OctoInformBeta

OpenAI preprint, September 2026 · §1. Introduction

§1 (setting): the energy and the full Sobolev comparison class: Local and global minimality for coefficient-one Bernoulli energy on balls

The energy and its Sobolev competitors.

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

The variational setting for the one-phase Bernoulli problem is comparison on balls, with boundary values understood in the Sobolev sense. The library defines the energy and its comparison class using a function together with an explicit weak-gradient representative.

The Definition

Definition 1. For a nonnegative integer dd, the ambient space is the real Euclidean space ℝd\mathbb R^d in (1), with its usual inner product and Lebesgue measure. ℝd={(x1,…,xd):xi∈ℝ},x⋅y=∑i=1dxiyi.\mathbb R^d=\{(x_1,\ldots,x_d):x_i\in\mathbb R\}, \qquad x\cdot y=\sum_{i=1}^d x_i y_i.(1)✦ Let D⊆ℝdD\subseteq\mathbb R^d. A smooth test function φ\varphi on DD is a globally smooth function whose support is a compact subset of DD; write 𝒟(D)\mathcal D(D) for the class in (2). 𝒟(D)={φ∈C∞(ℝd):suppφ is compact,suppφ⊆D}.\mathcal D(D)= \left\{\varphi\in C^\infty(\mathbb R^d): \operatorname{supp}\varphi\text{ is compact},\quad \operatorname{supp}\varphi\subseteq D\right\}.(2)✦

For a function u:ℝd→ℝu:\mathbb R^d\to\mathbb R, a weak gradient on DD is a vector field G:ℝd→ℝdG:\mathbb R^d\to\mathbb R^d satisfying (3), where ∂hφ=∇φ⋅h\partial_h\varphi=\nabla\varphi\cdot h is the directional derivative. ∫Du∂hφdx=−∫D(G⋅h)φdx(φ∈𝒟(D),h∈ℝd).\int_D u\,\partial_h\varphi\,dx =-\int_D (G\cdot h)\varphi\,dx \qquad (\varphi\in\mathcal D(D),\ h\in\mathbb R^d).(3)✦ The Sobolev space H1(D)H^1(D) consists of pairs (u,G)(u,G) with the integrability and weak gradient conditions in (4). (u,G)∈H1(D)quad⇔u∈L2(D),G∈L2(D;ℝd),G is a weak gradient of u on D.(u,G)\in H^1(D)quad\Longleftrightarrow\quad u\in L^2(D),\quad G\in L^2(D;\mathbb R^d),\quad G\text{ is a weak gradient of }u\text{ on }D.(4)✦ The zero-trace space H01(D)H^1_0(D) consists of Sobolev pairs approximable by smooth test functions and their gradients as specified in (5). (u,G)∈H01(D)⇔(u,G)∈H1(D) and, for every ε>0,there is φ∈𝒟(D) such that∥u−φ∥L2(D)<ε,∥G−∇φ∥L2(D;ℝd)<ε.\begin{aligned} (u,G)\in H^1_0(D)\quad\Longleftrightarrow\quad &(u,G)\in H^1(D)\text{ and, for every }\varepsilon>0,\\ &\text{there is }\varphi\in\mathcal D(D)\text{ such that}\\ &\|u-\varphi\|_{L^2(D)}<\varepsilon,\qquad \|G-\nabla\varphi\|_{L^2(D;\mathbb R^d)}<\varepsilon. \end{aligned}(5)✦ Write B(c,r)B(c,r) for the open ball with center cc and radius rr. The local Sobolev space Hloc1(D)H^1_{\mathrm{loc}}(D) consists of the pairs in (6). (u,G)∈Hloc1(D)⇔(u,G)∈H1(B(c,r))for every c∈ℝd,r>0 with B(c,r)¯⊆D.(u,G)\in H^1_{\mathrm{loc}}(D)\quad\Longleftrightarrow\quad (u,G)\in H^1(B(c,r)) \quad\text{for every }c\in\mathbb R^d,\ r>0 \text{ with }\overline{B(c,r)}\subseteq D.(6)✦

For a Sobolev pair (v,V)(v,V), write ∇v=V\nabla v=V. The coefficient-one Bernoulli energy J(v;B)J(v;B) of a Sobolev pair (v,V)(v,V) on a set BB is the integral in (7), where 𝟏{v>0}\mathbf 1_{\{v>0\}} is the indicator of the positive phase. J(v;B)=∫B(|∇v|2+𝟏{v>0})dx.J(v;B)=\int_B\bigl(|\nabla v|^2+\mathbf 1_{\{v>0\}}\bigr)\,dx.(7)✦

A local minimizer in DD is a function u:ℝd→ℝu:\mathbb R^d\to\mathbb R belonging to the class ℳ(D)\mathcal M(D) defined by (8). u∈ℳ(D)⇔D is open and there is G:ℝd→ℝd such that(u,G)∈Hloc1(D),u≥0 a.e. on D,J(u;B(c,r))≤J(v;B(c,r))for every c∈ℝd,r>0 with B(c,r)¯⊆D,and every (v,V)∈H1(B(c,r)) withv≥0 a.e. on B(c,r),(v−u,V−G)∈H01(B(c,r)).\begin{aligned} u\in\mathcal M(D)\quad\Longleftrightarrow\quad &D\text{ is open and there is }G:\mathbb R^d\to\mathbb R^d \text{ such that}\\ &(u,G)\in H^1_{\mathrm{loc}}(D),\qquad u\geq0\text{ a.e. on }D,\\ &J(u;B(c,r))\leq J(v;B(c,r))\\ &\quad\text{for every }c\in\mathbb R^d,\ r>0 \text{ with }\overline{B(c,r)}\subseteq D,\\ &\quad\text{and every }(v,V)\in H^1(B(c,r))\text{ with}\\ &\qquad v\geq0\text{ a.e. on }B(c,r),\qquad (v-u,V-G)\in H^1_0(B(c,r)). \end{aligned}(8)✦ A global minimizer u:ℝd→ℝu:\mathbb R^d\to\mathbb R is a function in the class ℳglob\mathcal M_{\mathrm{glob}} of (9). ℳglob=ℳ(ℝd).\mathcal M_{\mathrm{glob}}=\mathcal M(\mathbb R^d).(9)✦

What Is Checked

The comparison classes in (8) and (9) are stated over scalar functions and explicit weak-gradient fields on ℝd\mathbb R^d, with the Sobolev and zero-trace conditions of (4) and (5).