1 Low-temperature thermal properties of an hyperaged geological glass Tomás Pérez-Castañeda 1 *, Rafael J Jiménez-Riobóo 2 and Miguel A Ramos 1 (1) Laboratorio de Bajas Temperaturas, Departamento de Física de la Materia Condensada, Condensed Matter Physics Center (IFIMAC) and Instituto Nicolás Cabrera, Universidad Autónoma de Madrid, E-28049 Madrid, SPAIN (2) Instituto de Ciencia de Materiales de Madrid, Consejo Superior de Investigaciones Científicas (ICMM-CSIC), E-28049 Madrid, Spain. *Corresponding author: [email protected]Abstract. We have measured the specific heat of amber from the Dominican Republic, an ancient geological glass about 20 million years old, in the low-temperature range 0.6 K ≤ T ≤ 26 K, in order to assess the effects of its natural stabilization (hyperageing) process on the low-temperature glassy properties, i.e. boson peak and two-level systems. We have also conducted Modulated Differential Scanning Calorimetry experiments to characterize the thermodynamic state of our samples. We found that calorimetric curves exhibit a huge ageing signal ΔH ≈ 5 J/g in the first upscan at the glass transition T g = 389 K, that completely disappears after heating up (rejuvenating) the sample to T = 395 K for three hours. To independently evaluate the phonon contribution to the specific heat, Brillouin spectroscopy was performed in the temperature range 80 K ≤ T ≤ 300 K. An expected increase in the Debye level was observed after rejuvenating the Dominican amber. However, no significant change was observed in the low-temperature specific heat of glassy amber after erasing its thermal history: both its boson peak (i.e., the maximum in the C p /T 3 representation) and the density of tunnelling two-level systems (i.e., the C p ∼ T contribution at the lowest temperatures) remained essentially the same. Also, a consistent analysis using the soft- potential model of our C p data and earlier thermal conductivity data found in the literature further supports our main conclusion, namely, that these glassy “anomalous”
26
Embed
Low-temperature thermal properties of a hyperaged geological glass
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
1
Low-temperature thermal properties of an hyperaged geological glass
Tomás Pérez-Castañeda1*, Rafael J Jiménez-Riobóo2 and Miguel A Ramos1
(1) Laboratorio de Bajas Temperaturas, Departamento de Física de la Materia
Condensada, Condensed Matter Physics Center (IFIMAC) and Instituto Nicolás
Cabrera, Universidad Autónoma de Madrid, E-28049 Madrid, SPAIN
(2) Instituto de Ciencia de Materiales de Madrid, Consejo Superior de
With all above data, one is finally able to determine the Debye phonon contribution to the low-
temperature specific heat, CDebye ≡ cD · T 3. We obtain Debye coefficients cDprs = 15.5 ± 0.8
μJ/g·K4 and cDrej = 18.0 ± 0.9 μJ/g·K4, for the pristine and rejuvenated amber, respectively. A
small but clear increase in the Debye level is thus observed after rejuvenation, as a result of
driving the substance back to a conventional, non-densified glass state. This is directly related to
the corresponding decrease in both sound velocity and mass density when thermally
rejuvenating the sample (see figure 2(a)), as a consequence of the produced softening in the
molecular lattice.
14
3.3 Effects of hyperageing on the low-temperature specific heat
Our main goal is to study the influence of deep stabilization and densification processes, due to
extremely long ageing in Dominican amber, on the universal low-temperature glassy properties.
For this purpose, low-temperature specific heat was first measured for a pristine amber sample
in the range 0.7 K ≤ T ≤ 26 K. After this, a thermal rejuvenation process as described above was
applied to the sample, and again the specific-heat curve was obtained for comparison in the
range 0.6 K ≤ T ≤ 26 K. Data are shown in figure 3, in the usual reduced Cp(T) / T 3 vs T plot.
Corresponding Debye levels, determined from the measured sound velocities and mass densities
as detailed in the previous section, are also shown in figure 3 by thick solid lines. As expected, a
modest increase in the Debye coefficient after rejuvenation is observed. The most remarkable
result is however that both Cp(T) / T 3 curves are strikingly similar, even exhibiting the same
position for the minimum, at Tmin=1.3 K, and the maximum (boson peak), at Tmax = 3.7 K, the
only difference being a small but clear decrease in the height of the boson peak after
rejuvenation.
In order to account quantitatively for any possible differences between the two samples, the
experimental data for the low-temperature specific heat are analysed and discussed in the
framework of the SPM [22–28]. For the sake of simplicity, we will only consider the most basic
version of the SPM [24–27], which does not account for the whole of the boson peak and other
glassy properties at higher temperatures, that has been however addressed in further extensions
of this model [29,30]. In the SPM, low-frequency dynamics of glasses is assumed to be
governed by the coexistence of ordinary acoustic (phonon-like) lattice vibrations and some
additional quasilocalised vibrations (usually called ‘soft modes’). The latter contribution is
described within the SPM in terms of asymmetric quartic potentials, from which the specific
heat is found to be approximately linear in temperature, Cp ~ T, for kBT << W (in the region
dominated by double-well potentials), W being the average characteristic energy of the soft
modes, whereas Cp ~ T 5 dominates at kBT > W (single-well potentials) [24–28]. Hence in the
lowest-temperature limit T < 1 K, the TLS contribution (Cp ~ T) is dominant, and the SPM
15
essentially agrees [28] with the TM [18–20]. In the typical representation Cp /T 3, this TLS
region is thus seen as an upturn at the lowest temperatures, followed with increasing
temperature by a minimum Tmin and a rapid increase Cp ~ T 5 [28] produced by the quasi-
harmonic soft-mode vibrations responsible for the lower-energy side, T < Tmax, of the boson
peak.
1 100
10
20
30
40
50
60
70
T max
= 3.
7 K
Pristine Rejuvenated
C P / T
3 ( µJ
/ g
· K4 )
Temperature (K)
CDebye pristine
CDebye rejuvenated
SPM
T min=
1.3
K
Figure 3. Cp(T) / T 3 versus T plot for pristine (circles) and rejuvenated (squares) states of the same
Dominican amber sample. Both curves exhibit the same position for the minimum, Tmin=1.3 K, and
maximum (boson peak), Tmax = 3.7 K, of Cp(T) / T 3, the only difference being a small decrease in the
height of the boson peak after rejuvenation. The corresponding Debye levels, obtained from the measured
elastic constants, are indicated by thick solid lines. As expected, a small increase in the Debye coefficient
after rejuvenation is observed. The two (almost indistinguishable between them) dashed lines are the
corresponding total SPM specific-heat curves below the boson peak, according to equation (5), obtained
as the sum of the elastoacoustically measured Debye levels for each case (thick solid lines) and the SPM
linear fits (see text).
16
Therefore, the simplest but still reasonable fit of the low-temperature specific heat of glasses is
just a quadratic polynomial fit in a Cp / T versus T 2 representation, as suggested elsewhere
[36,37]:
Cp (T) = cTLS T + cD T 3 + csm T 5 (5)
where the coefficients in equation (5) correspond to the contribution of the TLS, Debye lattice
vibrations and quasiharmonic soft modes, respectively. In order to produce meaningful
estimations of the coefficients using this fit, the temperature range to be applied must be
approximately 0 < T < 3/2 Tmin, [36,37] where Tmin is the temperature of the minimum in the
Cp/T 3 representation.
Previously, in section 3.2, we were able to determine the elastic Debye coefficient cD, which
should coincide with the corresponding coefficient obtained from calorimetric measurements
[37] by fitting data to equation (5). Indeed, the elastic and calorimetric determinations of cD are
found to be equal in our samples within experimental error. Nonetheless, since we have a
model-independent elastic determination of the cubic Debye coefficient cD, we will better
employ the procedure followed in [38] for butanol glasses and conduct the SPM analysis with
only two fitting parameters, cTLS and csm. In brief, one fixes the elastoacoustically measured cD
coefficient for each glass, and the excess specific heat is directly manifested after subtraction of
the corresponding cubic Debye contribution CDebye = cD T 3, by plotting (Cp−CDebye)/T vs T
4 in
order to obtain the missing T and T
5 coefficients of the SPM from simple straight-linear fits, as
shown in figure 4 and Table 1. Despite the larger data scatter due to the very sensitive chosen
axes, the expected linear behaviour is observed indeed below 2 K, what supports the validity of
the made SPM assumptions and analysis. In addition, we also show in the Cp/T 3 representation
in figure 3 the total SPM specific heat (dashed lines) below the boson peak, for the amber glass
in both cases, that is the sum of the elastoacoustically measured Debye contribution for each
glass (thick solid lines) and the SPM fitted contributions from figure 4, according to equation
(5).
17
0 2 4 6 8 10 12 140
20
40
60
80
100
(CP-
C Deby
e) / T
( µJ
/ g
· K2 )
T 4 (K4)
Pristine Rejuvenated
Figure 4. (Cp – CDebye)/T versus T
4 plot for pristine (circles) and rejuvenated (squares) Dominican amber,
in the low temperature range. A least-squares linear fit to the SPM is done (dashed lines) to obtain the
SPM parameters cTLS and csm. Fitting curves are (Cp – CDebye)/T = 14.7 μJ/g·K2 + 5.7 μJ/g·K6·T 4 for the
pristine amber and (Cp – CDebye)/T = 13.8 μJ/g·K2 + 5.1 μJ/g·K6·T 4 for the rejuvenated one. The observed
linear behaviour in the temperature range 0 K ≤ T ≤ 1.95 K supports the validity of the SPM assumptions.
Table 1. Comparison of the low-temperature parameters for the Dominican amber samples, both pristine
and rejuvenated. ρ(0 K): zero-temperature mass-density extrapolation using equation (3); vD: Debye
averaged velocity; cD: Debye phonon contribution to the specific heat determined from HRBS; Tmin, Tmax:
respective temperatures where the minimum and the maximum in Cp/T 3 occur; cTLS, csm: TLS (∝T) and
soft modes (∝T
5) coefficients of the specific heat calculated from the linear fits in figure 4 with the SPM.
Amber
glass
ρ(0 K)
(g/cm3)
vD
(m/s)
cD
(μJ/g·K4)
Tmin
(K)
Tmax
(K)
cTLS
(μJ/g·K2)
csm
(μJ/g·K6)
Pristine 1.090 1926 15.5 1.3 3.7 14.7 5.7
Rejuvenated 1.031 1865 18.0 1.3 3.7 13.8 5.1
18
From both figures 3 and 4, it is very clear that the highly densified, hyperaged amber retains all
glassy excitations at low temperatures (i.e. TLS and “boson-peak” soft modes), even
quantitatively. They appear in our case to be even slightly above those of the canonical,
rejuvenated amber, what is certainly striking.
Finally, we will make use of this SPM analysis of the low-temperature specific heat to
concurrently address and discuss earlier thermal-conductivity measurements at low
temperatures, also conducted in Dominican amber by Love and Anderson [31]. In figure 5, we
depict their thermal-conductivity data κ (T) [31] for pristine (as received) and rejuvenated (at
370 K) Dominican amber samples. Although their samples and ours could be somewhat
different (indeed, their reported glass-transition temperature is lower than that of our samples),
the thermal conductivity of glasses is not very much sensitive to material parameters [19,20]
and a joint analysis can be done. As it has been discussed in detail [27,28], the SPM main
parameter W/kB marks the crossover temperature below which the TLS dominate the low-energy
glassy dynamics and above which quasilocalised soft modes (eventually piling up at the boson
peak) do. To be more precise, this crossover is seen in the specific heat as the minimum Tmin in
Cp/T 3 (with W/kB ≈ 1.6 Tmin analytically [24], although also W/kB
≈ 1.8 Tmin was postulated from
some numerical calculations [28,39]), and in the thermal conductivity as a bend from the
quadratic temperature dependence at the lowest temperatures to the plateau region, hence as a
maximum Tmax,κ in κ/T (with W ≈ 1.6 Tmax,κ) [27]. Therefore, the main qualitative prediction of
the SPM is that the minimum Tmin in Cp/T 3 and the maximum Tmax,κ of κ/T should occur [27] at
(almost) the same temperature. The inset in figure 5 shows that for the pristine Dominican
amber one has Tmax,κ = 1.3 K (hence W = 2.1 K), which nicely agrees with our minimum Tmin =
1.3 K in Cp/T 3. Furthermore, it can also be seen in figure 5 that the thermal conductivity for the
rejuvenated amber (unfortunately not provided [31] in the whole temperature range) is again
very similar to that of the pristine sample. This supports our statement that the hyperaged glass
fully retains the universal glassy properties at low temperatures, which are not depleted by this
dramatic structural and thermodynamic relaxation.
19
0.01 0.1 1 10
10-4
10-3
10-2
10-1
κ (W
/m·K
)
T (K)
as receivedrejuvenated 370 K
SPM
0.01 0.1 1 100.00
0.01
0.02
κ/T
(W/m
·K2 )
T (K)
(= 1.3 K)Tmax
Figure 5. Thermal conductivity κ data versus temperature in a double logarithmic scale, measured by
Love and Anderson [31] for other pristine (as received) and rejuvenated (at 370 K) Dominican amber
samples. The inset shows the same data plotted as κ/T vs T, following the SPM method [27,28] to
determine a maximum Tmax,κ of κ/T and hence the basic SPM parameter W ≈ 1.6 Tmax,κ, which marks the
crossover from the TLS-dominated region (κ ∝ T 2) to the soft-mode one (plateau).The solid line is the
SPM [27,28] predicted curve for the fitted W.
4. Discussion
As briefly reminded in the Introduction, the two most characteristic types of “anomalous” low-
energy excitations in glasses and disordered solids, which pervade all their physical properties
at low temperatures and frequencies, are: (i) TLS (or tunnelling) states, and (ii) low-frequency
vibrations producing the “boson peak”. The former are well accepted to exist and account for
thermal, acoustic and dielectric glassy properties at temperatures below 1 K and energies below
1 meV, following the Tunnelling Model, though a full microscopic understanding is still lacking
[40]. The very nature of the latter is nowadays a vivid matter of debate and controversy, and
many theories and models [29,30,41–46] compete to explain low-frequency dynamics of glasses
featured by the boson peak. Furthermore, during the last forty years many different attempts
20
have been made (with unclear conclusions [47–52]) to know whether these low-temperature
glassy properties significantly depend or not on the impurities and the thermal history, and
therefore could be depleted by annealing or ageing processes, or rather are essential and
intrinsic properties of non-crystalline solids. Alternatively, some inelastic neutron-scattering
experiments in a mineral hyperquenched glass [53], seemed to point out that the boson peak is
greatly enhanced in the hyperquenched glass over that of the standard and the annealed glass.
The same trend was found in computer simulations of binary mixtures [6,41], where the
observed total maximum in g(ω)/ω
2 was found to decrease and shift to higher frequencies with
slower cooling rates.
Very recently, a much debated issue is whether or not the boson peak scales with the Debye
frequency and hence with the elastic medium constants, where again opposite views have been
published: in some cases, the Debye-reduced vibrational density of states plotted versus the
Debye-reduced frequency merge into a master curve, either when comparing hyperquenched to
annealed glasses [54], or measuring at different temperatures [55], or even during
polymerization of a reactive mixture [56], whereas in other cases [38,57,58] this scaling has
been found to fail. Moreover, some authors [59] have claimed that the boson peak in glasses is
the counterpart of the acoustic van Hove singularity in crystals, shifted to lower energy by
force-constants disorder. Then, similar to the transverse acoustic singularity in crystals, the
boson peak originates from the piling up of the acoustic states near the boundary of the pseudo-
Brillouin-zone.
The first key question is then: do excess glassy properties (i.e. TLS and the boson peak) strongly
depend on the thermal history of the glass, and hence on the reached excess enthalpy, and
eventually would disappear for ideal, extremely annealed or stabilized glasses, or rather are they
intrinsic properties of the glassy, non-crystalline state? Our findings in the extreme case of an
hyperaged geological glass, which retains −if not enhances− both its boson peak and its density
of TLS at low temperatures, undoubtedly state that these universal low-temperature glassy
21
excitations are robust against the thermal history, and intrinsic to the non-crystalline lattice. The
SPM analysis conducted to quantitatively assess those low-temperature glassy excitations in the
specific heat, as well as earlier thermal-conductivity data, further support our conclusion.
Such a clear statement cannot be made in relation to the Debye scaling, since both the Debye
level and the observed Cp/T3 boson peak do not vary strongly after the thermal rejuvenation.
Still, the slightly lower boson peak observed after rejuvenation, whereas the Debye level has
increased around 15%, points out that a Debye scaling of the boson peak is not fulfilled. One
may speculate that the unexpected decrease of the boson-peak height with thermal rejuvenation
could be related to some residual curing happening around 390 K, as said in section 3.1. The
fact that almost every physical property remains constant or behaves in the opposite direction,
suggests that its influence, if any, is not significant.
The remarkable enthalpy reduction observed in our MDSC characterization experiments, ΔH ≈
5 J/g, is comparable to those found in ultrastable thin film glasses [1–5] obtained by physical
vapour deposition, and provides further evidence that the expected huge structural relaxation
has indeed occurred in our pristine amber samples, and that our results are meaningful.
The modest but clear variations earlier found [47–52] in the low-temperature specific heat of
some glasses with different thermal history could be attributed, in our view, to the concurrent
variation of the elastic constants and the Debye contribution below the boson peak, which in
most cases were not measured. As a matter of fact, in the case of pure and dry B2O3 glass, very
different maxima in Cp/T3 for different thermal treatments were found to collapse in a single
curve after their corresponding Debye levels were subtracted [51].
Finally, during the writing of this paper, we became aware of a forthcoming publication by
Zhao et al. [60], who have investigated the dynamics of similar 20 million year old Dominican
amber via stress relaxation experiments above and below the glass-transition temperature.
Among other results, by fitting the relaxation time curve in the equilibrium liquid state to the
Vogel-Tamman-Fulcher function, they obtained a dynamic fragility index of m = 90. It is worth
22
emphasizing that Sokolov et al. [61,62] found, after analysing thirteen glass-forming
substances, that the relative weight of vibrational over relaxational excitations (i.e. the boson
peak) is larger for stronger than for fragile glass formers, in the well-known Angell’s
classification. One of the ways used by them to assess this relative weight was through the ratio
R of Debye-reduced specific-heat data at Tmin over that at Tmax, R = (Cp/T 3)min / (Cp/T
3)max. For
glass formers so fragile as to have m ∼ 90 (e.g. o-terphenyl or CKN), R was as large as 1, that is
a very shallow boson peak, whereas strong glass formers (low m indices) exhibited strong and
well-defined boson peaks, with R below 0.5. The boson peak of our pristine amber sample gives
R ≈ 0.52, which lies between the values of strong glass formers B2O3 and sodium-oxide silica,
with m ∼ 40. Therefore, amber does not follow the proposed correlation [61,62], its liquid being
dynamically fragile but its glass showing a strong vibrational boson peak.
5. Conclusions
We have studied the effects of the extremely long ageing process (an estimated period of time
between 15 and 45 million years) naturally occurred in Dominican amber, on the universal low-
temperature properties of this geological glass. With such an aim, we have measured the
specific heat at low temperatures of the same sample before and after a thermal rejuvenation
process to erase its dramatic thermal history. MDSC experiments confirmed the very large
enthalpy reduction geologically occurred in the pristine sample, in comparison to the
rejuvenated one. Brillouin-scattering experiments allowed us to determine both temperature-
dependent (longitudinal and transverse) sound velocities and refractive indices, and hence the
elastic Debye contributions to the specific heat needed to make a thorough analysis of the low-
temperature glassy excitations.
Somewhat strikingly, we have found that the pristine, hyperaged amber exhibits a strong boson
peak and a high density of tunnelling states, as high (if not even higher) as those of the
rejuvenated sample. Given the extreme structural and thermodynamic relaxation produced in
23
this glass prior to its rejuvenation, our observations should be robust enough as to reach a high
level of generality. The main conclusion is that the ubiquitous glassy excitations at low
frequencies or temperatures (TLS and the boson peak) are intrinsic to the non-crystalline state
and they are not depleted by long annealing or structural relaxation processes. These can
however influence the Debye-like elastic properties, and hence indirectly the total heat capacity.
Our results in this paradigmatic geological glass are also at odds with the previously proposed
Debye scaling of the boson peak and the correlation of the latter with the inverse of the fragility
index.
In order to test the reproducibility and general character of the presented results, similar
experiments on other amber glasses from different sources (i.e. with different chemical and
thermal histories) are currently underway.
Acknowledgements
The Laboratorio de Bajas Temperaturas (LBT-UAM) is an Associated Unit with the ICMM-
CSIC. This work was partially supported by the Spanish MINECO (FIS2011-23488 project, and
Consolider Ingenio Molecular Nanoscience CSD2007-00010 program) and by the Comunidad
de Madrid through program Nanobiomagnet (S2009/MAT-1726). T. P.-C. acknowledges
financial support from the Spanish Ministry of Education through FPU grant AP2008-00030 for
his PhD thesis. María José de la Mata is gratefully acknowledged for her technical support and
kind advice with MDSC measurements conducted at SIdI−UAM. M.A.R. is grateful to Greg
McKenna for sending us their article [60] prior to its publication.
24
References
[1] Swallen S F, Kearns K L, Mapes M K, Kim Y S, McMahon R J, Ediger M D, Wu T, Yu L and Satija S 2007 Science 315 353
[2] Kearns K L, Swallen S F, Ediger M D, Wu T and Yu L 2007 J. Chem. Phys. 127 154702
[3] Kearns K L, Swallen S F, Ediger M D, Wu T, Sun Y and Yu L 2008 J. Phys. Chem. B 112 4934
[4] Leon-Gutierrez E, Sepulveda A, Garcia G, Teresa Clavaguera-Mora M and Rodriguez-Viejo J 2010 Phys. Chem. Chem. Phys. 12 14693
[5] Sepúlveda A, Leon-Gutierrez E, Gonzalez-Silveira M, Rodríguez-Tinoco C, Clavaguera-Mora M and Rodríguez-Viejo J 2011 Phys. Rev. Lett. 107 025901
[6] Singh S, Ediger M D and De Pablo J J 2013 Nat. Mat. 12 1
[7] Stillinger F H 1995 Science 267 1935
[8] Debenedetti P G and Stillinger F H 2001 Nature 410 259
[9] Brawer S 1985 Relaxation in viscous liquids and glasses (Columbus, Ohio: Amer. Ceram. Soc.)
[10] Chang S S and Bestul A B 1972 J. Chem. Phys. 56 503
[11] Chang S S and Bestul A B 1974 J. Chem. Therm. 6 325
[12] Larmagnac J P, Grenet J and Michon P 1981 J. Non-Cryst. Solids 45 157
[13] Wang P, Song C and Makse H A 2006 Nat. Phys. 2 526
[14] Hay J N 1995 Pure and Appl. Chem. 67 1855
[15] Corezzi S, Fioretto D and Rolla P 2002 Nature 420 653
[16] Lambert J B and Poinar G O 2002 Accounts Chem. Research 35 628
[17] Grimaldi David A 1996 Amber, Resinite, and Fossil Resins ACS Symposium Series vol 617, Chapter 11 pp 203−217
[18] Phillips W A 1981 Amorphous solids: Low-temperature properties (Berlin: Springer)
[19] Anderson P W, Halperin B I and Varma C M 1972 Phil. Mag. 25 1
[20] Phillips W A 1972 J. Low Temp. Phys. 7 351
25
[21] See, for instance, the specific sessions devoted to this topic in the 6th International Discussion Meeting on Relaxations in Complex Systems (IDMRCS), held in Rome (Italy), and their selected papers published in: J. Non-Cryst. Solids 357 (2011) pp 501−556
[22] Karpov V, Klinger M and Ignat’ev F 1983 Zh. Eksp. Teor. Fiz 84 439
[23] Ill'in M A, Karpov V G and Parshin D A 1987 Sov. Phys. − JETP 65 165
[24] Buchenau U, Galperin Y M, Gurevich V L and Schober H R 1991 Phys. Rev. B 43 5039
[25] Buchenau U, Galperin Y M, Gurevich V L, Parshin D A, Ramos M A and Schober H R 1992 Phys. Rev. B 46 2798
[26] Parshin D A 1994 Phys. Solid State 36 991
[27] Ramos M A and Buchenau U 1997 Phys. Rev. B 55 5749
[28] Ramos M A and Buchenau U 1998 Tunnelling systems in amorphous and crystalline solids ed P Esquinazi (Berlin: Springer), Chapter 9 pp 527−569
[29] Gurevich V L, Parshin D A and Schober H R 2003 Phys. Rev. B 67 094203
[30] Parshin D A, Schober H R and Gurevich V L 2007 Phys. Rev. B 76 064206
[31] Love M S and Anderson A C 1991 J. Low Temp. Phys. 84 19
[32] Wert C A, Weller M, Schlee D and Ledbetter H 1989 J. Appl. Phys. 65 2493
[33] Pérez-Enciso E and Ramos M A 2007 Thermochim. Acta 461 50
[34] Krüger J K 1989 Optical Techniques to Characterize Polymer Systems, Studies in Polymer Science vol 5, ed K Seki & H Bässler (Amsterdam: Elsevier) pp 429–534
[35] Krüger J, Baller J, Britz T, Le Coutre A, Peter R, Bactavatchalou R and Schreiber J 2002 Phys. Rev. B 66 012206
[36] Ramos M A, Talón C, Jiménez-Rioboó R and Vieira S 2003 J. Phys.: Cond. Matt. 15 S1007
[37] Ramos M A 2004 Phil. Mag. 84 1313
[38] Hassaine M, Ramos M A, Krivchikov A I, Sharapova I V., Korolyuk O A and Jiménez-Riobóo R J 2012 Phys. Rev. B 85 104206
[39] Ramos M A, Gil L, Bringer A and Buchenau U 1993 Phys. Stat. Sol. (a) 135 477
[40] Esquinazi P 1998 Tunnelling systems in amorphous and crystalline solids (Berlin: Springer)
[41] Grigera T, Martin-Mayor V, Parisi G and Verrocchio P 2003 Nature 422 289
[42] Schirmacher W 2006 Eur. Phys. Lett. 73 892
26
[43] Ruocco G, Sette F, Di Leonardo R, Monaco G, Sampoli M, Scopigno T and Viliani G 2000 Phys. Rev. Lett. 84 5788
[44] Götze W and Mayr M 2000 Phys. Rev. E 61 587
[45] Grigera T, Martín-Mayor V, Parisi G and Verrocchio P 2001 Phys. Rev. Lett. 87 085502
[46] Taraskin S, Loh Y, Natarajan G and Elliott S 2001 Phys. Rev. Lett. 86 1255
[47] Lasjaunias J C, Penn G and Vandorpe M 1980 Phonon Scattering in Condensed Matter ed H Maris (New York: Plenum) pp 25–28
[48] Löhneysen H v, Rüsing H and Sander W 1985 Z. Phys. B - Condensed Matter 60 323
[49] Ahmad N, Hutt K and Phillips W 1986 J. Phys. C: Sol. State Phys. 19 3765
[50] Pérez-Enciso E, Ramos M A and Vieira S 1997 Phys. Rev. B 56 32
[51] Inamura Y, Arai M, Yamamuro O, Inaba A, Kitamura N, Otomo T, Matsuo T, Bennington S M and Hannon A C 1999 Phys. B 263−264 299
[52] Calemczuk R, Lagnier R and Bonjour E 1979 J. Non-Cryst. Solids 34 149
[53] Angell C, Yue Y, Wang L, Copley J R, Borick S and Mossa S 2003 J. Phys.: Cond. Matt. 15 1051
[54] Monaco A, Chumakov A, Yue Y-Z, Monaco G, Comez L, Fioretto D, Crichton W and Rüffer R 2006 Phys. Rev. Lett. 96 1051
[55] Baldi G, Fontana A, Monaco G, Orsingher L, Rols S, Rossi F and Ruta B 2009 Phys. Rev. Lett. 102 195502
[56] Caponi S, Corezzi S, Fioretto D, Fontana A, Monaco G and Rossi F 2009 Phys. Rev. Lett. 102 27402
[57] Rufflé B, Ayrinhac S, Courtens E, Vacher R, Foret M, Wischnewski A and Buchenau U 2010 Phys. Rev. Lett. 104 067402
[58] Hong L, Begen B, Kisliuk A, Alba-Simionesco C, Novikov V N and Sokolov A P 2008 Phys. Rev. B 78 134201
[59] Chumakov A I et al. 2011 Phys. Rev. Lett. 106 225501
[60] Zhao J, Simon S L and McKenna G B 2013 Nat. Comm. (accepted)
[61] Sokolov A P, Rössler E, Kisliuk A and Quitmann D 1993 Phys. Rev. Lett. 71 2062
[62] Sokolov A, Kisliuk A, Quitmann D, Kudlik A and Rössler E 1994 J. Non-Cryst. Solids 172−174 138