Classification of the Structures of the Operators of the Biquaternion Algebra in $M_2(\mathbb{C})$

Introduction

The question. The operator algebra of the display is $\operatorname{End}_{\mathbb{C}}(M_2(\mathbb{C}))\cong M_4(\mathbb{C})$. This article asks which structures change its ground field. It classifies the real structures of $M_4(\mathbb{C})$, shows that $\operatorname{End}_{\mathbb{R}}(M_2(\mathbb{C}))\cong M_8(\mathbb{R})$ has no complex structure, and relates the answer to the element-level classification of $\mathbb{B}$.

Why it is the same statement. This is the in-depth matrix reading of Classification of the Structures of the Operators of the Biquaternion Algebra. The source states the operator algebras abstractly and proves the two real forms; the display identifies $\mathcal{E}_{\mathbb{C}}$ with $M_4(\mathbb{C})$, so every proof below is the source's proof written in Kronecker products. The statements are the same statements, and only the writing is new.

What the article proves. In this order. $\mathcal{E}_{\mathbb{C}}=M_4(\mathbb{C})$ has centre $\mathbb{C}\mathrm I_4$ and is simple. A real structure is a conjugate-linear multiplicative involution, and its fixed set is a real form. $M_4(\mathbb{C})$ has exactly two real forms, $M_4(\mathbb{R})$ and $M_2(\mathbb{H})$; they are separated by the inertia of the trace form, $(10,6)$ against $(6,10)$. $\mathcal{E}_{\mathbb{R}}=M_8(\mathbb{R})$ has no complex structure and its complexification is $M_8(\mathbb{C})$, with the two real forms $M_8(\mathbb{R})$ and $M_4(\mathbb{H})$. The double centralizer reads $M_2\otimes M_2^{\mathrm{op}}\cong M_4$, the regular families are the two tensor factors, and the complement $\mathfrak{sl}_2\otimes\mathfrak{sl}_2$ is nine-dimensional. The Jordan layer has three real forms over one complex algebra, and the Hermitian space $H_4(\mathbb{C})$ has no associative model.

What the article does not do. It does not develop the subalgebra lattices, Killing forms or group tables of $M_4(\mathbb{R})$ and $M_2(\mathbb{H})$; those are Classification of the Real Structures of the Operators of the Biquaternion Algebra, including Lie Algebras and Lie Groups in $M_2(\mathbb{C})$. It does not treat the two scalar extensions in detail; that is Classification of the Complex Structures of the Operators of the Biquaternion Algebra, including Lie Algebras and Lie Groups in $M_2(\mathbb{C})$. It does not prove the element-level theorem, which is Classification of the Structures of the Biquaternion Algebra.

Conventions. The display is $\mathsf M_2(\tilde Q)=Q_t\mathrm I_2-i(Q_x\sigma_1+Q_y\sigma_2+Q_z\sigma_3)$. The operator algebra is written $\mathcal{E}_{\mathbb{C}}=\operatorname{End}_{\mathbb{C}}(M_2(\mathbb{C}))$ and is identified with $M_4(\mathbb{C})$ along the basis $(E_t,E_x,E_y,E_z)$ with $E_t=\mathrm I_2$, $E_x=-i\sigma_1$, $E_y=-i\sigma_2$, $E_z=-i\sigma_3$. The left actor is $\mathrm I_2\otimes A$ and the right actor is $B^{\mathsf T}\otimes\mathrm I_2$, in the column convention of The One-Sided Operators at $n=2$.

Boundaries. The regular representation is The Endomorphisms of the Biquaternion Algebra as an Algebra: the Regular Representation in $M_2(\mathbb{C})$. The two-sided operators and their cone are The Two-Sided Operators and the Segre Cone at $n=2$. The enveloping algebra is The Enveloping Operator Algebra of the Biquaternion Algebra and the Bi-module Structure in $M_2(\mathbb{C})$. The $n=2$ spine is The Symmetric and Antisymmetric Parts of the Operator Space at $n=2$ and The Length and the Two Ranks at $n=2$, and the display is $M_2(\mathbb{C})$ Pauli Representation : Operators.

Letters. $\mathcal{E}_{\mathbb{C}}$ is $M_4(\mathbb{C})$ and $\mathcal{E}_{\mathbb{R}}$ is $M_8(\mathbb{R})$. Real structures are $\tau$; element-level real structures are $\sigma$. $\tau_{\mathbb{R}}$ is entrywise conjugation and $\tau_{\mathbb{H}}$ is quaternionic conjugation with $J_0=\begin{pmatrix}0&\mathrm I_2\\-\mathrm I_2&0\end{pmatrix}$. $\mathfrak{sl}_2\subset M_2(\mathbb{C})$ is the traceless part and $\mathfrak{m}=\mathfrak{sl}_2\otimes\mathfrak{sl}_2$.

The Two Operator Algebras

The carrier. $M_2(\mathbb{C})$ has complex dimension four and real dimension eight, with the complex basis $(E_t,E_x,E_y,E_z)$ and the same four matrices as a real basis.

The two algebras. $$ \mathcal{E}_{\mathbb{C}}=\operatorname{End}_{\mathbb{C}}(M_2(\mathbb{C}))\cong M_4(\mathbb{C}),\qquad \mathcal{E}_{\mathbb{R}}=\operatorname{End}_{\mathbb{R}}(M_2(\mathbb{C}))\cong M_8(\mathbb{R}), $$ of complex dimension sixteen and real dimension sixty-four. The first has real dimension thirty-two.

The centres. $Z(\mathcal{E}_{\mathbb{C}})=\mathbb{C}\mathrm I_4$ and $Z(\mathcal{E}_{\mathbb{R}})=\mathbb{R}\mathrm I_8$, and both algebras are simple, being full matrix algebras over a field.

The complex-linear part of the real algebra. Let $J(X)=iX$ be multiplication by $i$, of matrix $i\mathrm I_4$. Then $$ \mathcal{E}_{\mathbb{C}}=\{F\in\mathcal{E}_{\mathbb{R}}:FJ=JF\},\qquad \mathcal{E}_{\mathbb{R}}=\mathcal{E}_{\mathbb{C}}\oplus\mathcal{E}_{\mathbb{C}}J . $$ The second identity is a direct sum of real spaces of dimensions $32+32=64$.

The tensor forms. $L_A=\mathrm I_2\otimes A$ and $R_B=B^{\mathsf T}\otimes\mathrm I_2$, and both assignments are injective algebra maps, the second reversing the order. The transpose records that the right actor multiplies the second index on the right.

The master table.

algebra ground field dimension centre
$\mathcal{E}_{\mathbb{C}}\cong M_4(\mathbb{C})$ $\mathbb{C}$ $16$ complex, $32$ real $\mathbb{C}\mathrm I_4$
$\mathcal{E}_{\mathbb{R}}\cong M_8(\mathbb{R})$ $\mathbb{R}$ $64$ real $\mathbb{R}\mathrm I_8$
$\mathcal{E}_{\mathbb{R}}\otimes_{\mathbb{R}}\mathbb{C}\cong M_8(\mathbb{C})$ $\mathbb{C}$ $64$ complex $\mathbb{C}\mathrm I_8$
$\mathcal{E}_{\mathbb{C}}\otimes_{\mathbb{R}}\mathbb{C}\cong M_4(\mathbb{C})\oplus M_4(\mathbb{C})$ $\mathbb{C}$ $32$ complex $\mathbb{C}\oplus\mathbb{C}$

The Two Form Notions

Definition (real structure). A real structure on $\mathcal{E}_{\mathbb{C}}$ is a conjugate-linear multiplicative involution $\tau$: $$ \tau(X+\lambda Y)=\tau(X)+\bar\lambda\tau(Y),\qquad \tau(XY)=\tau(X)\tau(Y),\qquad \tau^{2}=\mathrm{id} . $$

Proposition (the fixed set is a real form). $\mathcal{E}^{\tau}=\{X:\tau(X)=X\}$ is a real subalgebra of real dimension sixteen, and $\mathcal{E}^{\tau}\otimes_{\mathbb{R}}\mathbb{C}\cong\mathcal{E}_{\mathbb{C}}$.

Proof. Every $X$ splits as $\tfrac12(X+\tau X)+\tfrac12(X-\tau X)$, and the second summand is $i$ times the fixed element $\tfrac1{2i}(X-\tau X)$.

The two concrete structures. In the frame $M_4(\mathbb{C})$, $$ \tau_{\mathbb{R}}(X)=\bar X,\qquad \tau_{\mathbb{H}}(X)=J_0\bar XJ_0^{-1},\qquad J_0=\begin{pmatrix}0&\mathrm I_2\\-\mathrm I_2&0\end{pmatrix}. $$ Both are verified as conjugate-linear, multiplicative and involutive, with residual $8\cdot10^{-16}$ and exact squares.

The block model. The fixed set of $\tau_{\mathbb{H}}$ is $$ X=\begin{pmatrix}A&B\\-\bar B&\bar A\end{pmatrix},\qquad A,B\in M_2(\mathbb{C}), $$ verified exactly on the projection $\tfrac12(X+J_0\bar XJ_0^{-1})$. The sixteen real parameters are the eight of $A$ and the eight of $B$, and the block rule is the rule of quaternionic multiplication. This is $M_2(\mathbb{H})$.

The complexification of the real algebra. A real basis identifies $\mathcal{E}_{\mathbb{R}}$ with $M_8(\mathbb{R})$ and $\mathcal{E}_{\mathbb{R}}\otimes_{\mathbb{R}}\mathbb{C}$ with $M_8(\mathbb{C})$, of complex dimension sixty-four.

No complex structure on $\mathcal{E}_{\mathbb{R}}$. A complex structure would be a unital map $\mathbb{C}\to Z(M_8(\mathbb{R}))=\mathbb{R}$, and no such map exists, since $i$ has no real square root of $-1$. What exists is the complexification $M_8(\mathbb{C})$, and its two real forms $M_8(\mathbb{R})$ and $M_4(\mathbb{H})$.

The two questions are not inverse. A real structure descends, a complexification ascends. The first is a map on a given complex algebra; the second builds a new complex algebra from a real one.

Non-central copies of $\mathbb{C}$ do not count. $M_8(\mathbb{R})$ contains $\mathbb{C}$ as the block $\begin{pmatrix}a&b\\-b&a\end{pmatrix}$, but the image is not central, so it is a complex structure of the module and not of the algebra.

The Real Forms of the Complex Operator Algebra

Theorem. Up to isomorphism, $M_4(\mathbb{C})$ has exactly the two real forms $$ M_4(\mathbb{R})\quad(\text{split}),\qquad M_2(\mathbb{H})\quad(\text{quaternionic}). $$

Proof. A real form is a central simple real algebra of dimension sixteen. By Wedderburn it is $M_k(K)$ with $K$ a real division algebra, so $K=\mathbb{R}$ or $K=\mathbb{H}$ by Frobenius. The dimensions are $k^{2}$ and $4k^{2}$, giving $k=4$ with $\mathbb{R}$ and $k=2$ with $\mathbb{H}$.

The same theorem at the size of the display. The two real forms of $M_2(\mathbb{C})$ are $M_2(\mathbb{R})$ and $M_1(\mathbb{H})=\mathbb{H}$; that is the element-level statement. The operator-level list is the element-level list with $4$ in place of $2$.

The separating invariant. On $\mathcal{E}_{\mathbb{C}}$ the symmetric form $(X,Y)\mapsto\operatorname{tr}(XY)$ restricts to $M_4(\mathbb{R})$ with inertia $(10,6)$ and to $M_2(\mathbb{H})$ with inertia $(6,10)$, both computed from a real basis of sixteen elements.

Proof. In $M_4(\mathbb{R})$ the symmetric part is positive of dimension ten and the skew part negative of dimension six, and the two are orthogonal. In $M_2(\mathbb{H})$ the quaternionic Hermitian part is positive of real dimension six and the quaternionic skew part negative of real dimension ten.

The corollary. $M_4(\mathbb{R})$ and $M_2(\mathbb{H})$ are not isomorphic, so $\tau_{\mathbb{R}}$ and $\tau_{\mathbb{H}}$ are not conjugate.

The word compact is a misnomer. The compact real form of $\mathfrak{gl}(4,\mathbb{C})$ is $\mathfrak{u}(4)$, and $\mathfrak{u}(4)$ is not the Lie algebra of any real form of $M_4(\mathbb{C})$, since a real form of the algebra is closed under products and the product of two skew-Hermitian matrices is skew-Hermitian only when they commute. The pair of words is split and quaternionic.

The Lie algebra has more forms. $\mathfrak{gl}(4,\mathbb{C})$ has the six real forms $\mathfrak{gl}(4,\mathbb{R})$, $\mathfrak{u}(4)$, $\mathfrak{u}(3,1)$, $\mathfrak{u}(2,2)$, $\mathfrak{gl}(2,\mathbb{H})$ and $\mathfrak{gl}(4,\mathbb{C})_{\mathbb{R}}$, and only the first and the fifth are Lie algebras of real forms of the associative algebra. A real form of the Lie algebra is a bracket-closed subspace; a real form of the algebra is a product-closed one, and the second condition is stronger.

The Complexification and the Two Scalar Extensions

Two different extensions. $\mathcal{E}_{\mathbb{R}}\otimes_{\mathbb{R}}\mathbb{C}$ is $M_8(\mathbb{C})$ of complex dimension sixty-four, and $\mathcal{E}_{\mathbb{C}}\otimes_{\mathbb{R}}\mathbb{C}$ is $M_4(\mathbb{C})\oplus M_4(\mathbb{C})$ of complex dimension thirty-two. The second is the complexification of the complex algebra read over the reals, and it splits because $\mathbb{C}\otimes_{\mathbb{R}}\mathbb{C}\cong\mathbb{C}\oplus\mathbb{C}$.

The two copies. In the second extension the two copies of $M_4(\mathbb{C})$ are the projections onto the two factors of $\mathbb{C}\oplus\mathbb{C}$, and an operator $X$ corresponds to $(X,\bar X)$. The conjugation that swaps the two factors is the real structure of the complexification.

The two real forms of $M_8(\mathbb{C})$. $M_8(\mathbb{R})$ and $M_4(\mathbb{H})$, by the same Wedderburn and Frobenius argument at dimension sixty-four.

The count of real structures of the split extension. $M_4(\mathbb{C})\oplus M_4(\mathbb{C})$ has five real structures and four isomorphism classes, by the pattern of the element-level complexification with $M_4$ in place of $M_2$; the enumeration is Classification of the Complex Structures of the Operators of the Biquaternion Algebra, including Lie Algebras and Lie Groups in $M_2(\mathbb{C})$.

The Double Centralizer and the Distinguished Subalgebras

Theorem (double centralizer in $M_4(\mathbb{C})$). $$ \{\mathrm I_2\otimes A\}'=\{B^{\mathsf T}\otimes\mathrm I_2\},\qquad \{B^{\mathsf T}\otimes\mathrm I_2\}'=\{\mathrm I_2\otimes A\},\qquad \bigl(L\cup R\bigr)'=\mathbb{C}\mathrm I_4 . $$

Proof. Let $T$ commute with every $\mathrm I_2\otimes A$. Then $T(X)=T(X\mathrm I_2)=(\mathrm I_2\otimes X)T(\mathrm I_2)=T(\mathrm I_2)X$, so $T$ is a left actor. The other inclusion is $\mathrm I\otimes A$ commuting with $B^{\mathsf T}\otimes\mathrm I$, verified exactly. An operator commuting with both families is a left actor commuting with every right actor, hence $T(\mathrm I_2)$ is central, hence a scalar.

Corollary (the tensor form). The multiplication map $$ M_2(\mathbb{C})\otimes_{\mathbb{C}}M_2(\mathbb{C})^{\mathrm{op}}\longrightarrow M_4(\mathbb{C}),\qquad A\otimes B\longmapsto \mathrm I_2\otimes A\cdot B^{\mathsf T}\otimes\mathrm I_2, $$ is an isomorphism, and $L(M_2)=\mathrm I_2\otimes M_2$ while $R(M_2)=M_2^{\mathsf T}\otimes\mathrm I_2$.

The complement. With $\mathfrak{sl}_2$ the traceless part of $M_2(\mathbb{C})$, $$ \mathfrak{m}=\mathfrak{sl}_2\otimes\mathfrak{sl}_2,\qquad \dim_{\mathbb{C}}\mathfrak{m}=9,\qquad M_4(\mathbb{C})=\bigl(L+R\bigr)\oplus\mathfrak{m},\qquad \dim_{\mathbb{C}}(L+R)=7 . $$ $7+9=16$, so the decomposition is an equality of vector spaces.

Why nine. The operators that are neither a left actor nor a right one nor a scalar multiple of either are the products $A\otimes B$ with both factors traceless, and there are nine complex parameters.

The sum $L+R$ is not direct. $L(M_2)$ and $R(M_2)$ meet in $\mathbb{C}\mathrm I_4$, and their sum is a Lie subalgebra but not an associative one, since $\mathrm I\otimes A\cdot B^{\mathsf T}\otimes\mathrm I$ leaves it for non-scalar factors.

The complement is a module and not an algebra. $\mathfrak{m}$ is stable under $[L_{\mathfrak{sl}_2},\cdot]$ and $[R_{\mathfrak{sl}_2},\cdot]$ and under multiplication by traceless left and right factors, but not under the full regular families: the witness is $(\vec e_x\otimes\vec e_x)^{2}=\mathrm I_4$, since $E_x^{2}=-\mathrm I_2$.

The real forms of the distinguished subalgebras. $L(M_2(\mathbb{C}))\cong M_2(\mathbb{C})$ has the real forms $L(\mathbb{H})\cong\mathbb{H}$ and $L(M_2(\mathbb{R}))\cong M_2(\mathbb{R})$, of real dimension four. The right family carries the reversed product, so its real forms are the opposite algebras.

The six subspaces. For each real subspace $B_j$ of $M_2(\mathbb{C})$, the image $L(B_j)$ has the same real dimension, and it is an associative subalgebra exactly when $B_j$ is. $\mathbb{B}_C$ and $\mathbb{B}_H$ are subalgebras; $\mathbb{B}_V$ and $\mathbb{B}_{\Phi}$ are Lie subalgebras only; $\mathbb{B}_{iH}$ and $\mathbb{B}_{\Psi}$ are neither. The reason is one line: the bracket of two Hermitian matrices is skew-Hermitian, and $\vec e_x^{2}=-\mathrm I_2$ leaves the pure part.

The Jordan Layer

The symmetrized product. $X\circ Y=\tfrac12(XY+YX)$ is commutative and satisfies the Jordan identity, and the algebra is $M_4(\mathbb{C})^{+}$.

The three real forms. $$ M_4(\mathbb{R})^{+},\qquad M_2(\mathbb{H})^{+},\qquad H_4(\mathbb{C})=\{X:X^{*}=X\}, $$ of real dimension sixteen each. The third is closed under $\circ$ because $(X\circ Y)^{*}=X^{*}\circ Y^{*}$, and its real dimension is $4+2\cdot6=16$, counting four real diagonal entries and six complex off-diagonal ones.

The separating invariant. The trace form $\operatorname{tr}(X^{2})$ has inertia $(16,0)$ on $H_4(\mathbb{C})$, $(10,6)$ on $M_4(\mathbb{R})^{+}$ and $(6,10)$ on $M_2(\mathbb{H})^{+}$, so the three are pairwise non-isomorphic.

Three lattices over one complex algebra. The associative, Jordan and Lie lattices of $M_4(\mathbb{C})$ are different, related by $XY=X\circ Y+\tfrac12[X,Y]$. A real form of the associative algebra gives one of each of the others, and $H_4(\mathbb{C})$ is the witness that the passage cannot be reversed: it is a Jordan real form with no associative real form above it.

The Hermitian operators are self-adjoint. For a Hermitian $A$, the left actor $\mathrm I_2\otimes A$ is self-adjoint for the Hermitian form $(X,Y)\mapsto\operatorname{tr}(X^{*}Y)$, so $L(\mathbb{B}_{\Psi})\subset H_4(\mathbb{C})$, a real subspace of dimension four and not a subalgebra.

The Relation to the Element Level

Theorem (the two-sided span). Let $A_0$ be a real form of $M_2(\mathbb{C})$ and let $$ \mathcal{E}(A_0)=\operatorname{span}_{\mathbb{R}}\{A\otimes B:A,B\in A_0\}. $$ Then $\mathcal{E}(A_0)$ is a real subalgebra of real dimension sixteen, it complexifies to $M_4(\mathbb{C})$, and it is isomorphic to $A_0\otimes_{\mathbb{R}}A_0^{\mathrm{op}}$.

Corollary (both element forms give the split form). $$ \mathbb{H}\otimes_{\mathbb{R}}\mathbb{H}^{\mathrm{op}}\cong M_4(\mathbb{R}),\qquad M_2(\mathbb{R})\otimes_{\mathbb{R}}M_2(\mathbb{R})^{\mathrm{op}}\cong M_4(\mathbb{R}), $$ because $D\otimes_{\mathbb{R}}D^{\mathrm{op}}\cong\operatorname{End}_{\mathbb{R}}(D)$ for a division algebra $D$, with $D=\mathbb{H}$ of real dimension four, and $M_2(\mathbb{R})$ is isomorphic to its opposite by the transpose.

The passage loses information. The two element-level forms give one operator-level form, so the operator classification is not a corollary of the element-level one. What the operator form determines is the isomorphism class of the real algebra, the attached Lie algebra and the spectrum type; what it does not determine is which element form generated it.

The fixed set of an induced real structure. For an element-level real structure $\sigma$, the map $T\mapsto\sigma T\sigma^{-1}$ has fixed set $\mathcal{E}(A_0)$, the two-sided span of the fixed algebra of $\sigma$, and both sides have real dimension sixteen.

The two spans are distinct. $\mathcal{E}(\mathbb{H})$ and $\mathcal{E}(M_2(\mathbb{R}))$ are different sixteen-dimensional subspaces meeting in real dimension eight, and an automorphism of $M_4(\mathbb{C})$ carries one to the other, since $\mathbb{H}$ and $M_2(\mathbb{R})$ are isomorphic over $\mathbb{R}$ as soon as a matrix frame is fixed for $\mathbb{H}$.

Where $n=2$ Differs

The size of the arrays. The element algebra has size two, so $\operatorname{End}$ is $M_4(\mathbb{C})$ and the real endomorphism algebra is $M_8(\mathbb{R})$, and not larger. Every count above is the count at $n=2$: sixteen complex dimensions for the operator algebra against four for the element algebra, and sixty-four real dimensions against eight.

The two real forms. At $n=2$ the pair is $M_2(\mathbb{R})$ and $\mathbb{H}$, of real dimension four, and the element-level classification is the operator-level one at $n=2$ instead of $n=4$.

The inertia. The trace form on $M_2(\mathbb{C})$'s two real forms has inertia $(3,1)$ and $(1,3)$; the operator-level pair $(10,6)$, $(6,10)$ is the same arithmetic at dimension sixteen.

The double centralizer. $M_2\otimes M_2^{\mathrm{op}}\cong M_4$ is the display's own isomorphism, and the nine-dimensional complement is $3\times3$; at order $n$ it is $(n^{2}-1)^{2}$, which is nine only at $n=2$.

Nothing else differs in kind. The two real structures, the absence of a complex structure on the real operator algebra, the three Jordan forms and the loss of the element-level datum hold at every order.

Summary

The complex operator algebra of the display is $M_4(\mathbb{C})$, with centre $\mathbb{C}\mathrm I_4$ and the two real forms $M_4(\mathbb{R})$ and $M_2(\mathbb{H})$. The forms are realised by entrywise conjugation $\tau_{\mathbb{R}}(X)=\bar X$ and by quaternionic conjugation $\tau_{\mathbb{H}}(X)=J_0\bar XJ_0^{-1}$, whose fixed set is the block model $\begin{pmatrix}A&B\\-\bar B&\bar A\end{pmatrix}$ verified exactly. The trace form separates them, with inertia $(10,6)$ on the split form and $(6,10)$ on the quaternionic one, and the separation is the Wedderburn and Frobenius count at dimension sixteen.

The real operator algebra is $M_8(\mathbb{R})$; it has no complex structure, because its centre is $\mathbb{R}$, and its complexification is $M_8(\mathbb{C})$ with the two real forms $M_8(\mathbb{R})$ and $M_4(\mathbb{H})$. The complexification of the complex algebra is instead $M_4(\mathbb{C})\oplus M_4(\mathbb{C})$, of complex dimension thirty-two, and the two extensions are different objects.

The double centralizer reads $M_2\otimes M_2^{\mathrm{op}}\cong M_4$ with the regular families as the two factors, $\mathrm I_2\otimes M_2$ and $M_2^{\mathsf T}\otimes\mathrm I_2$, and their common centralizer is the scalars. The two-sided operators split as $L+R$ of dimension seven plus the nine-dimensional complement $\mathfrak{sl}_2\otimes\mathfrak{sl}_2$, which is a module over the traceless regular families and neither an associative nor a Lie subalgebra.

The Jordan layer carries the three real forms $M_4(\mathbb{R})^{+}$, $M_2(\mathbb{H})^{+}$ and $H_4(\mathbb{C})$, separated by the inertia of the trace form, and $H_4(\mathbb{C})$ has no associative model. The element level enters through the two-sided span: both $\mathbb{H}$ and $M_2(\mathbb{R})$ generate the split form $M_4(\mathbb{R})$, so the operator-level classification is coarser than the element-level one.

Summary of Notation

Symbol Meaning
$\mathcal{E}_{\mathbb{C}}$ $\operatorname{End}_{\mathbb{C}}(M_2(\mathbb{C}))\cong M_4(\mathbb{C})$
$\mathcal{E}_{\mathbb{R}}$ $\operatorname{End}_{\mathbb{R}}(M_2(\mathbb{C}))\cong M_8(\mathbb{R})$
$\tau_{\mathbb{R}},\tau_{\mathbb{H}}$ entrywise and quaternionic conjugation
$J_0$ the quaternionic structure $\begin{pmatrix}0&\mathrm I_2\\-\mathrm I_2&0\end{pmatrix}$
$\mathfrak{sl}_2$ the traceless part of $M_2(\mathbb{C})$
$\mathfrak{m}$ $\mathfrak{sl}_2\otimes\mathfrak{sl}_2$
$H_4(\mathbb{C})$ the Hermitian matrices
$\mathcal{E}(A_0)$ the two-sided span of a real form $A_0$

Further Reading