OctoInformBeta

OpenAI preprint, September 2026 · §3. A regularized matrix field · §3

Lemma 3.4 [Directional comparison]: Comparison inequalities for the Hermitian field and scalar primitive alongside isolated-row evaluations and the endpoint jump of the matrix field primitive JJ

State Jensen bounds and diagonal matrix field evaluations.

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 regularized matrix-field approach to one-dimensional Lieb-Thirring inequalities for Schrödinger operators, real parameters σ<1\sigma < 1 (in applications σ=γ−1/2∈(0,1)\sigma = \gamma - 1/2 \in (0, 1) for the spectral exponent γ∈(1/2,3/2)\gamma \in (1/2, 3/2)) and δ>0\delta > 0 determine a regularized matrix field on N×NN \times N real symmetric matrices and its matrix field primitive, together with a Hermitian matrix field on n×nn \times n complex matrices built from a regularized scalar primitive. The questions are how positive unital ℝ\mathbb{R}-linear functionals compare the Hermitian field and the regularized scalar primitive against scalar expectations, and how the real field and the matrix field primitive evaluate on diagonal configurations including isolated coordinates and opposite endpoints.

Definitions

For real parameters σ\sigma and δ\delta, the regularized scalar primitive fσ,δ:ℝ→ℝf_{\sigma,\delta} : \mathbb{R} \to \mathbb{R} is the function given for y∈ℝy \in \mathbb{R} by fσ,δ(y)=∫0y(t++δ)σ−3/2dt,f_{\sigma,\delta}(y) = \int_0^y (t_+ + \delta)^{\sigma - 3/2}\,dt,(1)✦ where t+=max⁡(t,0)t_+ = \max(t, 0). Associated with a real parameter σ\sigma, the field normalization constant cσ∈ℝc_\sigma \in \mathbb{R} is defined by cσ=σ∫ℝ(1+s2)σ−3/2ds.c_\sigma = \frac{\sigma}{\int_{\mathbb{R}} (1 + s^2)^{\sigma - 3/2}\,ds}.(2)✦ From fσ,δf_{\sigma,\delta} in (1) and cσc_\sigma in (2), real parameters σ\sigma and δ\delta determine the scalar function μσ,δ:ℝ→ℝ\mu_{\sigma,\delta} : \mathbb{R} \to \mathbb{R}, given for z∈ℝz \in \mathbb{R} by μσ,δ(z)=cσ∫ℝ(fσ,δ(s2+z)−fσ,δ(s2))ds.\mu_{\sigma,\delta}(z) = c_\sigma \int_{\mathbb{R}} \bigl(f_{\sigma,\delta}(s^2 + z) - f_{\sigma,\delta}(s^2)\bigr)\,ds.(3)✦

For a finite index set nn, let Mn(ℂ)M_n(\mathbb{C}) denote the space of n×nn \times n complex matrices with identity II. A state on Mn(ℂ)M_n(\mathbb{C}) is an ℝ\mathbb{R}-linear functional φ:Mn(ℂ)→ℝ\varphi : M_n(\mathbb{C}) \to \mathbb{R} satisfying φ(I)=1\varphi(I) = 1 and φ(A)≥0\varphi(A) \ge 0 for every positive semidefinite matrix A∈Mn(ℂ)A \in M_n(\mathbb{C}).

Given C,B∈Mn(ℂ)C, B \in M_n(\mathbb{C}) and s∈ℝs \in \mathbb{R}, the complex matrix pencil XC,B(s)∈Mn(ℂ)X_{C,B}(s) \in M_n(\mathbb{C}) is the matrix s2I+C−2sBs^2 I + C - 2sB.✦ For a square matrix AA and real parameters σ\sigma and δ\delta, the functional calculus evaluation fσ,δ(A)f_{\sigma,\delta}(A) is the matrix given by the continuous functional calculus of fσ,δf_{\sigma,\delta} at AA. For σ,δ∈ℝ\sigma, \delta \in \mathbb{R}, C,B∈Mn(ℂ)C, B \in M_n(\mathbb{C}), and s∈ℝs \in \mathbb{R}, the paired field integrand Pσ,δ(C,B;s)∈Mn(ℂ)P_{\sigma,\delta}(C, B; s) \in M_n(\mathbb{C}) is defined by Pσ,δ(C,B;s)=fσ,δ(XC,B(s))+fσ,δ(XC,B(−s))−2fσ,δ(s2)I.P_{\sigma,\delta}(C, B; s) = f_{\sigma,\delta}\bigl(X_{C,B}(s)\bigr) + f_{\sigma,\delta}\bigl(X_{C,B}(-s)\bigr) - 2 f_{\sigma,\delta}(s^2) I.(4)✦ The Hermitian matrix field ℋσ,δ(C,B)∈Mn(ℂ)\mathcal{H}_{\sigma,\delta}(C, B) \in M_n(\mathbb{C}) for σ,δ∈ℝ\sigma, \delta \in \mathbb{R} and C,B∈Mn(ℂ)C, B \in M_n(\mathbb{C}) is the integral of Pσ,δP_{\sigma,\delta} from (4), defined by ℋσ,δ(C,B)=cσ2∫ℝPσ,δ(C,B;s)ds.\mathcal{H}_{\sigma,\delta}(C, B) = \frac{c_\sigma}{2} \int_{\mathbb{R}} P_{\sigma,\delta}(C, B; s)\,ds.(5)✦

For a natural number NN, let MN(ℝ)M_N(\mathbb{R}) denote the space of N×NN \times N real matrices with identity II, and for any vector v∈ℝNv \in \mathbb{R}^N let diag⁡(v)∈MN(ℝ)\operatorname{diag}(v) \in M_N(\mathbb{R}) denote the diagonal matrix with ii-th diagonal entry viv_i. Within MN(ℝ)M_N(\mathbb{R}), the symmetric matrix space Sym⁡N\operatorname{Sym} N is the subspace of N×NN \times N real symmetric matrices.✦ For k∈ℝNk \in \mathbb{R}^N, let Dk∈MN(ℝ)D_k \in M_N(\mathbb{R}) be the diagonal matrix with ii-th diagonal entry ki2k_i^2; for B∈MN(ℝ)B \in M_N(\mathbb{R}) and s∈ℝs \in \mathbb{R}, the real matrix pencil Xk,B(s)X_{k,B}(s) is the matrix Dk−2sB+s2ID_k - 2sB + s^2 I in MN(ℝ)M_N(\mathbb{R}). With N∈ℕN \in \mathbb{N}, σ,δ∈ℝ\sigma, \delta \in \mathbb{R}, k∈ℝNk \in \mathbb{R}^N, B∈MN(ℝ)B \in M_N(\mathbb{R}), and a cutoff R∈ℝR \in \mathbb{R}, the truncated matrix field Mσ,δ,k(R)(B)∈MN(ℝ)M_{\sigma,\delta,k}^{(R)}(B) \in M_N(\mathbb{R}) is defined using fσ,δf_{\sigma,\delta} from (1) and cσc_\sigma from (2) by Mσ,δ,k(R)(B)=cσ∫−RR(fσ,δ(Xk,B(s))−fσ,δ(s2)I)ds.M_{\sigma,\delta,k}^{(R)}(B) = c_\sigma \int_{-R}^R \Bigl(f_{\sigma,\delta}\bigl(X_{k,B}(s)\bigr) - f_{\sigma,\delta}(s^2) I\Bigr)\,ds.(6)✦ The regularized matrix field Mσ,δ,k(B)∈MN(ℝ)M_{\sigma,\delta,k}(B) \in M_N(\mathbb{R}) for N∈ℕN \in \mathbb{N}, σ,δ∈ℝ\sigma, \delta \in \mathbb{R}, k∈ℝNk \in \mathbb{R}^N, and B∈MN(ℝ)B \in M_N(\mathbb{R}) is the infinite-cutoff limit of (6), given by Mσ,δ,k(B)=limR→+∞Mσ,δ,k(R)(B).M_{\sigma,\delta,k}(B) = \lim_{R \to +\infty} M_{\sigma,\delta,k}^{(R)}(B).(7)✦

When N∈ℕN \in \mathbb{N}, σ<1\sigma < 1, δ>0\delta > 0, and k∈ℝNk \in \mathbb{R}^N, the symmetric matrix field Mσ,δ,ksym:Sym⁡N→Sym⁡NM_{\sigma,\delta,k}^{\mathrm{sym}} : \operatorname{Sym} N \to \operatorname{Sym} N is the map sending each B∈Sym⁡NB \in \operatorname{Sym} N to Mσ,δ,k(B)∈Sym⁡NM_{\sigma,\delta,k}(B) \in \operatorname{Sym} N from (7).✦ For N∈ℕN \in \mathbb{N}, real parameters σ\sigma and δ\delta, k∈ℝNk \in \mathbb{R}^N, and B∈MN(ℝ)B \in M_N(\mathbb{R}), the matrix field primitive Jσ,δ,k(B)∈ℝJ_{\sigma,\delta,k}(B) \in \mathbb{R} is defined from Mσ,δ,kM_{\sigma,\delta,k} in (7) by Jσ,δ,k(B)=−∫01tr⁡(BMσ,δ,k(tB))dt.J_{\sigma,\delta,k}(B) = -\int_0^1 \operatorname{tr}\bigl(B M_{\sigma,\delta,k}(tB)\bigr)\,dt.(8)✦

The Claim

Theorem 1. Let σ,δ∈ℝ\sigma, \delta \in \mathbb{R} satisfy σ<1\sigma < 1 and δ>0\delta > 0; suppress σ\sigma and δ\delta from the subscripts of ff, μ\mu, and ℋ\mathcal{H}, and for any N∈ℕN \in \mathbb{N} and k∈ℝNk \in \mathbb{R}^N write M=Mσ,δ,kM = M_{\sigma,\delta,k}, Msym=Mσ,δ,ksymM_{\mathrm{sym}} = M_{\sigma,\delta,k}^{\mathrm{sym}}, and J=Jσ,δ,kJ = J_{\sigma,\delta,k}.

  1. If σ>0\sigma > 0, nn is a finite index set, φ\varphi is a continuous state on Mn(ℂ)M_n(\mathbb{C}), and C,B∈Mn(ℂ)C, B \in M_n(\mathbb{C}) are Hermitian, then φ(ℋ(C,B))≤μ(φ(C)−φ(B)2).\varphi\bigl(\mathcal{H}(C, B)\bigr) \le \mu\bigl(\varphi(C) - \varphi(B)^2\bigr).(9)✦

  2. If nn is a finite index set, φ\varphi is a state on Mn(ℂ)M_n(\mathbb{C}), and A∈Mn(ℂ)A \in M_n(\mathbb{C}) is Hermitian, then φ(f(A))≤f(φ(A)).\varphi\bigl(f(A)\bigr) \le f\bigl(\varphi(A)\bigr).(10)✦

  3. For any N∈ℕN \in \mathbb{N} and k,b∈ℝNk, b \in \mathbb{R}^N, M(diag⁡(b))=diag⁡((μ(ki2−bi2))1≤i≤N).M\bigl(\operatorname{diag}(b)\bigr) = \operatorname{diag}\Bigl(\bigl(\mu(k_i^2 - b_i^2)\bigr)_{1 \le i \le N}\Bigr).(11)✦

  4. If N∈ℕN \in \mathbb{N}, k∈ℝNk \in \mathbb{R}^N, B∈Sym⁡NB \in \operatorname{Sym} N, and i∈{1,…,N}i \in \{1, \dots, N\} is a coordinate with Bij=0B_{ij} = 0 for every j≠ij \neq i, then (Msym(B))ii=μ(ki2−Bii2).\bigl(M_{\mathrm{sym}}(B)\bigr)_{ii} = \mu(k_i^2 - B_{ii}^2).(12)✦

  5. For any N∈ℕN \in \mathbb{N} and k,b∈ℝNk, b \in \mathbb{R}^N, J(diag⁡(b))=−∑i=1N∫0biμ(ki2−s2)ds.J\bigl(\operatorname{diag}(b)\bigr) = -\sum_{i=1}^N \int_0^{b_i} \mu(k_i^2 - s^2)\,ds.(13)✦

  6. For any N∈ℕN \in \mathbb{N} and k∈ℝNk \in \mathbb{R}^N, J(−diag⁡(k))−J(diag⁡(k))=∑i=1N∫−kikiμ(ki2−s2)ds.J\bigl(-\operatorname{diag}(k)\bigr) - J\bigl(\operatorname{diag}(k)\bigr) = \sum_{i=1}^N \int_{-k_i}^{k_i} \mu(k_i^2 - s^2)\,ds.(14)✦

In the paper, the result reads: writing μ=μσ,δ\mu = \mu_{\sigma,\delta}, M=Mσ,δ,kM = M_{\sigma,\delta,k}, J=Jσ,δ,kJ = J_{\sigma,\delta,k}, and K=diag⁡(k)K = \operatorname{diag}(k), for every B∈Sym⁡NB \in \operatorname{Sym} N and every real unit vector ee, eTM(B)e≤μ(eTK2e−(eTBe)2)e^T M(B) e \le \mu(e^T K^2 e - (e^T B e)^2); if a standard basis vector eie_i is an eigenvector of BB with eigenvalue bib_i, then M(B)ei=μ(ki2−bi2)eiM(B) e_i = \mu(k_i^2 - b_i^2) e_i; and for the primitive, J(−K)−J(K)=∑i=1N∫−kikiμ(ki2−s2)dsJ(-K) - J(K) = \sum_{i=1}^N \int_{-k_i}^{k_i} \mu(k_i^2 - s^2)\,ds.

What Is Checked

The claims are stated over real parameters σ<1\sigma < 1 and δ>0\delta > 0 (with σ>0\sigma > 0 and continuity of φ\varphi required in addition for (9)), states φ\varphi and Hermitian matrices in Mn(ℂ)M_n(\mathbb{C}) for a finite index set nn, the regularized scalar primitive fσ,δf_{\sigma,\delta} in (1), the field normalization constant cσc_\sigma in (2), the scalar function μσ,δ\mu_{\sigma,\delta} in (3), the Hermitian matrix field ℋσ,δ\mathcal{H}_{\sigma,\delta} in (5) formed from the complex pencil XC,BX_{C,B} and paired integrand Pσ,δP_{\sigma,\delta} in (4), the symmetric matrix space Sym⁡N\operatorname{Sym} N and vectors k,b∈ℝNk, b \in \mathbb{R}^N for N∈ℕN \in \mathbb{N}, the regularized matrix field Mσ,δ,kM_{\sigma,\delta,k} in (7) obtained as the cutoff limit of the truncated field Mσ,δ,k(R)M_{\sigma,\delta,k}^{(R)} in (6) with real pencil Xk,BX_{k,B}, the symmetric matrix field Mσ,δ,ksymM_{\sigma,\delta,k}^{\mathrm{sym}} on matrices B∈Sym⁡NB \in \operatorname{Sym} N at a coordinate i∈{1,…,N}i \in \{1, \dots, N\} with vanishing off-diagonal entries Bij=0B_{ij} = 0 for j≠ij \neq i, and the matrix field primitive Jσ,δ,kJ_{\sigma,\delta,k} in (8).