Verblunsky Coefficients in OPUC Theory

• Verblunsky coefficients are key parameters in orthogonal polynomials on the unit circle, uniquely defining the measure and enabling the Szegő recurrence.
• They construct CMV matrices and facilitate transfer matrix formalism, thereby bridging spectral theory with dynamical systems and quantum walks.
• Applications include ergodic theory, time series analysis, and statistical mechanics, with explicit formulas yielding insights into localization and gap labeling.

The Verblunsky coefficients are fundamental parameters in the theory of orthogonal polynomials on the unit circle (OPUC), determining both the underlying measure and the canonical unitary matrices (CMV matrices) associated with these polynomials. They serve as the unitary analogue of Jacobi parameters for orthogonal polynomials on the real line and are central to modern spectral theory, ergodic theory, mathematical physics, and time series analysis.

1. Definition and Foundational Properties

Given a nontrivial probability measure $\mu$ on the unit circle $\mathbb{T} = \{z \in \mathbb{C} : |z| = 1\}$, the sequence of monic orthogonal polynomials $\{P_n(z)\}_{n \geq 0}$ is obtained via the Gram–Schmidt process applied to $\{1, z, z^2, \ldots\}$ in $L^2(\mathbb{T}, d\mu)$. The reversed (Szegő dual) polynomial is $P_n^*(z) = z^n \overline{P_n(1/\overline{z})}$. The Szegő recurrence takes the form

$$P_{n+1}(z) = z P_n(z) - \overline{\alpha_n} P_n^*(z)$$

with the normalization $|\alpha_n| < 1$ for all $n \geq 0$; the $\alpha_n \in \mathbb{D}$ are the Verblunsky coefficients. Verblunsky's theorem establishes a one-to-one correspondence between sequences $\{\alpha_n\}_{n \geq 0}$ in the open unit disk and nontrivial probability measures $\mu$ on $\mathbb{T}$ (Li et al., 2021). This sequence determines, and is determined by, the measure and its spectral features (Lin et al., 2024).

2. Role in CMV Matrices and Transfer Matrix Formalism

The Verblunsky coefficients $\{\alpha_n\}$ enter naturally into the construction of the CMV matrix, a canonical five-diagonal unitary operator acting on $\ell^2(\mathbb{Z})$ (or $\ell^2(\mathbb{N})$ for the half-line case). The blocks of the CMV matrix are explicitly built from $\alpha_n$ and $1-|\alpha_n|^2$. The multiplication-by-$z$ operator in $L^2(\mathbb{T}, \mu)$ is unitarily equivalent to the CMV matrix, and the spectrum $\sigma(\mathcal{C})$ of the CMV matrix depends entirely on the sequence $\{\alpha_n\}$ (Li et al., 2021).

In the transfer matrix formalism, the Verblunsky coefficients determine the Szegő cocycle: $$S(\alpha, z) = \frac{1}{\rho} \begin{pmatrix} z & -\overline{\alpha} \ - \alpha z & 1 \end{pmatrix} \in SU(1,1), \quad \rho = (1 - |\alpha|^2)^{1/2}$$ Iteration of these matrices encodes dynamical and spectral properties of the OPUC and CMV operator (Li et al., 2021, Lin et al., 2024).

3. Spectral Theory, Gap Labelling, and Cantor Spectra

Spectral features such as the presence of Cantor spectra, gap labeling, and eigenvalue localization are governed by the Verblunsky sequence:

• In settings where the $\alpha_n$ are quasi-periodic or almost periodic (e.g., generated by analytic functions over a torus rotation or a skew product), the spectrum of the CMV matrix exhibits a fractal, Cantor-set structure. All gaps allowed by the gap-labelling theorem are generically open; gap labeling is governed by the rotation number associated with the Szegő cocycle (Li et al., 2021).
• For dynamically defined Verblunsky coefficients via strongly mixing systems like hyperbolic toral automorphisms, the associated Szegő cocycle has strictly positive Lyapunov exponents over large spectral arcs. This positivity, combined with large deviation estimates for the transfer matrix norms, leads to Anderson localization for the CMV operator—pure point spectrum and exponentially localized eigenfunctions (Lin et al., 2024).
• For substitution sequences such as Fibonacci Verblunsky coefficients, the essential spectrum is a dynamically characterized zero-measure Cantor set, with Hausdorff dimension depending sensitively on the parameters of the substitution (Damanik et al., 2012).

4. Explicit Formulas, Representations, and Parametrizations

Verblunsky coefficients admit several explicit representations and parametrizations:

• Real-sequence parametrization: For any OPUC, there exist real sequences $\{c_n\}, \{d_n\}$ (the latter is a positive chain sequence with minimal parameter sequence $\{m_n\}$) such that $$\alpha_{n-1} = \overline{\tau}_{n-1} \frac{1 - 2m_n - i c_n}{1 - i c_n}$$, where $\tau_n$ is a product of rotations determined by $c_k$ (Bracciali et al., 2016).
• Asymptotic results: For measures with varying exponential weights, the leading asymptotic behavior of the Verblunsky coefficients is controlled by the equilibrium measure, with higher-order corrections linked to derivatives of the weight function (Poplavskyi, 2010).
• PACF/Time series: In stationary stochastic processes, the Verblunsky coefficients coincide with the partial autocorrelation function (PACF). They can be explicitly calculated in terms of Fourier coefficients of the phase function, allowing precise asymptotics for FARIMA and short-memory processes (Bingham et al., 2011).
• Block/matrix generalization: For $m \times m$ matrix-valued measures on $\mathbb{T}$, the Verblunsky coefficients generalize to contractive matrices and appear as parameters in the block Szegő recurrence. They completely determine the spectral measure and the corresponding block CMV operator (Clark et al., 2010).
• Multivariate case: On the unit sphere $\partial \mathbb{B}^d \subset \mathbb{C}^d$, multivariate Verblunsky coefficients $\gamma_{\alpha, \beta}$ are defined via the Gram matrix of monomials with respect to the underlying measure, and participate in recurrence relations reminiscent of the classic Szegő formalism. Product formulas involving these coefficients (Szegő–Verblunsky theorems) hold under suitable holomorphicity/outer-function conditions, but can fail for more general weights, highlighting significant differences from the univariate setting (Gauntlett et al., 11 Dec 2025).

5. Applications: Dynamical Systems, Mathematical Physics, and Perturbation Theory

Verblunsky coefficients feature prominently in a range of applications:

• Dynamical sequences: When $\{\alpha_n\}$ are defined by iterating smooth functions over measure-preserving transformations, spectral properties such as pure point spectrum and singular continuous spectrum can be rigorously established. For example, the use of a hyperbolic toral automorphism induces strong mixing properties in $\{\alpha_n\}$, leading to Anderson localization for the associated CMV matrix (Lin et al., 2024).
• Quantum walks and spin chains: The spectral theory of time-homogeneous quantum walks with coin operators specified by Verblunsky coefficients translates directly into the analysis of CMV matrices, with consequences for wavepacket spreading and anomalous transport exponents. In quasicrystalline settings (e.g., Fibonacci sequences), the resulting spectra are singular continuous and support anomalous dynamical properties (Damanik et al., 2013).
• Statistical mechanics: In the 1D Ising model with complex field, the Lee–Yang zeros of the partition function in the thermodynamic limit are supported exactly on the essential spectrum of the CMV matrix whose Verblunsky coefficients are constructed from the coupling constants (Damanik et al., 2013).
• Point perturbations and spectral gaps: The effect of adding a pure point mass to a spectral gap of a measure is explicitly described at the level of the Verblunsky coefficients; explicit formulas quantify the rate and variation of the resulting perturbation, with power-law and non-exponential regimes possible (Wong, 2010). The minimal-rank decoupling of CMV matrices by altering one Verblunsky coefficient (as opposed to the rank-1 decoupling for Jacobi matrices) reflects the richer unitary structure (Clark et al., 2010).

6. Periodic, Quasi-periodic, and Substitutional Regimes

In periodic and quasi-periodic settings, the Verblunsky sequence dictates spectral bands, gap structure, and universality phenomena:

• Periodic sequences: For $\alpha_{n+p} = \alpha_n$, the measure's support consists of a finite-band set characterized by the discriminant $\Delta(z)$ constructed from the period-$p$ transfer matrix. There exist universal explicit closed-form expressions for the orthogonal polynomials in terms of Chebyshev polynomials and $\Delta(z)$, with consequences for singular points in the bands and universality limits (Simanek, 2018).
• Quasi-periodic and almost periodic cases: Analytic quasi-periodic or almost-periodic Verblunsky coefficients generically produce Cantor spectra with all allowed spectral gaps open, as guaranteed by small-divisor KAM schemes controlling the Szegő cocycle (Li et al., 2021).
• Substitution sequences: Fibonacci Verblunsky coefficients yield essential spectra with fractal (Cantor) structure, singular continuous spectral measures, power-law asymptotics in the norms of orthogonal polynomials, and are amenable to trace map and dynamical systems analysis (Damanik et al., 2012).

7. Generalizations and Verblunsky-type Coefficients

Verblunsky-type coefficients generalize the standard setting to discrete Dirac and canonical systems:

• Discrete Dirac systems: In self-adjoint discrete Dirac systems, the potential matrices $C_k$ are $j$-unitary and positive definite, and admit Halmos extension parametrizations where the off-diagonal blocks $\rho_k$ serve as Verblunsky-type coefficients. These parameters provide a one-to-one correspondence with the Weyl function, obey similar norm constraints, and decay to zero for rational Weyl functions, paralleling the behavior of classical Verblunsky coefficients in finite Szegő theory (Roitberg et al., 2018).
• Toeplitz and Hankel matrices: In the context of positive-definite Toeplitz and Hankel matrix families, Verblunsky-type coefficients arise naturally via Halmos extension or Cholesky factorization, connect to discrete Dirac or canonical systems, and capture the inverse spectral structure (Sakhnovich, 2017). The infinite sequence uniquely characterizes the spectral measure, echoing the original theorem of Verblunsky for OPUC.

---

References:

• (Li et al., 2021): Cantor spectrum for CMV matrices with almost periodic Verblunsky coefficients
• (Lin et al., 2024): Anderson localization for CMV matrices with dynamically defined Verblunsky coefficients
• (Damanik et al., 2012): Orthogonal polynomials on the unit circle with Fibonacci Verblunsky coefficients
• (Bracciali et al., 2016): OPUC: Verblunsky coefficients via real sequences and chain sequences
• (Poplavskyi, 2010): Asymptotic analysis of Verblunsky coefficients for varying weights
• (Bingham et al., 2011): Explicit representation of Verblunsky coefficients in time-series analysis
• (Clark et al., 2010): Weyl-Titchmarsh theory for CMV with matrix-valued Verblunsky coefficients
• (Gauntlett et al., 11 Dec 2025): Multivariate Verblunsky coefficients on the unit sphere
• (Damanik et al., 2013): CMV matrices and physics applications of Fibonacci Verblunsky coefficients
• (Wong, 2010): Point perturbations and asymptotic behavior of Verblunsky coefficients
• (Simanek, 2018): OPUC with periodic Verblunsky coefficients and universality
• (Clark et al., 2010): Minimal rank decoupling by changing a Verblunsky coefficient
• (Roitberg et al., 2018): Verblunsky-type coefficients in self-adjoint discrete Dirac systems
• (Sakhnovich, 2017): Verblunsky-type coefficients for Toeplitz and Hankel matrices