction is also very important in the problems of rubbers, in which the configurations of composite chains are severely restricted by the topological constraints of other chains. This gives an additional contribution to the elasticity of rubbers.°°! For linear polymers, the topological constraints do not affect the static problems at all since all configurations are accessible. However, the topological interaction seriously affects the dynamical properties since it imposes constraints on the motion of polymers. Indeed it is a crucial factor in the dynamics of polymer solutions above the overlap concentration. O0 Q) <> ( Fig. 5.4. Two loops in various topological states. + The topological constraints exist in the single chain problem, but the effect is generally considered not to be serious because the properties in dilute solutions are usually dominated by the external modes (translation and rotation of the chain as a whole) for which the topological constraints are not important. This conjecture is supported by the scaling theory and results of a computer simulation; both indicate that the topological constraints affect the numerical factor only, although the conclusion is not yet definitive.TOPOLOGICAL INTERACTION IN POLYMER DYNAMICS 157 4 4 PSt-Toluene 40°C PSt-Toluene 40°C 3 8 3 20% ° 18% ° ° 14% o 7. ge 2 ° a 2 3 ° 9.7% é 3 é & ° 2 1 2} s 6% 3 2 o 0 4% “1 “1 2! 1 1 1 -2 L 1 4 _L 0.0 05 1.0 15 2.0 5.0 55 6.0 65. 7.0 @ log p (g/100m)) o) log M Fig. 5.5. Steady state viscosity at zero shear rate of polystyrene in toluene is plotted against (a) concentration and (b) molecular weight. Reproduced from ref. 66. This has been well realized in the study of mechanical properties, which become quite conspicuous above the overlap concentration.°™*! Firstly, the viscosity of polymer solutions increases steeply (roughly in Proportion to p*~*) above the overlap concentration, and becomes much larger than the solvent viscosity. An example is shown in Fig. 5.5a. Notice that for high molecular weight polymers, even at the concentra- tion of only 10%, the viscosity is several orders of magnitude larger than the solvent viscosity, which is about 0.01 poise. Such high viscosity is a result of molecular entanglement in the state shown in Fig. 5.16. It can be easily imagined that the viscosity in such a state will depend strongly on the chain length. Indeed experiments indicate that at constant Fcentration Pp, the viscosity depends on the molecular weight as (see ig. 5.5b) n=M* (5.81) where the exponent x is about 3.4. This behaviour is rather universal, independent of temperature, solvent and the molecular species (as long as the polymer is linear and flexible), which indicates that the phenomena are governed by the general nature of polymers.158 MANY CHAIN SYSTEMS Secondly, the polymeric liquids show conspicuous elasticity. For example, if one stretches a polymeric liquid (say chewing gum) quickly and holds it, one will feel a restoring force which decreases with time. If one releases the specimen before the force vanishes completely, it shrinks like rubber. The characteristic time for such elastic behaviour to be observed (i.e., the longest relaxation time tax of the restoring force) depends strongly on the molecular weight of the polymer in the form Timex® M* with x=3.4 (5.82) which again indicates that the relaxation of the molecular conformation is extremely retarded by the molecular entanglement. There is much experimental evidence which indicates the dominant role of the topological constraints in the dynamical properties of polymeric liquids, a comprehensive discussion on which is given in refs 30 and 31. 5.4.2 Rigorous approach Although experimental results on the viscoelasticity of entangled poly- mers were established quite early, theoretical explanation was not successful for a long time. (Some of the early theories are reviewed in ref. 31). This is due to the difficulty of handling the topological constraints. From the theoretical point of view, the topological constraints provide a unique class of problems. This interaction is quite singular: for example, in linear polymers its effect is null for static properties but quite serious for dynamical properties. The interaction has no parameter which characterizes the strength. Rigorous theoretical treatment of such inter- action is quite difficult. This may be seen in a simple example. Consider the problem of calculating the second virial coefficients A, of the ring polymers shown in Fig. 5.4. The first step in such a calculation is to find a mathematical expression which distinguishes the non-entangled state (such as (a)) from the entangled state (such as (b)). How can we do that? A way of doing it is to use topological invariants which remain the same as long as the polymers do not cross each other. A classical quantity found by Gauss is the integral for two loops:** = Fe fanpam| (FE) x (=) oe (5.83) where the integral is done over the closed contour of the loops 1 and 2. The integral is 0 for the configuration (a) in Fig. 5.4, and 1 for (6) and remains the same if the loops do not cross each other. (This can be proved by using Ampére’s law in magnetism: the integral is equal to theTOPOLOGICAL INTERACTION IN POLYMER DYNAMICS 159 number of times that the (directed) loop 1 passes through loop 2.) However, the Gaussian invariant (5.83) is not enough to specify the topological class of loops because, for example, the configuration (c) gives the same value of J as the configuration (a). A better specification is achieved by the use of Alexander polynomials, which amounts to an analysis of the order of crossing of the projections of the curves on the surfaces, but the expression becomes too complicated to be used for analytical calculation. Despite the difficulty, progress has been made in the static problems (e.g., the calculation of A, of ring polymers” and the rubber elasticity“), by using the Gauss integral as a principal index for the topological class. Rigorous results can be obtained for the entanglement between a polymer on the plane and a line standing on it,**“* and the two-dimensional problem can generally be solved using Riemann surfaces. ** In the case of linear polymers, the entanglement effect appears only in the dynamical properties, and it appears at first sight one does not know where to start. Edwards” suggested that the entanglement effect can be described by the Smoluchowski equation, ow Qo Soo ape Hom (ho bm + ow aU ORom ORom ¥), 6.84) because the mobility tensor H,,., is zero for the pair of segments which are going to cross each other. To see how this works, consider the one-dimensional motion of two particles, 1 and 2 (see Fig. 5.6). If we x x Oo Oo @ x % =X. My (b) Fig. 5.6. (a) One-dimensional Brownian motion of two particles which cannot cross each other, (b) phase space of the system.160 MANY CHAIN SYSTEMS assume that the diffusion constant D, =k,TH, is a function of lx —x,| only, the Smoluchowski equation is written as ow a ow a “Ot Ox, DOT, xz DOs a wi a ow + 3y, D2) 5, + 5y, Pa ZT (5.85) Using 1 1 Xa +x2), %= 7 i — 2) (5.86) eqn ee is rewritten as & = (00) + D&) + OO - Dewy) se (5.87) at a The underlined diffusion constant, which corresponds to the relative motion of the particles, vanishes when X,=0, i.e., when x,;=x. This guarantees that the flux of the representative point across the line x, = x2 is zero, i.e., the particles cannot cross each other. Notice that this conclusion holds for any functional form of D (provided it is finite at X,=0). An attractive feature of this approach is that it guarantees that the equilibrium properties with the topological constraints are the same as the ideal chains. However, rigorous solution of eqn (5.84) is not easy, and approximate treatment such as the preaveraging approximation does not give the correct molecular weight dependence for 7 and Tmax. At this moment, this idea still remains to be explored. 5.4.3 The tube model Though rigorous theory for the topological interaction is extremely difficult, it has been shown that a highly entangled state can be treated by an effective model, the tube model. This model assumes that due to the topological constraints, the motion of the chain is essentially confined in a tube-like region made of the surrounding polymers (see Fig. 1.6). This model, originally proposed for the problem of rubbers,“ offers a basis for the dynamics of chains in a network,*” which has been quite successful in explaining many dynamical properties of entangled polymers.** We shall give a detailed account of it in the subsequent chapters. Though the tube model is successful, our present understanding of the dynamics in entangled systems is still incomplete. Agreement between theory and experiments is not yet complete as we shall discuss later. More seriously, the tube model does not describe all aspects of the dynamics: it describes properties which depend on a single chainDYNAMICS OF CONCENTRATION FLUCTUATIONS 161 dynamics, but cannot so far handle problems which involves the collective motion of many chains. Progress can be made on aspects of this problem, particularly in concentrated and semidilute solutions, as against melts, and this we shall study in the rest of this chapter. 5.5 Dynamics of concentration fluctuations 5.5.1 Kinetic equation In Section 5.2, it was shown that the Gaussian approximation for the collective coordinates c, is quite useful for describing the static properties of polymers in concentrated solutions. It is a natural temptation to generalize this approach to dynamical problems. The central assumption to be made in this approach is that the set of coordinates {c,} are good variables for characterizing the state of the system, and that therefore a closed equation can be constructed for their time evolution. This approach is quite analogous to the critical dynamics for binary solutions of low molecular weight,” where the dynamics of the system is described by the phenomenological Langevin equation Butte), Bcy: (5.88) a a = 2 — Ly where U({c,}) is the free energy, Lj, are phenomenological kinetic coefficients and 7, are Gaussian random variables satisfying (nm) =0, (re (t)re(t')) = 2kpTLyy-d(t - t'). (5.89) In the polymer problem, the validity of this equation is not obvious since the description of the polymeric system by {c,} disregards the chain connectivity and, therefore, neglects the entanglement effect. However, as far as the dynamics in the short time-scale is concerned, this will not be a serious problem} since, as we shall discuss later, the topological constraints are not important in the short time-scale dynamics, Indeed it will be shown that the initial slope in the dynamical structure factor is correctly described in this approach. In the long time-scale, on the other hand, the validity of eqn (5.88) is not clear, and it may well be that the theory has to be modified in future. Fortunately, many experiments Telated to concentration fluctuations are concerned with the short time-scale motion, so that it is worthwhile to pursue the idea in detail. t Here short time means that it is shorter than the reptation time t, which will be discussed in Chapter 6. For the phenomena which involve long time-scale, or large length-scale motion of polymers (as happens in the problem of phase separation™” and @ regimes’), the Present theory is not valid.162 MANY CHAIN SYSTEMS Now for the polymer problems, U({c,}) has been calculated in Section 5.2.1. To determine L,,-, we use the Langevin equation for R,, (see eqn (4.42). op - 3U[Ren] ay Ren = = Hantm (- ORom + hn) =_ _ 2U [Rand = Haro * Sp Tn (5.90) where Hontm = H(Ran — Rom) and Ton = 2 Hono * Son (5.91) whose time correlation function is (Fan (t)) = 0, (Tan (t)Fom(t")) = 2k e THanomd(t — t'). (5.92) Equations (5.13) and (5.90) give a 1y., IRan . arch) = 7 Dik exp ik - Ran) U[Ran] Rom aly ih (—Hoaom + in) exPCiK * Ran). (5.93) V an,bm Comparing eqn (5.93) with eqn (5.88), we have ly. . nt) = pm ik * Tan (t)exp(ik * Ran). (5.94) From eqn (5.92) and (5.94), it follows that Cralyrel@')) = Fa Be (lll) BE X exp(ik * Roy + ik’ « Rym)) (5.95) = 55D 2a TO — 1 RK: Hanon an,bm x exp(ik + Ran + ik’ + Rom): Using Haan = | aes a ttenptig = (Rax—Rim)) (5:96)DYNAMICS OF CONCENTRATION FLUCTUATIONS 163 we have —26 k-k’—(k-@)(k'- trade) = a | EE EOE “5 oxi +q)+Ra, t+ i(k’ —q)+ Rom] og kek’ -(k- OK) = —25(t— ‘ykeT | Gay nea? a (5.97) From eqns (5.89) and (5.97) we have Lye = —[ Ot KDR) (5.98) (any ng? Cueeeme Equations (5.88) and (5.98) give a nonlinear equation for c,. To proceed further, we use the preaveraging approximation, i.e., replace Ly, by its average in equilibrium _dg_k-k—(k-Ak'-4 Lye (Lie eq = 2 4) (Cergle'—g)eq: (2m)° 14 G.99) Since (Cy+gr'~g)eq is given by (c/V)64_«(k +), (Lax')eq is written as (Lit eq = On-0 Le (5.100) where _dq B-(k- ay Ly S GY ng? g(k+q). (5.101) Therefore the Langevin equation becomes 9 yey, Wud) my o(t)=—-Ly 3c +7(t). (5.102) The savivaen form of the Smoluchowski equation is = W((c})= Lae Stal kat woth)! 5.103) —k In the Gaussian oo an is given by U({cx}) = Xkp Tap ok) CRE-k (5.104) a(k)164 MANY CHAIN SYSTEMS Thus the Langevin equation (5.102) becomes 2 Cy = Tee, + 4 (t) (5.105) where Tyo kyT EE (5.106) why " or using eqn (5.101) 2 2. Ty=keT og glk+ qk Gy (6.107) ay gk) ng? Equation (5.105) is the basic equation for the dynamics of concentration fluctuation. We now study the consequence of this equation. 5.5.2 Dynamic light scattering Equation (5.105) is a linear Langevin equation which has been studied in Section 3.5. The time correlation function (c;,(t)c_,(0)) is thus easily calculated as (ce(t)e—4(0)) = (cxc-x)exP(—Tit). (5.108) Hence the dynamical structure factor is wth =F (ele) =e erp(—Te). 6.109) In this approximation, the decay of g(k, ¢) is single exponential. Note that I, given by eqn (5.107) is precisely the same as the exact initial decay rate given in Chapter 4 (eqn (4.110)). Thus eqn (5.109) is exact for t— 0. The explicit form of I, is evaluated from eqns (5.39) and (5.107) (see eqn (4.111)*) as __keT f 1418? (eee, kal The 4n°n, | da pe 1+ get (z 1g | 1] AeT Re RF 5.110 - Gant (ks) (5.110) where F(x)=> 1+eY + (x? - Dtan7“(x)). (5.111) The function is plotted in in Fig. 5.7.DYNAMICS OF CONCENTRATION FLUCTUATIONS 165 FidDecopk® 0.1 1 10 ke Fig. 5.7. F(x) =T;/Dcopk” is plotted against x = ké. The expression is simplified in the two limiting cases (i) For x >> 1, F(x) =32x/8, whence, _ ket 3 Te=T69, © KE 1. (5.112) This is precisely the same as eqn (4.113). This must be so since in the region of k&>>1, the dynamics is dominated by the single chain behaviour. The condition of k&>>1 is not easily attainable by light scattering, but the result has been confirmed by neutron scattering.” (ii) For x «1: F(0) =1, so that kT Te Gane The k dependence of I, is again the same as in dilute solutions (eqn (4.99)). In fact this k dependence holds quite generally in the small k Tegion, for consider the Taylor series expansion of I’, with respect to k. In the isotropic state, the most general form is Ty=anta,k’+a(ePPt.... Since T, must vanish at k=0, a)=0, whence the above equation is written as Ro kEKI. (5.113) Ty =Dappk? (k>0). (5.114) This defines an apparent diffusion constant D,,, for the entire concentra- tion region.166 MANY CHAIN SYSTEMS In dilute solutions, it was shown that D,,, agrees with the self-diffusion constant defined by Do = lim (Ro(0 ~ Ro(0)?) (6.115) where Rg is the position of a test polymer. On the other hand, in concentrated solutions, eqn (5.113) indicates that D,,, is given by’? kT 627n,€ which is called the cooperative diffusion constant. It should be noticed that D.oop has no relation to Dg. Indeed Dg should decrease with concentration (because of the molecular entanglement), while Daoop increases with the concentration. The increase Of D.oop With concentra- tion results from the fact that the restoring force for the concentration fluctuations is larger at higher concentration. The concentration dependence of Dzoop is obtained from eqns (5.36) and (5.116): Deoop = (5.116) Doop *c'” (Gaussian approximation). (5.1174) On the other hand, in the semidilute regime, use of eqn (5.65) gives Decop® 0%. (5.117) From experiments, the exponent in D.oop*c* has been found to be between 0.5 and 0.75°**” and this is discussed in detail by Schaefer and Han. It must be mentioned that although the behaviour of the dynamic light scattering in the short time-scale is well understood, theory is still lacking for the behaviour in the whole time-scale. Experimentally it has been observed" that in some systems the structure factor does not decrease in a single exponential manner and has a long tail. The long time-scale behaviour is considered to be related to the topological interaction, but quantitative theory is not yet given. 5.5.3 Form birefringence As mentioned in Chapter 4, the birefringence of polymer solutions has two origins: the intrinsic birefringence which arises from the orientation of polymer segments and the form birefringence which comes from the anisotropy in the correlation of the segment density. In dilute solutions, the relative contribution of each of these terms depends both on polymer molecules and solvent molecules. In concentrated solutions, the form birefringence decreases with increasing concentration,” while the intrin-DYNAMICS OF CONCENTRATION FLUCTUATIONS 167 sic birefringence increases steeply like the viscosity. (Note that the intrinsic birefringence is proportional to the stress (see eqn (4.238).) This leads to an important relation called the stress optical law® which states that the stress and birefringence are proportional to each other. Detailed discussion of this relationship is given in Chapter 7. Here, as an application of the Langevin equation (5.105), we shall calculate the form birefringence and show that it actually decreases with the concentration. According to eqn (4.237), the form birefringence is given by on\? 1 n=—25 (Z) aap | teCEaky—Wenlek) 6.118) where g,(k) is the (static) structure factor at time t, gh) =~ (exlte-4(0)- 6.119) To calculate g,(k), we first consider the Langevin equation under the macroscopic velocity field Ualt, t) = Kap(trp- (5.120) The macroscopic velocity causes a flow of c(r, ¢), 2D) 8 -r0(4, 0). (5.121) Ot | acite or The Fourier transform of which gives the drift for c,(r) as Sex) =k-K() Sali. (5.122) drift Adding this to eqn (5.105), we have the Langevin equation in the presence of the velocity gradient, é c(t) = —Tycx(t) + y(t) +k w(t) + Sal. (5.123) The equation for g,(k) is obtained from eqn (5.123) as a 50) = 2 (P00) + (00 *S9)] =~ gh) + © (ree-a(O)) + (eaCOr-0(0)1 t+k-K(Q)- Ae (&). (5.124)168 MANY CHAIN SYSTEMS The underlined terms can be calculated in the same way as in eqn (4.141), and gives 2k,TL,. Thus, using eqn (5.106) we have 3 gk) = = 20 u(8:(k) — Seq(k)) + & + w(t) alk) (5.125) where g.q(k) is the equilibrium structure factor. Since eqn (5.125) is a first-order differential equation, it can be solved by a standard method,} but here we consider the simple case of «(t) small and independent of time. The steady-state solution of eqn (5.125) is written as B(K) = Beq(k) + 81(k) (5.126) with 3 Kap ke Bull) = 5p KH Se Bal) =e KE gah). (5-127 Substituting this into ™ (5.118), we have 9 = ~2£ Eire p= 25 (ZY | abby ~ 1509) EM ES gah). 6.128) The integral over the direction & can be calculated by using x | dkk ks =450p (5.129) & | hk hy hy ky = 48(Sap5yv + Sau. 6v + Sav pu). (5-130) The result is written as 19, = AM Kap + Koa) (5.131) with any? 1 aQ= -£(2) in ) sant sas [awe 7 3 FBealk). (5.132) t Let Fig(t, t’) be the solution of the equation a Bao lt 0) = Fault, )Kpu(t) with the boundary condition F,9(¢, t) = 5,9, then the solution of eqn (5.125) is : . Ver { ar a(h)=2 0 kyr f at exp| 2] aT pe a]briey I tSCALING THEORY—DYNAMICS 169 In semidilute solutions, I, and g.q(k) are given by the following scaling form (see 5.3 and 5.6): T= ne Per, UKE), Beq(k) = c8°F,(KE). (5.133) Equations (5.132) and (5.133) give 3 40 = nea aa (2 ay" (5.134) If On/dc is regarded as independent of concentration, A{? depends on the concentration as AQ x 783 a eo 4 (5.135) which decreases with increasing concentration. On the other hand, for concentrated solutions, substitution of eqns (5.39) and (5.110) gives 1_(8 £ (2) = poth (By = npr) § 3c) kgTun \de (6.136) which also decreases with concentration since 9n/dc decreases with concentration. Therefore in both the semidilute and concentrated solutions, the contribution from the form birefringence decreases with the concentra- tion. Experimental results are summarized in refs 62 and 63. 5.6 Scaling theory—dynamics The dynamical scaling law discussed in Section 4.3 can also be used for semidilute solutions.'~* The hypothesis is that any physical quantities characterizing the dynamics satisfy the scaling transformation A> HA (5.137) when the basic parameters are changed as N>N/A, b> ba’, cel A. (5.138) It is worthwhile to ask whether dynamical scaling is valid in a system in which the topological constraints are important since it is well known that topology is extremely dimension dependent; for example, vorticity is conserved in two dimensions, not in three dimensions; knots exist in three dimensions but not in four dimensions, etc. Such effects do not appear in the renormalization group calculations which are quite con- tinuous in variation of the dimensionality of the system. On the other170 MANY CHAIN SYSTEMS hand, one can argue that since the topological constraints introduce no new length-scale, it would not upset the scaling arguments; rather the scaling argument would give a common restriction on the theoretical results for systems both with and without topological constraints (the latter of course being a hypothetical system in which chains can pass through each other like phantoms.) Here we shall take this viewpoint. Various experimental results are quite consistent with the scaling assumption in good solvents, though the validity is not yet established in © solvents. As an example let us consider the apparent diffusion constant D,p, measured by dynamic light scattering (see eqn (5.114)). First we use a dimensional analysis. The relevant parameters in the problem are b, c, N, kgT, and n,, 80 Dapp is written as Dapp = , N). (5.139) This must be invariant under the transformation (5.138), whence ; F(cb, N) ag hcorn, N/A) (5.140) which leads to = kT 3ay—14+3 Dapp = aN F(cb°N7**9"). (5.141) Using the diffusion coefficient in the dilute limit D9’ ~k,T/N’b and c* = N/(N’b)’, eqn (5.141) is written as Dapp = DYF(cie*) (dilute and semidilute). (5.142) This equation is valid both in dilute and semidilute solutions. Unlike with dilute solutions, the scaling argument does not predict molecular weight dependence explicitly. (This must be so since one can derive the same formula for the self-diffusion constant Dg. Further information is needed to distinguish the functional form of D,,, from that of Dg.) If one assumes that D,p, is independent of the molecular weight, then one has =H ( B3)"r-D ~The (5.143) D, ‘app which agrees with eqn (5.117b). On the other hand, Dg depends on the molecular weight. In the next section we shall show that D « N~? by the tube model. For this to be written as DF (c/c*), Dg must be written as kgT De= nN b(cleye OD * ee N72¢7™ (5.144)SCALING THEORY—DYNAMICS 171 which decreases with concentration. The self-diffusion coefficient in dilute and semidilute solutions has been measured by forced Rayleigh scattering.” The result is shown in Fig. 5.8, which includes the data of Deoop and displays the different concentration dependence of Dg and coop: Likewise the viscosity 7 of the solution is shown to be n =,F(cle*). (5.145) Again, the molecular weight dependence is not determined by the scaling argument alone, but eqn (5.145) predicts that if 7/7, is plotted against «x @ 10% Dg, Dogo (em? s°') ay SLOPE -1.75 \ y, +My = 78300 \ ©” = 245000 x = 598000 ¥ y= 754000 x I 10° F © from ORS. 10" 10° 10? 107 10° Weight fraction of polymer Fig. 5.8. The apparent diffusion constant D,,, obtained by the dynamic light scattering® and the self-diffusion constant D, obtained by the forced Rayleigh scattering of the chains for four different molecular weights. Reproduced from ref. 64.172 MANY CHAIN SYSTEMS c/c* = p/p*, then the viscosity curves of various molecular weight can be superimposed. Such superposition had been known prior to the scaling theory.* If one uses p*[n] = constant, eqn (5.145) can be rewritten as n =1,F(oe[n)). (5.146) This form also has been established experimentally for a long time.® 5.7 Effective medium theory 5.7.1 Failure of the scalar field description The collective coordinate formalism given in Section 5.5.1 is not capable of describing the whole aspect of polymer dynamics. For example, the quantities which are related to the orientation of the bond vector OR,,/dn, such as stress, birefringence, and electric displacement, are not expressed by {c,(t)}, and therefore cannot be dealt with in this framework. This is in contrast to the critical dynamics of binary solutions® in which the dynamics is entirely specified by an equation of motion for the concentration fluctuation. To discuss the orientation dynamics of polymers, other collective coordinates are needed. A possible collective coordinate is® =1y cos(e™ Vexpti Cap(t) = 750 a expCik Run) (5.147) which can describe orientation of bonds. For example, the tensor Rana od) 2 ( on On 6.148) is expressed as er) (etn) x ( N/ \dky kp | (5.149) However, the usefulness of such coordinates has not yet been fully worked out. An alternative idea proposed by Edwards and Freed” is to consider the motion of a single chain in an effective medium which includes the effect of the other chains. The property of the effective medium is determined self-consistently from the single chain dynamics. Though this method fails to describe the entanglement effect appropriately, it indicates an important aspect of the hydrodynamic interaction in the concentrated system, which is the screening of the hydrodynamic interaction.EFFECTIVE MEDIUM THEORY 173 5.7.2. Hydrodynamic screening The concept of hydrodynamic screening is seen if one considers the velocity field created by a point force in a fluid (see Fig. 5.9). If the fluid is a pure solvent of viscosity y,, the velocity field u(r) is long range, decreasing like 1/n,r as predicted by the Oseen formula. On the other hand if there are polymers in the fluid, the situation is different. On a small length-scale, the velocity profile will still be like 1/n,7 since it is not disturbed by the polymers. However, on a large length-scale, the velocity profile will obey the macroscopic hydrodynamics, so that the velocity must decrease like 1/nr, where n is the macroscopic viscosity of the solution. If 7 >> n,, the velocity field falls quickly and the effect caused by the point force becomes very weak beyond a certain characteristic length &,, called the hydrodynamic screening length (see Fig. 5.9). In such a situation, the hydrodynamic interaction becomes negligible between i c 7 r @) () al (a) pure solvent (ce) Fig. 5.9. (a), (b) Velocity profiles caused by a point force in a pure solvent and a polymer solution. (c) The velocity v,(r) on the axis shown in (a), (6) is plotted against 7.174 MANY CHAIN SYSTEMS segments whose distance apart is larger than &,. This is called the hydrodynamic screening. As a consequence of the hydrodynamic screen- ing the dynamical behaviour of a single chain becomes Rouse-like as against the Zimm-like behaviour in dilute solution. Though the idea of hydrodynamic screening is easy to understand, the actual formulation needs somewhat elaborate methods. In the following, we shall explain the method using the crudest approximation. The reader who is interested in a complete form of the formulation is advised to see the original papers.'°7”7 5.7.3 Effective medium theory First let us consider how the velocity profile caused by a point force is affected when a small number of polymers are present in solution. For simplicity we consider the steady state in the velocity field, though this assumption is not essential.”"”” Let F.. be the point force acting at the origin. For pure solvent, the velocity perturbance is given by vo(r) = H(r) * Foxe (5.150) where H(r) is the Oseen tensor given by eqn (3.106). By the Fourier transform, Uy = | drup(r)exp(ik +r). (5.151) eqn (5.150) is written as 1-kk Von Te Fox. (5.152) Our problem is how this field is affected by the polymers. In principle the answer is given by solving the Smoluchowski equation ow aU a 2 aR. [> Ham . (ssa t oR. ) - w(Ra) =0. (5.153) The hydrodynamic interaction is included in Myjom = H(Ran — Rom). If ¥ is obtained for a given u(r), the average velocity field is calculated as B= w+ DD (H(r ~ Ram) * Fam) (5.154) where a Fam =~ 5p— (koT In B+ U) (5.155) and the average {...) is taken for the distribution function WV.EFFECTIVE MEDIUM THEORY 175 Now it is difficult to solve the diffusion equation (5.153) which includes polymer-polymer interactions. So we first focus our attention on a particular polymer, which we shall call the test polymer, assuming that the velocity field created by the other polymers is known. Let 0,(r) be the velocity field created by the external force and the polymers excluding the test polymer, i.e., Bale) = vole) + DD (HU — Ron) Fon). (5.156) The effect of the test polymer is solved by the Smoluchowski equation for the single chain: a aw au a _ 2 oR. [Hom . («sR ™ an) | ~ 2 5p,” PaRan)]=0. (5.157) The velocity perturbation created by the test polymer is then calculated by 50,(r) = > (H(r — Ram) * Fam): (5.158) Since 6%,(r) depends on @,(r) linearly, it must be written using a spatial response function =(r), 60,(r) = - 5 | dr’S(r—r')a,(r’). (5.159) Equation (5.156) is then written as B_(r) = vo(r) + D, 50,(r) = w(r) -i5 J dr'Z(r — r')v,(r'). (5.160) b&b 6 Now we assume that the perturbation 5%,(r) is not large, and replace 0,(r) and @,(r’) by #(r) and v(r’) respectively: a(n) =v) -55 | dr'2(r— r')5(r’) = nr) -£ | dr’E(r— r')0(r’) (5.161) where c/N is the number of polymers in unit volume. Equation (5.161) is easily solved by the Fourier transform as Vow. v(k)= c 1+55(%) = Ak): Fox176 MANY CHAIN SYSTEMS where (5.162) and (Kk) = | drz(re*"t (5.163) The function A(k) represents the effect of the point force in the presence of the polymers. Comparing eqn (5.162) with eqn (5.152) we may interpret nea(k)=n [1+ £20] (6.164) as an effective viscosity of the polymer solution on the length-scale of 1/|k|. The macroscopic viscosity is given in the limit of k> 0 as n=n[1+<=)| So far we have been considering the dilute solution. This allowed us to use the Oseen tensor H(r) in solving the single chain problem, eqn (5.157). In the concentrated solution, however, the hydrodynamic interactions among the segments are affected by the surrounding poly- mers. Thus we have to replace Hyram in eqn (5.157) by the effective interaction Ajram Which is derived from A(k) given by eqn (5.162). (5.165) k=0 dk . - Ham > Rana = A (Ran ~ Ran) = | ns DLR * (Ran — Ran). (5.166) This gives a closed equation for the effective hydrodynamic interaction. 5.7.4 Example As an example, we consider a solution in @ condition. For simplicity of notation we omit the suffix a to denote the quantity of the test polymer. Since the general solution of eqn (5.157) is involved, we use a crude approximation: we assume that in the velocity field 0(r), the test polymer behaves as a rigid chain, and moves with constant velocity V (the rotational motion being neglected for simplicity.). The equations deter- mining V are V=> Aum? Fn + 0(Rm) (5.167)EFFECTIVE MEDIUM THEORY 177 id an DF, =0. (5.168) ™ Equation (5.167) indicates that the fluid velocity at R,, must be V and eqn (5.168) represents the condition that the total force acting on the polymer is zero. To solve the equations we use the preaveraging approximation Fm ~ (Bln a= Bl = J as (exBCik (RyRy) og R)- (5.169) From eqns (5.162) and (5.169) it follows that 2 Pum =| 3 qe (ex CIK + (Ra — Rm) eq. (5-170) (ny anh 5 (k) “ Equations (5.167) and (5.168) then give Fn = — > Snnd (Rr) (5.171) where San =F Dan = (EEMaE ms) (EE). a7) Thus the perturbed flow is given by eqns (5.158) and (5.171) as 88 (7) = —B (Hr Rn) - SnD (Rn) eq =~ fae [arr Hie P)8(0" — Rp) 8(r' — Ry) exon * (0) = AE for fae aH (OO Rade RD oak) (5.173) Comparing eqn (5.173) with eqn om we get 2) =VE far HH U0 ~Rn)B(Rs))oabam 5.174) or in Fourier transform, 3(6)= | dro Be) = 57a D (emit Rn RDaabame (5.175) on (5.170), (5.172), and (5.175) give a nonlinear equation for (k). To proceed further, we use the diagonal approximation for the normal178 MANY CHAIN SYSTEMS coordinate representation of vr where oe = “wk, 0 Joos?) (5.176) ins ” Py 00s ES cos( PF Vn (5.177) In this approximation, eqn (5.172) gives Brn = wo oos( 777 )eos(™) (5.178) (Note that the term of p = 0 is omitted from the summation.) Thus in the ® condition, eqn (5.175) becomes ~ 11 Per) (=) 2 Lp2p pam k OPC 37k? |n ml arg, oo w oS . (5.179) The sum over n and m are calculated as in eqn (4.52) > cos(P5"Joos( PE Jexp(—ab?&? |n-m)) nm ~ fan *cos( p (Fos (= (n +) exp(—36°%? |m'l) dm NT b?k716 ay J , vos expt 367K? |’) => + pal CGF (5.180) Then ayy = 1 1 50) - SW omnes ORR, 6.181) Similarly, h, is calculated from eqns (5.170) and (5.177), - 1 2 4=3D | oF Noum J Qz) 3nde(1 + pe) x exp(—4b7k? |n — ml)oos( >= eos ‘4 pm) 2 2(b7k2/6) ) (pxINy? + (B16)? =! ow Om anae(1+£20)EFFECTIVE MEDIUM THEORY 179 2 b? ¢ dk 1 1 (6.182) Equations (5.181) and (5.182) determine E(k). Since this is a nonlinear integral equation the general solution is quite hard to obtain,” but in special cases, the characteristic results can be discussed. Without going into the detail, we quote the main results. (i) In dilute solution, eqns (5.181) and (5.182) can be solved by expanding E(k) with respect to c: E(k) = Eq(k) +c, (k). (5.183) The first term gives the intrinsic viscosity [q] and the second term gives the Huggins coefficient (see eqn (5.2)). The result is ANa(VN OY (Sas (=< View (2? ) 18) and 3,0 -2 bua (Se) =0.757... (5.185) Equation (5.184) is larger than eqn (4.149) by a factor V2. Equation (5.185) is slightly larger than the experimental value which is between 0.4 and 0.6.” Thus the numerical factors differ by about 50%, but this can be improved by removing the rigid chain approximation used in the calculation. (ii) Given the effective hydrodynamic interaction, one can calculate the relaxation time 1, for the normal mode as — ep (1 + ac(2)” +. ) (5.186) where t{ is the longest relaxation time in the dilute limit (c— 0) and p is about 0.5 in © condition. Equation (5.186) explicitly shows that the p dependence of 1, changes from the Zimm-like (1, «p~*”) to Rouse-like (1, «p~) as the concentration increases. This result has been compared with experiments on dynamic flow birefringence.” _ (iii) At higher concentration, over a wide region of k, H(k) is expressed as” l-kk AW) = Es Ee) (5.187)180 MANY CHAIN SYSTEMS or in the r spacet A(r) = exp(—r/En)/nr (5.188) which explicitly shows the screening of the hydrodynamic interaction. The length &,, is called the hydrodynamic screening length, and given by”? &y = 2/acb?. (5.189) In this regime the steady-state viscosity is obtained’? as 7b° n= (1 + SN) (5.190) This again indicates that the molecular weight dependence of the viscosity becomes Rouse-like (7 « M) for high concentration. Appendix 5.1 Transformation to collective coordinates The free energy U, is formally calculated from exp(—Un({ea) ko) = | TT 8Ran TT 8[ e473 exptik- Re) a k>o x exp(—Us[Ran]/kaT). (5.1.1) From the definition of the complex integral given in eqn (2.1.28), it is shown that = doe) lid (ey — Tae-4n~ J (I B)ew(id(-Ave+) 612) where @_, means @,*. Using this, eqn (5.1.1) can be rewritten as exp(—Us({oa})/kaT)* f TT doef TT OR on k>0 a x enp(id @-a6e- 5 Dx explik - Re.) k+0 k+0 x exp(—Up[Ran]/ksT). (5.1.3) To evaluate the integral on R,,, let or) = “5 » o-.8". (6.14) k#t0 + Note that eqn (5.188) is not correct for 7 >> &, since the asymptotic behaviour of H(r) must be 1/yr, but the important region of integration in (5.181) and (5.182) is not this region, so (5.187) is a good self-consistent solution.TRANSFORMATION TO COLLECTIVE COORDINATES 181 Then eqn (2.46) gives N 2 [on exp| E 60Ron)~ 553 [an (Bet) | = far dr'Gee #', Ns) " 0 where G(r, r’, N;[@]) satisfies 2 [2 -Fe-omlowr.wten=ae-r900y. 6.15) Thus exp(-Unl(ea) ke) = [TT deeexp|id 6-24] k>0 +0 Np x [ | ar | ar'G(r, ', N; top| (5.1.6) where N,=cV/N is the total number of polymers in the system. Equation (5.1.5) can be solved in a power series of # as G(r’, Ns) = Golr= 2's) + [an 0 N n; x | dr, Gor — 43 N —2,)9(4)Go(r, — r'31) + | dn, | dn, 0 0 x [an f ars Gale— nsN — m1) O(R)Gilr~ 05m — m2) H(IGOLOa— sm) fees 6.17) Using the relation [arcice -r;N)=1 we have N n far arco, rN: [oD = v+N{ dro(e) + [anf dng ar, 0 0 x | drab (n)o()G(r, 95m —M). 6.18) Using eqn (5.1.4) and the definition of go(k) (see eqn (2.70)), this is written as farfaroe.r. nstop=v[i-3 Zbeb-s8k)] 6.19)182 MANY CHAIN SYSTEMS Thus exp(—Ual(ex))lka)= | TT aoe exp] id, #-1ce +N, tog(1 - x = $480k). (5.1.10) Assuming that the fluctuation in @, is small, we expand the logarithm and evaluate the integral over @,: ths = | Tam ex0[ iS, eee - ey $1$-180(0)) ~ | Thao exo] d (ities +iguct ~£ ox6-s80(h))| |- Da (5.1.11) which gives eqn (5.31). Appendix 5.11 Osmotic pressure in concentrated solution Using the formula in Appendix 2.1, we have {z Tl dc, exp(- Vv = Eat) exp(—(A — Ao)/kaT) = exo] —F ve? | Hdexeno(-V 2, = G5) = exp[-F ve] TT @tayent) (&) =exp| —7 ue? + 5 Hog & (5.11.1) He [-7 2 aot voaiv kag? (A~AbllkeT = 700 +5505 | dk og(<+F-), (5.11.2) The osmotic pressure is thus given by =£ _ (A= Ao) WksT = 7, av T,cV constant Va 24 &? Vv 3g? + ; say | a[toe(* Pa =) +R? ov c uv “N 2° < 5¢ eH 3G a | seo [loe(1 + gp) - FFI: (5.11.3)PERTURBATION CALCULATION OF (R?) 183 The underlined integral is performed by the integration by parts: [onr[re(t+ 7 2) nel fe ws x [roe( + 3) - sel -%. (14) l+x Hence T=ksT [E+5e - Tap (5.11.5) Appendix 5.111 Perturbation calculation of (R7) Let (...)o be the average for the ideal chain and U, be defined by He aa 5(R,— Rn) (5.11.1) then the perturbation method described in Appendix 2.III gives (R?) = (Rw —Ro))of1 + (D,)o] — (Ry —Roy’G,)o. (5.11.2) The first term gives N m (Ry ~ Re) + (O;)o] =O 1+ | dm ano, ~Rn))0) 8 (5.11.3) and the second term becomes (By -Ro)?0:)o= j aim | dn, ~ Ry (V ~m)b2-+ (Ry — Ry? ° nb?})o. (5.11.4) Hence (R°) = Nb?+ [ dm dn(0(R, — Rn )[ Gm — 28? — (Rn — Ra'Do N m = Nb? + | dm [ dn { dri(r)((m —n)b?—P)Ge(r, m—n) (5.1115) where Golr, 2) = (3/2nnb2)22 exp(- (5.11.6) an)184 MANY CHAIN SYSTEMS Since the integrand of eqn (5.III.5) depends only m —n, this is rewritten as N (R®) = Nb? + [ dn(N—n) | dri(r)(nb?— F)Gy(r, n). (5.111.7) 0 By use of the Fourier transform Gor, n)(nb? — P= lan yo and eqn (5.53), eqn (5.III.7) is written as dk b*h? , OR \ a, aL exp(- Z nje (5.1118) (ee (5.11.9) Since the underlined part decreases quickly with n, the integral over 7 is evaluated for large N as an (e)=Nnot+u a] N °° fanqv-n)...=Nfan.... (5.11.10) 0 0 Hence, we finally get 2 2 dk 267k? / 6 \> (Rt) = Nb [ela > (es) aE ool = Nor[1+ (5.11.11) References 1. de Gennes, P. G., Scaling Concepts in Polymer Physics, Chapter 3. Cornell Univ. Press, Ithaca, New York (1979). 2. Flory, P., J. Chem. Phys. 10, 51 (1942); see also Principles of Polymer Chemistry, Chapter 12. Cornell Univ. Press, Ithaca, New York (1953). . Huggins, M. L., J. Phys. Chem. 46, 151 (1942); J. Am. Chem. Soc. 64, 1712 (1942). . A review on polymer collapse is given by Williams, C., Brochard, F., and Frisch, H. L., Ann. Rev. Phys. Chem. 32, 433 (1981). . Daoud, M., and Jannink G., J. Phys. (Paris) 37, 973 (1976). fen” D. W., Joanny J. F., and Pincus, P., Macromolecules 13, 1280 1980). . Moore, M. A., J. Phys. (Paris) 38, 265 (1977). XN AW F w10. 11. 13. BS RRSB RB ss 38, 41. 42. 43. REFERENCES 185 . Schafer, L., Macromolecules 47, 1357 (1984). . Edwards, S. F., Proc. Phys. Soc. London 88, 265 (1966). Edwards, S. F., and Muthukumar, M., Macromolecules 17, 586 (1984). Leibier, L., Macromolecules 13, 1602 (1980). . Okano, K., Wada, E., Kurita, K., Hiramatsu, H., and Fukuro, H., J. Physique Lett. 40, L171 (1979). Okano, K., Wada, E., Kurita, K., and Fukuro, H., J. Appl. Cryst. 11, 507 (1978). . Muthukumar, M., and Edwards, S. F., J. Chem. Phys. 76, 2720 (1982). . Des Cloizeaux J., and Jannink G., Les Polyméres en Solution: Modelisation et leur structures, Chapter 13, §2.5.5. Edition de Physique, Paris (1986 or 7). . Edwards, S. F., J. Phys. A8, 1670 (1975); Edwards, S. F., and Jeffers, E. F., J. C. S. Faraday Trans. 11 75, 1020 (1979). . Flory, P., J. Chem. Phys. 17, 303 (1949); see also Principles of Polymer Chemistry, Chapter 12. Cornell Univ. Press, Ithaca, New York (1953). . Richards, R. W., Maconnachie, A., and Allen, G., Polymer 19, 266 (1978). . Oono, Y., Adv. Chem. Phys. 61, 301 (1985). Freed, K. F., Accounts Chem. Res., to be published. . Daoud, M., Cotton, J. P., Farnoux, B., Jannink, G., Sarma, G., Benoit, H., Duplessix, R., Picot, C., and de Gennes, P. G., Macromolecules 8, 804 (1975). Noda, I., Kato, N., Kitano, T., and Ngasawa, M., Macromolecules 14, 668 (1981). des Cloizeaux J., and Noda, I., Macromolecules 15, 1505 (1982). Schafer, L., Macromolecules 15, 652 (1982). Ohta, T., and Oono, Y., Phys. Lett. 89A, 460 (1982). . Wiltzius, P., Haller, H. R., Cannell, D. S., and Schaefer, D. W., Phys. Rev. Lett, 51, 1183 (1983). Nakanishi, A., and Ohta., T., J. Phys. A18, 127 (1985). - Vologodskii, A., Lukashin, A., Frank-Kamenetskii, M. F., and Anshelevich, V., Sou. Phys. JETP 39, 1059 (1975); 40, 932 (1975). Roovers, J., and Toporowski, P. M., Macromolecules 16, 843 (1983). (1980) J. D., Viscoelastic Properties of Polymers (3rd edn). Wiley, New York 1980). . Graessley, W. W., Adv. Polym. Sci. 16, 1 (1974). Baumgartner, A., Polymer 22, 1308 (1981). . Edwards, S. F., J. Phys. Al, 15 (1968). }. Alexander, J. W., Trans. Am. Math. Soc. 30, 275 (1928). . Ball, R., and Mehta, M. L., J. Physique. 42, 1193 (1981). . Brereton, M. G., and Shah, S., J. Phys. A14, L51 (1981). . des Cloizeaux, J., J. Phys. (Paris) 42, LA33 (1981). Iwata, K., and Kimura, T., J. Chem. Phys. 74, 2039 (1981); Iwata, K., J. Chem. Phys. 78, 2778 (1983); Macromolecules 18, 115 (1985). . Tanaka, F., Prog. Theor. Phys. 68, 148, 164 (1982). . Edwards, S. F., Proc. Phys. Soc. 92, 9 (1967). Iwata, K., J. Chem. Phys. 76, 6363, 6375 (1982). Edwards, S. F., Proc. Phys. Soc. 91, 513 (1967). Prager, S., and Frisch, H. L., J. Chem. Phys. 46, 1475 (1967). . Saito, N., and Chen, Y., J. Chem. Phys. 59, 3701 (1973).186 45. 47. 49. SASH BAS s a) 70. Th. MANY CHAIN SYSTEMS Ito, K., and McKean, H. P., Diffusion Processes and their Simple Paths. Springer, Berlin (1965). . Edwards, S. F., Proc. R. Soc. Lond. A385, 267 (1982). de Gennes, P. G., J. Chem. Phys. 55, 572 (1971). - Doi, M., and Edwards, S. F., J. C. $. Faraday Trans. II 74, pp. 1789, 1802, and 1818 (1978). Kawasaki, K., Annals Phys. 61, 1 (1970); Kawasaki, K., and Gunton, J. D., Critical dynamics, in Progress in Liquid Physics (ed. C. Croxton). Wiley, New York (1978). . Pincus, P., J. Chem. Phys. 75, 1996 (1981). . Brochard, F., and de Gennes, P. G., Macromolecules 10, 1157 (1977); Brochard, F., J. Physique 44, 39 (1983). . Richter, D., Hayter, J. B., Mezei, F., and Ewen, B., Phys. Rev. Lett. 41, 1484 (1978). . de Gennes, P. G., Macromolecules 9, 587, 594 (1976). . King, T. A., Knox, A., and McAdam, J. D. G., Polymer 14, 293 (1973). . Ford, N. C., Karasz, F. E., and Owen, J. E. M., Discuss. Faraday Soc. 49, 228 (1970). . Adam, M., and Delsanti, M., Macromolecules 10, 1229 (1977). . Nose, T., and Chu, B., Macromolecules 12, pp. 590, 599, and 1122 (1979); 13, 122 (1980); Nemoto, N., Makita, Y., Tsunashima, Y., and Kurata, M., Macromolecules 17, 2629 (1984). . Schaefer, D. W., and Han, C. C., in Dynamic Light Scattering (ed. R. Pecora). Plenum Press, New York (1985). Amis, E. J., and Han, C. C., Polymer 23, 1403 (1982); see also Amis, E. J., Janmey, P. A., Ferry, J. D., and Yu, H., Macromolecules 16, 441 (1983). Eisele, M., and Burchard, W., Macromolecules, 17, 1636 (1984). Brown, W., Macromolecules 17, 66 (1984). Tsvetkov, V. N., in Newer Methods of Polymer Characterization (ed. B. Ke). Interscience, New York (1964). . Janeschitz-Kriegl, H., Adv. Polym. Sci. 6, 170 (1969). Léger, L., Hervet, H., and Rondelez, F., Macromolecules 14, 1732 (1981). . Debye, P., J. Chem. Phys. 14, 636 (1946); see also ref. 75. . Onogi, S., Kimura, $., Kato, T., Masuda, T., and Miyanaga, N., J. Polym. Sci. C 381 (1966). Onogi, S., Masuda, T., Miyanaga, N., and Kimura, S., J. Polym. Sci. A2 5, 899 (1967). . See for example, Dreval, V. E., Malkin, A. Y., and Botvinnik, G. O., J. Polym. Sci. Phys. ed. 11, 1055 (1973). . Edwards, S. F., Faraday Symp. Chem. Soc. 18, 145 (1983). Edwards, S. F., and Freed, K. F., J. Chem. Phys. 61, 1189 (1974), Freed, K. F., and Edwards, S. F., J. Chem. Phys. 61, 3626 (1974); ibid 62, 4032 (1975). Freed, K. F., Polymer dynamics and the hydrodynamics of polymer solutions, in Progress in Liquid Physics (ed C. Croxton). Wiley, New York (1978). . Muthukumar, M., and Freed, K. F., Macromolecules 10, 899 (1977); 11, 843 (1978). . Perico, A., and Freed, K. F., J. Chem. Phys. 81, 1466, 1475 (1984).REFERENCES 187 74. A discussion for the solution is given by Yoshimura, T., J. Math. Phys. 24, 2056 (1983). 75. Frisch, H. L., and Simha, R., The viscosity of colloidal suspensions and macromolecular solutions, in Rheology (ed. F. R. Eirich) Vol. 1, Chapter 14. Academic Press, New York (1956); see also Bohdanecky, M., and Kovar, J. Viscosity of Polymer Solutions, Chapter 3. Elsevier, Amsterdam (1982). 76. Martel, C. J. T., Lodge, T. P., Dibbs, M. G., Stokich, T. M., Sammler, R. L., Carriere C. J., and Schrag, J. L., Faraday Symp. Chem. Soc. 18, 173 (1983).6 DYNAMICS OF A POLYMER IN A FIXED NETWORK 6.1 Tube model 6.1.1 Tube model in crosslinked systems As mentioned in Section 5.4.3, the highly entangled state of polymers can be effectively described by the tube model. The idea of the tube model originated in studying the problem of rubber elasticity."? A rubber is a huge molecular network which is formed when a polymeric liquid is crosslinked by chemical bonds.? An important problem in the theory of rubber elasticity is to calculate the entropy, which is essentially the number of allowed conformations of the chains constituting the rubber. The topological constraints play an important role in such a problem. Consider a lightly crosslinked rubber which consists of long strands of polymers between crosslinks. A strand in such a rubber is schematically shown in Fig. 6.1. In Fig. 6.16 the strand is placed on a plane and the cross-sections of other strands are shown by dots. Due to the topological constraints, the strand cannot cross the dots, so that the number of conformations allowed for the strand is much less than that in free space. How can we estimate it? Suppose for a moment that the other chains are frozen, then the dots can be regarded as fixed obstacles. One can see that the allowed conformation of the strand is almost confined in a tube-like region shown by the dotted lines: the conformations which go outside the tube are likely to violate the topological constraints. The axis of the tube can be defined as the shortest path connecting the two ends of the strand with the same topology as the strand itself relative to the obstacles. Such a path represents a group of conformations which are accessible to each other without violating the topological constraints imposed by the other chains, and is called the primitive path.” If the topological constraints are replaced by the tube, the number w of the allowed conformations can be calculated easily by the method described in Chapter 2 (see Section 6.4). In real rubbers, the situation is more complicated since the other strands are mobile. However, even in such a case, a self-consistent picture will be that the range in which each part of the strand can move around will remain finite. The range is perhaps larger than the mean separation between the frozen strands discussed above. What diameter one should assign to the tube is a question which has not been answered with absolute certainty. However, as long as the strand is long enough, the diameter is determined by local conditions, and will be independentTUBE MODEL 189 © Fig. 6.1. (a) A strand in a rubber. A and B denote the crosslinks. (b) Schematic picture of (a): the strand under consideration is placed on a plane and the other strands intersecting the plane are shown by dots. (c) The tube model. of the length of the strand. Detailed discussions'** on rubber elasticity based on this idea and its appraisals”® are given in the literature. An important point here is the proposition that the tube concept will be a self-consistent picture in the system of topologically interacting system. 6.1.2 Tube model in uncrosslinked systems The idea of the tube is intuitively appealing, and one can imagine that the same picture will be useful for uncrosslinked system such as polymer melts. However, one has to face a new problem that, in melts, the tube itself changes with time because all conformations in a melt are accessible. A key concept to solve this problem was introduced by de Gennes’ who discussed the Brownian motion of an unattached chain190 DYNAMICS OF A POLYMER IN A FIXED NETWORK primitive chain {b) Fig. 6.2. (a) A chain in a fixed network of obstacles (denoted by dots). (b) The defects and the primitive chain (dashed line). moving through a fixed network (see Fig. 6.2). His idea was that in the situation shown in Fig. 6.2, the motion of the chain is almost confined in a tube-like region denoted by dotted lines in Fig. 6.2b. Since the chain is rather longer than the tube, the slack will constitute a series of ‘defects’ which can flow up and down the tube (Fig. 6.2b). De Gennes visualized this as a gas of non-interacting defects running along like the arch in a caterpillar (see Fig. 6.3a). As a result of such motion, the tube itself changes with time (Fig. 6.3b): for example if the chain moves right, the part B)B can choose a random direction, and create a new part of the tube which will be a constraint for the rest of the chain, while the part of the previous tube AgA becomes empty and disappears. This type of motion was called reptation by de Gennes after the Latin reptare, to creep. Our current understanding of the dynamics of the highly entangled state is based on the concept of reptation. This picture is rigorously correct for the system that has been presented, i.e., a single chain in aREPTATION 191 SN —_N\___ NL f@ () Fig. 6.3. (a) Motion of a single defect (b) Motion of the tube. fixed network. Whether the picture does indeed hold for concentrated polymer solutions or melts still remains a matter for debate, but many experimental results suggest that reptation is the dominant mechanism for the dynamics of a chain in the highly entangled state. Leaving the detailed discussion of this problem to the next chapter, we shall first consider the simple situation of a chain moving in a fixed network. 6.2 Reptation 6.2.1 Primitive chain Let us consider a polymer moving in a fixed network of obstacles. For the convenience of later discussion, we shall specify the problem in slightly192 DYNAMICS OF A POLYMER IN A FIXED NETWORK different terms from those used by de Gennes. We assume that the intrinsic properties of the polymer are represented by the Rouse model consisting of N segments with bond length b and friction constant €. The obstacles are assumed to be thin lines, so they have no effect on static properties, but have a serious effect on dynamical properties by imposing topological constraints. The characteristic feature of the dynamics can be visualized by the tube model. For a given conformation of the polymer, we can draw the primitive path, i.e., the shortest path connecting the two ends of the chain with the same topology as the chain itself relative to the obstacles (see the dashed line in Fig. 6.2b). In the short time-scale the motion of the polymer is regarded as wriggling around the primitive path. On a longer time-scale, the conformation of the primitive path changes as the polymer moves, creating and destroying the ends of the primitive path. Even though such a picture is clear, the mathematical treatment of the problem is still complicated since the time evolution of the primitive path is governed by the wriggling motion of the polymer, and the wriggling motion itself is limited by the primitive path. However, if we are interested in the large-scale motion of long chains, we may disregard the small-scale fluctuations, and discuss only the time evolution of the primitive path. Since the primitive path at any moment represents the conformation of the chain with the small-scale fluctuations omitted, we shall use the term ‘primitive chain’ to denote the dynamical equivalent of the primitive path. At this level of description, the details of the wriggling motion are irrelevant, and we can start with a simpler model. To denote a point on the primitive chain, we use the contour length s measured from the chain end and call this the primitive chain segment s. If R(s, £) is its position at time ¢, the vector u(s, t)= ZR, t) (6.1) is the unit vector tangent to the primitive chain. The dynamics of the primitive chain is characterized by the following assumptions. (i) The primitive chain has constant contour length L. (ii) The primitive chain can move back and forth only along itself with a certain diffusion constant D.. (iii) The correlation of the tangent vectors u(s, t) and u(s’,t) de- creases quickly with |s — s’|. The first assumption corresponds to neglecting the fluctuations of the contour length. The second states that the motion of the primitive chain is reptation. The third guarantees that the conformation of the primitiveREPTATION 193 chain becomes Gaussian} on a large length-scale. This assumption introduces a new parameter into the problem. Since the mean square distance between two points on the Gaussian chain is proportional to |s —s'|, it is written as ((R(s, 1) — R(s', DY) =als—s'| for |s—s’|>>a. (6.2) The length a is called the step length of the primitive chain. The primitive chain is thus characterized by three parameters L, D, and a, which must be expressed by the Rouse model parameters N, b, € and the parameters characterizing the network. The parameter D, can be identified as the diffusion coefficient of the Rouse model ke NE because the motion of the primitive chain corresponds to the overall translation of the Rouse chain along the tube. The length L is expressed by a since the mean square end-to-end vector of the primitive chain, which is La according to eqn (6.2), must be the same as that of the Rouse chain Nb?. Thus D,= (6.3) Nb? =. (6.4) We are left with a single parameter a, which depends on the statistical nature of the network. Though precise calculation of this parameter is difficult, it is obvious that @ is of the order of the mesh size of the network and much less than L. This knowledge is enough for the purpose of the present discussion. 6.2.2 Simple application We now study the dynamics of the primitive chain and show that certain time correlation functions can be calculated by a straightforward method.’ For example, consider the time correlation function of the end-to-end vector P(t) = R(L, t)— R(0, t). Figure 6.4 explains the prin- ciple of calculating this correlation function. At t = 0, the chain is trapped in a certain tube. As time passes, the primitive chain reptates and at a certain later time (Fig. 6.4d), the part of the chain CD remains in the original tube while the parts AC and DB are in a new tube. To calculate t It must be remembered that despite the Gaussian behaviour with a large length-scale, the Primitive chain cannot be modelled bya continuous Gaussian chain since the contour length of the continuous Gaussian chain is infinite and has no physical significance, while the contour length of the primitive chain has a definite physical significance and appears in various dynamical results.194 DYNAMICS OF A POLYMER IN A FIXED NETWORK @) a) OY © < Fig. 6.4. Four successive situations of a reptating chain. (a) The initial conforma- tion of the primitive chain and the tube which we call the original tube. (b) and (c) As the chain moves right or left, some parts of the chain leave the original tube, The parts of the original tube which have become empty of the chain disappear (dotted line). (d) The conformation at a later time t. The tube segment vanishes when it is reached by either of the chain ends: €.g., the tube segment P and Q vanish at the instance (6) when &(t) =sp and at (c) when &(t)=sg—L, Tespectively. (P(t) P(0)), we express P(t) and P(0) as P(0) =A,C + CD + DB, (6.5) P(t)=AC+CD +DB (6.6) Since the vectors AC and DB are uncorrelated with P(0), (P(t) - P(0)) will have the form (PQ) PO) = (CD*) = a(o(t)) (6.7) where o(t) is the contour length of CD, i.e., the part in the original tube. To calculate (o(t)) we focus attention on a certain segment s of the original tube. This tube segment disappears when it is reached by either end of the primitive chain. Let p(s, ¢) be the probability that this tubeREPTATION 195 segment remains at time ¢. The average (o(t)) is calculated as L (o(9) = fasy6s, 0. (68) 0 Let ¥(E,t;s) be the probability that the primitive chain moves the distance & while its ends have not reached the segment s of the original tube. The probability satisfies the one-dimensional diffusion equation ow ay a7 Diag (6.9) with the initial condition W(E, 0; 5) = 6(&). (6.10) When §=s, the tube segment s is reached by the end of the primitive chain and W(&, t;s) vanishes (see Fig. 6.4). Similarly when § =5 — L, the tube segment is reached by the other end and W(E, t;s) vanishes.} Thus WE, ts)=0 at E=s and E=s-L. (6.11) The solution of eqn (6.9) with these boundary conditions is WE, t5)= > 2 sin( PE 7) sin(2*™S— 5) as L PHS 9) exp(—pitit,) (6.12) where tg = L?/D,n. (6.13) For the tube segment s to remain, & can be anywhere between s — L and 5, so that 5 v,0= | awe, 45) sok => S sin(P *)exp(—p't!t4). (6.14) podd PIE Thus from eqns (6.7), (6.8), and (6.14) (P(t): P(O)) = Lay(t) = Nb?y(0) (6.15) where 1f 8 , vOrT | dsy(s, t)= ae pe OPP tI). (6.16) + Strictly speaking this argument is valid in the limit of a—>0. If a is finite, the boundary condition is not written in a simple form, but the correction is of the order of a/L.196 DYNAMICS OF A POLYMER IN A FIXED NETWORK The longest relaxation time of (P(#) + P(0)) is given by ty. This is called the reptation or disengagement time, since it is the time needed for the primitive chain to disengage from the tube it was confined to at t= 0. Equation (6.15) can be compared with eqn (4.35) for the Rouse chain without constraints: (P(t) - P(0)) =e? Soa exp(—p"t/T,r) (6.17) where Tz is the Rouse relaxation time, TR= ce (6.18) On the other hand, eqn (6.13) is rewritten by eqns (6.3) and (6.4) as 1 EN*o" 6.19) w=sa lS: 4° kpTa’ Note that t, is proportional to N® and becomes much larger than tp for large N. This demonstrates the crucial effect of topological constraints on the conformational change of polymers. Let us define the number of steps in a primitive chain by Z=7=7. (6.20) Then the ratio between t, and tp is written as Tal Tr = 3Z. (6.21) Equation (6.19) has been confirmed by computer simulation.” The function y(s, t) will appear frequently in the subsequent discus- sions. This function has been defined as the probability that the original tube segment s remains at time ¢. As will be shown in Section 6.3.3, (s, t) also represents the probability that the primitive chain segment s is in the original tube at time ¢. (Note the distinction between the tube segment and the primitive chain segment; the former is fixed in space, while the latter moves with the primitive chain.) The behaviour of p(s, t) is shown in Fig. 6.5. The tube segments in the middle (s = L/2) have long lifetimes of order tz, while the tube segments near the chain ends have very short lifetimes: the end segment is almost instantaneously replaced. This fact will be used in the subsequent discussions. The function w(t) represents the average fraction of the original tube that remains at time 4. This function is also equal to the average fraction of the primitive chain contour that remains in the original tube.REPTATION DYNAMICS 197 0.01 w(st) ae 0 tL s Fig. 6.5. The probability y(s, t) that the tube segment s is remaining at time t. This is also equal to the probability that the primitive chain segment s remains in the original tube at time ¢. 6.3 Reptation dynamics 6.3.1 Stochastic equation for reptation dynamics Although the above probabilistic description is quite useful in under- standing the essence of reptation dynamics, it becomes progressively more difficult to proceed with the calculation for other types of time correlation function. For example, it is not easy to calculate the mean square displacement of a primitive chain segment ((R(s, t) — R(s, 0))?) by this method. In this section we shall describe a convenient method’! for calculating general time correlation functions. First we derive a simple mathematical equation for reptation dynamics. Let A&(t) be the distance that the primitive chain moves in a time interval between ¢ and ¢ + At, then R(s,t+ At) =R(s + A&(2), 0), (6.22) which states simply that if the primitive chain moves the distance A&(#) along itself, the segment s comes to the point where the segment s + AE was at time ¢ (see Fig. 6.6). The variable A&(¢) takes random values, the198 DYNAMICS OF A POLYMER IN A FIXED NETWORK [Ree “qa Ag(t) Rieteaty—/ Fig. 6.6. When the primitive chain moves a distance A&(#) along itself, the segment s comes to the point where the segment s + A was at time t. distribution of which is Gaussian characterized by the moments (A&())=0, — (AE(t)?) = 2D. At, (6.23) ie., 2 W(AE) = (42D. At)"exp ( - i) . (6.24) Equation (6.22) is not correct for all s: if s + A&(t) is not between 0 and L, R(s, t+ At) should be on the newly created part of the tube (see Fig. 6.6). Writing down this condition in a formal mathematical equation is cumbersome. However, such an expression is not needed in practice since the condition can usually be accounted for by the fact that the distribution of the tangent vectors at the chain ends u(0, ¢) and u(L, t) are independent of the previous conformation of the primitive chain since their correlation time is infinitesimal.} 6.3.2 Segmental motion To see how eqn (6.22) works, we calculate the mean square displacement of a primitive chain segment s((R(s, t) — R(s, 0))*). In this case it is more convenient to calculate the following time correlation function, $(s, 8°58) = (ROS, 1) - RG’, OY). (6.25) The time evolution equation for (s, 5’; ¢) is written as o(s, s'; t+ At) = ([R(s + AE(1), 2) -— R(s', OP) = (p(s + A(t), 55 1)). (6.26) The bracket in the last expression represents the average over A&(t). To tit must be remembered, however, that the dynamics of the chain end is extremely important in reptation dynamics since eqn (6.22) only describes the one-dimensional motion, and the three-dimensional property of the primitive chain is entirely governed by the dynamics of the chain ends.REPTATION DYNAMICS 199 calculate this average, we expand the right-hand side of eqn (6.26) and use eqn (6.23): hy = a AR ' ) (o(s + A&,5'32)) =((1 +ags+S SF Sa)ol,s a) = (1408) 2 + FOZ )965, 59 = (140.4 3) 9(6,5°s d. (6.27) From eqns (6.26) and (6.27) we have 3 # 3p POS 85) = De za O(S, 8°31). (6.28) To solve this, we need the initial condition and the boundary condition: (i) The initial condition is given by eqn (6.2), $(s, 53 Dleo= [3 —s’] a. (6.29) (ii) To obtain the boundary condition, we write 3/ds at s = L as Zs] =2uL,0- (RE) -RO', 0) as enk =2(u(L, t) - (R(L, t) — R(s", t))) + 2(u(L, t) + (R(s", t)— R(s', 0))) (6.30) where s” is an arbitrary value betwen 0 and L. Since the correlation time of u(L, t) is infinitesimal, the underlined part gives (u(L, t) - (R(s", t) — R(s', 0))) = (u(Z, 8) - ((RG", 1) — RG’, 0))) = 0. (6.31) The first term in eqn (6.30) gives 2(u(L, (RU, )- Rs", D)) = 2 (RG, - RE") s= a Ww = 3,408 —s") on a. (6.32) Hence 8 una = 3s o(s,s';t)=a at s=L. (6.33)200 DYNAMICS OF A POLYMER IN A FIXED NETWORK Similarly a 35 p(s, s';Q=-a at s=0. (6.34) The solution of eqn (6.28) with the boundary conditions (6.33) and (6.34), and the initial condition (6.29) is $(s, 53 t) =|s—s'|a +25 Dit Se 4a p=1P (pms x cos( 2H PT cos PA) — exp(—ip*/1.) (6.35) Hence $66.83 yazne+ S seo EE 4) —exp(—w2l2)}. (636) The expression becomes simple in two extreme cases: (i) For t«t,, $(s,5:t) is dominated by the terms with large p. Replacing cos*(pxs/L) by the average 1/2, and converting the sum into the integral, we have 1. 4La (8,830) [dp S24 exp(-w'lea) 0 = 2a(D,t/n)", (6.37) This result is easily derived, for if t 7,, the first term in eqn (6.36) dominates, so that G(s, 53t) = 2D,t/Z. (6.39) Hence the diffusion constant Dg of the centre of mass is given as =n POO) _ keTa? De lim a D,/3Z = aN7EbE" (6.40) Thus the self-diffusion constant is proportional to N~*, which has also been confirmed by computer experiment.”REPTATION DYNAMICS 201 6.3.3 Correlation function of the tangent vectors Next let us consider the time correlation function of the tangent vector u(s, t): G(s, s'; t) = (u(s, t) + u(s’, 0)) (6.41) which can be related to $(s, 5‘; £) by eqns (6.1) and (6.25): cone A a 1 # = 2a ((R(s, t) ~ R(s", 0))?) 1 = 355 3s" p(s, s';t). (6.42) Substituting eqn (6.35), one gets G(s, st) =ad(s —s')- 2 Pisin PE sin PE) ~ exp(—tp’/tz)] — 3 24., (pas (pms 2 = L = sin (22 Pe sin( P= “)exp(- (p"!t4). (6.43) An important conclusion is derived from eqn (6.43) by a geometrical interpretation of (u(s, t)- u(s’, 0)). First note that if u(s) and u(s') are the tangent vectors of a Gaussian chain of step length a, the integral 4 [=> | ds(u(s)-u(s’)) — (61>52) (6.44) is equal to 1 when s’ is between s, and s2. Now consider L V6", =3 | ds(a(s, 1) mls’, 0). (6.48) 0 This is equal to unity at time =0. As time passes, the original tube segment s’ becomes empty of the primitive chain. If this happens u(s', 0) becomes totally uncorrelated to u(s, t) and the contribution from such case to p(s’, t) is zero. Therefore ~(s’, t) represents the probability that the original tube segment s’ is remaining at time ¢. Indeed from eqns (6.43) and (6.45), it follows that vO", = SS sin( PE esp(—p'ur,) (6.46) which agrees with eqn 6.14),202 DYNAMICS OF A POLYMER IN A FIXED NETWORK Now, one can show by a similar argument that the integral L ~ 1 HG, )=7 [ ds"(u(s,£)-(s', 0)) (6.47) 0 Tepresents the probability that the primitive chain segment s is in the original tube. From eqns (6.43), (6.45), and (6.47) it follows that HG, )= Ys.) (6.48) i.e., the probability that the original tube segment s remains is equal to the probability that the primitive chain segment s is in the original tube that remains. 6.3.4 Dynamic structure factor As the final example, we shall consider the dynamic structure factor of a single chain: L L ab =X fas dor(explik-(R(.)-RG',091). 649) To calculate this quantity, we again consider $(s, s‘; t) defined by $(s, 53 t) = (explik - (R(s, 2) — R(s’, 0))]). (6.50) By the same trick as that used in Section 6.3.2, we can show that £96, 850=D F960. (6.51) Since the distribution of R(s, 0) — R(s’, 0) is Gaussian with the variance a|s—s'|, the initial condition for $(s,s';t) is obtained as (see eqn (2.79) 2 os, $s Dlno=exp(— a \s -s'\) (6.52) The boundary condition for $(s,s';t) is again obtained by the same method as in eqn (6.30): 206 8'30|_ sik (w(L, fexplik - (RL, 2)~ RG’, 0))]) =ik- (u(L, Dexplik- (RL) - RG", 8)] x explik - (R(s", t) — R(s", 0))1) (6.53) where s” is again an arbitrary value between 0 and L. If s” is taken to beREPTATION DYNAMICS 203 close to L, the average of the underlined part can be taken separately from the rest since its correlation time is very short. Hence ZO6.5's0] ike (w(L, Dexplik RL, RE", NY) x (explik -(R(s", RG", O))) = 5; (empl R(6,)—RG" OM] 6" 5'30) = — 2a exp[ - 3k’a(L—- s")]9(6", 532). (6.54) In the limit of s"— L, this gives a o(s, 8's o| = —Wag(L, 5’). (6.55) S s=L Similarly 246s 85 o| =2ka9(0, 5‘; t). (6.56) S s=0 The solution of eqn (6.51) under these conditions is 0950-3 [eral t (°-3)] cool (0 ~5) oP) + sy (cine cl e(—E) lee E) 22] cs n= x La= £ Nb 4(KR,)* (6.58) where and @, and §, are the positive solutions of the equations a, tan a, =p, B, cot Bp = —p. (6.59) From eqn (6.57) we get” 2uN 2 ( “Duse) =. ETF - . (6. g(k, th= 3S at ab ep pia ‘a, exp| ie (6.60) The expressions are simplified for the two limiting cases. @) u<1, ie., KRg<«K1. Here eqn (6.60) is dominated by the term204 DYNAMICS OF A POLYMER IN A FIXED NETWORK including a, = Vu <1, thus a(k; t) = N exp(—4D,tu/L?) = N exp( - Peat R)= N exp(—Dok’t) (6.61) which is natural since, for KR, <1, the dynamical structure factor is governed by the diffusion of the centre of mass. (ii) w>>1, kR,>>1. In this case, the @, are approximated as (p — 1/2), and eqn (6.60) reduces to 96N ak, th= Pala aL 2p J exp(- tp?!) = mi vl). (6.62) This result is also derived by a simple argument. In the limit AR, >> 1, the average of L 7] &exolik- RG", ~ RG, 0) is equal to 12/k?b? if R(s', t) remains in the original tube, and zero if it has already disengaged from it. Thus using the probability (s, t) that it remains, we have L ds 12 12 8(k, t)= | Lp VS = pape VO- (6.63) 0 Notice that the decay rate of g(k, t) is 1/Tz and depends strongly on the molecular weight. This behaviour is entirely different from the result of the Rouse dynamics, according to which g(k, t) becomes independent of the chain length for kR, >>1 (see eqn (4.III.12)). 6.3.5 General time correlation functions Time correlation functions of more complicated form can be calculated in a similar way. For example, the time correlation function C(s, s', 5"; t) = (RG, t) ~ R(s", 0) - (RGs', 1) — R(s", 0))) (6.64) can be calculated from a 2 Zc= p(2+ +a) (6.65) under appropriate boundary conditions. In general the time correlationCONTOUR LENGTH FLUCTUATION 205 function C(s,,52,...55m,t) which depends on the m-points on the primitive chain is obtained from a m™ 9\? Sc= v{ 55) Cc (6.66) with appropriate boundary conditions. 6.4 Contour length fluctuation 6.4.1 Statistical distribution of the contour length In the previous sections we regarded the primitive chain as an inexten- sible string of contour length L. In reality, the contour length of the primitive chain fluctuates with time, and the fluctuation sometimes plays an important role in various dynamical processes. First we consider the statistical distribution of the contour length. Since the primitive chain represents a set of conformations of the Rouse chain, the probability that a certain conformation of the primitive chain is realized is proportional to w, the number of the conformations of the Rouse chain which are represented by that primitive chain. The simplest hypothesis is to take the polymer as a random walk confined within a tube. Then @ is calculated by the method described in Section 2.3.2 (see Appendix 6.1). The result is 2 2 3 Nb ) 6.61) w(L) = @ exp( — ine? oo where Wo is the number of the configuration in the free space, ao, the diameter of the tube, and a, a certain numerical factor which depends on the shape of the cross-section of the tube. The logarithm of w gives the entropy of the primitive chain 2. ast) 2Nb? °° ad At first sight eqn (6.68) may seem to imply a paradoxical result: since the entropy increases with decreasing L, the chain will contract to the state of L=0 if its ends are not fixed. This argument is wrong. The collapse happens if the chain is confined in an infinitely long tube of given conformation, but does not happen in the case of a network, where the Rouse chain explores many tubes. To calculate the statistical distribution of L in the network, one has to take into account the multiplicity of the State specified by L. Let Q(L) be the number of primitive paths which have length L, then the probability that a primitive chain has a contour S(L) = 5 ko (6.68)206 DYNAMICS OF A POLYMER IN A FIXED NETWORK length L is W(L) & w(L)Q(L). (6.69) Q(L) can be estimated by the number of random walks consisting of L/ao steps: Q(L) = exp(a *) (6.70) where «, is a certain numerical constant which depends on the structure of the network. From eqns (6.67), (6.69), and (6.70) W(L) x exp( - Bieta) « «exp| — a3 (L- DF (6.71) which has a maximum at - _ a, Nb? L= 3 ap : (6.72) From eqns (6.4) and (6.72) 3 a= a, Q. (6.73) This shows explicitly that the step length a of the primitive chain is of the same order as the tube diameter ao. The average of the fluctuation is calculated from eqn (6.71) as AL= (AL?) = [ j dL Y(L)(L— oy" =(Nb?/3)'? for L>>VNb. (6.74) This statistical property of the primitive chain has been studied by computer simulation.’® Under certain conditions, the statistical distribu- tion of the primitive chain can be calculated analytically.'7""* Both studies show the relations (L)*N and (AL?)«N (6.75) in agreement with eqns (6.72) and (6.74). 6.4.2 Dynamics of the contour length fluctuation Having studied the static distribution of the contour length, we now examine the dynamics of the contour length fluctuation. As before, we use the Rouse model for the dynamics (see Fig. 6.7). Let s, be the curvilinear coordinate of the n-th Rouse segment measured from aCONTOUR LENGTH FLUCTUATION 207 iN Sy Fig. 6.7. A Rouse model describing the contour length fluctuation. The point O denotes the origin of the curvilinear coordinate s,. certain fixed point on the tube. The contour length of the primitive chain is defined by L(t) = sy(t) — So(t)- (6.76) The dynamics of s, are described by the Langevin equation for the Rouse model, a 3kpT # C59 GE Byatt) that) (6.7) where (fit) =O and (f,()fm(t')) = 26keT5(n — m)d(t— 1"). (6.78) Since eqn (6.77) is the same linear equation as appeared in the Rouse dynamics, analysis of this equation is straightforward. An important difference, however, must be mentioned. In the present case, the equilibrium average of the contour length is L: (sv ~50) = £, (6.79) while in the Rouse model the corresponding quantity (Ry — Ro) is zero. To get eqn (6.79) we have to modify the boundary condition which corresponds to eqn (4.11). This is obtained as follows. If we take the average of both sides of eqn (6.77) for the equilibrium state, we have BS (sq) = AES (a0) + 600). (6.80) Both 3({s,,)/dt and (f,(t)) vanish at equilibrium, so that & (Sn) = 0. (6.81)208 DYNAMICS OF A POLYMER IN A FIXED NETWORK For this to be consistent with eqn (6.79), the boundary condition must be a L ay at n=OandN. (6.82) The boundary condition (6.82) can be visualized as a hypothetical tensile force acting on the chain ends (see Fig. 6.82): _3ksT , _ 3keT “a Nb? a The origin of this force is again the multiplicity of the tube. One can intuitively understand it by considering the dynamical process at the chain end. Suppose in a time At the chain end moves one step in a random direction (see Fig. 6.85). If it moves to the (z — 1) positions Ai, A2,...,Az-1, the contour length increases, whilst if it moves to Ao, the contour length decreases. Thus there is an imbalance in the change, which tends to increase the contour length and causes the force (6.83). For eqn (6.82) to be satisfied, the normal mode is now defined as (6.83) N 1 hay J dns, (6.84) 17 7 LD = pan\( nh = Y= anoos( Fi") 4) for p=1,2,..., (6.85) es e ° ° @ (b) Fig. 6.8. (2) Equilibrium tension acting on the primitive chain. (6) Physical origin of the force F,,. The case of z = 4 is shown.CONTOUR LENGTH FLUCTUATION 29 or = : pan) nb m= %+2 D Yoo oos( 2) + (6.86) The coordinate Yq represents the position of the ‘curvilinear center of mass’, N sol) =% [ anss()= (0, a) 0 while the other coordinates Y, (p >0) represents fluctuations along the tube. The equation of motion for Y, is precisely the same as eqn (4.19): 9Y, 62 ri =-k,Y, +h (6.88) with So=NE, = 2NE for p=1,2,...5 6.89) 67k, p= ge Ps (6.90) and ($A) =0, (M(t!) = 26pkT6p,5(t — t'). (6.91) Thus the time correlation functions for Y, are obtained as ((%) — %O)?) = 2at t (6.92) Nb? «¥5(0%(0)) = Zaz exP(—1P Vt), (6.93) where Tz is the Rouse relaxation time given by eqn (6.18). Equation (6.92) justifies the previous assumption that the diffusion coefficient D, is equal to kgT/N€, the diffusion constant of the Rouse chain. From eqn (6.86) the contour length of the primitive chain is written as LQ) = sy(t) ~ So(t) (6.94) =L-4 a Y,(t). (6.95) The time correlation function of L(t) is calculated from eqns (6.93) and (6.95). The result is 8Nb? 1 = = exp(—tp”/ tp). (6.96) (HOLO) = B+ Se So210 DYNAMICS OF A POLYMER IN A FIXED NETWORK In particular, 8Nb? 1 _ Nb? ==> 6.97 aa pgp? 3 0) which agrees with eqn (6.74). The characteristic time of the contour length fluctuation is tg, the Rouse relaxation time. Since teil Ab (2). ta 2 L NL vz’ the effect of the contour length fluctuation can be neglected for Z >> 1. It is at this limit that the dynamics described in Section 6.3 is valid. (AL?) = (12) == (6.98) 6.4.3 Effect of the contour length fluctuation on reptation Though the contour length fluctuation becomes negligible for very large Z, its effect is not negligible for usual values of Z, which are typically less than 100. An important effect is found in the disengagement time t,.'° Consider the two situations shown in Fig. 6.9, one represents the motion of the chain with fluctuations, and the other without. Obviously the life time of the tube becomes shorter for the chain with fluctuation than for the chain without, i.e., the contour length fluctuation reduces the disengagement time. This effect is estimated as follows. If the contour length fluctuation is neglected, the disengagement time is estimated as the time necessary for the chain to move the distance L, 7") = [/D, (6.99) Oe, CIEE, momma aeE LLL SOLE, time Fig. 6.9, The Brownian motion of a primitive chain with (a) fixed contour length, and (b) fluctuating contour length. The oblique lines denote the region that has not been reached by either end of the primitive chain. Obviously the length of this region o(t) decreases faster in (b) than in (a). Reproduced from ref. 15.CONTOUR LENGTH FLUCTUATION 211 (The superscript (NF) stands for ‘no fluctuation’.) On the other hand if there are fluctuations, the chain can disengage from the tube when it moves the distance L — AL because the chain ends are fluctuating rapidly over the distance AL. Hence the disengagement time is estimated as oP = (L- AL)/D,. (6.100) (The superscript (F) stands for ‘fluctuation’.) From eqns (6.98) and (6.100), it follows that TP) = (1 - xy (6.101) where X is a certain numerical constant. Precise calculation of + requires the first passage problem in multidimensional phase space. A variational calculation’® for the Rouse model shows that X is larger that 1.47. Hence the effect of the contour length fluctuation is significant even if Z is as large as 100. The effect of the contour length fluctuation has been studied for slightly different models'”"* and it has been reported that the effect is less significant than in the case of the Rouse model. The discrepancy is perhaps due to the difference in the dynamics of the fluctuation, especially the short-time dynamics, which is quite important in the first passage time problem.’?” 6.4.4 Other small-scale fluctuations and their effects on the segmental motion Due to the small-scale fluctuations around the primitive path, the actual dynamics of the Rouse segment in a network is much more complicated than that described in Section 6.3.2. As an example, let us consider the mean square displacement of a Rouse segment. alt) = ((Ralt) — Rn(0)))- (6.102) The precise calculation of ¢,,(t) is difficult, but its characteristic features can be inferred easily. (i) For a very short time the segment does not feel the constraints of the network, so that ¢,(t) is the same as that calculated for the Rouse model in free space. Hence a is given by eqn (4.III.6) as 4Nb? (pan alt) = 6D.1+ 22 S 5 c0s*( PZ) 1 — exp(—tp*/re)). (6.103) p=1P Since t<< tr, eqn (6.103) is approximated in the same way as in eqn212 DYNAMICS OF A POLYMER IN A FIXED NETWORK (6.37), on | dp H(1~ exp(-9°/=) = ONB*CI tg)!” = (kpTb7t/£)'. (6.104) This formula is correct when the average displacement is much less than the tube diameter. Let 1, be the time at which the segmental displace- ment becomes comparable to a: a? = y(t) = (kp Tb7/E)'Vr, (6.105) ie., t= aC lkpTb?. (6.106) The time 1, denotes the onset of the effect of tube constraints: for t< t., the chain behaves as a Rouse chain in free space, while for t>, the chain feels the constraints imposed by the tube, (ii) For t>+, the motion of the Rouse segment perpendicular to the primitive path is restricted, but the motion along the primitive path is free. The mean square displacement along the tube is calculated from eqns (6.86)—(6.93) as Cat) ss(O)?) = ((Hle)— YAO?) +4 3 cos "CHC — HOP) kpT sn? a a2 ae pees oos?(P= a "a — exp(-tp?/tp)) =f" St, 6.107 keTt/NE tt. (6.108) From this, $,(¢) is estimated as in Section 6.3.2. If the segment moves 5,(¢t)—s,(0) along the tube, the mean square displacement in the three-dimensional space is a |s,(¢) — s,(0)|. Hence n(t) = a s(t) — 5n(0)|) =a ((n(2) — sn)". (6.109) From eqns (6.107)—(6.109) a(kpTb7t/t)"* 4,StStp, nll) (eaenney” tp StS Ty. (6.110) Note that ¢,(¢) is proportional to ¢” in the time region 1, St tp. This specific diffusion behaviour, first predicted by de Gennes,’ is a conse- quence of the two effects, the Rouse-like diffusion equation (6.107), and the tube constraints equation (6.109). On the other hand, the behaviourCONTOUR LENGTH FLUCTUATION 213 log<(R(t)—Fa(0) }?> log t Fig. 6.10. Mean square displacement of a chain segment is plotted logarithmically against time. for tt, agrees with that predicted by the primitive chain dynamics equation (6.37). (iii) For t= 1,4, the dynamics is governed by the reptation process. As discussed in the previous section 2. kpyTa? ¢,(t) = ze = NAEp! tet (6.111) Equations (6.104), (6.110), and (6.111) are summarized as follows” Nb*(t/tp)? tst, Nb(t/Z?tz)"* 1 St St, PO ~4 np2e/24)! n*” On the other hand there is much experimental evidence which indicates that reptation is the dominant motion governing the dynamics in the highly entangled state. Clear evidence comes from the study of diffusion. Klein ef al.®° measured the diffusion constant of a deutrated polyethylene in a polyethylene matrix, and found Dg *M~? (7.1) in agreement with the result of reptation dynamics (eqn 6.40). The self-diffusion constant of a labelled polymer in solutions and melts has been measured by other techniques (forced Rayleigh scattering’® and field gradient Nuclear Magnetic Resonance,""”) and eqn (7.1) is fully confirmed in these systems. (A comprehensive review is given by Tirrell.>) More evidence comes from the study of viscoelasticity, which has been done extensively in the past and established the characteristic aspects common to all flexible polymers.”> The reptation model has succeeded in explaining many of these features and also predicting some of the behaviour in nonlinear viscoelasticity. In this chapter we shall describe the reptation theory'*> for viscoelasticity in detail, and discuss the validity of the reptation model in solutions and melts. The viscoelastic properties of polymers are quite important in polymer technology. A great deal of experimental work has been done, and at the Same time phenomenological theories have been developed to a highly sophisticated level. These are summarized in various monographs.”!*1® An overview of this field can be obtained in the excellent textbook by Bird et al.’ Here we shall limit ourselves to the molecular aspects of the Problem, i.e., how the viscoelastic properties are related to the molecular dynamics and how they depend on molecular parameters such as molecular weight, concentration, and molecular structure.220 MOLECULAR THEORY OF POLYMERIC LIQUIDS 7.2 Microscopic expression for the stress tensor 7.2.1 Stress in polymeric liquids To understand the mechanical properties of materials, it is important to consider first the microscopic origin of stress. In the usual gas or liquid of small molecules, the stress comes from the momentum transfer due to the intermolecular collision. In polymeric liquids, the stress is mainly due to the intramolecular force, and is directly related to the orientation of the bond vectors of the polymer. This idea, originated from the theory of rubber elasticity,” is fundamental in the physics of polymeric materials.+ To demonstrate the point, let us first consider a dilute polymer solution. As shown in Section 4.5.2, the stress tensor is written as Cap (t) = (Kap (t) + Kpa(t))ns + OF}(t) + POap (7.2) where 7, is the viscosity of the solvent, x, is the velocity gradient tensor and o® represents the contribution from the polymer = 2 aE (Hn Ree) oC) N 2 b? an an | (7-3) This is directly related to the orientation of the vector OR,,/dn, and it is this term that gives the viscoelasticity of dilute solutions. In the dilute solution, the viscoelasticity has only a small effect; the major contribution to the stress is the purely viscous stress given by the first term in eqn (7.2). The situation changes with increasing polymer concentration. The contribution of o®} increases steeply (because the polymers become more easily oriented due to the entanglement), while the contribution of the first term remains essentially unchanged. When of} dominates the first term, the total stress is simply related to the orientation of the bond vectors: cm EE OE), a Note that eqn (7.4) is valid even if the excluded volume effect is accounted for since, as shown in eqn (4.135), the pseudo-potential described by the delta function does not change the expression for the stress tensor (apart from the isotropic stress, i.e., the pressure). + Historically, the molecular theory for the mechanical properties of polymeric materials was first constructed for rubbers, and it was realized that the stress is related to the orientation of the polymer segments. That the stress in polymer melts has the same molecular origin as in rubbers was noted by Green and Tobolsky,”' who regarded the polymer melt as a kind of network with temporary junctions. This idea has been the base of successful phenomenological theories'®- and it is indeed used in the present theory.MICROSCOPIC EXPRESSION FOR THE STRESS TENSOR 221 The above argument is essentially for semidilute solutions, and may not apply to concentrated solutions and melts in which the intermolecular interaction cannot be described by the pseudo-potential. However, since the intermolecular interaction does not have any specific polymeric effect (for example the associated relaxation times will be as short as in a liquid monomer), one can argue that their effect will be only to give a constant viscous stress. Indeed eqn (7.4) is derived from the fact that even in the highly entangled state the short-time dynamics of polymers is described by the Rouse model. For, as was discussed in Sections 3.7.4 and 3.8.4, the expression of the stress tensor is entirely determined by the short-time dynamics: for example, the elastic stress is given by the change in the free energy for an instantaneous deformation. Therefore if the short-time dynamics is given by the Rouse model, the expression for the elastic stress must be the same as for the Rouse model, which is eqn (7.4). Thus eqn (7.4) holds quite generally in polymeric liquids in which the viscous stress is negligibly small.+ Since the short-time dynamics of polymers is unaffected by the global molecular structure, such as branching or crosslinks, the same argument will also hold for branched polymers and gels. In these systems, the stress is written as _ aT (Rael) OR p(t) onl) = all aoe b on on ) (7-5) in unit volume where R, is the position of the segments which are chosen small compared to the distance between the branching points (or crosslink points). 7.2.2 Stress optical law The above argument is crucially supported by the stress optical law,”** which is obtained by comparing eqn (7.4) with the formula for the intrinsic birefringence (eqn (4.238)) n, = Coze(t) (7.6) where 2n(n? + 2)? c 27kpTn 7.7) That the viscous stress of polymeric liquids is negligibly small compared to the elastic Stress is well established experimentally. Indeed this fact has been taken as a basic postulate in Coleman’s phenomenological theory” (see also Chapter 4, ref. 17). It must be noted that this is the result of the elastic stress contribution becoming so large relative to the viscous Stress: the absolute magnitude of the viscous stress will not differ considerably between simple liquids and polymeric liquids.222 MOLECULAR THEORY OF POLYMERIC LIQUIDS In concentrated solutions and melts, the intrinsic birefringence n{) can be regarded as the total birefringence since the form birefringence is negligibly small as was shown in Chapter 5. Thus eqn (7.6) is written as Nap = COng + isotropic tensor. (7.8) This is the stress optical law, which has been found experimentally to hold for rubbers and the more general problem of polymeric liquids.” Two important aspects of eqn (7.8) must be mentioned. (i) The simple linear relation between the stress and the birefringence holds even when the relation between the stress and the velocity gradient becomes quite complex (generally a nonlinear functional). Thus eqn (7.8) has much deeper physical significance than the similar relation for amorphous solids, which is derived by the symmetry argument and is limited to the Hookian range of elasticity. Gi) The proportional constant C is a function of the local condition such as the temperature, solvent, and polymer concentration but is quite independent of the features of molecular structure on a large length-scale such as the molecular weight, molecular weight distribution, branching, and degree of crosslinking.t These results are fully confirmed by experiment. The only exptanation for such a general relation is that both the stress and the birefringence have the same physical origin, i.e., the orientational ordering in the polymer segments. In deriving eqn (7.8) we assumed that (a) the relation between the orientation of the bond vectors and that of the end-to-end vector of the Rouse segments is linear (eqn (4.1V.13)) and that (b) the form birefringence is neglected. The stress optical law breaks down when either of these conditions is not met.”4 For example, the first condition is Not satisfied when the stress is very large, as often happens in experi- ments near the glass transition temperature. The second condition is not satisfied when the sample includes a spatial inhomogeneity as in the case of block copolymers or polymer blends which include microphase separation. Except for these situations, the stress optical Jaw holds quite generally in polymeric liquids. 7.3 Linear viscoelasticity 7.3.1 Background of phenomenological theory The viscoelastic properties of a material are entirely characterized by the constitutive equation which relates the stress tensor 0,¢(¢) to the velocity + Though the excluded volume interaction (and other Van der Waals’ type of interaction among the segments) do not violate the validity of eqns (7.4) and 7.8)" they do affect the stress optical coefficient C. For example C is sensitive to the nematic-like interaction which tends to orient the neighbouring segments in the same direction.”*?LINEAR VISCOELASTICITY 223 gradient tensor x,g(t). In usual fluids the relationship is simple, Oap(t) = (Kap(t) + Kpa(t))n- (7.9) In polymeric liquids the constitutive equation becomes quite complicated: the stress depends on the previous values of the velocity gradient tensor and the dependence is generally not linear; hence the relation is written only by a nonlinear functional relation: Oap(t) = Fg[Kap] for Kag(t') with t' <¢. (7.10) However, in the special case that the perturbation by the velocity gradient is small, the relation between the stress and the velocity gradient becomes linear and is written as””? Oup(t) = | dt'G(t — t’)(Kap(t’) + Kga(t'))- (7.11) This constitutive equation includes only one material function G(t), called the shear relaxation modulus. Though the applicability of eqn (7.11) is limited, linear viscoelasticity has been studied in great detail as it represents a property of the material in a well-defined limit. For shear flow v,=K(t)y, vy=0, vu, =0, (7.12) equation (7.11) reduces to the form given in Chapter 4 (eqn (4.116)), O,y(t) = { dt’G(t — t')k(t'). (7.13) Let y(t) be the shear strain measured from the state at t= 0: y= fareey (-@0, (7.17) for which, eqn (7.16) gives O,(t) = YoG(*). (7.18) This provides a direct determination of G(t). (b) Oscillatory shear (Fig. 7.1b). This has already been discussed in Chapter 4. The shear strain is give by Y(t) = Yo cos (wt) = yo Re(e). (7.19) The response defines the storage modulus G'(w), loss modulus G’(w), and the complex modulus G*(w), Ory(t) = yo(G" cos( wt) — G"(w)sin(wt)) = yo Re(G*(w)e'”) (7.20) where G*(@) = G'(w) + iG"(w). (7.21) It is easy to show that eqn (7.16) gives G*() =io | dte“"G(t) (7.22) 0LINEAR VISCOELASTICITY 225 and i = G'(o)=0f dt sin(wt)G(0) Grw)=0f dt cos(wt)G(). (7.23) (c) Creep (Fig. 7.1c). When a constant shear stress 0 is applied to the system at equilibrium, the system starts to flow. The shear strain y(t) is obtained from the integral equation: oo= j aro — 2) 2, (7.24) dt This can be solved by the Fourier—Laplace transform: -id+0 do el n= 00 |e act@ (7.25) where 6 is an arbitrary positive constant. The behaviour of y(t) for r—> © is governed by the contribution from the pole at w = 0. Since for small w, G*(@) is given by G*(@)= iw axc1 —iot+...)G(t)=iwgotw’git... (7.26) oO with B= [dG and g.= [aGey, (7.2) oO 0 the contribution from the pole at w = 0 is given by 81 = 0+ +8). 1.28 (0) = 00) ao ge (7.28) For large t, eqn (7.28) gives o)=gody/dt, whence gy can be identified with the steady state viscosity mo. The constant term g,/g2 is called the steady-state compliance and written as J{. Thus o= j 1G (0)= fin, FO, JO= | dG (t / [ j aio] =i Jim a oor (7.30) (7.29) 7.3.2 Calculation by Rouse model We now study the linear viscoelasticity from the molecular viewpoint. First we consider Rouse dynamics. This corresponds to the case of short226 MOLECULAR THEORY OF POLYMERIC LIQUIDS polymers in melts. The basic equations and their development are precisely the same as in Sections 4.5.3 and 4.5.4. The only distinction is that the viscous stress is now negligibly small, so that G(s) in eqn (4.158), which corresponds to the polymer contribution to the relaxation modulus in Chapter 4, is now regarded as the relaxation modulus of the system. Thus Gi)= cnet > exp(—2tp?/ tp) (7.31) pol where __EN%? ET (7.32) The viscosity and the steady-state compliance are calculated from G(¢t), _x = (1), cf ww) ta 5g NO? (7.33) no= [ AGU) = yKoT He 3,2 Oo and JO= 5 facor=— = Noy #(Se)(Ze7) = wer (7:34) ckpT \m1 These are written in terms of the molecular weight M, the weight of polymers in unit volume p (=cM/NN,), and the gas constant R (=N,kg) as Te & M2, (7.35) x” (pR N0= (=) TR* pM, (7.36) IO= oar (137) These results are confirmed for polymer melts with low molecular weight.2> 7.3.3 Calculation by reptation model Next we consider a polymer melt of high molecular weight in which entanglement is very important. To calculate G(f), it is convenient to consider the stress relaxation after a step strain. Suppose at t= 0 a shear strain y is applied to the system in equilibrium. The strain causes the deformation of the molecular conformation, and creates the stress, which Telaxes with time as the conformation of polymers goes back toLINEAR VISCOELASTICITY 227 equilibrium. Though description of this process in the general case is slightly involved (see Section 7.5), the relaxation modulus in the linear regime can be obtained by a simple argument. (A) For small ¢ (¢<1,), the dynamics is described by the Rouse model and the relaxation modulus is given by eqn (7.31). Since t, «tz, eqn (7.31) is approximated as 2 Tr\'? Gt) ha | do cat 20p*Ite) = say ko ot(=) (tt, is only due to the disengagement. This can be evaluated as follows (see Fig. 7.2). At t=+t,, the whole polymer is confined in a deformed tube. As time passes, the polymer reptates, and at time ¢ the parts of the polymer near the ends have disengaged from the deformed tube, while the part in the middle is still confined in the tube. Since only the segments in the deformed tube are oriented and contribute to the stress, the stress is Proportional to the fraction w(t) of the polymers still confined in the LE LL “Lily Fig. 7.2. Explanation of the stress relaxation after small step strain. (a) Before deformation, the conformation of the tube is in equilibrium. (6) Immediately after the deformation, the whole tube is deformed. The deformed part is indicated by the oblique lines. For small strain, the contour length of the tube is unchanged. (c) At a later time #, the chain is partly confined in a deformed tube. The average of the contour length o(s) of this part is equal to Ly(t).228 MOLECULAR THEORY OF POLYMERIC LIQUIDS deformed tube, i.e., G(t)=GPy) zt) (7.39) where G® is a certain constant and *(t) has already been calculated in Section 6.2.2 (see eqn (6.16)), vor S, ap eor(-Pt!t) (7.40) with ‘N36? (b = (°) . (7.41) To obtain G®, we utilize the fact that at t= t,, the Rouse-like behaviour (eqn (7.38)) smoothly crosses over to the reptation behaviour (eqn (7.39), ie., ” GY ~ G(s.) = 5 ket (=) (7.42) Te or using eqn (6.106) b? CP =F ka®. (7.43) Equation (7.38) is then written as G(t)= ae(%)" for t! (see Fig. 7.4). Maximum relaxation time, viscosity, and steady-state compliance. The longest relaxation time Tmax Of G(#) is ta: “N35? /b\? Nb? ax = a= Seat (a) “3° te (7.46) The viscosity and the steady-state compliance are calculated from eqns (7.29) and (7.30). Since the contribution to the integral from the region t°5 An example of the viscosity in melts is given in Fig. 7.6. The reason for the discrepancy will be discussed later. 7.3.5 Tube diameter in melts Given the general agreement in the shape of the relaxation modulus, it is possible to determine the step length a of the primitive chain’ Though various ways are conceivable, a direct way is to use the plateau modulusLINEAR VISCOELASTICITY 231 log J. (om?/dyne) log My, Fig. 7.5. Steady-state compliance of polystyrene at 160°C. Dotted line represents the result of the Rouse model. O Plazek-O’Rourke @ Mills—Nevin © O’Reilly-Prest ® Murakami et al. © Tobolsky et al. @ Nemoto ® Akovali ® Onogi et al. © Mieras-Rijn Reproduced from ref. 35. G®, which is related to a by eqn (7.43). Since cb? = £0? = 9M no", (7.50) eqn (7.43) is written as 2 GP= eRt we (7.51) Experimentally G® is often expressed by a characteristic molecular weight M,, called the molecular weight between entanglements, which is defined by RT M,= eo . (7.52)) |» —T T_T Poly(di-methyl siloxane) Constant + log 19 T Poly(iso-butylene) |. Poly(ethylene) Poly(butadiene) Poly(tetra-methyl p-silphenylene [ _ siloxane) Poly(methyl | methacrylate) x Poly(ethylene glycol) Poly(vinyl acetate) Poly(styrene) log 10'7x, ' Fig. 7.6. Steady-state viscosity of various polymers in melt, where X, is a parameter proportional to M,. The curves are shifted vertically so as not to overlap each other. Reproduced from ref. 33.LINEAR VISCOELASTICITY 233 It follows from eqns (7.51) and (7.52) that M, wm. ax (Gene?) =u, (7.53) where Ry, is the root mean square of the end-to-end distance of the polymer with the molecular weight M,. The precise numerical coefficient in eqn (7.53) is not given by the simple argument given in Section 7.3.3, To determine the coefficient, a further assumption is needed about the deformation of the tube under strain. A specific model described in the next section gives a= ait No =0.8R%, (7.54) which we shall now use. The value of M, in polymer melts is tabulated in the literature,” from which a is found to be 82 A for polystyrene and 34 A for polyethylene.** This distance is much larger than the correlation length measured by neutron scattering or X-ray scattering. At present, it is not fully understood what factors determine a, but various semiempirical relations are available for M,.°>°78 Here we shall proceed regarding @ as an adjustable parameter. Graessley®* showed that the above estimation of a is consistent with the results of diffusion experiments. According to the reptation theory, the self diffusion constant Dg is given by (see eqn (6.40)) thet °°" 3Nb? NE- To eliminate , it is convenient to consider the zero shear rate viscosity calculated by the Rouse dynamics n§(M) = one? = EN NER, (7.56) (7.55) where R%, is the mean square end-to-end distance of the polymer of molecular weight M. Experimentally, {°(M) can be obtained by extracting the viscosity for low molecular weight according to the following equation (see Fig. 7.7) nM) = no Mas) (7.57) where M,,¢ is an arbitrary molecular weight in the Rouse regime.234 MOLECULAR THEORY OF POLYMERIC LIQUIDS og 0 log M Fig. 7.7. Viscosity curve. Solid line: experimental formula eqns (7.71) and (7.72). Dashed line: theoretical curve, eqn (7.73). From eqns (7.54)-(7.56), it follows that 1 a(R us) M,_pRT_ 6735 M n®(M)" According to Graessley, eqn (7.58) gives values in reasonable agree- ment with experimental results*”? (e.g. for the case of polyethylene the error being only about 30%). Considering the different nature of the experiments, the agreement is remarkable. That the length a is rather large seems to be consistent with the fact that both neutron scattering® and computer simulation’! do not find it easy to detect reptation: the characteristic time-scale in these experiments is often shorter than t,. Recently, however, it has been reported that indications of reptation are observed in some situations.“ (7.58) 7.3.6 Semidilute and concentrated solutions So far we have been considering polymer melts. It is expected that the same picture will hold in semidilute and concentrated solutions. Though this is generally believed to be the case, a few remarks must be made. (i) In semidilute solutions, the excluded volume effect and the hydrodynamic interaction become important for dynamics on a length- scale shorter than the correlation length & (or the hydrodynamic screening length &,). The problem of how this affects the reptationLINEAR VISCOELASTICITY 235 picture is delicate. However, in the case of a good solvent, certain conclusions can be drawn from scaling arguments.“*+ Consider for example the maximum relaxation time Tnx, the steady state viscosity mo, and the plateau modulus G{. If one imposes the condition that they are invariant under the scaling transformation given by eqn (5.138), one can show that _te N25? . Fax = Ailcle*), (7.59) No = nsh(e/e*), (7.60) and Ge =< keTfi(c/c*), (7.61) where c* is the overlap concentration 1-3v ctx B (7.62) The concentration dependence can be determined if one imposes a further condition that eqns (7.59)-(7.61) must be consistent with the reptation prediction: Tmax ®N?, No N2, GY N°, (7.63) then N32"? / c \O-2GV—D ton" (5) = Mp™, (7.64) c\¥Gr-0) No= 1s >) = M*p%4, (7.65) and Bi d GQ =LkyT(£) ep (7.66) NO? \ct eo . (The last expressions indicate the result when v is equal to 3/5). Experimentally, eqns (7.64)-(7.66) seem to hold at least approximately in the semidilute regime,***’ though the exponents in the molecular weight dependence of tna, and 19 are slightly larger than 3, and the exponent in the concentration dependence of G{? is somewhat higher than 9/4 depending on the quality of the solvent. + It must be mentioned that whether dynamical scaling holds for systems dominated by the topological interaction is still a matter of debate, though eqns (7.64)-(7.66) seem to be in agreement with experimental results.236 MOLECULAR THEORY OF POLYMERIC LIQUIDS (ii) At higher concentration, the effect of the excluded volume and the hydrodynamic interaction becomes less important. If we disregard these effects, the dynamics is described by the same model as in melts except that a and € now depend on concentration. Experimentally it has been found*“*° that G® « p? (7.67) which implies axp-'?, (7.68) The concentration dependence Of Tmax and Mo is delicate since the friction constant £ depends on the concentration in a nontrivial way. 7.4 Other relaxation modes 7.4.1 Discrepancy between the theory and experiments Although the theory given in the previous section is largely in agreement with experiments on linear viscoelasticity, there remain certain discrepancies. (i) The theoretical molecular weight dependence TloX MP and Tmax M? is weaker than the experimental one; the measured exponent is higher than 3, ranging from 3 to 3.7.>* (ii) The theoretical relaxation modulus G(t) is too close to a single exponential compared to the experimental modulus.” For example, the experimental value of JG, which measures the deviation from the single exponential behaviour of G(t), is between 2 and 3**" as against 6/5 for the theoretical result of eqn (7.48). Part of these discrepancies can be attributed to the molecular weight distribution, which seriously affects the value of J&.> On the other hand detailed comparison with experimental results indicates that not all the discrepancy can be resolved by the molecular weight distribution.” The discrepancy in the exponent of the viscosity has been a matter of debate which is not yet settled. Various modifications of the reptation picture have been proposed. For example, Wendel and Noolandi® argued that if the polymers are trapped by some tight knots with an extremely long lifetime, the diffusion along the tube becomes non- Fickian and this gives a higher exponent in 7) M*. However, such tight knots, if they exist, would create a rubbery plateau with extremely long +A similar problem exists in the melt, where it is often observed that the viscosity does not follow the Rouse behaviour at small molecular weights much less than M,. This is attributed to the fact that for short polymers the segmental friction constant € depends on the molecular weight.*?OTHER RELAXATION MODES 237 relaxation time, which no viscoelastic data seem to support. Curtiss and Bird™ suggested that the segmental friction constant may depend on the molecular weight, but this hypothesis seems to contradict the result of the diffusion experiment. Another possibility suggested by Ball® takes into account the long-range correlation in the motion of vacancies needed for the reptation motion to take place. This has been studied by computer simulation,** but the result is not yet conclusive. At present, a more consistent explanation seems to be that given by Graessley,° who pointed out that the observed viscosity and the relaxation time are smaller than the calculated ones. Using eqns (7.32), (7.33), (7.46), and (7.47), one can show that 3 2 nol) =" no) ) = nay 2) (7.68) and 3 Tan aoe MA(gr) GEM). (7.70) On the other hand, a widely accepted empirical formula is no(M.) aM for MM., (7.72) where M, is a certain molecular weight which is two or three times larger than M,. Using M,=2M., and no(M.) = no(Mc)(M./M-), eqn (7.69) is written as nst(M) = 15 M(H)» 7.73) nf**(M) is about 15 times larger than n{*°)(M) at M,, but the discrepancy decreases with increasing M and diminishes at M =(15)'°*M, ~ 800M, (see Fig. 7.7). Graessley thus conjectured that although the Pure reptation behaviour will be observed for very large molecular Weight, there is a large cross-over region in the viscosity from the Rouse-like behaviour to the pure reptation behaviour, which gives an apparent exponent larger than 3. Unfortunately since no data are available for molecular weights higher than 800M,, which is 3 x 10’ for a polystyrene melt, the crucial test of Graessley’s conjecture has not been given. However, it is obvious that238 MOLECULAR THEORY OF POLYMERIC LIQUIDS any relaxation process which occurs concurrently with reptation de- creases the viscosity and hence reduces the discrepancy between the theory and experiment. 7.4.2 Contour length fluctuation and tube reorganization Two relaxation processes have been suggested to alter the pure reptation behaviour discussed in Section 7.3.3. Contour length fluctuation. As was shown in the previous chapter, the contour length fluctuation reduces the disengagement time ty significantly. From eqns (6.20), (6.101), and (7.54), the disengagement time +? of a chain with fluctuation is given by ‘M, 12,2 P= o7(1 -x(F) ) (7.74) M where X’ is estimated as 1.47+* V(4/5) =~ 1.3.°’ For M/M, =50 the ratio between 1 and r{" is about 0.67 which displays the considerable effect of the contour length fluctuation in the region where the polymers are usually regarded as ‘fully entangled’. A crude calculation’”* indicates that the discrepancy in the viscosity and the steady state compliance is significantly improved if the contour length fluctuation is taken into account. Tube reorganization. So far, it has been assumed that the tube is fixed in the material and its conformational change occurs only at the ends. It is conceivable that the conformational change of the tube can occur in the middle.“ For example: (i) Constraint release: The topological constraints for a polymer can be teleased (or created) by the reptation of the surrounding polymers as shown in Fig. 7.8. This will cause the conformational change of the tube in the middle. A model describing this process is to regard the conformational change as a local jump of the primitive chain. Since the jump rate is of the order 1/t,, this process has a negligible effect on the longest relaxation time. However, the process gives an additional relaxation to G(t) in the plateau region, and improves the value of IOGY. (ii) Tube deformation: If a polymer is in a strained conformation, it will tend to relax the strain by creating the deformation of the surrounding polymers. This effect may be handled by considering the deformation of a strained chain placed in a viscoelastic medium. So far NO quantitative estimation of this effect has been done. Though these processes are conceivable, estimation of their effect isSTRESS RELAXATION AFTER LARGE STEP STRAIN 239 ©) Fig. 7.8. Release and creation of the topological constraints. (a) The topological constraints imposed on the chain A by C is released and recreated by the motion of C. (6) In the two-dimensional representation, this process can be represented by the disappearance and reappearance of the obstacle C. The process causes the deformation of the tube in the middle. still at the level of conjecture, and there are other theoretical treatments. Experimentally, in linear polymers with narrow molecular weight distribution, it seems that the major difficulty of the theory can be Tesolved by including the contour length fluctuation. On the other hand the tube reorganization is believed to be important for polymers with broader molecular weight distribution,” or long branches, which will be discussed later. 7.5 Stress relaxation after large step strain 7.5.1 Experimental setup Having seen the characteristic features of the linear viscoelasticity, we shall now study the nonlinear viscoelasticity. Before studying the general situation, we shall first consider a simple case, the stress relaxation after Stepwise deformation.“ Suppose that at time ¢=0, a polymeric liquid is suddenly deformed homogeneously. The deformation creates a stress which gradually relaxes with time. Our problem is to find how this telaxation takes place. For a homogeneous deformation, we may assume without loss of240 MOLECULAR THEORY OF POLYMERIC LIQUIDS fa) (b) Fig. 7.9. (a) Shear and (6) elongation. generality that a point r in the material is displaced to ror=E-r. (7.75) The tensor E is called the deformation gradient.t Two particular cases are often studied experimentally. (i) Shear (Fig. 7.92), for which the material is deformed to Tea het My Ty Hy TE Te (7.76) This deformation is characterized by a single parameter y, the shear strain. Since the deformation has a reflection symmetry with respect to the xy plane, the stress components 0,,(=0,,) and o,,(=0,,) vanish identically. However, the shear stress o,, = 0,, and the diagonal stresses Ox, yy and o,, generally do not vanish. Since the isotropic part of the stress has no significance, only two of the diagonal components have meaning. The stresses N, = Ox — Oyy (7.77) and Ny = Oy — 02, (7.78) are called the first and the second normal stress differences. Thus the tesponse of the shear deformation is characterized by three stress components 0,,, N, and N;, each of which are nonlinear functions of y and t. (ii) Uniaxial elongation (Fig. 7.9b). Here the sample is stretched in the + A general deformation is described by the function r’ = r’(r) which connects the position vectors r and r’ before and after the deformation. In such a case Ez, is given by 4rg/ dr.STRESS RELAXATION AFTER LARGE STEP STRAIN 241 z direction by factor A. Since the volume is unchanged, this causes a contraction in the x and y directions by a factor 1/VA. Hence, the deformation is described by iat rt ' mate WA Tyte ee ar. (7.79) In this deformation, the off-diagonal components of the stress vanish and two of the diagonal components o,, and 0,, are equal to each other. Thus the response is characterized by a component Or = Oz, — One (7.80) which is called the tensile stress. 7.5.2 Calculation by Rouse model First we calculate the stress relaxation using the Rouse model. The stress tensor (eqn 7.4) is expressed by the normal coordinates X, of the Rouse model (see Section 4.5.2, eqn (4.137)) ce Oat = Fy D Ky Xpelt)Xoa(0)) (7.81) p=1 where 6x7kyT k= ge? (P=1,2,--.). (7.82) According to Rouse dynamics, the position of the segments are changed in the same way as the macroscopic point;} thus if R,(—0) and R,(+0) are the positions of the segment before and after the deformation, R,(+0) = E+ R,(—0) (7.83) X,(+0) = E+ X,(-0). (7.84) Since the system is in equilibrium for «<0 (Xpa(—0)Xpa(-0)) = "27 5,,, (7.85) kp t This i is often called the affine deformation assumption, but it is actually derived from the equation for the Rouse model (eqn (4.139)). For an instantancous deformation, the velocity gradient Kap (t) becomes so large that it dominates the other terms on the right- hand side of eqn (4. 139) ‘The equation for X, then becomes the same as that for the macroscopic Point (see eqn (7.151)), and therefore the change of X, becomes affine.w2 MOLECULAR THEORY OF POLYMERIC LIQUIDS From eqns (7.84) and (7.85), it follows that (Xpa(-+0)Xp(+0)) = Ea Epe(Xou(—0)Xo(—0)) = Ey Er Spe “at ‘P = Byg(E) “2 (7.86) where ° Bap (E) = Eau E py (7.87) which is called the Finger strain. For t>0, (X,a(t)Xpp(t)) satisfies eqn (4.141) with Kap =0: a 2p kp > (Xpalt)Xpp(t)) = —=— ( (Xpul%o) — Sap}. (7.88) ot tr k, Equation (7.88) is solved with the initial condition (7.86) by kpT (Xpe(t)Xpp(t)) = 7 [Bap(E)exp(—2pt/tp) ‘P + Sap(1—exp(—2p7t/t,))]. (7.89) Substituting eqn (7.89) into eqn (7.81) and dropping the isotropic term we have Gap (t) = kp TBap(E) D exp(—2p"t/ tx) (7.90) N ie = Baa{E)G(t) (7.91) where G(t) is the linear relaxation modulus given by eqn (7.31). For a shear deformation, E is given by 1 y 0 E=|0 1 0 (7.92) 001 so that 1+y? y 0 B= Y 10 0 o1 and Ory = YG(t), (7.93) M=/G), (7.94) Np =0. (7.95) Note that the shear stress is a linear function of y even if y is large.STRESS RELAXATION AFTER LARGE STEP STRAIN 243, For the uniaxial elongation, the tensile stress is given by or= (2? -7)60. (7.96) The Rouse-like behaviour is expected to be seen in the initial stage of the relaxation (t< t,). 7.5.3 Calculation by reptation model Now we consider the behaviour for t > t, using the reptation model. To calculate the stress, we have to know (i) How the conformation R(s, t) of the primitive chain is changed by the macroscopic deformation, and (ii) How the stress is calculated for a given conformation R(s, t). We shall discuss these problems separately. Expression for the stress tensor. The microscopic expression for the stress tensor can be obtained by taking the average of eqn (7.4) for a given conformation of the primitive chain.“ Alternatively, it can be derived by an elementary argument explained in Fig. 7.10. In both cases the result is F(t) Fig. 7.10. Consider a plane of area A, normal to the z axis. The stress component o,, is given by the force (per area) S,/A acting through the plane. Consider a part of the primitive chain between s and s + As. If this part is within the distance u,(s,t)As from the plane, it penetrates the plane and gives a contribution of F,(s, t) = F(t)u.(s, t) to S,, where F(t) is the tensile force acting along the primitive chain. Since the number of the primitive chain in unit volume is (c/N), the number of such part is (c/N)Au,(s, t)As, whence s.=5 (ZAu(s, )ask(s, )= Lal [arouc, outs. ). Thus the force per area S,/A gives eqn (7.97).244 MOLECULAR THEORY OF POLYMERIC LIQUIDS written as L Oup(t) =<( | dsF(0)ua(s, t)up(s, 0) (7.97) 0 where u(s, t) = AR(s, t)/s is the unit vector tangent to the primitive chain and F(t) is the tensile force acting along the primitive chain. In the equilibrium state F(t) is given by (see eqn (6.83)) 3kaT ; aE L, (7.98) while in the non-equilibrium state in which the contour length is L(t), F(t) is given byt F(t)= F(t)= ae Th). (7.99) From eqns (7.97) and (7.99) ) apt) = SE (fare 1p (6s #)—38ap)). (7-10038 This formula shows that the stress is determined by two quantities, the contour length L(t) and the orientation of u(s, t). + Here it is assumed that the tensile force is constant along the chain. This assumption is not correct immediately after the deformation because initially the Rouse segments are stretched or compressed along the primitive path depending on their direction (see Fig. 7.11). However, such local imbalance in the segment density is adjusted in the time t,, and for t>t,, the tensile force F(s, t) can be regarded as independent of s. + Curtiss and Bird™ derived a slightly different stress formula, which includes an adjustable parameter called the link tension coefficient «. However, this formula is not consistent with the stress optical law unless ¢=0. In the case of ¢=0, the formula becomes essentially equivalent to eqn (7.120) which is a special case of eqn (7.100). Equation (7.100) is also derived from the principle of virtual work. The free energy per unit volume is = £ 3keT(LO") a Under a virtual deformation 6€,,, L(t) changes by L oL= J ds5eqgua(S, t)ug(s, t) ° whence La) dat = PERF (LOL) = = beag £3kat (Le) | dsu,(s, t)up(s, ») Oo which gives eqn (7.100).STRESS RELAXATION AFTER LARGE STEP STRAIN 25 Deformation of the primitive path. Our next question is how the conformation R(s, t) of the primitive chain is deformed by the macro- scopic strain. The simplest assumption is that the deformation is affine, ie., that the primitive chain (or the central axis of the tube) is deformed in the same way as the macroscopic deformation. Thus the point R(s, —0) on the primitive chain is displaced as R(s, -0) > E R(s, -0). (7.101) Let us now study how this changes the contour length L(t) and the orientation of u(s, t). This is illustrated in Fig. 7.11. (i) The change of the contour length: The transformation (7.101) changes the length As of a line segment on the primitive chain to As = As |E + u(s, —0)|. (7.102) The length AS can be larger or smaller than the original length As depending on u(s, —0). Since the distribution of u(s, —0) is isotropic, the average ratio between AS and As is given by a(E) = (|E-ul)o (7.103) a(e)L mt 0 Fig. 7.11. Deformation of the primitive chain by macroscopic strain. Here for the Purpose of explanation, the primitive chain is represented by randomly connected ime segments. According to the affine deformation assumption, a line segment Tepresented by r is transformed to E + r. Thus the length As = |r| and the direction u=r/|r| are changed as As—>|E + u| As and u— E- u/|E- ul.246 MOLECULAR THEORY OF POLYMERIC LIQUIDS where {...)o denotes the average of u over the isotropic state, du seslo=] ares 104) (.do= | S (7.104) For any deformation which conserves the volume (det |E|= 1), it can be shown that’? a(E)>1. (7.105) Thus the contour length of the primitive chain immediately after the deformation is given by (L(+0)) = (EL (7.106) which is always larger than L. (ii) The change of the orientation: To denote the orientation of the primitive path, we define the orientational tensor Sap (S, t) = (ua(s, t)ug(s, t) — 356) (7.107) which vanishes in the isotropic state, but does not vanish in the oriented state. Since the distribution of u(s, —0) is independent of s, S,g(s, +0) will also be independent of s and can be written as Sup(s, +0) = Qup(E). (7.108) To calculate Q,(£), let us consider the probability distribution function f(u, s, t) for the tangent vector u(s, t). (f(u, s, t) is the probability that the tangent vector at s and t is in the direction u.) Obviously 1 f(u, s, -0) "ke (7.109) By the deformation, the unit vector u changes as ._ E-u w= TE (7.110) The probability that an arbitrarily chosen point of the deformed primitive chain is in the direction of u’ = E - u/|E - ul is proportional to [E - u|, the length of such a part. Thus flu’, s, +0)= cf du |E + ul 3(w Ea s,-0) (7.111) where C is the normalization constant which is determined from the condition 1=C | du'f(u’, s, t) (7.112)STRESS RELAXATION AFTER LARGE STEP STRAIN 247 to be Ct=(4x) | du [E+ ul = (|E-ul)o. (7.113) Then flu, s +0 = [du |e -w'| 6(u- =~") (7.114) ” 4n(|E-ul)o |E-'| so that Qne(E)= { dutagflin, 5, +0)~ 3806 = (EEO) Eula Wops (715) Stress relaxation. Now it is easy to calculate the stress relaxation. According to the model described in Section 6.4, the relaxation of L(t) occurs on the time-scale of tz, while that of orientation occurs on the time-scale tz. Thus the stress relaxation for t > t, occurs in two steps (see Fig. 7.12). (i) Contour length relaxation: In the time-scale of tz, S.g(s, t) can be regarded as equal to the initial value Q,,(E), so that eqn (7.100) is (@) (b) (o) @ Fig. 7.12. Explanation of the stress relaxation after large step strain. (2) Before deformation the conformation of the primitive chain is in equilibrium (¢ = —0). (6) Immediately after deformation, the primitive chain is in the affinely deformed conformation (t= +0). (c) After time tz, the primitive chain contracts along the tube and recovers the equilibrium contour length (t= tz). (d) After the time t,, the primitive chain leaves the deformed tube by reptation (f= 1,). The oblique lines indicates the deformed part of the tube. Reproduced from ref. 107.28 MOLECULAR THEORY OF POLYMERIC LIQUIDS written as Oop = EBT (1(0))*Qap(E) for e tp, L(t) is at the equilibrium value L so that eqn (7.100) is written as £ € 3kT - apt) = Rts L| ds (uals, tps) —4800) 0 £ 1 =65 J dsSz(s, 1). (7.120) Now S,g(s, t) is equal to Q,.6(E) if the primitive chain segment s in the deformed tube, and is zero if it has left the tube. Since the probability that the primitive chain segment s is in the deformed tube is 1(s, t) (see Section 6.2.2) Saa(s, t) = Qup(E)Y(s, t). (7.121) Hence op) = G.Qan(E)z | d5¥65, 0) =GQu(E\Y(t) > tr. (7.122) Here y(t) is given by eqn (7.40). Combining eqns (7.118) and (7.122) we finally have Oap(t) = G.Qap(E)(1 + (a(E) — lexp(-t/tp))° YO) (> 4,). (7.123)STRESS RELAXATION AFTER LARGE STEP STRAIN 249 In the case of a shear deformation, we may write a(E) and Q.,(E) as a(y) and Q.9(y), respectively. For small y, Q,,(y) and a(y) are easily calculated. a(y) = (1+ 2yuguy + y7u5)'?)o = (14 yuu, — dy?uzuy + 4y7u5)o=1+ Ky? + O(y*) — (7.124) and a (__ et ry Qo) = a(y) (a + 2yu,uy + aay), = (uy — uus)o + O(Y°) = By + O(Y’). (7.125) Thus eqn (7.123) becomes Oxy (t) = YG. WY) + O(y’). (7.126) Hence, the relaxation modulus in the linear viscoelasticity is given by GO =8G.y(). (7.127) This determines the plateau modulus to be 2 GP =48G, =f Skat (7.128) which gives eqn (7.54). We shall now compare the results for large strain with experimental results. 7.5.4 Comparison with experimental results Shear, Extensive experiments on the stress relaxation for shear deforma- tion have been done by Osaki et al. For convenience of comparison, we shall represent the relaxation of the shear stress by the nonlinear relaxation modulus defined by 1 Gi y) = 7 alt y)- (7.129) In the limit of y—>0, this reduces to the relaxation modulus of linear viscoelasticity. Equation (7.123) gives Gt 1) = G, 22D + (aC) ~ 1pexp(—ra WO =A(y)GO(1 + (ey) — Yexp(—t/ta)? (7.130)Git. WG, 10? 107" 10° 10 the Fig. 7.13. Theoretical curve of the nonlinear relaxation modulus G(r, y). The case of T,/Tp = 100 is shown. Reproduced from ref. 64. 10! G(ty) (Pa) 10° 107 10° 10! 10° 10° t(sec) Fig. 7.14. Nonlinear relaxation modulus G(t, y) for polystyrene solution of chlorinated biphenyl at 30°C. The molecular weight of the polymer is 8.42 x 10° and the concentration is 0.06 g/cm. Magnitudes of shear y are <0.57, 1.25, 2.06, 3.04, 4.0, 5.3, and 6.1, from top to bottom. Reproduced from ref. 69.STRESS RELAXATION AFTER LARGE STEP STRAIN 251 where G(t) is the relaxation modulus of linear viscoelasticity and (7) = Q,,(y)/ Gy). (7.131) Equation (7.130) is plotted in Fig. 7.13. For small y, G(t, y) decays roughly in a single exponential manner with relaxation time t,. For large y, Gt y) shows another relaxation characterized by tz corresponding to the relaxation of the contour length. Such behaviour has actually been observed experimentally® as shown in Fig. 7.14. Detailed comparison teveals good agreement between the theory and experiment. (i) Osaki et al® found that at large t, curves in Fig. 7.14 for various y can be superimposed by a vertical shift (see Fig. 7.15). This implies that, for large t, G(t, y) can be written as a product of two functions, one depending on time and the other on strain. This agrees with eqn (7.130), which is written for t > tz as GO y) =hA()GO. (7.132) The function A(y), called the damping function, is found to be independent of the molecular weight and concentration over a wide range.7° Figure 7.16 shows the comparison between the theoretical damping function and the experimental one. The agreement is very good considering that h(y) includes no adjustable parameters. 10° 107 Gn'Pa) 10! 10° 10" 10° 10° t (sec) Fig. 7.15. Reduced relaxation modulus G(s, y)/h(y) derived from Fig. 7.14. Each curve for y > 1.25 in Fig. 7.14 is shifted vertically by an amount —log h(y) so that it superposes on the top curve in the long time region. t? indicates the longest relaxation time, and t, the characteristic time below which the superposi- tion is not possible. Reproduced from ref. 69.252 MOLECULAR THEORY OF POLYMERIC LIQUIDS AY) 10" 107! 10" 10 10" y Fig. 7.16. h() determined from the procedure explained in Fig. 7.15. Filled circles represent polystyrene of molecular weight 8.42 x 10° and the unfilled circles of 4.48 x 10°. Directions of pips indicate concentrations which range from 0.02 gcm’ to 0.08gcm*. The solid curve represents the theoretical value (eqn (7.131)), and the dashed curve the result of the independent alignment approximation (eqn 7.187). Reproduced from ref. 69. (ii) Experimentally, the first relaxation can be characterized by the time 1, below which the factorization of G(t, y) is not possible (see Fig. 7.15). Osaki et al. found that the ratio between t, and the Rouse relaxation time tg is about 4.5 and essentially independent of the molecular weight and concentration. (iii) The relaxation of the other stress components N,(t, 7) and Nt, y), measured by birefringence, have precisely the same time dependence as g,,(f, y), and their ratio depends only on y. This agrees with eqn (7.123), according to which MG y= Galt) rc Ooty) (7.133) (7) Q»(y) — Q.2(Y) NXt, Y) = O.(y) O,,(t, Y). (7.134) Since it can be proved™* that Qux(¥) — Qy(Y) = YOxy(Y), (7.135)STRESS RELAXATION AFTER LARGE STEP STRAIN 253 107 —Nelt,yV/Nilty) 10% 10° 10° 10° y Fig. 7.17. Quantity —N,/N, is plotted against magnitude of shear y. Sample: polystyrene solution in chlorinated biphenyl (M,, = 6.7 x 10°, p =0.40gcm™°). The number of entanglements Z corresponds to about 14. The solid line is the theoretical value, —(Q,,(y) — Q..(7))/(Qux(7) — Q,,(y)). The dashed line is the Tesult of the independent alignment approximation. Reproduced from ref. 67. eqn (7.133) is written as MG Y) = 78n(t Y)- (7.136) This relation, first found by Lodge and Meissner”! using a phenomenolo- gical argument, has been well confirmed.’ The ratio between the second normal stress difference N,(t,y) and the first normal stress difference N,(t, y) is shown in Fig. 7.17. The experimental values are again in reasonable agreement with the theory. Uniaxial elongation. For uniaxial elongation, the tensile stress is given by Or (t, 4) =[1 + (@(A) — Dexp(—t/te) PF(A)G (0). (7.137) Here @(A) denotes a(E) of uniaxial elongation and f@)=4(@..A) - Qn). (7.138) Explicit formulae for w(A) and f(A) can be calculated analytically,” a(A) =44(1 + A(A)) (7.139) and _ 15(43+1/2) 1 4-1 fO=TGi-y Ts4@) (1-F74@) 140)254 MOLECULAR THEORY OF POLYMERIC LIQUIDS where sinh™[(a? — 1)"7] eae— HP? The stress relaxation for uniaxial elongation has been studied by Ferry et al.,” and the result has been well fitted by eqn (7.137) if the effect of the molecular distribution is taken into account.” A(A) = (7.141) 7.5.5 Discussion As we have seen the theory has predicted many aspects of the nonlinear stress relaxation. However, there are some experimental results which are not in accordance with the theory and need some discussion. Anomalous stress relaxation in shear flow. Osaki et al.® found that the nonlinear relaxation modulus G(f, y) of polystyrene solutions does not agree with the theory for very-high-molecular-weight samples for which Mp > 10° g/cm?. (7.142) For these samples, the stress near t, decreases much more steeply than predicted by the theory, and shows complex M, p dependence. A similar anomaly is also reported by Vrentas ef al.” This result is puzzling since the theory should become most valid in the high molecular weight limit. A possible explanation, however, was given by Marrucci and Grizzuti,”* who pointed out that the theoretical damping function h(y) has a region where the differential rigidity modulus d(yh(y))/dy is negative. In this tegion, the elastic energy can be lowered by microscopic phase separa- tion, each phase having different local shear strains. Indeed the observed anomalous behaviour can be reproduced with suitable assumptions.”>”° Experimental evidence of such microphase separation is, however, lacking and further study is expected. Tube reorganization. The theory described in Section 7.5 includes two essential assumptions; (i) the conformation of the tube remains un- changed and (ii) the contour length of the tube returns to the equilibrium value L even if the environment is not in equilibrium. The validity of those assumptions is not established and it is worthwhile to study the consequences of the theory based on other assumptions. Marrucci et al.”7’ assumed that the volume of the tube remains constant by deformation, and derived a result which has the same time dependence as eqn (7.122) but different strain dependence. Though an experiment on PMMA” seems to fit with the modified formula, caution is needed in accepting the modification since critical experiments needNONLINEAR VISCOELASTICITY 255 data over a large time-scale and for a monodisperse sample, but the quoted experiment does not meet these conditions. Viovy et al.” argued that as the contour length of the surrounding primitive chain contracts, there will be an extra relaxation by the release of the topological constraints. They proposed a theory which gives a slightly different relaxation behaviour for ¢> M,. In practice, the condition (7.169) is not always satisfied, and the elongation of the contour length can be important, but this will not be considered here. For the sake of simplicity, we shall use a slightly different notation in this and the following two sections: the equilibrium contour length will be denoted by L (because L and L need not be distinguished for the inextensible model), and the segments of the primitive chain are labelled from —L/2 to L/2. (Thus the segment 0 corresponds to the middle of the chain.) 7.7.1 Deformation of the primitive chain First we express the transformation rule of the inextensible primitive chain in mathematical terms. Let R(s) and R(s) be the conformations of the primitive chain before and after the deformation. The transformation tule is explained in Fig. 7.21, i.e., (a) The segment in the middle changes its position affinely, i.e., RO) =E-R(0). (7.170) (b) The segment § lies on the curve E - R(s), so that R(5) = E-R(s) (7.171) Fig. 7.21. The deformation of an inextensible primitive chain by a macroscopic strain. The new conformation A’O’B’ is on the curve A"O”B”, which is the affine transformation of AOB. The new position of a segment, say C, is determined from the condition that the contour length O'C' is equal to OC, where O’ is the affine transformation of O and coincides with O”. Reproduced from ref. 108.262 MOLECULAR THEORY OF POLYMERIC LIQUIDS where § is the contour length along the curve E- R(s’) from s’=0 to s'=S, ie., s= fas’ |E-u(s’)|. (7.17) From eqn (7.171), the transformation rule for the direction u(s) is obtained as a) =A) =-2e- RG) =e. a(S) = gph O =57F R(s) aps © R(s) (7.173) os =F -u(s). (7.174) Equation (7.172) gives 1 -Se - u(s)}. (7.175) From eqns (7.174) and (7.175) it follows that E-u(s) |E-u(s)|" This is equivalent to the transformation given in Fig. 7.11. a6) = (7.176) 7.7.2 Independent alignment approximation According to eqn (7.172) s and § are not equal to each other. This leads to a constitutive equation of a rather complicated form (see Section 7.9). If we disregard the difference between s and § and assume the following transformation rule E-u(s) Ewe) (7.177) the constitutive equation is obtained in a simple form. This prompts the study of the approximation (7.177), which we call the independent alignment approximation (IA approximation).'*!5 Though the physical justification of this approximation is not clear, its error usually turns out to be small except for a few cases which will be discussed later. Therefore we shall first proceed using this approximation. According to eqn (7.100) the stress for the inextensible model is givenCONSTITUTIVE EQUATION FOR REPTATION MODEL 263 by c 3kpTL’ 1 NW NB? if ds (ua(s, t)up(s, t) — 4506) Capt) = L2 =G.z | as(uals, Qupls, 1) —45ap) -L2 L2 1 Gr , J dsSap(5, t)- (7.178) ay To calculate S,,(s,t) we need to obtain the probability distribution function f(u,s,t) that the tangent vector at the segment s is in the jirection u at time t. The time evolution equation for f(u, s, t) is obtained n the same way as in Section 6.3. Suppose that in the time interval between ¢ and ++ At, the chain egment s moves to the position at which the chain segment s + A& was ocated at time ¢, then u(s, t + Af) is given by E(t + At, t)- u(s + AG, t) ms, 6+ At) = JE(e+ At, t)- u(s + AE, 1)" (7.179) since the distribution function of u(s+A&,t) and A& are given by "u, s + AE, t) and W(A8&), respectively (see eqn (6.24)), the distribution function of u(s, t + At) is given by _ re(,, Et At teu" f(u,s, 1+ At)= faAew(as) | du 5(w Bera al a) x f(u', s + AE, t). (7.180) To assess the accuracy of the IA approximation, let us consider the stepwise deformation E imposed at t=0. In this case the orientational distribution before the deformation is fu, s, -0) = = (7.181) Since (A) becomes 6(A&) for an infinitesimally small time-interval At, eqn (7.180) gives fu, s, +0) = ia a(u - Eni) (7.182)264 MOLECULAR THEORY OF POLYMERIC LIQUIDS Hence S,¢(s, +0) is given by Sag(s, +0) = | du(ugity —48ap)f(us 8, +0) du’ E-u' = [eu Sy (tetia~480p)8(u 7g 7) du (E+ u)a(E + u)p 4a |E- ul? (E-u).(E + u), «era, = OME). (7.183) On the other hand the correct value of S.g(s, +0) is Q.g(E) given by eqn (7.115). Thus in the case of stepwise deformation, the LA approximation amounts to a decoupling approximation —4ap (E-u)a(E-u)\ 1 _ (E+ u)a(E-u)p ( E-ul ) de-anc=4 E-u? ). (7-184) It can be seen that the error of this approximation will not be large for any form of E. Indeed for the case of shear deformation, the IA approximation gives OM)=3y «KD, (7.185) OS) - OY) = BY « 1). (7.186) The damping function is thus given by AA” y) = OSM(y) /' (2). (7.187) This is shown by the dashed line in Fig. 7.16. It is seen that the error of the IA \ approximation is small over a wide range of y. A useful formula for Q%9(E) for general E is given in ref. 87. 7.7.3 Constitutive equation To obtain f(u, s,¢) in the general case, we rewrite eqn (7.180) in a differential form. For small At, E(t + At, t) is written as 1+ «(t)At. Hence E(t+At,th-u ut K(t)-udt |E(@ + Az, t)- val |u + «(0) - wAt| =u + (K(t)+u — (uu: K(t))u)At=u+ AM(u, t) (7.188) T(u, t) = K(@) + u — (uu: K(t))u. (7.189) whereCONSTITUTIVE EQUATION FOR REPTATION MODEL 265 Thus if we neglect the terms of order (Af)*, eqn (7.180) is rewritten as flu s, + At) = | dAEW(AE) du’d(u—u' — AMT(w’, )f(u', s+ AE, 1) = | angw(agy(1 - are Ty, )ftw, s+A6&,0) 2 = | dagw(agy(1 + age +4(AE)? S)(1 - are Tu, 0) fw st) =(1 + AID, 25 z )(a - are E(u, 0) ) fl 50). (7.190) Comparing the terms of order At, we have a # a 3 5,0 =D, gpfu 5, t)- aa" T(u, Of(u,s,t). (7.191) The boundary condition is that the tangent vector at the chain end is isotropic, 1 _ = . A f(u,s, Dare at s=+L/2. (7.192) Equation (7.191) can be rigorously solved to give’® t = [{ar(2 wo, 1-19) { (py - FOO flu s, 1) J at ( <.w,1-1) | a 5(u oP 1) (7.193) where p(s, t) is given by eqn (6.14). This solution could also have been arrived at by physical argument. Suppose that a tube segment is created in the direction u' at either of the chain ends between time ¢’ and ¢’ + dr’. If this still survives at time ¢ and is now occupied by the primitive chain segment s, u(s, t) must be E(t, t’)-u’/|E(t, t')-u’| (this is the result of the IA approximation). Since the probability that this happens is pe t—t’')/at'] dt’ and the distribution of u’ is 1/4a, we get eqn (7.193). From eqn (7.193), the orientation of the primitive chain segment is given by Sag(s, t) = J a(S w(s, t- 1) f auf E(,t’)-u' X (Ugg — 4809)(u ieeryel I) ~ fa (ye Viotot OPEC, r"). (7.194)266 MOLECULAR THEORY OF POLYMERIC LIQUIDS Substituting this into eqn (7.178), we finally have Cap(t) = Ge | ar(2 wt -1)) OPEC, r) (7.195) where p(t) is given by eqn (7.40). Equation (7.195) is the constitutive equation which comes out of the IA approximation. We shall now compare this equation with experimental results. 7.7.4 Comparison with experiments General feature. Equation (7.195) can be written in a more convenient form. Consider that a stepwise strain E is applied at time 0, then E(¢, t') is given by — ift>Oand?'<0, t, t’)= . Fe) ti otherwise. (7.196) Let Pap(t, E) be the stress caused by this deformation at a positive time ft. Equations (7.195) and (7.196) give beplt 5-6. a(S ve-) ogre) = Gy (NOvp(E). (7.197) This is very similar to eqn (7.122) except that Q,(E) is now replaced by Q%(E). In fact eqn (7.197) can be derived much more simply by the reasoning given in Section 7.5.3, if it is noted that S,,(s, +0) is given by Q%#)(E) in the independent alignment approximation. Using p(t, E), the stress for an arbitrary flow history is given by t Oxp(t) = | a(S bap (t—t', 6) . (7.198) 2, ot E=£(,"') Equation (7.198) agrees with the empirical equation proposed by Bernstein, Kearseley, and Zapas (BKZ),* who found that the stress response for various flow histories can be predicted by eqn (7.198) using the stress relaxation function @,g(t, E) determined experimentally. Subsequent experiments done by many authors®-** revealed that the BKZ equation is one of the most successful empirical constitutive equations. As was shown in Section 7.5, the empirical stress relaxation function gap(t, E) is in good agreement with the reptation theory for linear polymers of narrow molecular weight distribution. This, together with the success of the BKZ equation, indicates that the constitutive equationCONSTITUTIVE EQUATION FOR REPTATION MODEL 267 derived by the reptation theory works well for general flow histories. Indeed eqn (7.198) reproduces many characteristic features of the nonlinear viscoelasticity. Leaving detailed comparison to refs 51 and 94-97, we shall study the main features briefly. Nonlinear viscoelastic behaviour. To see the characteristic features of the constitutive equation (7.195), we approximate w(t) by vo~{ 1 for tty. (7.199) Then eqn (7.195) gives Fup ~ G-OT~(E(t, t—ta)) (7.200) which simply says that the stress is given by the elastic deformation caused between ¢ — ty and ¢. (i) Steady shear flow. In the steady shear flow, eqn (7.200) gives the shear stress G,y(K) = G.QGM(Kt4) ~ Gxt gh (Kr4). (7.201) Thus the viscosity becomes n(x) = 10h (Kt,). (7.202) Equation (7.202) indicates the shear thinning occurs at kK ~1/t,, which becomes very small for large molecules. This explains why the nonlinear response is important in polymeric liquids. The first normal stress coefficient is also estimated as W(k) = G_t7h (Kt) (7.203) which again decreases with the shear rate as shown in Fig. 7.18. The second normal stress coefficient W(x) = —G.tah(Kta), where WS") =~ (OM) OY), (7.204) is negative and again decreases with x, in agreement with experiments. The ratio W,(0)/'¥,(0) is precisely evaluated as™*+ 20) _ O97) - OY) Hi) OL) - OF(y) 1-0 The experimental value is between —0.1 and —0.3.%°-" tIf the IA approximation is not used, the precise value of W,(0)/¥,(0) becomes -1/7=-0.14. =-3=-~0.3. (7.205)268 MOLECULAR THEORY OF POLYMERIC LIQUIDS (ii) Stress growth in shear flow. When a shear flow is started with constant shear rate x at time ¢=0, eqn (7.200) gives 1 GK) Gt t— Tah tt). (7.206) Since Kt for tt,, (7.207) therefore GthM(kt) for tty. ( ) Since yh(y) has a maximum at y=2, n*(¢;«) shows a maximum at t=2/Kk provided Kt, 2. The height of the maximum decreases as 1/x. These features are in agreement with the experimental results shown in Fig. 7.19. (iii) Steady elongational flow. The steady elongational viscosity is given as ne(e) =~ (OM(Er.) - OW(Et) (7.208) which first increases slightly with é and then decreases with é. This was in contradiction with earlier data® for low density polyethylene, which indicated a sharp rise of nz(é), but recent data for monodisperse linear polymers®*™ are consistent with the theory. 7.7.5 Discussion Though the theoretical constitutive equation (7.195) explains many features of nonlinear viscoelasticity, there are some discrepancies which are worth discussing. Steady shear flow. The predicted steady state viscosity n(x) depends on the shear rate x too strongly. In fact eqn (7.195) predicts that at high shear rate of Kt, >>1™* n(x) =~ n(O)(«r,) >? (7.210) i.e., the shear stress o,,(K) = (k)k decreases with increasing shear rate, which means that the shear flow is not stable at high shear rate. At first sight this conclusion may seem to contradict the many experiments whichCONSTITUTIVE EQUATION FOR REPTATION MODEL 269 show stable shear flow up to very high shear rate. However, the theoretical prediction is not entirely ruled out for various reasons: (i) Equation (7.195) indicates that the form of (x) is sensitive to the relaxation spectra of the linear relaxation modulus G(t): the broader the relaxation spectra is, the smaller the exponent x in n(x) «x~* becomes. If the sample has broader molecular weight distribution, the relaxation spectra of G(t) becomes broad and the anomalous behaviour of n(x) disappears. Also, even for the monodisperse sample, the various relaxa- tion processes discussed in Section 7.4.2 broaden the relaxation spectra and weaken the shear rate dependence of n(x). (ii) Equation (7.210) is derived under the condition WtyKK K 1/tr. (7.211) If Ktg becomes of the order of unity, the contour length L(x) increases with the shear rate, and the stress starts to increase according to Oy & L(K)* (7.212) (see eqn (7.100)). Therefore if ty/tg=M/M, is not sufficiently large, which is the case in many experiments with monodisperse samples, the minimum of g,,(x) will not be observed and the flow will be stable. (iii) On the other hand if the system is monodisperse and if M/M, is large enough, the theory predicts the shear stress shown in Fig. 7.22. Such behaviour has indeed been proposed by Vinogradov and by Ball and McLeish,” to interpret the finding that the flow rate of polymers through pipes changes abruptly as the shear rate is raised if the molecular weight of the polymer is high and has a narrow distribution. Though the log oy, (x) Ke = Vi log x Fig. 7.22. Shear stress predicted by the theory for monodisperse systems with M/M, large.270 MOLECULAR THEORY OF POLYMERIC LIQUIDS detailed analysis has not been done, it should be worthwhile to study the phenomenon under well-controlled conditions. Stress overshoot. According to eqn (7.195), the stress maximum at the start of the shear flow appears in the shear stress, but not in the first normal stress difference N,(t, K) = Oz,(t; K) — Gyy(t; K),* whilst experi- mentally the maximum is often observed in N,(t;K). This is possibly due to the elongation of the contour length. Indeed the overshoot in the normal stress appears at a higher shear rate than in the shear stress." That the elongation of the contour length is important under usual flow conditions is indicated by the stress relaxation after the steady shear flow. It has been observed™ that when the shear rate becomes larger than 1/tg, the relaxation curves begin to show a short-time component which corresponds to the relaxation in the contour length. 7.8 Stress relaxation after double step strain Though the BKZ-type constitutive equation has been quite successful in many phenomena, it has been reported that under certain flow history, the equation gives unsatisfactory predictions. One such experiment is the stress relaxation after application of double step strain.'*" The flow history of this experiment is illustrated in Fig. 7.23. Two step shears y, and y, are applied with time interval ¢,, one at time —t, and the other at time 0. The BKZ equation (7.198) predicts Oap(t) = Paplt, Y2) + Paplt +t, Y2t 1) ~ Paplt +t, Y2)- (7.213) wt) Fig. 7.23. Double step strain experiments.STRESS RELAXATION AFTER DOUBLE STEP STRAIN 271 According to Osaki et al.,%° eqn (7.213) predicts the stress with reasonable accuracy when 7,72 > 0, i.e, when the sense of the two shears are the same, while when y,y,<0, a large discrepancy is found. It has been shown’ that this discrepancy is caused by the IA approximation and the rigorous analysis of the model gives good agreement with experiments. We consider the inextensible chain model. Figure 7.24 explains the change of polymer conformation under the double step strain. Figure 7.24a shows the undeformed state just before the first deformation. Figure 7.24 represents the state immediately after the deformation: the primitive chain is deformed by the shear y;. Figure 7.24c indicates the state just before the second deformation; the inner part AB still remains in the deformed tube, while the outer parts are in the undeformed tube. Now when the second deformation is applied, the inner part AB is deformed by the shear y, + y2 from the equilibrium state, while the outer part is deformed by the shear 2. It is important to note that the second shear stretches the contour length of the outer part by the factor a(y2), but that of the inner part by the factor B= (yi + Y2)/a(y), (7.214) since the inner part is already stretched by the factor a(y,). If the coordinates of A and B are s, and sz, respectively, the coordinates of A’ and B’ are fs, and Bs2. Hence the probability that a primitive chain segment s is between A’B’ is equal to the probability that it is between time @) -t-0 o -.. LY [jo Zn BR Bn Fig. 7.24. Microscopic process in the relaxation of double step strain. The deformation of the primitive chain at various times are shown. The deformation of the primitive chain relative to the equilibrium state is shown by oblique lines.272 MOLECULAR THEORY OF POLYMERIC LIQUIDS AB at time t= —0, and is given by (s/f, t,). Hence the average of Sap(s, +0) is given byt p(s/B, t1)Qap (ys + ¥2) + [1 — (s/B, 4,)]Qap(y2) Sup(s, +0) = |s/B| < L/2, als, +0) {onus \sie|>Lr2. ©7219) For t>0, S,p(s, £) satisfies a e 9 500 (5) 1) = Dea Seals, 6) (7.216) and the boundary condition Sap(s, t)=O at s= +e. (7.217) Hence S,9(s, ¢) is given by 12 Sap(s, t) = | ds’G(s, 5’, t)Sap(s', +0) (7.218) -L2 where 25... (pa *))si pal, )) 2 jak pr & px & — G(s, s', t) Le sin( L (s+ 2 sin( L (s + 3 jexp(—p7t/ta). (7.219) From eqns (7.178), (7.218), and (7.219), the stress at time ¢(t>0) is given by Ln “Le Oup(t) = Get i ds’ | dsG(s, 5’, t)Sap(s", +0) ~L2 LR ~L2 =G.z [| As’ V6s', DS", +0). (7.220) -La Substituting eqn (7.215) we have Oap(t) = GeQap(Y2) P(t) + GelQaa(¥1 + ¥2) — Qaa(y2)]Y'(t, 4, B) (7.221) +Here it is assumed that if |s/8| is larger than L/2, the primitive chain segment s is deformed by y2. Strictly speaking this is only approximately correct. Actually there is a small correction term to eqn (7.215),!”” which is neglected here because it is numericallySTRESS RELAXATION AFTER DOUBLE STEP STRAIN 273 where ne i [ ave. )y(s/B,t1) for B<1, VG B=4 Ain (7.222) 1 [ ave. )ws/B,)) for B>1. L -in Using eqn (6.14) and doing the integral, we get 2 _yp+q-aya Sinl(/2)(q — PB)] Pea yn pq’ - p’B’) —(p7i 2, al fe 1, ven Bad “MMO Payed for P< 32 > (—)ote-2y Sin /2)(0 — 4/8) prod qe" - 7/6?) x exp[—(p7t+ q7t)/ta] for B>1. To simplify the equation we consider the case of large t and t,. If t> ty and t, > ty, only the first term in the sum of eqn (7.223) is important and '(t, t1, B) is approximated by SE exol-t +t)/ta] for B<1 v'G ty B= (7.224) a exp[-(¢+4)/ta] for B>1 or it may be written as ¥'(b tr, B)= AB) VE +4) (7.225) where ee for B<1, A(B) = (7.226) 4 12, ieee for B>1. Although eqn (7.225) is obtained under the condition t> t, and t, > ty, it turns out that eqn (7.225) is actually a good approximation for the entire regime of ¢ and t,.'7 If eqn (7.225) is used, eqn (7.221) is written as Fup (t) = Paplts Y2) + ACB) Pop(t + tr» Ya + ¥2)— Paplt +t, ¥2)). (7.227) Equation (7.227) has been thoroughly checked by Osaki et al. An example is given in Fig. 7.25. This indicates that the theory described in274 MOLECULAR THEORY OF POLYMERIC LIQUIDS 107| 10° 10' 10° t (sec) Fig. 7.25. Shear stresses for double step shear deformation. —y, = y. = 11.6, and t, is indicated in the figure. Sample polystyrene solution in diethyl phthalate M=3.10x 10° and p =0.221gcm *. The heavy lines represent stress for single step deformation. The light solid line represents eqn (7.227) and the light broken lines the result of the BKZ equation (eqn (7.213)). Reproduced from ref. 68. Section 7.7 correctly reflects the reality of polymer dynamics in an entangled state. 7.9 Rigorous constitutive equation for reptation model Having seen that the IA approximation causes a serious error in certain situations, we now derive a constitutive equation without using the IA approximation. "°° In a small time-interval At, E(t + At, ¢) is given by E(t + At, t)=1+ «(t)At. (7.228) Thus the transformation rule described by eqn (7.171) is written as R(, t+ Ad) =R(s, t) + K(t)- R(s, DAt=R(s, t) + x(t) RG, At (7.229)CONSTITUTIVE EQUATION FOR REPTATION MODEL 275 The second equality holds since s — § is of order At. Similarly eqn (7.172) becomes s 5= [av |a(s’, t) + x(t) + u(s', t)At| oO | ds'(1 + «(t):u(s', t)u(s’, t)At) + O(AP) =stAt | ds’x(t):u(s', u(s', 2). (7.230) 0 Therefore, to the order of At, s is expressed by § as & s=5— At} ds'x(t):u(s’, t)u(s', t) 0 =5— At&(, t) (7.231) where &s,)= | ds'x(t):u(s’, t)u(s’, t). (7.232) 0 From eqns (7.229) and (7.230), the change in the tangent vector u(s, ) becomes u(5, t+ At) ~2RG, t+ At)= 2 (R(s, t) + K(f) RG, t)At) os 3 3 = peas RO t+K(o- ag. tat =(1- ars & 0))ms, t)+x(t)-u(% NAL — (7.233) Using eqn (7.232), u(S, t + At) =u(s, t) —[K(t): a(S, Ou, tu(s, 0) — x(t) - a, O)At =u(s, t) —[a(t):u(5, Dus, Du, t)— KQ)- u(G, At + O(A) =u(s, t) +1 (u(5, 1), At (7.234) where I'(u, t) is given by eqn (7.189) In eqn (7.234) the effect of Brownian motion was not taken into account. If this is included, the final equation becomes u(S, t+ At)=u(s + AE, t)+T(u(s, 1), OAt =u(5 — £5, At + AE, )+T(u(s, 1), At, (7.235)276 MOLECULAR THEORY OF POLYMERIC LIQUIDS or replacing 5 by s u(s, t+ At)=u(s — E(s, t)At+ AE, t)+T(u(s, ), At. (7.236) This is the time-evolution equation for the tangent vector u(s, t) of the inextensible primitive chain. Thus the equation for f(u, s, t) is f(u, s, t+ At) = (d[u —u(s, t+ A2)]) = (6[u —u(s — E(s, t)At + A&(t), t) —T(u(s, t), At) = (d[u — u(s — E(s, t)At + AE(2), £)]) ~ are - (u, Of, 0). (7.237) The first term is written X= (d[u—u(s — &(s, t)At + AED, D]) =([1 +ag 2488 2 at? lou -u(s, )) e a = [: + pea |fu, s,t)—-At (2 d[u —u(s, ol). (7.238) The underlined term is rewritten as Y= (é2 d[u—u(s, a) -2 (E6{u—a(s, ))) - (ou als, 2)] 38), (7.239) Since the correlation between u(s, t) and u(s', t) decreases quickly with an increase in |s ~ s‘|, the first average in eqn (7.239) becomes (&d[u — u(s, t)]) = [asrxey: (u(s', thu(s', t)d[u — u(s, t)]) m= fase: (u(s', t)u(s’, )) (d[u —u(s, 2))) 0 = [asrac dats’, t)u(s’, t))f(u, s, t). (7.240) 0 From eqn (7.232) it follows that 3 =K(t):u(s, Du(s, t). (7.241)CONSTITUTIVE EQUATION FOR REPTATION MODEL 277 From eqns (7.239)-(7.241), one has y=2 [favrace: cue t)u(s’, t))f(u, s, | asl} , , ” — (x(t): u(s, t)u(s, t)d6[e — u(s, t)]) = a():(uls, Quls, D)f(e, 5,4 (EG, 0) Zhlw, 5, —K(t): uuf(u, s, t). (7.242) Hence the time evolution equation for f(u, s, t) is obtained ast a 2 (0. 5- (86.9) 2\pte,5,)-Z (ew, ftw, 5.) + x(t): (ua — (u(s, thu(s, t)))f(u, s, t). (7.243) The average in eqn (7.243) can be expressed by f(u, s, t) as (6, )) = [ ds" | dux(t):wuf(a, s', 0) (7.244) and ° (ug(s, t)up(s, t)) = | duu, ugf(u, s, t). (7.245) Hence eqn (7.243) is a nonlinear integro-differential equation for Fu, s, t). Equation (7.243) can be rewritten into more tractable form. By a similar technique described in ref. 108, eqn (7.243) can be transformed into a closed equation for S.g(s, t): Sup(S, 1) = | a'(2 KG, 11°) Qap(EC 1) (7.246) where K(s, t, t’) is the solution of the differential equation a e 3 , (5- DeSa+ (Es,9) <)KG. tr)=0 (7.247) with the initial condition K(s,t,’)=1 at t=0’ (7.248) +In ref. 108 the terms in the last parenthesis are erroneously omitted. This gives a constitutive equation which includes Q°)(E) instead of Q(E) in eqn (7.246). Since the difference in QE) and Q(E) is small, the error caused by this is not serious.278 MOLECULAR THEORY OF POLYMERIC LIQUIDS and the boundary condition K(s,t,t')=0 at s=—-L/2 and s=L/2. (7.249) Finally (&(s, 2) is given by (86.9) = | ds" )Su066'. 1). (7.250) 0 Equations (7.246)-(7.250) determine S,,(s, t). Given S,g(s, t), the stress can be calculated by eqn (7.178). In the special case of step strain, one can solve the set of equations (7.246)-(7.250) rigorously and obtain the results given in eqn (7.122). In the general case, the solution of the equation needs numerical calcula- tion. It turns out that the difference between eqns (7.194) and (7.246) is not large for the usual flow history discussed in Section 7.6. For such flows, the simple constitutive equation will be useful. The effect of the IA approximation has also been examined for large amplitude oscillatory shear deformation.’ In this case the result of the constitutive equation without using the IA approximation is shown to be in better agreement with experimental results.""° It has been shown by Marrucci"!™ and Marrucci and Grizzuti that analysis at the level of accuracy of this section is required to derive Weissenberg effect correctly. 7.10 Further applications Here we shall briefly discuss some pending problems which have not been discussed in the previous sections. 7.10.1 Branched polymers As discussed in Section 6.4.5, reptation is severely suppressed if the polymer has long branches. Indeed it has been observed that the dynamical properties of branched polymers are quite distinct from those of linear polymers. So far studies have been done for branched polymers of the simplest type, the star-shaped polymer in which f chains are connected to a centre. The observed phenomena are: (i) The diffusion constant Dg of a star polymer in a high molecular weight matrix is much smaller than that of a linear polymer of the same molecular weight,'" and the molecular weight dependence of Dg is much stronger than that of linear polymers. This is consistent with the prediction of the reptation theory’? (eqn (6.118)). +The same constitutive equation has recently been derived by G. Marrucci (J. Non- Newtonian Fluid Mech, to appear) by a different method.FURTHER APPLICATIONS 279 LE Wipes imal (a) (o) (2) Fig. 7.26. Relaxation of a star polymer. Figures show the states (a) before the deformation, (b) immediately after the deformation, (c) at a later time ¢. The deformed part of the tube which contributes to the stress is denoted by oblique lines. plateau region and the steady-state compliance J® is Rouse like (i.e., proportional to M) even if the molecular weight becomes quite high.’" The anomalous behaviour in the linear viscoelasticity has been ex- plained by the tube model."2""56 Figure 7.26 shows schematically how the stress relaxation takes place in star polymers. In the crude theory,'’* it is assumed that the centre of the star is fixed during the viscoelastic relaxation time and that the relaxation takes place only by the contour length fluctuation, i.e., by the process that the polymer retracts its arm down the tube and evacuates from the deformed tube as shown in Fig. 7.26. Let y(s,f) be the probability that the tube segment s which is separated from the centre by the contour length s still remains at time ¢, then the relaxation modulus is written as GW) = GY x [ove. ) (7.251) 7% where Z, is the equilibrium length of the tube for the arm of the star polymer, and the constant G{? can be identified, in a first approximation, with the plateau modulus for linear polymers: pRT GM = (7.252) A simple approximation for #(s, ¢) is w(s, t) = exp(—t/r(s)) (7.253) where t(s) is the average time at which the chain end first reaches the tube segment s, i.e., the contour length L,(¢) first becomes equal to s. As280 MOLECULAR THEORY OF POLYMERIC LIQUIDS was discussed in Section 6.4.5, the motion of L,(t) can be regarded as a Brownian motion of a particle in a harmonic potential 3kpT 2N,b” where N, is the number of Rouse segments in the arm. Hence the time 1(s) is estimated as U(L,) = (L,-L,) (7.254) (£. -s) t(s) = D ° exp((U(L, =s) — U(L,)\VkpT). (7.255) The maximum relaxation time is given by rc 3. 5N2 b4 b\? Tmax = 15 = 0) =" exe 5552 i) kya exp[3n.(2) | (7.256) Equation (7.255) is then written as b 2 1(s) = Tmax ~ 8)? exp(3N.(2) (2-28) (7.257) where E=s/L,. (7.258) Consider the case 2 a=in,(2) >1, (7.259) then the viscosity is evaluated as ~ Le 1 no= | JG) = GPz- | dst(s) = G9 tax | dE (1 - §)* exp(-2a8 + a8%) = GO tex d& exp(—2aé) = x G2 an. (7.260) 0 Similarly, the steady-state compliance is obtained as 1 a (0) — =— = | a(n = 5. (7.261) Since the molecular weight of an arm is M/f, it follows from eqns (7.54) and (7.259) 15M oa (7.262)FURTHER APPLICATIONS 281 Thus eqns (7.260) and (7.261) are written as No® (#) ox (Sa) (7.263) and O) =e SER (7.264) The results of eqns (7.263) and (7.264) are in qualitative agreement with experimental results: the viscosity increases steeply because of the exponential factor, and the steady state compliance is proportional to M. However, the quantitative agreement is not satisfactory. The observed viscosity is smaller than the calculated one, and the best fit with experiments is obtained only when the numerical coefficient in the exponential of eqn (7.263) is replaced by a smaller number (about 1/2) instead of 15/8.16 This suggests that relaxation mechanisms other than the contour length fluctuations are important for star polymers. Indeed it has been pointed out” that in the case of star polymers the constraint release, and perhaps other tube reorganization processes, are as impor- tant as the contour length fluctuation. That the tube reorganization is important for star polymers is indicated by another experiment. Kan et al.’’ found that the relaxation time of a star polymer dispersed in a crosslinked system is by orders of magnitude larger than that in the melt, while for linear polymers the former is larger only by a factor of 2 or 3. The constraint release or other mechanisms of the tube reorganization are supposed to be important in other branched polymers such as H-shaped polymers’ or ring polymers." Theoretical prediction for the theological properties of these polymers is interesting and challenging. 7.10.2 Molecular weight distribution Various experimental data suggest that the tube reorganization is im- portant in linear polymers with molecular weight distribution. (i) The diffusion constant of a polymer (of molecular weight M) in a matrix (of molecular weight P) has been found to be essentially independent of P if P is larger than a certain value P, which is between M and M,.°!° This indicates that the tube reorganization is weak in monodisperse systems (M = P). On the other hand, if P becomes smaller than P., the diffusion constant increases with decreasing P.'?°!71 (ii) The linear viscoelasticity of a mixture of two polymers of the same282 MOLECULAR THEORY OF POLYMERIC LIQUIDS Fig. 7.27. Discrepancy between experimental results and eqn (7.265). species, but different molecular weights M, and Mg (M,> Ms) is not explained by the model which includes only reptation. According to the fixed tube model, the relaxation modulus of the mixture is the weight average of that of the pure melts of individual components: GO (0) = Wa Ga(t) + Wa Galt) (7.265) where Pa PB w=——, =. 7.266) “pats “ Pat PB ( ) The discrepancy between the experimental results” and eqn (7.265) is schematically explained in Fig. 7.27: G™™(t) shows two characteristic telaxations, each corresponding to the disengagement of polymer A and B. Though this feature is in agreement with eqn (7.265), the relaxation time of the larger component tg, is shorter than that in the pure A component 1}, and the plateau modulus for the larger polymer is lower than expected from eqn (7.265). Kurata’ suggested that the experimen- tal data can be fitted by GOD) = (1 - w)Ga(t) + WAGa(t/ ma). (7.267) These results clearly indicate that the tube constraint for a polymer becomes weaker if it is made of shorter polymers. The weakening of the tube can be expressed either by an increase in the step length," or by an increase in the constraint release process,°* or both.» 7° However, the interpretation seems to be still at a tentative level. 7.10.3 Future problems The reptation model has been applied to various problems other than the problems of viscoelasticity and diffusion that have been discussed. TheseREFERENCES 283 include dielectric relaxation,” spinodal decomposition," polymer— polymer welding,“°*!_ diffusion controlled _ reaction,“ and crazing."%*"55 A concise review of various applications is given by de Gennes and Léger.'* On the whole the reptation model works well qualitatively, and for several problems it gives quantitatively successful predictions. However, many problems remain unsolved. Perhaps the most important problem is the tube reorganization. We have seen that the tube reorganization is important in branched polymers and in linear polymers with polydispersity. It will also be important in a nonuniform system such as polymer mixtures. So far the reptation theory is based on the assumption that there is a tube which is characterized by a single parameter a, the step length of the tube. Though the outcome of this simple assumption is quite fruitful, one could ask: to what extent is this picture correct? A complete answer to this question will be given when the tube is derived from more basic equations such as eqn (5.84) by a kind of mean field approximation. This will require a new development of statistical mechanics since the tube is a dynamical concept rather than static. (Notice that the mean force acting on the polymer vanishes if it is averaged over a time longer than ty, so that the average of the surrounding field must be taken over a finite time.) Perhaps the tube is better understood as representing the effect of dynamical correlation of the environment rather than the usual mean field. A slightly different, but closely related, problem is rubber elasticity. Here the dynamical problem does not arise since the topological constraints are permanent. However, the correlation plays an essential role in the problem. Indeed it is the correlation in the topological structure between the undeformed state and the deformed state that gives rise to the rubber elasticity. In the modern theory of rubber elasticity, this correlation is neatly handled by the replica method.'°7** Generali- zation of this method to dynamical problems might be quite useful. On the other hand, apart from that purely theoretical approach, it will be quite promising to develop a theory by closely studying experimental results. Collaboration between experiment and theory will be essential for further progress. References 1. (19), J. D., Landel, R. F. and Williams, M. L., J. Appl. Phys. 26, 359 2. Ferry, J. D., Viscoelastic Properties of Polymers (3rd edn). Wiley, New York (1980). 3. Graessley, W. W., Adv. Polym. Sci. 16, 1 (1974).aus Seo 11. 12. 13. 14. 15. 16. 17. 18. 19. MOLECULAR THEORY OF POLYMERIC LIQUIDS . Baur, M. E., and Stockmayer, W. H., J. Chem. Phys. 43, 4319 (1965). . Higgins, J. S., Nicholson, L. K., and Hayter, J. B., Polymer 22, 163 (1981). . Richter, D., Hayter, J. B., Mezei, F., and Ewen, B., Phys. Rev. Lett. 41, 1484 (1978). . Baumgartner, A., Kremer, K., and Binder, K., Faraday Symp. Chem. Soc. 18, 37 (1983), . Klein, J., Nature (London), 271, 143 (1978); Phil. Mag. 443, 771 (1981). . Klein, J., and Briscoe, B. J., Proc. R. Soc. London A365, 53 (1979). . Hervet, H., Léger, L., Rondelez, F., Phys. Rev. Lett. 42, 1681 (1979); Léger, L., Hervet, H., and Rondelez, F., Macromolecules 14, 1732 (1981). Tanner, J. E., Macromolecules 4, 748 (1971); Tanner, J. E., Liu, K. J., and Anderson, J. E., Macromolecules 4, 586 (1971). Bachus, R., and Kimmich, R., Polymer 24, 964 (1983). Tirrell, M., Rubber Chem. Tech. 57, 523 (1984). Doi, M., and Edwards, S. F., J. C. S. Faraday Trans. 274, 1802 (1978). Doi, M., and Edwards, S. F., J. C. S. Faraday Trans. 274, 1818 (1978). Walters, K., Rheometry. Chapman & Hall, London. New York (1975). Astarita, G., and Marrucci, G., Principles of Non-Newtonian Fluid Mechanics, McGraw-Hill, London (1974). Lodge, A. S., Elastic Liquids. Academic Press, London (1964). Bird, R. B., Armstrong, R. C., Hassager, O., and Curtiss, C. F., Dynamics of Polymeric Liquids, Vols. 1, 2. Wiley, New York (1977). . See for example Treloar, L. R. G., Rep. Prog. Phys. 36, 755 (1973); and Treloar, L. R. G., The Physics of Rubber Elasticity, (3rd edn). Clarendon Press, Oxford (1975). . Green, M. S., and Tobolsky, A. V., J. Chem. Phys. 14, 80 (1946). . Yamamoto, M., J. Phys. Soc. Jpn 11, 413 (1956), 12, 1148 (1957); 13, 1200 (1958). Lodge, A. S., Rheol. Acta 7, 379 (1968). - Coleman, B. D., Arch. Ratl. Mech. 17, 1 (1964), 17, 230 (1964). Truesdell, C., and Noll, W., The nonlinear field theories of mechanics, in Encyclopedia of Physics III/3. Springer (1965). |. Janeschitz-Kriegl, H., Polymer Melt Rheology and Flow Birefringence. Springer, New York (1983). . Janeschitz-Kriegl, H., Adv. Polym. Sci. 6, 170 (1969). . Historical development of the stress optical law is described in refs 20 and 24. A recent study is given by Wales J. L. S., The Application of Flow Birefringence to Rheological Studies of Polymer Melts. Delft Univ. Press, Rotterdam (1976). . DiMarzio, E. A., J. Chem. Phys. 3, 1563 (1962). . Fukuda, M., Wilkes, G. L., and Stein, R. S., J. Polym. Sci. A2, 9, 1417 (1971). . Jarry, J. P., and Monnerie, L., Macromolecules 12, 316 (1979). . Tobolsky, A. V., Properties and Structure of Polymers. Wiley, New York (1960). . Onogi, S., Masuda, T., and Kitagawa, K., Macromolecules 3, 109 (1970). . Doi, M., Chem. Phys. Lett. 26, 269 (1974). . Berry, G. C., and Fox, T. G., Adv. Polym. Sci. 5, 261 (1968). . Casale, A., Porter, R. S., and Johnson, J. F., J. Macromol. Sci. Rev. Macromol. Chem. CS, 387 (1971).35. 36. 37. 38. 39. 40. 51, 52. 33. RESVSSSSIRAL & REFERENCES 285 Odani, H., Nemoto, N., and Kurata, M., Bull. Inst. Chem. Res. Kyoto Univ. 50, 117 (1972). Graessley, W. W., J. Polym. Sci. 18, 27 (1980). van Krevelen, D. W., Properties of Polymers, p. 338. Elsevier, Amsterdam (1976). Graessley, W. W., and Edwards, S. F., Polymer 22, 1329 (1981). Bartels, C. R., Crist, B., and Graessley, W. W., Macromolecules 11, 2702 (1984). Kremer, K., Macromolecules 16, 1632 (1983). . A recent review on the computer simulation for polymer dynamics is given by Baumgirtner, A., Ann. Rev. Phys. Chem. 35, 419 (1984). . Deutsch, J. M., Phys. Rev. Lett. 49, 926 (1982). . Higgins, J. §., and Roots, J. E., J. C. S. Faraday Trans II 81, 757 (1985). . de Gennes, P. G., Macromolecules 9, 587, 594 (1976). - Adam, M., and Delsanti, M., J. Phys. (Paris) 44, 1185 (1983). . Raju, V. R., Menezes, E. V., Marin, G., Graessley, W. W., and Fetters, L. J., Macromolecules 14, 1668 (1981). . Onogi, S., Masuda, T., Miyanaga, N., and Kimura, Y., J. Polym. Sci. Part A2 5, 899 (1967); Onogi, S., Kimura, S., Kato, T., Masuda, T., and Miyanaga, N., J. Polym. Sci. C15, 381 (1966). . Masuda, T., Toda, N., Aoto, Y., and Onogi, S., Polymer J. 3, 315 (1972). . Nemoto, N., Ogawa, T., Odani, H., and Kurata, M., Macromolecules 5, 641 (1972). ). See for example, Osaki, K., Fukuda, M., and Kurata, M., J. Polym. Sci. Phys. ed. 13, 775 (1975). Recent data for melt are given by Lin, Y. H., Macromolecules 17, 2846 (1984); J. Rheol. 28, 1 (1984). Graessley, W. W., Faraday Symp. Chem. Soc. 18, 7 (1983). Bernard, D. A., and Noolandi, J., Macromolecules 15, 1553 (1982); 16, 548 (1983). Wendel, H., and Noolandi, J., Macromolecules 15, 1318 (1982); Wendel, H., Colloid Polym. Sci. 259, 908 (1981). Curtiss, C. F., and Bird, R. B., J. Chem. Phys. 74, 2016, 2026 (1981). . Ball R., private communication. Deutsch, J. M., Phys. Rev. Lett. 54, 56 (1985). Doi, M., J. Polymer Sci. 21, 667 (1983); J. Polym. Sci. Lett. 19, 265 (1981). Lin, Y. H., Macromolecules 19, 159, 168 (1986). Klein, J., Macromolecules 11, 852 (1978). Daoud, M., and de Gennes, P. G., J. Polym. Sci. Phys ed. 17, 1971 (1979). Graessley, W. W., Adv. Polym. Sci. 41, 67 (1982). Marrucci, G., J. Polym. Sci. Phys. 23, 159 (1985). Viovy, J. L., J. Physique Lett. 46, 847 (1985). . Doi, M., J. Polym. Sci. 18, 1005 (1980). . Einaga, Y., Osaki, K., Kurata, M., Kimura, S., and Tamura, M., Polymer. J. 2, 580 (1971); Fukuda, M., Osaki, K., and Kurata, M., J. Polym. Sci. Phys. 13, 1563 (1975). Osaki, K., Bessho, N., Kojimoto, T., and Kurata, M., J. Rheol. 23, 617 1979). . Osaki, K., Kimura, S., and Kurata, M., J. Polym. Sci. Phys. ed. 19, 517 (1981).Be $8 MOLECULAR THEORY OF POLYMERIC LIQUIDS Osaki, K., and Kurata, M., Macromolecules 13, 671 (1980). Osaki, K., Nishizawa, K., and Kurata, M., Macromolecules 15, 1068 (1982). . Takahashi, M., Nakamura, H., Masuda, T., and Onogi, S., Polymer Preprints Japan 30, 1970 (1981). . Lodge, A. S., and Meissner, J., Rheol. Acta, 11, 351 (1972); Lodge, A. S., ibid. 14, 664 (1975). . Marrucci, G., and de Cindio, B., Rheol. Acta, 19, 68 (1980). . Taylor, C. R., Greco, R., Kramer, O., and Ferry, J. D., Trans. Soc. Rheol. 20, 141 (1976); Noordermeer, J. M., and Ferry, J. D., J. Polym. Sci. 14, 509 (1976). . Vrentas, C. M., and Graessley, W. W., J. Rheol. 26, 359 (1982); Pearson, D. S., IUPAC Proceedings, 28th Macromolecular Symposium, July, p. 866 (1982). . Marrucci, G., and Grizzuti, N., J. Rheol. 27, 433 (1983). . McLeish, T. C. B., and Ball, R. C., J. Polym. Sci., to be published. . Marrucci, G., and Hermans, J. J., Macromolecules 13, 380 (1980). . Viovy, J. L., Monnerie, L., and Tassin, J. F., J. Polym. Sci. Phys. ed. 21, 2427 (1983); Viovy, J. L., J. Polym. Sci. Phys. ed. 23, 2423 (1985). . Boué, F., Adv. Polym. Sci. to be published; Boué, F., Nierlich, M., and B82 8 S$ BSR R SSRAS 98. Osaki, K., Faraday Symp. Chem. Soc. 18, 83 (1983): Boué, F., Nierlich, M., Jannink, G., and Ball, R. C., J. Phys. (Paris) 43, 137 (1982); J. Phys. Lett. 43, L585, L593 (1982). Bastide, J., Herz, J., and Boué, F., J. Physique 46, 1967 (1985). Sekiya, M., and Doi, M., J. Phys. Soc. Jpn 51, 3672 (1982). Noolandi, J., and Hong, K. M., J. Physique Lett, 45, L149 (1984). Boué, F., Osaki, K., and Ball, R. C., J. Polym. Sci. 23, 833 (1985). Takahashi, M., Masuda, T., Bessho, N., and Osaki, K., J. Rheol. 24, 517 (1980). Takahashi, M., Masuda, T., Oono, H., and Onogi, S., Polymer Preprints Japan 33, 871 (1984). Meissner, J., Rheol. Acta, 10, 230 (1971). Currie, P. K., J. Non-Newtonian Fluid Mech. 11, 53 (1982). on B., Kearsley, E. A., and Zapas, L. J., Trans. Soc. Rheol. 7, 391 1963). Osaki, K., Ohta, S., Fukuda, M., and Kurata, M., J. Polym. Sci. A14, 1701 (1976). Chang, W. V., Bloch, R., and Tschoegl, N. W., Rheol. Acta 15, 367 (1976); J. Polym. Sci. A18, 923 (1977). . Wagner, M. H., Rheol. Acta 15, 136 (1976); 16, 43 (1977). Phillips, M. C., J. Non-Newtonian Fluid, Mech. 2, 109, 123, 139 (1977). . Osaki, K., Proceeding of the 7th International Congress on Rheology (eds C. Klason and J. Kubat). Chalmers University of Technology, Gothenberg, p. 104 (1976). . Doi, M., and Edwards, S. F., J. C. S. Faraday Trans. 275, 38 (1979). . Bird, R. B., Saab, H. H., and Curtiss, C. F., J. Phys. Chem. 86, 1102 (1982); J. Chem. Phys. 7, 4747, 4758 (1982). Osaki, K., and Doi, M., Polym. Eng. Rev. 4, 35 (1984). . Marrucci, G., Adv. Transport Processes § (eds. A. S. Mujumdar and R. A. Mashelkar). New York (1985). Camachandran, S., Gao, H. W., and Christiansen, E. B., J. Rheol. 25, 213 981).101. 102. 103. 104. 105. 106, 107. 108. 109. 110. REFERENCES 287 . Tanner, R. I., Trans. Soc. Rheol. 17, 365 (1973). 100. Laun, H. M., and Munstedt, H., Rheol. Acta 18, 427 (1979); Munstedt, H., J. Rheol. 23, 421 (1979). Vinogradov, G. V., Rheol. Acta 12, 273 (1973); see also Lin, Y. H., J. Rheol. 29, 65 (1985). Menezes, E. V., and Graessley, W. W., J. Polym. Sci. Phys. 20, 1817 (1982). Takahashi, M., Masuda, T., and Onogi, S., Polymer Preprints Japan 29, 1807 (1980). Osaki, K., and Kurata, M., J. Polym. Sci. Phys. 18, 2421 (1980). Zapas, L. J., Deformation and Fracture of High Polymers (eds H. H. Kausch, J. A. Hassell, and R. I. Jaffe). Plenum Press, New York (1974); see also McKenna, G. B., and Zapas, L. J., J. Rheol. 23, 151 (1979); 24, 367 (1980). Osaki, K., Kimura, S., and Kurata, M., J. Rheol. 25, 549 (1981). Doi, M., J. Polym. Sci. 18, 1891 (1980). Doi, M., J. Polym. Sci. 18, 2055 (1980). Helfand, E., and Pearson, D. S., J. Polym. Sci. 20, 1249 (1982). Pearson, D. S., and Rochefort, W. E., J. Polym. Sci. Phys. ed. 20, 83 (1982). 110a. Marrucci, G., J. Non-Newtonian Fluid Mech. 21, 329-36 (1986). 110b. Marrucci, G. and Grizzuti, N., J. Non-Newtonian Fluid Mech, 21, 319-28 111. 112. 113. 114. 115. 116. 117. 118. 119. . Smith, B. A., Samulski, E. T., Yu, L. P., and Winnik, M. A., Phys. Rev. 121. 125. 126. (1986). Klein, J., Fletcher, D., and Fetters, L. J., Faraday Symp. Chem. Soc. 18, 159 (1983). de Gennes, P. G., J. Phys. (Paris) 36, 1199 (1975). Kraus, G., and Gruver, J. T., J. Polym. Sci. A3, 105 (1965); J. Appl. Polym. Sci. 9, 739 (1965). Graessley, W. W., Masuda, T., Roovers, J. E. L., and Hadjichristidis, N., Macromolecules 9, 127 (1976); Graessley, W. W., and Roovers, J., Macromolecules 12, 959 (1979); Raju, V. R., Menezes, E. V., Marin, G., Graessley, W. W., and Fetters, L. J., Macromolecules 14, 1668 (1981). Doi, M., and Kuzuu, N., J. Polym. Sci. Lett. 18, 775 (1980). Pearson, D. S., and Helfand, E., Macromolecules 17, 888 (1984). Kan, H. C., Ferry, J. D., and Fetters, L. J., Macromolecules 13, 1571 (1980). Roovers, J., Macromolecules 17, 1196 (1984). Roovers, J., J. Polym. Sci. 23, 1117 (1985). Lett, 52, 45 (1984). Green, P. F., Mills, P. J., Palmstrgm, C. J., Mayer, J. W., and Kramer, E. J., Phys. Rev. Lett. 53, 2145 (1984). . Masuda, T., Takahashi, M., and Onogi, S., Appl. Polym. Symp. 20, 49 (1973); Bogue, D. C., Masuda, T., Einaga, Y., and Onogi, S., Polym. J. 1, 563 (1970). ba Kurata, M., Macromolecules 17, 895 (1984). Montfort, J. P., Marin, G., and Monge, P., Macromolecules 17, 1551 (1984). Watanabe, H., and Kotaka, T., Macromolecules 17, 2316 (1984). Masuda, T., Yoshimatsu, S., Takahashi, M., and Onogi, S., Polymer Preprints Japan 33, 2699 (1984).288 127, 128. 129. 130. 131, 132. 133. 134. 135, 136. 137. 138. MOLECULAR THEORY OF POLYMERIC LIQUIDS Adachi, K., and Kotaka, T., Macromolecules 17, 120 (1984); 18, 466 (1985). de Gennes, P. G., J. Chem. Phys. 72, 4756 (1980). Pincus, P., J. Chem. Phys. 75, 1996 (1981). de Gennes, P. G., C. R. Acad. Sci. Paris B291, 219 (1980). Prager, S., and Tirrell, M., J. Chem. Phys. 75, 5194 (1981); Adolf, D., Tirrell, M., and Prager, S., J. Polym. Sci. 23, 413 (1985). Tulig, T. J., and Tirrell, M., Macromolecules 14, 1501 (1981). de Gennes, P. G., J. Chem. Phys. 16, 3316, 3322 (1982). Kramer, E. J., Adv. Polym. Sci. 52/53, 1 (1983). Evans, K. E., and Donald, A. M., Polymer 26, 101 (1985). de Gennes, P. G., and Léger, L., Ann. Rev. Phys. Chem. 33, 49 (1982). Deam, R. T., and Edwards S. F., Phil. Trans. R. Soc. A280, 317 (1976). Ball, R. C., Doi, M., Edwards S. F., and Warner, M., Polymer 22, 1010 (1981).8 DILUTE SOLUTIONS OF RIGID RODLIKE POLYMERS 8.1 Rodlike polymers Though many polymers are flexible and take a random coil structure, there is a large class of polymers which are not flexible and assume a rodlike structure. For example, some polypeptides or polynucleotides form a helix structure which can be regarded effectively as a rigid rod. If the chemical bonds in the backbone chain consist of double bonds or phenylene rings, the internal rotation of the polymer is severely restricted and the polymer takes an elongated form. These latter type of rodlike polymers are quite important in polymer technology because of their capability of creating very strong fibres, and an increasing amount of research is being done as a result. The physical properties of the rodlike polymers differ from those of flexible polymers in many respects. Firstly, an obvious distinction is that rodlike polymers are much larger than flexible polymers with the same molecular weight. If the polymer is a straight rod, its radius of gyration R, is proportional to the contour length of the polymer, or the molecular weight M, as compared to the relation R,« M” (v~0.6) for flexible polymers. The elongated form of the polymer is reflected in various dilute solution properties such as the larger intrinsic viscosity, larger relaxation time, or smaller diffusion constant as compared to those of flexible polymers.’ Secondly, due to the large molecular anisotropy, rodlike polymers are much more easily oriented by an external field and show large birefrin- gence. This enables us to use electric or magnetic birefringence as a practical tool to study the rotational motion of these polymers.” Thirdly, the distinction between rodlike polymers and flexible polymers becomes more pronounced as concentration increases. Due to their larger size, the interaction of the rodlike polymers becomes important at a much lower concentration than with flexible polymers, and, as we shall show later, the effect of the entanglement is much more remarkable. Fourthly, but not least, when the concentration becomes sufficiently high, rodlike polymers spontaneously orient towards some direction, and form a liquid crystalline phase.° It is this capability of forming a highly ordered phase that produces strong fibres. In this and the following two chapters, we shall discuss the physical Properties of such polymers. Although real polymers have finite rigidity and can bend to some extent, we shall mainly consider the extreme290 DILUTE SOLUTIONS OF RIGID RODLIKE POLYMERS situation, i.e. tigid rodlike polymers. The effect of the flexibility will be discussed only briefly. Theoretical treatment of rodlike polymers is much easier than for flexible polymers since rodlike polymers can have only two kinds of motion, i.e., translation and rotation. Once the basic equation is set up, mathematical analysis is easy. However, important physics is included in an essential way in the problems of rodlike polymers. In particular, the importance of the orientational degrees of freedom and the peculiar nature of the topological constraints will be seen clearly in this system. In this chapter we shall discuss the properties of dilute solutions. The properties at higher concentrations will be discussed in later chapters. 8.2 Rotational diffusion 8.2.1 Rotational Brownian motion Rodlike polymers do two kinds of Brownian motion, translation and rotation. The translational Brownian motion is the random motion of the position vector R of the centre of mass, and the rotational Brownian motion is the random motion of the unit vector u which is parallel to the polymer. To visualize the rotational Brownian motion we imagine the trajectory of u(t), which is on the surface of the sphere |u| =1 (see Fig. 8.1). For short times, the random motion of u(t) can be regarded as Brownian motion on a two-dimensional flat surface, and the mean square displace- Fig. 8.1. Rotational diffusion.ROTATIONAL DIFFUSION 291 ment of u(¢) in time ¢ is written as ((u(@) — u))?) =4D,t (for D,t «1). (8.1) The coefficient D, is called the rotational diffusion constant. Note that the dimension of D, is (time)', and is not the same as that of the translational diffusion constant, which is (length)?/(time). Equation (8.1) is correct only for D,t <1. To discuss the general case, we have to study the Smoluchowski equation for the rotational Brownian motion. This equation can be derived straightforwardly according to the Kirkwood theory* described in Section 3.8. Such a derivation is given in Appendix 8.1. Here we derive it by an elementary method to clarify the underlying physics. 8.2.2. Hydrodynamics of rotational motion As was discussed in Chapter 3, the first step in deriving the Smoluchow- ski equation is to obtain the phenomenological relation between the force and flux by using the hydrodynamics of the problem. Consider a rod placed in a quiescent viscous fluid. If an external field exerts a torque N on the rod, the rod will rotate with certain angular velocity w. For thin rod, we may neglect the rotation around u, and assume that both w and N are perpendicular to u. If N is small, @ is linear in N, and by symmetry, parallel to N. 1 o E N. (8.2) The coefficient €, is called the rotational friction constant. A simple estimation of & is done for the ‘shish-kebab model’ illustrated in Fig. 8.2: the rod is regarded as made up of N = L/b ‘beads’, which are numbered from —N/2 to N/2, When the rod rotates with angular velocity w, the bead n which is separated from the centre by the distance nb moves with velocity V, =(@ X nbu). If the hydrodynamic interaction is neglected, the frictional force acting on the segment n is —€)V,, where €=3,7,6 is the translational friction constant of the bead. Thus the total torque due to the hydrodynamic friction is given by NI2 Nevction=— 2, _nbu X £oV, n=—-NI2 NI =- > nbuX (foo X nbu) n=-N2 NZ 2 3 aL? =- 2620 = — 22 (NV yan Ze . bo 3 n'b?@ = —Gan,b)6?=(X) o=-n wo (83) n=-Ni2292 DILUTE SOLUTIONS OF RIGID RODLIKE POLYMERS ue Fig. 8.2. Rodlike polymer and shish-kebab model, which consists of N = L/b beads of diameter b placed along a straight line. which must balance with the external torque N. Hence ansL? t,=2be 4 If the hydrodynamic interaction among the beads is taken into account, €, is shown to be (see Appendix 8.1): (8.4) an,L? 3in(L/2b) More precise hydrodynamic calculation for the cylinder gives a correction y to the denominator:+ c= (8.5) p= * 3(in(L/b) — y)” + For a prolate elipsoid, an exact calculation can be done and the result is” 16x -( yl 2p-1 (ete vn) T = 0,°(1-4)|—*— -1 beng me 3p(p?= 1p P=) where 2a is the length of the long axis and p is the aspect ratio. For p >2, the above equation is approximated as (8.6) te 1627,a* * 3{2tn(2p) — 1] which agrees with eqn (8.6)ROTATIONAL DIFFUSION 293 The calculation based on the point force approximation® gives y =0.8, while more recent calculation” indicates that y weakly depends on L/b. The torque N is now expressed by the potential U(u) of the external field. Consider a small rotation Sw which changes u to u+ dw Xu. The work needed for this change is —N- dy, which must be equal to the change in U, i.e., -N- by = Ulu + by Xu) — Ulu) = (69 Xu) “2 U=5y- (ux) Hence 87 N=-RU 8. wheret 68) R=ux 2 : (8.9) ou The operator ®, called the rotational operator, plays the role of the gradient operator 9/AR in translational diffusion. An important property of ® is the formula of integration by parts, i.e., for the integral over the entire surface of the sphere of |u| = 1, | duA(u)RB(u) = — | du[RA(u)]B(u). (8.10) (In quantum mechanics, —i@ corresponds to the angular momentum operator so that eqn (8.10) is equivalent to the Hermitian property of this operator.) Now if the fluid surrounding the rod is flowing with a certain velocity gradient, there will be an additional angular velocity w» of the rod, which again can be calculated by hydrodynamics.} For a slender rod, Wp is obtained by a simple geometrical reasoning explained in Fig. 8.3: Oo =UuX Kou. (8.11) tIn eqn (8.9) 3/u, means the partial derivative in which u,, u,, u, are regarded as indent variables. Since w is a unit vector, there are many ways to express U. For example, consider the following three quantities = = uy =(1— 22-42) Rate Boca B= G-a- un All represent the same quantity for the unit vector w. It is easily checked that, though OF,/du,, 9F,/du, and 3F,/3u, are not equal to each other, RA, RE, and RF, are all equal. Thus the derivative RF has no ambiguity. +For a spheroid of aspect ratio p = a/b, wy is given by** =ux p Ke -age-s) won mT Marre “#)- In the limit of p>, this reduces to eqn (8.11).294 DILUTE SOLUTIONS OF RIGID RODLIKE POLYMERS Fig. 8.3. Geometrical meaning of eqn (8.11). If the rod follows the macroscopic velocity gradient, its direction changes as = K - u — (uu:K)u. Hence the angular velocity @o is given by @ =u X a =u X(K-u). Therefore the angular velocity of a rod immersed in a fluid with velocity gradient « and subject to an external potential U(u) is given as o= —-lausuxnen. (8.12) é This consequence of the hydrodynamics corresponds to eqn (3.118) for translational motion. 8.2.3 Smoluchowski equation for rotational motion Now it is easy to give an account of the Brownian motion. If Y(u; t) is the probability distribution function of u, the Brownian motion is included by adding the ‘Brownian potential’ k,T In W to U. The angular velocity @ is now given by w= ~ EM kel In W +) + ux Kw (8.13) For given , u changes with the velocity w Xu, and the equation for the conservation of the probability becomes ow 9 a = -Z-(oxuy= -(w x=) ‘aU =-R-(W). (6.14) From eqns (8.13) and (8.14), we have the Smoluchowski equation forTRANSLATIONAL DIFFUSION 295 rotational diffusion SEER: [kpTRY + RU] — R(X nu) = DR [RY + RU] - R-(wXHe-u) (8.15) where D, is defined by kpT _ 3kgT(in(L/b) — y) & a,b? We shall later show that D, agrees with the rotational diffusion constant defined by eqn (8.1). Note the formal similarity between the rotational diffusion equation and the usual translational diffusion equation: if the gradient operator 9/AR in the translational diffusion equation is replaced by the operator &, the rotational diffusion equation is obtained. The rotational Brownian motion can also be described by the Langevin equation, but it is rarely used in the problem of rodlike polymers because it is less convenient for calculation than the Smoluchowski equation. D,= . (8.16) 8.3 Translational diffusion 8.3.1 Hydrodynamics of translational motion It is straightforward to include the translational motion into the Smol- uchowski equation. Again the hydrodynamics is considered first. Suppose the rod is moving with the velocity V in a quiescent fluid (see Fig. 8.4). If the rod moves along u, the rod will feel a hydrodynamic drag, which is parallel to V and is written as €V. On the other hand if V is perpendicular to u, the drag is again parallel to V and is written as €, V. AY @F=CV (oe) F=C.V OFHCV,+0V, Fig. 8.4. Anisotropy in the translation friction constant. (a) V||u, (b) V Lu, and (c) general direction V= V+ Vi.296 DILUTE SOLUTIONS OF RIGID RODLIKE POLYMERS In general the coefficients £, and €, are not equal to each other. They are called the parallel and perpendicular components of the translational friction constant, respectively. Given €, and ¢,, the hydrodynamic drag for a translational motion in a general direction is obtained as follows. Since the Stokes equation (eqn (3.102)) is a linear equation, the hydrodynamic drag must be linear in V. Thus if Vj, and V, are the parallel and the perpendicular components of V, the drag is written as FaQVjtoih. 6.17) Substituting Vi=(V-uu and V,=V-Y, (8.18) we have F=Cyuu-V+6.(l—uu)-V. (8.19) The calculation of £, and £, based on the Kirkwood theory is given in Appendix 8.I. The result is 2an,L Si inci) (8.20) 6. = 26). (8.21) Equation (8.19) is solved for V as v= le uu +e- wu)| “F. (8.22) If there is a macroscopic flow v(r)=« +r, there is an additional velocity xR for the rod at point R, and eqn (8.22) becomes Vm aut Om) eR, (8.23) This is the result of hydrodynamic calculation.} 8.3.2 Smoluchowski equation including both translational and rotational diffusion We can now write down the Smoluchowski equation which includes both the rotational and translational motions. Let W(R, u; t) be the probabil- +Note that in the case of a rod, V is independent of the torque N acting on it. In the general case, this is not true. (Consider for example a screw: the torque turns the screw and. causes a translational motion.) The formula for the general case is given in refs 6 and 11, and an example of its application to the Brownian dynamics is given in ref. 12TRANSLATIONAL DIFFUSION 297 ity distribution function for the rod in the configuration (R, u). The velocity V is given by v=-[pmet ze (-m)|- ZeTinw +0) +K- +R. (8.24) The angular velocity is again given by eqn (8.13). Substituting this in the continuity equation awa a — 5p VE) BR (a) (8.25) we get aw aw WOU ae aR’ Pame + DG uw) (Rt ee an RY) +DQR- (av + er?) R-(uXn«-u), (8.26) where _keT _keT n(L/b) ‘oT I“) 8.27 = ty lanl (8.27) and _kpT_keT in(L/b) (628) 1b, 4am The constants Dy and D, characterize the diffusion parallel and perpendicular to the rod axis: if the rod is along the z axis, then the displacement of R in a small time interval At is given as ((Rx(At) — R(0))?) = ((Ry(At) — Ry(0))?) = 2D At {(R,(At) — R,(0))*) = 2D, Az. Since Dy >D_, the rod can move more easily in the direction parallel to the axis than that perpendicular. Due to this anisotropy in the diffusion constant, translational and rotational motions of a rod are generally coupled with each other. For example, a concentration gradient of the rodlike polymer can induce an anisotropy in the orientational distribution. However, the reverse is not true: in a homogeneous system (in which the Positional distribution is uniform), the translation—rotation coupling has No effect: if the system is homogeneous, it will remain homogeneous even if the orientational distribution is not isotropic. Thus in a homogeneous system, one can discuss the rotational diffusion using eqn (8.15) instead of the full Smoluchowski equation (8.26). (8.29)298 DILUTE SOLUTIONS OF RIGID RODLIKE POLYMERS 8.4 Brownian motion in the equilibrium state Having obtained the Smoluchowski equation, we now study the charac- teristic features of the Brownian motion of a free polymer (U =0 and «=0). 84.1 Vector correlation function (u(t) -u(0)) To see the rotational motion, let us consider the time correlation function (u(t)-u(0)). According to the general prescription given in Chapter 3, this is calculated by (u(t) w(0)) = du du'a - 'G(u, u's 1).,(u"), (8.30) where W,, is the equilibrium distribution function 1 Weq(u) = re (8.31) and G(u, u';t) is the conditional probability that the polymer is in the direction of u at time ¢, given that it was in the direction u’ at time t= 0. This probability is the Green function of the diffusion equation 2G, u'; t)=D,R’G(u, u’; t) (8.32) with the initial condition G(u, u';¢=0) = 5(u—u'). (8.33) Though the explicit form of G(u, u’; t) is available (see for example ref. 13 Chapter 7), the time correlation function can be calculated directly as before. The time derivative of (u(t) - u(0)) becomes 3 (u(e)-0(0)) = feu duu -w'| 2 Gu, ws 0) ea(u’) =D, [du du'u + w'[R°G(w, u's eu’). (8:34) Using eqn (8.10) for the right-hand side, we get 2 (u(t) « u(0)) = D, i du du'[Ru- u']G(u, w'; (uw). (8.35) By a straightforward calculationt Relig = — CapyUly. (8.36) t Here e,gy is Levi Civita’s symbol, i.e. e.g, =e, * (eg X€,), where e, is the unit vector in the direction of the a axis. oer (eo erBROWNIAN MOTION IN THE EQUILIBRIUM STATE 299 Applying this twice, we have Ritts = = — C apy Rally = Capyl ayully = —2ilp- , (8.37) Hence eqn (8.35) is written as Suto -u(0)) = ~2b, {du du'u-u'G(u, u'; t)Y,,(u') = —2D, (u(t) - u(0)). (8.38) Since (u(t) + u(0)) is equal to 1 at time t= 0, eqn (8.38) gives (u(t) - u(0)) = exp(—2D,t). (8.39) The rotational correlation time 1, is thus given by 1, =1/2D, (8.40) From eqn (8.39), it follows that ((u(t) — u(0))?) = 2— 2(u(?) - u(0)) = 2(1 — exp(—2D,t)). (8.41) For tD, <1, eqn (8.41) reduces to eqn (8.1), which gives a clear physical meaning of D,. In the same way, one can show? 3([u(e) - (0)P —3)) = exp(—6D,2) (8.42) or in general (P,(u(f) + u(0))) = exp(—D,n(n + 1)t) (8.43) where P,(x) is the Legendre polynomial of n-th order 1 da n Pa) = eq gyn DY (8.44) 8.4.2 Translational diffusion Consider the mean square displacement of the centre of mass: 9) = ((R() - RO)’) (8.45) This is calculated by essentially the same method as before. Let G(R, u, R', a t) be the Green function for the configuration (R, 4), c-[p, wR? + * (Dyuu + DI ua) ale (8.46) with the initial condition G(R, u, R', u’;t=0)=6(R—R')6(u -w'). (8.47)300 DILUTE SOLUTIONS OF RIGID RODLIKE POLYMERS Then ¢(¢) is calculated from e()= far du dR’ du’(R — R'PG(R, u, R', u’;t)Y.,(R’, u’). (8.48) We again evaluate the time derivative of #(t) using eqn (8.46). The resulting equation is written, after integration by parts, as a - o < o()= [aR du ar du’GY,, a a x [Dw + 5 -(Djuu + Dy (I~ wu) 5 |(R = RP = far du dR' du’GW,,(2D, + 4D) = 2(D, + 2D,). (8.49) Hence $(t) = 2D, + 2D, )t. (8.50) Thus $(f) increases linearly with ¢ and the diffusion constant Dg defined by Dg =lim x ((RQ) - RO))) (8.51) is given by Dg = Dit2Ps _ (Lib), (8.52) 3 32n,L This formula can also be directly derived from the Kirkwood formula for the diffusion constant eqn (4.102). It must be noted that although $(#) increases linearly with time, the diffusion of R is not Fickian because of the translation—rotation coupling. Indeed as will be shown below, G(R, u, R', u';t) is not Gaussian in R-—R’'. The Fickian diffusion is recovered if the relevant length-scale is much larger than L. 8.4.3 Dynamic light scattering As before, the Brownian motion of the polymer can be studied by dynamic light scattering.’* If we take the shish-kebab model shown in Fig. 8.2, the dynamical structure factor is given by a D= pS explik R= ROM). 65) (Note the normalization of eqn (8.53) is chosen so that g(0, 0) = 1, whichBROWNIAN MOTION IN THE EQUILIBRIUM STATE 301 is different from that chosen for flexible polymers.) Since R,=R+nbu (8.54) the sum over n is NR NIZ > exp(ik- R,) = exp(ik +R) | dn exp(ik + unb) n==NI2 -Ne - «py Sin(k uNB/2) _ py si(k + wL/2) 2exp(ik - R) rr Nexp(ik +R) aL (8.55) Thus g(k, f) is expressed by R(t) and u(t) as _ loan tpi) sin[K - u(s)] sin[K - u(0)] ak, #) = (expGk [R(® - RCO) SEE TA where ) (8.56) K=kL/2. (8.57) Let G,(u,u’;t) be the Fourier transform of the Green function G(R, u, R’, u';t), which depends only on R — R’, G,(u, ust) = | dRe*®G(R, u, 0, u’;1) (8.58) then sin(K + w) sin(K « u’) = du’ " y a(h 1) = [edu fau’Gy(u, u's) SESE Hy wy (8.59) From eqns (8.46) and (8.58), G,(u, u'; t) satisfies (2+ F)G., ust) =0 (8.60) with F=—D,R?+ Dy(k-u)?+ Di[- (ku) (8.61) The solution of eqn (8.60) is involved, so here we shall briefly describe its characteristic aspects. The limiting cases are: (i) |K| K1, ie., [AI LL. In this case, it is intuitively obvious that g(k, t) is described by the Fickian diffusion with the diffusion constant Dg, so that gk, t) = exp(—Dgk’). (8.62) A formal justification of this is made by considering the eigenfunction expansion of G,. Let , and A, be the eigenfunctions and the corresponding eigenvalues of I. Pp = op (8.63)302. +=DILUTE SOLUTIONS OF RIGID RODLIKE POLYMERS Then Ge= z exp(—A,1) (4) ¥p(u')- (8.64) Substituting this into eqn (8.59) and using K «1, we have Bh, =D enol —Agt( day, cw) A): -23 ents Jawy,(u) (8.65) If k=0, the eigenfunctions are given by spherical harmonics Y,,,(u) with the eigenvalues A{2) = D,J(/ +1) and only the term of the lowest eigen- function Yoo(u) = 1/V(4s) remains in the sum of eqn (8.65). If k is small, the relaxation is still dominated by the eigenfunction of the lowest eigenvalue 8(k,t) = exp(—Aot). (8.66) Ag is obtained by perturbation theory: Jo= | duYool Yon / | duY2,= PUA ee Do (8.67) which justifies eqn (8.62). For the above perturbation calculation to be justified, Aj must be much smaller than the next smallest eigenvalue AS, = 2D,, i.e., Dok? «2D, (8.68) which is rewritten using eqns (8.16) and (8.52) as |A|L<«1. (8.69) Equation (8.69) indicates that if the relevant length-scale is much larger than L, the diffusion can be regarded as Fickian with the diffusion constant Dg. Gi) [K|>>1, ie., |k| L>>1. In this case, g(k, t) is not expressed by a single exponential. However, the initial decay rate is calculated easily: TP =-Ling(&, D)loa “75 ~ fost _sin(K8) sink) sin( KE) sin) / I ag (ae (8.70) sink »f pee Dy ¥.,(u)ORIENTATION BY AN ELECTRIC FIELD 303 where £=K-u/|K| and K=|k|. (8.71) By using the relation OF | uXKOoF 72) RF(E)= (RE) BE TK] 29€ (8.72) we have, after some senate [ae =[-p SE 20-92 3et DPE + Dude e|\e®). (8.73) The integral over & is carried out analytically by using the fact that K is large: for example, 2B) = fan(S24) = fon) = jae KE K “Kia 7) K Straightforward calculation gives finally’ 2 TP=Dke+ 5 DR. (8.74) More detailed studies are given in the literature.'* 16 Since rodlike polymers have a large optical anisotropy, they have a significant depolarized light scattering, which is particularly suitable for studying rotational diffusion. In the small-angle regime |k| L<1, the dynamic structure factor is written as’? Saep(k, t) « exp(—Dgk"t — 6D,t), (8.75) the decay of which is mainly determined by D,,. 8.5 Orientation by an electric field 85.1 The effect of an electric field An electric or magnetic field can orient the polymer, and measurement of this process gives information on the rotational motion of polymers in solution.’*° Using the Smoluchowski equation, we shall consider the orientation caused by an electric field. Elementary electrostatics says that if an object with dipole moment p is placed in an electric field E(t), it feels a torque N=pXE. (8.76) The dipole moment p consists of two parts, the permanent dipole p, and304 DILUTE SOLUTIONS OF RIGID RODLIKE POLYMERS the induced dipole p;. For thin, rodlike polymers the permanent dipole is always parallel to u, the direction of the polymer, and is written as Pp = Hu (8.77) where y is the magnitude of the permanent dipole moment. On the other hand, the induced dipole moment is written using the parallel and perpendicular polarizability a and a, as Pi= a(u + E)u+ w(E —(u-E)u) = [utes 14+ (q- a,)(uu -%)| -E=a-E. (8.78) Hence the torque acting on the polymer is written as N=(p, +p) XE =puxE+Aa(E-uuxeE (8.79) where Aa =a - a. (8.80) The potential which gives such a torque is = —pu- E—4Aa(E -u)*. (8.81) The orientation of the polymers can be studied by measurement of the dipole moment (p) = (ws +a +E) (8.82) or the birefringence. The refractive index tensor A, is written as Aap = NS ap + A( Ua — 48ap) (8.83) where A is a constant which includes contributions from both the form birefringence and the intrinsic birefringence. 8.5.2 Dielectric relaxation We now calculate the average dipole moment under a given electric field E(t). The diffusion equation to be solved is ow w = =px-[av+ au). (8.84) To obtain (u), we multiply both sides of eqn (8.84) by wu and integrate over w. 5 (u) =D, [auu( 2 + 8S. ev) (8.85)ORIENTATION BY AN ELECTRIC FIELD 305 Integration by parts leads to 2 uya 2) -—_ (gu - ) S (uy =D,(( 1) ~ papi Bu BU). (8.86) As before, Ru gives —2u, and (Ru-RU) is calculated from eqn (8.81). Hence 2 yy =D,(- ~ 5p (u-B- 8) - 2% ((E-uy'u-E-w)B)) = (u)=0,( 2(u) — pg (EE) — Fo ((E “uu (Eu). (8.87) We shall consider the linear response, in which case we can neglect the third term and evaluate the average in the second term for the isotropic distribution function of u: ((u-B)u—E)y= | 2 [(u- Bu E]=— 8. (8.88) Hence apy 2D, 5 = 2D.u) +30 FE. (8.89) The solution of this equation is { 2, (u)= f ae’ exp(—20.0--9) Tepe) (8.90) From eqns (8.82) and (8.90), the dipole moment is given by i ip) -2e | dt’ exp(-2D,(t = #*)E(") +7 t 1 gp, (8.91) B For an oscillating electric field where E(t) = Re{E exp(iot)], (8.92) the dipole moment is calculated as (p) = Re[a*(w)E exp(iat)] (8.93) where 2 * __# 1 ay t 2a, ()=FeTitior* 3 654) with an,L? 6kyT(In(L/b) — 7)" @*(w) is called the complex polarizability. Note that the induced dipole t=1/2D,= (8.95)306 DILUTE SOLUTIONS OF RIGID RODLIKE POLYMERS Moment shows no frequency dependence. Dielectric relaxation experi- ments have been carried out for various polymers." 8.5.3 Electric birefringence Calculation of the electric birefringence is slightly more complicated because it is a second-order effect in the electric field. Let us consider the case that a time-dependent, weak electric field E(t) is applied in the z direction. Straightforward perturbation calculation gives**”* : (ub 02) =3D,25 | des expl-6D.(¢~ A IEC? B +402(45) | deexpl-60,¢— se) x ie exp[-2D,(ti — )JE(4). (8.96) The difference in the refractive index An =,, —f,, is proportional to (u2- uz). The response of An for various histories is immediately calculated from eqn (8.96). (i) Steady state: If the electric field is constant, eqn (8.96) gives 2_ 2) aa { A, (HL ‘| 2 (u2 — u2) alet+ (4) Ee, (8.97) Thus An = KoE” with =Aal 425 (4 ‘| Ko= ase (Zz) 6.98) which is called the Kerr constant. It consists of a permanent and an induced dipole term. If the polymer has a dipole moment, their ratio 2 r= AakgT is much larger than unity since both 4 and Aq increase in proportion to the molecular weight. (8.99) (ii) Decay: a constant field E is switched off at t=0, ie., E(t)= E@(—1), and An(t) = KoE? exp(—6D,t). (8.100)LINEAR VISCOELASTICITY 307 (iii) Rise: a constant field is applied at t = 0, i.e., vO = EQ(t), and An(t)= KoE*(1- exp(—2D,t) + exp(—6D, .t)). (8.101) (iv) Reversal: the direction of a constant field is changed at t = 0, i.e., E=-—E@(-—t)+ EQ@(t), and _3R_ 2(R +1) aes 5 An(t)= Koe|1 + a (exp(-6D,1) — exp(~20,1))} . (8.102) (v) Oscillation: E(t) = E cos(wt). The response for this field is written as An(t) = E?[Ka() + Re Kt,(o)exp(2iot)] (8.103) where K R Ka“ oR + 1) G +(atpt 1) (6.104) and a a Se Kul) =F R40 la + 2iwt/3\(1+ ior) * a+ iim (8.105) where t is given by eqn (8.95). 8.6 Linear viscoelasticity 86.1 Expression for the stress tensor We now consider the viscoelasticity of a solution of rodlike polymers. As was discussed in Chapter 3, the stress tensor consists of two terms, the elastic stress a) and the viscous stress 0. The elastic stress is related to the change in the free energy of (per volume) for a virtual deformation dé,, as 6A = CDSE qp- (8.106) Since the free energy is given by =v | du(keTY In ¥ + WU) (8.107) (where v is the number of polymers in unit volume), the change in & is written as b= du(kgTS¥ In V + kgTSW + 5WU). (8.108)308 DILUTE SOLUTIONS OF RIGID RODLIKE POLYMERS The change 5 is calculated using the Smoluchowski equation (8.15). For the instantaneous deformation, the velocity gradient Kg = d&q/6t (6t being the duration time of the deformation) dominates the time evolution of WY, so that = =-R-(uxXK -uW) during the deformation. _(8.109) Hence bY =-R-(uXK-uP)dt=—-R-(ux de -uV). (8.110) Substituting eqn (8.110) and integrating by parts, we have b= v felu(ksT(u x Be + u) > OE + (u x be -ul)- RU) =v faw(-kaT «(ux be +n) +(uX de +u)+ RU). (8.111) The underlined term can be calculated using eqn (8.36): Rs (uX 5E + u) = —35Ey6(Uglig — 45g). (8.112) Hence 65 = vbExg | du'Y (3k eT (Untip — 45 ap) — ( X RU) qtg) = Vbeap(3kgT (Ualls — 4605) — (UX RU)qutg)). (8.113) Thus OGD = 3vkgT (Ugg -45ap) — V((u X RU)aup). (8.114) If U=0, the stress is given by o® =3vkp TS xp (8.115) Sop = (Ualtp - 45ep), (8.116) which is called the orientational tensor. Note that this stress comes entirely from the Brownian potential. That the Brownian motion of a rod can produce a stress may be understood in the following way. Suppose a rod is placed at the origin along the z axis (see Fig. 8.5). If the rod rotates by Brownian motion, it will create flow of the surrounding fluid as in (a) or (6) of Fig. 8.5, depending on the direction of the rotation. Whichever direction the rod rotates, the fluid around the z axis comes towards the rod while the fluid in the x — y plane goes away from the rod as shown in (c). This fluid motion is equivalent to the appearance of the stress 0,, — 0,,- The viscous stress is related to the hydrodynamic energy dissipation W by (see Section 3.8.4) with W = kapoSe (8.117)LINEAR VISCOELASTICITY 309 om °° yf Mw _ a ws A a a @ 0) ©) Fig. 8.5. Explanation of the stress expression (8.115). The rotation of the rod causes fluid flow as in (a) or (b). The direction of the rotation is random, but on average, the fluid moves as shown by thick arrows in (c). A crude estimation of W is done easily again by neglecting the hydrodynamic interaction in the shish-kebab model. Under the velocity gradient «, the rod rotates with the angular velocity @»)=u X (m+). Hence the velocity of the n-th bead relative to the fluid is Var = nb (Wo X u — K+ u) = nb([u X (K+ u)] Xu — K+ u) = —nbu(K: un). (8.118) The frictional force acting on the segment is F, = €)V,,. Hence the work done by the frictional force in unit time and unit volume is WH vd (Ey Vir) =v 3 bn? (na?) = veg (0 ay?) n n=-N/2 (8.119) where NiI2 3 bar= Dont? = TL (8.120) n=-NR2 4 Hence OR = VE secKuv (Uy ty taltg (8.121) Note that in this case, €,., is equal to the rotational friction constant 6, (see eqn (8.4)). If the hydrodynamic interaction is taken into account, Ey, is given by £,/2 (see Appendix 8.1). We shall consider only this case,310 DILUTE SOLUTIONS OF RIGID RODLIKE POLYMERS whence OS) =F Ekne (atl tty) (8.122) Thus the stress due to the polymer is of = off + off = 3vkpTSyp — v((6X RU)aup) + 5 E,Kyy (Ue Uiplty tly) (8.123) and the total macroscopic stress in the solution is up = 09} + 0, (Kap + Kpa)- (8.124) 86.2 Calculation for weak velocity gradient Let us now calculate the stress tensor under a weak velocity gradient. If we consider the case U =0, then eqn (8.15) becomes ow a DRWV-R+ (ux n-u). (8.125) To calculate S,, we again multiply u,ug — 54/3 by eqn (8.125) and integrate over u. The equation is then rewritten by integration by parts as a F Sug = [du PDB (usp — $59) + Rlitatig — HBug) «(UX H- W)) (8.126) The terms in the bracket is calculated directly giving 3 3500 = —6D,(Uqttg — 35 ap) + Kay (Upp) + Key (Upla) — 2Kyy (Ua lipUyptly ) = —6D,Sop +4(Kap + Kpa) + Kau Spu + Kpp Say — 2Kuy (Ugg ly Uy) (8.127) To calculate the first order in x, the coefficients of kas can be replaced by their equilibrium values. Since at equilibrium Sap =O and (ugtg,t) = i(Sap Suv + San Spy + SavOpn) (8.128) eqn (8.127) is approximated by 2 Sap = —6D,Sap + 4(Kap + Kpa) — %Kuv(Sap Suv + Sap Spy + Sav5pu) = -6D,Sap +}(Kap + Kea)» (8.129)LINEAR VISCOELASTICITY 311 (Here the incompressible condition x... =0 has been used.) This is solved by Sap(t) = i | dr’ exp[—6D,(t — t')(Kap(t') + Kpa(t')). (8.130) From eqns (8.123), (8.128), and (8.130), the stress is obtained, to the first order in Kgg, aS OC) = SvkaT | ae’ expl-60.(¢— 1 )Ifkap(t?) + Kpa()] + 55 S(Kaa(0) + Kpa()- (8.131) For the shear flow x ={ro for w=x and B=y, lo otherwise. Equations (8.124) and (8.131) give (8.132) Ony(t) = mex(t) + ivkyT | ai’ expl-6D-(¢= FeO) + C x(0). (8.133) Let us consider two special cases. (i) Steady shear flow: x(t) is constant, for which eqn (8.133) gives a(n, + veer A) - Oxy = (x. + 10D, +39 )K= (n. + &VE,)K. (8.134) Thus the viscosity of the solution is 2an,L> = dvl= —_— nen, +ive, "+ S5Gn(Liby= 1)" (8.135) The intrinsic viscosity is defined by 1 = lim —(n- n,. 8.136 [n] tim on” ns) (8.136) where p is the weight of polymer in unit volume, which is expressed in terms of v and the molecular weight M of the polymer as M pax. (8.137)312 DILUTE SOLUTIONS OF RIGID RODLIKE POLYMERS Thus [n]= 28 Na 1) 45(In(L/b) — y) M* (ii) Oscillatory flow: x(t) = Ko Re(e'™). The stress for this flow defines the complex viscosity n*(w) by Ory (t) = Ko Re(n*(w)e'). (8.139) (Note: The complex viscosity is related to the complex modulus by n*(w) = iwG*(w).) From eqn (8.133), it is easy to show that (8.138) ety =n 4 VEBT (__1 1) 10) = 16+ Top, (Fon (8.140) where a1. mt? 7 6D, 18(In(L/b) — y)keT” 8.141) The intrinsic complex viscosity is defined by 1 *(w)] = lim *(@) — ns 8.142) [n*(@)] tim) 7s) (8.142) which is given by 3001 1 “(oy =tal(> 1 "l=" Frsioe ta): 143) Note that [n*()] has a finite value as @—» ©, which arises from the rigid constraints of the rods. The relaxation time for (n*(@)] is one third of the relaxation time in the dielectric relaxation. Since L is proportional to M, [n] and t, depend on the molecular weight, [n] C x82, =bvkeT( a) . (8.171) Thus the intrinsic viscosity decreases in proportion to x-**. On the other hand, rigorous asymptotic analysis”? and the numerical calculation? show that the intrinsic viscosity decreases as x". 8.8 Effect of flexibility So far we have been considering the dynamics of rigid rods. Real polymers have greater or lesser flexibility. This can be modelled by a rod which has an elastic energy for bending. For the polymer which has constant contour length, the simplest possible model will be the follow- ing. Let R(s) be the position of a point on the chain at the contour length s. The vector oR u(s)= =a (8.172) is a unit vector tangent to the chain. The straight rod corresponds to u(s)=constant, or du/ds=0. Thus the bending energy must be a quadratic of Ou/ds. Since u + du/ds =0, the only quadratic form is =1E fa(2ey (8.173) where E is a constant. The conformational distribution of the polymer isEFFECT OF FLEXIBILITY 317 thus given by the Boltzmann distribution for this s energy: warson( St) eoml -Zfa(@)] avo where keT ==. 5.17: I= (8.175) This is called the Kratky-Porod model, and the length (2A)~' referred to as the persistence length. Equation (8.173) indicates the analogy between the change of u(s) of the Kratky-Porod model and the time evolution of u(t) in rotational Brownian motion: both processes are Gaussian with the constraint u’ = 1,> From eqn (8.174) it can be shown that for small s ((u(s) — #(0))) = 4a (8.176) which indicates that A plays the role of D, in the rotational Brownian motion. Thus the correlation functions of u(s) are immediately obtained from the result of eqn (8.43). In particular (u(s) + u(0)) = exp(—2As) (8.177) from which the mean square end-to-end distance is calculated as R= (RL) - RO)? = [sf d5"(u(s)-w(")) 1) 0 L s Loi =] as [a exp(-2i(s - 5!) =F sal —exp(-2AL)]- 8.178) The two limiting cases are: (i) LA>>1 (the random flight limit), R?=L/A (8.179) (ii) LA «1 (the rigid rod limit), =L’, (8.180) To develop a complete theory for the dynamical properties is difficult, but various approximate treatments have been proposed. Crudely speak- ing, the flexibility affects the dynamical properties in two respects. Firstly, the flexibility changes the transport coefficients. As the size of the chain decreases with the increase in the flexibility, the diffusion constant Dg, which is roughly proportional to R;', increases, and the318 DILUTE SOLUTIONS OF RIGID RODLIKE POLYMERS intrinsic viscosity [n] («R3) decreases. The results of various approaches are summarized in refs 1 and 36. Secondly, the flexibility gives a relaxation in the high-frequency region. For example, according to the rigid rod model, the complex intrinsic viscosity [)*(@)] approaches the finite value [a2] = lim [n*(@)] =4[7]. (8.181) For the flexible polymer, [7*(w)] decreases as the frequency becomes comparable to that of the bending mode.”* The experimental situation is summarized in ref. 29 and theoretical analysis is in ref. 37. Appendix 8.1 Derivation of the Smoluchowski equation by the Kirkwood theory In this appendix we derive the kinetic equation based on the Kirkwood theory described in Section 3.8, and using the model shown in Fig. 8.2. The position vector of the n-th segment (—N/2 Him * Fon (8.1.3) From the definition of H,,, (eqn (3.106)) and eqn (8.1.1), it follows that Hy, = (uu t+ Diam nm (8.1.4) with liam = 1/820, |n — ml b. (8.1.5) In eqn (8.1.4), we neglected the term h,,,. The validity of this approxima- tion is discussed later. Now the total force acting on the centre of mass is given by 2, F- Equating this with that given by the thermodynamic potential we get > — 2 (kgT in¥+ U). (8.1.6) a OR Similarly the balance in torque is written as > nbu X F, = -—R(kgT In W + U). (8.1.7)SMOLUCHOWSKI EQUATION 319 Our aim is to express V and @ in terms of the quantity on the right-hand side of eqns (8.1.6) and (8.1.7). For that purpose we solve eqn (8.1.3) for Fi B= D(H am Ym = Rr) (8.1.8) where (H™") um is the inverse of Hymn, i.e., 2 Ham * (H7")mic = Snel (8.1.9) From eqn (8.1.4), (H~)am is written as (Ham = CD am( Ht -3) (8.1.10) where (h7)am is the inverse of Aam, 2 Fam (h Dnt = Sut (8.1.11) We substitute eqn (8.1.8) into eqns (8.1.6) and (8.1.7), and use eqn (8.1.2) to obtain ZAM (V+ mb Xu) —K-(R + mbu)]= -SlkeT In W+U) (8.1.12) > nbu X (H™")am+ [V+ mbw X u— K+ (R+mbu)| = —RksT In W + V). “m (8.1.13) Using eqn (8.1.10), the right-hand side is rewritten as 2 Stan 1 +(V-K-R)= -ZTInw+ U) (81.14) DO agrmb?u x (1-2). (@X u— K+ u) = —R(kyT In ¥ + U). (8.1.15) (Here we used the property D,m1(h~")am=0, which follows from hom = h—n—m-) Defining =D ams (8.1.16) ¢.=0? > (A74),,,nm, (8.1.17)320 DILUTE SOLUTIONS OF RIGID RODLIKE POLYMERS we rewrite eqns (8.1.14) and (8.1.15) as tI -=). (V-K-R)= - Flo in +U) (8.1.18) C.uXx(oxXu—K+u)=6(w—uXK-u)=—-RksTIn¥+U), (8.1.19) which is solved for V and o giving V=k- ee a 2 (kgT In ¥ + U), (8.1.20) ¢ o=uX K-u—1 RkpT In + U). (8.1.21) Comparing eqn (8.1.20) with eqn (8.24), we have C120, by =6,/2. (8.1.22) Substituting this into the conservation equation (8.25), we finally get = DR: [av += av] - R- (ux k-uP) kpT ow A oO Ds +(i+uu)- [Set eetorl~ ore “RY (8.1.23) where D,=ksT/6, and D,=kgT/€,. (8.1.24) Equation (8.1.23) agrees with eqn (8.26). Next we obtain the expression for the stress tensor. According to eqn (3.134), of =—v >) (FraeRns) + (8.1.25) Substituting eqns (8.1.1), (8.1.2), and (8.1.10) into eqn (8.1.8), we obtain the force F, as F,=DOm(I-Z)- [Vtmbaxu—K-(R+mbu)]. (8.1.26) Here we consider the case when the total force acting on the polymer is zero, DF,=0, (8.1.27) which gives V=K-R, and = Le "San( =) « [mboxXu—K-mbul. (8.1.28)SMOLUCHOWSKI EQUATION 321 From egns (8.1.25) and (8.1.28), it follows that 08 = —v >) (kD amnmb> (00 X ut) qt — (K+ U)qUg + 4U,ugu-K-u). (8.1.29) Finally, substituting eqns (8.1.17) and (8.1.21) we get 09} = V{(CouvtyRulkaT In ¥ + U))ug) +> 5 EA otgt tly Kyv)- (8.1.30) The first term on the right-hand side of eqn (8.1.30) is rewritten using integration by parts (eqn (8.10) to give (Up aut, R, In VW) = ete Fy = Cane [du —W)Ry pty = 3(uglts —35ap). (8.1.31) Hence Of = 3vkgT (alg — 45ap) — v{(u X RU) attp) +3 ” iyy (Ualtpltly)- (8.1.32) The first two terms represent the elastic stress, and the last term is the viscous stress. Finally we calculate the friction constants ¢, and , using eqns (8.1.16) and (8.1.17). Since Kym decreases quickly with |n—m|, we may ap- proximate it by Fram = AS nm (8.1.33) with “e In(N/2) B=2 | don =" (8.1.34) (Aum = Sam he (8.1.35) Therefore from eqns (8.1.16), (8.1.17), and (8.1.35) N _ 4an.Nb _ Ann L b= a “in(N/2) ~In(L/2b) 6.1.36) and m? an (Nby _ an, L? = 2° { dm “31n(N/2) 3 In(L/2b)° 6.1.37)322 DILUTE SOLUTIONS OF RIGID RODLIKE POLYMERS References wn SN awe 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 24, 27, . Yamakawa, H., Modern Theory of Polymer Solutions. Harper & Row, New York (1971). . Tsvetkov, V. N., and Andreeva, L. N., Adv. Polym. Sci. 39, 95 (1981). . See for example, Blumstein, A., (ed.) Liquid Crystalline Order in Polymers. Academic Press, New York (1978); and Ciferri, A., Krigbaum, W. R., and Meyer, R. B., (eds.) Polymer Liquid Crystals. Academic Press, New York (1982). . Kirkwood, J. G., and Auer, P. L., J. Chem. Phys. 19, 281 (1951). . Jeffery, G. B., Proc. R. Soc. London A102, 161 (1922). Happel, J., and Brenner, H., Low Reynolds Number Hydrodynamics, Chap. 5. Prentice-Hall, Englewood Cliffs, New Jersey (1965). Perrin, F., J. Phys. Radium. 5, 497 (1934); 7, 1 (1936). , Burgers, J. M., Ver. Kon. Ned. Akad. Wet. (1). 16, 113 (1938). See also Second Report on Viscosity and Plasticity, pp. 113. North-Holland, Amster- dam (1938). . Broersma, S., J. Chem. Phys. 32, 1626 (1960). Yoshizaki, T., and Yamakawa, H., J. Chem. Phys. 72, 57 (1980). Bremner, H., Int. J. Multiphase Flow 1, 195 (1974). Rallison, J. M., J. Fluid Mech. 84, 237 (1978). Beme, B. J., and Pecora, R., Dynamic Light Scattering. Wiley, New York, (1976). Maeda, H., and Saito, N., J. Phys. Soc. Jpn 27, 984 (1969); Polymer J. 4, 309 (1972). Schaefer, D. W., Benedek, G. B., Schofield, P., and Bradford, E., J. Chem. Phys. 55, 3884 (1971). Wilcoxon, J., and Shurr, J. M., Biopolymers 22, 849 (1983). Maeda, T., and Fujime, S., Macromolecules 17, 1157 (1984); Kubota, K., Urabe, H., Tominaga, Y., and Fujime, S., Macromolecules 17, 2096 (1984). Rallison, J. M., and Leal, L. G., J. Chem. Phys. 74, 4819 (1981). A classical textbook on dielectric relaxation is Debye, P., Polar Molecules. Dover, New York (1945). A recent review on dielectric properties including rodlike polymers is given by Mandel, M., and Odijk, T., Ann. Rev. Phys. Chem. 35, 75 (1984). Fredericq, E., and Houssier, C., Electric Dichroism and Electric Birefrin- gence. Oxford Univ. Press, London (1973). A recent review is given by Watanabe, H., and Morita, A., Adv. Chem. Phys. 56, 255 (1984). Wada, A., J. Chem. Phys. 30, 329 (1959); 31, 495 (1959). Sakamoto, M., Kanda, H., Hayakawa, R., and Wada, Y., Biopolymers 15, 879 (1976); 18, 2769 (1979). . Bur, A. J., and Roberts, D. E., J. Chem. Phys. 51, 406 (1969). Peterlin, A., and Stuart, H. A., Hand und Jahrbuch der Chemischen Physik, Vol 81. Akademische Verlagses, Leipzig (1943). . Benoit, H., Ann. Phys. 6, 561 (1951); Tinoco, I. Jr., and Yamaoka, K., J. Phys. Chem. 63, 423 (1959). . Doi, M., J. Polym. Sci. Phys. ed. 19, 229 (1981). Nemoto, N., Schrag, J. L., Ferry, J. D., and Fulton, R. W., Biopolymers 14, 409 (1975).31. 32. 33. . Kratky, O., and Porod, G., Rec. Trav. Chim. 68, 1106 (1949). . Saito, N., Takahashi, K., and Yunoki, Y., J. Phys. Soc. Jpn 22, 219 (1967). . Yamakawa, H., Ann Rev. Phys. Chem. 25, 179 (1974); Pure Appl. Chem. 46, 37. REFERENCES 323 . Ookubo, N., Komatsubara, M., Nakajima, H., and Wada, Y., Biopolymers 15, 929 (1976). . Ferry, J. D., Viscoelastic Properties of Polymers (3rd edn), Chap. 9. Wiley, New York (1980). . See for example Bird, R. B., et al. Dynamics of Polymeric Liquids Vol 2, Chap. 11. Wiley, New York (1977). Hinch, E. J., and Leal, L. G., J. Fluid Mech. 52, 683 (1972). Hinch, E. J., and Leal, L. G., J. Fluid Mech. 76, 187 (1976). Stewart, W. E., and Sgrensen, J. P., Trans. Soc. Rheol. 16, 1 (1972). 135 (1976). Yamakawa, H., Ann. Rev. Phys. Chem. 35, 23 (1984).9 SEMIDILUTE SOLUTIONS OF RIGID RODLIKE POLYMERS 9.1 Semidilute and concentrated solutions of rodlike polymers In Chapter 8, we discussed the dynamics of a single rodlike polymer. Let us now consider the interaction between the polymers at finite concentration. Solutions of slender rodlike polymers of length L and diameter b may be classified into four concentration regimes (Fig. 9.1). Let p be the weight of polymers in unit volume of the solution, then the number of polymers per volume is given by v =a (9.1) where M is the molecular weight. (i) Dilute solution (Fig. 9.12). A dilute solution is defined as one having a sufficiently low concentration that the average distance between -13 5 the polymers v~*° is much larger than L, i.e., vsy=UL? (9.2) In such a solution each polymer can rotate freely without interference by other polymers. The effect of the interaction can be expressed by a power series expansion with respect to v as in the case of flexible polymers. (ii) Semidilute solution (Fig. 9.1b). If v>>¥v, the rotation of each polymer is severely restricted by other polymers, so that the dynamics of the polymers will be entirely different from that in dilute solution. However, the static properties will not be affected seriously until the concentration reaches another characteristic concentration v2. This is easily seen if one considers that the polymers are mathematical lines with no thickness. The equilibrium distribution of such polymers is entirely independent of each other at all concentrations. Thus the effect of the interaction becomes important in static properties only for polymers with finite diameter. Indeed the excluded volume of rigid rods is shown to be of the order of bL? (see Fig. 9.2)), so that the static properties are unaffected if vbL? is small. We call the concentration regime VS VK V2= 1/bL? (9.3)SOLUTIONS OF RODLIKE POLYMERS 325 (@) Fig. 9.1, Four concentration regimes of rodlike polymers: (a) dilute solution, (b) semidilute solution, (c) isotropic concentrated solution, and (d) liquid crystalline solution. semidilute. (Note that this definition of semidilute is different from that used for flexible polymers.) In this concentration regime, the effect of intermolecular interaction can be neglected in the static properties, while the dynamical properties are completely changed by the constraint that polymers cannot cross each other. We call such an interaction the entanglement interaction, although rods do not entangle with each other literally. (iii) Concentrated (isotropic) solution (Fig. 9.1c). The static correla- tion of the polymers becomes important at a higher concentration v > v2. Here, as will be shown later, polymers tend to orient in the same Fig. 9.2. Excluded volume between the rods in the direction u and uw’. For a given position of the rod in the direction u, the centre of mass of the other rod is not allowed in the parallelepiped region shown in the figure. The volume of this region is 2bL” |sin ©| where @ is the angle between u and u’.326 SEMIDILUTE SOLUTIONS direction as their neighbours. If the concentration becomes larger than a certain critical value v*, which is of the order of 1/bL”, the polymers align on a macroscopic scale in equilibrium, and the solution becomes an anisotropic liquid. In the concentration regime veSvsv" (9.4) the solution is still isotropic, but the excluded volume interactions among the polymers are important in both the static and dynamic properties. Such a solution is called a concentrated isotropic solution. (iv) Liquid crystalline solution (Fig. 9.1d). The anisotropic solution above v* is called a liquid crystalline solution. This solution shows a range of interesting properties quite distinct from those of isotropic solutions, and will be discussed in the next chapter. In this chapter we shall discuss the semidilute solution. This is in fact an ideal system for studying the entanglement effect since the dynamics of rodlike polymers is much simpler than that of the flexible polymers and the excluded volume effect is negligibly small. 9.2 Entanglement effect in rodlike polymers 9.2.1 Tube model In the semidilute region, the dominant interaction is caused by the topological constraint that the polymers cannot cross each other. This effect can be treated’” by the same model as that for flexible polymers. In the situation shown in Fig. 9.1b, the motion along a polymer is almost aN, a aN wt B B OX T a yy oe os a e NA Le Y A @ ©) Fig. 9.3. (2) Tube model for the rodlike polymer. (b) The mechanism of rotation in semidilute solution. The polymer can change its direction when it disengages from a tube, e.g., by moving from AB to A’B’ and then to A”B”. Reproduced from ref. 10.ENTANGLEMENT EFFECT IN RODLIKE POLYMERS 327 free, while the motion perpendicular to the polymer is severely limited by surrounding polymers. Such a characteristic feature of the Brownian motion can be again represented by a tube which surrounds the polymer (Fig. 9.3). The tube radius a corresponds to the average distance that the polymer can move perpendicularly to its own axis without being hindered by other polymers. Let us now study how the tube constraint affects the translational and the rotational Brownian motion. 9.2.2 Translational diffusion The translational motion is easily analysed. The motion of the polymer parallel to the tube axis, which is nearly parallel to u, is not hindered by the tube, so that the parallel component Dj will be nearly equal to the diffusion constant in dilute solution Djo.t On the other hand, the perpendicular motion is limited to within the distance a, so that the perpendicular component D, can be regarded as zero provided the small-scale motion of order @ is neglected. Thus Di=Dyo, _D, = 0. (9.5) 9.2.3 Rotational diffusion The rotational diffusion can be discussed using the following model (Fig. 9.3b). As long as the polymer stays in a certain tube, its direction is essentially fixed in the direction of the tube axis. If the polymer moves the distance L/2 along itself, it disengages from the old tube and goes into a new tube which is generally tilted from the old tube with an angle of about e~a/L. If the polymer leaves this tube, it again changes direction by order ¢. Thus overall rotation of the polymer is attained by repetition of this process. If ty is the average time necessary for the polymer to leave a certain tube, the polymer repeats the process ¢/tz times in a time interval ¢. Since the direction u(t) changes about € per step, the mean square displacement of u(t) is estimated as ({u)-wOP) == (0.6) Comparing this with eqn (8.1), the rotational diffusion constant is estimated as d,=£=(2) Ire @.7) + Strictly speaking, Dy is different from Do due to the hydrodynamic interaction among polymers. However, this effect gives only a logarithmic correction.?*328 SEMIDILUTE SOLUTIONS T c c oct fe, KH @ ) () Fig. 9.4. Constraint release process. The constraint on the test polymer T (thick line) imposed by a surrounding polymer A can be released if the polymer A moves away as shown in (6), or a new constraint can be created by an incoming polymer D as shown in (c). The time t, is estimated as the time necessary for the polymer to move the distance L: t= LD, ~ L/D yp. (9.8) Since L?/Dyo = D7o' (see eqns (8.16) and (8.27)), eqn (9.7) is written as D ~D.(2) (9.9) i.e., the rotational diffusion constant in the semidilute region becomes smaller than that in dilute solution by a factor (a/L)*. In the above estimation, it is implicitly assumed that the tube is fixed during the time t,. Actually this assumption is not correct because during the time t,, the surrounding polymers constituting the tube will also move a distance of the order of L, and the constraints imposed by them will be released. At the same time, new constraints are created by incoming rods (see Fig. 9.4). Thus the tube itself changes with correlation time tz. However, this effect only changes the numerical coefficients and does not change the functional form in the result of eqn (9.9) because the characteristic time and the step length associated with this process are again tg and e.' (Note that this situation is entirely different from that of the flexible polymers, for which the release of tube constraints has negligible effect if the polymer is sufficiently long.) 9.2.4 Estimation of the tube radius and the rotational diffusion constant To complete the analysis, we express the tube radius a in terms of v and L. A crude estimation of a is done as follows. We consider a tube of radius r around the test polymer and calculate the average number N(r)ENTANGLEMENT EFFECT IN RODLIKE POLYMERS 329 ) Fig. 9.5. (a) Tube enveloping the test polymer (thick line). (b) The polymers in the direction u’ intersect the area AS if their centres of mass are in the region shown here. The volume of this region is ASL |u’ -s|, where s is the unit vector normal to the region AS. of the surrounding polymers which intersect this tube. If r=a, the number N(r) will be of order unity. Hence a is estimated by N(a)=1. (9.10) To calculate N(r), we consider a smail region of area AS on the surface, and count the number AN(r) of surrounding polymers which penetrate this area (see Fig. 9.5). Let s be a unit vector normal to this region. As shown in Fig. 9.5, the surrounding polymers in the direction u' intersect the region AS if their centres of mass are in the region of volume LAS |u’ - s|. Hence AN = | W,(w/WLAS |’ «s| du’ (9.11) where W,(u’) is the orientational distribution function of the surrounding polymers. | First we consider that the distribution of the surrounding polymers is isotropic. In this case AN is independent of s, and given by x AN= vias [ do} |cos 6| sin 8 ="Zas. (9.12) o Summation of AN over the entire surface element gives vLS/2, (S= 2zxrL being the total surface area). This is twice the number of polymers330 SEMIDILUTE SOLUTIONS penetrating the tube since most polymers intersect the surface of the tube twice. Hence N(r) is obtained as 2 M(t) =3VLS = “=e . (9.13) From eqns (9.10) and (9.13), it follows that a=1/vL?, (9.14) Equation (9.14) indicates that in the semidilute region 1/L?« v «1/bL’, the tube radius a is much smaller than L, but much larger than b. Substituting eqn (9.14) into eqn (9.9) we finally get D, = D(VL?). (9.15) Thus D, is smaller than Dj» by a factor (vL*)~?. In the above estimation we neglected the numerical factor entirely. To write eqn (9.15) as an equality, we have to put in a numerical factor B D, = BD,(vL’)?. (9.16) The precise value of the numerical factor B is not known. Various data suggest that B is rather large (of the order of 10°) as will be shown later in comparisons with experiments. If the distribution of the surrounding polymers is not isotropic, the tube radius becomes a function of u, and so does the rotational diffusion constant, which will be written as D,(u). This is calculated in Appendix 9.1, B,(u) = p|4 | du! [us X 1'| ww] (9.17) where D, is the rotational diffusion constant in the isotropic environment, eqn (9.16). 9.3 Brownian motion in equilibrium 9.3.1 Time correlation functions Having obtained the translational and rotational diffusion constants, we can write down a kinetic equation for the probability Y(R, wu; t) that a given test polymer is in the configuration (R, u) at time t.+ ay = =D u-—) V+ R-D,- RV. (9.18) a Di(t 3p) OR: +D, must be placed between the two rotational operator %’s because (i) the equilibrium distribution of W must be isotropic and (ii) the integral of 9W/dt over u and R must vanish.BROWNIAN MOTION IN EQUILIBRIUM 331 Equation (9.18) describes the Brownian motion of a test polymer in a given environment whose distribution is specified by W,(u). In the isotropic solution, D, becomes a constant given by eqn (9.16). The conditional probability G(R, u, R’, u'; t) that the test polymer which was in the configuration (R’, w’) at time ¢ = 0 is in the configuration (R, u) at time ¢ satisfies 8G 9 a7 i aR Since eqn (9.19) has the same form as eqn (8.46), the time correlation functions for the test polymer are immediately obtained from the result of Section 8.4. yo + DRG. (9.19) (i) Translational motion. The mean square displacement of the centre of mass is given by ((R() — R(D)?) = 2Dyt ~ 2Dyot. (9.20) Thus the diffusion constant Dg is Da = lim = {(R(0) ~ R(O)??) = Dye (9.21) Hence the ratio between Dg and Dgo, the diffusion constant in dilute solution, is given by De __ Py 1 Doo Djo+2Dio 2° (6.22) Note that in the rodlike polymers, the entanglement does not change the molecular weight dependence of Dg. This is in contrast to the result for flexible polymers, in which the molecular weight dependence changes from Dgo* M~” to Dg «M7. (ii) Rotational motion The correlation function of u(t) is calculated as (u(t) - u(0)) = exp(—2D,t) (9.23) Thus the rotational relaxation time t, = 1/2D, is larger than that in dilute solution by 1, lt = (vLP/B. * (9.24) This shows a strong entanglement effect. It must be mentioned that in eqns (9.21) and (9.23) we have neglected the motion inside the tube. As in the case of flexible polymers, the tube332 SEMIDILUTE SOLUTIONS constraint is ineffective in a very short time-scale, so that {RO — R))’) = 6Deot (9.25) and (u(t) - u(0)) = exp(—2Dyot). (9.26) These equations hold if the time-scale is less than a ¢ le 9.27 Da De’ 27) In Section 9.6, we shall show how to describe both short-time (¢ < t.) and long-time (t¢ > t,) behaviour. 9.3.2 Dynamic light scattering Now we shall consider the dynamic light scattering in semidilute solution. In general the light scattering from polymer solutions of finite concentra- tions includes both the intramolecular and the intermolecular inter- ferences. In the semidilute regime, however, even though the average distance between the polymers is smaller than the polymer size, the intermolecular interference is negligible because there is no correlation between the configurations of different polymers. Therefore the dynamic structure factor is again given by eqn (8.59) ah, 1)= fan {aw (S8E2)) (RA) Gu, ws Naga!) (9.28) K-u where K=kL/2 (9.29) and 3 G,(u, u’; t) = (D,R? — Dy(k > u))G,(u, u'; 1). (9.30) The distinction between the dilute and semidilute regimes is in the magnitude of the ratio between the translational term and the rotational term in the equation for the Green function G,(u, u’; t). Let r be defined by: r= Dgk?/D,. (9.31) In dilute solutions, r is given by 19 = Dgok"/Dyo = (KL? (9.32) which is usually of the order of unity. In the semidilute regime, on the other hand, r can be quite large because 1 = Dgk?/D, = (kLP(vL?? >> (RLY for vL?>> 1. (9.33)ORIENTATION BY EXTERNAL FIELDS 333 To see the characteristic feature of the semidilute regime, let us consider the extreme case of r = © (i.e., D, = 0). Equation (9.30) is then solved by G,(u, u’; t) = 6(u — u’)exp(—Dj(k + u)’t). (9.34) Hence atk, = [oH (OY exp(—Dyck- 0) K| -| a(t 1) exp(—D 78). (9.35) Thus the dynamical structure factor has a very broad distribution of decay rates ranging from 0 to D,k’. Experimental study of dynamic light scattering has been carried out by several groups.** Though some qualitative features of the above predic- tions are indeed observed, clear interpretation of experimental results has been hindered by various factors inherent in real polymers, such as polydispersity, partial flexibility, and association. These effects will be discussed later in connection with rotational motion. An important factor which will not affect the rotational motion, but will be important in the dynamic light scattering is the effect of weak, long-range repulsive force.° As in the case of flexible polymers, such interaction tends to keep the segment density homogeneous, and increases the decay rate of g(k, t) with the concentration. This is indeed observed in several systems.>* 9.4 Orientation by external fields 9.4.1 Linear regime So far we have been considering the motion of a test polymer in an isotropic environment. We now consider a slightly different problem: how does the orientational distribution function of polymers change under external fields such as a potential field U,(u) or a velocity gradient x. Let W(u; t) be the probability that an arbitrarily chosen polymer is in the direction w. Since each polymer feels the external field as in eqn (8.15), the time evolution of Y(u; t) can be described by*"? ow A= R-D, [avs auyy] - R-[ux(K-wV)]. (9.36) An important point here is that in this problem, the environmental distribution function , is the same as Y(u; ¢) itself, and D, is now given334 SEMIDILUTE SOLUTIONS by 4 2 b,=D|* | du'V(u'; t) |u xu'|| . (9.37) Equations (9.36) and (9.37) give a nonlinear equation for W. The nonlinearity indicates a mean field character of the present theory: it comes from identifying the distribution of the surrounding polymers with that of the test polymer. Equation (9.36) is thus different from the usual Smoluchowski equation, which is always linear in WV. In the linear response regime, however, the nonlinearity of the kinetic equation is not important because there D, can be replaced by D, since the change in D, appears only in the higher order perturbation. Therefore the linear response function is given by the same form as that in dilute solution except that D, is much smaller than D,o. For example, consider a todlike polymer which has permanent dipole moment and isotropic polarizability (a = a, =«a.). The complex polarizability and the dyna- mic Kerr constant (per polymer) are given in the same form as eqns (8.94) and (8.104), wool (= STi + ior (9.38) aKo_ 1 Ke) =F 7 Fon? (9.39) with +=1/2D,. (9.40) The rotational diffusion constant can be obtained from these expressions. 9.4.2 Nonlinear regime—tube dilation The nonlinearity in the kinetic equation becomes important if the external perturbation is large. In this case precise mathematical analysis becomes quite difficult. A convenient approximation is to replace D, by the average D, 2 b,=B,=0]* few du"B(u; E(u’; 1) \uxw'l] . @41) Since D, is independent of u, the kinetic equation can be written as pe. [aw + au] —R-[ux(w-w)) (9.42) ot kgT This can be handled much more easily than eqn (9.36) (see refs 2 and 10). Though the approximation (9.41) is crude, it takes into account theORIENTATION BY EXTERNAL FIELDS 335 (a) (b) Fig. 9.6. Tube dilation. The tube radius increases when the surrounding polymers are oriented in the same direction as the test polymer. following effect. As the polymers orient in the same direction, the average diameter of the tube becomes larger, so that the average rotational diffusion constant increases (see Fig. 9.6). This effect is called the tube dilation. The tube dilation may be seen in, for example, the relaxation of birefringence from the highly oriented state; the initial relaxation rate is larger than the final one. A theoretical analysis of this effect has been done in ref. 2 9.4.3 Experimental study of the rotational diffusion constant The rotational diffusion constant has been measured by relaxation of the Kerr effect,'? and by dynamic light scattering.” The experimental results are in accordance with the theoretical predictions D,|Djo= B(VL*)? « pM, (9.43) or, since Dj «In(M)/M°, D,« p~?M~" In(M). (9.44) However, the absolute magnitude of D, has turned out to be quite large: experimental values of 6 range from 10° to 10*. Such large values of B have also been found by a computer simulation for thin polymers (with no thickness).'*15 Figure 9.7 shows the concentration-dependence of D, obtained by dynamic electric birefringence and computer simulation. The solid line indicates eqn (9.43) with B = 1.3 x 10°, which was obtained by Hayakawa et al."? by detailed study of the tube statistics. Though the preciseness of this value can be questioned, it is clear that the numerical336 SEMIDILUTE SOLUTIONS 1 © 00.8 ge oe Bh LN é a o \ 01 y X \ 0.01 pe woh 04 1 10 100 1000 vl? Fig. 9.7. Reduced rotational diffusion constant D,/D, is plotted against reduced concentration vL*. Filled circles: the experimental results’? of dynamic electric birefringence of poly(y-benzyl-L-glutamate, molecular weight ranging from 7.3 x 10* to 1.5 x 10°). Open circles: the result of computer simulation.’* The solid line is the theoretical result.’* (Courtesy of Prof. Hayakawa, Tokyo University.) factor is large and that the semidilute regime starts at rather large values of vL. As has already been mentioned, precise comparison with experiments is hindered by various effects such as polydispersity and bending of polymers. Quantitative theory for these effects is difficult, but the qualitative aspects have been discussed in refs 16-18 9.5 Viscoelasticity 9.5.1 Expression for the stress tensor To discuss the viscoelastic properties in semidilute solutions, we first consider the expression for the stress tensor. This is obtained from the principle of virtual work as in Section 8.6.1. In the semidilute region, the expression for the free energy is the same as that in the dilute solution since the excluded volume effect is negligible: A= J du'U(kpT InW + U,). (9.45) The change in W by instantaneous deformation dé, is again given by OW =-R- (ux de-u). (9.46)VISCOELASTICITY 337 Hence the elastic stress is given by precisely the same form as eqn (8.114) 0G = 3vkgT (uglip — 40a) — V{(u X RU, )atip )- (9.47) The viscous stress is obtained from the hydrodynamic energy dissipa- tion under a given velocity gradient. If we assume that the hydrodynamic interaction is completely screened, this is calculated in the same manner as in eqn (8.119): OG) = VEcKuy (Up ly Ua lp ) (9.48) with an, L3 be (9.49) On the other hand if we neglect the hydrodynamic screening entirely, €.., is given by the same formula as eqn (8.122) __an,L? ~ 6In(L/b)* The actual form will be between the two. Indeed the effective medium theory‘ indicates that Cote (9.50) fog numbers? (9.51) se In(1/vbL?)* . It is important to note that €,, is not affected as seriously as D,. This is because the viscous stress reflects very fast motions, for which the tube constraint is not effective.+ For simplicity we proceed using eqn (9.50). The stress tensor is given in precisely the same manner as for dilute solutions (see eqn (8.123)). Ong = 3VK pT (Ugttg — 36g) — V{(u X RU, )altp ) F VE acKuy (Up ly alls) + Is(Kap + Kea)» (9.52) 9.5.2 Linear viscoelasticity To calculate the stress, we have to solve aw FRB, RV-R- UX Kw). (9.53) 7 It has been suggeted””° that better agreement with some experiments can be obtained if baz is replaced by kpT ID, = n,L(vL*)*/In(L/b). However, this agreement is perhaps superficial, caused by non ideal effects of real polymers such as molecular weight distribution, or flexibility. Theoretically there is no reason to believe that f,,, is given by k,TID,-338 SEMIDILUTE SOLUTIONS In the linear viscoelastic region, DB, can be replaced by D,, and the solution is obtained in the same way as in Section 8.6. In steady-state shear flow with shear rate x, the terms in eqn (9.52) are given by (the case of U, =0 is being considered) vkpT _ a (vL°) EB) = = o®) = 3vkpT (ugly) 10D, K 30B in(L/b) nsK; (9.54) 3 Vv) — 242) e eck a _VE OG) = Vbsur( wey) K = HVE uK ~ op inhi ™ (9.55) and of = nx. Hence their ratio is oF): oS): $9 = B-'(vL?)?: (vL?):1. (9.56) In the semidilute regime vL*>>1, the contribution of the elastic stress is much larger than the viscous stress and the solvent stress. Thus in the ideal semidilute region of 1/L?« v<« 1/bL?, the stress can be written as Cap = 3Vk pT (Ugtts — 45ap) = 3vk5TSap- (9.57) Since the characteristic time of S,g is very large (being of the order of 1/D,), the viscoelastic behaviour becomes quite pronounced in the semidilute region. From egns (9.54) and (9.57), the steady-state viscosity is given by _vkeT_ (vL39° = F0D, 306" in(L/b) which depends on the molecular weight and concentration as n= p°M*/In(N). (9.59) The strong molecular weight dependence of 7 is in accordance with experiments,”"?> though the precise exponent is difficult to extract because of the nonideal effect discussed previously. The numerical factor B can also be obtained from the viscosity and has again turned out to be rather large ranging from 10° to 10*. (Comparison between the theoreti- cal prediction and the experimental results are given in refs 1, 19, 24, and 26) In some cases the concentration dependence of 7 is stronger than predicted by eqn (9.59) at higher concentration.~%75 One possible explanation for this is the rod jamming effect.”” The complex viscosity defined by eqn (8.139) is calculated as 3pRT_ ¢ 5M 1+iot (9.58) n*(@) = (9.60)VISCOELASTICITY 339 16? ms gt gy S O6h 04 o.2t- bev eit ii ei 0 05 1.0 15 Fig. 9.8. Real part of the complex viscosity n'(w) of poly(y-benzyl-L-glutamate) in m-cresole is plotted against concentration p at various frequencies (from top to bottom, 0, 2.2, 6.6, 20, 58, 202, and 525 KHz). Here n, is the viscosity of the pure solvent. Reproduced from ref. 28. with t= 1/6D,. (9.61) Figure 9.8 shows the real part of the complex viscosity at various frequencies.% As the concentration increases, the viscosity at low frequency increases sharply, while the viscosity at high frequency increases only in proportion to the concentration. This is in accordance with eqns (9.54) and (9.55). 9.5.3 Nonlinear viscoelasticity Since D, becomes small in the semidilute region, the nonlinear visco- elasticity becomes quite important. To handle the nonlinear visco- elasticity, we need the full solution of eqn (9.53). This has been done by340 SEMIDILUTE SOLUTIONS solving eqn (9.53) numerically’°”® with the approximation (9.41). It turns out that the qualitative features of the nonlinear viscoelasticity are quite similar to those of flexible polymers, for example: (i) Shear thinning: In the steady shear flow, the viscosity (x), the first normal stress coefficient Y,(x), and the absolute value of the second normal stress coefficients W(x) (which is negative) all decrease with the shear rate. The characteristic shear rate for the shear thinning is about D,. (ii) Stress overshoot: When a constant shear flow is started, the shear stress shows overshoot if the shear rate is sufficiently high. (iii) Elongational viscosity: The elongational viscosity first increases slightly with the elongational rate and then decreases. These features agree at least qualitatively with the observed behaviour.74-?43°% Unlike the case of flexible polymers the constitutive equation cannot be written in a simple closed form unless we use the decoupling approximation. Since the equation for Sg is given by the same equation as that in dilute solution, eqns (8.149) and (9.57) give a closed equation for Oz: a = 1 > Cap = —6D,0ng + Kay Ogu + Key Fay + 3 G.(Kap + Kpa) at — 20 ,0Krv( P+ 3808) (9.62) where G, = 3vkpT. (9.63) The qualitative features described above can be checked by this approximate constitutive equation. 9.6 Short time-scale motion 9.6.1 Chopstick model So far we have neglected the small-scale fluctuation that the polymer makes inside the tube. To include such motion, we consider the following model™ (see Fig. 9.9). We separate the direction of the polymer and the direction of the tube, and consider two vectors wu representing the direction of the polymer and a the direction of the tube. The diffusion constant of n is D,, while the diffusion constant of w is Dj» since the polymer can move freely inside the tube. The condition that wu is fluctuating inside the tube is represented by the coupling potentialSHORT TIME-SCALE MOTION 341 Fig. 9.9. The chopstick model. between u and a. U(u —n) = Su —ny. (9.64) Equation (9.64) guarantees that at a ticum, the deviation between u and n is of order e: (aay Jomese(—“s2") The kinetic equation is now given for the two-vector distribution function ®(u, n; t) as (@-ay)= =2e, (9.65) = DI: (20+; RyU-(u— ”) + Do, * (2. o+e = Ra Ue(u — n) + U .(u))) (9.66) where ®, and &, are written as a a R, =ux oa" R, =nX oa (9.67) and for simplicity the tube dilation is neglected. The model described by eqn (9.66) is easy to visualize: the two vectors u and m move together like a pair of chopsticks. The external field, represented by the potential U,(u), affects the rapidly moving vector u, which drags the slowly moving vector a through the coupling potential U.(u — n).342 SEMIDILUTE SOLUTIONS An important property of the kinetic equation (9.66) is that at equilibrium, the distribution of u is not affected by the vector n. In fact, the equilibrium solution of eqn (9.66) is ®,,(u, n) « exp[—(U.(u) + U.(u — n))/kpT). (9.68) Hence the equilibrium distribution of u is Wea(u) = [dn q(u, 9) exp[-U-(u)/koT] (0.68) which is the same as the isolated polymer without the tube constraint. This must be so since the entanglement interaction does not affect the static properties. 9.6.2 Local equilibrium approximation Although the kinetic equation (9.66) is conceptually simple, it is not easily handled mathematically. However, if we focus our attention on slow motion, a simple treatment is possible. If the external field changes slowly compared to t,~€7/D,o, the distribution of the vector u can be assumed to be in a local equilibrium for given n. Then ®(u, n; t) can be written as O(u, n; 1) = Bn; t)peq(u, 2) (9.70) where ., represents the local equilibrium distribution of u for given n, ice., _Usu—n) + Ulu) elu ye ar on) “~ fou [Hea + Ho) . xP ket arf _ Ulu —n) + U,(u) - O(n) = exp| ee | OD) Here O.(n) = —kyT in fu exo| - Rigen (9.73) B and Weq(u, m) is normalized such that fama n)=1. (9.74) To determine W(n;¢) we substitute eqn (9.70) into eqn (9.66). The