OctoInformBeta

OpenAI preprint, September 2026 · §3. An augmented system and analytic estimates

Proposition 3.2: A negative-order forcing norm with arbitrary imaginary phase and nonzero real phase

A supported Carleman estimate for augmented fields.

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

In the augmented elasticity system, a scalar component dd accompanies the displacement u=(u1,u2,u3)u=(u_1,u_2,u_3) in the field U=(d,u1,u2,u3)U=(d,u_1,u_2,u_3). The analytic question concerns control of this field under a large complex exponential weight when its support is restricted to a fixed ball.

The spatial domain is the Euclidean space X=ℝ3X=\mathbb R^3✦, and the augmented target is Y=ℂ4Y=\mathbb C^4.✦ For k=1,2,3k=1,2,3, the coordinate vector eke_k has coordinate 11 at kk and 00 elsewhere.✦ Directional differentiation along w∈Xw\in X, coordinate differentiation, and the Laplacian are the operators in (1), acting componentwise on smooth YY-valued fields: ∂wf(x)=df(x)[w],∂k=∂ek,Δf=∑k=13∂k2f.\partial_w f(x)=df(x)[w],\qquad \partial_k=\partial_{e_k},\qquad \Delta f=\sum_{k=1}^3\partial_k^2 f.(1)

The Carleman test space 𝒮=𝒮(ℝ3)\mathcal S=\mathcal S(\mathbb R^3) consists of complex-valued smooth functions whose derivatives of every order decay faster than every inverse polynomial.✦ On 𝒮\mathcal S, the directional derivatives and Laplacian are complex linear maps. For R∈ℝR\in\mathbb R, a test function f∈𝒮f\in\mathcal S is supported in the closed ball BR¯={x∈X:∥x∥≤R}\overline{B_R}=\{x\in X:\|x\|\le R\} when f(x)=0f(x)=0 for every ∥x∥>R\|x\|>R.✦

The Semiclassical Norms

The continuous complex linear embedding ι2:𝒮→L2(X)\iota_2:\mathcal S\to L^2(X) sends a test function to its class for Lebesgue measure; write ∥f∥2=∥ι2f∥L2\|f\|_2=\|\iota_2 f\|_{L^2}. Let ℱ\mathcal F denote the Fourier transform. For real h,sh,s, the Bessel weight whs:X→ℝw_h^s:X\to\mathbb R and its multiplication operator Mhs:𝒮→𝒮M_h^s:\mathcal S\to\mathcal S are defined in (2). The Bessel potential Jhs:𝒮→𝒮J_h^s:\mathcal S\to\mathcal S is the Fourier multiplier in the same display: whs(ξ)=(1+∥2πhξ∥2)s/2,(Mhsg)(ξ)=whs(ξ)g(ξ),Jhsf=ℱ−1(Mhs(ℱf)).\begin{aligned} w_h^s(\xi)&=(1+\|2\pi h\xi\|^2)^{s/2}, & (M_h^s g)(\xi)&=w_h^s(\xi)g(\xi),\\ J_h^s f&=\mathcal F^{-1}\bigl(M_h^s(\mathcal F f)\bigr). \end{aligned}(2) For f∈𝒮f\in\mathcal S, ∥f∥2\|f\|_2 denotes the L2L^2 norm of its image under the embedding ι2\iota_2. The semiclassical Sobolev norm of a finite tuple W=(Wi)i=1nW=(W_i)_{i=1}^n of Schwartz functions is the quantity in (3): ∥W∥Hhs=(∑i=1n∥JhsWi∥22)1/2.\|W\|_{H_h^s} =\left(\sum_{i=1}^n\|J_h^s W_i\|_2^2\right)^{1/2}.(3)✦

The Conjugated Operators

A complex direction is a vector θ∈ℂ3\theta\in\mathbb C^3.✦ Its dot product is complex bilinear, without complex conjugation. The linear phase φθ\varphi_\theta, directional operator DθD_\theta, and exponential weight EzE_z are defined in (4), for z∈ℂz\in\mathbb C: φθ(x)=θ⋅x=∑k=13θkxk,Dθ=∑k=13θk∂k,Ez(x)=exp⁡(zθ⋅x).\varphi_\theta(x)=\theta\cdot x=\sum_{k=1}^3\theta_k x_k, \qquad D_\theta=\sum_{k=1}^3\theta_k\partial_k, \qquad E_z(x)=\exp\bigl(z\theta\cdot x\bigr).(4) A null direction θ∈ℂ3\theta\in\mathbb C^3 is a nonzero vector satisfying θ⋅θ=0\theta\cdot\theta=0. For the large real parameter τ>0\tau>0, set h=τ−1h=\tau^{-1}.

For complex matrix fields Ak,A0:X→End⁡ℂ(Y)A_k,A_0:X\to\operatorname{End}_{\mathbb C}(Y), the normalized augmented operator ℒ\mathcal L is the differential operator on smooth YY-valued fields defined by (5): ℒf=Δf+∑k=13Ak∂kf+A0f.\mathcal L f=\Delta f+\sum_{k=1}^3 A_k\partial_k f+A_0 f.(5)✦

A smooth scalar coefficient c:X→ℂc:X\to\mathbb C has temperate growth when every derivative of cc is bounded by some polynomial in ∥x∥\|x\|. For such a coefficient c:X→ℂc:X\to\mathbb C, coefficient multiplication Tc:𝒮→𝒮T_c:\mathcal S\to\mathcal S is the continuous complex linear map (Tcf)(x)=c(x)f(x)(T_c f)(x)=c(x)f(x).✦

The negative conjugated Laplacian Ph,α,βP_{h,\alpha,\beta} and scaled shifted derivatives Qh,α,β,kQ_{h,\alpha,\beta,k} on 𝒮\mathcal S are given by (6) for real hh and a phase θ=α+iβ\theta=\alpha+\mathrm i\beta, with α,β∈X\alpha,\beta\in X; here α⋅β\alpha\cdot\beta is the real Euclidean inner product: Ph,α,βf=−h2Δf−2h∂αf−2hi∂βf−(∥α∥2−∥β∥2+2iα⋅β)f,Qh,α,β,kf=h2∂kf+h(αk+iβk)f.\begin{aligned} P_{h,\alpha,\beta}f &=-h^2\Delta f-2h\partial_\alpha f-2h\mathrm i\partial_\beta f\\ &\quad-\bigl(\|\alpha\|^2-\|\beta\|^2 +2\mathrm i\alpha\cdot\beta\bigr)f,\\ Q_{h,\alpha,\beta,k}f &=h^2\partial_k f+h(\alpha_k+\mathrm i\beta_k)f. \end{aligned}(6)

For scalar coefficients b=(bk)k=13b=(b_k)_{k=1}^3 and cc of temperate growth, the conjugated lower-order operator Lh,α,β(b,c)L_{h,\alpha,\beta}(b,c) is defined in (7). The conjugated augmented system 𝒫h,α,β(a,v)\mathcal P_{h,\alpha,\beta}(a,v) on 𝒮4\mathcal S^4 is the differential operator defined in (7) for derivative coefficient matrices ak=(akij)a_k=(a_{kij}) and potential matrix v=(vij)v=(v_{ij}) whose scalar entries have temperate growth; the component indices i,ji,j run from 00 to 33: Lh,α,β(b,c)f=∑k=13TbkQh,α,β,kf+h2Tcf,(𝒫h,α,β(a,v)U)i=Ph,α,βUi+∑j=03Lh,α,β((akij)k=13,vij)Uj.\begin{aligned} L_{h,\alpha,\beta}(b,c)f &=\sum_{k=1}^3 T_{b_k}Q_{h,\alpha,\beta,k}f+h^2T_c f,\\ \bigl(\mathcal P_{h,\alpha,\beta}(a,v)U\bigr)_i &=P_{h,\alpha,\beta}U_i +\sum_{j=0}^3L_{h,\alpha,\beta} \bigl((a_{kij})_{k=1}^3,v_{ij}\bigr)U_j. \end{aligned}(7)

The Claim

Theorem 1. Let Ak,A0A_k,A_0 be complex matrix fields on XX. For every smooth f:X→Yf:X\to Y, complex vector θ\theta with θ⋅θ=0\theta\cdot\theta=0, and z∈ℂz\in\mathbb C, the normalized conjugation identity (8) holds: ℒ(Ezf)=Ez(ℒf+2zDθf+z∑k=13θkAkf).\mathcal L(E_z f) =E_z\left(\mathcal L f+2zD_\theta f +z\sum_{k=1}^3\theta_k A_k f\right).(8)✦

Let R≥1R\ge1 and α,β∈X\alpha,\beta\in X, with α≠0\alpha\ne0. Suppose all entries of aka_k and vv have temperate growth and satisfy the uniform bounds |akij(x)|≤M|a_{kij}(x)|\le M and |vij(x)|≤V|v_{ij}(x)|\le V for some M,V≥0M,V\ge0, all xx, and all indices. There exist constants C,h0>0C,h_0>0, independent of hh and UU, such that for 0<h≤h00<h\le h_0 and U∈𝒮4U\in\mathcal S^4 with each component supported in BR¯\overline{B_R}, the four-component Carleman estimate (9) holds: ∥U∥Hh1≤Ch∥𝒫h,α,β(a,v)U∥Hh−1.\|U\|_{H_h^1} \le\frac{C}{h} \bigl\|\mathcal P_{h,\alpha,\beta}(a,v)U\bigr\|_{H_h^{-1}}.(9)✦

In the paper, the result reads: for a bounded ball BB, a field U∈Cc∞(B;ℂ4)U\in C_c^\infty(B;\mathbb C^4), a fixed nonzero null direction θ\theta, and all sufficiently small h>0h>0, with τ=h−1\tau=h^{-1}, the normalized augmented operator satisfies (10), where CC may depend on the fixed coefficients, ball and direction, but not on hh: ∥Eτ−1U∥Hh1≤Ch∥Eτ−1ℒU∥Hh−1.\|E_\tau^{-1}U\|_{H_h^1} \le Ch\|E_\tau^{-1}\mathcal L U\|_{H_h^{-1}}.(10)

What Is Checked

The estimate (9) is stated over supported four-component Schwartz fields, bounded coefficient entries of temperate growth, and real phase vectors with α≠0\alpha\ne0; the identity (8) is stated over smooth YY-valued fields and arbitrary complex matrix fields, with θ⋅θ=0\theta\cdot\theta=0.