Three-dimensional spin glasses resist clean analytical solutions, and numerical methods that probe their critical behavior typically require comparing many replicas — an expensive proposition. The spectral approach constructs overlap matrices from two-dimensional cross-sections and examines their eigenvalue distributions. At high temperatures, the distribution follows the Wigner semicircle law — the universal signature of uncorrelated random matrices. As temperature drops toward the spin-glass critical point, the distribution deforms continuously into a Gaussian, passing through intermediate shapes described by Tsallis q-statistics with the entropic index q sliding from -1 to 1.
The transition in spectral shape is not just a diagnostic convenience. It reveals that the paramagnetic phase — often treated as structurally featureless — already contains statistical signatures of the approaching transition. The spectral density carries information about incipient correlations that bulk thermodynamic quantities miss. Local level statistics remain GOE-universal throughout; temperature reshapes only the global envelope, leaving the fine structure untouched.
Note (2026-11-07): the abstract says the Gaussian distribution can potentially serve as a robust spectral indicator of criticality and presents the spectral approach as a computationally efficient complement to multi-replica correlator methods; it does not say the spectrum is a more sensitive thermometer than the traditional quantities, that it detects structure missed by bulk thermodynamic quantities, or that transitions can be read before they appear in the order parameter. Those claims are mine.
The principle: critical transitions can be read from the shape of a spectrum before they manifest in the order parameter. The global distribution of eigenvalues functions as a more sensitive thermometer than the quantities traditionally used to locate phase boundaries — detecting structure in what appears, by other measures, to be noise.
(arXiv:2603.03513)