TL;DR
This paper explores the surprising connection between the distribution of values of L-functions associated with elliptic curves and the statistical properties of random matrices. The authors propose conjectures that link the Fourier coefficients of half-integral weight modular forms to these L-functions, offering a new perspective on their behavior and providing numerical evidence to support their claims.
What Problem Does This Paper Address?
The distribution of values of L-functions, particularly at the center of the critical strip, is a fundamental question in number theory. For integral weight modular forms, the celebrated Sato-Tate conjecture, relating these distributions to those of random unitary matrices, provides a powerful framework. However, the analogous question for half-integral weight modular forms and their associated L-functions has remained largely open. This paper aims to bridge this gap by extending the insights from random matrix theory to the Fourier coefficients of these half-integral weight forms.
Methods
The authors leverage a known connection between L-functions of elliptic curves and Fourier coefficients of half-integral weight modular forms, established by Waldspurger, Shimura, Kohnen, and Zagier. They then combine this with conjectures from random matrix theory, specifically those by Montgomery, Katz, and Sarnak, which relate the distribution of L-function zeros and values to the eigenvalues of random matrices. The core of their approach involves translating conjectures about the distribution of L-function values into conjectures about the distribution of Fourier coefficients. They develop several conjectures concerning moments, logarithms, and the distributions of these coefficients, and then provide numerical evidence to support them.
Datasets
The paper utilizes a database of elliptic curves and their quadratic twists, specifically drawing from a database of ternary quadratic forms provided by Tornaria and Rodriguez-Villegas. Numerical experiments were conducted for 2398 elliptic curves and millions of quadratic twists with |d| < 10⁸.
Key Insights and Ideas
- Bridging the Gap: The paper successfully extends the random matrix theory paradigm from integral weight modular forms to half-integral weight forms, offering a unified perspective.
- L-functions as Random Matrices: The core insight is that the statistical behavior of L-functions associated with elliptic curves, and consequently their related Fourier coefficients, can be modeled by the properties of random matrices.
- Conjectures for Fourier Coefficients: The authors formulate precise conjectures for the distribution of Fourier coefficients, including central limit theorems and moment conjectures, which are analogous to existing conjectures for L-functions.
- Numerical Verification: Extensive numerical computations provide strong support for the proposed conjectures, demonstrating a remarkable agreement between theoretical predictions and empirical data.
Key Results
The paper presents several key conjectures:
- Conjectures 6, 7, and 8: These conjectures propose central limit theorems and moment properties for the Fourier coefficients of half-integral weight modular forms, directly motivated by the random matrix theory conjectures for L-functions.
- Conjectures 9 and 10: These address the behavior of these coefficients under specific conditions and for integer moments, further solidifying the connection to random matrix theory.
- Conjecture 11: This conjecture predicts the frequency of vanishing of L-functions, relating it to a power of X and log X, and is supported by numerical evidence showing relatively flat graphs.
- Conjecture 12 and 13: These conjectures refine the predictions for the distribution of specific ratios related to L-functions, showing improvement when considering additional terms.
The numerical experiments show that the proposed conjectures are remarkably well-supported, with histograms exhibiting expected behaviors (e.g., t⁻¹/² near the origin) and comparisons to Gaussian distributions showing good agreement.
Limitations and Future Work
- Arithmetic Factors: The authors acknowledge that the numerical comparisons often require adjustments for "arithmetic factors" that are not fully understood within the random matrix model. These factors can lead to slow convergence and variability in the observed distributions.
- Torsion Subgroup: Figure 8 suggests a dependence of the constants on the order of the torsion subgroup of elliptic curves, a phenomenon that is not yet fully understood or incorporated into the model.
- Constant c in Conjecture 11: The constant c in Conjecture 11 is noted to be potentially zero, and its arithmetic explanation is still an area of investigation.
- Future Work: The paper suggests that further investigation into the arithmetic factors and the role of the torsion subgroup is needed. Incorporating these elements could lead to a more complete understanding and better predictions. The authors also mention that the numerical evidence supports the conjectures to the accuracy to which the moment polynomials have been computed.
Summary updated