THE JOURNAL OF CHEMICAL PHYSICS 122, 144712 !
2005"
Local dynamic mechanical properties in model free-standing
polymer thin films
Kenji Yoshimoto, Tushar S. Jain, Paul F. Nealey, and Juan J. de Pabloa!
Department of Chemical and Biological Engineering, University of Wisconsin-Madison, Madison,
Wisconsin 53706-1691
!Received 15 December 2004; accepted 25 January 2005; published online 14 April 2005"
High-frequency sinusoidal oscillations of a coarse-grained polymer model are used to calculate the
local dynamic mechanical properties !DMPs" of free-standing polymer thin films. The storage
modulus G! and loss modulus G" are examined as a function of position normal to the free surfaces.
It is found that mechanically soft layers arise near the free surfaces of glassy thin films, and that their
thickness becomes comparable to the entire film thickness as the temperature approaches the glass
transition Tg. As a result, the overall stiffness of glassy thin films decreases with film thickness. It
is also shown that two regions coexist in thin films just at the bulk Tg; a melt-like region !G!
! G"" near the free surfaces and a glass-like region !G! " G"" in the middle of the film. Our findings
on the existence of a heterogeneous distribution of DMPs in free-standing polymer thin films
provide insights into recent experimental measurements of the mechanical properties of glassy
polymer thin films. © 2005 American Institute of Physics. #DOI: 10.1063/1.1873732$
I. INTRODUCTION results suggest that a mobile layer exists near the free sur-
face, and that a substrate can have a strong influence in the
The thermal and mechanical properties of polymeric dynamics of polymer thin films.
nanostructures are of critical importance in a wide variety of It is intuitively attractive to anticipate a correlation be-
technological applications. In advanced lithography, the stiff- tween a localized, enhanced polymer motion and a local me-
ness of photoresist patterns is an important parameter in the chanical weakening of a material. To the best of our knowl-
capillary force induced collapse that arises as rinsing sol- edge, however, the local mechanical properties of polymer
vents evaporate.1,2 The glass transition temperature Tg of thin films have not been measured as a function of position
chemically amplified photoresist thin films is also a key pa- within the film. It is also important to note that the mechani-
rameter for optimization of lithographic processes.3 cal deformation of polymeric nanostructures observed in ad-
A number of experimental4–7 and theoretical8,9 studies vanced lithography1,2 takes place at room temperature, i.e.,
have shown that Tg of thin polymer films can change with 20%–30% below the Tg of amorphous polymers such as PS
thickness h. Mattsson, Forrest, and Börjesson, for example, and poly!methyl methacrylate" !PMMA". The mechanical
report that the Tg of free-standing polystyrene !PS" thin films properties of amorphous polymers are nearly independent of
decreases monotonically below thicknesses of 100 nm.4 In temperature in the glass region, but they change significantly
contrast, past work by our group suggests that the Tg of as Tg is approached.10,11 The extent of softening in glassy
supported PS thin films can be either larger or lower than the thin films will therefore depend not only on thickness but
bulk Tg for h ! 100 nm, depending on the chemical interac- also on the proximity to Tg.
tions between the substrate and the polymer matrix.5,7 Recent experimental studies suggest that the overall
Torres, Nealey, and de Pablo have demonstrated in mo- stiffness of glassy polymer thin films remains comparable to
lecular dynamic simulations that the diffusivity of polymer that of the bulk for film thicknesses in the range 30–40
segments is highly heterogeneous in polymer thin films and nm.12,13 The Young’s modulus E of PS thin films, for ex-
that it is strongly correlated with deviations of Tg from the ample, is comparable to that of the bulk for h " 30 nm.12
bulk.8 Ellison and Torkelson have recently quantified dy- Also, the propagation of acoustic waves through PMMA thin
namic heterogeneities in supported PS thin films by using a films exhibits the same velocity as in the bulk for h
fluorescence-multilayer technique, which is capable of mea- " 40 nm.13 Similar observations emerge from molecular
suring local Tg at different locations in the film.6 For rela- simulations of polymer nanostructures.14,15 The overall bend-
tively thick films !h = 270 nm", the Tg near the free surface is ing modulus of the glassy nanostructures remains bulk-like
shown to be %30 K lower than the bulk Tg !=374 K", until a critical dimension !e.g., line width" approaches 20#.
whereas no change of Tg is found in the the middle or near The symbol # represents the diameter of the coarse-grained
the bottom of the film. For thinner films !h = 24, 36 nm", polymer segments in simulations. A length of 20# would
however, a decrease of the local Tg is observed across the translate into 30–40 nm for amorphous PS or PMMA films.16
films, but its magnitude becomes smaller, i.e., 4–14 K. These Experiments and simulations therefore suggest that the
onset of changes of mechanical properties does not occur
a"
Electronic mail: [email protected] until thicknesses on the order of 30–40 nm are attained. It
0021-9606/2005/122"14!/144712/6/$22.50 122, 144712-1 © 2005 American Institute of Physics
Downloaded 05 Jun 2007 to 128.104.198.120. Redistribution subject to AIP license or copyright, see http://jcp.aip.org/jcp/copyright.jsp
144712-2 Yoshimoto et al. J. Chem. Phys. 122, 144712 "2005!
should be noted, however, that most available mechanical
measurements have been performed on supported films; it is
unclear whether the absence of any changes of the mechani-
cal properties is indicative of a homogeneous distribution of
the bulk stiffness, or of a cancellation of the softened free
surface with the stiffened layers near the substrate.
In this work, we present results of nonequilibrium mo-
lecular dynamics !NEMD" simulations using a coarse-
grained polymer model. The local dynamic mechanical prop-
erties !DMPs" of free-standing polymer thin films are
reported as a function of thickness and temperature. By lim-
iting our study to free-standing films, we are able to elimi-
nate any convolution of effects arising from the presence of a FIG. 1. Schematic view of model free-standing polymer thin film. The ar-
free surface and a substrate. Dynamic properties, such as G! row on top indicates the oscillatory shearing direction. Each sphere repre-
and G", are given special attention in this work. Previous sents the monomer segment.
studies have already examined the purely elastic response of
amorphous polymer glasses well below Tg;14,15 near Tg, how-
ever, the loss component of the modulus becomes significant. shifted LJ potential in Eq. !1". Such a choice of parameters
More importantly, near a free surface, where segmental mo- inhibits crystallization and facilitates the formation of an
bility can be highly enhanced, it is of interest to consider amorphous material.9 In what follows, all quantities are re-
viscous losses, as the local Tg might be much smaller than ported in reduced units,17 e.g., distance r* = r / #, number den-
that of the bulk. sity &* = , temperature T* = kBT / %, where kB is the Boltz-
We begin our discussion by describing the model and mann constant, and stress '* = '#3 / %.
methods employed in this work. We then report results for The bulk system contains 22 polymer chains under peri-
the local viscoelastic properties, which show how thin films odic boundary conditions in all three directions. The films
become mechanically and dynamically heterogeneous in consist of 22 or 44 chains, depending on thickness. Figure 1
free-standing geometries. We also discuss the formation of provides a schematic representation of our thin film systems.
soft layers near free surfaces, the effect of these soft surface Periodic boundary conditions are implemented in the x and y
layers on the overall stiffness of glassy thin films, and the directions; the film thickness h is defined in the z direction.
coexistence of melt-like regions and glass-like regions in All systems !bulk and thin films" are first equilibrated at
thin films near the glass transition temperature. temperatures well above Tg !T* = 0.60", and then cooled
gradually to a temperature well below Tg !T* = 0.10". In all
simulation runs, the time step was set at 0.001 in reduced
II. SIMULATION MODEL AND METHOD units. For the bulk system, the pressure was fixed at zero and
the volume of the system was allowed to fluctuate. For thin
A. Model
films, the box length in the x and y directions was fixed at 8#
We have adopted a coarse-grained polymer model that !see Fig. 1" and the pressure normal to the free surfaces was
has been used in previous studies of glasses.9 Each polymer maintained at zero. The same cooling procedure was re-
chain consists of 32 interactive sites !i.e., monomer par- peated with six to ten statistically independent configurations
ticles". The interaction energy function u!r" between non- prepared at high temperatures !T* = 0.60". The resulting en-
bonded monomers is given by a shifted Lennard-Jones !LJ" ergy, density, and structure factor at each temperature were
potential,17 found to be the same !within statistical errors" for all simu-
' (
&
dv!r" lation runs. By plotting the energy !or the density" as a func-
v!r" − v!rc" − !r − rc" r $ rc tion of temperature, T*g was estimated to be 0.37 for the thin-
u!r" = dr r=rc , !1" ner films of 22 chains and 0.40 for the bulk and the thicker
0 r " rc films of 44 chains.9
where r is the distance between the two monomers and rc is
the cutoff distance !=2.5#". The function v!r" is a simple
Lennard-Jones potential energy function,
v!r" = 4%#!#/r"12 − !#/r"6$, !2" B. Method
where # and % represent the monomer diameter and the in- To calculate the loss and storage modulus of our sys-
teraction energy parameter, respectively. For bonded pairs, tems, we implemented a SLLOD algorithm with Lee–
u!r" is represented by a simple harmonic potential, Edwards periodic boundary conditions.17–20 NEMD simula-
tions were performed under an oscillatory shear flow.17,18
u!r" = k!r − r0"2/2, !3"
The temperature was kept constant via a Nose–Hoover chain
where k and r0 are the spring constant !=2000% / #2" and the thermostat.19,20 A double thermostat of unit mass was em-
bond length !=1#", respectively. Note that r0 is slightly ployed. The time-dependent shear strain imparted to our sys-
smaller than the distance at the energy minimum of the tems, (xy!t", is given by a simple sinusoid of the form
Downloaded 05 Jun 2007 to 128.104.198.120. Redistribution subject to AIP license or copyright, see http://jcp.aip.org/jcp/copyright.jsp
144712-3 Local dynamic mechanical properties in model free-standing polymer thin films J. Chem. Phys. 122, 144712 "2005!
(xy!t" = (max sin )t, !4"
where (max and ) are the amplitude and the frequency of an
oscillation, respectively. Superscript xy denotes the two Car-
tesian components parallel and normal to the shearing direc-
tion !see Fig. 1". The shear stress corresponding to (xy!t" can
be calculated from the following expression:17
' !t" =
xy −1
V
') N
mivxi !t"viy!t"
(
i=1
)* * dU!r" rxij!t"rijy!t"
− , !5"
i"j dr r=rij!t" rij!t"
where V is the volume of the system and N is the total num-
ber of monomers. The mass and the x !or y" velocity com-
ponent of monomer i are denoted by mi !=1" and vxi !or viy",
respectively. In the same manner, the magnitude and the x !or
y" component of the vector which connects the positions of
monomers i and j are denoted by rij and rxij !or rijy", respec-
tively. The summation runs over all possible monomer pairs.
In a homogeneous bulk system, the local stresses at any
spatial point are expected to be the same as that obtained
from Eq. !5". In thin films, however, the local stresses de-
pend strongly on position !e.g., near the free surfaces or at
the middle of the film" and direction !e.g., tangential or nor-
mal to the free surfaces".21 To consider such a heterogeneous
FIG. 2. DMPs of the bulk at !a" )* = 0.3, !b" )* = 0.06, !c" )* = 0.03 vs
stress distribution explicitly, we have applied a one- temperature. In each figure, G!*, G"*, and tan * are shown in !, ", and #,
dimensional version of the Irving–Kirkwood stress tensor:21
+
respectively.
+
N
−1
'xy!t,z" =
A
) mivxi !t"viy!t"*#z − rzi !t"$ 'xy!t"/(max = G!!)"sin )t + G"!)"cos )t
. !8"
i=1 'xy!t,z"/(max = G!!),z"sin )t + G"!),z"cos )t
)* * dU!r" rxij!t"rijy!t" In the fitting, the sum of the squared differences between the
−
r=rij!t" rij!t" calculated stress and that predicted from Eq. !8" was mini-
(-
i"j dr
' ('
mized by optimizing the two independent parameters, G!!)"
1 z − rzi !t" rzj !t" − z and G"!)" for the bulk, or G!!) , z" and G"!) , z" for thin
+ , , , !6"
,rzij!t", rzij!t" rzij!t" films. Note that G! and G" must be independent of the strain
amplitude (max at a given frequency as shown in Eq. !8". We
where , is the Heaviside step function. Note that 'xy !t , z", verified this independence by performing the same calcula-
represents the average local stress over the area of the xy tions described above with several different amplitudes. In
plane, denoted by A. The kinetic term in Eq. !6" and the local what follows, we show the DMPs averaged over different
density, initial configurations and/or different strain amplitudes.
N
1
&!t,z" = )
A i=1
mi*#z − rzi !t"$, !7" III. RESULTS AND DISCUSSION
A. DMPs for the bulk
were estimated with a technique developed by Daivis, In Fig. 2, the DMPs for the bulk are shown as a function
Travis, and Todd.22 The two velocity components, vxi and viy of temperature. The frequencies )* employed in this study
are independent, so the kinetic terms in Eqs. !5" and !6" are 0.30, 0.06, and 0.03. We should note that, in order to
become approximately zero. obtain the DMPs at )* = 0.03, we have run the NEMD simu-
Over 50 oscillation cycles were applied to all systems. In lations over 108 time steps; this required one month of com-
each cycle, instantaneous stresses were determined from Eq. puter time on a 2 GHz CPU.
!5" and !6" at regular time intervals. The dominant frequency At )* = 0.3 #Fig. 2!a"$, it is observed that the decrease of
of the stress data was analyzed by a discrete Fourier trans- the stiffness G!* as the glass transition is traversed is only a
form. As expected, all results were consistent with the fre- factor of 10. The loss modulus G"* does not exhibit a clear
quency of the applied strain, namely, ) in Eq. !4". The stress peak in the vicinity of the glass transition. In addition, the
data were subsequently fitted to an expression of the form:10 intersection point of G!* and G"*, which can be viewed as
Downloaded 05 Jun 2007 to 128.104.198.120. Redistribution subject to AIP license or copyright, see http://jcp.aip.org/jcp/copyright.jsp
144712-4 Yoshimoto et al. J. Chem. Phys. 122, 144712 "2005!
FIG. 4. Local DMPs of thicker films !h* % 22#" at )* = 0.03 as a function of
FIG. 3. DMPs of the bulk at different frequencies vs temperature; !a" G!* vs the z* coordinate: G!*!z*" in bold solid line, G"*!z*" in bold dashed line, and
T*, !b" G"* vs T*, !c" tan * vs T*. In each figure, !, ", and # represent the &*!z*" in thin solid line. The scale on the left axis is for G!*!z*" and G"*!z*",
data obtained at )* = 0.3, 0.06, and 0.03, respectively. and that on the right axis is for &*!z*". From the top to the bottom figure,
temperature is increased constantly by 0.1 in reduced units. The insets at
T* = 0.05 and 0.6 show local DMPs near the middle of film.
another definition of T*g, is %20% larger than the one esti-
mated from the density-temperature curve in the cooling pro-
cedure !=0.40". As the frequency is decreased #Figs. 2!b" and different temperatures. The left and right edges of G!*!z*",
2!c"$, however, the glass transition becomes more pro- G"*!z*", and &*!z*" are zero beyond the polymer-vacuum in-
nounced, as evidenced by a larger decrease of G!* and a terface. The film thickness h* is defined as the distance be-
sharper appearance of the peak in G"*. The intersection of tween the two points of z* at which &*!z*" is equal to half of
G!* and G"* also moves towards T* = 0.40 #see Figs. 2!a" and the bulk density. For the case of Fig. 4, h* is %21–22 de-
2!c"$. pending on the temperature.
To clarify the DMPs dependence of frequency, each Far below the bulk T*g !i.e., T* = 0.1, 0.2", G!*!z*" exhib-
property in Fig. 2 is combined into one figure !Fig. 3". Far its a sharp increase near the free surfaces and a fluctuation
below T*g, it is observed that G!* is nearly constant #Fig. 3!a"$ around a bulk value in the middle of the film. In contrast,
whereas G"* decreases with )* #Fig. 3!b"$. Consequently, G"*!z*" is vanishingly small over the entire film thickness.
tan * becomes smaller at lower frequencies #Fig. 3!c"$ and Most of the film !except the immediate vicinity of the free
the system becomes more elastic. On the other hand, above
surfaces" may be assumed rigid and elastic. As the tempera-
T*g, both G!* and G"* decrease with )* by the same magni-
ture approaches the bulk T*g !i.e., T* = 0.3, 0.4", however,
tude, and therefore tan * remains nearly constant; our bulk
G!*!z*" rises more gently near the free surfaces and attains
systems are clearly viscoelastic in the melt regime. Note that
the bulk value deep inside the film. In addition, the magni-
the - peak is not observable in Fig. 3!c". !Since the - peak is
tude of G!*!z*" becomes comparable to that of G"*!z*" in the
broadened at higher frequencies, the highest temperature
considered here is not large enough to see a clear peak." middle of the film. These results reveal a gradual transition
Unfortunately, our calculations become inaccurate for the of the thin film from the glass to the melt. Note that, just at
melt at higher temperatures !"0.6"; the stiffness is too small the bulk T*g !i.e., T* = 0.4", a small portion of G"*!z*" becomes
to generate a detectable stress response. larger than G!*!z*" near the free surfaces, i.e., a phenomenon
akin to “surface melting”6 occurs. Meanwhile, a glass-like
region #G!*!z*" " G"*!z*"$ is still maintained in the middle of
B. Local DMPs for thin films
the film. This finding is consistent with the experimental ob-
For thin films, we focus our attention on results at the servations on PS thin films mentioned above; a smaller Tg at
smallest frequency, namely, )* = 0.03. Figure 4 shows the the free surface,6,23 but no change in Tg deep inside the film.6
local G!*, G"*, and density as a function of z* !see Fig. 1" at By further increasing the temperature above the bulk T*g !i.e.,
Downloaded 05 Jun 2007 to 128.104.198.120. Redistribution subject to AIP license or copyright, see http://jcp.aip.org/jcp/copyright.jsp
144712-5 Local dynamic mechanical properties in model free-standing polymer thin films J. Chem. Phys. 122, 144712 "2005!
TABLE I. Ratio of the soft surface layers hs* to the entire film thickness h*.
T* hs* h* !hs* + 2" / h*
0.3 2.5–3.0 22 0.23–0.27
3.0–3.5 11 0.55–0.64
of the thicker films; we also detect the emergence of a soft
surface layer and surface melting. In order to compare the
thickness of soft surface layers hs* in Figs. 5 and 4, we define
hs* as a surface region for which G!*!z*" ! Gbulk!* . The results
at T = 0.3 are summarized in Table I. As shown in Table I, hs*
*
for h* = 11 is found to be between 3.0 and 3.5, approximately
the same as hs* for h* = 22. Consequently, the ratio of the soft
surface layers !hs* + 2" to the overall film thickness h* for the
11# thick film becomes 55%–64%, which is more than
double that for the 22# film. With such a large surface ratio,
it is clear how significant surface effects become on the me-
chanical properties of free-standing glassy thin films.
C. Overall DMPs for thin films
By integrating G!!z" and G"!z" over the film thickness,
one can also calculate the overall DMPs for free-standing
polymer thin films:
FIG. 5. Local DMPs of thinner films !h* % 11#" at )* = 0.03 as a function of
z* coordinate. The symbols and scales are the same as those in Fig. 4. G! =
1
h
.
h
G!!z"dz, G" =
1
h
.
h
G"!z"dz. !9"
T* = 0.5, 0.6", G"*!z*" becomes larger than G!*!z*" for most The results are shown in Fig. 6. In the glass region, the G!*
values of z*; the entire film turns into a highly viscoelastic curve for the 22# film !# with solid line" almost overlaps
melt. with that for the bulk !! with solid line"; however, that for
It is also interesting to observe in Fig. 4 that the hetero- the 11# film !" with solid line" exhibits a systematic depar-
geneous behavior of G!*!z*" and G"*!z*" does not simply ture from the other two. As discussed in the preceding sec-
correspond to the local density profile, particularly near the tion, a decrease of G! in thinner films is due to the greater
free surfaces. Far below the bulk T*g, there exists a small lag contribution !greater surface to volume ratio" of soft surface
between G!*!z*" and &*!z*"; &*!z*" reaches the bulk density
first, and then G!*!z*" rises to the bulk G!*. This delay of
G!*!z*" is mainly due to the lower density near the free sur-
faces. Note that the transition of &*!z*" from zero to a bulk
density is completed within 1–2# at all temperatures, and the
shoulder slope of &*!z*" is only weakly dependent on tem-
perature. Meanwhile, the shoulder slope of G!*!z*" becomes
milder at higher temperatures !i.e., T* = 0.3, 0.4", and several
of the surface underlying layers exhibit a G!*!z*" below the
bulk value, even if a bulk density has been attained !see, for
example, z* = 4.0–6.0 and z* = 25.5–24.0 at T* = 0.3, z*
= 3.5–6.5, and z* = 25.0–22.0 at T* = 0.4". This decrease of the
stiffness at bulk-like densities is consistent with observations
that the motion of polymer segments near interfaces is highly
cooperative,9 and that the activation energy for segmental
motions near free surfaces is much lower.24 Faster segmental
motions near the free surfaces can initiate a global rotational
motion that includes many segments,9 and the mobility of
polymer segments in the film can be enhanced even under
bulk-like conditions.24 Above the bulk T*g, we could not find
a clear relationship between &*!z*" and G!*!z*" or between FIG. 6. Overall DMPs of thin films at )* = 0.03 vs temperature; ! for the
bulk, # for 22# thick film, " for 11# thick film. G!* and G"* are shown
&*!z*" and G"*!z*", mainly as a result of relatively large fluc-
with solid and dashed lines, respectively. An arrow indicates the intersecting
tuations. point of the two curves of G!* and G"*. The inlet shows the ratio of G!* for
The behavior of 11# films !see Fig. 5" is similar to that the thin films to the bulk G!* as a function of temperature.
Downloaded 05 Jun 2007 to 128.104.198.120. Redistribution subject to AIP license or copyright, see http://jcp.aip.org/jcp/copyright.jsp
144712-6 Yoshimoto et al. J. Chem. Phys. 122, 144712 "2005!
1
layers to the overall behavior of the system. The magnitude M. P. Stoykovich et al., Adv. Mater. !Weinheim, Ger." 15, 1180 !2003".
2
K. Yoshimoto et al., J. Appl. Phys. 96, 1857 !2004".
of the decrease in G! is further clarified in the inset of Fig. 6. 3
L. F. Thompson, C. Willson, and M. J. Bowden, Introduction to Microli-
The G! for the 11# film is found to be 60%–80% of the bulk thography, 2nd ed. !American Chemical Society, Washington, D. C.,
G! in the glass region !T* ! 0.4". Such a large decrease of 1994".
4
stiffness could become a large limiting factor for the semi- J. Mattsson, J. A. Forrest, and L. Börjesson, Phys. Rev. E 62, 5187
!2000".
conductor industry in its attempts to produce sub-50-nm 5
R. S. Tate et al., J. Chem. Phys. 115, 9982 !2001".
pitched patterns. 6
C. J. Ellison and J. M. Torkelson, Nat. Mater. 2, 695 !2003".
7
D. S. Fryer et al., Macromolecules 34, 5627 !2001".
IV. CONCLUSIONS 8
J. A. Torres, P. F. Nealey, and J. J. de Pablo, Phys. Rev. Lett. 85, 3221
!2000".
In our calculations of the local DMPs for free-standing 9
T. S. Jain and J. J. de Pablo, J. Chem. Phys. 120, 9371 !2004".
polymer thin films with oscillatory deformations, it is shown 10
R. S. Lakes, Viscoelastic Solids !CRC, Boca Raton, FL, 1998".
11
that soft layers are formed near the free surfaces of glassy D. W. Van Krevelen, Properties of Polymers, 3rd ed. !Elsevier, Amster-
thin films, and the ratio of the soft layers to the entire film dam, Netherlands, 1990".
12
C. M. Stafford et al., Nat. Mater. 3, 545 !2004".
thickness becomes larger as the temperature becomes closer 13
Y. C. Lee, K. C. Bretz, F. W. Wise, and W. Sachese, Appl. Phys. Lett. 69,
to the bulk Tg or as the film thickness is decreased. For 1692 !1996".
example, at T* = 0.3 !i.e., T / Tg = 0.75", which is equivalent to 14
T. R. Böhme and J. J. de Pablo, J. Chem. Phys. 116, 9939 !2002".
15
room temperature for PS and PMMA, we found that the ratio K. Van Workum and J. J. de Pablo, Nano Lett. 3, 1405 !2003".
16
M. Rubinstein and R. H. Colby, Polymer Physics !Oxford University
of soft surface layers becomes more than 50% of the entire Press, New York, 2003".
film thickness and G! becomes %70% of the bulk G! for 11# 17
M. P. Allen and D. J. Tildesley, Computer Simulation of Liquids !Oxford
thick films. We are currently working on a similar problem in University Press, New York, 1987".
18
supported thin films; we anticipate that the free surface soft- H. Komatsugawa and S. Nosé, Phys. Rev. E 51, 5944 !1995".
19
C. J. Mundy, J. I. Siepmann, and M. L. Klein, J. Chem. Phys. 103, 10192
ening might be canceled by the interfacial stiffening that can !1995".
arise near a substrate. 20
Z. Xu, J. J. de Pablo, and S. Kim, J. Chem. Phys. 102, 5836 !1995".
21
F. Varnik, J. Baschnagel, and K. Binder, J. Chem. Phys. 113, 4444 !2000".
ACKNOWLEDGMENTS 22
P. J. Daivis, K. P. Travis, and B. D. Todd, J. Chem. Phys. 104, 9651
!1996".
This work was supported by the Semiconductor Re- 23
N. Satomi, K. Tanaka, A. Takahara, and T. Kajiyama, Macromolecules 34,
search Corporation !SRC" and the National Science Founda- 6420 !2001".
tion !NSF" NIRT program. 24
T. S. Jain and J. J. de Pablo, Phys. Rev. Lett. 92, 155505 !2004".
Downloaded 05 Jun 2007 to 128.104.198.120. Redistribution subject to AIP license or copyright, see http://jcp.aip.org/jcp/copyright.jsp
