The Trace Form and Its Inertia on the Two Halves of $M_2(\mathbb{R})$

Introduction

The form $\langle X,Y\rangle=\operatorname{tr}(XY)$ on $M_2(\mathbb{R})$ is symmetric, nondegenerate and of inertia $(3,1)$, positive definite on the symmetric half and negative definite on the antisymmetric line, so the real matrix algebra carries the geometry of a four-dimensional Minkowski space.

The article polarises the quadratic form $\operatorname{tr}(X^{2})$, exhibits an orthogonal basis with the signs $+1,+1,+1,-1$, describes the isotropic cone and its cone of light, reads the form through Cayley–Hamilton as $(\operatorname{tr}X)^{2}-2\det X$, and identifies the Lie algebra of the isometry group with the image of the adjoint action. It assumes The Two Halves as Orthogonal Complements at $n=2$, The Derivation $X\mapsto[A,X]$ and Its Anti-Self-Adjointness for the Trace Form and The Elementary Symmetric Functions: $\operatorname{tr}$ and $\det$ in Dimension Two.

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 four forms with their norms to The Four Forms and Their Algebraic Norms at $n=2$.

The Form and Its Polarisation

Definition. The trace form on $M_2(\mathbb{R})$ is $\langle X,Y\rangle=\operatorname{tr}(XY)$, the polar form of the quadratic form $Q(X)=\operatorname{tr}(X^{2})$.

Proposition. The form is symmetric and nondegenerate.

Proof. $\operatorname{tr}(XY)=\operatorname{tr}(YX)$ by cyclicity. If $\operatorname{tr}(XY)=0$ for every $Y$ then taking $Y=E_{ji}$ gives $x_{ij}=0$ for every pair of indices, so $X=0$.

Corollary. The polarisation $\langle X,Y\rangle=\tfrac12\bigl(Q(X+Y)-Q(X)-Q(Y)\bigr)$ recovers the form from the quadratic form, so the two carry the same information.

Proof. Expand $Q(X+Y)=\operatorname{tr}(X^{2})+\operatorname{tr}(Y^{2})+2\operatorname{tr}(XY)$ and compare.

The form is the derivative of the determinant along the diagonal. At the identity one has $\det(I+tX)=1+t\operatorname{tr}(X)+t^{2}\det(X)$, so the linear term is the trace, and the quadratic term is the determinant.

Proof. Expand the determinant of $I+tX$ in the two eigenvalues, or use the Cayley–Hamilton identity at $I$.

The Gram Matrix and the Signs

Theorem. In the basis $I/\sqrt2$, $\sigma_{3}/\sqrt2$, $\sigma_{1}/\sqrt2$, $J/\sqrt2$ the Gram matrix of the trace form is $\operatorname{diag}(1,1,1,-1)$.

Proof. $\operatorname{tr}(I^{2})=\operatorname{tr}(\sigma_{3}^{2})=\operatorname{tr}(\sigma_{1}^{2})=2$ and $\operatorname{tr}(J^{2})=-2$, while every mixed pairing vanishes by The Four Multiplication Tables: $SS$, $SK$, $KS$, $KK$ at $n=2$.

Corollary. The inertia is $(3,1)$: the form is positive definite of rank $3$ on the symmetric half and negative definite of rank $1$ on the antisymmetric line.

Proof. The Gram matrix in these coordinates is diagonal with three entries $+1$ and one entry $-1$, and the two halves are the spans of the first three and of the last basis vector.

Corollary. The symmetric half is positive definite and the antisymmetric line is the unique timelike direction, up to the sign of the vector.

Proof. The restriction of the form to $\operatorname{Sym}_2$ is positive definite with eigenvalues $1,1,2$ in the Pauli basis, as computed in The Two Halves as Orthogonal Complements at $n=2$, and the restriction to the line is $-2c^{2}$ at $cJ$.

The Isotropic Cone

Theorem. A matrix $X=S+cJ$ with $S$ symmetric is isotropic for the trace form exactly when $\operatorname{tr}(S^{2})=2c^{2}$.

Proof. $\operatorname{tr}((S+cJ)^{2})=\operatorname{tr}(S^{2})+2c\operatorname{tr}(SJ)+c^{2}\operatorname{tr}(J^{2})=\operatorname{tr}(S^{2})-2c^{2}$, since $SJ$ is antisymmetric and traceless.

Corollary. The isotropic cone is a double cone over the ellipsoid $\operatorname{tr}(S^{2})=2$ in the symmetric half, and its two nappes are separated by the sign of $c$.

Proof. The equation is $c^{2}=\tfrac12\operatorname{tr}(S^{2})$, which gives two values of $c$ for each direction $S$ and one point $S=0$, $c=0$.

Corollary. The cone contains no nonzero symmetric matrix, and every nonzero isotropic matrix has a nonzero antisymmetric part.

Proof. On the symmetric half the form is positive definite, so $\operatorname{tr}(S^{2})=0$ only for $S=0$, and the equation then forces $c=0$.

The cone is the light cone of a Lorentz form. The level sets $\operatorname{tr}(X^{2})=1$ are the hyperboloids of a Minkowski space of dimension four, and the two halves are the space directions and the time direction.

Proof. The Gram matrix is $\operatorname{diag}(1,1,1,-1)$, which is the standard Lorentz form in four dimensions after a linear change of coordinates.

The Cayley–Hamilton Reading

Proposition. For every two-by-two matrix

$$ \operatorname{tr}(X^{2})=\bigl(\operatorname{tr}X\bigr)^{2}-2\det X . $$

Proof. The eigenvalues $\lambda_{1},\lambda_{2}$ give $\operatorname{tr}(X^{2})=\lambda_{1}^{2}+\lambda_{2}^{2}$ and $(\operatorname{tr}X)^{2}-2\det X=(\lambda_{1}+\lambda_{2})^{2}-2\lambda_{1}\lambda_{2}=\lambda_{1}^{2}+\lambda_{2}^{2}$.

Corollary. The sign of $Q$ is the sign of $(\operatorname{tr}X)^{2}-2\det X$, so $Q$ vanishes on the cone $\det X=\tfrac12(\operatorname{tr}X)^{2}$ and is negative on the matrices whose determinant is large with respect to their trace, such as $J$.

Proof. The identity writes the form as a difference of the two invariants, and $\det J=1$ with $\operatorname{tr}J=0$ gives $Q(J)=-2$.

Corollary. On the traceless plane the form is $-\operatorname{tr}(X^{2})=2\det X$ up to sign, so the form and the determinant carry the same data there.

Proof. With $\operatorname{tr}X=0$ the identity gives $\operatorname{tr}(X^{2})=-2\det X$.

The Isometry Group

Theorem. The isometry group of the trace form is $\mathrm{O}(3,1)$, of dimension $6$.

Proof. The isometry group of a nondegenerate form of inertia $(3,1)$ on a space of dimension four is the orthogonal group of that inertia, and the dimension of the orthogonal group of a form in $n$ dimensions is $n(n-1)/2=6$ for $n=4$.

Corollary. The Lie algebra of the isometry group is the space of endomorphisms $T$ with $\langle TX,Y\rangle+\langle X,TY\rangle=0$, of dimension $6$.

Proof. The Lie algebra of a linear group is the tangent space at the identity, which consists of the endomorphisms $T$ with $e^{tT}$ an isometry to the first order, and that condition is the display.

Theorem. The adjoint action embeds $\mathfrak{sl}_2(\mathbb{R})$ into the isometry algebra, as the copy of $\mathfrak{so}(2,1)$.

Proof. Every $\operatorname{ad}_A$ is anti-self-adjoint for the form by The Derivation $X\mapsto[A,X]$ and Its Anti-Self-Adjointness for the Trace Form, the kernel of the adjoint action is the scalars, and the traceless algebra has dimension $3$; the compact dual of $\mathfrak{sl}_2(\mathbb{R})$ is $\mathfrak{so}(3)$ and the split one is $\mathfrak{so}(2,1)$, so the image is the split copy.

Corollary. The conjugation $X\mapsto gXg^{-1}$ by $g\in\mathrm{SL}_2(\mathbb{R})$ is an isometry, and these are the isometries produced by the adjoint action.

Proof. $\operatorname{tr}(gXg^{-1}\,gYg^{-1})=\operatorname{tr}(XY)$ for every invertible $g$, and the derivative at the identity is $\operatorname{ad}_A$ with $A$ traceless.

The one-sided multiplications are not isometries. The map $X\mapsto JX$ does not preserve the form, since $\langle JI,JI\rangle=\operatorname{tr}(J^{2})=-2$ while $\langle I,I\rangle=2$; so the isometry algebra is not obtained by multiplying on one side.

Proof. The value at the identity matrix is computed directly, and an isometry cannot change the squared length of a vector.

Corollary. The isometry algebra is strictly larger than the image of the adjoint action, by three dimensions, and it is not generated by the one-sided multiplications of the algebra.

Proof. The image has dimension $3$ and the algebra has dimension $6$; and the multiplication on one side already fails at the single vector $I$.

Summary

The trace form $\langle X,Y\rangle=\operatorname{tr}(XY)$ on $M_2(\mathbb{R})$ is symmetric and nondegenerate, with Gram matrix $\operatorname{diag}(1,1,1,-1)$ in the normalised Pauli basis, hence of inertia $(3,1)$: positive definite on the symmetric half and negative definite on the antisymmetric line.

The form is the polar form of $\operatorname{tr}(X^{2})$, which Cayley–Hamilton writes as $(\operatorname{tr}X)^{2}-2\det X$; its isotropic cone is the double cone $c^{2}=\tfrac12\operatorname{tr}(S^{2})$ over the ellipsoid in the symmetric half, and the isometry group is $\mathrm{O}(3,1)$ of dimension $6$, whose Lie algebra receives the adjoint action of the algebra.

Summary of Notation

Symbol Meaning
$\langle X,Y\rangle$ $\operatorname{tr}(XY)$, the trace form
$Q(X)$ $\operatorname{tr}(X^{2})$, the quadratic form
$\sigma_{3},\sigma_{1}$ traceless symmetric Pauli matrices
$J$ the antisymmetric unit, the timelike direction
$\mathrm{O}(3,1)$ the isometry group of the form
$\mathfrak{so}(2,1)$ the three-dimensional part of the isometry algebra

Further Reading