The phase space structure of supercooled liquids: Evidence of a dynamical critical temperature

Francesco Sciortino, Srikanth Sastry, P. TartagliaarXiv 1998 · DOI · arXiv

This 1998 classic in condensed matter physics provides a bridge between the world of liquids and the abstract mathematics of spin glasses. By analyzing the "topography" of the energy landscape in a simulated liquid, Sciortino and colleagues found a clear signature of a dynamical transition that had previously been mostly theoretical.

TL;DR

The researchers identified a "dynamical critical temperature" ($T_{MCT}$) in supercooled liquids by studying how their potential energy relaxes into local minima (inherent structures). They demonstrated that above this temperature, the liquid's phase space is dominated by unstable saddle points, a finding that aligns liquid behavior with mean-field spin glass models.

What Problem Does This Paper Address?

When a liquid is cooled rapidly enough to avoid crystallization, it becomes "supercooled" and eventually forms a glass. Understanding this transition is one of the deepest problems in physics. Two major frameworks attempt to explain it: Mode Coupling Theory (MCT), which focuses on the slowing down of particle dynamics, and mean-field $p$-spin glass models, which describe the transition in terms of a complex energy landscape.

Both theories predict a specific temperature ($T_{MCT}$ or $T_D$) where the liquid should theoretically "freeze" into a specific region of phase space. However, in real liquids, this transition is "smeared out" because particles can "hop" over energy barriers (activated processes). The challenge addressed here was to find a static, structural signature of this elusive dynamical temperature within the potential energy landscape of a molecular model.

Methods

The authors performed Molecular Dynamics (MD) simulations of a model for ortho-terphenyl (OTP), a well-known glass-forming liquid.

  1. Simulation: They equilibrated the system at various temperatures ($T$) and for different system sizes ($N = 343$ to $9261$ molecules).
  2. Inherent Structure (IS) Mapping: From the MD snapshots, they used a conjugate gradient minimization technique. This "quenches" the system, stripping away thermal vibrations to find the nearest local potential energy minimum, known as an "inherent structure."
  3. Landscape Analysis: Crucially, they didn't just look at the final energy; they analyzed the path the system took during minimization and how the resulting energy ($E_{min}$) changed with the starting temperature and the number of molecules in the box.

Datasets

The study relied on generated data from 12 different state points (temperatures) ranging from 255 K to 450 K. The primary "dataset" consisted of the potential energy trajectories and the resulting inherent structure energies across three different system sizes ($N=343, 1000, 9261$).

Key Insights and Ideas

  • The Two-Step Relaxation: Above a certain temperature, the energy minimization happens in two stages. First, a fast drop to a "quasi-stable" energy ($E_s$), followed by a slower search for a deeper minimum ($E_{min}$). This suggests the system is starting near a "saddle point" (a mountain pass) and eventually finding a "valley."
  • System Size Matters: The ability of the system to find that deeper minimum ($E_{min}$) is highly sensitive to the number of particles. In larger systems, the system is more likely to get "trapped" at the higher energy $E_s$. This is a key signature of the landscape's complexity.
  • Saddles vs. Minima: The authors argue that above $T_{MCT}$, the phase space is dominated by saddle points (points with at least one unstable direction). As the system size increases, these saddles effectively act as traps during a local search, providing evidence for the "threshold energy" ($E_{th}$) predicted by spin glass theory.
  • The Landscape "Break": They found that below ~285 K, the inherent structure energy depends strongly on the temperature at which the liquid was equilibrated. Above this temperature, the inherent structure energy becomes roughly constant.

Key Results

  • Identification of $T_{MCT}$: The researchers located a crossover temperature at $T \approx 285$ K for their OTP model. This matches the dynamical critical temperature predicted by other methods.
  • Evidence for Spin Glass Analogy: The behavior of the inherent structure energies as a function of temperature and system size provides strong evidence that supercooled liquids belong to the same universality class as $p$-spin glass models.
  • Thermodynamic Limit: By extrapolating $E_{min}$ to an infinite system size ($1/N \to 0$), they showed that the energy of the "disordered solution" (the liquid) corresponds to the highest-energy stable basins in the landscape.

Limitations and Future Work

  • Model Simplicity: While the 3-site OTP model captures essential physics, it is still a simplification of real molecular interactions.
  • Constant Density: The simulations were performed at constant density. The authors noted that exploring constant pressure paths would be a valuable next step to see if the results hold under different experimental conditions.
  • Dynamics vs. Statics: While the paper links the energy landscape (statics) to $T_{MCT}$ (dynamics), the exact mechanism of how the "onset of stretched exponential relaxation" relates to the energy $E_s$ remained a hypothesis for further investigation.

Summary updated

Abstract

We calculate local potential energy minima (inherent structures) for a simple model of orthoterphenyl (OTP), from computer simulations over a wide temperature range. We show that the results are very sensitive to the system size. We locate, from the temperature and size dependence of inherent structure energies, a well defined cross-over temperature that corresponds to the dynamical critical temperature predicted by mode coupling theory and in mean field spin glass models. Comparing the known phase space structure of mean field $p$-spin models and the present results for the OTP model, we find evidence supporting the proposed similarity between glass-forming liquids and a class of spin glass models.
BibTeX
@misc{sciortino1998phase,
  author = {Francesco Sciortino and Srikanth Sastry and P. Tartaglia},
  title = {{The phase space structure of supercooled liquids: Evidence of a dynamical critical temperature}},
  year = {1998},
  eprint = {cond-mat/9805040},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/cond-mat/9805040},
}

Loading related content...

Figure