Fredholm Determinants

• Fredholm determinants are analytic functions defined via eigenvalue products or cumulant expansions for trace-class operators.
• They are fundamental in integrable systems and random matrix theory, linking tau-functions to gap probabilities and non-perturbative phenomena.
• Accurate numerical evaluation using methods like Nyström discretization and regularized determinants underpins practical spectral and stability analyses.

A Fredholm determinant is an analytic function associated with a trace-class integral operator, and it plays a central role in spectral theory, integrable systems, random matrix theory, statistical mechanics, quantum field theory, quantum gravity, and numerics. The Fredholm determinant encodes subtle non-perturbative information about operator spectra, gap probabilities in determinantal processes, tau-functions of integrable hierarchies, and spectral statistics of quantum systems.

1. Formal Definitions and Analytic Foundations

Let $K$ be a trace-class integral operator acting on a separable Hilbert space $H$ (e.g., $L^2(a,b)$ or $L^2(\mathbb{R})$), with kernel $K(x,y)$. The Fredholm determinant is defined, for $z \in \mathbb{C}$, via either of two equivalent constructions:

• Product over eigenvalues: If $\{\lambda_n\}$ are the eigenvalues of $K$ (counted with algebraic multiplicity),

$$\det \left(I - z K\right) = \prod_{n=1}^\infty (1 - z \lambda_n).$$

• Cumulant (trace) expansion:

$$\det(I - z K) = \exp\left(-\sum_{m=1}^\infty \frac{z^m}{m} \mathrm{Tr}\left[K^m\right]\right),$$

where $H$0 denotes the $H$1-fold trace, i.e.

$H$2

For operators with a kernel satisfying suitable regularity and decay, the determinant may also be developed as a convergent power series (Fredholm expansion) in the coupling parameter $H$3: $H$4 This series underlies the connection to determinantal point processes and gap probabilities (0804.2543).

Generalizations to matrix-valued kernels, infinite domains, and regularized determinants (e.g. 2-modified for Hilbert–Schmidt class) are well-established and necessary when dealing with Birman–Schwinger operators, integrable systems, or multiparametric models (Britz et al., 2020, Gallo et al., 30 Jul 2025, Gallo et al., 30 Mar 2026).

2. Fredholm Determinants and Integrable Systems

Fredholm determinants are intrinsically linked to tau-functions of integrable hierarchies and isomonodromic deformations (Krajenbrink, 2020, Bothner, 2022, Adler et al., 2012, Blower et al., 17 Dec 2025). The determinant often serves as a generating object (tau-function) for hierarchy solutions, spectral invariants, and conserved quantities.

Example: Painlevé Structure

• The gap probability of the Airy kernel process is the Tracy–Widom distribution,

$H$5

which satisfies a Painlevé II equation for $H$6: $H$7 with a representation

$H$8

(Adler et al., 2012, Cafasso et al., 2019, Bothner, 2022).

• Similar correspondences occur for higher integrable hierarchies (Painlevé I, II, V, etc.), with Fredholm determinants of generalized Airy, Bessel, or sine kernels yielding gap/largest eigenvalue distributions in random matrix ensembles, or special solutions of nonlinear PDEs (KP, KdV, NLS) (Krajenbrink, 2020, Blower et al., 17 Dec 2025, Nagai, 10 Nov 2025).

The integral operator structure underlying these hierarchies is typically "integrable" in the sense of Its–Izergin–Korepin–Slavnov, often with kernels admitting Christoffel–Darboux or Wronskian representations.

3. Random Matrix Theory, Statistical Physics, and Quantum Gravity

Fredholm determinants are fundamental in expressing probabilities for random matrix spectra (e.g., gap and level statistics) and physical correlation functions:

• Gap probabilities: For determinantal point processes with kernel $H$9, the probability of no eigenvalue in $L^2(a,b)$0 is $L^2(a,b)$1. Explicitly, in Wigner-Dyson ensembles (GUE, GOE, etc.), those determinants encode edge/bulk/soft/hard edge universality (Cafasso et al., 2019, Nagai, 10 Nov 2025, Gamayun et al., 2024).
• Ising Model: Diagonal spin-spin correlation of the Ising model can be written as a Fredholm determinant of an operator with Appell function kernel, with exact agreement to a discrete Fredholm determinant with hypergeometric kernel. This equivalence is shown via Riemann–Hilbert and scattering-theoretic approaches (Witte et al., 2011).

Jackiw–Teitelboim gravity and black hole microstates: In 2d quantum gravity, the non-perturbative spectral statistics of JT-gravity are determined by a kernel derived from the string-equation potential, and its Fredholm determinant provides direct access to the microstate energy-level statistics, resolving the transition from the smooth, semiclassical density to a discrete, statistical regime (Johnson, 2021).

4. Riemann–Hilbert Problems, Tau-Functions, and Nonlinear PDEs

The connection with Riemann–Hilbert problems is fundamental and ties the Fredholm determinant to isomonodromic tau-functions:

• For many kernels (Hankel, Airy, Bessel, etc.), there is a 2×2 (or higher) matrix-valued Riemann–Hilbert problem whose solution's asymptotics encode derivatives of the Fredholm determinant (Bothner, 2022, Cafasso et al., 2019).
• By integrable operator methods, the logarithmic derivatives of the determinant satisfy nonlinear PDEs in the boundary parameters (e.g., interval endpoints), typically of Painlevé or even more general form.
• For models such as the KP and NLS equations, the tau-function arising as a Fredholm determinant of a Hankel (Lyapunov) operator yields classical solutions as $L^2(a,b)$2 (Blower et al., 17 Dec 2025).

5. Numerical Evaluation and Practical Algorithms

Modern computational approaches for Fredholm determinants are grounded in projection and quadrature methods:

• Nyström discretization: Select quadrature points $L^2(a,b)$3 and weights $L^2(a,b)$4, approximate $L^2(a,b)$5 by a matrix $L^2(a,b)$6, and compute $L^2(a,b)$7 as an approximation to $L^2(a,b)$8 (0804.2543, Gallo et al., 30 Jul 2025).
• Exponential convergence is achieved for analytic kernels typical in random matrix theory; Clenshaw–Curtis and Gauss–Legendre quadrature achieve high accuracy.
• Recent extensions accurately handle matrix-valued kernels, unbounded domains, and trace class and Hilbert–Schmidt operators, with rigorous error bounds (Gallo et al., 30 Jul 2025, Gallo et al., 30 Mar 2026). Regularized determinants (e.g. $L^2(a,b)$9, 2-modified) are used for Hilbert–Schmidt class operators when the standard determinant is ill-defined.

Algorithmic steps (abbreviated for matrix-valued kernel $L^2(\mathbb{R})$0 on the real line) (Gallo et al., 30 Jul 2025):

1. Truncate the domain to $L^2(\mathbb{R})$1 using exponential decay estimates for the kernel.
2. Discretize $L^2(\mathbb{R})$2 adaptively (e.g., composite Simpson rule).
3. Form the block matrix $L^2(\mathbb{R})$3.
4. Compute the determinant (regular or modified).
5. Analyze and control truncation and quadrature errors.

This methodology enables the computation of gap probabilities, eigenvalue statistics, and spectral stability problems for linearized nonlinear wave equations, even when direct use of Evans functions or analytical solutions is intractable (Gallo et al., 30 Mar 2026).

6. Regularized Determinants, Product Formulas, and Spectral Shift

Product and composition formulas for Fredholm and modified determinants are crucial in spectral theory and perturbation analysis:

• Trace-class case: $L^2(\mathbb{R})$4 for $L^2(\mathbb{R})$5 trace class.
• Modified determinants: For Hilbert–Schmidt $L^2(\mathbb{R})$6,

$L^2(\mathbb{R})$7

Generalizations for higher Schatten classes involve explicit correction terms given by traces of noncommutative polynomials in $L^2(\mathbb{R})$8 (Britz et al., 2020).

These product and composition formulae are essential for perturbation theory, scattering theory, functional determinants in quantum field theory, and spectral shift functions.

Further, subspace perturbation theory expresses overlap determinants (Anderson orthogonality), overlap of spectral projectors, or spectral shift as explicit integrals or functionals of the shifted spectra via Fredholm determinants (Gebert, 2017).

7. Asymptotics, Universality, and Connections to Tau-Functions

In large size, scaling, or finite temperature limits, Fredholm determinants display rich asymptotic structures:

• Sine kernel and Painlevé V: In the bulk scaling limit, the determinant becomes a tau-function of Painlevé V, with expansion involving Bernoulli numbers and Riemann zeta-values organizing global and local asymptotics (Nagai, 10 Nov 2025).
• Finite temperature and Toeplitz/Szegő theory: At finite temperature, deformed sine kernels and their determinants are closely connected with Toeplitz determinants, Riemann–Hilbert analysis, and yield subleading corrections including Fisher–Hartwig and Borodin–Okounkov terms, with explicit formulas involving deformed phase shifts and winding numbers (Gamayun et al., 2024).
• Physical regimes in quantum gravity: The Fredholm determinant resolves the crossover from geometric/topological expansion to discrete, statistical microphysics in black-hole entropy, bridging classical and quantum regimes in gravitational models (Johnson, 2021).

Table: Fredholm Determinant Structures and Applications

Domain / Structure	Kernel Type	Governing Equation / Object
Random Matrix Edge (e.g., GUE, GOE)	Airy / Bessel / sine	Painlevé II / Painlevé V
Quantum gravity, JT, microstates	JT kernel (numerical)	Nonlinear ODE (integrable, Painlevé-type)
Integrable PDEs (KdV, KP, NLS, KPZ)	Hankel / composition	Tau-function, integrable hierarchies
Ising model correlations	Appell/hypergeometric	Scattering theory, Riemann–Hilbert
Subspace / projector overlap	Spectral projections	Spectral shift, determinant formula
Spectral stability (Birman-Schwinger)	Matrix-valued, HS	Evans function, modified determinant
Numerical evaluation	General/analytic	Nyström (quadrature) methods

• Quantum gravity microstates and nonperturbative JT analysis: (Johnson, 2021)
• Noncommutative Painlevé II, matrix Airy, RHP: (Tarricone, 2020)
• Painlevé/correlation and determinantal point processes: (Cafasso et al., 2019)
• General structure in integrable systems and stochastic PDEs: (Krajenbrink, 2020)
• Hankel composition, matrix RHP, large deviation: (Bothner, 2022)
• Singular, Toeplitz, and finite temperature spectral asymptotics: (Gamayun et al., 2024, Nagai, 10 Nov 2025)
• Numerical schemes for general and matrix-valued kernels: (Gallo et al., 30 Jul 2025, 0804.2543)
• Product formulas for regularized determinants: (Britz et al., 2020)
• Spectral shift, projections, perturbation overlap: (Gebert, 2017)
• KP tau-functions from linear systems: (Blower et al., 17 Dec 2025)
• Gap statistics and Schrödinger operator kernels: (Kimura et al., 2024)
• Ising model correlation Fredholm representation: (Witte et al., 2011)
• Evans function/Fredholm determinant connection: (Karambal et al., 2011)

---

The Fredholm determinant provides an indispensable analytic invariant and computational tool across mathematical physics, integrable probability, spectral theory, and nonlinear dynamics, unifying divergent methods through operator theory, tau-functions, and Riemann–Hilbert technology.