OctoInformBeta

OpenAI preprint, September 2026 · Reciprocal Green functions and a squared-distance kernel

Theorem 5.1: Finite restrictions and the full circle kernel under a subcritical weighted Dirichlet bound

Conditional negativity of a boundary distance kernel.

Hover over a blue or green term in the text, or a node in the graph, to find it in the other. Select it to read its declaration both in Lean and in words.

The Setting

For a weighted Dirichlet problem on the unit disk, the Green and Poisson kernels specify an interior evaluation kernel and a boundary interaction kernel. The distance kernel compares two boundary points through their evaluation at a fixed interior point. The open unit disk and its boundary in the Euclidean plane are D={x∈ℝ2:|x|<1},S1={s∈ℝ2:|s|=1}.D=\{x\in\mathbb R^2:|x|<1\},\qquad S^1=\{s\in\mathbb R^2:|s|=1\}.(1) Let dxdx denote area measure, and let ν\nu be a nonnegative measure satisfying ν≤Cdx|D\nu\leq C\,dx|_D for a finite constant C≥0C\geq0.

A test function φ∈Cc∞(D)\varphi\in C_c^\infty(D) is a smooth real function on the plane with compact closed support contained in DD.✦ For a direction e∈ℝ2e\in\mathbb R^2, its directional derivative ∂eφ\partial_e\varphi and the vector field φe\varphi e are regarded as scalar and vector L2(D,dx)L^2(D,dx) classes. A square-integrable vector field gg is the weak gradient of a square-integrable real function vv when ∫Dv∂eφdx=−∫D(g⋅e)φdx(φ∈Cc∞(D),e∈ℝ2).\int_D v\,\partial_e\varphi\,dx =-\int_D (g\cdot e)\varphi\,dx \qquad(\varphi\in C_c^\infty(D),\ e\in\mathbb R^2).(2)✦ The Sobolev space H1(D)H^1(D) is the weak gradient graph of (2) in L2(D,dx)×L2(D,dx;ℝ2)L^2(D,dx)\times L^2(D,dx;\mathbb R^2), with the Euclidean product norm.✦ The Sobolev class of a pair (v,g)(v,g) in the weak gradient graph on DD is the pair of their L2L^2 classes; the value and gradient maps on H1(D)H^1(D) are the continuous linear maps selecting vv and gg.

The Sobolev embedding of test functions sends φ\varphi to its value class and classical gradient.✦ The Dirichlet Sobolev space H01(D)H_0^1(D) is the closure in H1(D)H^1(D) of the pairs consisting of a test function and its classical gradient: H01(D)={(φ,∇φ):φ∈Cc∞(D)}¯H1(D).H_0^1(D)=\overline{\{(\varphi,\nabla\varphi): \varphi\in C_c^\infty(D)\}}^{\,H^1(D)}.(3) The Dirichlet value and gradient maps are the restrictions of the Sobolev maps to (3). The weighted L2L^2 inclusion J:L2(D,dx)→L2(ν)J:L^2(D,dx)\to L^2(\nu) sends an area class to its ν\nu class✦, and the weighted Sobolev value map W:H01(D)→L2(ν)W:H_0^1(D)\to L^2(\nu) is its composition with the Dirichlet value map.✦ A subcritical bound for the weighted Sobolev value map W:H01(D)→L2(ν)W:H_0^1(D)\to L^2(\nu) with constant dd is the following inequality, where 0≤d<10\leq d<1: ∥Wu∥L2(ν)2≤d∥∇u∥L2(D,dx)2(u∈H01(D)).\|Wu\|_{L^2(\nu)}^2\leq d\,\|\nabla u\|_{L^2(D,dx)}^2 \qquad(u\in H_0^1(D)).(4)

The Boundary Kernels

The complex coordinate isometry identifies x1+ix2x_1+ix_2 with (x1,x2)∈ℝ2(x_1,x_2)\in\mathbb R^2.✦ Under this identification, the complex disk Green and Poisson kernels and the complex boundary interaction give their planar counterparts GG, PP and RR: G(x,y)=14πlog⁡(1+(1−|x|2)(1−|y|2)|x−y|2)(x,y∈D,x≠y),P(s,x)=1−|x|22π|s−x|2(s∈S1,x∈D),R(s,t)=1π|s−t|2(s,t∈S1,s≠t).\begin{aligned} G(x,y)&=\frac1{4\pi}\log\!\left( 1+\frac{(1-|x|^2)(1-|y|^2)}{|x-y|^2}\right) &&(x,y\in D,\ x\ne y),\\ P(s,x)&=\frac{1-|x|^2}{2\pi|s-x|^2} &&(s\in S^1,\ x\in D),\\ R(s,t)&=\frac1{\pi|s-t|^2} &&(s,t\in S^1,\ s\ne t). \end{aligned}(5) The diagonal values of GG and RR are zero.

The area Green integral operator AA on L2(D,dx)L^2(D,dx) is the continuous kernel operator associated with GG in (5).✦ The weighted Green operator TT on L2(ν)L^2(\nu) and the Neumann series resolvent 𝒮(B)\mathcal S(B) of a continuous real linear operator BB on a Banach space are (Af)(x)=∫DG(x,y)f(y)dy,T=JAJ*,𝒮(B)=∑n=0∞Bn,(Af)(x)=\int_D G(x,y)f(y)\,dy,\qquad T=JAJ^*,\qquad \mathcal S(B)=\sum_{n=0}^{\infty}B^n,(6) where J*J^* is the Hilbert adjoint of JJ.

The smoothed Poisson kernel P1:S1×D→ℝP_1:S^1\times D\to\mathbb R is the Green integral of the Poisson kernel PP with respect to ν\nu.✦ For p∈Dp\in D and s∈S1s\in S^1, the Green row rpr_p and smoothed Poisson row bsb_s are the L2(ν)L^2(\nu) classes specified by P1(s,x)=∫DG(x,y)P(s,y)dν(y),rp=[G(⋅,p)],bs=[P1(s,⋅)].P_1(s,x)=\int_D G(x,y)P(s,y)\,d\nu(y),\qquad r_p=[G(\cdot,p)],\qquad b_s=[P_1(s,\cdot)].(7) The full boundary evaluation kernel K:D×S1→ℝK:D\times S^1\to\mathbb R and the full boundary interaction kernel N:S1×S1→ℝN:S^1\times S^1\to\mathbb R are the functions defined by K(p,s)=P(s,p)+P1(s,p)+⟨rp,𝒮(T)bs⟩L2(ν),N(s,t)=R(s,t)+∫DP(s,x)P(t,x)dν(x)+∫DP(s,x)P1(t,x)dν(x)+⟨bs,𝒮(T)bt⟩L2(ν).\begin{aligned} K(p,s)&=P(s,p)+P_1(s,p) +\langle r_p,\mathcal S(T)b_s\rangle_{L^2(\nu)},\\ N(s,t)&=R(s,t)+\int_D P(s,x)P(t,x)\,d\nu(x)\\ &\quad+\int_D P(s,x)P_1(t,x)\,d\nu(x) +\langle b_s,\mathcal S(T)b_t\rangle_{L^2(\nu)}. \end{aligned}(8) For a fixed p∈Dp\in D, the boundary distance kernel Dp:S1×S1→ℝD_p:S^1\times S^1\to\mathbb R is the quotient formed from (8), with zero diagonal: Dp(s,t)={K(p,s)K(p,t)N(s,t),s≠t,0,s=t.D_p(s,t)= \begin{cases} \displaystyle\frac{K(p,s)K(p,t)}{N(s,t)},&s\ne t,\\[4pt] 0,&s=t. \end{cases}(9)✦

For a finite index set II and a real matrix MM indexed by II, its bilinear form on ℝI\mathbb R^I is x𝖳My=∑i,j∈IxiMijyjx^{\mathsf T}My=\sum_{i,j\in I}x_iM_{ij}y_j for x,y∈ℝIx,y\in\mathbb R^I.✦ A real matrix MM indexed by a finite set II is conditionally negative when it is symmetric and x𝖳Mx≤0for every x∈ℝI with ∑i∈Ixi=0.x^{\mathsf T}Mx\leq0\qquad \text{for every }x\in\mathbb R^I\text{ with }\sum_{i\in I}x_i=0.(10) A real kernel HH on a set XX is a conditionally negative distance kernel when it is symmetric, vanishes on the diagonal, and satisfies the finitely supported version of (10): ∑s,t∈Xc(s)H(s,t)c(t)≤0for every finitely supported c:X→ℝ with ∑s∈Xc(s)=0.\sum_{s,t\in X}c(s)H(s,t)c(t)\leq0 \quad\text{for every finitely supported }c:X\to\mathbb R \text{ with }\sum_{s\in X}c(s)=0.(11)✦

The Claim

Theorem 1. Let ν≤Cdx|D\nu\leq C\,dx|_D be a nonnegative measure with a finite domination bound, and suppose that a constant 0≤d<10\leq d<1 satisfies (4). Fix p∈Dp\in D. For every finite index set II and every family (si)i∈I(s_i)_{i\in I} in S1S^1, the matrix in (12) is conditionally negative: (Dp(si,sj))i,j∈Iis conditionally negative.\bigl(D_p(s_i,s_j)\bigr)_{i,j\in I} \quad\text{is conditionally negative}.(12)✦ On the full circle, DpD_p is a conditionally negative distance kernel in the sense of (11): Dpis a conditionally negative distance kernel on S1.D_p\quad\text{is a conditionally negative distance kernel on }S^1.(13)✦

In the paper, the result reads: if dν=adxd\nu=a\,dx on DD, with a∈L∞(D)a\in L^\infty(D) nonnegative, and some 0<d<10<d<1 satisfies ∫D|ψ|2dν≤d∫D|∇ψ|2dx\int_D|\psi|^2\,d\nu\leq d\int_D|\nabla\psi|^2\,dx for every ψ∈H01(D)\psi\in H_0^1(D), then for the associated kernels K,NK,N and every p∈Dp\in D, the kernel Dp(s,t)=K(p,s)K(p,t)/N(s,t)D_p(s,t)=K(p,s)K(p,t)/N(s,t) for s≠ts\ne t, with Dp(s,s)=0D_p(s,s)=0, is conditionally negative semidefinite on S1S^1.

What Is Checked

The assertions concern the open disk and circle in (1), the weighted Sobolev value map under (4), and the resolvent kernels used in (9).

The Green rows G(x,⋅)G(x,\cdot) belong to L2(D,dx)L^2(D,dx) for almost every x∈Dx\in D✦, and their L2L^2 row norms are uniformly bounded on DD. The L2L^2 kernel integral construction sends every L2L^2 datum to a square-integrable function.✦