← All papers
First page of Mean Value for Random Ideal Lattices

Mean Value for Random Ideal Lattices

Nihar Gargava, Maryna Viazovska

math.NT Nov 22, 2024 · v3
Theorem 7, the mean-value bound for ideal lattices, was formally verified in Lean with AI assistance and published to the Palomar Registry.
We investigate the average number of lattice points within a ball for the $n$th cyclotomic number field, where the lattice is chosen at random from the set of unit determinant ideal lattices of the field. We show that this average is nearly identical to the average number of lattice points in a ball among all unit determinant random lattices of the same dimension. To establish this result, we apply the Hecke integration formula and subconvexity bounds on Dedekind zeta functions of cyclotomic fields. The symmetries arising from the roots of unity in an ideal lattice allow us to prove the existence of ideal lattice packings of dimension $\varphi(n)$ and density $n\cdot 2^{-\varphi(n)}(1+o(1))$ as $n$ goes to infinity.

Siegel's mean value theorem gives the expected number of lattice points in a ball for Haar-random unimodular lattices. The question is whether an analogous result holds for random unit-covolume ideal lattices of cyclotomic number fields, drawn from the Arakelov class group.

The authors apply the Hecke integration formula to express the average over the Arakelov class group in terms of Dedekind zeta functions. They bound the error term with subconvexity estimates for cyclotomic Dedekind zeta functions, using the Petrow-Young bounds for Dirichlet L-functions. Roots-of-unity symmetries are then used to obtain packings. The main theorem (Theorem 7) was later formally verified in Lean with AI assistance, which led to corrections of minor details.

The expected lattice point count is 1+V plus an error term that decays rapidly in the degree of the field. This yields ideal lattice packings of dimension φ(n) with density n·2^{-φ(n)}(1+o(1)). The Lean verification is publicly available in the Palomar Registry.