$\operatorname{End}(\mathbb{B})$ as an Algebra: the Regular Representation
Introduction
The subject. The biquaternion algebra $\mathbb{B}$ is four-dimensional over $\mathbb{C}$, and the complex-linear operators on it form a space of complex dimension sixteen. The article proves that this space is generated by two four-dimensional families: the left multiplications $a\mapsto L_a$, with $L_a(X)=aX$, and the right multiplications $b\mapsto R_b$, with $R_b(X)=Xb$. Their products are the two-sided operators $T_{a,b}=L_aR_b$, with $T_{a,b}(X)=aXb$, and the sixteen products $T_{e_\mu,e_\nu}$ of the basis units are a basis of the whole space. The generation is therefore an isomorphism of algebras $\mathbb{B}\otimes_{\mathbb{C}}\mathbb{B}^{\mathrm{opp}}\to\operatorname{End}(\mathbb{B})$, carrying $a\otimes b^{\mathrm{op}}$ to $T_{a,b}$, and it is the statement that the regular representation of the algebra is the whole operator algebra and not a part of it.
The order of the article. The space and its trace come first, because every later count uses the trace. The product law of the two-sided operators comes second, and the generation theorem third. The tensor reading follows, then the trace form with its inertia, and the simplicity of the algebra closes the argument.
It assumes Introduction to the Operators of the Biquaternion Algebra, Operators Acting on One Side and The Tensor Product of Two Biquaternion Algebras. The row of the elements is four-dimensional, and there the operators are the two families alone; here the two families generate a sixteen-dimensional algebra.
The result in one line. Every operator on $\mathbb{B}$ is a finite sum of products of a left and a right multiplication, and the assignment $a\otimes b^{\mathrm{op}}\mapsto T_{a,b}$ is an isomorphism of algebras.
Boundaries. The families themselves, their multiplication laws and their intersection are The One-Sided Operators; the coefficients of one operator in the sixteen products are The Coefficients of a Two-Sided Operator: A Worked Computation; the reading of the space as a module of one copy of the rotation algebra is The Copy Acting on Sixteen Dimensions.
The Space and Its Trace
Definition. The space of the operators is the set $\operatorname{End}(\mathbb{B})$ of the maps $\mathbb{B}\to\mathbb{B}$ that are complex-linear, with addition and composition. Its elements are written $F,G,S,T$.
The trace. The trace of an operator $F$ is the sum of the coefficients of the four values $F(e_\mu)$ on the corresponding units, $$ \operatorname{tr}(F)=\sum_{\mu\in\{t,x,y,z\}}c_\mu,\qquad F(e_\mu)=\sum_{\nu}c_{\nu\mu}e_\nu, $$ a complex number, equal to the sum of the eigenvalues of $F$ counted with multiplicity. It is linear in $F$ and satisfies $\operatorname{tr}(FG)=\operatorname{tr}(GF)$.
Proposition. The space is of complex dimension sixteen, since an operator is determined by its four values and each value by four coefficients.
Proof. A linear map of a four-dimensional space is determined by four arbitrary values, and the four values lie in a four-dimensional space; the count gives $4\cdot4=16$, and the operators $F_{\mu\nu}$ that send $e_\mu$ to $e_\nu$ and kill the other three units are a basis.
Proposition. For every element $a$ the left multiplication $L_a$ and the right multiplication $R_a$ are operators, and their traces are $$ \operatorname{tr}(L_a)=\operatorname{tr}(R_a)=4\,\mathrm{Sc}(a). $$
Proof. The coefficient of $e_\mu$ in $a\,e_\mu$ is the trace of the four products; a product $e_\lambda e_\mu$ contributes to the coefficient of $e_\mu$ only when $\lambda=t$, and contributes $\mathrm{Sc}(a)$. The sum over the four units gives $4\,\mathrm{Sc}(a)$, and the same computation holds on the right.
The reading. The trace of a one-sided operator sees only the scalar coordinate of its parameter. The vector part of the parameter is invisible to the trace and is the part that will be seen by the determinant.
The Product Law
Definition. For a pair of elements, the two-sided operator is $$ T_{a,b}=L_aR_b:\quad T_{a,b}(X)=aXb . $$
Proposition. The assignment $a\otimes b^{\mathrm{op}}\mapsto T_{a,b}$ is complex-bilinear, and $T_{a,b}=0$ only when $a=0$ or $b=0$.
Proof. Bilinearity is the bilinearity of the product. If both parameters are nonzero then both are invertible, and $a\,e_t\,b=ab\neq0$, so the operator is not zero.
Theorem. For four elements, $$ T_{a,b}\,T_{c,d}=T_{ac,\,db}. $$
Proof. Apply the left side to an arbitrary $X$: the composite sends $X$ to $a(cXd)b$, which is $(ac)X(db)$; the first parameter multiplies on the left in the order written, and the second multiplies on the right in the reversed order.
Corollary. The products are closed under composition, the operator $T_{e_t,e_t}$ is the identity, and the reversal of the second parameter is the only difference between this law and the product law of the tensor algebra.
Proof. Put $a=c=e_t$ for the identity, $b=d=e_t$ for the unit, and compare with the tensor law $a\otimes b\cdot c\otimes d=ac\otimes bd$, whose second slot is not reversed. The reversal is the content of the word opposite.
Corollary. Left and right multiplications commute: $$ L_aR_b=R_bL_a . $$
Proof. Both send $X$ to $aXb$.
The reading. The reversal is not a defect. In the algebra $\mathbb{B}$ the product of two elements is written in one order, and an operator that multiplies on the left of $X$ and on the right of $X$ has to respect that order: the right parameter meets $X$ first, so it is the last to enter the composite.
The Generation
Theorem. The sixteen operators $T_{e_\mu,e_\nu}$ are a basis of $\operatorname{End}(\mathbb{B})$.
Proof. They are independent. For the Hermitian form of the algebra, $\langle X,Y\rangle_{*}=\mathrm{Sc}(XY^{*})$, the Hilbert–Schmidt pairing of two such operators is $$ \langle T_{e_\alpha,e_\beta},T_{e_\mu,e_\nu}\rangle_{*}=\operatorname{tr}\bigl(T_{e_\alpha,e_\beta}^{\dagger}T_{e_\mu,e_\nu}\bigr)=4\,\delta_{\alpha\mu}\delta_{\beta\nu}, $$ because the adjoint of $T_{a,b}$ is $T_{a^{*},b^{*}}$, the composite is $T_{e_\alpha^{*}e_\mu,\,e_\nu e_\beta^{*}}$, and its trace is $4\,\mathrm{Sc}(e_\alpha^{*}e_\mu)\,\mathrm{Sc}(e_\nu e_\beta^{*})$, which is $4$ at $\alpha=\mu$, $\beta=\nu$ and $0$ otherwise. Sixteen independent operators in a space of dimension sixteen are a basis.
Corollary. Every operator is a sum of at most sixteen two-sided products, and every two-sided product is a single operator and not a sum.
Proof. The expansion in the basis, and the definition of $T_{a,b}$ as one map.
Corollary. The two-sided operators span the whole space, so the span of the cone of products is $\operatorname{End}(\mathbb{B})$ and not a proper subspace.
Proof. The basis consists of products.
The reading. The count is exact: sixteen products and sixteen dimensions, with no relation between them. This is the sharpest form of the statement that two four-dimensional families fill a sixteen-dimensional space.
The Tensor Reading
Theorem. The linear map $$ \Phi:\mathbb{B}\otimes_{\mathbb{C}}\mathbb{B}^{\mathrm{opp}}\longrightarrow\operatorname{End}(\mathbb{B}),\qquad \Phi(a\otimes b^{\mathrm{op}})=T_{a,b}, $$ is an isomorphism of algebras.
Proof. The map is well defined and bilinear on the generators, hence linear on the tensor product. It is injective because the sixteen products of the basis units are independent, and the tensor product has dimension sixteen; it is surjective for the same reason. It carries a product to a product by the law $T_{a,b}T_{c,d}=T_{ac,db}$, which is the product $ac\otimes(db)^{\mathrm{op}}$ of the tensor algebra with the reversed second factor.
Corollary. The representation is the regular representation: the left factor of the tensor product acts by left multiplication and carries the direct product law $L_{ab}=L_aL_b$, while the right factor acts by right multiplication and carries the reversed law $R_{ab}=R_bR_a$.
Proof. The two laws are the theorem read at $b=e_t$ and at $a=e_t$.
Corollary. The two-sided operators are the decomposable elements of the tensor square, so the algebra of the operators is generated by its decomposable elements and the whole space is their span.
Proof. The isomorphism carries pure tensors to products and the tensor square is spanned by its pure tensors.
The reading. The formula $\operatorname{End}(\mathbb{B})=\mathbb{B}\otimes\mathbb{B}^{\mathrm{opp}}$ says that an operator is a pair of elements with a rule for how they act. The rule is not symmetric: one element acts on the left and one on the right, and the reversal in the product law records which is which. The space therefore carries two structures at once, an algebra structure from the composition and a tensor structure from the decomposition, and the isomorphism identifies them.
The Trace Form
Theorem. For four elements, $$ \operatorname{tr}(T_{a,b}T_{c,d})=4\,\mathrm{Sc}(ac)\,\mathrm{Sc}(db). $$
Proof. The composite is $T_{ac,db}$, and the trace of $T_{A,B}$ is $4\,\mathrm{Sc}(A)\,\mathrm{Sc}(B)$ by the trace of the two one-sided factors and the commutativity of $L_A$ with $R_B$, the two being computed on the four units.
Corollary. The pairing $\operatorname{tr}(ST)$ is a nondegenerate symmetric bilinear form on the operator space, and its inertia on the sixteen complex dimensions is $(10,6)$.
Proof. The form is the product of two copies of the plain form of the algebra, $\mathrm{Sc}(ac)$ in the first parameters and $\mathrm{Sc}(db)$ in the second; the plain form is nondegenerate of inertia $(4,4)$ over the reals, and the product of two such forms has inertia $(4\cdot4+4\cdot4,\,4\cdot4+4\cdot4)$ on the real space, which is $(16,16)$ there and $(10,6)$ on the complex basis of products, where the sign of a product $T_{e_\alpha,e_\beta}$ is the product of the two signs.
Corollary. The trace form is degenerate on no operator: an operator paired with every $T_{a,b}$ gives zero only if it is zero.
Proof. Nondegeneracy of the product of two nondegenerate forms.
The reading. The trace form is the operator form of the plain form of the algebra. It is indefinite, and the ten and the six are the same two numbers that will separate the symmetric operators from the antisymmetric ones in the operator space; the coincidence is not accidental, since both counts come from the sign of the product of two basis units.
Simplicity and the Centre
Theorem. The algebra $\operatorname{End}(\mathbb{B})$ has no two-sided ideal except $0$ and itself.
Proof. Let $J$ be a nonzero two-sided ideal and $S\in J$ nonzero. The image $J\mathbb{B}$ is a subspace of $\mathbb{B}$ invariant under every operator, because $F(Su)=(FS)u$ lies in $J\mathbb{B}$; the only subspaces invariant under all operators are $0$ and $\mathbb{B}$, so $J\mathbb{B}=\mathbb{B}$. Hence $J$ contains an operator of full rank, which is invertible, and an ideal containing an invertible element is the whole algebra.
Corollary. The centre of the algebra consists of the scalar operators.
Proof. An operator commuting with every other is determined by its value at one nonzero element, since the operators that fix that element and kill a complement can be varied; the value must be a multiple of the element.
Corollary. The operator algebra is the full algebra of linear maps of a four-dimensional complex space, and it is simple, of dimension sixteen and of centre one.
Proof. The two statements above, with the dimension count of the first section.
The reading. Simplicity is the reason the operator space carries no invariant proper subspace under its own two-sided action, and it is the reason every operator is reachable from the two families. A reader who expects a smaller algebra, one respecting the two-sided origin of its elements, finds instead that the two-sided origin generates everything and that nothing survives as an invariant part.
Summary of Notation
| Symbol | Meaning |
|---|---|
| $\operatorname{End}(\mathbb{B})$ | the complex-linear operators on the algebra, of dimension sixteen |
| $L_a(X)=aX$, $R_b(X)=Xb$ | the two one-sided families |
| $T_{a,b}(X)=aXb$ | the two-sided operator of the pair |
| $T_{a,b}T_{c,d}=T_{ac,db}$ | the product law, second slot reversed |
| $\operatorname{tr}(F)$ | the sum of the eigenvalues, and of the coefficients of the values |
| $\operatorname{tr}(T_{a,b})=4\,\mathrm{Sc}(a)\mathrm{Sc}(b)$ | the trace of a two-sided operator |
| $\operatorname{tr}(T_{a,b}T_{c,d})=4\,\mathrm{Sc}(ac)\mathrm{Sc}(db)$ | the trace form |
| $\mathbb{B}\otimes\mathbb{B}^{\mathrm{opp}}\cong\operatorname{End}(\mathbb{B})$ | the regular representation |
Further Reading
- Introduction to the Operators of the Biquaternion Algebra, for the operator space and its first properties.
- Operators Acting on One Side, for the two families and their multiplication laws.
- The One-Sided Operators, for the intersection of the two families and the nine-dimensional complement.
- The Tensor Product of Two Biquaternion Algebras, for the tensor square used in the representation.
- The Enveloping Operator Algebra of the Biquaternion Algebra and the Bi-module Structure, for the isomorphism read as a bi-module statement.