By OpenAISourceGitHub

Manuscripts from openai/math

Fork
Clone
git clone https://openversions.com/git/index/openai-math.git
Cite
Version link
https://openversions.com/index/openai-math/commit/dfd8c46870b48511e697cbb43a3c8c91d375fa2f
Version identifier (commit hash)
dfd8c46870b48511e697cbb43a3c8c91d375fa2f

Fork to create your own copy with credit to the source. Clone to get a Git copy on your computer.

Latest version

Checks complete

Sync OpenAI Math manuscripts from 4eaee87548250b2682affda7d05d947c05554bae

Committed

All papers

Claims

Findings

Sharp one-dimensional Lieb–Thirring constants

For every 1/2<γ<3/21/2 < \gamma < 3/2 and every nonnegative W∈Lγ+1/2(R)W \in L^{\gamma+1/2}(\mathbb{R}), Tr⁡(H−W)−γ≤2(γ−1/2γ+1/2)γ−1/2Lγ,1cl∫RWγ+1/2\operatorname{Tr}(H_{-W})_-^\gamma \le 2 \left(\frac{\gamma - 1/2}{\gamma + 1/2}\right)^{\gamma - 1/2} L_{\gamma,1}^{\mathrm{cl}} \int_{\mathbb{R}} W^{\gamma+1/2}. The constant is optimal and equals Lγ,1(1)L_{\gamma,1}^{(1)}, with equality attained by W(x)=(r+1)sech⁡2(rx)W(x) = (r+1)\operatorname{sech}^2(rx) for r=(γ−1/2)−1r = (\gamma - 1/2)^{-1}.

Let u=(u1,…,uN)T∈H01(I;RN)u = (u_1, \dots, u_N)^{\mathsf{T}} \in H_0^1(I; \mathbb{R}^N) satisfy ∫IuuTdx=IN\int_I u u^{\mathsf{T}} dx = I_N, and let ki>0k_i > 0. Then sup⁡C lower triangularEK(Cu)≥∑i=1N∫−kiki(ki2−s2)σds\sup_{C \text{ lower triangular}} \mathcal{E}_K(Cu) \ge \sum_{i=1}^N \int_{-k_i}^{k_i} (k_i^2 - s^2)^\sigma ds, where the supremum is over all real lower triangular N×NN \times N matrices.

For every B∈Sym⁡NB \in \operatorname{Sym}_N and every a∈RNa \in \mathbb{R}^N, M(B)=aaT  ⟹  aT(K2−B2)a≥∣a∣2qM(B) = aa^{\mathsf{T}} \implies a^{\mathsf{T}}(K^2 - B^2)a \ge |a|^{2q}.

For every ε>0\varepsilon > 0, there are a real lower triangular matrix CC and a path B∈C1(I‾;Sym⁡N)B \in C^1(\overline{I}; \operatorname{Sym}_N) such that, with a=Cua = Cu, εB′=M(B)−aaT\varepsilon B' = M(B) - aa^{\mathsf{T}}, B(x−)=KB(x_-) = K, B(x+)=−KB(x_+) = -K, satisfying ∥B(x)∥op≤∥K∥op\|B(x)\|_{\mathrm{op}} \le \|K\|_{\mathrm{op}} for x∈Ix \in I, with CC bounded uniformly for 0<ε≤10 < \varepsilon \le 1.

Let VV be real measurable with V−∈Lp(R)V_- \in L^p(\mathbb{R}), and suppose DV={v∈H1(R):∫RV+∣v∣2<∞}\mathcal{D}_V = \{v \in H^1(\mathbb{R}) : \int_{\mathbb{R}} V_+ |v|^2 < \infty\} is dense in L2(R)L^2(\mathbb{R}) (which holds in particular if V+∈Lloc1(R)V_+ \in L_{\mathrm{loc}}^1(\mathbb{R})). Then the form hV[v]=∫R∣v′∣2+∫RV+∣v∣2−∫RV−∣v∣2h_V[v] = \int_{\mathbb{R}} |v'|^2 + \int_{\mathbb{R}} V_+|v|^2 - \int_{\mathbb{R}} V_-|v|^2 on DV\mathcal{D}_V is closed and bounded below, and its associated operator HVH_V has discrete negative spectrum and satisfies Tr⁡(HV)−γ≤ℓσ∫RV−p\operatorname{Tr}(H_V)_-^\gamma \le \ell_\sigma \int_{\mathbb{R}} V_-^p.

Methods

Construct a locally Lipschitz regularized matrix field Mδ:Sym⁡N→Sym⁡NM_\delta : \operatorname{Sym}_N \to \operatorname{Sym}_N by integrating f(K2−2sB+s2IN)−f(s2)INf(K^2 - 2sB + s^2 I_N) - f(s^2)I_N over symmetric cutoffs [−R,R][-R, R] and taking R→∞R \to \infty, together with a C1C^1 trace primitive J(B)=−∫01tr⁡(BM(tB)) dtJ(B) = -\int_0^1 \operatorname{tr}(BM(tB))\,dt.

Source excerpts

“X_s(B) = K^2 − 2sB + s^2 I_N, M_R(B) = c ∫_{-R}^R (f(X_s(B)) − f(s^2)I_N) ds, M(B) = lim_{R→∞} M_R(B). (3.2)” · Show in paper

“J(B) = − ∫_0^1 tr(BM(tB)) dt (3.3)” · Show in paper

Compare Xs(B)=K2−2sB+s2INX_s(B) = K^2 - 2sB + s^2 I_N with Ys=(sI−B)2+dIY_s = (sI - B)^2 + dI using a resolvent-based tangent inequality on positive matrices, and express the integrated tangent map P\mathcal{P} as an average of unitary conjugations ∫ReitBZe−itBdν(t)\int_{\mathbb{R}} e^{itB} Z e^{-itB} d\nu(t) controlled by the generalized Klein trace inequality.

Source excerpts

“M(B) ≤ mI + μ'(d)P(D − dI), P = (c / μ'(d)) ∫_R T_s ds. (4.7)” · Show in paper

“There is an even probability measure ν on R such that P(Z) = ∫_R eitBe^{itB} Z e−itBe^{−itB} dν(t). (4.9)” · Show in paper

Find a lower triangular coefficient matrix CC and a connecting matrix path BB solving εB′=M(B)−(Cu)(Cu)T\varepsilon B' = M(B) - (Cu)(Cu)^{\mathsf{T}} from KK to −K-K by applying a relative mod-two continuation principle to a homotopy of sections on the quotient bundle (L××Sym⁡N)/G→L×/G(\mathcal{L}^\times \times \operatorname{Sym}_N)/G \to \mathcal{L}^\times / G modulo row signs.

Source excerpts

“Lemma 5.2 (Continuation modulo two). Let Q be a smooth d-dimensional manifold without boundary, and let E → Q be a smooth real vector bundle of rank d. Suppose s_θ, 0 ≤ θ ≤ 1, is a continuous homotopy of sections and {(q, θ) ∈ Q × [0, 1] : s_θ(q) = 0} is compact. If s_0 is smooth and has an odd number of zeros, all transverse, then s_1 has a zero.” · Show in paper

“E = (L^× × Sym_N)/G −→ Q, (C, Z) ∼ (SC, SZS). By (5.9), the maps Φ_θ descend to a continuous homotopy of sections of this bundle.” · Show in paper

Pass from finite-interval Dirichlet eigenvalue bounds to the full-line Schrödinger operator H−WH_{-W} by multiplying finite orthonormal families of eigenfunctions by smooth compactly supported cutoffs χR\chi_R and applying the form-domain min-max principle as R→∞R \to \infty.

Source excerpts

“Choose smooth cutoffs χ_R equal to one on [−R, R], supported in I_R = (−2R, 2R), and satisfying 0 ≤ χ_R ≤ 1 and ||χ'_R||_∞ ≤ c/R. Then vj,Rv_{j,R} = χ_R φ_j −→ φ_j in H^1(R).” · Show in paper

“Applying the form-domain min–max principle [20, Theorem 4.10 and the following form-domain extension] to the span of v1,Rv_{1,R}, . . . , vj,Rv_{j,R} gives, for every j ≤ N, λ_j(I_R) ≤ λ_j + η_R,” · Show in paper