The Lie Algebra $\operatorname{Skew}_n\cong\mathfrak{so}(n)$

Introduction

The antisymmetric half is a Lie algebra of dimension $n(n-1)/2$ and rank $\lfloor n/2\rfloor$, it is simple for $n\ge3$ except at $n=4$, where it splits into two three-dimensional ideals, and it is the Lie algebra of the rotation group.

The article proves closure under the bracket, identifies the half with the exterior square $\Lambda^{2}\mathbb{R}^{n}$, computes the dimension and the rank, gives the simplicity statement with the exception at $n=4$ verified by exhibiting the self-dual and anti-self-dual ideals, computes the Killing form $(n-2)\operatorname{tr}(XY)$ with its definiteness, and relates the algebra to $\mathrm{SO}(n)$ through the exponential. It assumes The Four Multiplication Tables: $SS$, $SK$, $KS$, $KK$ for the bracket table and The Two Halves over $\mathbb{R}$: Symmetric and Antisymmetric for the split. It defers the Killing form to The Killing Form, Its Radical and the Triangular Case, the rotations to The Exponential of the Antisymmetric Part: One-Parameter Rotations, and the group to The Cartan Decomposition of $\mathrm{GL}_n(\mathbb{C})$.

The Algebra

The antisymmetric half is closed under the bracket, and the bracket is the commutator. For $X,Y\in\operatorname{Skew}_n$,

$$ [X,Y]^{\mathsf T}=(XY-YX)^{\mathsf T}=Y^{\mathsf T}X^{\mathsf T}-X^{\mathsf T}Y^{\mathsf T}=YX-XY=-[X,Y], $$

so $[X,Y]\in\operatorname{Skew}_n$. The bracket is bilinear, antisymmetric, and satisfies the Jacobi identity because it is the commutator in an associative algebra:

$$ [[X,Y],Z]+[[Y,Z],X]+[[Z,X],Y]=0 . $$

So $\operatorname{Skew}_n$ is a Lie algebra, and it is the Lie subalgebra $\mathfrak{so}(n)$ of $\mathfrak{gl}_n(\mathbb{R})$.

The dimension is $n(n-1)/2$. An antisymmetric matrix has zero diagonal and $n(n-1)/2$ free entries above it, so

$$ \dim\mathfrak{so}(n)=\frac{n(n-1)}{2}=1,3,6,10,15\quad(n=2,3,4,5,6), $$

which was checked by counting the basis elements at those orders. The complement has dimension $n(n+1)/2$, and the two add to $n^{2}$.

The standard basis is the family $X_{ij}=E_{ij}-E_{ji}$. For $i

$$ [X_{ij},X_{kl}]=\delta_{jk}X_{il}-\delta_{ik}X_{jl}-\delta_{jl}X_{ik}+\delta_{il}X_{jk}, $$

which was verified on all index combinations with $i\neq j$ and $k\neq l$ at $n=3,4,5$. In particular $[X_{ij},X_{jk}]=X_{ik}$ and $[X_{ij},X_{kl}]=0$ when the four indices are distinct.

The Identification with $\Lambda^{2}$

An antisymmetric matrix is a two-form, and the identification is canonical. The map

$$ u\wedge v\longmapsto uv^{\mathsf T}-vu^{\mathsf T} $$

sends a decomposable two-vector to an antisymmetric matrix of rank two, and it extends by linearity to an isomorphism $\Lambda^{2}\mathbb{R}^{n}\to\operatorname{Skew}_n$. Both sides have dimension $\binom{n}{2}=n(n-1)/2$, and the map is injective because $uv^{\mathsf T}-vu^{\mathsf T}=0$ forces $u$ and $v$ to be dependent.

Under the identification the bracket is the commutator of the two operators. For $X=u\wedge v$ and $Y=w\wedge z$ the matrix $X$ acts on $\mathbb{R}^{n}$ by $p\mapsto u\langle v,p\rangle-v\langle u,p\rangle$, which is the infinitesimal rotation in the plane spanned by $u$ and $v$. The commutator of two such operators is again of the form $\sum u_{k}\wedge v_{k}$, and the identity

$$ [X,Y]=XY-YX $$

is the statement that $\Lambda^{2}\mathbb{R}^{n}$ with the commutator is $\mathfrak{so}(n)$. The Killing form of the Lie algebra is then the form induced by the Euclidean form on $\Lambda^{2}$, up to the factor $n-2$.

The rank-two elements are the simple rotations. An element of $\operatorname{Skew}_n$ has even rank, and rank two exactly when it is decomposable, $X=u\wedge v$. Such an element generates a one-parameter group of rotations in a single plane, which is the subject of The Exponential of the Antisymmetric Part: One-Parameter Rotations. The higher-rank elements are sums of several plane rotations and are treated in Simple and Non-Simple Rotations: One Angle at $n\le3$, Two Angles at $n\ge4$.

Simplicity and the Exception at $n=4$

For $n\ge3$ the algebra has no proper ideal, except at $n=4$. A nonzero ideal $\mathfrak{a}$ of $\mathfrak{so}(n)$ is invariant under the adjoint action, and the adjoint action of $\mathfrak{so}(n)$ on $\Lambda^{2}\mathbb{R}^{n}$ is irreducible for $n\ge3$, $n\neq4$; so $\mathfrak{a}$ is everything. The irreducibility is the standard statement that $\mathrm{SO}(n)$ acts transitively on the decomposable two-vectors and their span. As a numerical check, the ideal generated by a random element together with its first brackets spans the whole algebra: the span has dimension $3,6,10,15$ at $n=3,4,5,6$.

At $n=4$ the algebra splits into two three-dimensional ideals. The Hodge star on $\Lambda^{2}\mathbb{R}^{4}$ satisfies $\star^{2}=1$, so $\Lambda^{2}\mathbb{R}^{4}$ splits into the self-dual and anti-self-dual two-forms, each of dimension three. In the basis

$$ X_{01}\pm X_{23},\qquad X_{02}\mp X_{13},\qquad X_{03}\pm X_{12}, $$

the two eigenspaces are each closed under the bracket, of dimension three, and the brackets between them vanish, so

$$ \mathfrak{so}(4)\cong\mathfrak{so}(3)\oplus\mathfrak{so}(3),\qquad 6=3+3 . $$

The closure of each three-dimensional space and the dimension of the algebra they span were verified numerically on the explicit basis above, and the eigenvalue list of $\star$ is $+1,+1,+1,-1,-1,-1$.

At $n=2$ the algebra is one-dimensional and abelian. The single basis element is $J=E_{12}-E_{21}$, and $[J,J]=0$, so $\mathfrak{so}(2)$ is commutative, of dimension one and rank one. The simplicity statement therefore needs $n\ge3$, and the exception at $n=4$ shows that $\mathfrak{so}(n)$ is not simple for every larger $n$ either: it is simple for $n\ge3$, $n\neq4$.

The Rank and the Cartan Subalgebra

The rank is $\lfloor n/2\rfloor$. A maximal torus is spanned by the $\lfloor n/2\rfloor$ commuting rotations in the disjoint coordinate planes,

$$ \mathfrak{h}=\langle X_{12},X_{34},\dots\rangle, $$

and no larger commutative subalgebra of semisimple elements exists, since an antisymmetric matrix has pure imaginary eigenvalues occurring in conjugate pairs and each pair needs its own plane. The ranks at $n=2,3,4,5,6$ are $1,1,2,2,3$.

The roots are the differences and the sums of the plane coordinates. Writing an element of the torus as $\mathrm{diag}(i\theta_1,-i\theta_1,\dots)$, the roots of the complexified algebra are $\pm\epsilon_k\pm\epsilon_l$ with $k\neq l$, and each root space is one-dimensional, spanned by a complex combination of $E_{kl}$ and $E_{lk}$. Their number is

$$ \dim\mathfrak{so}(n)-\operatorname{rank}=2,4,8,12\quad(n=3,4,5,6), $$

which was checked by subtracting the rank from the dimension at those orders. Over $\mathbb{R}$ the root spaces pair into two-dimensional real planes, one pair per complex root, which is why the real form has no real root decomposition.

The compact form is the real algebra itself. Since $\mathfrak{so}(n)$ has negative definite Killing form, it is the compact real form of the complex Lie algebra $\mathfrak{so}(n,\mathbb{C})=\mathfrak{so}(n)\otimes\mathbb{C}$. The complexification is the algebra of complex antisymmetric matrices, the fixed points of the conjugation $X\mapsto\overline{X}$, and it is the Lie algebra of $\mathrm{SO}(n,\mathbb{C})$.

The Exponential and the Group

The exponential of an antisymmetric matrix is orthogonal. For $X^{\mathsf T}=-X$,

$$ \big(e^{X}\big)^{\mathsf T}=e^{X^{\mathsf T}}=e^{-X}=\big(e^{X}\big)^{-1}, $$

so $e^{X}\in\mathrm{O}(n)$, and since $t\mapsto e^{tX}$ is continuous with value $\mathrm{I}$ at $t=0$ and $\det e^{X}=e^{\operatorname{tr}X}=1$, the image lies in $\mathrm{SO}(n)$. The map $\exp\colon\mathfrak{so}(n)\to\mathrm{SO}(n)$ is surjective, with kernel containing $2\pi$ times the integral points of the torus, which is the statement that every rotation is a rotation in a family of planes.

The exponential is a local diffeomorphism and a covering for $n\ge3$. Its differential at the origin is the identity, so it is a local diffeomorphism there; the group $\mathrm{SO}(n)$ is connected and has fundamental group $\mathbb{Z}/2$ for $n\ge3$, so the exponential is the universal covering map only after passing to $\mathrm{Spin}(n)$. The double cover is the subject of The Rotor as a Square Root of a Rotation: the Two Signs and the Double Cover.

The Killing form is $(n-2)\operatorname{tr}(XY)$ and it is negative definite for $n\ge3$. Since $\mathfrak{so}(n)$ is simple for $n\ge3$, $n\neq4$, its Killing form is a multiple of the trace form, and the multiple is $n-2$:

$$ B(X,Y)=(n-2)\operatorname{tr}(XY),\qquad B(X,X)=-(n-2)\operatorname{tr}(XX^{\mathsf T})<0 . $$

At $n=2$ the factor vanishes and the algebra is abelian, so the Killing form is identically zero there; at $n=4$ the form is the sum of the two Killing forms of the two ideals and is still negative definite. The values of the factor at $n=3,4,5$ are $1,2,3$, and the computation of the form on the standard basis reproduces them.

The Case $n=2$

At $n=2$ the algebra is the line $\mathbb{R}J$ and the bracket vanishes. With $J=\begin{pmatrix}0&-1\\1&0\end{pmatrix}$ and $J^{2}=-\mathrm{I}$,

$$ \mathfrak{so}(2)=\mathbb{R}J,\qquad [J,J]=0,\qquad \exp(\theta J)=\cos\theta\,\mathrm{I}+\sin\theta\,J, $$

so the group is the circle and the algebra is the tangent line at the identity. The element $J$ is a complex structure: it satisfies $J^{2}=-\mathrm{I}$ and it turns $\mathbb{R}^{2}$ into a one-dimensional complex space.

The identification with $\Lambda^{2}\mathbb{R}^{2}$ is an isomorphism of one-dimensional spaces. There is one basis two-vector $e_1\wedge e_2$, and it maps to $E_{12}-E_{21}=J$. So the whole exterior square is the line of the rotation, and the statement that the bracket vanishes is the statement that $\Lambda^{2}$ of a two-dimensional space has no decomposable products left to bracket with.

The biquaternion reading is the statement that the rotation direction is one line. In the biquaternion algebra $\mathbb{B}\cong M_2(\mathbb{C})$ the antisymmetric part is the line $\mathbb{C}J$, and the idempotents $\tilde\Pi_\pm=\tfrac12(e_t\pm\tilde{Ζ})$ are symmetric. The rotation direction is therefore the unique complement of the Jordan algebra $\operatorname{Sym}_2$ inside $M_2(\mathbb{C})$, and the pair $(\operatorname{Sym}_2,\mathbb{R}J)$ is the smallest symmetric pair of the category.

Summary

The antisymmetric half is closed under the bracket and is the Lie algebra $\mathfrak{so}(n)$, of dimension $n(n-1)/2$ and rank $\lfloor n/2\rfloor$, isomorphic to $\Lambda^{2}\mathbb{R}^{n}$ with the commutator of operators. It is simple for $n\ge3$ except at $n=4$, where the Hodge star splits it into the self-dual and anti-self-dual ideals $\mathfrak{so}(3)\oplus\mathfrak{so}(3)$, and it is abelian at $n=2$. Its exponential is the rotation group, its Killing form is $(n-2)\operatorname{tr}(XY)$ and is negative definite for $n\ge3$, and its complexification is the algebra of complex antisymmetric matrices.

Summary of Notation

Symbol Meaning
$\mathfrak{so}(n)=\operatorname{Skew}_n$ the Lie algebra of antisymmetric matrices
$X_{ij}=E_{ij}-E_{ji}$ the standard basis elements, one per pair $i
$\Lambda^{2}\mathbb{R}^{n}$ the exterior square, isomorphic to $\mathfrak{so}(n)$ by $u\wedge v\mapsto uv^{\mathsf T}-vu^{\mathsf T}$
$\star$ the Hodge star, with $\star^{2}=1$ on $\Lambda^{2}\mathbb{R}^{4}$
$\lfloor n/2\rfloor$ the rank, the dimension of a maximal torus
$B(X,Y)=(n-2)\operatorname{tr}(XY)$ the Killing form of $\mathfrak{so}(n)$
$\mathrm{Spin}(n)$ the double cover of $\mathrm{SO}(n)$ for $n\ge3$

Further Reading