Exact finite-size scaling spectra in the quenched and annealed Sherrington-Kirkpatrick spin glass
Abstract
Fine resolution of the discrete eigenvalues at the spectral edge of an $N\times N$ random matrix is required in many applications. Starting from a finite-size scaling ansatz for the Stieltjes transform of the maximum likelihood spectrum, we show that the scaling function satisfies a first-order ODE of the Riccati type. Further transformation yields a linear second-order ODE for the characteristic function, whose nodes yield the exact locations of leading eigenvalues. With this new technique, we examine in detail the spectral crossover of the annealed Sherrington-Kirkpatrick (SK) spin glass model where a gap develops below a certain critical temperature. Our analysis yields analytic predictions for the finite-size scaling of the spin condensation phenomenon in both quenched and annealed SK models, which are then compared and validated in Monte Carlo simulations of the annealed model. More generally, rescaling of the spectral axis, adjusted to the distance of neighboring eigenvalues, offers a powerful strategy to deal with singularities that arise in the infinite size limit.
- Publication:
-
arXiv e-prints
- Pub Date:
- May 2024
- DOI:
- 10.48550/arXiv.2405.14215
- arXiv:
- arXiv:2405.14215
- Bibcode:
- 2024arXiv240514215W
- Keywords:
-
- Condensed Matter - Disordered Systems and Neural Networks;
- Condensed Matter - Statistical Mechanics
- E-Print:
- 17 pages, 8 figures, Submission to SciPost