A Local Bifurcation Theorem for McKean-Vlasov Diffusions
Shao-Qin Zhang
TL;DR
The work proves existence and local bifurcation results for a broad class of probability measure-valued equations that characterize stationary distributions of McKean-Vlasov diffusions with gradient-type drifts, even when coefficients are discontinuous in the distribution variable. It combines Lyapunov-based Schauder fixed-point arguments with a local Krasnosel'skii bifurcation framework, using a regularized Hilbert-Schmidt determinant to detect eigenvalue crossings. The main payoff is a concrete, finite-rank determinant criterion for bifurcation points, with applications to granular media and Vlasov-Fokker-Planck type equations; the approach also handles non-smooth interactions and degenerate SDEs. Overall, the paper provides a rigorous bridge between measure-valued fixed-point theory and phase-transition phenomena in nonlinear diffusions, offering practical criteria to locate bifurcations and multiplicity of stationary states.
Abstract
We establish an existence result of a solution to a class of probability measure-valued equations, whose solutions can be associated with stationary distributions of many McKean-Vlasov diffusions with gradient-type drifts. Coefficients of the probability measure-valued equation may be discontinuous in the weak topology and the total variation norm. Owing to that the bifurcation point of the probability measure-valued equation is relevant to the phase transition point of the associated McKean-Vlasov diffusion, we establish a local Krasnosel'skii bifurcation theorem. Regularized determinant for the Hilbert-Schmidt operator is used to derive our criteria for the bifurcation point. Concrete examples, including the granular media equation and the Vlasov-Fokker-Planck equation with quadratic interaction, are given to illustrate our results.
