betaFeedback

WorksPhysics and Astronomy

Growing length scales while approaching the glass transition

Abstract

Even after decades of research, the nature of the glass transition is controversially debated: Kinetic freezing phenomena, purely local strain-mediated effects, or collective phenomena including a hidden phase transition are usually discussed. In this work, we check reports on clusters of correlated particles in supercooled liquids which grow upon approaching the glass-transition. We specifically focus on dielectric studies providing approximate access to the four-point correlation function, which allows insight into the evolution of so-called amorphous order. Existing temperature-dependent data derived from the three-point correlation function or higher-order dielectric susceptibilities are modeled using the random first-order transition theory combined with the Vogel-Fulcher-Tammann law. This ansatz enables detailed predictions concerning normalized correlation lengths or the number of correlated particles as function of temperature and fragility. It provides a good fit to the experimental data for all materials investigated, including network and molecular glass formers and polymers. We conclude that length scales and corresponding particle numbers determined by the different methods used are at least similar and closely related to those considered by the theory. Moreover, the model predictions on the fragility dependence of the number of correlated molecules are at least approximately fulfilled. Existing criticism concerning an apparent lack of criticality in the temperature dependence of experimental correlation lengths is discussed in terms of this fragility dependence: All available data are well consistent with a hidden phase transition which is best revealed for the most fragile liquids.

Introduction

When avoiding crystallization upon cooling, the viscosities of supercooled liquids drastically increase and relaxation times slow down by many orders of magnitude without significant structural changes. At glassy freezing, on laboratory time scales the dynamics falls out of thermal equilibrium and the glass-transition temperature TgT_{\mathrm{g}} is canonically defined by a relaxation time τ=100 s\tau= 100\ \mathrm{s} or alternatively by a viscosity η=1012 Pa s\eta= 10^{12}\ \mathrm{Pa\ s}. Glass-transition phenomena are important for a variety of materials, including silicate glasses, polymers, molecular glasses, ionic liquids, bio-related matter, and amorphous metals, which often are of high technological relevance. Despite enormous theoretical and experimental efforts, a microscopic understanding of continuous solidification without crystallization is still missing and controversially discussed.[1, 2, 3, 4, 5] Notably, TgT_{\mathrm{g}} is based on a purely human-made definition, and it remains a mystery why this transition temperature is observed with good approximation at two-thirds of the melting temperature TmT_{\mathrm{m}}, a well-defined thermodynamic quantity. Over the years, different models and theoretical concepts have been developed to understand the experimentally observed dramatic increase of viscosity when approaching TgT_{\mathrm{g}} (for critical surveys see, e.g., Refs. [4, 5, 7]). Here we consider the proposal initiated by Adam and Gibbs[8] and later theoretically worked out in the random first-order transition (RFOT) theory by Wolynes, Kirkpatrick, Lubchenko and coworkers,[9, 10] guided by theoretical concepts developed for spin glasses.[11] In this theory, supercooled liquids consist of clusters or “droplets” of correlated molecules which grow on decreasing temperature. This growth, which is determined by competition of configurational entropy and surface energy, produces an increase of the apparent hindering barriers against relaxational motion: In each cluster, the relaxation of one molecule is only possible when all the remaining molecules in this cluster reorganize as well. In RFOT theory, these clusters are characterized by a growing length scale ξ\xi or, alternatively, by a growing number of correlated particles NcorrN_{\mathrm{corr}}. This increasingly correlated motion is the source of the significant super-Arrhenius behavior of the viscosity or relaxation time observed in many fragile glass formers:[3] Approaching the glass transition, it drastically enhances the effective energy barriers relevant for the molecular relaxations that are related to the viscosity of the supercooled liquid. The fragility mm, defined by the slope at TgT_{\mathrm{g}} in the Angell plot, log⁡10(τ)\log_{10}(\tau) vs. Tg/TT_{\mathrm{g}}/T,[12] is about 16 for strong glasses, which are characterized by pure Arrhenius behavior, and ∼170\sim170 for the most fragile non-polymeric solids.[13] Fragile liquids have relatively low viscosity at the melting temperature[6] and reveal a strong increase of η(T)\eta(T) or τ(T)\tau(T) towards the glass transition. As mentioned above, in the Adam-Gibbs scenario[8] or in the RFOT theory[14, 15] this strong temperature dependence of apparent activation energies is explained in terms of growing length scales when approaching a phase transition located below the glass-transition temperature, which is hidden by the kinetic freezing occurring around TgT_{\mathrm{g}}.

However, there are also alternative theoretical proposals, where glassy freezing does not rely on growing length scales. Facilitation-based theories[16, 17] and Dyre’s shoving-model[4] are prominent examples. Recently, the criticism on the importance of growing length scales was reinforced by Wyart and coworkers,[18, 19, 20] proposing that only a minor fraction of the increase of the apparent activation energies stems from correlation effects, while the main contributions result from purely local effects strongly connected with the high-frequency shear modulus. Hence, these latter authors questioned whether growing length scales control the glass transition. [18]

In the present work, we quantitatively analyze published experimental results on correlation lengths, determined from dielectric spectroscopy, using the RFOT theory combined with the Vogel-Fulcher-Tammann (VFT) equation. The latter represents a commonly used empirical description of τ(T)\tau(T) or η(T)\eta(T) in supercooled liquids [2, 6, 12] and was also derived in various theoretical approaches, including the Adam-Gibbs theory. [8] Combining RFOT and VFT enables quantitative predictions concerning the absolute value and the temperature dependence of the number of correlated particles. We find reasonable agreement of experimental data and model predictions, with remarkable results concerning their temperature dependence, which strongly depends on fragility. In our analysis of experimental results, here we only consider one theoretical view of glass formation, and we do not aim to judge the validity of the mentioned competing approaches. However, we document that the RFOT theory, combined with the VFT law, reasonably well explains a variety of experimental results on growing length scales, derived by different groups, utilizing different dielectric techniques. We hope this work will trigger similar quantitative checks, based on competing theories.

Model Calculations

As discussed above, it is intuitive to assume that growing time scales observed in glassy freezing are correlated with the growing lengths presumed in various models describing the dynamics of spin glasses. [11] Growing length scales of cooperatively coupled clusters are also predicted by RFOT theory [14, 21] where the correlation length ξ\xi is related to the configurational entropy scs_{\mathrm{c}} via [14]

ξa=5.1(sc)2/3(1)\frac{\xi}{a} = \frac{5.1}{(s_{\mathrm{c}})^{2/3}} \tag*{(1)}

Here aa is defined as molecular length and corresponds to an approximate nearest-neighbor distance. Within this theory, scs_{\mathrm{c}} can be expressed in terms of normalized relaxation times, sc=32/ln⁡(τ/τ0)s_{\mathrm{c}} = 32/\ln(\tau/\tau_0). [14] Assuming a VFT law, τ=τ0exp⁡[DTVF/(T−TVF)]\tau= \tau_0 \exp[D T_{\mathrm{VF}}/(T-T_{\mathrm{VF}})], [3, 12] for τ(T)\tau(T), scs_{\mathrm{c}} then depends on its exponent:

sc=32(T−TVF)DTVF(2)s_{\mathrm{c}} = \frac{32(T-T_{\mathrm{VF}})}{D T_{\mathrm{VF}}} \tag*{(2)}

Here τ0\tau_0 corresponds to the high-temperature limiting relaxation rate, often assumed to be around 10−1410^{-14} s, [12] DD is the strength parameter, [3, 12] and TVFT_{\mathrm{VF}} the Vogel-Fulcher temperature where the relaxation times would diverge if freezing could be avoided. [3] Combining Eqs. (1) and (2) and using the approximate relations TVF/Tg=(m−16)/m6T_{\mathrm{VF}}/T_{\mathrm{g}} = (m-16)/m^6 and D=590/(m−16)D = 590/(m-16), [12] one derives:

(ξ/a)3≈Ncorr=Ncorr0[1+m(T/Tg−1)16]−2(3)(\xi/a)^3 \approx N_{\mathrm{corr}} = N_{\mathrm{corr}}^0 \left[1 + \frac{m(T/T_{\mathrm{g}}-1)}{16}\right]^{-2} \tag*{(3)}

where Ncorr0=Ncorr(Tg)≈176N_{\mathrm{corr}}^{0}=N_{\mathrm{corr}}(T_{\mathrm{g}})\approx176 can be derived from Eq. (1) using sc(Tg)s_{\mathrm{c}}(T_{\mathrm{g}}) obtained from sc=32/ln⁡(τ/τ0)s_{\mathrm{c}}=32/\ln(\tau/\tau_{0}) with τ(Tg)=100\tau(T_{\mathrm{g}})=100 s. Surprisingly, it is constant for all supercooled liquids. (3) provides a description of the temperature dependence of the number of correlated particles/molecular units which only depends on fragility mm and glass-transition temperature TgT_{\mathrm{g}} for a given material. Both are known for most glass formers and documented in literature.[12, 22]

(3) implies a number of unexpected predictions, which all are accessible experimentally: (i) at the glass-transition temperature all supercooled liquids, independent of structural details, molecular size or bonding strength should reveal clusters with a universal number of correlated molecules Ncorr0≈176N_{\mathrm{corr}}^{0}\approx176; ii) the temperature dependence of NcorrN_{\mathrm{corr}} is a function of glass-transition temperature and fragility only; (iii) the use of the VFT law automatically guarantees that NcorrN_{\mathrm{corr}} diverges at TVFT_{\mathrm{VF}} and, hence, reveals true criticality, and iv) even the strongest glass formers with m=16m=16 are characterized by significant temperature dependence Ncorr=Ncorr0(Tg/T)2N_{\mathrm{corr}}=N_{\mathrm{corr}}^{0}(T_{\mathrm{g}}/T)^{2}. The latter seems counterintuitive as for m=16m=16 the relaxation time follows an Arrhenius law suggesting temperature-independent energy barriers. However, as mentioned by Berthier et al.[23], even strong glasses possibly develop non-trivial dynamic correlations.

It seems interesting and can serve as additional plausibility check to discuss the fragility dependence of the correlation lengths at melting assuming the validity of the two-thirds rule Tg≈2/3 TmT_{\mathrm{g}}\approx2/3\,T_{\mathrm{m}}.[6] According to (3), strong liquids with m=16m=16 are characterized by a number of correlated molecules Ncorr(Tm)≈78N_{\mathrm{corr}}(T_{\mathrm{m}})\approx78, while fragile liquids with m=170m=170 reveal a drastically smaller number, Ncorr(Tm)≈4N_{\mathrm{corr}}(T_{\mathrm{m}})\approx4. Hence, in the framework of RFOT theory the correlation number mirrors the experimentally observed strong fragility dependence of the viscosity of liquids close to TmT_{\mathrm{m}}.[6]

In addition, at the upper boundary of fragility for non-polymeric solids[13] the normalized correlation length ξ/a≈1.6\xi/a\approx1.6 just exceeds the typical distance to the next-nearest neighbor shell. Low-viscosity liquids with mm close to 170 are characterized by small correlation lengths at melting, and seemingly in these melts crystallization is favored compared to supercooling. The above considerations also make it obvious that, against common belief, fragile liquids should not feature stronger particle correlations or cooperativity than strong ones. Instead, within the RFOT framework, it is solely the strong temperature dependence of these correlations that characterizes a fragile glass former.

Length Scales Measured by Different Experimental Methods

In RFOT theory, supercooled liquids are described as mosaic of droplets, which are termed “correlated volumes”,[10] “cooperative regions”,[14, 21] or “rearranging units”.[21] Their temperature-dependent size is denoted “coherence”,[10] “correlation”,[10, 14, 21] or “cooperativity length”,[14, 15] Thus, these objects remind of the cooperatively rearranging regions (CRRs) as proposed by Adam and Gibbs,[8] although it was pointed out in [21] that their size should be larger and reveal stronger temperature dependence that predicted by that older theory. The droplets assumed within the RFOT theory imply a non-uniform, heterogeneous nature of supercooled liquids. Indeed, it was noted that “dynamic heterogeneities” are “intrinsic to the mosaic picture of the deeply supercooled liquid”.[14] Notably, dynamic heterogeneity is a well-known concept in glass physics since long: It was shown to explain one of the hallmark features of supercooled liquids,[4] namely their non-exponential relaxation dynamics, which is nowadays commonly ascribed to a distribution of relaxation times caused by heterogeneity.[24, 25] Non-exponentiality, which can be quantified, e.g., by the stretching exponent β\beta, is usually found to become stronger upon cooling.[26, 27] This may suggest an increase of the length scales associated with this heterogeneity when approaching the glass transition, reminding of the predicted growth of the droplet sizes invoked within RFOT theory to explain another typical feature of supercooled liquids – the mentioned non-Arrhenius behavior. In recent literature, there is no clear consensus concerning the terms cooperativity or heterogeneity and they are often used almost synonymously, despite in principle describing quite different concepts. While some interrelations may exist, it was pointed out in Ref. 28 that cooperativity seems to be less well defined.

Various experimental methods were used to detect length scales in supercooled liquids. However, in some cases their exact nature remains rather unclear. It is well accepted that multidimensional NMR, revealing values of ~ 3 nm, detects heterogeneity, related to relaxation-time distributions in supercooled liquids.[29, 30] In contrast, molecular cooperativity lengths ranging from 1.4 – 3.4 nm were estimated by analyzing the spectral shape of the boson peak, but their values were directly compared to the heterogeneity lengths from NMR.[31] Donth and coworkers[32, 33] determined lengths between 1.0 – 3.6 nm from heat-capacity measurements, identifying them with the sizes of CRRs, but the justification for this assignment was doubted in Ref. 28. Recently, many of these results were summarized and interpreted as “average correlation length of the CRR” by Schawe et al.[34] They found that, at TgT_{\mathrm{g}}, all these lengths for various classes of glass formers seem to fall into a narrow range of 2 – 4 nm. Notably, this finding is in accord with the RFOT theory prediction of a universal relative correlation length ξ(Tg)/a≈Ncorr03≈5.6\xi(T_{\mathrm{g}})/a \approx\sqrt[3]{N_{\mathrm{corr}}^{0}} \approx5.6 [see Eq. (3)]. Assuming a correlation length of 3 nm results in a nearest-neighbor distance of ~ 0.5 nm, which is compatible with the observation of the first-sharp diffraction peak in many supercooled liquids (see, e.g., Ref. 35).

Next, we turn to the determination of temperature dependent correlation lengths from dielectric spectroscopy, which is the main scope of this work. Theoretical concepts developed in the field of spin glasses[11] evidenced that conventional two-point correlation functions are blind to long-range “amorphous order”,[25, 23, 36] suggested by RFOT and that the respective lengths can only be detected via the four-point correlation function χ4\chi_4.[23] For supercooled liquids, the importance of four-point correlators, simultaneously probing correlations at two points and two times, and, hence, measuring correlations of fluctuations, was established by Parisi and others,[37, 38] Bennemann et al.[39] and Berthier et al.[40, 41] Experimentally, four-point correlation functions are hardly accessible, except in cases one can monitor trajectories of individual particles. However, it was shown that they can be approximately measured via non-linear dielectric spectroscopy.[42] Moreover, a lower bound can be obtained by evaluating the three-point correlation function χT\chi_T, which is defined via the temperature derivative of the timedependent two-point correlation function and, making some assumptions, can be estimated from derivatives of the linear susceptibility or relaxation time.[23, 28, 41] The relation between χ4\chi_4 and χT\chi_T is not trivial, and here we follow arguments provided in Refs. 28,43,44 and analyze the number of correlated particles assuming the relation χ4∼χT\chi_4 \sim\chi_T.

Several measurements of χT\chi_T were performed for different classes of materials Refs. 28,41,45,46,47, providing detailed results on the temperature dependence of NcorrN_{\mathrm{corr}}. Notably, the authors of Refs. 45,46,47 analyzed their results assuming χ4∼χT2\chi_4 \sim\chi_T^2. Hence, as explained above, in fitting their results we used the square root of their published NcorrN_{\mathrm{corr}} values, to stay consistent with the analysis of the remaining data. In the following years, triggered by the proposal that NcorrN_{\mathrm{corr}} can be detected via non-linear dielectric measurements,[42] numerous such experiments on disordered materials were performed Refs. 36,44,48,49,50,51,52,53,54,55,56. This includes aging experiments,[48] studies of plastic crystals,[50, 51, 55] non-linear measurements under hydrostatic pressure,[52] and even experiments of 5th order susceptibilities.[53, 54, 56]

In the present work, we do not aim to clarify the mentioned ambiguities about the true nature of the length scales detected by different methods and their relation to the correlation length ξ\xi (or number of correlated particles NcorrN_{\mathrm{corr}}) considered within the RFOT theory. Instead, we pragmatically provide a critical comparison of all these published results to the expectations from the combination of the RFOT theory with the VFT law, promoted in section II, and fit a representative selection utilizing Eq. (3).

Comparison of Predictions of RFOT Theory with Experimental Results

For a detailed comparison of results from different dielectric experiments and theoretical predictions we chose propylene carbonate (PC) and glycerol, two canonical molecular glass formers, with glass transition temperatures Tg=159T_{\mathrm{g}} = 159 and 185185 K[57] and fragilities m=104m = 104 and 5353,[12] respectively. The relaxation dynamics of both compounds is well characterized by broadband dielectric spectroscopy covering more than 17 decades in frequency.[58] In both materials, the VFT law specifically well describes the temperature dependence of relaxation times close to TgT_{\mathrm{g}}, which seems important for the present analysis. In Figure 1 we collected the existing dielectric results on the temperature dependence of the number of correlated molecules. We present the Ncorr(T)N_{\mathrm{corr}}(T) traces normalized to unity at the glass-transition temperature, as RFOT theory predicts a universal Ncorr0≈176N_{\mathrm{corr}}^0 \approx176. We normalized the results as the absolute values may be rather uncertain, containing a number of approximations.[14] For both compounds, Figure 1 compares Ncorr(T)N_{\mathrm{corr}}(T) from third- and fifth-order susceptibilities[53] with results from three-point susceptibilities[28, 45, 59] as indicated in the figure legend. For the fifth-order susceptibility χ5\chi_5, we assumed spherical compact clusters, leading to the relation of χ5∼Ncorr2\chi_5 \sim N_{\mathrm{corr}}^2.[53, 56] As mentioned above, for the three-point susceptibilities determined by Fragiadakis et al. 45,47 we used Ncorr∼χTN_{\mathrm{corr}} \sim\chi_T[28, 43, 44] and not ∼χT2\sim\chi_T^2 proposed in that work.

Normalized number of correlated molecules versus the $T_g$-normalized temperature

Figure 1. Normalized number of correlated molecules, Ncorr/Ncorr(Tg)N_{\mathrm{corr}}/N_{\mathrm{corr}}(T_g), vs. the TgT_g-normalized temperature: (a) propylene carbonate and (b) glycerol as derived from different groups, utilizing different dielectric techniques as indicated in the legend: fifth order (χ5\chi_5) (red squares) and third order (χ3\chi_3) (blue spheres) nonlinear susceptibilities from Albert et al.[53], three-point correlation function χT\chi_T from Dalle-Ferrier et al.[28] and Fragiadakis et al.[45, 47] (magenta rhombs). The lines represent fits using Eq. (3), combining the RFOT theory with the VFT law, with fragility mm as the only free parameter. In PC, we fitted the different measurement results separately, and in the case of χT\chi_T from Ref. 45 we restricted the fit to the shown temperature regime. In glycerol, revealing only minor variance of the different experimental results, we fitted all data sets by a single fragility value. The two data points farthest below TgT_g, which may be measured out of equilibrium, were ignored. For each fit, the mm values are indicated in the figure.

All the results documented in Figure 1 were fitted using Eq. (3) with fragility mm as the only free parameter. Of course, it would be striking if the mm values derived from these fits would coincide with those determined analyzing the relaxation dynamics. Apparently, this is not the case: In PC [Figure 1(a)] the different experimental techniques result in some spread of the obtained temperature dependences of NcorrN_{\mathrm{corr}}. First, all the determined fragilities as indicated in the figure are significantly too low, compared with the value m=104m = 104 reported from relaxation data.[12] In addition, the three-point susceptibilities reveal larger fragilities than those derived from non-linear dielectric results. In PC, Fragiadakis et al.[45] reported NcorrN_{\mathrm{corr}} values up to temperatures extending beyond 2Tg2T_g. A fit to the complete temperature range results in m=81m = 81, only slightly enhanced compared to the result shown in Figure 1(a) (m=77m = 77). We restricted the fit to low temperatures, as at higher temperatures a crossover to a power-law behavior may be expected.[28] In glycerol (m=54m = 54),[12] the results are more convincing, and the temperature dependent number of correlated molecules can be well described by a unique fragility, which, however, is by a factor ∼1.5\sim1.5 too low. Nevertheless, the glycerol data provide a convincing test that the different dielectric techniques probe the same observable and, in addition, that the RFOT theory provides a good description of the experimentally observed temperature dependence of the number of correlated particles.

In Figure 2, we check for a possible fragility dependence of Ncorr(T)N_{\mathrm{corr}}(T), which is predicted by RFOT theory [see Eq. (3)]. We selected three sets of measurements of glass forming compounds spanning the range from strong to fragile and analyzed results on B2O3\mathrm{B_2O_3}},[28] 3-fluoro aniline (FAN)[49] and decahydro-naphthalene (decaline).[28] B2O3\mathrm{B_2O_3}} is a covalent-network-forming borate, a relatively strong glass former with glass-transition temperature Tg=554T_g = 554 K and fragility $m = 32.1232.^{12} FAN, consisting of small rigid molecules, is rather fragile with a glass transition at 172 K[60] and fragility m=89m = 89, which we determined from the VFT fit of τ(T)\tau(T) shown in Ref. [49]. Finally, decaline is an extremely fragile molecular glass former with Tg=135T_{\mathrm{g}} = 135 K and m=147m = 147.[61] Ncorr(T)N_{\mathrm{corr}}(T) of B2_2O3_3 and decaline was determined by the three-point correlation function,[28] while for FAN it was deduced from third-order nonlinear dielectric spectroscopy.[49]

Normalized number of correlated particles versus normalized temperature for B2O3, FAN, and decaline

Figure 2. Normalized number of correlated particles, Ncorr/Ncorr(Tg)N_{\mathrm{corr}}/N_{\mathrm{corr}}(T_{\mathrm{g}}), versus T/TgT/T_{\mathrm{g}} as reported for B2_2O3_3,[28] 3-fluoro aniline (FAN),[49] and decaline.[28] The dielectric data of Dalle-Ferrier et al.[28] were derived via the analysis of the derivative of the two-point correlation function, while those by Bauer et al.[49] represent third-order non-linear results. The solid lines show fits of the experimental data by Eq. (3) with fragility as the only free parameter. Fragilities mm obtained from the fits and from relaxation data are indicated in the figure legend. In addition, we added the mm-averaged fit curve of PC [Figure 1(a)] and the global fit of glycerol shown in Figure 1(b).

The temperature dependences of the normalized number of correlated molecules of these three glass formers were fitted using Eq. (3), again with fragility mm being the only free parameter. The fits are shown by the solid lines in Figure 2, and the resulting mm values are indicated in the figure legend. For these materials, fragilities derived by fitting the number of correlated molecules and those deduced from the relaxation dynamics (indicated in brackets) compare reasonably well. For comparison, we also included the results of the fits for PC and glycerol, documented in Figure 1. For PC we plot an mm-averaged curve using m=56m = 56 and for glycerol we show the global fit to all data with m=35m = 35. As mentioned earlier, both fragilities are significantly too low compared to those determined from the relaxation dynamics. This also becomes apparent in Figure 2 by the sequence of the Ncorr(T)N_{\mathrm{corr}}(T) results: Glycerol being an intermediate glass former exhibits the weakest temperature dependence, although B2_2O3_3 is reported to have smaller fragility.[12] In the figure, PC is located at intermediate fragilities. However, usually it is characterized as fragile glass former (m=104m = 104)[12] and, in Figure 2, thus should be located well below FAN. Concerning PC, one must mention two points: i) The experimental results from different sources provided in Figure 1 reveal considerable deviations and ii) PC is the only conventional glass former that strongly deviates from the overall correlation of the Böhmer plot, mm vs. β\beta. [12] It exhibits a stretching exponent β(Tg)=0.73\beta(T_{\mathrm{g}})=0.73, which according to the general correlation of supercooled liquids, m=250(±30)−320βm=250(\pm30)-320\beta, [12] would characterize PC as significantly stronger glass former with m≤46m\le46.

Concluding Remarks

In this work, we have presented a critical review of published results on correlation lengths in various supercooled liquids approaching the glass transition. We have compared the experimental data with model predictions derived by a combination of the RFOT theory with the VFT law and overall found remarkable agreement: i) The deduced prediction concerning the temperature-dependence of the relative number of correlated molecules [Eq. (3)] is well fulfilled in all investigated glass formers. This indicates that the length scales and corresponding particle numbers determined by the vastly different methods employed are all somehow related to those considered in the RFOT theory. Concerning the theory’s expectation of a universal NcorrN_{\mathrm{corr}} (or ξ\xi) at TgT_{\mathrm{g}}, we want to point out that in literature there are several hints at its approximate validity. [34, 41] ii) The prediction that the (relative) temperature-dependence of ξ\xi or NcorrN_{\mathrm{corr}} should only be determined by the fragility [Eq. (3)] is at least approximately fulfilled. Figure 2 provides the expected and rather significant correlation of Ncorr(T)N_{\mathrm{corr}}(T) with fragility, except for PC. iii) The general criticism concerning the importance of growing length scales (or correlated particle numbers) when approaching glassy freezing, namely, that all their observed temperature dependences are far from criticality, is not convincing: While the analyzed Ncorr(T)N_{\mathrm{corr}}(T) data indeed reveal only a limited relative increase when approaching the glass transition (up to a factor of five) they can be well fitted by Eq. (3) which involves a divergence at the Vogel-Fulcher temperature. Therefore, all experimental correlation numbers and lengths are well consistent with a divergence around the Vogel-Fulcher temperature, roughly corresponding to the Kauzmann temperature. Clear traces of criticality can only be observed in very fragile liquids [see Ncorr(T)N_{\mathrm{corr}}(T) of decaline in Figure 2]. This supports the observation that for fragile liquids the Prigogine-Defay ratio approaches unity [62], indicating the closeness of a thermodynamic phase-transition temperature to TgT_{\mathrm{g}}.

AUTHOR DECLARATION

Conflict of interest

The authors have no conflict to disclose.

Author contributions

Alois Loidl: Writing – original draft (lead); Writing – review & editing (equal) Conceptualization (lead); Formal analysis (equal); Validation (equal); Visualization (supporting). Peter Lunkenheimer: Writing - review & editing (equal); Conceptualization (supporting); Formal analysis (equal); Validation (equal); Visualization (lead). Konrad Samwer: Writing - review & editing (supporting); Conceptualization (equal); Formal analysis (supporting); Validation (supporting).

DATA AVAILABILITY

All data that support the findings of this study are available within this article.

References

  1. [1]W. Kauzmann, The nature of the glassy state and the behavior of liquids at low temperatures, Chem. Rev. 43, 219-256 (1948).DOI
  2. [2]P. G. Debenedetti and F. H. Stillinger, Supercooled liquids and the glass transition, Nature 410, 259-267 (2001).pubmed.ncbi.nlm.nih.gov/11258381
  3. [3]C. A. Angell, Formation of glasses from liquids and biopolymers, Science 267, 1924-1935 (1995).pubmed.ncbi.nlm.nih.gov/17770101
  4. [4]J. C. Dyre, Colloquium: The glass transition and elastic models of glass-forming liquids, Rev. Mod. Phys. 78, 953-970 (2006).DOI
  5. [5]L. Berthier and G. Biroli, Theoretical perspective on the glass transition and amorphous materials, Rev. Mod. Phys. 83, 587-645 (2011).DOI
  6. [6]P. Lunkenheimer, K. Samwer, and A. Loidl, The 2/3 rule of glass physics implies universalities in crystal melting, Phys. Rev. B 113, 054108 (2026).arxiv.org/abs/2511.21368
  7. [7]G. Biroli and J.-P. Bouchaud, The random first-order transition theory of glasses: A critical assessment, in Structural glasses and supercooled liquids, edited by P. G. Wolynes and V. Lubchenko (Wiley, Hoboken, 2012), pp. 31-113.
  8. [8]G. Adam and J. H. Gibbs, On the temperature dependence of cooperative relaxation properties in glass-forming liquids, J. Chem. Phys. 43, 139-146 (1965).DOI
  9. [9]T. R. Kirkpatrick and D. Thirumalai, Dynamics of the structural glass transition and the p-spin–Interaction spin-glass model, Phys. Rev. Lett. 58, 2091-2094 (1987).pubmed.ncbi.nlm.nih.gov/10034647
  10. [10]T. R. Kirkpatrick, D. Thirumalai, and P. G. Wolynes, Scaling concepts for the dynamics of viscous liquids near an ideal glassy state, Phys. Rev. A 40, 1045-1054 (1989).pubmed.ncbi.nlm.nih.gov/9902230
  11. [11]K. Binder and A. P. Young, Spin glasses: Experimental facts, theoretical concepts, and open questions, Rev. Mod. Phys. 58, 801 (1986).DOI
  12. [12]R. Böhmer, K. L. Ngai, C. A. Angell, and D. J. Plazek, Nonexponential relaxations in strong and fragile glass formers, J. Chem. Phys. 99, 4201–4209 (1993).DOI
  13. [13]L.-M. Wang, C. A. Angell, and R. Richert, Fragility and thermodynamics in nonpolymeric glassforming liquids, J. Chem. Phys. 125, 074505 (2006).pubmed.ncbi.nlm.nih.gov/16942349
  14. [14]V. Lubchenko and P. G. Wolynes, Barrier softening near the onset of nonactivated transport in supercooled liquids: Implications for establishing detailed connection between thermodynamic and kinetic anomalies in supercooled liquids, J. Chem. Phys. 119, 9088–9105 (2003).DOI
  15. [15]V. Lubchenko and P. G. Wolynes, Theory of structural glasses and supercooled liquids, Annu. Rev. Phys. Chem. 58, 235–266 (2007).DOI
  16. [16]G. H. Fredrickson and H. C. Andersen, Kinetic Ising model of the glass transition, Phys. Rev. Lett. 53, 1244 (1984).DOI
  17. [17]J. P. Garrahan and D. Chandler, Coarse-grained microscopic model of glass formers, Proc. Natl. Acad. Sci. USA 100, 9710–9714 (2003).
  18. [18]M. Wyart and M. E. Cates, Does a Growing Static Length Scale Control the Glass Transition? Phys. Rev. Lett. 119, 195501 (2017).DOI
  19. [19]M. P. Ciamarra, W. Ji, and M. Wyart, Local vs. cooperative: Unraveling glass transition mechanisms with SEER, PNAS 121, e2400611121 (2024).DOI
  20. [20]M. P. Ciamarra, J. C. Dyre, E. Lerner, and M. Wyart, Theories of the Glass Transition Based on Local Excitations, preprint, arXiv: 2603.05209 (2026).arxiv.org/abs/2603.05209
  21. [21]X. Xia and P. G. Wolynes, Fragilities of liquids predicted from the random first order transition theory of glasses, PNAS 97, 2990–2994 (2000).DOI
  22. [22]P. Lunkenheimer, A. Loidl, B. Richers, A. Zaccone, and K. Samwer, Thermal expansion and the glass transition, Nat. Phys. 19, 694–699 (2023).DOI
  23. [23]L. Berthier, G. Biroli, J.-P. Bouchaud, W. Kob, K. Miyazaki, and D. R. Reichman, Spontaneous and induced dynamic fluctuations in glass formers. I. General results and dependence on ensemble and dynamics, J. Chem. Phys. 126, 184503 (2007).pubmed.ncbi.nlm.nih.gov/17508807
  24. [24]H. Sillescu, Heterogeneity at the glass transition: a review, J. Non-Cryst. Solids 243, 81–108 (1999).
  25. [25]M. D. Ediger, Spatially heterogeneous dynamics in supercooled liquids, Annu. Rev. Phys. Chem. 51, 99–128 (2000).pubmed.ncbi.nlm.nih.gov/11031277
  26. [26]A. Schönhals, F. Kremer, A. Hofmann, E. W. Fischer, and E. Schlosser, Anomalies in the saling of the dielectric α-relaxation, Phys. Rev. Lett. 70, 3459–3462 (1993).DOI
  27. [27]P. Lunkenheimer, M. Köhler, S. Kastner, and A. Loidl, Dielectric spectroscopy of glassy dynamics, in Structural glasses and supercooled liquids, edited by P. G. Wolynes and V. Lubchenko (Wiley, Hoboken, 2012), pp. 115–149.DOI
  28. [28]C. Dalle-Ferrier, C. Thibierge, C. Alba-Simionesco, L. Berthier, G. Biroli, J.-P. Bouchaud, F. Ladieu, D. L’Hôte, and G. Tarjus, Spatial correlations in the dynamics of glassforming liquids: Experimental determination of their temperature dependence, Phys. Rev. E 76, 041510 (2007).
  29. [29]K. Schmidt-Rohr and H. W. Spiess, Nature of nonexponential loss of correlation above the glass transition investigated by multidimensional NMR, Phys. Rev. Lett. 66, 3020–3023 (1991).pubmed.ncbi.nlm.nih.gov/10043678
  30. [30]U. Tracht, M. Wilhelm, A. Heuer, H. Feng, K. Schmidt-Rohr, and H. W. Spiess, Length scale of dynamic heterogeneities at the glass transition determined by multidimensional nuclear magnetic resonance, Phys. Rev. Lett. 81, 2727–2730 (1998).DOI
  31. [31]L. Hong, P. D. Gujrati, V. N. Novikov, and A. P. Sokolov, Molecular cooperativity in the dynamics of glass-forming systems: A new insight, J. Chem. Phys. 131, 194511 (2009).
  32. [32]E. Donth, The size of cooperatively rearranging regions at the glass transition, J. Non-Cryst. Solids 53, 325–330 (1982).DOI
  33. [33]E. Hempel, G. Hempel, A. Hensel, C. Schick, and E. Donth, Characteristic Length of Dynamic Glass Transition near Tg for a Wide Assortment of Glass-Forming Substances, J. Phys. Chem. B 104, 2460–2466 (2000).DOI
  34. [34]J. E. K. Schawe, K. Hallavant, A. Esposito, J. F. Löffler, A. Saiter-Fourcin. The universality of cooperative fluctuations in glass-forming supercooled liquids, J. Phys. Chem. Lett. 16, 12255–12265 (2025).DOI
  35. [36]G. Biroli, J.-P. Bouchaud, and F. Ladieu, Amorphous order and nonlinear susceptibilities in glassy materials, J. Phys. Chem. B 125, 7578 (2021).DOI
  36. [37]S. Franz and G. Parisi, On non-linear susceptibility in supercooled liquids, J. Phys. Condens. Matter 12, 6335–6342 (2000).DOI
  37. [38]C. Donati, S. Franz, S. C. Glotzer, and G. Parisi, Theory of non-linear susceptibility and correlation length in glasses and liquids, J. Non-Cryst. Solids 307-310, 215–224 (2002).arxiv.org/abs/cond-mat/9905433
  38. [39]C. Bennemann, C. Donati, J. Baschnagel, and S. C. Glotzer, Growing range of correlated motion in a polymer melt on cooling towards the glass transition, Nature 399, 246–249 (1999).DOI
  39. [40]L. Berthier, Time and length scales in supercooled liquids, Phys. Rev. E 69, 020201R (2004).pubmed.ncbi.nlm.nih.gov/14995420
  40. [41]L. Berthier, G. Biroli, J.-P. Bouchaud, L. Cipelletti, D. El Masri, D. L’Hote, F. Ladieu, M. Pierno, Direct experimental evidence of a growing length scale accompanying the glass transition, Science 310, 1797–1800 (2005).DOI
  41. [42]J.-P. Bouchaud and G. Biroli, Nonlinear susceptibility in glassy systems: A probe for cooperative dynamical length scales, Phys. Rev. B 72, 064204 (2005).DOI
  42. [43]M. Tarzia, G. Biroli, A. Lefèvre and J.-P. Bouchaud, Anomalous nonlinear response of glassy liquids: General arguments and a mode-coupling approach, J. Chem. Phys. 132, 054501 (2010)DOI
  43. [44]C. Crauste-Thibierge, C. Brun, F. Ladieu, D. L’Hote, G. Biroli and J.-P. Bouchaud, Evidence of Growing Spatial Correlations at the Glass Transition from Nonlinear Response Experiments, Phys. Rev. Lett. 104, 165703 (2010).DOI
  44. [45]D. Fragiadakis, R. Casalini, and C. M. Roland, Density Scaling and Dynamic Correlations in Viscous Liquids, Phys. Chem. B 113, 13134–13137 (2009).arxiv.org/abs/0908.0683
  45. [47]D. Fragiadakis, R. Casalini, R. B. Bogoslovov, C. G. Robertson, and C. M. Roland, Dynamic Heterogeneity and Density Scaling in 1,4-Polyisoprene, Macromolecules 44, 1149–1155 (2011).
  46. [48]C. Brun, F. Ladieu, D. L’Hote, G. Biroli, and J.-P. Bouchaud, Evidence of Growing Spatial Correlations during the Aging of Glassy Glycerol, Phys. Rev. Lett. 109, 175702 (2012).
  47. [49]Th. Bauer, P. Lunkenheimer, and A. Loidl, Cooperativity and the freezing of molecular motion at the glass transition, Phys. Rev. Lett. 111, 225702 (2013).DOI
  48. [50]M. Michl, T. Bauer, P. Lunkenheimer, and A. Loidl, Cooperativity and heterogeneity in plastic crystals studied by nonlinear dielectric spectroscopy, Phys. Rev. Lett. 114, 067601 (2015).
  49. [51]M. Michl, T. Bauer, P. Lunkenheimer, and A. Loidl, Nonlinear dielectric spectroscopy in a fragile plastic crystal, J. Chem. Phys. 144, 114506 (2016)DOI
  50. [52]R. Casalini, D. Fragiadakis, and C. M. Roland, Dynamic correlation length scales under isochronal conditions, J. Chem. Phys. 142, 064504 (2015).
  51. [53]S. Albert, Th. Bauer, M. Michl, G. Biroli, J.-P. Bouchaud, A. Loidl, P. Lunkenheimer, R. Tourbot, C. Wiertel-Gasquet, F. Ladieu, Fifth-order susceptibility unveils growth of thermodynamic amorphous order in glass-formers, Science 352, 1308–1311 (2016).
  52. [54]P. Gadige, S. Albert, M. Michl, Th. Bauer, P. Lunkenheimer, A. Loidl, R. Tourbot, C. Wiertel-Gasquet, G. Biroli, J.-P. Bouchaud, and F. Ladie, Unifying different interpretations of the nonlinear response in glass-forming liquids, Phys. Rev. E 96, 032611 (2017).
  53. [55]P. Lunkenheimer, M. Michl, Th. Bauer, and A. Loidl, Investigation of nonlinear effects in glassy matter using dielectric methods, Eur. Phys. J. Special Topics 226, 3157–3183 (2017).DOI
  54. [56]S. Albert, M. Michl, P. Lunkenheimer, A. Loidl, P. M. Déjardin, and F. Ladieu, Third and fifth harmonic responses in viscous liquids, J. Stat. Mech.: Theory and Experiment 124003 (2019).
  55. [57]P. Lunkenheimer, S. Kastner, M. Köhler, and A. Loidl, Temperature development of glassy α-relaxation dynamics determined by broadband dielectric spectroscopy, Phys. Rev. E 81, 051504 (2010).
  56. [58]P. Lunkenheimer, U. Schneider, R. Brand, and A. Loid, Glassy dynamics, Contemp. Phys. 41, 15–36 (2000).
  57. [59]D. Fragiadakis, R. Casalini, and C. M. Roland, Comparing dynamic correlation lengths from an approximation to the four-point dynamic susceptibility and from the picosecond vibrational dynamics, Phys. Rev. E 84, 042501 (2011).
  58. [60]K. Kudlik, S. Benkhoff, T. Blochowicz, C. Tschirwitz, E. Rössler, The dielectric response of simple organic glass formers, J. Molecular Structure 479, 201–218 (1999).DOI
  59. [61]K. Duvvuri and R. Richert, Dynamics of glass-forming liquids. VI. Dielectric relaxation study of neat decahydro-naphthalene, J. Chem. Phys. 117, 4414–4418 (2002).
  60. [62]A. Loidl, P. Lunkenheimer, and K. Samwer, Prigogine-Defay ratio at glassy freezing scales with liquid fragility, Phys. Rev. 111, 035407 (2025).

Paper details

vlnc:W_xnx7jppc
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents