OctoInformBeta

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

Lemma 5.6: Bordering by reciprocal interior evaluation values allows repeated boundary points

Subcritical boundary reciprocals have at most one positive square.

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

A weighted Dirichlet Green operator on the disk gives kernels for interaction between boundary points and evaluation at an interior point. Finite samples of these kernels supply the boundary block and the border of a real matrix.

The unit disk D⊂ℝ2D\subset\mathbb R^2 and unit circle S1⊂ℝ2S^1\subset\mathbb R^2 are the open ball and its boundary in (1): D={x∈ℝ2:‖x‖<1},S1={s∈ℝ2:‖s‖=1}.D=\{x\in\mathbb R^2:\lVert x\rVert<1\}, \qquad S^1=\{s\in\mathbb R^2:\lVert s\rVert=1\}.(1) Write dAdA for planar area measure. The measure ν\nu has the finite density bound (2) relative to disk area measure: ν≤CdA|D,0≤C<∞.\nu\le C\,dA|_D,\qquad 0\le C<\infty.(2)

Sobolev Data

For a planar set UU, a test function φ∈Cc∞(U)\varphi\in C_c^\infty(U) is a smooth real function on the plane whose closed support is compact and contained in UU.✦ For a planar set UU, a test function φ∈Cc∞(U)\varphi\in C_c^\infty(U) and a vector e∈ℝ2e\in\mathbb R^2, the test derivative ∂eφ=Dφ[e]\partial_e\varphi=D\varphi[e] and test vector field φe\varphi e are scalar and vector L2L^2 classes on UU, respectively. The ambient space ℋ(U)\mathcal H(U) of a planar set UU is the Hilbert product of scalar and vector L2L^2 spaces in (3), with area measure on UU: ℋ(U)=L2(U)×L2(U;ℝ2),‖(v,g)‖ℋ(U)2=‖v‖L2(U)2+‖g‖L2(U)2.\mathcal H(U)=L^2(U)\times L^2(U;\mathbb R^2), \qquad \lVert(v,g)\rVert_{\mathcal H(U)}^2 =\lVert v\rVert_{L^2(U)}^2+\lVert g\rVert_{L^2(U)}^2.(3)✦

A square-integrable vector field gg is the weak gradient of a square-integrable function vv on UU when the identity (4) holds for every test function φ∈Cc∞(U)\varphi\in C_c^\infty(U) and every e∈ℝ2e\in\mathbb R^2: ∫Uv∂eφdA=−∫U⟨g,e⟩φdA.\int_U v\,\partial_e\varphi\,dA =-\int_U\langle g,e\rangle\varphi\,dA.(4)✦ The Sobolev space H1(U)H^1(U) is the subspace of ℋ(U)\mathcal H(U) consisting of pairs whose vector component is the weak gradient of their scalar component. A function vv with weak gradient gg determines the Sobolev class ([v],[g])([v],[g]); the continuous linear Sobolev value map and gradient map extract [v][v] and [g][g], respectively.

For open UU, the Sobolev embedding of test functions sends φ\varphi to its class with its classical gradient.✦ For an open planar set UU, the Dirichlet space H01(U)H_0^1(U) is the closure in H1(U)H^1(U) of the Sobolev classes of test functions in Cc∞(U)C_c^\infty(U), with respect to the Sobolev norm. The Dirichlet value and gradient maps on H01(U)H_0^1(U) are the restrictions of the Sobolev value and gradient maps on H1(U)H^1(U), respectively. Under (2), the weighted Sobolev value map Jν:H01(D)→L2(ν)J_\nu:H_0^1(D)\to L^2(\nu) sends a function to its ν\nu value class.✦ The subcritical Dirichlet bound with constant d∈[0,1)d\in[0,1) is (5): ‖Jνu‖L2(ν)2≤d‖∇u‖L2(D;ℝ2)2for every u∈H01(D).\lVert J_\nu u\rVert_{L^2(\nu)}^2 \le d\,\lVert\nabla u\rVert_{L^2(D;\mathbb R^2)}^2 \qquad\text{for every }u\in H_0^1(D).(5)

The Kernels

The complex coordinate isometry z↦(Re⁡z,Im⁡z)z\mapsto(\operatorname{Re}z,\operatorname{Im}z) identifies ℂ\mathbb C with ℝ2\mathbb R^2.✦ Under this identification, the complex disk Green kernel, Poisson kernel and boundary interaction give the planar kernels GG, PP and RR in (6), where x,y∈ℝ2x,y\in\mathbb R^2 and s,t∈S1s,t\in S^1. All scalar reciprocals use 0−1=00^{-1}=0. G(x,y)=14πlog⁡(1+(1−‖x‖2)(1−‖y‖2)‖x−y‖2),P(s,x)=1−‖x‖22π‖s−x‖2,R(s,t)=1π‖s−t‖2.\begin{aligned} G(x,y)&=\frac1{4\pi}\log\!\left( 1+\frac{(1-\lVert x\rVert^2)(1-\lVert y\rVert^2)} {\lVert x-y\rVert^2}\right),\\ P(s,x)&=\frac{1-\lVert x\rVert^2}{2\pi\lVert s-x\rVert^2}, &R(s,t)&=\frac1{\pi\lVert s-t\rVert^2}. \end{aligned}(6)

The Green integral operator 𝒢\mathcal G is the continuous linear L2L^2 kernel operator on L2(D,dA)L^2(D,dA) in (7).✦ The weighted L2L^2 inclusion Iν:L2(D,dA)→L2(ν)I_\nu:L^2(D,dA)\to L^2(\nu) sends an area value class to its ν\nu class.✦ The weighted Green operator TT on L2(ν)L^2(\nu) is its composition with 𝒢\mathcal G and the Hilbert adjoint Iν*I_\nu^*, as specified in (7): (𝒢f)(x)=∫DG(x,y)f(y)dA(y),T=Iν𝒢Iν*.(\mathcal Gf)(x)=\int_D G(x,y)f(y)\,dA(y), \qquad T=I_\nu\mathcal G I_\nu^*.(7) The Neumann series resolvent 𝒮\mathcal S of TT is the operator series (8): 𝒮=∑n=0∞Tn.\mathcal S=\sum_{n=0}^{\infty}T^n.(8)✦

The smoothed Poisson kernel P1P_1 on S1×ℝ2S^1\times\mathbb R^2 is the function given by the integral in (9).✦ Every Green column G(⋅,p)G(\cdot,p), p∈Dp\in D, is square integrable with respect to ν\nu✦; every smoothed Poisson function P1(s,⋅)P_1(s,\cdot), s∈S1s\in S^1, is square integrable whenever ν\nu is finite.✦ For p∈Dp\in D and s∈S1s\in S^1, the Green row gpg_p and smoothed Poisson row hsh_s are the L2(ν)L^2(\nu) classes of G(⋅,p)G(\cdot,p) and P1(s,⋅)P_1(s,\cdot), respectively, under (2), as specified in (9). P1(s,x)=∫G(x,y)P(s,y)dν(y),gp=[G(⋅,p)]ν,hs=[P1(s,⋅)]ν.\begin{aligned} P_1(s,x)&=\int G(x,y)P(s,y)\,d\nu(y),\\ g_p&=[G(\cdot,p)]_\nu, &h_s&=[P_1(s,\cdot)]_\nu. \end{aligned}(9)

The full boundary evaluation kernel K:D×S1→ℝK:D\times S^1\to\mathbb R and full boundary kernel N:S1×S1→ℝN:S^1\times S^1\to\mathbb R are the functions in (10); the inner products are those of L2(ν)L^2(\nu): K(p,s)=P(s,p)+P1(s,p)+⟨gp,𝒮hs⟩,N(s,t)=R(s,t)+∫P(s,x)P(t,x)dν(x)+∫P(s,x)P1(t,x)dν(x)+⟨hs,𝒮ht⟩.\begin{aligned} K(p,s)&=P(s,p)+P_1(s,p)+\langle g_p,\mathcal S h_s\rangle,\\ N(s,t)&=R(s,t)+\int P(s,x)P(t,x)\,d\nu(x)\\ &\quad+\int P(s,x)P_1(t,x)\,d\nu(x) +\langle h_s,\mathcal S h_t\rangle. \end{aligned}(10)

Finite Matrices

For a real matrix AA indexed by a finite set JJ, its bilinear form qAq_A is (11): qA(x,y)=x𝖳Ay=∑i,j∈JxiAijyj,x,y∈ℝJ.q_A(x,y)=x^{\mathsf T}Ay =\sum_{i,j\in J}x_iA_{ij}y_j, \qquad x,y\in\mathbb R^J.(11)✦ Property (P)(P) for AA means at most one positive square, with the precise condition (12): 𝖯(A)⇔{A=A𝖳,qA(y,y)≤0 whenever qA(x,x)>0 and qA(x,y)=0,x,y∈ℝJ.\mathsf P(A)\quad\Longleftrightarrow\quad \left\{ \begin{array}{l} A=A^{\mathsf T},\\ q_A(y,y)\le0\text{ whenever }q_A(x,x)>0 \text{ and }q_A(x,y)=0,\quad x,y\in\mathbb R^J. \end{array} \right.(12)✦

For a real matrix BB indexed by a finite set II and a column ℓ∈ℝI\ell\in\mathbb R^I, the bordered matrix A(B,ℓ)A(B,\ell) is the matrix on I⊔{*}I\sqcup\{*\} given by (13): A(B,ℓ)=(Bℓℓ𝖳0).A(B,\ell)= \begin{pmatrix} B&\ell\\ \ell^{\mathsf T}&0 \end{pmatrix}.(13)✦ For p∈Dp\in D and a family (si)i∈I(s_i)_{i\in I} in S1S^1, define the reciprocal boundary block BB and evaluation column ℓ\ell by (14); reciprocation is entrywise: Bij={0,si=sj,N(si,sj)−1,si≠sj,ℓi=K(p,si)−1.B_{ij}=\begin{cases} 0,&s_i=s_j,\\ N(s_i,s_j)^{-1},&s_i\ne s_j, \end{cases} \qquad \ell_i=K(p,s_i)^{-1}.(14)

The Claim

Theorem 1. Let ν\nu satisfy the finite density bound (2) and the subcritical Dirichlet bound (5) for some 0≤d<10\le d<1. Let p∈Dp\in D, and let (si)i∈I(s_i)_{i\in I} be any finite family of points of S1S^1, with repetitions allowed. For the data in (14), 𝖯(A(B,ℓ)).\mathsf P\bigl(A(B,\ell)\bigr).(15)✦

In the paper, the result reads: under the hypotheses of Theorem 5.1, for distinct s1,…,sm∈S1s_1,\ldots,s_m\in S^1 and p∈Dp\in D, the bordered matrix (Bℓℓ𝖳0)\left(\begin{smallmatrix}B&\ell\\\ell^{\mathsf T}&0\end{smallmatrix}\right), with Bij=1/N(si,sj)B_{ij}=1/N(s_i,s_j) for i≠ji\ne j, Bii=0B_{ii}=0 and ℓi=1/K(p,si)\ell_i=1/K(p,s_i), has property (P)(P).

What Is Checked

The assertion (15) is stated over the weighted disk Green operator and its full kernels, under finite measure domination and the Dirichlet bound (5).

The area Green rows G(⋅,p)G(\cdot,p) are square integrable with respect to dAdA for almost every p∈Dp\in D.✦ There is a nonnegative constant that bounds the L2(D,dA)L^2(D,dA) norm of every square-integrable area Green row.✦

For an L2L^2 kernel operator associated with a measurable kernel HH and measures α,β\alpha,\beta, every f∈L2(β)f\in L^2(\beta) has a square-integrable kernel integral x↦∫H(x,y)f(y)dβ(y)x\mapsto\int H(x,y)f(y)\,d\beta(y) with respect to α\alpha.✦