On the fractional diffusion for the linear Boltzmann equation with drift and general cross-section
Dahmane Dechicha
TL;DR
This work establishes a fractional diffusion limit for the linear Boltzmann equation with heavy-tail equilibrium and a cross-section that depends on space and degenerates at high velocities, without assuming symmetry. By adapting the auxiliary problem method to this non-symmetric setting and incorporating a drift term j^ε_F, the authors show that, under the scaling θ(ε) = ε^γ with γ = (α−β)/(1−β), the density ρ(x,t) satisfies a macroscopic equation ∂_t ρ + κ L(ρ) = 0, where L is a spatial elliptic operator of order γ defined by a weighted PV integral. The operator L is expressed with a kernel η(x,y) that remains bounded and self-adjoint, reducing to a multiple of the fractional Laplacian when σ and ν are x-independent. The result extends prior symmetric-case diffusion limits, demonstrating that anomalous diffusion can arise in broad kinetic settings with non-symmetric cross-sections and degenerating collision frequencies. This has implications for modeling transport in media with heavy-tailed equilibria and velocity-dependent interactions, where fractional diffusion governs macroscopic behavior.
Abstract
This paper is devoted to the hydrodynamic limit for the linear Boltzmann equation, in the case of a heavy tail equilibrium and a cross section which depends on the space variable and which degenerates for large velocities, without symmetry assumptions. For an appropriate time scale, a macroscopic equation with an elliptic operator, which is equivalent to the fractional Laplacian, is obtained. This problem has been addressed in [Mellet, Mischler and Mouhot, ARMA, 2011] for a space-independent cross section, using a Fourier-Laplace transform, where a fractional diffusion equation has been obtained, and revisited in [Mellet, Indiana Univ., 2010] for a space-dependent but a bounded cross section, using the moments method by introducing an auxiliary problem. In this work, we will adapt the latter method to generalize both results.
