Molecular dynamics simulations of

concentrated polymer solutions in thin film

geometry. I. Equilibrium properties near the
glass transition
Cite as: J. Chem. Phys. 131, 014902 (2009); https://doi.org/10.1063/1.3158608
Submitted: 13 January 2009 . Accepted: 30 May 2009 . Published Online: 01 July 2009

S. Peter, H. Meyer, and J. Baschnagel

ARTICLES YOU MAY BE INTERESTED IN

Molecular dynamics simulations of concentrated polymer solutions in thin film geometry. II.
Solvent evaporation near the glass transition
The Journal of Chemical Physics 131, 014903 (2009); https://doi.org/10.1063/1.3158607

Dynamics of entangled linear polymer melts:  A molecular-dynamics simulation

The Journal of Chemical Physics 92, 5057 (1990); https://doi.org/10.1063/1.458541

Molecular dynamics simulation of a polymer chain in solution

The Journal of Chemical Physics 99, 6983 (1993); https://doi.org/10.1063/1.465445

J. Chem. Phys. 131, 014902 (2009); https://doi.org/10.1063/1.3158608 131, 014902

© 2009 American Institute of Physics.

 THE JOURNAL OF CHEMICAL PHYSICS 131, 014902 共2009兲

Molecular dynamics simulations of concentrated polymer solutions in thin

film geometry. I. Equilibrium properties near the glass transition
S. Peter, H. Meyer, and J. Baschnagela兲
Institut Charles Sadron, 23 Rue du Loess, BP 84047, 67034 Strasbourg Cedex 2, France
共Received 13 January 2009; accepted 30 May 2009; published online 1 July 2009兲

We report on results of molecular dynamics simulations for supported polymer films with explicit
solvent. The simulation represents the polymers by bead-spring chains and the solvent particles by
monomers. The interaction between polymer and solvent favors mixing. We find that the solvent
acts as a plasticizer. The glass transition temperature Tg is reduced relative to the pure polymer film.
Near Tg we explore equilibrium properties as a function of temperature and solvent concentration.
We find that the structure and dynamics of the films are spatially heterogeneous. The solvent density
is enriched at the supporting wall and at the free surface where the film is in equilibrium with
solvent vapor. At both interfaces the solvent dynamics is fast, but smoothly crosses over to bulk
dynamics when moving from the interfaces toward the center of the film. A smooth gradient from
enhanced dynamics at the interfaces to bulk behavior in the film center is also found for the
monomers. We show that the same formula used to parametrize the spatial gradient of the dynamics
in the pure polymer film may also be applied here. Furthermore, we determine the concentration
dependence of the relaxation time of the solvent in the center of film and compare this dependence
to models proposed in literature. © 2009 American Institute of Physics. 关DOI: 10.1063/1.3158608兴

I. INTRODUCTION evidence from experiment8–10 and theory13 that they control

Spin coating is a powerful technique that allows one to
produce polymer films of nanoscopic dimensions.1,2 In the
spin-coating process, a drop of a dilute polymer solution is
deposited onto a substrate and spun at high angular velocity
to spread the solution and to evaporate the solvent. As the
solvent evaporates, the glass transition temperature 共Tg兲 of
the solution increases until it reaches the ambient tempera-
ture at which the film vitrifies. At this point, however, there
is still a considerable amount of solvent left in the film, typi-
cally 10%–20%.3–5 In the glassy state, solvent evaporation
continues, but at a much slower rate6 because further loss of
solvent requires the glassy matrix to contract, and this costs a
high energy.3,6,7 To remove residual solvent and to allow the
chains to relax annealing of the film at temperatures above
the Tg of the bulk polymer melt is usually performed after
spin coating. The duration for this treatment is, however,
limited because the film tends to dewet. Therefore, it is un-
clear to what extent annealing is able to eliminate the non-
equilibrium states created by the film preparation method.3,4
One viewpoint is that spin-coated polymer films “repre-
sent highly metastable forms of matter”7 with large8 “re-
sidual stresses”7–10 that relax very slowly with time.9 The
molecular origin of these stresses is not well understood, but
it has been suggested that a contributing factor are out-of-
equilibrium chain conformations which are quenched by the
solvent evaporation in late stages of the spin-coating
process.7,9 While it may be argued11 that these stresses are
unlikely to be responsible for the shifts of Tg 共Refs. 3, 4, and
12兲 of thin polymer films relative to the bulk value, there is
a兲
Author to whom correspondence should be addressed. Electronic mail:
baschnag@ics.u-strasbg.fr.
the propensity of the film for dewetting and so its stability.
Apparently, the residual stresses correlate differently to local
properties 共such as Tg兲 than to global properties 共such as
dewetting兲 of a film. The reason for this difference is not
understood. This motivates the need for further investiga-
tions attempting to elucidate how the solvent evaporation
process affects the properties of polymer films.
Computer simulations can contribute to this research. Al-
though a simulation of the entire spin-coating process rang-
ing from the spreading of a dilute polymer solution to the
final dry film is currently beyond computational possibilities,
an aspect of this process, the solvent evaporation from qui-
escent polymer solutions, can be studied. Two such studies
have recently been carried out.14,15 Tsige and Grest14 per-
formed molecular dynamics 共MD兲 simulations of coarse-
grained models of polymer-solvent mixtures in supported
film geometry. They found that the evaporation process leads
to a sharp increase in the polymer density at the retracting
film-vapor interface. A similar increase was also observed by
Müller and Smith15 in MD and single-chain-in-mean-field
simulations of solvent evaporation from freestanding poly-
mer films, but only for fast evaporation rates. Slowing down
the mass transport in the vapor phase suppresses the enrich-
ment of polymer at the surface of the film.15
Both simulation studies were concerned with tempera-
tures 共T兲 far above the bulk Tg. The present work extends the
simulations to lower T. To approach the experimental situa-
tion an ideal starting point would be to simulate solvent
evaporation from concentrated solutions 共solvent content
⬃25%兲 at a temperature below the Tg of the pure polymer
but above the Tg of the solution. Then, it would be possible
to observe the vitrification of the solution while there is still

0021-9606/2009/131共1兲/014902/7/$25.00 131, 014902-1 © 2009 American Institute of Physics

014902-2 Peter, Meyer, and Baschnagel
FIG. 1. Left panel: schematic drawing of the various LJ interactions in a
solvent 共S兲-polymer 共P兲 mixture that is supported by a wall 共W兲. y denotes
the direction perpendicular to the wall. Right panel: schematic drawing of
the film geometry used to simulate a supported polymer solution in equilib-
rium with its solvent vapor. The right wall represent a barrier whose position
is adjusted to fix the vapor pressure 共see Sec. II B for details兲.
some solvent trapped in the film, and the subsequent shrink-
age of the film as the residual solvent evaporates. Such a
simulation could be able to capture the putative trapping of
the chains in out-of-equilibrium states and the attendant cre-
ation of residual stresses. However, we found that it is very
challenging to run long enough simulations to observe com-
plete solvent evaporation at low temperatures. Therefore, we
restrict the present study to temperatures larger than or equal
to the Tg of the pure polymer melt.
Our work is split into two parts, the present and subse-
quent paper, hereafter referred to as parts I and II, respec-
tively. Part I deals with the structure and dynamics of sup-
ported polymer solutions that are in equilibrium with a
solvent vapor phase. It represents an extension of our work
on pure supported and freestanding polymer films.12,16–18
Part II then extends the present work to the study of solvent
evaporation.
Our article is organized as follows. Section II presents
details about the simulation model and simulation technique.
The properties of the supported polymer solution are de-
scribed in Sec. III, where we determine the glass transition
temperature Tg and discuss the relaxation behavior of the
films as a function of the distance from the interfaces. There,
we suggest a parametrization of the spatial dependence of the
local relaxation time. This parametrization will be an impor-
tant ingredient in the analysis of the solvent evaporation in
part II.19 The final section 共Sec. IV兲 summarizes our main
results.
II. SIMULATION WITH EXPLICIT SOLVENT: MODEL
AND TECHNIQUE
A. Simulation model
Our simulation model has been described in Ref. 16 for
pure polymer films and in Ref. 20 for bulk polymer-solvent
mixtures. For conciseness, only its main features are reported
here. The polymers are represented by monodisperse bead-
spring chains with monomers of mass m PP, and the solvent is
modeled as a monomer of mass mSS. Bonded monomers
along the polymer chain are connected by a harmonic poten-
tial. Nonbonded monomer-monomer interactions, monomer-
solvent interactions, and solvent-solvent interactions are
given by a Lennard-Jones 共LJ兲 potential that is truncated at
twice the position of its minimum and shifted to zero there.
The LJ potential introduces various parameters: The
monomer diameter 共␴ PP兲, the solvent diameter 共␴SS兲, and the
J. Chem. Phys. 131, 014902 共2009兲
strengths of attraction between the species 共see Fig. 1兲, that
is, ⑀ PP 共monomer-monomer兲, ⑀ PS 共monomer-solvent兲, and ⑀SS
共solvent-solvent兲. In the following, we will report our data in
LJ units, i.e., we set ⑀ PP = 1, ␴ PP = 1, and m PP = 1. Then, tem-
perature T is measured in units of ⑀ PP / kB 共Boltzmann con-
stant kB = 1兲, and time t in units of ␶LJ = 共m PP␴2PP / ⑀ PP兲1/2.
In these units, we set mSS = 1 and choose the LJ parameters
of the solvent-solvent and solvent-polymer interactions as
␴SS = ␴ PS = 1, ⑀SS = 0.5, and ⑀ PS = 0.8. This choice favors
mixing between polymer and solvent because 共⑀ PP + ⑀SS兲 / 2
− ⑀ PS ⬍ 0.20
With this model we simulate polymer films supported by
a substrate. To this end, we introduce a wall at y = 0 in the
xz-plane of the simulation box and model the monomer-wall
and solvent-wall interactions by a 共nontruncated兲 9-3 LJ po-
tential
冋冉 冊 冉 冊 册
9 3
1 1
A
Uwall 共y兲 = ⑀AW − 共A = P,S兲. 共1兲
y y
Here y is the distance perpendicular to the wall and ⑀AW is
the depth of the attractive minimum of the potential. Follow-
ing previous work16–18 we take a value close to the wetting
transition, ⑀ PW = 3, for the monomer-wall interaction. For the
solvent-wall interaction we choose a weaker attraction, ⑀SW
= ⑀ PW冑⑀ PP⑀SS = 3 / 冑2, to avoid strong segregation of the sol-
vent at the wall.
B. Simulation technique
We examine polymer-solvent mixtures in a supported
film geometry by MD simulations. The equations of motion
are integrated by the velocity-Verlet algorithm21 with a step
size of ⌬t = 0.005. The simulations are performed at constant
temperature 共DPD thermostat22兲. The supported film is in
contact with a solvent vapor at vapor pressure pv. The vapor
pressure is controlled by a barrier wall at distance D from the
supporting wall of the film 共Fig. 1兲. We model this barrier
wall by the repulsive part of the 9-3 LJ potential,12
冉 冊
9
1
Ubarrier wall共y兲 = . 共2兲
D−y
By varying D the solvent concentration in the vapor phase
and along with that pv can be adjusted. This allows us to
study polymer films in equilibrium with solvent vapor at
pressure pv. For the temperatures and solvent concentrations
in which we are interested here—low T close to the glass
transition and concentrated polymer solutions—pv is always
close to 0. For example, the vapor pressure at T = 0.5 in a film
containing 21% of solvent is pv = 0.03. Therefore, we can
compare the simulation results for the films to those of bulk
simulations at pressure p = 0 for the same composition and
temperature.
The simulated systems contain a total of Ntot = N P + NS
particles where NS is the number of solvent particles and
N P = Nn with N as the chain length and n as the number of
chains. We study chains of length N = 64 共some results for
N = 10 will also be presented兲. As the polymers are not vola-
tile, the number of chains in the film is constant 共n = 96 cor-
responding to N P = 6144 monomers兲, whereas the solvent
014902-3 Simulation of polymer solutions in thin films
T=0.5
wall T=0.4 N=10
1.4
1.2 region 1 region 2 region 3
1.0
ρx(y) 0.8 polymer
0.6
0.4 solvent
0.2
0 at the wall. This accumulation occurs because the solvent has
0 5 10 15 20 25 30
y a lower surface tension than the polymer. Both the mono-
FIG. 2. Density profiles ␳x共y兲 of the solvent 共x = S; dashed lines兲 and of the
monomers of the polymers 共x = P; solid lines兲 vs distance y from the wall for
chains of length N = 10 at T = 0.5 and 0.4. Three regions are indicated by
vertical dotted lines: A region encompassing the main part of the film up to
the film-vapor interface 共region 1兲, the film-vapor interface 共region 2兲, and
the vapor phase 共region 3兲. The solvent concentration in the center of film
depends on T, but is about ␾S ⯝ 24%.
particles disperse between the film and vapor phases. Thus,
not the solvent concentration in the film, but only NS is fixed.
We study systems with NS = 768, NS = 1536, NS = 2304, NS
= 3072, and NS = 4608.
To prepare the polymer-solvent mixtures we start from
equilibrated configurations of the pure polymer film at T
= 0.5 and “dissolve” a number of 共randomly chosen兲 chains
by redefining their monomers as solvent particles so as to
obtain the desired solvent concentration. The resulting sys-
tems are then equilibrated at T = 0.5 until the solvent distri-
bution stabilizes. 共We choose this fairly low temperature—Tg
of the pure bulk melt is ⯝0.404; the freezing temperature of
the pure solvent is T ⬇ 0.35—because the film cannot be
equilibrated at higher T without considerable reduction of the
solvent concentration in the film. This is due to the strong
increase of pv for T Ⰷ 0.5.兲 Equilibration takes longer than in
the bulk20 since the film after the dissolution of the chains
contains all solvent particles, some of which have to evapo-
rate into the vacuum phase above the film until equilibrium
is reached. Runs of 100 000␶LJ are necessary to relax the
monomer and solvent densities in the film and achieve equi-
librium with the vapor phase. This time 共100 000␶LJ兲 is also
sufficient for the polymers to diffuse over a distance compa-
rable to their radius of gyration. Starting from the equili-
brated configurations at T = 0.5 the film is cooled to lower
temperatures with a rate of ⌫T = 2 ⫻ 10−5 according to the
cooling schedule T共t兲 = 0.5− ⌫Tt.16 Further runs of
50 000␶LJ – 200 000␶LJ, depending on temperature and sol-
vent concentration, are carried out in the films to re-
equilibrate the density profiles and vapor pressure.
III. RESULTS AND DISCUSSION
The presence of solvent modifies the structure and dy-
namics of the polymer films in comparison to the undiluted
systems studied before.12,16–18 To obtain insight into these
effects we discuss in this section density profiles, the glass
transition temperature, mean-square displacements 共MSDs兲,
and relaxation times from a layer-resolved analysis.
J. Chem. Phys. 131, 014902 共2009兲
A. Monomer and solvent density profiles
Figure 2 shows the density profiles of the monomers
共N = 10兲 and solvent particles at T = 0.5 and 0.4. It is possible
to distinguish three regions, a region extending from the wall
to the beginning of the film-vapor interface 共region 1兲, the
region around the interface 共region 2兲, and the vapor phase
共region 3兲. In region 1 we find that the solvent accumulates
mers and the solvent exhibit oscillating density profiles. The
amplitude of the oscillations decreases with increasing dis-
tance from the wall so that the profiles become constant in
the center of the film. These features of the density profiles
are typical of liquids close to smooth walls23 and are found
in many other studies.24–28
Upon cooling the solvent concentration in the film in-
creases. This has two consequences. First, since the total
number of solvent particles 共NS兲 is constant, the solvent con-
centration in the vapor phase 共region 3兲 decreases. This leads
to a decrease in pv with decreasing T. Second, the higher
solvent concentration in the film reinforces the packing con-
straints. This gives rise to an amplification of the density
oscillations, their decay becoming thus more long ranged on
cooling.
Contrary to the oscillatory behavior at the impenetrable
wall the monomer density profile smoothly vanishes in the
transition zone between the film and vapor phases 共region 2兲,
while the solvent density exhibits a peak there. Segregation
of solvent at the film-vapor interface is not uncommon in
polymer-solvent mixtures. For instance, this phenomenon is
analyzed in detail in Ref. 29 which explains that solvent
enrichment occurs due to interfacial wetting and may be in-
terpreted as a precursor of the phase behavior at the triple
line where a polymer-rich liquid, a solvent-rich liquid, and a
solvent vapor coexist.
Qualitatively identical results are also found for N = 64
共cf. Figure 4兲. This similarity between N = 10 and 64 also
carries over to dynamic features and to the behavior during
the evaporation process 共see Ref. 19兲. Therefore, we focus on
N = 64 in Sec. III B.
B. Glass transition temperature
Reference 20 examines the influence of solvent on Tg of
bulk polymer solutions. Here we want to extend this analysis
to polymer films. Therefore, we employ the same method as
in Ref. 20 to determine Tg. That is, we perform continuous
cooling runs with rate ⌫T = 2 ⫻ 10−5 to monitor the T depen-
dence of the nonbonded LJ energy per monomer,
N N
1 P tot AB
pol
ENB = 兺 兺 U 共兩ri − r j兩兲
N P i=1 j⬎i LJ
共A,B = P,S兲, 共3兲
AB
where N P is the total number of monomers and ULJ denotes
the LJ potential. Similar to the equilibrium case discussed in
Sec. III A, the solvent concentration in the film also in-
creases on cooling in the present nonequilibrium simulations.
However, the uptake of solvent from the vapor phase re-
mains small, increasing the 共number兲 concentration ␾S of
014902-4 Peter, Meyer, and Baschnagel J. Chem. Phys. 131, 014902 共2009兲

-4.5 0 5 10 15 20 25 30
0% ∆Tg(φs) T=0.5 region 1 region 2
10% 1
φs=21%

ρx(y)
15% polymer (N=64) region 3
-5.0 0.5
solvent
ENB
pol
4
10 0
3 region 1
-5.5 10 solvent 4Dt
2 region 2
bulk
Tg 10 region 3
film 1
Tg 10

g0,x(t)
0
-6.0 10 0.5
0.2 0.3 0.4 0.5 -1 ~t
10
T -2 2Tt
2

10 polymer (0%) polymer (21%)

FIG. 3. LJ energy of nonbonded monomers 共ENB pol
兲 vs temperature 共T兲 for
three number fractions of solvent, ␾S = 0%, 10%, and 15%. ENB pol
is the en-
ergy per monomer, and all results refer to N = 64. Filled symbols show the
results for bulk polymer solutions 共Ref. 20兲, open symbols 共of the same
type兲 show the film results for the same ␾S as in the bulk. The film results
are shifted downward to coincide with the bulk energies at high T. This
difference in energy between film and bulk arises due to missing interactions
at the interfaces. The dashed lines for the film at ␾S = 0 indicate the method
used to determine Tg by the intersection of linear extrapolations from the
low-T 共glass兲 and high-T 共liquid兲 branches of ENB pol
. The vertical lines show
the resulting Tg’s. Film: Tg共␾S = 0%兲 ⯝ 0.376, Tg共␾S = 10%兲 ⯝ 0.339, and
Tg共␾S = 15%兲 ⯝ 0.313. Bulk: Tg共␾S = 0%兲 ⯝ 0.404, Tg共␾S = 10%兲 ⯝ 0.378, and
Tg共␾S = 15%兲 ⯝ 0.365.
solvent in the film center by less than 1%. Thus, we only
report values for ␾S that are averaged over all T in the fol-
lowing.
Figure 3 depicts ENB pol
共T兲 for bulk and film systems at
different ␾S. We see that the energy continuously decreases
on cooling. From this decrease we determine Tg by the inter-
section of linear extrapolations from the low-T 共glass兲 and
high-T 共liquid兲 branches of the LJ energy 共dashed lines in
Fig. 3兲. Certainly, this procedure leads to systematic errors in
Tg. These errors can be rather large16 but do not affect the
qualitative dependence of Tg on ␾S provided the same T
intervals on the liquid and glass branches are chosen for all
systems. In Ref. 20 we confirmed by an analysis of equilib-
rium data that the qualitative trend obtained from the ex-
trapolation procedure is reliable. Here we find the following
results from Fig. 3. As in the bulk, the solvent acts as a
plasticizer; it reduces Tg. For all ␾S the glass transition in the
films is depressed relative to the bulk. This depression has
also been found in previous work on undiluted polymer
films16–18 where the reduction in Tg was related to an inter-
face effect. The same mechanism also operates here; we
elaborate this point by a layer-resolved analysis of the film
-3
10
-4
10 -2 -1 0 1 2 3 4
10 10 10 10 10 10 10
t
FIG. 4. Log-log plot of the MSD of the solvent 共symbols兲 parallel to the
wall in three different regions of a film. The regions are indicated in the
inset which shows the density profiles of the monomers 共x = P兲 and solvent
particles 共x = S兲. The vertical dashed lines represent the boundaries of the
regions for which g0,S共t兲 in the main figure is calculated. All data refer to a
polymer film at T = 0.5 containing polymers of length N = 64 and a solvent
content of ␾S = 21%. For comparison also the monomer MSDs, g0,P共t兲, for
␾S = 0% 共dotted line兲 and ␾S = 21% 共dashed line兲 are shown. At late times the
monomer MSD displays a power law 共⬃t1/2兲 due to chain connectivity. In
region 1 the solid line through the solvent data represents a fit to the late
time behavior of g0,S共t兲, where g0,S共t兲 ⬃ t, to determine the diffusion coeffi-
cient DS 共=0.006兲. The solid lines through the solvent data in regions 2 and
3 are fits to Eq. 共5兲. These fits yield DS = 0.097 in region 2 and DS = 1.457 in
region 3 共with DS = T / ␥ and the values for ␥ given in the text兲. All data
display ballistic motion, g0,x共t兲 = 2Tt2, at early times.
Figure 4 depicts these MSDs for N = 64 at T = 0.5 in the
regions identified before in Fig. 2. In region 1 the solvent
data 共crosses兲 are compared to the monomer MSD of the
diluted 共␾S = 21%; dashed line兲 and pure 共␾S = 0%; dotted
line兲 polymer films. The MSD of the pure film has the fol-
lowing features. After a steep initial rise due to ballistic mo-
tion 共g0,P共t兲 = 2Tt2兲 the increase in the MSD strongly slows
down if g0,P共t兲 is of the order of 10% of the monomer diam-
eter 共g0,P共t兲 ⬃ 0.04兲. Upon cooling this slowing down will
become more pronounced and g0,P共t兲 almost plateaus 共cf.
Fig. 5 and Refs. 12, 16–18, and 20兲. This intermittency of
large-scale motion reflects the temporary caging of the
2
10 T=0.4 bulk
solvent
1 φs=25%
10
g0,x(t,y)

0
10 1 bulk
dynamics in Sec. III D. polymer
-1
10
-2
center
10 wall
C. Mean-square displacements of polymer free surface
and solvent -3
10
-1 0 1 2 3 4 5
10 10 10 10 10 10 10
The plasticization of the polymer by the solvent is also t
visible in the dynamic properties of the films. We illustrate
this effect by an analysis of the MSDs of the monomers and FIG. 5. Log-log plot of g0,x共t , y兲 for the monomers 共filled symbols兲 and
solvent particles 共open symbols兲 in a supported film of a polymer solution
solvent particles parallel to the wall. The MSD is defined by 共N = 64兲 at T = 0.4 and ␾S = 25%. The MSDs are shown for different positions
N in the film, for monomers, and solvent in the center 共filled and open circles兲,
1 x
g0,x共t兲 = 兺 具兩si共t兲 − si共0兲兩2典
at the supporting wall 共filled and open triangles兲, and at the free surface
共x = P,S兲, 共4兲 共filled and open squares兲. For comparison the solid lines indicate g0,x共t兲 for
Nx i=1
the solvent and monomers in a bulk polymer solution 共Ref. 20兲 at ␾S
= 25% and T = 0.4 共multiplied by 2/3 to account for the difference in dimen-
where si共t兲 = 共xi共t兲 , zi共t兲兲 is the position of particle i at time t
sions兲. The horizontal dotted line shows the criterion used to extract the
parallel to the wall and the sum runs either over all mono- relaxation time ␶x by g0,x共␶x , y兲 = 1. The dashed line through the solvent data
mers 共N P兲 or solvent particles 共NS兲. at the free surface represents Eq. 共5兲 with ␥ = 5.5.
014902-5 Simulation of polymer solutions in thin films J. Chem. Phys. 131, 014902 共2009兲

monomer by their neighbors at low T. At long times the polymer 0%

5%
monomer eventually escapes its neighbor cage. Then, we 10
2
10%
find g0,P共t兲 ⬃ t1/2 for displacements of the order of the mono- 15%

τx(y,φS)
1 21%
mer diameter and beyond. This subdiffusive motion is caused 10 27%
by chain connectivity and agrees with the prediction of the 0
10 solvent
Rouse model30,31 for displacements less than the chain size. T=0.5
Figure 4 reveals that the solvent weakens the neighbor -1
10 0 5 10 15 20 25
cage and plasticizes the polymer; the monomer MSD in- y
creases faster than for the pure film. In turn, the solvent also 5 0%
10 polymer 6%
experiences caging by the monomers. In comparison to the 10
4
12%
3
MSDs in regions 2 and 3 the solvent dynamics is strongly 10 18%

slowed down. After the ballistic regime g0,S共t兲 closely fol-
lows the monomer MSD, although g0,S共t兲 is always larger
than g0,P共t兲. With increasing time the solvent and monomer
displacements separate from each other. Due to chain con-
nectivity we have g0,P共t兲 ⬃ t1/2, while the solvent MSD di-
rectly crosses over to diffusive motion, g0,S共t兲 ⬃ t.
In the interfacial and gas regions 共regions 2 and 3兲 a
direct crossover from ballistic motion to free diffusion oc-
curs. Since the density is low in these regions, we may in-
terpret this behavior by the theory of gases. In the Enskog
approximation, the velocity autocorrelation function of a gas
particle decays exponentially with a time constant ␥−1.32 This
leads to
4T
g0,S共t兲 = 2 关␥t − 共1 − e−␥t兲兴.
␥
Figure 4 reveals that Eq. 共5兲 provides a good description of
the MSDs for the solvent. From the fits we find ␥ = 5.153 in
region 2 and ␥ = 0.343 in region 3.
D. Layer-resolved dynamics
Previous work on pure polymer films reveals that the
films are dynamically heterogeneous due to the influence of
the interfaces.12,16–18 Monomers at the supporting wall and
free surface have strongly enhanced mobility relative to the
bulk. With increasing distance from the interfaces the en-
hanced mobility reduces and smoothly crosses over to bulk
behavior in the center of the film.
Similarly, we expect to find a smooth gradient from
interface-induced fast dynamics to bulklike dynamics for the
supported polymer solutions studied here. To test this expec-
tation we perform a layer-resolved analysis of the MSD. That
is, we determine the MSD in a layer as a function of the
position of this layer from the supporting wall. Following
τx(y,φS)
2 25%
10 35%
1
10
0
10 solvent
-1
10 T=0.4
-2
10 0 5 10 15 20 25
y
FIG. 6. Relaxation time ␶x共y , ␾S兲 for monomers 共filled symbols, x = P兲 and
solvent particles 共open symbols, x = S兲 vs the distance y from the supporting
wall for different number fractions of solvent in the center of the film 共0% is
the undiluted polymer film兲. The upper panel shows the data for T = 0.5, the
lower panel for T = 0.4. For clarity ␶S共y , ␾S兲 is divided by 10 for T = 0.5 and
by 100 for T = 0.4. The chain length of the polymers is N = 64.
ter, both the monomers and solvent display bulk dynamics;
the MSDs increase in two steps, as it is typical of cold liq-
共5兲 uids close to the glass transition. This two-step relaxation is
absent at either interfaces where the dynamics is strongly
enhanced relative to the bulk.
A more complete overview of the layer and composition
dependences of the film dynamics may be obtained by an
analysis of a local relaxation time ␶x共y , ␾S兲. We determine
this relaxation time by the criterion
g0,x共t = ␶x共y, ␾S兲,y兲 = 1 共x = P,S兲. 共7兲
␶x measures the time a monomer or solvent particle take to
move across a distance of their size parallel to the interfaces.
Figure 6 depicts ␶x共y , ␾S兲 at T = 0.5 共upper panel兲 and T
= 0.4 共lower panel兲. For both the monomer and solvent the
relaxation time smoothly increases with increasing distance
from the supporting wall toward the center of the film and
smoothly decreases again when crossing the middle and
moving toward the free surface. The position of the free
surface shifts to larger y at higher solvent concentration be-
cause the film thickness increases with ␾S. With increasing
␾S and T the smooth gradient between ␶x共y , ␾S兲 at the inter-
faces and in the center of the film diminishes. At T = 0.5 we

冓 冔
Refs. 12, 16, and 17 we define the layer-resolved MSD by find that the difference between center and interface relax-
t
ation times is about an order of magnitude, whereas it may
1 be more than three orders of magnitude at T = 0.4 共⬇ bulk Tg;
g0,x共t,y兲 = 兺 兿 ␦关y − yi共t⬘兲兴兩si共t兲 − si共0兲兩2
nx,t i t =0
. 共6兲
cf. Fig. 3兲.
⬘
This definition only takes into account the nx,t monomers
1. Parametrization of ␶S„y , ␾S…
共x = P兲 or solvent particles 共x = S兲 which are at all times t⬘
⬍ t within the layer that is centered at y and is of width ⌬y Our previous studies on undiluted polymer films16
共=2兲. As before, we only consider the dynamics in the un- showed that it is possible to parametrize the layer-resolved
constrained directions parallel to the wall. relaxation time by an empirical formula suggested in Refs.
As an example for this analysis Fig. 5 depicts the results 33 and 34. Here we apply the same parametrization to the
for T = 0.4 and ␾S = 25%. The figure supports the expectation relaxation time of the solvent at the free surface. 关We focus
of spatially heterogeneous dynamics in the film. In the cen- on ␶S共y , ␾S兲 because the modeling by Eq. 共8兲 will be utilized
014902-6 Peter, Meyer, and Baschnagel J. Chem. Phys. 131, 014902 共2009兲

6% TABLE I. Parameters used in the fit of Eq. 共8兲 to ␶S共y , ␾S兲 at T = 0.4 共Fig. 7兲.
12%
10
0
18% ␶Sc is the relaxation time in the center of the film and y + is the position of the
25% film-vapor interface.

τS(y,φS)/ττS(φS)
35%
-1 0
c 10 10

ln{A/ln[ττS/τS(y)]}/y+
␾S 6% 12% 18% 25% 35%
-1
-2 10
10 ␶cS

c
5400 2200 750 275 120
10
-2 y+ 15.55 17.12 19.78 22.61 25
-3 0 0.5 1 y/y+
10
0 0.5 1
y/y+
Diffusion in bulk polymer solutions is an experimentally
FIG. 7. Main figure: relaxation time ␶S共y , ␾S兲 for the solvent particles vs y well studied problem, and various theoretical models have
at T = 0.4. The data are the same as in the lower panel of Fig. 6, but the axes been proposed to describe the experimental results.35 These
are rescaled: The ordinate by the relaxation time ␶cS共␾S兲 in the center of the
film and the abscissa by the position of the Gibbs dividing surface y +. The
models invoke different physical concepts involving obstruc-
solid and dashed lines show Eq. 共8兲 for ␾S = 6% and ␾S = 34%, respectively. tion effects, hydrodynamic interactions, or ideas from free
Inset: same data, but rescaled according to Eq. 共8兲. That is, volume theory. Their strengths and weaknesses in rationaliz-
ln关A / ln共␶cS / ␶S共y , ␾S兲兲兴 / y + vs y / y + 共A = 8; ␶cS and y + are given in Table I兲. The ing and predicting diffusion in polymer solutions are criti-
solid line shows Eq. 共8兲 after this rescaling, i.e., 共1 − y / y +兲 / ␰ with ␰ = 1.75.
cally reviewed in Ref. 35. From this review it is clear that a
broadly accepted model applicable over a large range of con-
in part II;19 the same modeling could also describe ␶S共y , ␾S兲 centration does currently not exist. For concentrated polymer
at the supporting wall and ␶ P共y , ␾S兲 at both interfaces.兴 This solutions, however, the discussion in Ref. 35 suggests that
parametrization reads the Fujita–Yasuda model could provide a good description of

␶S共y, ␾S兲 = ␶Sc共␾S兲exp − A exp − 冋 冉 y +共 ␾ S兲 − y

␰
冊册 , 共8兲
solvent diffusion. This model proposes the following ansatz
for ␶xc共␾S兲:

冋 册
where ␶Sc共␾S兲 is the relaxation time in the center of the film,
y +共␾S兲 is the position of the film-vapor interface, and A and E0x 共T兲
␶xc共T, ␾S兲 = a0x 共T兲exp 共x = P,S兲. 共9兲
␰ are adjustable parameters. The results of this analysis are ␾0共T兲 + ␾S
exemplified in Fig. 7. The figure shows a plot of ␶S共y , ␾S兲 / ␶Sc
versus y / y + for T = 0.4. Choosing A = 8 and ␰ = 1.75—that is, Figure 8 reveals that Eq. 共9兲 can indeed fit the relaxation
values compatible with those for pure polymer films16—we times of the monomers and of the solvent at both tempera-
see that Eq. 共8兲 provides a good description of the y depen- tures 共dashed lines兲. The parameters of the fit, a0x 共T兲, E0x 共T兲,
dence of the solvent relaxation time. This agrees with our and ␾0共T兲, are compiled in Table II.
results for pure polymer films12,16 and with those for other However, other approaches are also possible. In Ref. 20
confined glass-forming liquids.33,34 we found that the monomer relaxation time in bulk solutions
is well described by the power law
2. Dependence of ␶Sc on ␾S and T
Figure 8 depicts the dependence of the relaxation time in a P共 ␾ S兲
␶ Pc共T, ␾S兲 = 共␥ = 2兲, 共10兲
the center of the film on solvent concentration for the mono- 关T − Tc共␾S兲兴␥
mers 共␶ Pc兲 and solvent 共␶Sc兲 at T = 0.4 and 0.5. For both tem-
peratures ␶cP共␾S兲 and ␶cS共␾S兲 agree with the results for bulk
with a P共␾S兲 = 3.073− 4.385␾S and Tc共␾S兲 = 0.4165− 0.270␾S.
polymer solutions determined previously 共filled symbols兲.20
Figure 8 displays Eq. 共10兲 as a dotted line for T = 0.5. Indeed,
this bulk result provides a description of similar quality as
5 N=64 T=0.4 Eq. 共9兲, not only for ␶ Pc but also for ␶Sc since ␶Sc = ␶ Pc / 2.3 at
10 T=0.5
T = 0.5.
4
10 For T ⬍ Tc共0兲 共=0.4165兲, however, Eq. 共10兲 cannot be
τx(φS)

3 polymer
polymer
applied because the denominator diverges at some ␾S. For
10
c

T = 0.4 this problem occurs at ␾S ⯝ 0.061. Nevertheless,

10
2
prompted by linear dependence of T − Tc共␾S兲 on ␾S we at-
1
solvent tempt to fit ␶Sc共␾S兲 at T = 0.4 by the power law
10
0 0.1 0.2 0.3
φS
TABLE II. Survey of the values for the parameters obtained from adjusting
FIG. 8. Relaxation time ␶xc in the center of the film, determined according to Eq. 共9兲 to the relaxation of the monomers 共P兲 and solvent 共S兲 in the center
Eq. 共7兲 vs ␾S at T = 0.5 共squares兲 and T = 0.4 共circles兲 for the monomers and of the film.
solvent as indicated. The filled symbols show the values obtained from bulk
simulations of polymer solutions at the same ␾S and p = 0 共Ref. 20兲. All data T a0P E0P a0S E0S ␾0
refer to a chain length of N = 64. The dashed lines indicate a fit to Eq. 共9兲.
For T = 0.5 the dotted line shows Eq. 共10兲 with parameters determined 0.4 0.220 5.170 0.420 4.125 0.377
previously 共see text for details兲. For T = 0.4 the dotted line represents a fit to 0.5 2.288 2.682 0.995 2.682 0.490
Eq. 共11兲.
014902-7 Simulation of polymer solutions in thin films J. Chem. Phys. 131, 014902 共2009兲

2
ãS D. W. Schubert and T. Dunkel, Mater. Res. Innovations 7, 314 共2003兲.
␶Sc共␾S兲 = 共␥ = 2兲. 共11兲
3
M. Alcoutlabi and G. B. McKenna, J. Phys.: Condens. Matter 17, R461
共 ␾ S + ␾ c兲 ␥ 共2005兲.
4
C. B. Roth and J. R. Dutcher, in Soft Materials: Structure and Dynamics,
Figure 8 shows that such a fit, although worse than Eq. 共9兲, is edited by J. R. Dutcher and A. G. Marangoni 共Dekker, New York, 2005兲,
also possible 共ãS = 22.5 and ␾c = 0.01兲. We will employ this pp. 1–38.
5
latter parametrization of ␶Sc in part II where we investigate H. Richardson, M. Sferrazza, and J. Keddie, Phys. Rev. E 70, 051805
共2004兲.
solvent evaporation from thin polymer films. 6
H. Richardson, M. Sferrazza, and J. Keddie, Eur. Phys. J. E 12, S87
共2003兲.
7
IV. SUMMARY G. Reiter and P. G. de Gennes, Eur. Phys. J. E 6, 25 共2001兲.
8
S. Al Akhrass, G. Reiter, S. Y. Hou, M. H. Yang, Y. L. Chang, F. C.
We have performed MD simulations of a coarse-grained Chang, C. F. Wang, and A. C.-M. Yang, Phys. Rev. Lett. 100, 178301
model for polymer-solvent mixtures in a supported film ge- 共2008兲.
9
ometry. These simulations, an extension of recent work on G. Reiter, M. Hamieh, P. Damman, S. Sclavos, S. Gabriele, T. Vilmin,
and E. Raphaël, Nature Mater. 4, 754 共2005兲.
bulk polymer solutions,20 focus on equilibrium properties for 10
H. Bodiguel and C. Fretigny, Eur. Phys. J. E 19, 185 共2006兲.
temperatures above the glass transition of the films. The 11
C. Ellison, M. Mundra, and J. Torkelson, Macromolecules 38, 1767
main results obtained may be summarized as follows. 共2005兲.
12
As in the bulk,20 the solvent acts as a plasticizer. The J. Baschnagel and F. Varnik, J. Phys.: Condens. Matter 17, R851 共2005兲.
13
T. Vilmin and E. Raphaël, Phys. Rev. Lett. 97, 036105 共2006兲.
glass transition temperature Tg is reduced relative to the un- 14
M. Tsige and G. S. Grest, Macromolecules 37, 4333 共2004兲.
diluted polymer films. The reduction in Tg increases with 15
M. Müller and G. D. Smith, J. Polym. Sci., Part B: Polym. Phys. 43, 934
increasing solvent concentration ␾S 共Fig. 3兲. 16
共2005兲.
We find that the structure and dynamics in the films are S. Peter, H. Meyer, and J. Baschnagel, J. Polym. Sci., Part B: Polym.
Phys. 44, 2951 共2006兲.
spatially heterogeneous. The solvent density is enriched at 17
S. Peter, H. Meyer, J. Baschnagel, and R. Seemann, J. Phys.: Condens.
the supporting wall and at the free surface where the film is Matter 19, 205119 共2007兲.
in equilibrium with a vapor of solvent 共Fig. 2兲. At both in-
18
S. Peter, S. Napolitano, H. Meyer, M. Wübbenhorst, and J. Baschnagel,
terfaces the solvent dynamics is fast, but slows down toward Macromolecules 41, 7729 共2008兲.
19
S. Peter, H. Meyer, and J. Baschnagel, J. Chem. Phys. 131, 014903
the center of the film where it is bulklike 共Figs. 4–6兲. A 共2009兲.
20
smooth gradient from enhanced dynamics at the interfaces to S. Peter, H. Meyer, and J. Baschnagel, Eur. Phys. J. E 28, 147 共2009兲.
21
bulk behavior in the center of the film is also found for the K. Binder, J. Horbach, W. Kob, W. Paul, and F. Varnik, J. Phys.: Con-
dens. Matter 16, S429 共2004兲.
monomers 共Figs. 5 and 6兲. This spatial variation of the dy- 22
T. Soddemann, B. Dünweg, and K. Kremer, Phys. Rev. E 68, 046702
namics qualitatively agrees with previous results for pure 共2003兲.
polymer films12,16–18 and may also be parametrized by the 23
H. T. Davis, Statistical Mechanics of Phases, Interfaces and Thin Films
same formula as for the undiluted system 共Fig. 7兲. This pa- 共Wiley-VCH, New York, 1995兲.
24
A. L. Frischknecht and J. G. Curro, J. Chem. Phys. 121, 2788 共2004兲.
rametrization, together with the modeling of the composition 25
M. Müller, L. G. MacDowell, and A. Yethiraj, J. Chem. Phys. 118, 2929
dependence of the solvent relaxation time in the center of the 共2003兲.
film 共Fig. 8兲, will be utilized in part II of this work which is 26
A. Yethiraj, J. Chem. Phys. 109, 3269 共1998兲.
27
devoted to the study of solvent evaporation from polymer S. Sen, J. M. Cohen, J. D. McCoy, and J. G. Curro, J. Chem. Phys. 101,
9010 共1994兲.
films.19 28
C. Mischler, J. Baschnagel, S. Dasgupta, and K. Binder, Polymer 43, 467
共2002兲.
29
ACKNOWLEDGMENTS Y.-C. Lin, M. Müller, and K. Binder, J. Chem. Phys. 121, 3816 共2004兲.
30
M. Rubinstein and R. H. Colby, Polymer Physics 共Oxford University
We acknowledge financial support from the European Press, Oxford, 2003兲.
31
Community’s “Marie-Curie Actions” under Contract No. M. Doi and S. F. Edwards, The Theory of Polymer Dynamics 共Oxford
MRTN-CT-2004-504052 共POLYFILM兲. J.B. acknowledges University Press, Oxford, 1986兲.
32
J. P. Hansen and I. R. McDonald, Theory of Simple Liquids 共Academic,
financial support by the IUF. London, 1986兲.
33
P. Scheidler, W. Kob, and K. Binder, Europhys. Lett. 59, 701 共2002兲.
1 34
K. Norrman, A. Ghanbari-Siahkali, and N. B. Larsen, Annu. Rep. Prog. P. Scheidler, W. Kob, and K. Binder, J. Phys. Chem. B 108, 6673 共2004兲.
Chem., Sect. C: Phys. Chem. 101, 174 共2005兲. 35
L. Masaro and X. X. Zhu, Prog. Polym. Sci. 24, 731 共1999兲.