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

Positive lower density of large prime gaps

For every fixed real C>0C > 0, there are constants c(C)>0c(C) > 0 and N0(C)N_0(C) such that, for every integer N≥N0(C)N \ge N_0(C), #{1≤n≤N:pn+1−pn>Clog⁡pn}≥c(C)N\#\{1 \le n \le N : p_{n+1} - p_n > C \log p_n\} \ge c(C)N.

The set {n≥1:pn/n<pn+1/(n+1)}\{n \ge 1 : p_n/n < p_{n+1}/(n+1)\} has positive lower asymptotic density, answering a question of Erdős and Prachar affirmatively.

For each fixed λ>0\lambda > 0 there is a finite constant Kλ>0K_\lambda > 0 such that for every ε>0\varepsilon > 0, there are weights WX(m)≥0W_X(m) \ge 0 for integers X<m≤2XX < m \le 2X satisfying EXWX→1\mathbb{E}_X W_X \to 1, EX(WXVJ)→λ\mathbb{E}_X(W_X V_J) \to \lambda, lim sup⁡X→∞EX(WXVI)≤ε\limsup_{X\to\infty} \mathbb{E}_X(W_X V_I) \le \varepsilon, lim sup⁡X→∞EX(WXVJ2)≤Kλ\limsup_{X\to\infty} \mathbb{E}_X(W_X V_J^2) \le K_\lambda, and EXWX2=Oλ,ε(1)\mathbb{E}_X W_X^2 = O_{\lambda,\varepsilon}(1), with KλK_\lambda independent of ε\varepsilon.

If h≥1h \ge 1 and for at least δX\delta X integers m∈(X,2X]m \in (X, 2X], the interval (m,m+h](m, m+h] contains a prime and (m+h,m+2h](m+h, m+2h] contains none, then at least δX/h\delta X/h distinct primes p∈(X,2X+h]p \in (X, 2X+h] have a consecutive successor p+p^+ satisfying p+−p>hp^+ - p > h.

For fixed λ,k\lambda, k, and (fj)0≤j≤k(f_j)_{0\le j\le k}, as X→∞X \to \infty, EXZ2→w:=∑j=0kαj∥fj∥22\mathbb{E}_X Z^2 \to w := \sum_{j=0}^k \alpha_j \|f_j\|_2^2, EXZ4=Oλ,k,f(1)\mathbb{E}_X Z^4 = O_{\lambda,k,\mathbf{f}}(1), EX(Z2VJ)→λw\mathbb{E}_X(Z^2 V_J) \to \lambda w, EX(Z2VI)→v:=λ∑j=0kαj∥fj+Tfj+1∥22\mathbb{E}_X(Z^2 V_I) \to v := \lambda \sum_{j=0}^k \alpha_j \|f_j + T f_{j+1}\|_2^2, and lim sup⁡X→∞EX(Z2VJ2)≤C∗w\limsup_{X\to\infty} \mathbb{E}_X(Z^2 V_J^2) \le C_* w where C∗=λ+λ2cG2C_* = \lambda + \lambda^2 c_G^2 is independent of kk and (fj)(f_j).

For every fixed λ>0\lambda > 0, there are families of the type used in Proposition 3.1, with increasing finite maximum dimension kk, for which w→1w \to 1 and v→0v \to 0.

Under the support budgets ∑rρr<1\sum_r \rho_r < 1 if δ=0\delta = 0 and ∑rρr<1/2\sum_r \rho_r < 1/2 if δ=1\delta = 1, EX[χδ(m)∏rDFr(m;b(r))]=L−t{S(H)C+O(L−1/2(log⁡L)A)}\mathbb{E}_X[\chi_\delta(m) \prod_r D_{F_r}(m; \mathbf{b}^{(r)})] = L^{-t}\{\mathfrak{S}(\mathcal{H})\mathcal{C} + O(L^{-1/2}(\log L)^A)\} uniformly in the allowed shifts.

For fixed s≥0s \ge 0 and K≥1K \ge 1, if I1,…,IsI_1, \dots, I_s are intervals of hh consecutive integers contained in an interval of diameter at most KhKh, then ∑ai∈Ii, distinctS({a1,…,as})=hs(1+o(1))\sum_{a_i \in I_i, \text{ distinct}} \mathfrak{S}(\{a_1, \dots, a_s\}) = h^s(1 + o(1)) as h→∞h \to \infty.

Methods

Constructing a nonnegative weight WX(m)=Z(m)2/wW_X(m) = Z(m)^2/w as the normalized square of a signed sum of smooth divisor sums over subsets of the second interval I={h+1,…,2h}I = \{h+1, \dots, 2h\} to detect a prime in J={1,…,h}J = \{1, \dots, h\} while suppressing primes in II.

Source excerpts

“Define the signed sum Z(m) = X k j=0 X S⊂I |S|=j DFj (m; S), αj = λj j! .” · Show in paper

“Freeze this family and define WX(m) = Z(m)2 w .” · Show in paper

Achieving cancellation between adjacent dimensions in the marked second moment vv by choosing alternating-sign product test functions fj(t)=(−1)jrαjQj(t)χ(Sj)f_j(\mathbf{t}) = \frac{(-1)^j}{\sqrt{r}\sqrt{\alpha_j}} Q_j(\mathbf{t}) \chi(S_j) supported on a simplex over r=⌊k⌋r = \lfloor \sqrt{k} \rfloor neighboring levels.

Source excerpts

“Take r = ⌊ √ k⌋ and a = k − r + 1. Choose once and for all a smooth cutoff χ with 0 ≤ χ ≤ 1, equal to one on [0, β] and zero on [γ,∞). At the r indices a ≤ j ≤ k, set fj (t) = (−1)j √ r √αj Qj (t)χ(Sj ), (9)” · Show in paper

Evaluating uniform mixed moments of smooth divisor sums with at most one prime mark via the Bombieri–Vinogradov theorem with uniform shifted endpoints (Lemma 5.2) and Fourier inversion of the finite density sum.

Source excerpts

“Lemma 5.2 (Weighted distribution with uniform shifted endpoints). Assume the Bombieri–Vinogradov theorem...” · Show in paper

“Choose a smooth compact extension Fe r of each positive-dimensional function and define its Fourier transform by Φr(u) = (2π) −jr Z Rjr e P ℓ vℓFe r(v)eiu·v dv.” · Show in paper

Averaging the prime-tuple singular series over distinct shifts in boxes of length hh by truncating the Euler product at y=(1/4)log⁡hy = (1/4)\log h.

Source excerpts

“We give a self-contained proof by truncating the Euler product; no statement about the distribution of primes in individual short intervals is assumed.” · Show in paper