Spectral Curves with Complex Multiplication in Hermitian Matrix Models

Ali Nassar · 2025

Read Spectral Curves with Complex Multiplication in Hermitian Matrix Models by Ali Nassar (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2509-16997.txt.

Introduction

Ali Nassar, working alone, finds number theory inside random matrix models in a 2025 paper. A Hermitian matrix model in a two-cut phase yields a spectral curve of genus one. Such a curve is an elliptic curve. Nassar works out a closed form for its key invariant. The form depends on a coupling in a quartic potential.

At specific coupling values, the curve gains extra symmetry, called complex multiplication. That trait means its endomorphism ring is larger than the integers. The result ties number-theoretic structure directly to spectral data of random matrices. Bridges like this often enrich both sides, giving each field new questions and tools.

Catalog record

CREATORAli Nassar
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE30 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2509-16997

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Spectral Curves with Complex Multiplication in Hermitian Matrix Models Ali Nassar Abstract We show that elliptic curves with complex multiplication (CM) naturally emerge in the spectral geometry of Hermitian one-matrix models in the two-cut phase. Focusing on a symmetric quartic potential, we derive the corresponding genus-one spectral curve and compute its modular j j -invariant in closed form as a function of the quartic coupling g g . We identify specific values of g g for which the elliptic curve exhibits C ​ M CM , i.e., its endomorphism ring is larger than ℤ \mathbb{Z} . This establishes a direct connection between number-theoretic structures and the spectral data of random matrix ensembles. 1 Introduction Random matrix models provide a unifying framework for a wide range of problems in physics and mathematics, from quantum gravity and string theory to number theory, statistical mechanics, and quantum chaos Mehta2004 ; GinspargMoore1993 ; Bourgade2013 ; Haake2001 ; DiFrancesco1995 ; Klebanov:1991qa . Random matrix ensembles of large dimension were first introduced by Wigner in 1951 Wigner51 to study the spectral properties of complex systems exhibiting chaotic behavior. In this framework, the Hamiltonian of a chaotic system is modeled as a large matrix with random entries. One of the most remarkable aspects of matrix models is the emergence of a spectral curve, capturing the large- N behavior of eigenvalues and encoding the analytic structure of the theory in geometric terms Brezin:1977sv ; EynardOrantin2007 ; Eynard1992 ; Eynard2004 (see Eynard:2015aea for a detailed account). In particular, multi-cut solutions of matrix models give rise to higher-genus spectral curves, with the two-cut phase corresponding to a genus-one elliptic curve. In this work, we explore how spectral curves can exhibit complex multiplication (CM), a deep arithmetic property of elliptic curves ParshinShafarevich1992 ; Cox1989 ; ShimuraTaniyama1961 ; Lang1991 (For a comprehensive introduction, as well as applications of number - theoretic ideas in string theory, see Moore:1998pn ). For the Hermitian matrix model with quartic potential V(x)=-\tfrac{1}{2}x^{2}+\tfrac{g}{4}x^{4}, (1) we focus on the range of g that produces a two-cut phase, corresponding to a genus-one spectral curve. We explicitly determine this curve and compute its j(g) -invariant as a function of g . In particular, we find five admissible values of g at which the spectral curve acquires complex multiplication, meaning its endomorphism ring is larger than \mathbb{Z} . At these special values of g , the automorphism group of the curve is enhanced. For instance, for j=1728 the usual group of symmetries of the elliptic curve which is of order 2 is promoted to a symmetry group of order 4 , with similar enhancements occurring at the other CM points. Elliptic curves with complex multiplication have been studied by various authors in different contexts. For example, in Gukov:2002nw ; Hosono:2002yb the authors formulated a criterion for the rationality of c=2 conformal field theories on an elliptic curve E , with the essential requirement that E possess complex multiplication. The organization of this paper is as follows. In Section 2 , we review the basics of Hermitian matrix models and their spectral curves in the large - N limit. Section 3 develops the construction of the spectral curve and introduces its genus - one form in the two - cut phase. In Section 3.1 , we provide a concise exposition of elliptic curves, focusing on the algebraic formulation needed for our analysis, while Section 3.2 discusses the notion of complex multiplication. Section 4 contains the main results: we compute the j -invariant of the spectral curve in the quartic matrix model and identify the special values of the coupling constant for which complex multiplication occurs. Finally, Section 5 summarizes our findings and outlines possible directions for future research. 2 Hermitian Matrix Models Random matrix theory studies ensembles of matrices whose entries are drawn from a specified probability distribution. Here we will consider the Gaussian Unitary Ensemble (GUE) of Hermitian N\times N matrices M . The matrix elements M_{ij} are chosen according to the following probability distribution: P(M_{ii})=\frac{1}{\sqrt{\pi}}e^{-\frac{1}{2}M_{ii}^{2}},\qquad P(M_{ij})=\frac{1}{\pi}e^{-|M_{ij}|},\quad i<j. (2) The probability of choosing a matrix M is therefore given by P(M)=\prod_{i=1}^{N}P(M_{ii})\prod_{1\leq i<j\leq N}^{N}P(M_{ij})=A_{N}\prod_{i,j=1}^{N}e^{-\frac{1}{2}|M_{ii}|^{2}}=A_{N}e^{-\frac{1}{2}\text{Tr}(M^{2})}, (3) where A_{N} is a constant. The partition function is Z=\int_{\mathcal{H}_{N}}dMe^{-\frac{1}{2}\text{Tr}(M^{2})}, (4) where \mathcal{H}_{N} is the space of N\times N Hermitian matrices. The Lebesgue measure dM on the space \mathcal{H}_{N} is given by: dM=\prod_{i=1}^{N}dM_{ii}\prod_{1\leq i<j\leq N}d\,\Re M_{ij}\,d\,\Im M_{ij}, (5) where \Re M_{ij}\,d\,\Im M_{ij} denote the real and imaginary parts of M_{ij} , respectively. The measure dM is invariant under unitary conjugation. We can generalize the preceding construction by replacing the quadratic term M^{2}/2 with a general potential V(M) , where V is a polynomial (or, more generally, analytic function). The Gaussian matrix model, characterized by the quadratic potential V(M)=M^{2}/2 , is the simplest example in random matrix theory. It serves as a starting point for understanding the universal features of eigenvalue statistics, such as Wigner’s semicircle law Wigner:1967qdh . Moreover, it admits a complete analytic solution and establishes connections to various physical systems, such as quantum chaos, nuclear spectra, and statistical mechanics . However, it corresponds to a single-cut solution in the large- N limit and fails to capture the richer phase structure or critical behavior present in more complex systems. To explore phenomena such as multi-cut solutions, phase transitions, and connections to two-dimensional quantum gravity or string theory, it is necessary to go beyond the Gaussian potential and consider higher degree polynomials, for example, V(M)=\sum_{k}\frac{g_{k}}{k}M^{k}, (6) where g_{k} are coupling constants. These deformations introduce new critical points and allow for the realization of multiple eigenvalue supports (multi-cut phases) Brezin:1977sv ; Eynard:2003zu . Such generalizations are crucial for accessing non-trivial universality classes, studying double-scaling limits, and constructing a bridge between matrix models and continuum theories of random surfaces Brezin:1990rb ; Douglas:1989ve ; Gross:1989ni ; Gross:1989aw . We consider the partition function Z_{N}=\int_{\mathcal{H}_{N}}dMe^{-N\text{Tr}V(M)}, (7) where we have introduced an explicit factor of N in the exponent of the matrix integral. This normalization is motivated by the large- N limit, where N plays a role analogous to 1/\hbar in quantum mechanics or inverse temperature in statistical mechanics. This normalization also guarantees that the eigenvalues of M remain of order one and do not scale with N in the large- N limit. To analyze the partition function ( 7 ), we use the fact that M can be diagonalized by a unitary transformation M=U\Lambda U^{\dagger} , with \Lambda=\mathrm{diag}(\lambda_{1},\dots,\lambda_{N})) . The measure becomes dM=\prod_{i=1}^{N}d\lambda_{i}\prod_{1\leq m<n\leq N}(\lambda_{m}-\lambda_{n})^{2}\,d\mu(U), (8) where d\mu(U) is the Haar measure on the unitary group U(N) . The eigenvalue representation of the partition function is Z_{N}=C_{N}\int_{\mathbb{R}^{N}}\prod_{i=1}^{N}d\lambda_{i}\,\prod_{1\leq i<j\leq N}(\lambda_{i}-\lambda_{j})^{2}\,e^{-N\sum_{i=1}^{N}V(\lambda_{i})}, (9) where C_{N} is a normalization constant that will play no role in the subsequent analysis and will therefore be omitted. We can express Z_{N} in terms of an effective action: Z_{N}=\int_{\mathbb{R}^{N}}\prod_{i=1}^{N}d\lambda_{i}\,e^{-N^{2}S_{\text{eff}}(\lambda_{1},\cdots,\lambda_{N})}, (10) where S_{\text{eff}}(\lambda_{1},\dots,\lambda_{N})=\frac{1}{N}\sum_{i=1}^{N}V(\lambda_{i})-\frac{1}{N^{2}}\sum_{1\leq i<j\leq N}\log|\lambda_{i}-\lambda_{j}|. (11) In the large- N limit, the partition function is dominated by configurations that minimize the effective action S_{\text{eff}} . This motivates the use of the saddle-point approximation to evaluate the matrix integral. The saddle-point equations are \frac{\partial S_{\text{eff}}}{\partial\lambda_{i}}=0, (12) which leads to V^{\prime}(\lambda_{i})=\frac{2}{N}\sum_{j\neq i}\frac{1}{\lambda_{i}-\lambda_{j}}. (13) To solve the saddle point equations, we define the microscopic eigenvalue density. \rho(\lambda)=\frac{1}{N}\sum_{i=1}^{N}\delta(\lambda-\lambda_{i}),\qquad\int\rho(\lambda)=1. (14) We also define the resolvent: W_{N}(x)=\frac{1}{N}\mathrm{Tr}\frac{1}{x-M}=\frac{1}{N}\sum_{i=1}^{N}\frac{1}{x-\lambda_{i}},\qquad x\in\mathbb{C}. (15) The above equation can be written as: W(x)=\int\frac{\rho(\lambda)}{x-\lambda}d\lambda.

Read the complete text