Worked Examples in the Split-Quaternion Algebra

Introduction

This article collects explicit computations in the split-quaternion algebra $\mathbb{H}_{\mathrm{spl}}$, in the order in which the theory is built: the basis products; the three involutions and their eigenspaces; the idempotents and the two minimal left ideals; explicit pairs of zero divisors; and the unit criterion. Every number below is computed from the multiplication table of Split-Quaternion Algebra.

The article is a companion to the structural articles of the category. Its purpose is to put the abstract statements on concrete elements, so that the reader can carry each of them back to a computation. It introduces no new result; the statements it illustrates are those of Split-Quaternion Algebra, Split-Quaternion Idempotents and Projections, Split-Quaternion Zero Divisors, Split-Quaternion Norm and Invertibility and Split-Quaternion Rotations and the Lorentz Group.

The Basis Products

The multiplication table, whose entry in row $\mu$ and column $\nu$ is $e_\mu e_\nu$, is

$1$ $\vec e_x$ $\vec e_y$ $\vec e_z$
$1$ $1$ $\vec e_x$ $\vec e_y$ $\vec e_z$
$\vec e_x$ $\vec e_x$ $-1$ $\vec e_z$ $-\vec e_y$
$\vec e_y$ $\vec e_y$ $-\vec e_z$ $+1$ $-\vec e_x$
$\vec e_z$ $\vec e_z$ $\vec e_y$ $\vec e_x$ $+1$

Example (a product worked in full). The entry $\vec e_z^2 = +1$ follows from $\vec e_z = \vec e_x \vec e_y$ and anticommutativity:

$$ \vec e_z^2 = \vec e_x \vec e_y \vec e_x \vec e_y = -\vec e_x^2 \vec e_y^2 = -(-1)(+1) = +1 . $$

The product of two general elements, worked from the table, is

$$ (q_{1t} e_t + q_{1x} \vec e_x + q_{1y} \vec e_y + q_{1z} \vec e_z)(q_{2t} e_t + q_{2x} \vec e_x + q_{2y} \vec e_y + q_{2z} \vec e_z) = (q_{1t} q_{2t} - q_{1x} q_{2x} + q_{1y} q_{2y} + q_{1z} q_{2z}) + (q_{1t} q_{2x} + q_{2t} q_{1x} - q_{1y} q_{2z} + q_{2y} q_{1z})\vec e_x + (q_{1t} q_{2y} + q_{2t} q_{1y} + q_{1x} q_{2z} - q_{2x} q_{1z})\vec e_y + (q_{1t} q_{2z} + q_{2t} q_{1z} + q_{1x} q_{2y} - q_{2x} q_{1y})\vec e_z . $$

The product is associative and distributive, so the multiplication table of the four basis elements determines it completely; the table is collected in the summary. The products of the vector basis elements are

$$ \boxed{\; \vec e_x^2 = -1, \quad \vec e_y^2 = 1, \quad \vec e_z^2 = 1, \qquad \vec e_x \vec e_y = \vec e_z, \quad \vec e_y \vec e_x = -\vec e_z, \qquad \vec e_x \vec e_z = -\vec e_y, \quad \vec e_z \vec e_x = \vec e_y, \qquad \vec e_y \vec e_z = -\vec e_x, \quad \vec e_z \vec e_y = \vec e_x \;}. $$

Example. For $\tilde q = 2 + \vec e_x$ and $\tilde p = 1 + \vec e_y + \vec e_z$, the table gives

$$ (2 + \vec e_x)(1 + \vec e_y + \vec e_z) = 2 + 2\vec e_y + 2\vec e_z + \vec e_x + \vec e_x \vec e_y + \vec e_x \vec e_z = 2 + \vec e_x + 2\vec e_y + 2\vec e_z - \vec e_y + \vec e_z = 2 + \vec e_x + \vec e_y + 3\vec e_z , $$

so the product is $\tilde q \tilde p = 2 + \vec e_x + \vec e_y + 3\vec e_z$; the coefficient of $\vec e_x$ also follows from the general product formula displayed above.

The Three Involutions and Their Eigenspaces

Three involutions act on the algebra. The conjugation ${}^{\natural}$ negates the vector part, the principal involution $\alpha$ is the automorphism

$$ \alpha(\vec e_x) = -\vec e_x, \qquad \alpha(\vec e_y) = -\vec e_y, \qquad \alpha(\vec e_z) = \vec e_z, $$

and the reversal $\rho$ is the anti-automorphism

$$ \rho(\vec e_x) = \vec e_x, \qquad \rho(\vec e_y) = \vec e_y, \qquad \rho(\vec e_z) = -\vec e_z . $$

On a general element they read

$$ \tilde{q}^{\natural} = q_t e_t - q_x \vec e_x - q_y \vec e_y - q_z \vec e_z, \qquad \alpha(\tilde q) = q_t - q_x \vec e_x - q_y \vec e_y + q_z \vec e_z, \qquad \rho(\tilde q) = q_t + q_x \vec e_x + q_y \vec e_y - q_z \vec e_z, $$

and the three are related by $\tilde{q}^{\natural} = \alpha(\rho(\tilde q)) = \rho(\alpha(\tilde q))$.

Example (eigenspaces). Each involution is diagonalisable with eigenvalues $\pm 1$.

involution $+1$ eigenspace $-1$ eigenspace dimensions
${}^{\natural}$ $S = \mathbb{R}\cdot 1$ $V = \operatorname{span}\{\vec e_x,\vec e_y,\vec e_z\}$ $1, 3$
$\alpha$ $\operatorname{span}\{1, \vec e_z\}$ $\operatorname{span}\{\vec e_x, \vec e_y\}$ $2, 2$
$\rho$ $\operatorname{span}\{1, \vec e_x, \vec e_y\}$ $\mathbb{R} \vec e_z$ $3, 1$

Example (the common eigenspaces). Since $\alpha$ and $\rho$ commute, they are simultaneously diagonalisable, and the common eigenspaces of the pair $(\alpha, \rho)$ are

$$ \mathbb{H}_{\mathrm{spl}} = \mathbb{R}\cdot 1 \oplus \operatorname{span}\{\vec e_x, \vec e_y\} \oplus \mathbb{R} \vec e_z, $$

on which the sign pair $(\alpha, \rho)$ is $(+,+)$, $(-,+)$ and $(+,-)$ respectively. The plane $\operatorname{span}\{\vec e_x,\vec e_y\}$ is not split further: on it the conjugation also acts as $-1$, so no eigenspace of any of the three involutions cuts it. This is the finest decomposition produced by the involutions, and it is recorded in Split-Quaternion Subspaces and the Involutions.

The Idempotents and the Two Minimal Left Ideals

Example (the standard idempotents). With $\tilde\pi_1 = \tfrac{1}{2}(1 + \vec e_y)$ and $\tilde\pi_2 = \tfrac{1}{2}(1 - \vec e_y)$,

$$ \tilde\pi_1^2 = \tfrac{1}{4}(1 + 2\vec e_y + \vec e_y^2) = \tfrac{1}{4}(2 + 2\vec e_y) = \tilde\pi_1, $$

and $\tilde\pi_2^2 = \tilde\pi_2$ likewise. Moreover

$$ \boxed{\; \tilde\pi_1 \tilde\pi_2 = \tfrac{1}{4}(1 - \vec e_y^2) = 0, \qquad \tilde\pi_1 + \tilde\pi_2 = 1, \qquad \langle\tilde\pi_1\rangle_{\natural} = \langle\tilde\pi_2\rangle_{\natural} = 0 \;}. $$

They are not central, since $\vec e_x \tilde\pi_1 = \tfrac{1}{2}(\vec e_x + \vec e_z)$ while $\tilde\pi_1 \vec e_x = \tfrac{1}{2}(\vec e_x - \vec e_z)$.

Example (the two minimal left ideals). The reductions $\vec e_y \tilde\pi_1 = \tilde\pi_1$ and $\vec e_z \tilde\pi_1 = \tfrac{1}{2}(\vec e_z + \vec e_x) = \vec e_x \tilde\pi_1$ give

$$ \mathbb{H}_{\mathrm{spl}} \tilde\pi_1 = \operatorname{span}\{\tilde\pi_1, \vec e_x \tilde\pi_1\} = \operatorname{span}\big\{\tfrac{1}{2}(1 + \vec e_y),\, \tfrac{1}{2}(\vec e_x + \vec e_z)\big\}, $$

$$ \mathbb{H}_{\mathrm{spl}} \tilde\pi_2 = \operatorname{span}\{\tilde\pi_2, \vec e_x \tilde\pi_2\} = \operatorname{span}\big\{\tfrac{1}{2}(1 - \vec e_y),\, \tfrac{1}{2}(\vec e_x - \vec e_z)\big\}, $$

each of real dimension $2$, with $\mathbb{H}_{\mathrm{spl}} = \mathbb{H}_{\mathrm{spl}} \tilde\pi_1 \oplus \mathbb{H}_{\mathrm{spl}} \tilde\pi_2$. The four spanning elements are the vectors $\tilde\pi_1$, $\tilde p = \tfrac{1}{2}(\vec e_x+\vec e_z)$, $\tilde\pi_2$, $-\tilde q = \tfrac{1}{2}(\vec e_x-\vec e_z)$ of the matrix-unit basis $\{\tilde\pi_1, \tilde q, \tilde p, \tilde\pi_2\}$ of Split-Quaternion Ideals and Peirce Decomposition.

Example (the split-complex subalgebra). The idempotents lie in the subalgebra $\operatorname{span}\{1, \vec e_y\} \cong \mathbb{C}_{\mathrm{spl}}$, and in that commutative algebra they are central; in $\mathbb{H}_{\mathrm{spl}}$ they are not, and the decomposition they give is a decomposition of modules and not of algebras, because $\tilde\pi_1 \vec e_z \tilde\pi_2 = \tfrac{1}{2}(\vec e_z - \vec e_x) \neq 0$.

Explicit Zero-Divisor Pairs

Example (two split-complex pairs). Let $\lambda, \mu$ be real. Then

$$ (1 + \vec e_y)(1 - \vec e_y) = 1 - \vec e_y^2 = 0, \qquad (1 + \vec e_z)(1 - \vec e_z) = 1 - \vec e_z^2 = 0, $$

so $(1+\vec e_y)$ with $(1-\vec e_y)$, and $(1+\vec e_z)$ with $(1-\vec e_z)$, are explicit pairs of nonzero zero divisors.

Example (a nilpotent pair). The element $\vec e_x + \vec e_z$ is nonzero and

$$ (\vec e_x + \vec e_z)^2 = \vec e_x^2 + \vec e_x \vec e_z + \vec e_z \vec e_x + \vec e_z^2 = -1 + (-\vec e_y) + \vec e_y + 1 = 0, $$

so $\vec e_x + \vec e_z$ is a nonzero nilpotent and the pair $(\vec e_x+\vec e_z, \vec e_x+\vec e_z)$ is a zero-divisor pair.

Every nilpotent lies in the vector subspace $V$ and on the level set $N = 0$, $q_x^2 = q_y^2 + q_z^2$; the element $1 + \vec e_y$ shows that the zero divisor set is strictly larger than the nilpotent set, since $(1+\vec e_y)^2 = 2(1+\vec e_y) \neq 0$.

Example (a mixed pair). The idempotent $\tilde\pi_2$ annihilates $1 + \vec e_y$:

$$ (1 + \vec e_y) \tilde\pi_2 = 2 \tilde\pi_1 \tilde\pi_2 = 0, $$

a pair in which neither factor is a nilpotent.

The Invertibility Criterion Worked

Criterion (from Split-Quaternion Norm and Invertibility). A nonzero $\tilde q$ is invertible if and only if $\langle\tilde q\rangle_{\natural} \neq 0$, and then $\tilde q^{-1} = \tilde{q}^{\natural}/\langle\tilde q\rangle_{\natural}$. It is recalled here only to be applied.

Example (a unit of positive norm). For $\tilde q = 2 + \vec e_x$, the split-quaternion norm is $\langle\tilde q\rangle_{\natural} = 4 + 1 = 5$, and

$$ \boxed{\; \tilde q^{-1} = \frac{\tilde{q}^{\natural}}{\langle\tilde q\rangle_{\natural}} = \frac{2 - \vec e_x}{5}, \qquad (2 + \vec e_x)\frac{2 - \vec e_x}{5} = \frac{4 - \vec e_x^2}{5} = \frac{5}{5} = 1 \;}. $$

Example (a unit of norm one). For $\tilde p = 1 + \vec e_x + \vec e_y$, the split-quaternion norm is $\langle\tilde p\rangle_{\natural} = 1 + 1 - 1 = 1$, so $\tilde p^{-1} = \tilde p^{\natural} = 1 - \vec e_x - \vec e_y$; indeed $\tilde p \tilde p^{\natural} = (1+\vec e_x+\vec e_y)(1-\vec e_x-\vec e_y) = 1$. Since $\langle\tilde p\rangle_{\natural} = 1$, the element is invertible split-quaternion.

Example (a unit of negative norm). For $\tilde q_1 = \vec e_y$, the split-quaternion norm is $\langle\tilde q_1\rangle_{\natural} = -1$, and $\tilde q_1^{-1} = \bar{\tilde q_1}/\langle\tilde q_1\rangle_{\natural} = (-\vec e_y)/(-1) = \vec e_y$, consistent with $\vec e_y^2 = 1$.

Example (a non-invertible). For $w = 1 + \vec e_z$, the split-quaternion norm is $\langle w\rangle_{\natural} = 1 - 1 = 0$, so $w$ is not invertible; and indeed $(1+\vec e_z)(1-\vec e_z) = 0$.

The four examples realise the three-way classification of Split-Quaternion Norm and Invertibility by the value of $\langle\cdot\rangle_{\natural}$: a positive value ($N = 5$ and $N = 1$), a negative value ($N = -1$), and the value zero ($N = 0$, the zero divisors).

The Action of the Group of Invertibles

The action of an invertible element on the vector subspace by conjugation, the elliptic and hyperbolic one-parameter subgroups it produces, and the classification of the vectors of $V$ by their orbits, use the multiplication table above and nothing else; they are computations of the Geometry group and are carried out in Split-Quaternion Rotations and the Lorentz Group. They are not repeated here, and the multiplication table of §The Basis Products is what they consume.

Summary

The basis products are collected in one table, and the multiplication table determines the algebra. The three involutions — conjugation, the principal involution $\alpha$ and the reversal $\rho$ — are diagonalisable; their individual eigenspaces and their common refinement $\mathbb{R}\cdot 1 \oplus \operatorname{span}\{\vec e_x,\vec e_y\} \oplus \mathbb{R} \vec e_z$ are listed, and the plane $\operatorname{span}\{\vec e_x,\vec e_y\}$ does not split further.

The idempotents $\tilde\pi_{1,2}$ are verified to be orthogonal, complete and of zero norm, and they give the two minimal left ideals $\mathbb{H}_{\mathrm{spl}} \tilde\pi_{1,2} = \operatorname{span}\{\tilde\pi_{1,2}, \vec e_x \tilde\pi_{1,2}\}$. Explicit zero-divisor pairs are $(1+\vec e_y)(1-\vec e_y) = 0$, $(1+\vec e_z)(1-\vec e_z) = 0$, the nilpotent $(\vec e_x+\vec e_z)^2 = 0$, and the mixed pair $(1+\vec e_y)\tilde\pi_2 = 0$. The unit criterion is carried through on four elements: $2+\vec e_x$ (positive norm), $1+\vec e_x+\vec e_y$ (norm one), $\vec e_y$ (negative norm), and the non-invertible $1+\vec e_z$ (zero norm). The action of the group of invertibles on the vector subspace is not worked here; it is the subject of Split-Quaternion Rotations and the Lorentz Group.

Summary of Notation

Symbol Meaning Article
$\mathbb{H}_{\mathrm{spl}}$ the split-quaternion algebra Split-Quaternion Algebra
$\tilde q = q_t e_t + q_x \vec e_x + q_y \vec e_y + q_z \vec e_z$ a general split-quaternion Split-Quaternion Algebra
${}^{\natural}$, $\alpha$, $\rho$ conjugation, principal involution, reversal Split-Quaternion Algebra
$\tilde\pi_{1,2} = \tfrac{1}{2}(1\pm \vec e_y)$ the standard idempotents Split-Quaternion Idempotents and Projections
$\mathbb{H}_{\mathrm{spl}} \tilde\pi_{1,2}$ the two minimal left ideals Split-Quaternion Idempotents and Projections
$\langle\tilde q\rangle_{\natural} = \tilde q\tilde{q}^{\natural} = q_t^2+q_x^2-q_y^2-q_z^2$ the central product, formed and evaluated algebraically Split-Quaternion Algebra
$\tilde q^{-1} = \tilde{q}^{\natural}/\langle\tilde q\rangle_{\natural}$ the inverse of an invertible element Split-Quaternion Norm and Invertibility

Further Reading

  • Pertti Lounesto, Clifford Algebras and Spinors, 2nd ed. (Cambridge University Press, 2001), for worked computations with the coquaternions and their matrix models.
  • John H. Conway and Derek A. Smith, On Quaternions and Octonions (A K Peters, 2003), for the explicit idempotents, zero divisors and split forms.
  • Ian R. Porteous, Clifford Algebras and the Classical Groups (Cambridge University Press, 1995), for the action of $\mathrm{SL}_2(\mathbb{R})$ on the $(2,1)$-form and the Lorentz group.
  • T. Y. Lam, Introduction to Quadratic Forms over Fields (American Mathematical Society, 2005), for the isotropic forms and the geometry of their null cones.