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.

All papers

Claims

Findings

The irrationality exponent of pi is 2

The irrationality exponent of π\pi is 22; more precisely, for every real ν>2\nu > 2, there is an integer Q(ν)Q(\nu) such that ∣π−p/q∣≥q−ν|\pi - p/q| \ge q^{-\nu} for all p∈Zp \in \mathbb{Z} and q∈Zq \in \mathbb{Z} with q≥Q(ν)q \ge Q(\nu), whether or not p/qp/q is reduced.

The classical Flint–Hills series ∑n≥11n3sin⁡2n\sum_{n \ge 1} \frac{1}{n^3 \sin^2 n} converges, with angles in radians.

For fixed real numbers a,b>0a, b > 0, the generalized Flint–Hills series ∑n=1∞1na∣sin⁡n∣b\sum_{n=1}^{\infty} \frac{1}{n^a |\sin n|^b} with angles in radians converges if and only if a>max⁡{1,b}a > \max\{1, b\}.

Given integers m,K≥1m, K \ge 1 and positive rationals w0,v0,θw_0, v_0, \theta with 0<θ<10 < \theta < 1, Kθm<1K\theta^m < 1, and K(w0/v0)θm<1K(w_0/v_0)\theta^m < 1, there exist successive lower thresholds for positive rational weights w1,…,wmw_1, \dots, w_m (depending only on these fixed data and earlier weights, not on the centers) such that for any KK points aj=(1,cj1,…,cjm)∈Cm+1a_j = (1, c_{j1}, \dots, c_{jm}) \in \mathbb{C}^{m+1} with c0i,…,cK−1,ic_{0i}, \dots, c_{K-1,i} pairwise distinct for each ii, the jet evaluation map PW(H)→⨁j=0K−1JV(H)\mathcal{P}_W(H) \to \bigoplus_{j=0}^{K-1} \mathcal{J}_V(H) along Y=1+tY = 1+t, Xi=cji+ui+log⁡(1+t)X_i = c_{ji} + u_i + \log(1+t) is surjective for all sufficiently large, sufficiently divisible integers HH.

For 0<A<10 < A < 1 and 0<η<10 < \eta < 1, with g=ν(A(1−η)−θ)−(1−θ)g = \nu(A(1-\eta)-\theta) - (1-\theta) and L=η2Kθm(m+1)v0AmL = \frac{\eta^2 K \theta^m}{(m+1)v_0 A^m}, the rational approximation hypotheses (3.1) and the interpolation hypotheses of Theorem 2.1 cannot hold simultaneously if g>0g > 0, Ear+Ean<gE_{\mathrm{ar}} + E_{\mathrm{an}} < g, and cL>1+Ear+EancL > 1 + E_{\mathrm{ar}} + E_{\mathrm{an}}.

If α>0\alpha > 0 is irrational and ∥qα∥≥cq1−ν\|q\alpha\| \ge c q^{1-\nu} for all integers q≥1q \ge 1 with some c>0c > 0 and 1<ν<5/21 < \nu < 5/2, then ∑q=1∞1q3∥qα∥2<∞\sum_{q=1}^{\infty} \frac{1}{q^3 \|q\alpha\|^2} < \infty.

Methods

Separated-weight multivariable interpolation on logarithmic curves using a weighted transverse multiplicity comparison, a curve inequality via the derivative-ideal method, and ordinary blow-up positivity with Serre vanishing.

Source excerpts

“We prove the interpolation statement used in the determinant argument. Its essential uniformity is that the weights can be chosen before the interpolation points: each coordinate weight has a lower threshold depending on the earlier weights, but not on the locations of the points.” · Show in paper

“Let p : X′ → X be the ordinary blow-up of I, namely the relative Proj of its Rees algebra, and write IOX′ = OX′(−E).” · Show in paper

Interpolation-determinant method comparing an arithmetic lower bound (via Gaussian-integer scaling and LCM denominator clearing) against an analytic upper bound (via exact row translation to entire functions and Taylor-degree collision savings).

Source excerpts

“We estimate one full-row minor from below arithmetically, and then translate its rows to exact logarithmic periods to obtain an analytic upper bound.” · Show in paper

“The Taylor expansion of interpolation determinants eliminates repeated degrees [10, Theorem 1]. Our rows fall into different transverse-index groups, so this cancellation is used only inside a fixed group.” · Show in paper

Dyadic spacing estimate on the circle R/Z combined with nearest-multiple grouping to prove convergence of Flint–Hills-type trigonometric series.

Source excerpts

“For K ≤ q < 2K, the points qα on the circle R/Z have distance at least d from zero. Two distinct such points also have circle distance at least d: apply the hypothesis to their nonzero integer difference, whose absolute value is less than 2K, and use 1 − ν < 0.” · Show in paper

“For each positive integer n, choose a nearest nonnegative integer q to n/π, and group the integers n by q.” · Show in paper