← All papers
First page of Modular Forms with Only Nonnegative Coefficients

Modular Forms with Only Nonnegative Coefficients

Paul Jenkins, Jeremy Rouse

math.NT Jul 23, 2025 · v3
Lean's linarith tactic is used to compute the exact nonnegativity bound A(k) for weights 36 ≤ k ≤ 88 by checking implications among linear inequalities.
We study modular forms for $\textrm{SL}_2(\mathbb{Z})$ with no negative Fourier coefficients. Let $A(k)$ be the positive integer where if the first $A(k)$ Fourier coefficients of a modular form of weight $k$ for $\textrm{SL}_2(\mathbb{Z})$ are nonnegative, then all of its Fourier coefficients are nonnegative, so that $A(k)$ can be interpreted as a “nonnegativity Sturm bound”. We give upper and lower bounds for $A(k)$, as well as an upper bound on the $n$th Fourier coefficient of any form with no negative Fourier coefficients.

The paper asks how many leading Fourier coefficients of a modular form of weight k for SL2(Z) must be checked as nonnegative to guarantee that all coefficients are nonnegative. This quantity, A(k), is a 'nonnegativity Sturm bound'.

The set of normalized forms with nonnegative coefficients is shown to be a bounded finite convex polytope, which proves that A(k) is well-defined. A lower bound comes from positivity of the Poincaré series coefficients, using Bessel function estimates. An upper bound comes from bounds on cusp form coefficients in terms of the first floor(k/12) coefficients. For small weights, A(k) is computed exactly; for 36 ≤ k ≤ 88, Lean's linarith tactic finds the smallest N whose first N coefficient inequalities imply the remaining ones up to a computed bound B(k).

The paper proves (k-1)^2/(16π^2) < A(k) ≤ k^4(log k + log log k)^2/7316. It also gives an upper bound on the nth coefficient of any form with nonnegative coefficients. Exact values of A(k) are computed for 12 ≤ k ≤ 88.

kL(k)A(k)U(k)
121232
1623128
48141919807
64263371508
884859292773
Lower bound L(k), exact A(k), and upper bound U(k) for selected weights