---
title: Noncommutative Bi-orthogonal Polynomial Systems
url: https://www.emergentmind.com/topics/non-commutative-multiple-bi-orthogonal-polynomial-systems
type: topic
---

# Noncommutative Bi-orthogonal Polynomial Systems

Non-commutative multiple bi-orthogonal polynomial systems are polynomial systems over a non-commutative coefficient algebra, typically a division ring or a matrix algebra, subject simultaneously to multiple orthogonality constraints and left-right bi-orthogonality. In the formal framework introduced by Adam Doliwa, the basic data are \(r\) families of formal bi-moments \((\nu_{ij}^{(k)})_{i,j\ge 0}\), \(k=1,\dots,r\), in a free division ring \(\mathcal{R}\), from which one defines \(r\) bilinear forms and monic polynomials indexed by multi-indices \(s=(s_1,\dots,s_r)\). These polynomials satisfy \((x^i,Q_s(x))_{(k)}=0\) for \(i=0,\dots,s_k-1\), admit quasideterminantal expressions, and their normalization functions satisfy non-commutative Hirota equations; after a Hankel-type reduction and a standard discrete-time shift of moments, the same formalism yields a non-commutative multidimensional discrete-time Toda system [2510.02207]. In this sense, the subject unifies three previously distinct lines of development: non-commutative orthogonality, bi-orthogonality, and multiple orthogonality.

## 1. Genealogy of the subject

The non-commutative multiple theory rests on an earlier algebraic theory of non-commutative bi-orthogonality. Emily Sergel formulated noncommutative biorthogonal polynomials over a division ring \(R\) with center \(C\), using a biadditive pairing
\[
\langle \cdot,\cdot\rangle: R[x]\times R[y]\to R,
\]
determined by bimoments \(I_{i,j}=\langle x^i,y^j\rangle\). In that setting, a biorthogonal system is a pair \(\{p_n(x)\},\{q_n(y)\}\) with \(\deg p_n=\deg q_n=n\) and \(\langle p_n,q_m\rangle=0\) for \(n\neq m\), with quasideterminants replacing determinants throughout. Sergel also proved a broad extension of Favard’s theorem: any two sequences of degree \(n\) polynomials can be made biorthogonal for a uniquely constructed system of bimoments [1009.5472].

That framework was not yet “multiple” in the classical sense. This distinction is explicit in the study of deformed complex Hermite polynomials in noncommutative quantum mechanics: the \(GL(2,\mathbb{C})\)-deformed families \(h_{k,l}^g\) and \(\tilde h_{k,l}^g\) are biorthogonal with respect to a single measure on \(L^2(\mathbb C,dv)\), but the construction is stated not to be of the multiple orthogonal type, because it uses one measure and two dual families rather than several functionals or weights [1309.4163].

The genuinely multiple non-commutative setting therefore emerges only when several bilinear forms or moment functionals are imposed simultaneously. Doliwa’s formal construction makes that unification explicit by combining the multi-indexed constraints of multiple orthogonality with the left-right asymmetry of bi-orthogonality and the quasideterminantal machinery required by non-commutativity [2510.02207].

## 2. Formal algebraic structure

In the formal theory, one starts from \(r\) matrices of formal bi-moments
\[
(\nu_{ij}^{(k)})_{i,j\ge 0}, \qquad k=1,\dots,r,
\]
with entries in a free division ring \(\mathcal R\). These define \(r\) bilinear forms on \(\mathcal R[x]\):
\[
(f,g)_{(k)}=\sum_{i,j} a_i \nu_{ij}^{(k)} b_j,
\qquad
f=\sum a_i x^i,\quad g=\sum b_j x^j.
\]
For a multi-index \(s=(s_1,\ldots,s_r)\in\mathbb N_0^r\), with \(|s|=s_1+\cdots+s_r\), one forms a block matrix \(\mathcal M_s\) from the first \(s_k\) rows of the \(k\)-th bi-moment array. The monic non-commutative multiple bi-orthogonal polynomial is then defined by a quasideterminant
\[
Q_s(x)=
\left|
\begin{matrix}
\mathcal M_s \\
1~ x~ x^2~ \cdots~ \boxed{x^{|s|}}
\end{matrix}
\right|,
\]
and satisfies the orthogonality conditions
\[
(x^i,Q_s(x))_{(k)}=0,
\qquad
i=0,\dots,s_k-1,\quad k=1,\dots,r
\]
[2510.02207].

The formalism also has a dual right-action version: one may define polynomials \(P_s(x)\) by requiring \((P_s(x),x^i)_{(k)}=0\) for the corresponding ranges of \(i\), with dual normalization potentials \(\pi_s^{(k)}\) [2510.02207]. This left-right asymmetry is intrinsic. It is the direct analogue, in a multiple setting, of the separate polynomial families in the earlier noncommutative biorthogonal construction over \(R[x]\) and \(R[y]\) [1009.5472].

Conceptually, the structure generalizes three distinct theories at once. It is “multiple” because the constraints are indexed by \(k=1,\dots,r\); it is “bi-orthogonal” because the bilinear forms are not assumed symmetric; and it is “non-commutative” because coefficients and moments take values in a non-commutative algebra, so order matters and quasideterminants replace determinants [2510.02207].

## 3. Quasideterminants and normalization data

The normalization functions are as central as the polynomials themselves. For each \(k\), Doliwa defines
\[
\rho_s^{(k)}=
\left|
\begin{matrix}
\mathcal M_s \\
\nu_{s_k,0}^{(k)}~ \ldots~ \boxed{\nu_{s_k,|s|}^{(k)}}
\end{matrix}
\right|,
\qquad
(x^{s_k},Q_s(x))_{(k)}=\rho_s^{(k)}.
\]
These functions play the role that squared norms, leading principal minors, or \(\tau\)-functions play in commutative theories [2510.02207].

The use of quasideterminants is structurally unavoidable in non-commutative settings. Sergel’s original construction of noncommutative biorthogonal polynomials is already phrased entirely in quasideterminantal form, with the polynomials \(p_n(x)\) and \(q_n(y)\) extracted from moment matrices by noncommutative analogues of Cramer-type formulas [1009.5472]. Matrix-valued variants show the same pattern in a more concrete realization: matrix-valued Cauchy bi-orthogonal polynomials and matrix-valued \(\theta\)-deformed bi-orthogonal polynomials are both represented by quasideterminants built from block moment matrices [2212.14512; 2305.17962].

A distinguished reduction occurs when the bi-moments are of Hankel type,
\[
\nu_{ij}^{(k)}=\nu_{i+j}^{(k)}.
\]
Then the formal multiple bi-orthogonal system reduces to a theory of non-commutative multiple orthogonal polynomials, with bilinear forms
\[
\langle f,g\rangle_{(k)}=\sum_{i,j} a_i^*\,\nu_{i+j}^{(k)}\,b_j,
\]
and the quasideterminants become Hankel-type expressions in the moments [2510.02207]. This reduction is the precise non-commutative analogue of passing from general bi-moment arrays to moment sequences in commutative orthogonal polynomial theory.

## 4. Integrable-system content

The normalization functions \(\rho_s^{(k)}\) satisfy the non-commutative Hirota system
\[
[\rho_s^{(i)}]^{-1}\rho_{s+e_j}^{(i)}+[\rho_s^{(j)}]^{-1}\rho_{s+e_i}^{(j)}=0,
\]
together with the three-index relation
\[
[\rho_s^{(i)}]^{-1}\rho_{s+e_j}^{(i)}
+
[\rho_s^{(j)}]^{-1}\rho_{s+e_k}^{(j)}
+
[\rho_s^{(k)}]^{-1}\rho_{s+e_i}^{(k)}=0,
\]
for distinct \(i,j,k\). The associated linear system is
\[
Q_{s+e_i}(x)-Q_{s+e_j}(x)=Q_s(x)\,[\rho_s^{(i)}]^{-1}\rho_{s+e_j}^{(i)},
\qquad i\neq j.
\]
The proof is based on quasideterminant versions of the Sylvester and homological identities, so integrability is not appended externally; it is encoded in the same algebraic identities that define the polynomials [2510.02207].

Under the Hankel reduction, one may introduce a discrete-time variable by shifting moments,
\[
\nu_{j;t}^{(k)}=\nu_{j+t}^{(k)}.
\]
Then the polynomials satisfy
\[
xQ_{s;t+1}(x)=Q_{s+e_j;t}(x)+Q_{s;t}(x)A_{s;t}^{(j)},
\qquad
A_{s;t}^{(j)}=[\rho_{s;t}^{(j)}]^{-1}\rho_{s;t+1}^{(j)},
\]
and the compatibility conditions become the non-commutative multidimensional discrete-time Toda lattice
\[
A_{s;t+1}^{(j)}-A_{s;t+1}^{(k)}
=
A_{s+e_k;t}^{(j)}-A_{s+e_j;t}^{(k)},
\qquad
A_{s;t}^{(k)}A_{s+e_k;t}^{(j)}
=
A_{s;t}^{(j)}A_{s+e_j;t}^{(k)}
\]
[2510.02207].

This integrable interpretation is consistent with several earlier non-commutative bi-orthogonal models. Matrix-valued Cauchy bi-orthogonal polynomials satisfy a four-term recurrence whose coefficients obey a noncommutative C-Toda lattice with a Lax pair built from fractional differential operators with non-abelian variables [2212.14512]. Matrix-valued \(\theta\)-deformed bi-orthogonal polynomials generate non-commutative Toda-type hierarchies, and Wronski quasi-determinants are used there as non-commutative \(\tau\)-functions; moment modification implements Bäcklund transformations [2305.17962]. Adjacent families of matrix-valued \(\theta\)-deformed bi-orthogonal polynomials likewise produce a fully discrete non-commutative hungry Toda lattice, which is then used as a pre-precessing algorithm for block Hessenberg matrices [2404.13492].

## 5. Model realizations and representative families

One major source of non-commutative bi-orthogonal systems is the asymmetric simple exclusion process. In the tensor-algebraic formulation of ASEP, a linear map \(\mathcal L\) on a tensor algebra generated by non-commuting symbols \(d,e\) gives a bi-moment matrix
\[
B_{n,m}=\mathcal L(d^n e^m).
\]
From the determinant and \(LDU\) decomposition of this matrix, one constructs monic polynomials \(P_n(d)\) and \(Q_n(e)\) satisfying
\[
\mathcal L(P_n(d)Q_m(e))=\Lambda_n\delta_{n,m}.
\]
In one version, the polynomials satisfy first-order recurrence relations and the second moment defines a tridiagonal matrix linked to Chebyshev-like orthogonal polynomials; in the five-parameter case, the boundary basis diagonalizes the bi-moment matrix, while \(\mathbf d+\mathbf e\) becomes tridiagonal and yields a family having the same moments as the Askey–Wilson polynomials [1412.7235; 1902.06373].

A very different realization comes from noncommutative quantum mechanics. The \(GL(2,\mathbb C)\)-deformation of complex Hermite polynomials produces families \(h_{k,l}^g\) and \(\tilde h_{k,l}^g\) with
\[
\int_{\mathbb C}\tilde h_{L-k,k}^g(z,\bar z)\,h_{M-m,m}^g(z,\bar z)^*\,dv(z,\bar z)
=
\delta_{LM}\delta_{km}.
\]
This is a biorthogonal family adapted to non-unitary mode mixing, but it is explicitly distinguished from multiple orthogonality because the construction still uses a single measure [1309.4163].

Another direction is provided by non-commutative Laurent bi-orthogonal polynomials. Bao Wang and Shi-Hao Li define a moment pairing
\[
\langle a z^i,b z^j\rangle = a\,m_{i-j}\,b,
\]
construct \(P_n(z)\) and \(Q_n(z)\) by quasideterminants, and derive a non-commutative three-term recurrence. Their recurrence coefficients realize the non-commutative relativistic Toda dynamics underlying the non-commutative leapfrog map, whose integrability is expressed by a discrete zero-curvature equation and a compatible network Poisson structure [2310.01993].

## 6. Distinctions, reductions, and significance

A recurrent source of confusion is the relation among “matrix-valued,” “bi-orthogonal,” and “multiple.” These notions overlap but are not identical. Matrix-valued orthogonal or bi-orthogonal polynomials are non-commutative because coefficients and weights are matrices, yet they need not be multiple. Bi-orthogonal systems involve two dual families or asymmetric pairings, yet they need not involve several functionals. The deformed complex Hermite families provide a clear example of bi-orthogonality without multiple orthogonality [1309.4163], while Sergel’s theory provides noncommutative biorthogonality without a multi-indexed family of constraints [1009.5472].

The formal theory of non-commutative multiple bi-orthogonal polynomials is significant because it subsumes all three aspects in a single algebraic object. Its normalization functions solve non-commutative Hirota equations, its polynomials furnish the associated linear problem, and its Hankel reduction gives a standard route to a multidimensional non-commutative Toda system [2510.02207]. This establishes a direct bridge between quasideterminantal polynomial algebra and discrete integrable hierarchies.

The subject also has concrete analytic and computational ramifications. In matrix-valued settings, non-commutative bi-orthogonal recurrences lead to block Hessenberg operators and to integrable numerical algorithms such as the generalized block qd-algorithm [2404.13492]. In statistical mechanics, the same machinery explains matrix product representations and boundary bases in ASEP [1412.7235; 1902.06373]. In discrete geometry and integrable maps, Laurent bi-orthogonal polynomials encode the leapfrog dynamics and its zero-curvature structure [2310.01993].

The present theory remains explicitly formal in its most general multiple form, being built from formal bi-moments in a free division ring [2510.02207]. This suggests that analytic realizations—measure-theoretic, spectral, or Riemann–Hilbert theoretic—are likely to remain model-dependent. A plausible implication is that future work will proceed by importing the formal multiple framework into concrete non-commutative models, much as matrix orthogonal polynomials have already been connected to IIKS kernels, Riemann–Hilbert problems, and non-commutative Painlevé equations in the orthogonal, rather than bi-orthogonal, setting [1301.2116].

Source: https://www.emergentmind.com/topics/non-commutative-multiple-bi-orthogonal-polynomial-systems