Systematic corrections to Flory–Huggins theory: Polymer–solvent–void systems

and binary blend–void systems

M. G. Bawendi and Karl F. Freed

Citation: The Journal of Chemical Physics 88, 2741 (1988); doi: 10.1063/1.454005
View online: http://dx.doi.org/10.1063/1.454005
View Table of Contents: http://scitation.aip.org/content/aip/journal/jcp/88/4?ver=pdfcov
Published by the AIP Publishing

Articles you may be interested in

Flory-Huggins swelling of polymer Bragg mirrors
Appl. Phys. Lett. 89, 164104 (2006); 10.1063/1.2358811

A theory of polymer solutions without the mean-field approximation in Flory-Huggins theory

J. Chem. Phys. 121, 4968 (2004); 10.1063/1.1781091

Kinetic theory and the Flory–Huggins approximation

J. Chem. Phys. 105, 1237 (1996); 10.1063/1.471964

Lattice theory of polymer blends and liquid mixtures: Beyond the Flory–Huggins approximation
J. Chem. Phys. 90, 2017 (1989); 10.1063/1.456044

A lattice model for selfavoiding polymers with controlled length distributions. II. Corrections to Flory–Huggins
mean field
J. Chem. Phys. 84, 7036 (1986); 10.1063/1.450625

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
213.119.64.124 On: Sun, 18 May 2014 21:32:37
 Systematic corrections to Flory-Huggins theory: Polymer-solvent-void
systems and binary blend-void systems
M. G. Bawendi and Karl F. Freed
The James Franck Institute and the Department of Chemistry, The University of Chicago, Chicago, Illinois
60637

(Received 28 May 1987; accepted 5 September 1987)

A field theory, presented earlier by us, which is formally an exact mathematical solution of the
Flory-Huggins lattice model, is used to evaluate corrections to Flory-Huggins mean field
theory in a systematic series expansion in the inverse of the lattice coordination number and in
the nearest-neighbor interaction energies. We explicitly determine the first few corrections to
the free energy of mixing for polymer-solvent-void systems and for systems containing two
kinds of polymers and voids (binary blend-void systems). Applications of the theory to the
calculation of equations of state and effective Flory X parameters are discussed. We compare
the result of our theory with the lattice Monte Carlo data of Dickman and Hall for the chain
insertion probability and for the pressure in a system of athermal chains and voids. Good
quantitative agreement is found. We discuss shortcomings of the lattice model in representing
real polymers as well as possible extensions of the model to remedy these deficiencies.

I. INTRODUCTION
Concentrated polymer solutions, melts, and blends are
systems in which the volume fraction occupied by polymers
is approaching unity. It follows then that any full statistical
mechanical theory of concentrated polymer systems must
properly describe the volume occupied by the polymer mole-
cules. Delta function pseudopotential models of polymer ex-
cluded volume, which are useful for describing dilute and
semidilute solutions l,2 but which do not incorporate the vol-
ume occupied by the polymers, thus appear to be inappro-
priate at high polymer concentrations. Lattice models, on
the other hand, by their very nature are well suited to proper-
ly account for both the volume occupied by the monomer
units on the polymer molecules and the large number of con-
figurations available to long flexible chains. Thus, concen-
trated polymer solutions and polymer blends are often de-
scribed theoretically using Flory-Huggins (FH) lattice
theory.2
In the FH model the polymer and solvent molecules are
placed on a lattice such that each lattice site is exclusively
occupied by either a polymer segment or a solvent molecule,
and nonbonded nearest neighbors of species i and j are as-
signed attractive interaction energies € ij' Flory, 3-5 Huggins, 6
and other practitioners ofthe lattice moder- l l do not solve
this lattice model. Instead, mean field approximations are
introduced, and it is this mean field approximation which is
usually termed "Flory-Huggins theory.,,2 Deficiencies of
the mean field Flory-Huggins theory are well known, as dis-
cussed below for concentrated polymer systems, and the sit-
uation becomes much worse at infinite dilution where a de-
tailed analysis has been made of the lattice model to which
Flory-Huggins theory is an approximation. Flory-Huggins
mean field treatments at infinite dilution do not properly
describe the swelling of polymers in good solvents, while
Monte Carlo treatments of the same lattice model provide a
good representation of observed behavior. 2 For concentrat-
ed systems, a fundamental question is whether the deficien-
cies ofFH theory, as partially described below, are deficien-
cies ofthe underlying lattice model or just of the mean field
approximation to that model.
In the standard FH mean field approximations the strict
single occupancy constraint is replaced by site occupancy
probability arguments which yield a mean field free energy
of mixing consisting of a combinatorial entropy and an ener-
gy of mixing. For example, the Helmholtz free energy of
mixing for a mixture of two kinds of polymers (a blend) has
the well-known form 12
M'FH,p1 ,p2
- - = --In,pl + --In,p2 + g12,p1,p2'
N/kBT NI +1 N2 +1
(1.1 )
where N/ is the total number of lattice sites, kB is Boltz-
mann's constant, T is the absolute temperature, ,pi is the
fraction of sites occupied by species i (usually referred to as
the segment fraction), and Ni + 1 is the number oflattice
sites occupied by a chain of type i. The interaction parameter
g 12 is obtained from nearest-neighbor interaction energies € ij
as 2
( 1.2)
where z is the lattice coordination number. In practice, g 12 is
treated phenomenologically in Eq. (1.1) and the free energy
of mixing [Eq. (1.1)] becomes independent of any lattice
parameters. 13
In the original formulation ofFH mean field theory the
interaction parameter gl2 as given in Eq. (1.2) is indepen-
dent of concentration, proportional to T - I, and energetic in
origin. Comparisons with experiments, however, show that
g12' treated phenomenologically, depends on polymer con-
centrations and contains both energetic and entropic contri-
butions. 13 •14-18 These empirical findings are in conflict with
the original theory as developed through Eqs. (1.1) and
( 1.2) and imply limitations with either the mean field ap-
proximation, the lattice model, or both. Theories developed
to correct this deficiency fall mainly into two categories. The
first type6--9,19 refines the probability arguments leading to

J. Chern. Phys. 88 (4). 15 February 1988 0021-9606/88/042741-16$02.10 @ 1988 American Institute of Physics 2741

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
213.119.64.124 On: Sun, 18 May 2014 21:32:37
 2742 M. G. Bawendi and K. F. Freed: Corrections to Flory-Huggins theory

Eq. (1.1) and, for example, replaces the segment fractions rPi
in the interaction term g12rPlrP2 ofEq. (1.1) with more accu-
rate site occupancy probabilities. 19 While these theories be-
gin to correct for local correlations due to chain connectivity
and successfully explain why the empirical gl2 depends on
polymer concentration, they are incapable of producing a
rigorous and systematic analysis of corrections to the origi-
nal FH mean field approximation to the lattice model. The
other type of theory, 14-16 which includes, for example, theor-
ies based on equation of state arguments, introduces con-
cepts which, although physically sensible, are unrelated to
the original lattice model or some more general continuum
model. These theories lack the rigor of being derived from a
model by statistical mechanical methods which permit the
systematic computation of either the parameters of the the-
ory or the corrections to the treatment as well as the delinea-
tion of conditions under which the approximations are valid.
Neither type of theory addresses the fundamental problem
of determining methods to provide a more rigorous descrip-
tion of the FH lattice model.
The fact that FH mean field theory, even in its most
refined versions, is not an exact solution to the FH lattice
model is clearly demonstrated in our recent work in which a
systematic treatment of the FH model beyond the mean field
approximation yields a free energy of mixing of the form of
Eq. (1.1) with an effective interaction parameter (g12)
which contains concentration dependent entropic terms and
concentration dependent energetic terms, including some
formally proportional to T -2 (Refs. 20 and 21). Such con-
tributions are for the most part not accessible through the
usual FH probability arguments. While the lattice model
itself may seem rather unphysical, it is important to be able
to distinguish between deficiencies which are due to math-
ematical approximations in the solution of the model from
those which are inherent to the lattice assumption. It is our
intent in this paper then to show that the FH lattice model
mean field approximation from the entropic and energetic
corrections to mean field. Applications of the theory to equa-
tions of state and effective Flory X parameters are also dis-
cussed in these sections. In addition, Sec. III contains a com-
parison of our theory with Dickman and Hall's Monte Carlo
data on athermal chains. 25 Details of the calculation, which
are not previously presented, are included in the Appen-
dices.
II. THEORY
The model is the same lattice model as that used by
Flory and Huggins. The polymer and solvent molecules lie
on a lattice, and each lattice site is occupied either by a poly-
mer segment, a solvent molecule, or a void. Interaction ener-
gies are assumed present between nonbonded nearest-neigh-
bor sites occupied by solvent molecules or polymer
segments. Solvent molecules are simply treated as polymers
consisting only of a single segment. This allows the identical
formalism to be used for both blends and polymer-solvent
systems.
The lattice field theory which exactly describes the FH
lattice model for these ternary systems (two types of poly-
mers and voids) is a slight generalization of that presented
previously.2o,22 The field theoretic grand partition function
Z for a system of flexible, purely self-avoiding (athermal)
chains on a lattice, with the chains of type" 1" and "2" made
up of (N I + 1) or (N2 + 1) segments and NI or N2 bonds
connecting the segments, is given by
Z = J D [rP]D [rPt] I;I (1 + XJI) +X?»
xexp { - ~ [Nt/f(q)] [~: rP~,~trP~,I~
+ N~
a=O
",(2)t
~
I
Y'q.a Cf'q,a ,
",(2) ] }
(2.1 )

does indeed contain several physically relevant contribu- where the X Y) are defined by
tions which are missed in the usual FH mean field formalism
and which are observed in a range of experiments. X Jil = H (i) L exp( - iq·r t )rP~:6t
q
We have previously introduced a field theory that is for-
N,-I
mally an exact mathematical solution of the FH lattice mod-
el. 20-24 Here a solution to the theory is provided as a system- + L Lexp(i(q-q')'rdrP~:~-lrP~!!J
a= I qq'
atic expansion beyond mean field. The first few corrections
to the mean field approximation are evaluated for polymer- +! H (ilt L exp(iqor )rP~:1/ _
t I' (2.2)
solvent-void systems and systems consisting of two kinds of q

polymers and voids (blends). The voids in our model de-
scribe free volume and are necessary to treat the changes in
volume observed upon mixing polymers or polymers and
solvent. The inclusion of voids also allows us to relax the
usual incompressibility condition, often imposed on poly-
mer mixtures, and thereby obtain equations of state.
Section II briefly summarizes the field theory which ex-
actly describes the FH lattice model. This section contains a
generalization of the field theory previously described and
can be skipped by readers more interested in the results than
in the formulation of the theory. Sections III and IV treat
The q sums in (2.1) and (2.2) are over the wave vectors in
the first Brillouin zone, Nt is the total number oflattice sites,
and the product over I is over all lattice sites, with rt the
position vector of site t. While Eq. (2.2) is simpler in real
space where the exponential factors and the sums over wave
vectors do not appear, both Eq. (2.1) and the calculations
which follow are cleaner in Fourier representation. The
functionf(q) is the nearest-neighbor structure factor,
f(q) = Lexp( -iqoap ), (2.3)
p

polymer-solvent-void systems and binary blend-void sys- where the sum is over the z lattice directions and ap is the
tems, respectively. Each of these sections separates contribu- nearest-neighbor lattice vector in the directionp. The fields
tions to the interaction parameters which arise within a (rP~:~t, rP~:~) are complex with both real (rP + rPt) and imagi-

J. Chern. Phys., Vol. 88, No.4, 15 February 1988

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
213.119.64.124 On: Sun, 18 May 2014 21:32:37
 M. G. Bawendi and K. F. Freed: Corrections to Flory-Huggins theory 2743

nary [(</J - </Jt)/i) parts ranging from - 00 to + 00. The terms ofthe averaging in Eq. (2.5) by
subscript a on the fields </J is an internal symmetry index
which ranges from 0 to N; - 1 (corresponding to the N;
bonds connecting the N; + 1 segments making up the
I(</J,</Jt) = (exp [ ~ 1];(1 + X J') + X J2»1]1 It (2.7)

chains) and which enables us to control the chain length The integration over Grassman fields is defined 21 ,26 by the
distribution [taken as monodisperse in Eq. (2.1)]. The su- relations
perscript (i) indicates that the field is associated with a chain
of type "i." The measure D [</J] incorporates the proper nor-
malization so that Z would yield unity if the X Ji) were absent
and
in Eq. (2.1). The factor of 1/2 in front of H (i)t in Eq. (2.2) is
a symmetry factor present for polymers but absent if the
species described by the X V) is a solvent molecule.
f d1]; 1]; = f d1]r 1]1 = 1.

The expansion of the product III (1 + X?) + X?) ) The average I(</J,</Jt) in Eq. (2.7) is expanded using cumu-
gives a sum of terms. A given contribution to this sum has, lants to yield
say, ~ factors of the X (j), which are associated with seg-
ments of type j, and NI - M, - M2 factors of unity which I(</J,</Jt) = exp{~ 1];(1 +XJI) +XJ 2 »1]1 + (1/2)
correspond to void occupancy. The product form III assures
that each site is occupied either by a segment or a void, while
the integral in Eq. (2.1) and the exponential weight factors X (£;> 1]11]11];" 1]m [it, jt, [exp(Eij)
serve to connect adjacent lattice sites into chains with N; + 1
segments and N; bonds. Each polymer of type "i" is associat- -1]XJi)X}.()] +O(~)}, (2.8)
ed with a factor of H (i)tH (;) in the grand partition function
in Eq. (2.1), so the n I' n2-polymer partition function Zn"n, is where the expression (1m) means that sites I and m are near-
simply obtained as est neighbors.
It is straightforward to formally generalize the theory to
treat systems with more components, such as systems con-
taining two types of polymers, solvent, and voids, three types

X an, I of polymers and voids, or solutions of block copolymers and

a [H (2)tH(2)] n, Z H(l)t=H"'=H(2)t=H(21=O
. voids. Adding more types of polymers or solvents is incorpo-
(2.4)
Nearest-neighbor interaction energies Eij, in units of
kB T, are incorporated into Eq. (2.1) by defining an averag-
ing ( ... ) I as
(X Ji)X }.(» I
=X(i)X(j){l,l,mnotnearestneighbors , (2.5)
I m exp (Eij ) , I,m nearest neighbors
with higher moments a straightforward extension of Eq.
(2.5) which is explained in Ref. 22. The interaction energies
Eij correspond to attractive energies between segments of
type i and segments oftypej. By including sequentially bond-
ed neighboring sites, the definition (2.5) overcounts the en-
ergy by an amount - N1Elln i - N 2E22n 2, where n l and n 2
are the numbers of polymers of type 1 and 2, respectively.
This extra energy is subtracted off at the end.
The definition (2.5) is incorporated into the grand par-
tition function Z in a computationally convenient fashion by
introducing a set of anticommuting Grassman fields
{q;,1]J,2I,26 for which
1]11]J + 1]J1]1 = 1];1]j + 1]j1]; = 1]l1]j
O. + 1]j1]1 =
This is shown in Ref. 21 to yield the total grand partition
function ZI for a system of interacting chains as
ZI = f D [1]]D [1]t]D [</J]D [</Jt]I(</J,</Jt)exp{'''},
(2.6)
where the quantity in curly brackets is the same as that in the
exponential of Eq. (2.1), and where I(</J,</Jt) is defined in
rated in the formalism by letting the index i on the </J~:~ , H (i) ,
and theXJi) in Eqs. (2.1) and (2.2) run over the additional
polymers or solvents.
The mean field approximation consists of retaining only
theq = ocomponents ofthe</J fields in Eqs. (2.1), (2.6), and
(2.8). Corrections to the mean field approximation are clas-
sified by the number of nonzero wave vectors q which range
over the whole first Brillouin zone. 20,23,24 The spatial indices
on 1];" and 1]m are necessarily retained to generate mean field
energies to lowest order in Eij from Eq. (2.8) and thus to
replicate the traditional mean field approximation of Eqs.
( 1.1) and (1.2). Higher order contributions in the interac-
tion energies Eij are also present when a mean field approxi-
mation is used for the fields </J, and these higher terms in Eij
provide additional corrections to the standard FH mean
field theory.21 We only provide a brief summary of the de-
tails and rules to obtain corrections in the Appendices and
refer to previous works for a more complete discus-
sion. 2o,21,23,24 New higher order contributions, which are
presented in the text of the paper, are, however, evaluated
and summarized in the Appendices.
The next two sections provide the computed Helmholtz
free energies of mixing for a polymer-solvent-void system
and for a blend-void system, respectively. Inclusion of voids
makes the systems slightly compressible and allows us to
derive equations of state. While we provide the full depend-
ence of the free energies on polymer lengths for complete-
ness, the examples are then specialized to very long polymers
in order to simplify the equations. The inverse coordination
number Z-I and the interaction energies Eij are treated as

J. Chern. Phys., Vol. 88, No.4, 15 February 1988

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
213.119.64.124 On: Sun, 18 May 2014 21:32:37
 2744 M. G. Bawendi and K. F. Freed: Corrections to Flory-Huggins theory
expansion parameters in our computation scheme. The free
energies are calculated to order Z-3, Eij' andzCy, so the lead-
ing contributions neglected are ofordersz- 2, EijZ-I, Cy, and
z~. This corresponds to formally arbitrarily taking Eij to be
of order z - 1.5 and calculating corrections to order Z-2. The
lattice is nominally assumed to be hypercubic in z/2 dimen-
sions. While free energies for different types of lattices yield
slightly different numerical factors, the qualitative aspects of
the results should remain unchanged.
III. POLYMER-SOLVENT-VOID SYSTEM
The Helmholtz free energy of mixing for a polymer-
solvent-void mixture is customarily written in the general
phenomenological form 27.28
M' /N/kBT = {,60 In {,60 + {,61 1n {,61 + [{,62/(N + 1)]In {,62
+ gOI ({,6 ){,60t,61 + g02({,6 ){,60t,62
(3.1)
where{,6o, {,61' and {,62 correspond to the site fractions occupied
by the voids, the solvent, and the polymers, respectively. The
functionsgij denote the interaction parameters between spe-
cies i andj. We use the definition
fgl({,6)=Z[E~I- ~I (,6(1-{,6)+O(E~I)]'
jg2({,6) =Z [E~2 - ~2 (,6(1-{,6) + O(E~2)]'
.rt2({,61,{,62) =Z [
E Ei I ~2 Ei2 ( - ~II
2'+4'+4'-2+ ~2
-2+EII€12+2
Note that theru are functions of the three independent inter-
action energy is € II' E22 , and € 12' In the limit of negligible free
volume ({,61 + (,62 = 1), Eq. (3.5) describes a binary system
with only a single effective interaction energy €
.rt2 ({,61,{,62 = 1 - (,61)
(3.6)
The interaction energy E of (3.6) is identical to that appear-
ing in classical Flory theory. The chain polymerization in-
dex does not appear in these mean field contributions be-
cause the effect of the mean field approximation is to treat
the positions of the monomers as uncorrelated and thus un-
connected. As a result, Eqs. (3.5) and (3.6) apply both to
polymer-solvent-void mixtures and to binary blend-void sys-
tems. The terms in Cy in Eqs. (3.5) and (3.6) are not general-
ly obtained within the traditional mean field methods. How-
ever, the mean field approximation to our lattice field theory
J. Chern. Phys., Vol. 88, No.4, 15 February 1988
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
213.119.64.124 On: Sun, 18 May 2014 21:32:37
(3.2)
throughout the paper to represent the volume fraction occu-
pied by molecules ({,6 + (,60 = 1). The remainder 1 - {,6 = (,60
is the free volume fraction.
Contributions togij arise from Eq. (2.6) which are both
energetic and entropic in origin. We define
gij=f~+f~-sij' (3.3)
where ru
is a mean field energetic contribution, f~ is an
energetic contribution due to corrections to the mean field
approximation, and sij is an entropic contribution also aris-
ing from corrections to the mean field approximation.
A. Mean field contributions
As is noted in Sec. II, corrections to the traditional FH
theory of Eqs. (1.1) and (1.2) already appear within our
definition of the mean field approximation. These correc-
tions arise from higher order terms in Eij in Eq. (2.8) and
provide concentration dependent contributions. Defining
€=EII+E22-2EI2' (3.4)
with € again in units of k B T, we obtain from Eq. (2.4) (see
Ref. 21),
(3.5a)
(3.5b)
Ei2 ) 'f'1
"'"
(3.5c)
is conveniently defined, as explained following Eqs. (2.8),
with these correction terms included.
B. Energy contributions beyond mean field
Energy contributions to the interaction parameters also
appear as corrections to the mean field approximation and
arise from the fact that the chain configurations have corre-
lations between bond spatial positions. Hence, the energy of
these configurations are not exactly described by the mean
field approximation. The corrections emerge from Eq. (2.6)
as an expansion in powers of Eij and z- I. We calculate only
the first energetic correction to the mean field approxima-
tion which is of order €ij and which gives (see Appendix B)
fbI ({,6) = 0, (3.7a)
(3.7b)
 M. G. Bawendi and K. F. Freed: Corrections to Flory-Huggins theory 2745

1:2 (rpl,rp2) Expansion ofEq. (3.9) in powers ofz- I and truncation to

order ZO produces
= (~) [ - 2€22 + €22rpl + 2€22rp2 - €lIrpl
N+ 1 gtK=.!....[Z-2(~)
+ 2€12(1 - rp2)] + O(~'€ijZ-I). (3.7c) 2 N+l
Before considering entropic contributions, it is interest- + 2rp2 ( N: 1 ) + O(Z-I) ] . (3.10)
ing to specialize Eqs. (3.1)-(3.7) to the limit of zero free
volume or voids (rpo --> 0) and to compare our energetic con- For comparison, the limit rp--> I of our expressions in Eq.
tributions to the interaction parameter gl2 with those from (3.6) and (3.7) yields
refinements of the original FH mean field approximation in
the manner of Orofino and FlorylO or Koningsveld and
Kleintjens (KK). 19 The FH mean field approximation
gl2 = {; [z - 2 (N: 1) + 2¢J2 J
~: + O(Z-I)]
yields the free energy of mixing in our notation as - z~ rp2(1 - rp2) + O(Z~)} . (3.11)
Z€
t::.F INlkB T = rpl In rpl + [rp2/(N + 1) ]In rp2 + - rp1rp2' The first corrections in Eq. (3.11) are identical to those in
2
(3.8) Eq. (3.10) and are directly related to the presence of the
local correlations discussed above. These first corrections in
The product of the site fractions rp1 and rp2 in the last term of Eq. (3.11) thus arise due to chain connectivity. Our last
Eq. (3.8) is supposed to represent the probability of having a term in Eq. (3.11),
polymer segment and a solvent molecule as nearest neigh-
bors. Equation (3.8) assumes that the polymer segments
zc rp2(1 -
-""4 rp2)' (3.12)
and the solvent molecules are randomly placed on the lattice.
This assumption disregards the connectivity of the polymer which arises from Eq. (2.6) and the treatment of the full
chains; polymer segments other than those at the ends of a spatial dependence of the field 1]1 is, however, inaccessible to
chain necessarily have two nearest neighbors which are also the type of refined mean field theories leading to expressions
polymer segments, and the two end segments on the polymer such as Eq. (3.9).
chains necessarily each have one nearest neighbor which is a
polymer segment. This local correlation can be incorporated
c. Entroplc contributions
into Eq. (3.8) by replacing the site fractions rpl and rp2 in the We have also calculated contributions to the interaction
last term of (3.8) with nearest-neighbor site occupancy parameters to order Z-2 using the diagrammatic technique
probabilities, a procedure which yields a concentration de- described in previous papers and summarized in the Appen-
pendent interaction parameter gl2 (Ref. 19), dices. 20,23,24 These contributions are corrections to the mean
field approximation of Eq. (2,6), While below we present
the full dependence of the sij on the polymerization index N
(assuming N> 2), the large N limit is sufficient for most
purposes. The calculations outlined in Appendix A yield

S01 (rp) = 0, (3.13a)

4 3
S02(rp) = -=....!.. [N I(N + 1)]2 _ 1 5N + 2N - 12N2 - 12N + 3
z 3z2 (N + 1)4
_ rp - 4N4 + 2N 3 + 12N2 2¢J2 4-3
3z2 (N + 1)4 -7 [N I(N + 1)] + O(z ), (3.13b)

S12(rpl,rp2) = -=....!.. [N I(N + 1)]2 _ J.. 3N 2 - 6N + 1 + iL N 2( ION - 12)

z r (N+1)2 3r (N+l)3
'),/,. N 2 ( ION 12) ~,/,. ,/,. '),/,.2
+_~_2 - __
V'f'_1'f'_2 [NI(N+l)]4--~_1 [NI(N+1)]4
3r (N+ 1)3 r Z2

- ~i [N I(N + 1)]4 + O(Z-3). (3.13c)

In the long chain limit Eqs. (3.13b) and (3.13c) simplify to SI2(rpl,rp2) = - J.. _ 2. + 10rfJ1 + 20rfJ22 _ ~
z r 3r 3z r
S02(rp) = - J.. _ 2.... + 4rfJ _ 2rp2 + O(Z-3,N -I)
Z 3r 3r r ' X (3rplrp2 + 3rpi + rpi) + O(Z-3,N-I).
(3.14a) (3.14b)

J. Chem. Phys., Vol. 88, No.4, 15 February 1988

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
213.119.64.124 On: Sun, 18 May 2014 21:32:37
 2746 M. G. Bawendi and K. F. Freed: Corrections to Flory-Huggins theory
D. Equation of state and Flory X parameter
An equation of state can be obtained from the Helm-
holtz free energy of mixing, corrected to include contribu-
tions beyond the mean field approximation. The derivation
follows simply from
p _ _ aMi' (3.15)
- av'
where P is the pressure and V is the total volume of the
system. Changing the volume in our model corresponds to
changing the number of voids while keeping the number of
polymers and solvent molecules constant. Denoting the vol-
ume of a lattice cell by a3 and the number of voids by no, Eq.
(3.15) is equivalent to
(3.16)
In order to carry out the differentiation, we simply re-
write Mi' of Eq. (3.1) in terms of numbers of segments of
each species rather than in terms of segment fractions. The
quantity N/ is replaced by no + n I + nz where n i is the num-
ber of segments of species i, and the segment fraction ,pi is
replaced by n;l(no + n l + n z ). For a polymer melt (,pI = 0,
,pz = ,p) and long polymers, the equation of state is calculat-
ed, including all available corrections to the mean field ap-
proximation, as
kBT
- {In(1-,p) _I -,p-,p z
P= 3
a Xeff =,pz-2{ln
X[(Z-4)E 22
2
+ z~z
4
+2.+2.]
z2
Z
3(
-,p 2E22-Z~2 - 20
-) -,p 3Z~2 6 )
4(--+-
3zz 4 z2
+ O(N -1,z-3'E2zZ-I'~2)}' (3.17)
Walsh and Deez9 have recently obtained PVT data for
high density polyethylene and have fit their data with our
Eq. (3.17) in the limit (N ..... 00). While the agreement be-
tween theory and experiment is reasonable, it is not much
different than what is obtained by simply using Flory-Hug-
gins theory. The lattice model treats the lattice framework as
rigid, with all the compressibility accounted for by the local
movements of voids. However, distortions of the lattice
structure as a consequence of phonon propagation may be
important. These distortions can be indirectly introduced
into the model by allowing the voids to interact among them-
selves through a phonon mediated effective void-void inter-
action energy. This would also prevent large void bubbles
which do not occur in real systems from forming in the lat-
tice model. Void-void interaction energies can be intro-
duced in the theory by simply treating the voids as solvent
molecules where the solvent-monomer interaction energy
E 12 is set to zero. Such a scheme may provide for better com-
parisons between experiment and theory. Although includ-
ing higher order corrections and void-void interactions may
remedy some of the discrepancies between theory and exper-
iment, the discussion in Sec. IV C on the possible shortcom-
ings in comparing the lattice theory to real experiments is
J. Chern. Phys., Vol. 88, No.4, 15 February 1988
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
213.119.64.124 On: Sun, 18 May 2014 21:32:37
applicable here also. In the next section we find excellent
agreement between our theory and Monte Carlo data 25 on
the compressibility factor. This agreement seems to imply
that it is the lattice model itself that requires improvement
rather than the need for more corrections in our theory. The
introduction of void-void interactions and different volumes
for the voids and monomers may provide for a better model.
By fixing the pressure (at 1 atm, for example) and by
solving for the site fraction of voids in terms of the segment
fractions of polymer and solvent and the temperature, it is
possible, in principle, to obtain theoretical predictions for
phase diagrams. Critical solution temperatures could then
be calculated and compared with experimental data. Al-
though these procedures are straightforward, albeit tedious,
we do not pursue them here but leave them for future works.
Instead our emphasis lies in deriving the basic free energies
of mixing which are required for these and other analyses.
An effective Flory X parameter for these polymer-sol-
vent-void systems is traditionally defined through the chem-
ical potential for solvent molecules,ul by14.16
A..ul = ~ Mi' = In(1 - ,pz)
kBT anI kBT
+(N: J,pz+Xeff,pL (3.18 )
and is readily calculated from Eq. (3.1) as
[1 ~~iz,pz] +gOI(,p)[,pO(,pO+,p2)]
a
+ nl,po -a gOI (,p) - (,potP2)gOZ(,p)
nl
a
+nz,po-a goz(,p) +glz(,pl',pz)
nl
X [,pz(,po + ,pz)] + nl,pz a~1 gl2(,pI',pZ)} . (3.19)
This produces a Xeff which is a function of both the site frac-
tions of polymers and of voids (the excess volume). In prac-
tice, effective X parameters are measured at constant pres-
sure where the excess volume of mixing is important: The
volume of the mixture is generally different from the sum of
the volumes of the individual species. 16 By solving the equa-
tion of state (3.17) for the free volume in terms of the pres-
sure and the polymer segment fraction,pz and by substituting
the result into Eq. (3.19), a Xeff parameter is determined
which is only a function of the polymer segment fraction,pz
and the pressure P. This process is probably best left for
numerical evaluation which we do not pursue here since the
main intent of the paper is to evaluate corrections to the FH
mean field approximation systematically and to note several
applications.
E. Comparison with Monte Carlo data
Comparison of the theory with data from Monte Carlo
simulations based on the same underlying model as the the-
ory is, at this point, the best test of the theory and the math-
ematical approximations involved. Computer simulations of
lattice polymers employ parameters identical to those in our
 M. G. Bawendi and K. F. Freed: Corrections to Flory-Huggins theory 2747

theory. Comparisons with experimental results, on the other After some algebra and using the results of Eqs. (3.1) and
hand, are subject to ambiguities (summarized below in Sec. (3.13c), our theory, for N> 2 and to O(Z-2), produces
IV C) as to the physical interpretations of parameters such
as z, the lattice coordination number, or (N + 1), the num- (N + l)- l lnp «(J,N) = In(1- (J) + ~[N I(N + 1) F
ber of sites occupied by the polymer. In addition, it is often z
unrealistic to model solvent, monomers, and voids as occu- 2(J 3N 4
- 8N 2 - 4N + 1
pying the same volumes; the void volumes must always be +7 (N+ 1)4
the smallest. 30,31 These problems of interpretation of param- 2(J2 5N 4 - N 3 - 6N 2
eters and limitations of the model obscure the determination
of which discrepancies between theory and experiment are
-7 (N+ 1)4
due to mathematical approximations in the theory and + 8(J3 [~]4 + O(Z-3).
which are due to the underlying lattice model itself. Once Z2 N+ 1
good agreement with Monte Carlo data is found for poly- (3.26)
mers of many lengths with different interaction energies,
While the theories of HMO, and later those of Kurata et
then a systematic comparison of the theory with experimen-
al., 9 are interesting, useful, and yield closed form expressions
tal data should resolve the physical ambiguity in interpreting
which contain corrections to the Flory approximation, their
the parameters of the theory and should help in indicating
approaches lack a rigorous basis which would permit ap-
limitations of the lattice model.
proximations to be systematically improved by a well de-
In this subsection we compare our results with the
fined successive sequence of computations. The approach of
Monte Carlo data of Dickman and Hall 25 for the chain inser-
Kurata et al., 9 for example, is developed in analogy with
tion probability and for the compressibility factor for sys-
quasiclassical or cluster variational methods and is probably
tems of athermal chains (E = 0) and voids on square lattices
the most sophisticated attempt at improving the counting
(z = 4). A later Sec. (IV D) explores the ambiguities which
procedures of Flory and Huggins. The method seeks to in-
arise when comparing the theory with experimental data.
troduce the effects of correlation between groups or seg-
The chain insertion probability p(n,N) is defined as the
ments by considering occurrence probabilities of compact
probability that a randomly generated self-avoiding chain of
clusters. Kurata et al., however, discuss the problems in de-
length (N + 1), randomly placed into a configuration of n
ciding the best clusters to use and the fact that prior studies
polymers also of length (n + 1), does not overlap or inter-
indicate that there is no guarantee of improvement as the
sect any of the n chains. The insertion probability p(n,N) is
variational cluster size is increased. It is also unclear whether
written in terms of the n-polymer partition functions Zn of
any of these methods could easily be generalized to treat
Eq. (2.4) for chains oflength (N + 1) as
ternary systems as we do here.
(n + 1) Z n+ 1 (3.20)
Expansion of the logarithm
p(n,N) = Z Z
n 1

A limiting probability p«(J,N) is assumed to exist in the ther-

- [N I(N + 1) ]In [ 1- 2; N I(N + 1) ]

modynamic limit of infinite volume,

=~ [N I(N + 1)F + ~2 [N I(N + 1)]3 + O(z3)
N['~ 00, n- 00, (N + l)nIN/-(J. (3.21) z z
(3.27)
We can relate our calculated free energies to Eq. (3.20) by
use of the definitions in the theories of HMO in Eq. (3.25) yields a term of order
F Z-I identical to ours in Eq. (3.26). However, the term of
-=lnZn (3.22) order Z-2 in this expansion differs from ours, implying that
kBT
and HMO theory is really rigorously valid only to order Z-I.
Dickman and Hall present Monte Carlo data for the
(N + 1)- l lnp «(J,N) = limit insertion probability as a function of the polymer volume
NJ-oo
fraction for chains with N = 4, 9, and 19 bonds on a two-
(N+ l)nIN, -4> dimensional square lattice (z = 4).25 They compare their
(3.23 ) simulation data with predictions using the Flory approxima-
for comparison with the simulations. tion and the HMO theory. We reproduce their data along
The Flory mean field approximation to the FH lattice with our predictions in Figs. 1, 2, and 3 for chains with
theory predicts3 ,25 N = 4, 9, and 19 bonds, respectively, in plots of
(N + 1)- l l nPF «(J,N) = In(1- (J) (3.24) - (N + 1)- l lnp «(J,N) as a function of the volume frac-
6
while the theories of Huggins, Miller, and Ouggenheim8 7 tion (J. In all three figures the solid line represents the Flory
(HMO), which contain some corrections to the Flory ap- approximation [Eq. (3.24)], the dashed line is the result of
proximation, yield 25 using the HMO theory [Eq. (3.25)], and the dot-dashed
line is our theory [Eq. (3.26)]. The simulation data points
(N + 1)-llnpHMG «(J,N) are represented by " + " signs. As reported by Dickman and
Hall, the Flory approximation consistently underestimates
=In(1-(J) - [NI(N+ l)]ln[l- 2ifJN ].
(N + l)z the insertion probability for the three chain lengths. The
(3.25) HMO theory, on the other hand, works well for N = 4.

J. Chern. Phys., Vol. 88, No.4, 15 February 1988

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
213.119.64.124 On: Sun, 18 May 2014 21:32:37
 2748 M. G. Bawendi and K. F. Freed: Corrections to Flory-Huggins theory

1.8 N=4 .9 N=19

1.6 .8

1.1 .7
0.
..s .6
M

.!-..1. 0 LIS
+ +
~ .8 ~.4
I I

.6 .3

.4 .2

.1
.2

.05 .10 .15 .20 .25 .30 .35 .~o .~5 .50 .55 .60 .65
.2 .3 .4 .5 .6 .7 .8 .9 Polymer Volume Fraction, rp
Polymer Volume Fraction. rp
FIG. 3. Chain insertion probability - (N + 1) -1 In p vs volume fraction tP
FIG. 1. Chain insertion probability - (N + 1) -1 Inpvs volume fraction tP for athermal chains on a square lattice (z = 4) with N = 19 bonds. (-) is
for athermal chains on a square lattice (z = 4) with N = 4 bonds. (-) is the prediction of the Flory approximation, (- - -) is the prediction of HMG
the prediction of the Flory approximation, (- - -) is the prediction of HMG theory, and (- . -) is the prediction through order Z-2 from the theory of
theory, and (- . -) is the prediction through order z- 2 from the theory of this paper. The + are the Monte Carlo data points of Dickman and Hall
this paper. The + are the Monte Carlo data points of Dickman and Hall (Ref.2S).
(Ref. 25).

1.5% at t/J = 0.2, the HMG theory, there, is offby a substan-

However, as N grows to 9 and 19, the HMG theory begins to tially larger 45%. The apparent trend of the HMG theory to
depart from the simulation data. The predictions from our increasingly underestimate the insertion probability as the
theory fit the simulation data reasonably well for all N, al- chain length grows while our theory stays more faithful to
though at N = 19 our theory also begins to underestimate the data is probably because of the terms of O(Z-2) which
the insertion probability. Thus, for N = 4 where the HMG are missing in the HMG theory and which involve terms of
theory works well, little difference can be seen between our order t/J, t/J2, and t/J3. This reinforces the need for evaluating
results and those of the HMG theory. However, while our systematic corrections to the Flory approximation in an un-
theory differs from the simulation data for N = 19 by only ambiguous manner that may, in principle, be made to arbi-
trary accuracy. It is of interest to note that our calculated
insertion probability applies also to d = 3 and higher dimen-
sional hypercubic lattices, and since these lattices have
z = 2d, it would appear that our predictions should then be
1.4 N=9
more accurate for d = 3.
Another quantity which can be compared with Monte
1.2
Carlo data is the compressibility factor W defined through
a(tlF IkB T) Pa\N + 1)
W= - (3.28)
M ano t/JkBT
.!-.. .8
Flory theory 25 predicts a compressibility factor
+
Z
1. 6 WF = 1- (N + 1) [In(1-t/J) +t/JJ. (3.29)
t/J
.1 and the refinements of the HMG theory yield 25
.2 2(N + 1)
W HMG = WF + t/J

X(ln(1-~~) +~~]. (3.30)

Polymer Volume Fraction. rp 2N+1 2N+1
FIG. 2. Chain insertion probability - (N + I) -1 In p vs volume fraction tP The compressibility factor with corrections to the mean field
for athermal chains on a square lattice (z = 4) with N = 9 bonds. (-) is of order Z-2 included gives
the prediction of the Flory approximation, (- - -) is the prediction of HMG
theory, and (- . -) is the prediction through order Z-2 from the theory of
24N2 - 12N + 3 }
4
W _ _ {~~ _1_ 9N -
this paper. The + are the Monte Carlo data points of Dickman and Hall
(Ref. 25). - w F
t/J z N + 1 + 3z2 (N + 1)3

J. Chern. Phys., Vol. 88, No.4, 15 February 1988

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
213.119.64.124 On: Sun, 18 May 2014 21:32:37
 M. G. Bawendi and K. F. Freed: Corrections to Flory-Huggins theory 2749

_ </i{_I_ ( - 20N 4 + 4N 3 + 24N2) } 16

3r (N + 1)3
14
_¢3{~Z2 (N
N
4

+ 1)3
} + O(Z-3). (3.31 )
~ 12
,.;
Figure 4 shows a comparison of the Monte Carlo data of ...~ 10
0

Dickman25 for chains with N = 19 with Eqs. (3.29 )-( 3.31).

r...
As with the comparison involving the insertion probability,
we see that while Flory-Huggins mean field (solid line) con-
...;::::
:>,
8
:0
....
sistently overestimates the compressibility factor, the HMG rn
rn
theory (dashed line) and our theory (dot-dashed line) are Q)
r-.
6
0-
much closer to the simulation data. It is interesting to note S
that the HMG theory and our theory with corrections to the u
0 'I
mean field approximation up to Z-2 included are indistin-
guishable for volume fractions of ¢ = 0.6 and higher. This is 2
because in the limit of complete coverage of the lattice by
polymer molecules, the HMG theory and our theory ap- 00 .1 .2 .3 .5 .7 .8
proach the Flory-Huggins mean field approximation as- Polymer Volume Fraction, rp
ymptotically. Higher order corrections to the mean field ap- FIG. 4. Compressibility factor w vs the volume fraction r/J for athermal
proximation thus become irrelevent in the limit ¢ = 1 for the chains on a square lattice (z = 4) with N = 19 bonds. (-) is the prediction
compressibility factor. of the Flory approximation, (- - -) is the prediction of HMG theory, and
While the results of this section show good agreement (- . -) is the prediction through order Z-2 from the theory of this paper. The
between our theory and Monte Carlo data for relatively
+ are the Monte Carlo data points of Dickman and Hall (Ref. 25) and
Dickman (Ref. 25).
short athermal chains, there is a need to compare the theory
with simulations which include monomer-monomer inter-
actions and possibly longer chains. The theory is written as
8"t =Ito+ltl-st, (4.2)
an expansion in powers of E and Z-l, and it is important to where Ito and/t I are, respectively, the mean field and cor-
understand how many terms need to be retained in the ex- rections to mean field energy contributions, and st is the
pansions. The comparison with Monte Carlo data shows a entropic contribution arising from corrections to the mean
growing error with higher ¢ and N and suggests that at even field approximation. The polymer-void interaction param-
larger volume fractions there is a need for additional terms in eters gogl and gg2 have the same functional form as the poly-
the z- I expansion. Monte Carlo data at higher segment frac- mer-void interaction parameter g02 ofEq. (3.1) for the poly-
tions and in higher dimensions would thus be helpful in un- mer-solvent-void system, but with all energies and polymer
derstanding where the expansion may be truncated. lengths referring to the proper polymer species. Thus gogl is
equivalent tog02 with all E22 and Ningo2 replaced by El1 and
N 1, etc., for gogl .
IV. BINARY BLEND-VOID SYSTEM
The phenomenological free energy of mixing for mix- A. Energy contributions
tures of two kinds of polymers and voids is of the form 27,28
As noted in Sec. III, the mean field interaction param-
!:Ji'/N/kBT =¢oln¢o+ [¢I/(N1 + 1)]ln¢1 eters are independent of the chain lengths. The mean field
+ [¢2/(N2 + 1)]In ¢2 interaction parameters for blends thus have the same form as
in the polymer-solvent-void mixture, so we have
+ gogl (¢ )¢o¢l + gg2 (¢ )¢o¢2
Iff =i?2' (4.3)
+ gf2 (¢1'¢2)¢1¢2' (4.1 )
where/?2 is as defined in Eqs. (3.Sc).
We again write the interaction parameters as the sum of The polymer-polymer interaction parameter Ifi is
mean field contributions and corrections to the mean field more complicated than in the polymer-solvent-void system,
approximation through the definitions and the calculations outlined in Appendix B yield

+ ¢1[2E l1 (~) - El1 (~) + E22 (~) - 2EI2 (~)]

Nl + 1 N2 + 1 N2 + 1 NI + 1

+ ¢2 [2E22 (N:~ J- E22 (N~~ J + El1 (N~~ J- 2EI2 (N:~ J] + O(Z-IEij'Cy). (4.4)

J. Chern. Phys., Vol. 88, No.4, 15 February 1988

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
213.119.64.124 On: Sun, 18 May 2014 21:32:37
 2750 M. G. Bawendi and K. F. Freed: Corrections to Flory-Huggins theory

A number of limits are of interest. In the absence of voids, the polymer-polymer interaction parameter ftl becomes a
function of the single energy parameter E ofEq. (3.4), the chain lengths, and the segment fractions <P;,

ftl(<pI'<p = 1) = - E (~) + E<PI [(~) - (~)] + O(E2,EZ- I ). (4.5)

NI + 1 NI + 1 N2 + 1
Only when NI = N 2, does Eq. (4.4) reduce, in the presence of voids, to a functional form forftl which depends only on the ef-
fective interaction energy E of Eq. (3.4),

ftl(<pI,<P2,N I = N 2) = - 2E (~) + E (~) <P + O(Z-IE,C). (4.6)

NI + 1 NI + 1

B. Entroplc contributions
The entropic contribution St2 to the polymer-polymer interaction parameter gt2 to order Z-2 is a rather complicated
function of the chain lengths N; and the polymer segment fractions <PI and <P2' The lengthy calculations of Appendix A yield

I: (A. A. ) _ (N I - N2)2 { 1 + 1 [ 10
12 'f'1''f'2 - (N
I
+ 1)2(N2 + 1)2 - -; Z2 -

+ 1 [( 18NIN 2 + ¥ NI +.If N 2)<p1 + (18NIN 2 + ¥ NI +.If N 2)<P2]

(N I + 1)(N2 + 1)

21 2 [(12NiN~ + 16NiN2 + 8NIN~ + 6Ni + 4NIN2 + 2N~ )<pi

(N I + 1) (N2 + 1)
+ (12NiN~ + 16NIN~ + 8NiN2 + 6N~ + 4NIN2 + 2Ni )<p~ + 6(2NIN 2 + NI + N 2)2<P1<P2]]) .
(4.7)

When polymers 1 and 2 are of the same length (N I = N 2 ), pursue this possibility here, instead we focus directly on a
this entropic contribution st2 in Eq. (4.7) must and does discussion of effective Flory X parameters for blends.
vanish. When the two polymers are of the same lengths, the In the limit of negligible free volume or void site frac-
only way to distinguish them is through their "color" or tion, (<PI + <P2 = 1), the expressions (4.3), (4.4), and (4.7)
type. The entropy of mixing for these two sets of objects of for the interaction parameters 8t
become quite tractable,
different "colors" is the combinatorial entropy that is al- and an effective X parameter can be computed. The Xeff for
ready fully included in the mean field and thus cannot ap- blends is experimentally defined in terms of the structure
pear in the corrections. It follows then that the expressions function Seq) through the q ..... Olimit1
derived here for the purely entropic corrections to Flory-
Huggins mean field are equally applicable to systems of S(O)=~~
a<pi N/kBT
block copolymers. A larger set of interaction energies that
serves to distinguish the two ends of the copolymer and thus
= 1 + 1 _ 2Xeff'
"colors" them would yield energetic corrections to the mean (N I + 1)<P1 (N2 + 1) (1 - <PI)
field approximation which can be calculated in the same (4.9)
manner as presented here for linear homopolymers. In the long chain limit and to order (N;- I) Eq. (4.9) yields
For long chains, Eq. (4.7) simplifies quite a bit; keeping
only terms to order N ;- 2 gives
Xeff =
E(Z -
2
2) + '"4
zc [2 1]
+ E NI + 1 - N2 + 1
St2 (<P1,<P2) = (N I-I - N 2-1)2
+ (NI-I_N2-1)2(Z~4)
X{ - ~
Z
+~
Z2
(- 10 + 18<p - 12<P2)}

+ O(z-3,N- 3). (4.8) _ 3E<p1 [_1 ___1_] _3zc2

NI + 1 N2 + 1
<P1(1-<PI)

+ O(Z-3 ,z-IE,c,N -3). (4.10)

c. Flory X parameter for blends
An equation of state for binary blend-void systems can
be obtained in the same manner as in Sec. III D. The calcula-
tion, apart from some minor additional algebraic complexity
follows similarly and needs not be repeated here. As for poly-
mer-solvent-void systems, phase diagrams can also be calcu-
lated from the theory by fixing the pressure and by solving
for the void site fraction in terms of the segment fractions of
the two types of polymers and the temperature. We do not
The contribution proportional to (z + 4/Z2) in Eq.
( 4.10) is purely entropic (temperature independent). The
term proportional to zc
is nominally proportional to T -2,
although we must remember that TE, while assumed tem-
perature independent within the model, may in comparisons
with experiments develop some temperature dependence to
account for inadequacies of the lattice model. The constant
contribution [E(z - 2)/2] in Eq. (4.10) disappears for a

J. Chem. Phys., Vol. 88, No.4, 15 February 1988

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
213.119.64.124 On: Sun, 18 May 2014 21:32:37
 M. G. Bawendi and K. F. Freed: Corrections to Flory-Huggins theory
one-dimensional linear lattice (z = 2). On such a one-di-
mensional lattice, polymers can interact with each other
only through their terminal monomers, yielding an energetic
contribution to Xeft" which is at most proportional to the in-
verse of the polymer lengths. The - 2 part in [€(z - 2)/2]
is a correction to the mean field approximation of order ;/J
which arises, as explained in Sec. III B, because polymer
segments other than those at the ends of the chains must be
bonded to two neighboring polymer segments of the same
type. It is then reasonable to expect that corrections of order
;/Jc would also yield the replacement ofzc in Eq. (4.10) by
(z - 2)c.
Higher order corrections to the mean field approxima-
tion provide a yet richer structure to the g interaction param-
eters and to Xeff' However, the corrections to FH theory
included in this paper do reveal physically relevant contribu-
tions to both the g's and Xefl" contributions which are com-
pletely lost in a usual FH mean field approximation but
which are present in the lattice model itself.
In comparisons of the theory with experimental data, a
proper choice for the model parameters NI and N2 is some-
what uncertain. In practice, the assumption of equal seg-
ment sizes inherent in the lattice model is violated31 and this
may introduce discrepancies between model predictions and
experimental data. Thus, for polymers with monomer vol-
umes VI and V 2 (V I >V2)' monomer lengths a l and a2
(a l > a 2 ) and polymerization indices n l and n 2 , a number of
reasonable choices exist for the parameters NI and N 2 • A few
of these choices follow.
(a) We may assume that NI and N2 are to be associated
with monomer volumes. If the smaller monomers are taken
to define a unit lattice volume, then we can take
(4.11 )
(b) On the other hand, polymer lengths along the chain
backbone may determine the N parameters. If the shorter
monomers define the length of a unit cell, then the relation-
ships
(4.12)
apply.
( c ) The lattice model considers completely flexible
polymers but real polymers generally require several mon-
omers to produce an equivalent flexible segment. Assuming
that nA monomers define a flexible segment length the re-
sulting N parameters are those of (b) divided by nA • A more
general approach in this case may be to introduce our recent
generalization of the lattice field theory to semiflexible poly-
mers.24
A wide range of choices of values for NI and N2 is thus
theoretically available for a given experimental system, and
we have not yet begun discussing choices for the coordina-
tion number z. In addition, as pointed out by Dickman and
Hall,30 the void size must generally be considered smaller
than the monomer size in the continuum. The uncertainties
inherent in a quantitative comparison of theory and experi-
ment at this stage illustrate the need to first compare the
theory with Monte Carlo data for the identical lattice model.
Then, extensive comparisons between experiment and the-
ory can aid in establishing a useful relation between the mod-
J. Chern. Phys., Vol. 88, No.4, 15 February 1988
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
213.119.64.124 On: Sun, 18 May 2014 21:32:37
2751
el parameters and empirical quantities. There is also a need
for extending the theory to cases where monomer (and void)
volumes need not be taken as identical. It should be straight-
forward to extend the model to general lattices and to semi-
flexible chains to introduce greater correspondence with real
polymers. Nevertheless, the results already provided by our
model calculations can provide semiquantitative guides to
the nature of corrections to the Flory-Huggins mean field
approximation.
V. CONCLUSION
The leading corrections to Flory-Huggins mean field
free energies are evaluated in this paper and provide clear
entropic and concentration dependent contributions to the
FH interaction parameters. These corrections are evaluated
using a field theory which exactly describes the Flory-Hug-
gins lattice model, a model which is only treated approxi-
mately by Flory-Huggins theory. We find that even within
our mean field approximation, concentration dependent en-
ergetic contributions to the interaction parameters appear.
We present results of calculations for polymer-solvent-void
systems as well as for binary polymer-void systems (blends).
The voids model the "free volume" found experimental-
lyI4.16: The volumes of the mixtures are generally different
from the sum of the volumes of the individual components.
Including voids also allows for the relaxation of the incom-
pressibility condition, often assumed for polymer mixtures,
and enables the derivation of equations of state.
Elsewhere, we have generalized our theory to treat semi-
flexible polymers. 24 While the algebra becomes quite in-
volved and computer assistance may be required to provide
computations to the same order as presented here for flexible
polymers, it should be possible to repeat the analysis of this
paper with semiftexible polymer systems. The use of a semi-
flexible model may impart a closer correspondence between
individual monomers and lattice volumes. The same type of
field theory can also probably be applied to polymers in the
presence of boundary walls, yielding a theory similar in spirit
to that of Scheutjens and Fleer. II Such a theory should re-
cover their results in mean field and should be capable of
providing corrections as well.
The lattice assumption is a restrictive one. Quantitative
agreement with specific experiments should not be expected.
Comparison of the theory with Monte Carlo simulations of
the same underlying model should provide a better guide as
to the validity of the truncation of our series. The Monte
Carlo data of Dickman and Ha1l 25 indicate the accuracy of
our computations for low and moderate polymer volume
fractions with the probable need for further terms at higher
volume fractions of two-dimensional polymers. Once the
problems due to mathematical approximations are under-
stood by comparing the theory with Monte Carlo experi-
ments, remaining deficiencies due to the nature of the lattice
assumption can then be addressed. Comparisons with exper-
iment provide a strong motivation for extending the theory
to permit different volumes to be occupied by monomers of
different polymers, solvent molecules, or voids. 31
Many interesting quantities may readily be evaluated
from our results, including equations of state, phase dia-
 2752 M. G. Bawendi and K. F. Freed: Corrections to Flory-Huggins theory

grams, and effective X parameters of concentrated polymer chain. Thus, the correlations are a direct consequence of
solutions and melts. By calculating many observables it both the connectivity of the chains and the nonoverlapping
should be possible to eliminate the parameters of the theory nature of the segments.
and obtain expressions relating the observables among each The dashed lines in Fig. 5 designate bonds between mon-
other and thereby provide a parameter independent check of omers at nearest-neighbor lattice sites, and filled circles des-
the theory. As these procedures are straightforward, albeit ignate monomers belonging to sequences of correlated
algebraically tedious, they are left for future works which successive monomers along a chain. The wavy lines indicate
consider comparisons with Monte Carlo and experimental that the monomers connected by these lines have at least one
data. uncorrelated (i.e., q = 0 contribution) intervening mon-
omer. Entropy contributions through z - n involve diagrams
ACKNOWLEDGMENTS with up to 2n correlating bonds (dashed lines), and all such
This research is supported, in part, by NSF Grant No. contributing diagrams are presented in Fig. 5. We summa-
DMR 86-14358 (condensed matter theory). One of us rize here the rules for evaluating the diagrams. Previous pa-
(MGB) gratefully acknowledges the support of a University pers 20,23,24 contain more details.
of Chicago McCormick Fellowship. We thank R. Dickman Assume that the diagram to be evaluated has n vertices
and C. K. Hall for providing us with their Monte Carlo data, andm bonds.
D. Walsh and G. Dee for sharing their experimental results ( 1) There is first an overall factor corresponding to the
with us, and T. Russell and A. M. Nemirovsky for stimulat- number of distinct ways of extracting the sequences of bonds
ing discussions. denoted by the diagram from the np polymer chains, keeping

TABLE I. Values for the diagrams needed in Appendix A to calculate en-

APPENDIX A: COMPUTATION OF CORRECTIONS TO tropic corrections to the mean field free energy to O(Z-2).
ENTROPY THROUGH O(z-Z)
Here we provide the diagrams needed to calculate cor- Diag. A-~t/J2+0(NI-I)
N+l
rections to order Z-2 for the entropy of mixing and summa-
rize the rules to evaluate them. Since the mean field contri- Diag.B- _ N-I t/J2 (N1 +3) + N-I t/J2+ 0 (N 1- 1)

N+I Z N+I
bution to Eq. (2.1) emerges from retaining only the q = 0
Diag. C- (N - I)(N - 2) t/J2 + O(N -I)
contributions to the product III (1 + X ~ 1) + X ~2», the cor- (N+ I) z 1

rections arise from the q;i:O terms in Eq. (2.2). These q;i:O 2 2
corrections arise by virtue oflocal correlations in the system. Diag. D- N 2 t/Ji [NI + 6] _ N t/J2
(N + 1) z (N + I) z
As described in detail in our previous papers,20,23,24 these 2
corrections may be represented in terms of a cluster expan- N .J. 2 +O(N- ' )
2(N + 1)2 '1'2 1
sion which bears some similarities to the Mayer cluster ex-
pansion that is produced from an expansion of the form
·
Dlag. N-2 N 1t/J2 °
E-------::2+0(N1)
N+I z-
IIi>j (1 + fij) with/;j the Mayerf functions. As depicted in
Fig. 5, the corrections for a system consisting of a single type
· 4N(N-I) N1t/Ji
Dlag. F - - (N + 1)2
.
r + (Dlag. . °
A)(Dlag. B) + O(N I)
of polymer are represented in terms of diagrams much like
Mayer cluster diagrams. Our mean field corrections arise Diag. G-
2N(N - 2) Nlt/J~ . . ° 3
2 ----::2 + (Dlag. A) (Dlag. E) + O(N I'Z- )
(N + I) z-
due to correlations between segments which may lie on dif-
ferent chains (diagram D), or which may be placed sequen- Dia. H- (N_1)2 N 1t/J2 (Nt/J -I) _ (N_I)2
g 2(N + 1)2 r 12 (N + 1)2
tially (diagram B), or nonsequentially (diagram C) along a 2
+ 4(N - 1)2 NIt/J + O(N0 Z-3)
• + 1)2 r
N
'j> 'j> 'i' /' P (N
I I I' , I I I' I
I.

b
I
00
,
00 bb cI '.1 Dia . 1- (N - 3)(N - 4) N 1t/J2 + O(N0)
A B c D E g 2(N+ I) r 1
3
· 8 N Nlt/J~ °
Dlag. J-3" (N + 1)3 r + O(N 1)
I·,
, I
00 'j> I·'
,, I·'
,, I
I·' , •,
,- I ,
I I I I I
2 2
I
Dia . K- 3N (N - I) Nlt/J~ _ N (N - I) Nlt/J~ (N + 9)
I , I I
00 • b 00 00 00 00 cf '00""'<:/ b
G H I g (N + 1)2 r (N + 1)3 r 1

2
+ N (N - I; t/J~ + (Diag. A) (Diag. B) + O(N~,z-3)
2(N + I) z
N ? 'i'1
'i' 'j> 'j> 'i' ? 9 'j> 'j> 'j>
1 1 1 ,- ,
I·,
I I I I I I 1
t/J~N: _ 3N Nlt/J~ + 8N
4 4 4
b b b 00
I
'0 b b b b b b 6 b Dia . L- N NIt/J1
J K L M g 2(N+I)4 (N+I)3 r (N+I)4 r r
+ (Diag. A)2(Diag. D) + (Diag. A) (Diag. 1) + O(N~,z-3)
2
Dia . M- 3N (N - I)(N - 2) N 1t/J3 + O(N0)
g (N + 1)3 r 1

FIG. 5. Diagrams used to calculate entropic corrections to the mean field

Dia . N __ 2N(N - 2)(N - 3) N12 + O(N~)
g (N + 1)2 z
free energy to order Z-2 and N 1 • Diagram A is also used in Appendix B in
the calculation of energetic corrections to the mean field free energy.

J. Chern. Phys., Vol. 88, No.4, 15 February 1988

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
213.119.64.124 On: Sun, 18 May 2014 21:32:37
 M. G. Bawendi and K. F. Freed: Corrections to Flory-Huggins theory 2753

in mind that the chains have a definite direction running polymers is computed from the mean field canonical parti-
from a = 0 to N. This is divided by the number of ordered tion function and corrections to it which are written in terms
ways of arranging the n vertices on the lattice. of a diagram expansion of the form
(2) Then choose a particular direction along the chain
S IkB = In ZMF + In{1 + diagrams}, (AI)
for each connected set of vertices and label both the vertices
separately as i l through in and the bonds connecting them where the diagrams are described above. When the term con-
sequentially as ql through qm . Each qc5' which connects ver- taining diagrams in Eq. (AI) is expanded using
tices iy and iy+ I' contributes a factor of
x2 x3
exp[ -i~'(fiy - fiy +,)] [f(qc5)/z]. In(1 +x) =x --+-- ... (A2)
2 3 '
(3) There is then an overall summation I i ,;06 ... ;06 in
l:'q""qm' where in the first sum the ix range over all lattice considerations of the extensivity of the entropy in the infinite
sites with the omission of configurations where two or more volume (thermodynamic) limit (N/ --+ (0) lead us to retain
ix represent the same site. The second sum is over the first only terms proportional to N/ in Eq. (A2). Terms in higher
Brillouin zone and excludes all qc5 = 0 terms. powers of N/ have been explicitly demonstrated to cancel
We restrict ourselves to presenting the values ofindivid- identically in lowest orders. 23
ual diagrams in Table I. The interested reader may consult If the system contains only one type of polymer, the
prior works for additional details, and the examples below diagrams contributing to Eq. (A2) to order N/lz andN/lz2
and in Table I should aid in understanding and using these are depicted in Fig. S. We give details for the evaluation of
rules further. diagrams F and I and simply provide final results for the
The entropy for a system containing one or two kinds of others in Table I. Thus, diagrams F and I are calculated as
I

(Diag. F) = N(N - I )n2(n 2 - 1) L L ' exp[ - iql' (f i , - f ,) -

i
iq2' (f i , - f )]
i
N/(N/ -I)"'(N/ -4) i,;06···;06i,q,q,q,

X exp[ - iq3' (f i• - f i, ) V(ql) f (q2) f (q3)/z3

--+ _ 4 N(N - I) tP~N/ _ N(N - 1) tP~N/ + O(N 0 ) (A3)

(N+ 1)2 Z2 (N+ 1)2 z / ,

(Diag.1) = (N - 3)(N - 4)n 2 L L ' exp[ - iql' (f i , - f i, ) - iq2' (f i , - f i ,)]

2N/ (N/ - I)'" (N/ - S) i,;06i,;06i,;06i4 ;06i,;06i. q,q,q,q.

xexp[ - iq3' (f i4 - f i,) - iq4' (f i, - fiJ V(ql) f (q2) f (q3) f (q4)/z4

(N - 3)(N - 4) tP2N/ + O(N 0 ).

--+ (A4)
2(N+I) Z2 /
The notation in Eqs. (A3) and (A4) and in Table I is as follows: The quantity (N + 1) designates the number oflattice sites
occupied by a polymer chain, N/ is the total number oflattice sites, n2 is the number of polymer chains, and tP2 = (N + 1 )n 21
N/ is the fraction of lattice sites occupied by the polymers. The arrow indicates that only the part of the diagram which
contributes to the corrections has been written. The individual factors in both Eqs. (A3) and (A4) arise straightforwardly
from the rules set forth above. For example, the prefactor
[N(N - 1)n2(n2 - 1) l/[ N[ (N/ - 1) ... (N/ - 4)]
in (A3) is a direct consequence of rule (1). The exponentials and the factors of [f(q)/z] arise from rule (2), and the
summations are presented in rule (3).
The entropy to order Z-2 in the infinite volume limit, then emerges in the form

SIN/k B = -tPolntPo-tPllntPl-~ln~+ NtP2 lnz

N+l N+l N+l
2 4 3
_ ~tP + tP tP [_ ~ N __1_ SN + 2N - 12N2 - 12N + 3
N+ 1 2 ° 2 z (N + 1)2 3z 2 (N + 1)4
_L 4 3
-4N +2N + 12N2 _ 2tjJ2 N 4 ]
3z2 (N + 1)4 Z2 (N + 1)4
1 N2 1 3N 2 - 6N + 1 tPI N 2( ION - 12)
+ tPltP2 [ - -; (N + 1)2 - r (N + 1)2 + 3z2 (N + 1)3

2tP2 N 2 ( ION - 12) N4

6¢JltP2 2tPi N4 6¢J~ N4 ]
+ 3r (N+ 1)3 -~ (N+ 1)4 - 7 (N+ 1)4 - 7 (N+ 1)4
.J. [~ 1 + _1_ N 4 - 2N 3 + 12N2 + 30N + 27] + O(Z-3). (AS)
+ 'fJ2 z (N + 1)2 6r (N + 1)4
J. Chern. Phys., Vol. 88, No.4, 15 February 1988
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
213.119.64.124 On: Sun, 18 May 2014 21:32:37
 2754 M. G. Bawendi and K. F. Freed: Corrections to Flory-Huggins theory

Contributions to the entropy for a mixture of two kinds of polymers are obtained from the diagrams in Fig. S by taking
each disconnected piece in a given diagram as belonging to either kind of polymer. Thus, for example, the values of diagrams J
and N for the blend become

D· J) 8 ~ Ni~Nk A. A. A. N O(N 0 (A6)

(lag...... i,jt! 1 (N; + 1)(Nj + 1)(Nk + 1) 'l'i'l'j'l'k I + I)'

N(N-2)(N-3)
(Diag. N)-
2
L2
I J J ¢i¢jN/ + O(N~). (A7)
",j= I (N. + 1)(Nj + 1)

The sums in Eqs. (A6) and (A7) run from 1 to 2 and there- F IkBT= -In Z _lnZMF -In(l + diagrams).
by allow each disconnected piece of the diagrams J and N to (BS)
belong to either type of polymer; all possible combinations
An analysis of the second average in Eq. (B4) follows simi-
are thus included. The extension to ternary polymer mix-
larly as for the first average leading to a new class of dia-
tures is therefore straightforward.
grams. 31 We do not pursue the analysis here since it is almost
identical to that for athermal chains but simply describe the
APPENDIX B: COMPUTATION OF CORRECTIONS TO
new diagrams that arise from it and which we denote by E
THE MEAN FIELD ENERGY
diagrams.
In this appendix we briefty describe how to calculate We thus write
corrections to the mean field energy terms systematically.
F IkBT -In ZMF
We introduce new diagrams necessary for this purpose and
explicitly evaluate the first correction to the mean field free
energy. For simplicity, we consider only a simple mixture of -In[ 1 + diagrams + ~ (€ diagrams)
monodisperse polymers and voids. Generalizations such as
those in the text to more complicated systems follow simply. +O(c) ), (B6)
The grand partition function ZI for this system is taken
from Eq. (2.6) as with the term E diagrams consisting of the sum of all topolo-

ZI J D [1]]D [¢]I(¢)
gically distinct diagrams which are described below. The
rules for constructing and evaluating the new diagrams (E
diagrams) are almost identical to those described previously
for athermal chains. Here we do not repeat all the rules and
(Bl)
the details in expressing the value of the diagrams but simply
where, as in Eq. (2.7), note the differences and give examples.
In addition to the two types of vertices used in the terms
I(¢) = (exp [ ~ 1])0 + X 1 )1]/]) I (B2) diagrams of Appendix A, we now introduce a third kind of
vertex: the cross vertex, denoted by X (see Fig. 6, for exam-
and the average ( ... >I is the same as in Eq. (2.S). We addi- pIe), which corresponds to monomers with interaction ener-
tionally define an average ( ... ) over the normalized distri- gies "turned on." The new E diagrams each contain one and
bution

rexp { - ~ [N/lf(q» :~~ ¢:,a¢q,a} (B3)

with r the normalization. €- diagrams =

) + +
To first order in the monomer-monomer interaction en-
ergy E, after evaluating the integral over the Grassman fields
1], the grand partition function Z I can be shown to become21 •,,
D+ /
/
)
ZI = {IT, (1 +X1») + EZ)' L (X/Xm
2 t;:, q
n
,¥l
(1 +X;)}
+ c:I '0

i¥m

xexp{ - iq' (r/ - rm )}f(q)/(N1z) + O(c). (B4)

0'
/
/~,,)
'
/
/
~
, ,
The first average on the right-hand side of Eq. (B4) corre-
+ + 0' '0
sponds to the grand partition function for a system of ather-
mal chains and voids and has been previously analyzed. Dia-
grams to evaluate entropic corrections to mean field have 0/
/
/~
,,
+ x/
/
J*, ,
'x + •••
been previously described (Refs. 20, 23, and 24) and used in +
Appendix A. For athermal chains we have already written FIG. 6. Diagrammatic expansion for the E diagrams in Eq. (B6). Only the
the relation [seeEq. (AI)] first few diagrams are shown as an illustration.

J. Chern. Phys., Vol. 88, No.4, 15 February 1988

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
213.119.64.124 On: Sun, 18 May 2014 21:32:37
 M. G. Bawendi and K. F. Freed: Corrections to Flory-Huggins theory 2755

only one pair of these cross vertices. A cross vertex may

replace a closed or an open vertex or may simply stand on its
own and not be connected to either a bond or a wiggly link
(again see Fig. 6, for example) . Cross vertices are now joined
together with a solid line to indicate that they must be near-
est neighbors on the lattice in order for them to interact ener-
getically. For counting purposes, the solid line is given a ?I
specific directionality.
6
The expression corresponding to the new E diagrams, as
with the diagrams for athermal chains, contains a prefactor
and a set of restricted sums. The prefactor now corresponds FIG. 7. Diagrams used to calculate the lowest order energetic corrections to
to the number of ways of extracting the sequences of bonds the mean field free energy.
and isolated monomers represented in the diagram by bonds
and cross vertices, respectively, from the n polymer chains,
keeping in mind that both the chains and the interaction (B7)
links are directional. This factor is divided by the number of q

ordered ways of arranging all the vertices in the diagram on
the lattice. Then, in addition to the factors and restricted
sums already present in the evaluation of diagrams for ather-
mal chains, if the cross vertices are labeled im and i/ and if
the interaction link connecting them is labeled with the wave
vector q, we must now also add the factor
which ensures that cross vertices correspond to nearest
neighbor monomers on the lattice (not necessarily on the
same chain).
Figure 6 shows the first few terms in the expansion of E
diagrams. The diagrams which contribute to lowest order
are pictured in Fig. 7 and we evaluate them as

(Diag. AE) = [(N+ 1)n]2- (N+ l)n I N/- 1 Iexp[ -iq'(ri, -ri,)].f(q)/z
N/(N/ - 1) i,#i, q

-+N/¢i + O(N7), (B8)

(Diag. BE)=Nn{[(N+1)n]2- 5 (N+l)n+6} I I'N/-1Iexp[-iq'(ri, -ri,)]

N/(N/ - l)(N/ - 2)(N/ - 3) i,#i,#i,#i. q' q

xexp[ - jq" (r i, - r i4 ) ]f(q)f(q')/z 2

3
-+ [N(N + 1)] [N/rfJ3 + 2N/rfJ /z + O(N7)], (B9)

(Diag, CE) = 4Nn[ (N + 1)n - 2] I '2,,'N/- 1 Iexp[ - iq· (ri, - r i,) - iq" (ri, - r i,) ].f(q)f(q')/r
N/(N/ - l)(N/ - 2) i,#i,#I, q q

2
-+ [N I(N + 1) l[ - 4N/rfJ /z + O(N7)] , (BlO)

(Diag. DE) = 2Nn I I ' N/- 1 Iexp[ -i(q+q')'(ri, -ri,)]f(q)f(q')lr

N/(N/ - 1) i,#i, q' q

-+ [N I(N + 1) l[2N/rfJlz + O(N7)]. (Bll)

The prefactor in diagram CE, for example, arises from the
rules above as follows: the factor of 4 is present because of the
directionality of the links (placing arrows along the bonds,
there are four distinct ways of linking the two arrows), the
factor of Nn arises from the Nn bonds present, and the factor
[(N + 1)n - 2] isthenumberofmonomersleftfrom which
to choose the lone cross vertex.
Inserting the results of Eqs. (B8 )-(Bll) into the loga-
rithm in Eq. (B6) and formally expanding as in Eq. (A2),
we find that the term N/ rfJ3 in diagram BE is canceled by the
term arising from the multiplication of diagram A of Fig. 5
in Appendix A and diagram AE of Fig. 7. Cancellation of
powers of N/ higher than unity can also be assumed using the
extensivity property of the free energy as discussed previous-
ly (Refs. 20, 23, and 24). Thus, to order E, we obtain the
main result of this appendix as
In Zj = In ZMF + (entropic corrections)
+ EZ rfJ2 + E(~)rfJ(1 _ rfJ)2 + O(EZ-1,zC).
2 N+l
(B12)
This equation, generalized to ternary systems, is the basis of
Eqs. (3.5) and (4.4) in the text for the energetic corrections
to the mean field free energies of mixing. Higher order cor-
rections follow analogously,
IS. F. Edwards, Proc. Phys. Soc. London 88, 265 (1966); 1. Phys. A 10,
1670 (1975). See also K. F. Freed, Renormalization Group TheoryofMa-
cromolecules (Wiley, New York, 1987); T. Ohta and A. Nakanishi,l.
Phys. A 16,4155 (1983), and references cited in the above.
2See for example P.-G. de Gennes, Scaling Concepts in Polymer Physics
(Cornell University, Ithaca, 1979), and references cited therein.

J. Chem. Phys., Vol. 88, No.4, 15 February 1988

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
213.119.64.124 On: Sun, 18 May 2014 21:32:37
 2756 M. G. Bawendi and K. F. Freed: Corrections to Flory-Huggins theory

3p. J. Flory, J. Chem. Phys. 9, 660 (1941); 10, 51 (1942).
'P. J. Flory, Proc. R. Soc. London Ser. A 234,60 (1956).
'Po J. Flory, Proc. Nat!. Acad. Sci. USA 79, 4510 (1982).
6M. L. Huggins, J. Chern. Phys. 9, 440 (1941); J. Phys. Chern. 46, 151
(1942); Ann. N.Y. Acad. Sci. 44, 431 (1943).
7A. R. Miller, Proc. Cambridge Philos. Soc. 38,109 (1942).
8E. A. Guggenheim, Proc. R. Soc. London Ser. A 183, 203 (1944).
9M. Kurata, R. Kikuchi, and T. Watari, J. Chem. Phys. 21, 434 (1953); M.
Kurata, M. Tamura, and T. Watari, ibid. 23, 991 (1955).
'''T. A. Orofino and P. J. Flory, J. Chem. Phys. 26, 1067 (1957).
"1. M. H. M. Scheutjens and G. J. Fleer, 1. Phys. Chem. 83,1619 (1979);
Macromolecules 18, 1882 (1985).
'2See, for example, the review by R. M. Masegosa, M. G. Prolongo, and A.
Horta, Macromolecules 19, 1478 (1986).
'3Polymer Handbook, edited by J. Brandup and E. H. Immergut (Wiley,
New York, 1975), p. IV-131 and references therein.
I·P. J. Flory, Principles ofPolymer Chemistry (Cornell University, Ithaca,
1953).
ISB. E. Eichenger and P. J. Flory, Trans. Faraday Soc. 64, 2035, 2053, 2061,
2066 (1968).
16p. J. Flory, Discuss. Faraday Soc. 49, 7 (1970).
17Th. G. Scholte, J. Polymer Sci. A-2 8, 841 (1970).
'"e. T. Murray, J. W. Gilmer, and R. S. Stein, Macromolecules 18, 996
(1985).
19R. Koningsveld and L. A. Kleitjens, Macromolecules 4,637 (1971).
20M. G. Bawendi, K. F. Freed, and U. Mohanty, J. Chem. Phys. 84, 7036
(1986).
21M. G. Bawendi, K. F. Freed, and U. Mohanty, J. Chem. Phys. 87, 5534
(1987).
22K. F. Freed, J. Phys. A 18, 871 (1985).
23M. G. Bawendi and K. F. Freed, J. Chern. Phys. 85, 3007 (1986).
24M. G. Bawendi and K. F. Freed, 1. Chern. Phys. 86, 3720 (1987).
2SR. Dickman and C. K. Hall, J. Chem. Phys. 85, 3023 (1986); R. Dickman
(preprint). In this paper we employ the convention of Ref. 2 and refer to
the mean field approximation as the Flory-Huggins mean field. We refer
to the more sophisticated counting arguments of Huggins (Ref. 6), Miller
(Ref.7), and Guggenheim (Ref. 8) as the HMG theory. In contrast,
Dickman and Hall refer. to the mean field approximation as the Flory
theory while they describe the more involved expressions of HMG as the
Flory-Huggins mean field.
26For an introduction to Grassman fields and Grassman algebra see, for
example, C.ltzykson and J.-B. Zuber, Quantum Field Theory (McGraw-
Hill, New York, 1980), or P. Ramond, Field Theory, A Modern Primer
(Benjamin/Cummings, Reading, MA, 1981).
27R. Koningsveld, H. A. G. Chermin, and M. Gordon, Proc. R. Soc. Lon-
don Ser. A 319,331 (1970).
2sT. Fukuda, M. Nagata, and H. Imagaki, Macromolecules 19, 1411
(1986).
29D. Walsh and G. Dee (private communication).
3OR. Dickman and e. K. Hall {J. Chem. Phys. 85, 4108 (1986) 1use data
from off-lattice simulations to explore the intluence of the void volume in
comparisons between continuum and lattice theories.
31A. M. Nemirovsky, M. G. Bawendi, and K. F. Freed, J. Chem. Phys. 87,
7272 (1987).

J. Chern. Phys., Vol. 88, No.4. 15 February 1988

This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
213.119.64.124 On: Sun, 18 May 2014 21:32:37