Novel Monte Carlo Molecular Simulation Scheme Using

Identity-Altering Elementary Moves for the Calculation of Structure

and Thermodynamic Properties of Polyolefin Blends
Loukas D. Peristeras,

Anastassia N. Rissanou,

Ioannis G. Economou,

and
Doros N. Theodorou*
,,
Molecular Thermodynamics and Modelling of Materials Laboratory, Institute of Physical Chemistry,
National Center for Scientific Research Demokritos, GR-153 10 Aghia ParaskeVi, Attikis, Greece,
and Department of Materials Science and Engineering, School of Chemical Engineering, National
Technical UniVersity of Athens, 9 Heroon Polytechniou Street, GR-157 80 Athens, Greece
ReceiVed NoVember 25, 2006; ReVised Manuscript ReceiVed February 4, 2007
ABSTRACT: In this work, the semi-grand statistical ensemble formalism proposed by Pant and Theodorou
(Macromolecules, 1995, 28, 7224) for polydisperse melts of linear chains is extended to mixtures of linear and
branched chains in order to calculate stability of binary polymer blends based on Monte Carlo simulation.
Furthermore, a new elementary Monte Carlo move is introduced that consists of chain identity-altering between
branched and linear chain species and thus allows fluctuations in the blend composition. The new move, together
with previously developed moves for long chain molecules, are shown to relax efficiently both the branched and
the linear molecules in binary linear polyethylene-triarm polyethylene blends of various macromolecular sizes.
Subsequently, they are used for the calculation of the microscopic structure and the thermodynamic properties of
these systems. All blends examined are shown to be fully miscible.
1. Introduction
Polymer blends have attracted considerable interest in recent
years both in the research community and in industry.
1,2
The
superior mechanical, electrical, thermophysical, optical, rheo-
logical, and other properties of many polymer blends and their
favorable cost in comparison to their corresponding polymer
components have encouraged the use of these materials in a
wide range of industrial applications. In all these applications,
a better understanding and an ability to predict blend properties
from chemical constitution and composition is crucial. An
important issue is the miscibility of the polymers in a blend.
For most applications, it is desirable that the phase behavior of
the polymer blend be accurately known, since it will affect its
physical properties and, consequently, its processability and
performance in the end-use environment. In parallel, the
molecular engineering design of new polymeric materials
based on blends with tailored properties is of great interest. This
work focuses on polyolefin blends, a widely used and exten-
sively studied but still incompletely understood class of polymer
blends.
A large number of experimental groups worldwide has
devoted considerable effort in studying the thermodynamics and
phase behavior of blends of polymers that differ in macromo-
lecular architecture, polarity, macromolecular size, and composi-
tion. A wide range of state-of-the art methods, including small-
angle neutron scattering (SANS), small-angle X-ray scattering
(SAXS), small-angle light scattering (SALS), solid-state NMR,
and atomic force microscopy (AFM) were used for such
purposes.
3-10
It is clear today that binary polyolefin blends
exhibit all different types of fluid phase equilibria, such as upper
critical solution temperature (UCST), lower critical solution
temperature (LCST), and even both UCST and LCST at different
temperatures.
11
Despite extensive work by different research
groups, the phase behavior of many common polymer blends,
as for example high-density polyethylene-low-density poly-
ethylene (HDPE-LDPE), is not fully understood yet.
12-14
Such
difficulty is rooted in experimental uncertainties, due to the
similarity in the constituent groups of the polymers that does
not allow clear contrast in scattering experiments. Partial
deuteration of one of the polymers does not alleviate the problem
fully.
10
Molecular simulation is a reliable tool for the elucidation of
microscopic mechanisms that control macroscopic behavior and
subsequent calculation of thermodynamic properties and phase
behavior. Molecular dynamics (MD) is a convenient method
for the simulation of non-polymeric systems. However, for
polymers that exhibit long characteristic times for relaxation
that cannot be captured with MD, Monte Carlo (MC) simulation
is the preferable method. In this case, smart elementary moves
are needed, so that important configurations are sampled
efficiently. In recent years, a number of elementary MC moves
were proposed specifically for such purposes. Some of the most
widely used moves for polymers include: reptation,
15
configu-
rational bias (CB),
16-18
concerted rotation (CONROT),
19-20
flip,
21
end-bridging (EB),
20,22
and double-bridging (DB).
23
All
of these moves were designed primarily for linear molecules
and relax efficiently an internal or a terminal part of the chain
molecule.
Recently, two new elementary moves were proposed for
branched chain molecules:
24
Branch point flip (BPF) results in
a displacement in space of a branch point and its neighboring
atoms in the molecule without change in the connectivity. Wick
and Siepmann have proposed a similar elementary move named
self-adapting fixed-end-point configurational-bias for the simu-
lation of small branched and cyclic alkanes.
25
An additional
* Author to whom all correspondence should be addressed. E-mail:
doros@central.ntua.gr.

Molecular Thermodynamics and Modelling of Materials Laboratory,

Institute of Physical Chemistry, National Center for Scientific Research
Demokritos.

Department of Materials Science and Engineering, School of Chemical

Engineering, National Technical University of Athens.
2904 Macromolecules 2007, 40, 2904-2914
10.1021/ma0627121 CCC: $37.00 2007 American Chemical Society
Published on Web 03/16/2007
move proposed is the so-called branch point slithering (BPS)
that allows for branch point displacement along the backbone;
24
in this case, the size of the branches involved is changed. In
this work, we further extend these moves by introducing the
identity-altering (IA) elementary move for systems consisting
of linear and branched molecules. The IA move is an intermo-
lecular bridging move where part of or an entire branch of a
linear or a branched chain is cut and appended to a different
chain (either linear or branched). In this way, the local
configuration changes dramatically and, in addition, the mixture
composition (linear vs branched) is altered. It is shown that the
new move results in efficient relaxation of linear and branched
chain molecules.
In a typical phase equilibrium calculation using MC simula-
tion, molecules of the various components involved are moved
from one phase to another.
26
In some cases, the chemical
potential of one or more of the components needs to be
evaluated.
27
Both operations become very cumbersome, if not
impossible, for polymer systems. A number of alternative
methodologies have been proposed to overcome such
restrictions.
28-30
A useful approach is the use of a semi-grand
statistical ensemble allowing interconversion between different
components, wherein appropriate differences in chemical po-
tential between the components are specified.
In this work, the new move is developed in the semi-grand
statistical ensemble of Pant and Theodorou
20
extended to
pseudobinary mixtures. The chemical potential difference
between the two components is specified and the system
composition is calculated. In this way, a single phase system is
studied and no chemical potential calculation is needed. The
proposed method is applied to mixtures of linear polyethylene
and triarm polyethylene of the same molecular weight. Micro-
scopic structure, chain sizes, and thermodynamic properties are
reported. For all of the mixtures examined, the two components
exhibit similar characteristics and so are fully miscible. A similar
methodology has been proposed by Kofke and Glandt for simple
Lennard-Jones binary mixtures.
31,32
2. Development of the Statistical Ensemble
2.1. Definition of the System. The system under consider-
ation consists of N polymer chains of which N
l
are linear and
N
b
are branched, so that ) N
l
+ N
b
. The total number of
atoms is n. A branch (or an arm) is defined as the linear part of
a chain that connects a chain end or a branch point with another
chain end or branch point of the chain. Consequently, a linear
chain is considered as a single branch and a branched chain
consists of at least three branches. In general, every component
k has b
k
branch points. The number of atoms in each branch q
is denoted by
q
. Every branch point (bp) of the chain is
characterized by the functionality index, f
bp
, which is the number
of branches attached to it. All branched chains in the system
are assumed to have the same macromolecular architecture (they
can be stars, H-shaped, brushes, etc.) with N
br
ch
branches per
chain. Therefore, b
k
) 0 for linear chains and b
k
) b for
branched chains, and
The branched chains that we will examine here are assumed to
have f
bp
) 3, so that N
br
ch
) 1 + 2b.
2.2. The Statistical Ensemble. The statistical ensemble is a
generalization of the semi-grand [NnPT*] canonical ensemble
of Pant and Theodorou
20
for a mixture of linear and branched
polymers with branches of variable mean molecular weight. For
this mixture, changes in the Helmholtz free energy (A) are given
by the following expression:
where S, T, P, and V are the entropy, temperature, pressure,
and volume of the system, respectively, N
k
and
k
are the number
of molecules and the chemical potential of component k,
respectively, and m is the total number of components in the
system.
The partition function of the canonical ensemble for the
system is
where Z(V,T,N
1
,N
2
,..,N
m
) is the configurational integral, equal
to
with r
i
being the Cartesian coordinates of atom i, ) 1/k

T
(k
B
is the Boltzmann constant), the monomer (atom) thermal
wavelength, and V the potential energy function.
In the system under consideration, the total numbers of chains
and of atoms are calculated from the expression
with n
k
being the number of atoms in a chain of species k.
Selecting an arbitrary pair of components i and j as reference
components, variables N
i
and
j
can be eliminated by invoking
eqs 5 and 6 in favor of n and N. In this way, eq 2 results in
where
A Legendre transformation of the Helmholtz free energy with
respect to V and N
k
, k ) 1, 2, ..., m: k * i, j, results in the
thermodynamic potential Y:
dA ) - S dT - P dV +

k)1
m

k
dN
k
(2)
Q(V,T,N
1
,N
2
,...,N
m
) )
1
N
1
!N
2
!...N
m
!
1

3n
Z(V,T,N
1
,N
2
,...,N
m
)
(3)
Z(V,T,N
1
,N
2
,...,N
m
) )

d
3n
r exp[- V(r
1
,r
2
,...,r
n
)]
(4)

k)1
m
N
k
) N (5)

k)1
m
n
k
N
k
) n (6)
dA ) - S dT - P dV +

k)1
k*i,j
m

k
/
dN
k
+
(n)
dn +
(N)
dN
(7)

k
/
)
k
+
(
n
k
- n
j
n
j
- n
i
)

i
-
(
n
k
- n
i
n
j
- n
i
)

j
(8)

(n)
)

i
-
j
n
i
- n
j
(9)

(N)
)
n
i

j
- n
j

i
n
i
- n
j
(10)
N
br
ch
) 1 + (f
bp
- 1)b (1)
Macromolecules, Vol. 40, No. 8, 2007 Polyolefin Blends 2905
The partition function Y of the statistical ensemble that
corresponds to this potential is related to Y through the
expression
and is equal to
Consequently, the probability density function of this ensemble
is
2.3. Size Distribution of the Chain Branches. Given the
fact that that the number of branch points per branched chain
remains constant, one may invoke the conformal solution
principle for the configurational integral:
20
According to eq 16, the configurational integral depends on the
number-averaged molecular weight, density of monomers and
mole fraction of branched (or linear) chains but not on the details
of the chain length distribution of linear and branched species.
This assumption has been shown to be satisfactory for systems
of sufficiently long linear polyethylene chains.
20
Equation 16
allows us to treat our system as a pseudobinary, consisting of
linear and branched chains.
In the thermodynamic limit (N
i
f , i ) 1, 2, ..., m, all
intensive properties of the system being kept constant) by
invoking the maximum term approximation and using eqs 5, 6,
and 16, eq 14 becomes
Maximization of Y is equivalent with maximization of the term
in brackets in eq 17 with the constraints that the total number
of molecules N and the total number of monomers n remain
constant. The method of Lagrange multipliers is used for this
purpose. The expressions for the partial derivatives of the term
under consideration, where , are the Lagrange multipliers
that correspond to the constraints of eqs 5 and 6, respectively,
are
where s ) 1, 2, ..., m. Equation 18 for the various species s
becomes
where c ) exp[], y ) exp[], z ) exp[], and ) (ln(Z))/
(N
b
). Coefficients c and y can be calculated from eqs 5 and 6
for given z and
s
/
.
Manipulating the composition of the system is possible by
using different chemical potential values for each species. One
of the possible sets of values for the statistical ensemble under
consideration is
which imposes the same chemical potential for all linear
components, and the same relative chemical potential with
respect to linear components for all branched components. Such
an approach is meaningful for the case where all branched
species have the same macromolecular architecture.
In eq 22, we have to specify the ranges of sizes for which
the chemical potential differs from negative infinity and are
therefore allowed in the system. Our intention is to impose a
uniform distribution on the size of branches of the species
present in the mixture, within a preselected range of values.
This range is defined according to the expression:
where j
l
is the mean chain size for linear chains and j
q
is the
mean branch size for branch q of a branched chain. Taking into
account that the total number of segments in a branched chain
is b +
q)1
1+2b

q
, one obtains
Y ) A + PV -

k)1
k*i,j
m

k
/
N
k
(11)
dY ) - S dT + V dP -

k)1
k*i,j
m
N
k
d
k
/
+
(n)
dn +
(N)
dN
(12)
Y ) - k
B
T ln(Y) (13)
Y )

{N
1
,N
2
,...,N
m
}:

k)1
m
N
k
)N
exp[

k)1
k * i,j
m

k
/
N
k
]
1
V
0

dV
exp[- PV]Q(V,T,N
1
,N
2
,...,N
m
) (14)
F
NnPT*
(V,r
1
,r
2
,...,r
n
;connectivity) exp[

k)1
k*i,j
m

k
/
N
k
-
PV - V(r
1
,r
2
,...,r
n
;connectivity)] (15)
Z(V,T,N
1
,N
2
,...,N
m
) ) Z(V,T,N,n,N
b
) (16)
Y exp[

k)1
k*i,j
m

k
/
N
k
- PV -

k)1
m
N
k
ln N
k
+

k)1
m
N
k
- 3nln + ln Z(V,T,N,n,N
b
)] (17)

N
s
[term of eq 17 in brackets + (

k)1
m
N
k
- N) +
(

k)1
m
n
k
N
k
- n)] ) 0 (18)

s
/
- ln(N
s
) +
ln(Z)
N
b
N
b
N
s
+ + n
s
)

s
/
- ln(N
s
) +
b
s
b
+ + n
s
) 0 w
N
s
) cy
n
s
z
b
s
/b
exp[
s
/
], s ) 1,2,...,m: s * i, j (19)
- ln(N
i
) +
b
i
b
+ + n
i
) 0 wN
i
) cy
n
i
z
b
i
/b
(20)
- ln(N
j
) +
b
j
b
+ + n
j
) 0 wN
j
) cy
n
j
z
b
j
/b
(21)

k
/
)
{
0 for b
k
) 0 [linear species with a given range of sizes]
* for b
k
) b [branched species with a given range of branch sizes]
- all other cases
}
(22)

{
[j
l
(1 -
l
), j
l
(1 +
l
)] for b
q
) 0
[j
q
(1 -
q
), j
q
(1 +
q
)] for b
q
) b, q ) 1, 2, ..., (1 + 2b)
}
(23)
2906 Peristeras et al. Macromolecules, Vol. 40, No. 8, 2007
Given the fact that y f 1 w f 0, this is equivalent with
relaxing the constraint of the total number of monomer units
being constant, which is possible if the branch size distribution
is well-defined. In this way one may define the mean chain
size and, since the total number of chains in the system is
constant, then the total number of atoms will also be constant.
For this reason, eqs 24 and 25 should have a solution for y f
1 and so
If the mean size of linear chains (Xh
l
) is set equal to the mean
size of branched chains (Xh
b
), then:
and the two equations become linearly dependent. In this way,
eq 26 is solved for c:
2.4. Calculation of Mole Fractions. The c value calculated
from eq 29 can be used for the evaluation of the number of
molecules and consequently for the mole fraction of linear and
branched species in the system under consideration:
and the mole fractions of linear and branched species become
2.5. Molecular Simulation Strategy. In our semi-grand
canonical simulations we specify a profile of relative chemical
potentials according to eqs 22 and 23. At equilibrium, the system
responds with a certain composition (distribution of molecular
species present). If the conformal solution assumption, eq 16,
holds, then the size distributions of linear species and of
branches belonging to branched species should be uniform
within the ranges defined by eq 23. (This will be tested). Also,
the equilibrium composition should conform to eqs 32 and 33.
Having set * and having obtained x
b
as an output from
molecular simulation, we may calculate z, which is a measure
of the nonideality of mixing, through eq 32 or eq 33:
or
If the mixture is ideal, then (compare also eq 41 below):
and so z is independent of the composition. For this value of z,
the composition dependence of * is identical to the depen-
dence for an ideal solution:
Molecular simulation provides the dependence of * on the mole
fraction of branched species, *(x
b
). From this dependence, the
activity coefficients of the components can be evaluated. For
the binary linear chain-branched chain mixture under consid-
eration, the Gibbs-Duhem equation under constant temperature
and pressure results in
N ) c

k
)j
l
(1-
l
)
j
l
(1+
l
)
y

k
+ czy
b
exp[*]

q)1
1+2b

q
)j
q
(1-
q
)
j
q
(1+
q
)
y

q
(24)
n ) c

k
)j
l
(1-
l
)
j
l
(1+
l
)

k
y

k
+ czy
b
exp[*]
{b

q)1
1+2b
(

q
)j
q
(1-
q
)
j
q
(1+
q
)
y

q
) +

q)1
1+2b
(

q
)j
q
(1-
q
)
j
q
(1+
q
)

q
y

q
)

q)1
q*q
1+2b
(

q
)j
q
(1-
q
)
j
q
(1+
q
)
y

q
)} (25)
N ) c

k
)j
l
(1-
l
)
j
l
(1+
l
)
[1] + cz[1]
b
exp[*]

q)1
1+2b

q
)j
q
(1-
q
)
j
q
(1+
q
)
[1]
) c(2j
l

l
+ 1) + c z exp[*]

q)1
1+2b
(2j
q

q
+ 1)
(26)
n ) c

k
)j
l
(1-
l
)
j
l
(1+
l
)

k
+ cz[1] exp[*]{b

q)1
1+2b

q
)j
q
(1-
q
)
j
q
(1+
q
)
[1] +

q)1
1+2b
j
q
q)1
1+2b

q
)j
q
(1-
q
)
j
q
(1+
q
)
[1]}
) cj
l
(2j
l

l
+ 1) + cz exp[*](b +

q)1
1+2b
j
q
)

q)1
1+2b
(2j
q

q
+ 1) (27)
Xh ) Xh
l
) Xh
b
)
n
N
; Xh
l
) j
l
; Xh
b
) b +

q)1
1+2b
j
q
(28)
c )
N
(2j
l

l
+ 1) + z exp[*]

q ) 1
1+2b
(2j
q

q
+ 1)
(29)
N
l
) c(2j
l

l
+ 1) (30)
N
b
) cz exp[*]

q ) 1
1+2b
(2j
q

q
+ 1) (31)
x
l
)
2j
l

l
+ 1
2j
l

l
+ 1 + z exp[*]

q ) 1
1+2b
(2j
q

q
+ 1)
(32)
x
b
)
z exp[*]

q ) 1
1+2b
(2j
q

q
+ 1)
2j
l

l
+ 1 + z exp[*]

q ) 1
1+2b
(2j
q

q
+ 1)
(33)
eq 33 wz )
x
b
1 - x
b
2j
l

l
+ 1
exp[*]

q ) 1
1+2b
(2j
q

q
+ 1)
* ) ln
[
2j
l

l
+ 1
z

q ) 1
1+2b
(2j
q

q
+ 1)
]
+ ln
[
x
b
1 - x
b
]
(34)
)
ln(Z)
N
b
|
V,T,N,n
) constant wz ) exp[constant]
(35)
*
,id
) (
b
-
l
) ) (
b,pure
-
l,pure
) + ln
[
x
b
1 - x
b
]
(36)
{
(1 - x
b
)d
l
|
T,P
+ x
b
d
b
|
T,P
) 0
d
l
|
T,P
- d
b
|
T,P
) d*|
T,P
}
w
{
d
b
|
T,P
) (1 - x
b
) d*|
T,P
d
l
|
T,P
) - x
b
d*|
T,P
}
(37)
Macromolecules, Vol. 40, No. 8, 2007 Polyolefin Blends 2907
Taking the reference state for each component as the pure
component at the mixture temperature and pressure, we can
write for the activity coefficients
b
,
l
of the branched and
linear species, respectively:
The thermodynamic eqs 37, 38 are perfectly general for a two-
component system.
According to the definition of z:
and the standard thermodynamic expression:
z can be expressed as a function of the activity coefficients:
where P
ig
) Fk
B
T, F is the total molecular density of the system
and Z
i
intra
is the intramolecular configurational integral over all
orientations and over all conformations of species i in the ideal
gas state. Equation 41 indicates that z depends on the ratio of
activity coefficient values.
Finally, the system stability can be assessed; for a stable
single-phase mixture, we must have
33
In Figure 1, a schematic representation of the variation of *
as a function of composition, x
b
, is shown for a fully miscible
and a partially miscible mixture.
3. Identity-Altering Elementary Monte Carlo Moves
The systems examined in this work are binary mixtures of
linear chains and branched chains with a single trifunctional
branch point (triarm chains), so that f
bp
) 3 and b ) 1. In a
recent work,
24
a number of elementary MC moves were
proposed specifically for long-chain branched molecules. Here,
we further extend this approach by developing moves that allow
IA, that is conversion from a linear chain to a branched chain
and vice versa. As in previous publications,
20,24
the derivation
is formulated in the (N,n,P,T,
*
) semi-grand statistical ensemble.
These moves allow efficient convergence of the mole fraction
to the value that corresponds to the equilibrium under the set
thermodynamic conditions (chemical potential difference, tem-
perature, pressure). The new moves combine previously devel-
oped moves for chain molecules, such as CB and EB, so that a
branch, or part of it, is transferred from one chain to another
chain. In Figure 2, the four different ways by which the identity
of chains is altered according to the moves developed here are
shown. Every branch that is involved in a move may contain
additional smaller branches, so that this set of moves can be
further extended to more complex systems. The details of these
moves are presented below.
3.1. Branch Creation (BC) Move. A branch is created
through a chain cut and connection to a linear chain. This move
(BC, in short) is shown schematically in Figure 3. In all cases,
a trimer is relocated in the spirit of the CONROT move.
19-20
The BC move is implemented in two steps:
(a) Initially, the end-bridging move is used so that the branch
is connected to its new position. One of the terminal four atoms
(a, b, c, d) of the donor chain attacks atom iv that was
selected from a list of candidates (fulfilling the MW distribution
constraints) as the new branch point on the acceptor chain.
For the connection of the new branch, the following strategy is
used: The donor chain is cut at a randomly selected point
(in Figure 3, it is cut between atoms 5 and 6). The atoms
necessary to form the first atom of the new branch and to bridge
it with the branch point are taken from the vicinity of the cut
point. For example, if atom b attacks atom iv, then atoms 6
and 7, along with atom a, should be used to construct a trimer
bridge between atom iv and atom b.
(b) In order to relax the nonbonded interactions in the vicinity
of the cut point, a number of atoms are repositioned using the
CB algorithm. For example, with reference to Figure 3, if atom
c attacks atom iv, then atom 6 is cut and moved to assist the
{

b
)
b,pure
+ k
B
T ln(
b
x
b
)

l
)
l,pure
+ k
B
T ln(
l
x
l
)
}
w
{
d
b
|
T,P
)
dx
b
x
b
+ d(ln(
b
))
d
l
|
T,P
) -
dx
b
1 - x
b
+ d(ln(
l
))
}
w
{
d(ln(
b
))|
T,P
) (1 - x
b
) d*|
T,P
-
dx
b
x
b
d(ln(
l
))|
T,P
) - x
b
d*|
T,P
+
dx
b
1 - x
b
}
w
{
d ln(
b
)
dx
b
|
T,P
) (1 - x
b
)
d*
dx
b
|
T,P
-
1
x
b
d ln(
l
)
dx
b
|
T,P
) -x
b

d*
dx
b
|
T,P
+
1
1 - x
b
}
w
{
ln(
b
) )

x
b
1
[
1
x
b
- (1 - x
b
)
d*
dx
b
|
T,P
]
dx
b
ln(
l
) )

0
x
b
[
1
1 - x
b
- x
b

d*
dx
b
|
T,P
]
dx
b
}
(38)
z ) exp[] ) exp
[
(ln(Z))
N
b
|
V,T,N,n
]
(39)
* )
b
-
l
)
b,pure
-
l,pure
+
1

ln
(

l
)
+
1

ln
[
x
b
1 - x
b
]
(40)
z )

l

b
exp[- (
b,pure
(T,P) -

l,pure
(T,P))] exp[(
b,pure
ig
(T,P
ig
) -
l,pure
ig
(T,P
ig
))]
Z
b
intra
Z
l
intra
(41)
x
b
d(ln(
b
))
dx
b
|
T,P
+ 1 g 0 w
d*
dx
b
|
T,P
g 0 (42)
Figure 1. Schematic representation of the dependence of the relative
chemical potential of branched chains, *, as a function of the
composition, xb, for a binary mixture of linear and branched chains.
For a stable mixture * varies monotonically with the composition,
whereas for an unstable mixture * goes through a local maximum
and a local minimum value (points on the spinodal) that define the
region of instability of the one-phase mixture.
2908 Peristeras et al. Macromolecules, Vol. 40, No. 8, 2007
new branch bridging to the acceptor chain, while two atoms
in the vicinity of the cut are repositioned. In Table 1, the various
options concerning the attack atoms, the moving atoms and the
repositioned atoms are shown.
3.2. Branch Destruction (BD) Move. Cutting a branch of a
branched chain and connecting it with the end of another chain
results in the creation of a linear chain. This move (BD, in short)
is in essence the opposite of the BC move (three atoms are
repositioned again) and takes place in two steps that are briefly
presented here (see Figure 4):
(a) An end-bridging move is used initially in order to connect
one of the four terminal atoms of the donor branch with a
randomly selected atom from the acceptor chain (the latter
atom should be located close to the acceptor chain end).
Consequently, a number of atoms of the donor branch are
relocated from the vicinity of the branch point to the bridging
area.
(b) A number of atoms are repositioned in the vicinity of the
branch point of the donor chain using the CB algorithm. In
this case, this is necessary since the nonbonded interactions near
the branch point are unrealistically high because of the increased
local density.
3.3. Acceptance Criteria. The elementary moves presented
above proceed as follows: A chain is randomly selected,
followed by the random selection of one of its atoms close to
a branch end (one of the four terminal atoms of a branch). This
will be the attack atom. Furthermore, it is decided randomly
whether the move will be the creation or destruction of either
a linear or a branched chain. Finally, a tentative acceptor chain
is selected.
The following microscopic reversibility equation is the
starting point for the derivation of the acceptance criterion of a
MC move:
where a
(ifj)
is the conditional attempt probability of going from
state i to state j, A
(ifj)
is the conditional acceptance probability
of going from state i to state j and P
(i)
is the equilibrium
probability of the system being in state i. Subscripts o and n
denote old and new states, respectively.
In a recent work on the development of elementary MC
moves for the branch point of chain molecules,
24
the detailed
derivation of acceptance probability for the branch point flip
(BPF) and branch point slithering (BPS) moves were presented.
The derivation of acceptance probability for BC or BD is based
to a large extent on the same principles. In the bridge step and
the atom reposition step of the BC move statistical weights,
w
bridge
and w
cb
(cb here stands for configurational bias),
respectively, are introduced. Consequently, the conditional
attempt probability for the system to move from state i to state
j is
where N is the number of chains, n
ep
ch
is 2 for a linear chain and
3 for a branched chain, and n

is the number of terminal atoms

in the chain that are possibly involved in the move (here n

)
4). Finally, a
(ifj)
accounts for the random selection of the type
of chain (linear or branched) through the 1/2 term. Incorporation
of eq 44 into the general expression for the acceptance criterion
24
results in the acceptance criterion for the branched chain creation
move:
and the acceptance criterion for linear chain creation move:
Figure 2. Schematic representation of IA elementary moves using
intermolecular rebridging: In (a), a linear chain attacks another linear
chain in order to form a branch. In the reverse move (b), a linear chain
attacks a branched chain in order to incorporate one of its branches.
In (c), a branched chain attacks another branched chain in order to
incorporate one of its branches, converting it to a linear chain. In the
reverse move (d), a branched chain attacks a linear chain, providing
to it part of one of its branches and converting it to a branched chain.
In all cases, the cross defines the point where a branch is cut. In the
accompanying chemical equations, the moves are represented as
bimolecular reactions.
Figure 3. Schematic representation of the branch creation (BC) move.
The donor chain may be a linear or a branched chain. Here, the donor
chain is cut between atoms 5 and 6 and the new branch point in the
acceptor chain is atom iv.
Table 1. Attack Atoms, Moving Atoms, and Repositioned Atoms for
the BC Move (With Reference to Figure 3)
a
options for the atom repositioning
attack atom moving atoms 1st 2nd 3rd 4th
a 6, 7, 8 none none none none
b 6, 7 5 8
c 6 4, 5 5, 7 7, 8
d none 3, 4, 5 4, 5, 6 5, 6, 7 6, 7, 8
a
The repositioned atoms in bold belong to the newly formed branch.
a
(ofn)
A
(ofn)
P
(o)
) a
(nfo)
A
(nfo)
P
(n)
(43)
a
(ifj)
)
1
N
1
n
ep
ch
1
2
1
n

w
bridge(ifj)
w
cb(ifj)
(44)
A
(ofn)
) min
(
1,

s)1
no. of steps
a
(nfo),s
F
n
s)1
no. of steps
J
n,s

s)1
no. of steps
a
(ofn),s
F
o
s)1
no. of steps
J
o,s
)
(45)
A
(ofn)
)
min
(
1,
n
,o
w
bridge(nfo)
w
cb(nfo)
exp(- * - V
n
)J
br,n
J
cb,n
n
,n
w
bridge(ofn)
w
cb(ofn)
exp(- V
o
)J
br,o
)
(46)
Macromolecules, Vol. 40, No. 8, 2007 Polyolefin Blends 2909
The two Jacobians J
br
and J
cb
can be found in refs 34 and 16-
18, respectively, and are not repeated here.
4. Simulation Details
In order to assess the proper implementation of the statistical
ensemble, the correctness of the new elementary moves
introduced and the overall efficiency of the method, ideal
mixtures of phantom triarm symmetric polyethylene chains with
their linear analogs were examined. In phantom chains only
bond angle bending interactions are nonzero, whereas dihedral
angle torsional potentials and all nonbonded intramolecular and
intermolecular interactions are set equal to zero. Bond lengths
were fixed in all calculations. Subsequently, the same mixtures
were examined using full intra- and intermolecular interactions.
The mean number of carbon atoms per chain (Xh) was 61 (j
q
)
20 carbon atoms per branch and a branch point), 121 (j
q
) 40
carbon atoms per branch and a branch point), or 301 (j
q
) 100
carbon atoms per branch and a branch point). The branch
molecular weight distribution for the triarm chains and the chain
molecular weight distribution for the linear chains were uniform
in the interval [Xh(1 - ), Xh(1 + )] with
l
)
b
) ) 0.5.
The simulation box contained a total number of 80 chains
for the C
61
mixture, 40 chains for the C
121
mixture and 12 chains
for the C
301
mixture. In all cases, the temperature was set equal
to 450 K and the pressure at 0.1 MPa. The relative chemical
potential values, *, were varied between -10k
B
T and 0k
B
T
for the phantom chain mixtures and between -4k
B
T and 5k
B
T
for the real mixtures. A typical MC simulation run, unless
otherwise indicated, consisted of the following elementary
moves: 10% reptation, 4% end-rotation, 19% CONROT, 25.5%
EB, 0.5% volume fluctuation, 3% DF, 4% BPF, 4% BPS, and
30% chain IA. All simulations were performed on AMD Athlon
MP 1800+ processors. A typical run consisted of 5 10
7
MC
moves. The acceptance rate for the chain IA move was on the
order of 0.02% which is a typical value for such complex
elementary moves that result in drastic changes of the config-
uration.
In all cases the force field employed was TraPPE-UA, an
accurate united atom (UA) force field developed for thermo-
dynamic properties of linear and branched alkanes and later
extended to other classes of organic compounds.
35,36
In TraPPE
bond lengths are kept constant, whereas bond angle bending
and dihedral angle torsions are calculated using appropriate
expressions. Finally, nonbonded intra- and intermolecular
interactions are calculated from the Lennard-Jones potential with
parameters fitted to saturated liquid density and vapor pressure
experimental data.
5. Results and Discussion
5.1. Mixtures of Phantom Chains. In order to assess the
proper implementation of the new statistical ensemble with the
new chain IA MC move, ideal mixtures of phantom triarm and
phantom linear chains of the same mean molecular weight were
examined. For a given relative chemical potential value, the
run consisted of 5 10
6
MC moves and the CPU time was 2.6
10
4
s. The moves used were 10% reptation, 4% end-rotation,
10% flip, 10% CONROT, 24.5% EB, 0.5% volume fluctuation,
3% DF, 4% BPF, 4% BPS, and 30% chain IA.
In Figure 5, the normalized distribution of the two types of
bond angles for the C
121
triarm chains is shown. For comparison,
the theoretical distribution ( sin() exp[-U
bend
()]) calculated
from the TraPPE-UA force field is plotted. The two sets of
calculations (molecular simulation and model predictions)
coincide over the entire spectrum of bond angle values.
Furthermore, the normalized distribution of torsion angles for
these chains is shown in Figure 6, together with the theoretically
expected uniform distribution. In a phantom chain, all torsion
angle values are equally probable. Indeed, the moves used do
not impose any bias in the torsion angle distribution; this
Figure 4. Schematic representation of the branch destruction (BD)
move. The acceptor branch may be part of a linear or a branched chain.
Here, the branch is cut at the cross and bridging occurs between atoms
b and iv of the two different branches.
Figure 5. Normalized distribution of the bond angles centered at the
-CH2- and the -CH- UA, respectively, of phantom C121 triarm (3
40 + 1) chains in a mixture with their linear analogs at 450 K, 0.1
MPa and * ) -5. The solid lines display the theoretical distributions
based on the TraPPE-UA force field, and the points are simulation
results.
Figure 6. Normalized distribution of the torsion angles around the
-CH2-CH2- and the -CH-CH2- bonds, respectively, of phantom
C121 triarm (3 40 + 1) chains in a mixture with their linear analogs
at 450 K, 0.1 MPa and * ) -5. There is no torsional potential and
so a uniform distribution is expected (solid line), which also results
from simulation.
A
(ofn)
)
min
(
1,
w
bridge(nfo)
w
cb(nfo)
exp(- V
n
)J
br,n
w
bridge(ofn)
w
cb(ofn)
exp(- * - V
o
)J
br,o
J
cb,o
)
(47)
2910 Peristeras et al. Macromolecules, Vol. 40, No. 8, 2007
provides a strong indication that the moves have been imple-
mented properly.
In Figure 7, the ability of the elementary moves used to
provide uniform sampling of all arm lengths allowed, based on
the chemical potential values imposed (eq 22), is shown. The
mean carbon number per arm is 40 and ) 0.5, so that the
allowable arm size is between 20 and 60 carbon atoms. A
uniform distribution (within a statistical uncertainty) is obtained,
ensuring the proper implementation of the move.
The probability distribution of the size of the entire phantom
triarm chain is calculated based on the distribution of the
individual arms (Figure 8). The shape of this distribution is close
to a Gaussian and the exact mathematical expression for it was
presented in eq 21 of ref 24. Simulation results follow very
closely the theoretical curve. In Figure 8, the probability
distribution for the size of linear chains of the same mixture is
shown. A uniform distribution is obtained within the predefined
range of sizes, thus ensuring a proper implementation of the
method.
The statistical ensemble examined in this work allows
accurate estimation of the mixture composition for a given
relative chemical potential value. Consequently, MC simulation
of binary mixtures at various chemical potential values can used
to evaluate the entire x
b
vs * curve and thus estimate the
mixture stability. In Figure 9, simulation results for three ideal
phantom linear chain-triarm chain mixtures differing in the size
of the chains are shown. The mean carbon number of linear
and triarm chains in the first mixture is 61, in the second mixture
it is 121, while in the third mixture it is 301. In all cases,
mixtures exhibit a monotonic dependence of * on x
b
and thus
the mixtures are fully stable, as expected of course. For such
ideal mixtures, z is composition independent (eq 35) and thus
the first term in eq 34 assumes a constant value based on the
mixture chain characteristics (mean arm size, arm size distribu-
tion, and linear chain size distribution). Indeed, in Figure 9,
the calculated * curves are shown for the three mixtures,
assuming that z is composition independent (compare eq 41).
The coefficients shown in the legend result from best fits to
the simulation data. Note that simulation results for all pairs
(i.e., C
61
-C
121
, C
61
-C
301
and C
121
-C
301
) of mixtures differ by
a constant value for all compositions. These offsets are
calculable from eq 34, if one assumes that z is the same in all
three ideal gas phantom chain mixtures:
On the other hand, the offsets between *(x
b
) curves can be
calculated analytically for such phantom chain ideal gas
mixtures. Using Mathematica, the following values were
obtained:
Figure 7. Probability distribution of the length (in carbon atoms) of
each of the three arms of phantom C121 triarm (3 40 + 1) chains in
a mixture with their linear analogs at 450 K, 0.1 MPa and * ) -5.
Figure 8. Probability distribution of the molecular size (in carbon
atoms) of phantom C121 triarm (3 40 + 1) chains and of linear C121
linear chains in a mixture at 450 K, 0.1 MPa and * ) -5. The solid
curve is the exact mathematical expression for this distribution.
24
Figure 9. Relative chemical potential, *, as a function of composi-
tion, xb, for the mixtures: (a) phantom linear C61-phantom triarm C61
(3 C20 + 1), (b) phantom linear C121-phantom triarm C121 (3 C40
+ 1), and (c) phantom linear C301-phantom triarm C301 (3 C100 +
1). Points are simulation results, and lines are best fits to the data
assuming a composition independent z.

C
121
/
-
C
61
/
) ln
[
2 120 0.5 + 1
(2 40 0.5 + 1)
3
]
-
ln
[
2 60 0.5 + 1
(2 20 0.5 + 1)
3
]
) -1.3 (48a)

C
301
/
-
C
61
/
) ln
[
2 300 0.5 + 1
(2 100 0.5 + 1)
3
]
-
ln
[
2 60 0.5 + 1
(2 20 0.5 + 1)
3
]
) -3.1 (48b)

C
301
/
-
C
121
/
) ln
[
2 300 0.5 + 1
(2 100 0.5 + 1)
3
]
-
ln
[
2 120 0.5 + 1
(2 40 0.5 + 1)
3
]
) -1.8 (48c)
Macromolecules, Vol. 40, No. 8, 2007 Polyolefin Blends 2911
which are in very good agreement (within statistical error) with
both the estimates of eq 48 above and with the best fits to the
simulation results, given in the legend to Figure 9.
5.2. Real Polyolefin Blends. The three binary blends were
further studied in the semi-grand statistical ensemble introduced
here using TraPPE-UA, a fully realistic intra- and intermolecular
potential.
35,36
Initially, the molecular size characteristics of the
two components were examined to ensure proper implementa-
tion of the method. For the C
61
mixture, the probability
distribution of the arm length for the triarm molecules and the
probability distribution of the overall chain size for the linear
and triarm molecules, both measured in carbon atoms, are shown
in Figures 10 and 11, respectively. Both distributions follow
the theoretically expected curves based on the conformal
solution assumption. Furthermore, the molecular size distribution
of the triarm polyethylene is considerably less noisy than the
size distribution of the same species in the C
121
ideal mixture
(Figure 7 for the arms and Figure 8 for the chain). This should
be attributed to the larger number of chains in the C
61
mixture
(80 chains) compared to the C
121
mixture (40 chains) and to
the longer run times.
The spatial extent of the molecules was examined for the
mixtures of linear and triarm polyethylene. For the root-mean-
square radius of gyration of polymer chains, the following
scaling equation holds
37
where X is the number of carbon atoms in the polymer and V is
0.5 for the case of unperturbed chains and 0.6 for the case of
good solvent conditions. In general, polymer melts exhibit
unperturbed chain behavior. In Figure 12, a double logarithmic
plot of R
g
2
vs carbon number is shown for the two components
in the C
301
mixture at 450 K, 0.1 MPa and x
b
) 0.5. Results for
both polymers agree well with eq 49 with V ) 0.5, thus verifying
unperturbed chain behavior. Furthermore, the linear chains
exhibit a higher R
g
compared to the branched chains. The ratio
g ) R
g
2

b
/R
g
2

l
for this mixture is 0.729 ( 0.035. Furthermore,
this ratio is equal to 0.725 ( 0.027 for the C
121
mixture and
0.740 ( 0.018 for the C
61
mixture. For comparison, the Zimm-
Stockmayer
38-39
theory based on the random coil hypothesis
predicts a value for g equal to 0.778.
An important consideration for the chain IA elementary move
introduced here is whether it is able to relax the long chain
molecules efficiently. Most of the elementary moves proposed
in the past, including CB, CONROT, EB, DB, BPS, and BPF,
were shown to be successful in this respect for the systems on
which they were tested. In order to assess the new moves
capability to relax long branched chains, three vectors are
introduced to characterize the geometry of the triarm chains
(Figure 13). These vectors are as follows: (a) The branch point
vector, v
bp
, that is the unit vector along the vectorial sum of
the three bonds emanating from the branch point (with reference
to Figure 13: v
bp
) (s
1
+ s
2
+ s
3
)/(|s
1
+ s
2
+ s
3
|)), (b) the arm
end-to-branch point unit vector, v
eb
, and (c) the arm end-to-
arm end unit vector, v
ee
. An autocorrelation function is defined
according to the product v
R
(0)v
R
(i), where v
R
(0) and v
R
(i)
are the unit vectors of type R (here, R ) bp, eb, or ee) at the
beginning of the simulation and after a certain number i of MC
moves.
In Figure 14, the autocorrelation functions for the three
characteristic unit vectors of a triarm C
61
polyethylene in the
mixture with a linear C
61
polyethylene at 450 K, 0.1 MPa, and
* ) 1 are shown. The blend of elementary MC moves used

C
121
/
-
C
61
/
) - 1.2,
C
301
/
-
C
61
/
) - 2.9,

C
301
/
-
C
121
/
) - 1.7
R
g
2

1/2
X
V
(49)
Figure 10. Probability distribution of the length (in carbon atoms) of
each of the three arms of C61 triarm (3 20 + 1) polyethylene in a
mixture with its linear analog at 450 K, 0.1 MPa and * ) 1.
Figure 11. Probability distribution of the molecular size (in carbon
atoms) of C61 triarm (3 20 + 1) polyethylene and of linear C61
polyethylene in mixture at 450 K, 0.1 MPa and * ) 1. The solid
curve is the exact mathematical expression for this distribution.
24
Figure 12. Double logarithmic plot of the mean squared radius of
gyration, Rg
2
, as a function of carbon number for the linear C301-
triarm C301 mixture at 450 K, 0.1 MPa and equimolar mixture. The
dotted line has a slope of 1 (V ) 0.5).
2912 Peristeras et al. Macromolecules, Vol. 40, No. 8, 2007
for these calculations relaxes very efficiently both the triarm
chain ends as well as the branch point.
An autocorrelation function is defined also for the end-to-
end unit vector of the linear chains, v
ee
(0)v
ee
(i). In Figure
15, v
ee
(0)v
ee
(i) for the linear C
61
polyethylene in the mixture
with a triarm C
61
polyethylene at 450 K and 0.1 MPa is shown
for different relative chemical potential values. In all cases, the
elementary moves used relax the chains efficiently. The *
value has a relatively small effect on the chain relaxation; the
autocorrelation function decreases more slowly for larger *
values. This should be attributed to the fact that larger *
values correspond to lower linear chain content (x
l
) and so linear
chain involvement in MC moves is less frequent.
MC simulations at 450 K and 0.1 MPa and different chemical
potential values were performed for the three polyolefin blends
and the composition was recorded, as shown in Figure 16. In
all cases, * varies monotonically with composition and so
all three blends are fully miscible. This is an expected result,
since the components in all blends are of relatively low mean
molar mass (856 g/mol for the C
61
blend, 1696 g/mol for the
C
121
blend and 4216 g/mol for the C
301
blend) and the
temperature is relatively high. At these conditions, experimental
data for high-density polyethylene-low-density polyethylene
blend reveal a fully miscible system.
40
Here again, as in Figure
9, theoretical calculations assuming a constant z value are in
reasonably good agreement, within the statistical uncertainty,
with simulation results. The mixtures close to ideal behavior
can be further assessed from the mass density variation with
composition. Results in Figure 17 for the case of C
121
blend at
450 K reveal that the mass density is practically constant, within
Figure 13. Schematic representation of the characteristic vectors in a
triarm chain: (a) The branch point vector, vbp, is the vectorial sum of
the three bond vectors emanating from the branch point, (b) the arm
end-to-branch point vector, veb, and (c) the arm end-to-arm end vector,
vee.
Figure 14. Relaxation of the characteristic vectors of triarm C61 (3
20 + 1) polyethylene in a mixture with linear C61 polyethylene at 450
K, 0.1 MPa and * ) 1.
Figure 15. Relaxation of the end-to-end characteristic vector of linear
C61 polyethylene in a mixture with triarm C61 (3 20 + 1) polyethylene
at 450 K, 0.1 MPa and various * values.
Figure 16. Relative chemical potential, *, as a function of
composition, xb, for the mixtures: (a) linear C61 polyethylene-triarm
C61 (3 C20 + 1) polyethylene, (b) linear C121 polyethylene-triarm
C121 (3 C40 + 1) polyethylene, and (c) linear C301 polyethylene-
triarm C301 (3 C100 + 1) polyethylene at 450 K and 0.1 MPa. Points
are simulation results and lines are the theoretically predicted curves
from eq 34. The coefficients shown in the legend are obtained by best
fits to the simulation data.
Figure 17. Mass density, F, as a function of composition, xb for the
mixture linear C121 polyethylene-triarm C121 (3 C40 + 1) polyeth-
ylene at 450 K and 0.1 MPa.
Macromolecules, Vol. 40, No. 8, 2007 Polyolefin Blends 2913
the statistical uncertainty of the simulation, over the entire
composition range.
In order to evaluate the blend nonideality at the microscopic
level, the radial distribution function between different UA pairs
was calculated. The vast majority of UA interaction sites in
both linear and triarm polyethylenes examined here are -CH
2
-.
In Figure 18, g(r) for CH
2
(linear)
- CH
2
(triarm)
pairs is shown at
various chemical potential values (and, consequently, various
compositions) for the C
61
blend. Results for the different
mixtures are almost indistinguishable, which indicates that
correlations between segments belonging to linear and triarm
chains do not change as the composition changes. In other
words, one may conclude that for these systems z remains
composition-independent. This is also verified by the fact that
by setting the first term on the right-hand side of eq 34 constant
and equal to 1.7 for the C
61
blend, a very good description of
the simulation results is obtained (Figure 16).
For considerably higher molecular weight and lower tem-
perature values, blends are expected to deviate from ideality.
Such systems require substantially higher computing power.
Preliminary results for polyolefin blends at 350 K and for blends
of linear and branched chains of different molecular weights
reveal some deviations from ideal mixture behavior which,
however, are not large enough to result in partially miscible
blends. Work is still underway and will be presented in a
forthcoming publication.
6. Conclusions
A new, efficient molecular simulation scheme in the frame-
work of the semi-grand statistical ensemble was proposed for
the calculation of structure, thermodynamic and phase equilib-
rium properties of binary linear-long branched polyolefin
blends. At the same time, a new elementary MC move was
developed that allows chain identity alterations (IA), from linear
to branched and vice versa. In this way, the mixture composition
converges to the value characteristic of the state point defined
by the semi-grand ensemble and the conformations of the long
chains relax efficiently. The new methodology was applied to
three linear polyethylene-symmetric triarm polyethylene blends
of the same mean molecular weight at 450 K and 0.1 MPa. All
systems were shown to be fully miscible and to be within error
bars of ideal solution thermodynamics. Future work will focus
on polyolefin blends that vary greatly in molecular size,
asymmetry between chain branches, temperature, and pressure.
Acknowledgment. Financial support for this work provided
by the General Secretariat of Research and Technology, Greece
through the PENED99 and Collaborative Research Projects with
Third Countries (Project No. -010) programs to IGE is
gratefully acknowledged. The CINECA Supercomputer Center
in Bologna, Italy, is acknowledged for a visiting post-graduate
fellowship and generous allocation of supercomputing time to
LDP through the MINOS program. IGE is thankful to the
Technical University of Denmark, Department of Chemical
Engineering, IVC-SEP for a visiting professorship during the
last part of this work, funded by the Danish Research Council
for Technology and Production Sciences.
References and Notes
(1) Manson, J. A.; Sperling, L. H. Polymer Blends and Composites;
Plenum Press: New York, 1976.
(2) Larson, R. G. The Structure and Rheology of Complex Fluids; Oxford
University Press: New York, 1999.
(3) Bates, F. S.; Wignall, G. D.; Koehler, W. C. Phys. ReV. Lett. 1985,
55, 2425.
(4) Banaszak, M.; Petsche, I. B.; Radosz, M. Macromolecules 1993, 26,
391.
(5) Reichart, G. C.; Graessley, W. W.; Register, R. A.; Krishnamoorti,
R.; Lohse, D. J. Macromolecules 1997, 30, 3363.
(6) White, J. L.; Brant, P. Macromolecules 1998, 31, 5424.
(7) White, J. L.; Lohse, D. J. Macromolecules 1999, 32, 958.
(8) Wolak, J.; Jia, X.; Gracz, H.; Stejskal, E. O.; White, J. L.; Wachowicz,
M.; Jurga, S. Macromolecules 2003, 36, 4844.
(9) Jeon, H. S.; Lee, J. H.; Balsara, N. P.; Newstein, M. C. Macromolecules
1998, 31, 5340.
(10) Stephens, C. H.; Hiltner, A.; Baer, E. Macromolecules 2003, 36, 2733.
(11) Rubinstein, M.; Colby, R. H. Polymer Physics; Oxford University
Press: New York, 2003.
(12) Puig, C. C.; Odell, J. A.; Hill, M. J.; Barham, P. J.; Folkes, M. J.
Polymer 1994, 35, 2452.
(13) Schipp, C.; Hill, M. J.; Barham, P. J.; Cloke, V. M.; Higgins, J. S.;
Oiarzabal, L. Polymer 1996, 37, 2291.
(14) Agamalian, M.; Alamo, R. G.; Kim, M. H.; Londono, J. D.;
Mandelkern, L.; Wignall, G. D. Macromolecules 1999, 32, 3093.
(15) Vacatello, M.; Avitabile, G.; Corradini, P.; Tuzi, A. J. Chem. Phys.
1980, 73, 548.
(16) Frenkel, D.; Mooij, G. C. A. M.; Smit, B. J. Phys.: Condens. Matter
1992, 4, 3053.
(17) Siepmann, J. I.; Frenkel, D. Mol. Phys. 1992, 75, 59.
(18) de Pablo, J. J.; Laso, M.; Siepmann, J. I.; Suter, U. W. Mol. Phys.
1993, 80, 55.
(19) Dodd, L. R.; Boone, T. D.; Theodorou, D. N. Mol. Phys. 1993, 78,
961.
(20) Pant, P. V. K.; Theodorou, D. N. Macromolecules 1995, 28, 7224.
(21) Mavrantzas, V. G.; Theodorou, D. N. Macromolecules, 1998, 31, 6310.
(22) Mavrantzas, V. G.; Boone, T. D.; Zervopoulou, E.; Theodorou, D. N.
Macromolecules, 1999, 32, 5072.
(23) Karayiannis, N. Ch.; Mavrantzas, V. G.; Theodorou, D. N. Phys. ReV.
Lett. 2002, 88, 105503.
(24) Peristeras, L. D.; Economou, I. G.; Theodorou, D. N. Macromolecules
2005, 38, 386.
(25) Wick, C. D.; Siepmann, J. I. Macromolecules 2000, 33, 7207.
(26) Panagiotopoulos, A. Z. Mol. Phys. 1987, 61, 813.
(27) Widom, B. Mol. Phys. 1963, 39, 2808.
(28) Spyriouni, T.; Economou, I. G.; Theodorou, D. N. Phys. ReV. Lett.
1998, 80, 4466.
(29) Zervopoulou, E.; Mavrantzas, V. G.; Theodorou, D. N. J. Chem. Phys.
2001, 115, 2860.
(30) Boulougouris, G. C.; Economou, I. G.; Theodorou, D. N. J. Chem.
Phys. 2001, 115, 8231.
(31) Kofke, D. A.; Glandt, E. D. J. Chem. Phys. 1987, 87, 4881.
(32) Kofke, D. A.; Glandt, E. D. Mol. Phys. 1988, 64, 1105.
(33) Modell, M.; Reid, R. C. Thermodynamics and its Applications; Prentice
Hall: New York, 1983.
(34) Karayiannis, N. Ch.; Giannousaki, A. E.; Mavrantzas, V. G.; The-
odorou, D. N. J. Chem. Phys. 2002, 117, 5465.
(35) Martin, M. G.; Siepmann, J. I. J. Phys. Chem. B 1998, 102, 2569.
(36) Martin, M. G.; Siepmann, J. I. J. Phys. Chem. B 1999, 103, 4508.
(37) de Gennes, P.-G. Scaling Concepts in Polymer Physics; Cornell
University Press: Ithaca, NY, 1979.
(38) Zimm, B. H.; Stockmayer, W. H. J. Chem. Phys. 1949, 17, 1301.
(39) Zimm, B. H.; Kilb, R. W. J. Polym. Sci. 1959, 37, 19.
(40) Walsh, D. J.; Graessley, W. W.; Datta, S.; Lohse, D. J.; Fetters, L. J.
Macromolecules 1992, 25, 5236.
MA0627121
Figure 18. Radial distribution function for linear methylene-branched
methylene pairs for a mixture of linear C61 polyethylene-triarm C61
(3 20 + 1) polyethylene at 450 K, 0.1 MPa and various relative
chemical potential values.
2914 Peristeras et al. Macromolecules, Vol. 40, No. 8, 2007