Reformulating mean-field glass theory beyond the Gaussian cage: consistent statics and dynamics of high-dimensional hard spheres, and the finite-dimensional fluctuation regime it enables

Simulations of hard-sphere glass formers in spatial dimensions $d=4$ – $9$ showed that the caging order parameter — the distribution of a particle's displacement inside the cage formed by its neighbors — remains persistently non-Gaussian deep into the mean-field high- $d$ regime, while mode-coupling theory (MCT) predicts dynamical transition packing fractions $\phi_d$ , exponents $\gamma$ , and cage shapes that are not merely quantitatively wrong but deteriorate with increasing $d$ , and replica theory and density functional theory assume a Gaussian cage from the outset. Exact $d\to\infty$ statics and dynamics now exist: the Gaussian ansatz is thermodynamically exact at leading order, the exact dynamics gives a Gaussian self-correlator at fixed $\phi<\phi_d$ , and the non-Gaussian cage has been traced to divergent single-particle critical fluctuations at the dynamical transition. Nevertheless, a controlled many-body theory that quantitatively predicts the observed non-Gaussian self van Hove profile and the finite- $d$ corrections to $\phi_d$ and $\gamma$ is still missing, and no high-dimensional test of dynamical facilitation has been performed. Formulating and quantitatively validating a mean-field theory of the glass transition built on a non-Gaussian caging order parameter, and reconciling its dynamical and static implementations, remains an open problem; such a theory would in turn serve as the controlled starting point for the study of finite-dimensional fluctuations below the upper critical dimension $d_u=8$ .

Background

When a liquid is cooled or compressed fast enough to avoid crystallization, it becomes a supercooled liquid whose structural-relaxation time grows by many orders of magnitude before the material falls out of equilibrium as a glass. A minimal model for this phenomenon is the hard-sphere fluid, identical particles with purely excluded-volume interactions, whose state is controlled by the packing fraction $\phi$ (fraction of volume covered by the spheres); simulations are conveniently done as a function of spatial dimension $d$ because glassy slowdown sharpens as $d$ grows.

Two classes of mean-field (MF) descriptions dominate the theory. On the dynamical side, mode-coupling theory (MCT) derives from microscopic equations of motion an approximate closed equation for density correlations; it predicts an ideal dynamical glass transition at $\phi_d$ , where the diffusion constant $D$ vanishes as $D\sim(\phi_d-\phi)^{\gamma}$ with a critical exponent $\gamma$ . On the static side, the random first-order transition (RFOT) scenario describes the glass transition by analogy with spin glasses: replica theory (RT) computes the thermodynamics of amorphous glass states by coupling replicas of the liquid, predicting, beyond the dynamical transition, a thermodynamic Kauzmann transition at $\phi_K>\phi_d$ where the liquid entropy equals that of the glass states; density functional theory (DFT) pursues the same goal from the free-energy functional of the density field. In the MF limit $d\to\infty$ these descriptions are expected to become exact and mutually consistent.

The physical mechanism shared by all of them is self-caging: close to arrest, each particle rattles in a transient cage formed by its neighbors, so the distribution of its displacement within the cage — equivalently the long-time self part $G_s(r,t)$ of the van Hove function, the probability distribution of single-particle displacements — serves as the order parameter of the transition; its width is the cage size, often denoted $A$ . RT and DFT assume from the outset that this cage distribution is Gaussian (the Gaussian-cage or Gaussian ansatz), a belief so standard that a Gaussian cage was taken to define the MF limit. A convenient diagnostic time is $\tau_{\mathrm{MP}}$ , the logarithmic mid-point of the intermediate caging ( $\beta$ -relaxation) window, and the kurtosis of $G_s(r,t)$ defines the non-Gaussian parameter $\alpha_2(t)$ , which measures how far single-particle displacements are from Gaussian (dynamically homogeneous) behavior.

Charbonneau, Ikeda, Parisi, and Zamponi (Dimensional study of the caging order parameter at the glass transition, PNAS 109, 13939 (2012)) simulated 8000 hard spheres in $d=4$ – $9$ and found that the cage does not become Gaussian: at $\tau_{\mathrm{MP}}$ the self van Hove function develops a fat tail that is not reduced with increasing $d$ and departs ever further from MCT predictions; individual particle cages are non-Gaussian in isoconfigurational runs (same configuration, different initial momenta), so the effect is not a mere superposition of Gaussian cages with different relaxation speeds. They also confirmed that MCT predictions for $\phi_d$ , $\gamma$ , and the cage shape worsen with $d$ , and that both the dynamical (MCT-like) and static (RT/DFT) implementations of RFOT therefore fail precisely where MF theory should be at its best. Two further threads frame the problem. First, the dynamical transition is a critical phenomenon with an upper critical dimension $d_u=8$ (from the replica field theory of Franz, Parisi, Ricci-Tersenghi, and Rizzo and the RFOT renormalization-group analysis of Cammarota, Biroli, Tarzia, and Tarjus), so a corrected MF theory is expected to be the controlled starting point for fluctuation corrections below $d_u$ ; the paper found non-MF corrections to remain small even below $d_u$ but could not quantify the regime where they dominate. Second, the alternative dynamical facilitation picture — relaxation proceeding by local, hierarchically facilitated events — predicts a smooth dependence on $d$ consistent with the data, a consistency the source paper itself notes; the high-dimensional robustness of facilitation-based descriptions, however, was left as an open evaluation.

Problem Statement

Construct a corrected mean-field theory of the glass transition for hard spheres (and simple glass formers more generally) in which the caging order parameter is non-Gaussian, and validate it quantitatively. Relative to the benchmarks established by the dimensional study of the caging order parameter (Charbonneau–Ikeda–Parisi–Zamponi, PNAS 109, 13939 (2012): $\phi_d$ , $\gamma$ , cage size $A$ , and the self van Hove profile $G_s(r,\tau_{\mathrm{MP}})$ for $d=4$ – $9$ ), the reformulation comprises three scientifically linked requirements: (1) reformulate the dynamical mean-field description so that, unlike standard MCT, it correctly predicts the numerically observed dynamical transition packing fraction $\phi_d$ , the critical exponent $\gamma$ , and the non-Gaussian long-time self van Hove profile in high but finite dimensions, while reducing to the exact $d\to\infty$ dynamics; (2) modify the static mean-field treatments (replica theory and density functional theory) so they no longer rest on the Gaussian-cage ansatz, and determine quantitatively how the non-Gaussian caging order parameter changes their predictions for $\phi_d$ , the Kauzmann packing fraction $\phi_K$ , and other thermodynamic quantities at finite dimension; and (3) establish the detailed theoretical origin of the persistent non-Gaussian single-particle cage observed at $\tau_{\mathrm{MP}}$ in high but finite $d$ , going beyond the reproduction of the profile required by (1) to derive its explicit closed-form shape and its analytic scaling with $d$ and with the distance to $\phi_d$ , within a mean-field treatment that goes beyond the leading Gaussian order and is consistent with the known exact $d\to\infty$ results (Gaussian leading-order statics, and a Gaussian exact self-correlator at fixed $\phi<\phi_d$ with divergent single-particle fluctuations at $\phi_d$ ). The theory must make the dynamical and static descriptions mutually consistent on a non-Gaussian-cage basis and must not retain the Gaussian cage. All three requirements are jointly required, and the theory must reproduce the benchmark quantities above ( $\phi_d$ , $\gamma$ , cage size $A$ , and $G_s(r,\tau_{\mathrm{MP}})$ ) quantitatively across the studied dimensions $d=4$ – $9$ . A corrected mean-field theory would in addition provide the controlled starting point for a refined renormalization-group analysis of finite-dimensional fluctuations below the upper critical dimension $d_u=8$ , and would allow the high-dimensional robustness of descriptions based on dynamical facilitation to be evaluated; these are downstream directions enabled by the theory, not components of the demanded reformulation.

References

Patrick Charbonneau, Atsushi Ikeda, Giorgio Parisi, Francesco Zamponi, Dimensional study of the caging order parameter at the glass transition, Proc. Natl. Acad. Sci. USA 109, 13939 (2012), DOI: 10.1073/pnas.1211825109, arXiv:1205.4057, https://arxiv.org/abs/1205.4057
Jorge Kurchan, Giorgio Parisi, Francesco Zamponi, Exact theory of dense amorphous hard spheres in high dimension. I. The free energy, J. Stat. Mech. (2012) P10012, DOI: 10.1088/1742-5468/2012/10/P10012, arXiv:1208.0421, https://arxiv.org/abs/1208.0421
Jorge Kurchan, Giorgio Parisi, Pierfrancesco Urbani, Francesco Zamponi, Exact theory of dense amorphous hard spheres in high dimension. II. The high density regime and the Gardner transition, J. Phys. Chem. B 117, 12979 (2013), DOI: 10.1021/jp402235d, arXiv:1303.1028, https://arxiv.org/abs/1303.1028
Patrick Charbonneau, Jorge Kurchan, Giorgio Parisi, Pierfrancesco Urbani, Francesco Zamponi, Exact theory of dense amorphous hard spheres in high dimension. III. The full replica symmetry breaking solution, J. Stat. Mech. (2014) P10009, DOI: 10.1088/1742-5468/2014/10/P10009, arXiv:1310.2549, https://arxiv.org/abs/1310.2549
Thibaud Maimbourg, Jorge Kurchan, Francesco Zamponi, Solution of the dynamics of liquids in the large-dimensional limit, Phys. Rev. Lett. 116, 015902 (2016), DOI: 10.1103/PhysRevLett.116.015902, arXiv:1507.03421, https://arxiv.org/abs/1507.03421
Grzegorz Szamel, Simple theory for the dynamics of mean-field-like models of glass-forming fluids, Phys. Rev. Lett. 119, 155502 (2017), DOI: 10.1103/PhysRevLett.119.155502, arXiv:1705.00036, https://arxiv.org/abs/1705.00036
Jorge Kurchan, Thibaud Maimbourg, Francesco Zamponi, Statics and dynamics of infinite-dimensional liquids and glasses: a parallel and compact derivation, J. Stat. Mech. (2016) 033210, DOI: 10.1088/1742-5468/2016/03/033210, arXiv:1512.02186, https://arxiv.org/abs/1512.02186
Chen Liu, Giulio Biroli, David Reichman, Grzegorz Szamel, Dynamics of liquids in the large-dimensional limit, Phys. Rev. E 104, 054606 (2021), DOI: 10.1103/PhysRevE.104.054606, arXiv:2108.02378, https://arxiv.org/abs/2108.02378
Giulio Biroli, Patrick Charbonneau, Giampaolo Folena, Yi Hu, Francesco Zamponi, Local dynamical heterogeneity in simple glass formers, Phys. Rev. Lett. 128, 175501 (2022), DOI: 10.1103/PhysRevLett.128.175501, arXiv:2109.11822, https://arxiv.org/abs/2109.11822
Corentin C. L. Laudicina, Patrick Charbonneau, Yi Hu, Liesbeth M. C. Janssen, Peter K. Morse, Ilian Pihlajamaa, Grzegorz Szamel, Simple fluctuations in simple glass formers, J. Phys. Chem. B 128, 12237 (2024), DOI: 10.1021/acs.jpcb.4c05447, arXiv:2408.06933, https://arxiv.org/abs/2408.06933
Ludovic Berthier, Patrick Charbonneau, Joyjit Kundu, Finite dimensional vestige of spinodal criticality above the dynamical glass transition, Phys. Rev. Lett. 125, 108001 (2020), DOI: 10.1103/PhysRevLett.125.108001, arXiv:1912.11510, https://arxiv.org/abs/1912.11510
S. Franz, G. Parisi, F. Ricci-Tersenghi, T. Rizzo, Field theory of fluctuations in glasses, Eur. Phys. J. E 34, 102 (2011), DOI: 10.1140/epje/i2011-11102-0, https://doi.org/10.1140/epje/i2011-11102-0
Chiara Cammarota, Giulio Biroli, Marco Tarzia, Gilles Tarjus, Renormalization group analysis of the random first-order transition, Phys. Rev. Lett. 106, 115705 (2011), DOI: 10.1103/PhysRevLett.106.115705, arXiv:1007.2509, https://arxiv.org/abs/1007.2509
Patrick Charbonneau, Yuliang Jin, Giorgio Parisi, Francesco Zamponi, Hopping and the Stokes–Einstein relation breakdown in simple glass formers, Proc. Natl. Acad. Sci. USA 111, 15025 (2014), DOI: 10.1073/pnas.1417182111, arXiv:1407.5677, https://arxiv.org/abs/1407.5677
Hugo Jacquin, Francesco Zamponi, Systematic expansion in the order parameter for replica theory of the dynamical glass transition, J. Chem. Phys. 138, 12A542 (2013), DOI: 10.1063/1.4792641, arXiv:1211.3468, https://arxiv.org/abs/1211.3468
Giulio Biroli, Jean-Philippe Bouchaud, Diverging length scale and upper critical dimension in the mode-coupling theory of the glass transition, Europhys. Lett. 67, 21 (2004), DOI: 10.1209/epl/i2004-10044-6, arXiv:cond-mat/0401260, https://arxiv.org/abs/cond-mat/0401260
Corentin C. L. Laudicina, Liesbeth M. C. Janssen, Grzegorz Szamel, Microscopic theory of a fluctuation-induced dynamical crossover in supercooled liquids, Phys. Rev. Lett. (2026), DOI: 10.1103/8lw2-352q, arXiv:2512.13082, https://arxiv.org/abs/2512.13082