Download as pdf or txt
Download as pdf or txt
You are on page 1of 49

Single-chain statistics in polymer systems

A. Aksimentiev
a
, R. Hoyst
b,
*
a
Computational Science Department, Performance Materials R&D Center, Mitsui Chemicals, Inc., 1190 Kasama-cho,
Sakae-ku 247-8567 Yokohama, Japan
b
Institute of Physical Chemistry and College of Science, Polish Academy of Sciences, Kasprzaka 44/52, 01-224 Warsaw,
Poland
Received 14 June 1999; accepted 14 July 1999
Abstract
In this review we study the behavior of a single labelled polymer chain in various polymer systems: polymer
blends, diblock copolymers, gradient copolymers, ring copolymers, polyelectrolytes, grafted homopolymers, rigid
nematogenic polymers, polymers in bad and good solvents, fractal polymers and polymers in fractal environments.
We discuss many phenomena related to the single chain behavior, such as: collapse of polymers in bad solvents,
protein folding, stretching of polymer brushes, coilrod transition in nematogenic main-chain polymers, knot
formation in homopolymer melts, and shrinking and swelling of polymers at temperatures close to the bulk
transition temperatures. Our description is mesoscopic, based on two models of polymer systems: the Edwards
model with Fixman delta interactions, and the LandauGinzburg model of phase transitions applied to polymers.
In particular, we show the derivation of the LandauGinzburg model from the Edwards model in the case of
homopolymer blends and diblock copolymer melts. In both models, we calculate the radius of gyration and relate
them to the correlation function for a single polymer chain. We discuss theoretical results as well as computer
simulations and experiments. 1999 Elsevier Science Ltd. All rights reserved.
Keywords: Radius of gyration; Copolymer; LandauGinzburg model; One-loop calculations; Critical point; Orderdisorder
transition
Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1046
1.1. Linear polymers in good solvents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1047
1.2. Scaling in the Edwards model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1047
1.3. Membranes, gels and fractal polymers in good solvents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1049
1.4. Polymer melts, ideal chains and knots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1049
1.5. Polymers in bad solvents and the Q point . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1050
Prog. Polym. Sci. 24 (1999) 10451093
0079-6700/99/$ - see front matter 1999 Elsevier Science Ltd. All rights reserved.
PII: S0079- 6700( 99) 00023- 4
* Corresponding author. Tel.: 48-22-632-43-77; fax: 48-39-120-238.
E-mail address: holyst@saka.ichf.edu.pl (R. Hoyst)
1.6. Protein folding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1050
1.7. Grafted polymers and polymer brushes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1051
1.8. Polyelectrolytes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1052
1.9. Stiff chains and coilrod transition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1052
1.10. Polymers near bulk phase transitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1053
2. LandauGinzburg model for dense polymer systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1054
2.1. Denition of the LandauGinzburg model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1054
2.2. Single chain properties in the LandauGinzburg model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1058
2.3. Upper wave-vector cutoff in dense polymer mixtures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1060
3. Homopolymer blends near the critical point . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1061
3.1. Equation for the radius of gyration of a polymer chain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1063
3.2. Swelling and shrinking of polymer chains in homopolymer blends . . . . . . . . . . . . . . . . . . . . . . 1066
4. Copolymer melts near the orderdisorder transition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1072
4.1. Equations for the size of a diblock copolymer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1074
4.2. Single-chain conformations in the copolymer melts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1078
Appendix A. Partition function of homopolymer blends with a labeled chain . . . . . . . . . . . . . . . . . . . . . 1085
Appendix B. Ideal average of microscopic density operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1086
Appendix C. Partition function of the diblock copolymer melt with a labeled chain . . . . . . . . . . . . . . . . . 1088
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1090
1. Introduction
The history of polymer science dates back to the end of the 19th century when August von Kekule in
1877 expressed the idea that molecules of life such as cellulose are composed of long, chain-like
molecules. The idea of molecules that are much longer than a unit cell in a crystal was not easily
accepted by the German crystallographers and until 1930 when a polymer molecule was pictured as
an aggregate of smaller molecules [1,2].
Hermann Staudinger, the man who coined the term macromolecules for polymers, believed that
macromolecules are stretched rod-like structures. His idea was based on the linear relation between
the viscosity of polymers in solutions and their molecular mass. This idea was dispelled by Kuhn and by
Flory [3], who based their calculation of the shape of a macromolecule on the rotation around the
covalent bonds in the linear polymer. From their calculations emerged a new picture of the shape and
size of a macromolecule, namely, the coiled structure with the conguration resembling the trajectory of
the Brownian particle. In particular, the size of the coil structure R in this picture scales with a molecular
mass as M
0.5
or with a polymerization index (number of monomers), N as N
0.5
and the coil follows
Gaussian statistics. It should be noted that the Brownian trajectory is only a crude approximation of the
polymer conguration. They are similar only when viewed at a large distance scale when all the
molecular details of the system are washed out. Moreover, the analogy is not quite correct since, as
observed already in 1934 by W. Kuhn and independently by E. Guth and H. Mark, a polymer chain
should avoid itself, while a Brownian trajectory can cross itself. The new and important physical effect
introduced at this point was the effect of excluded volume. At the molecular level in vacuum the
excluded volume for two monomers in a polymer chain is the volume in space inaccessible to one
monomer due to the presence of the other. For example, an excluded volume for two spheres is a sphere
of the radius twice as big as the radius of a single sphere. In a solution an excluded volume does not
follow from the quantum mechanics steric interactions described above, but is related to the
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1046
thermodynamics of a solvent. Therefore the models should take into account not the bare interactions
between the monomers but rather the effective interactions between the monomers following from the
Gibbs free energy of the surrounding solvent.
1.1. Linear polymers in good solvents
The analogy between the Brownian trajectory and the polymer chain brought into the science of
polymers two new concepts: the Feynmann path integrals [4,5] and scaling [5]. The renormalization
group calculations followed these concepts [5,6]. Another theoretical method which followed from the
path integral formulation of the polymer problem was a self-consistent eld theory (SCFT) [4,5]. The
scaling relations and SCFT with excluded volume included in the Hamiltonian of the polymer chain
gave a Flory exponent for the radius of gyration in 3d, i.e. R N
n
n = 0:6. In general n = 3=(d 2),
where d is the dimensionality of space. The Flory value of the exponent is close to the one obtained by
the renormalization group technique [6], n = 0:588, Computer simulations [8] gave an exponent
n = 0:59. In 2d space the exponent [6] is exactly equal to 3/4. In the four-dimensional space the linear
chains assume the Gaussian shape (with the exponent 0.5) since the self-avoidance effects are negligible
in this case. The scaling analysis of the Edwards Hamiltonian [6] with the Fixman delta interaction [7]
showed that although in principle one should have many-body effective interaction potential in a
polymer chain, only the two-body potential (and in some cases three-body potential) does not vanish
in the limit of N - when rescaled by the appropriate power of N. We discuss the scaling in the next
section.
1.2. Scaling in the Edwards model
The simplest model of a polymer molecule is the Gaussian model. Each monomer in the chain has an
extension given by the Gaussian probability distribution. In the continuum limit the conformation
distribution for such a chain is given by [9]:
W[r] = const: exp
3
2l
2
_N
0
dn
2r(n)
2n
_ _
2
_ _
; (1)
where l is the Kuhn length, r(n) is the location of the nth monomer and N is the total number of
monomers. Here n is assumed to be a continuous variable measuring the length along the chain and
W is the functional of all the positions of the monomers r in space. The Gaussian chain does not describe
correctly the local structure of the polymer, but does correctly describe its large length-scale properties.
In particular it gives the scaling R = AN
0:5
. The exponent is correctly predicted (e.g. for polymer melts)
only the constant A has to be calculated for any particular local structure of the polymer. The examples of
calculations of A can be found in e.g. Ref. [10]. For the scaling between R and the molecular mass M one
nds R = aM
0:5
(R is given in Angstrom units) where a is usually of the order of 0.10.3 [10]. This
scaling is valid for polymer melts (to be discussed in the following subsection) but not in good solvents
where polymers experience excluded volume interactions. The part of the conformation distribution due
to the interactions is modeled in the Edwards model by the following equation:
W
int
[r] = const: exp
bv
2
_N
0
dn
1
_N
0
dn
2
d(r(n
1
) r(n
2
))
_ _
(2)
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1047
Here b = (k
B
T)
1
is the Boltzmann factor. The delta interactions introduced by Fixman [7] represent
simply the fact that on the mesoscopic scale of the polymer chain the interactions are short-range and can
be approximated by the delta function. The coefcient v follows from the integration of the solvent
degrees of freedom in the full partition function for the solvent and polymer. This effective interaction is
related to the second virial coefcient of the monomers or in general to the direct correlation function for
two monomers integrated over the volume of the system. From the discussion it follows that the model is
mesoscopic and not microscopic. It should describe correctly the properties of the interacting chain on
large length scales much larger than the Kuhn length l. The model includes two-body interactions and
neglects the effective three-body and higher terms. The total conformation distribution including
interactions between the monomers of the chain is simply
W
tot
[r] = const: exp(H
tot
=k
B
T) = W[r]W
int
[r] (3)
where H
tot
is the mesoscopic Hamiltonian for the polymer molecule. Let us consider the properties of the
model in the long-length limit N - . If we do the following rescaling of our variables r(n) = N
n
R(n
/
)
and n = Nn
/
we nd the Hamiltonian in the following form:
H
tot
= N
2n1
3
2l
2
_1
0
dn
/
2R(n
/
)
2n
/
_ _
2
N
2dn
v
2
_1
0
dn
/
1
_1
0
dn
/
2
d(R(n
/
1
) R(n
/
2
)) (4)
where d is the dimension of space. Now let us consider the order of both terms with respect to N. When
N - the dominant term in the Hamiltonian will have a higher exponent in N. If
2n 1 2 dn (5)
then the Gaussian term dominates over the interaction term and the chain should be ideal. If we set n =
0:5 we nd that inequality (5) is satised for the space dimension d greater than 4. For d 4 the
interactions scale out in the limit of N - . Now if
2n 1 = 2 dn (6)
both terms in the Hamiltonian have the same order of magnitude and this equation gives the exponent for
the radius of gyration n. We nd for d = 3 n = 0:6 and for d = 2, n = 0:75, which are the Flory
exponents for the radius of gyration of linear polymers in good solvents. This simple scaling allows
us to consider the effect of many-body interactions on the exponent n. Let us consider an m-body term in
the Hamiltonian. Such term scales with N as N
mdn(m1)
: For d = 3 and n = 0:6 we nd that the two-
body term dominates over the many-body interactions and the latter scale out in the long-chain limit.
Therefore they are not relevant for the large scale properties of the polymer chain. Only when the
exponent n is 1/3 all the terms in the Hamiltonian are relevant for the structure of the polymer coil.
The exponent n = 1=3 corresponds to the state at which the polymer chain is densely packed into a small
globule. This happens below the Q point (discussed in one of the following sections). Finally we note
that so far we have tacitly assumed that the interaction parameter bv is of the order of unity. However at
high temperatures or at the Qpoint it can be much smaller than unity. For example in the case of polymer
blends above the critical point we nd the Gaussian behavior of the chains since the critical temperature
T
c
= (k
B
b)
1
scales in this case linearly with N and consequently vb 1=N.
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1048
1.3. Membranes, gels and fractal polymers in good solvents
The Edwards Hamiltonian has been also applied to polymers that are not linear, but form, for example,
a two-dimensional membrane [1115]. Let us consider a polymer as a D-dimensional object (D = 1 for
a linear chain, 2 for a membrane and 3 for a gel) in a d-dimensional space. The total mass of the D-
dimensional object is given by L
D
= N, where L is the linear size of the object. One can use the results of
the previous section to obtain the Flory exponents, by replacing the one-dimensional integrals by the D-
dimensional integrals. The Flory exponent n = (2 D)=(2 d); relates R and L i.e. R L
n
, therefore
for a linear polymer we have 0.6, for a membrane 0.8 and for a gel 1. In a gel, additionally, we have the
usual Flory exponent 0.6 for the scaling of the distance between the crosslinking points with the number
of monomers between the points. Below dimension d
1
= 4D=(2 D) the self avoiding effects are
relevant. For the space dimension d d
1
the self-avoiding effects can be neglected and the ideal
exponent follows: n = (2 D)=2. For linear chains (D = 1) d
1
= 4 as calculated in the previous section,
but for membranes (D = 2) for any dimension d of the space we have to include the effect of self-
avoidance. Another problem comes from the analysis of the many-body interactions. For linear chains
the many-body interactions are irrelevant in the long-chain limit. This is not the case for membranes.
The m-body interaction term is relevant for d 2mD=(2(m 1) D), therefore for D = 2 and d = 3
the three-body, four-body and ve-body interactions are relevant [12].
Let us now discuss the size of the polymer chain in a fractal pore of the fractal dimension d
f
. The
distance traveled by the random walker in the fractal should scale with time as r
2
t
d
s
=d
f
where d
s
is the
spectral dimension of the fractal describing its connectivity independently of its spatial conguration
[1517]. The equilibrium conguration of the self-avoiding polymer chain in the fractal pore should
depend on both d
s
and d
f
. In particular for the ideal chain in the fractal we have n = d
s
=2d
f
, according to
the denition of d
s
and analogy between the random walk trajectory and the ideal polymer chain
conformation. For a self avoiding polymer chain in a fractal pore one nds using the SCFT [18]
n = 3d
s
=d
f
(2 d
s
). As we see if d
s
= d
f
= 3 we retrieve the well-known exponent 0.5 for the ideal
chain and 0.6 exponent for the self-avoiding walk in the ordinary three-dimensional space.
One can also imagine more complicated structures of fractal polymers. Suppose that we have a fractal
polymer of spectral dimension d
s
. All the above results concerning the size of the fractal polymer in the
three-dimensional space can be easily generalized if instead of D we use d
s
[15,16]. Let us consider an
example of the trajectory of the Brownian particle moving on another trajectory of the Brownian
particle. It can be proven rigorously [19] that in this case 6=11 d
s
=d
f
2=3 and since d
f
= 2 for a
random walk trajectory we nd 12=11 d
s
4=3. The best estimate [19] for d
s
is 6/5.
1.4. Polymer melts, ideal chains and knots
The case of a dense polymer system, without any solvent is particularly simple. In a homogeneous
polymer melt the number of contacts between the monomers does not change with swelling or shrinking
of the polymer chains. The surroundings of a given chain in the melt is the same regardless of the shape
of this chain. Although with the swelling the number of contacts of the monomers belonging to the
swelling chain decreases, the number of contacts with the monomers from the other chains increases
exactly matching the former decrease. It follows that the chain should assume the Gaussian statistics
with the ideal exponent 0.5. Howcan we reconcile this fact with the interactions between the monomers?
As already noted by Flory and calculated by Edwards [20,21,24] the effective interactions between two
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1049
monomers belonging to the same chain are screened due to the interactions with other chains. If we
integrate out all the degrees of freedom of all the chains except for one we nd that the effective
interactions between two monomers in that chain are zero when averaged over the volume of the system.
Thus the chains are ideal. Although the chains are ideal they are also strongly entangled and the
nontrivial element of the chain statistics in the melts is their topology related to the knots formation.
From the standpoint of the topology a ring molecule represents a knot. Since in a melt the distance
between the ends of a linear polymer chain is much smaller than the contour length of the chain, it can be
considered as quasi-closed and its topology can be discussed from the point of view of the knot theory
(for a recent review see Ref. [22]). The problem of knots in polymer chains is known as the Delbruck
problem. One nds that the probability that the ideal chain will form a knot is 1 in the limit of N - .
The relevant length in this case is the topological persistence length N
T
which is the average
distance along the chain between two subsequent knots. For a knotted chain N
T
N. The knots
affect, for example, the dynamics. When the system is rapidly quenched into the glassy state and
next the temperature is raised the long-time dynamics is associated with the untightening of tight
knots formed in the quench process [22]. The knots are also important in biological systems (see for
example Ref. [23]).
1.5. Polymers in bad solvents and the Q point
So far, we have discussed a behavior of a polymer chain in a good solvent when the effective
interactions between the monomers are repulsive. However as we lower the temperature the effective
two-body interactions will become zero at the Q temperature. At this temperature the chain will assume
the size of the Gaussian chain with an exponent 0.5 [2426]. The three-body interactions give only
logarithmic corrections to the chain statistics. Below the Q temperature the two-body interactions
become attractive and the chains shrink very fast assuming a globular shape with R N
1=3
. This
phenomenon is called a coilglobule transition. The compact structure of the globule leads to many-
body interactions between the monomers which do not scale out in the limit of the innite chain length.
The globule consists of a dense nucleus and a relatively thin surface layer. The size of the globule settles
at such a value that the osmotic pressure in the nucleus is zero [27]. The coilglobule transition is located
in a region close to the Q point. The size of the region on a temperature scale is of the order of N
0.5
.
Thus in the limit of N - the transition takes place exactly at the Q temperature and can be identied
with a true phase transition. The sharpness of the transition depends on the chain stiffness i.e. stiffer
chains have the sharper transition analogous to the rst-order phase transitions [27]. The shape of the
globule also depends on the stiffness of the chain. For stiff chains with large persistent length the shape
of a globule is toroidal [27].
1.6. Protein folding
The coilglobule transition is closely related to the long-standing problem in biology known as
protein folding. Proteins are linear heteropolymers which are built of 20 different amino acids. At
physiological conditions (near neutral pH at 2040C) the proteins assume a native folded structure.
The structure is unique for each protein and determines its biological functions, therefore the problem of
the protein folding received a name of the problem of the second genetic code. It should be noted that a
small protein of 100 residues has 10
100
possible conformations, yet only one of them is the native
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1050
conformation. At high temperatures or acidic or basic pH the unique folded structure unfolds (denatures)
often reversibly (i.e. when cooled it returns to the same native conformation). Under such conditions the
global conformation of a globular protein is a random coil having low segment density similar to an
ordinary polymer above the Q temperature. The origin of the coilglobule transition in proteins is
hydrophobic interactions with water. In the native structure the hydrophobic part of the protein is
sequestered into a compact core separated from the water solvent. The detailed role of different inter-
actions in the nal architecture of the folded protein is still a controversial issue. There are a number of
problems studied in connection with protein folding: why proteins fold into a unique state despite the
fact that the number of possible congurations is enormous; how do they fold and why the folding
transition from the open coil to the nal globular (for globule proteins) state is so fast (Levinthal
paradox)? With the development of new computational techniques physicists and chemists are quite
close to the solution of these problems [2830]. In particular, it is nowpossible to predict the detailed 3D
architecture of the folded protein chain from the knowledge of the linear sequence of its chemical
constituents using Monte Carlo simulations, but so far the reliable results have been restricted to
short chains [3133]. It should be noted that despite the fact that we know more than 400 000 proteins,
the 3D folded structure is known for only 400 of them.
1.7. Grafted polymers and polymer brushes
Polymers grafted to the surface either by physical or chemical means are important from the tech-
nological point of view. For example, in the preparation of masks for the production of microchips or for
colloidal stabilization, chromatography, adhesion, spreading and wetting properties. A polymer brush
consists of long polymer chains attached to the surface with a sufciently high coverage density (number
of polymers per unit area) so that chains stretch away from the surface. The behavior of the grafted
chains as a function of the quality of the solvent (good or bad solvents) surrounding the grafted layer and
the surface density of chains (coverage) has been studied by Alexander [34] and de Gennes [35] using
scaling arguments and blob models. They have obtained the following results. In the limit of low
coverage in a good solvent the chains have the same statistics as in the bulk solvent. In particular R
N
0:6
l: This scaling is correct providing that the chains at the surface do not overlap i.e. when the coverage
sR
2
1 or simply sl
2
N
6=5
: The average concentration prole near the surface is f(z) = sl
2
(z=l)
2=3
;
where z is the distance from the wall and f(z) is the density of monomers. When sl
2
N
6=5
the chains
start to stretch and the thickness of the grafted layer L = Nl(sl
2
)
1=3
: The prole f(z) according to de
Gennes is given by z
2=3
up to the distance z = s
1=2
from the wall. From that point the prole is at and
f(z) = (sl
2
)
2=3
: However, as shown by Milner et al. [3640] using a self-consistent eld theory, the
prole in the regime of long chains and moderate coverage is not at but parabolic. The chain can be
pictured as a linear chain of blobs stretched normal to the wall. In the direction parallel to the wall the
typical distance covered by a single chain is of the order of N
1=2
: In a bad solvent the thickness of the
grafted layer is Nsl
3
[34]. The small-angle neutron scattering experiments [41] conrmed these predic-
tions concerning the height of the grafted layer in good and bad solvents. In the case of a polymer of
length Nl grafted at the surface in the melt of the same polymers of length Pl(P N
1=2
) the size of the
grafted chains scales as N
0:5
provided that sl
2
N
1
[35]. Stretching occurs when sl
2
PN
3=2
and in
the intermediate regime N
1
sl
2
PN
3=2
the grafted chains overlap but this overlap does not lead to
stretching. Finally for the coverage sl
2
P
1=2
the mobile chains are expelled from the grafted layer
[35]. The brushes that consist of two types of the chains exhibit a phase separation phenomenon [42]. A
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1051
brush subjected to the shear ow increases its height since the ow stretches the chains [43]. The solvent
owing past the grafted layer can penetrate into the layer quite deeply due to the parabolic monomer
density prole [44]. The polymer brush can also enhance the spreading [45]. Polymer brushes have been
also studied in computer simulations [4649].
1.8. Polyelectrolytes
So far we have discussed the polymer systems where the local microscopic scale does not affect the
scaling properties at the mesoscopic scale. In polyelectrolytes the short-range and long-range interac-
tions are simultaneously present and details of the local chain structure can be altered by the long-range
interactions. This nontrivial coupling between the microscopic and mesoscopic length scale is further
complicated by the Debye screening of electrostatic interactions which introduce an intermediate length
scale into the problem. Therefore the predictions of various mesoscopic models may differ considerably
when applied to polyelectrolytes [50]. In particular the conformations of the charged polymer depend on
the fraction of charged monomers, concentration of salt and in the case of weakly charged macromo-
lecules also on short-range interactions [24,27]. In a salt-free solvent a polymer molecule carrying a total
charge fN (i.e. f is the fraction of charged monomers, the rest is neutral) has the size R Nf
2=3
(l
b
l
2
)
1=3
;
where l
b
is the Bjerrum length which characterize the strength of the electrostatic potential in the solvent
(for water at room temperature l
b
= 7

A). For the nite fraction of the charged monomers the charged
macromolecule assumes a rod-like conguration, while for f N
3=4
(l=l
b
)
1=2
it is a coil with the Gaus-
sian radius of gyration (neglecting the short-range interactions). In the latter case the chains behave as
neutral chains. This simple scaling follows from the comparison between the electrostatic interactions
and the Gaussian bead-spring Hamiltonian (Eq. (1)). When the salt is present two effects come into play:
the DebyeHuckel screening length l, which can vary from 10 to 1000 A

, and the ion condensation


(known as Manning condensation, although the phenomenon was predicted by Onsager in 1947 [27]).
The importance of l follows from the fact that on the length scale smaller than l the chains are
essentially rigid and only on the larger scale the chains assume the coil conguration (of size N
3=5
in
a good solvent). However as shown by Odijk, Skolnick and Fixman, the electrostatic screened interac-
tions may induce a rod behavior on length scale larger than the screening length [50]. This effect
introduces a new length scale, known as the SkolnickFixmanOdijk length. The detailed discussion
of various effects in polyelectrolytes are contained in the review article by Barrat and Joanny [50] and
the book by Grosberg and Khokhlov [27]. Therefore, we shall not discuss it further.
1.9. Stiff chains and coilrod transition
In the exible molecule modeled by Eq. (1) the bond angle between neighboring segments is not
restricted and bending of the molecule does not cost any energy. As we have seen however in the case of
polyelectrolytes we may have an induced stiffness due to the electrostatic repulsion. We can also have a
natural stiffness arising from the structure of the chain. In a DNA molecule two interwoven helices
provide a mechanical stiffness. On a scale given by the persistence length the chain is stiff and start to
deect from the straight conguration on a larger length scale. For example, the DNA molecules has the
persistence length of 500 A

. Such chains are quite well described by the KratkyPorod worm chain
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1052
model [51]. The worm-chain conguration of an ideal (noninteracting) chain follows:
W[r] exp
1
2
k
1
k
B
T

N
n=1
u
n1
u
n
_ _
(7)
where k
1
is the bending elastic constant, which is the energy penalty for the change of angle between the
nearest monomers and u
n
= (r
n1
r
n
)=l is the unit vector tangent to the chain and Nl = L is the length
of the chain. The continuous version of this is dened as follows [52]:
W[r] exp
1
2
k
1
k
B
T
_N
0
dn
2u(n)
2n

2
_ _
(8)
subject to the constraint u
2
= 1: The mean-square end-to-end distance is simply
R
2
endend
= l
2
_N
0
dn
_N
0
dn
/
u(n)u(n
/
) (9)
where the average is taken with respect to the distribution given by Eq. (8). One nds
R
2
endend
= l
2
N
3D
1
1
ND
(exp(2DN) 1))
_ _
(10)
where D = (2k
1
=k
B
T)
1
; hence P = 2l=D is the persistence length. At high temperatures P 1 and the
chain assume the Gaussian statistics with R
endend


N
_
: At sufciently low temperatures P Nl and
the chain assumes the rigid rod congurations [5154] with R
endend
Nl: The stiff chains usually
interact with the typical interactions favoring the orientational ordering. Such interactions favor, at
low temperature, the orientationally ordered nematic phase. At some temperature we have the isotro-
picnematic phase transition. The transition is accompanied by the expansion of the chain length along
the direction of the orientational ordering [53,55]. In fact just below the transition temperature the chains
stiffen exponentially with temperature. Therefore it is legitimate to say that the coilrod transition
accompanies the isotropicnematic phase transition in the system of the semi-exible chains [56].
Actually we may have a positive feed-back: stiffening of the chains induces the isotropicnematic
phase transition and the phase transition induces stiffening [57]. The conguration of the polymer in
the nematic phase is not however a linear straight rod [53,55]. The chain makes a number of rapid U-
turns which are called hairpins. Only at very low temperatures the hairpins disappear and the chains are
straight, long molecules gently uctuating around the straight conguration. The review of the experi-
mental studies of the conformation of stiff chains in various ordered liquid crystalline phases is given by
Cotton and Hardouin [58]. The discussion of various theoretical aspects of the conformations of the
worm-like chain with the application to DNA molecules is given by Odijk [59]. The discussion of the
various liquid crystalline phases formed by DNA is done by Livolant and Leforestier [60]. Finally we
note that recent experiments allow to determine various elastic constants of the single DNA molecule
[61,62].
1.10. Polymers near bulk phase transitions
The phase transitions in dense polymer systems usually take place at high temperatures. For example
in the homopolymer blend the critical temperature scales linearly with the polymerization index N.
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1053
Similarly the orderdisorder transition temperature in diblock copolymer melts scales linearly with N.
Hence, the values of the excluded volume interaction, v, at which the phase separation in a polymer
mixture occurs scales as (1/N). Under this condition, the mesoscopic Hamiltonian H (Eq. (4)) is domi-
nated by the part which describe the ideal chain conformations. It means that the interactions divided by
k
B
T are small above the transition temperature and the chains should be Gaussian. However near the
phase transition the uctuation effects become strong and they can change the shape of the Gaussian
chain. Below we present the detailed analysis of the corrections to the Gaussian shape of the chain
arising from the critical uctuations near phase transitions for homopolymer blends and copolymer
melts. The model appropriate for the study of phase transitions is the LandauGinzburg mesoscopic
model. We will present the detailed derivation of the LandauGinzburg model for polymer systems
from the Edwards Hamiltonian.
2. LandauGinzburg model for dense polymer systems
In order to describe a dense polymer mixture on a mesoscopic length scale one should introduce order
parameter elds and relate them to density operators which specify the microscopic structure of the
system. The model is often specied by the Hamiltonian describing the interactions between the
monomers, and the density distribution function of the monomers within a polymer molecule. In the
mean-eld approximation the effects of the long-range concentration uctuations on a single chain are
neglected. The inuence of the critical uctuations near the phase transition can be taken into account
within the one-loop approach as will be described in the next sections.
2.1. Denition of the LandauGinzburg model
The information that N monomers are connected to form a chain is specied by the distribution
function W[r]: To describe the exible polymer, the chain model is used, in which atoms are described
as being joined by freely rotating bonds of xed length l. The normalized distribution function for N
atoms in such a chain is given by [63]:
W[r] =

N
n=1
d(u
n
l)
4pl
2
; (11)
and is normalized as
_
DrW[r] = 1 (12)
with
Dr =
1
V
dr
0
dr
1

dr
N
: (13)
Here u
n
= r
n
r
n1
is the vector specifying the orientation of the monomer and r
n
is the location of n
point between the subsequent monomers. d is the Dirac delta function. Instead of the previous model one
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1054
can use the Gaussian model in the discrete version:
W[r] exp
3
2l
2

N
n=1
r
n
r
n1

2
_ _
; (14)
or a continuous version of the model,
W[r] exp
3
2l
2
_N
0
dn
2r(n)
2n
_ _
2
_ _
: (15)
In both cases one normalizes the distributions. The same quantitative results within negligible correc-
tions of the order of (1=N
a
); a 1 are obtained at the level of the LandauGinzburg free energy when
one uses the distribution functions (14), (15) or (11).
A exible diblock copolymer molecule with N
A
and N
B
monomers in each block can also be described
by the distribution function (Eq. (11)) one assumes the equal segment lengths in both of the blocks (and
N = N
A
N
B
).
The rigid polymers are described as rigid rods, or needles, in which all bonds have not only the same
length, but also the same direction. In this case
W[r] =
d(u
1
l)
4pl
2

N
j=2
d(u
j
u
j1
): (16)
Finally, for a diblock copolymer in which a exible chain consisting of N atoms has been joined with a
rod containing M atoms,
W[r] =

N
i=1
d(u
i
l)
4pl
2

N M
j=N 1
d(u
j
u
j1
): (17)
The case of semiexible polymers [52] has been discussed in Section 1.9.
To describe the interactions between monomers, one rst notes that the typical length scale in the melt
of exible chains is given by the radius of gyration i.e. the size of the region occupied by the chain in the
melt. In the simplest Gaussian approximation one nds that it is proportional to

Nl
_
and is much larger
than the monomer size l. All the interesting phenomena take place at the length scale proportional to the
radius of gyration. On the other hand the range of the potential is proportional to l, and therefore is not
relevant to the phenomena occur on a mesoscopic length scale. Guided by this simple observation the
following short-range effective interaction potential has been proposed by Fixman [7] and used by
Edwards in his model [4]
v
ab
ij
(r
i
; r
j
) = w
ab
d(r
i
r
j
); (18)
where d is the Dirac delta function and w
ab
is the effective interaction parameter for a and b type
monomers (e.g. A, B monomers in the binary homopolymer mixture). The effective interaction para-
meter is given by the integral [4,64] of the direct correlation function [65]. In the rst approximation for
the rigid molecules, such integral, does not depend on temperature and is equal to the excluded volume,
the volume inaccessible to one molecule when the other is xed in space. Apart from the repulsive forces
there is also an attractive potential e.g. van der Waals potential. The direct correlation function is, in the
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1055
rst approximation proportional to this potential. Thus w
ab
contains two contributions: one associated
with the excluded volume and one with the attractive potential. Summarizing, the interactions given by
Eq. (18) are related to the microscopic interaction potential via the direct correlation function. The delta
function signies the extreme short-range character of the potential as measured in terms of the length
given by the radius of gyration.
The interaction potential for rigid, elongated molecules, depends not only on the positions but also on
the orientations of the molecules. In the case of two rods [66] with xed orientations the excluded
volume has roughly the shape of a rectangular box. If we expand the excluded volume in terms of the
Legendre polynomials of the cosine of the relative angle of two rods we obtain the constant term,
proportional to w
ab
, plus the second term related to the second Legendre polynomial, P
2
. If higher
order terms are neglected the anisotropic part of the potential is modeled as [67,68]:
v
ab
i;j
(r
i
; u
i
; r
j
; u
j
) = v
ab
d(r
i
r
j
)P
2
u
i
u
j
u
i
u
j

_ _
: (19)
The total potential is given by the sum of Eqs. (18) and (19). Please note that in general the parameters
w
ab
and v
ab
are not independent, since they follow from the same direct correlation function.
For the sake of clarity we consider further a mixture of polymers which consists of the monomers of
two types (A and B).The density operators needed to specify the density distribution of the monomers of
type A and B in the system are as follows:
^
f
A
(r) =
1
r
0

n
g=1

i={A}
d(r r
g
i
) (20)
^
f
B
(r) =
1
r
0

n
g=1

i={B}
d(r r
g
i
); (21)
where

{A}
or

{B}
denotes the summation over all A- or B-type monomers in the gth chain, r
0
is the
number density of all monomers in the system and n is a number of chains in the system. These two
operators,
^
f
A
(r) and
^
f
B
(r), represent the microscopic number fraction at point r of A and B monomers,
respectively. If we also consider the interactions between the molecules which depend on their orienta-
tions the next two operators should be introduced:
^
Q
(A)
ab
(r) =
1
r
0

n
g=1

i={A}
d(r r
g
i
)
3
2
(u
g
i
)
a
(u
g
i
)
b
u
g
i

2

d
ab
2
_ _
; (22)
^
Q
(B)
ab
(r) =
1
r
0

n
g=1

i={A}
d(r r
g
i
)
3
2
(u
g
i
)
a
(u
g
i
)
b
u
r
i

2

d
ab
2
_ _
(23)
They represent the microscopic nematic tensor order parameters [66] at point r for the A and B
monomers, respectively. These tensors are symmetric and of zero trace, thus each has ve independent
components.
The total interaction Hamiltonian is obtained by summing all the interactions between monomers.
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1056
Using the density operators it can be rewritten in the following simple form:
H = H
I
H
O
; (24)
where
H
I
= r
0
_
dr[w
AA
^
f
2
A
(r) w
BB
^
f
2
B
(r) w
AB
^
f
A
(r)
^
f
B
(r)]; (25)
and
H
O
= r
0
_
dr[
1
3
v
AA
^
Q
(A)
ab
(r)
^
Q
(A)
ba
(r)
1
3
v
BB
^
Q
(B)
ab
(r)
^
Q
(B)
ba
(r)
2
3
v
AB
^
Q
(A)
ba
(r)]: (26)
Here summation over repeated a and b indices is implied. Only the rst part, H
I
, of the Hamiltonian H is
relevant in the case of the exible polymers since the orientation dependent part H
O
in this case is
negligibly small after averaging over all possible conformations of the polymer chains. The part of the
Hamiltonian, H
I
, leads to the macrophase separation in homopolymer blends and to the microphase (or
mesophase) separation in copolymer melts providing w
AA
w
BB
2w
AB
= 2k
B
Tx 0: Here x is
the FloryHuggins parameter. The orientation dependent part of the Hamiltonian, H
O
, is important for
the systems which contain macromolecules with rigid, elongated parts providing nematic ordering in the
systems [67,68]. There is no general recipe for the choice of the microscopic operators. The hint as to the
right choice is provided by the form of the interactions between monomers and the chain architecture.
Now we can specify the mesoscopic model. Let P
i
be the average values of the microscopic operators
^
P
i
over the mesoscopic volume with the characteristic length L. The conditional partition function,
Z[P
i
]; is the partition function for the system subjected to the constraint that the microscopic operators
^
P
i
are xed at some prescribed values [69] of P
i
, i.e.
Z[P
i
] =

i
d(
^
P
i
P
i
); (27)
where the average is calculated as follows:

=
1
Z
0

n
a=1
_
Dr
a

W[r
a
]exp
H
k
B
T
_ _
: (28)
Here Z
0
is the canonical partition function, Dr
a
denotes the measure Eq. (13). The LandauGinzburg
(LG) free energy (in the mean-eld approximation), V[P
i
], is given by
V[P
i
] = k
B
T ln Z[P
i
]: (29)
The conditional partition function and the LG free energy are functionals of the order parameters P
i
. The
partition function of the system is given by the summation of the conditional partition function over all
possible congurations of the elds P
i
:
Z =

i
_
DP
i
Z[P
i
]; (30)
For the microscopic density operators
^
f
A
(r) and
^
f
B
(r) their respective mesoscopic functions f
A
(r)
and f
B
(r) give the averaged values of the number fraction at point r of A and B monomers, respectively.
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1057
It is convenient to introduce two new elds [7072]:
C
A
(r) = f
A
(r) f
A
(r); C
B
(r) = f
B
(r) f
B
(r); (31)
which describe the excess of the fraction of A or B monomers at the point r over their average values in
the system. It is usually assumed that the system is incompressible, i.e.
f
A
(r) f
B
(r) = 1; (32)
so
C
A
(r) = C
B
(r) = C(r): (33)
For exible polymers the eld C is the only order parameter in the system. It vanishes in the
homogeneous state but is nonzero at the coexistence curve of the homopolymer blend and in the ordered
region of copolymer melts.
2.2. Single chain properties in the LandauGinzburg model
The local structure of a dense polymer mixture can be studied within the described above model by
introducing additional microscopic operators which contain information about the single chain proper-
ties [73,74]. For example, the radius of gyration of a single polymer chain
R
2
=
1
2N
2

N
i=1

N
j=1
(r
(1)
i
r
(1)
j
)
2
; (34)
can be found by introducing a single chain microscopic operator
^
f
(1)
(q) =
1
r
0

N
i=1
exp(qr
(1)
i
): (35)
The radius of gyration (Eq. (34)) is obtained from the single-chain correlation function
S
(1)
(q; q) =
^
f
(1)
(q)
^
f
(1)
(q); (36)
by differentiating twice with respect to q and taking the limit of q - 0. In the system composed of
different types of polymer molecules the single chain microscopic operators should be introduced for
each type of the macromolecules. If the system contains polymer molecules which have chemically
different parts, the single-chain microscopic operators can be specied for each of the parts. One can also
investigate orientational correlations between different macromolecules or their parts considering
single-chain microscopic operators which describe orientational properties of the macromolecules.
For example, the average cosine of the angle between the segments of a polymer chain,
1
2N
2

N
i=1

N
j=1
cos u
ij
=
1
2N
2

N
i=1

N
j=1

u
i
u
j
u
i
u
j

; (37)
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1058
can be studied by considering the following microscopic operator
^
Q
(1)
(r) =
1
r
0

N
i=1
1
u
i

d(r u
i
): (38)
The average cosine of the angle between the segments is obtained from the correlation function
S
(1)
u
(q; q) =
^
Q
(1)
(q)
^
Q
(1)
(q) (39)
by differentiating twice with respect to q and taking the limit of q -0. In a similar way, the average
value of the quantity (cos u
ij
)
2
can be obtained from the correlation function
S
(1)
J
(q; q) =

3
a=1

^
J
(1)
a
(q)
^
J
(1)
a
(q); (40)
where the microscopic operators are
^
J
(1)
a
(r) =
1
r
0

N
i=1
(u
i
)
a
u
i

2
d(r u
i
): (41)
One should not be misled by the microscopic character of the operators which have been used to
specify the local structure of the system. The physical quantities which can be determined in this
approach give only the information about their values which are averaged over a characteristic meso-
scopic length scale. This length scale, L, appears in the model as a cutoff for certain integrals describing
the long-range uctuations.
The next step to nd out the single-chain properties is to calculate the correlation functions such as
Eqs. (36) and (39), etc. For this purpose one can introduce an additional external eld Uwhich couples to
the single chain microscopic operator
^
f
(1)
only and consequently add the term
_
dr
^
f
(1)
(r)U(r) to the
interaction Hamiltonian H. The single chain correlation function is obtained as
S
(1)
(q; q) =
d
2
Z[U]
Z[U]dU(q)dU(q)

U=0
; (42)
where:
Z[U]
Z[U = 0]
= exp
_
dq
(2p)
3
^
f
(1)
A
(q)U(q)
_ _ _ _
: (43)
The average is given by Eq. (28). This partition function, Z[U], can be conveniently evaluated using the
cumulant expansion and the Legendre transform [73,74].
At the end of the calculation one gets the following equation for the single chain correlation function:
S
(1)
(q; q) =
^
f
(1)
(q)
^
f
(1)
(q)
0
corrections; (44)
where

0
denotes the ideal average:

0
=

n
a=1
_
Dr
a

W[r
a
]: (45)
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1059
The corrections in the RHS of Eq. (44) are given as an innite expansion of Eq. (42) with increasing
numbers of integrals over q. The natural small parameter in this expansion appears in connection with
the upper wave-vector cutoff 2p=L which limits the range of the integration. In the case of the homo-
polymer blends [75,76] and diblock copolymer melts [77] L

N
_
and therefore the expansion is in the
powers of 1=

N
_
: For long chains the expansion can be limited to the rst term. These corrections depend
on the order parameter correlation functions, in particular, on the collective structure factor. Therefore,
in order to describe the single-chain statistics near the phase transitions one should supply Eq. (44) by the
set of equations which include long-range uctuation corrections to the order parameter correlation
function. If one restricts the expansion of the single-chain correlation function to the rst-order terms
the convenient framework to describe the order parameter correlation function is the one-loop self-
consistent (Hartree) approximation [7881].
At the rst step of the calculation of S
(1)
(q; q) we assume a certain ideal statistics for a single
polymer chain by specifying the distribution function, W[r], which gives zero-order terms in these
expansions (independent from the cutoff). The higher-order corrections perturb the ideal chain statistics
due to the long-range uctuations (cutoff dependent). It is important to note that all the correlations at the
length scales smaller than the characteristic mesoscopic length scale L have not been ignored. They
have been taken into account within the ideal chain statistics. We have used the fact that the interactions
between the monomers in a dense polymer system are too weak to affect strongly the ideal chain
conformations and assume that the change of the single chain statistics near the phase transition is
determined only by the long-range concentration uctuations. In the following sections we describe in
more detail some applications of this approach to the homopolymer blends and copolymer melts.
2.3. Upper wave-vector cutoff in dense polymer mixtures
As was mentioned in the previous section the corrections to the radius of gyration due to the long-
range uctuations depend on the upper wave vector cutoff. This quantity is important, because the
corrections to the mean-eld theory would dominate the properties of the system if the upper-wave
cutoff is not introduced. Although the cutoff is ubiquitous in statistical mechanics there are no ready
recipes for its choice. In low molecular mass liquids we usually have one natural length scale, which
corresponds to the size of a molecule and the cutoff usually corresponds to this length scale. In polymer
mixtures the problem is more complicated since we have three different length scales: the total length of
a polymer molecule, Nl, the size of the region occupied by a polymer molecule (proportional to the
radius of gyration),

Nl
_
; and nally, the microscopic length scale l, corresponding to the size of a
single monomer. We assume that L is proportional to the radius of gyration, that is to

N
_
. Here we
repeat the same arguments as given elsewhere [75]. At high temperatures, when the interactions between
the monomers are irrelevant and the monomermonomer correlation function decays exponentially with
the characteristic correlation length proportional to the radius of gyration. Moreover, in all our calcula-
tions the specic structure of monomers has not been taken into account. Therefore the microscopic
length scale is irrelevant in the model. Finally, if we chose the microscopic length scale as the cutoff the
uctuation corrections would survive in the limit of N - and the mean-eld FloryHuggins model
and RPA model for the scattering intensity would not be the correct description of the system in this
limit. We believe this is not the case. We note, that since the size of the polymer molecule in the blend is
roughly proportional to

N
_
, this choice is also in accordance with the prescription known from low
molecular mass systems, where the cutoff corresponds to the size of the molecule. Since in the polymer
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1060
system this size changes with temperature and concentration, it would be desirable to determine the
cutoff consistently from the theory. The equation for the radius of gyration, R, obtained from the single-
chain correlation function (Eq. (44)) offers a simple way for a self-consistent determination of the cutoff
if we set it equal to 2pC=R where C is some constant which can only be determined from the full
microscopic theory. Thus, apart from this constant, the cutoff is determined from the mesoscopic theory.
Therefore, the equation for the collective structure factor S
c
(1=V)C(q)C(q) (taken within the
one-loop approximation) and the single-chain structure factor S
(1)
(q; q) are coupled via the self-
consistently determined radius of gyration and the cutoff.
This prescription is good for the symmetric homopolymer blend. In the case of the asymmetric blends
we postulate that the cutoff should be equal to (2p=[L(R
A
; R
B
)]); where L is the symmetric function of
the two radii of gyration for A and B chains. We assume the scaling form for L : L(R
A
; R
B
) = R
A
f (x)
where x = R
A
=R
B
: From the symmetry properties of L, we nd the equation for f(x):
1
x
f (x) = f
1
x
_ _
: (46)
We choose the following form for f (x) from the set of possible solutions of that equation:
f (x) =

C
1
1 x
3
1 x
C
2
x
_
: (47)
Here C
1
and C
2
are two constants which cannot be determined from the mesoscopic theory since they
depend on the microscopic details of the system. In general they are additional phenomenological
parameters in the mesoscopic theory. We assume here that they are of the order of unity.
In the case of the diblock copolymer melts there are three independent contributions to the total size of
a diblock polymer molecule: the size of block A, R
AA
, the size of block B, R
BB
, and the distance between
the centers of mass of the blocks, R
AB
. We postulate that the cutoff is equal to q
upper
=
2p=L
M
(R
AA
; R
BB
; R
AB
); where L
M
is the symmetric function of the block sizes R
AA
and R
BB
. We assume
the scaling form for L
M
:
L
M
(R
AA
; R
BB
; R
AB
) = R
BB
h(R
AA
=R
BB
; R
AB
=R
BB
) = R
AA
h(R
BB
=R
AA
; R
AB
=R
AA
): (48)
From the symmetry properties of L
M
, we nd the equation for h:
h(y; x) = xh
y
x
;
1
x
_ _
; (49)
where x = R
AA
=R
BB
and y = R
AB
=R
BB
: The simplest solution of this functional equation is h(x; y) =
C(xy)
1=3
; and the cutoff now has the following form:
q
upper
=
2pC

R
AA
R
BB
R
AB
3
_ ; (50)
where C is the constant which depends on the microscopic details of the system.
3. Homopolymer blends near the critical point
A homopolymer blend consists of two types of linear macromolecules (say A and B) which are mixed
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1061
together without a solvent. The state of this system is xed by two parameters: the concentration of the
A-type polymer molecules and temperature. At high temperatures the system is a homogeneous viscous
polymer liquid. If one decreases the temperature below the critical temperature the difference in the
interactions between the monomers A and B drives the system to the phase separation (macrophase
separation). The system separates into the A-rich and B-rich phases.
The critical behavior of the homopolymer blend belongs to the Ising universality class [82,83].
Nonetheless in the limit of very long chains the Ginzburg region around the critical point decreases
and eventually in the limit of innitely long chains the mean-eld behavior is recovered even innitely
close to the critical point. Despite the well-known facts about the global properties of the system (such as
critical behavior) there are still some unresolved issues concerning the local structure of the blend. In
particular it is interesting to know how do the single chain conformations change when the system
approaches the phase separation?.
The blend where the polymerization index of the rst component is much smaller than the polymer-
ization index of the other component resembles the system of polymer chains in a solvent. Indeed, a
single polymer chain of N A-type monomers in an incompatible melt of B-type chains with P monomers
in each chain (P pN) exhibit a coilglobule transition [84] similar to a collapse of a single polymer
chain in a poor solvent. At high temperatures, when the difference between the interactions of distinct-
type monomers are negligible in comparison to k
B
T, the polymer blend should have the same properties
as a melt of the same-type polymers. In the blends of long and short chemically identical polymer chains
the long chains are swollen if their index of polymerization N P
2
, where P is the polymerization index
of the short chains [85]. A single polymer ring in a melt of linear polymer chains has also swollen
conformations due to the topological constraint of being unknotted with itself [86,87]. Moreover, rings
of A monomers can be compatible with linear chains of B monomers, even when linear chains of A and
B are not compatible. It is also known that the melt of ring polymers have statistics intermediate between
those of collapsed and Gaussian chains [86]. A FloryHuggins treatment suggests that the radius of
gyration of such rings scales with the index of polymerization as R N
2=5
.
The conformations of the macromolecules in blends and solutions are studied experimentally by
light-, X-ray-, or neutron-scattering. These experimental techniques allow to measure the radius of
gyration, R, molecular weight, and the FloryHuggins interaction parameter, x, (or the second virial
coefcient, A
2
) [85,8890]. Small angle neutron scattering (SANS) experiments [58] typically involve
the measurement of scattering from dilute labeled (deuterated) chains in a matrix of unlabeled (proto-
nated) chains. For example, the radius of gyration, R, is determined from the asymptotic form of the
scattering intensity in the limit of small q:
S(q) r
0
N(1 q
2
R
2
=3); (51)
where r
0
is the monomer number per unit volume. The whole form factor
P(q) =
1
N
2

N;N
i;j
exp iq(r
i
r
j
) (52)
allows to determine the chain conformation by tting the experimental data to various models of P(q)
(for a recent review see Ref. [58]).
The calculation of the single-chain conformations in a dense homopolymer blend on a microscopic
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1062
level is a difcult task. Therefore a number of mesoscopic models [20,72,9193] and computer
simulation techniques [9499] are used to study this problem. In both cases a very complicated inter-
action potential between the monomers is approximated by some simple, tractable form.
The FloryHuggins lattice model is often used in the Monte-Carlo computer simulations. In this
model the polymer chains are arranged on a lattice. Each site of the lattice can be occupied by the A- or
B-type monomer. The chains are represented by a self-avoiding random walks with the number of steps
equal to the polymerization index. The interactions between the monomers are modeled within some
simple approximation. For example, the system gains the energy 1 if two neighboring sites of the lattice
are occupied by the monomers of the same type [94]. There are three different ways of modeling the
Monte-Carlo motions of the lattice chains [95]. The rst one is applicable for the system where there is a
signicant number of non-occupied sites. This method includes a drawing of a chain segment and a
nearest neighbor position, and, if the nearest site position is vacant, the motion is performed. If the
motion is not possible, the old conformation is retained. At higher densities the randomly chosen
nearest neighbor position is rarely vacant, and it is better to draw a vacancy and a nearest neighbor
position. If this nearest neighbor is a suitably arranged chain segment, the motion is performed. Finally,
when all sites on the lattice are occupied, only a collective rearrangement motions can take place
[97,100].
Despite the great benets of the computer simulations [96], they can be practically performed only for
the blends with small polymerization indices. On the other hand, the mesoscopic descriptions for
polymer blends which start from the Gaussian model are practical in the limit of long chains. An
example of such mesoscopic model is the Random Phase Approximation (RPA) developed by
de Gennes [24]. The theories based on the generalization of the reference interaction-site model
(RISM) integral-equation theory to homopolymer systems [64,101,102] also use the Gaussian
model for single chain conformations as in RPA. The large concentration uctuations in the
system with nite chain lengths near the phase separation may affect the Gaussian statistics. In order
to study this problem we derive the equation for the radius of gyration of a single polymer chain with
the explicit inclusion of the uctuations. As we show in the following sections the critical uctuations
do not perturb the Gaussian statistics in the homopolymers blends near the critical point. Large
deviations from the Gaussian statistics occur at low temperatures and for the disparate sizes of the A
and B chains.
3.1. Equation for the radius of gyration of a polymer chain
Let us consider a mixture of n
A
A-type polymers, with N
A
monomers in each molecule, and n
B
B-type
polymers, with N
B
monomers in each molecule, inside a volume V. We will use the chain model in which
monomers are joined by freely rotating bonds of xed length (exible chains). The distribution function
for N monomers in such a chain is given by Eq. (11). The part of the interaction Hamiltonian corre-
sponding to the mixture is given by the following expression [4,71] (see also Eq. (25)):
H
I
= r
0
_
dq
(2p)
3
1
2
w
AA

^
f
A
(q)
2

1
2
w
BB

^
f
B
(q)
2
w
AB
^
f
A
(q)
^
f
B
(q)
_ _
; (53)
where r
0
= (n
A
N
A
n
B
N
B
)=V is the density of monomers in the system,
^
f
A
(q) and
^
f
B
(q) are Fourier
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1063
transforms of the microscopic concentration operators:
^
f
A
(q) =
1
r
0

n
A
a=1

N
A
i=1
exp(qr
a
i
); (54)
^
f
B
(q) =
1
r
0

n
B
b=1

N
B
i=1
exp(qr
b
i
): (55)
In the case of the homopolymer blend the conditional partition function, Eq. (27), is the partition
function for the system subject to the constraint that the microscopic operators
^
f
g
(q) (g = A; B) are
xed at some prescribed values [69] f
g
(q); i.e.
Z[f
A
; f
B
] = N
0

n
A
a=1

n
B
b=1
_
Dr
a
Dr
b
W
A
(r
a
)W
B
(r
b
)

g=A;B
d[f
g
(q)
^
f
g
(q)] exp
H
I
k
B
T
_ _
: (56)
Here N
0
is a constant, Dr
g
denotes the measure (13), the interaction Hamiltonian H
I
is given by Eq. (53),
W
A
and W
B
are given by Eq. (11), and f
g
(q) (g = A; B) are Fourier transforms of the concentrations
f
A
(r) and f
B
(r); respectively.
The radius of gyration of an A-type polymer chain [9],
R
2
A
=
1
2N
2
A

N
A
i=1

N
A
j=1
(r
(1)
i
r
(1)
j
)
2
; (57)
can be found from the single chain correlation function S
AA
(q; q) by differentiating twice with respect
to q and taking the limit of q - 0 (see Section 2.2). the single chain correlation function,
S
AA
(q
1
; q
2
) =
^
f
(1)
A
(q
1
)
^
f
(1)
A
(q
2
); (58)
with the single chain microscopic operator dened as
^
f
(1)
A
(q) =
1
r
0

N
A
i=1
exp(qr
(1)
i
) (59)
is obtained from the partition function Z[U
A
] (Eq. (43)) which can be conveniently written in the
following form [73,74,76]:
Z[U
A
] =
_
Df
A
_
Df
B
exp
H
I
k
B
T
_ _
_
DJ
A
_
DJ
B
exp i
_
dq
(2p)
3
f
A
J
A
i
_
dq
(2p)
3
f
B
J
B
F
(n
A
)
A
[J
A
; U
A
] F
(n
B
)
B
[J
B
]
_ _
: (60)
F
(n
g
)
g
is the free-energy for the system of n
g
non-interacting Gaussian chains in the external elds J
g
, U
A
,
and is given in the form of a cumulant explanation [71,75] (see also Appendix A). One can integrate out
the elds J
A
, J
B
in a saddle point [75], take the Legendre transform to the elds (31) and write the
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1064
partition function Z[U
A
] as
Z[U
A
] =
_
DC
A
_
DC
B
exp
H
I
k
B
T
G
A
[C
A
; U
A
] G
B
[C
B
]
_ _

g=A;B
1

det[d
2
F
(n
g
)
(J

g
)=dJ

g
(q
1
)dJ

g
(q
2
)]
_ ; (61)
where J

g
denes the saddle point:
dF
n
g
(J
g
)
dJ
g
(q)
= iC
g
(q): (62)
Now we take the functional derivative of Z[U
A
] twice with respect to U
A
(q), put U
A
(q) = 0, and then
eliminate all the terms that vanish in the thermodynamic limit. Additionally we impose the condition of
incompressibility C
A
= C
B
= C (see Eq. (32)). In this way the approximate equation for the single
chain correlation function is obtained, i.e.:
f
(1)
A
(q)f
(1)
A
(q) = f
(1)
A
(q)f
(1)
A
(q)
0

1
2
_
dk
(2p)
3
G
A
2
(k; k) G
A
2
(k; k)
1
V
C(k)C(k) 1
_ _
[
^
f
(1)
A
(q)
^
f
(1)
A
(q)
^
f
(1)
A
(k)
^
f
(1)
A
(k)
0

^
f
(1)
A
(q)
^
f
(1)
A
(q)
0

^
f
(1)
A
(k)
^
f
(1)
A
(k)
0
]:
(63)

means the average with partition function (56), and

0
means the average (45). The RHS of Eq.
(44) contains the zero-order term f
(1)
A
(q)f
(1)
A
(q)
0
and the rst-order corrections. These corrections
are of the order of (1=

N
_
) since the upper wave-vector cutoff has been chosen as (const:=R); were R is the
radius of gyration, proportional to

N
_
.
In order to obtain the correct description of the collective structure factor C(q)C(q) =
V=(G
2
(q; q)) near the critical point the set of the self-consistent one-loop equations [75] have been
used, i.e.:
G
2
(q; q) = G
(0)
2
(q; q) DG
(0)
2
(q; q)
1
2
_
dk
(2p)
3
G
(0)
4
(q; q; k; k)
G
2
(q; q)

1
2
_
dk
(2p)
3
[G
3
(q; k; q k)]
2
G
2
(q; q)G
2
(q k; q k)
; (64)
G
3
(q; p; q; p) = G
(0)
3
(q; p; q p) DG
(0)
3
(q; p; q p)

1
2
_
dk
(2p)
3
G
3
(q; k; q k)G
3
(p; k; p k)G
3
(q p; p k; q k)
G
2
(k; k)G
2
(q k; q k)G
2
(p k; p k)

1
2
_
dk
(2p)
3
G
(0)
5
(k; k; p; q; p q)
G
2
(k; k)
: (65)
They have been derived using the loop expansion [78] of the partition function of the system, Eq. (30).
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1065
Due to the cutoff the loop expansion is also the expansion in the small parameter (1=

N
_
): Here we denote
the ideal term corrections coming from the denominator of Eq. (61) as DG
(0)
2
and DG
(0)
3
: The long-range
concentration uctuations are taken into account in the set of equations (64) and (65) as one-loop
corrections [75]. The critical temperature and concentration determined from these equations are shifted
in comparison to the mean-eld values, but in the limit of N - the FloryHuggins theory and RPA
results are recovered.
The relevant change of the radius of gyration of an A-type polymer chain R
A
in comparison to the
Gaussian value R
0
=

N
A
_
l
A
=

6
_
is given by the equation:
R
2
A
R
2
0
= 1
0:062

N
A
_ 1
1

f

f
(l
A
=l
B
)
3
_ _
_
R
0
=Ldx
A
f (x
A
)
g
D
(x
A
)
G
(A)
2
(x
A
)
G
2
(x
A
)
1
_ _
; (66)
where

f = n
A
N
A
=(n
A
N
A
n
B
N
B
) is the average fraction of A monomer in the system, x
A
=

N
A
_
ql
A
=

6
_
; and L is a symmetric function of the two radii of gyration for A and B chains which
determine the upper wave vector cutoff (see Section 2.3). Our assumption that the mixture is incom-
pressible is expressed in the following form:
V
n
A
N
A
l
3
A
n
B
N
B
l
3
B
= 1: (67)
The functions g
D
(x), f(x) and the other ones needed to calculate the vertex functions in Eqs. (64) and (65)
are given in Appendix B.
3.2. Swelling and shrinking of polymer chains in homopolymer blends
Although as we shall see the effect of swelling and shrinking is very small in homopolymer blends of
long chains, it is still instructive to nd out where and why it occurs. We will also see that even a very
small change of the single chain statistics can lead to a large increase of the scattering intensity.
Fig. 1 shows the behavior of the radius of gyration as a function of the FloryHuggins parameter x
(proportional to the inverse of the temperature) and the fraction of the A-type monomers

f =
n
A
N
A
=(n
A
N
A
n
B
N
B
) for the mixture with equal indices of polymerization (N
A
= N
B
= 1000):The
solid lines indicate the cross-over between the system with swollen and the system with shrunk chains.
On these lines, the chains of A-type (line A) and B-type (line B) have the Gaussian radii of gyration.
These lines have been obtained from the self-consistent solutions of Eqs. (64)(66) for A and B-type
chains. The mean-eld spinodal curve
1
N
A

f

1
N
B
(1

f)
2x = 0; (68)
is plotted as a dotted line. The minimum value of x on the spinodal curve corresponds to the mean-eld
critical point (shown as an empty triangle). If one takes into account the long-range uctuations, its true,
shifted position can be found [75] (shown as a lled triangle). This location is dened by the self-
consistent system of Eqs. (64) and (65) i.e. by G
2
(0; 0) = 0 and G
3
(0; 0; 0) = 0:
The general features of the behavior of the radius of gyration behavior are as follows. In the limit of an
innite temperature, when the interaction parameter x approaches zero, the chains of both types are
swollen. Then, when we lower the temperature, the size of the chains decreases. The chains less
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1066
abundant in the blend pass through the Gaussian size rst. So we can have the chains of one type swollen
and of the other shrunk. The chains of both types are shrunk when we approach the spinodal. There is one
point on the phase diagram where the chains of both types are Gaussian (similar to the Q point in the
theory of polymer solutions). This point is located at the intersection of the solid lines.
The relative shrinking of the chains changes as 1=

N
_
which follows immediately from the Eq. (66)
since the corrections to the Gaussian chain are proportional to 1=

N
_
: This result has been conrmed by
the SCFT calculations and Monte-Carlo simulations [93]. The change of the end-to-end distance in the
symmetric mixture (

f = 0:5) with the temperature and the index of polymerization resulting from these
calculations are shown in Fig. 2.
The calculations of the radii of gyration based on the self-consistent determination of G
2
(q; q) are
valid in some region around the critical point where the uctuation corrections are important. Far away
from the critical point at high temperatures this is not the case and the RPA prediction for G
2
(q; q)
G
mf
2
(q; q) =
r
0
N
A

f
1
N
A

f
N
B
(1

f)
_ _
2xN
A

f 1
l
2
B

f
l
2
A
(1

f)
_ _
N
A
l
2
A
q
2
6
_ _
; (69)
is used in Eq. (66). From the substitution of this expression in the equation for the radius of gyration (66)
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1067
Fig. 1. Swelling and shrinking of polymer chains in the symmetric mixture (N
A
= 1000; N
B
= 1000): The lines A and lines
B show where the polymer chains of A- and B-type have the Gaussian size R
2
0
(A)
= N
A
l
2
A
=6 or R
2
0
(B)
= N
B
l
2
B
=6: Above these
lines chains are shrunk, and below are swollen. These lines have been computed using the self-consistent one-loop system of
equations. The mean-eld spinodal is plotted as the dotted line. The empty triangle presents the mean-eld critical point, lled
triangle presents the real critical point determined by the self-consistent one-loop system of equations.
follows that the chains at x = 0 are always swollen and have the same radii of gyration independently of
the concentration for N
A
= N
B
: This result is quite reasonable, because the swelling at x = 0 is caused
only by the constraint of incompressibility. If one calculates the radius of gyration directly as an average
of the squared inter-monomer distances with the constraint of incompressibility given by Eq. (12), i.e.
R
2
=

n
A
a=1

n
B
b=1
_
Dr
a
Dr
b
W
A
(r
a
)W
B
(r
b
)d(
^
f
A
(q)
^
f
B
(q))
1
2N
2
A

N
A
i=1

N
A
j=1
(r
(1)
i
r
(1)
j
)
2
1=

n
A
a=1

n
B
b=1
_
Dr
a
Dr
b
W
A
(r
a
)W
B
(r
b
)d(
^
f
A
(q)
^
f
B
(q)); (70)
one exactly obtains Eq. (66) without the term G
A
2
(x
A
)=G
2
(x
A
) in square brackets under integration. In the
limit of innite chain length, N - , the Gaussian size for the polymer chains is obtained.
The changes of the radii of gyration in the symmetric mixture with different concentration of A- and
B-type monomers depend on the FloryHuggins parameter x as shown in Fig. 3. All calculations have
been performed using the self-consistent set of equations (64)(66). The change of the radius of gyration
directly depends on the value of the monomer concentration. The chains less abundant in the blend have
bigger relative shrinking. When the concentration of one component is very small, the shrinking of
chains of this type can be very large at low temperatures. For the systems with concentrations different
from

f = 0:5 one can observe some pretransitional effects. In the regions where the chains exhibit a
sharp change of their size, the scattering intensity I(q) 1=G
2
also increases very sharply. In Fig. 4, we
present such a sharp increase of the scattering intensity. For the same system with fraction of A-
monomers different from

f = 0:5 (here we choose

f = 0:2) we change FloryHuggins parameter x
(from bottom to top on Fig. 4:x = 0.88, 0.89, 0.90, 0.91, 0.92). The jump occurs near some special point,
which we call a local demixing point (in our case for

f = 0:2 and x = 0:91). At this point the correlation
length jumps to a higher value. It is preferable for the chains to be surrounded by the chains of the same
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1068
Fig. 2. The end-to end distance at

f = 0:5 as a function of the temperature and chain length. The lines present the results of the
SCFT calculations for chain length N = 16; 32, 64, 128, 256 and 512 (from bottom to top), while the symbols display the results
of the Monte-Carlo simulations for chain length N = 16: Reprinted with permission from Macromolecules 1998;31:904457.
1998 American Chemical Society [93].
type, so they can demix locally. This effect is weak when the concentrations of A- and B-type monomers
are comparable, and becomes stronger when we increase the difference between the A- and B-type
monomer concentrations.
The relative shrinking of chains with N = 1000 is very small even close to the spinodal line. The more
pronounced shrinking of the chains with minor number fraction has been observed in the computed
simulations based on the soft ellipsoid model for polymer melts and mixtures [99] and performed for
short chains (N = 50): The averaged squared radius of gyration of both the minority A chains and
majority B chains (

f = 0:1) as a function of the excess AB interaction parameter d (proportional to the
FloryHuggins parameter) is shown on Fig. 5. The chains A exhibit signicantly larger shrinking.
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1069
Fig. 3. The radius of gyration in the symmetric mixture (N
A
= 1000; N
B
= 1000) depends on the monomer concentration
(f
a


f); and the FloryHuggins parameter. Please note the regions where the chains exhibit sharp points (although small)
shrinking (see Fig. 4).
Fig. 4. A jump of the scattering intensity I(q) for the symmetric mixture (N
A
= N
B
= 1000) at the xed concentration,

f =
0:2: The solid lines correspond to the different FloryHuggins parameters: from bottom to top: x = 0:88; 0.89, 0.90, 0.91, 0.92.
Please note the sudden increase in I(q) for x 0:90 0:91: This is the indication of the local demixing point characterized by
the sharp shrinking of chains (see also Fig. 3).
The case of asymmetric mixtures (N
A
= 400; N
B
= 2000) is shown in Fig. 6. The general features
are the same as in the case of symmetric mixtures, but the intersection of the Gaussian lines is moved to
the right on the gure. The incompleteness of the solid lines is caused by some technical difculties in
the numerical computations. In the limit of x - 0; the longer chains swell more than the shorter ones.
This resembles the well-known situation when the chains consist of the same type monomers, and the
short chains can be considered as a solvent for the long ones with the partially screened excluded volume
interactions [24,103].
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1070
Fig. 5. The averaged squared radius of gyration of both the minority A chains (f
A
= 0:1) and the majority B chains (f
B
= 0:9)
as a function of the excess AB interaction parameter d. The simulation volume contained 4000 particles representing chains of
N = 50: According to Ref. [99].
Fig. 6. Swelling and shrinking of polymer chains in the asymmetric mixture (N
A
= 400; N
B
= 2000): The lines A and lines
B showwhere the polymer chains of A- and B-type have the Gaussian size R
2
0
(A)
= N
A
l
2
A
=6 or R
2
0
(B)
= N
B
l
2
B
=6 (see the legend of
Fig. 1).
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1071
Fig. 7. Swelling and shrinking of polymer chains in the mixture with equal indices of polymerization (N
A
= 1000; N
B
=
1000) but different segment lengths l
A
and l
B
. The lines A and lines B showwhere the polymer chains of A- and B-type have
Gaussian size R
2
0
(A)
= N
A
l
2
A
=6 or R
2
0
(B)
= N
B
l
2
B
=6: Above these lines the chains are shrunk, and below are swollen. The solid lines
correspond to l
A
= l
B
; dashed to l
A
= 2l
B
; dotted to l
A
= 5l
B
: The empty squares present the intersection of the Gaussian lines
computed from the self-consistent determination of G
2
, The mean-eld spinodal is plotted as the dotted line.
Fig. 8. The magnitude of the chains shrinking as a function of the asymmetry of the indices of polymerization. The lines Ra or
Rb correspond to the relative differences of the radii of gyration of A- or B-type chains in the critical points, R
cr
, in
comparison to its Gaussian radii of gyration, R
0
. We change the number of segments in B-type chains, N
B
, while the number
of segments in A-type chains, N
A
, is constant (N
A
= 200):
The case of the mixture with the same indices of polymerization (N
A
= N
B
= 1000) but different
monomer volumes is shown in Fig. 7. This means, that the bond lengths l
A
and l
B
for each type of chains
are different in Eq. (11). The solid lines correspond to l
A
= l
B
; the long-dashed lines correspond to l
A
=
2l
B
; and the dotted lines correspond to l
A
= 5l
B
: The mean-eld spinodal is shown as a dot-dashed line.
Q-points are plotted as empty squares. The increase of the ratio l
A
/l
B
moves the Gaussian point toward
higher fractions of monomers with the longer segments.
In Fig. 8 it is shown how the shrinking of the chains at the critical point depends on the asymmetry of
the indices of polymerization. The points on the lines Ra or Rb correspond to the relative changes of
the radii of gyration of A- or B-type chains at the critical point, R
cr
, in comparison with the Gaussian radii
of gyration R
0
. In the symmetric case these changes are equal. When we increase the index of poly-
merization in one-type (B-type) chains, shrinking of the shorter (A-type) chains becomes smaller, but
shrinking of the longer (B-type) chains increases. When asymmetry is very big our system behaves as a
system of polymer chains mixed with a solvent. If, additionally, the concentration of the longer chains is
very small, they shrink very rapidly as x increases. This phenomenon is similar to the collapse of a
polymer chain in a strongly incompatible melt [84].
Summarizing, the critical uctuations have a very weak inuence on the radius of gyration in homo-
polymer blends. The Gaussian statistics is a very good approximation for the polymer statistics near the
critical point. We can expect signicant deviations from the Gaussian statistics when the sizes of the
molecules are very disparate and/or the temperature is very low.
4. Copolymer melts near the orderdisorder transition
The simplest copolymer system is a diblock AB copolymer melt. It composes of the diblock
copolymer molecules which contain a sequence of A-type monomers chemically joined to the sequence
of B-type monomers. The state of the system of mixed diblock copolymer molecules with a given
architecture is determined by temperature, only. In contrast to the homopolymer blend, in the copolymer
melt A-homopolymer and B-homopolymer are chemically joined in a block, so the system of diblocks
cannot undergo a macroscopic phase segregation. Instead, a number of orderdisorder phase transitions
take place in the system between the isotropic phase and spatially ordered phases in which A- and B-rich
domains, of the size of a diblock copolymer, are periodically arranged in space. One can expect
the following types of ordering in such a system [70,79,104112]: the lamellar mesophase
(LAM); hexagonally packed cylinders (HEX); spheres arranged at the sites of a bcc lattice
(BCC); bicontinuous double-gyroid structure with Ia

3d space group symmetry (G); hexagonally


modulated lamellae (HML) and hexagonally perforated layers (HPL). The size of the domains in
the ordered structures can vary from tens up to hundreds of angstrom and is controlled by the
size of the diblock macromolecules.
The phase diagram of the AB diblock copolymers is often divided into two regions: the weak
segregation limit (where the boundaries between the A- and B-rich domains are diffuse, i.e.
comparable to the size of the domains) and the strong segregation limit (ordered mesophases
with sharp boundaries between A- and B-rich domains). Deep in the disordered region the
conformations of a single chain are the same as the conformations of an isolated noninteracting
chain (Gaussian coil) and the size of a polymer chain scales with the polymerization index as
R lN
1=2
: The scattering experiments manifest the characteristic correlation length [70]
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1072
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1073
Fig. 9. A schematic illustration of the proposed real-space morphology of a symmetric diblock copolymer melts. GST and ODT
correspond to the Gaussian- to stretched-coil transition and orderdisorder transition, respectively. Taken from Ref. [116].
Fig. 10. The dependence of the SANS peak position q

on the polymerization index N. According to Ref. [116].


(correlation hole) in the disordered melt due to the connectivity of the two blocks and low
compressibility of such melts. In the strong segregation limit the chains are signicantly stretched
and lN
2=3
scaling for the radius of gyration is valid instead of the Gaussian scaling [113115].
Therefore, the crossover between the Gaussian- and stretched-coil conformations is expected to
be somewhere around the orderdisorder transition (ODT) (see illustration in Fig. 9).
The rst direct observation of the chains stretching in the disordered region was performed by Almdal
et al. [116] in SANS experiments. They observed two different scaling behaviors of the peak position
q

2p=R in the scattering intensity with the index of polymerization of the diblock macromolecules at
a xed temperature (see Fig. 10). They found a slope d = 0:49 ^0:02(q

N
d
) in the disordered
region characteristic for the Gaussian conformation of the chains, but near the ODT the scaling behavior
changed to d = 0:8 ^0:04:
It is important to note that concentration uctuations play a decisive role in the nature of the ODT in
the diblock copolymer system. For example, the mean-eld theory [70] predicts a second-order phase
transition from the disordered state to the lamellar structure in the symmetric melt (the lengths of block
A and block B are equal). However, as it was shown [79,80], increasing the volume in the reciprocal
space of important eld uctuations modify the nature of ODT and instead of the second-order transition
the uctuations induce the weak rst-order transition. Also the single-chain conformation in the disor-
dered state are strongly affected by large concentration uctuations as observed in the experiments
[116,117].
In Section 4.1 we derive the set of equations which describe the single chain conformations in the
diblock copolymer melts. In Section 4.2 we discuss how the local structure of the diblock melt depends
on the architecture of the chains and temperature in the disordered region before the phase transition to
the ordered state. In Section 4.2 we also discuss the single-chain properties of different copolymer
systems such as random copolymers [118120], gradient copolymers [97,121] and the melts of the
copolymer rings [122,123].
4.1. Equations for the size of a diblock copolymer
We consider a mixture of n polymer molecules inside a volume V. Each of the molecules consists of
one block of N
A
monomers of type A and the second block of N
B
monomers of type B. The architecture
of the polymer molecules is specied by the composition, f = N
A
=(N
A
N
B
) and distribution function,
W[r]; (r = {r
1
; ; r
N
}) (Eq. (11)) where N = N
A
N
B
is the index of polymerization of the entire
diblock macromolecule. Here we assume the equal segments lengths in both blocks of different species.
The interaction Hamiltonian with the specied short-range interactions between the monomers (18) is
given by Eq. (53), where r
0
= n(N
A
N
B
)=V is the number density of monomers in the system,
^
f
A
(q)
and
^
f
B
(q) are the Fourier transforms of the microscopic concentration operators
^
f
A
(q) =
1
r
0

n
a=1

N
A
i=1
exp(qr
a
i
);
^
f
B
(q) =
1
r
0

n
b=1

N
A
N
B
i=N
A
1
exp(qr
b
i
): (71)
Due to the connectivity of the blocks this part of the Hamiltonian leads to the microphase (mesophase)
separation in the diblock copolymer melt providing w
AA
w
BB
2w
AB
= 2k
B
Tx 0; where x is the
usual FloryHuggins parameter.
The conditional partition function, Z[f
A
; f
B
]; is specied in the similar way as for a homopolymer
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1074
blend (Eq. (56)) with the microscopic density operators (71). Two additional elds (31) have the
following form:
C
A
(r) = f
A
(r) f ; C
B
(r) = f
B
(r) (1 f ); (72)
and describe the excess of the fraction of A or B monomers at the point r over their average values in the
system, f or 1 f (f is the composition). In the incompressible system C
A
= C
B
= C:
It was shown [75,77] that the partition function of the incompressible copolymer melt, Z[c], could be
rewritten as follows:
Z[C] = exp

n=2
1
n!
_
dq
1
(2p)
3

_
dq
n
(2p)
3
G
(0)
n
(q
1
; ; q
n
)d(q
1

q
n
)C(q
1
)

C(q
n
)
_ _
; (73)
where the vertex functions G
(0)
n
are known functions arising from the ideal chain conformations and the
interaction Hamiltonian, H
I
, is included in G
(0)
2
(for more details see Refs. [70,73,75]), C(q) is a Fourier
transform of C(r): The integrals in Eq. (73) diverge if the upper limit of integration q
up
- ; therefore
the nite upper wave-vector cutoff must be introduced (see Section 2.3).
In order to characterize the local structure of the diblock copolymer melt we use the following
physical quantities: the radius of gyration of a single copolymer chain, i.e.
R
2
=
1
2N
2

N
i=1

N
j=1
(r
(1)
i
r
(1)
j
)
2
; (74)
the radii of gyration of blocks A and B:
R
2
AA
=
1
2N
2
A

N
A
i=1

N
A
j=1
(r
(1)
i
r
(1)
j
)
2
; (75)
R
2
BB
=
1
2N
2
B


N
i=N
A
1

N
j=N
A
1
(r
(1)
i
r
(1)
j
)
2
; (76)
and the interblock distance:
R
2
AB
=
1
2N
A
N
B

N
A
i=1

N
j=N
A
1
(r
(1)
i
r
(1)
j
)
2
; (77)
which is the average distance between the monomers of different blocks and is proportional to the
distance between the centers of mass of the blocks. Here and below

means the statistical average.


It is easy to see that
R
2
= f
2
R
2
AA
(1 f )
2
R
2
BB
2f (1 f )R
2
AB
: (78)
We dene single chain correlation functions S
gg
(g = A; B) as
S
(1)
gg
(q; q) =
^
f
(1)
g
(q)
^
f
(1)
g
(q); (79)
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1075
where
^
f
(1)
A
=
1
r
0

N
A
i=1
exp(qr
(1)
i
); (80)
^
f
(1)
B
=
1
r
0

N
i=N
A
1
exp(qr
(1)
i
): (81)
Now the quantities of our interest given by Eqs. (74)(77) can be obtained from the single-chain
correlation functions given by Eqs. (79)(81) by differentiating twice with respect to q and taking
the limit of q - 0.
We calculate S
(1)
AA
(q; q) using the same method as described for the homopolymer blends. And
nally nd the equation for the radius of gyration of a block A in the following form:
_
R
AA
R
(0)
AA
_
2
= 1
9

3
2
_
4f
3
p
2

N
_
V
nNl
3
_
Q
cutoff
dkk
2
__
C(k)C(k)
S
2
(k; k)

1
d(k)
_
[S
BB
(k; k)S
//
AAAA
(k; k; 0; 0) S
AA
(k; k)S
//
BBAA
(k; k; 0; 0)
S
BA
(k; k)S
//
ABAA
(k; k; 0; 0) S
AB
(k; k)S
//
BAAA
(k; k; 0; 0)] [S
//
AAAA
(k; k; 0; 0)
S
//
BBAA
(k; k; 0; 0) S
//
ABAA
(k; k; 0; 0) S
//
BAAA
(k; k; 0; 0)]
C(k)C(k)
d(k)
_
;
(82)
where C(k)C(k) is the scattering intensity of the system; S
2
is the mean-eld scattering intensity at
innitely high temperature [70] ((x = 0):
S
2
(k; k) =
[S
AA
(k; k)S
BB
(k; k) S
AB
(k; k)S
BA
(k; k)]
2
S
AA
(k; k) S
BB
(k; k) S
AB
(k; k) S
BA
(k; k)
; (83)
d(k) for convenience denotes [S
AA
(k; k)S
BB
(k; k) S
AB
(k; k)S
BA
(k; k)] and
S
gg
(k; k) =
^
f
g
(k)
^
f
g
(k)
0
(84)
S
//
ggbb
(k; k; 0; 0) = lim
q-0
2
2
2q
2

^
f
g
(k)
^
f
g
(k)
^
f
b
(q)
^
f
b
(q)
0
_ _
; (85)
k =

N=6
_
(lk); the average

0
is taken with respect to the ideal chain statistics (Eq. (11)). The upper
wave-vector cutoff is given by:
Q
cutoff
=
pC

R
AA
R
(0)
AA
R
BB
R
(0)
BB
R
AB
R
(0)
AB
3
_
1

f (1 f )
6
_ ;
(86)
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1076
where R
(0)
AA
; R
(0)
BB
and R
(0)
AB
are the values of radii of gyration (Eqs. (75)(77)) evaluated with the
Gaussian statistics. We have varied C between C = 0:75 and C = 1:2 and found almost no inuence
on the nal results.
Eq. (57) has been obtained from the innite expansion in the small parameter, 1=

N
_
; of Eq. (42), and
limited to rst-order corrections. Our approximations break down in the case of very asymmetric melts
(f p0:5), in the ordered region far away from the spinodal (the strong segregation limit), and for the
melts which consist of short chains. The Gaussian size, R
0AA
, is the zero-order term, while the rst-order
correction is due to the long-range uctuations (cutoff dependent).
The equations for the radius of gyration of block B, R
BB
, and for the interblock distance, R
AB
have the
similar structures and are given in the Appendix C. The only quantity needed to complete the set of
equations is the expression for the scattering intensity, C(k)C(k): In the framework of the one-loop
self-consistent (Hartree) approximation [75,78] one can obtain the equation for the inverse of the
scattering intensity T
2
(k; k) = 1=C(k)C(k); i.e.
G
2
(q; q) = G
(0)
2
(q; q) DG
(0)
2
(q; q)
1
2
_
dk
(2p)
3
G
(0)
4
(q; q; k; k)
G
2
(k; k)

1
2
_
dk
(2p)
3
[G
(0)
3
(q; k; q k)]
2
G
2
(k; k)G
2
(q k; q k)
; (87)
if the system, is described by the partition function Z(C) (Eq. (73)). In the mean-eld approximation [70]
the expression for the scattering intensity is as follows:
[G
(0)
2
(k; k)
1
] = C(k)C(k)
mf
=
1
1
S
2
(k; k)
2xN
:
(88)
In the simplest approximation this function is used in the equations for the radii of gyrations (Eqs. (82),
(C9) and (C10)). In general, one can evaluate the characteristic sizes of a diblock copolymer molecule by
the substitution of an experimentally obtained scattering intensity into Eqs. (82), (C9) and (C10).
The equations for R
AA
, R
BB
and R
AB
together with the expressions for the scattering intensity (Eq. (87))
and for the cutoff (Eq. (86)) make a complete system of equations from which all these quantities can be
determined in a self-consistent way. To solve the integral equation (Eq. (87)) we use the following
approximation for G
2
(q; q) [70,73]:
G
2
(q; q) = a b(q q

)
2
; (89)
and the three parameters a, b, q

can be determined as a solution of the system of three equations:


a = G
2
(q

; q

); 0 = lim
q-q

2
2q
[G
2
(q; q)]; 2b = lim
q-q

2
2
2q
2
[G
2
(q; q)]; (90)
where G
2
(q; q) in the RHS of Eqs. (90) denotes the RHS of Eq. (87); q

is the peak wave-vector of the


scattering intensity [70].
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1077
4.2. Single-chain conformations in the copolymer melts
First we consider the single-chain conformations of the symmetric diblock copolymer melt (f = 0:5):
In such a system the blocks A and B of a diblock copolymer chain are equivalent and the shape of a
polymer coil can be described by the radius of gyration of the entire polymer chain, R, and the interblock
distance, R
AB
(the distance between their centers of mass). It has been found in the computer simulations
[124,125], LandauGinzburg eld-theory calculations [73,77] and the liquid-state polymer reference
interaction site model theory (PRISM) [126] that the size of a polymer chain, R increases with the
temperature. It has also been found that this increase is dominated by the change of the interblock
distance. The change of the radius of gyration R and of the interblock distance R
AB
with the monomer
monomer energy interaction parameter eN (equivalent to the FloryHuggins interaction parameter),
obtained from the MC simulation [124] is shown in Fig. 11. One can see that the shape on the polymer
coil begins to change in the disordered region. In the transition region (estimated in this paper to be
around 7.59eN) the coils are strongly stretched which is manifested by the pronounced increase of the
interblock distance. The ODT in the symmetric diblock copolymer melt and the stretching of the chains
should lead to the orientational ordering in the system [91,125]. In the disordered state the second
Legendre polynomial (i.e. the orientational order parameter) f
or
,
f
or
=
1
2
(3cos
2
u 1); (91)
uctuates around zero, while at a certain temperature this orientational parameter reaches a nite value
(Fig. 12). The angle u in Eq. (91) is the angle between an arbitrary chosen directions and the vector
connecting the center of mass of the two blocks. It is interesting to note at this point that the instanta-
neous conguration of the diblock copolymers even in a homogeneous melt without any interactions
resembles a cigar-like shape characteristic for linear polymer chains [127]. Of course, the single chain
statistics (averaged over all congurations) remains Gaussian.
The shape of the diblock copolymer molecule in the slightly asymmetric melt (N = 1000 and
composition f = 0:45) has been studied within the one-loop approximation by solving the system of
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1078
Fig. 11. The change of the radius of gyration, R and the distance between the blocks R
AB
with the monomermonomer
interaction parameter e and index of polymerization N in the symmetric diblock copolymer melt resulting from the MC
simulation. (N = 24): According to Ref. [124].
equations (82), (C9), (C10), (86) and (87) in order to nd out the characteristic quantities R, R
AA
, R
BB
,
R
AB
and q

. The results are shown in Fig. 13. The spinodal value of the interaction parameter is shifted
towards the higher values due to the inuence of the long-range uctuations [75] in comparison to the
mean-eld transition value (x
mf
= 10:69): At high temperatures (small x) the chains are almost Gaus-
sian. As x increases, the local structure changes. The deviations from the Gaussian size are considerable
already in the region which is far from the spinodal. The sizes of the blocks, R
AA
and R
BB
, depend on x in
a similar way but their magnitudes are different.
The one-loop calculation has been performed also for the asymmetric melts with different values of
the composition. In Fig. 14 R
AA
(Fig. 14a), R
BB
(Fig. 14b) and the interblock distance, R
AB
(Fig. 14c), are
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1079
Fig. 12. The orientational parameter f
or
(Eq. (91)) reaches a nite value at the ordering temperature, when the system is cooled
down. These results have been obtained from the MC computer simulations with zero vacancy concentration. The transition
region has been identied to be k
B
T=1N 0:45 0:5: The lines are only guides to the eye. According to Ref. [125].
Fig. 13. The radii of gyration of the entire diblock copolymer molecule, R, its blocks, R
AA
and R
BB
, and the interblock distance,
R
AB
as functions of the interaction parameter, x, in the nearly symmetric melt. These lines represent the results of the one-loop
calculations.
shown as functions of the interaction parameter x. When the melt is nearly symmetric, 0:35 f 0:5,
the blocks of both types shrink a little initially and then, at some value of x, they begin to swell. The
same qualitative behavior in the case of symmetric and asymmetric melts has been detected. In the more
asymmetric melts, f 0:35, the block of a macromolecule which consists of the monomers of the
minority type (block A) shrinks monotonically, while the other block monotonically swells. From the
results one can conclude that the interblock distance (Fig. 14c) in more asymmetric melts changes more
sharply than in the symmetric case. Qualitatively similar results for a chain behavior in the asymmetric
melts have been obtained in the PRISM theory [126] and MC simulations [128].
Summarizing, one can say that two different qualitative behaviors of the local structure of the
copolymer melts in the disordered state take place depending on the composition. In the rst case,
the blocks of both type behave similarly, which imply that the ordered phase which occur after the ODT
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1080
Fig. 14. The radius of gyration of the (a) block A, R
AA
, (b)block B, R
BB
and (c) the interblock distance, R
AB
, for the specied
composition, f, as functions of the interaction parameter, x from the one-loop calculations. The sizes of the block A (minority
type monomers) of the diblock macromolecules with the composition in the range of 0:2 f 0:35 and in the range of 0:4
f 0:5 behave differently which reects the different ordered morphologies which occur in the system after the ODT.
will have the symmetry irrelevant to the interchange of the blocks [77]. This suggests that in the nearly
symmetric melt the ordered phase will have the lamellar or bicontinuous (e.g. gyroid) morphology. In
the second case the sizes of the blocks change in a completely different way. The minority block shrinks
and the majority block swells. This suggests that in the more asymmetric melts, the ordered phase with
the curved interface should occur: e.g.. the hexagonally packed cylinders. In this case the cylinders
consist of the monomers of the minority type (shrunk blocks) surrounded by the monomers of the
majority type (swollen blocks). The inverted structure (cylinders which consist of the majority
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1081
Fig. 15. The change of the radius of gyration of the entire diblock macromolecule, R, with the specied composition, f, in
comparison with its size at innitely high temperature, R
(0)
x=0
; as a function of the interaction parameter, x. These functions have
been obtained from the one-loop calculations.
Fig. 16. The peak wave-vector, q

of the scattering intensity as a function of the interaction parameter, x, for the melt with the
specied composition, f (one-loop calculations).
monomers (swollen blocks) surrounded by minority monomers (shrunk blocks)) is energetically and
entropically unfavorable and does not occur in real systems. These assumptions about the ordered
morphologies which take place after the ODT are in accordance with the recent experiments [109]
where the lamellar (gyroid) morphology was established in the same range of composition
0:38 f 0:65, and hexagonal and bcc structures for f 0:38 and f 0:65. Thus the local structure
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1082
Fig. 17. The change of the radius of gyration of the entire diblock macromolecule, R, with the specied degree of polymeriza-
tion, N, in comparison with its Gaussian value as a function of the interaction parameter, x. The empty diamonds show the
locations of the ODT. These results have been obtained from the one-loop calculations.
Fig. 18. The peak wave-vector position in the scattering intensity in the melt with the specied degree of polymerization, N, as a
function of the interaction parameter, x. The dashed line shows the position of the peak wave-vector in the mean-eld scattering
intensity, and the empty diamonds represent the locations of the ODT.
of the melt in the disordered phase gives an indication about the type of the ordered structure which
emerges after the ODT.
The change of the size of the entire polymer molecule is mostly determined by the change of the
interblock distance, R
AB
, and the polymer coil is stretched near the spinodal (where the calculated
scattering intensity diverges). The size of the macromolecules, R, as a function of x for is shown on
Fig. 15 for different compositions. The curves shown on this gure increase sharply at different values of
x because for each value of the composition the location of the ODT (or spinodal) occurs at different
value of x.
Another physical quantity which describes the local structure is the value of the peak wave-vector, q

,
in the scattering intensity. This quantity tells us about the characteristic average size of the domains in
the system. In the mean-eld theory the peak wave-vector has the specic value for a given composition,
which does not depend on x. But recent experiments [116,129,130] showed that q

decreases as the
system approaches the spinodal structure, and the magnitude of this decrease can be of the order of 25%
of its initial value. In Fig. 16 the dependence of the peak wave-vector, q

, on the interaction parameter x


is shown for different values of the composition. In the nearly symmetric melt the peak wave-vector
decreases signicantly close to the spinodal, while in the more asymmetric melt such a dependence is
rather weak. As was just shown, the size of the polymer molecules increase as x approaches the spinodal,
so the characteristic size also increases and consequently q

decreases.
We change the number of polymerization, N, of the diblock macromolecule with the xed composi-
tion, f = 0:45; in order to nd out how the local structure of the melt depends on the length of the
macromolecules. In Fig. 17 the dependences of the sizes of the macromolecules on x are given for the
different numbers of monomers, N. The empty diamonds show the location of the ODT. The melt with
the smaller polymerization index has more stretched chains, because the corrections from the uctua-
tions which are responsible for the stretching are proportional to 1=

N
_
: At the same time, the shift of the
spinodal is also proportional to 1=

N
_
; so longer chains start to stretch signicantly at smaller values of x.
At the ODT the chains are already stretched, but the magnitude of this stretching is small. The depen-
dence of the peak wave-vector in the scattering intensity on the degree of polymerization of the melt and
on the interaction parameter x is given in Fig. 18. The dashed line shows the mean-eld position of the
peak wave-vector. At innitely high temperature (x = 0) the peak wave vector has slightly bigger values
then in the mean-eld theory. This is a consequence of the initial size of the chains at x = 0 (see Fig. 14).
The chain conformations in the complex copolymer systems are directly related to the material
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1083
Fig. 19. Bridge and loop congurations of blocks from a triblock copolymer in the lamellar phase. Reprinted with permission
from Ref. [131].
properties of the melt and its phase behavior. For example, mechanical properties of the lamellar ordered
multiblock copolymer melt depend on the number of loops and bridges which are formed by the chains
[131] (Fig. 19). The topological constraints could also affect the properties of the copolymer system. The
melt of the block copolymer rings [122,123] exhibit the similar properties as a block copolymer melt
[122]. However, the single-chain conformations are quite different. The relative increase of the
interblock distance R
AB
at low temperatures is more pronounced [123] for the ring diblock copolymers
(about 2025% for N = 16; ; 48) than for the linear diblock copolymers (15%). At the same time, the
blocks A and B shrink in the ring diblock copolymers (up to 5%) while in the linear copolymer blend
they exhibit only small changes of size.
Another type of copolymer systems are random copolymers [118120]. They are produced by a
polymerization reaction in a medium which contains two different types of monomers [3]. The
architecture of these chains is conditioned by the reaction rates and can be conveniently
described by a probability matrix. For example, the conditional probability that an A-type mono-
mer is followed by a B-type monomer in the chain is given by the element p
BA
of this matrix. If
the polymerization process takes place under the steady-state conditions the elements of the
matrix p
XY
do not depend on the position of the monomers in the chain. A quenched disordered
architecture of such a chain is determined by the average composition of, say, A-type monomers
and a certain combination of two elements of the matrix p
XY
. The multiblock copolymers are also
obtained in such processes. One can also consider a copolymer melt which has been produced by
the polymerization reaction under the non-steady-state conditions and assume the ideal situation
in which reaction rates can be precisely controlled by the change of external parameters. If,
additionally, the characteristic time at which a new monomer is joined to the chain during the
polymerization process is comparable with the time needed by the reaction rates to adapt to the
change of the external conditions, the chains would have the same distribution of the monomers
averaged over a few monomers length scale. The architecture of these chains can be described by
a distribution function [97,121], and such a melt of the linear AB copolymer chains with a
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1084
Fig. 20. The composition distribution of comonomers along polymer chains in the model system generated for the simulation
[97]. The line D corresponds to the diblock copolymer system; R to the random copolymers; G1, G2, and G3 represent the
gradient copolymer chains with the distribution function (x=N)
1
; (x=N)
2
and (x=N)
3
respectively. x counts the mono-
mers along the chain contour and changes in this simulations between 1 and N = 40:
specied distribution of the monomers along the chains is called the gradient copolymer melt. It
has been shown theoretically that the complex ordered structures analogous to the ordered
structures exhibited by the diblock copolymer melts also exist in such systems [121]. The
single-chain properties have been investigated in the MC computer simulations [97]. In Fig.
20 the different distribution functions which have been used in the simulations are shown.
Line D corresponds to the diblock copolymer system; R to the random copolymers; G1, G2,
and G3 represent the gradient copolymer chains with the distribution function (x=N)
1
; (x=N)
2
and (x=N)
3
, respectively. The radius of gyration for each type of the melt is shown in Fig. 21
as a function of temperature. The vertical dotted lines indicate the location of the ODT. Diblock
and gradient copolymer chains swells as the system approaches the ODT. The ordered structures
observed in the simulations in all these systems have the lamellar morphology. On the contrary,
the random copolymer chains (R) do not change their sizes with the temperature and the ODT
does not occur in this system.
Summarizing, the uctuations near the ODT not only change the order of the phase transition but also
strongly affect the single-chain properties in the diblock copolymer melts.
Appendix A. Partition function of homopolymer blends with a labeled chain
The free-energy F
(n
g
)
g
for the system of n
g
noninteracting Gaussian chains in the external elds J
g
and
U
A
(U
A
couples to only one chain of A monomers), given by
F
(n
A
)
A
[J
A
; U
A
] =

n=1
(i)
n
n!
_
dq
1
(2p)
3

_
dq
n
(2p)
3
S
A
n
(q
1
; ; q
n
)d(q
1

q
n
)J
A
(q
1
)J
A
(q
n
)

n=1
1
n!
_
dq
1
(2p)
3

_
dq
n
(2p)
3

^
f
(1)
A
(q
1
)
^
f
(1)
A
(q
n
)
c
0
U
A
(q
1
)U
A
(q
n
); (A1)
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1085
Fig. 21. The radii of gyration as a function of temperature for various copolymer melts (see Fig. 20). The vertical dotted lines
indicate the ODT temperatures. According to Ref. [97].
S
A
n
(q
1
; ; q
n
)d(q
1

q
n
) =
^
f
(n
A
)
A
(q
1
)
^
f
(n
A
)
A
(q
n
)
c
0

m=1
1
m!
_
dk
1
(2p)
3

_
dk
m
(2p)
3

^
f
(n
A
)
A
(q
1
)
^
f
(n
A
)
A
(q
n
)
^
f
(1)
A
(k
1
)
^
f
(1)
A
(k
m
)
c
0
U
A
(k
1
)U
A
(k
m
);
(A2)
F
(n
B
)
B
[J
B
] =

n=1
(i)
n
n!
_
dq
1
(2p)
3

_
dq
n
(2p)
3

^
f
(n
B
)
B
(q
1
)
^
f
(n
B
)
B
(q
n
)
c
0
J
B
(q
1
)J
B
(q
n
): (A3)
The average
c
0
(superscript c stands for the cumulant) is taken with respect to the chain statistics
which has been taken to be the Gaussian statistics.
The formulas for G
A
[C
A
; U
A
] and G
B
[C
B
] are:
G
A
[C
A
; U
A
] =

n=2
(i)
n
n!
_
dq
1
(2p)
3

_
dq
n
(2p)
3
G
A
n
(q
1
; ; q
n
; U
A
) C
A
(q
1
; U
A
)C
A
(q
n
; U
A
); (A4)
G
B
[C
B
] =

n=2
(i)
n
n!
_
dq
1
(2p)
3

_
dq
n
(2p)
3
G
B
n
(q
1
; ; q
n
) C
B
(q
1
)C
B
(q
n
); (A5)
and
C
A
(q; U
A
) = fA(q)
^
fA
(n
A
)
(q)
0

m=1
1
m!
_
dk
1
(2p)
3

_
dk
m
(2p)
3

^
f
(n
A
)
A
(q
1
)
^
f
(1)
A
(k
1
)
^
f
(1)
A
(k
m
)
c
0
U
A
(k
1
)U
A
(k
m
): (A6)
The expressions for G
A
[C
A
; U
A
] and G
B
[C
B
] in terms of S
A
n
(q
1
; ; q
n
) and S
B
n
(q
1
; ; q
n
) can be
obtained from the general properties of the Legendre transform [75,78].
Appendix B. Ideal average of microscopic density operators
In order to solve the system of equations (64), (65) and (66) for A- and B-type chains we need to know
the analytical expression for the moments
^
f
(g)
(q
1
);
^
f
(g)
(q
n
)
0
up to the fth order. Let us calculate

^
f
(g)
(k
1
);
^
f
(g)
(k
2
);
^
f
(g)
(k
3
);
^
f
(g)
(k
4
);
^
f
(g)
(k
5
)
0
:

^
f
(g)
(k
1
);
^
f
(g)
(k
2
);
^
f
(g)
(k
3
);
^
f
(g)
(k
4
);
^
f
(g)
(k
5
)
0
=

n
A
a=1

n
B
b=1
_
Dr
a
Dr
b
W
A
(r
a
)W
B
(r
b
)
1
r
5
0

n
A
a
1
;a
2
;a
3
;a
4
;a
5
=1

N
A
i;j;l;m;n=1
exp[k
1
r
a
1
i
k
2
r
a
2
j
k
3
r
a
3
l
k
4
r
a
4
l
k
5
r
a
5
n
];
(B1)
We are interested only in the part of Eq. (B1) that is proportional to d(q
1

q
5
):. Thus, Eq. (B1)
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1086
reduces to
(B1) =
n
g
r
5
0
V
d(q
1
q
2
q
3
q
4
q
5
)

N
A
i;j;l;m;n=1

N
A
s=1
du
g
s
d(u
g
s
l
g
)
4pl
2
g
exp[u
g
s
(k
1
u(i s) k
2
u(j s)
k
3
u(l s) k
4
u(m s) k
5
u(n s))]; (B2)
where u(i s) is the Heaveside step function. After integrating out u
g
s
in Eq. (B2) one obtains:

N
A
i;j;l;m;n=1

N
A
s=1
sin[l
A
k
1
u(i s) k
2
u(j s) k
3
u(l s) k
4
u(ms) k
5
u(n s)]
l
A
k
1
u(i s) k
2
u(j s) k
3
u(l s) k
4
u(m s) k
5
u(n s)

n
g
r
5
0
V
d(q
1
q
2
q
3
q
4
q
5
): (B3)
Now, if we replace the summation in Eq. (B3) by the integration and use an approximate relation
(sin x=x) exp(x
2
=3!) for x - 0 (in our case in such a relation we neglect some terms of the order
of (1=N
n
); n 1), we obtain:
(B3)
n
g
N
5
g
r
5
0
V
_1
0
di
_1
0
dj
_1
0
dl
_1
0
dm
_1
0
dn exp
_

N
g
l
2
g
6
_1
0
ds[k
1
u(i s) k
2
u(j s) k
3
u(l s)
k
4
u(m s) k
5
u(n s)]
2
_
: (B4)
This integral is divided into 5! = 120 parts, each of which has the same structure, i.e.
_1
0
di
_i
0
dj
_j
0
dl
_l
0
dm
_m
0
dn exp[(i j)p
1
(j l)p
2
(l m)p
3
(mn)p
4
]; (B5)
and can be easily calculated. The sequence (p
1
, p
2
, p
3
, p
4
) is generated by the argument of the exponent
function in Eq. (B4) for the given sequence (i j l m n): Thus, the algorithm is as follows:
generate 120 permutations in the sequence (i j l m n); calculate 120 sequences of (p
1
, p
2
, p
3
,
p
4
) and substitute them into Eq. (B5).
^
f
g
(k);
^
f
g
(k);
^
f
g
(p);
^
f
(g)
(q);
^
f
(g)
(p q) can be calcu-
lated using the following function in the argument of the exponent function in Eq. (B4) that generates the
sequences of (p
1
, p
2
, p
3
, p
4
):

N
g
l
2
g
6
_1
0
ds[k(u(i s) u(j s)) p(u(l s) u(n s)) q(u(m s) u(n s))]
2
_ _
: (B6)
To calculate the fourth-order momentum we must eliminate integration over one variable from Eqs.
(B5) and (B4). In such a way
^
f
g
(q);
^
f
g
(q);
^
f
g
(k);
^
f
g
(k) is obtained if we eliminate l from (B5)
and put p = 0 into (B6). Analogically,
^
f
g
(q);
^
f
g
(k);
^
f
g
(p);
^
f
g
(q k p) is obtained if we
eliminate j from Eq. (B5) and put j = n into Eq. (B6), and
^
f
g
(q);
^
f
g
(k);
^
f
g
(q p);
^
f
g
(k p)
is obtained if we eliminate l from Eq. (B5) and j = m and then l = j in Eq. (B6), etc.
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1087
Function g
D
(x) is the Debye function, which originates from the second momentum, i.e.
g
D
(x) =
2[x
2
exp(x
2
) 1]
x
4
; (B7)
and f(x) is given by
f (x) =
2
3
[120 120 exp(x
2
) 96x
2
24 exp(x
2
)x
2
24x
4
12 exp(x
2
)x
4
4 exp(x
2
)x
6
x
8
]=[exp(x
2
)x
10
]: (B8)
Appendix C. Partition function of the diblock copolymer melt with a labeled chain
The partition function (Eq. (43)) can be conveniently expressed in the following form [70,73,77]:
Z[U
A
] =
_
Df
A
_
Df
B
exp
H
I
[f
A
; f
B
]
k
B
T
_ _
_
DJ
A
_
DJ
B
exp i
_
dq
(2p)
3
f
A
J
A
i
_
dq
(2p)
3
f
B
J
B
F
(n)
[J
A
; J
B
; U
A
]
_ _
(C1)
where F
(n)
[J
A
; J
B
; U
A
] is the free-energy for the system of n non-interacting Gaussian diblock chains
in the external elds J
A
; J
B
; U
A
(U
A
couples to the block A of one chain only), and is given in the form
of the cumulant expansion [71,75], i.e.
F
(n)
[J
A
; J
B
; U
A
] =

m=1
(i)
m
m!
_
dq
1
(2p)
3

_
dq
m
(2p)
3

g
1
=A;B


g
m
=A;B
S
g
1
g
m
(q
1
; ; q
m
; U
A
)d(q
1

q
m
)J
g1
(q
1
)J
gm
(q
m
)

m=1
1
m!
_
dq
1
(2p)
3

_
dq
m
(2p)
3

^
f
(1)
A
(q
1
)
^
f
(1)
A
(q
m
)
c
0
U
A
(q
1
)U
A
(q
m
); (C2)
where
S
g
1
g
m
(q
1
; ; q
m
; U
A
)d(q
1

q
m
) =
^
f
(n)
g
1
(q
1
)
^
f
(n)
g
m
(q
m
)
c
0

1=1
1
1!
_
dk
1
(2p)
3

_
dk
1
(2p)
3

^
f
(n)
g
1
(q
1
)
^
f
(n)
g
m
(q
m
)
^
f
(1)
A
(k
1
)
^
f
(1)
A
(k
1
)
c
0
U
A
(k
1
)U
A
(k
1
);
(C3)
where the average
c
0
(superscript c stands for the cumulant) is taken with respect to the ideal chain
statistics.
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1088
We integrate out the elds J
A
, J
B
in the saddle point [71,75],
dF
(n)
[J
A
; J
B
; U
A
]
dJ
g
(q)

J
g
=J

g
= if
g
(q); (C4)
and take the Legendre transform to the new elds C
A
, C
B
(Eq. (72)). The partition function has now the
following form:
Z[U
A
] =
_
Df
A
_
Df
B
exp
H
I
[f
A
; f
B
]
k
B
T
G[C
A
; C
B
; U
A
])
_ _

d
2
F
(n)
[J
A
; J
B
; U
A
]
dJ

A
(q
1
)dJ

A
(q
2
)

d
2
F
(n)
[J
A
; J
B
; U
A
]
dJ

B
(q
1
)dJ

B
(q
2
)

d
2
F
(n)
[J
A
; J
B
; U
A
]
dJ

B
(q
1
)dJ

A
(q
2
)
d
2
F
(n)
[J
A
; J
B
; U
A
]
dJ

A
(q
1
)dJ

B
(q
2
)
_ _ _ _
1=2
;
(C5)
where
G[C
A
; C
B
; U
A
] =

m=1
(i)
m
m!
_
dq
1
(2p)
3

_
dq
m
(2p)
3

g
1
=A;B


g
m
=A;B
G
g
1
g
m
(q
1
; ; q
m
; U
A
)d(q
1

q
m
)C
g
1
(q
1
; U
A
)C
gm
(q
m
; U
A
)

m=1
1
m!
_
dq
1
(2p)
3

_
dq
m
(2p)
3

^
f
(1)
A
(q
1
)
^
f
(1)
A
(q
m
)
c
0
U
A
(q
1
)U
A
(q
m
); (C6)
and
C
g
(q; U
A
) = C
g
(q)

1=1
1
1!
_
dk
1
(2p)
3

_
dk
1
(2p)
3

^
f
(n)
g
(q)
^
f
(1)
A
(k
1
)
^
f
(1)
A
(k
1
)
c
0
U
A
(k
1
)U
A
(k
1
):
(C7)
To express G
g
1
g
m
(q
1
; ; q
m
; U
A
) in terms of S
g
1
g
m
(q
1
; ; q
m
; U
A
) one can use the general
rules of such transformations [78]. The expressions we found coincide with the ones found
previously [70]. Correction DG
(0)
2
(q; q) in Eq. (87), which arises from the ideal chain conformations
[75] and has not been included in the RPA calculation, in the case of the diblock copolymer melt has the
following form:
DG
(0)
2
(q; q) =
1
2
_
dk
(2p)
3

a;b;v;1=A;B
{S
abv1
(k; k; q; q)G
ab
(q; q) 2S
ab
(k; k)}[G
1A
(k; k)
G
1B
(k; k)][G
vA
(k; k) G
vB
(k; k)]: (C8)
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1089
The equation for the size of the block B, R
BB
, is:
_
R
BB
R
(0)
BB
_
2
= 1
9

3
2
_
4(1 f )
3
p
3

N
_
V
nNl
3
_
Q
cutoff
dkk
2
__
C(k)C(k)
S
2
(k; k)

1
d(k)
_
[S
BB
(k; k)S
//
AABB
(k; k; 0; 0) S
AA
(k; k)S
//
BBBB
(k; k; 0; 0)
S
BA
(k; k)S
//
ABBB
(k; k; 0; 0) S
AB
(k; k)S
//
BABB
(k; k; 0; 0)] [S
//
AABB
(k; k; 0; 0)
S
//
BBBB
(k; k; 0; 0) S
//
ABBB
(k; k; 0; 0)S
//
BABB
(k; k; 0; 0)]
C(k)C(k)
d(k)
_
; (C9)
and for the interblock distance, R
AB
:
_
R
AB
R
(0)
AB
_
2
= 1
3

3
2
_
4f (1 f )p
3

N
_
V
nNl
3
_
Q
cutoff
dkk
2
__
C(k)C(k)
S
2
(k; k)

1
d(k)
_
[S
BB
(k; k)S
//
AAAB
(k; k; 0; 0) S
//
AABA
(k; k; 0; 0)) S
AA
(k; k)(S
//
BBAB
(k; k; 0; 0)
S
//
BBBA
(k; k; 0; 0)) S
BA
(k; k)(S
//
ABAB
(k; k; 0; 0) S
//
ABBA
(k; k; 0; 0)) S
AB
(k; k)
(S
//
BAAB
(k; k; 0; 0) S
//
BABA
(k; k; 0; 0))] [S
//
AAAB
(k; k; 0; 0) S
//
BBAB
(k; k; 0; 0)
S
//
ABAB
(k; k; 0; 0) S
//
BAAB
(k; k; 0; 0) S
//
AABA
(k; k; 0; 0) S
//
BBBA
(k; k; 0; 0)
S
//
ABBA
(k; k; 0; 0) S
//
BABA
(k; k; 0; 0)]
C(k)C(k)
d(k)
_
: (C10)
References
[1] Oldy R. The path to the double helix. The discovery of DNA. New York: Dover Publications Inc, 1994. p. 40.
[2] Eisenberg H. Biophys Chem 1996;59:24757.
[3] Flory P. Principles of polymer chemistry. Ithaca NY: Cornell University Press, 1953.
[4] Edwards SF. Proc Phys Soc 1966;18:26580.
[5] Freed KF. Renormalization group theory. New York: Wiley, 1987.
[6] Duplantier B. In: van Beijeren H, editor. Fundamental problems in statistical mechanics VII. Amsterdam: Elsevier
Science, 1990. p. 171223.
[7] Fixman M. J Chem Phys 1955;23:16569.
[8] Ladd AJC, Frenkel D. Macromolecules 1992;25:34358.
[9] Doi M, Edwards SF. The theory of polymer dynamics. Oxford: Clarendon Press, 1986. p. 157.
[10] Zhou Z, Yan D. Macromol Theory Simul 1997;6:597611.
[11] Kardar M, Nelson DR. Phys Rev A 1988;38:96682.
[12] Hwa T. Phys Rev A 1990;41:17516.
[13] Duplantier B, Hwa T, Kardar M. Phys Rev Lett 1990;64:20225.
[14] David F, Duplantier B, Guitter E. Phys Rev Lett 1994;72:3115.
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1090
[15] Gompper G, Kroll DM. J Phys Condens Matter 1997;9:8795834.
[16] Gefen Y, Aharony A, Alexander S. Phys Rev Lett 1983;50:7780.
[17] Cates ME. Phys Rev Lett 1984;53:9269.
[18] Chen Y, Guyer RA. J Phys A Math Gen 1988;21:417381.
[19] Burdzy K, Lawler GF. J Phys A Math Gen 1990;23:L23L28.
[20] Edwards SF. J Phys A Math Gen 1975;8:167080.
[21] Oono Y. J Phys Soc Jpn 1976;41:22836.
[22] Kholodenko AL, Vilgis TA. Phys Rep 1998;298:251372.
[23] Bates AD, Maxwell A. DNA topology. Oxford: Oxford University Press, 1993.
[24] de Gennes PG. Scaling concept in polymer physics. Ithaca and London: Cornell University Press, 1979. p. 299313.
[25] des Cloizeaux J, Jannink G. Polymers in solution. Their modeling and structure. Oxford: Clarendon Press, 1990. p. 649
710.
[26] Doi M. Introduction to polymer physics. Oxford: Clarendon Press, 1997. p. 146.
[27] Grosberg AYu, Khokhlov AR. Statistical physics of macromolecules. New York: AIP Press, 1994. p. 13040.
[28] Cieplak M, Vishveshwara S, Banavar JR. Phys Rev Lett 1996;77:36814.
[29] Cieplak M, Banavar JR. Folding and Design 1997;2:23545.
[30] Chan HS. Nature 1998;392:7613.
[31] Skolnick J, Kolinski A. Annu Rev Phys Chem 1989;40:20735.
[32] Skolnick J, Kolinski A. Science 1990;250:11215.
[33] Ortiz AR, Kolinski A, Skolnick J. Proc Natl Acad Sci USA 1998;95:10205.
[34] Alexander S. J Physique 1977;38:9837.
[35] de Gennes PG. Macromolecules 1980;13:106975.
[36] Milner ST, Witten TA, Cates ME. Macromolecules 1988;21:26109.
[37] Milner ST, Witten TA, Cates ME. Europhys Lett 1988;5:4138.
[38] Milner ST, Witten TA, Cates ME. Macromolecules 1989;22:85361.
[39] Ball RC, Marko JF, Witten ST, Witten TA. Macromolecules 1991;24:693703.
[40] Milner ST. Science 1991;251:90514.
[41] Auroy P, Auvray L, Leger L. Phys Rev Lett 1991;66:71921.
[42] Marko JT, Witten TA. Phys Rev Lett 1991;66:15414.
[43] Barrat J-L. Macromolecules 1992;25:8324.
[44] Milner ST. Macromolecules 1991;24:37045.
[45] Martin JI, Wang Z-G, Schick M. Langmuir 1996;12:49509.
[46] Pakula T, Zhulina EB. J Chem Phys 1991;95:46917.
[47] Lai P-Y, Binder K. J Chem Phys 1991;95:928899.
[48] Lai P-Y, Binder K. J Chem Phys 1992;97:58695.
[49] Lai P-Y, Binder K. Makromol Chem Macromol Symp 1993;65:18998.
[50] Barrat J-L, Joanny J-F. Adv Chem Phys 1996;XCIV:166.
[51] Kratky O, Porod G. Recl Trav Chim Pays-Bas 1949;68:1106445.
[52] Saito N, Takahashi K, Yunoki Y. J Phys Soc Jpn 1967;22:21926.
[53] Warner M, Gunn JMF, Baumgartner AB. J Phys A Math Gen 1985;18:300726.
[54] Wang XJ, Warner M. J Phys A Math Gen 1986;19:22152227.
[55] de Gennes PG. In: Cifferri A, Krigbaum WR, Meyer RB, editors. Polymer liquid crystals. New York: Academic, 1982
[Chapter 5].
[56] Pincus P, de Gennes PG. J Polym Sci Polym Symp 1978;65:8590.
[57] Matsuyama A, Sumikawa Y, Kato T. J Chem Phys 1997;107:47118.
[58] Cotton JP, Hardouin F. Prog Polym Sci 1997;22:795828.
[59] Odijk T. J Chem Phys 1996;105:127086.
[60] Livolant F, Leforestier A. Prog Polym Sci 1996;21:111564.
[61] Bensimon D, Simon AJ, Croquette V, Bensimon A. Phys Rev Lett 1995;74:47547.
[62] Strick TR, Allemand J-F, Bensimon D, Bensimon A, Croquette V. Science 1996;271:18357.
[63] Freed KF. Renormalization group theory of macromolecules. New York: Wiley, 1987. p. 22.
[64] Schweizer KS, Curro JG. Phys Rev Lett 1988;60:80912.
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1091
[65] Hansen J-P, MacDonald IR. Theory of simple liquids, 2nd edn. New York: Academic Press, 1986.
[66] de Gennes PG, Prost J. The physics of liquid crystals. Oxford: Clarendon Press, 1993. p. 57.
[67] Hoyst R, Schick M. J Chem Phys 1992;96:7309.
[68] Hoyst R, Schick M. J Chem Phys 1992;96:7219.
[69] Landau LD, Lifshitz EM. Statistical physics, Part 1, 3rd edn. Oxford: Pergamon Press, 1980. p. 479.
[70] Leibler L. Macromolecules 1980;13:160217.
[71] Ohta T, Kawasaki K. Macromolecules 1986;19:262131.
[72] Hoyst R, Vilgis TA. Macromol Theory Simul 1996;5:573643.
[73] Barrat JL, Fredrickson GH. J Chem Phys 1991;95:12819.
[74] Hoyst R, Vilgis TA. Phys Rev E 1994;50:208792.
[75] Hoyst R, Vilgis TA. J Chem Phys 1993;99:483544.
[76] Aksimentiev A, Hoyst R. Macromol Theory Simul 1998;7:44756.
[77] Aksimentiev A, Hoyst R. Macromol Theory Simul 1999;8:32842.
[78] Amit DJ. Field theory, the renormalization group, and critical phenomena. Singapore: World Scientic, 1984. p. 1039.
[79] Fredrickson GH, Helfand E. J Chem Phys 1987;87:697.
[80] Brazovskii SA. Sov Phys JETP 1975;41:81.
[81] Fredrickson GH, Binder K. J Chem Phys 1989;91:7265.
[82] Schwahn D, Mortensen K, Yee-Madeira H. Phys Rev Lett 1987;58:15446.
[83] Bates FS, Rosedale JH, Stepanek P, Lodge TP, Wiltzius P, Fredrickson GH, Hjelm Jr. RP. Phys Rev Lett 1990;65:
18936.
[84] Joanny JF, Brochard F. J Physique 1981;42:114550.
[85] Landry MR. Macromolecules 1997;30:750010.
[86] Cates ME, Deutsch JM. J Physique 1986;47:21218.
[87] Brereton MG, Vilgis TA. J Phys A Math Gen 1995;28:114967.
[88] Bates FS, Wignall GD, Koehler WC. Phys Rev Lett 1985;55:24258.
[89] Briber R, Bauer BJ, Hammouda B. J Chem Phys 1994;101:25929.
[90] Theobald W, Sans-Penninckx A, Meier D, Vilgis TA. Phys Rev E 1997;55:572330.
[91] Vilgis TA, Weyersberg A, Brereton MG. Phys Rev E 1994;49:30317.
[92] Garaz GE, Kosmas MK. J Chem Phys 1995;103:107909.
[93] Muller M. Macromolecules 1998;31:904457.
[94] Sariban A, Binder K. J Chem Phys 1987;86:585973.
[95] Reiter J, Eding T, Pakula T. J Chem Phys 1990;93:83744.
[96] Binder K. Makromol Chem Macromol Symp 1993;69:21327.
[97] Pakula T, Matyjaszewski K. Macromol Theory Simul 1996;5:9871006.
[98] Muller M, Binder K. J Phys II (France) 1996;6:18794.
[99] Murat M, Kremer K. J Chem Phys 1998;108:43408.
[100] Pakula T. Recent Res Devel Polymer Sci 1996;1:10118.
[101] Schweizer KS, Curro JG. J Chem Phys 1989;91:505981.
[102] Weinhold JD, Kumar SK, Szleifer I. Europhys Lett 1996;35:695700.
[103] Muller M. Macromol Theory Simul, in press.
[104] Anderson DM, Thomas EL. Macromolecules 1988;21:3221.
[105] Marques CM, Cates ME. Europhys Lett 1990;13:267.
[106] Olvera M. de la Cruz. Phys Rev Lett 1991;67:85.
[107] Disko MM, Liang KS, Behal SK. Macromolecules 1993;26:2983.
[108] Matsen MW, Schick M. Phys Rev Lett 1994;72:2660.
[109] Schulz MF, Khandpur AK, Bates FS, Almdal K, Mortensen K, Hajduk DS, Gruner SM. Macromolecules 1996;29:2857.
[110] Matsen MW, Bates FS. Macromolecules 1996;29:7641.
[111] Matsen MW, Bates FS. J Chem Phys 1997;106:2436.
[112] Laradji M, Shi A-C, Noolandi J, Desai R. Macromolecules 1997;30:3242.
[113] Hashimoto T, Shibayama M, Kawai H. Macromolecules 1980;13:123747.
[114] Richards RW, Thomason JT. Macromolecules 1983;16:98292.
[115] Hasegawa H, Tanaka H, Yamasaki K, Hashimoto T. Macromolecules 1987;20:165162.
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1092
[116] Almdal K, Rosendale JH, Bates FS, Wignall GD, Fredrickson GH. Phys Rev Lett 1990;65:11125.
[117] Bartels VT, Stamm M, Aberz V, Mortensen K. Europhys Lett 1995;31:816.
[118] Fredrickson GH, Milner ST. Phys Rev Lett 1991;67:835.
[119] Dobrynin AV, Leibler L. Macromolecules 1997;30:4756.
[120] Potemkin II, Panyukov SV. Phys Rev E 1998;57:6902.
[121] Aksimentiev A, Holyst R. J Chem Phys 1999;111:2329-39.
[122] Marko JF. Macromolecules 1993;26:14424.
[123] Weyersberg A, Vilgis TA. Phys Rev E 1994;49:3097101.
[124] Fried H, Binder K. J Chem Phys 1991;94:834966.
[125] Weyersberg A, Vilgis T. Phys Rev E 1993;48:37790.
[126] David EF, Schweizer KS. J Chem Soc Faraday Trans 1995;91:241125.
[127] Pakula T. Makromol Chem Theory Simul 1993;3:23943.
[128] Binder K, Fried H. Macromolecules 1993;26:687883.
[129] Wolff T, Burger C, Ruland W. Macromolecules 1993;26:170711.
[130] Sakamoto N, Hashimoto T. Macromolecules 1995;28:682534.
[131] Matsen MW. J Chem Phys 1995;102:38847.
A. Aksimentiev, R. Hoyst / Prog. Polym. Sci. 24 (1999) 10451093 1093

You might also like