---
title: Generalized Laurent Multiple Orthogonal Polynomials
url: https://www.emergentmind.com/topics/generalized-laurent-multiple-orthogonal-polynomials
type: topic
---

# Generalized Laurent Multiple Orthogonal Polynomials

Generalized Laurent multiple orthogonal polynomials (Laurent MOPs) extend the framework of classical orthogonal polynomials on the unit circle (OPUC) to settings involving several measures or linear functionals, and to simultaneous orthogonality at both $z=0$ and $z=\infty$. These objects are characterized as solutions to a generalized two-point Hermite–Padé approximation problem and encompass various constructions such as Angelesco and AT systems. The resulting theory unifies and significantly generalizes foundational aspects of OPUC, multiple orthogonality on the real line (OPRL), and Hermite–Padé approximation on the unit circle, providing a comprehensive setting for the study of multi-measure and operator-theoretic phenomena on $\mathbb T$ and $\mathbb R$ [2601.04783][2410.12094].

## 1. Formal Definition and Orthogonality Conditions

Let $r \in \mathbb N$, and consider two multi-indices $\bm n = (n_1,\dots,n_r)$ and $\bm m = (m_1,\dots,m_r)$, with $n_j, m_j \ge 0$ and $n_j + m_j \ge 0$ for $j=1,\dots,r$. Associate to each $j$ a linear functional $L_j: \operatorname{Span}\{z^k\}_{k\in\mathbb Z} \to \mathbb C$ specified by moments $L_j[z^{-k}] = c_{k,j}$. In the case $L_j[f] = \int_{|w|=1} f(w) d\mu_j(w)$, the $c_{k,j}$ are Fourier moments of the probability measure $\mu_j$ on $\mathbb T$.

**Type II Laurent MOPs** of index $(\bm n;\bm m)$ are Laurent polynomials
\[
\Phi_{\bm n;\bm m}(z) \in \operatorname{Span}\{z^{-m_j},\dots, z^{n_j}\}, \quad \forall j
\]
satisfying the simultaneous orthogonality
\[
L_j\bigl[\Phi_{\bm n;\bm m}(w)w^{-k}\bigr]=0, \quad k=-m_j,\ldots, n_j-1, \quad j=1,\ldots,r.
\]
Uniqueness (up to multiplication by a constant) and the existence of a monic normalization are guaranteed if a certain block Toeplitz normality determinant $\det T_{\bm n;\bm m}$ is nonzero.

Dually, **type I Laurent MOPs** are vector-valued Laurent polynomials
\[
\bm\Xi_{\bm n;\bm m}(z) = (\Xi_1,\dots,\Xi_r), \quad \Xi_j \in \operatorname{Span}\{z^{-n_j},\dots,z^{m_j-1}\}
\]
subject to
\[
\sum_{j=1}^r L_j[\Xi_j(w)w^{-k}] =
\begin{cases}
1, & k=-\sum n_j \\
0, & k=-\sum n_j+1, \ldots, \sum m_j-1
\end{cases}
\]
[2601.04783][2410.12094].

## 2. Hermite–Padé Approximation and Characterization

Associated to each $L_j$ is a pair of Carathéodory-type expansions:
\[
F_j^{(0)}(z) = c_{0,j} + 2\sum_{k=1}^\infty c_{k,j}z^k, \qquad F_j^{(\infty)}(z) = -c_{0,j} - 2\sum_{k=1}^\infty c_{-k,j}z^{-k}
\]
The **generalized two-point Hermite–Padé problem of type II** seeks Laurent polynomials $\Phi_{\bm n;\bm m}$ and $\Psi_j$ such that, for all $j$,
\begin{align*}
\Phi_{\bm n;\bm m}(z) F_j^{(0)}(z) + \Psi_j(z) &= \mathcal O(z^{n_j}) \quad (z\to 0) \\
\Phi_{\bm n;\bm m}(z) F_j^{(\infty)}(z) + \Psi_j(z) &= \mathcal O(z^{-m_j-1}) \quad (z\to \infty)
\end{align*}
Any solution $\Phi_{\bm n;\bm m}$ of this system is a type II Laurent MOP; conversely, every such Laurent MOP solves the Hermite–Padé problem [2601.04783].

Dually, type I Laurent MOPs are characterized by solutions to a related type I Hermite–Padé problem under complementary asymptotic requirements.

This framework subsumes classical orthogonal polynomials and their multiple orthogonality analogs, recovering both unit circle and real-line multiple orthogonality as special or limiting cases [2601.04783][2410.12094].

## 3. Angelesco and AT Systems: Normality and Existence

**Angelesco systems** on $\mathbb T$ are collection of measures $\mu = (\mu_1, \dots, \mu_r)$ with $\,\operatorname{supp}\mu_j \subseteq I_j\,$ for disjoint arcs $I_1,\dots,I_r \subset \mathbb T$, and $I_j\cap I_k$ at most two points for $j\ne k$. For Angelesco systems, the moment matrix $M_{\bm n}$ is non-singular for all $\bm n$, ensuring normality for any multi-index [2410.12094].

**AT systems** are characterized by measures absolutely continuous on a single arc and such that their associated trigonometric families (sines and cosines up to degree determined by $\bm n$) form an extended Chebyshev (T-)system. Normality similarly follows via Wronskian-type determinants and the Chebyshev property.

These concepts guarantee that for large classes of measures, the corresponding generalized Laurent MOPs are uniquely defined for all admissible multi-indices [2410.12094].

## 4. Szegő Mapping and the Circle–Line Correspondence

Given measures $\nu_1, \dots, \nu_r$ supported on $[-2,2]$, their Szegő images on the unit circle $\mu_j$ are defined by
\[
\int_{\mathbb T} g(z+z^{-1})\,d\mu_j(z) = \int_{-2}^2 g(x)\,d\nu_j(x)
\]
for all integrable $g$. Let $P_{\bm n}(x)$ be the real-line type II multiple orthogonal polynomial, and let $\Phi_{2\bm n}(z)$ be the unit-circle Laurent MOP for $\mu_j$, with doubled multi-index.

The core relations are
\begin{align*}
\Phi_{2\bm n}(z) + \Phi_{2\bm n}(1/z) &= (1+a_{2\bm n})\,P_{\bm n}(z+z^{-1}) \\
z\Phi_{2\bm n}(z) + z^{-1}\Phi_{2\bm n}(1/z) &= \sum_{j=1}^r \kappa_{\bm n, j} A_{\bm n + \bm e_j}(z+z^{-1})
\end{align*}
where $A_{\bm n + \bm e_j}$ are type I real-line MOPs and the $\kappa_{\bm n,j}, a_{2\bm n}$ are explicit constants. Thus, even Laurent MOPs on $\mathbb T$ correspond to MOPs on $[-2,2]$ under $x = z + z^{-1}$ [2601.04783][2410.12094].

Further, via the extended Szegő mapping and Geronimus relations, Jacobi-type recurrence coefficients for OPRL can be expressed in terms of the multi-index Szegő parameters of generalized Laurent MOPs. This provides a comprehensive duality and constructive correspondence between multi-orthogonality on $\mathbb T$ and $\mathbb R$ [2601.04783].

## 5. Recurrence Relations and Determinantal Structures

Laurent MOPs satisfy multi-parameter Szegő-type nearest-neighbour recurrence relations. For each increment in $\bm n$ or $\bm m$ along a standard basis direction, the relations are
\[
\Phi_{\bm n+\bm e_k;\bm m}(z) = z\,\Phi_{\bm n;\bm m}(z) - \alpha_{\bm n;\bm m} \Phi^*_{\bm n;\bm m}(z) - \sum_{j=1}^r \rho_{\bm n;\bm m,j} z \Phi_{\bm n-\bm e_j;\bm m}(z)
\]
and the corresponding relations for $\bm m$-shifts involve parameters $\beta,\sigma$. Here, $\Phi^*(z)$ denotes the reversed polynomial, and the recurrence coefficients generalize classical Verblunsky/Szegő–Baxter parameters to the multi-index setting, subject to nontrivial commutation and compatibility relations, such as
\[
\alpha_{\bm n;\bm m} \beta_{\bm n;\bm m} + \sum_{j=1}^r \rho_{\bm n;\bm m,j} = 1.
\]
All polynomials and recurrence coefficients admit explicit Heine-type determinantal representations in terms of block Toeplitz matrices constructed from the moments $c_{k,j}$. These formulas generalize the determinantal theory of scalar OPUC to block multi-index settings [2601.04783].

## 6. Christoffel–Darboux Kernels and Duality

Given a path in the multi-index lattice, Christoffel–Darboux-type formulas for Laurent MOPs on the unit circle have the structure
\[
(\xi-z)\sum_{k=0}^{N-1} \Phi_{\bm n_k;\bm m}(z)\,\Xi_{\bm n_{k+1};\bm m}(\xi) = \Phi^*_{\bm n_N;\bm m}(z)\Xi^*_{\bm n_N;\bm m}(\xi) - z\xi\sum_{j=1}^r \rho_{\bm n_N;\bm m,j} \Phi_{\bm n_N-\bm e_j;\bm m}(z)\Xi_{\bm n_N+\bm e_j;\bm m}(\xi)
\]
This identity and its duals encode reproducing kernel properties, allow analysis of spectral measures, and supply central tools for further operator-theoretic developments [2601.04783].

## 7. Synthesis, Applications, and Outlook

The framework of generalized Laurent multiple orthogonal polynomials provides a unified platform bridging OPUC, Hermite–Padé approximation, and MOPs on the real line. Key achievements include:

- Block determinant representations for all polynomials and recurrence data.
- Szegő-type recurrences and compatibility relations governing multi-parameter orthogonality.
- Christoffel–Darboux kernels extending the scalar theory to multiple measures.
- Explicit Szegő-Geronimus correspondences yielding real-line recurrence parameters from circle-side data.
- Existence and uniqueness theory encompassing Angelesco and AT systems.

This extended theory has implications in spectral theory, random matrix models, and the analysis of integrable systems with multi-component measure structures. The circle–line duality established by this construction completes and generalizes the scalar Szegő and Geronimus theorems, offering new perspectives and concrete tools for further exploration of multiple orthogonality on both compact and unbounded domains [2601.04783][2410.12094].

Source: https://www.emergentmind.com/topics/generalized-laurent-multiple-orthogonal-polynomials