Spectral theory of diffusion in partially absorbing media

TL;DR

This paper develops a spectral approach to analyze single-particle diffusion in partially absorbing media, extending encounter-based methods to reactive substrates and illustrating the theory with a one-dimensional example.

Contribution

It introduces a spectral decomposition framework for the Laplace-transformed propagator in diffusion with reactive substrates, generalizing previous boundary cases.

Findings

Spectral decomposition of Dirichlet-to-Neumann operators enables computation of the propagator.

Inverse Laplace transform with respect to the absorption rate z is addressed.

Application demonstrated through a 1D example with scalar Dirichlet-to-Neumann operators.

Abstract

A probabilistic framework for studying single-particle diffusion in partially absorbing media has recently been developed in terms of an encounter-based approach. The latter computes the joint probability density (generalized propagator) for particle position $\X_t$ and a Brownian functional ${\mathcal U}_t$ that specifies the amount of time the particle is in contact with a reactive component $\calM$. Absorption occurs as soon as $\calU_t$ crosses a randomly distributed threshold (stopping time). Laplace transforming the propagator with respect to $\calU_t$ leads to a classical boundary value problem (BVP) in which the reactive component has a constant rate of absorption $z$, where $z$ is the corresponding Laplace variable. Hence, a crucial step in the encounter-based approach is finding the inverse Laplace transform. In the case of a reactive boundary $\partial \calM$, this can be achieved by solving a classical Robin BVP in terms of the spectral decomposition of a Dirichlet-to-Neumann operator. In this paper we develop the analogous construction in the case of a reactive substrate $\calM$. In particular, we show that the Laplace transformed propagator can be computed in terms of the spectral decomposition of a pair of Dirichlet-to-Neumann operators. However, inverting the Laplace transform with respect to $z$ is more involved. We illustrate the theory by considering a 1D example where the Dirichlet-to-Neumann operators reduce to scalars.