← All papers
First page of Class polynomial obstructions to canonical congruences for the smallest parts function

Class polynomial obstructions to canonical congruences for the smallest parts function

Larry Rolen

math.NT Oct 19, 2026 · v1
The paper's proofs were formally verified in Lean by AxiomProver, with the formalization files posted in a public GitHub repository.
We resolve an open conjecture by proving that the only "canonical” Ramanujan congruence primes for the Andrews' smallest parts function, $\operatorname{spt}(n)$, are $\ell= 5, 7, 13$. Building on work of Ahlgren and Kim, we show that any such congruence forces square-index coefficients of certain weakly holomorphic forms of weight $1/2$ to vanish modulo the congruence prime. A normalized Borcherds product converts this into the vanishing of the derivative of a quotient of two oriented CM polynomials on $X_0(6)$. A quaternionic volume estimate in an Eichler order of level $6$ prevents that vanishing when a suitable inert discriminant is small. Treviño's character estimate supplies such a discriminant for every prime at least $10^{11}$. The remaining range is covered by the previous calculations of Ahlgren and Kim. These results have been verified in Lean by AxiomProver.

Andrews proved Ramanujan-type congruences for the smallest parts function spt(n) modulo 5, 7 and 13. It was a long-standing conjecture that these are the only 'canonical' congruence primes.

Building on Ahlgren and Kim, a canonical congruence is shown to force square-index coefficients of certain weight 1/2 weakly holomorphic forms to vanish modulo ℓ. A normalized Borcherds product turns this into the vanishing of the derivative of a quotient of oriented CM class polynomials on X_0(6). A quaternionic volume estimate in an Eichler order of level 6 rules out that vanishing when a small inert discriminant exists. Treviño's character bound supplies such a discriminant for all primes ℓ ≥ 10^11. The proofs were then verified in Lean by AxiomProver.

The canonical congruence primes for spt(n) are exactly 5, 7 and 13. Primes below 10^11 are covered by the earlier computations of Ahlgren and Kim, and the arguments were formally verified in Lean.