The Symmetric and Antisymmetric Parts of the Operator Space at $n=2$

Introduction

Transposition is an involution of the matrix space, conjugation by it is an involution of the operator algebra, and the operator algebra splits into the sixteen-dimensional symmetric part and the sixteen-dimensional antisymmetric part, with the two-sided operator $T_{X,Y}$ symmetric exactly when $Y=X^{\mathsf T}$.

The article defines the transposition operator, transports it to an involution of the endomorphism algebra, computes the two eigenspaces and their dimensions, determines which one-sided and two-sided operators lie in each part, compares the transposition involution with the adjoint, and reads the decomposition on the unitary and the self-adjoint operators. It assumes The Adjoints and the Polar Decomposition of an Operator at $n=2$, The Two-Sided Operators and the Segre Cone at $n=2$ and The Two Halves over $\mathbb{R}$: Symmetric and Antisymmetric at $n=2$.

It follows the biquaternion article The Endomorphisms of the Biquaternion Algebra for the same involution on the eight real coordinates.

The Transposition Operator

Transposition is a conjugate-linear involution of the matrix space, and it is an isometry of the Hilbert–Schmidt form.

Definition. Let $\tau(Z)=Z^{\mathsf T}$ be the transposition operator on the matrix space.

Proposition. The operator $\tau$ is conjugate-linear, involutive and an isometry: $\langle\tau Z,\tau W\rangle=\langle Z,W\rangle$.

Proof. Transposition reverses the order of a product and conjugates a scalar product, and applying it twice returns the matrix; the form of a product of transposes is the conjugate of the reversed form.

Corollary. The fixed points of $\tau$ are the symmetric matrices, and the anti-fixed points are the antisymmetric matrices.

Proof. A matrix equal to its transpose is symmetric, and one equal to minus its transpose is antisymmetric.

Corollary. The transposition operator is not complex-linear, so it is not an element of the endomorphism algebra.

Proof. The complex scalar $i$ is carried to $-i$ under transposition of a scalar product, which is conjugate-linearity.

The Involution on the Operator Algebra

Theorem. The map $\theta(T)=\tau T\tau$ is an involutive automorphism of the endomorphism algebra, and it is complex-linear.

Proof. The two factors of the conjugation are conjugate-linear, so the composite is linear; and $\theta(TS)=\tau TS\tau=\tau T\tau\,\tau S\tau=\theta(T)\theta(S)$ because $\tau^{2}$ is the identity, so the map is an automorphism, and it is involutive for the same reason.

Corollary. The operator algebra splits as the direct sum of the two eigenspaces of $\theta$, of dimensions sixteen each.

Proof. An involution has the eigenvalues $+1$ and $-1$, and each eigenspace has half the dimension because the two projectors $\tfrac12(1\pm\theta)$ are complementary.

Corollary. The identity operator is symmetric, and the transposition operator composed with the identity is symmetric as an operator on the algebra.

Proof. The identity operator commutes with $\tau$, and $\tau$ is itself fixed by the conjugation when it is regarded as an operator.

The Two Families

Proposition. The conjugate of a two-sided operator is $\theta(T_{X,Y})=T_{Y^{\mathsf T},X^{\mathsf T}}$.

Proof. Applying $\tau$ on both sides sends $Z$ to $(XZ^{\mathsf T}Y)^{\mathsf T}=Y^{\mathsf T}ZX^{\mathsf T}$.

Corollary. The two-sided operator $T_{X,Y}$ is symmetric exactly when $Y=X^{\mathsf T}$, and antisymmetric exactly when $Y=-X^{\mathsf T}$.

Proof. The conjugate is $T_{Y^{\mathsf T},X^{\mathsf T}}$, which equals $T_{X,Y}$ exactly when the two parameters agree as stated, and equals its negative in the other case.

Corollary. The left multiplication by $X$ is symmetric exactly when $X$ is symmetric, and the right multiplication by $Y$ is symmetric exactly when $Y$ is symmetric.

Proof. The left multiplication is the two-sided operator with the second parameter the identity, so the condition is $I=X^{\mathsf T}$; the right multiplication is the case of the first parameter the identity.

The Relation with the Adjoint

Proposition. The transposition involution and the adjoint commute: $\theta(T^{*})=\theta(T)^{*}$.

Proof. Both are computed by conjugating and transposing the four-by-four matrix of the operator, and conjugation and transposition of a matrix commute with its conjugate transpose.

Corollary. The adjoint of a symmetric operator is symmetric, and the adjoint of an antisymmetric operator is antisymmetric.

Proof. The involution preserves the eigenspaces, and the adjoint is compatible with it.

Corollary. The symmetric self-adjoint operators are exactly the operators $T_{H,H}$ with $H$ Hermitian, a real space of dimension four.

Proof. A symmetric operator has $Y=X^{\mathsf T}$, and self-adjointness forces both parameters to be Hermitian; the Hermitian half has real dimension four.

The Antisymmetric Part

Proposition. The antisymmetric operators are exactly those of the form $T_{X,-X^{\mathsf T}}$ together with the sums of such operators, and they form a Lie subalgebra.

Proof. The bracket of two antisymmetric operators is the difference of two products, each of which is symmetric, so the bracket is antisymmetric; the space is closed under the bracket.

Corollary. The commutator with the transposition operator is $[\tau,T]=(\theta(T)-T)\tau=-2T_{-}\tau$, so it measures the antisymmetric part.

Proof. Since $\tau T=\theta(T)\tau$, the commutator is $(\theta(T)-T)\tau$, and $\theta(T)-T$ is minus twice the antisymmetric part.

Corollary. The bracket of the antisymmetric part with the whole algebra preserves the antisymmetric part, so the antisymmetric part is an ideal for the bracket action.

Proof. The involution is an automorphism, so it acts on the bracket by the bracket of its actions, which preserves the two eigenspaces.

The Two Halves inside the Operator Space

Proposition. The symmetric operators form a subalgebra of the endomorphism algebra, and the antisymmetric part is a module over it.

Proof. The product of two symmetric operators is symmetric because the involution is an automorphism, and the product of a symmetric with an antisymmetric operator is antisymmetric for the same reason.

Corollary. The symmetric part is the fixed algebra of the involution, and its unitary group is the group of operators that commute with transposition up to the involution.

Proof. The fixed algebra is by definition the symmetric part, and its unitary group consists of the unitaries fixed by the involution.

Corollary. The antisymmetric part is an ideal of the Lie algebra, since the bracket of an antisymmetric operator with any operator is antisymmetric.

Proof. The involution is an automorphism of the algebra and hence of the bracket, so it acts on a bracket by the brackets of its actions; the sign product of the two eigenspaces shows that a bracket with an antisymmetric factor stays antisymmetric.

Corollary. The dimension count is sixteen plus sixteen, matching the thirty-two real dimensions of the operator algebra.

Proof. The two eigenspaces are complex subspaces of dimension sixteen each, and their real dimensions add to thirty-two.

Summary

Transposition is a conjugate-linear involution of the matrix space, and conjugation by it is an involution of the operator algebra splitting it into two parts of dimension sixteen each. The two-sided operator $T_{X,Y}$ is symmetric exactly when $Y=X^{\mathsf T}$ and antisymmetric exactly when $Y=-X^{\mathsf T}$, so the left and right multiplications are symmetric exactly when their parameters are symmetric.

The transposition involution commutes with the adjoint, the symmetric part is a subalgebra, and the antisymmetric part is a Lie subalgebra and a module over the symmetric one.

Summary of Notation

Symbol Meaning
$\tau(Z)=Z^{\mathsf T}$ the transposition operator
$\theta(T)=\tau T\tau$ the involution of the operator algebra
$T_{X,Y}$ the two-sided operator
$+1,-1$ the two eigenspaces of the involution
$T^{*}$ the adjoint for the Hilbert–Schmidt form

Further Reading