[Paper Review] Differential properties of matrix orthogonal polynomials

Maŕıa José Cantero, Leandro del Moral Ituarte+1 more|ArXiv.org|May 9, 2002

Mathematical functions and polynomialsMathematics11 references21 citations

TL;DR

This paper develops a general theory of semi-classical matrix orthogonal polynomials by characterizing them via a distributional equation $D(uA) = uB$, where $A$ and $B$ are matrix polynomials. It establishes equivalences between quasi-orthogonality of derivatives, structure relations, and second-order differo-differential equations, extending scalar semi-classical theory to the matrix case and constructing non-diagonalizable examples with explicit Rodrigues-type formulas and differential equations.

ABSTRACT

In this paper a general theory of semi-classical matrix orthogonal polynomials is developed. We define the semi-classical linear functionals by means of a distributional equation $D(u A) = u B,$ where $A$ and $B$ are matrix polynomials. Several characterizations for these semi-classical functionals are given in terms of the corresponding (left) matrix orthogonal polynomials sequence. They involve a quasi-orthogonality property for their derivatives, a structure relation and a second order differo-differential equation. Finally we illustrate the preceding results with some non-trivial examples.

Motivation & Objective

• To extend the theory of semi-classical orthogonal polynomials from the scalar to the matrix case.
• To characterize matrix orthogonal polynomials through quasi-orthogonality of their derivatives, structure relations, and second-order differo-differential equations.
• To construct non-diagonalizable semi-classical matrix functionals and demonstrate their differential properties.
• To generalize the scalar semi-classical framework to matrix orthogonal polynomials using distributional equations of Pearson type.
• To provide explicit examples of matrix orthogonal polynomials satisfying Rodrigues-type formulas and differential equations.

Proposed method

• Define semi-classical linear functionals via the distributional equation $D(uA) = uB$, where $A$ and $B$ are matrix polynomials.
• Introduce and analyze quasi-orthogonality for matrix polynomials as a key characterization tool.
• Derive a structure relation for matrix orthogonal polynomials by linking it to the distributional equation.
• Establish a second-order differo-differential equation satisfied by the matrix orthogonal polynomials.
• Use Rodrigues-type formulas to construct explicit matrix orthogonal polynomial sequences.
• Verify orthogonality and differential properties through integration by parts and matrix weight analysis.

Experimental results

Research questions

• RQ1 How can the concept of semi-classical orthogonal polynomials be generalized from the scalar to the matrix case?
• RQ2 What are the equivalent characterizations of matrix semi-classical functionals in terms of derivative quasi-orthogonality, structure relations, and differo-differential equations?
• RQ3 Can non-diagonalizable matrix orthogonal polynomial sequences be constructed, and what differential equations do they satisfy?
• RQ4 What conditions ensure that a matrix weight function gives rise to a semi-classical functional via a Pearson-type equation?
• RQ5 How do Rodrigues-type formulas and differential equations relate in the matrix orthogonal polynomial setting?

Key findings

• The matrix orthogonal polynomial sequence $ (P_n) $ defined by $ P_n(x) = (-2)^{-n} e^{x^2} S^{-1}(x) \frac{d^n}{dx^n} (e^{-x^2} S(x)) $ satisfies a structure relation $ P_n(x) = n P_{n-1}(x) $, confirming its semi-classical nature.
• The sequence $ (P_n) $ satisfies the second-order differo-differential equation $ P_n''(x) + B^t(x) P_n'(x) = -2n P_n(x) $, where $ B(x) = \begin{pmatrix} -2x & 1 \\ 0 & -2x \end{pmatrix} $, and is orthogonal with respect to the non-positive definite weight $ u = e^{-x^2} R(x) $, $ R(x) = \begin{pmatrix} 0 & 1 \\ 1 & x \end{pmatrix} $.
• The matrix weight $ u = (1 - x^2) \begin{pmatrix} 1 & x \\ x & 2 - x^2 \end{pmatrix} $ on $[-1,1]$ satisfies a Pearson-type equation $ D(u(1 - x^2)I) = uB(x) $ with $ B(x) = \begin{pmatrix} -9x & 2 + x^2 \\ 1 & -11x \end{pmatrix} $, and generates a non-diagonalizable semi-classical functional.
• The matrix orthogonal polynomials associated with this weight satisfy a second-order differo-differential equation of the form $ (1 - x^2)^2 P_n''(x) - x(1 - 2x) P_n'(x) = \sum_{k=-4}^{2} \Lambda_{nk} P_{n+k} $, confirming the general theory.
• The functional defined by $ u = e^{-x^2} R(x) $ is quasi-definite but not positive definite, and the corresponding orthogonal polynomial sequence is shown to satisfy the required differential and structural properties via integration by parts and Rodrigues-type construction.
• The paper proves that the three characterizations—quasi-orthogonality of derivatives, structure relation, and second-order differo-differential equation—are equivalent to the distributional equation $ D(uA) = uB $, extending scalar semi-classical theory to the matrix case.