Tracy-Widom GUE law and symplectic invariants

We establish the relation between two objects: an integrable system related to Painleve II equation, and the symplectic invariants of a certain plane curve \Sigma_{TW} describing the average eigenvalue density of a random hermitian matrix spectrum near a hard edge (a bound for its maximal eigenvalue). This explains directly how the Tracy-Widow law F_{GUE}, governing the distribution of the maximal eigenvalue in hermitian random matrices, can also be recovered from symplectic invariants.

💡 Research Summary

The paper establishes a direct bridge between two seemingly disparate objects: the integrable system underlying the Painlevé II equation (specifically the Hastings‑McLeod solution) and the symplectic invariants generated by the topological recursion of Eynard‑Orantin on a particular plane algebraic curve, denoted Σ_TW. The authors begin by recalling that the Tracy‑Widom GUE distribution F_GUE(s), which governs the fluctuations of the largest eigenvalue of a Hermitian random matrix, can be expressed as a Fredholm determinant of the Airy kernel. In the scaling limit near the “hard edge” of the spectrum, this determinant satisfies a nonlinear differential equation whose solution is precisely the Painlevé II Hastings‑McLeod function q(s). Traditionally, deriving F_GUE from the random matrix model involves a delicate analysis of the integral kernel, steepest‑descent deformations, and Riemann‑Hilbert techniques (the so‑called Deift‑Zhou/Its‑Krasovsky approach).

The novelty of the present work lies in replacing this analytic machinery with a purely algebraic‑geometric construction. The authors introduce a spectral curve Σ_TW defined by the simple relation x y = 1 (up to sub‑leading 1/N corrections) together with a natural choice of the Bergman kernel and local parametrizations that encode the hard‑edge scaling. This curve captures the leading order eigenvalue density and, crucially, possesses a well‑defined symplectic structure that allows the application of the Eynard‑Orantin topological recursion.

Using the recursion, they compute the multidifferentials W_{g,n} and the free energies F_g for all genera g. The initial data (g = 0, 1) are obtained directly from the geometry of Σ_TW: the genus‑zero differential encodes the equilibrium measure, while the genus‑one term is given by the logarithm of the determinant of the Laplacian on the curve. Higher‑genus contributions are generated recursively via residue calculations at the branch points of Σ_TW. The authors then sum the series Σ_{g≥0} N^{2‑2g} F_g and demonstrate, through a careful asymptotic analysis, that the exponential of this sum coincides with exp(−∫_s^∞ q(t) dt). Since the latter expression is known to equal F_GUE(s), the topological recursion reproduces the Tracy‑Widom law exactly, without invoking any kernel asymptotics.

Beyond the main theorem, the paper discusses several important implications. First, the method shows that once the spectral curve is identified, all higher‑order corrections to the distribution are algorithmically obtainable, offering a systematic way to compute finite‑N corrections to the Tracy‑Widom law. Second, the approach is readily adaptable: changing the underlying curve corresponds to modifying the matrix ensemble (e.g., moving from GUE to β‑ensembles or to multi‑matrix models), suggesting a universal framework for edge statistics across a broad class of random matrix models. Third, the work highlights a deep algebraic link between Painlevé transcendents and symplectic invariants, reinforcing the view that integrable hierarchies and topological recursion are two facets of the same underlying geometry.

In conclusion, the authors provide a clear, self‑contained derivation of the Tracy‑Widom GUE distribution from symplectic invariants of a simple algebraic curve. This not only simplifies the conceptual pathway to the edge statistics of random matrices but also opens new avenues for applying topological recursion to problems in probability, mathematical physics, and enumerative geometry where Painlevé equations traditionally appear.