Table of Contents
Fetching ...

Beurling-Deny formula for Sobolev-Bregman forms

Michał Gutowski, Mateusz Kwaśnicki

TL;DR

The paper develops a full Beurling--Deny-type decomposition for Sobolev--Bregman forms $\mathscr{E}_p$ associated with a regular Dirichlet form $\mathscr{E}$ and its symmetric Markov semigroup $T_t$. It establishes a domain characterization $\mathscr{D}(\mathscr{E}_p)=\{u\in L^p(E): u^{p/2}\in \mathscr{D}(\mathscr{E})\}$ and a precise Beurling--Deny expansion of $\mathscr{E}_p$ into local, jump, and killing parts with explicit constants, extending known cases without translation-invariance or strong regularity assumptions. The approach hinges on elementary bounds for the heat kernel approximation, a detailed $L^2$-to-$L^p$ comparison, and a limit argument that recovers the energy contributions from $\mathscr{E}^{\mathrm{c}}$, $J$, and $k$; a Hardy--Stein identity follows as a natural corollary. The Euclidean-space example illustrates how the local gradient term and nonlocal jump term combine in the $p$-form, connecting energy measures to Bregman divergences and broadening applicability to nonlinear PDEs and stochastic analysis for general symmetric Markov processes.

Abstract

For an arbitrary regular Dirichlet form $\mathscr{E}$ and the associated symmetric Markovian semigroup $T_t$, we consider the corresponding Sobolev-Bregman form $\mathscr{E}_p(u) = -\tfrac{1}{p} \frac{d}{d t}\bigr\vert_{t = 0} \|T_t u\|_p^p$, where $p \in (1, \infty)$. We prove a variant of the Beurling-Deny formula for $\mathscr{E}_p$. As an application, we prove the corresponding Hardy-Stein identity. Our results extend previous works in this area, which either required that $\mathscr{E}$ is translation-invariant, or that $u$ is sufficiently regular.

Beurling-Deny formula for Sobolev-Bregman forms

TL;DR

The paper develops a full Beurling--Deny-type decomposition for Sobolev--Bregman forms associated with a regular Dirichlet form and its symmetric Markov semigroup . It establishes a domain characterization and a precise Beurling--Deny expansion of into local, jump, and killing parts with explicit constants, extending known cases without translation-invariance or strong regularity assumptions. The approach hinges on elementary bounds for the heat kernel approximation, a detailed -to- comparison, and a limit argument that recovers the energy contributions from , , and ; a Hardy--Stein identity follows as a natural corollary. The Euclidean-space example illustrates how the local gradient term and nonlocal jump term combine in the -form, connecting energy measures to Bregman divergences and broadening applicability to nonlinear PDEs and stochastic analysis for general symmetric Markov processes.

Abstract

For an arbitrary regular Dirichlet form and the associated symmetric Markovian semigroup , we consider the corresponding Sobolev-Bregman form , where . We prove a variant of the Beurling-Deny formula for . As an application, we prove the corresponding Hardy-Stein identity. Our results extend previous works in this area, which either required that is translation-invariant, or that is sufficiently regular.
Paper Structure (5 sections, 4 theorems, 97 equations, 2 figures)

This paper contains 5 sections, 4 theorems, 97 equations, 2 figures.

Key Result

Theorem 1.1

Let $\mathscr{E}$ be a regular Dirichlet form and $p \in (1, \infty)$. Then the domain $\mathscr{D}(\mathscr{E}_p)$ of the Sobolev--Bregman form $\mathscr{E}_p$ is characterised by and for every $u \in \mathscr{D}(\mathscr{E}_p)$ we have Furthermore, for every $u \in \mathscr{D}(\mathscr{E}_p)$ we have the following analogue of the Beurling--Deny formula: where $\tilde{u}$ is the quasi-continuo

Figures (2)

  • Figure 1: Functions defined in Lemma \ref{['lem:main']} (not to scale).
  • Figure 2: Regions considered in the proof of Lemma \ref{['lem:main']} (not to scale).

Theorems & Definitions (9)

  • Theorem 1.1: Beurling--Deny formula for Sobolev--Bregman forms
  • Corollary 1.2
  • Lemma 2.1
  • proof
  • Lemma 2.2
  • proof
  • proof : Proof of Theorem \ref{['thm:main']}, part I
  • proof : Proof of Theorem \ref{['thm:main']}, part II
  • proof : Proof of Theorem \ref{['thm:main']}, part III