Cooperative and non-cooperative dynamics in ultra-thin lms

of polystyrene studied by dielectric spectroscopy and

capacitive dilatometry
Veronica Lupas cu
a
, Stephen J. Picken
a
, Michael Wu bbenhorst
b,
*
a
Department of Chemical Technology, Delft University of Technology, Julianalaan 136, 2628 BL Delft, The Netherlands
b
Department of Physics and Astronomy, Katholieke Universiteit Leuven, Celestijnenlaan 200D, B-3001 Leuven, Belgium
Received 16 June 2006; received in revised form 29 August 2006
Available online 19 October 2006
Abstract
The eect of thickness reductions on the glass transition dynamics in ultrathin lms of polystyrene has been studied by dielectric spec-
troscopy (DS) and capacitive dilatometry (CD). Upon reduction of the lm thickness, a systematic decrease in the dilatometric glass
transition temperatures, T
g
(dil), was observed via CD, while DS revealed a continuous speed-up and broadening of the a-process,
accompanied by only minor reductions in the fragility index. A good agreement between spectroscopic and the dilatometric glass
transition temperatures was found for lms thicker than 20 nm, while for thinner lms both quantities diverge increasingly. A likely
explanation for this discrepancy is the presence of another dynamic process showing Arrhenius-behavior (E
a
72 kJ/mol) with a pre-
exponential factor of 10
12
s being indicative for non-cooperative dynamics. Such a new process might be assigned to distinct surface
dynamics in polystyrene lms as suggested in recent papers.
2006 Elsevier B.V. All rights reserved.
PACS: 61.30.v; 77.22.Gm; 64.70.Pf
Keywords: Dielectric properties; Relaxation; Electric modulus; Films and coatings; Glass transition; Fragility; Structural relaxation
1. Introduction
Since the pioneering work of Keddie and Jones [1] in
1994, who observed substantial reductions in the glass tran-
sition temperature T
g
compared to the bulk value in ultra-
thin lms of polystyrene (PS), glass transition eects in
ultrathin polymer lms have been studied by many
research groups with the aim to improve our understanding
of the glass transition as a general phenomenon in con-
densed matter physics [28]. Starting from ellipsometric
measurements on polystyrene in the original work by Ked-
die et al. [1], systematic investigations comprising various
polymer systems (e.g., PS, PMMA, P2VP), dierent sample
geometries (freely-standing lms, supported and capped
lms) and various measurement techniques were performed
in the last decade. Though most of the studies in the earlier
days were based on density related methods such as ellips-
ometry [1,9], X-ray reectivity [2,10], Brillouin light scatter-
ing [11] and positron annihilation lifetime spectroscopy
(PALS) [12], recent studies utilise more and more tech-
niques that are directly sensitive to molecular uctuations,
such as dielectric spectroscopy [1317], calorimetric meth-
ods [7,18,19], and shear-modulated scanning force micros-
copy [8].
Despite the diversity in the experimental approaches
and, sometimes, contradicting results, one can identify
some clear trends from the experimental work:
1. Polymers having weak interactions with the substrate,
e.g., atactic polystyrene (a-PS) on Si, show typically a
reduction in the glass transition temperatures below a
0022-3093/$ - see front matter 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.jnoncrysol.2006.09.004
*
Corresponding author. Tel.: +32 16 327148; fax: +32 16 327984.
E-mail address: wubbenhorst@fys.kuleuven.be (M. Wu bbenhorst).
www.elsevier.com/locate/jnoncrysol
Journal of Non-Crystalline Solids 352 (2006) 55945600
thickness of roughly 40 nm, regardless to the chemical
nature of the substrate (e.g., SiO
x
, H-passivated
Si(111), glass, aluminium oxide). For PS, these T
g
-
reductions are also maintained in the absence of a free
surface (capped vs. uncapped lms) [1,4], and have been
conrmed consistently by dierent experimental tech-
niques, (ellipsometry, positron annihilation and X-ray
reectivity).
2. In contrast, polymers with strong specic interactions to
the substrate like poly(methyl methacrylate) (PMMA)
on SiO
x
might show either depressions or elevations in
the glass transition temperature, however, the specic
T
g
(L) trends depends often on subtle details in the poly-
mer (micro)structure such as the stereoregularity of the
polymer.
Dielectric relaxation spectroscopy (DRS) is particularly
useful to assess the dynamics of ultra-thin polymer lms
in a wide dynamic range and thus provides detailed spectral
information about the dynamic glass transition and other
relaxation processes present in both the glassy and the melt
state. DRS measurements have been performed on capped
and supported layers of PS [14,19,20], PVAc [21], PMMA
[1517,21], cis-polyisoprene [22] and Poly(2-vinylpyridine)
P2VP [23,24], and have provided detailed insights into the
glass transition dynamics, secondary relaxation linked to
local conformations [16,25], and even whole chain mobility
[22] by virtue of dielectric normal process. Regarding DRS
results on ultrathin PS lms, Fukao was the rst one who
studied capped thin polystyrene lms supported on an alu-
minium (Al) coated glass substrate using Dielectric Spec-
troscopy [14,26]. His results are largely consistent with
results obtained with the other methods in view of the trend
that the observed T
g
decreases with decreasing the sample
thickness. In his experiments, glass transition temperatures
were determined using the temperature change in the high-
frequency permittivity, a technique known as capacitive
dilatometry (CD). It was also shown that the thickness
dependence in T
g
is directly correlated to the width of the
a-process in the temperature representation and thus to
the distribution of relaxation times of the a-process.
Another, interesting nding was the observation of a
second, weak relaxation peak (labelled a
l
) at temperature
below the main transition [14]. Though the authors could
not determine the thermal activation parameters of this
process, a link was suggested between the observed a
l
-peak
and a possibly distinct surface dynamics in PS lms in the
context of a three-layer model. Such layer model assumes
that a thin lm having one free surface (supported lm)
likely consists of three thickness regions, a dead layer of
immobilised polymer chains at the substrate, a bulk like
core layer and a surface layer next to the free surface, each
of which characterised by specic dynamics. In this picture,
the a-process described the cooperative dynamics in the
core layer, which become broadened upon thickness reduc-
tion due to its interconnection to a layer of enhanced
mobility (surface) and a layer of reduced mobility (dead
layer). As a consequence, a net shift in the glass transition
dynamics could be rationalised by an asymmetric eect of
the outer layers on the overall dynamics. Though such a
layer model is a useful aid for the interpretation of aver-
aged mobility data (like peak shape and position of the
a-relaxation), one has to realise the actual mobility prole
in ultrathin lms might be more complicated as shown
recently by a Ellison et al. by using multi-layer PS samples
labelled with uorescent probes [23].
A systematic study by dielectric spectroscopy of the
glass transition dynamics in ultra-thin atactic polystyrene
lms is presented in this paper. A specic aim is to compare
the volumetric glass transition temperature as obtained by
capacitive dilatometry, with thickness eects on the core
dynamics manifested by the dielectric a-relaxation. For this
purpose, simultaneous CD and dielectric relaxation spec-
troscopy (DRS) measurements on PS samples in the thick-
ness range from 8 < L < 285 nm were performed.
2. Experimental
Atactic polystyrene (Pressure Chemical Company) with
a weight average molecular mass of 160.000 g/mol and low
polydispersity (<1.06) was used in this study. Ultra-thin
PS-lms were prepared by spin coating of the polymertol-
uene solution on aluminium deposited glass slides (DRS
samples) for 20 s at a rotation speed of 3000 rpm. The
resulting lms were annealed for about 12 h at
T
ann
= 100 C in order to remove residual solvent and to
allow for the relaxation of internal stresses. After this
annealing procedure, a patterned top electrode was depos-
ited by ash evaporation (deposition time <5 s) of puri-
ed aluminum from a tungsten lament, a process that
resulted in ultra-smooth Al-layers between 50 and 100 nm
in thickness.
Polymer lms of dierent thickness were obtained by
varying the concentration of the polymer in solution, while
all other parameters (spin coating) were kept constant. In
this study, (semi)-dilute PS solutions between 0.22 and
5 wt% yielded samples in the thickness range from 8.7 to
285 nm.
Dielectric measurements were performed in the fre-
quency range from 10
1
to 10
7
Hz using a high-resolution
dielectric analyzer (ALPHA Analyzer, Novocontrol Tech-
nologies). The thin-lm samples were rst heated to
150 C, annealed for 1 h and subsequently measured upon
cooling to 20 C at an eective rate of 0.5 K/min. All
measurements were done under an N
2
atmosphere that pre-
vented moisture uptake and possible oxidation during the
experiments. A typical result is given in Fig. 1 that shows
the dielectric loss of a PS lm of 50 nm in thickness in a
three-dimensional representation e
00
(f, T).
The thickness of the lms was evaluated from the value
of capacitance at room temperature using the relation for
the capacitance (C
0
) of a parallel-plate capacitor,
C
0
= e
0
e
o
S/L, where e
0
, e
0
, S and L are the permittivity of
V. Lupas cu et al. / Journal of Non-Crystalline Solids 352 (2006) 55945600 5595
bulk PS, the vacuum permittivity, the electrode area
(S = 4 mm
2
) and lm thickness.
3. Results
Owing to the weak dielectric activity of polystyrene,
dielectric measurements on PS-lms oer the determina-
tion of the glass transition temperature in two ways. The
rst method, capacitive dilatometry, relies on the analysis
of the temperature dependence of the permittivity e
0
(T) as
shown in Fig. 2 for bulk-PS and various thin-lms samples.
At suciently high frequencies, i.e. outside the appearance
of the a-relaxation, the permittivity e
0
(T) solely depends on
the density of the PS lm and thus allows the evaluation of
the linear thermal expansivity in the lm thickness direc-
tion from the e
0
(T)-curves. From the kink in the e
0
(T)
dependence, indicating the dierent volume expansivities
in the glass state and the amorphous melt, the dilatometric
glass transition temperature T
g
(dil) can be obtained [19].
The second way is the evaluation of the spectroscopic
T
g
from the temperature dependent shift of the a-relaxa-
tion peak. This method requires a full analysis of complete
dielectric loss spectra at dierent temperatures; the details
will be discussed later in this paper.
Fig. 2 displays the temperature dependence of the nor-
malized permittivity for various lm thicknesses. As
expected, the value of T
g
(dil) decreases systematically
upon lowering the lm thickness, ranging from 100 C
for bulk PS and the 285 nm thick lm sample to about
60 C for the thinnest lm. Furthermore, the glass transi-
tion region for the thinnest lms (15 and 8.7 nm) is tremen-
dously broadened compared to the bulk sample that is
characterized by a narrow kink in e
0
(T).
The thickness dependence of the dilatometric glass
transition temperature as obtained from all experimental
e
0
(T)-curves is summarized in Fig. 3. For comparison with
previous literature data, T
g
-values from Fukao et al. [26]
obtained on PS-lms of two dierent molecular weights
(1.8 10
6
g/mol, 2.8 10
5
g/mol) are shown as well (open
symbols).
Furthermore, there are two lines plotted in Fig. 3 that
represent a best t to T
g
-data from the literature obtained
by dierent experimental techniques. Here, the left line
refers to glass transition values for supported lms, the
Fig. 1. Three-dimensional representation of the dielectric loss e
00
(f, T) of a
PS lm of 50 nm thickness.
0 50 100 150
T [C]
0.95
1.00
1.05

'
(
T
)
/

'
(
0

C
)

+

bulk
285 nm
20 nm
15 nm
8.7 nm
Fig. 2. Temperature dependence of the normalized permittivity at
100 kHz for bulk-PS and for four lms at various thicknesses from
8.7 nm to 285 nm. For clarity, all curves have been normalized and
vertically shifted by a constant D = n 0.01 (n = 0, 1, 2, 3, 5).
0 20 40 60 80
L [nm]
40
60
80
100
T
g
(
d
i
l
)

[

C
]
M
w
= 1.610
5
M
w
= 1.810
6
[14]
M
w
= 2.810
5
[14]
Fig. 3. Thickness dependence of the dilatometric glass transition temper-
ature T
g
(dil) for ultrathin lms of PS. The two lines represent typical T
g
-
trends compiled from extensive literature referring to freely standing lms
(right line) [38] and supported lms (left line) [28]. Solid symbols show our
T
g
-data for PS of M
w
= 1.6 10
5
g/mol. For comparison, literature data
obtained by capacitive dilatometry [14] are indicated by open symbols.
5596 V. Lupas cu et al. / Journal of Non-Crystalline Solids 352 (2006) 55945600
right line represents the trend for freely standing lms
[1,3,5,20,26,27]. From Fig. 3 we can make two observa-
tions: First, both our T
g
(dil) values and those from Ref.
[26] fall in between the upper limit (supported lms) and
the lower limit (freely-standing lms) despite the fact that
the PS sandwich geometry has no true free surface. Second,
it is obvious that our T
g
reductions are generally lower than
Fukaos [14] values, however, they are still close to the limit
for supported lms [28].
In order to obtain the spectral glass transition temper-
ature T
g
(a) from the dielectric a-process of PS, the follow-
ing strategies were applied. For thick lms showing narrow
loss peaks in the frequency spectra, the dielectric relaxation
time s(T) was determined by a t of the dielectric loss spec-
tra e(x) to the empirical HavriliakNegami (HN) relaxa-
tion function (Eq. (1)) [29]:
e
00
Im
De
1 ixs
a

b
( )

r
e
v
x
1
where De corresponds to the relaxation strength, while the
two shape parameters a and b represent the logarithmic
slope of the low frequency loss tail (a) and the high fre-
quency loss tail (a b). The second term in Eq. (1) ac-
counts for ohmic conduction, more details about the t
procedure for dielectric data are given in [30,31]. The relax-
ation time data for the 285 nm thick PS sample, obtained in
this way, are plotted in the activation diagram, Fig. 4. For
thinner samples, where loss peaks became noisier and
strongly broadened, peak relaxation time data were ob-
tained by analyzing isochronal loss curves e
00
(T) at various
frequencies (cf. open triangles and open circles in Fig. 4).
In a next step, these relaxation time data s(T) were tted
to the VogelFulcherTammann equation
s s
1
exp
E
V
RT T
V

; 2
which typically describes very well the temperature depen-
dence of the structural relaxation time s
a
(T) of glass form-
ing materials. Here, E
V
and T
V
are the Vogel-activation
energy and the Vogel temperature [3234]. The other
parameters R and s
1
have the usual meaning. For the thin-
nest lm (L = 8.7 nm), a two-dimensional t procedure of
the e
00
(f, T) data to a combined HN-VFT-function [30,31]
was applied that yielded the VFT-parameters (rightmost
solid line in Fig. 4) in a direct way.
The knowledge of the VFT-parameter nally allows us
to evaluate two other important quantities that character-
ize the glass transition: The steepness index m (fragility)
according to a denition by Bo hmer et al. [35] (Eq. (3)),
and the operationally dened glass transition temperature
by assuming s(T
g
) = 100 s.
m
d loghsi
dT
g
=T

TTg

E
V
2:303R
T
g
T
g
T
V

2
3
The resulting T
g
(a) values are displayed in Fig. 5 together
with the corresponding dilatometric data T
g
(dil) and the
steepness index (inset gure). All VFT-parameters are sum-
marized in Table 1.
From Figs. 4 and 5 we can make the following
observations:
2.0 2.5 3.0
1000/T [K
-1
]
-8
-6
-4
-2
0
2
l
o
g
(

[
s
]
)
8.7 nm
15 nm
20 nm
286 nm
T
g
(

=100s)
Fig. 4. Activation plot of the relaxation time s
a
determined from the peak
maxima in the dielectric loss e
00
(T) for a-PS lms of various thickness. The
lines represent ts according to the VFT equation. For the thinnest lm
(L = 8.7 nm), a two-dimensional t procedure of the e
00
(f, T) data to a HN-
VFT-function [30,31] was applied that yields the VFT-parameters (shown
as solid line) instead of discrete s
a
values.
5 10 100 500
L [nm]
40
50
60
70
80
90
100
T

[

C
]
T
g
(dil)
T
g
()
5 10 100 500
L [nm]
0
50
100
150
200
s
t
e
e
p
n
e
s
s

in
d
e
x

m
Fig. 5. Thickness dependence of the dilatometric glass transition temper-
ature T
g
(dil) (open circles) in comparison to T
g
(a) (lled triangles),
evaluated from the s
a
(T)-data. The dashed line is only a guide for the eyes.
The inset gure displays the corresponding thickness dependence of the
steepness index m.
V. Lupas cu et al. / Journal of Non-Crystalline Solids 352 (2006) 55945600 5597
1. The dielectric a process, which was detectable even in
the thinnest PS lms, shows a systematic speed-up
towards lower thicknesses.
2. While the entire cooperative dynamics is obviously
aected by a thickness reduction, no substantial changes
in the steepness index as seen from Fig. 5 can be noticed,
a nding which is in contradiction to earlier observa-
tions by other authors [14].
3. Comparing the dilatometric T
g
[T
g
(dil)] with the glass
transition values obtained from the VFT-parameters
[T
g
(a)] reveals a good agreement between the two quan-
tities for thicknesses above 20 nm, while there is an
increasing discrepancy between T
g
(dil) and T
g
(a) for
lms thinner than 20 nm. This discrepancy approaches
is maximum for the thinnest sample (L = 8.7 nm), that
exhibits a dierence by more than 15 C between the
dilatometric T
g
(61 C) and the T
g
(a) being 77 C.
Further inspection of the dielectric spectra of the sample
with the lowest thickness reveals another peculiar feature
that is clearly seen in Fig. 6. Here, the temperature depen-
dences of both the permittivity (left Figure) and the loss
(right gure) are displayed in isochronal representation
for four dierent frequencies between 0.7 Hz and 260 Hz.
Besides the a-relaxation being located around 100 C at
these frequencies, a second relaxation process shows up at
lower temperatures. This second relaxation process is char-
acterized by both a weaker intensity and lower thermal
activation than the a-process as indicated by the remark-
able peak shift within the frequency range displayed in
Fig. 6. In order to classify this new process the data were
analyzed in more detail, again by applying a two-dimen-
sional t procedure as mentioned before, based on loss
data of dierent frequencies and temperatures. The result
is given in Fig. 7 showing the temperature dependence of
Table 1
VFT-parameters and steepness index for all PS lm corresponding to
Fig. 5
L (nm) E
V
(kJ/mol) log s
1
T
V
(K) T
g
(s
a
= 100 s) (C) m
285 10.9 12.0 330.9 95.8 146
20 11.4 12.0 318.3 84.6 136
15 10.2 12.0 320.0 82.2 151
8.7 11.3 12.8 295.8 77.3 111
-50 0 50 100 150
T [C]
2.45
2.50
2.55
2.60

'
(
T
)
260 Hz
16 Hz
2.8 Hz
0.7Hz
-50 0 50 100 150
T [C]
0.00
0.01
0.02
0.03

"
K
K
(
T
)

Fig. 6. Isochronal presentation of the permittivity e

0
(T) (left) and the corresponding dielectric loss e
0 0
(T) (right) at four frequencies for a 8.7 nm thick
polystyrene sample. The symbol a
s
denotes the second relaxation process.
2.0 2.5 3.0 3.5 4.0
1000/T [K
-1
]
-8
-6
-4
-2
0
2
l
o
g
(

[
s
]
)
285 nm
8.7 nm
8.7 nm

s
Fig. 7. Activation plot of the relaxation time data s
a
(T) of bulk PS
(L = 285 nm) together with s
a
(T) data obtained for the 8.7 nm thick
sample (cf. Fig. 6). The dashed curve represents a VFT-t to the bulk
data while the two solid curves show the temperature dependence of the a
and a
s
process of the ultrathin lm sample according the VFT or
Arrhenius law.
5598 V. Lupas cu et al. / Journal of Non-Crystalline Solids 352 (2006) 55945600
the a-process together with that of the new sub-T
g
process.
The s
a
(T)-data for the bulk like sample (L = 285 nm) are
given as a reference. As seen from the gure, the new pro-
cess clearly shows Arrhenius behavior (Eq. (4)), i.e. the
relaxation time obeys the equation for simple thermal
activated processes with the activation parameters
E
a
= 71 kJ/mol and s
1
= 10
12
s.
s s
1
exp
E
a
RT

4
4. Discussion
The results described in the previous section have dem-
onstrated the unique strength of dielectric techniques for
the study of ultra-thin PS lms: From a single experiment,
measuring the frequency-and temperature dependent com-
plex permittivity spectra e
*
(f, T), information is gathered
about both volume relaxation (thickness changes) and the
segmental dynamics in a wide frequency range. Though
both the dilatometric and the spectroscopic [s
a
(T)]
response contain specic information on the dynamic glass
transition, the two (sub)-experiments are performed simul-
taneously on the same sample, which is an essential advan-
tage over the common practice to compare experimental
results from dierent samples (geometry, preparation tech-
nique, thermal and solvent history, etc.) and dierent mea-
surement techniques. As a consequence, both the T
g
(dil)
and T
g
(a) values should coincide within the experimental
accuracy for thick, bulk-like PS samples, which is validated
by the results Fig. 5 for L > 20 nm.
Another main nding that deserves further discussion
(cf. Fig. 3) is the observation that the volumetric T
g
-reduc-
tions (from CD) resemble well the behavior known for
truly supported PS lms. Such agreement is not evident,
since both the lower and the upper aluminum electrodes
are in intimate contact with the polymer surface. However,
some authors discussed the possibility of partial delaminat-
ing of the polymermetal interface during the harsh depo-
sition process of the upper electrode onto the polymer lm
[36,37]. We think that the existence of dimensionally stable
blisters, which mimic a free surface on local scale, is
highly unlikely in view of the high polymer mobility above
T
g
and the strength of attractive dispersion forces across
the polymermetal gap. Anyhow, one should realize, that
the roughness of the two polymermetal interfaces within
the Al-polymer-Al sandwich samples might be quite dier-
ent. While the interfacial roughness of the lower alumi-
num electrode is in the range of 1.22 nm rms, the upper
electrode usually forms a much smoother interface
(0.5 nm rms) to the polymer lm, resulting in a priori
asymmetric lm boundary conditions.
Another potential source for asymmetry in the thin
polymer lm samples is the fact, that after spin-casting
onto the lower metal electrode, the lms were annealed
while they had a free upper surface. Subsequent deposition
of a weakly interacting metal like aluminum onto the PS
surface will probably not aect the once formed, specic
surface structure, which might govern the resulting T
g
-
reduction of the nal sandwich sample.
The most striking result of the present study is the appar-
ent divergence of the volumetric glass transition temperature
fromT
g
(a), derived fromthe relaxation time of the dielectric
a-process, for lms below20 nmin thickness. Such behavior
can be rationalized by basically two arguments:
1. Ultra-thin polymer lms exhibit a strong mobility gradi-
ent across the lm thickness that gives rise to a wide dis-
tribution in relaxation times of the cooperative
dynamics as manifested by an increasing peak width of
the a-process as well as in the broadened dilatometric
transition range. Since both responses average over the
mobility gradients in a dierent way, slight shifts in
the characteristic transition temperature are expected.
2. Though T
g
(dil) and T
g
(a) are deduced from the same
experimental data set, the characteristic frequency of
both quantities is dierent. While the equivalent fre-
quency of the CD-experiment is around 10
3
Hz,
depending on the eective sample cooling rate
(0.5 K/min) and the apparent activation energy of
the a-process, reliable dielectric relaxation data were
only available down to 1 Hz. However, to explain an
increasing discrepancy in the T
g
-values one needs to pos-
tulate a qualitative change in the temperature depen-
dence of the glass transition dynamics below 1 Hz, an
assumption that is not in agreement with previous
DRS studies on various ultra-thin polymer lms.
While a qualitative change in the VFT behavior of the
segmental dynamics is not a likely mechanism, we think
that the unusually low dilatometric glass transition temper-
ature, including its tremendous broadening, could be the
eect of a hidden, second relaxation process as we have
found for the 8.7 nm sample (Figs. 6 and 7). From the Acti-
vation parameters it was concluded that this additional
relaxation process refers to a non-cooperative molecular
motions, i.e. some local dynamics that do not involve
entropy uctuations. Since this relaxation process becomes
prominent in the thinnest PS lms, it is a straightforward
idea to assign this a
s
-process to the dynamics of a distinct
surface layer as suggested in recent publication [19]. As a
consequence, this surface layer, which does not contribute
to the cooperative a-process, has a faster dynamics than the
lm core layer at temperatures below the merging temper-
ature with the a-process (cf. Fig. 7).
Given the fact that surface and bulk dynamics manifest
themselves in two separate dielectric relaxation processes, it
is obvious that both dynamic processes should contribute
to the dilatometric response as well. We think that the tre-
mendous broadening and downshift of the expansivity
curves as measured by CD and other dilatometric tech-
niques might largely be the result of a superposition of a
distinct bulk-type dynamics with the specic response of
a structurally and dynamically distinct surface-near region.
V. Lupas cu et al. / Journal of Non-Crystalline Solids 352 (2006) 55945600 5599
5. Conclusions
The eect of lm thickness on the glass transition
dynamics in ultrathin polystyrene lms was investigated
by both dielectric spectroscopy, yielding direct kinetic
information about the segmental dynamics (a-process),
and capacitive dilatometry, probing the volumetric glass
transition upon cooling.
Systematic reductions in the dilatometric glass transition
temperatures, T
g
(dil), were found, which were slightly lar-
ger than those reported in earlier dielectric studies and
which resemble quantitatively T
g
-depressions known for
supported PS lms.
From the dielectric spectra, a systematic speed-up of the
a-process was found towards lower thicknesses, which was
accompanied by increasing broadening of the a-peak.
Despite the substantial shifts in the relaxation time, no
major changes in the fragility (or steepness index) were
found. Comparing the spectroscopic glass transition tem-
perature T
g
(a) = T(s
a
= 100 s) with the dilatometric glass
transition temperature reveals a good agreement between
T
g
(a) and T
g
(dil) for lms thicker than 20 nm, while below
20 nm both quantities diverge increasingly. A likely expla-
nation for this apparent discrepancy between volumetric
response and the temperature dependence of the coopera-
tive dynamics (a-process) is the presence of another, hid-
den, dynamic process as found for the thinnest lm in
this study (L = 8.7 nm). This new dielectric relaxation pro-
cess shows Arrhenius-behavior (E
a
72 kJ/mol), a pre-
exponential factor being typical for non-cooperative
dynamics (log[s
1
] = 12) and might be assigned to a dis-
tinct surface dynamics in polystyrene lms as discussed ear-
lier [14,19].
Acknowledgement
Veronica Lupas cu gratefully acknowledges the Dutch
Organization for Fundamental Research on Matter
(FOM) for founding this research project.
Appendix A. Supplementary data
Supplementary data associated with this article can be
found, in the online version, at doi:10.1016/j.jnoncrysol.
2006.09.004.
References
[1] J.L. Keddie, R.A.L. Jones, R.A. Cory, Europhys. Lett. 27 (1994) 59.
[2] W.E. Wallace, J.H. van Zanten, W.L. Wu, Phys. Rev. E 52 (1995)
3329.
[3] J.A. Forrest, K. Dalnoki-Veress, J.R. Stevens, J.R. Dutcher, Phys.
Rev. Lett. 77 (1996) 2002.
[4] J.A. Forrest, K. Dalnoki-Veress, J.R. Dutcher, Phys. Rev. E 56 (1997)
5705.
[5] K. Fukao, Y. Miyamoto, J. De Physique IV 10 (2000) 243.
[6] S. Kawana, R.A.L. Jones, Phys. Rev. E 63 (2001) 021501.
[7] M.Y. Efremov, E.A. Olson, M. Zhang, L.H. Allen, Thermochim.
Acta 403 (2003) 37.
[8] S. Sills, R.M. Overney, W. Chau, V.Y. Lee, R.D. Miller, J. Frommer,
J. Chem. Phys. 120 (2004) 5334.
[9] C.B. Roth, J.R. Dutcher, Eur. Phys. J. E 12 (2003) S103.
[10] E. Cecchetto, N.R. de Souza, B. Jerome, J. De Physique IV 10 (2000)
247.
[11] J.A. Forrest, K. Dalnoki-Veress, J.R. Dutcher, Phys. Rev. E 58 (1998)
6109.
[12] G.B. DeMaggio, W.E. Frieze, D.W. Gidley, M. Zhu, H.A. Hristov,
A.F. Yee, Phys. Rev. Lett. 78 (1997) 1524.
[13] K. Fukao, Y. Miyamoto, Europhys. Lett. 46 (1999) 649.
[14] K. Fukao, Y. Miyamoto, Phys. Rev. E 61 (2000) 1743.
[15] L. Hartmann, W. Gorbatschow, J. Hauwede, F. Kremer, Eur. Phys.
J. E 8 (2002) 145.
[16] M. Wu bbenhorst, C.A. Murray, J.A. Forrest, J.R. Dutcher, in: R.
Fleming (Ed.), 11th International Symposium on Electrets,13
October, IEEE Service Center, Piscataway N.J., Melbourne, Austra-
lia, 2002, p. 401.
[17] M. Wu bbenhorst, C.A. Murray, J.R. Dutcher, Eur. Phys. J. E direct
12 (2003) S109.
[18] D.S. Fryer, P.F. Nealey, J.J. de Pablo, Macromolecules 33 (2000)
6439.
[19] V. Lupascu, H. Huth, C. Schick, M. Wu bbenhorst, Thermochim.
Acta 432 (2005) 222.
[20] J.S. Sharp, J.A. Forrest, Phys. Rev. Lett. 91 (2003) 235701-1.
[21] K. Fukao, S.B. Uno, Y. Miyamoto, A. Hoshino, H. Miyaji, J. Non-
Cryst. Solids 307 (2002) 517.
[22] A. Serghei, F. Kremer, Phys. Rev. Lett. 91 (2003) 165702-1.
[23] C.J. Ellison, S.D. Kim, D.B. Hall, J.M. Torkelson, Eur. Phys. J. E 8
(2002) 155.
[24] V. Lupascu, M. Wu bbenhorst, manuscript in preparation (2006).
[25] M. Wu bbenhorst, V. Lupascu, ISE-12, IEEE, Brazil, 2005, p. 4.
[26] K. Fukao, Y. Miyamoto, Phys. Rev. E 64 (2001) 051807.
[27] K. Dalnoki-Veress, J.A. Forrest, C. Murray, C. Gigault, J.R.
Dutcher, Phys. Rev. E 63 (2001) 031801.
[28] J.A. Forrest, K. Dalnoki-Veress, Adv. Colloid Interf. Sci. 94 (2001)
167.
[29] S. Havriliak, S. Negami, Polymer 8 (1967) 161.
[30] J. van Turnhout, M. Wu bbenhorst, J. Non-Cryst. Solids 305 (2002)
50.
[31] M. Wu bbenhorst, J. van Turnhout, J. Non-Cryst. Solids 305 (2002)
40.
[32] H.Z. Vogel, Physics 22 (1921) 645.
[33] G. Tammann, G.Z. Hesse, Anorg. Allg. Chem. 156 (1926) 245.
[34] G.S. Fulcher, J. Am. Ceram. Soc. 8 (1925) 339.
[35] R. Bo hmer, K.L. Ngai, C.A. Angell, D.J. Plazek, J. Chem. Phys. 99
(1993) 4201.
[36] F. Faupel, R. Willecke, A. Thran, Mater. Sci. Eng. R-Reports 22
(1998) 1.
[37] T. Strunskus, V. Zaporojtchenko, K. Behnke, C. von Bechtolsheim,
F. Faupel, Adv. Eng. Mater. 2 (2000) 489.
[38] J.A. Forrest, J. Mattsson, Phys. Rev. E 61 (2000) R53.
5600 V. Lupas cu et al. / Journal of Non-Crystalline Solids 352 (2006) 55945600