no. 3 · Tuesday 27 October 2026
Martingale dimensions for a class of metric measure spaces
Abstract
We study the martingale dimension of diffusion processes associated with strongly local regular Dirichlet forms on metric measure spaces. The martingale dimension is an intrinsic quantity describing the probabilistic complexity of a diffusion, but its precise determination in non-smooth settings is highly nontrivial. In this paper, we establish a purely analytic criterion, formulated in terms of energy measures and relative capacities, which guarantees that the AF-martingale dimension equals one. Our approach does not rely on self-similarity, heat kernel estimates, or global geometric assumptions, and applies to settings where such structures are unavailable. The proof is based on a blow-up analysis of harmonic functions combined with a push-forward argument for energy measures, which excludes the coexistence of multiple independent martingale directions. As applications, we verify the assumptions of the theory for a class of inhomogeneous Sierpinski gaskets, including genuinely non-self-similar examples, and for the two-dimensional standard Sierpinski carpet, where the criterion yields a new capacity-based proof of the one-dimensionality of the martingale structure. In these examples, self-similarity or local symmetry enters only in verifying the hypotheses; the mechanism that forces the one-dimensionality is the analytic balance condition itself.
- comments
- 57 pages, 4 figures. Major revision
- fields
- Probability