Numerical determination of the exponents controlling the relationship between time, length, and temperature in glass-forming liquids

Chiara Cammarota, Andrea Cavagna, Giacomo Gradenigo, Tomás S. Grigera, P. VerrocchioThe Journal of Chemical Physics 2009 · DOI · arXiv

Abstract

There is a certain consensus that the very fast growth of the relaxation time $\tau$ occurring in glass-forming liquids on lowering the temperature must be due to the thermally activated rearrangement of correlated regions of growing size. Even though measuring the size of these regions has defied scientists for a while, there is indeed recent evidence of a growing correlation length $\xi$ in glass formers. If we use Arrhenius law and make the mild assumption that the free-energy barrier to rearrangement scales as some power $\psi$ of the size of the correlated regions, we obtain a relationship between time and length, $T \log \tau \approx \xi^\psi$. According to both the Adam-Gibbs and the random first order theory the correlation length grows as $\xi \approx (T-T_k)^{-1/(d-\theta)}$, even though the two theories disagree on the value of $\theta$. Therefore, the super-Arrhenius growth of the relaxation time with the temperature is regulated by the two exponents $\psi$ and $\theta$ through the relationship $T \log \tau \approx (T-T_k)^{-\psi/(d-\theta)}$. Despite a few theoretical speculations, up to now there has been no experimental determination of these two exponents. Here we measure them numerically in a model glass former, finding $\psi=1$ and $\theta=2$. Surprisingly, even though the values we found disagree with most previous theoretical suggestions, they give back the well-known VFT law for the relaxation time, $T \log \tau \approx (T-T_k)^{-1}$.

Introduction

When a liquid is cooled toward its glass transition, its relaxation time (viscosity) increases by many orders of magnitude over a small temperature range. This "super-Arrhenius" behavior is widely believed to be caused by the growth of "cooperatively rearranging regions" (CRRs)—pockets of the liquid that must move together to allow the system to relax. While scientists have long suspected a link between the size of these regions ($\xi$) and the time it takes to relax ($\tau$), the exact nature of this link remained theoretical. Specifically, two exponents were unknown: $\psi$, describing how the energy barrier to movement scales with the size of the region ($\Delta \sim \xi^\psi$), and $\theta$, describing how the energy cost of the interface between different "phases" of the liquid scales with size ($E_{surface} \sim R^\theta$).

Methods

The authors used a binary mixture of soft particles—a standard model for a "fragile" glass-former. To measure the exponents, they developed a novel "Amorphous Excitation" protocol:

  • Creating Droplets: They took two independent, equilibrium configurations of the liquid ($\alpha$ and $\beta$). They then cut a sphere of radius $R$ from configuration $\alpha$ and swapped it into the same location in configuration $\beta$.
  • Measuring Energy ($\theta$): By calculating the energy cost of this artificial "droplet" interface, they could determine how the surface energy scales with $R$.
  • Measuring Time ($\psi$): They monitored how long it took for these artificial excitations to relax back to equilibrium. By varying the radius $R$, they could extract the scaling between the energy barrier and the length scale.
Schematic of the Amorphous Excitation protocol. A sphere of radius $R$ is swapped between two independent equilibrium configurations $C_\alpha$ and $C_\beta$.

Simulation techniques included Swap Monte Carlo for efficient thermalization (reaching low temperatures that standard methods can't) and standard Metropolis Monte Carlo to observe the relaxation dynamics. The study relied on generated numerical data from a system of $N = 16,384$ particles across five distinct temperatures.

Key Results and Insights

The most striking finding is that while the individual exponents $\psi=1$ and $\theta=2$ disagree with the specific values predicted by major theories like Adam-Gibbs (AG) and Random First Order Theory (RFOT), their combination perfectly yields the empirical Vogel-Fulcher-Tamman (VFT) law. In 3D ($d=3$), the relaxation law follows $T \log \tau \sim (T - T_k)^{-\psi/(d-\theta)}$. Plugging in the measured values gives an exponent of $1/(3-2) = 1$, which is exactly what the VFT law requires.

Scaling of the relaxation time $\tau$ with droplet radius $R$. The linear fit of $T \log \tau$ vs $R$ yields $\psi = 1.01 \pm 0.04$.

Specifically, the study found:

  • Exponent $\psi \approx 1$: This suggests the energy barrier scales linearly with the length of the region, rather than its volume (as AG suggests) or its surface area (as RFOT suggests).
  • Exponent $\theta \approx 2$: This indicates that in three dimensions, the interface energy scales like a standard surface ($R^2$). While the interface is "rough" (with a roughening exponent $\gamma \approx 0.62$), the disorder isn't strong enough to change the leading-order scaling.
  • Violation of the Fisher-Huse Inequality: In spin-glass theory, it is assumed that $\psi \ge \theta$. Here, the authors find $\psi < \theta$ ($1 < 2$). They argue this is possible in structural glasses because of the massive "entropic degeneracy"—there are exponentially many states the system can relax into, which effectively lowers the free-energy barrier.

Discussion and Future Work

While the results are clear for the sizes studied ($R$ from 3 to 8), the authors acknowledge that $\theta$ is an asymptotic exponent. Whether these values hold for much larger, experimental length scales remains a point of discussion. The paper opens a new door for investigating how the "entropy of target configurations" specifically lowers the relaxation barrier, potentially leading to a more refined version of RFOT that doesn't rely on simple nucleation analogies. Future work may focus on extending these measurements to larger systems and exploring the temperature dependence of these exponents more deeply.

Summary updated

Abstract

There is a certain consensus that the very fast growth of the relaxation time tau occurring in glass-forming liquids on lowering the temperature must be due to the thermally activated rearrangement of correlated regions of growing size. Even though measuring the size of these regions has defied scientists for a while, there is indeed recent evidence of a growing correlation length xi in glass formers. If we use Arrhenius law and make the mild assumption that the free-energy barrier to rearrangement scales as some power psi of the size of the correlated regions, we obtain a relationship between time and length, T log tau approximately xi(psi). According to both the Adam-Gibbs and the random first order theory the correlation length grows as xi approximately (T-T(k))(-1/(d-theta)), even though the two theories disagree on the value of theta. Therefore, the super-Arrhenius growth of the relaxation time with the temperature is regulated by the two exponents psi and theta through the relationship T log tau approximately (T-T(k))(-psi/(d-theta)). Despite a few theoretical speculations, up to now there has been no experimental determination of these two exponents. Here we measure them numerically in a model glass former, finding psi=1 and theta=2. Surprisingly, even though the values we found disagree with most previous theoretical suggestions, they give back the well-known VFT law for the relaxation time, T log tau approximately (T-T(k))(-1).
BibTeX
@article{cammarota2009numerical,
  author = {Chiara Cammarota and Andrea Cavagna and Giacomo Gradenigo and Tomás S. Grigera and P. Verrocchio},
  title = {{Numerical determination of the exponents controlling the relationship between time, length, and temperature in glass-forming liquids}},
  journal = {The Journal of Chemical Physics},
  year = {2009},
  doi = {10.1063/1.3257739},
}
Figure