OctoInformBeta

OpenAI preprint, September 2026 · Section 3. A localized virial identity

Lemma 3.1: Absolute interaction flux integrability for positive solutions with p,q>0p,q>0 in dimension n≥2n\ge2

Source energies satisfy a localized virial identity.

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 localized virial identity compares nonlinear source energies with a two-point flux of a radial vector field. The cutoff enters both through its value and through its radial derivative.

Let n≥2n\ge2 be an integer, and let ℝn\mathbb R^n be real Euclidean coordinate space with Lebesgue measure dxdx.✦ For real parameters p,q,A,Bp,q,A,B, a positive solution (u,v)(u,v) is a pair of continuous, strictly positive functions on ℝn\mathbb R^n, of class C2C^2 off the origin, satisfying −Δu(x)=|x|Av(x)p,−Δv(x)=|x|Bu(x)q(x≠0).-\Delta u(x)=|x|^A v(x)^p, \qquad -\Delta v(x)=|x|^B u(x)^q \quad (x\ne0).(1)✦ The weighted power sources associated with the solution of (1) are the functions f,g:ℝn→ℝf,g:\mathbb R^n\to\mathbb R defined by f(x)={|x|Bu(x)q,x≠0,0,x=0,g(x)={|x|Av(x)p,x≠0,0,x=0.f(x)= \begin{cases}|x|^B u(x)^q,&x\ne0,\\0,&x=0,\end{cases} \qquad g(x)= \begin{cases}|x|^A v(x)^p,&x\ne0,\\0,&x=0.\end{cases}(2)✦

Cutoffs and Interaction

For r∈ℝr\in\mathbb R and an integer exponent j≥0j\ge0, the positive-part power r+jr_+^j is the real number defined by r+j=(max⁡{r,0})jr_+^j=(\max\{r,0\})^j.✦ The auxiliary cutoff parameters m,k,Nm,k,N are the integers m=n+3,k=n(m+1),N=k+n+2.m=n+3,\qquad k=n(m+1),\qquad N=k+n+2.(3) For the exponents in (3), the radial cutoff profile η0\eta_0 and the compact radial cutoff η:ℝn→ℝ\eta:\mathbb R^n\to\mathbb R are the functions η0(r)=(1−(r−1)+3)+N,η(x)=η0(|x|).\eta_0(r)=\bigl(1-(r-1)_+^3\bigr)_+^N, \qquad \eta(x)=\eta_0(|x|).(4) Writing Dη(x)[x]D\eta(x)[x] for the derivative of η\eta at xx applied to the vector xx, the virial vector field X:ℝn→ℝnX:\mathbb R^n\to\mathbb R^n and the radial derivative weight w:ℝn→ℝw:\mathbb R^n\to\mathbb R associated with (4) are X(x)=η(x)x,w(x)=−Dη(x)[x].X(x)=\eta(x)x, \qquad w(x)=-D\eta(x)[x].(5)

Let σn−1\sigma_{n-1} be the finite area of the unit sphere for the unnormalized surface measure induced by Lebesgue measure.✦ The Newton normalization constant dnd_n and the Newton kernel KnK_n on ℝn×ℝn\mathbb R^n\times\mathbb R^n are defined by dn=(n−2)σn−1,Kn(x,y)={dn−1|x−y|2−n,n≥3 and x≠y,0,otherwise.d_n=(n-2)\sigma_{n-1}, \qquad K_n(x,y)= \begin{cases} d_n^{-1}|x-y|^{2-n},&n\ge3\text{ and }x\ne y,\\ 0,&\text{otherwise}. \end{cases}(6) For real functions F,GF,G on ℝn\mathbb R^n, their Newton interaction measure πF,G\pi_{F,G} is the measure with nonnegative density dπF,G(x,y)=F(x)+G(y)+Kn(x,y)+dxdy,t+=max⁡{t,0}.d\pi_{F,G}(x,y) =F(x)_+G(y)_+K_n(x,y)_+\,dx\,dy, \qquad t_+=\max\{t,0\}.(7)✦ The density and its integrals are understood as nonnegative extended quantities. For the sources in (2), write π=πf,g\pi=\pi_{f,g}.

The localized flux factor Φ:ℝn×ℝn→ℝ\Phi:\mathbb R^n\times\mathbb R^n\to\mathbb R associated with the vector field in (5) is Φ(x,y)={⟨x−y,X(x)−X(y)⟩|x−y|2,x≠y,0,x=y.\Phi(x,y)= \begin{cases} \displaystyle \frac{\langle x-y,X(x)-X(y)\rangle}{|x-y|^2},&x\ne y,\\ 0,&x=y. \end{cases}(8)✦

For a real weight χ\chi and a real function UU on ℝn\mathbb R^n, the weighted nonlinear energy ℰχ[U;s,C]\mathcal E_\chi[U;s,C] with power ss and weight exponent CC is the real integral ℰχ[U;s,C]=∫ℝnχ(x)hU,s,C(x)U(x)dx,hU,s,C(x)={|x|CU(x)s,x≠0,0,x=0.\mathcal E_\chi[U;s,C] =\int_{\mathbb R^n}\chi(x)h_{U,s,C}(x)U(x)\,dx, \qquad h_{U,s,C}(x)= \begin{cases}|x|^C U(x)^s,&x\ne0,\\0,&x=0.\end{cases}(9) For the two sources and the weights η,w\eta,w, use the notation Ef=ℰη[u;q,B],Eg=ℰη[v;p,A],Wf=ℰw[u;q,B],Wg=ℰw[v;p,A],a=1q+1,b=1p+1.\begin{aligned} E_f&=\mathcal E_\eta[u;q,B],& E_g&=\mathcal E_\eta[v;p,A],\\ W_f&=\mathcal E_w[u;q,B],& W_g&=\mathcal E_w[v;p,A], \end{aligned} \qquad a=\frac1{q+1},\quad b=\frac1{p+1}.(10)✦

The Claim

Theorem 1 (Localized virial identity). Let n≥2n\ge2, let p,q>0p,q>0 and A,B∈ℝA,B\in\mathbb R, and let (u,v)(u,v) be a positive solution of (1). For the interaction measure (7) and flux factor (8), the flux integral is absolutely convergent: ∫ℝn×ℝn|Φ(x,y)|dπ(x,y)<∞.\int_{\mathbb R^n\times\mathbb R^n}|\Phi(x,y)|\,d\pi(x,y)<\infty.(11)✦ With the energies and coefficients of (10), a(n+B)Ef+b(n+A)Eg−aWf−bWg=(n−2)∫ℝn×ℝnΦ(x,y)dπ(x,y).a(n+B)E_f+b(n+A)E_g-aW_f-bW_g =(n-2)\int_{\mathbb R^n\times\mathbb R^n}\Phi(x,y)\,d\pi(x,y).(12)✦

In the paper, the result reads: for n≥3n\ge3, p,q>0p,q>0, A,B>−2A,B>-2 and pq>1pq>1, a solution with these sources and cutoffs, with uu and vv equal to the Newton potentials of gg and ff, respectively, has an absolutely convergent flux integral and satisfies a(n+B)Ef+b(n+A)Eg−aWf−bWg=(n−2)∫Φdπa(n+B)E_f+b(n+A)E_g-aW_f-bW_g=(n-2)\int\Phi\,d\pi.

What Is Checked

The claims (11) and (12) are stated over positive solutions on Euclidean space, the specified radial cutoff and its derivative weight, and the Newton interaction measure of the two weighted power sources.