The partition function and elliptic curves
Ken Ono
math.NT
Aug 13, 2025 · v4
math.CO
TL;DR
AxiomProver produced sorry-free Lean 4 formalizations of the paper's two key algebraic identities, which are released in a public repository.
Abstract
The Bruinier-Ono formula expresses the partition number $p(n)$ as a trace of `non-holomorphic' singular moduli of discriminant $Δ_n:=1-24n$ CM points on $X_0(6).$ We interpret this trace through the geometry of CM points. Each nonholomorphic contribution is the value of the weight-two completion $E_2^*$ at a CM point, which is a canonical invariant of the underlying elliptic curve, determined by the diagonal `tangent' of the CM isogeny relation. This turns the trace into a quantity that can be reduced to the supersingular locus that is organized by Deuring-Eichler multiplicities and a Brandt-module pairing. For primes $ \ell\geq 5$ that are nonsplit in $\mathbb{Q}(\sqrt{Δ_n})$, we obtain a supersingular trace formula on $X_0(6)$ over $\overline{\mathbb{F}}_{\ell}$. For the special primes $\ell=5,7,11$, this sheds new light on Ramanujan's classical partition congruences. These primes are special because they are the only ones for which the supersingular locus of $X_0(6)$ lies over $j\in \{0, 1728\}.$ This perspective offers a moduli-theoretic framework for Ramanujan's congruences modulo powers of these primes, organized through elliptic curves. The two new algebraic identities at the heart of this framework, as opposed to the classical results it builds on, were formalized and verified in Lean by AxiomProver.
Problem
The Bruinier–Ono formula expresses the partition number p(n) as a trace of nonholomorphic singular moduli at CM points on X_0(6). The goal is to interpret this trace geometrically and relate it to Ramanujan's partition congruences.
Approach
Each nonholomorphic contribution is identified as the value of the completed Eisenstein series E_2^* at a CM point, which is expressed through the diagonal tangent of the CM isogeny relation. The trace is reduced to the supersingular locus using Deuring–Eichler multiplicities and a Brandt-module pairing. The two new algebraic identities underlying the argument were treated as field identities over complex ground data. These are the Maass/Serre operator splitting and the symmetry reduction of the CM tangent, and both were formalized and verified in Lean 4.28.0 by AxiomProver.
Results
A supersingular trace formula on X_0(6) is obtained for primes ℓ ≥ 5 that are nonsplit in Q(√Δ_n). The primes 5, 7 and 11 are shown to be exactly those for which the supersingular locus lies over j ∈ {0, 1728}, which gives a moduli-theoretic framework for Ramanujan's congruences. The Lean proofs of the two identities are sorry-free and released at github.com/AxiomMath/PartitionElliptic.