The Involutions of $M_2(\mathbb{C})$ over $\mathbb{R}$ and over $\mathbb{C}$
Introduction
An involution of $M_2(\mathbb{C})$ is one of six maps up to conjugacy, and each fixes one of the spaces the rest of the series uses.
The article classifies the involutions of order two, both the complex-linear and the conjugate-linear ones, computes the fixed space of each in explicit matrices, and identifies the fixed spaces with $\operatorname{Sym}_2$, $\operatorname{Herm}_2$, the line of scalars, the diagonal algebra and the quaternion algebra. It assumes The Matrix Algebra $M_2(\mathbb{C})$ for the product and the transpose, and The Two Base Fields: $M_2(\mathbb{R})$ inside $M_2(\mathbb{C})$ for the two real forms. It defers the two halves themselves to The Two Halves over $\mathbb{R}$: Symmetric and Antisymmetric at $n=2$ and The Two Halves over $\mathbb{C}$: Hermitian and Anti-Hermitian at $n=2$.
The biquaternion article to read alongside is The Element: The Coordinates and the Conjugations, where the three conjugations of $\mathbb{B}$ play the role of the three complex-linear involutions below. The dictionary is exact: the reversion, the conjugation and the reversion-conjugation of $\mathbb{B}$ are the transpose, the symplectic involution and conjugation by a reflection.
The Three Kinds of Involution
Definition. An involution is a bijective $\mathbb{R}$-linear map $\theta$ of $M_2(\mathbb{C})$ with $\theta^{2}=\mathrm{id}$ that is either an automorphism or an antiautomorphism, and that is either $\mathbb{C}$-linear or conjugate-linear.
Proposition. There are four kinds, and all four occur.
Proof. The map is multiplicative or antimultiplicative, and it respects or reverses the scalars. Each of the four combinations has an example: the identity for linear automorphism, the transpose for linear antiautomorphism, the conjugation of entries for conjugate-linear automorphism, and the conjugate transpose for conjugate-linear antiautomorphism. Each of those four maps has square the identity.
The fixed space is a real subspace. For a conjugate-linear involution the fixed set is not a complex subspace; it is only real, of dimension $4$, since the eigenspaces of a real-linear involution with eigenvalues $\pm1$ partition the eight real dimensions, and $+1$ has dimension equal to that of $-1$ when exactly half the eigenvalues are $+1$.
Every conjugate-linear involution fixes a four-real-dimensional space. That number is the dimension of $M_2(\mathbb{R})$ and of $\operatorname{Herm}_2$ and of $\mathbb{H}$, which are the spaces that appear.
The Two Real Forms
Proposition. The conjugation of entries $\sigma(A)=\bar A$ is a conjugate-linear involution of the algebra, and its fixed space is $M_2(\mathbb{R})$, of real dimension $4$.
Proof. Conjugation is additive, multiplicative and reverses scalars, and it has square the identity; a matrix is fixed exactly when all its entries are real.
Proposition. With $J=\begin{pmatrix}0&1\\ -1&0\end{pmatrix}$ the map $\sigma_{J}(A)=J\bar AJ^{-1}$ is a conjugate-linear involution of the algebra, and its fixed space is the quaternion algebra
$$ \mathbb{H}=\left\{\begin{pmatrix}\alpha&\beta\\ -\bar\beta&\bar\alpha\end{pmatrix}\right\}, $$
of real dimension $4$.
Proof. $\sigma_J$ is the composition of the automorphism $A\mapsto JAJ^{-1}$ with the entrywise conjugation, so it is a conjugate-linear automorphism, and $\sigma_J^{2}=\mathrm{id}$ because $\bar J=J$ and $J^{2}=-I$ cancels between the two steps. The fixed condition $A=J\bar AJ^{-1}$ is equivalent to $\bar A=J^{-1}AJ=-JAJ$, which reads $\bar a=d$, $\bar b=-c$.
The two forms are the ones of the base-field article. That $M_2(\mathbb{R})$ and $\mathbb{H}$ are the two real forms of $M_2(\mathbb{C})$, and that no third exists, is The Two Base Fields: $M_2(\mathbb{R})$ inside $M_2(\mathbb{C})$.
The biquaternion reading. The conjugation $\sigma$ restricted to $\mathbb{B}$ is the conjugation of the corpus that fixes the four Peirce coordinates and negates the mixed corners; on the matrix side it is the entrywise conjugation, whose fixed algebra is the real matrices.
The Transpose and the Symmetric Space
Proposition. The transpose $\tau(A)=A^{\mathsf T}$ is a $\mathbb{C}$-linear antiautomorphism of order two, and its fixed space is $\operatorname{Sym}_2$, of complex dimension $3$ and real dimension $6$.
Proof. $\tau(AB)=B^{\mathsf T}A^{\mathsf T}=\tau(B)\tau(A)$ and $\tau^{2}=\mathrm{id}$; a matrix is fixed exactly when $a_{12}=a_{21}$.
The transpose is the reversion of the corpus. In the biquaternion language the transpose is the reversion, which fixes the real scalars, the root of one and the imaginary units and negates the mixed corner coordinates; on the matrix side those are exactly the conditions $a_{12}=a_{21}$ in the Peirce basis. The correspondence is spelled out in The Peirce Basis of the Biquaternion Algebra.
The $-1$ eigenspace is the antisymmetric line. The antisymmetric matrices form the one-dimensional space $\mathbb{C}J$, and the transpose acts on it as $-1$.
Corollary. $M_2(\mathbb{C})=\operatorname{Sym}_2\oplus\Lambda^{2}\mathbb{C}^{2}$ is the eigenspace decomposition of the transpose, and the two summands have complex dimensions $3$ and $1$.
The Conjugate Transpose and the Hermitian Space
Proposition. The conjugate transpose $\theta(A)=A^{*}=\bar A^{\mathsf T}$ is a conjugate-linear antiautomorphism of order two, and its fixed space is $\operatorname{Herm}_2$, of real dimension $4$.
Proof. $\theta(AB)=B^{*}A^{*}$ and $\theta^{2}=\mathrm{id}$; a matrix is fixed exactly when $\bar a_{ji}=a_{ij}$, that is when it is Hermitian.
The anti-Hermitian space is the other eigenspace. The matrices with $A^{*}=-A$ form the space $i\operatorname{Herm}_2$ of real dimension $4$, and
$$ M_2(\mathbb{C})=\operatorname{Herm}_2\oplus i\operatorname{Herm}_2 $$
is the real eigenspace decomposition of $\theta$.
The trace is real on the fixed space. For $A$ Hermitian the diagonal entries $a_{ii}$ are real, so $\operatorname{tr}A\in\mathbb{R}$; for $A$ anti-Hermitian the trace is purely imaginary. The two halves are therefore distinguished by the trace as well as by the involution, and that is the start of The Two Halves over $\mathbb{C}$: Hermitian and Anti-Hermitian at $n=2$.
The Symplectic Involution and the Line of Scalars
Proposition. With $J$ as above the map $\tau_{J}(A)=JA^{\mathsf T}J^{-1}$ is a $\mathbb{C}$-linear antiautomorphism of order two whose fixed space is the line of scalars $\mathbb{C}I$.
Proof. It is the composition of the transpose with the automorphism $A\mapsto JAJ^{-1}$, hence an antiautomorphism, and its square is the identity because $J^{\mathsf T}=-J$ and $J^{-1}=-J$. Writing $A=\begin{pmatrix}a&b\\ c&d\end{pmatrix}$ one computes $JA^{\mathsf T}J^{-1}=\begin{pmatrix}d&-b\\ -c&a\end{pmatrix}$, so the fixed condition is $b=c=0$ and $a=d$.
The name. $\tau_J$ is the symplectic involution: it is the involution attached to the alternating form $J$, and its $-1$ eigenspace is the three-dimensional space of traceless matrices, which is the symplectic Lie algebra $\mathfrak{sp}_2=\mathfrak{sl}_2$.
Corollary. $\tau_J$ acts as $+1$ on $\mathbb{C}I$ and as $-1$ on $\mathfrak{sl}_2(\mathbb{C})$.
The two antiautomorphic involutions are the two sorts of form. $\tau$ is attached to the symmetric form $I$ and $\tau_{J}$ to the alternating form $J$; these are the two types of involution of the first kind, orthogonal and symplectic, and at order two both exist because a nondegenerate alternating form needs an even dimension.
The Linear Automorphism of Order Two
Proposition. Let $D=\operatorname{diag}(1,-1)$. The map $\rho(A)=DAD^{-1}$ is a $\mathbb{C}$-linear automorphism of order two, and its fixed space is the algebra of diagonal matrices, of complex dimension $2$.
Proof. Conjugation by an invertible matrix is an automorphism, and $D^{2}=I$ gives $\rho^{2}=\mathrm{id}$; the condition $DAD^{-1}=A$ reads $a_{12}=-a_{12}$ and $a_{21}=-a_{21}$.
Every $\mathbb{C}$-linear automorphism of order two is of this form. By Skolem–Noether an automorphism is $A\mapsto gAg^{-1}$, and $g^{2}$ must be central for the square to be the identity; so $g$ is a scalar multiple of an involution, and a non-scalar involution of order two is a reflection, all of which are conjugate to $D$.
The fixed algebra is $\mathbb{C}\times\mathbb{C}$. Diagonal matrices multiply entrywise, so the fixed algebra is the product of two copies of $\mathbb{C}$; its two minimal idempotents are $E_{11}$ and $E_{22}$.
This is one point of the gauge, not a datum. A reflection $R$ with $R^{2}=I$ and $R$ not scalar is a choice inside the conjugacy class of $D$; the biquaternion frame of The Frame, the Simple Nilpotent and the Gauge of $M_2(\mathbb{C})$ does not depend on it, and no article of the corpus assumes such an $R$ as part of the definition.
The Classification
Theorem. Up to conjugacy there are exactly six nontrivial involutions:
$$ \tau(A)=A^{\mathsf T},\qquad \tau_{J}(A)=JA^{\mathsf T}J^{-1},\qquad \rho(A)=DAD^{-1},\qquad \sigma(A)=\bar A,\qquad \sigma_{J}(A)=J\bar AJ^{-1},\qquad \theta(A)=A^{*},\qquad \theta_{J}(A)=JA^{*}J^{-1}. $$
Proof. A $\mathbb{C}$-linear antiautomorphism of $M_2(\mathbb{C})$ is $A\mapsto SA^{\mathsf T}S^{-1}$ for some invertible $S$, by the Skolem–Noether argument applied to the opposite algebra, and the square is the identity exactly when $S^{\mathsf T}=\pm S$; conjugating $S$ by $T$ replaces it by $T^{\mathsf T}ST$. A symmetric invertible matrix is congruent to $I$ and an anti-Hermitian one to $J$, so the two cases give $\tau$ and $\tau_{J}$. A $\mathbb{C}$-linear automorphism of order two is a conjugation by an involution, and all non-scalar involutions are conjugate to $D$. A conjugate-linear involution is $\sigma$ composed with one of the linear ones, giving $\sigma$, $\sigma_J$, and the two conjugate transposes $\sigma\circ\tau=\theta$ and $\sigma\circ\tau_{J}=\theta_{J}$.
The fixed spaces are the six spaces of the series.
| Involution | Kind | Fixed space | Real dimension |
|---|---|---|---|
| $\tau$, transpose | linear antiautomorphism | $\operatorname{Sym}_2$ | $6$ |
| $\tau_J$, symplectic | linear antiautomorphism | $\mathbb{C}I$ | $2$ |
| $\rho$, reflection | linear automorphism | diagonal algebra | $4$ |
| $\sigma$, entrywise conjugation | antilinear automorphism | $M_2(\mathbb{R})$ | $4$ |
| $\sigma_J$, quaternionic | antilinear automorphism | $\mathbb{H}$ | $4$ |
| $\theta_J$, symplectic | antilinear antiautomorphism | $\alpha$, $\beta$ form | $4$ |
The last row is the fourth conjugate-linear fixed space. The map $\theta_{J}$ fixes the matrices $\begin{pmatrix}\alpha+i\beta&i\gamma\\ i\delta&\alpha-i\beta\end{pmatrix}$ with real $\alpha,\beta,\gamma,\delta$, of real dimension $4$.
The pairs are the two halves. The three linear involutions cut $M_2(\mathbb{C})$ into $\operatorname{Sym}_2\oplus\Lambda^{2}$, into $\mathbb{C}I\oplus\mathfrak{sl}_2$, and into the diagonal algebra plus its complement; the two conjugate-linear ones cut the real algebra into $M_2(\mathbb{R})$ or $\mathbb{H}$ and their complements. These cuts are the decompositions of The Fine Decomposition against the Diagonal at $n=2$.
Summary
An involution of $M_2(\mathbb{C})$ is an automorphism or an antiautomorphism of order two, linear or conjugate-linear. Up to conjugacy there are six nontrivial ones: the transpose, the symplectic involution $JA^{\mathsf T}J^{-1}$, the conjugation by a reflection, the entrywise conjugation, the quaternionic conjugation $J\bar AJ^{-1}$, and the conjugate transpose with its symplectic companion. The fixed spaces are $\operatorname{Sym}_2$ of complex dimension $3$, the line $\mathbb{C}I$, the diagonal algebra of complex dimension $2$, and the two real forms $M_2(\mathbb{R})$ and $\mathbb{H}$ of real dimension $4$; the conjugate transpose fixes $\operatorname{Herm}_2$, also of real dimension $4$.
| Involution | Kind | Fixed space | $\dim_{\mathbb{R}}$ |
|---|---|---|---|
| $\tau$ | linear antiautomorphism | $\operatorname{Sym}_2$ | $6$ |
| $\tau_J$ | linear antiautomorphism | $\mathbb{C}I$ | $2$ |
| $\rho$ | linear automorphism | diagonal | $4$ |
| $\sigma$ | antilinear automorphism | $M_2(\mathbb{R})$ | $4$ |
| $\sigma_J$ | antilinear automorphism | $\mathbb{H}$ | $4$ |
| $\theta$ | antilinear antiautomorphism | $\operatorname{Herm}_2$ | $4$ |
Summary of Notation
| Symbol | Meaning |
|---|---|
| $\tau$ | the transpose $A\mapsto A^{\mathsf T}$ |
| $\tau_J$ | the symplectic involution $JA^{\mathsf T}J^{-1}$ |
| $\rho$ | the conjugation by $D=\operatorname{diag}(1,-1)$ |
| $\sigma$ | the entrywise conjugation $A\mapsto\bar A$ |
| $\sigma_J$ | the quaternionic conjugation $J\bar AJ^{-1}$ |
| $\theta$ | the conjugate transpose $A^{*}$ |
| $J$ | $\begin{pmatrix}0&1\\ -1&0\end{pmatrix}$, the alternating form |
| $\operatorname{Sym}_2$ | the fixed space of $\tau$ |
Further Reading
- Max-Albert Knus, Alexander Merkurjev, Markus Rost and Jean-Pierre Tignol, The Book of Involutions (AMS, 1998), for the classification of the involutions of a central simple algebra into orthogonal and symplectic types.
- Israel Nathan Herstein, Noncommutative Rings (Carus Mathematical Monographs, 1968), for the Skolem–Noether theorem in the form used here.
- The Two Base Fields: $M_2(\mathbb{R})$ inside $M_2(\mathbb{C})$, for the two real forms.
- The Element: The Coordinates and the Conjugations, for the three conjugations of the biquaternion algebra.
- The Peirce Basis of the Biquaternion Algebra, for the reversion in the Peirce coordinates.