OctoInformBeta

OpenAI preprint, September 2026 · A variational principle for tangent vector fields

Section 3: algebraic replacement for Lemma 3.2: The algebraic boundary defect satisfies a Lipschitz multiplier identity on bounded open sets

Vector energy splits into Cartesian and boundary terms.

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

Let Ω⊂ℝ2\Omega\subset\mathbb R^2, and let e1,e2e_1,e_2 be the standard orthonormal basis of the Euclidean plane. All L2L^2 spaces below use area measure restricted to Ω\Omega. The ambient space of scalar-vector pairs is the product Hilbert space ℋΩ\mathcal H_\Omega with the Euclidean product norm in (1): ℋΩ=L2(Ω)×L2(Ω;ℝ2),∥(v,g)∥ℋΩ2=∥v∥L22+∥g∥L22.\mathcal H_\Omega=L^2(\Omega)\times L^2(\Omega;\mathbb R^2), \qquad \|(v,g)\|_{\mathcal H_\Omega}^2 =\|v\|_{L^2}^2+\|g\|_{L^2}^2.(1)✦

A test function φ\varphi on Ω\Omega is a smooth real function on ℝ2\mathbb R^2 with compact closed support contained in Ω\Omega.✦ For a test function φ\varphi on Ω\Omega, write Dφ(x)D\varphi(x) for its differential at xx. For a direction e∈ℝ2e\in\mathbb R^2, the test derivative of φ\varphi in direction ee is ∂eφ(x)=Dφ(x)e\partial_e\varphi(x)=D\varphi(x)e, and the associated test vector field is φe\varphi e; these belong to L2(Ω)L^2(\Omega) and L2(Ω;ℝ2)L^2(\Omega;\mathbb R^2), respectively. A weak gradient gg of a real function vv is a square-integrable vector field such that vv is square-integrable and (2) holds for every such φ\varphi and every ee: ∫Ωv∂eφ=−∫Ω(g⋅e)φ.\int_\Omega v\,\partial_e\varphi =-\int_\Omega (g\cdot e)\varphi.(2)✦

The Sobolev space H1(Ω)H^1(\Omega) is the weak gradient graph, the linear subspace of ℋΩ\mathcal H_\Omega consisting of pairs satisfying (2).✦ The value map val⁡:H1(Ω)→L2(Ω)\operatorname{val}:H^1(\Omega)\to L^2(\Omega) and gradient map ∇:H1(Ω)→L2(Ω;ℝ2)\nabla:H^1(\Omega)\to L^2(\Omega;\mathbb R^2) of the Sobolev space H1(Ω)H^1(\Omega) are the continuous linear maps val⁡(v,g)=v\operatorname{val}(v,g)=v and ∇(v,g)=g\nabla(v,g)=g. A function vv with weak gradient gg determines the Sobolev class formed from their L2L^2 equivalence classes.✦

The Energy Forms

The space L2(Ω;ℝ2)L^2(\Omega;\mathbb R^2) consists of square-integrable vector fields.✦ For i=1,2i=1,2, the coordinate projection pip_i to scalar L2L^2 and the coordinate embedding aia_i from scalar L2L^2 are continuous linear maps. The actions of the coordinate maps, the weak coordinate derivatives, and the Sobolev vector value 𝑿\mathbf X are given in (3) for Z∈L2(Ω;ℝ2)Z\in L^2(\Omega;\mathbb R^2), v∈L2(Ω)v\in L^2(\Omega), u∈H1(Ω)u\in H^1(\Omega), and a field X=(X1,X2)∈H1(Ω)2X=(X_1,X_2)\in H^1(\Omega)^2: piZ=Z⋅ei,aiv=vei,∂iu=pi∇u,𝑿=a1val⁡X1+a2val⁡X2.\begin{aligned} p_iZ&=Z\cdot e_i,& a_iv&=ve_i,& \partial_i u&=p_i\nabla u,\\ \mathbf X&=a_1\operatorname{val}X_1+a_2\operatorname{val}X_2. \end{aligned}(3) For a field X∈H1(Ω)2X\in H^1(\Omega)^2, its Sobolev divergence div⁡X\operatorname{div}X and curl curl⁡X\operatorname{curl}X are the scalar functions in L2(Ω)L^2(\Omega) defined by (4): div⁡X=∂1X1+∂2X2,curl⁡X=∂1X2−∂2X1.\operatorname{div}X=\partial_1X_1+\partial_2X_2, \qquad \operatorname{curl}X=\partial_1X_2-\partial_2X_1.(4)

Write (⋅,⋅)0(\cdot,\cdot)_0 for the appropriate real L2L^2 inner product. The Cartesian Dirichlet energy CC and divergence-curl energy QQ on H1(Ω)2H^1(\Omega)^2 are the bilinear forms with arguments X,Y∈H1(Ω)2X,Y\in H^1(\Omega)^2 given in (5); write Q(X)=Q(X,X)Q(X)=Q(X,X) for the corresponding quadratic energy: C(X,Y)=∑i=12(∇Xi,∇Yi)0,Q(X,Y)=(div⁡X,div⁡Y)0+(curl⁡X,curl⁡Y)0.\begin{aligned} C(X,Y)&=\sum_{i=1}^2(\nabla X_i,\nabla Y_i)_0,\\ Q(X,Y)&=(\operatorname{div}X,\operatorname{div}Y)_0 +(\operatorname{curl}X,\operatorname{curl}Y)_0. \end{aligned}(5) For a real parameter λ\lambda, the shifted Cartesian form CλC_\lambda and vector form QλQ_\lambda are the bilinear forms on H1(Ω)2H^1(\Omega)^2 defined by (6): Cλ(X,Y)=C(X,Y)−λ∑i=12(val⁡Xi,val⁡Yi)0,Qλ(X,Y)=Q(X,Y)−λ(𝑿,𝒀)0.\begin{aligned} C_\lambda(X,Y)&=C(X,Y) -\lambda\sum_{i=1}^2(\operatorname{val}X_i,\operatorname{val}Y_i)_0,\\ Q_\lambda(X,Y)&=Q(X,Y)-\lambda(\mathbf X,\mathbf Y)_0. \end{aligned}(6)

The quarter-turn rotation J:ℝ2→ℝ2J:\mathbb R^2\to\mathbb R^2 is the continuous linear map J(x1,x2)=(−x2,x1)J(x_1,x_2)=(-x_2,x_1); its action on L2L^2 vector fields is pointwise rotation. The skew gradient energy SS on H1(Ω)H^1(\Omega) and the algebraic boundary defect BB on H1(Ω)2H^1(\Omega)^2 are the bilinear forms defined by (7): S(u,v)=(∇u,J∇v)0,B(X,Y)=−S(X1,Y2)+S(X2,Y1).\begin{aligned} S(u,v)&=(\nabla u,J\nabla v)_0,\\ B(X,Y)&=-S(X_1,Y_2)+S(X_2,Y_1). \end{aligned}(7)

On a bounded open set Ω\Omega, Lipschitz multiplication by a real Lipschitz function ff on ℝ2\mathbb R^2 is the map Mf:H1(Ω)→H1(Ω)M_f:H^1(\Omega)\to H^1(\Omega) whose value is fval⁡uf\operatorname{val}u for u∈H1(Ω)u\in H^1(\Omega).✦ On fields, MfM_f acts on each component; write fX=MfXfX=M_fX and f2X=Mf(MfX)f^2X=M_f(M_fX).

The Claim

Theorem 1. Let Ω⊂ℝ2\Omega\subset\mathbb R^2 be a set, let λ\lambda be a real parameter, and let X,Y∈H1(Ω)2X,Y\in H^1(\Omega)^2 be vector fields. Then the energy decomposition is Qλ(X,Y)=Cλ(X,Y)+B(X,Y).Q_\lambda(X,Y)=C_\lambda(X,Y)+B(X,Y).(8)✦ If Ω\Omega is bounded and open, f:ℝ2→ℝf:\mathbb R^2\to\mathbb R is Lipschitz, and X∈H1(Ω)2X\in H^1(\Omega)^2, then B(fX,fX)=B(X,f2X).B(fX,fX)=B(X,f^2X).(9)✦

In the paper, the result reads as (10) for tangent fields XX, with unit tangent τ\tau and curvature κ=det⁡(τ,∂ττ)\kappa=\det(\tau,\partial_\tau\tau), positive on a convex circle: Q(X)=∫Ω|∇X|2+∫∂Ωκ|X|2.Q(X)=\int_\Omega|\nabla X|^2 +\int_{\partial\Omega}\kappa|X|^2.(10)

What Is Checked

The statements concern the weak gradient graph H1(Ω)2H^1(\Omega)^2 and its L2L^2 energy forms, with an arbitrary real shift in (8) and Lipschitz multiplication on bounded open sets in (9). Lipschitz multiplication on a bounded open set gives a Sobolev class whose weak gradient satisfies the product rule.