Basis-free solution to general linear quaternionic equation

Changpeng Shao, Hongbo Li, Lei Huang

TL;DR

This paper presents the first general, "basis-free" solution to linear quaternionic equations of the form $\sum a_i q b_i = c$. By leveraging Clifford algebra embeddings, the authors derive an explicit formula for the variable $q$ using only the input coefficients and their conjugates, successfully generalizing a problem that has remained partially solved since Sylvester first tackled it in 1884.

What Problem Does This Paper Address?

Quaternions are four-dimensional numbers widely used in computer graphics, robotics, and physics to represent rotations and orientations. A common task is solving a linear equation where the unknown is a quaternion $q$.

The "easy" way to solve this is to break $q$ and the coefficients into their four real components (using the basis $1, \mathbf{i, j, k}$) and solve a system of four real linear equations. However, this "basis-dependent" approach is algebraically messy and loses the geometric intuition of the quaternionic structure.

The "hard" way—and the goal of this paper—is to find a basis-free solution. This is a single formula for $q$ expressed only in terms of the coefficients ($a_i, b_i, c$) and their conjugates. While solutions for equations with two or three terms (like the Sylvester equation $aq + qb = c$) have existed for over a century, a general solution for an arbitrary number of terms remained elusive until now.

Methods

The authors employ two distinct mathematical strategies to bridge the gap between real-valued matrix algebra and quaternionic logic:

  1. The Matrix Adjugate Approach: They first convert the quaternionic equation into a $4 \times 4$ real linear system. By calculating the determinant and the adjugate matrix of this system, they obtain a raw solution. The "magic" happens when they use brute-force symbolic manipulation to map these massive real-valued polynomials (the determinant alone has 196 terms!) back into compact quaternionic expressions.
  2. The Clifford Algebra "Lift": To provide a more elegant proof, the authors embed the algebra of quaternions into a higher-dimensional Clifford algebra $CL(\mathbb{R}^4)$. They "lift" the quaternionic equation into this Clifford space, solve it using geometric algebra identities, and then project the result back down to the quaternionic domain. This method explains why certain complex terms appear in the solution and provides a much cleaner derivation.

Datasets

Not discussed in the paper. As a theoretical work in pure and applied mathematics, the paper relies on algebraic proofs and symbolic verification rather than empirical datasets.

Key Insights and Ideas

  • Operator Notation: The authors use a clever notation $(u|v)q := uqv$ to treat left and right quaternionic multiplications as linear operators. This allows them to treat the entire equation as a single operator acting on $q$.
  • The Power of Embedding: The most significant "aha moment" is the realization that solving the problem in the standard quaternionic space is unnecessarily difficult. By moving to $CL(\mathbb{R}^4)$, the complex interactions between coefficients become manageable geometric operations.
  • Handling Conjugates: The paper demonstrates that once you can solve the general linear equation, solving equations that include the conjugate $\overline{q}$ (e.g., $\sum a_i q b_i + \sum c_j \overline{q} d_j = e$) becomes a straightforward extension.

Key Results

  • The General Formula: The paper provides an explicit, closed-form quaternionic rational function for $q$ (Equation 4.11). This formula involves summations of coefficients, their conjugates, and "bracket" terms that represent the determinant of the system.
  • Determinant Identity: They derived a basis-free expression for the determinant of the associated $4 \times 4$ real system (Theorem 4.1), which is essential for determining if a unique solution exists (the "non-degenerate" case).
  • Unified Framework: The solution holds for any number of terms $n$, effectively closing a chapter of quaternionic algebra that began in the 19th century.

Limitations and Future Work

  • Non-degenerate Constraint: The solution assumes the associated real linear system is non-singular (i.e., the determinant is non-zero). The authors do not deeply explore "degenerate" cases where there might be infinitely many solutions or no solution at all.
  • Computational Complexity: While the solution is "explicit," the number of terms in the formula grows as the number of terms in the original equation increases. For very large systems, the basis-dependent (numerical) approach might still be more computationally efficient, even if it is less mathematically "pure."
  • Extension to Higher Dimensions: A natural next step would be investigating whether similar basis-free techniques can be applied to octonions or other higher-dimensional non-commutative algebras.

Summary updated

Abstract

A linear quaternionic equation in one quaternionic variable q is of the form a1qb1+a2qb2+⋯+amqbm=c, where the ai,bj,c are given quaternionic coefficients. If introducing basis elements i,j,k of pure quaternions, then the quaternionic equation becomes four linear equations in four unknowns over the reals, and solving such equations is trivial. On the other hand, finding a quaternionic rational function expression of the solution that involves only the input quaternionic coefficients and their conjugates, called a basis-free solution, is non-trivial.In 1884, Sylvester initiated the study of basis-free solution to linear quaternionic equation. He considered the three-termed equation aq+qb=c, and found its solution q=(a2+bb¯+a(b+b¯))−1(ac+cb¯) by successive left and right multiplications. In 2013, Schwartz extended the technique to the four-termed equation, and obtained the basis-free solution in explicit form.This paper solves the general problem for arbitrary number of terms in the non-degenerate case.
BibTeX
@inproceedings{shao2018basis,
  author = {Changpeng Shao and Hongbo Li and Lei Huang},
  title = {{Basis-free solution to general linear quaternionic equation}},
  booktitle = {Linear},
  year = {2018},
  doi = {10.1080/03081087.2018.1508404},
}

Loading related content...

Figure