A Gaussian integral formula for the Hermite polynomials: Combinatorics, Asymptotics and Applications

Mihai Nica, Janosch OrtmannarXiv 2025 · DOI · arXiv

Abstract

The Hermite polynomials are ubiquitous but can be difficult to work with due to their unwieldy definition in terms of derivatives. To remedy this, we showcase an underappreciated Gaussian integral formula for the Hermite polynomials, which is especially useful for generalizing to multivariable Hermite polynomials. Taking this as our definition, we prove many useful consequences, including:

  1. Combinatorial interpretations for the Hermite polynomials, including a proof of orthogonality.
  2. A more elementary proof of Plancherel--Rotach asymptotics that does not involve residues.
  3. Limit theorems for GUE random matrices and Dyson's Brownian motion, including bulk convergence to the semi-circle law and edge convergence to the Airy limit/Tracy-Widom law.
  4. An analysis of a phase transition in the spiked GUE random matrix as the top eigenvalue goes from well-separated to attached to the bulk, analogous to the BBP phase transition.
  5. Elementary derivations of Edgeworth expansions and multivariable Edgeworth expansions. This article is primarily expository and features many illustrative figures.

Introduction: Beyond the Derivative Definition

Hermite polynomials are fundamental tools across physics, statistics, and probability. Traditionally, they are defined via the Rodrigues' formula—involving $n$-th order derivatives—or through recurrence relations. While functional, these classical definitions often obscure the intuitive probabilistic nature of the polynomials. This "derivative-first" approach makes proofs for properties like orthogonality or asymptotic behavior mathematically heavy, frequently requiring complex analysis or residue calculus. Furthermore, generalizing these definitions to multiple variables becomes notationally cumbersome.

The Gaussian Expectation Formula

This paper champions an alternative, underappreciated starting point: the Gaussian expectation formula. We define the probabilist's Hermite polynomial as: $$He_n(x) = \mathbb{E}_{Z \sim \mathcal{N}(0,1)}[(x + \imath Z)^n]$$ where $\imath$ is the imaginary unit. This definition offers a powerful "imaginary shift" intuition: a Hermite polynomial is simply what results when a standard power $x^n$ is perturbed by an imaginary Gaussian noise component. This framework generalizes effortlessly to $d$ dimensions by replacing the scalar Gaussian $Z$ with a multivariable Gaussian vector $\vec{Z}$ and a covariance matrix $V$.

Combinatorial Interpretations

By treating Hermite polynomials as expectations, we can leverage Wick’s Theorem (Isserlis’ Theorem) to interpret their terms through the lens of "perfect matchings" and pairings. This combinatorial perspective allows for elegant, bijection-based proofs of properties like orthogonality—results that typically require heavy calculus.

Combinatorial expansion of the fourth Hermite polynomial $\tilde{H}_4(x,t)$ using the Gaussian expectation formula, illustrating how terms emerge from pairings.

As shown in the figure above, the expectation of the shifted power $(x + \imath \sqrt{t} Z)^4$ naturally decomposes into terms representing different pairing configurations. For instance, the $-6tx^2$ term arises from the six ways to pair two of the four available factors, with each pair contributing a factor of $-t$ (the variance of the imaginary shift).

Simplifying Asymptotics and Random Matrix Theory

The integral-first approach significantly streamlines the study of large-scale behavior ($n \to \infty$). Instead of employing residues and Cauchy’s integral formula, we apply Laplace’s method (steepest descent) directly to the Gaussian integral. This yields cleaner proofs for the Plancherel–Rotach asymptotics across three distinct regimes: the oscillatory bulk, the monotone tails, and the Airy-type edge.

These asymptotic results are critical for analyzing the eigenvalues of Gaussian Unitary Ensemble (GUE) random matrices. The paper provides an elementary derivation of the BBP phase transition in spiked random matrices. This transition, where the top eigenvalue moves from a Gaussian distribution (when well-separated) to a Tracy-Widom distribution (when attached to the bulk), can be visualized as a physical heuristic involving a stretching rubber band reaching a boundary.

Applications and Future Directions

Beyond theoretical physics, these multivariable formulas are increasingly relevant in modern AI research. Specifically, they help model how signals propagate through the layers of deep neural networks via Edgeworth expansions.

While this work focuses on the elegance of the $He_n$ form, the authors suggest that this integral-first methodology could be extended to other families of orthogonal polynomials, such as Laguerre or Jacobi polynomials. Such extensions may reveal similar combinatorial and asymptotic simplifications, further unifying the study of orthogonal systems through the lens of probability.

Summary updated

Abstract

The Hermite polynomials are ubiquitous but can be difficult to work with due to their unwieldy definition in terms of derivatives. To remedy this, we showcase an underappreciated Gaussian integral formula for the Hermite polynomials, which is especially useful for generalizing to multivariable Hermite polynomials. Taking this as our definition, we prove many useful consequences, including: 1. Combinatorial interpretations for the Hermite polynomials, including a proof of orthogonality. 2. A more elementary proof of Plancherel--Rotach asymptotics that does not involve residues. 3. Limit theorems for GUE random matrices and Dyson's Brownian motion, including bulk convergence to the semi-circle law and edge convergence to the Airy limit/Tracy-Widom law. 4. An analysis of a phase transition in the spiked GUE random matrix as the top eigenvalue goes from well-separated to attached to the bulk, analogous to the BBP phase transition. 5. Elementary derivations of Edgeworth expansions and multivariable Edgeworth expansions. This article is primarily expository and features many illustrative figures.
BibTeX
@misc{nica2025gaussian,
  author = {Mihai Nica and Janosch Ortmann},
  title = {{A Gaussian integral formula for the Hermite polynomials: Combinatorics, Asymptotics and Applications}},
  year = {2025},
  eprint = {2508.13910},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2508.13910},
  howpublished = {arXiv},
}

Loading related content...

Figure