Abstract
We consider a Langevin dynamics associated with a d-dimensional Edwards-Anderson model having Gaussian coupling variables, and show that the averaged law of the empirical process satisfies a large-deviation principle according to a good rate functional Ia having a unique minimizer Qx. The asymptotic dynamics Q∞ may be characterized as the unique weak solution corresponding to a non-Markovian system of interacting diffusions having an infinite range of interaction. We then establish that the quenched law of the empirical process also obeys a large-deviation process, according to a (deterministic) good rate functional Iq satisfying Iq ≥ Ia, so that, for a typical realization of the disorder variables, the quenched law of the empirical process also converges exponentially fast to a Dirac mass concentrated at Q∞.
Original language | English (US) |
---|---|
Pages (from-to) | 921-954 |
Number of pages | 34 |
Journal | Bernoulli |
Volume | 9 |
Issue number | 6 |
DOIs | |
State | Published - Dec 2003 |
Keywords
- Disordered systems
- Interacting diffusion processes
- Large deviations
- Statistical mechanics
ASJC Scopus subject areas
- Statistics and Probability