Convergence of clock processes and aging in Metropolis dynamics of a truncated REM (Q261966)

Many experiments show that glassy systems are never in equilibrium states on laboratory time scales; their dynamics is very slow as time elapses. It is called aging and can be described for example by Bouchaud's trap models which give the long-time behaviour of spin glass dynamics in terms of thermally activated barrier crossing in a state space reduced to the configurations of lowest energy. A Glauber dynamics is one of the main examples of microscopic systems behaviour that is described by trap models. However, there is a main question: Which Glauber dynamics should be taken into account: Metropolis, heath-bath dynamics or random hopping dynamics? Beyond model-based analysis, there is a second possible approach called ``clock process'', linking aging to the arcsine law for subordinators through the asymptotic behaviour of a partial sum process. The whole paper is devoted to considerations of possible usage of this approach by the introduction of some expansions, especially by the front-end and back-end clock processes. They are useful to prove some theorems in this paper, which is organized as follows. After the introduction in Section 1, we have a presentation of random graph properties of the random energy model landscape as a consequence of the proposed approach. Section 3 gives a formal definition of the front-end and back-end clock processes. In Section 4, there is an analysis of the increments of the front-end process \(\hat{S}_n\), especially its upper bound of the tail distribution. A similar analysis is carried out in Section 5 for the increments of the back-end clock process \(\hat{S}^\dagger_n\); an explicit expression is also obtained for the distribution of the sojourn times. The next section shows the attributes of the symmetric random walk. Sections 7--9 show other necessary proofs.