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 nodal length on smooth surfaces

Let (M,g)(M, g) be a closed connected C∞C^\infty Riemannian surface. There is a finite constant C=C(M,g)C = C(M, g) such that every nonzero real Laplace eigenfunction satisfying −Δgu=λu-\Delta_g u = \lambda u, λ>0\lambda > 0, obeys Hg1(Zu)≤Cλ\mathcal{H}^1_g(Z_u) \le C\sqrt{\lambda}, where Hg1\mathcal{H}^1_g is Hausdorff measure for the Riemannian distance. Together with the known lower bound, this proves Yau's conjecture for smooth closed surfaces.

On a fixed planar disk Ω\Omega, if real smooth functions Uj,pjU_j, p_j satisfy ΔUj+Kj2pjUj=0\Delta U_j + K_j^2 p_j U_j = 0 with Sj→∞S_j \to \infty, Kj/Dj→0K_j/D_j \to 0, sup⁡j∥pj∥∞<∞\sup_j \|p_j\|_\infty < \infty, and Kj2∥∇pj∥∞/(SjDj)→0K_j^2 \|\nabla p_j\|_\infty / (S_j D_j) \to 0 (where Dj=Sj+∣Bj∣D_j = S_j + |B_j|), and if dμj=cje−2Bj⋅x∣Uj(x)∣2dxd\mu_j = c_j e^{-2 B_j \cdot x} |U_j(x)|^2 dx have mass at most 11 on Ω\Omega, then every upper semicontinuous logarithmic profile VV with the compact, open, and weighted bounds of Lemma 2.1 is subharmonic on Ω\Omega, with the identically −∞-\infty function allowed.

Let (M,g)(M, g) be a fixed smooth closed connected surface, and let G⊂MG \subset M be a fixed nonempty open coordinate disk. There is a finite constant C=C(M,g,G)C = C(M, g, G) such that every nonzero real eigenfunction −Δgu=k2u-\Delta_g u = k^2 u with k>0k > 0 satisfies ∥u∥L2(G)>0\|u\|_{L^2(G)} > 0 and log⁡(∥u∥L2(M)/∥u∥L2(G))≤Ck\log(\|u\|_{L^2(M)} / \|u\|_{L^2(G)}) \le Ck.

There are an integer A>10RA > 10R and constants C0,C1>0C_0, C_1 > 0, depending only on the fixed coordinate coefficient bounds and RR, such that for any admissible square QQ of side s≤1s \le 1, K=ks>0K = ks > 0, b∈R2b \in \mathbb{R}^2, and S>0S > 0 with B=KbB = Kb satisfying S≥C0S \ge C_0, E(Q,b)≤SE(Q,b) \le S, K/(S+∣B∣)≤1/C0K/(S+|B|) \le 1/C_0, and K2s/(S(S+∣B∣))≤1/C0K^2 s/(S(S+|B|)) \le 1/C_0, real vectors bqb_q can be chosen for the children q≺Qq \prec Q so that avgq≺Q∣bq−b∣≤C1S/K\mathrm{avg}_{q \prec Q} |b_q - b| \le C_1 S/K and avgq≺Qm(q,bq)≤12S/K\mathrm{avg}_{q \prec Q} m(q, b_q) \le \frac{1}{2} S/K.

There are constants P≥1P \ge 1 and σ>0\sigma > 0 such that if a square II and a tilt bIb_I satisfy ∣bI∣≥P|b_I| \ge P and m(I,bI)≤σ∣bI∣m(I, b_I) \le \sigma |b_I|, then at least half the paths from II to the terminal level satisfy N(Q)>LN(Q) > L at every visited square QQ, including II and the terminal square.

For the finite trees considered in Section 5, with Tterm\mathcal{T}_{\mathrm{term}} the collection of terminal squares and sterms_{\mathrm{term}} their common side length, there is a constant CC, independent of the eigenfunction and the number of levels, such that avgq∈TtermN(q)≤C\mathrm{avg}_{q \in \mathcal{T}_{\mathrm{term}}} N(q) \le C and avgq∈TtermE(q,0)≤Cksterm\mathrm{avg}_{q \in \mathcal{T}_{\mathrm{term}}} E(q,0) \le C k s_{\mathrm{term}}.

There is an ε>0\varepsilon > 0 depending only on the fixed coordinate coefficient bounds and neighborhood from Section 3 such that, for every admissible square II of side ss with ks≤εks \le \varepsilon, H1(I∩Zu)≤Cs(1+E(I,0))\mathcal{H}^1(I \cap Z_u) \le Cs(1 + E(I,0)), with constants independent of uu, kk, and ss.

Methods

Extraction of upper semicontinuous logarithmic profiles from sequences of nonnegative Borel measures with exponential weights via diagonalization on a countable basis of relatively compact open sets.

Source excerpts

“Lemma 2.1 (Extraction of a logarithmic profile). Let X be an open subset of R2 or a compact smooth surface. Suppose that Sj → ∞ and that μj are nonnegative Borel measures on X with μj (X) ≤ 1. There are a subsequence and an upper semicontinuous function V : X → [−∞, 0] such that, for every compact F ⊂ X and every open G ⊂ X, lim sup_{j→∞} log μj(F)/(2Sj) ≤ sup_F V, lim inf_{j→∞} log μj(G)/(2Sj) ≥ sup_G V.” · Show in paper

Localized two-dimensional Carleman estimate with an affine tilt Bj⋅xB_j \cdot x that enlarges the weight gradient scale Dj=Sj+∣Bj∣D_j = S_j + |B_j| without enlarging the curvature scale SjS_j.

Source excerpts

“Put Tj(x) = −Bj · x − Sjχ(x). There are a neighborhood Ω0 of x0, a constant c > 0, and an index j0 such that ∥eTje^{Tj}(Δ + K_j^2 pj)w∥_2 ≥ c√(Sj Dj) ∥eTje^{Tj}w∥_2 for all j ≥ j0 and all real w ∈ C_c^∞(Ω0).” · Show in paper

Affine correction of subharmonic logarithmic profiles on AA-adic child squares using a truncated Riesz potential gradient cy=∇h(y)+∫∣z−y∣>2Rρy−z∣y−z∣2dν(z)c_y = \nabla h(y) + \int_{|z-y|>2R\rho} \frac{y-z}{|y-z|^2} d\nu(z).

Source excerpts

“Define cy = ∇h(y) + ∫_{z∈D4, |z−y|>2Rρ} (y − z)/|y − z|^2 dν(z). Then sup_{DRρD_{Rρ}(y)}(V(x) − cy · x) − sup_{D_ρ(y)}(V(x) − cy · x) ≤ C(ρ^2 + ∫_{D4} min{1, ρ^2/|z − y|^2} dν(z)).” · Show in paper

Stopped descent along AA-adic subdivision trees with tilt updates to control the expected sum of residual growth and tilt increments.

Source excerpts

“Lemma 4.2 (Stopped descent). Start at I with a prescribed tilt bI. Whenever a step is made, choose the child tilts by Lemma 4.1, and suppose that its applicability condition holds at every square from which the procedure continues. Stop at a time T, no later than the terminal level.” · Show in paper

Geometric packing of grid squares from nearby witness sets YQ⊂DQY_Q \subset DQ by bounding ancestry-membership mismatches within boundary collars.

Source excerpts

“Lemma 5.4 (Packing from nearby witnesses). Let F be a finite family of squares in the A-adic subdivision tree of a square Q0, where A ≥ 2. Suppose there are fixed D ≥ 1, δ > 0, and an integer J ≥ 0 such that every Q ∈ F admits a measurable set YQ ⊂ DQ satisfying |YQ| ≥ δ|Q|, #{F ∈ F : s(F) < s(Q), y ∈ F} ≤ J for almost every y ∈ YQ. Then ∑_{Q∈F} |Q| ≤ C(A, D, δ, J)|Q0|.” · Show in paper