Infinite Ginibre Point Process

• Infinite Ginibre point process is a determinantal process defined as the thermodynamic limit of eigenvalue distributions from non-Hermitian complex Gaussian matrices.
• It exhibits strong local repulsion, rigidity, and hyperuniform fluctuations, providing exact solvability of its correlation functions.
• Its framework underpins applications in random matrix theory, planar statistical mechanics, and spatial modeling through precise analytic expressions.

The infinite Ginibre point process is a paradigmatic example of a determinantal point process (DPP) on the complex plane, arising as the thermodynamic limit of the eigenvalue distributions of non-Hermitian complex Gaussian matrices. Its intrinsic translational invariance, exact solvability of correlation functions, rigidity phenomena, and hyperuniform fluctuation properties have established it as a canonical model of repulsive random point fields in probability, random matrix theory, and spatial statistics.

1. Origin and Determinantal Structure

The infinite Ginibre point process, denoted $\mathcal{X}_\infty$, is the weak limit of the empirical eigenvalue distribution of $n \times n$ complex Ginibre matrices $G_n$ with i.i.d. standard complex Gaussian entries as $n \to \infty$. The joint density of the eigenvalues $\{z_1, \dots, z_n\}$ is

$$p_n(z_1,\dots,z_n) = \frac{1}{\pi^n \prod_{k=1}^n k!} \exp\left(-\sum_{j=1}^n |z_j|^2\right) \prod_{1 \le i < j \le n} |z_i - z_j|^2.$$

In the limit $n \to \infty$, the process $\mathcal{X}_\infty$ is determinantal with respect to the planar Lebesgue measure $\mu(dz) = d^2z$, with correlation kernel

$$\mathbb{K}(z, w) = \frac{1}{\pi} \exp\left(z\overline{w} - \frac{1}{2}|z|^2 - \frac{1}{2}|w|^2\right).$$

Alternatively, relative to the Gaussian measure $n \times n$0, the kernel can be written as $n \times n$1 (Adhikari et al., 2016, Ghosh et al., 2012, Miyoshi et al., 2016, Katori, 2022, Ghosh, 2012).

2. Correlation Functions and Repulsion

For any $n \times n$2, the $n \times n$3-point correlation function of $n \times n$4 is given by

$n \times n$5

The first and second correlation functions are particularly simple: $n \times n$6 The pair correlation function $n \times n$7 for $n \times n$8 signals strong local repulsion: $n \times n$9 (point-exclusion), and $G_n$0 as $G_n$1 (Miyoshi et al., 2016, Ghosh et al., 2012).

3. Rigidity and Tolerance Properties

A remarkable feature of the infinite Ginibre process is number rigidity: for any bounded open set $G_n$2 with negligible boundary, the configuration of points outside $G_n$3 determines the exact number of points inside $G_n$4 almost surely. Formally, there exists a measurable function $G_n$5 so that $G_n$6 almost surely under $G_n$7 (Ghosh et al., 2012, Ghosh, 2012).

The process also exhibits quantitative tolerance: conditional on the outside configuration, the probability density for the inside points is mutually absolutely continuous with respect to Lebesgue measure on $G_n$8, comparable to a squared Vandermonde factor: $G_n$9 for almost every $n \to \infty$0 in $n \to \infty$1, where $n \to \infty$2 (Ghosh, 2012). Thus, even after conditioning on the outside, the points inside continue to repel quadratically.

4. Hole Probabilities and Potential Theory

For a bounded open set $n \to \infty$3 and scaling factor $n \to \infty$4, the "hole probability"—the probability that $n \to \infty$5 is empty of points—decays as

$n \to \infty$6

where $n \to \infty$7 is the minimum of

$n \to \infty$8

taken over probability measures $n \to \infty$9 supported on $\{z_1, \dots, z_n\}$0 (Adhikari et al., 2016). For $\{z_1, \dots, z_n\}$1, $\{z_1, \dots, z_n\}$2, given by the uniform law on the unit disk.

Several explicit computations are available for $\{z_1, \dots, z_n\}$3 in simple domains:

Region $\{z_1, \dots, z_n\}$4	$\{z_1, \dots, z_n\}$5	Explicit $\{z_1, \dots, z_n\}$6
Disk $\{z_1, \dots, z_n\}$7	$\{z_1, \dots, z_n\}$8	$\{z_1, \dots, z_n\}$9
Annulus $$p_n(z_1,\dots,z_n) = \frac{1}{\pi^n \prod_{k=1}^n k!} \exp\left(-\sum_{j=1}^n |z_j|^2\right) \prod_{1 \le i < j \le n} |z_i - z_j|^2.$$0	$$p_n(z_1,\dots,z_n) = \frac{1}{\pi^n \prod_{k=1}^n k!} \exp\left(-\sum_{j=1}^n |z_j|^2\right) \prod_{1 \le i < j \le n} |z_i - z_j|^2.$$1	$$p_n(z_1,\dots,z_n) = \frac{1}{\pi^n \prod_{k=1}^n k!} \exp\left(-\sum_{j=1}^n |z_j|^2\right) \prod_{1 \le i < j \le n} |z_i - z_j|^2.$$2
Ellipse $$p_n(z_1,\dots,z_n) = \frac{1}{\pi^n \prod_{k=1}^n k!} \exp\left(-\sum_{j=1}^n |z_j|^2\right) \prod_{1 \le i < j \le n} |z_i - z_j|^2.$$3, $$p_n(z_1,\dots,z_n) = \frac{1}{\pi^n \prod_{k=1}^n k!} \exp\left(-\sum_{j=1}^n |z_j|^2\right) \prod_{1 \le i < j \le n} |z_i - z_j|^2.$$4	$$p_n(z_1,\dots,z_n) = \frac{1}{\pi^n \prod_{k=1}^n k!} \exp\left(-\sum_{j=1}^n |z_j|^2\right) \prod_{1 \le i < j \le n} |z_i - z_j|^2.$$5	$$p_n(z_1,\dots,z_n) = \frac{1}{\pi^n \prod_{k=1}^n k!} \exp\left(-\sum_{j=1}^n |z_j|^2\right) \prod_{1 \le i < j \le n} |z_i - z_j|^2.$$6
Half-disk of radius $$p_n(z_1,\dots,z_n) = \frac{1}{\pi^n \prod_{k=1}^n k!} \exp\left(-\sum_{j=1}^n |z_j|^2\right) \prod_{1 \le i < j \le n} |z_i - z_j|^2.$$7	$$p_n(z_1,\dots,z_n) = \frac{1}{\pi^n \prod_{k=1}^n k!} \exp\left(-\sum_{j=1}^n |z_j|^2\right) \prod_{1 \le i < j \le n} |z_i - z_j|^2.$$8	$$p_n(z_1,\dots,z_n) = \frac{1}{\pi^n \prod_{k=1}^n k!} \exp\left(-\sum_{j=1}^n |z_j|^2\right) \prod_{1 \le i < j \le n} |z_i - z_j|^2.$$9

The underlying proof leverages a large deviation principle for the empirical measure and the potential-theoretic energy minimization, with equilibrium measures on $n \to \infty$0 (Adhikari et al., 2016).

5. Fluctuations: Laws of Large Numbers and Hyperuniformity

For the point count $n \to \infty$1, the mean and variance are $n \to \infty$2 and $n \to \infty$3 as $n \to \infty$4. The law of the single logarithm holds: $n \to \infty$5 almost surely, and similarly for the liminf with a negative sign. Thus, fluctuations are sharply concentrated at a scale $n \to \infty$6, refining the central limit theorem and revealing highly regular spatial statistics (Buraczewski et al., 2023).

The variance of $n \to \infty$7 grows as $n \to \infty$8—Class I hyperuniformity in the sense of diminishing density fluctuations: $n \to \infty$9 (Katori, 2022).

6. Generalizations and Simulation

The infinite Ginibre process admits a one-parameter generalization, the $\mathcal{X}_\infty$0-Ginibre process, with kernel

$\mathcal{X}_\infty$1

interpolating between $\mathcal{X}_\infty$2 (Ginibre) and $\mathcal{X}_\infty$3 (homogeneous Poisson). Repulsion decreases as $\mathcal{X}_\infty$4 (Miyoshi et al., 2016).

Simulation methods exploit spectral decompositions or thinning-and-scaling representations. On a large disk, restriction and eigenvalue sampling allows efficient and accurate numerical realization, with convergence to the infinite-volume law as the disk enlarges (Miyoshi et al., 2016).

7. Applications and Significance

The infinite Ginibre process plays a foundational role in random matrix theory, planar statistical mechanics, and spatial modeling. Its determinantal structure provides closed-form expressions for all-order correlation functions, Palm kernels, and void statistics, underpinning the rigorous understanding of nontrivial spatial repulsion phenomena. In wireless network modeling, it captures the sub-Poissonian (repulsive) nature of certain spatial configurations, outperforming independent (Poisson) benchmarks (Miyoshi et al., 2016).

The robust rigidity, sharp control on hole probabilities, hyperuniform fluctuations, and generalizability to $\mathcal{X}_\infty$5-Ginibre and higher-dimensional analogues (Katori, 2022) have made the infinite Ginibre ensemble a central object in the theory of point processes and its applications.