Flux large deviations of weakly interacting jump processes via well-posedness of an associated Hamilton-Jacobi equation (Q2040073): Difference between revisions

The paper under review studies a large deviation principle for a pair of processes in the context of time periodic rates, the path-space large deviation principle in the context of the dynamic Curie-Weiss-Potts model and the comparison principle for the Hamilton-Jacobi equation. The author focuses on time-inhomogeneous interacting jump processes with weak interaction, time-periodic with decreasing period size, converging rates and Lipschitz rates. The work is a natural continuation of [\textit{P. Dupuis} et al., ``Large deviation principle for finite-state mean field interacting particle systems'', Preprint, \url{arXiv:1601.06219}] and [\textit{R. I. A. Patterson} and \textit{D. R. M. Renger}, Math. Phys. Anal. Geom. 22, No. 3, Paper No. 21, 32 p. (2019; Zbl 1480.60063)] for the same time jumps and [\textit{R. Kraaij}, J. Stat. Phys. 164, No. 2, 321--345 (2016; Zbl 1347.60019)] for jumps with flux. Large deviations for weakly interacting jump processes have been studied before for spatial structures or random fields by direct evaluation of the asymptotics or by tilting arguments based on Sanov's theorem, Varadhan's lemma and the contraction principle. Section 2 begins with preliminaries on notations and basic concepts; Definition 2.1 deals with large deviation principles; Definition 2.2 introduces the Martingale problem for time-inhomogeneous processes and Definition 2.3 establishes the comparison principle. Section 3 states the main results: Theorem 3.3 on a large deviation principle for the pair processes of the empirical measure and the weakly interacting jump process, and Theorem 3.8 on the comparison principle for the Hamilton-Jacobi equation. Section 4 introduces the machinery to establish the large deviation principle by three main inputs: (1) exponential tightness which reduces the large deviation principle on the Skorokhod space to the one of the finite dimensional distributions, (2) convergence of operators which reduces the large deviation principle for the finite dimensional distribution convergence to the one at time 0 and the convergence of conditional generating functions, and (3) the well-posedness of the Hamilton-Jacobi equation which treats a functional analytic framework to form a semigroup. Theorem 3.3 follows from the the application of Theorem 7.10 of Kraaij [loc. cit.] with technical verifications provided in Section 5 and Section 6. Section 5 is devoted to the verification of the comparison principle with the main technical input being Lemma 5.3 for the proof of Theorem 3.8, using the semi-continuity properties of all functions involved and the compact level sets. The proof of Theorem 3.8 (comparison principle) follows with a good containment function given in Lemma 5.8, and proceeds by constructing a pair of good penalization function in Lemma 5.9, and verifies the explicit conditions in Proposition 5.6 for the comparison principle with a previous containment function and good penalization functions with various cases. Section 6 reduces the problem to technical estimates to apply the dominated convergence theorem in order to obtain the convergence of operators in the sense of ex-LIM, and verifies the exponential tightness for Theorem 4.4 with a standard martingale argument, and the last condition on rate functions in Lagrangian form given by the combination of [\textit{J. Feng} and \textit{T. G. Kurtz}, Large deviations for stochastic processes. Providence, RI: American Mathematical Society (AMS) (2006; Zbl 1113.60002)] with the Legendre transform and a duality argument. Appendix repeats arguments for the completeness.