week ending

PRL 101, 115501 (2008) PHYSICAL REVIEW LETTERS 12 SEPTEMBER 2008

Thickness Induced Structural Changes in Polystyrene Films

M. K. Mukhopadhyay,1 X. Jiao,2 L. B. Lurio,3 Z. Jiang,1,2 J. Stark,3 M. Sprung,2 S. Narayanan,2
A. R. Sandy,2 and S. K. Sinha1
1
Department of Physics, University of California San Diego, La Jolla, California 92093, USA
2
Advanced Photon Source, Argonne National Laboratory, Argonne, Illinois 60439, USA
3
Department of Physics, Northern Illinois University, DeKalb, Illinois 60115, USA
(Received 5 September 2007; revised manuscript received 17 July 2008; published 10 September 2008)
Changes to the structure of polystyrene melt films as measured through the spectrum of density
fluctuations have been observed as a function of film thickness down to the polymer radius of gyration
(Rg ). Films thicker than 4Rg show bulklike density fluctuations. Thinner films exhibit a peak in SðqÞ near
q ¼ 0 which grows with decreasing thickness. This peak is attributed to a decreased interpenetration of
chains resulting in an enhanced compressibility. Measurements were made using small angle x-ray
scattering in a standing wave geometry designed to enhance scattering from the interior of the film
compared to interface scattering.

DOI: 10.1103/PhysRevLett.101.115501 PACS numbers: 61.05.cf, 68.60.
p, 82.35.Gh

A fundamental question in the theory of molecular found experimentally using other techniques. Tretinnikov
liquids is how is their structure affected by confinement et al. indirectly measured free volume changes in very thin
on a scale comparable to the molecular size. This is par- films from the shift in infrared vibration bands [10].
ticularly relevant to polymer melts since their molecular Furthermore, free volume may play a role in the observa-
size can extend to tens of nanometers. Improved under- tions of a reduced glass transition temperature in thin films
standing of the structure of molten polymers is of techno- [11–14].
logical interest for nanoscale polymeric materials. It is also To address the shortcomings in previous work we have
of fundamental interest as a test of basic theories of poly- developed a new method using x rays to study density
mer fluids, and to evaluate how well these theories predict fluctuations in the interior of the polymer film. By setting
structural modifications in response to perturbations which up an x-ray standing wave within the film, scattering from
vary over molecular length scales. the interior can be excited to the exclusion of substrate and
Silberberg [1] advanced a hypothesis that the main effect surface scattering. We find that films thicker than 4Rg show
of an interface on a polymer fluid is to reflect conforma- density fluctuations nearly identical to the bulk, while
tions of the polymer chains when they intersect the inter- thinner films exhibit a peak in SðqÞ near q ¼ 0 which
face. More detailed calculations made using a range of grows with decreasing thickness, indicating an enhanced
techniques including both analytic methods [2,3] as well as compressibility and a longer correlation length for density
Monte Carlo simulations and molecular dynamics [2–5] fluctuations. These observations are explained by reduced
support Silberberg’s basic conjecture. Silberberg’s conjec- interpenetration of the polymer chains in thin films, result-
ture predicts no other change to the chain conformations. ing in a concurrent increase in free volume. We also find
However, in films whose thickness approaches Rg reflec- that the density of thin films decreases consistent with such
tion of a chain at an interface significantly increases the lo- a free volume increase. Note that x-ray scattering, unlike
cal density of the coil and thus reduces the degree of neutron scattering, measures collective density fluctuations
entanglement. Since the chain conformation is a function rather than the structure of individual chains.
of entanglement, there must be other changes beyond We studied polystyrene (PS) films (Mw ¼ 129 K,
Silberberg’s prediction. Specifically, a reduced entangle- Mw =Mn  1:05) that were spun cast onto 20  20 mm2
ment in thin films will make the chain conformation more HF etched and polished silicon substrates and annealed
like a self-avoiding random walk at short length scales in vacuum for 24 h at 160  C. The thickness of the
relative to the random walk approximated by a bulk films ranged from 10 to 114 nm. The expected Rg of
polymer. 129 K PS is 9.6 nm yielding a range in thickness from
Neutron experiments have tried to explore this issue by 1Rg to 12Rg . Measurements were performed at beam line
directly measuring the structure factor of an individual 8-ID at the Advanced Photon Source. The sample was
chain in a thin film. Some neutron experiments have ob- held in a vacuum <10
3 torr during the measurement to
served an expansion of the chain structure [6,7], while minimize parasitic scattering and prevent x-ray damage.
other experiments find no change [8,9]. The conflicting The exposure was limited to 5 min to minimize dam-
results of these measurements may be due to the difficulty age. Data were collected over a temperature range from
of separating surface from interior scattering. Some sup- 50  C to 160  C and as a function of incident angle
from
port for reduced interpenetration in thin films has also been 0.13 to 0.26.

0031-9007=08=101(11)=115501(4) 115501-1 Ó 2008 The American Physical Society

 week ending
PRL 101, 115501 (2008) PHYSICAL REVIEW LETTERS 12 SEPTEMBER 2008

In Fig. 1, we have schematically shown the diffuse by the spatial variation of the electric field in the sample;
scattering geometry. The scattered beam was collected however, as long as the length scales associated with the
using a CCD area detector, and the intensity was normal- sample structure are sufficiently distinct from the length
ized to 10% via a secondary standard. For each value of scales associated
R with the electric field variation, and

a range of scattering vectors could be simultaneously ~
provided SðqÞdq=q 2 converges sufficiently fast, this ef-

measured comprising a range of in-plane scattering com- fect can be neglected.

ponents from 0:5 < qk < 3 nm
1 and out of plane from The solid lines in Fig. 2 display fits to the
dependence
0:4 < qz < 2:3 nm
1 . of the diffuse scattering where interior scattering is calcu-
Figure 2 shows the diffuse scattering as a function of
lated using Eq. (1). The diffuse scattering from the interior
at a fixed value of total wave vector q. Two different thick- of the film can be extracted by fitting the variation of
ness films are shown: D ¼ 20 nm (2Rg ) and 114 nm the total diffuse scattering as a function of
with three
(12Rg ). The data in each of the panels of Fig. 2 represent adjustable parameters: the intensity from the surface, sub-
a combination of scattering from the surface, substrate, and strate, and the interior of the film. The intensity of the
the interior of the films. The effect of varying
on the surface scattering agreed with the capillary wave theory
intensity is particularly important over the range of angles prediction based on the temperature and the bulk surface
between the critical angle for total external reflection of PS tension [18]. Thus, the fits involved only two parameters:

c;PS  0:17 and that of Si
c;Si  0:24 at an energy the intensity of the interior diffuse scattering from the film
7.35 keV. In this range there is interference inside the PS and the substrate scattering. The latter is negligible com-
film between the incident beam and the beam reflected by ~ but becomes important in the
pared to the former at large q,
the Si. Some particular values of
yield a standing wave small q~ region. We exclude data where the substrate scat-
condition resulting in a single peak in Fig. 2(a) and four tering exceeds the contribution of the interior of the film at
peaks in Fig. 2(b). At these conditions nearly all the an
corresponding to the strongest peak of the standing
intensity comes from the interior of the film; the surface wave inside the film. The relative contributions of surface,
and the substrate scattering are suppressed. substrate, and interior scattering of the film are shown in
A formalism for calculating diffuse surface scattering in Fig. 2.
reflection geometry using the distorted wave Born approxi-
mation was worked out by Sinha and collaborators [15]
and is described in detail by Tolan [16]. We have extended
1/A dσ/dΩ X 104

this formalism to a thin film geometry by including an 2.0

integral over the electric fields throughout the film. The (a)
details of this will be presented in full in a second paper 1.5
[17]. The interior scattering is then given by 1.0


d






r2 m fm Axy



 0 SðqÞYð
; Þ: (1) 0.5
d

interior 2
0.0
Here
is the incident angle,  the exit angle, r0 is the
Thompson radius, m is the monomer density, Axy is the (b)
1.5
1/A dσ/dΩ X 103

illuminated surface area, SðqÞ is the structure factor as a

function of the magnitude of wave vector transfer, and
Yð
; Þ is an amplitude correction term that accounts for 1.0
the integral of the electric field through the sample. We
ignore the structure of the monomer units and take the form
factor as the number of electrons fm ¼ 56. The interpre- 0.5
tation of the scattering intensity can be further complicated
0.0
0.15 0.20 0.25
α (deg.)

FIG. 2 (color online). Diffuse scattering cross section at q~ k ¼

1:525 nm
1 and qz ¼ 0:616 nm
1 as a function of incident
angle for (a) 20 nm, (b) 114 nm thick PS films on Si at
160  C. Symbols are experimental data and the solid line is
the fit. The surface (red dash-dotted line), substrate (magenta
FIG. 1 (color online). Schematic of the diffuse scattering dot-dot-dashed line), and interior (blue dotted line) scattering
setup. contributions are shown separately.

115501-2
 week ending
PRL 101, 115501 (2008) PHYSICAL REVIEW LETTERS 12 SEPTEMBER 2008

0.20 0.08 first term in Eq. (3). The quadratic term represents the
quasistatic component. This model then has three adjust-
0.06 able parameters, T , , and A. The dashed lines in Fig. 3

S(q)
0.15 represent fits using this model. For brevity we focus on the
0.04
data taken at T ¼ 160  C, well above Tg . Data at other
0.02
temperatures above bulk Tg  100  C of PS are similar.
S(q)

0.10 10 20 30 40 50 The thickness dependence of T and  are shown in

tan−1(qz/qr) (deg.)
Fig. 4 for 160  C. The compressibility shows no change for
D=Rg > 6 and then shows a sharp increase with decreasing
0.05 D up to about 7 times for the thinnest film. The thickness
S(0)bulk dependence of the compressibility can be reasonably well
described by the functional form:
0.00
0 1 2 3 T ðDÞ ¼ bulk
T ½1 þ C1 expð
D=D0 Þ: (4)
q (nm−1)
The extrapolated value of T for an infinitely thick film
FIG. 3 (color online). SðqÞ for film thicknesses of 10 nm (red is consistent with the bulk value. The temperature depen-
diamonds solid line), 20 nm (blue circles dot-dot-dashed line), dence of T (not shown) is also consistent with that of the
38 nm (black triangles dash-dotted line), 57 nm (cyan crosses bulk. The values required to fit the data of compressibility
dotted line), and 105 nm (magenta squares dashed line) at as a function of thickness are D0 ¼ 18:9  6:6 nm and
160  C. Lines represent fits, symbols represent the measured
C1 ¼ 13:6  3:8. 2 multiplied by a factor 8.0 is also
data. The arrow marks the bulk Sð0Þ. Inset: Variation of SðqÞ with
orientation at q ¼ 2 nm
1 for film thicknesses of 10 nm (red plotted in the same scale to show the similar thickness
squares solid line), 20 nm (blue crosses dashed line), and 38 nm dependence as compressibility.
(black triangles dotted line). We can qualitatively understand our results as follows.
For thin films the polymer chains fold over when they
reach the surface leading to a reduced entanglement. This
Figure 3 shows the interior contribution to SðqÞ averaged reduced entanglement then leads to an increase in com-
over the orientation of q.~ The orientation dependence of pressibility. Furthermore, as the chains become less en-
~ at q ¼ 2 nm
1 for various film thicknesses is shown
SðqÞ
in the inset of Fig. 3. The scattering was found to vary by 10
around 20% with orientation in the thinnest film, with little
variation in thicker films. In bulk PS the thermal diffuse 1.05
scattering [19] approaches a constant at small q given by
ρ (gm/cc)

1.00
Sð0Þ ¼ m fm kB TT : (2)
0.95

(ξD)2 (nm2)
κT(D) (GPa−1)

Here kB is the Boltzmann constant, T is the temperature,

0.90
and T is the isothermal compressibility. For thick films
5
Eq. (2) holds at small q and yields the expected value of 0.85
T . At larger q there is a rise in SðqÞ which has also been 50 100 150
seen previously for bulk PS by Roe and Curro [19]. They Temperature
interpreted the rise as resulting from quasistatic correla-
tions between the monomers. Roe and Curro fit their data
to the form SðqÞ ¼ Sð0Þ expðbq2 Þ which would also work
for our thick film data, but fails for films thinner than 60 nm
where there is an additional rise in intensity at small q. This 0
0 5 10
behavior is qualitatively similar to the structure factor D/Rg
calculated in Monte Carlo simulations of thin polymer
films by Binder [20], and also to the predictions of the FIG. 4 (color online). Thickness dependence of fitting parame-
Gaussian thread model of Schweizer and Curro [2,21]. ters. The black circles show the measured film compressibility.
To account for this scattering we employ a functional Thicknesses are normalized by the bulk radius of gyration Rg .
form inspired by the Gaussian thread model: Red diamonds show the square of the correlation length obtained
from the fitting and are multiplied by a factor of 8 to appear on
Sð0Þ the same scale as the compressibility. Lines depict fits using
SðqÞ ¼ þ Aq2 : (3)
1 þ 2 q2 Eq. (4). Inset: Density versus temperature for thicknesses of
20 nm (blue circles), 38 nm (black triangles), 57 nm (cyan
Here  represents a density fluctuation screening length. crosses), 70 nm (red diamonds), 78 nm (green rotated crosses),
The Gaussian thread model [2,21] would only predict the and 105 nm (magenta squares).

115501-3
 week ending
PRL 101, 115501 (2008) PHYSICAL REVIEW LETTERS 12 SEPTEMBER 2008

tangled excluded volume effects become more pronounced technical work, and A. Habenschuss for sharing data on
at small length scales. This leads to a stiffening of the bulk PS melts. ’This work is supported by NSF Grant
chains, and suppresses short length scale (but not long No. DMR-0209542. Use of the Advanced Photon Source
length scale) fluctuations leading to the observed increase at Argonne National Laboratory was supported by the
in . A similar mechanism was suggested by Shuto and co- U.S. Department of Energy, Office of Science, Office of
workers [6] and was also recently predicted in theoretical Basic Energy Sciences, under Contract No. DE-AC02-
studies by Peter et al. [22]. 06CH11357.
We can also understand these results in the context of the
Gaussian thread model [2,21]. This model predicts that
there is a relationship between T and  given by T ¼
12ð=Þ2 =kB T. This relation does not yield the correct [1] A. Silberberg, J. Colloid Interface Sci. 90, 86 (1982).
value for T for the known bulk statistical segment length [2] K. S. Schweizer, in Advances in Chemical Physics, edited
(  0:68 nm [9]). However the prediction that T  2 is by I. Prigogine and S. A. Rice (John Wiley & Sons, New
satisfied. This can be seen in Fig. 4. York, 1997), Vol. 98, p. 1.
[3] F. T. Oyerokun and K. S. Schweizer, J. Chem. Phys. 123,
Oyerokun and Schweizer [3] have calculated that the
224901 (2005).
backfolding of polymer chains in thin films leads to an [4] J. Baschnagel, H. Meyer, F. Varnik, S. Metzger, M.
ordering of the orientation of the chains parallel to the Aichele, M. Muller, and K. Binder, Interface Sci. 11,
surface. They calculate that such orientation would in- 159 (2003).
crease the compressibility by a factor of 3 for films with [5] A. Cavallo, M. Muller, J. P. Wittmer, A. Johner, and K.
Rg ¼ D. This is, however, smaller than the value found Binder, J. Phys. Condens. Matter 17, S1697 (2005).
here of 7 times enhancement. This effect should be aniso- [6] K. Shuto, Y. Oishi, T. Kajiyama, and C. C. Hant,
tropic leading to a dependence of SðqÞ~ on the orientation of Macromolecules 26, 6589 (1993).
q~ with respect to the surface. This effect is seen in the [7] A. Brulet, F. Boue, A. Menelle, and J. P. Cotton,
present data as shown in Fig. 3. Macromolecules 33, 997 (2000).
[8] R. L. Jones, S. K. Kumar, D. L. Ho, R. M. Briber, and T. P.
An increase in  should also expand the polymer chains
Russell, Nature (London) 400, 146 (1999).
and consequently reduce the film density. This prediction [9] R. L. Jones, S. K. Kumar, D. L. Ho, R. M. Briber, and T. P.
contradicts the neutron reflectivity results of Wallace et al. Russell, Macromolecules 34, 559 (2001).
[23], who found no change in the polymer film density with [10] O. N. Tretinnikov and R. G. Zbankov, Macromolecules 37,
thickness. By contrast, we do find a decrease in the density 3543 (2004).
for thinner films. These results are shown in the inset of [11] J. Keddie, R. A. L. Jones, and R. A. Cory, Europhys. Lett.
Fig. 4. The decrease in the density is most apparent for thin 27, 59 (1994).
films and only at temperatures well above Tg , which may [12] J. A. Forrest, K. Dalnoki-Veress, J. R. Stevens, and J. R.
explain the difference between our measurements and Dutcher, Phys. Rev. Lett. 77, 2002 (1996).
theirs. [13] J. A. Forrest, K. Dalnoki-Veress, and J. R. Dutcher, Phys.
Rev. E 56, 5705 (1997).
In conclusion, we have shown that the structure factor of
[14] C. J. Ellison and J. M. Torkelson, Nature Mater. 2, 695
thin PS melt films develops a small q peak as the film (2003).
thickness approaches Rg . This peak is predicted both from [15] S. K. Sinha, E. B. Sirota, S. Garoff, and H. B. Stanley,
Monte Carlo simulations and analytic Gaussian thread Phys. Rev. B 38, 2297 (1988).
calculations, and can be interpreted as the signature of an [16] M. Tolan, X-Ray Scattering from Soft-Matter Thin Films
increasingly large excluded volume as characterized by an (Springer, Berlin, 1999).
increasing correlation length . Furthermore, the Gaussian [17] M. Mukhopadhyay, Z. Jiang, S. K. Sinha, and L. B. Lurio
thread model predicts that the compressibility should in- (unpublished).
crease as the square of  which is also observed. While [18] L. Lurio, H. Kim, A. Ruhm, J. Basu, J. Lal, S. Sinha, and
present theories qualitatively predict this result, there are S. G. J. Mochrie, Macromolecules 36, 5704 (2003).
[19] R. Roe and J. Curro, Macromolecules 16, 428 (1983).
no predictions which yield the correct magnitude of the
[20] K. Binder, Physica (Amsterdam) 200A, 722 (1993).
observed effect nor the observed scaling with thickness. [21] K. S. Schweizer and J. G. Curro, Chem. Phys. 149, 105
We hope that these results will motivate further develop- (1990).
ment of theoretical understanding of the effects of confine- [22] S. Peter, H. Meyer, J. Baschnagel, and R. Seemann,
ment on polymer melts. J. Phys. Condens. Matter 19, 205119 (2007),
We would like to acknowledge K. Binder and K. [23] W. E. Wallace, N. C. B. Tan, W. L. Wu, and S. Satija,
Schweizer for useful discussions, H. Gibson for his expert J. Chem. Phys. 108, 3798 (1998).

115501-4