Full Paper

Molecular Dynamics Simulation of the

Formation of Polymer Networks

Wen Yang, Dongshan Wei, Xigao Jin, Qi Liao*

The random end linking of different amounts of trifunctional crosslinkers with 3 000 pre-
polymer linear chains, with length varying from 10 to 30 monomers, to form networks at
different system number densities was dynamically simulated by the molecular dynamics
method. Investigation of the crosslinking kinetics
shows that, with a stoichiometric number of cross-
linkers present, the time evolution of free ends frac-
tion decays as a power law in time t
3/4. This scaling
behavior is different from the one in a dense polymer
melt. Structural analyses of the resulting networks
indicate that the imperfections of the network struc-
ture strongly depend on the initial synthesis condi-
tions including the initial system number density and
the ratio of crosslinkers to precursor polymer chain
ends.

Introduction in synthetic macromolecules. They are the basic structural

elements for different systems such as rubbers and
Network polymers made by the stepwise polymerization polymer gels that have extensive applications in industrial
of polyfunctional monomers now play an important role and novel biological materials.[1–3] However, our under-
standing of the relationship between their structures and
the properties is far from complete.
W. Yang, D. Wei, X. Jin, Q. Liao Since the complicated topology and the ultimate
Beijing National Laboratory for Molecular Sciences, State Key
physical and chemical properties of polymer networks
Laboratory of Polymer Physics and Chemistry, Joint Laboratory of
heavily depend on how the polymerization reaction is
Polymer Science and Materials, Institute of Chemistry, Chinese
Academy of Sciences, Beijing 100080, P. R. China
carried out, many researchers are trying to establish the
E-mail: [email protected] relations to predict their microscopic structure with the
W. Yang, D. Wei given synthesis conditions. The first gelation theory
Graduate University of Chinese Academy of Sciences, Beijing discussing the critical condition for the formation of
100049, P. R. China infinite networks was propounded by Flory,[4] who put

Macromol. Theory Simul. 2007, 16, 548–556

548 ß 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim DOI: 10.1002/mats.200700011
Molecular Dynamics Simulation of the Formation of Polymer Networks

forward the most important molecular parameter known Trautenberg et al.[18,19] simulated the formation and
as the gel point. Later, he and Stockmayer[5,6] calculated the swelling of end-linked polymer networks with strand
distribution of molecular weight during a nonlinear lengths varying from 11 to 83 segments using the
polymerization. Although their theories provided a basic bond fluctuation model in an athermal lattice. They found
mechanism for the generation of ideal polymer networks, that the soluble fraction and network imperfections
they did not quantitatively agree with experimental only depend on the precursor polymer chain length.
results[7,8] at the gelation threshold because these theories Meanwhile, they discovered that the kinetics of cross-
are based on the hypothesis of equal and independent linking exhibited a crossover from reaction-controlled
reactivity of all functional groups and no loop formation. to diffusion-limited behavior. Other researchers[18,20,21]
Theoretical approaches to chemical gels involving the also employed the bond fluctuation model to study
percolation theory,[9] taking into account the cyclizations the evolution of the network as well as determining the
and branches, described the gelation threshold with more optimum synthetic conditions that minimized the amount
accurate estimates of the critical exponents. However, of soluble fraction and network defects by varying the
such models do not consider realistic kinetics of reaction initial parameters for the end-linked network formation.
and the multiplicity of bonds formation. Therefore, Applying the discontinuous molecular dynamics tech-
understanding the evolution of such complex networks nique to the hard-chain model, Kenkare et al.[22] obtained a
from a real polymerization process is one of the most network crosslinked from a melt of linear chains without
difficult problems in both the experimental and theoretical using explicit crosslinker molecules. However, their work
communities. was mainly focused on the structure and relaxation of
With the development of computer science, computer end-linked trifunctional networks. To investigate the
simulations have provided an important complement to formation process and the structure of continuum-space
experimental and theoretical analyses. Since simulations polymeric networks, a few studies based on coarse-grained
can exert better control over the crosslinking process than models and atomistic models performed by Grest
experiments, we can easily quantify the effects of various et al.[23–26] have resorted to the molecular dynamic (MD)
macroscopic parameters, such as the strand length, extent method. Comparing this with the Monte Carlo method, the
of reaction and crosslinking density on the microscopic MD method could provide direct information on the
structure, and obtain information about the dynamics by kinetics of crosslinking process because of the inclusion of
choosing proper molecular models. Early contributions to the actual internal dynamic evolution of reactants. They
the work on the simulation of network formation was prepared the network by carrying out the end-linking
made by Eichinger et al.[10–13] They employed a static reaction in an equilibrated melt of the linear polymer, the
crosslinking method to obtain the structural properties of ends of which were attached with a fraction of polyfunc-
trifunctional and tetrafunctional end-linked networks and tional crosslinkers in advance, and analyzed the network
evaluated the defects of the networks. However, the structure in detail. However, they did not take into account
crosslinking reactions were not been conducted by a the influence of initial system density and the motion of
molecular dynamics simulation method, but instead by crosslinkers which have many effects on the kinetics of
resorting to changing a capture radius for static molecules crosslinking and network structure.
to crosslink. Utilizing the method introduced by Leung and At a constant extent of crosslinking, the mechanical
Eichinger,[10,11] Mark illustrated the relationship between properties of polymer networks such as modulus are
the structures and properties of elastomeric polymer dependent not only on the concentration of use but also on
networks in his review.[14] Investigation of the formation the preparation state concentration.[27] However, few
of poly(dimethylsiloxane) gels and the comparison of the works have focused on the influence of the preparation
experimental data with the simulation results was state concentration on the network formation. In this
performed by Braun et al.[15] Their simulations gave a work, we systemically study the kinetics of end-linked
very good account of the extent of reaction at the gel point, network formation with a range of lengths of precursor
but overestimated the maximum extent attainable; this polymer chains, ratios of crosslinkers to polymer chains
may have been due to the simulation method ignoring the ends and number densities for all systems. The effects of
dynamic properties of the reaction system. In order to intramolecular reactions on gelation threshold and the
study the dynamics of cross-linking, several simulations of imperfects of resulting network structure for different
gelation have been performed based on Monte Carlo preparation state conditions are discussed in detail.
algorithms.[16–21] Considering the mobility of monomers The rest of the manuscript is structured as follows:
and microgels on a simple cubic lattice, some research- firstly, we introduce the simulation model and algorithms
ers[16,17] focused on the reaction kinetics for both for the polymer network formation; secondly, the effects of
irreversible and reversible gelation processes. These the synthesis condition on the gelation threshold as well as
studies are, however, restricted to small reactant sizes. the kinetic of cross-linking process are discussed and we

Macromol. Theory Simul. 2007, 16, 548–556

ß 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.mts-journal.de 549
 W. Yang, D. Wei, X. Jin, Q. Liao

evaluate the structural properties of resulting networks;

and, finally, our key conclusions are presented.

Simulation Methodology

Using the MDs method, polymer networks are generated

by carrying out the copolymerizing reactions in the
solution of difunctional prepolymers having N0 Lennard-
Jones (LJ) monomers of mass m0 by means of crosslinking
agents with functionality f. The precursor chain, modeled
as a free jointed bead spring chain, and the crosslinker,
represented by a single LJ particle, are randomly dis-
tributed in a three-dimensional periodic cubic simulation
box. The functionality introduced by Carothers[28,29]
indicates how many particles can be bound by chemical
bonds. For example, a crosslinker molecule with a
functionality of 3 means that such a molecule can react
with at maximum three chain ends. We employ a shifted
LJ potential to describe the pure repulsive excluded volume
interactions between any pair of particles for the condition
of good solvent,
Figure 1. Snapshot of the end-linked polymer network in the
8 
  system with N0 ¼ 10, f ¼ 0.1 and r ¼ 1.0 after the gel point. The
< 12
6
S 4"LJ rs0
rs0 þ "LJ r0  21=6 s sol molecules are shown by the gray color and the gel molecule is
ULJ ðr0 Þ ¼ (1) by the black color.
:
0 r0 > 21=6 s

where the parameter eLJ is the interaction energy which chain ends, defined as the ratio of the number of newly
controls the strength of the short-range interactions, r0 is formed bonds to the total number of bonds formed
the distance between any pair of particles and s is the LJ theoretically, is calculated before the next reaction
unit of length. Additionally, monomers interact with their proceeds. Simulation of end-linking polymerization is
chemical neighbors along the chain by introducing the stopped at a certain time that depends on the number
finite extension nonlinear elastic (FENE) potential density of each system. All of the processes are carried out
  with periodic boundary condition using the Langevin
1 r2 thermostat at constant temperature T ¼ 1.0e/kB, where
UFENE ðr0 Þ ¼
kR20 ln 1
02 (2)
2 R0 kB is the Boltzmann factor, using the DL_POLY Version
2.12 software package.[30] A velocity Verlet algorithm is
where k ¼ 7.0e/s2 is the spring constant, e is the LJ unit of used to integrate the equations of motion with a time step
energy and R0 ¼ 2.0s is the maximum bond length at equal to Dt ¼ 0:005t LJ .
which the elastic energy of the bond becomes infinite. In the present study, the number density f of the initial
We perform the end-linking polymerization according system is defined as:
to the following procedure. Checking
pffiffiffiffiffiffiffiffiffiffiffi
ffi the whole system for
every 5tLJ, where t LJ ¼ s m0 =" is the standard time unit MN0 þ C
f¼ (3)
for LJ fluid, if the distance between a crosslinker which is V
not fully reacted and an unreacted chain end is less than
the reaction radius d ¼ 1.8s, which is smaller than the where N0 is the number of monomers per precursor chain,
maximum bond length R0 ¼ 2.0s, the reaction takes place M is the number of precursor chains, C is the number of
and a new bond is formed. The FENE potential between crosslinkers and V is the volume of the simulation box.
this pair of monomers is added and maintained in the For the trifunctional end-linked system, the ratio r of
following runs. A snapshot of the end-linked network is the number of crosslinkers available for the reaction to the
shown in Figure 1. During the end-linking process, number of precursor polymer chains ends is defined as
information about newly formed bonds and the molecular r ¼ 3C/2M. All systems with different lengths of the
weight of every cluster formed in the system are recorded. precursor chains are generated with number density
The reaction extent, p, which is the fraction of reacted varying from 0.1 to 0.4. For all the chain lengths and

Macromol. Theory Simul. 2007, 16, 548–556

550 ß 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim DOI: 10.1002/mats.200700011
Molecular Dynamics Simulation of the Formation of Polymer Networks

f values, simulations are performed at various values of Table 1. Gel points for different N0 with M ¼ 3 000 and C ¼ 2 000
at f ¼ 0.3.
r: 0.9 corresponding to the lack of crosslinkers; 1.0 for
stoichiometric mixture composition; and, 1.1 correspond-
ing to an excess of crosslinkers. The monodisperse chains N0 pc ac a)
are constructed by a sequence of self-avoiding random
10 0.770 0.593
walks and the crosslinkers then are distributed randomly
in the simulation box. Subsequently, the process of 15 0.766 0.587
polymerization begins, with all the initial conformations 20 0.767 0.588
are relaxed. 30 0.755 0.570

a) 
ac is calculated from Equation (4).
Results and Discussion
According to the theory proposed by Flory,[4] for the
Determination of the Gel Point
polymerization of a bifunctional unit A2 and a trifunc-
To estimate the gel point via the computer simulations, tional unit B3 with equal reactivity of A and B groups, the
general methods have been investigated by Shy et al.[12] branching coefficient a is calculated by
who made use of the weight-average degree of polymer-
ization (DPw) and the maximum reduced average cluster a ¼ p2B =r ¼ rp2A (4)
size (RDPw)[31] to locate the upper and lower bound of the
gel point. Whenever there are two or more large particles where r is the functionality ratio of A to B groups initially
formed in the system, RDPw reaches a maximum value present in the system, and pA and pB are the fractions of
during the cross-linking process, which gives a lower the A and B groups that have reacted, respectively.
bound to the true gel point. When these particles are Therefore, the gel point ac is the critical value of a at
incorporated into a single large particle, there is a rapid which the formation of an infinite network becomes
increase in the slope of the plot of the weight-average possible. The tree-like model[4] of branching theory gives a
degree of polymerization DPw with respect to the reaction general expression to the critical condition of gelation:
extent p, and the maximum of the slope is an upper bound.
The average between the upper and lower bounds gives 1
ac ¼ (5)
the critical reaction extent pc at the gel point. Using this f
1
upper-lower bound method to estimate the gel point, they
obtained results with very good agreement between where f is the functionality of the branching unit. In such a
simulations[32] and measurements.[33] Figure 2 shows system of our investigation, the gel point ac is equal to
the reaction extent dependence of DPw and RDPw for 0.5 because f ¼ 3. However, the results obtained above are
f ¼ 0.3, r ¼ 1 and N0 ¼ 20. The results of different simula- based on the assumption, introduced as an approximation,
tion systems for the critical extent of reaction pc using the that no intramolecular reactions occur. Actually, as will be
method described above are shown in Table 1 and 2. shown later, the intramolecular reactions are inevitable
and influence the reaction extent at the gel point during
the polyfunctional polymerization. In our simulations, the
intramolecular reaction means that a new bond is formed
while the number of molecules in the system is
unchanged. We use a parameter m to denote the fraction

Table 2. Gel points for systems of r ¼ 1, N0 ¼ 30, M ¼ 3 000 and

C ¼ 2 000 at different number densities.

Number Density f pc ac a)

0.1 0.784 0.615

0.2 0.770 0.593
0.3 0.755 0.570
Figure 2. DPw and RDPw distribution for the trifunctional system 0.4 0.756 0.572
at f ¼ 0.3, N0 ¼ 20 and r ¼ 1.0. The solid lines are for visual
a) 
reference. ac is calculated from Equation 4.

Macromol. Theory Simul. 2007, 16, 548–556

ß 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.mts-journal.de 551
 W. Yang, D. Wei, X. Jin, Q. Liao

of intramolecular reactions, which is defined as the ratio of extent of loop formation. Therefore, with the increase in
the number of newly formed bonds assigned to the the probability of the intramolecular reactions the speed
intramolecular reaction to the total number of newly of molecule growth will decrease, which leads to a shift
formed bonds during the crosslinking process. of the critical gel point from short precursor polymer chain
As the length of the precursor polymer chain varies from to the long one. As Table 2 shows, for the same molecular
10 to 30, the critical value ac obtained in the system with weight, ac increases as the reaction system becomes more
number density of 0.3 shifts from 0.593 to 0.570, as shown dilute. In Figure 3b, it is obvious that the intramolecular
in Table 1. The difference between ac obtained from the reactions become much more significant when the initial
simulation and the theoretical value results from the number density of the system is lower. Therefore, chains
ignorance of intramolecular reactions. Actually, both in have a higher probability of forming loops with increasing
experiments and simulations, the intramolecular reactions dilution. Similar results have also been reported in Shy’s
are inevitable. Figure 3a shows that the fraction of work.[12] Meanwhile, Figure 3a and b also indicates that
intramolecular reactions m increases as the reactions the precursor chain length and initial system number
evolve, and increases suddenly with a similar linear trend density have no influence on the reaction extent
just beyond the critical point for various lengths of dependence of intramolecular reaction fraction after the
prepolymer chains at r ¼ 1 and f ¼ 0.3. The plot also network formed.
indicates that for lower molecular weight there is a greater In order to study the kinetics of the end-linking process,
we plotted Figure 4 to show the fraction of unreacted chain
N
ends n calculated by n ¼ 2Mf , where Nf is the free end
number, as a function of time for the studied parameters
(N0, r, f). Firstly, considering the stoichiometric case shown
in Figure 4a, the time dependence of the fraction of
unreacted chain ends at r ¼ 1.0 and f ¼ 0.3 decays as a
power law in time t
n with n  3/4 for three precursor
chain lengths. This end-linking process is similar to the
problem studied by Redner et al. and Toussaint et al.[34–37]
to investigate the time evolution of the irreversible dif-
fusive recombination process in which two kinds of free
particle combine and annihilate or form an inert species.
They found that the decay law predicted from the
rate-equation approach to describe the kinetics of the
irreversible diffusive recombination process has to be
altered due to the spatial fluctuation of the concentration
of the reactants arising from the random distribution of
the particles at the initial state in the low spatial
dimension dc  4. In a region of linear dimension j, the
spatial fluctuations of the reactant are of the order of
rð0Þjdc  ½rð0Þjdc 1=2 , where r(0) is the initial density of the
reactant. After a time tj  j2, the particles will diffuse
across a domain of size j and have a chance to interact.
Therefore, for short times, t < tj, local fluctuations in the
density difference will determine the decay rate and the
particle density is predicted to decay as t
n with n ¼ dc/
4.[36] For equal initial densities, this power-law decay
should hold for all times. For unequal initial densities, the
decay rate at long times is governed by the spatial
fluctuations of the majority species, rather than density-
difference fluctuations, and can be expressed as a quasi-
exponential decay. However, the scaling exponent we
obtained is different from the value of 0.5 gained by Grest
et al., which described the reaction kinetics of the network
Figure 3. Dependence of the fraction of intramolecular reaction m
on the reaction extent at r ¼ 1.0 for: (a) different N0 at f ¼ 0.3, and formation in the melts.[23–26] Since the relaxation time of
(b) different f with N0 ¼ 20. The solid lines are for visual refer- the polymer chain in the dilute solution is much shorter
ence. than that in the melts, the motion of a monomer on a chain

Macromol. Theory Simul. 2007, 16, 548–556

552 ß 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim DOI: 10.1002/mats.200700011
Molecular Dynamics Simulation of the Formation of Polymer Networks

is diffusive (t  j2) for early time in our polymerization

process while it is Rouse-like (t  j4) in the melts. Therefore,
due to the higher mobility of the crosslinkers and chains in
the solution after the gel point, the effective speed of the
intermolecular reaction in the solution is faster than in the
melt. In Figure 4b, the results for the fraction of unreacted
chain ends are compared for different values of r. For
r < 1.0, since there are more free chain ends than available
crosslinkers in the final state, the surplus free ends re-
mained have been subtracted in the calculation. The plot
shows that the decay of unreacted chain ends falls on a
common curve which follows the power law of t
3/4 for
early time while deviate at larger time. This is due to the
more complex process at later time for non-stoichmetric
case. For r < 1.0, except for a few free chain ends, the
system contains some dangling ends attached to the
infinite network or on free chain. For r > 1.0, the final state
contains some unsaturated crosslinkers and some dan-
gling ends. As a result, the decay of unreacted chain ends
for r 6¼ 1.0 is faster than a power law for long time and can
be described approximately by an exponential decay
which is governed by the spatial fluctuations of the majo-
rity species. From Figure 4c, we can easily deduce that in all
systems considered the crosslinking process becomes
slower as their number density decreases before the gel
point, because there are fewer unreacted chain ends
around the unsaturated crosslinkers in the reaction region.
Moreover, one can say that the influence of f on the speed
of the cross-linking process is more considerable, com-
pared to the effect of the growth of the precursor chain
length. This also agrees with the results from the shift of
the gel point with the change of the system number
density. To sum up this section, the main parameters that
influence the speed of the molecules growth turn out to be
r and f, especially when the length of prepolymer chain is
not too long.

Network Structure
As observed from the above studies, the synthesis
conditions significantly affect the network formation. It
is necessary to evaluate the structure properties of the
resulting networks for different synthesis conditions.
Close to the gel point, since the system consists of a
highly polydisperse distribution of polymers, one of the
most important features of gelation is the dependence
of number density of the polymers on the degree of
polymerization. Both the mean-field theory and scaling
Figure 4. (a) The fraction of unreacted chain ends vs. tLJ for model indicate that the polymer number density decays as
different N0 at f ¼ 0.3 and r ¼ 1.0. (b) The fraction of unreacted
a power of the number of monomers in a polymer,
chain ends which is minus the excess chain ends vs. tLJ for the
30-mer networks with different ratios r at f ¼ 0.3. (c) The fraction expressed as:[27]
of unreacted chain ends vs. tLJ for the 30-mer networks with
different number densities at r ¼ 1.0.
nðp; NÞ  N
t f ðN=N  Þ (6)

Macromol. Theory Simul. 2007, 16, 548–556

ß 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.mts-journal.de 553
 W. Yang, D. Wei, X. Jin, Q. Liao

where n( p,N) is defined as the number of N-mers divided

by the total number of monomers, and f ðN=N  Þ is the
cutoff function truncating at the characteristic degree of
polymerization N  for branched polymers with different
forms for the two theories. However, the exponent
t describing the mean-field gelation is 5/2 while in
three-dimensional percolation the critical exponent is
t  2.18. Figure 5 shows the number density distribution
function nðp; NÞ at different reaction extents below the gel
point ( pc ¼ 0.71) for the systems at different preparation
states. From simulation results, we obtain the exponent
corresponding to critical percolation in three dimensions
since the number of monomers between the branch points
is not too large and the chains are not strongly overlapped.
Moreover, the initial system number density has no
influence on the distribution of the molecular weight near
Figure 6. The dependence of gel fraction on the number density f
the gel point.
at different r values for two precursor polymer chain lengths N0.
Owing to the dependence of gelation on the number The solid lines are for visual reference.
density, simulation runs at different system number
densities are ceased at different reaction time; for example, the lack of crosslinkers such as in the case of r ¼ 0.9, the size
the reaction time is 600tLJ for the simulation runs at f ¼ 0.1 of the ultimate network prominently decreases irrespec-
to reach the final reaction extent. With the increase of the tive of what values f and N0 take. Furthermore, for the
system number density, the reaction time will continually number density f > 0.1, the size of the gel becomes
decrease. approximately the same even with different values of r
In order to analyze the structures of the resulting and N0, except for the case of the lack of crosslinkers. This
network, we first determine the gel fraction of the reacting implies that the influence of r and N0 on the size of the
system that is defined as ultimate network becomes weaker with the number
density of the system increasing.
mN0 þ c Based on the connectivity of the particles recorded in
wgel ¼ (7)
MN0 þ C simulations during the polymerization process, we can
determine the fraction of pendent materials and the
where m is the number of strand chains in the network amount of loops in the reaction systems. Here the fraction
and c is the number of crosslinkers in the network. Figure 6 of the elastically effective materials wel is determined as
represents the gel fraction for the considered variants of r,
f and N0 in the crosslinking process. It reveals that, due to wel ¼ wgel
wpend
wloop (8)

where wpend is the fraction of pendent material consisting

of long connected pendent chains and branched pendent
structures, and wloop is the fraction of single-chain loops
calculated by determining the number of chains connected
at both ends with the same crosslinkers. From Figure 7a,
summarizing the fraction of elastic materials, single-chain
loops and pendent materials in the resulting system as a
function of number density f at different r values, one can
see that the lack of crosslinkers (r ¼ 0.9) makes the network
have more dangling chains, longer pendent structures, less
elastic materials. As the reacting systems become much
denser, the fraction of single-chain loops decreases, while
the fraction of elastic materials and pendent materials
does not change so much, especially for the case of r ¼ 1.0
and 1.1. It implies that the denser the reacting system and
Figure 5. Number density distribution function nðp; NÞ at differ- the more crosslinkers present, the fewer defects the
ent reaction extents p below the gel point ( pc ¼ 0.71) for f ¼ 0.2, ultimate network structure has. Figure 7b, which com-
N0 ¼ 30 and f ¼ 0.3, N0 ¼ 20, 30 at r ¼ 1.0. pares the fraction of elastic materials, single-chain loops

Macromol. Theory Simul. 2007, 16, 548–556

554 ß 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim DOI: 10.1002/mats.200700011
Molecular Dynamics Simulation of the Formation of Polymer Networks

varying from 10 to 30 at different number densities of the

reacting system in a three-dimensional periodic cubic box.
Our simulations showed that with a stoichiometric
number of crosslinkers present, the time dependence of
free ends fraction decays as a power law of t
3/4. This is in
an excellent agreement with the scaling prediction for
the irreversible diffusive recombination process. How-
ever, the scaling parameter is different from the one
shown in the simulations of crosslinking in the polymer
melts. This discrepancy is attributed to faster relaxation in
the solution than in the melts. We also discovered that
the ratio of crosslinkers to prepolymer chain ends and the
initial system number density significantly influenced the
polymer network formation process and the resulting
network structure. Our simulation results also indicated
that the abundance of intramolecular connections in a real
network is an essential structural feature. Because of the
significant dependence of network formation on the
preparation state, work on the effects of such conditions
on the mechanical properties is ongoing, which will help
us greatly in making a wider application of polymer gels in
the future.

Acknowledgements: This work was financially supported by

NSFC grant nos. 20474075, and 20674091, Innovation Funding
grant no. KJCX2-SW-H07 from the Chinese Academy of Sciences,
and 973 grant no. 2003CB615604 from the Chinese Ministry of
Science and Technology.

Received: January 29, 2007; Revised: March 24, 2007; Accepted:

March 26, 2007; DOI: 10.1002/mats.200700011

Keywords: end-linked polymer networks; molecular dynamics;

network structure; reaction kinetics
Figure 7. (a) The fraction of single-chain loops wloop, pendent
materials wpend and elastic materials wel for the system of the
precursor polymer chain length N0 ¼ 10 as a function of number
density f for different r values. (b) The fraction of single-chain
loops wloop, pendent materials wpend and elastic materials wel as a [1] R. Dagani, Chem. Eng. News 1997, 75, 26.
function of number density f at r ¼ 1.0 for the systems of two [2] L. R. G. Treloar, ‘‘The physics of rubber elasticity’’, 3rd Ed.,
precursor polymer chain length N0 ¼ 10 and 20. The solid lines are Clarendon Press, Oxford 1975.
for visual reference. [3] A. O’Halloran, F. O’Malley, ‘‘Topics in Bio-Mechanical Engin-
eering’’, P. J. Prendergast, P. E. McHugh, Eds., Trinity Centre
and pendent materials of the system as a function of for Bioengineering and the National Centre for Biomedical
Engineering Science, Dublin & Galway 2004.
number density f at r ¼ 1.0 for two precursor polymer [4] P. J. Flory, ‘‘Principles of Polymer Chemistry’’, Cornell Univer-
chain lengths N0, indicates that the network structure sity Press, Ithaca, NY 1953.
weakly depends on the precursor polymer chain length. [5] W. H. Stockmayer, J. Chem. Phys. 1943, 11, 45.
This observation is consistent with the conclusion we have [6] W. H. Stockmayer, J. Chem. Phys. 1943, 12, 125.
made from part a. [7] E. M. Valles, C. W. Macosko, Macromolecules 1979, 12, 521.
[8] D. Durand, F. Naveau, J. P. Busnel, M. Delsanti, M. Adam,
Macromolecules 1989, 22, 2011.
[9] P. G. de Gennes, ‘‘Scaling Concepts in Polymer Physics’’, Cornell
Conclusion University Press, Ithaca and London 1979.
[10] L. Yu-Kwan, B. E. Eichinger, J. Chem. Phys. 1984, 80, 3885.
[11] L. Yu-Kwan, B. E. Eichinger, J. Chem. Phys. 1984, 80, 3877.
In this work, we have employed the MD method to
[12] L. Y. Shy, Y. K. Leung, B. E. Eichinger, Macromolecules 1985,
investigate systematically the formation of both stoichio- 18, 983.
metric and nonstoichiometric end-linked trifunctional [13] K.-J. Lee, B. E. Eichnger, Polymer 1990, 31, 406.
networks of precursor polymer chains with the length [14] J. E. Mark, Macromol. Symp. 2001, 171, 1.

Macromol. Theory Simul. 2007, 16, 548–556

ß 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.mts-journal.de 555
 W. Yang, D. Wei, X. Jin, Q. Liao

[15] J. L. Braun, J. E. Mark, B. E. Eichinger, Macromolecules 2002, [25] E. R. Duering, K. Kremer, G. S. Grest, Prog. Colloid Polym. Sci.
35, 5273. 1992, 90, 13.
[16] Y. Liu, R. B. Pandey, J. Phys. II 1994, 4, 865. [26] D. R. Heine, G. S. Grest, C. D. Lorenz, M. Tsige, M. J. Stevens,
[17] Y. Liu, R. B. Pandey, Phys. Rev. B: Condens. Matter Mater Phys. Macromolecules 2004, 37, 3857.
1997, 55, 8257. [27] M. Rubinstein, R. H. Colby, ‘‘Polymer Physics’’, Oxford Uni-
[18] H. L. Trautenberg, J. U. Sommer, D. Goritz, Macromol. Symp. versity Press, New York 2003.
1994, 81, 153. [28] W. H. Carothers, Chem. Rev. 1931, 8, 353.
[19] H. L. Trautenberg, J. U. Sommer, D. Goritz, J. Chem. Soc., [29] W. H. Carothers, Trans. Faraday Soc. 1936, 32, 39.
Faraday Trans. 1995, 91, 2649. [30] T. R. Forester, W. Smith, ‘‘The DL_POLY_2 Reference Manual’’,
[20] N. Gilra, C. Cohen, A. Z. Panagiotopoulos, J. Chem. Phys. 2000, Daresbury Laboratory, Daresbury 2000.
112, 6910. [31] J. Hoshen, R. Kopelman, Phys. Rev. B: Condens. Matter Mater
[21] A. G. Balabanyan, E. Y. Kramarenko, I. A. Ronova, A. R. Phys. 1976, 14, 3438.
Khokhlov, Polymer 2005, 46, 4248. [32] L. Y. Shy, B. E. Eichinger, Br. Polym. J. 1985, 17, 200.
[22] N. R. Kenkare, S. W. Smith, C. K. Hall, S. A. Khan, Macro- [33] J. L. Stanford, R. F. T. Stepto, Br. Polym. J. 1977, 9, 124.
molecules 1998, 31, 5861. [34] I. Ispolatov, P. L. Krapivsky, S. Redner, Phys. Rev. E: Stat.
[23] E. R. Duering, K. Kremer, G. S. Grest, J. Chem. Phys. 1994, 101, Nonlin. Soft Matter Phys. 1995, 52, 2540.
8169. [35] D. Toussaint, F. Wilczek, J. Chem. Phys. 1983, 78, 2642.
[24] G. S. Grest, K. Kremer, E. R. Duering, Europhys. Lett. 1992, 19, [36] K. Kang, S. Redner, Phys. Rev. Lett. 1984, 52, 955.
195. [37] K. Kang, S. Redner, Phys. Rev. [Sect.] A 1985, 32, 435.

Macromol. Theory Simul. 2007, 16, 548–556

556 ß 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim DOI: 10.1002/mats.200700011