Stability of quaternion matrix polynomials

Basic Information

• Paper ID: 2407.16603
• Title: Stability of quaternion matrix polynomials
• Authors: Pallavi Basavaraju, Shrinath Hadimani, Sachindranath Jayaraman
• Classification: math.SP (Spectral Theory)
• Publication Date: July 2024 (arXiv preprint, latest version October 2025)
• Paper Link: https://arxiv.org/abs/2407.16603

Abstract

This paper investigates the stability of right quaternion matrix polynomials $P(\lambda) = \sum_{i=0}^{m} A_i \lambda^i$, where $A_i$ are $n \times n$ quaternion matrices and $A_m \neq 0$. The article aims to determine the location of right eigenvalues of $P(\lambda)$ relative to certain subsets of the quaternion set. The main contributions include: (1) establishing (hyper)stability relationships between quaternion matrix polynomials and their complex companion matrix polynomials; (2) proving that stability of quaternion balls centered at complex numbers is equivalent to stability of their intersection with the complex set; (3) proving that right eigenvalues lie between two concentric spheres of specific radii near the origin. As an application, a generalization of the Eneström-Kakeya theorem to quaternion matrix polynomials is provided.

Research Background and Motivation

Problem Background

1. Importance of polynomial root localization: Although the fundamental theorem of algebra establishes the existence and number of roots of complex coefficient polynomials, algebraic methods become intractable for polynomials of degree exceeding four. Therefore, determining the region containing roots is crucial for iterative methods.
2. Challenges in matrix polynomials: For complex matrix polynomials, eigenvalues equal the zeros of the determinant (scalar complex polynomial). However, when the coefficient matrices are large-scale, computing determinants becomes difficult, making eigenvalue localization for matrix polynomials an interesting research problem.
3. Non-commutativity of quaternions: Quaternion matrix polynomials present new challenges due to the non-commutativity of quaternion multiplication. The definition of quaternion determinants differs from the complex case, and distinctions exist between left and right eigenvalues.
4. Limitations of existing methods: Existing stability theories for complex matrix polynomials cannot be directly extended to the quaternion case, requiring specialized theories and methods.

Research Motivation

• Extend the stability concept of complex matrix polynomials to quaternion matrix polynomials
• Establish connections with the complex case by utilizing the special structure of quaternions (expressible as extensions of complex numbers)
• Provide effective methods for eigenvalue localization of quaternion matrix polynomials

Core Contributions

1. Defined stability and hyperstability concepts for quaternion matrix polynomials and proved their equivalence with eigenvalue location properties
2. Established correspondence relationships between the stability of quaternion matrix polynomials and their complex companion matrix polynomials (Theorem 4.8)
3. Proved a simplification theorem for stability of quaternion balls centered at complex numbers (Theorem 4.12), reducing quaternion stability problems to the complex case
4. Provided annular region localization results for eigenvalues of quaternion matrix polynomials (Theorem 4.19), analogous to the Higham-Tisseur results for complex matrix polynomials
5. Generalized the Eneström-Kakeya theorem to quaternion matrix polynomials (Theorem 4.23)
6. Identified classes of quaternion matrix polynomials where stability and hyperstability are equivalent
7. Derived hyperstability in the univariate case from stability of multivariate quaternion matrix polynomials

Detailed Methodology

Basic Definitions

Quaternion matrix polynomial: A right quaternion matrix polynomial is defined as $P(\lambda) = \sum_{i=0}^{m} A_i \lambda^i$ where $A_i \in M_n(\mathbb{H})$, $A_m \neq 0$, and the variable $\lambda$ appears on the right side of the matrix coefficients.

Right eigenvalue: $\lambda_0 \in \mathbb{H}$ is a right eigenvalue of $P(\lambda)$ if there exists a nonzero vector $y \in \mathbb{H}^n$ such that $\sum_{i=0}^{m} A_i y \lambda_0^i = 0$

Stability Definitions

Stability (Definition 4.1): $P(\lambda)$ is stable with respect to $\Omega \subseteq \mathbb{H}$ if for any nonzero vector $y \in \mathbb{H}^n$ and any $\mu \in \Omega$, there exists a nonzero vector $z \in \mathbb{H}^n$ such that $z^* A_m y \mu^m + z^* A_{m-1} y \mu^{m-1} + \cdots + z^* A_1 y \mu + z^* A_0 y \neq 0$

Hyperstability (Definition 4.2): $P(\lambda)$ is hyperstable with respect to $\Omega$ if for any nonzero vector $y \in \mathbb{H}^n$, there exists a nonzero vector $z \in \mathbb{H}^n$ such that the above inequality holds for all $\mu \in \Omega$.

Complex Companion Matrix

Complex companion matrix (Definition 4.6): For $A \in M_n(\mathbb{H})$, expressible as $A = A_1 + A_2 j$ where $A_1, A_2 \in M_n(\mathbb{C})$, the complex companion matrix is defined as $\chi_A := \begin{bmatrix} A_1 & A_2 \\ -\overline{A_2} & \overline{A_1} \end{bmatrix}$

Complex companion matrix polynomial (Definition 4.7): $P_\chi(\lambda) = \sum_{i=0}^{m} \chi_{A_i} \lambda^i$

Key Theorems

Theorem 4.8: $P(\lambda)$ is stable (hyperstable) with respect to $\Omega \cap \mathbb{C}$ if and only if its complex companion matrix polynomial $P_\chi(\lambda)$ is stable (hyperstable) with respect to $\Omega \cap \mathbb{C}$.

Theorem 4.12: Let $\Omega = B(a;r)$ be an open quaternion ball centered at $a \in \mathbb{C}$ with radius $r > 0$. Then $P(\lambda)$ is stable with respect to $\Omega$ if and only if $P(\lambda)$ is stable with respect to $\Omega \cap \mathbb{C}$.

Experimental Setup

Theoretical Verification

This paper is primarily theoretical research, with results verified through mathematical proofs. Multiple concrete examples are provided to illustrate theorem applications:

Example 4.9: Demonstrates the distinction between stability and hyperstability Example 4.10: Shows that stability does not necessarily imply hyperstability Example 4.21: Verifies that the bounds given in Theorems 4.14 and 4.18 are tight

Application Verification

The validity of the methods is verified through:

1. Comparison with known results for complex matrix polynomials
2. Generalization of classical theorems (such as the Eneström-Kakeya theorem)
3. Construction of concrete examples to verify theoretical predictions

Experimental Results

Main Results

Eigenvalue localization (Theorem 4.19): For quaternion matrix polynomials with invertible leading and constant coefficient matrices, the right eigenvalues lie in the annular region $\{q \in \mathbb{H} : r \leq |q| \leq R\}$, where:

• $r$ is the unique positive zero of $\|A_m\|z^m + \cdots + \|A_1\|z - \|A_0^{-1}\|^{-1}$
• $R$ is the unique positive zero of $\|A_m^{-1}\|^{-1}z^m - \|A_{m-1}\|z^{m-1} - \cdots - \|A_0\|$

Eneström-Kakeya generalization (Theorem 4.23): For quaternion matrix polynomials with coefficients satisfying $A_m \succeq A_{m-1} \succeq \cdots \succeq A_0 \succ 0$, each right eigenvalue $\lambda_0$ satisfies $\frac{\lambda_{\min}(A_0)}{2\lambda_{\max}(A_m)} \leq |\lambda_0| \leq 1$

Equivalence Results

Theorem 4.28: For upper triangular quaternion matrix polynomials with leading coefficient equal to the identity matrix, stability and hyperstability are equivalent.

Multivariate Extension

Theorem 4.35: Provides sufficient conditions for deriving hyperstability in the univariate case from stability of multivariate quaternion matrix polynomials.

Related Work

Complex Matrix Polynomials

• Higham and Tisseur 15: Pioneering work providing eigenvalue bounds through associated block matrices and scalar polynomials
• Szymański and Wojtylak 31: Stability theory for matrix polynomials, the direct reference for this paper's extensions

Quaternion Matrix Theory

• Zhang 33: Foundational theory of quaternion matrices and complex companion matrices
• Ahmad et al. 2,3,4: Eigenvalue localization work using classical methods such as Geršgorin's theorem
• Rodman 28: Monograph on systematic theory of quaternion linear algebra

Polynomial Root Theory

• Classical Eneström-Kakeya theorem and its matrix generalizations
• Various eigenvalue localization and perturbation theories

Conclusions and Discussion

Main Conclusions

1. Successfully extended the stability theory of complex matrix polynomials to the quaternion case, overcoming technical difficulties arising from non-commutativity
2. Established a bridge between quaternion and complex cases by transforming quaternion problems into complex problems through complex companion matrices
3. Provided practical eigenvalue localization methods that avoid direct computation of quaternion determinants
4. Identified important special cases, such as the equivalence of stability and hyperstability for upper triangular matrix polynomials

Limitations

1. Primary focus on right eigenvalues: Due to incomplete theory of left eigenvalues, the paper primarily addresses right eigenvalues and right quaternion matrix polynomials
2. Restrictions of Theorem 4.12: Requires the ball center to be a complex number; does not hold for balls centered at general quaternions (as shown in Remark 4.13)
3. Conservatism of certain results: The provided bounds may not be optimal, although Example 4.21 shows that in certain cases the bounds are tight
4. Limitations in multivariate cases: Due to non-commutativity, some results for multivariate quaternion matrix polynomials cannot be fully generalized

Future Directions

1. Development of left eigenvalue theory: Improve the theory of left eigenvalues for quaternion matrices
2. More general localization results: Study balls with non-complex centers or other geometric regions
3. Computational methods: Develop efficient algorithms for computing eigenvalues of quaternion matrix polynomials
4. Application extensions: Apply the theory to signal processing, computer graphics, and other quaternion application domains

In-Depth Evaluation

Strengths

1. Strong theoretical innovation: First systematic generalization of matrix polynomial stability theory to the quaternion case, filling a theoretical gap
2. Clever methodology: Establishes connections between quaternions and complex numbers through complex companion matrices, skillfully leveraging existing complex theory
3. Good completeness of results: Provides not only main theorems but also abundant applications and analysis of special cases
4. Rigorous technical treatment: Adequately addresses technical difficulties arising from quaternion non-commutativity with detailed and rigorous proofs
5. High practical value: The provided eigenvalue localization methods avoid computing quaternion determinants and have practical application value

Weaknesses

1. Limited application verification: As pure theoretical research, lacks verification in practical problems
2. Relatively strong restrictions: Restrictions such as requiring the ball center to be a complex number in Theorem 4.12 may cause inconvenience in applications
3. Insufficient integration with numerical methods: How theoretical results translate into efficient numerical algorithms requires further research
4. Missing left eigenvalue theory: The completeness of the theory is somewhat affected by the incomplete theory of left eigenvalues

Impact

1. Significant academic contribution: Provides important theoretical tools for quaternion matrix theory, expected to advance related fields
2. Methodological value: The method of establishing connections through complex companion matrices may be applicable to other non-commutative algebraic structures
3. Broad application prospects: The widespread applications of quaternions in 3D graphics, robotics, quantum mechanics, and other fields give this theory potential practical value

Applicable Scenarios

1. Theoretical research: Quaternion matrix theory, matrix analysis in non-commutative algebra
2. Engineering applications: Quaternion matrix polynomial problems in 3D rotation, robot control, and signal processing
3. Numerical computation: Preprocessing and initial value selection for quaternion matrix polynomial eigenvalue problems

References

The paper cites 33 important references, primarily including:

• 15 N. J. Higham and F. Tisseur: Classical work on eigenvalue bounds for complex matrix polynomials
• 31 O. J. Szymański and M. Wojtylak: Direct reference for matrix polynomial stability
• 33 F. Zhang: Foundational theory of quaternion matrices
• 28 L. Rodman: Monograph on quaternion linear algebra
• 2,3,4 S. S. Ahmad et al.: Related work on quaternion matrix eigenvalue localization