🔢 mathematics

Inhomogeneous Geometric Last Passage Percolation and the $q$ -deformed Jacobi Unitary Ensemble

This paper investigates inhomogeneous geometric last passage percolation by establishing an exact link to the $q$ -deformed Jacobi unitary ensemble, thereby deriving explicit limit shapes and fluctuation limits governed by Airy or Bessel kernels depending on the inhomogeneity strength.

Original authors: Sung-Soo Byun, Yeong-Gwang Jung, Nick Simm

Published 2026-10-30

📖 4 min read🧠 Deep dive

CC BY 4.0

Original authors: Sung-Soo Byun, Yeong-Gwang Jung, Nick Simm

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine a vast grid of points, like a checkerboard stretching out in every direction. On each square of this board sits a number, a random weight that represents a cost or a delay. Now, picture a traveler who must move from the bottom-left corner to the top-right, stepping only up or to the right. The traveler's goal is to find the path where the sum of these numbers is as large as possible. This is a problem of last passage percolation, a mathematical model used to understand how things grow, spread, or accumulate over time. In a perfectly uniform world, where every square has the same statistical chance of holding a number, mathematicians have long known the rules of the game. They know exactly how the total sum grows and how it wobbles around that average. But the real world is rarely uniform. In nature, growth happens in messy, uneven environments where the rules change from place to place. When the landscape is uneven, the old rules break down, and the traveler's path becomes much harder to predict.

This is the challenge that Sung-Soo Byun, Yeong-Gwang Jung, and Nick Simm set out to solve. They studied a version of this traveler's journey where the random numbers on the grid are not the same everywhere. Instead, the likelihood of finding a large number depends on the location, creating a landscape that is either gently sloping or sharply concentrated near the starting point. The researchers wanted to know two things: what the average shape of the traveler's path would look like over a long distance, and how much the actual path would wiggle away from that average. To answer this, they did something unexpected. Instead of trying to track the traveler directly, they translated the entire problem into a different language: the language of random matrices.

Random matrices are grids of numbers where the entries are chosen at random, but they follow specific statistical laws. These matrices often appear in physics and number theory, and they have a special property: the numbers inside them, called eigenvalues, tend to arrange themselves in a very specific, predictable way. The authors discovered a precise, mathematical bridge connecting the traveler's journey on the uneven grid to the smallest number in a special family of these random matrices. This family is known as the q-deformed Jacobi unitary ensemble. By building this bridge, the researchers could use the powerful, well-understood tools of random matrix theory to solve the difficult problem of the uneven traveler's path.

The study revealed that the behavior of the traveler depends entirely on how strong the unevenness of the environment is. When the variation in the landscape is mild, the traveler's path fluctuates in a way that follows a pattern known as the Airy kernel. This is the same pattern seen in many other growing systems, suggesting a deep universality in how things expand in a slightly messy world. However, when the unevenness is extreme—when the weights are heavily concentrated right at the start—the behavior changes completely. In this strong inhomogeneity regime, the fluctuations follow a different pattern, governed by the Bessel kernel. This is a much rarer phenomenon, usually associated with boundaries or hard walls that stop growth. The researchers showed that the transition between these two behaviors is not a guess or a simulation, but a rigorous mathematical fact derived from their new connection between the two models.

Beyond just the traveler, the paper also uncovered new details about the random matrices themselves. The authors calculated the large-scale distribution of the numbers inside these matrices, finding a new shape that acts as a distorted version of a classic distribution known as the Wachter distribution. They also proved that the smallest number in these matrices behaves in the same two ways as the traveler: it follows the Airy pattern when the environment is mild, and the Bessel pattern when the environment is harsh. This work is significant because it provides one of the very few explicit formulas for how growth behaves in an uneven environment. It confirms that even when the rules of the game change across space, the underlying mathematics remains orderly and predictable, provided one knows how to look at it through the right lens. The findings offer a complete picture of how randomness and structure interact in a non-uniform world, turning a complex, messy problem into a clear, solvable equation.

Technical Summary: Inhomogeneous Geometric Last Passage Percolation and the q-Deformed Jacobi Unitary Ensemble

Problem Statement
This paper investigates directed last passage percolation (LPP) with geometric weights in a spatially inhomogeneous environment. While homogeneous LPP models are well-understood through their duality with random matrix ensembles (specifically the Meixner orthogonal polynomial ensemble, as established by Johansson), explicit dualities for inhomogeneous systems have remained scarce. The authors address the challenge of identifying an appropriate dual model for inhomogeneous geometric LPP that retains sufficient algebraic structure for exact solvability while accommodating spatially varying parameters. The study focuses on a model where the geometric weights $\omega_{ij}$ at lattice site $(i,j)$ follow a distribution determined by parameters $x_i y_j$ , specifically $x_i = \alpha q^{i-1}$ and $y_j = \alpha q^{j-1}$ . The analysis distinguishes between two regimes of inhomogeneity: "weak" (fixed $\alpha \in (0,1)$ ) and "strong" ( $\alpha = e^{-\eta/N}$ ), where $N$ is the macroscopic scale.

Methodology
The core methodological contribution is the establishment of an exact duality between the inhomogeneous geometric LPP and the $q$ -deformed Jacobi unitary ensemble ( $q$ -JUE).

Exact Duality: The authors prove that the distribution of the last passage time $G_{n,m}$ is equivalent to the distribution of the smallest particle (smallest eigenvalue) of a suitably parameterized $q$ -JUE. This is achieved via a logarithmic transformation and a correspondence involving Schur polynomials and the Robinson–Schensted–Knuth (RSK) correspondence.
Asymptotic Analysis of the $q$ -JUE: Since the asymptotic theory for the $q$ $q$ -JUE was largely unexplored, the authors develop a comprehensive asymptotic theory for this ensemble independently. This involves:
- Macroscopic Limit: Deriving the limiting spectral density (the "limit shape") using a constrained logarithmic equilibrium problem (energy minimization with an upper density constraint arising from the $q$ -lattice structure).
- Edge Statistics: Analyzing the local scaling limits of the smallest particle. The analysis distinguishes between the soft edge (where the spectral support is bounded away from zero) and the hard edge (where the support touches zero). This is performed using the Riemann–Hilbert method applied to the associated little $q$ -Jacobi orthogonal polynomials, specifically utilizing contour integral representations and the method of steepest descent.
Transfer of Results: The asymptotic results derived for the $q$ -JUE are then transferred back to the LPP model via the established duality to obtain limit shapes and fluctuation distributions for the passage times.

Key Contributions and Results

Explicit Limit Shapes: The paper derives explicit limit shape functions for the inhomogeneous geometric LPP.
- In the weak inhomogeneity regime, the passage time scales linearly with $N$ ( $O(N)$ ). The limit shape is given by an explicit function $\omega(\gamma, \alpha, \lambda)$ .
- In the strong inhomogeneity regime, the passage time scales as $O(N \log N)$ . The limit shape is independent of the aspect ratio $\gamma$ , reflecting a "locality" phenomenon where the passage time is dominated by weights near the origin.
Fluctuation Limits: The paper establishes the fluctuation limits for the passage time, showing they depend on the strength of the inhomogeneity:
- Weak Inhomogeneity: Fluctuations are governed by the Airy kernel, leading to the GUE Tracy–Widom distribution ( $F_{TW}$ ). This corresponds to the soft-edge regime of the $q$ -JUE.
- Strong Inhomogeneity: Fluctuations are governed by the Bessel kernel, leading to a hard-edge Bessel distribution. This corresponds to the hard-edge regime of the $q$ -JUE.
Theory of the $q$ -JUE:
- Limiting Spectral Density: The authors derive the macroscopic limiting density of the $q$ -JUE, identifying it as a natural $q$ -deformation of the classical Wachter distribution. They identify a "band-to-saturated" phase transition in the density, where a saturated region (attaining the upper constraint $1/(\lambda x)$ ) appears under specific parameter conditions.
- Edge Universality: The paper proves that despite the global $q$ -deformation, the edge fluctuations of the $q$ -JUE are universal. Specifically, the smallest particle exhibits Bessel kernel statistics at the hard edge ( $c_1=0$ ) and Airy kernel statistics at the soft edge ( $c_1>0$ ).
- Degenerations: The results recover known limits: as $q \to 1$ , the $q$ -JUE degenerates to the classical JUE (Wachter distribution) or the Meixner ensemble, consistent with Johansson's homogeneous results.
Exponential LPP: As a corollary, the authors derive results for inhomogeneous exponential LPP by taking a scaling limit of the geometric model. They establish a duality between this model and the classical Jacobi unitary ensemble (JUE), recovering known Bessel-type fluctuations.

Significance and Claims
The paper claims to provide a unified random matrix framework for understanding inhomogeneous LPP. By identifying the $q$ -JUE as the dual model, the authors bridge the gap between stochastic growth in heterogeneous media and $q$ -deformed random matrix theory.

The work extends the classical homogeneous picture (Johansson's correspondence) to inhomogeneous settings, demonstrating that the nature of the inhomogeneity dictates whether the system falls into the soft-edge (Airy) or hard-edge (Bessel) universality class.
The derivation of the limiting spectral density and edge statistics for the $q$ -JUE is presented as a result of independent interest, filling a gap in the literature regarding the asymptotic theory of $q$ -deformed ensembles.
The paper emphasizes that the explicit duality allows for a complete asymptotic analysis of both the limit shape and the fluctuations, which are often difficult to determine in inhomogeneous settings using hydrodynamic or variational methods alone.

The authors remain modest regarding applications, framing the work as a contribution to integrable probability and random matrix theory, specifically extending the class of exactly solvable stochastic systems to include spatially inhomogeneous environments.

More like this