Poisson kernels on the half-plane are bell-shaped
Mateusz Kwaśnicki · 2025
Read Poisson kernels on the half-plane are bell-shaped by Mateusz Kwaśnicki (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-2501-16068.txt.
Introduction
Mateusz Kwaśnicki proves that a family of kernels central to diffusion in a half-plane always has a bell shape. The Poisson kernel for an elliptic operator gives the distribution of where a wandering diffusion first hits the boundary. Its coefficients here depend only on the height above that boundary.
Bell-shaped means the kernel rises once and falls once, with the sign pattern of its derivatives fully ordered. In particular it has exactly two inflection points, curving down, then up, then down again. Such shape theorems are prized because they hold for every operator in the class at once. Probability and partial differential equations both gain a tidy, universal fact.
Catalog record
| CREATOR | Mateusz Kwaśnicki |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 51 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2501-16068
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Poisson kernels on the half-plane are bell-shaped Thanks: Work supported by the National Science Centre, Poland, grant no. 2023/49/B/ST1/04303 Mateusz Kwaśnicki Abstract . Consider a second-order elliptic operator L L in the half-plane ℝ × ( 0 , ∞ ) \mathbb{R}\times(0,\infty) with coefficients depending only on the second coordinate. The Poisson kernel for L L is used in the representation of positive L L -harmonic functions, that is, solutions of L u = 0 Lu=0 . In probabilistic terms, the Poisson kernel is the density function of the distribution of the diffusion in ℝ × ( 0 , ∞ ) \mathbb{R}\times(0,\infty) with generator L L at the hitting time of the boundary. We prove that the Poisson kernel for L L is bell-shaped: its n n th derivative changes sign n n times. In particular, it is unimodal and it has two inflection points (it is concave, then convex, then concave again). 2020 Mathematics Subject Classification 26A51, 35J25, 35J70, 60J60 1. Introduction and statement of the result Since the influential work of Caffarelli and Silvestre [ 3 ] , the harmonic extension technique has become a standard method of tackling nonlocal problems using local PDE methods. Identification of the square root of the Laplace operator with the Dirichlet-to-Neumann operator in half-space is a classical result, and [ 3 ] provides a similar description of other fractional powers of the Laplace operator. While similar ideas appeared earlier, also in a probabilistic context (the link between corresponding jump-type Markov processes and diffusions goes back to the work of Molchanov and Ostrovski [ 27 ] ; see also [ 4 , 5 , 9 , 26 , 31 ] ), Caffarelli and Silvestre stated the method clearly, thereby making it readily accessible to the PDE community. They were also the first to give an explicit and surprisingly simple expression for the corresponding Poisson kernel, see ( 1.1 ) below. Certain more general nonlocal operators are known to allow a similar harmonic extension technique; that is, they are linked to solutions of appropriate PDEs in a half-space. An explicit expression for the Poisson kernel is, however, rarely available. Below we consider a class of translation-invariant nonlocal operators, for which the harmonic extension technique was developed in [ 18 ] (see also [ 19 ] for a probabilistic counterpart). Using the results of [ 22 ] , we prove a certain ‘geometric’ property of the corresponding Poisson kernel: it is bell-shaped , that is, its n th derivative changes sign exactly n times for n=1,2,\ldots We state our main result, Theorem 1.1 , in Section 1.3 . First, however, we give a short introduction to the harmonic extension technique in Section 1.1 , and we discuss the elliptic equations covered by our methods in Section 1.2 . Probabilistic interpretation of our work is presented in Section 1.4 . We mention here that a parallel development in [ 34 ] covers the case of discrete variables, and that our main theorems contain as a special case the main results of [ 13 , 30 , 35 , 36 ] . 1.1. Harmonic extensions By the classical Poisson representation formula, given any bounded Borel function f on \mathbb{R}^{d} , the function u given by \displaystyle u(x,y) \displaystyle=\frac{\Gamma(\tfrac{d+1}{2})}{\pi^{(d+1)/2}}\int_{\mathbb{R}^{d}}\frac{y}{(\lvert x-x^{\prime}\rvert^{2}+y^{2})^{(d+1)/2}}\,f(x-x^{\prime})dx^{\prime} is harmonic in the upper half-space \mathbb{R}^{d}\times(0,\infty) and it converges to f(x) in an appropriate sense as y\to 0^{+} . Conversely, every bounded harmonic function u in the upper half-space is of the above form. We say that u is the harmonic extension of f , and it is now well-understood that various fine properties of f , such as its regularity, can be conveniently expressed in terms of u ; see, for example, Stein’s monograph [ 32 ] . The integral kernel \displaystyle P_{(x,y)}(x^{\prime}) \displaystyle=\frac{\Gamma(\tfrac{d+1}{2})}{\pi^{(d+1)/2}}\,\frac{y}{(\lvert x-x^{\prime}\rvert^{2}+y^{2})^{(d+1)/2}} is known as the Poisson kernel for the upper half-space. In [ 3 ] , Caffarelli and Silvestre generalised the above construction by considering \alpha\in(0,2) and \displaystyle u(x,y) \displaystyle=\frac{\Gamma(\tfrac{d+\alpha}{2})}{\alpha\pi^{d/2}\Gamma(\tfrac{\alpha}{2})}\int_{\mathbb{R}^{d}}\frac{y^{\alpha}}{(\lvert x-x^{\prime}\rvert^{2}+y^{2})^{(d+\alpha)/2}}\,f(x-x^{\prime})dx^{\prime}. They proved that u is a solution of the elliptic PDE Lu=0 , where \displaystyle Lu(x,y) \displaystyle=\nabla_{x,y}\cdot(y^{1-\alpha}\nabla_{x,y}u)(x,y), and again u(x,y) converges to f(x) in an appropriate sense as y\to 0^{+} . This enabled the use of standard (local) PDE techniques in the study of a nonlocal operator, the fractional Laplace operator \displaystyle-(-\Delta)^{\alpha/2}f(x) \displaystyle=\frac{2^{\alpha}\Gamma(\tfrac{d+\alpha}{2})}{\pi^{d/2}\lvert\Gamma(-\tfrac{\alpha}{2})\rvert}\operatorname{pv}\int_{\mathbb{R}^{d}}\frac{f(x^{\prime})-f(x)}{\lvert x^{\prime}-x\rvert^{d+\alpha}}\,dx^{\prime}, because it coincides with the corresponding Dirichlet-to-Neumann operator : \displaystyle-(-\Delta)^{\alpha/2}f(x) \displaystyle=\frac{\lvert\Gamma(-\tfrac{\alpha}{2})\rvert}{\alpha 2^{\alpha}\Gamma(\tfrac{\alpha}{2})}\lim_{y\to 0^{+}}\frac{u(x,y)-u(x,0)}{y^{\alpha}}\,. By analogy, the kernel function (1.1) \displaystyle P_{(x,y)}(x^{\prime}) \displaystyle=\frac{\Gamma(\tfrac{d+\alpha}{2})}{\alpha\pi^{d/2}\Gamma(\tfrac{\alpha}{2})}\,\frac{y^{\alpha}}{(\lvert x-x^{\prime}\rvert^{2}+y^{2})^{(d+\alpha)/2}} is said to be the Poisson kernel for L , and the method described above is now known as the Caffarelli–Silvestre extension technique. We refer to Section 10 in [ 11 ] or Section 6 in [ 16 ] for further discussion and references. More generally, one can replace L with an arbitrary second-order elliptic operator in the upper half-space, and use an analogous construction to study the corresponding Dirichlet-to-Neumann operator K . It turns out that if we require L to be invariant under translations and rotations in the x variable, then -K is a complete Bernstein function of -\Delta , and, conversely, every complete Bernstein function of -\Delta is the Dirichlet-to-Neumann operator -K for an essentially unique L of this kind. This is a consequence of Krein’s spectral theory of strings, and the details were given by Mucha and the author in [ 21 ] ; see [ 1 , 10 , 12 , 33 ] for closely related research. However, with very few exceptions, no explicit formula is known for the Poisson kernel for L and, to the best knowledge of the author, it has not even been proved that the Poisson kernel is unimodal. When d=1 , there is an analogous description for operators L which are invariant under translations in the x variable, but not necessarily symmetric. In this case the corresponding Dirichlet-to-Neumann operators K have a kernel function which is completely monotone on (0,\infty) and absolutely monotone on (-\infty,0) , and again there is a one-to-one correspondence between nonlocal operators K of this form and the corresponding elliptic operators L (up to certain natural transformations). This result was given in [ 18 ] , and its probabilistic variant is described in [ 19 ] ; the crucial ingredient is, however, due to Eckhardt and Kostenko, and de Branges; see [ 6 ] and the references therein. The Poisson kernel for these operators L is again described only in terms of Fourier transform (see ( 1.4 ) and ( 1.5 )), and except in a few special cases, explicit expressions are not available. 1.2. Elliptic equations In the present paper we prove that in the context of the previous paragraph, the Poisson kernel P_{(x,y)}(x^{\prime}) for the upper half-plane is a bell-shaped function of x^{\prime} . This means that it is unimodal (increasing and then decreasing), it has two inflection points (it is convex, then concave, then again convex), and so on: the n th derivative of the Poisson kernel changes sign exactly n times. Before we state this result rigorously in Theorem 1.1 , let us describe in more detail the class of operators L that it covers. Following [ 18 ] , we consider functions u defined either on the upper half-plane \mathbb{R}\times(0,\infty) , or a strip \mathbb{R}\times(0,R)