An Aronson-Bénilan / Li-Yau estimate in the JKO scheme in small dimension

Fanch Coudreuse · 2026

Read An Aronson-Bénilan / Li-Yau estimate in the JKO scheme in small dimension by Fanch Coudreuse (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-04169.txt.

Introduction

Coudreuse proves a new estimate for a step-by-step scheme that simulates diffusion equations in one and two dimensions. On simple shapes such as cubes, half-spaces and the torus, his argument rests on a maximum principle for the Hessian determinant of Brenier potentials, applied one step at a time. The result gives density bounds uniform in the time step, and settles the optimality conditions for fast diffusion.

Catalog record

CREATORFanch Coudreuse
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE86377
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.04169

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

An Aronson-Bénilan / Li-Yau estimate in the JKO scheme in small dimension Fanch Coudreuse Abstract . We derive an Aronson-Bénilan / Li-Yau estimate in the JKO scheme associated to the porous-medium, heat, and fast-diffusion equations, in dimensions 1 1 and 2 2 , and on simple domains (cubes, quarter-space, half-spaces, whole space, and the torus). Our method is based on a maximum principle for the determinant of the Hessian of Brenier potentials, iterated as a one-step improvement along the scheme. As a consequence, we obtain local L ∞ L^{\infty} bounds on the density, uniform in the time step, consistent with the continuous-time result. As a byproduct, we rigorously derive the optimality conditions in the fast-diffusion case, filling a gap in the literature. 1. Introduction This paper focuses on the JKO scheme approximation to the porous-medium, heat and fast-diffusion equations \partial_{t}\rho=\Delta\rho^{m} on a convex domain \Omega , or in the torus \mathbb{T}^{d} , in the space of probability measures with finite second moment, with Neumann boundary condition and given initial data. The different names for this equation refer to the different regimes for the values of m : • The case m>1 corresponds to the porous-medium equation. It exhibits slow-diffusion-type behavior, and a free-boundary can appear if one starts from a compactly supported initial datum. We refer to the monograph [ 42 ] by Vazquez for the general theory of this equation. • The linear case m=1 corresponds to the classical heat equation, where strong smoothing effects occur, and solutions are instantaneously strictly positive. • Finally, the case 0<m<1 corresponds to the so-called fast-diffusion regime. Solutions are again strictly positive for positive time. In this setting, one can distinguish further regimes for the parameter m . In our analysis, two of them will be relevant: – The regime m>m_{c}^{1}:=1-\frac{2}{d} to ensure well-posedness in the space of probability measures. – The regime m>m_{c}^{2}:=1-\frac{2}{d+2} corresponding to existence of solutions in the space of probability measures with finite second moment. In both cases, one can check using Barenblatt profiles (corresponding to Dirac initial datum) that this threshold is sharp. We refer to the survey [ 6 ] by Bonforte and Figalli for more results on this equation. Note that all these equations are particular instances of the general filtration equation \partial_{t}\rho=\Delta\phi(\rho) . Since the seminal work of Jordan, Kinderlehrer and Otto [ 26 ] , it is now well understood that a large class of diffusion equations posed on the space of probability measures can be interpreted as the gradient flow on the Wasserstein space. While this can be made rigorous using the general theory of gradient flows in metric spaces developed by Ambrosio, Gigli and Savaré in [ 1 ] , another classical approach to tackle this interpretation is to use the JKO (Jordan-Kinderlehrer-Otto) scheme, which can be seen as an implicit Euler scheme in the Wasserstein space. This scheme takes the following form: given a time step \tau and a functional \mathscr{E} over the space of probability measures with finite second moment, one constructs a sequence (\rho_{k}^{\tau})_{k\geq 0} by fixing \rho_{0}\in\mathcal{P}_{2}(\Omega) and iterating \rho_{k}^{\tau}\in\rm{argmin}_{\eta\in\mathcal{P}_{2}(\Omega)}\mathscr{E}[\eta]+\frac{1}{2\tau}W_{2}^{2}(\eta,\rho_{k}^{\tau}) Then one expects that the curve obtained by interpolation of the values of the sequence converges to a weak solution to the equation \partial_{t}\rho=\nabla\cdot\left(\rho\nabla\frac{\delta\mathscr{E}}{\delta\rho}[\rho]\right) with \rho_{0} as initial datum, and suitable Neumann boundary conditions, where \frac{\delta\mathscr{E}}{\delta\rho} is the first variation of \mathscr{E} with respect to linear perturbation of the measures. This convergence can be made rigorous using the general theory of [ 1 ] , or by [ 40 , Chapter 8] , provided that the scheme exists, the initial datum is of finite energy (i.e. \mathscr{E}[\rho_{0}]<+\infty ), and the functional \mathscr{E} admits \lambda -convexity with respect to Wasserstein geodesics. For our setting of interest, the functional \mathscr{E} is taken to be of the form \mathscr{E}_{m}[\rho]=\int_{\Omega}f_{m}(\rho)\differential{\cl{L}^d} for \rho\ll\mathcal{L}^{d} , where f_{m}(z)=\frac{1}{m-1}z^{m} for m\neq 1 , and f_{1}(z)=z\log z . This functional falls into the general theory of geodesically convex functionals provided that m\geq m_{c}^{\rm{geo}}=1-\frac{1}{d} , in which case f_{m} satisfies the McCann conditions [ 35 ] [ 40 , Section 7.3.2] . A natural question in the study of the JKO scheme is whether the qualitative and quantitative properties known for the continuous flow can be recovered at the discrete level, uniformly in the time step \tau . Such results are desirable for at least two reasons: they provide robustness of the scheme in recovering the behavior of the continuous in time equation, and they can be leveraged to improve convergence rates. Several properties of the continuous flow have been investigated in this direction: Lipschitz and continuity estimates [ 29 ] [ 20 ] [ 11 ] , BV and Sobolev estimates with applications to L^{2}_{t}H^{2}_{x} -convergence [ 38 ] [ 17 ] [ 19 ] [ 39 ] , or comparison principle and L^{1} -contraction [ 24 ] [ 31 ] [ 23 ] . One particularly desirable estimate is the Aronson-Bénilan / Li-Yau estimate. It states that, for m>m_{c}^{1} , and for any positive solution \rho to \partial_{t}\rho=\Delta\rho^{m} , the pressure variable —defined by p=\frac{m}{m-1}\rho^{m-1} for m\neq 1 , and p=\log\rho for m=1 —, satisfies the sub-harmonic lower bound: (1.1) \Delta p\geq-\frac{\alpha_{d,m}}{t}\qquad\alpha_{d,m}=\frac{d}{d(m-1)+2} In the context of the heat equation, this inequality was proved by Li and Yau [ 32 ] [ 45 ] and bears their name. We note that in this context, on \mathbb{R}^{d} or \mathbb{T}^{d} , Hamilton [ 22 ] derived the stronger semi-convexity estimate D^{2}\log\rho\succeq-\frac{1}{2t} , called the Li-Yau-Hamilton matrix inequality. The remaining cases were tackled by Aronson and Bénilan [ 3 ] [ 2 ] (see [ 42 , Chapter 9] for a review of this estimate in the context of the porous-medium equation). Subsequently, this estimate has been extended to more general frameworks: smooth manifolds [ 34 ] , filtration equations [ 16 ] , with extension to L^{p} -version, eventually with a source term [ 4 ] , or to the Keller-Segel model in [ 18 ] . This estimate is fundamental, as it is the cornerstone for deriving L^{1} - L^{\infty} -regularization effects for the equation (see Lemma 2.12 ), and for studying the regularity of the solution [ 2 ] [ 9 ] and its free-boundary [ 8 ] . The first study of such an inequality in the JKO scheme was performed by P.W. Lee [ 30 ] for the heat equation in the torus, where he proved that a version of the Li-Yau-Hamilton inequality holds at the level of the JKO scheme, at least for regular initial data. This was then extended by the author in [ 15 ] , still in the torus, for a more general class of equations of granular-medium type: \partial_{t}\rho=\Delta\rho+\nabla\cdot(\rho[\nabla V+\nabla W*\rho]) , under no assumptions on the initial datum; this estimate was then used to derive L^{2}_{t,loc}H^{2}_{x} -strong convergence of the scheme. In this paper, we take a first step toward a proof of an Aronson-Bénilan estimate in the JKO scheme by focusing on small dimension and simple domains. 1.1. Main Result We consider an iteration of the JKO scheme starting from some measure \rho_{0}\in\mathcal{P}_{2}(\Omega) : \rho_{k+1}^{\tau}\in\argmin_{\eta\in\mathcal{P}_{2,ac}(\Omega)}\mathscr{E}_{m}[\eta]+\frac{1}{2\tau}W_{2}^{2}(\eta,\rho_{k}^{\tau}) assuming m>m_{c}^{1} , and additionally m>m_{c}^{2} if \Omega is unbounded. Although the super-linear case m\geq 1 is well understood and can be found

Read the complete text