The Kronecker Factorisation $M_4(\mathbb{C})\cong M_2(\mathbb{C})\otimes M_2(\mathbb{C})$
Introduction
The algebra $M_4(\mathbb{C})$ is the tensor product of two copies of $M_2(\mathbb{C})$:
$$ M_4(\mathbb{C})\cong M_2(\mathbb{C})\otimes_\mathbb{C}M_2(\mathbb{C}),\qquad (A\otimes B)(C\otimes D)=AC\otimes BD. $$
The isomorphism is not a curiosity. It exhibits two four-dimensional subalgebras, $M_2\otimes I$ and $I\otimes M_2$, that commute elementwise, generate the whole algebra, and are each the commutant of the other. It also produces the flip $F(x\otimes y)=y\otimes x$, a symmetric unitary involution whose conjugation is the swap of the two factors; the swap is an inner automorphism of $M_4$, and the fixed algebra of the swap is the centraliser of $F$, isomorphic to $M_3(\mathbb{C})\oplus\mathbb{C}$.
This article states the isomorphism, the two factors, the double centraliser theorem, the flip and its eigenspaces, and the Brauer reading. The two-sided operators on the biquaternion algebra are The Two-Sided Operators Method to Build $M_4(\mathbb{C})$; the warning that the ambient algebra is the endomorphism algebra of $\mathbb{B}$ and not $\mathbb{B}$ itself is repeated in §The Biquaternion Case. The rank-two split $M_2\oplus M_2$, which is a different decomposition, is The Rank-Two Split of $M_4(\mathbb{C})$.
The Isomorphism
Definition. Let $\mathbb{C}^4=\mathbb{C}^2\otimes \mathbb{C}^2$ with basis $f_i\otimes g_j$, $i,j\in\{1,2\}$, ordered lexicographically. For $A,B\in M_2(\mathbb{C})$ the Kronecker product $A\otimes B$ is the $4\times4$ matrix of the map $x\otimes y\mapsto Ax\otimes By$.
Proposition. The Kronecker product is bilinear and satisfies
$$ (A\otimes B)(C\otimes D)=AC\otimes BD,\qquad I\otimes I=I_4. $$
Proof. On a decomposable vector $x\otimes y$,
$$ (A\otimes B)(C\otimes D)(x\otimes y)=(A\otimes B)(Cx\otimes Dy)=ACx\otimes BDy, $$
and both sides extend by linearity. The unit is the identity on each factor.
Theorem. The map $A\otimes B\mapsto A\otimes B$ is an isomorphism of algebras
$$ M_2(\mathbb{C})\otimes_\mathbb{C}M_2(\mathbb{C})\longrightarrow M_4(\mathbb{C}). $$
Proof. The map is a homomorphism by the proposition. It is injective because the sixteen matrices $E_{ij}\otimes E_{kl}$ are linearly independent: $E_{ij}\otimes E_{kl}$ has a single nonzero entry, in the position determined by the pair $(i,k)$ of rows and $(j,l)$ of columns, so the sixteen occupy sixteen distinct positions. Both sides have dimension $16$, so the injective map is an isomorphism.
The matrix units. Under the lexicographic ordering,
$$ E_{ij}\otimes E_{kl}=E_{2(i-1)+k,\,2(j-1)+l}. $$
The other basis. The sixteen $E_{ij}$ are the matrix units of the abstract algebra and the basis of the statement above. The tensor split is the split of the two-sided operators on the biquaternion algebra $\mathbb{B}=\mathbb{C}\otimes_{\mathbb{R}}\mathbb{H}$: the left multiplications $L_{\mathrm{axf},e_\mu}$ form one factor and the right multiplications $R_{\mathrm{axf},e_\nu}$ the other, and the sixteen products $T_{\mu\nu}=L_{\mathrm{axf},e_\mu}R_{\mathrm{axf},e_\nu}$ are the same basis in the coefficient basis $e_t,\vec e_x,\vec e_y,\vec e_z$, each a signed permutation matrix. The basis and the four left matrices are $M_4(\mathbb{C})$ Axis-Free Representation; the sixteen products are $M_4(\mathbb{C})$ Axis-Free Representation: Operators.
The sixteen Kronecker products are the sixteen matrix units, reindexed by pairs.
The Two Factors
Definition. The two factors of the decomposition are
$$ \mathcal{A}=\{A\otimes I:A\in M_2(\mathbb{C})\},\qquad \mathcal{B}=\{I\otimes B:B\in M_2(\mathbb{C})\}. $$
Proposition. $\mathcal{A}$ and $\mathcal{B}$ are subalgebras of $M_4$, each isomorphic to $M_2(\mathbb{C})$ and of dimension $4$. Every element of $\mathcal{A}$ commutes with every element of $\mathcal{B}$, and
$$ \mathcal{A}\cap\mathcal{B}=\mathbb{C}I. $$
Proof. The product formula gives $(A\otimes I)(A'\otimes I)=AA'\otimes I$, so $\mathcal{A}$ is a subalgebra with unit $I\otimes I$; the map $A\mapsto A\otimes I$ is injective, so $\mathcal{A}\cong M_2$. The same holds for $\mathcal{B}$. Commutation is the computation $(A\otimes I)(I\otimes B)=A\otimes B=(I\otimes B)(A\otimes I)$. An element of the intersection is $A\otimes I=I\otimes B$; comparing the entries, $A$ is scalar and $B$ is the same scalar.
The factors generate the algebra. The products $ (A\otimes I)(I\otimes B)=A\otimes B$ span all sixteen dimensions, so the algebra generated by $\mathcal{A}\cup\mathcal{B}$ is $M_4$. The two factors are small, but together they are everything.
The factors are not ideals. $\mathcal{A}$ is not a two-sided ideal: for $X=A\otimes I$ and $Y=C\otimes D$ with $D$ not scalar, $XY=AC\otimes D$ is not in $\mathcal{A}$. The algebra $M_4$ is simple, so it has no nontrivial two-sided ideals at all.
The double centraliser. The commutant of $\mathcal{A}$ in $M_4$ is $\mathcal{B}$, and the commutant of $\mathcal{B}$ is $\mathcal{A}$.
Proof. $\mathcal{B}\subseteq\mathcal{A}'$ by commutation. For the reverse inclusion, identify $M_4=\operatorname{End}(\mathbb{C}^{2}\otimes\mathbb{C}^{2})$. The algebra $\mathcal{A}=M_2\otimes I$ is the full algebra of endomorphisms acting on the first factor alone. A map $X$ commuting with all $A\otimes I$ commutes in particular with all the matrix units $E_{ij}\otimes I$; writing $X$ in the basis $f_i\otimes g_j$, the relation $X(E_{ij}\otimes I)=(E_{ij}\otimes I)X$ for all $i,j$ forces $X$ to preserve each $f_i\otimes \mathbb{C}^2$ and to act on it by the same matrix, so $X=I\otimes B$. Hence $\mathcal{A}'=\mathcal{B}$, and symmetrically $\mathcal{B}'=\mathcal{A}$.
The dimensions. The theorem is an instance of the double centraliser theorem: $\dim\mathcal{A}=4$, $\dim\mathcal{B}=4$ and $\dim M_4=16=4\cdot4$.
The Flip
Definition. The flip is the linear map
$$ F:\mathbb{C}^{2}\otimes\mathbb{C}^{2}\longrightarrow\mathbb{C}^{2}\otimes\mathbb{C}^{2},\qquad F(x\otimes y)=y\otimes x. $$
In the basis $f_i\otimes g_j$ its matrix has entries $F_{(i,j),(k,l)}=\delta_{il}\delta_{jk}$.
Proposition. $F$ is a symmetric unitary involution:
$$ F^{2}=I,\qquad F^{\mathsf T}=F,\qquad F^{*}=F. $$
Its trace is $2$, its determinant is $-1$, and its eigenvalues are $1$ with multiplicity three and $-1$ with multiplicity one.
Proof. Applying $F$ twice returns $x\otimes y$, so $F^{2}=I$. The matrix of $F$ in the lexicographic basis is symmetric: the entry at $((i,j),(k,l))$ is $\delta_{il}\delta_{jk}$, symmetric in the two pairs. A real symmetric involution is self-adjoint, hence unitary. The trace is the number of basis vectors fixed by $F$, namely $f_1\otimes g_1$ and $f_2\otimes g_2$, so $\operatorname{tr}F=2$. The eigenvalues are $\pm1$; from $\operatorname{tr}F=2$ and $\operatorname{tr}F^{2}=4$ the multiplicities are three and one, and the determinant is $1^{3}(-1)=-1$.
The eigenspaces. On $\mathbb{C}^2\otimes \mathbb{C}^2$ the $+1$ eigenspace of $F$ is $\operatorname{sym}^{2}\mathbb{C}^{2}$, of dimension $3$; the $-1$ eigenspace is $\Lambda^{2}\mathbb{C}^{2}$, of dimension $1$. On the algebra, the automorphism $\sigma$ has a fixed space of dimension $10$ and an anti-fixed space of dimension $6$: the two dimensions add to $16$, and the fixed space is the subalgebra $M_3\oplus\mathbb{C}$ computed below.
The swap. Let $\sigma(X)=FXF$. Then $\sigma$ is an automorphism of $M_4$ of order two, and
$$ \sigma(A\otimes B)=B\otimes A. $$
Proof. $F$ is invertible and conjugation by $F$ is an automorphism; $F^{2}=I$ gives $\sigma^{2}=\mathrm{id}$. For the formula, apply both sides to $x\otimes y$: $\sigma(A\otimes B)(x\otimes y)=F(A\otimes B)(y\otimes x)=F(Ay\otimes Bx)=Bx\otimes Ay=(B\otimes A)(x\otimes y)$.
The swap is inner. Every automorphism of $M_4(\mathbb{C})$ is inner, by Skolem–Noether, so $\sigma$ is conjugation by an invertible matrix; the matrix is $F$ itself, which is unitary. The flip is not in $SL_4$, since $\det F=-1$, but the automorphism is insensitive to the scalar: $F$ and $-F$ give the same $\sigma$.
The fixed algebra. The fixed algebra of $\sigma$ is the centraliser of $F$,
$$ M_4^{\sigma}=\{X:FXF=X\}=\{X:FX=XF\}. $$
Because $F$ is diagonalisable with eigenvalues $1$ of multiplicity three and $-1$ of multiplicity one, its centraliser is
$$ M_4^{\sigma}\cong M_3(\mathbb{C})\oplus\mathbb{C}, $$
of dimension $10$. The $-1$ eigenspace of $\sigma$ has dimension $6$; it is not a subalgebra, since the product of two anti-fixed elements is fixed.
The Brauer Reading
The class. In the Brauer group $Br(\mathbb{C})$ the class of $M_n(\mathbb{C})$ is the identity for every $n$: every complex matrix algebra is split. The isomorphism $M_4\cong M_2\otimes M_2$ is the statement
$$ [M_2]\,[M_2]=[M_4]=1 $$
in $Br(\mathbb{C})$, which is trivial because the group is trivial. Over $\mathbb{R}$ the same product is not trivial: $[M_2(\mathbb{R})]$ is the class of the quaternion algebra, of order two, and $M_2(\mathbb{R})\otimes M_2(\mathbb{R})\cong M_4(\mathbb{R})$ is split. The complex case is the split case of that statement, and the quaternionic structure of The Quaternionic Structure and the Real Forms of $M_4(\mathbb{C})$ is the real shadow of it.
The tensor product of modules. The isomorphism is the algebra side of the tensor product of the two standard modules: $\mathbb{C}^4=\mathbb{C}^2\otimes \mathbb{C}^2$, and $M_2$ acts on the first factor, $M_2$ on the second. The four-dimensional module of $M_4$ is the tensor product of the two two-dimensional modules, and the flip is the comparison isomorphism of the two orderings.
The Biquaternion Case
The endomorphism algebra. The biquaternion algebra is $\mathbb{B}\cong M_2(\mathbb{C})$. The algebra generated by the two-sided operators $L_{a}R_{b}(x)=axb$ is $\operatorname{End}_\mathbb{C}(\mathbb{B})$, and
$$ \operatorname{End}_\mathbb{C}(\mathbb{B})\cong\mathbb{B}\otimes_\mathbb{C}\mathbb{B}_{\mathrm{opp}}\cong M_2(\mathbb{C})\otimes M_2(\mathbb{C})\cong M_4(\mathbb{C}). $$
The two factors are the left multiplications $L_a$ and the right multiplications $R_b$. The factor of left multiplications is a copy of $\mathbb{B}$; the factor of right multiplications is a copy of $\mathbb{B}_{\mathrm{opp}}$, which is isomorphic to $\mathbb{B}$ because $\mathbb{B}$ is a matrix algebra. The sixteen operators $L_\mu R_\nu$ are the products of the two factors, and they are the matrix units of the tensor product.
The terminology. $M_4(\mathbb{C})$ is the matrix representation of the algebra generated by the two-sided operators on $\mathbb{B}$. It is not $\mathbb{B}$ itself, and it is not a matrix representation of $\mathbb{B}$ in the sense of a homomorphism $\mathbb{B}\to M_4$: the left regular representation is the inclusion of the factor $\mathbb{B}\otimes I$, of dimension four, and the full algebra is the endomorphism algebra of the four-dimensional carrier. The distinction is the one stated in The Matrix Algebra $M_4(\mathbb{C})$, §The Full Algebra and the Regular Model.
The rank-two split is different. The decomposition $M_4=M_2\oplus M_2$ of The Rank-Two Split of $M_4(\mathbb{C})$ is a block-diagonal subalgebra of dimension $8$, attached to a rank-two idempotent. The Kronecker factorisation is an isomorphism of the whole algebra with a tensor product of two four-dimensional algebras, and the two factors are not subalgebras of the block-diagonal split. The first is a grading, the second is a tensor structure.
Summary
$M_4(\mathbb{C})$ is $M_2(\mathbb{C})\otimes M_2(\mathbb{C})$ by the Kronecker product. The two factors $M_2\otimes I$ and $I\otimes M_2$ commute, generate the algebra, meet in the scalars, and are mutual commutants. The flip $F(x\otimes y)=y\otimes x$ is a symmetric unitary involution of trace $2$ and determinant $-1$, with eigenspaces $\operatorname{sym}^{2}\mathbb{C}^{2}$ of dimension $3$ and $\Lambda^{2}\mathbb{C}^{2}$ of dimension $1$; its conjugation is the swap of the factors, an inner automorphism whose fixed algebra is $M_3\oplus\mathbb{C}$ of dimension $10$ and whose anti-fixed space has dimension $6$. The factorisation is the complex split case of the Brauer statement $[M_2][M_2]=[M_4]=1$, and it is the algebra side of $\operatorname{End}_\mathbb{C}(\mathbb{B})\cong\mathbb{B}\otimes\mathbb{B}_{\mathrm{opp}}$.
| Quantity | Value |
|---|---|
| $\dim M_2\otimes M_2$ | $16$ |
| $\dim(M_2\otimes I)$ | $4$ |
| $\dim(I\otimes M_2)$ | $4$ |
| $\dim(M_2\otimes I)\cap(I\otimes M_2)$ | $1$ |
| $\operatorname{tr}F$ | $2$ |
| $\det F$ | $-1$ |
| $\dim\operatorname{sym}^2$ | $10$ |
| $\dim\Lambda^2$ | $6$ |
| $\dim M_4^{\sigma}$ | $10$ |
| $M_4^{\sigma}$ | $M_3(\mathbb{C})\oplus\mathbb{C}$ |
Summary of Notation
| Symbol | Meaning |
|---|---|
| $A\otimes B$ | the Kronecker product, the matrix of $x\otimes y\mapsto Ax\otimes By$ |
| $\mathcal{A}=M_2\otimes I$ | the first factor, the left action on the first tensor slot |
| $\mathcal{B}=I\otimes M_2$ | the second factor, the left action on the second slot |
| $F$ | the flip $x\otimes y\mapsto y\otimes x$ |
| $\sigma(X)=FXF$ | the swap automorphism of order two |
| $M_4^{\sigma}$ | the fixed algebra of the swap, the centraliser of $F$ |
| $[M_n]$ | the Brauer class of $M_n$ |
Further Reading
- $M_4(\mathbb{C})$ Axis-Free Representation, for the coefficient basis, the four left matrices and the transpose as the quaternion conjugation.
- $M_4(\mathbb{C})$ Axis-Free Representation: Operators, for the sixteen two-sided operators in the coefficient frame, their signs and their composition.
- The Sylvester Operator and the Kronecker Sum of $M_4(\mathbb{C})$, for the operators $L_A$ and $R_B$ and the difference $L_A-R_B$.
- The Numerical Range and the Field of Values of $M_4(\mathbb{C})$, for the rule $W(A\otimes I)=W(I\otimes A)=W(A)$.
- The Matrix Algebra $M_4(\mathbb{C})$, for the full algebra and the regular model.
- The Two-Sided Operators Method to Build $M_4(\mathbb{C})$, for the sixteen operators $L_\mu R_\nu$ and their product rule.
- The Enveloping Operator Algebra of the Biquaternion Algebra and the Bi-module Structure, for $\mathbb{B}\otimes\mathbb{B}_{\mathrm{opp}}\cong\operatorname{End}_\mathbb{C}(\mathbb{B})$.
- The Rank-Two Split of $M_4(\mathbb{C})$, for the block-diagonal split, which is not the tensor factorisation.
- The Involution and the Hermitian Structure of $M_4(\mathbb{C})$, for the unitary group and the Hermitian structure.
- The Quaternionic Structure and the Real Forms of $M_4(\mathbb{C})$, for the real shadow of the Brauer statement.
- Richard S. Pierce, Associative Algebras (Springer, 1982), for the double centraliser theorem and the Brauer group.