Spectral and Phase Structure of a Unitary Matrix Model with Fisher-Hartwig Singularities

1. Problem Statement

The authors investigate a Unitary Matrix Model defined on the group $SU(N)$ with a complex action containing Fisher-Hartwig singularities. The model is motivated by its relevance to:

• Weakly coupled gauge theories.
• The Ising model spin correlation function.
• Low-temperature two-level Quantum Chromodynamics (QCD) models.
• Systems exhibiting the sign problem, where standard Monte Carlo methods fail due to a complex action.

The core objective is to analyze the spectral properties (eigenvalue distribution) and phase structure of this model at both finite $N$ and the large-$N$ limit, specifically focusing on how the complex potential and singularities influence phase transitions.

The action is given by:
$$S(U) = -\text{Tr}\left( \ln(1 + \alpha U) + \ln\left(1 + \frac{1}{\gamma} U^\dagger\right) \right)$$
where $\alpha, \gamma > 0$ are parameters, and the coupling is $g$ (with $\tau = 1/g$). The singularities occur at branch points $z = -1/\alpha$ and $z = -1/\gamma$.

2. Methodology

The authors employ distinct mathematical techniques for the finite-$N$ and large-$N$ regimes:

A. Finite-$N$ Analysis

• Fourier Decomposition: The constrained $SU(N)$ integral is converted into an unconstrained $U(N)$ integral plus an infinite sum over Fourier modes ($k$) to enforce $\det U = 1$.
• Toeplitz Determinants: The partition function is expressed using Toeplitz determinants of the symbol $f(z) = (1+\alpha z)^\tau (1 + \frac{1}{\gamma z})^\tau$.
• Hypergeometric Functions: The Fourier coefficients of the symbol are calculated explicitly using Gauss hypergeometric functions ($_2F_1$) and Appell functions ($F_1$).
• Singularity Analysis: The authors analyze the behavior of these coefficients as the branch points cross the unit circle. They determine the order of phase transitions by examining the discontinuities in the derivatives of the free energy with respect to the coupling parameters.

B. Large-$N$ Analysis (Double Scaling Limit)

• Saddle Point Method: In the limit $N \to \infty$ with $\sigma = 1/(Ng)$ fixed, the integral is evaluated using the saddle point approximation.
• Eigenvalue Density: The problem is mapped to finding the spectral density $\rho(z)$ of the eigenvalues $z = e^{i\theta}$.
• Riemann-Hilbert Problem: The saddle point equation is treated as a Riemann-Hilbert problem to solve for the spectral function and its support (contour $C$) in the complex plane.
• Ansatz Construction:
  • Ungapped Phases: The support $C$ is a closed contour. An ansatz for $\rho(z)$ is constructed based on the poles of the potential.
  • Gapped Phase: The support $C$ is an open arc. The resolvent method ($R(z)$) is used, employing the Sokhotski-Plemelj formula to relate the potential to the spectral density.
• Constraints: The $SU(N)$ constraint ($\sum \theta_i = 0$) is enforced via a Lagrange multiplier $N$ (distinct from the matrix size $N$), determined self-consistently.

3. Key Contributions and Results

A. Finite-$N$ Phase Transitions

• Coupling-Dependent Order: The model exhibits phase transitions at finite $N$ whenever the branch points cross the unit circle.
• Order of Transition: The order of the transition is determined by the coupling $\tau = 1/g$. Specifically, the order is $k = 1 + \lceil \tau \rceil$.
  • If $\tau$ is an integer, the partition function is smooth (no transition).
  • If $\tau$ is non-integer, the $k$-th derivative of the free energy is discontinuous.
• Disappearance at Large-$N$: As $N \to \infty$, $\tau \propto N$, causing the transition order to approach infinity. Consequently, these specific finite-$N$ transitions vanish in the large-$N$ limit.

B. Large-$N$ Phase Structure

In the large-$N$ limit, the model exhibits a rich phase diagram characterized by Gross-Witten-Wadia (GWW) type transitions. The phases are defined by the location of the branch points ($z = -1/\alpha, -1/\gamma$) relative to the support contour $C$.

1. Four Ungapped Phases:
The spectral support $C$ is a closed loop enclosing specific poles. The phases are:

• Phase A: Small $\alpha$, large $\gamma$. Encloses $z = -1/\gamma$.
• Phase B: Large $\alpha$, large $\gamma$. Encloses both poles.
• Phase C: Large $\alpha$, small $\gamma$. Encloses $z = -1/\alpha$.
• Phase D: Small $\alpha$, small $\gamma$. Encloses no poles.
• Result: Analytic expressions for the spectral density $\rho(z)$, Wilson loops, and free energy were derived for all ungapped phases.

2. One Gapped Phase:

• The support $C$ becomes an open arc (a gap opens on the negative real axis).
• The spectral density vanishes at the endpoints of the gap.
• Result: Observables in this phase (Wilson loops, free energy) depend on transcendental equations involving the gap endpoints and cannot be expressed in simple closed forms; they require numerical solution.

3. Nature of Transitions:

• Third-Order Transitions: Transitions between ungapped phases and the gapped phase are third-order (discontinuity in the third derivative of the free energy).
• No Direct Ungapped-Ungapped Transitions: The model does not allow direct transitions between different ungapped phases (e.g., A to B). Any transition between ungapped phases must proceed through the intermediate gapped phase.
• Complex Saddle Points: Unlike standard unitary models with real actions where eigenvalues lie on the unit circle, the saddle points in this complex model lie off the unit circle in the complex plane for both gapped and ungapped phases.

C. Connection to QCD

The authors map the model parameters to a low-temperature, weakly coupled QCD model:

• $\alpha = e^{(\mu-\epsilon)/T}$, $\gamma = e^{(\mu+\epsilon)/T}$.
• Phases: The ungapped phases correspond to confined phases, and the gapped phase corresponds to the deconfined phase.
• Silver Blaze Phenomenon: The Wilson loop expectation values exhibit behavior analogous to the Silver Blaze phenomenon, where thermodynamic quantities remain independent of the chemical potential $\mu$ in the confined phase despite $\mu$ appearing in the action.

4. Significance

• Resolution of Sign Problem Dynamics: The study provides a rigorous analytical framework for understanding matrix models with complex actions, which are notoriously difficult to simulate numerically due to the sign problem.
• Generalization of GWW Transitions: It extends the classic Gross-Witten-Wadia transition (typically a single gapped-to-ungapped transition) to a system with multiple singularities, revealing a complex landscape of four distinct ungapped phases separated by a single gapped phase.
• Finite vs. Large-$N$ Distinction: The work highlights a crucial distinction: finite-$N$ transitions are driven by the non-analyticity of the coupling (order depends on $1/g$), while large-$N$ transitions are driven by the topology of the eigenvalue support (third-order GWW type).
• QCD Analogies: By successfully mapping the model to a two-level QCD system, the authors provide a solvable toy model that captures essential features of confinement/deconfinement transitions and the behavior of Wilson loops in the presence of a chemical potential.

In conclusion, the paper establishes that Fisher-Hartwig singularities in complex unitary matrix models induce a rich phase structure where the interplay between coupling constants and singularity locations dictates the spectral support, leading to third-order phase transitions and complex eigenvalue distributions that mirror phenomena in high-energy physics.