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 Poisson-Dirichlet law for prime predecessors

For every fixed k≥1k \ge 1 and every bounded continuous function F:[0,1]k→RF : [0, 1]^k \to \mathbb{R}, lim⁡x→∞1π(x)−1∑3≤p≤x, p primeF(V1(p),…,Vk(p))=EF(L1,…,Lk)\lim_{x \to \infty} \frac{1}{\pi(x) - 1} \sum_{3 \le p \le x, \, p \text{ prime}} F(V_1(p), \dots, V_k(p)) = \mathbb{E} F(L_1, \dots, L_k), where Vj(p)=log⁡qj(p)log⁡(p−1)V_j(p) = \frac{\log q_j(p)}{\log(p-1)} for the decreasing prime factors qj(p)q_j(p) of p−1p-1 with multiplicity and (L1,L2,… )(L_1, L_2, \dots) has the Poisson–Dirichlet distribution PD(1)\mathrm{PD}(1).

For fixed δ,C,D∗>0\delta, C, D_* > 0 and q∈(0,1)q \in (0, 1), Hm,Hn≥xδH_m, H_n \ge x^\delta with X=HmHn≍xX = H_m H_n \asymp x, ∣F(h)∣≤1|F(h)| \le 1 invariant under group primes, αm=miv(1m prime−1m roughV(W)log⁡m)\alpha_m = m^{iv}(\mathbf{1}_{m \text{ prime}} - \frac{\mathbf{1}_{m \text{ rough}}}{V(W)\log m}) with ∣v∣≤LC|v| \le L^C, and βn\beta_n supported on rough integers in [Hn,2Hn][H_n, 2H_n] with ∣βn∣≤LC|\beta_n| \le L^C, if KK is sufficiently large in terms of δ,C,D∗,q\delta, C, D_*, q (independently of the band exponents aia_i), then ∣∑m,nαmβnF(mn−1)W(mn−1)∣≪XL−D∗\left|\sum_{m,n} \alpha_m \beta_n F(mn-1)\mathcal{W}(mn-1)\right| \ll X L^{-D_*}.

For every fixed E0>0E_0 > 0, choosing A0A_0 and then KK sufficiently large in terms of E0,qE_0, q and the fixed parameters of Theorem 3.1 yields 1UV∑P∈Ω primitive⟨uP,(AA∗)RuP⟩≤L−E0N\frac{1}{UV} \sum_{P \in \Omega \text{ primitive}} \langle \mathbf{u}_P, (\mathcal{A}\mathcal{A}^*)^R \mathbf{u}_P \rangle \le L^{-E_0 N}, and consequently ∣QY−QYmaj∣≪AXYL−A|\mathcal{Q}_Y - \mathcal{Q}_Y^{\mathrm{maj}}| \ll_A XY L^{-A} for any prescribed fixed A>0A > 0.

For every fixed D>0D > 0, the unrestricted major sum QYmaj\mathcal{Q}_Y^{\mathrm{maj}} in (5.1) satisfies ∣QYmaj∣≪DXYL−D|\mathcal{Q}_Y^{\mathrm{maj}}| \ll_D XY L^{-D} uniformly in the coefficient intervals, ∣v∣≤LC|v| \le L^C, and the admissible scale YY.

For fixed d≥1d \ge 1, ε>0\varepsilon > 0, intervals I1,…,IdI_1, \dots, I_d with lower endpoints ≥xε\ge x^\varepsilon and product of upper endpoints ≤x1−ε\le x^{1-\varepsilon}, Fx(n)=W(n)(∑l1l∣n−Cx)F_x(n) = \mathcal{W}(n)(\sum_{\mathbf{l}} \mathbf{1}_{l|n} - C_x), and G(n)=G(n∗)G(n) = G(n_*) with ∣G(n)∣≤LCτ(n∗)C|G(n)| \le L^C \tau(n_*)^C, for every fixed smooth compactly supported Ψ:(0,∞)→C\Psi : (0, \infty) \to \mathbb{C} and A>0A > 0, ∑w≥1Ψ(w/x)wFx(w)G(w+1)W1(w)W1(w+1)≪AL−A\sum_{w \ge 1} \frac{\Psi(w/x)}{w} F_x(w) G(w+1) W_1(w) W_1(w+1) \ll_A L^{-A} uniformly over admissible active groups and slot intervals.

For fixed d≥1d \ge 1, ε>0\varepsilon > 0, prime intervals Ii\mathcal{I}_i with lower endpoints ≥xε\ge x^\varepsilon and product of upper endpoints ≤x1−ε\le x^{1-\varepsilon}, and every nonnegative Φ∈Cc∞((0,∞))\Phi \in C_c^\infty((0, \infty)), ∑p primeΦ((p−1)/x)(fx(p−1)−Cx)=o(x/log⁡x)\sum_{p \text{ prime}} \Phi((p-1)/x)(f_x(p-1) - C_x) = o(x/\log x) as real x→∞x \to \infty.

For fixed 0<τ<η<10 < \tau < \eta < 1 and C,A>0C, A > 0, there is B0=B0(τ,η,C,A)B_0 = B_0(\tau, \eta, C, A) such that uniformly for xτ/2≤N≤xηx^\tau/2 \le N \le x^\eta, q≤LCq \le L^C, LB0≤∣t∣≤x2L^{B_0} \le |t| \le x^2, every Dirichlet character χ(modq)\chi \pmod q and interval I⊂[N,2N]I \subset [N, 2N] satisfy ∣∑p∈Iχ(p)p−1+it∣≪τ,η,C,AL−A\left|\sum_{p \in I} \chi(p) p^{-1+it}\right| \ll_{\tau,\eta,C,A} L^{-A}.

Methods

Signed-memory determinant moment method: averaging long signed moments of a small-determinant relation between primitive lattice vectors using a primitive-root residue formula, independent projective residue lines, and an exact primewise memory expansion that records a prime across gaps between its active uses.

Source excerpts

“Cauchy's inequality, applied after dividing out one tuple of marks, leads to a small-determinant relation between primitive lattice vectors. We average its long signed moments using a root-residue formula and an exact memory expansion. The memory records a prime across gaps between its uses, so its divisibility probability is charged only once.” · Show in paper

“The proof constructs operators which remember a prime between two of its uses.” · Show in paper

Rank-based global birth distinctness restoration: restoring global distinctness of performed prime births in the memory expansion by decomposing the distinctness indicator over collections of equality constraints grouped by their rank (number of independent equality constraints), bounding high ranks absolutely and low ranks via circle orthogonality and signed contraction.

Source excerpts

“Global distinctness is then restored by grouping equalities between newly drawn prime labels according to their rank, the number of independent equality constraints. High ranks supply many small point-probability factors, whereas low ranks alter only a vanishing fraction of the signed edge contractions.” · Show in paper

“A union bound over one repeated prime would not suffice against the absolute cost of a long path; we instead separate equality constraints by their rank.” · Show in paper

Two-sided marked correlation via dilation-graph transference and bilinear comparison multipliers: applying dilation-graph operator transference and residual ideal-kernel theorems to a centered long-prime divisor statistic, then bounding the resulting comparison correlations using bilinear Fourier multipliers, local Mellin energy, and a truncated divisor model.

Source excerpts

“The second estimate, Theorem 6.1, uses marks on both endpoints. We apply the general graph inputs of S to a centered long-prime divisor statistic and prove the required new endpoint Fourier estimates. A small mark restricts difficult frequencies to a sparse set; a long-prime polynomial controls the positive tuple term there, and a finite divisor model controls the constant term.” · Show in paper

Presieving, slot replacement, and disjoint marking array averaging: using a block sieve to presieve u+1u+1, replacing composite prime slots by rough weights via the one-sided Type II estimate, applying the two-sided correlation estimate, and removing the small-band marks by averaging over finitely many disjoint marking arrays.

Source excerpts

“Presieving u + 1 leaves both primes and composites in the average of W(u)(fx(u) − Cx). We will show that the composite contribution is negligible in two stages. The one-sided estimate first replaces its prime factors by rough variables; additional divisor marks then allow the two-sided estimate to bound the resulting centered sum. This isolates a prime average carrying W. Finally, a mean-square bound for averages of disjoint marking arrays removes that weight.” · Show in paper

Size-biased sampling and factorial-measure passage to the Poisson–Dirichlet law: converting interior factorial-moment asymptotics into probability densities on the open simplex via sequential size-biased sampling without replacement, controlling boundary escape via total mass one, and passing to ranked factors via sorting and dyadic decomposition.

Source excerpts

“Finally, sequential size-biased sampling turns the interior factorial measures into probability densities of total mass one. This excludes escape to the boundary and yields the stick fragments in (1.1). The expected undrawn mass controls sorting, and a finite dyadic decomposition gives all real upper endpoints.” · Show in paper