The Matrix Algebra $M_2(\mathbb{C})$

Introduction

The algebra $M_2(\mathbb{C})$ is the algebra of $2\times2$ complex matrices, and everything in it is decided by four matrix units and one product rule.

This article fixes the algebra that the whole category reads: the matrix units and the rule $E_{ij}E_{kl}=\delta_{jk}E_{il}$, the trace, the determinant and the rank, the centre and the simplicity, the real dimension $8$, the group of invertibles and the projective group, and the identification with the biquaternion algebra $\mathbb{B}$. It assumes only the linear algebra of $\mathbb{C}^{2}$. It defers the element to The Element: Rank, Trace, Determinant, Characteristic Polynomial at $n=2$, the real matrices to The Two Base Fields: $M_2(\mathbb{R})$ inside $M_2(\mathbb{C})$, and the ideal theory to The Ideals, the Centre and the Simplicity of $M_2(\mathbb{C})$.

No quadratic form is used in the proofs. The trace pairing is recorded as symmetric and nondegenerate, nothing more; its inertia is The Trace Form and Its Inertia on the Two Halves of $M_2(\mathbb{R})$.

The Algebra

Definition. $M_2(\mathbb{C})$ is the set of $2\times2$ arrays of complex numbers with the entrywise sum and the product $(AB)_{ij}=\sum_{k}a_{ik}b_{kj}$. The matrix units $E_{ij}$ have a $1$ in position $(i,j)$ and $0$ elsewhere.

Proposition (the product rule). The four units are a $\mathbb{C}$-basis, and

$$ \boxed{\;E_{ij}E_{kl}=\delta_{jk}E_{il}.\;} $$

Proof. The $(p,q)$-entry of $E_{ij}E_{kl}$ is $\sum_{r}\delta_{ip}\delta_{jr}\delta_{kr}\delta_{lq}=\delta_{jk}\delta_{ip}\delta_{lq}$, which is that of $\delta_{jk}E_{il}$. Independence holds because the $(i,j)$-entry of a combination is its own coefficient.

Theorem. With this product the algebra is associative, unital, non-commutative, of complex dimension $4$ and real dimension $8$.

Proof. Both $(AB)C$ and $A(BC)$ have $(i,l)$-entry $\sum_{j,k}a_{ij}b_{jk}c_{kl}$. The unit is $I$. Non-commutativity: $E_{12}E_{21}=E_{11}$ while $E_{21}E_{12}=E_{22}$.

The product is composition. The map $A\mapsto(x\mapsto Ax)$ is an isomorphism onto $\operatorname{End}_{\mathbb{C}}(\mathbb{C}^{2})$: injective because $Ae_{j}$ is the $j$-th column, surjective because a map is fixed by the images of a basis. The transpose is an anti-isomorphism onto the opposite algebra, since $(AB)^{\mathsf T}=B^{\mathsf T}A^{\mathsf T}$.

Trace and Determinant

Definition. For $A=\begin{pmatrix}a&b\\c&d\end{pmatrix}$, the trace is $\operatorname{tr}A=a+d$, the determinant is $\det A=ad-bc$, and the characteristic polynomial is

$$ \chi_A(\lambda)=\det(\lambda I-A)=\lambda^{2}-(\operatorname{tr}A)\lambda+\det A . $$

Proposition (the trace). The trace is linear, cyclic, $\operatorname{tr}(AB)=\operatorname{tr}(BA)$, and it is the only linear functional vanishing on all commutators, up to a scalar.

Proof. The cyclic identity is the equality $a_{11}b_{11}+a_{12}b_{21}+a_{21}b_{12}+a_{22}b_{22}=\operatorname{tr}(BA)$. A functional vanishes on commutators exactly when it is invariant under transposing the two factors. The commutator subspace is the traceless part, of dimension $3$, so its annihilator in the dual is one-dimensional, and the trace spans it.

Proposition (the determinant). The determinant is multiplicative, $\det(AB)=\det A\det B$, and it is the only multiplicative polynomial of degree two, normalised to $1$ at $I$, that vanishes on singular matrices.

Proof. Multiplicativity is that of the determinant of a linear map. The normalisation is the Leibniz formula over the two permutations of $S_2$.

Cayley–Hamilton. Substituting $A$ gives $A^{2}-(\operatorname{tr}A)A+(\det A)I=0$. At order two this is a four-entry computation: the diagonal identities are $a^{2}+bc=(a+d)a-ad$ and $cb+d^{2}=(a+d)d-ad$. Whenever $\det A\neq0$ the identity gives the inverse

$$ A^{-1}=\frac{1}{ad-bc}\begin{pmatrix}d&-b\\-c&a\end{pmatrix}. $$

Worked example. For $A=\begin{pmatrix}1&2\\3&4\end{pmatrix}$ one has $\operatorname{tr}A=5$ and $\det A=-2$, so $\chi_A=\lambda^{2}-5\lambda-2$, with roots $(5\pm\sqrt{33})/2$, and $A^{-1}=-\tfrac12\begin{pmatrix}4&-2\\-3&1\end{pmatrix}$.

Rank and the Singular Set

The rank is the dimension of the image and takes the three values $0,1,2$. It is $2$ exactly off the determinant hypersurface and $0$ only at the origin, so the rank-one matrices are the remaining stratum.

Proposition (the strata). The rank-$r$ set has complex dimension $r(4-r)$: that is $0,3,4$ for $r=0,1,2$.

Proof. A rank-one matrix is $uv^{\mathsf T}$ with $u,v\neq0$, and the pair is fixed up to $u\mapsto\lambda u$, $v\mapsto\lambda^{-1}v$, a count of $2+2-1=3$. The rank-two set is open in the four-dimensional space.

The singular set is the determinant hypersurface. It has complex dimension $3$ and real codimension $2$, and it is singular only at $0$, because the gradient of the determinant is the cofactor matrix, which vanishes only at rank $0$. It is not the null cone of a form on the algebra: the determinant is quadratic, but the singular set is its zero set inside the algebra, and the algebra is read with a positive definite form.

The Centre and the Simplicity

Theorem (the centre). $Z(M_2(\mathbb{C}))=\mathbb{C}I$.

Proof. If $A$ commutes with $E_{11}$ then $a_{12}=a_{21}=0$, so $A$ is diagonal; commuting with $E_{12}$ then gives $a_{11}=a_{22}$; conversely a scalar matrix is central.

Theorem (simplicity). The only two-sided ideals are $0$ and the algebra.

Proof. Let $I\neq0$ and $0\neq A\in I$ with $a_{ij}\neq0$. Then $E_{ii}AE_{jj}=a_{ij}E_{ij}\in I$, so $E_{ij}\in I$, and $E_{ki}E_{ij}E_{jl}=E_{kl}\in I$ for all $k,l$.

Corollary. The algebra is central simple. Every automorphism is inner and every derivation is inner, by Skolem–Noether and Noether; the proofs are The Ideals, the Centre and the Simplicity of $M_2(\mathbb{C})$.

The algebra is not a division algebra. Every off-diagonal unit is a nonzero nilpotent, $E_{12}^{2}=0$, so there are zero divisors and the non-invertibles are exactly the singular matrices.

The Group of Invertibles and the Projective Group

Proposition. $GL_2(\mathbb{C})=\{\det\neq0\}$ is a complex algebraic group of complex dimension $4$. The determinant maps it onto $\mathbb{C}^{\times}$, with kernel $SL_2(\mathbb{C})$ of complex dimension $3$.

Proof. The group of invertibles is the complement of a hypersurface in $\mathbb{C}^{4}$; the determinant is polynomial, multiplicative and onto, since $\operatorname{diag}(d,1)$ has determinant $d$; its kernel is a hypersurface, of dimension $3$.

Proposition. $PGL_2(\mathbb{C})=GL_2(\mathbb{C})/\mathbb{C}^{\times}I\cong SL_2(\mathbb{C})/\{\pm I\}$, of complex dimension $3$.

Proof. Multiply a class by $\lambda$ with $\lambda^{2}=(\det A)^{-1}$ to reach determinant $1$; two such representatives differ by a scalar of square $1$.

The action. $PGL_2$ acts by $[g]\cdot A=gAg^{-1}$, and the orbits are the conjugacy classes, classified by $(\operatorname{tr},\det)$ in The Spectrum of a $2\times2$ Matrix. On the projective line the action is the Möbius action, which is strictly three-transitive.

The Reading over the Real Numbers

Every matrix splits into a Hermitian and an anti-Hermitian part. Put $A^{*}=\bar A^{\mathsf T}$; the Hermitian matrices satisfy $H^{*}=H$ and the anti-Hermitian ones $S^{*}=-S$.

Proposition. Every $A$ splits uniquely as

$$ A=H+S,\qquad H=\tfrac12(A+A^{*}),\quad S=\tfrac12(A-A^{*}), $$

so $M_2(\mathbb{C})=\mathfrak{u}(2)\oplus i\mathfrak{u}(2)$ over $\mathbb{R}$, with $\dim_{\mathbb{R}}\mathfrak{u}(2)=4$ and $\dim_{\mathbb{R}}\mathfrak{su}(2)=3$.

Proof. The star is conjugate-linear and involutive, so the two parts have the required symmetry and sum to $A$; they meet only at $0$ because $H=-H$ forces $H=0$. A Hermitian matrix has two real diagonal entries and one free off-diagonal complex entry, hence $4$ real parameters.

Worked example. For $A=\begin{pmatrix}1&2\\3&4\end{pmatrix}$,

$$ H=\tfrac12(A+A^{*})=\begin{pmatrix}1&\tfrac52\\ \tfrac52&4\end{pmatrix}, \qquad S=\tfrac12(A-A^{*})=\begin{pmatrix}0&-\tfrac12\\ \tfrac12&0\end{pmatrix}, $$

with $\operatorname{tr}H=5$ and $\operatorname{tr}S=0$.

The product preserves neither half. For anti-Hermitian $S,T$ one has $(ST)^{*}=TS$, which equals $-ST$ only when $S$ and $T$ anticommute. The split is one of the real vector space and of the real Lie algebra.

The Identification with the Biquaternion Algebra

The algebra is the biquaternion algebra. The following is the dictionary between the two languages.

Theorem. There is an isomorphism of $\mathbb{C}$-algebras $\mathsf M_2:\mathbb{B}\to M_2(\mathbb{C})$ with

$$ \mathsf M_2(e_t)=I,\qquad \mathsf M_2(\vec e_x)=-i\sigma_1,\qquad \mathsf M_2(\vec e_y)=-i\sigma_2,\qquad \mathsf M_2(\vec e_z)=-i\sigma_3 . $$

Proof. The images are independent and reproduce the relations: $\mathsf M_2(\vec e_k)^{2}=-I$ because $\sigma_k^{2}=I$, and $(-i\sigma_1)(-i\sigma_2)=-\sigma_1\sigma_2=-i\sigma_3$ is the image of $\vec e_z$. A linear map agreeing with the products on a basis is a homomorphism, and it is bijective because both spaces have dimension four. The details are $M_2(\mathbb{C})$ Pauli Representation.

Corollary. $\operatorname{tr}\mathsf M_2(\tilde Q)=2Q_t$ and $\det\mathsf M_2(\tilde Q)=\langle\tilde Q\rangle_{\natural}=Q_t^{2}+Q_x^{2}+Q_y^{2}+Q_z^{2}$.

The determinant is the algebraic norm $N$. Multiplicativity of the determinant is multiplicativity of $N$, and the zero divisors of $\mathbb{B}$ are the singular matrices. This is what the dimension adds: at $n=2$ the determinant is a quadratic form with a genuine cone of real dimension $6$, while at $n\ge3$ the singular set is a hypersurface. The cone is N-Regularity and N-Degeneracy in the Biquaternion Algebra.

Corollary (the four corners). A complete family of two idempotents of $\mathbb{B}$ maps to a pair of complementary rank-one projections $P_{+}$ and $P_{-}=I-P_{+}$, and the Peirce lines $\mathbb{B}_{st}$ map onto the matrix corners $A_{kl}=P_kM_2(\mathbb{C})P_l$, each one-dimensional. The companion nilpotents $\vec\Upsilon_{+-},\vec\Upsilon_{-+}$ of The Companion Nilpotents of the Biquaternion Algebra map onto $E_{12},E_{21}$, and a Peirce basis onto the four matrix units.

What the identification does not carry. It is $\mathbb{C}$-linear, so it does not send the real form $\mathbb{H}$ to a real subalgebra; the image is a four-real-dimensional subspace. The two real forms are $M_2(\mathbb{R})$ and $\mathbb{H}$, treated in The Two Base Fields: $M_2(\mathbb{R})$ inside $M_2(\mathbb{C})$. The transpose is an anti-automorphism of the matrix algebra and not the quaternion conjugation of $\mathbb{B}$.

Summary

$M_2(\mathbb{C})$ is the associative unital non-commutative $\mathbb{C}$-algebra of $2\times2$ complex matrices, with

$$ \boxed{\;E_{ij}E_{kl}=\delta_{jk}E_{il},\qquad \dim_{\mathbb{C}}=4,\qquad \dim_{\mathbb{R}}=8.\;} $$

Its centre is $\mathbb{C}I$ and it is simple, hence central simple. The trace is the only commutator-vanishing functional up to scale, the determinant is multiplicative, and the two are the coefficients of $\chi_A=\lambda^{2}-(\operatorname{tr}A)\lambda+\det A$. The rank takes the values $0,1,2$ on strata of dimensions $0,3,4$, and the singular set is the determinant hypersurface of real codimension two. The group of invertibles is $GL_2(\mathbb{C})$ of dimension $4$, and $PGL_2(\mathbb{C})\cong SL_2(\mathbb{C})/\{\pm I\}$ has dimension $3$. Over $\mathbb{R}$ the algebra is $\mathfrak{u}(2)\oplus i\mathfrak{u}(2)$. It is the biquaternion algebra under $\mathsf M_2(\vec e_k)=-i\sigma_k$, with $\operatorname{tr}=2Q_t$ and $\det=N$.

Property Value for $M_2(\mathbb{C})$
Complex dimension $4$
Real dimension $8$
Centre $\mathbb{C}I$
Simple yes
Division algebra no; $E_{12}^{2}=0$
Group of invertibles $GL_2(\mathbb{C})$, dimension $4$
Projective group $PGL_2\cong SL_2/\{\pm I\}$, dimension $3$
Real reading $\mathfrak{u}(2)\oplus i\mathfrak{u}(2)$, real dimensions $4+4$
Biquaternion model $\vec e_k\mapsto-i\sigma_k$; $\det=N$, $\operatorname{tr}=2Q_t$

Summary of Notation

Symbol Meaning
$M_2(\mathbb{C})$ the algebra of $2\times2$ complex matrices
$E_{ij}$ the matrix unit with $1$ in position $(i,j)$
$I$ the identity matrix
$\operatorname{tr}$, $\det$ the trace and the determinant
$\chi_A$ the characteristic polynomial $\det(\lambda I-A)$
$\mathrm{rank}\,A$ the dimension of the image
$A^{*}=\bar A^{\mathsf T}$ the conjugate transpose
$\mathfrak{u}(2)$, $\mathfrak{su}(2)$ the anti-Hermitian matrices, and the traceless ones
$GL_2(\mathbb{C})$, $SL_2(\mathbb{C})$ the invertible matrices, and those of determinant $1$
$PGL_2(\mathbb{C})$ $GL_2(\mathbb{C})/\mathbb{C}^{\times}I$
$\mathsf M_2$ the isomorphism $\mathbb{B}\to M_2(\mathbb{C})$
$N$ the algebraic norm, equal to $\det$

Further Reading

  • Roger A. Horn and Charles R. Johnson, Matrix Analysis (Cambridge University Press, 2013), for the matrix units, the trace and the determinant at order two.
  • Richard S. Pierce, Associative Algebras (Springer, 1982), for central simple algebras and the centre.
  • $M_2(\mathbb{C})$ Pauli Representation, for the identification with the biquaternion algebra in full.
  • The Biquaternion Algebra, for the algebra $\mathbb{B}$ identified with this one.