← All problems

GL(3)×GL(2) shifted convolution problem

Let π\pi be a Hecke--Maass cusp form on SL(3,Z)\mathrm{SL}(3,\mathbb{Z}), let a(n)a(n) be a GL(2)\mathrm{GL}(2)-type coefficient sequence, and let h≠0h\neq 0 be an integer. For X≥1X\geq 1, study the shifted convolution sum

Sπ,a(X,h)=∑X<n≤2Xλπ(1,n) a(n+h).S_{\pi,a}(X,h)=\sum_{X<n\leq 2X}\lambda_\pi(1,n)\,a(n+h).

The problem is to obtain nontrivial power-saving estimates for the full family of such sums, namely bounds of the form

Sπ,a(X,h)≪π,a,εX1−δ+εS_{\pi,a}(X,h)\ll_{\pi,a,\varepsilon}X^{1-\delta+\varepsilon}

for some δ>0\delta>0, with appropriate uniformity in the shift and the parameters. Important cases include a(n)=d(n)a(n)=d(n) and a(n)=r(n,Q)a(n)=r(n,\mathcal Q), where Q\mathcal Q is a positive definite integral binary quadratic form, but the full GL(3)×GL(2)\mathrm{GL}(3)\times\mathrm{GL}(2) family is not resolved.

References

Source: arXiv

Progress summary

Refreshed
Open · progress reported

A new paper gives stronger estimates for some cases, but the full problem remains open.

The problem concerns shifted correlations between GL(3)\mathrm{GL}(3) and GL(2)\mathrm{GL}(2) automorphic coefficients. Existing results settle important special configurations, but not the entire family.

Known results

  • Xi obtained a fixed-shift bound of X1−1/22+εX^{1-1/22+\varepsilon} for the general coefficient configuration.
  • Weighted averages over shifts yield bounds such as X3/4+ε∥a∥2/∣H∣X^{3/4+\varepsilon}\|a\|_2/\sqrt{|\mathcal H|}.
  • The divisor-function configuration was reported with a uniform bound of order N41/42+εN^{41/42+\varepsilon} and applications to twisted LL-values.

October 2026 stronger bounds

Guangwei Hu and Guangshi Lü improve a divisor-function estimate and prove the corresponding estimate for representations by a positive-definite binary quadratic form. The advance provides stronger power savings for selected configurations, but does not resolve the full shifted-convolution problem.

Current status (as of October 2026): Selected GL(3)×GL(2)\mathrm{GL}(3)\times\mathrm{GL}(2) configurations have stronger estimates, while the full problem remains open.

Sources

Solutions 0

No solutions have been posted yet.