Large deviations for trajectory observables of diffusion processes in dimension $d>1$ in the double limit of large time and small diffusion coefficient
Cecile Monthus
DOI: https://doi.org/10.1088/1742-5468/ad1390
2023-11-07
Abstract:For diffusion processes in dimension $d>1$, the statistics of trajectory observables over the time-window $[0,T]$ can be studied via the Feynman-Kac deformations of the Fokker-Planck generator, that can be interpreted as euclidean non-hermitian electromagnetic quantum Hamiltonians. It is then interesting to compare the four regimes corresponding to the time $T$ either finite or large and to the diffusion coefficient $D$ either finite or small. (1) For finite $T$ and finite $D$, one needs to consider the full time-dependent quantum problem that involves the full spectrum of the Hamiltonian. (2) For large time $T \to + \infty$ and finite $D$, one only needs to consider the ground-state properties of the quantum Hamiltonian to obtain the generating function of rescaled cumulants and to construct the corresponding canonical conditioned processes. (3) For finite $T$ and $D \to 0$, one only needs to consider the dominant classical trajectory and its action satisfying the Hamilton-Jacobi equation, as in the semi-classical WKB approximation of quantum mechanics. (4) In the double limit $T \to + \infty$ and $D \to 0$, the simplifications in the large deviations in $\frac{T}{D}$ of trajectory observables can be analyzed via the two orders of limits, i.e. either from the limit $D \to 0$ of the ground-state properties of the quantum Hamiltonians of (2), or from the limit of long classical trajectories $T \to +\infty$ in the semi-classical WKB approximation of (3). This general framework is illustrated in dimension $d=2$ with rotational invariance.
Statistical Mechanics,Probability