Asymptotic average solutions for second order hypoelliptic PDEs
Alessia E. Kogoj · 2025
Read Asymptotic average solutions for second order hypoelliptic PDEs by Alessia E. Kogoj (2025) 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-2504-19240.txt.
Introduction
Alessia Kogoj introduces a new idea in this 2025 paper on a class of partial differential equations whose diffusion spreads unevenly. For such equations, exact average-value formulas may not exist. Kogoj defines a weaker notion, an asymptotic average solution, that still lets a classic Poisson equation with continuous data be solved point by point. The approach follows a line of work that traces back to a 1909 paper by Pizzetti.
Catalog record
| CREATOR | Alessia E. Kogoj |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 20 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2504-19240
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Asymptotic average solutions for second order hypoelliptic PDEs Alessia E. Kogoj Abstract . Following an analogous procedure with that used in [ 19 ] , in turn inspired by a 1909 paper by Pizzetti [ 30 ] , we introduce the notion of asymptotic average solutions for hypoelliptic linear partial differential operators with non-negative characteristic form. This notion makes every Poisson equation ℒ u ( x ) = − f ( x ) {\mathcal{L}}u(x)=-f(x) with continuous data f f pointwise solvable. 2020 Mathematics Subject Classification 35K65, 35K70, 35H10, 35H20, 35D99, 35B05, 35D30 1. Introduction We are dealing with partial differential operators of the type (1.1) {\mathcal{L}}=\sum_{i,j=1}^{n}\partial_{x_{i}}(a_{ij}(x)\partial_{x_{j}})+\sum_{i=1}^{n}b_{i}(x)\partial_{x_{i}}\mbox{ in }{{\mathbb{R}}^{{n}}}, where the coefficients a_{ij} and b_{i} are smooth functions in {{\mathbb{R}}^{{n}}} . The matrix A=(a_{ij}),i,j=1,\ldots,n is supposed to be symmetric and non-negative definite at any point and not totally degenerate, that is \inf_{{\mathbb{R}}^{{n}}}a_{11}>0 . We also assume that the operator {\mathcal{L}} is hypoelliptic and endowed with a non-negative smooth fundamental solution \Gamma:\{(x,y)\in{{\mathbb{R}}^{{n}}}\times{{\mathbb{R}}^{{n}}}\ |x\neq y\}\longrightarrow{\mathbb{R}}, such that (i) for any fixed x\in{{\mathbb{R}}^{{n}}} , \Gamma(\cdot,x) and \Gamma(x,\cdot) belong to L^{1}_{\rm loc}({{\mathbb{R}}^{{n}}}) ; (ii) for every \varphi\in C_{0}^{\infty}({{\mathbb{R}}^{{n}}}) and x\in{{\mathbb{R}}^{{n}}} , {\mathcal{L}}\int_{{{\mathbb{R}}^{{n}}}}\Gamma(x,y)\varphi(y)\ dy\ =\ \int_{{{\mathbb{R}}^{{n}}}}\Gamma(x,y){\mathcal{L}}\varphi(y)\ dy\ =\ -\varphi(x); (iii) \displaystyle\limsup_{x\rightarrow y}\Gamma(x,y)=\limsup_{y\rightarrow x}\Gamma(x,y)=\infty; (iv) \displaystyle\lim_{|x|\rightarrow\infty}\left(\sup_{|y|\leq M}\Gamma(x,y)\right)=\lim_{|y|\rightarrow\infty}\left(\sup_{|x|\leq M}\Gamma(x,y)\right)=0 , for every real positive M. Then, given x\in{{\mathbb{R}}^{{n}}} and r>0 , we define the {\mathcal{L}} -ball of center x and radius r as (1.2) \Omega_{r}(x)=\left\{y\in{{\mathbb{R}}^{{n}}}\,\,:\,\,\Gamma(x,y)>\frac{1}{r}\right\}, and we assume that the following Poisson–Jensen-type representation formula holds for any function u\in C^{2}(\Omega) , where \Omega is an open subset of {{\mathbb{R}}^{{n}}} . For every {\mathcal{L}} -ball \Omega_{r}(x) such that \overline{\Omega_{r}(x)}\subseteq\Omega we have (1.3) \displaystyle u(x) \displaystyle= \displaystyle\frac{1}{r}\int_{\Omega_{r}(x)}\!\!\!u(y)\frac{\langle A(y)\nabla_{y}\Gamma(x,y),\nabla_{y}\Gamma(x,y)\rangle}{(\Gamma(x,y))^{2}}\ dy \displaystyle-\frac{1}{r}\int_{0}^{r}\left(\int_{\Omega_{\rho}(x)}\left(\Gamma(x,y)-\frac{1}{\rho}\right){\mathcal{L}}(u)(y)\ dy\right)d\rho. We will write (1.4) \displaystyle u(x)=M_{r}(u)(x)-N_{r}({\mathcal{L}}u)(x). All the properties we require are satisfied, in particular, by the classical heat operator, the heat operators on Carnot groups, the Kolmogorov operators studied in [ 23 ] , the operators constructed with the link procedure introduced in [ 17 ] and the more general class of hypoelliptic ultraparabolic operators introduced and studied in [ 22 ] , which contains all the previous ones. For this wide class of operators, a fundamental solution with good properties containing the ones required in this work has been constructed in [ 22 ] and representation formulae of Poisson–Jensen-type have been proved in [ 5 ] . We also have to remark that Kogoj and Tralli have given asymptotic characterizations of the subsolutions in [ 21 ] . Moreover, we mention the works [ 29 ] , [ 4 ] , [ 18 ] , [ 6 ] , [ 7 ] , [ 8 ] , [ 20 ] , [ 16 ] , where Riesz and Poisson-Jensen representation formulae, Gaussian estimates of the fundamental solution, Harnack-type inequalities, Liouville-type theorems, the Cauchy and the Dirichlet problems have been studied for this class of operators. In this paper, by using the above Poisson–Jensen formula ( 1.3 ), we give at first the following definition of asymptotic average solution to {\mathcal{L}}u=f . Definition 1.1 . Let us define Q_{r}(x)=N_{r}(1)(x) . Let u and f be continuous functions in an open set \Omega\subseteq{{\mathbb{R}}^{{n}}} . We say that u is an asymptotic average solution to {\mathcal{L}}u=f if {\mathcal{L}}_{a}u(x):=\lim_{r\rightarrow 0}\frac{M_{r}(u)(x)-u(x)}{Q_{r}(x)} exists in {\mathbb{R}} at every point x\in\Omega and {\mathcal{L}}_{a}u(x)=f(x)\mbox{ for every }x\in\Omega. This definition completes the one given by Kogoj and Lanconelli in the recent paper [ 19 ] - where operators of the second order just in principal part are considered - and the one given by Gutierrez and Lanconelli in [ 13 ] for a class of second order operators with non negative characteristic form with drift in the case f=0 . When {\mathcal{L}} is the classical Laplacian, it coincides with the one given by Pizzetti in 1909 in in his paper on the average of the values that a function takes on the surface of a sphere [ 30 ] . Furthermore, we will observe that if u is a compactly supported continuous function and is an asymptotic average solution to {\mathcal{L}}_{a}u=f\mbox{ in }{{\mathbb{R}}^{{n}}} , then, {\mathcal{L}}_{a}u=f\mbox{ in }{{\mathbb{R}}^{{n}}} if and only if {\mathcal{L}}u=f in the weak sense of the distributions. In recent years, mean value formulae that characterize classical or viscosity solutions to both linear and nonlinear second order PDEs have been studied by several authors. We refer to the works [ 13 , 26 , 25 , 15 , 11 , 12 , 24 , 14 , 28 , 27 , 1 ] and the references therein. Our definition of asymptotic average solution enables us to solve the Poisson problem {\mathcal{L}}_{a}u=-f\mbox{ in }{{\mathbb{R}}^{{n}}}. We state and prove the following Pizzetti-type result. Theorem 1.2 . For every f:{{\mathbb{R}}^{{n}}}\rightarrow{\mathbb{R}} compactly supported continuous function, the function u_{f}(x):=\int_{{\mathbb{R}}^{{n}}}\Gamma(x,y)f(y)\ dy,\qquad x\in{{\mathbb{R}}^{{n}}}. is an asymptotic average solution to {\mathcal{L}}_{a}u=-f\mbox{ in }{{\mathbb{R}}^{{n}}}. Pizzetti proved that the Newtonian potentials of continuous bounded functions are solutions of the classical Poisson-Laplace equations in an asymptotic sense ( [ 30 , Equation (8)] , see also [ 19 , Theorem A] ). He deduced this result from his celebrated formula ( [ 30 , Equation (A)] , see also [ 9 , Part IV, Chapter 3, pp. 287–288] ) where the spherical integral mean in {\mathbb{R}}^{3} of a smooth function over an Euclidean ball is expressed as a power series with respect to the radius of the ball having the iterated Laplacians as coefficients. While Da Lio and Rodino, [ 10 ] , generalized Pizzetti’s formula to the heat operator, to the best of our knowledge the result we are considering has not been extended so far in the parabolic setting, even in the case of the heat operator. To prove our result, we follow an analogous procedure with that used in [ 19 ] . In Section 2 we describe some properties of the {\mathcal{L}} -balls. We will exploit these properties in Section 3 to prove a Poisson-Jensen type representation formula for the \Gamma -potentials u_{f} . Thanks to this representation formula, we will be able to prove in Section 4 that the Poisson equations relating to the {\mathcal{L}} operators are always solvable by the \Gamma -potentials u_{f} according to our definition of asymptotic average solution. In Section 5 we compare the notions of classical, weak and asymptotic average solutions; in particular we show that our asymptotic average solutions are weak solutions and vice versa. 2. The {\mathcal{L}} -balls: some properties The notion of asymptotic average solution is based on some representation formulae on the superlevel set of \Gamma , the {\mathcal{L}} -balls defined in ( 1.2 ). From the properties of the fundamental solution, it follows that \Omega_{r}(x) is a nonempty bounded open set of {{\mathbb{R}}^{{n}}} . In addition, \bigcap_{r>0}\Omega_{r}(x)=\{x\} and \frac{|\Omega_{r}(x)|}{r}\longrightarrow 0\mbox{ as }r\longrightarrow 0, where |\cdot| denotes its Lebesgue measure. Moreover, we prove some lemmata involving {\mathcal{L}} -ball useful in the subsequent sections. Lemma 2.1 . Given a compact subset G of {{\mathbb{R}}^{{n}}} and a positive radius r , the set {G}_{r} G_{r}:=\bigcup_{x\in G}\Omega_{r}(x), is bounded. Proof. Let’s argue by contradiction and assume G_{r} unbounded. We can find a sequence (z_{n}) in G_{r} , |z_{n}|\longrightarrow\infty. As z_{n} belongs to G_{r} , there exists x_{n}\in G such that z_{n}\in\Omega_{r}(x_{n})\iff\Gamma(x_{n},z_{n})>\frac{1}{r}. It follows that for every n\in{\mathbb{N}} \frac{1}{r}<\Gamma(x_{n},z_{n})\leq\sup_{x\in G}\Gamma(x,z_{n}). If now we let