Crossover in dynamics in the Kob-Andersen binary mixture glass-forming liquid

Pallai Das, Srikanth SastryJournal of Non-Crystalline Solids X 2022

This paper investigates a fundamental mystery in condensed matter physics: why do some liquids turn into glasses in a "fragile" way (slowing down exponentially fast) while others are "strong" (slowing down predictably)? By pushing simulations of the classic Kob-Andersen model to unprecedentedly low temperatures, the authors discover a "crossover" where the liquid's behavior shifts from fragile to strong, revealing that the way particles cluster and move changes fundamentally even when the underlying thermodynamics appear smooth.

TL;DR

The researchers demonstrate that the standard Kob-Andersen glass model undergoes a "fragile-to-strong" dynamic crossover at low temperatures. While particle movement and cluster shapes change significantly at this point, the classic Adam-Gibbs thermodynamic relation remains valid—provided you account for the "messy" anharmonic vibrations of atoms in their local energy wells.

What Problem Does This Paper Address?

In the study of glass-forming liquids, "fragility" describes how rapidly a liquid's viscosity increases as it cools toward the glass transition. Strong liquids (like silica) follow an Arrhenius law (a straight line on a log-plot), while fragile liquids (like many polymers or the Kob-Andersen model) show "super-Arrhenius" behavior, slowing down much faster than expected.

Some materials, like water, are known to "crossover" from fragile to strong as they cool. This paper asks: Does the Kob-Andersen (KA) model—the "standard workhorse" of computational glass physics—also exhibit this crossover? If it does, is this change driven by a thermodynamic phase transition (like a heat capacity spike) or is it purely a change in how particles move (dynamical heterogeneity)?

Methods

The authors used Molecular Dynamics (MD) simulations of the Kob-Andersen 80:20 binary Lennard-Jones mixture.

  • Scale: They simulated 4,000 particles using the LAMMPS package.
  • Depth: They accessed temperatures well below the "mode coupling temperature" ($T_{MCT} \approx 0.435$), where simulations usually become prohibitively slow. Their run lengths extended up to $2 \times 10^{10}$ integration steps.
  • Analysis: They calculated relaxation times ($\tau_\alpha$), the non-Gaussian parameter ($\alpha_2$), and the dynamical susceptibility ($\chi_4$).
  • Novelty: A key part of their method involved calculating anharmonic corrections to the vibrational entropy. Usually, researchers assume atoms vibrate like simple harmonic oscillators; this paper goes further to account for more complex, real-world vibrations to get a more accurate measure of "configurational entropy."

Datasets

The study uses the Kob-Andersen (KA-BMLJ) model, a standard theoretical benchmark in the field. No external experimental datasets were used; the "data" consists of the trajectories generated by their high-performance MD simulations.

Key Insights and Ideas

  • The Dynamic Crossover is Real: Below $T_{MCT}$, the relaxation times stop following the super-Arrhenius curve and settle into a simpler Arrhenius behavior. The activation energy barrier for this low-T regime is approximately 15 (in reduced units).
  • Morphology Shift: The "shape" of how particles move changes. At higher temperatures, mobile particles move in string-like clusters (fractal dimension $d_f \approx 2.0$). Below the crossover, these clusters become more compact and branched ($d_f \approx 2.5$).
  • Thermodynamics vs. Dynamics: Interestingly, while the movement (dynamics) shows a sharp change, the thermodynamics (like specific heat $C_v$) do not show the peak often associated with fragile-to-strong transitions in materials like silica or water. This suggests the crossover in this model might be purely dynamical or structural.
  • Saving the Adam-Gibbs Relation: The Adam-Gibbs (AG) relation connects a liquid's relaxation time to its configurational entropy. Previous studies suggested AG might fail at very low temperatures. This paper shows that AG actually holds perfectly across the crossover, but only if you include the anharmonic corrections to the entropy. The "failure" of AG in other studies may just be an artifact of using an oversimplified (harmonic) entropy model.

Key Results

  • Relaxation Time: Confirmed a crossover from super-Arrhenius to Arrhenius behavior around $T \approx 0.435$.
  • Dynamical Heterogeneity: The peak of the dynamical susceptibility ($\chi_4^{peak}$) saturates at low temperatures, mirroring the dynamic crossover.
  • Entropy: By including anharmonic vibrational corrections, the authors showed the Adam-Gibbs relation remains valid for roughly three decades of relaxation time below the crossover temperature.
  • No Heat Capacity Peak: Unlike some previous (and perhaps under-equilibrated) studies, they found no evidence of a specific heat maximum, suggesting the transition isn't a simple thermodynamic one.

Limitations and Future Work

The authors acknowledge that finite-size effects could potentially influence the results, though they believe their system is large enough to capture the essential physics.

A major remaining question is the structural origin of the crossover. While they showed that cluster geometry changes, they haven't yet identified a specific "locally preferred structure" (like icosahedra or tetrahedra) that triggers the shift. Future work will likely focus on mapping these dynamic changes to specific geometric arrangements of atoms to provide a complete structural explanation for why the liquid suddenly changes its "slowing down" strategy.

Summary updated

Abstract

Glass-forming liquids are broadly classified as being fragile or strong, depending on the deviation from Arrhenius behavior of their relaxation times. A fragile to strong crossover is observed or inferred in liquids like water and silica, and more recently also in metallic glasses and phase change alloys, leading to the expectation that such a crossover is more widely realised among glass formers. We investigate computationally the well-studied Kob-Andersen model, accessing temperatures well below the mode coupling temperature $T_{MCT}$. We find that relaxation times exhibit a crossover in dynamics around $T_{MCT}$, and discuss whether it bears characteristics of the fragile to strong crossover. Several aspects of dynamical heterogeneity exhibit behavior mirroring the dynamical crossover, whereas thermodynamic quantities do not. In particular, the Adam-Gibbs relation describing the relation between relaxation times and configurational entropy continues to hold below the dynamical crossover, when anharmonic corrections to the vibrational entropy are included.
BibTeX
@article{das2022crossover,
  author = {Pallai Das and Srikanth Sastry},
  title = {{Crossover in dynamics in the Kob-Andersen binary mixture glass-forming liquid}},
  journal = {Journal of Non-Crystalline Solids X},
  year = {2022},
}
Figure