Wiki
Wiki

Research notes on every problem, and a library of the papers behind them. Built from the open erdos repository.

Updated


Source. Theorem 1.3, p. 3, of Victor Y. Wang, Sums of cubes and the Ratios Conjectures, arXiv:2108.03398v2 (19 April 2023), the edition named on the source card. The hypotheses are Conjecture 1.2 (p. 3), Conjecture 1.4 (p. 4) and Conjecture 1.5 (p. 4).

Read depth. Claims checked: the statement, the three conjectures it assumes and the notation of §1 were read clause by clause on pp. 2--4; the proof in §10.2 (p. 55) was read for its structure. Nothing here is independently reviewed.

Statement

Write F0(x,y,z)=x3+y3+z3F_0(x,y,z)=x^3+y^3+z^3 and, for a cubic form FF in mm variables,

NF(X)=#{x∈[−X,X]m:F(x)=0}N_F(X)=\#\{\mathbf x\in[-X,X]^m: F(\mathbf x)=0\}

(display (1.2), p. 2), counting integer points. For a nonzero integer vector c\mathbf c, VcV_{\mathbf c} is the variety F(x)=c⋅x=0F(\mathbf x)=\mathbf c\cdot\mathbf x=0 in Pm−1\mathbb P^{m-1}, Δ(c)\Delta(\mathbf c) is the discriminant polynomial of §2 (display (2.1), p. 8), and S1\mathcal S_1 is the set of c∈Zm\mathbf c\in\mathbb Z^m with Δ(c)≠0\Delta(\mathbf c)\ne0 (display (1.6), p. 3).

Theorem 1.3 (p. 3). Let F=x13+⋯+x63F=x_1^3+\cdots+x_6^3, and assume Conjectures 1.2, 1.4 and 1.5. Then

NF(X)≪X3N_F(X)\ll X^3

(display (1.8)). Moreover, "Let S⊆Z≥0S\subseteq\mathbb Z_{\ge0}. If SS has positive lower density in Z≥0\mathbb Z_{\ge0}, then so does F0(S3)F_0(S^3)." (p. 3).

The three hypotheses, as the paper states them:

  • Conjecture 1.2 (HW2) (p. 3). For each c∈S1\mathbf c\in\mathcal S_1 and each of the Hasse--Weil LL-functions L(s,Vc)L(s,V_{\mathbf c}), L(s,Vc,⊗2)L(s,V_{\mathbf c},\otimes^2), L(s,Vc,Sym2)L(s,V_{\mathbf c},\mathrm{Sym}^2), L(s,Vc,∧2)L(s,V_{\mathbf c},\wedge^2), ζ(s)\zeta(s) and L(s,V)L(s,V) of list (1.7), where VV is the hypersurface F=0F=0: there are an integer d≥1d\ge1 and an isobaric automorphic representation Π\Pi of GLd(AQ)\mathrm{GL}_d(\mathbb A_{\mathbb Q}) whose local factors agree with those of the LL-function at every place, gamma factor included, and L(s,Π)L(s,\Pi) has no zeros in Re⁡(s)>1/2\operatorname{Re}(s)>1/2.
  • Conjecture 1.4 (R2′') (p. 4). For even mm, with Φc,1(s)=1/ζ(2s)L(s+1/2,V)L(s,Vc)\Phi^{\mathbf c,1}(s)=1/\zeta(2s)L(s+1/2,V)L(s,V_{\mathbf c}) (display (1.9)): for every entire ff with f(s)≪f,b(1+∣Im⁡(s)∣)−bf(s)\ll_{f,b}(1+|\operatorname{Im}(s)|)^{-b} on the strip 0≤Re⁡(s)≤20\le\operatorname{Re}(s)\le2 for all b∈Z≥1b\in\mathbb Z_{\ge1}, all reals Z,N≥1Z,N\ge1 with N≤Z3N\le Z^3, and every σ0∈(1,2)\sigma_0\in(1,2),
∑c∈S1∩[−Z,Z]m∣∫(σ0)ds Φc,1(s)f(s)Ns∣2≪FZmNsup⁡0≤σ≤2∫Rdt (1+∣t∣)2∣f(σ+it)∣2\sum_{\mathbf c\in\mathcal S_1\cap[-Z,Z]^m}\Bigl|\int_{(\sigma_0)}ds\,\Phi^{\mathbf c,1}(s)f(s)N^s\Bigr|^2 \ll_F Z^mN\sup_{0\le\sigma\le2}\int_{\mathbb R}dt\,(1+|t|)^2|f(\sigma+it)|^2

(display (1.10)), the contour running from σ0−i∞\sigma_0-i\infty to σ0+i∞\sigma_0+i\infty.

  • Conjecture 1.5 (SFSCp,3_{p,3}) (p. 4). There is a real η0=η0(Δ)>0\eta_0=\eta_0(\Delta)>0 such that for all reals Z,P≥1Z,P\ge1 with P≤Z3/2P\le Z^{3/2},
#{c∈Zm∩[−Z,Z]m: p2∣Δ(c) for some prime p∈[P,2P]}≪ΔZmP−η0.\#\{\mathbf c\in\mathbb Z^m\cap[-Z,Z]^m:\ p^2\mid\Delta(\mathbf c)\text{ for some prime }p\in[P,2P]\}\ll_\Delta Z^mP^{-\eta_0}.

All three are unproved, so both conclusions are conditional. Unconditionally the paper recalls NF(X)≪ϵX7/2/(log⁡X)5/2−ϵN_F(X)\ll_\epsilon X^{7/2}/(\log X)^{5/2-\epsilon} for X≥2X\ge2 when m=6m=6 and FF is diagonal (p. 3, citing Vaughan), and, assuming automorphy and GRH for L(s,Vc)L(s,V_{\mathbf c}) (Conjecture 1.2 for that function), the Hooley--Heath-Brown bound NF(X)≪ϵX3+ϵN_F(X)\ll_\epsilon X^{3+\epsilon} of Theorem 1.1 (p. 3). Theorem 1.3 removes the ϵ\epsilon.

Proof pointer

§10.2, p. 55. Write NF(X)N_F(X) as the sixth moment of the cubic Weyl sum over ∣x∣≤X|x|\le X (display (10.20)), split the sum dyadically and apply Hölder ((10.21)--(10.22)), which reduces (1.8) to bounding a smoothed count NF,w(X)N_{F,w}(X) with a weight ww supported away from the coordinate hyperplanes ((10.23)). The delta-method identity (2.10) splits that count over c∈S0\mathbf c\in\mathcal S_0 and c∈S1\mathbf c\in\mathcal S_1; Theorem 2.5 (p. 10, quoted from Wang's earlier work) bounds the S0\mathcal S_0 part unconditionally, and Theorem 10.5 (p. 52) with ξ=0\xi=0 bounds the S1\mathcal S_1 part, its moment hypotheses Conjectures 9.6 and 9.8 being supplied by Proposition 9.7 and, under Conjecture 1.5, by Proposition 9.9 (p. 46). The density statement follows from (1.8) by a Cauchy--Schwarz argument on the number of representations, which the paper calls standard and does not write out.

Dependencies

Conjectures 1.2, 1.4 and 1.5 as hypotheses; within the paper, Theorem 2.5, Propositions 9.7 and 9.9, and Theorem 10.5.

Bears on

  • Problem 940: with S=Z≥0S=\mathbb Z_{\ge0}, the second conclusion gives the sums x3+y3+z3x^3+y^3+z^3 with x,y,z≥0x,y,z\ge0 positive lower density. Each positive one is a sum of at most three positive cubes, hence of at most three 33-powerful numbers, so under Conjectures 1.2, 1.4 and 1.5 the sums of at most three 33-powerful numbers do not have density 00: a conditional negative answer to the density question at r=3r=3 only. The theorem says nothing about the infinitude question or about r≥4r\ge4, and it settles nothing unconditionally.
  • Problem 325: with S=Z≥0S=\mathbb Z_{\ge0}, the same conclusion gives f3,3(x)≫xf_{3,3}(x)\gg x for large xx, the bound the problem asks for at k=3k=3, under the same three unproved conjectures. The theorem says nothing about k≥4k\ge4.