The Antisymmetric Sector, the Pfaffian and the Klein Quadric of $M_4(\mathbb{C})$

Introduction

The antisymmetric matrices of order four form a six-dimensional subspace $\Lambda^{2}\mathbb{C}^{4}$ of $M_4(\mathbb{C})$, closed under the commutator and isomorphic to the Lie algebra $\mathfrak{so}(4,\mathbb{C})$. On this sector the determinant has a square root: the Pfaffian

$$ \mathrm{Pf}(A)=a_{12}a_{34}-a_{13}a_{24}+a_{14}a_{23}, $$

a quadratic form in the entries with $\mathrm{Pf}(A)^{2}=\det A$. The vanishing of the Pfaffian is the condition for the corresponding two-form to be decomposable, and the six coordinates $p_{ij}$ of a decomposable two-form are the Plücker coordinates of a line in $\mathbb{P}^{3}$. The Pfaffian quadric in $\mathbb{P}^{5}$ is therefore the Grassmannian of lines in $\mathbb{P}^{3}$, the Klein quadric, a smooth quadric of dimension four with two rulings.

This article states the identification, the Pfaffian and its invariance, the rank criterion, the Hodge star and the self-dual splitting, the Klein quadric and the Plücker relation, and the two rulings. The determinant and the singular set of the full algebra are The Forms, the Norms and the Singular Set of $M_4(\mathbb{C})$; the two chiral $\mathfrak{su}_2$'s of the frame are The Peirce $\mathfrak{su}_2$ and the Chiral $\mathfrak{su}_2\times\mathfrak{su}_2$ of $M_4(\mathbb{C})$, and the relation between them and the splitting below is stated, not reproved.

The Antisymmetric Sector

Definition. Let $\Lambda^{2}\mathbb{C}^{4}$ be the space of antisymmetric bilinear forms on $\mathbb{C}^{4}$, identified with the matrices $A$ with $A^{\mathsf T}=-A$.

Proposition. The space $\Lambda^{2}\mathbb{C}^{4}$ has dimension $6$, with basis $e_{i}\wedge e_{j}$ for $1\le i

Proof. An antisymmetric matrix is determined by its entries above the diagonal, of which there are $\binom{4}{2}=6$; the matrices $e_{i}\wedge e_{j}=E_{ij}-E_{ji}$ are linearly independent and span the space.

The axis-free form. In the coefficient basis $e_t,\vec e_x,\vec e_y,\vec e_z$ of the biquaternion algebra $\mathbb{B}=\mathbb{C}\otimes_{\mathbb{R}}\mathbb{H}$ the antisymmetric sector is spanned by the six axis-free operators $T_{tx},T_{ty},T_{tz},T_{xt},T_{yt},T_{zt}$, and the symmetric sector by the other ten, $T_{tt},T_{xx},T_{yy},T_{zz},T_{xy},T_{xz},T_{yx},T_{yz},T_{zx},T_{zy}$, each of which is a symmetric signed permutation matrix. The six are exactly the mixed time-space pairs, so the split of $M_4(\mathbb{C})$ into symmetric and antisymmetric matrices is a coordinate split in that basis, with no diagonalisation. The sixteen matrices are $M_4(\mathbb{C})$ Axis-Free Representation: Operators; the trace pairing that separates the two halves is The Forms, the Norms and the Singular Set of $M_4(\mathbb{C})$.

The Lie algebra. The commutator of two antisymmetric matrices is antisymmetric:

$$ [A,B]^{\mathsf T}=(AB-BA)^{\mathsf T}=B^{\mathsf T}A^{\mathsf T}-A^{\mathsf T}B^{\mathsf T}=BA-AB=-[A,B]. $$

So $\Lambda^{2}\mathbb{C}^{4}$ is a Lie subalgebra of $\mathfrak{gl}_4(\mathbb{C})$, isomorphic to $\mathfrak{so}(4,\mathbb{C})$, of dimension $6$.

The exponentials. For antisymmetric $A$ the exponential $\exp A$ is orthogonal, $\exp(A)^{\mathsf T}\exp(A)=I$: the series commutes with transposition and $A^{\mathsf T}=-A$ gives $\exp(A)^{\mathsf T}=\exp(-A)=\exp(A)^{-1}$. The image is the identity component $SO(4,\mathbb{C})$.

The Pfaffian

Definition. For an antisymmetric $A$ the Pfaffian is

$$ \mathrm{Pf}(A)=a_{12}a_{34}-a_{13}a_{24}+a_{14}a_{23}. $$

Theorem. $\mathrm{Pf}(A)^{2}=\det A$.

Proof. Expand $\det A$ by the Leibniz formula and collect the terms. Of the $24$ permutations, the twelve that pair the indices $\{1,2\}$ with $\{3,4\}$ contribute $a_{12}a_{34}$, those that pair $\{1,3\}$ with $\{2,4\}$ contribute $-a_{13}a_{24}$, and those that pair $\{1,4\}$ with $\{2,3\}$ contribute $a_{14}a_{23}$; each group enters with the same sign pattern in the two squares, and the cross terms cancel by antisymmetry. Equivalently, the identity is the case $n=2$ of the general Pfaffian expansion, and it is checked directly on the basis: $\mathrm{Pf}(e_{1}\wedge e_{2}+e_{3}\wedge e_{4})=1$ and $\det=1$.

The invariance. For $U\in GL_4(\mathbb{C})$,

$$ \mathrm{Pf}(UAU^{\mathsf T})=\det(U)\,\mathrm{Pf}(A). $$

Proof. Both sides are quadratic in $A$ and multiplicative in $U$; the identity is checked on the basis, where $U(e_{i}\wedge e_{j})U^{\mathsf T}=(Ue_{i})\wedge(Ue_{j})$ and the Pfaffian of a decomposable form $u\wedge v$ is the determinant of the $2\times2$ matrix of coordinates of $u$ and $v$ in a completion to a basis, which scales by $\det U$. For $U\in SO(4)$ the Pfaffian is invariant; for $U$ with $\det U=-1$ it changes sign.

The rank criterion. An antisymmetric matrix has even rank. For $A\in\Lambda^{2}\mathbb{C}^{4}$,

$$ \operatorname{rank}A=0\iff A=0,\qquad \operatorname{rank}A=2\iff \mathrm{Pf}(A)=0,\ A\ne0,\qquad \operatorname{rank}A=4\iff \mathrm{Pf}(A)\ne0. $$

Proof. The rank is even because the matrix is anti-self-adjoint for the symmetric form, and the only even ranks in dimension four are $0,2,4$. If the rank is four, the matrix is invertible and $\det A\ne0$, so $\mathrm{Pf}(A)\ne0$. If the rank is two, the kernel has dimension two; choosing a basis adapted to the kernel, the matrix has the form $u\wedge v$ in the remaining two dimensions, of rank two and Pfaffian zero. If the rank is zero the matrix vanishes. The converses follow from the same normal form.

The normal form. Every nonzero antisymmetric matrix of rank two is $A=u\wedge v$ for two independent vectors $u,v$; every antisymmetric matrix of rank four is congruent to $e_{1}\wedge e_{2}+e_{3}\wedge e_{4}$, with Pfaffian $1$.

The Hodge Star and the Self-Dual Splitting

Definition. With the standard symmetric form on $\mathbb{C}^{4}$, the Hodge star on $\Lambda^{2}$ is the linear map defined on the basis by

$$ *(e_{1}\wedge e_{2})=e_{3}\wedge e_{4},\qquad *(e_{1}\wedge e_{3})=-e_{2}\wedge e_{4},\qquad *(e_{1}\wedge e_{4})=e_{2}\wedge e_{3}, $$

and extended antisymmetrically.

Proposition. $*^{2}=I$, and $*$ is symmetric for the form $\langle\omega,\eta\rangle$ defined by $\omega\wedge\eta=\langle\omega,*\eta\rangle\,e_{1}\wedge e_{2}\wedge e_{3}\wedge e_{4}$.

Proof. The three displayed values give $*^{2}=1$ on a basis, and the symmetry is the identity $\omega\wedge\eta=\eta\wedge\omega$ together with $*^{2}=1$.

The splitting. The eigenvalues of $*$ are $+1$ and $-1$, and the eigenspaces

$$ \Lambda^{2}_{+}=\{\omega:*\omega=\omega\},\qquad \Lambda^{2}_{-}=\{\omega:*\omega=-\omega\} $$

have dimension $3$ each. A basis of $\Lambda^{2}_{+}$ is

$$ e_{1}\wedge e_{2}+e_{3}\wedge e_{4},\qquad e_{1}\wedge e_{3}-e_{2}\wedge e_{4},\qquad e_{1}\wedge e_{4}+e_{2}\wedge e_{3}, $$

and a basis of $\Lambda^{2}_{-}$ is obtained by changing the signs of the second terms.

The Lie algebra. The splitting is a splitting of Lie algebras:

$$ \mathfrak{so}(4,\mathbb{C})=\Lambda^{2}_{+}\oplus\Lambda^{2}_{-}\cong\mathfrak{sl}_2(\mathbb{C})\oplus\mathfrak{sl}_2(\mathbb{C}). $$

Proof. The star commutes with the bracket, because it is an isometry of the form and the bracket is built from the metric; hence the two eigenspaces are Lie subalgebras. Each is three-dimensional, and the only three-dimensional simple complex Lie algebra is $\mathfrak{sl}_2$. The two summands commute, since $[\Lambda^{2}_{+},\Lambda^{2}_{-}]\subseteq\Lambda^{2}_{+}\cap\Lambda^{2}_{-}=0$.

The Pfaffian on the two halves. For $\omega\in\Lambda^{2}_{+}$,

$$ \omega\wedge\omega=\langle\omega,*\omega\rangle\,e_{1}\wedge e_{2}\wedge e_{3}\wedge e_{4}=\langle\omega,\omega\rangle\,e_{1}\wedge e_{2}\wedge e_{3}\wedge e_{4}, $$

and the form $\langle\cdot,\cdot\rangle$ is nondegenerate on the three-dimensional space. The restriction of the Pfaffian quadric to $\mathbb{P}(\Lambda^{2}_{+})$ is therefore a smooth conic, and the same holds on $\mathbb{P}(\Lambda^{2}_{-})$. The two conics are the chiral conics of the antisymmetric sector.

The Klein Quadric

The Plücker coordinates. A decomposable two-form $\omega=u\wedge v$ has coordinates

$$ p_{ij}=u_{i}v_{j}-u_{j}v_{i},\qquad 1\le i

the Plücker coordinates of the line $\mathbb{P}(\langle u,v\rangle)\subset\mathbb{P}^{3}$. They satisfy the Plücker relation

$$ p_{12}p_{34}-p_{13}p_{24}+p_{14}p_{23}=0, $$

which is exactly the vanishing of the Pfaffian.

Theorem (the Klein quadric). The image of the map

$$ \operatorname{Gr}(2,4)\longrightarrow\mathbb{P}(\Lambda^{2}\mathbb{C}^{4})=\mathbb{P}^{5},\qquad \langle u,v\rangle\longmapsto[u\wedge v], $$

is the quadric $Q=\{\mathrm{Pf}=0\}$. The map is an isomorphism onto $Q$, and $Q$ is a smooth quadric of dimension $4$.

Proof. The map is well defined: a change of basis of $\langle u,v\rangle$ multiplies $u\wedge v$ by the determinant, so the class in $\mathbb{P}^{5}$ is unchanged. It is injective: the line is recovered as the projectivisation of the span of the vectors whose wedge with $\omega$ vanishes, a two-dimensional subspace for a nonzero decomposable $\omega$. Its image lies in $Q$ by the Plücker relation. Conversely, every nonzero $\omega$ with $\mathrm{Pf}(\omega)=0$ has rank two by the rank criterion, hence is decomposable, $\omega=u\wedge v$, and lies in the image. The quadric $Q$ is defined by a nondegenerate quadratic form in six variables, so it is smooth, and its dimension is $5-1=4$. The Grassmannian $\operatorname{Gr}(2,4)$ has dimension $2(4-2)=4$, consistent with the isomorphism.

The cone. The affine cone over $Q$ is the set of decomposable two-forms, including zero; it is the determinantal variety of rank-two antisymmetric matrices, of dimension $5$.

The two rulings. Through every point of $Q$ pass two planes contained in $Q$:

  • the plane of the lines through a fixed point $P\in\mathbb{P}^{3}$;
  • the plane of the lines contained in a fixed plane $H\subset\mathbb{P}^{3}$.

Each family is parametrised by a $\mathbb{P}^{3}$, the dual $\mathbb{P}^{3}$ for the second, and the two families are the rulings of $Q$. Two points of $Q$ lie on a common line of $Q$ exactly when the corresponding lines of $\mathbb{P}^{3}$ meet.

Proof. The lines through a fixed point $P$ form a $\mathbb{P}^{2}$ in $\mathbb{P}^{3}$, and their Plücker points satisfy the two linear equations $P\wedge\omega=0$, which define a plane in $\mathbb{P}^{5}$; that plane lies in $Q$ because every form in it is decomposable. The same argument applies to the lines in a fixed plane, with the dual equations. The two planes through a point of $Q$ are obtained by taking $P$ on the line and $H$ containing the line. The incidence statement is the Klein correspondence: two lines meet exactly when the corresponding two-forms have a common vector, which is the condition that their Plücker points are joined by a line lying in $Q$.

The Place of the Sector in $M_4(\mathbb{C})$

The trace form. On the antisymmetric sector the trace form is nondegenerate, and in the basis $e_{i}\wedge e_{j}$ it is diagonal:

$$ \operatorname{tr}\bigl((e_{i}\wedge e_{j})(e_{k}\wedge e_{l})\bigr)=-2\,\delta_{ik}\delta_{jl}, $$

so $\operatorname{tr}(A^{2})=-2\sum_{i

The determinant. The determinant of an antisymmetric matrix is the square of its Pfaffian, so on the sector the determinant is a perfect square. The singular set of the sector is the Pfaffian quadric, of dimension $4$ in $\mathbb{P}^{5}$, and not a hypersurface of the same type as the singular set $\{\det=0\}$ of the full algebra.

The chiral reading. The two three-dimensional halves $\Lambda^{2}_{\pm}$ are the two chiral pieces of the rotation algebra, and each is an $\mathfrak{sl}_2$. The frame of the algebra carries another pair of $\mathfrak{su}_2$'s, the Peirce chain and the chiral pair of The Peirce $\mathfrak{su}_2$ and the Chiral $\mathfrak{su}_2\times\mathfrak{su}_2$ of $M_4(\mathbb{C})$; the two pairs are different subalgebras. The star splitting is attached to the metric and the antisymmetric sector, the Peirce splitting to the idempotents and the frame.

The rank and the idempotents. An antisymmetric matrix is never an idempotent except zero, and its rank is even; the rank stratification of the sector is $0$, $2$ and $4$, against the five spectral strata of the full algebra. The two stratifications meet at the zero matrix only.

Summary

The antisymmetric sector $\Lambda^{2}\mathbb{C}^{4}$ is six-dimensional, is the Lie algebra $\mathfrak{so}(4,\mathbb{C})$, and carries the Pfaffian, a quadratic form whose square is the determinant. The Pfaffian vanishes exactly on the forms of rank at most two and is nonzero exactly on the invertible ones; its invariance under congruence is the invariance of the determinant up to the square of $\det U$. The Hodge star splits the sector into the self-dual and anti-self-dual three-spaces, and the Lie algebra into $\mathfrak{sl}_2\oplus\mathfrak{sl}_2$; the Pfaffian restricts to a smooth conic on each chiral plane. In $\mathbb{P}^{5}$ the Pfaffian quadric is the Klein quadric, isomorphic to the Grassmannian of lines in $\mathbb{P}^{3}$, with two rulings by planes. The sector is the even-rank part of the algebra, and its determinant is a square.

Quantity Value
$\dim\Lambda^{2}\mathbb{C}^{4}$ $6$
$\dim\Lambda^{2}_{\pm}$ $3$
Rank of an antisymmetric matrix $0,2,4$
$\mathrm{Pf}(e_{1}\wedge e_{2}+e_{3}\wedge e_{4})$ $1$
$\dim\operatorname{Gr}(2,4)=Q$ $4$
Ambient projective space $\mathbb{P}^{5}$
Rulings two, each parametrised by $\mathbb{P}^{3}$
Chiral conics two smooth conics

Summary of Notation

Symbol Meaning
$\Lambda^{2}\mathbb{C}^{4}$ the space of antisymmetric matrices, identified with two-forms
$e_{i}\wedge e_{j}$ the basis matrix $E_{ij}-E_{ji}$
$\mathrm{Pf}(A)$ the Pfaffian $a_{12}a_{34}-a_{13}a_{24}+a_{14}a_{23}$
$*$ the Hodge star on $\Lambda^{2}$
$\Lambda^{2}_{\pm}$ the self-dual and anti-self-dual eigenspaces of $*$
$p_{ij}$ the Plücker coordinates of a decomposable two-form
$Q=\{\mathrm{Pf}=0\}$ the Klein quadric in $\mathbb{P}^{5}$
$\operatorname{Gr}(2,4)$ the Grassmannian of two-planes in $\mathbb{C}^{4}$

Further Reading