The Derivation $X\mapsto[A,X]$ and Its Anti-Self-Adjointness for the Trace Form

Introduction

The map $\operatorname{ad}_A(X)=[A,X]$ is a derivation of $M_2(\mathbb{C})$ for every $A$, it is anti-self-adjoint for the trace form for every $A$, and it is anti-self-adjoint for the Hilbert–Schmidt form exactly when $A$ is antisymmetric.

The article proves the Leibniz rule, the representation property $\operatorname{ad}_{[A,B]}=[\operatorname{ad}_A,\operatorname{ad}_B]$, the anti-self-adjointness of $\operatorname{ad}_A$ for the trace form, the formula $\operatorname{ad}_A^{\mathrm{HS}*}=\operatorname{ad}_{A^{\mathsf T}}$, the spectrum of $\operatorname{ad}_A$ as the two differences of the eigenvalues, the identification of the kernel with the centre and the dimension of the space of inner derivations, and the nilpotency of $\operatorname{ad}_A$ for a nilpotent $A$. It assumes The Two Halves over $\mathbb{R}$: Symmetric and Antisymmetric at $n=2$ and The Lie Algebra $\operatorname{Skew}_2$: One Dimension, Hence Abelian.

It defers the Killing form to The Killing Form of $\mathfrak{sl}_2$ and Its Three Inertias, the Hilbert–Schmidt form to The Hilbert–Schmidt Form at $n=2$, and the operator viewpoint to The One-Sided Operators at $n=2$.

The Commutator Is a Derivation

Theorem. For every $A$ the map $\operatorname{ad}_A(X)=[A,X]$ satisfies the Leibniz rule

$$ [A,XY]=[A,X]\,Y+X\,[A,Y]. $$

Proof. Expand: $[A,XY]=AXY-XYA$ and the right side is $AXY-XAY+XAY-XYA$, which is the same.

Corollary. $\operatorname{ad}_A$ is a derivation of the associative algebra, and it is traceless as an operator on the algebra because its values lie in the traceless subspace.

Proof. The Leibniz rule is the derivation property; and $\operatorname{tr}([A,X])=0$ for every $X$, so every value of $\operatorname{ad}_A$ is traceless.

Corollary. $\operatorname{ad}_A$ annihilates the identity and preserves the traceless subspace $\mathfrak{sl}_2$.

Proof. $[A,I]=0$, and the traceless matrices are exactly the kernel of the trace, which $\operatorname{ad}_A$ respects since its images are traceless.

The Adjoint Representation

Theorem. The map $A\mapsto\operatorname{ad}_A$ is a representation of the Lie algebra $\mathfrak{gl}_2$:

$$ \operatorname{ad}_{[A,B]}=[\operatorname{ad}_A,\operatorname{ad}_B]. $$

Proof. Both sides applied to $X$ give $[[A,B],X]$ and $[A,[B,X]]-[B,[A,X]]$, which are equal by the Jacobi identity.

Corollary. The map $\operatorname{ad}$ is a homomorphism of Lie algebras with kernel the centre $\mathbb{C}I$, so its image is a Lie algebra of dimension $3$ over $\mathbb{C}$.

Proof. $\operatorname{ad}_A=0$ iff $A$ commutes with the whole algebra iff $A$ is scalar; the domain has dimension $4$ and the kernel has dimension $1$.

The kernel is the centre and nothing else. At order two the central simple algebra has a one-dimensional centre, so the adjoint representation is faithful exactly on the traceless part.

Proof. The scalar matrices commute with everything and no other matrix does, since the algebra is simple.

Anti-Self-Adjointness for the Trace Form

Theorem. For every $A$ and all $X,Y$,

$$ \operatorname{tr}\bigl([A,X]\,Y\bigr)+\operatorname{tr}\bigl(X\,[A,Y]\bigr)=0 . $$

Proof. $\operatorname{tr}([A,X]Y)=\operatorname{tr}(AXY)-\operatorname{tr}(XAY)=\operatorname{tr}(A[XY])$ using cyclicity, and $\operatorname{tr}(X[A,Y])=\operatorname{tr}(XAY)-\operatorname{tr}(XYA)=-\operatorname{tr}(A[XY])$ by the same cyclicity; the sum vanishes.

Corollary. Every $\operatorname{ad}_A$ is anti-self-adjoint for the trace form, so the image of $\operatorname{ad}$ lies in the Lie algebra of the orthogonal group of the trace form, which is $\mathfrak{so}(3,1)$.

Proof. The anti-self-adjointness is the theorem, and the Lie algebra of the orthogonal group of a form of inertia $(3,1)$ is the set of anti-self-adjoint endomorphisms.

The trace form cannot detect the anti-self-adjointness condition. Here the freedom is total: no hypothesis on $A$ is needed, because the trace is invariant under the cyclic permutation of three factors, and the invariance is what produces the anti-self-adjointness.

Proof. The proof above uses only cyclicity, with no symmetry of $A$.

Anti-Self-Adjointness for the Hilbert–Schmidt Form

Theorem. For the Hilbert–Schmidt form $\langle X,Y\rangle=\operatorname{tr}(X^{*}Y)$ the adjoint of $\operatorname{ad}_A$ is $\operatorname{ad}_{A^{\mathsf T}}$.

Proof. $\langle[A,X],Y\rangle=\operatorname{tr}((AX-XA)^{*}Y)=\operatorname{tr}(X^{*}A^{*}Y)-\operatorname{tr}(A^{*}X^{*}Y)$ and $\langle X,[A^{\mathsf T},Y]\rangle=\operatorname{tr}(X^{*}A^{\mathsf T}Y)-\operatorname{tr}(X^{*}YA^{\mathsf T})$; the two agree after cyclicity.

Corollary. $\operatorname{ad}_A$ is anti-self-adjoint for the Hilbert–Schmidt form exactly when $A$ is antisymmetric, and it is self-adjoint exactly when $A$ is symmetric.

Proof. $\operatorname{ad}_{A^{\mathsf T}}=-\operatorname{ad}_A$ iff $A^{\mathsf T}=-A$, since $\operatorname{ad}$ is injective on the traceless part and the antisymmetric matrices are traceless.

The two forms divide the algebra. The trace form sees every derivation as antisymmetric, while the Hilbert–Schmidt form separates the symmetric directions, where the derivation is self-adjoint, from the antisymmetric direction, where it is antisymmetric.

The Spectrum of $\operatorname{ad}_A$

Theorem. If $A$ has eigenvalues $\lambda_{1},\lambda_{2}$ then $\operatorname{ad}_A$ has eigenvalues $0,0,\lambda_{1}-\lambda_{2},\lambda_{2}-\lambda_{1}$, with the two zeros on the eigen-directions $E_{11}$ and $E_{22}$.

Proof. In the basis of matrix units, $\operatorname{ad}_A(E_{ij})=(\lambda_{i}-\lambda_{j})E_{ij}$, so the four eigenvalues are the four differences, two of which vanish on the diagonal units.

Corollary. The trace of $\operatorname{ad}_A$ is zero and the determinant of $\operatorname{ad}_A$ is zero, and the rank of $\operatorname{ad}_A$ is $2$ when the two eigenvalues differ and $0$ when they coincide.

Proof. The four eigenvalues sum to zero and contain two zeros; the rank is the number of nonzero eigenvalues.

Corollary. The spectrum of $\operatorname{ad}_A$ is real for every complex matrix $A$ whose eigenvalues are real, and it is purely imaginary for an anti-Hermitian $A$ with imaginary eigenvalues.

Proof. The differences of two real numbers are real, and the differences of two purely imaginary numbers are purely imaginary.

The Kernel and the Inner Derivations

Theorem. Every derivation of $M_2(\mathbb{C})$ is inner, and the space of derivations is $\mathfrak{sl}_2(\mathbb{C})$, of dimension $3$.

Proof. The algebra is central simple, so every derivation is of the form $\operatorname{ad}_A$; the kernel of $\operatorname{ad}$ is the centre, so the space of derivations is the quotient by the scalars, which is $\mathfrak{sl}_2$.

Corollary. The commutator subspace of $M_2(\mathbb{C})$ is $\mathfrak{sl}_2(\mathbb{C})$, of dimension $3$, and the commutator of two matrix units is another matrix unit or zero.

Proof. The commutators span the traceless part, and $[E_{ij},E_{kl}]=\delta_{jk}E_{il}-\delta_{li}E_{kj}$.

The derivations of the two halves. A derivation that preserves the symmetric half is the restriction of some $\operatorname{ad}_A$ with $A$ antisymmetric, and at order two that space is one-dimensional.

Proof. The antisymmetric matrices form the line $\mathbb{R}J$, and $\operatorname{ad}_J$ is the single derivation of the symmetric half into itself up to scale.

The Nilpotent Case

Theorem. If $A$ is nilpotent then $\operatorname{ad}_A$ is nilpotent, and at order two its index is $3$.

Proof. For $A=E_{12}$ the eigenvalues of $A$ coincide, so all four eigenvalues of $\operatorname{ad}_A$ vanish; a direct computation gives $\operatorname{ad}_{E_{12}}^{2}\neq0$ and $\operatorname{ad}_{E_{12}}^{3}=0$.

Corollary. The derivation attached to a nilpotent element lies in the nilpotent cone of $\mathfrak{gl}(\mathfrak{gl}_2)$, and its powers cut the algebra into the filtration of the associated graded space.

Proof. The vanishing of the cube is the statement, and the filtration by the kernels of the powers is the standard one.

The matrix of $\operatorname{ad}_{E_{12}}$. In the basis $I,\sigma_{3},\sigma_{1},J$ the operator is

$$ \operatorname{ad}_{E_{12}}=\begin{pmatrix}0&0&0&0\\ 0&0&1&-1\\ 0&-1&0&0\\ 0&-1&0&0\end{pmatrix}. $$

Proof. The columns are the coordinates of the images of the four basis elements, computed with $E_{12}I-IE_{12}=E_{12}$ and $E_{12}J-JE_{12}=-\sigma_{3}$.

Corollary. The image of $\operatorname{ad}_{E_{12}}$ is two-dimensional, spanned by $\sigma_{3}$ and $\sigma_{1}+J$, and its kernel is the same span together with $I$.

Proof. Read the column space and the null space of the displayed matrix.

Summary

$\operatorname{ad}_A(X)=[A,X]$ is a derivation for every $A$, the map $A\mapsto\operatorname{ad}_A$ is a representation of $\mathfrak{gl}_2$ with kernel the scalars, and every derivation of $M_2(\mathbb{C})$ is inner, so the derivations form $\mathfrak{sl}_2(\mathbb{C})$ of dimension $3$.

For the trace form every $\operatorname{ad}_A$ is anti-Hermitian, with no condition on $A$, while for the Hilbert–Schmidt form the adjoint is $\operatorname{ad}_{A^{\mathsf T}}$, which makes the derivation anti-self-adjoint for antisymmetric $A$ and self-adjoint for symmetric $A$. The spectrum of $\operatorname{ad}_A$ is $0,0,\pm(\lambda_{1}-\lambda_{2})$, and for a nilpotent element the derivation is nilpotent of index $3$.

Summary of Notation

Symbol Meaning
$\operatorname{ad}_A$ the derivation $X\mapsto[A,X]$
$[A,B]$ the commutator $AB-BA$
$\operatorname{tr}(X^{*}Y)$ the Hilbert–Schmidt form
$\operatorname{ad}_A^{\mathrm{HS}*}$ the adjoint for the Hilbert–Schmidt form
$\mathfrak{sl}_2$ the traceless matrices, the space of derivations
$E_{ij}$ the matrix units

Further Reading