Why many polymers are so fragile: A new perspective

C. Dalle-Ferrier, A. Kisliuk, L. Hong, G. CariniJr., G. Carini, G. D’Angelo, C. Alba-Simionesco, V. N. Novikov,

and A. P. Sokolov

Citation: J. Chem. Phys. 145, 154901 (2016); doi: 10.1063/1.4964362

View online: http://dx.doi.org/10.1063/1.4964362
View Table of Contents: http://aip.scitation.org/toc/jcp/145/15
Published by the American Institute of Physics

Articles you may be interested in

Perspective: Structure and dynamics of water at surfaces probed by scanning tunneling microscopy and
spectroscopy
J. Chem. Phys. 145, 160901 (2016); 10.1063/1.4964668

Perspective: Quantum mechanical methods in biochemistry and biophysics

J. Chem. Phys. 145, 140901 (2016); 10.1063/1.4964410

Capture zone area distributions for nucleation and growth of islands during submonolayer deposition
J. Chem. Phys. 145, 211911 (2016); 10.1063/1.4961264

The melting of stable glasses is governed by nucleation-and-growth dynamics

J. Chem. Phys. 144, 244506 (2016); 10.1063/1.4954327
 THE JOURNAL OF CHEMICAL PHYSICS 145, 154901 (2016)

Why many polymers are so fragile: A new perspective

C. Dalle-Ferrier,1 A. Kisliuk,2 L. Hong,3 G. Carini, Jr.,4 G. Carini,5 G. D’Angelo,5
C. Alba-Simionesco,1,a) V. N. Novikov,6 and A. P. Sokolov2,6,a)
1
Laboratoire Léon Brillouin, UMR 12, CEA-CNRS, 91191 Saclay, France
2
Chemical Sciences Division, Oak Ridge National Laboratory, Oak Ridge, Tennessee 37831, USA
3
Institute of Natural Sciences & Department of Physics and Astronomy, Shanghai Jiao Tong University,
Shanghai 200240, China
4
IPCF del CNR, UOS di Messina, I-98158 Messina, Italy
5
Dipartimento di Fisica e Scienze della Terra, Università di Messina, I-98166 Messina, Italy
6
Department of Chemistry and Joint Institute for Neutron Sciences, University of Tennessee, Knoxville,
Tennessee 37996, USA
(Received 15 May 2016; accepted 23 September 2016; published online 18 October 2016)

Many polymers exhibit much steeper temperature dependence of their structural relaxation time
(higher fragility) than liquids of small molecules, and the mechanism of this unusually high fragility
in polymers remains a puzzle. To reveal additional hints for understanding the underlying mechanism,
we analyzed correlation of many properties of polymers to their fragility on example of model poly-
mer polystyrene with various molecular weights (MWs). We demonstrate that these correlations work
for short chains (oligomers), but fail progressively with increase in MW. Our surprising discovery
is that the steepness of the temperature dependence (fragility) of the viscosity that is determined
by chain relaxation follows the correlations at all molecular weights. These results suggest that the
molecular level relaxation still follows the behavior usual for small molecules even in polymers,
and its fragility (chain fragility) falls in the range usual for molecular liquids. It is the segmental
relaxation that has this unusually high fragility. We speculate that many polymers cannot reach an
ergodic state on the time scale of segmental dynamics due to chain connectivity and rigidity. This
leads to sharper decrease in accessible configurational entropy upon cooling and results in steeper
temperature dependence of segmental relaxation. The proposed scenario provides a new important
insight into the specifics of polymer dynamics: the role of ergodicity time and length scale. At
the end, we suggest that a similar scenario can be applicable also to other molecular systems with
slow intra-molecular degrees of freedom and to chemically complex systems where the time scale
of chemical fluctuations can be longer than the time scale of structural relaxation. Published by AIP
Publishing. [http://dx.doi.org/10.1063/1.4964362]

I. INTRODUCTION dependence of their viscosity or τα(T) over a broad

The mechanism of the sharp slowing down of structural
dynamics upon cooling liquids close to their glass transition
temperature Tg remains one of the most challenging problems
in condensed matter physics. Most of the glass-forming liquids
exhibit strongly non-Arrhenius temperature dependence of
their structural relaxation time τα(T) and viscosity η(T),
with apparent activation energy increasing with temperature
decrease.1–5 The steepness of the temperature variation of η(T)
or τα(T) at Tg and the degree of non-Arrhenius behavior can
be characterized by the so-called fragility index or steepness
index m defined as6
temperature range above Tg leading to a fragility value
of m ∼ 20-227 (we note that the lowest value of fragility
m ∼ 14 was recently reported for low density amorphous water
and was ascribed to quantum effects8,9). Fragility increases
strongly in molecular liquids, reaching m ∼ 80-100 in many
van der Waals and ionic liquids.7 Among all the classes of
glass-forming systems, polymers exhibit a particular behavior:
the steepness of the temperature dependence of structural
(segmental) relaxation in many polymers can be extremely
high up to m ∼ 150-200.10,11 Such high values of fragility are
not observed in molecular systems, where usually m < 100.7
It is interesting to note that very short chains (oligomers)
of even extremely fragile polymers have typical fragility
∂ log ⌈τα (T)⌉ 

m=  T g   . (1) of van der Waals liquids, m ∼ 60-80, but then m increases
∂ T strongly with increase of molecular weight (MW) in many
T =T g

The least fragile known glass forming systems, covalent
bonded silica and BeF2, exhibit an Arrhenius temperature
a)Authors to whom correspondence should be addressed. Electronic
addresses: christiane.alba-simionesco@cea.fr and sokolov@utk.edu
0021-9606/2016/145(15)/154901/10/$30.00
polymers.11–13 This is usually explained through frustration in
packing: high-fragility polymers have either rigid backbone
or side groups, making the packing less efficient, that leads to
the increase in fragility.14–19
To unravel the mechanism that controls the steepness
of the temperature dependence of the structural relaxation
145, 154901-1 Published by AIP Publishing.
154901-2 Dalle-Ferrier et al. J. Chem. Phys. 145, 154901 (2016)

there were many attempts to correlate fragility to some other TABLE I. Sound velocities, density, Debye temperature, glass transition
properties of glass-forming systems above and even below temperature, and fragility for PS with various MW at 80 K. The data for
MW = 5100 are obtained by linear extrapolation from T = 100 K.
Tg.7,20–29 One of the puzzling correlations is the relationship
between fast dynamics and structural relaxation that are MW Vt (km/s) Vl (km/s) Density (g/cm3) θD (K) Tg (K) m
separated by more than 12 orders at T ∼ Tg.20,22–32 It has
been found that fragility also correlates to a non-ergodicity 580 1.380 2.908 1.080 85 253 68
1 100 1.323 2.771 1.087 81 294 95
parameter23,24 and elastic properties25,26,33 in a glassy state. It
5 100 1.310 2.720 1.095 81 360 124
is obvious that these correlations21 cannot provide direct
223 000 1.271 2.628 1.096 78 379 143
proof of the underlying mechanism controlling fragility.
However, they might offer additional hints to important
microscopic details influencing the fragility. We emphasize
that polymers with high fragility usually deviate from all temperature. Statistical and systematic errors are believed to
these correlations,13,31,34 which points to a polymer specific be less than 3%–4%.
behavior. Brillouin scattering spectra have been measured at
To better understand the origin of this polymer specific right angle in symmetric sample geometry. This geometry
contribution to fragility, we performed detailed studies of provides compensation of refractive index and its temperature
molecular weight dependence of several properties in a model dependence.36 We used a solid state laser with wavelength
polymer polystyrene (PS) and analyzed several correlations λ = 532 nm and ∼50 mW power on the sample. Tandem Fabri-
as a function of molecular weight (MW) in PS. Our analysis Perot Interferometer has been used to measure the spectra.
reveals that taking into consideration the molecular scale The Brillouin spectra were fitted to a Lorentzian function to
relaxation, which for polymers is the dynamics of the center estimate the peak frequency. The transverse and longitudinal
of mass of macromolecules determined by chain relaxation, sound velocities Vi were estimated from the transversal and
no difference is seen in the behavior of small molecules longitudinal Brillouin frequencies νi, respectively, using the
and polymers. It appears that PS properties obey most of following expression for the symmetric geometry at 90◦
the correlations observed for small molecules as long as scattering angle:36
fragility of viscosity (connected to the chain dynamics) is νi λ νi λ
considered instead of segmental dynamics. These observations Vi = = √ . (2)
2 sin(45◦) 2
are consistent with the idea of the critical role of intra-
molecular energy barriers in the behavior of segmental Brillouin data at the lowest measured temperature T = 80 K
relaxation,35 while these barriers play no role in the behavior of (Table I) show a clear decrease of longitudinal and transverse
chain modes with much longer relaxation times. We speculate sound velocity with increase in MW, consistent with earlier
that ergodicity of polymers is not completely restored on the studies.13
time and length scale of segmental relaxation, in contrast to
structural relaxation in small molecules. This leads to steeper
temperature variations of their segmental dynamics, while the
behavior of the chain dynamics (viscosity), that restore the
ergodicity at all scales, remains comparable to that of small
molecular systems.

II. EXPERIMENTAL
Polystyrene samples with several molecular weights and
low polydispersity were purchased from Scientific Polymer.
The samples were used without further purification. The
heat capacity was measured using an automated calorimeter
which is operated by the thermal relaxation method in a
4
He cryostat between 1.7 K and 20 K. Samples of about
10-20 mg were bonded to a silicon chip as a sample holder;
for solid polymers (MW > 1000), vacuum grease (Apiezon
N) was used to optimize the thermal contact. A SMD (surface
mounting device) resistor, as a temperature sensor, and a
constantan strain gauge, as a heater element, were attached to
the other surface of the chip using a thermosetting resin.
Temperatures were determined by calibrated commercial
germanium thermometers. The heat capacity of the sample
holder was measured in separate experiments by smearing on
the chip the same quantity of grease used to bond polymers. FIG. 1. Low temperature specific heat of PS with different molecular weight
The sample contribution to the total heat capacity varied from (shown by numbers in [g/mol]) scaled by T3 (a) and by the Debye specific
50% at the lowest temperature up to 80% at the highest heat CD (b).
154901-3 Dalle-Ferrier et al. J. Chem. Phys. 145, 154901 (2016)

Neutron scattering spectra have been measured using

disc chopper time-of-flight spectrometer at NIST. After the
background and empty cell corrections, the spectra were
summed up over all detectors to increase the statistics and
normalized by the total scattering.

III. RESULTS
Figure 1 presents the specific heat of PS at low T = 2–20 K
in normalized form, Cp/T3. This presentation stresses the
deviation of Cp from the simple Debye law expected for
homogeneous elastic continuum37

CDeb = (12π4/5)(T/θD)3, (3)

where kBθD = –hv̄(6π2n)1/3, θD is the Debye temperature,
n = N/V is the particle density, and v̄ = (3/(v−3 l + 2vt ))
−3 1/3

is the average sound velocity. As in all amorphous materials,

Cp/T3 of PS in this low-temperature range exhibits a peak
due to excess vibrations that form the so-called boson peak in
the THz frequency range.38–40 It is also visible in the neutron
scattering spectra of low MW sample (Fig. 2). The position
of the peak maximum, Tmax, shifts from 3.8 K for MW = 580
to the lower value 3.4 K for MW = 223 000, i.e., by about
11%. The amplitude of the peak in Cp/T3 increases by 20%
(Fig. 1(a)). To estimate the Debye temperature, we used
sound velocities estimated from the Brillouin scattering data FIG. 3. (a) Correlation between Cp/CD and fragility for several non-
at the lowest measured temperature T = 80 K (Table I) and polymeric glass forming systems (blue triangles) listed in the order of increas-
ing fragility, and PS with different MW (shown by numbers). (b) Correlation
between the fast relaxation intensity parameter R obtained from neutron
scattering spectra and fragility. Data for non-polymeric systems are taken
from Ref. 20.

density data from Refs. 41 and 42. The estimated values of θD

(T = 80 K) are given in Table I. If one normalizes the specific
heat by the so estimated CDeb, the change in the amplitude as
a function of MW reverses: the ratio Cp/CDeb at high MW is
lower by ∼10% than that for low MW (Fig. 1(b)). Thus, the
amplitude of the boson peak in PS shows only weak changes
with MW.
The neutron scattering spectra at Tg (Fig. 2(a)) show
a significant increase of the quasielastic scattering (QES)
contribution relative to the boson peak with increase in MW.
However, these spectra are measured at corresponding Tg
of the sample, which itself changes significantly with MW
(Table I). To remove the trivial temperature dependence,
we presented the data as the spectral density, Sn(ω)
= S(ω)/[ω∗n(ω)], where n(ω) = [exp(–hω/kT) − 1]−1 is the
Bose factor (Fig. 3(b)). For better comparison, we normalized
data at higher energy tail. The QES intensity at low energy
shows a strong increase with increase in MW at their respective
Tg. This suggests a strong increase in the amplitude of the
fast relaxation process, such as rattling in a cage, at Tg with
increasing MW in PS. This result is consistent with earlier
light scattering studies of fast dynamics in PS.13

FIG. 2. Neutron scattering spectra of PS with different molecular weight

shown by numbers (in [g/mol]). The spectra measured at temperatures slightly
below Tg of the samples: T = 230 K for MW = 540, T = 290 K for MW
= 1460, T = 340 K for MW = 6000, and T = 357 K for MW = 105 300. (a)
The dynamic structure factor summed up over all angles. (b) The same data
presented as the spectral density to remove the trivial temperature factor and
scaled at the high frequency tail of the boson peak.
IV. DISCUSSION
A. Failure of correlations with segmental fragility
It is well known that the glass transition temperature
Tg and fragility m in PS increase strongly with increase
154901-4 Dalle-Ferrier et al.
in MW.11 Similar behavior is observed for many other
relatively rigid polymers that have very high fragility,
e.g., poly(methyl methacrylate) (PMMA).43 It is also known
that many correlations of fragility with various materials
properties (e.g., non-ergodicity parameter, Poisson ratio,
change in entropy, etc.) observed for small molecular systems
and covalent-bonded glasses fail in all these polymers at
high MW.24–26,34,44,45 Using the presented results and earlier
literature data, we can provide a detailed analysis of these
correlations and unravel how they fail with increase in MW
of PS.
Let us start our analysis from the correlations between
fragility and fast dynamics in PS. It was shown that the
amplitude of the boson peak measured as the maximum of the
ratio Cp/CD is larger in stronger glass-formers and decreases
monotonously with increase in fragility.20 This correlation
was confirmed later using the density of vibrational states
measured by inelastic neutron scattering.26 The obtained data
for (Cp/CD)max in PS with different MW are plotted vs fragility
in Fig. 3(a) together with the earlier data for non-polymeric
materials. This comparison clearly reveals the failure of the
correlation for PS: data for PS oligomer with MW = 580
agree well with the trend known for non-polymeric systems;
however, the increase of molecular weight leads to strong
increase in fragility and only minor changes in the boson
peak amplitude (Fig. 3(a)). As a result, the correlation fails
progressively with the increase in MW.
Another characteristic part of the fast dynamics is the
fast picosecond relaxation which leads to broad quasielastic
scattering observed between the central line and the boson
peak in neutron (Fig. 2) and light scattering spectra.20,46–50
This QES contribution is usually ascribed to a rattling in a
cage, i.e., to the mode-coupling theory fast process, and exists
in all glass forming systems below and above Tg.51 It has
been shown that the QES intensity with respect to that of the
boson peak correlates with fragility—the higher the fragility,
the stronger is the QES intensity relative to the boson peak.20
A simple model-independent quantitative measure of the QES
intensity relative to the boson peak intensity is the ratio R
of the intensity of scattering in the minimum of the spectral
density between the central line and the boson peak to the
intensity of the maximum of the boson peak.20 Since this
ratio depends on temperature, the analysis is usually done
at Tg. In PS with higher molecular weights, the minimum
between the boson peak and the QES disappears (Fig. 2(b)).
However, qualitatively one can estimate R by taking the QES
intensity at a fixed frequency (e.g., the same as in the low MW
sample where the minimum is still visible) to the intensity
of the shoulder (the boson peak). The ratio R for PS with
different MW estimated from the data in Fig. 2(b) is shown
together with the respective data for non-polymeric systems
(Fig. 3(b)). This analysis demonstrates the correlations of
R with fragility also in the case of PS with different MW,
consistent with earlier analysis of light scattering data.20 It
means that the amplitude of the fast picosecond relaxation
(rattling in a cage) reflects the fragility even in polymers
with extremely high m. This result is consistent with the idea
formulated by generalized entropy theory that relates fragility
to frustration in molecular packing.14–17 The more frustrated
J. Chem. Phys. 145, 154901 (2016)
is packing, the larger will be amplitude of the rattling in a
cage (amplitude of the fast picosecond relaxation) and also
fragility. At the same time, the amplitude of the boson peak
changes very little when fragility of PS doubles (Fig. 3(a)).
The latter suggests that there is no direct connection of the
boson peak amplitude to fragility in polymers.
Let us consider now other correlations of material
properties with fragility for the case of PS with different
MW. It is known that in non-polymeric materials the jump
of specific heat at Tg, ∆Cp is small in strong systems and
large in fragile ones.1 The jump of the specific heat is the
basis for the so-called thermodynamic fragility.1 In non-
polymeric liquids, thermodynamic and dynamic fragilities are
in a reasonable agreement.52,53 In contrast to non-polymeric
systems, ∆Cp decreases with MW in PS, while fragility
increases.45 Thus, contrary to the correlations observed in
small molecular liquids and network glass formers, ∆Cp in PS
decreases with m and the relation between the thermodynamic
and kinetic fragilities is clearly violated in PS with different
MW.
Behavior of the PS fragility m(MW) violates other
known correlations for non-polymeric glass forming systems
discussed, e.g., in Ref. 34. It has been shown that the non-
ergodicity parameter in a glassy state correlates with fragility
of non-polymeric systems.23,24 The non-ergodicity parameter
f (Q,T) was analyzed in the glassy state using the inelastic
X-ray scattering. In the low-Q limit, temperature dependence
of the f (Q,T) in the glassy state can be approximated by a
simple relationship23
1
f (Q → 0,T) = , (4)
1 + α TTg
where α is a material-dependent constant. The analysis of
several glass-forming systems revealed23 a possible correlation
between fragility of a liquid and the parameter α that
characterizes the temperature variation of f (Q,T) in its glassy
state (Fig. 4(a)), although some systems (e.g., CKN and B2O3
in Fig. 4(a)) deviate from this correlation.24,26 This correlation
suggests that the non-ergodicity parameter, i.e., the amplitude
of the structural relaxation frozen at Tg, f0(Tg) ∼ 1/(1 + α),
correlates with the steepness of the temperature dependence
of τα(T) at Tg. The earlier PS studies24,26,34 revealed strong
violation of these correlations as MW increases (Fig. 4(a)).
Only data for PS oligomer agree well with this correlation.
However, fragility doubles with increasing MW, while the
non-ergodicity parameter α slightly decreases.24,26,34 The
same deviation is observed for many other polymers with
high fragility24 (Fig. 4(a)).
In a similar way, the observed earlier correlation between
the ratio of bulk to shear moduli K/G and fragility fails in
the case of PS with different MW (Fig. 4(b)).25,26 It has
been shown that materials with higher resistance to shear
deformation relative to compression (lower value of K/G)
are less fragile.25,26 This correlation has been observed for
molecular and network glass forming systems, although it
deviates in metallic glasses and fails in systems with complex
chemical composition (alloys).26,33,54,55 The same as for many
other properties (Figs. 3(a) and 4(a)), the value of K/G for PS
oligomer agrees well with the correlations observed for small
154901-5 Dalle-Ferrier et al. J. Chem. Phys. 145, 154901 (2016)

logτα(T)/τ0 = const/TSc(T). (5)

The latter with some assumptions can be estimated from calo-

rimetric measurements. Thus the relationship Eq. (5) suggests
a connection between two independent measurements. Direct
comparison of 1/TSc(T) with log[τα(T)] indeed confirms the
proposed relationship for many molecular systems.56–60 But
this relationship fails in the case of PS (Fig. 5(a)), PMMA, and
several other high fragility polymers.45 We did not find data
to analyze the Adam-Gibbs relationship for PS oligomers.
However, there are literature data for cumene;61 the latter can
be considered as a monomer of PS. Indeed, the Adam-Gibbs
relationship holds well for cumene (Fig. 5(c)). Thus again

FIG. 4. Correlation of segmental fragility with the non-ergodicity parameter

α (a) and with the ratio of bulk to shear moduli K/G (b). Red crosses show
data for non-polymeric systems from Refs. 23, 26, and 34. Data in (a) for
polymers are from Ref. 34. Data in (b) for PB and PIB are from Ref. 26, for
PMMA and PC—from Ref. 87.

molecules, but the PS data for K/G deviate progressively with

increase in MW. The ratio K/G decreases slightly with MW
(Fig. 4(b)), while fragility doubles.26 The same deviations are
observed for several other high-fragility polymers (Fig. 4(b)).
Comparison of Figs. 3 and 4 shows that the deviations
of PS with increase of MW follow more or less the same
scenario for all these properties: (i) data for PS oligomers
agree well with the correlations known for molecular and
covalent-bonded glass forming systems; (ii) however, while
the fragility increases drastically with increase in MW, all
other parameters do not change much (except the QES
intensity, Fig. 3(b)). This divergence in properties behavior
leads to the failure of the correlations with MW in PS. Similar
deviations are observed for PMMA, polycarbonate (PC) and
several other high fragility polymers.24,34,44 It is interesting
to note that more flexible polymers, such as polybutadiene
(PB) and polyisobutylene (PIB), do not show any significant
deviations from the correlations established for non-polymeric
systems. These polymers also do not show any strong change
of fragility with molecular weight, and m remains in the range
below ∼100, usual for molecular systems.7 They also have
much weaker change of Tg with MW.18,19,34
FIG. 5. Comparison of the Adam-Gibbs scaling in PS for the segmental
B. Adam Gibbs relationship and chain fragility relaxation (a) and viscosity (b). The data for S c are from Ref. 44, segmental
relaxation and viscosity are from Refs. 64 and 65. (c) The same comparison
According to the Adam-Gibbs theory,4 the structural for cumene (isopropylbenzene) that can be considered as a monomer of PS.
relaxation time is controlled by configurational entropy Sc(T) The data for viscosity are from Ref. 88, and for S c are from Ref. 61.
154901-6 Dalle-Ferrier et al.
FIG. 6. Fragility of chain relaxation estimated from the temperature depen-
dence of viscosity vs molecular weight in PS, triangles—data from Ref. 64,
circles—data from Ref. 65.
the property of short PS follows well the trends known for
small molecular glass forming systems. In that sense, PS
oligomers are indeed relatively small molecules with van der
Waals interactions. However, the strong increase of segmental
fragility with MW does not reflect in other properties (except
QES intensity). The same apparently is true for other high
fragility polymers.
In contrast to segmental relaxation, it has been found that
fragility in polymers determined from temperature variations
of viscosity or terminal relaxation time (both are related to
the center of mass motion, and for simplicity, in what follows
we will call it fragility of chain relaxation, mchain) does not
change much with molecular weight and remains in the limits
usual for small molecular systems.62,63 Figure 6 presents
mchain as a function of molecular weight for PS obtained from
two independent measurements.64,65 Although data slightly
disagree at low MW, they show the same trend—essentially
no significant change (maybe a slight decrease) of the chain
relaxation fragility with MW (Fig. 6). This differs strongly
with significant increase in m for segmental dynamics, as has
been already discussed in Ref. 62.
Direct comparison of the temperature dependence of
viscosity η(T) to the change in entropy Sc(T) reveals a
surprising result: The Adam-Gibbs relationship holds for the
chain relaxation (Fig. 5(b)), although it fails for segmental
dynamics in the same polymer (Fig. 5(a)). This is an
unexpected result because the glass transition is related to the
segmental relaxation, not to the chain dynamics. However, the
Adam-Gibbs relationship (Eq. (5)) holds for the chain modes
but fails for the segmental dynamics. Substitution of mchain
instead of segmental fragility in all the correlations discussed
earlier (Figs. 3-5) essentially restores all the correlations.
We note that in earlier studies the failure of the Adam-
Gibbs relationship and correlation of fragility with the non-
ergodicity parameter α were ascribed to a contribution
of a secondary relaxation.24,44 The authors argued that
the contribution of a secondary relaxation should be
subtracted from the non-ergodicity parameter and from the
configurational entropy.24,44 However, this cannot be a general
explanation. The secondary relaxation exists in PS and should
not change much with molecular weight. Moreover, there are
J. Chem. Phys. 145, 154901 (2016)
strong secondary relaxations in many small molecules that
show no deviations from the general correlations, the same
as PS oligomers. However, the deviation from the general
correlations in PS increases monotonously with increase in
MW for all the properties we discussed above (Figs. 3-5). This
emphasizes polymer specific violations of these correlations.
Moreover, similar deviations are observed for other high
fragility polymers, regardless of their secondary relaxations.
Thus, the presence of a secondary relaxation cannot justify the
failure of the correlations. The above discovered correlations
with the chain fragility (instead of segmental one) suggest
very different explanation that is discussed in Sec. IV C.
C. The role of ergodicity and intra-molecular energy
barriers in behavior of segmental dynamics
To interpret the presented above results, we need to
emphasize that there might be two important lengths and
time scales controlling structural relaxation and macroscopic
properties. There is a molecular scale, and there is a scale of
structural relaxation. The latter is a cooperative phenomenon
that involves a region with size Rcoop ∼ 1-2 nm4 that is larger
than or comparable to the molecular scale in the case of small
molecules. However, the molecular length scale is much larger
in the case of polymers with high MW. Also the time scale
of molecular motion (chain relaxation) in polymers is much
longer than their segmental relaxation time.
The fast picosecond relaxation happens on a time scale
much faster than the structural relaxation and involves local
fluctuations. So, the quasielastic intensity that reflects the
behavior of this fast local process correlates well with
the behavior of structural relaxation even in highly fragile
polymers (Fig. 3(b)). Apparently QES intensity might be
controlled by the same property, e.g., frustration in packing
on length and time scale of segmental dynamics, or even on
shorter scales. The segmental relaxation sets up the potential
landscape, or local structural environment, in which the local
fast relaxations take place.
On the other hand, macroscopic elastic constants and the
amplitude of their fluctuations (the latter defines the amplitude
of the boson peak39,66–68), the non-ergodicity parameter α, and
many other properties are defined by the molecular (or some
intermediate) length scale. The latter is comparable to the
scale of structural relaxation in small molecular systems,
covalent-bonded glasses, and oligomers. As a result, these
properties in molecular and covalent-bonded systems usually
show good correlations to the behavior of structural relaxation
(fragility) of the liquid, regardless whether we estimate it from
τα(T) or viscosity.
The situation is different in the case of polymers. Due to
chain connectivity, their molecular length scale can be much
larger than the scale of segmental relaxation. We assume
that the relevant for the glass transition length scale can be
determined from the molecular weight dependence of Tg.
This molecular weight dependence of segmental dynamics
saturates at some finite MW (Fig. 7). In the case of rigid
polymers, such as PS, Tg, and segmental dynamics show
molecular weight dependence up to MW ∼ 30k-100k.11,18,34
Fragility of segmental dynamics in PS also saturates at similar
154901-7 Dalle-Ferrier et al.
FIG. 7. Molecular weight dependence of Tg scaled by Tg of high molecular
weight samples for flexible polymer polydimethylsiloxane (PDMS) (open
circles—data from Ref. 89, closed circles—data from Ref. 69) and for rigid
polymer PS (open triangles—data from Ref. 45 and closed triangles—data
from Ref. 89). The upper X-axis shows R g of the PS polymer, and the vertical
dashed line marks the expected size of regions involved in the structural
relaxation.
MW,18 confirming the importance of this scale for the glass
transition. It is interesting that MW dependence of density,
mechanical modulus, and some other macroscopic properties
in PS also saturates at about the same MW.11 It has been
suggested in Ref. 69 that this MW corresponds to the chain
length when the molecule approaches the state of the Gaussian
coil. Once this state is achieved, molecular weight dependence
for most of the properties saturates. We note that long range
chain connectivity may influence Tg indirectly via influence on
static quantities (e.g., Kuhn segment, as in theory of Ref. 70,
or density, elastic constants), or in a more explicit dynamical
way, in particular related to intra-chain barriers, or both. At
present, we do not have an exact answer which mechanism is
responsible for the increase of Tg. However, whatever is the
mechanism, the sensitivity of the segmental relaxation time
to MW up to MW ∼ 30k-100k in PS suggests comparatively
large length scale affecting segmental dynamics, on the order
of several nanometers.
We emphasize that individual segmental relaxation events
usually happen on the lengths scale ∼2-3 nm.71–74 However, PS
with molecular weight ∼30k-100k has the radius of gyration
Rg ∼ 7-15 nm, far above the characteristic scale of segmental
relaxation event. It means that on the time and length scale
of segmental dynamics, the system might remain non-ergodic
because the chain cannot access all possible conformational
states. As a result, a part of configurational entropy in rigid
polymers corresponds to the states inaccessible on the time
scale of segmental dynamics. Following this idea, we assume
that the total configurational entropy Sc consists of two
parts, Sc = Sacc + Sin, where Sacc corresponds to the entropy
of accessible configurations and Sin to inaccessible ones.
This leads to a stronger decrease of the accessible entropy
Sacc in comparison with the decrease of total entropy with
decrease in temperature. As a result, segmental dynamics
shows much sharper temperature variation than one expected
J. Chem. Phys. 145, 154901 (2016)
according to the Adam-Gibbs relationship with total Sc .
The idea of subtraction of an excess contribution from the
configurational entropy is similar to the proposed in Ref. 44
idea to subtract the contribution of secondary relaxations.
However, we ascribe this excess entropy to inaccessible on
time scale of segmental dynamics part of the configurational
space. Increase of molecular weight of these rigid polymers
leads to an increase of the inaccessible configurational space
Sin in the total Sc , and to steeper temperature variations
of segmental dynamics (higher fragility). This explains the
strong increase in fragility of these relatively rigid polymers
with increase in MW.
Thermodynamic measurements usually present the total
entropy and this explains why the Adam-Gibbs relationship
fails for segmental relaxation in many rigid polymers.44
At the same time, dynamics on the chain time scale have
access to all the conformational states and explore the total
configurational space. As a result, the temperature dependence
of viscosity follows well the Adam-Gibbs relationship even
in rigid polymers (Fig. 5(b)). This also explains why many
other properties would correlate to the behavior of molecular
scale dynamics rather than to the behavior (fragility) of
segmental dynamics. The latter does not reflect the behavior
of completely ergodic system, while most of the discussed
properties are presented for ergodic conditions.
In the case of flexible polymers, such as polydimethyl-
siloxane (PDMS), PIB, and PB, molecular weight dependence
of Tg, modulus and segmental dynamics saturates already at
MW ∼ 2.5-5k (Fig. 7).18 This corresponds to molecular size
Rg ∼ 1.5-2.5 nm, comparable to the characteristic length scale
of segmental dynamics. Thus, in these flexible polymers the
nearly total relevant configurational space might be accessible
on the time and length scale of the segmental dynamics. As
a result, the behavior of segmental dynamics is similar to
the behavior of structural relaxation in small molecules and
most of the correlations hold, including the Adam-Gibbs
relationship (Eq. (5)).4 Moreover, there is no significant
difference between steepness of temperature dependence of
chain and segmental relaxations in these flexible polymers.63
To explore how the chain rigidity might affect temperature
dependence of the configurational entropy, we consider a
simple worm-like chain with Ns segments. Let a segment in a
flexible chain has p possible states. In a worm-like rigid chain
with a small bond angle α, a segment can actually explore
only a small part of this space, with the number of states
∼αp. For the whole chain, this number will be (αp) N s . The
total number of available states for a melt with the Nc chains
will be
Γ ∼ [k(αp) N s ] Nc , (6)
where k takes into account the degrees of freedom of the
chain center of mass. Note that Ns Nc = N is the total number
of segments. The configurational entropy of the system per
segment is equal to
Sc ∼ (1/N)ln Γ ∼ [ln p − ln(1/α) + (1/Ns )ln k]. (7)
Taking into account that a small bond angle α can be connected
to the characteristic ratio as C∞ ≈ 4/α2, the configurational
entropy per segment of a melt of rigid polymer chains can be
154901-8 Dalle-Ferrier et al.
rewritten in the form
Sc ∼ S0 + B/MW − (1/2)ln C∞, (8)
where MW = Ns M0 is the molecular weight of a chain,
M0 is the mass of a segment, B = M0ln k is a constant,
and S0 is the total configurational entropy per segment
available for infinitely long flexible chains. In this rough
estimate, we neglected the difference between the finite chain
length characteristic ratio Cn and C∞. The terms B/MW and
(1/2)ln C∞ reflect the contribution of the finite molecular
weight and the chain rigidity. The latter term is a rough
estimate of the difference in configuration entropy of rigid
chains on two different time scales—one is the segmental
relaxation time and another is the chain relaxation time or
time scale of viscosity. We assumed that in the latter case
chains are totally relaxed, so basically the static flexibility
and, respectively, C∞ are relevant. In more general case of
arbitrary times, the dynamical flexibility that depends on
the intra-chain barrier heights should be considered. We do
not study this general case in the present work. However,
we would like to mention that the dynamical and static
flexibilities correlate in many polymers, e.g., torsional barriers
are significantly smaller in flexible polydimethylsiloxane than
in rigid PS or PMMA.75–77 Moreover, model simulations
presented in Refs. 78–80 clearly revealed that increase in
chain rigidity increases both Tg and fragility of model
chains. In terms of the Adam-Gibbs Equation (5), the
derived expression means that for the description of the
viscosity’s temperature dependence one should use the
configuration entropy that takes into account all possible
chain states. In contrast, for segmental relaxation time, the
term of the order of (1/2)ln C∞ should be subtracted from
the configuration entropy due to inaccessibility of some
conformational states in a rigid chain on the segmental
relaxation time scale. We note that in polystyrene C∞ = 9.5, so
(1/2)Rln C∞ ∼ 1.12R ∼ 9.3 J/mol K, which is comparable to
the excess entropy estimated in PS at the Vogel-Fulcher-
Tammann (FVT) temperature T0 : Sc (T0) = 6.7 J/mol K.44
This simple consideration provides an explanation of how
the chain rigidity might affect the temperature dependence of
the segmental relaxation time. We note that the Adam-Gibbs
approach is not the only possible way to explain the high
fragility and other peculiarities of rigid polymers. We chose
it as an illustration of a possible role of the chain rigidity.
It also helps to rationalize the observation that relaxation on
the molecular length/time scale brings fragility of polymers
to the values usual for small molecules and restore many
correlations.
The proposed explanation is consistent with the recent
idea suggested by Colmenero35 on the specific role of chain
rigidity in the behavior of segmental dynamics (fragility).
Analysis of a bead-spring model with different intra-molecular
bending and torsional force constants78–80 revealed an increase
in Tg and fragility with increasing equilibrium rigidity
measured by the characteristic ratio C10.
Recently, the elastically collective nonlinear Langevin
equation (ECNLE) theory of dynamics in supercooled
molecular liquids was extended to polymers.70,81 It correctly
predicts fragilities ∼80 of molecular liquids and short chain
J. Chem. Phys. 145, 154901 (2016)
polymers. With no adjustable parameters, it reproduces
Tg(MW) and some other parameters in PS and other polymers.
The Kuhn length which depends on equilibrium chain stiffness
via chemistry and the MW dependent characteristic ratio
C(MW) is one of the key parameters of the theory that
determine the MW dependence of dynamics. The results of
the theory and their agreement with experimental data and with
the bead-spring simulations in Ref. 35 for Tg and fragility as a
function of C(MW) suggest that the MW dependence of Tg is
associated with an equilibrium measure of backbone rigidity,
which does not require any direct information about intra-
chain barriers. However, the current ECNLE theory fails to
describe very high fragilities of rigid polymers. We suspect that
the weakness of the current ECNLE theory is the consideration
of a polymer as a number of disconnected segments. So, the
theory does not consider chain connectivity which might be
the key to extremely high fragility of rigid polymers. We
expect that including explicitly chain connectivity will help
the ENCLE theory to describe high fragility of rigid polymers.
The difference in temperature dependence of the chain and
segmental modes in polymers is known for decades.82,83 It has
been ascribed to dynamic heterogeneities that are important
on segmental relaxation time and length scale and lead to
a decoupling of structural relaxation and diffusion.62,84 We
emphasize that the model proposed here does not contradict to
this earlier picture. The dynamic heterogeneities are obviously
averaged out on the length scale of the chain dynamics.62,84
However, how to combine these two approaches remains
unclear. We also note that in the presented model the high
fragility of rigid polymers is connected to a non-ergodic state
of the polymer melt on the segmental relaxation time scale.
As one of the consequences, a possibility of aging phenomena
might be expected even in the supercooled liquid state of
polymers. Indication of such effects indeed was observed in
polymeric melts, e.g., Ref. 85.
It is interesting also whether similar effects and higher
fragility can be observed in non-polymeric systems with
intra-molecular degrees of freedom that have relaxation
time longer than structural relaxation time. In that case,
a part of configurational space will not be accessible on
time scale of structural relaxation and specific behavior
might emerge. Another interesting extension of the scenario
proposed here can be glass-forming liquids that contain
mixtures of chemically different components, e.g., alkaline-
borate or alkaline-silicate systems. In that case, chemical
fluctuations on time and length scale larger than structural
relaxation will also affect the configurational entropy. As
a result, these liquids might be again non-ergodic on the
time scale of structural relaxation and their properties will
not follow the behavior characteristic for chemically simple
molecular systems.26,55
V. CONCLUSIONS
Our detailed analysis shows that many properties of
polystyrene do not have strong variations with increase
in molecular weight, while steepness of the temperature
dependence of segmental dynamics (fragility) almost doubles.
This leads to the failure of several correlations established
154901-9 Dalle-Ferrier et al.
earlier for non-polymeric systems. PS oligomers behave the
same as other small molecules and their properties follow the
correlations established for non-polymeric systems. However,
increase in molecular weight of PS leads to a progressive
failure of the correlations. This observation is not specific for
PS only, and the same violations of correlations are known for
several other high fragility (relatively rigid) polymers.
Our surprising discovery is that using the steepness of
the chain relaxation (viscosity) of PS (instead of segmental
fragility) essentially restores all the correlations, including
the Adam-Gibbs relationship to thermodynamic fragility
(Fig. 5(b)). It means that using relaxation behavior in polymers
on their molecular time scale leads to the same correlations
as has been observed for non-polymeric systems. Based on
this observation, we speculate that rigid polymers remain non-
ergodic on time and length scale of the segmental dynamics. As
a result, a part of configurational space is not accessible on this
time and length scale, and relevant for segmental dynamics
accessible configurational entropy varies much faster with
temperature than the total configurational entropy. This leads
to significantly higher fragility of these polymers, but it does
not reflect in many other properties defined by the ergodic state
of the material. Relaxations on much larger length and time
scales (e.g., chain relaxation) probe the total configurational
entropy and exhibit significantly lower fragility, which is
comparable to the fragility of molecular van der Waals liquids.
In the case of flexible polymers, the relevant for segmental
dynamics length scale (estimated, e.g., from the molecular
weight dependence of Tg) is much shorter and is comparable
to the size involving structural relaxation (Fig. 7). As a result,
essentially all the configurational states are accessible on
the time and length scale of segmental dynamics in flexible
polymers. This is the reason these polymers do not show
extremely high fragility of segmental dynamics, temperature
behavior of their segmental and chain dynamics do not differ
much, and most of the correlations known for non-polymeric
systems hold in the case of flexible polymers.
It would be important to provide similar analysis for
other rigid polymers and verify the proposed qualitative
picture that finally might answer the question: Why many
polymers are so fragile? Moreover, it would be important to
verify this naïve qualitative picture and speculations using
more rigorous theoretical approach, such as generalized
entropy theory14–17 or elastically collective nonlinear Langevin
equation theory.70,81,86
ACKNOWLEDGMENTS
US team acknowledges the financial support from NSF
Polymer program (Grant No. DMR-1408811).
1C. A. Angell, Science 319, 582 (2008).
2P. G. Debenedetti and F. H. Stillinger, Nature 410, 259 (2001).
3C. A. Angell, K. L. Ngai, G. B. McKenna, P. F. McMillan, and S. W. Martin,
J. Appl. Phys. 88, 3113 (2000).
4G. Adam and J. H. Gibbs, J. Chem. Phys. 43, 139 (1965).
5D. Kivelson and G. Tarjus, J. Non-Cryst. Solids 235-237, 86 (1998).
6C. A. Angell, J. Non-Cryst. Solids 73, 1 (1985).
7R. Böhmer, K. L. Ngai, C. A. Angell, and D. J. Plazek, J. Chem. Phys. 99,
4201 (1993).
J. Chem. Phys. 145, 154901 (2016)
8C. Gainaru, A. L. Agapov, V. Fuentes-Landete et al., Proc. Natl. Acad. Sci.
U. S. A. 111, 17402 (2014).
9A. L. Agapov, A. I. Kolesnikov, V. N. Novikov, R. Richert, and A. P. Sokolov,
Phys. Rev. E 91, 022312 (2015).
10D. Huang and G. B. McKenna, J. Chem. Phys. 114, 5621 (2001).
11P. G. Santangelo and C. M. Roland, Macromolecules 31, 4581 (1998).
12C. M. Roland and R. Casalini, J. Chem. Phys. 119, 1838 (2003).
13Y. Ding, V. N. Novikov, A. P. Sokolov, A. Cailliaux, C. Dalle-Ferrier, C.
Alba-Simionesco, and B. Frick, Macromolecules 37, 9264 (2004).
14E. B. Stukalin, J. F. Douglas, and K. F. Freed, J. Chem. Phys. 131, 114905
(2009).
15J. Dudowicz, K. F. Freed, and J. F. Douglas, J. Phys. Chem. B 109, 21285
(2005).
16J. Dudowicz, K. F. Freed, and J. F. Douglas, J. Phys. Chem. B 109, 21350
(2005).
17J. Dudowicz and K. F. Freed, J. Chem. Phys. 124, 064901 (2006).
18K. Kunal, C. G. Robertson, S. Pawlus, S. F. Hahn, and A. P. Sokolov,
Macromolecules 41, 7232 (2008).
19A. L. Agapov, V. N. Novikov, and A. P. Sokolov, “Fragility and other
properties of glass-forming liquids: Two decades of puzzling correlations,”
in Fragility of Glass-forming Liquids, edited by A. L. Greer, K. Kelton, and
S. Sastry (Hindustan Book Agency, Gurgaon, 2014), Chap. 4.
20A. P. Sokolov, E. Rössler, A. Kisliuk, and D. Quitmann, Phys. Rev. Lett. 71,
2062 (1993).
21K. Niss and C. Alba-Simionesco, Phys. Rev. B 74, 024205 (2006).
22K. Niss, C. Dalle-Ferrier, B. Frick, D. Russo, J. Dyre, and C. Alba-
Simionesco, Phys. Rev. E 82, 021508 (2010).
23T. Scopigno, G. Ruocco, F. Sette, and G. Monaco, Science 302, 849 (2003).
24T. Scopigno, D. Cangialosi, and G. Ruocco, Phys. Rev. B 81, 100202(R)
(2010).
25V. N. Novikov and A. P. Sokolov, Nature 431, 961 (2004).
26V. N. Novikov, Y. Ding, and A. P. Sokolov, Phys. Rev. E 71, 061501 (2005).
27J. C. Dyre, N. B. Olsen, and T. Christensen, Phys. Rev. B 53, 2171 (1996).
28J. C. Dyre, T. E. Christensen, and N. B. Olsen, J. Non-Cryst. Solids 352,
4632 (2006).
29J. C. Dyre, Rev. Mod. Phys. 78, 953 (2006).
30K. Niss, C. Dalle-Ferrier, V. M. Giordano, G. Monaco, B. Frick, and C. Alba-
Simionesco, J. Chem. Phys. 129, 194513 (2008).
31K. Niss, C. Dalle-Ferrier, V. M. Giordano, G. Monaco, B. Frick, and C. Alba-
Simionesco, J. Chem. Phys. 131, 249902 (2009).
32L. Larini, A. Ottochian, C. De Michele, and D. Leporini, Nat. Phys. 4, 42
(2008).
33V. N. Novikov and A. P. Sokolov, Phys. Rev. B 74, 064203 (2006).
34C. Dalle-Ferrier, K. Niss, A. P. Sokolov, B. Frick, J. Serrano, and C. Alba-
Simionesco, Macromolecules 43, 8977 (2010).
35J. Colmenero, J. Phys.: Condens. Matter 27, 103101 (2015).
36L. Hong, B. Begen, A. Kisliuk, C. Alba-Simionesco, V. N. Novikov, and
A. P. Sokolov, Phys. Rev. B 78, 134201 (2008).
37L. D. Landau and E. M. Lifshitz, Statistical Physics, 3rd ed. (Butterworth-
Heinemann, Oxford, 1980).
38V. K. Malinovsky and A. P. Sokolov, Solid State Commun. 57, 757 (1986).
39U. Buchenau, M. Prager, R. Nucker, A. J. Dianoux, N. Ahmad, and W. A.
Phillips, Phys. Rev. B 34, 5665 (1986).
40E. Courtens, M. Foret, B. Hehlen, and R. Vacher, Solid State Commun. 117,
187 (2001).
41R. Gray, G. Harrison, and J. Lamb, Proc. R. Soc. London, A 356, 77 (1977).
42P. Zoller and D. Walsh, Standard Pressure-Volume-Temperature Data for
Polymers (Technomic Publishing Co., Lancaster, PA, 1995).
43R. Casalini, C. M. Roland, and S. Capaccioli, J. Chem. Phys. 126, 184903
(2007).
44D. Cangialosi, A. Alegría, and J. Colmenero, Europhys. Lett. 70, 614 (2005).
45P. G. Santangelo and C. M. Roland, Phys. Rev. B 58, 14121 (1998).
46J. Jäckle, in Amorphous Solids: Low-Temperature Properties, edited by
W. A. Phillips (Springer, Berlin, 1981).
47V. Z. Gochiyaev, V. K. Malinovsky, V. N. Novikov, and A. P. Sokolov, Philos.
Mag. B 63, 777 (1991).
48A. P. Sokolov, U. Buchenau, W. Steffen, B. Frick, and A. Wischnewski, Phys.
Rev. B 52, R9815 (1995).
49S. V. Adichtchev, N. V. Surovtsev, J. Wiedersich, A. Brodin, V. N. Novikov,
and E. A. Rössler, J. Non-Cryst. Solids 353, 1491 (2007).
50S. V. Adichtchev, St. Benkhof, Th. Blochowicz, V. N. Novikov, E. Rössler,
Ch. Tschirwitz, and J. Wiedersich, Phys. Rev. Lett. 88, 055703 (2002).
51W. Götze, Complex Dynamics of Glass-Forming Liquids: A Mode-Coupling
Theory (Oxford University Press, Oxford, 2012).
52R. Richert and C. A. Angell, J. Chem. Phys. 108, 9016 (1998).
154901-10 Dalle-Ferrier et al.
53U. Mohanty, N. Craig, and J. T. Fourkas, J. Chem. Phys. 114, 10577 (2001).
54L. Battezzati, Mater. Trans. 46, 2915 (2005).
55S. N. Yannopolous and G. P. Johary, Nature 442, E7 (2006).
56P. Richet, Geochim. Cosmochim. Acta 48, 471 (1984).
57J. H. Magill, J. Chem. Phys. 47, 2802 (1967).
58C. M. Roland, S. Capaccioli, M. Lucchesi, and R. Casalini, J. Chem. Phys.
120, 10640 (2004).
59A. Scala, F. W. Starr, E. La Nave, F. Sciortino, and H. E. Stanley, Nature 406,
166 (2000).
60S. Sastry, Nature 409, 164 (2001).
61K. Kishimoto, H. Suga, and S. Seki, Bull. Chem. Soc. Jpn. 46, 3020 (1973).
62A. P. Sokolov and K. S. Schweizer, Phys. Rev. Lett. 102, 248301 (2009).
63Y. Ding and A. P. Sokolov, Macromolecules 39, 3322 (2006).
64K. L. Ngai, D. J. Plazek, and C. M. Roland, Macromolecules 41, 3925 (2008).
65A. Agapov, Y. Wang, V. N. Novikov, and A. P. Sokolov, Macromolecules
(unpublished).
66E. Duval, A. Boukenter, and T. Achibat, J. Phys. Condens. Matter 2, 10227
(1990).
67W. Schirmacher, B. Schmid, C. Tomaras, G. Viliani, G. Baldi, G. Ruocco,
and T. Scopigno, Phys. Status Solidi C 5, 862 (2008).
68D. Quitmann, M. Soltwisch, and G. Ruocco, J. Non-Cryst. Solids 203, 12
(1996).
69Y. Ding, A. Kisliuk, and A. P. Sokolov, Macromolecules 37, 161 (2004).
70S. Mirigian and K. S. Schweizer, Macromolecules 48, 1901 (2015).
71M. D. Ediger, C. A. Angell, and S. R. Nagel, J. Phys. Chem. 100, 13200
(1996).
72J. Korus, E. Hempel, M. Beiner, S. Kahle, and E. Donth, Acta Polym. 48,
369 (1997).
J. Chem. Phys. 145, 154901 (2016)
73U. Tracht, M. Wilhelm, A. Heuer, H. Feng, K. Schmidt-Rohr, and H. W.
Spiess, Phys. Rev. Lett. 81, 2727 (1998).
74M. D. Ediger, Annu. Rev. Phys. Chem. 51, 99 (2000).
75J. S. Smith, O. Borodin, and G. D. Smith, J. Phys. Chem. B 108, 20340
(2004).
76S. Arrese-Igor, A. Arbe, B. Frick, and J. Colmenero, Macromolecules 44,
3161 (2011).
77B. C. Laskowski, R. L. Jaffe, and A. Komornicki, in Applied Quantum
Chemistry, edited by V. H. Smith, Jr. et al. (D. Reidel Publishing Company,
Dordrecht, 1986), pp. 347–359.
78M. Bernabei, A. J. Moreno, and J. Colmenero, Phys. Rev. Lett. 101, 255701
(2008).
79M. Bernabei, A. J. Moreno, and J. Colmenero, J. Chem. Phys. 131, 204502
(2009).
80R. Kumar, M. Goswami, B. G. Sumpter, V. N. Novikov, and A. P. Sokolov,
Phys. Chem. Chem. Phys. 15, 4604 (2013).
81S. Mirigian and K. S. Schweizer, J. Chem. Phys. 140, 194507 (2014).
82K. L. Ngai and D. J. Plazek, Rubber Chem. Technol. 68, 376 (1995).
83D. J. Plazek, J. Phys. Chem. 69, 3480 (1965).
84M. D. Ediger and P. Harrowell, J. Chem. Phys. 137, 080901 (2012).
85P. G. Santangelo and C. M. Roland, J. Rheol. 45, 583 (2001).
86K. S. Schweizer and E. J. Saltzman, J. Phys. Chem. B 108, 19729
(2004).
87A. P. Sokolov, V. N. Novikov, and A. Kisliuk, Philos. Mag. 87, 613
(2007).
88A. C. Ling and J. E. Willard, J. Phys. Chem. 72, 1918 (1968).
89J. Hintermeyer, A. Herrmann, R. Kahlau, C. Goiceanu, and E. A. Rössler,
Macromolecules 41, 9335 (2008).