Probabilistic representation of solutions to the parabolic $p$-Laplace equation

Abstract: This work is concerned with the probabilistic representation of solutions to the $p$-Laplace evolution equation $\frac{\partial u}{\partial t}={\rm div}(|\nabla u|^{p-2}\nabla u)$ in $(0,\infty)\times\mathbb{R}^d$, $u(0,x)=u_0(x),$ $x\in\mathbb{R}^d$. One proves that, if $p\geq 4$, and if $u_0$ is a probability density with compact support and $u_0\in L^2$, $|\nabla u_0|\in L^\infty$, then $u$ can be represented as $u(t,x)dx=\mathscr L_{X(t)}(dx)$, where $\mathscr L_{X(t)}$ denotes the time marginal law of $X$ at time $t$ with $X$ being a probabilistically weak solution to a corresponding McKean-Vlasov stochastic differential equation. This result is based on a new second order global regularity result for the weak solutions to the parabolic $p$-Laplace equation.

• The paper establishes a probabilistic representation for solutions to the nonlinear parabolic p-Laplace equation, linking it to McKean–Vlasov SDEs.
• It employs variational methods and monotone operator theory to derive global regularity estimates and prove the existence and uniqueness of weak solutions.
• The study lays a framework for stochastic simulation and numerical approximation of nonlinear diffusion models in applications like mathematical physics and statistical mechanics.

Probabilistic Representation for the Parabolic $p$-Laplace Equation

Problem Setting and Motivation

The paper "Probabilistic representation of solutions to the parabolic $p$-Laplace equation" (2604.26719) addresses the nonlinear evolution equation

$$\frac{\partial u}{\partial t} = \operatorname{div}(|\nabla u|^{p-2}\nabla u), \quad t > 0, \;\; x \in \mathbb{R}^d,$$

with initial data $u(0,x) = u_0(x)$, where $u_0$ is a probability density of compact support. The focus is on the case $p \geq 4$, which is technically challenging due to the degeneracy and nonlinear structure of the operator. The objective is to construct a probabilistic representation for the solution $u$ by relating it to the time marginal law of an SDE of McKean–Vlasov type.

This research builds on the classical superposition principle, deepening the connection between nonlinear PDEs (especially of Fokker–Planck type) and stochastic processes. Practical implications include the ability to leverage stochastic analysis tools (Itô calculus, martingale techniques, Malliavin calculus) for PDE regularity and qualitative properties, relevant in mathematical physics, statistical mechanics, and stochastic modeling.

Existence and Regularity Theory

The paper rigorously establishes existence and uniqueness of weak solutions in $L^2$-based spaces, exploiting the monotonicity and maximal accretivity property of the operator

$$A u := -\operatorname{div}(|\nabla u|^{p-2}\nabla u)$$

with domain tailored to the integrability and regularity requirements dictated by $p$.

The analysis deploys variational methods and monotone operator theory, notably:

• Weak solutions are constructed in $p$0 for initial data $p$1.
• Key energy estimates:
  • $p$2 is non-increasing.
  • The $p$3 contraction and positivity are maintained under evolution.
  • Finite propagation speed is quantified: the support expands at a controlled, sublinear rate in time.

A pivotal contribution is new global second-order regularity estimates: for initial data $p$4 with additional $p$5 regularity,

$p$6

with explicit dependence of constants on $p$7, the support radius $p$8, and norms of $p$9. These estimates subsume and extend prior local results, ensuring the required regularity for applying the superposition principle.

Probabilistic Representation via McKean–Vlasov SDE

The parabolic $$\frac{\partial u}{\partial t} = \operatorname{div}(|\nabla u|^{p-2}\nabla u), \quad t > 0, \;\; x \in \mathbb{R}^d,$$0-Laplace equation is recast as a nonlinear Fokker–Planck equation. The probabilistic counterpart is the time-marginal law of a stochastic process $$\frac{\partial u}{\partial t} = \operatorname{div}(|\nabla u|^{p-2}\nabla u), \quad t > 0, \;\; x \in \mathbb{R}^d,$$1 governed by the McKean–Vlasov SDE:

$$\frac{\partial u}{\partial t} = \operatorname{div}(|\nabla u|^{p-2}\nabla u), \quad t > 0, \;\; x \in \mathbb{R}^d,$$2

where $$\frac{\partial u}{\partial t} = \operatorname{div}(|\nabla u|^{p-2}\nabla u), \quad t > 0, \;\; x \in \mathbb{R}^d,$$3 is a standard Brownian motion in $$\frac{\partial u}{\partial t} = \operatorname{div}(|\nabla u|^{p-2}\nabla u), \quad t > 0, \;\; x \in \mathbb{R}^d,$$4. The process $$\frac{\partial u}{\partial t} = \operatorname{div}(|\nabla u|^{p-2}\nabla u), \quad t > 0, \;\; x \in \mathbb{R}^d,$$5 is a probabilistic weak solution, its time marginal distribution coincides with the evolving $$\frac{\partial u}{\partial t} = \operatorname{div}(|\nabla u|^{p-2}\nabla u), \quad t > 0, \;\; x \in \mathbb{R}^d,$$6:

$$\frac{\partial u}{\partial t} = \operatorname{div}(|\nabla u|^{p-2}\nabla u), \quad t > 0, \;\; x \in \mathbb{R}^d,$$7

This is predicated on advanced regularity estimates, which guarantee well-posedness of the SDE, even with the strong nonlinearity and potential degeneracy. The construction relies on recent extensions of the superposition principle for nonlinear Fokker–Planck equations ([Barbu, Röckner, Ann. Probab. 2020]), and continuity/compactness methods for stochastic flows.

Numerical Implications and Contradictory Findings

A strong result is the explicit global control of second-order regularity, which enables probabilistic representation for an extensive class of initial data ($$\frac{\partial u}{\partial t} = \operatorname{div}(|\nabla u|^{p-2}\nabla u), \quad t > 0, \;\; x \in \mathbb{R}^d,$$8 with compact support and appropriate Sobolev regularity), in contrast with prior works that only obtained local estimates or treated the Barenblatt profile.

No contradictory claims are made. However, the global-in-time regularity estimate refines previous approaches to regularity, and the extension to the full class of distributional solutions is technically non-trivial.

Theoretical and Practical Implications

The probabilistic representation has major ramifications for both theory and applications:

• Enables the use of stochastic calculus for analyzing nonlinear PDEs beyond linear Fokker–Planck models.
• Facilitates analysis of pathwise uniqueness, regularity, and singularity formation via probabilistic methods.
• Opens avenues for stochastic simulation, numerical approximation, and uncertainty quantification in high-dimensional nonlinear diffusion models.
• Provides a bridge between kinetic theory, mean-field models, and nonlinear diffusion in statistical physics, epidemics, and other applied fields.

Future Directions in AI and Stochastic PDEs

Future research can exploit the probabilistic structure to study long-time behavior, invariant measures, and ergodic properties for nonlinear evolutions. In applied mathematics and AI, stochastic representations of nonlinear PDEs are key for probabilistic numerics, data-driven modeling, and efficient simulation in generative models and neural transport.

The methodology may generalize to:

• Lower or intermediate $$\frac{\partial u}{\partial t} = \operatorname{div}(|\nabla u|^{p-2}\nabla u), \quad t > 0, \;\; x \in \mathbb{R}^d,$$9 regimes ($u(0,x) = u_0(x)$0), where degeneracy is subtler.
• Systems with anisotropy, inhomogeneity, or interaction terms.
• Coupled PDE/SDE systems in mass transport, stochastic control, and reinforcement learning.

Conclusion

The paper develops a robust framework for representing solutions of the parabolic $u(0,x) = u_0(x)$1-Laplace equation in terms of the time marginal law of solutions to a McKean–Vlasov SDE. This is contingent on novel global regularity estimates, applicable for $u(0,x) = u_0(x)$2 and compactly supported initial probability densities. The outcome broadens the scope of probabilistic analysis for nonlinear degenerate parabolic equations, strengthening connections between stochastic processes and PDE theory, and is poised to impact both theoretical analysis and computational techniques in nonlinear diffusion and related fields.