$\operatorname{End}(M_4(\mathbb{C}))$ as an Algebra: the Regular Representation
Introduction
The algebra of complex-linear endomorphisms of the sixteen-dimensional space $M_4(\mathbb{C})$ is $M_{16}(\mathbb{C})$, of dimension $256$, and it is generated by the left and the right multiplications, which commute and together give an isomorphism $M_4(\mathbb{C})\otimes M_4(\mathbb{C})^{\mathrm{op}}\to\operatorname{End}(M_4(\mathbb{C}))$; the two images are mutual centralisers, and the trace of the left multiplication by $A$ is $4\operatorname{tr}A$.
The article computes the dimension of the endomorphism algebra, proves that the left and right multiplications are injective homomorphisms with opposite multiplication rules, proves the double centraliser theorem in this case, constructs the tensor isomorphism by dimension and by the product rule, computes the trace and the determinant of the regular representation, and writes the left and right families in the axis-free frame as signed permutations.
It assumes The Two-Sided Operators and the Segre Cone at $n=4$ and The One-Sided Operators at $n=4$. The model at order two is $\operatorname{End}(M_2(\mathbb{C}))$ as an Algebra: the Regular Representation; the structure theory is The Ideals, the Centre and the Simplicity of $M_4(\mathbb{C})$.
The Space of Endomorphisms
Proposition. $\operatorname{End}_{\mathbb{C}}(M_4(\mathbb{C}))$ has complex dimension $256$ and is isomorphic to $M_{16}(\mathbb{C})$; it is simple, with centre the scalars.
Proof. A linear map of a complex space of dimension sixteen is a $16\times16$ matrix in a chosen basis, so the algebra is $M_{16}(\mathbb{C})$, of dimension $16^{2}=256$; a full matrix algebra is simple with scalar centre.
Corollary. The real dimension is $512$, and every endomorphism has a trace and a determinant computed in any basis of the matrix space.
Proof. The real dimension is twice the complex one, and the trace and determinant are invariant under conjugation.
Corollary. The endomorphism algebra has exactly two two-sided ideals, and every nonzero element of it is invertible or a zero divisor.
Proof. Simplicity and the structure of a full matrix algebra.
The two scales. The algebra $M_4(\mathbb{C})$ has dimension $16$, its endomorphism algebra $256$, and the ratio $16^{2}$ is the square of the dimension of the space; the same ratio at order two is $4^{2}=16$.
Proof. The dimension of the endomorphism algebra of an $n$-dimensional space is $n^{2}$.
The Left and Right Multiplications
Definition. For a fixed $A$ let $L_A(X)=AX$ and $R_A(X)=XA$.
Proposition. $L_A$ and $R_A$ are linear, and the assignments are multiplicative with opposite conventions: $$ L_AL_B=L_{AB},\qquad R_AR_B=R_{BA}. $$
Proof. Compose: $L_A(L_B(X))=ABX=L_{AB}(X)$, and $R_A(R_B(X))=XBA=R_{BA}(X)$.
Corollary. $A\mapsto L_A$ is an injective algebra homomorphism of $M_4(\mathbb{C})$ into $\operatorname{End}(M_4(\mathbb{C}))$, and $A\mapsto R_A$ is an injective anti-homomorphism; both images have dimension $16$.
Proof. Injectivity: $L_A=0$ forces $A=L_A(\mathrm I_4)=0$; the dimension of the image is that of the source.
Corollary. $L_AR_B=R_BL_A$, so the two images commute, and $L_A=R_B$ only when $A=B$ is a scalar.
Proof. $L_AR_B(X)=AXB=R_BL_A(X)$; for the second part, evaluate at $X=\mathrm I_4$ to get $A=B$, and then $AX=XA$ for all $X$ forces $A$ to be a scalar.
The intersection. The two images meet in the scalars, so their sum has dimension $16+16-1=31$, not $256$.
Proof. The intersection is one-dimensional and the dimensions add with the intersection subtracted.
Corollary. The left and right multiplications alone do not exhaust the endomorphism algebra; the missing maps are the ones that mix the two sides.
Proof. $31<256$.
The Double Centraliser
Theorem. The centraliser of $L(M_4)$ in $\operatorname{End}(M_4)$ is $R(M_4)$, and the centraliser of $R(M_4)$ is $L(M_4)$.
Proof. A map $T$ commuting with every $L_A$ satisfies $T(AX)=AT(X)$, so $T$ is determined by $T(\mathrm I_4)=B$ through $T(X)=XB$, that is $T=R_B$; the converse is the commutation proved above. The second statement is the same computation.
Corollary. The two images are mutual centralisers and their intersection is the scalars, so each is the whole centraliser of the other.
Proof. The centraliser of one image is the other, and the scalars are the intersection.
Corollary. The algebra generated by the two images is their product, of dimension $256$, hence the whole endomorphism algebra.
Proof. The product of the two 16-dimensional images contains all maps $X\mapsto AXB$, and these span the whole algebra: their matrices in the basis of matrix units are the elementary matrices of the tensor product, and they are $256$ independent maps.
Corollary. $M_4(\mathbb{C})$ is its own bicommutant inside $\operatorname{End}(M_4)$: the double centraliser theorem holds in the strong form.
Proof. The centraliser of the centraliser is the original image.
The Tensor Form
Theorem. The map $$ \Phi:M_4(\mathbb{C})\otimes_{\mathbb{C}}M_4(\mathbb{C})^{\mathrm{op}}\longrightarrow\operatorname{End}_{\mathbb{C}}(M_4(\mathbb{C})),\qquad A\otimes B^{\mathrm{op}}\longmapsto\bigl(X\mapsto AXB\bigr), $$ is an isomorphism of algebras.
Proof. The map is well defined, because $B^{\mathrm{op}}$ multiplies in the reversed order; it is multiplicative, since $(AXB)(CYB')$ composes correctly; both sides have dimension $256$; and it is injective because the images of the pure tensors are the maps $T_{A,B}$ of the two-sided operators, which are independent when $A$ and $B$ range over a basis.
Corollary. $\operatorname{End}(M_4)\cong M_4\otimes M_4^{\mathrm{op}}\cong M_{16}(\mathbb{C})$, and the two factors of the tensor product are the two one-sided images.
Proof. The isomorphism and the identification of the factors with $L$ and $R$.
Corollary. The endomorphism algebra is simple because the tensor product of two central simple algebras is central simple.
Proof. $M_4$ is central simple, and the tensor product of central simple algebras is central simple.
The rank-one operators. The pure tensors $A\otimes B^{\mathrm{op}}$ map to the operators $T_{A,B}$ of rank one in the coefficient matrix, and these generate the algebra.
Proof. The coefficient of $T_{A,B}$ is the outer product of the coefficient vectors, of rank one, and the rank-one matrices span the matrix space.
The Trace of the Regular Representation
Proposition. $\operatorname{tr}L_A=4\operatorname{tr}A$, $\operatorname{tr}R_A=4\operatorname{tr}A$, and $\operatorname{tr}T_{A,B}=\operatorname{tr}A\,\operatorname{tr}B$.
Proof. In the basis of matrix units, $L_A$ acts on $E_{ij}$ by $AE_{ij}$, whose $ij$ entry is $A_{ii}$; summing over the four values of $j$ gives $4A_{ii}$ for each $i$, hence $4\operatorname{tr}A$. The same computation for $R_A$, and for $T_{A,B}$ the entry is $A_{ii}B_{jj}$, whose sum is $\operatorname{tr}A\operatorname{tr}B$.
Corollary. $\det L_A=(\det A)^{4}$ and $\det R_A=(\det A)^{4}$.
Proof. The matrix of $L_A$ in the matrix-unit basis is the Kronecker product $A\otimes\mathrm I_4$, whose determinant is $(\det A)^{4}$.
Corollary. The character of the regular representation at $A$ is $4\operatorname{tr}A$, which is four times the character of the defining representation.
Proof. The trace computation.
The module reading. Under the left action $M_4(\mathbb{C})$ is the direct sum of four copies of $\mathbb{C}^{4}$, and the multiplicity four is the factor in the trace.
Proof. The left action preserves each column space, acts on each as the defining representation, and the four columns are independent.
Corollary. The regular representation is faithful and its image has dimension $16$; the trace detects the defining trace up to the factor four.
Proof. Faithfulness of $L$ and the trace formula.
The Axis-Free Form
Theorem. In the axis-free frame the left and right families are the matrices $L^{\mathrm{axf}}$ and $R^{\mathrm{axf}}$, whose four elements are signed permutations of the coefficient basis; the endomorphism $L^{\mathrm{axf}}_a$ acts on the sixteen $T_{\mu\nu}$ by left multiplication of the index.
Proof. The coefficient basis is $e_t,\vec e_x,\vec e_y,\vec e_z$, and left multiplication by an invertible element permutes the four basis elements up to signs; the action on the two-sided basis operators is the corresponding signed permutation of the first index.
Corollary. In the basis $T_{\mu\nu}$ the regular representation is a group of $16\times16$ signed permutation matrices, and the two families are the two index permutations.
Proof. The product rule $T_{a,b}T_{c,d}=T_{ac,db}$ permutes the two indices separately.
Corollary. $\operatorname{tr}L^{\mathrm{axf}}_a=4Q_t(a)$, so the trace of the regular representation is four times the scalar coordinate, in agreement with $\operatorname{tr}A=4c_{tt}$ for the element.
Proof. The trace formula and the identification of the scalar coordinate.
The finite group. The group of order $32$ of the signed permutations acts on the algebra through the regular representation, and the two copies of $Q_8$ act as the two index permutations.
Proof. The elements $\pm T_{\mu\nu}$ act by left and right multiplication, hence by signed permutations of the two indices.
Summary
The endomorphism algebra of the sixteen-dimensional matrix space has dimension $256$ and is $M_{16}(\mathbb{C})$. The left multiplications form a copy of $M_4$ and the right multiplications a copy of the opposite algebra; the two commute, meet in the scalars, and their sum has dimension $31$, while their product is the whole algebra.
The tensor map $A\otimes B^{\mathrm{op}}\mapsto(X\mapsto AXB)$ is an isomorphism $M_4\otimes M_4^{\mathrm{op}}\to\operatorname{End}(M_4)$, and the two factors are mutual centralisers, so the double centraliser theorem holds in the strong form. The trace of $L_A$ and of $R_A$ is four times the trace of $A$, the trace of $T_{A,B}$ is the product of the two traces, and the determinant of $L_A$ is the fourth power of the determinant of $A$.
In the axis-free frame the two families are signed permutations of the coefficient basis, and the regular representation acts on the sixteen two-sided operators by independent permutations of the two indices.
Summary of Notation
| Symbol | Meaning |
|---|---|
| $\operatorname{End}(M_4)\cong M_{16}$ | the endomorphism algebra, dimension $256$ |
| $L_A(X)=AX$, $R_A(X)=XA$ | the one-sided multiplications |
| $M_4\otimes M_4^{\mathrm{op}}$ | the tensor form of the endomorphism algebra |
| $T_{A,B}(X)=AXB$ | the two-sided operator |
| $\operatorname{tr}L_A=4\operatorname{tr}A$ | the character of the regular representation |
| $(\det A)^{4}$ | the determinant of $L_A$ |
Further Reading
- Israel Nathan Herstein, Noncommutative Rings (MAA, 1968), for the double centraliser theorem.
- The One-Sided Operators at $n=4$, for the two families and the stitch.
- The Two-Sided Operators and the Segre Cone at $n=4$, for the sixteen operators and the coefficient matrix.
- The Ideals, the Centre and the Simplicity of $M_4(\mathbb{C})$, for the simplicity used in the tensor argument.
- $\operatorname{End}(M_2(\mathbb{C}))$ as an Algebra: the Regular Representation, for the case at order two.