Matrix Correlators as Discrete Volumes of Moduli Space I: Recursion Relations, the BMN-limit and DSSYK
A. Giacchetto, P. Maity, E. A. Mazenc · 2025
Read Matrix Correlators as Discrete Volumes of Moduli Space I: Recursion Relations, the BMN-limit and DSSYK by A. Giacchetto, P. Maity, E. A. Mazenc (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-2510-17728.txt.
Introduction
Alessandro Giacchetto, Pronobesh Maity, and Edward A. Mazenc link matrix models to surface geometry. Certain averages in one-matrix models define discrete volumes of moduli space. Moduli space collects every possible shape of a curved surface. Their work extends a known link between random matrices and gravity.
The team proves these volumes follow a recursion, a rule that builds big cases from smaller ones. The recursion echoes a famous one found by Mirzakhani. In a large-trace limit, their discrete rule becomes a smooth one. They also settle a conjecture about a related matrix integral. That integral gives a discrete analog of classic surface volumes.
Catalog record
| CREATOR | A. Giacchetto, P. Maity, E. A. Mazenc |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 139 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2510-17728
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Matrix Correlators as Discrete Volumes of Moduli Space I: Recursion Relations, the BMN-limit and DSSYK A. Giacchetto, P. Maity, E. A. Mazenc Abstract We show certain correlators in generic one-matrix models define a notion of “discrete” volumes of the moduli space of Riemann surfaces, generalizing the connection between random matrices and JT gravity. We prove they obey a discrete, Mirzakhani-like recursion relation. Their fundamental discreteness crucially relies upon studying these matrix integrals away from the usual double-scaling limit. In a BMN-like limit of large traces, this recursion universally goes over to a continuous one, and the correlators asymptote to the volumes of Kontsevich. Finally, we demonstrate that the ETH matrix integral for DSSYK furnishes a discrete, q q -analog of the Weil–Petersson volumes, thereby proving a conjecture due to K. Okuyama. 1 Overview & summary of results This first installment establishes three main results: A) a discrete, manifestly geometric recursion relation for correlators of pruned traces in generic one-cut matrix models, B) this recursion becomes universal and continuous in a BMN-like limit of large powers of the matrices, and C) a proof that the DSSYK matrix model computes a discrete q -analog of the Weil–Petersson volumes, thereby proving a conjecture by K. Okuyama Oku23 . 1.1 A discrete Mirzakhani recursion for pruned correlators Correlation functions of resolvents in large \mathsf{N} matrix models obey topological recursion Eyn04 ; AMM05 ; CEO06 ; EO07 . Essentially, a clever 1/\mathsf{N} expansion of the Schwinger–Dyson equations for the matrix integral ginsparg1993lectures ; DiFrancesco:1993cyw shows that the planar one-point function suffices to determine all correlators to all orders in 1/\mathsf{N} . From an entirely different perspective, certain volumes of the moduli space of Riemann surfaces, such as the Weil–Petersson volumes studied by Mirzakhani Mir07a ; Mir07b , were also found to follow from a recursion relation. This similarity was elucidated by EO07+ and provided the starting point of Saad–Shenker–Stanford (SSS) SSSJTmatrix , recasting JT gravity and its supersymmetric extensions as a matrix integral SW20 ; SaadTasi ; TuriaciLesHouches ; CJohnsonReview ; CliffMaciek ; CliffWPsuper . However, the geometric origins of the recursion kernels appearing in the work of Mirzakhani are somewhat obscure on the matrix model side. N_{g,n} \displaystyle=\sum_{m=2}^{n} B N_{g,n-1} + C \Biggl( N_{g-1,n+1} N_{h,1+J} N_{h^{\prime},1+J^{\prime}} \Biggr) \displaystyle+\sum_{\begin{subarray}{c}h,h^{\prime}\\ J,J^{\prime}\end{subarray}} Figure 1: A Pictorial Representation of the Discrete Recursion: The pruned correlators \langle\prod_{i=1}^{n}\frac{1}{b_{i}}\!:\mathrel{\Tr{M^{b_{i}}}}:\rangle_{g,\mathrm{c}} define a discrete notion of volume of the moduli space of Riemann surfaces, denoted N_{g,n}(b_{1},\ldots,b_{n}) . They satisfy a discrete recursion relation that parallels Mirzakhani’s formula for the Weil–Petersson volumes. The recursion kernels B and C can be computed directly from the matrix model potential. In this work, we derive a recursion relation directly for the connected correlators of pruned traces in a generic one-cut matrix model: N_{g,n}(b_{1},\ldots,b_{n})\coloneqq\left\langle\prod_{i=1}^{n}\frac{1}{b_{i}}:\mathrel{\Tr{M^{b_{i}}}}:\right\rangle_{\!\!g,\mathrm{c}},\qquad\qquad b_{i}\in\mathbb{Z}_{+}. (1) Pruning can be viewed as a planar analog of normal ordering: all planar one-point functions vanish, though not necessarily the higher-genus ones. In a Feynman diagram expansion, this normal ordering corresponds to deleting all petals from the diagrams. Pruned correlators were first introduced by Norbury and Scott NS13 in the abstract setting of topological recursion, independently of any matrix model realization. It is the correlator of pruned traces, and not the standard ones, which satisfies a Mirzakhani-like recursion relation. In the specific context of the double-scaled JT gravity matrix integral, they play the same role as the Weil–Petersson volumes. Our work makes a more general statement: it goes beyond the specific choice of matrix model and, most importantly, breaks away from the usual double-scaling limit. While our recursion ultimately follows from the Eynard–Orantin topological recursion for matrix model correlators EO07 , it possesses two fundamental features that distinguish it. First, it is inherently discrete, replacing the traditional residue calculus with sums over the powers of the matrices appearing in the traces. Second, it makes manifest the geometric content of the recursion, in direct parallel with the Mirzakhani recursion satisfied by the Weil–Petersson volumes, as illustrated in figure 1 . The first main result of this paper is: Theorem A . For 2g-2+n>1 , the pruned correlators in a generic one-cut matrix model satisfy the recursion relation N_{g,n}(b_{1},\dots,b_{n})=\sum_{m=2}^{n}\sum_{\beta>0}\beta\,B(b_{1},b_{m},\beta)N_{g,n-1}(\beta,b_{2},\dots,\widehat{b_{m}},\dots,b_{n})\\ +\frac{1}{2}\sum_{\beta,\beta^{\prime}>0}\beta\beta^{\prime}\,C(b_{1},\beta,\beta^{\prime})\Bigg(N_{g-1,n+1}(\beta,\beta^{\prime},b_{2},\dots,b_{n})\\ +\sum_{\begin{subarray}{c}h+h^{\prime}=g\\ J\sqcup J^{\prime}=\{2,\dots,n\}\end{subarray}}^{\textup{stable}}N_{h,1+|J|}(\beta,b_{J})N_{h^{\prime},1+|J^{\prime}|}(\beta^{\prime},b_{J^{\prime}})\Bigg), (2) where a caret as in \widehat{b_{m}} denotes omission. The recursion kernels B and C can be expressed in terms of a single building-block function H : \displaystyle B(b,b^{\prime},\beta) \displaystyle\coloneqq\frac{1}{2b}\Bigl(H(b+b^{\prime}-\beta)-H(-b-b^{\prime}-\beta) (3) \displaystyle+H(b-b^{\prime}-\beta)-H(-b+b^{\prime}-\beta)\Bigr), \displaystyle C(b,\beta,\beta^{\prime}) \displaystyle\coloneqq\frac{1}{b}\Bigl(H(b-\beta-\beta^{\prime})-H(-b-\beta-\beta^{\prime})\Bigr), which in turn is explicitly determined from the matrix model potential—see equation 51 . Together with the genus- 0 , 3 -point correlator N_{0,3} and genus- 1 , 1 -point correlator N_{1,1} , the recursion uniquely determines all correlators. This geometric version of topological recursion therefore suggests that the pruned correlators can be viewed as providing a discrete notion of volumes of the moduli space of Riemann surfaces. In this picture, they would compute a weighted count of Riemann surfaces with integer-length boundaries. The discrete boundary lengths correspond to the powers appearing in the traces of the dual matrix model correlators. We will make this picture precise using a well-established bijection between metrized ribbon graphs and points on the decorated moduli space of Riemann surfaces Str84 ; Kon92 . In a nutshell, we expand the correlators in terms of Feynman diagrams, and map each diagram to a point on the moduli space. This notion of discreteness exists at each order in 1/\mathsf{N} , and we argue it persists to any finite order in perturbation theory in the interaction coupling—see section 3 for further details. 1.2 The BMN-like limit and its Airy universality Our recursion relation also reveals the existence of a particularly interesting and universal limit of pruned correlators. It reflects the well-known Airy universality that governs the square-root vanishing of the matrix eigenvalue distribution near its endpoints TW94 ; Eyn16 ; BrezinHikamiAiry ; BrezinZee . In essence, this limit consists in taking the powers of the matrices inside each trace to be very large. This closely parallels the Berenstein–Maldacena–Nastase (BMN) limit in AdS/CFT BMN02 . Geometrically, it corresponds to sending the boundaries of the dual Riemann surfaces to infinity. Using the recursion relation, we will demonstrate that for any one-cut matrix model, the pruned correlators converge in this limit to the Kontsevich volumes, independently of the potential. Theorem B . For 2g-2+n>0 , the pruned correlators in a generic even 1 1 1 For simplicity, we have stated the result for an even potential. In the absence of definite parity, there are two separate scaling constants. one-cut matrix model satisfy \lim_{t\rightarrow 0^{+}}\ c^{2g-2+n}\,t^{2(3g-3+n)}\left\langle\prod_{i=1}^{n}\frac{1}{L_{i}/t}:\mathrel{\Tr{M^{L_{i}/t}}}:\right\rangle_{\!\!g,\mathrm{c}}=2\cdot V^{\textup{Kon}}_{g,n}(L_{1},\ldots,L_{n}) (4) whenever the sum of L_{i}/t\in\mathbb{Z}_{+} is even. Here V^{\textup{Kon}}_{g,n} are the Kontsevich volumes of the moduli space of Riemann surfaces and c is a scaling constant that depends on the matrix model spectral curve. In the BMN-like limit, the discrete recursion universally goes over to a continuous one, known to govern the celebrated Kontsevich volumes BCSW12 ; ABCGLW . For the precise definition of these volumes and their expression as integrals over \overline{\mathcal{M}}_{g,n} , see section 2 . The overall factor of 2 originates from the two endpoints of the large \mathsf{N} eigenvalue distribution, which characterize the one-cut phase of the underlying matrix model. To heuristically motivate this result from a diagrammatic point of view, as the powers b_{i} of