Rotors Acting on the Bivector Space: the Adjoint Action at $n=4$
Introduction
A rotation of four-space acts on the six-dimensional space $\Lambda^{2}\mathbb{C}^{4}$, which splits into the self-dual and anti-self-dual three-spaces, and the two factors of $\mathrm{Spin}(4)=SU(2)\times SU(2)$ act on the two three-spaces separately; on the self-dual space the eigenvalues are $1$ and $e^{\pm2i\theta_+}$ and on the anti-self-dual space $1$ and $e^{\pm2i\theta_-}$, so each three-space carries one angle doubled.
The article computes the second exterior power of a rotation, proves that the Hodge star splits the bivector space and that a rotation preserves the two halves, identifies the action on each half with the adjoint action of one summand of $\mathfrak{so}(4)$, computes the eigenvalues and the Casimir, and writes the six generators as the three left and the three right matrices of the axis-free frame.
It assumes The Spectral Theorem: the Anti-Hermitian Case and the Reality of the Spectrum at $n=4$ and The Rotor as a Square Root of a Rotation: the Two Signs and the Double Cover at $n=4$. The model at order two is Rotors Acting on the Bivector Space at $n=2$: the Adjoint Action; the operator version is The Two-Sided Operators and the Segre Cone at $n=4$.
The Second Exterior Power
Definition. For $Q\in SO(4,\mathbb{C})$ the second exterior power $\Lambda^{2}Q$ acts on $\Lambda^{2}\mathbb{C}^{4}$ by $u\wedge v\mapsto Qu\wedge Qv$.
Proposition. $\Lambda^{2}Q$ is linear, invertible with inverse $\Lambda^{2}(Q^{-1})$, and the assignment is a homomorphism: $\Lambda^{2}(QQ')=\Lambda^{2}Q\,\Lambda^{2}Q'$.
Proof. Bilinearity and the alternating property give the induced map, and the functoriality is the identity $Q Q'u\wedge QQ'v$.
Corollary. The six-dimensional space $\Lambda^{2}\mathbb{C}^{4}$ carries a representation of $SO(4,\mathbb{C})$, and through the covering also a representation of $\mathrm{Spin}(4)$.
Proof. The homomorphism property, and the covering map composes with it.
The trace. $\operatorname{tr}\Lambda^{2}Q=(\operatorname{tr}Q)^{2}-\operatorname{tr}(Q^{2})$, divided by two for the alternating convention.
Proof. On the basis $e_i\wedge e_j$ with $i Corollary. For a rotation the eigenvalues of $\Lambda^{2}Q$ are the pairwise products of the eigenvalues of $Q$. Proof. Triangularise $Q$ and read the diagonal in the corresponding bivector basis. Definition. The Hodge star is the map $\ast:\Lambda^{2}\mathbb{C}^{4}\to\Lambda^{2}\mathbb{C}^{4}$ sending $e_i\wedge e_j$ to $\pm e_k\wedge e_l$ with the complement in cyclic order; it satisfies $\ast^{2}=\mathrm I$. Proof. The complement of a pair in four indices is a pair, and complementing twice returns the original sign. Definition. $\Lambda^{2}_{+}$ and $\Lambda^{2}_{-}$ are the eigenspaces of $\ast$ for the eigenvalues $+1$ and $-1$, the self-dual and anti-self-dual spaces, each of dimension $3$. Proof. The eigenvalues are $\pm1$ and the two eigenspaces have equal dimension, the sum being six. Theorem. For $Q\in SO(4,\mathbb{C})$ the operator $\Lambda^{2}Q$ commutes with the Hodge star, so it preserves both three-spaces. Proof. The Hodge star is natural under an orientation-preserving isometry: $\ast(\Lambda^{2}Q\,\omega)=\Lambda^{2}Q(\ast\omega)$ for $\det Q=1$. Corollary. The representation of $SO(4,\mathbb{C})$ on $\Lambda^{2}$ is the direct sum of two three-dimensional representations, and the covering group acts through the same two. Proof. Invariance of the two summands, and the sum is the whole. The concrete basis. With $J_1=e_1\wedge e_2+e_3\wedge e_4$, $J_2=e_1\wedge e_3+e_4\wedge e_2$, $J_3=e_1\wedge e_4+e_2\wedge e_3$ the three are self-dual, and the three with the signs changed on the second term are anti-self-dual. Proof. $\ast(e_1\wedge e_2)=e_3\wedge e_4$, $\ast(e_1\wedge e_3)=e_4\wedge e_2$, $\ast(e_1\wedge e_4)=e_2\wedge e_3$. Theorem. The self-dual three-space carries the adjoint representation of the first summand of $\mathfrak{so}(4)$ and the trivial action of the second, and the anti-self-dual three-space the reverse. Proof. The summands act on $\Lambda^{2}$ by the commutator, which is the derivative of the conjugation action; the self-dual part of $\mathfrak{so}(4)$ fixes $\Lambda^{2}_{+}$ pointwise as a bivector identity and acts on it by its own adjoint action, and it annihilates $\Lambda^{2}_{-}$; the anti-self-dual part does the reverse. Corollary. A rotor which is a single left quaternion acts on $\Lambda^{2}_{+}$ as an element of $SO(3)$ and on $\Lambda^{2}_{-}$ as the identity; a single right quaternion does the reverse. Proof. The left quaternion generates the self-dual summand, and the exponential of a self-dual element acts trivially on the anti-self-dual space. Corollary. The two actions commute, and the representation of $\mathrm{Spin}(4)$ on $\Lambda^{2}$ is the sum of the two adjoint representations. Proof. The two summands commute and each acts trivially on the other's space. The isoclinic reading. A rotation acts trivially on one of the two three-spaces exactly when it is isoclinic in the corresponding family. Proof. Triviality on $\Lambda^{2}_{-}$ means $\theta_-=0$, that is $\theta_1=\theta_2$, and the other case likewise. Theorem. For $Q$ with angles $\theta_1,\theta_2$ the eigenvalues of $\Lambda^{2}Q$ are
$$
1,\ 1,\ e^{\pm2i\theta_+},\ e^{\pm2i\theta_-},
\qquad \theta_\pm=\tfrac{\theta_1\pm\theta_2}{2},
$$
and on $\Lambda^{2}_{+}$ they are $1,e^{\pm2i\theta_+}$ while on $\Lambda^{2}_{-}$ they are $1,e^{\pm2i\theta_-}$. Proof. The eigenvalues of $Q$ are $e^{\pm i\theta_1},e^{\pm i\theta_2}$; the pairwise products are $1=e^{i\theta_1}e^{-i\theta_1}$, $1=e^{i\theta_2}e^{-i\theta_2}$, $e^{i(\theta_1+\theta_2)}=e^{2i\theta_+}$ and $e^{i(\theta_1-\theta_2)}=e^{2i\theta_-}$ with their inverses. The self-dual and the anti-self-dual space separate the two nontrivial pairs: on $\Lambda^{2}_{+}$ the operator is a rotation of the three-space with angle $2\theta_+$, and on $\Lambda^{2}_{-}$ one with angle $2\theta_-$. Corollary. The angle on each three-space is twice the angle of the corresponding one-sided quaternion, and the isoclinic rotations are those with two eigenvalues equal to one on one of the spaces. Proof. The factor two is the factor of the pairwise product, and the rest is the previous corollary. Corollary. $\operatorname{tr}\Lambda^{2}Q=2+2\cos2\theta_++2\cos2\theta_-=2+4\cos\theta_1\cos\theta_2$, which at $\theta_1=\theta_2=\theta$ is $4+2\cos2\theta$. Proof. Sum the eigenvalues and use $2\cos2\theta_++2\cos2\theta_-=4\cos\theta_1\cos\theta_2$. The determinant. $\det\Lambda^{2}Q=1$, so the second exterior power map lands in $SL(\Lambda^{2})$. Proof. The eigenvalues multiply to $1$ in each three-space, since $e^{2i\theta}e^{-2i\theta}=1$. Proposition. On each three-space the Casimir of the corresponding summand is a scalar, and with the trace form as the invariant pairing the scalar is one on the adjoint three-space and $3/8$ on the defining two-space. Proof. The Casimir is invariant, so it acts as a scalar on an irreducible representation; the adjoint computation is $\operatorname{ad}(J_1)^{2}+\operatorname{ad}(J_2)^{2}+\operatorname{ad}(J_3)^{2}=-2\mathrm I$ with the normalisation of the trace form, giving $-\tfrac12(-2\mathrm I)=\mathrm I$; the defining computation is $\sum_k(i\sigma_k/2)^{2}=-\tfrac34\mathrm I$, giving $3\mathrm I/8$. Corollary. The two Casimirs of the two summands commute and are diagonal in the bivector basis, so the bivector space carries two commuting angular momenta. Proof. The scalar Casimirs of the two summands, each acting trivially on the other's space. Corollary. The Killing form of $\mathfrak{so}(4)$ is the sum of the two Killing forms, and on each three-space it is negative definite. Proof. The two summands are orthogonal ideals, and each is a copy of $\mathfrak{sl}_2$ with negative definite Killing form. Theorem. In the axis-free frame the three left matrices $L_x,L_y,L_z$ span $\Lambda^{2}_{+}$ and the three right matrices $R_x,R_y,R_z$ span $\Lambda^{2}_{-}$, and each triple obeys the quaternion brackets. Proof. The left multiplications by the three imaginary units rotate two planes in the same sense, hence are self-dual; the right ones rotate them in opposite senses, hence are anti-self-dual; and $[L_x,L_y]=2L_z$ in the normalisation of the frame. Corollary. The six-dimensional bivector space is spanned by the six signed permutations, and the Hodge star is the identity on the left triple and minus the identity on the right triple. Proof. The two triples and the definition of the star. Corollary. An element $\theta_1L_i+\theta_2R_i$ acts on $\Lambda^{2}_{+}$ with angle $2\theta_1$ and on $\Lambda^{2}_{-}$ with angle $2\theta_2$, so the two angles of the rotation are read directly from the two triples. Proof. The two triples are the two summands, and the theorem above. The finite shadow. The group of the sixteen $T_{\mu\nu}$ acts on $\Lambda^{2}$ through the two copies of $Q_8$, and the two actions are the two finite shadows of the two factors. Proof. $T_{\mu\nu}=L_\mu R_\nu$ acts as the product of the two one-sided actions, each of which is a signed permutation of the corresponding triple. A rotation acts on the six-dimensional bivector space by its second exterior power, which commutes with the Hodge star and therefore preserves the self-dual and the anti-self-dual three-spaces. Each three-space carries the adjoint action of one summand of $\mathfrak{so}(4)=\mathfrak{sl}_2\oplus\mathfrak{sl}_2$ and the trivial action of the other. The eigenvalues are $1$ and $e^{\pm2i\theta_+}$ on the self-dual space and $1$ and $e^{\pm2i\theta_-}$ on the anti-self-dual one, with $\theta_\pm=(\theta_1\pm\theta_2)/2$; the angles are doubled by passing to bivectors, and a rotation acts trivially on one three-space exactly when it is isoclinic in the corresponding family. The Casimir of each summand is a scalar on its three-space, one in the adjoint normalization; the Killing form restricts to a negative definite form on each. In the axis-free frame the left triple spans the self-dual space and the right triple the anti-self-dual one.The Two Three-Spaces
The Action of the Two Factors
The Eigenvalues and the Two Angles
The Casimir and the Killing Form
The Axis-Free Matrices
Summary
Summary of Notation
Symbol
Meaning
$\Lambda^{2}\mathbb{C}^{4}$
the six-dimensional bivector space
$\Lambda^{2}_{\pm}$
the self-dual and anti-self-dual three-spaces
$\ast$
the Hodge star
$\theta_\pm=(\theta_1\pm\theta_2)/2$
the two half-angles
$L_k,R_k$
the two triples in the axis-free frame
$J_1,J_2,J_3$
the self-dual bivector basis
Further Reading