OctoInformBeta

OpenAI preprint, September 2026 · §7. From the action inequality to spectral moments · §7

Proposition 7.2: Finite-interval trial bounds for σ∈(0,1)\sigma \in (0, 1) and whole-line spectral moment bounds for γ∈(1/2,3/2)\gamma \in (1/2, 3/2)

Action bounds and sharp Lieb-Thirring inequalities.

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

Consider a one-dimensional Schrödinger operator −d2/dx2−W-d^2/dx^2 - W with an almost everywhere nonnegative potential WW on ℝ\mathbb{R} and a spectral exponent γ∈(1/2,3/2)\gamma \in (1/2, 3/2), where σ=γ−1/2∈(0,1)\sigma = \gamma - 1/2 \in (0, 1). Given NN orthonormal eigenfunctions with negative eigenvalues −k12,…,−kN2-k_1^2, \dots, -k_N^2 (ki>0k_i > 0), or trial functions with Dirichlet boundary conditions on a compact interval [l,r][l, r], the question is how action inequalities for lower triangular combinations yield sharp bounds on the spectral moments ∑i=1Nki2γ\sum_{i=1}^N k_i^{2\gamma} in terms of ∫ℝW(x)γ+1/2dx\int_{\mathbb{R}} W(x)^{\gamma + 1/2} \, dx.

Definitions

Given a natural number NN, a real number σ\sigma, and vectors k,a,v∈ℝNk, a, v \in \mathbb{R}^N, the action density 𝒜σ(k,a,v)\mathcal{A}_\sigma(k, a, v) is the real number defined by (1): 𝒜σ(k,a,v)=∑i=1Nvi2+∑i=1Nki2ai2−(∑i=1Nai2)1+1/σ.\mathcal{A}_\sigma(k, a, v) = \sum_{i=1}^N v_i^2 + \sum_{i=1}^N k_i^2 a_i^2 - \left( \sum_{i=1}^N a_i^2 \right)^{1 + 1/\sigma}.(1)✦

For a natural number NN, the lower triangular matrix space ℒN\mathcal{L}_N is the space of real N×NN \times N matrices CC whose entries satisfy Cij=0C_{ij} = 0 whenever i<ji < j.✦

At any real number σ\sigma, the action Young constant c(σ)c(\sigma) is the real number given by (2): c(σ)=σσ(1+σ)1+σ.c(\sigma) = \frac{\sigma^\sigma}{(1 + \sigma)^{1 + \sigma}}.(2)✦

The beta action integral β(σ)\beta(\sigma) at a real number σ\sigma is the real number defined by (3): β(σ)=∫−11(1−t2)σdt.\beta(\sigma) = \int_{-1}^1 (1 - t^2)^\sigma \, dt.(3)✦

The C1C^1 Sobolev space C1(ℝ)∩H1(ℝ)C^1(\mathbb{R}) \cap H^1(\mathbb{R}) is the real vector space of continuously differentiable functions u:ℝ→ℝu : \mathbb{R} \to \mathbb{R} such that u∈L2(ℝ)u \in L^2(\mathbb{R}) and u′∈L2(ℝ)u' \in L^2(\mathbb{R}).✦

A function u∈C1(ℝ)∩H1(ℝ)u \in C^1(\mathbb{R}) \cap H^1(\mathbb{R}) satisfies the integrated eigenvalue equation for a function W:ℝ→ℝW : \mathbb{R} \to \mathbb{R} and a real parameter kk if (4) holds for all a,b∈ℝa, b \in \mathbb{R}: u′(b)−u′(a)=∫ab(k2−W(x))u(x)dx.u'(b) - u'(a) = \int_a^b \bigl(k^2 - W(x)\bigr) u(x) \, dx.(4)✦

At a spectral exponent γ∈ℝ\gamma \in \mathbb{R}, the semiclassical constant LγclL_\gamma^{\mathrm{cl}} is the real number defined in terms of the Euler gamma function Γ\Gamma by (5): Lγcl=Γ(γ+1)2πΓ(γ+3/2).L_\gamma^{\mathrm{cl}} = \frac{\Gamma(\gamma + 1)}{2 \sqrt{\pi}\, \Gamma(\gamma + 3/2)}.(5)✦

In terms of the semiclassical constant LγclL_\gamma^{\mathrm{cl}}, the sharp constant LγL_\gamma at γ∈ℝ\gamma \in \mathbb{R} is the real number given by (6): Lγ=2(γ−1/2γ+1/2)γ−1/2Lγcl.L_\gamma = 2 \left(\frac{\gamma - 1/2}{\gamma + 1/2}\right)^{\gamma - 1/2} L_\gamma^{\mathrm{cl}}.(6)✦

The Claim

Theorem 1. Given a natural number NN, a real number σ>0\sigma > 0, real endpoints l≤rl \le r, and k∈ℝNk \in \mathbb{R}^N, suppose a,v:ℝ→ℝNa, v : \mathbb{R} \to \mathbb{R}^N are continuous on [l,r][l, r], and let W:ℝ→ℝW : \mathbb{R} \to \mathbb{R} be integrable on [l,r][l, r] and nonnegative almost everywhere on ℝ\mathbb{R} with W1+σ∈L1(ℝ)W^{1+\sigma} \in L^1(\mathbb{R}). If aa, vv, kk, and WW satisfy the quadratic-form inequality (7), ∫lr(∑i=1Nvi(t)2+∑i=1Nki2ai(t)2)dt≤∫lrW(t)∑i=1Nai(t)2dt,\int_l^r \left( \sum_{i=1}^N v_i(t)^2 + \sum_{i=1}^N k_i^2 a_i(t)^2 \right) dt \le \int_l^r W(t) \sum_{i=1}^N a_i(t)^2 \, dt,(7) then the triangular action bound (8) holds: ∫lr𝒜σ(k,a(t),v(t))dt≤c(σ)∫ℝW(t)1+σdt.\int_l^r \mathcal{A}_\sigma\bigl(k, a(t), v(t)\bigr) \, dt \le c(\sigma) \int_{\mathbb{R}} W(t)^{1 + \sigma} \, dt.(8)✦

For N∈ℕN \in \mathbb{N}, σ∈(0,1)\sigma \in (0, 1), real endpoints l≤rl \le r, and k∈ℝNk \in \mathbb{R}^N with ki>0k_i > 0 for every ii, let u,v:ℝ→ℝNu, v : \mathbb{R} \to \mathbb{R}^N be continuous on [l,r][l, r] such that each component uiu_i is differentiable on (l,r)(l, r) with ui′(t)=vi(t)u_i'(t) = v_i(t), while u(l)=u(r)=0u(l) = u(r) = 0 and ∫lrui(t)uj(t)dt\int_l^r u_i(t) u_j(t) \, dt equals 11 for i=ji = j and 00 for i≠ji \neq j. Let W:ℝ→ℝW : \mathbb{R} \to \mathbb{R} be integrable on [l,r][l, r] and nonnegative almost everywhere on ℝ\mathbb{R} with W1+σ∈L1(ℝ)W^{1+\sigma} \in L^1(\mathbb{R}). If for every lower triangular matrix C∈ℒNC \in \mathcal{L}_N, the matrix-vector products Cu(t)C u(t) and Cv(t)C v(t) satisfy (9), ∫lr(∑i=1N(Cv(t))i2+∑i=1Nki2(Cu(t))i2)dt≤∫lrW(t)∑i=1N(Cu(t))i2dt,\int_l^r \left( \sum_{i=1}^N \bigl(C v(t)\bigr)_i^2 + \sum_{i=1}^N k_i^2 \bigl(C u(t)\bigr)_i^2 \right) dt \le \int_l^r W(t) \sum_{i=1}^N \bigl(C u(t)\bigr)_i^2 \, dt,(9) then the finite spectral trial bound (10) holds: ∑i=1N∫−kiki(ki2−s2)σds≤c(σ)∫ℝW(t)1+σdt.\sum_{i=1}^N \int_{-k_i}^{k_i} (k_i^2 - s^2)^\sigma \, ds \le c(\sigma) \int_{\mathbb{R}} W(t)^{1 + \sigma} \, dt.(10)✦

Suppose N∈ℕN \in \mathbb{N}, σ∈(0,1)\sigma \in (0, 1), and W:ℝ→ℝW : \mathbb{R} \to \mathbb{R} is locally integrable and nonnegative almost everywhere with W1+σ∈L1(ℝ)W^{1+\sigma} \in L^1(\mathbb{R}), and let u1,…,uNu_1, \dots, u_N be an orthonormal family in C1(ℝ)∩H1(ℝ)C^1(\mathbb{R}) \cap H^1(\mathbb{R}) with positive parameters k1,…,kN>0k_1, \dots, k_N > 0 such that each uiu_i satisfies the integrated eigenvalue equation (4) for WW at kik_i. If k1≥…≥kNk_1 \ge \dots \ge k_N, then the sorted eigenvalue action bound (11) holds: (∑i=1Nki1+2σ)β(σ)≤c(σ)∫ℝW(x)1+σdx.\left( \sum_{i=1}^N k_i^{1 + 2\sigma} \right) \beta(\sigma) \le c(\sigma) \int_{\mathbb{R}} W(x)^{1 + \sigma} \, dx.(11)✦ Without the ordering hypothesis k1≥…≥kNk_1 \ge \dots \ge k_N, the eigenvalue action bound (12) holds: (∑i=1Nki1+2σ)β(σ)≤c(σ)∫ℝW(x)1+σdx.\left( \sum_{i=1}^N k_i^{1 + 2\sigma} \right) \beta(\sigma) \le c(\sigma) \int_{\mathbb{R}} W(x)^{1 + \sigma} \, dx.(12)✦

When γ∈(1/2,3/2)\gamma \in (1/2, 3/2), let W:ℝ→ℝW : \mathbb{R} \to \mathbb{R} be locally integrable and nonnegative almost everywhere with Wγ+1/2∈L1(ℝ)W^{\gamma + 1/2} \in L^1(\mathbb{R}), let Λ\Lambda be a finite index set, and let (ui)i∈Λ(u_i)_{i \in \Lambda} be an orthonormal family in C1(ℝ)∩H1(ℝ)C^1(\mathbb{R}) \cap H^1(\mathbb{R}) with positive parameters (ki)i∈Λ(k_i)_{i \in \Lambda} such that each uiu_i satisfies (4) for WW at kik_i. Then the real Lieb-Thirring bound (13) holds: ∑i∈Λki2γ≤Lγ∫ℝW(x)γ+1/2dx.\sum_{i \in \Lambda} k_i^{2\gamma} \le L_\gamma \int_{\mathbb{R}} W(x)^{\gamma + 1/2} \, dx.(13)✦

In the paper, the result reads: for a bounded interval II, an exponent γ∈(1/2,3/2)\gamma \in (1/2, 3/2) with p=γ+1/2p = \gamma + 1/2 and σ=γ−1/2\sigma = \gamma - 1/2, and a potential 0≤W∈Lp(I)0 \le W \in L^p(I), if −k12,…,−kN2-k_1^2, \dots, -k_N^2 are the first NN negative Dirichlet eigenvalues of −d2/dx2−W-d^2/dx^2 - W ordered so that k1≥…≥kN>0k_1 \ge \dots \ge k_N > 0, then ∑j=1Nkj2γ≤Lγ∫IWp\sum_{j=1}^N k_j^{2\gamma} \le L_\gamma \int_I W^p, where LγL_\gamma stands for the paper’s constant ℓσ\ell_\sigma. Instead of a Dirichlet eigenvalue problem on II, the bounds (8) and (10) on [l,r][l, r] and (11) through (13) on ℝ\mathbb{R} are formulated for finite-interval trial families and finite orthonormal families in C1(ℝ)∩H1(ℝ)C^1(\mathbb{R}) \cap H^1(\mathbb{R}) satisfying (4).

What Is Checked

The bounds (8) through (13) are stated over real exponents σ>0\sigma > 0, σ∈(0,1)\sigma \in (0, 1), and γ∈(1/2,3/2)\gamma \in (1/2, 3/2), almost everywhere nonnegative potentials W:ℝ→ℝW : \mathbb{R} \to \mathbb{R} that are integrable on [l,r][l, r] (l≤rl \le r) or locally integrable on ℝ\mathbb{R} with W1+σ∈L1(ℝ)W^{1+\sigma} \in L^1(\mathbb{R}) (respectively Wγ+1/2∈L1(ℝ)W^{\gamma + 1/2} \in L^1(\mathbb{R})), continuous finite-interval functions and orthonormal Dirichlet trial families on [l,r][l, r] over lower triangular matrices in ℒN\mathcal{L}_N, and finite orthonormal families in C1(ℝ)∩H1(ℝ)C^1(\mathbb{R}) \cap H^1(\mathbb{R}) satisfying (4) at positive parameters (sorted in (11)), using the action density 𝒜σ\mathcal{A}_\sigma, action Young constant c(σ)c(\sigma), beta action integral β(σ)\beta(\sigma), semiclassical constant LγclL_\gamma^{\mathrm{cl}}, and sharp constant LγL_\gamma.