The Material-Informational Split as a Superselection Structure in Biquaternionic Form

Introduction

The biquaternion algebra $\mathbb{B} = \mathbb{C}\otimes_\mathbb{R}\mathbb{H}$ contains a four-real-dimensional subspace $\mathbb{M}_-$ whose elements are the anti-Hermitian biquaternions and a complementary four-real-dimensional subspace $\mathbb{M}_+$ whose elements are the Hermitian ones, $$ \mathbb{B} = \mathbb{M}_- \oplus \mathbb{M}_+, \qquad \mathbb{M}_\pm = \{\tilde{Q} : \tilde{Q}^{*} = \pm \tilde{Q}\}. $$ The two summands are the material and the informational subspaces of the foundational articles: $\mathbb{M}_-$ carries the four-vectors, the interval, and the configuration data of a relativistic system, while $\mathbb{M}_+$ carries the states and observables of its quantum description. This is a kinematic split, and the purpose of this article is to ask a structural question about it: does it define a superselection structure?

The question is sharp because the two things that a superselection structure is expected to supply are both present in a weak form and both absent in the strong form. Present: the split is the eigenspace decomposition of an involutive charge, no operation of the framework of the conjugation type maps one summand into the other, and the relative phase between the two summands is unobservable. Absent: the framework is central simple, its only central idempotents are $0$ and $e_0$, and it therefore admits no decomposition into non-interfering branches of the standard algebraic kind. The thesis of this article is that both statements are exact, that they do not conflict, and that the correct reading of the split is as a superselection structure of a real form — a grading carried by the algebra's antilinear real structure $\flat = -{}^{*}$ rather than by its centre.

The distinction matters for the relativistic quantum information that this subcategory develops. A superselection structure is the statement that some coherences are not observable; the material-informational grading says precisely which ones, namely the coherences generated by the central complex structure, and it thereby fixes what a change of frame can and cannot do to the information carried by a state. The relativistic qubit is the carrier of that information; frame-dependent entanglement and the Wigner rotation are the action of the frame group on it; spin entropy is its entropic face. The present article supplies the structural statement on which those four rest.

The article proceeds as follows. The standard algebraic notion of a superselection structure is recalled, in the form in which it will be tested. Then the two sectors are described as the eigenspaces of an involution, and the algebra's own operations are examined: which of them preserve the sectors and which mix them. Then the factor property is derived and its consequence stated: there is no central-projection superselection structure. Then the structure that does survive is isolated — the real-structure grading, its invariance under the conjugation action, and the unobservability of the relative phase. Then the same split is read on the Lie algebra, where it becomes the Cartan decomposition of $\mathrm{SL}(2,\mathbb{C})$ and acquires the sharpest algebraic content available. The physical reading and the framing of the subcategory close the article.

The notation is that of the foundational articles: $\mathbb{B} = \mathbb{C}\otimes_\mathbb{R}\mathbb{H}$, quaternion units $e_0 = 1, e_1, e_2, e_3$ with $e_k^2 = -e_0$ and $e_1e_2 = e_3$, central scalar imaginary $i$ with $i^2 = -1$, trace $\mathrm{Tr}(\tilde{Q}) = 2\,\mathrm{Sc}(\tilde{Q})$, Hermitian conjugation $\tilde{Q}^{*}$, quaternion conjugate $\tilde{Q}^{\natural}$, and the real structure $\flat = -{}^{*}$ with fixed space $\mathbb{M}_-$. The matrix model is $\mathbb{B}\cong M_2(\mathbb{C})$ with $e_0\mapsto I_2$, $e_k\mapsto -i\sigma_k$ and $i\mapsto iI_2$.

The companion articles supply the pieces: - Companion article The Anti-Hermitian Subspace M- as the Material Sector, for the four-vector content, the interval, and the anti-Hermitian sector. - Companion article The Hermitian Subspace M+ as the Informational Sector, for the trace pairing, the Hermitian sector, and the Born rule. - Companion article Quantum Mechanics in Biquaternionic Form, for the state space, the idempotents, and the conjugation action. - Companion article The Spinor Module in Biquaternionic Form and Its Lorentz Action, for the defining module and its irreducibility. - Companion article The Lorentz Group in Biquaternionic Form — Structure and Representations, for the Lie algebra, the Cartan decomposition, and the Wigner rotation.

Superselection Structures: The Standard Notion

A superselection structure for a quantum theory is a decomposition of its objects into sectors such that two conditions hold. First, no operation of the theory maps one sector into another: the observables preserve the sectors. Second, the relative phase between components drawn from different sectors is unobservable, so no coherence between them can be prepared or detected. The two conditions are equivalent in the standard algebraic setting, and it is worth recording why, because the equivalence is what makes the notion testable.

The algebraic form. Let $\mathcal{A}$ be the algebra of observables acting on a Hilbert space $\mathcal{H}$, and let $Z(\mathcal{A}) = \mathcal{A}\cap\mathcal{A}'$ be its centre. Every projection $\tilde{C}\in Z(\mathcal{A})$ is a central projection, and it decomposes the representation space, $$ \mathcal{H} = \tilde{C}\mathcal{H} \oplus (1-\tilde{C})\mathcal{H}, $$ with no observable connecting the two summands: since $\tilde{C}$ commutes with every observable $\tilde{A}$, $$ \tilde{C}\tilde{A}(1-\tilde{C}) = 0 . $$ A vector that is a superposition of a vector from each summand therefore carries a relative phase that no expectation value can see, because the interference term vanishes identically. The eigenvalues of the central charge label the sectors, and the charge itself is a superselection quantum number: it is conserved by every operation because it commutes with every operation.

Examples. Electric charge superselection forbids coherent superpositions of states of different total charge; the univalence (boson–fermion) rule forbids superpositions of integer and half-integer spin; the intrinsic-parity rule of Wick, Wightman, and Wigner forbids a relative phase between a particle and its antiparticle in the same spatial state when the charge-conjugation quantum number is superselected. In each case the obstruction is a central charge, and in each case the operational content is the same: a relative phase that cannot be measured.

The kinematic form. For the purpose of testing a candidate structure that is not presented as an algebra of operators on a fixed Hilbert space, it is convenient to record the kinematic form of the same statement. A decomposition $\mathcal{V} = \mathcal{V}_1\oplus\mathcal{V}_2$ of a complex vector space into two real subspaces is a superselection structure if there is an involutive charge $\theta$ with $\mathcal{V}_j$ the eigenspace of eigenvalue $(-1)^{j+1}$, if every operation of the theory commutes with $\theta$ and hence preserves the $\mathcal{V}_j$, and if the only complex structure available for forming a relative phase acts trivially on the rays of the theory. The first two conditions are algebraic and are checked by direct computation; the third is what converts the grading into an obstruction to coherence.

The remainder of the article tests the split $\mathbb{B} = \mathbb{M}_-\oplus\mathbb{M}_+$ against both forms.

The Two Sectors as the Eigenspaces of an Involution

The decomposition

Hermitian conjugation $\tilde{Q}\mapsto\tilde{Q}^{*} = \overline{\tilde{Q}}^{*}$ is an antilinear involution of $\mathbb{B}$: it is additive, it satisfies $(\lambda\tilde{Q})^{*} = \lambda^{*}\tilde{Q}^{*}$ for $\lambda\in\mathbb{C}$, and it satisfies $(\tilde{Q}^{*})^{*} = \tilde{Q}$. Its two eigenspaces are exactly the two sectors, $$ \mathbb{M}_+ = \{\tilde{Q} : \tilde{Q}^{*} = +\tilde{Q}\}, \qquad \mathbb{M}_- = \{\tilde{Q} : \tilde{Q}^{*} = -\tilde{Q}\}, $$ and the eigenspace property gives at once $$ \mathbb{M}_+\cap\mathbb{M}_- = \{0\}, \qquad \mathbb{B} = \mathbb{M}_+\oplus\mathbb{M}_- . $$ Both summands are four-real-dimensional. A general Hermitian element and a general anti-Hermitian element are $$ \tilde{H} = h_0\,e_0 + i\mathbf{h}, \qquad \tilde{Q} = i x_0\,e_0 + \mathbf{x}, \qquad h_0,\mathbf{h},x_0,\mathbf{x}\ \text{real}, $$ so that $\mathbb{M}_+$ is the span of $\{e_0, ie_1, ie_2, ie_3\}$ and $\mathbb{M}_-$ is the span of $\{ie_0, e_1, e_2, e_3\}$. The real structure $$ \flat = -{}^{*}, \qquad \flat(\tilde{Q}) = -\tilde{Q}^{*} , $$ is the involution whose fixed space is $\mathbb{M}_-$ and whose $(-1)$-eigenspace is $\mathbb{M}_+$.

The sector exchange is the complex structure

The subspaces are not complex subspaces of $\mathbb{B}$, because the algebra's complex structure does not preserve them. Multiplication by the central scalar, $$ i\,\mathbb{M}_+ = \mathbb{M}_-, \qquad i\,\mathbb{M}_- = \mathbb{M}_+, $$ is immediate from the definitions: multiplying a Hermitian element by $i$ produces an anti-Hermitian one and conversely. Hence the complex structure $i$ is exactly the operation that exchanges the sectors. Geometrically, $\mathbb{M}_+$ is a maximally real subspace of the complex vector space $(\mathbb{B}, i)$: $$ \dim_\mathbb{R}\mathbb{M}_+ = \tfrac12\dim_\mathbb{R}\mathbb{B}, \qquad \mathbb{M}_+\cap i\,\mathbb{M}_+ = \mathbb{M}_+\cap\mathbb{M}_- = \{0\}. $$ Every element of $\mathbb{B}$ therefore has a unique decomposition into a part in the real form and a part in its image, and the map that effects the decomposition is the real-linear projection $$ P_\pm = \tfrac12\bigl(\mathrm{id} \pm {}^{*}\bigr), \qquad P_+\mathbb{B} = \mathbb{M}_+, \quad P_-\mathbb{B} = \mathbb{M}_- . $$ These projections are not elements of $\mathbb{B}$. They are real-linear operators on the eight-dimensional real vector space $\mathbb{B}$, built from an antilinear involution, and they do not arise from left multiplication by any element:

Proposition. If $\tilde{A}\in\mathbb{B}$ satisfies $\tilde{A}\mathbb{M}_+ \subseteq \mathbb{M}_+$, then $\tilde{A}\in\mathbb{R}e_0$; the same holds with $\mathbb{M}_-$ in place of $\mathbb{M}_+$. Consequently no element of $\mathbb{B}$ implements a sector projection by left multiplication.

Proof. Let $\tilde{H}\in\mathbb{M}_+$, so that $\tilde{H}^{*} = \tilde{H}$. The condition reads $(\tilde{A}\tilde{H})^{*} = \tilde{A}\tilde{H}$, that is $\tilde{H}\tilde{A}^{*} = \tilde{A}\tilde{H}$. Taking $\tilde{H} = e_0$ gives $\tilde{A}^{*} = \tilde{A}$, so $\tilde{A}\in\mathbb{M}_+$, and then $\tilde{H}\tilde{A} = \tilde{A}\tilde{H}$ for every Hermitian $\tilde{H}$, which says $\tilde{A}$ commutes with the whole Hermitian basis and hence with $\mathbb{B}$: $\tilde{A}$ is central. The centre is computed in the next section to be $\mathbb{C}e_0$, and its Hermitian part is $\mathbb{R}e_0$. The argument for $\mathbb{M}_-$ is identical after $\tilde{A}\to i\tilde{A}$.

The proposition says that the sector decomposition, although it decomposes the algebra as a real vector space, is invisible to the algebra's own left action: there is no element whose module action measures the sectors, and the only elements that preserve a sector by left multiplication are the real scalars.

The sectors are not a splitting of the algebra

A superselection structure in its algebraic form requires the sectors to be the summands of an algebra decomposition, and that fails at the first product. The product of two Hermitian elements is Hermitian only when they commute, $$ \tilde{H}, \tilde{K}\in\mathbb{M}_+ \ \Longrightarrow\ (\tilde{H}\tilde{K})^{*} = \tilde{K}\tilde{H} = \tilde{H}\tilde{K} \iff [\tilde{H},\tilde{K}] = 0 , $$ and the idempotents provide an elementary counterexample. For the two pure-state idempotents $\tilde\Pi_+(\hat{\mu}) = \tfrac12(e_0+i\hat{\mu})$ and $\tilde\Pi_+(\hat{\nu}) = \tfrac12(e_0+i\hat{\nu})$ with $\hat{\mu}\cdot\hat{\nu} = 0$, $$ 4\,\tilde\Pi_+(\hat{\mu})\tilde\Pi_+(\hat{\nu}) = e_0 + i(\hat{\mu}+\hat{\nu}) - (i\hat{\mu})(i\hat{\nu}), $$ whose Hermitian conjugate differs: the commutator is $$ [\tilde\Pi_+(\hat{\mu}),\,\tilde\Pi_+(\hat{\nu})] = -\tfrac12\,\hat{\mu}\times\hat{\nu}, $$ a non-zero element of the anti-Hermitian sector. The product of two informational elements is therefore not informational; it has a material component. Equivalently, $$ \mathbb{M}_+\mathbb{M}_+ \not\subseteq \mathbb{M}_+, \qquad \mathbb{M}_+\mathbb{M}_- \not\subseteq \mathbb{M}_-\cup\mathbb{M}_+ . $$ The sectors are a grading of the vector space and not a splitting of the algebra, and multiplication mixes them. This is the first genuine obstruction, and it is the reason that the standard central-projection route cannot be taken for granted.

The Factor Property: There Is No Central-Projection Superselection

The standard route asks for central projections, so the centre must be computed.

The centre. An element $\tilde{C} = C_0e_0 + C_1e_1 + C_2e_2 + C_3e_3$ is central if and only if $[\tilde{C}, e_k] = 0$ for $k = 1,2,3$. Using $e_je_k = -\delta_{jk}e_0 + \epsilon_{jkl}e_l$ for $j,k\in\{1,2,3\}$, the bracket with $e_1$ gives $$ [\tilde{C}, e_1] = 2C_3e_2 - 2C_2e_3 , $$ and the two companion brackets give $$ [\tilde{C}, e_2] = 2C_1e_3 - 2C_3e_1, \qquad [\tilde{C}, e_3] = 2C_2e_1 - 2C_1e_2 . $$ All three vanish if and only if $C_1 = C_2 = C_3 = 0$, so $$ Z(\mathbb{B}) = \mathbb{C}e_0 = \{C_0e_0 : C_0\in\mathbb{C}\}. $$ The centre is two-real-dimensional and consists of the complex scalars. It splits along the sectors as $$ Z(\mathbb{B}) = \bigl(Z(\mathbb{B})\cap\mathbb{M}_+\bigr)\oplus\bigl(Z(\mathbb{B})\cap\mathbb{M}_-\bigr) = \mathbb{R}e_0 \oplus \mathbb{R}(ie_0), $$ so the centre itself carries the material-informational grading: the real scalar $e_0$ is informational, the imaginary scalar $ie_0$ is material. This is the only place in the algebra where a central element of each sector occurs, and the fact that the sector-carrying central element $ie_0$ is imaginary is the algebraic seed of the phase discussion below.

No central idempotents. A central idempotent is $\tilde{C} = c\,e_0$ with $\tilde{C}^2 = \tilde{C}$, that is $c^2 = c$, whose only complex solutions are $c = 0$ and $c = 1$. Hence $$ \{\text{central idempotents of }\mathbb{B}\} = \{0,\ e_0\}. $$ A non-trivial decomposition $\mathbb{B} = \mathbb{B}_1\oplus\mathbb{B}_2$ by central projections does not exist, and every non-zero element generates the whole algebra as a two-sided ideal, $$ \mathbb{B}\tilde{Q}\mathbb{B} = \mathbb{B} \qquad (\tilde{Q}\neq0), $$ which is the statement that $\mathbb{B}$ is a simple complex algebra. Its centre is exactly $\mathbb{C}$, so $\mathbb{B}$ is central simple: a factor in the algebraic sense.

Why this forbids the standard superselection structure. Two consequences follow, and both are relevant. First, on the algebra side, a decomposition into non-interfering branches labelled by central charges requires central projections, and there are none beyond $0$ and $e_0$. Second, on the module side, the defining module $S = \mathbb{C}^2$ on which $\mathbb{B}$ acts faithfully is the unique simple left module, and it is irreducible: a non-zero spinor generates $S$ under the action of the algebra. A superselection decomposition of the state space would split the module into invariant subspaces, and an irreducible module has none. The state space of the framework is therefore a single superselection sector: the framework has no intrinsic superselection structure in the central-projection sense.

The conclusion is not a defect of the split but a property of the algebra, and it must be stated before anything positive is said. Any superselection structure that the framework is to carry cannot come from its centre; if one is to be found, it must come from a different structure.

What Survives: The Real-Structure Grading

A structure different from the centre is available, and it is the one the split actually uses: the real structure $\flat = -{}^{*}$.

The charge is antilinear, not central. The sector grading is measured by $\flat$, whose $\pm1$ eigenspaces are the two sectors. Unlike a central charge, $\flat$ is not an element of the algebra, so it is not an observable and cannot be added to the algebra's list of quantities; it is a conjugation, and the sector eigenvalue is its fixed-space decomposition. In the standard framework a superselection charge is an observable that commutes with all observables; here the charge is an antilinear map that anticommutes with the complex structure, $$ \flat(i\tilde{Q}) = -i\,\flat(\tilde{Q}), $$ which is exactly what the sector exchange by $i$ requires. The grading is thus carried by the algebra's real structure rather than by its centre, and the generic feature of real-structure superselection rules — that the charge is a conjugation and the forbidden coherence is between an object and its conjugate — is what occurs here.

The conjugation action preserves the sectors. The operations of the framework act on the algebra by the rotor congruence $$ \tilde{Q} \ \longmapsto\ \tilde{\Lambda}\,\tilde{Q}\,\tilde{\Lambda}^{*} , \qquad \tilde{\Lambda}\in\mathbb{B}\ \text{invertible}. $$ For every such $\tilde{\Lambda}$ and every sector label, $$ \bigl(\tilde{\Lambda}\tilde{Q}\tilde{\Lambda}^{*}\bigr)^\dagger = \tilde{\Lambda}\tilde{Q}^{*}\tilde{\Lambda}^{*} = \pm\,\tilde{\Lambda}\tilde{Q}\tilde{\Lambda}^{*} , $$ so $$ \tilde{\Lambda}\,\mathbb{M}_\pm\,\tilde{\Lambda}^{*} \subseteq \mathbb{M}_\pm . $$ The Lorentz transformations on four-vectors, the unitary evolution of states, and the reversible operations of the quantum formalism are all congruences of this type, and none of them moves an element from one sector to the other. This is the precise sense in which the grading is a symmetry of the operations of the theory.

The module action does not preserve the sectors. The contrast with left multiplication is what makes the statement non-trivial. Left multiplication is the action of the algebra on its defining module, and it mixes the sectors: $$ e_1\,e_0 = e_1, \qquad e_1\in\mathbb{M}_-, \quad e_0\in\mathbb{M}_+, $$ so a material element carries an informational element to a material one. The two actions organise the framework differently, and the difference is worth naming: the congruence action — conjugation by invertible elements, under which the relativistic and the quantum transformations both fall — is sector-preserving, while the module action — left multiplication, which builds the algebra out of the module — is sector-mixing. A superselection statement concerns operations, and the operations of this framework are congruences; the module action is the algebraic structure that produces the objects on which the operations act, not an operation in the physical sense.

The two non-trivial central charges. The centre contains one element of each sector, $e_0$ and $ie_0$. The imaginary scalar $ie_0$ is central and lies in $\mathbb{M}_-$; it is the algebra's continuous superselection-type charge. As an operator on the module it is the scalar $i$ on every irreducible representation — in the matrix model it is $iI_2$ — so it acts trivially on every ray. Its expectation is the same in every state, and the one-parameter group it generates is the unobservable global phase. The sector-carrying central charge is thus a charge whose eigenvalue is the same everywhere: the strictest possible superselection charge, in the sense that it provides no label at all, and simultaneously the only non-trivial central charge the algebra has.

The Unobservable Relative Phase

The kinematic form of the superselection statement requires a third check beyond sector preservation: that the relative phase between the sectors cannot be observed. The check succeeds, and the reason is the identity of the algebra's complex structure with the sector exchange.

The relative phase is central. Let a general element be decomposed as $\tilde{Q} = \tilde{Q}_+ + \tilde{Q}_-$. The only way to give the two summands a relative phase in this algebra is to multiply the sum by a central phase, $$ \tilde{Q}(\theta) = e^{i\theta}\bigl(\tilde{Q}_+ + \tilde{Q}_-\bigr) = \bigl(\cos\theta\,\tilde{Q}_+ + \sin\theta\, i\tilde{Q}_-\bigr) + \bigl(\cos\theta\,\tilde{Q}_- + \sin\theta\, i\tilde{Q}_+\bigr), $$ using $e^{i\theta} = \cos\theta\,e_0 + \sin\theta\,(ie_0)$ and $i\tilde{Q}_\pm\in\mathbb{M}_\mp$. The central phase does mix the sectors — it is that mixing — but it is an element of the centre, and on the defining module it acts as the scalar $e^{i\theta}$: $$ \rho_S\bigl(e^{i\theta}\tilde{Q}\bigr)\,|u\rangle = e^{i\theta}\,\rho_S(\tilde{Q})\,|u\rangle . $$ It therefore fixes every ray of the state space. No state, and no expectation value of any observable in any state, distinguishes $\tilde{Q}$ from $e^{i\theta}\tilde{Q}$: the relative phase generated by the sector exchange is a gauge redundancy. The forbidden coherence is unobservable not because the algebra is too small to contain the exchange, but because the exchange is central and the centre acts trivially on rays.

The pairing separates the sectors. The same conclusion can be read off the Born pairing, which is the operational content of the framework. For a state $\tilde{\rho}\in\mathbb{M}_+$ and an element $\tilde{Q} = \tilde{Q}_+ + \tilde{Q}_-$, $$ \mathrm{Tr}\bigl(\tilde{\rho}\tilde{Q}\bigr) = \mathrm{Tr}\bigl(\tilde{\rho}\tilde{Q}_+\bigr) + \mathrm{Tr}\bigl(\tilde{\rho}\tilde{Q}_-\bigr), $$ and the two terms have definite reality types. Because $\tilde{\rho}$ and $\tilde{Q}_+$ are Hermitian, $\tilde{\rho}\tilde{Q}_+$ has a real scalar part, so $\mathrm{Tr}(\tilde{\rho}\tilde{Q}_+)\in\mathbb{R}$. Because $\tilde{\rho}$ is Hermitian and $\tilde{Q}_-$ is anti-Hermitian, $\tilde{\rho}\tilde{Q}_-$ is anti-Hermitian, so its scalar part is purely imaginary and $\mathrm{Tr}(\tilde{\rho}\tilde{Q}_-)\in i\mathbb{R}$. The observable, real part of the pairing therefore involves the informational component alone: $$ \mathrm{Re}\,\mathrm{Tr}\bigl(\tilde{\rho}\tilde{Q}\bigr) = \mathrm{Tr}\bigl(\tilde{\rho}\,\mathrm{Re}_{\mathbb{B}}\,\tilde{Q}\bigr) = \mathrm{Tr}\bigl(\tilde{\rho}\tilde{Q}_+\bigr). $$ The material component contributes only to the imaginary part, which no Hermitian observable reads out. The material component is invisible to every expectation value, exactly as a superselection charge is invisible to every observable that commutes with it. This is a statement about the pairing and not about the physical importance of the material sector: the material sector enters the theory through invariants — the interval $N(\tilde{Q}) = \tilde{Q}\tilde{Q}^{\natural}$, the four-velocity norm, the mass shell — and not through expectation values.

Summary of the test. Against the two conditions of a superselection structure: the operations preserve the sectors (verified by the conjugation identity), and the relative phase is unobservable (verified by the centre acting as a scalar on rays and by the reality type of the pairing). Against the standard algebraic realization: there is no central projection, and the module is irreducible, so the sectors are not the summands of a central decomposition. The split therefore satisfies the operational conditions of a superselection structure while violating its standard realization, and the charge that carries it is the antilinear real structure $\flat$.

The Lie-Algebra Grading: The Cartan Decomposition

The sharpest positive content of the split is not on the algebra but on its Lie algebra, where the material summand becomes a subalgebra and the split becomes a genuine $\mathbb{Z}_2$ grading.

Traceless parts. Let $\mathbb{M}_\pm^{0}$ denote the traceless elements of the two sectors. Since $\mathrm{Tr}(\tilde{Q}) = 2\,\mathrm{Sc}(\tilde{Q})$, the traceless elements are exactly those with vanishing scalar part along $e_0$: $$ \mathbb{M}_+^{0} = \mathrm{span}_\mathbb{R}\{ie_1, ie_2, ie_3\}, \qquad \mathbb{M}_-^{0} = \mathrm{span}_\mathbb{R}\{e_1, e_2, e_3\}. $$ These six elements span the traceless part of $\mathbb{B}$, and the traceless part is the Lie algebra $\mathrm{SL}(2,\mathbb{C})$, which acts as the Lorentz algebra. Hence $$ \mathrm{SL}(2,\mathbb{C}) = \mathbb{M}_+^{0}\oplus\mathbb{M}_-^{0}, $$ with $\mathbb{M}_+^{0}$ Hermitian (traceless) and $\mathbb{M}_-^{0}$ anti-Hermitian (traceless).

The brackets. With $e_je_k = -\delta_{jk}e_0 + \epsilon_{jkl}e_l$ for $j,k\in\{1,2,3\}$ and $j\neq k$, the commutators close on the two summands as $$ [\mathbb{M}_-^{0}, \mathbb{M}_-^{0}]\subseteq\mathbb{M}_-^{0}, \qquad [\mathbb{M}_-^{0}, \mathbb{M}_+^{0}]\subseteq\mathbb{M}_+^{0}, \qquad [\mathbb{M}_+^{0}, \mathbb{M}_+^{0}]\subseteq\mathbb{M}_-^{0}. $$ Explicitly, the rotation generators and the boost generators satisfy $$ [e_j, e_k] = 2\epsilon_{jkl}\,e_l, \qquad [ie_j, ie_k] = -2\epsilon_{jkl}\,e_l, \qquad [e_j, ie_k] = 2\epsilon_{jkl}\,ie_l . $$ The first line says that the material traceless sector is closed under commutation: it is the compact subalgebra $\mathrm{SU}(2)$ of rotations. The second says that the informational traceless sector is a symmetric complement: the bracket of two boosts is a rotation, and the bracket of a rotation with a boost is a boost. The split is therefore a $\mathbb{Z}_2$ grading of the Lorentz Lie algebra, $$ [\text{even},\text{even}]\subseteq\text{even}, \qquad [\text{even},\text{odd}]\subseteq\text{odd}, \qquad [\text{odd},\text{odd}]\subseteq\text{even}, $$ with even part $\mathbb{M}_-^{0}$ and odd part $\mathbb{M}_+^{0}$.

The Cartan involution. The involution that effects this grading is $\flat = -{}^{*}$ restricted to real-linear combinations: it acts as $+1$ on $\mathbb{M}_-^{0}$ and $-1$ on $\mathbb{M}_+^{0}$, and on the Lie algebra it is a linear involution because the combinations are real. It is the Cartan involution of $\mathrm{SL}(2,\mathbb{C})$, its fixed subalgebra is the compact $\mathrm{SU}(2)$ of rotations, and its $(-1)$-eigenspace is the non-compact complement of boosts. The material-informational split restricted to the generators is the Cartan decomposition of the Lorentz algebra, and the compact/non-compact dichotomy — the reason the Lorentz group is non-compact while the rotation group is compact — is the sector dichotomy.

Consequence for the framework. Two conclusions follow, one negative and one positive. Negatively, the subalgebra property holds only on the traceless part; it does not extend to the sectors themselves, because the trace directions $e_0\in\mathbb{M}_+$ and $ie_0\in\mathbb{M}_-$ are central and their products stay in the centre, while the products of the non-central parts leave the sectors. Positively, on the generators the split is exactly a $\mathbb{Z}_2$ grading with a compact even part, and this is the algebraic locus of the interaction between the relativistic and the quantum structures: a boost is an informational generator, a rotation is a material one, and the failure of the boosts to close — the bracket $[\text{boost},\text{boost}] = \text{rotation}$ — is the algebraic origin of the Thomas–Wigner rotation and of the non-commutativity of the boosts. The next three articles of this subcategory are computations of that failure and of its information-theoretic consequences.

The Physical Reading

The material-informational split is a superselection structure of a definite and restricted kind. It is worth stating the reading in one place.

Material and informational data. $\mathbb{M}_-$ carries the four-vectors: the four-position $\tilde{Q} = ict\,e_0+\mathbf{x}$, the four-velocity $\tilde{U} = \gamma(ic\,e_0+\mathbf{v})$, the four-momentum $\tilde{P} = m\tilde{U}$, and their invariants. $\mathbb{M}_+$ carries the states and observables: the idempotents that parametrise the pure states, the Bloch ball of mixed states, and the Hermitian observables paired with them by $\mathrm{Tr}(\tilde{\rho}\tilde{H})$. The two are complementary real forms of the same complex algebra, related by the complex structure $i$ that exchanges them.

What the superselection statement forbids. Because the conjugation action preserves the sectors and the relative phase is central, there is no operation of the framework that converts material data into informational data or the reverse, and there is no observable coherence between the two. A change of frame is a congruence, so it acts on both sectors simultaneously and preserves the grading: it rotates a four-vector in $\mathbb{M}_-$ and a state in $\mathbb{M}_+$ by the same rotor, and it never turns one into the other. This is the structural reason that a Lorentz transformation can rearrange the information carried by a system but cannot create it from kinematic data or destroy it into kinematic data.

What it does not forbid. Because the sectors are not central summands, the framework has no preferred basis in which the material and informational contents are separately stored, and no conserved charge that counts them. The split is a grading of the operations, not a decomposition of the state space; the state space is irreducible, and its single sector is the whole of it. A reader who expects the two sectors to behave like electric-charge sectors — with superselection labels, additive charges, and a decomposition of the Hilbert space — will not find them, and their absence is a property of the algebra rather than a missing ingredient of the physical description. What the two sectors supply is a structural distinction between the kinematic and the informational faces of one algebraic object, sharp enough to organise the transformations of the theory and honest enough to make no claim to a decomposition that the algebra cannot support.

Framing of the relativistic quantum information. With the split fixed, the informational sector is the carrier of the quantum information, and the questions of this subcategory become questions about a single object: how the frame group acts on the carrier. The relativistic qubit is the carrier itself, the defining module with its Lorentz action and its little group; frame-dependent entanglement is the action of a congruence on the informational content of a joint state; the Wigner rotation is the compact, material rotation generated by the non-compact, informational boosts; and spin entropy is the entropic measure of the carrier's mixedness. Each of those articles takes the grading established here as its starting point and asks what the congruence action does to it.

Summary

The decomposition $\mathbb{B} = \mathbb{M}_-\oplus\mathbb{M}_+$ into anti-Hermitian and Hermitian elements is the eigenspace decomposition of the real structure $\flat = -{}^{*}$, and it satisfies the two operational conditions of a superselection structure. The operations of the framework are the congruences $\tilde{Q}\mapsto\tilde{\Lambda}\tilde{Q}\tilde{\Lambda}^{*}$, and every one of them preserves both sectors, since $(\tilde{\Lambda}\tilde{Q}\tilde{\Lambda}^{*})^{*} = \tilde{\Lambda}\tilde{Q}^{*}\tilde{\Lambda}^{*}$. The relative phase between the two components is generated by the central scalar $i$, which exchanges the sectors and acts on every irreducible module as the scalar $e^{i\theta}$, hence fixes every ray; operationally, the informational component supplies the real part of the Born pairing $\mathrm{Tr}(\tilde{\rho}\tilde{Q})$ and the material component only its imaginary part, which no Hermitian observable reads.

The split does not satisfy the standard algebraic realization of a superselection structure. The centre is $Z(\mathbb{B}) = \mathbb{C}e_0$, its only central idempotents are $0$ and $e_0$, and $\mathbb{B}$ is central simple; the defining module is the unique simple module and is irreducible. There is consequently no central-projection decomposition into non-interfering branches, and the framework has no intrinsic superselection sectors of the standard kind. The centre itself is graded by the split, into the informational real scalar $e_0$ and the material imaginary scalar $ie_0$; the latter is the algebra's only non-trivial central charge, and because it acts as a scalar on the module it labels nothing.

The sharpest positive content of the split is on the Lie algebra. The traceless parts furnish the Cartan decomposition $$ \mathrm{SL}(2,\mathbb{C}) = \mathbb{M}_+^{0}\oplus\mathbb{M}_-^{0}, \qquad [\mathbb{M}_\pm^{0},\mathbb{M}_\pm^{0}]\subseteq\mathbb{M}_\mp^{0}, \qquad [\mathbb{M}_+^{0},\mathbb{M}_-^{0}]\subseteq\mathbb{M}_+^{0}, $$ with the material part the compact $\mathrm{SU}(2)$ of rotations and the informational part the non-compact complement of boosts, and with $\flat$ the Cartan involution. The bracket $[\text{boost},\text{boost}] = \text{rotation}$ is the algebraic origin of the Wigner rotation, and it is the point at which the two sectors of the relativistic and quantum structures meet.

Summary of Notation

Symbol	Meaning
$\mathbb{B} = \mathbb{C}\otimes_\mathbb{R}\mathbb{H}$	Biquaternion algebra
$e_0 = 1, e_1, e_2, e_3$	Quaternion basis, $e_k^2 = -e_0$
$i$	Scalar imaginary, $i^2 = -1$, central
${}^{*}$	Hermitian conjugation, antilinear involution
$\flat = -{}^{*}$	Real structure; fixed space $\mathbb{M}_-$
$\mathbb{M}_+$	Hermitian (informational) sector, fixed space of ${}^{*}$
$\mathbb{M}_-$	Anti-Hermitian (material) sector, fixed space of $\flat$
$\mathbb{B} = \mathbb{M}_+\oplus\mathbb{M}_-$	Hermitian decomposition
$P_\pm = \tfrac12(\mathrm{id}\pm{}^{*})$	Real-linear sector projections (not in $\mathbb{B}$)
$Z(\mathbb{B}) = \mathbb{C}e_0$	Centre; $\mathbb{R}e_0$ informational, $\mathbb{R}(ie_0)$ material
$\tilde{H} = h_0e_0 + i\mathbf{h}$	General Hermitian element
$\tilde{Q} = ix_0e_0 + \mathbf{x}$	General anti-Hermitian element
$\mathrm{Tr}(\tilde{Q}) = 2\,\mathrm{Sc}(\tilde{Q})$	Trace
$\tilde{Q}\mapsto\tilde{\Lambda}\tilde{Q}\tilde{\Lambda}^{*}$	Congruence (physical operation)
$\mathbb{M}_\pm^{0}$	Traceless parts, $\mathrm{SL}(2,\mathbb{C}) = \mathbb{M}_+^{0}\oplus\mathbb{M}_-^{0}$
$\mathbb{M}_-^{0}\cong\mathrm{SU}(2)$	Material traceless part, compact rotations
$\mathbb{M}_+^{0}$	Informational traceless part, non-compact boosts