Introduction to the Operators of the Biquaternion Algebra in $M_2(\mathbb{C})$

Introduction

The question. The four general products of $\mathbb{B}$ differ from one another by an insertion into one slot. In the display the element is a complex $2\times2$ matrix $A=\mathsf M_2(\tilde Q)$, and an insertion becomes a matrix operation: the quaternion conjugation becomes a transpose conjugated by the symplectic matrix, and the Hermitian conjugation becomes the conjugate transpose. This article asks what the four products and the four pairings look like as matrix statements, and which of their properties the display decides.

Why it is the same statement. This is the in-depth matrix reading of Introduction to the Operators of the Biquaternion Algebra. The map $\mathsf M_2$ is an isomorphism of complex algebras from $\mathbb{B}$ onto $M_2(\mathbb{C})$, so every product, every pairing and every insertion has one image in the display and one only. The isomorphism makes the statements identical: the four products are the four readings of one matrix product, the group of the readings is the same group, and the sign that the plain product loses is the same sign. Only the writing is new, and it is written out as arrays.

What the article proves. In this order. The display sends the four units to $\mathrm I_2,-i\sigma_1,-i\sigma_2,-i\sigma_3$, and the quaternion conjugation to $A\mapsto\sigma_2A^{\mathsf T}\sigma_2$ while the Hermitian conjugation is $A\mapsto A^{\dagger}$. The four general products are the four matrices $A_1A_2$, $\sigma_2A_1^{\mathsf T}\sigma_2A_2$, $A_1A_2^{\dagger}$ and $\sigma_2A_1^{\mathsf T}\sigma_2A_2^{\dagger}$, and their scalar parts are $\tfrac12$ of the corresponding traces. The readings form the group $\mathbb{Z}_2\times\mathbb{Z}_2$ generated by the two insertions, and the group acts on the four pairings and on their Gram matrices $E,\mathrm I_4,\mathrm I_4,E$. The fourth product alone has a central square. The four operators of $\vec e_x$ are four distinct arrays, and the plain product of two vector units is minus a reflection. The sign is invisible on the centre and on even parameters. The choice of which involution plays the sign is a choice of slot and not of algebra.

What the article does not do. It proves no adjoint identity beyond the matrix form of the four that the products decide, computes no product of a left and a right operator, and states nothing about the two-sided family, the enveloping algebra or the spin group. Those are Adjoint Pair of $L$ and $R$ Operators in $M_2(\mathbb{C})$, Operators Acting on Two Sides in $M_2(\mathbb{C})$ and The Enveloping Operator Algebra of the Biquaternion Algebra and the Bi-module Structure in $M_2(\mathbb{C})$.

Conventions. The algebra is $\mathbb{B}=\mathbb{C}\otimes_{\mathbb{R}}\mathbb{H}$ on $(e_t,\vec e_x,\vec e_y,\vec e_z)$ with $\vec e_k^2=-e_t$ and $\vec e_x\vec e_y=\vec e_z$ cyclically. The display is

$$ \mathsf M_2(\tilde Q)=Q_t\mathrm I_2-i\bigl(Q_x\sigma_1+Q_y\sigma_2+Q_z\sigma_3\bigr), $$

with $\sigma_1,\sigma_2,\sigma_3$ the Pauli matrices. The four units map to $\mathrm I_2,-i\sigma_1,-i\sigma_2,-i\sigma_3$. The trace is $\operatorname{tr}A=2Q_t$ and the determinant is $\det A=N(\tilde Q)=Q_t^2+Q_x^2+Q_y^2+Q_z^2$, both verified below. Write $\sigma_2=\begin{pmatrix}0&-i\\ i&0\end{pmatrix}$ and $J=-i\sigma_2=\begin{pmatrix}0&1\\-1&0\end{pmatrix}$, so that $\sigma_2A^{\mathsf T}\sigma_2=JA^{\mathsf T}J^{-1}$. Any $4\times4$ array in this article is the axis-free array, not the Pauli array: the axis-free family is the unique representative, among the $96$ signed-permutation bases, whose sixteen operator arrays are signed permutations, so its entries are the multiplication table read aloud and every identity checked in it is an integer check ($M_4(\mathbb{C})$ Axis-Free Representation : Operators). The Pauli and gamma arrays are a translation of it and are refused as primary, because one entry of one of their matrices is diagonal by privilege, which puts a spatial direction into the notation.

Boundaries. The abstract four products are the source article. The four forms and their norms are The Four Forms and the Four Algebraic Norms over the Biquaternion Algebra. The involutions as a group are The Group of Involutions of the Biquaternion Algebra. The display itself is $M_2(\mathbb{C})$ Pauli Representation : Operators, whose Conventions paragraph fixes the display. The $n=2$ spine is The Two-Sided Operators and the Segre Cone at $n=2$ and The Symmetric and Antisymmetric Parts of the Operator Space at $n=2$.

Letters. The matrix of $\tilde Q$ is $A$, the matrix of $\tilde Q_2$ is $B$, and their entries are $a_{ij},b_{ij}$. The conjugate transpose is $A^{\dagger}=\bar A^{\mathsf T}$. The symplectic matrix is $J=\begin{pmatrix}0&1\\-1&0\end{pmatrix}$, with $J^2=-\mathrm I_2$. The coefficient column is $[Q]=(Q_t,Q_x,Q_y,Q_z)^{\mathsf T}$. The four readings are labelled (none), $({}^{\natural})$, $({}^{*})$ and (both).

The Display and the Four Readings

The Display as One Matrix

The array. In the Peirce frame the display of an element is

$$ A=\mathsf M_2(\tilde Q) =\begin{pmatrix} Q_t-iQ_z & -iQ_x-Q_y\\ -iQ_x+Q_y & Q_t+iQ_z \end{pmatrix}. $$

The four entries carry the four coordinates linearly, and the inverse is read from them: $Q_t=\tfrac12\operatorname{tr}A$ and $Q_z=\tfrac{i}{2}(a_{11}-a_{22})$.

The invariants. Two numbers exhaust the spectral data at order two:

$$ \operatorname{tr}A=2Q_t,\qquad \det A=Q_t^2+Q_x^2+Q_y^2+Q_z^2=N(\tilde Q). $$

The determinant identity is the display of the norm, and it is the reason the norm is complex and not a real part: the coefficients are complex, the determinant is complex, and no conjugation enters either formula. The two identities were checked on a hundred complex Gaussian elements with residual below $2\cdot10^{-14}$.

The vector square. Write $B^2=Q_x^2+Q_y^2+Q_z^2$. Then

$$ \det A-\tfrac14(\operatorname{tr}A)^2=B^2,\qquad (\operatorname{tr}A)^2-4\det A=-4B^2 . $$

The second is the discriminant of the characteristic polynomial, $\Delta=(\lambda_1-\lambda_2)^2$. The eigenvalues are

$$ \lambda_\pm=Q_t\pm iB, $$

and the pair is distinct exactly when $B^2\neq0$, which is spectral regularity in the display.

The Four Insertions in Matrix Form

The two insertions. The quaternion conjugation is $\mathbb{C}$-linear and the Hermitian conjugation is conjugate-linear. In the display they are

$$ \mathsf M_2(\tilde Q^{\natural})=\sigma_2A^{\mathsf T}\sigma_2=JA^{\mathsf T}J^{-1}, \qquad \mathsf M_2(\tilde Q^{*})=A^{\dagger}. $$

Proof. For the second, $\mathsf M_2$ is complex-linear and $\mathsf M_2(e_k)=-i\sigma_k$, so conjugating the coefficients and negating the three vector coefficients gives $A^{\dagger}$ entry by entry. For the first, the same computation with the coefficients untouched gives $\sigma_2A^{\mathsf T}\sigma_2$; the identity was checked on a hundred elements with zero residual. The sign of $J$ is immaterial, since $J$ enters twice.

The four products. Composing the two insertions independently gives the four matrix products

$$ A_1A_2,\qquad \sigma_2A_1^{\mathsf T}\sigma_2\,A_2,\qquad A_1A_2^{\dagger},\qquad \sigma_2A_1^{\mathsf T}\sigma_2\,A_2^{\dagger}. $$

The table names them.

Reading Product Matrix Scalar part
(none) $\tilde Q_1\tilde Q_2$ $A_1A_2$ $\tfrac12\operatorname{tr}(A_1A_2)$
$({}^{\natural})$ $\tilde Q_1^{\natural}\tilde Q_2$ $\sigma_2A_1^{\mathsf T}\sigma_2A_2$ $\tfrac12\operatorname{tr}(\sigma_2A_1^{\mathsf T}\sigma_2A_2)$
$({}^{*})$ $\tilde Q_1\tilde Q_2^{*}$ $A_1A_2^{\dagger}$ $\tfrac12\operatorname{tr}(A_1A_2^{\dagger})$
(both) $\tilde Q_1^{\natural}\tilde Q_2^{*}$ $\sigma_2A_1^{\mathsf T}\sigma_2A_2^{\dagger}$ $\tfrac12\operatorname{tr}(\sigma_2A_1^{\mathsf T}\sigma_2A_2^{\dagger})$

The reading of the table. The four rules multiply the same two matrices; the difference is a transpose conjugated by $J$ in the first slot, a conjugate transpose in the second, or both. Nothing else changes, and in particular the order of the factors is the same in all four. That is the matrix form of the statement that the four products are one product with insertions.

The Four Pairings as Traces

The pairings. Taking the scalar part, which is $\tfrac12\operatorname{tr}$, gives the four pairings

$$ \langle\tilde Q_1,\tilde Q_2\rangle=\tfrac12\operatorname{tr}(A_1A_2),\quad \langle\tilde Q_1,\tilde Q_2\rangle_{\natural}=\tfrac12\operatorname{tr}\bigl(JA_1^{\mathsf T}J^{-1}A_2\bigr), $$ $$ \langle\tilde Q_1,\tilde Q_2\rangle_{*}=\tfrac12\operatorname{tr}(A_1A_2^{\dagger}),\quad \langle\tilde Q_1,\tilde Q_2\rangle_{\natural*}=\tfrac12\operatorname{tr}\bigl(JA_1^{\mathsf T}J^{-1}A_2^{\dagger}\bigr). $$

The second pairing is the Hilbert–Schmidt form. The identity

$$ \tfrac12\operatorname{tr}(A_1A_2^{\dagger})=\tfrac12\operatorname{tr}(A_2^{\dagger}A_1)=\langle A_2,A_1\rangle_{\mathrm{HS}} $$

identifies $\langle,\rangle_{*}$ with the Hilbert–Schmidt form of the display, the form used to take adjoints in The One-Sided Operators at $n=2$. It is positive definite on the real space of matrices, and its Gram matrix on the four units is $\mathrm I_4$.

The Gram matrices. Evaluated on the units $e_\mu$, the four pairings give

$$ E=\operatorname{diag}(1,-1,-1,-1),\qquad \mathrm I_4,\qquad \mathrm I_4,\qquad E, $$

in the order (none), $({}^{\natural})$, $({}^{*})$, (both). The sign $(1,-1,-1,-1)$ survives exactly when the two insertions agree, which is the pair (none), (both); it is replaced by $1$ when they differ. The computation is an integer check on four units: $\varepsilon_\mu\delta_{\mu\nu}$ with $\varepsilon=(1,-1,-1,-1)$.

The matrix reason for the alternation. The units satisfy $\mathsf M_2(e_\mu)^{\natural}=\varepsilon_\mu\mathsf M_2(e_\mu)$ and $\mathsf M_2(e_\mu)^{*}=\varepsilon_\mu\mathsf M_2(e_\mu)$, because the coordinates of a unit are $0$ or $1$ and the insertions only change their signs. Two insertions contribute $\varepsilon_\mu^2=1$ and cancel the sign; one insertion leaves it. The alternation is therefore a statement about the four basis matrices, and it disappears as soon as a general matrix is paired with another.

The Slot Group on the Display

The Group of the Readings

The four readings. Each insertion is an involution of the set of rules, the two insertions commute because they touch different slots, and their composite is the fourth reading. The four readings therefore form a group of order four and exponent two:

$\circ$ (none) $({}^{\natural})$ $({}^{*})$ (both)
(none) (none) $({}^{\natural})$ $({}^{*})$ (both)
$({}^{\natural})$ $({}^{\natural})$ (none) (both) $({}^{*})$
$({}^{*})$ $({}^{*})$ (both) (none) $({}^{\natural})$
(both) (both) $({}^{*})$ $({}^{\natural})$ (none)

Its matrix form. In the display the group is generated by two matrix operations,

$$ \Phi_{\natural}(A)=\sigma_2A^{\mathsf T}\sigma_2,\qquad \Phi_{*}(A)=A^{\dagger}, $$

and each is an involution of the space of matrices. The first is complex-linear, the second conjugate-linear, and they commute because transposition and entrywise conjugation commute. The group they generate is again $\mathbb{Z}_2\times\mathbb{Z}_2$, and its four elements applied to the identity matrix all give $\mathrm I_2$: the group acts on the rule, not on the value at the unit.

The group is not a group of matrices. The four readings are operations on rules. The four conjugations $\mathrm{id},\natural,\bar{\cdot},{}^{*}$ are operations on elements. The two groups are both elementary abelian of order four and they are different groups, and the display shows why: a reading multiplies two matrices, a conjugation replaces one matrix, and the two actions have different arities.

The Slots and the Table of Properties

The first slot. The first-slot insertion is complex-linear, so a property of the product that does not involve complex conjugation belongs to the first slot alone. In the display this is the statement that $\sigma_2A_1^{\mathsf T}\sigma_2$ is a complex-linear function of $A_1$.

The second slot. The second slot carries a conjugate-linear insertion when it is not the identity. The pairing $({}^{*})$ is therefore sesquilinear and the pairing (none) is bilinear, and no change of basis can turn one into the other.

The table. The properties of the four products sort by slot.

Property Decided by
the order of the matrix factors neither slot: the same in all four
the Gram matrix on the units the parity of the two insertions
bilinear against sesquilinear the second slot alone
the square is central the setting (both) alone
$\operatorname{tr}(A_1A_2)=\operatorname{tr}(A_2A_1)$ the setting (none) alone

The Square at One Setting

The statement. Among the four squares exactly one is a scalar matrix for every element: the square of the general quaternionic sesquilinear reading, the setting (both).

The computation. Using $\sigma_2A^{\dagger}\sigma_2=\overline{\sigma_2A^{\mathsf T}\sigma_2}$ and $\sigma_2\sigma_2=\mathrm I_2$, the middle block of the square collapses. The source identity is that the square of $\tilde Q^{\natural}\tilde Q^{*}$ is the norm times the unit, and the display turns it into

$$ \bigl(\sigma_2A^{\mathsf T}\sigma_2A^{\dagger}\bigr)^{2} =\det(A)\,\mathrm I_2 . $$

The other three squares are not scalar: $A^2$ is scalar only when $A$ is, the square of $({}^{\natural})$ is $\sigma_2(A^{\mathsf T})^2\sigma_2$, and the square of $({}^{*})$ is $AA^{\dagger}$, Hermitian and scalar only when $A$ is normal with a single eigenvalue modulus.

The reading. The setting (both) is the product $\tilde Q^{\natural}\tilde Q^{*}$, whose scalar part is the norm and whose square is central. In the display this is $\det A$ times the identity matrix, and it is the origin of the inverse formula

$$ A^{-1}=\frac{1}{\det A}\,\sigma_2A^{\mathsf T}\sigma_2, $$

which holds exactly when $\det A\neq0$. The inverse is the matrix statement of $\tilde Q^{-1}=\tilde Q^{\natural}/N(\tilde Q)$.

The Twisted Right Factor and the Switch

The two switches. The source builds each reading as a switch pair: a conjugation on the left factor and one on the right. In the display a switch pair is a pair of matrix operations $(\Phi_1,\Phi_2)$ and the product is $\Phi_1(A_1)\Phi_2(A_2)$. The four settings are the four pairs drawn from $\{\mathrm{id},\Phi_{\natural}\}\times\{\mathrm{id},\Phi_{*}\}$.

The closure. The set of switch pairs is closed under composition, because each factor is an involution and the two slots are independent, and the composition is componentwise. The twist on the left factor is the composite $\Phi_{\natural}\Phi_{*}$, which is neither generator: it is entrywise conjugation followed by transposition, that is $A\mapsto\sigma_2\bar A\sigma_2^{-1}$, the matrix form of $\tilde Q\mapsto\bar{\tilde Q}$.

The example at a unit. At $A=\mathrm I_2$ the four products of a fixed $B$ give $B$, $B$, $B^{\dagger}$ and $B^{\dagger}$. The first slot is silent at the identity and the second is not, and the example separates the two slots with one computation.

What the Products Decide

The Four Adjoints

The adjoints by name. Each of the four pairings has its adjoint, and the display names all four.

Pairing Adjoint of $A_1\mapsto A_1A_2$ Name
$\tfrac12\operatorname{tr}(A_1A_2)$ $B\mapsto A_2B$ plain bilinear adjoint $\approx$
$\tfrac12\operatorname{tr}(A_1A_2^{\dagger})$ $B\mapsto BA_2^{\dagger}$ Hermitian adjoint $*$
$\tfrac12\operatorname{tr}(JA_1^{\mathsf T}J^{-1}A_2)$ $B\mapsto JA_2^{\mathsf T}J^{-1}B$ quaternionic adjoint $\natural$
$\tfrac12\operatorname{tr}(JA_1^{\mathsf T}J^{-1}A_2^{\dagger})$ $B\mapsto BA_2^{\mathsf T}$ Krein adjoint $\natural*$

The reading of the table. The two bilinear pairings give adjoints that multiply on the same side, and the two sesquilinear pairings give adjoints that multiply on the other side. The four adjoints are computed in Adjoint Pair of $L$ and $R$ Operators in $M_2(\mathbb{C})$, where the same table is read on the operator families.

What the products do not decide. The four products decide the four adjoints and nothing about an operator. A product is a pairing of two matrices; an operator is a map of the matrix space, and a map needs a matrix beside the rule. The display makes the gap visible: the product $A_1A_2$ is one matrix, while the left multiplication $L_{A_1}(Y)=A_1Y$ is a map of the four-dimensional space, and the two are different objects of different dimensions.

One Element, Four Operators

The four operators of $\vec e_x$. Take $A=\mathsf M_2(\vec e_x)=-i\sigma_1=\begin{pmatrix}0&-i\\-i&0\end{pmatrix}$. The four readings of the element are four matrices,

$$ A,\qquad \sigma_2A^{\mathsf T}\sigma_2=-A,\qquad A^{\dagger}=i\sigma_1=-A,\qquad \sigma_2A^{\mathsf T}\sigma_2A^{\dagger}=A^2=-\mathrm I_2, $$

so for a real vector unit the two insertions collapse in pairs and the four readings give the three matrices $A$, $-A$ and $-\mathrm I_2$.

A generic complex element. For $\tilde Q=\vec e_x+i\vec e_y$ the matrix is $-i\sigma_1+\sigma_2$, and the four readings are

$$ A=\begin{pmatrix}0&-1-i\\ 1-i&0\end{pmatrix},\quad \sigma_2A^{\mathsf T}\sigma_2=\begin{pmatrix}0&1-i\\-1-i&0\end{pmatrix},\quad A^{\dagger}=\begin{pmatrix}0&1+i\\-1+i&0\end{pmatrix},\quad \sigma_2A^{\mathsf T}\sigma_2A^{\dagger}=\begin{pmatrix}-2&0\\0&-2\end{pmatrix}. $$

The four matrices are pairwise distinct. The collapse for a real vector is therefore a property of the coefficients and not of the display.

The four left operators. Applying the same four readings to the parameter of a left multiplication gives the four operators $L_A$, $L_{\sigma_2A^{\mathsf T}\sigma_2}$, $L_{A^{\dagger}}$ and $L_{\sigma_2A^{\mathsf T}\sigma_2A^{\dagger}}$. They are distinct for a generic complex parameter and coincide in pairs for a real vector unit. The distinction is the matrix form of the source's count of four operators for a generic element.

Rotation, Reflection and the Sign in the Display

The Plain Product Gives Minus the Reflection

The two vector products. In the display the product of two vector units is

$$ \mathsf M_2(\vec e_x)\mathsf M_2(\vec e_y)=(-i\sigma_1)(-i\sigma_2)=-\sigma_1\sigma_2=-i\sigma_3=\mathsf M_2(\vec e_z), $$

and the reversed product gives $-\mathsf M_2(\vec e_z)$. The product of two vector units is therefore a vector unit up to sign, and the sign records the order.

The reflection. The reflection in the plane orthogonal to $\vec e_z$ is the map $\vec e_z\mapsto-\vec e_z$ with $\vec e_x,\vec e_y$ fixed. In the display it is conjugation by $\sigma_3$:

$$ \sigma_3\mathsf M_2(\vec e_z)\sigma_3=-\mathsf M_2(\vec e_z),\qquad \sigma_3\mathsf M_2(\vec e_x)\sigma_3=\mathsf M_2(\vec e_x),\qquad \sigma_3\mathsf M_2(\vec e_y)\sigma_3=\mathsf M_2(\vec e_y). $$

The plain product $\mathsf M_2(\vec e_x)\mathsf M_2(\vec e_y)$ is $-i\sigma_3$, while the reflection of the product ray is $+i\sigma_3$: the plain product gives minus the reflection. The minus is the orientation, and it is carried by the antisymmetric part of the product and not by the display.

The determinant. Multiplication by a vector unit is an isometry of the algebra, and in the display it is a matrix of determinant $1$: $\det\mathsf M_2(\vec e_x)=\det(-i\sigma_1)=1$. No rotation-reflection marker survives on the matrix of the element, and the sign of the reflection is read on the product and not on the factor.

The Sign Is Invisible on the Centre and on Even Parameters

The centre. A central element is $\lambda\mathrm I_2$. Its four readings are $\lambda\mathrm I_2$, $\lambda\mathrm I_2$, $\bar\lambda\mathrm I_2$ and $\bar\lambda\mathrm I_2$: the first slot is invisible on the centre and the second is the entrywise conjugation. The sign that distinguishes the four products therefore does not act on the centre, which is the matrix statement that the four products agree on the centre.

The even parameters. An even parameter is a matrix with equal diagonal entries, $A=w\mathrm I_2+u$ with $u$ traceless. Its quaternion conjugate is $\sigma_2A^{\mathsf T}\sigma_2$, and for a traceless $u$ the two matrices differ by the sign of the whole traceless part only when the coefficients are real. The four readings remain distinct on a generic even parameter.

The determinant table. On a real vector the four readings are $A$, $-A$, $-A$ and $-B^2\mathrm I_2$, and the determinant of each is $1$. The determinant cannot see the sign, so the sign that the products carry is not a determinant sign.

Which Involution Plays the Sign

The choice. The sign of the product can be read through the quaternion conjugation or through the Hermitian conjugation, and the two give different operators. In the display the first is $\sigma_2A^{\mathsf T}\sigma_2$ and the second is $A^{\dagger}$; on a real matrix they agree, and on a complex matrix they differ by the entrywise conjugation.

The slot rule. The sign belongs to the slot in which the insertion is made. An insertion in the parameter and an insertion in the argument give the same product only on the centre, and the display separates them: the insertion in the parameter replaces $A$ by $\sigma_2A^{\mathsf T}\sigma_2$, the insertion in the argument replaces $Y$ by $Y^{\dagger}$ or by $\sigma_2Y^{\mathsf T}\sigma_2$, and the two replacements act on different factors of the matrix product.

The cost of multiplicativity. Only $\natural$ is an anti-automorphism; the entrywise conjugation is not multiplicative and $*$ is conjugate-linear. The display shows the cost: $\Phi_{\natural}(AB)=\Phi_{\natural}(B)\Phi_{\natural}(A)$ holds because transposition reverses the order and conjugation by $\sigma_2$ is a homomorphism, while $\Phi_{*}(AB)=B^{\dagger}A^{\dagger}$ holds as well but conjugates the coefficients. A multiplicative sign is therefore available on the coefficients and not on the complex structure, and the choice of which involution plays the sign is a choice of which of the two one is willing to lose.

Where $n=2$ Differs

The plane and the three axes. At $n=2$ the conjugation $\sigma_2A^{\mathsf T}\sigma_2$ uses one fixed matrix, the symplectic form of the two-dimensional module. In the biquaternion statement the conjugate is computed on four coordinates with three independent signs; in the display the same three signs are the three Pauli matrices, and the single matrix $\sigma_2$ generates all three through $\sigma_2\sigma_k^{\mathsf T}\sigma_2=-\sigma_k$. There is one symplectic form and not three, and it is unique up to scalar.

The traceless part. The traceless part is three-complex-dimensional, spanned by the three $-i\sigma_k$, and the four readings act on it as the identity, the negation, the entrywise conjugation and the composite. Nothing larger is available: $M_2(\mathbb{C})$ has no room for a second triple.

The discriminant. The spectral data are one trace and one determinant, and the discriminant $\Delta=(\operatorname{tr}A)^2-4\det A=-4B^2$ is a single complex number. At a higher order the discriminant is a polynomial in the coefficients and not a single number, so spectral regularity is a single equation here.

The operator space. The endomorphism algebra is $M_4(\mathbb{C})$, of complex dimension sixteen, because the carrier has complex dimension four. Every product of two elements is a matrix product in $M_2(\mathbb{C})$, and no product needs more than two factors.

Summary

The display sends $\tilde Q$ to $A=Q_t\mathrm I_2-i(Q_x\sigma_1+Q_y\sigma_2+Q_z\sigma_3)$, with $\operatorname{tr}A=2Q_t$ and $\det A=Q_t^2+Q_x^2+Q_y^2+Q_z^2$. The quaternion conjugation becomes $\sigma_2A^{\mathsf T}\sigma_2$ and the Hermitian conjugation becomes $A^{\dagger}$, so the four general products become $A_1A_2$, $\sigma_2A_1^{\mathsf T}\sigma_2A_2$, $A_1A_2^{\dagger}$ and $\sigma_2A_1^{\mathsf T}\sigma_2A_2^{\dagger}$.

The four pairings are $\tfrac12$ of the four traces, and their Gram matrices on the units are $E,\mathrm I_4,\mathrm I_4,E$. The readings form $\mathbb{Z}_2\times\mathbb{Z}_2$, generated by the two insertions, and the group acts on the pairings and on the Gram matrices by the parity of the insertions. The setting (both) alone has a central square, and the square is $\det A$ times the identity matrix, which gives the inverse formula.

The four operators of a real vector unit collapse in pairs and the four operators of a generic complex element are distinct. The plain product of two vector units is minus the reflection, and the sign is invisible on the centre and on the determinant. The choice of which involution plays the sign is a choice of slot, and only the quaternion conjugation is multiplicative.

Summary of Notation

Symbol Meaning
$A=\mathsf M_2(\tilde Q)$ the $2\times2$ matrix of the element
$\sigma_1,\sigma_2,\sigma_3$ the Pauli matrices
$J=-i\sigma_2$ the symplectic matrix, $J^2=-\mathrm I_2$
$\sigma_2A^{\mathsf T}\sigma_2$ the matrix of the quaternion conjugate
$A^{\dagger}$ the matrix of the Hermitian conjugate
$\Phi_{\natural},\Phi_{*}$ the two insertion operations on matrices
$E=\operatorname{diag}(1,-1,-1,-1)$ the Gram matrix of the plain pairing
$B^2=Q_x^2+Q_y^2+Q_z^2$ the square of the vector part
$\lambda_\pm=Q_t\pm iB$ the two eigenvalues
$N(\tilde Q)=\det A$ the norm

Further Reading