The Spin Carriers of the Quaternion Algebra
Introduction
A carrier, in the sense of this corpus, is a minimal left ideal of an algebra, the space on which the algebra acts by left multiplication and which no proper submodule of the algebra cuts. In the biquaternion block the carriers exist and they are the two spinor spaces $S_+(\vec u)$ and $S_-(\vec u)$ of a root of $-e_t$, the images of the two idempotents $\tfrac12(1\pm\vec u)$. In the quaternion algebra there is no such space at all, and the reason is the division property: every non-zero left ideal is the whole algebra.
The article proves that absence, shows what the algebra does have — the regular module, which is its own carrier and has no proper submodule — and then shows the carrier the matrix model lends it, the columns of $M_2(\mathbb{C})$, with the spinor as a column and the quaternionic structure that replaces the reality condition. It ends with the two carriers of the complexified block, which the quaternionic structure exchanges, and with the readings of Weyl and of Majorana, which are the two ways of asking for a real subspace of a spinor space and which the quaternionic structure grants only to their sum.
The modules over the algebra are The Category of Modules over the Quaternion Algebra; the idempotents are The Peirce Decomposition of the Quaternion Algebra and Its Collapse; the matrix model is The M2C-H Standard Representation.
A Carrier Is a Minimal Left Ideal
Definition. A left ideal of $\mathbb{H}$ is a real subspace $S$ with $\mathbb{H}S\subseteq S$; it is minimal if it is non-zero and contains no non-zero left ideal but itself. A carrier is a minimal left ideal, and the algebra carries a spinor space if it has one.
Proposition (the module language). The left ideals are the submodules of the regular module ${}_{\mathbb{H}}\mathbb{H}$, and a carrier is a simple submodule of it; the left multiplication by an idempotent $\tilde e$ has for its image the left ideal $\mathbb{H}\tilde e$, and every left ideal of a finite-dimensional algebra that is generated by an idempotent arises in this way.
Proof. The definitions coincide, since a submodule of the regular module is a subspace stable under the left action. The image of $L_{\tilde e}$ is $\mathbb{H}\tilde e$ and the restriction of $L_{\tilde e}$ to it is the identity, so it is a left ideal, and it is the ideal generated by $\tilde e$. Verified on the coefficient basis.
Remark (what a carrier is for). A carrier is the algebraic object behind a spinor: the algebra acts on it, the action is irreducible, and the vectors of the carrier are the spinors. The existence of the carrier is therefore exactly the existence of a non-trivial idempotent, and the Peirce calculus of the algebra is the calculus of the corners of that idempotent. No idempotent, no carrier; no carrier, no spinor of the algebra.
Why the Algebra Has None
Theorem (the left ideals). The only left ideals of $\mathbb{H}$ are $0$ and $\mathbb{H}$.
Proof. Let $S$ be a non-zero left ideal and let $\tilde x\in S$ be non-zero. The element is invertible, by the division property, so $e_t=\tilde x^{-1}\tilde x\in S$ and then $\mathbb{H}=\mathbb{H}e_t\subseteq S$. Verified on the coefficient basis.
Corollary (no carrier). $\mathbb{H}$ has no carrier: the only simple submodule of the regular module is the module itself, and it has no proper simple submodule.
Proof. A minimal left ideal is a non-zero left ideal that contains no smaller one; by the theorem the only non-zero left ideal is $\mathbb{H}$, and it contains the non-zero left ideal $\mathbb{H}$, so it is minimal only in the vacuous sense of having no proper part. The class of the algebra is the class of a division algebra, whose regular module is simple. Verified by the exhaustion of the ideals.
Remark (the carrier is the whole algebra). The reader who insists that a carrier is a left ideal that is minimal among the non-zero ones will say that $\mathbb{H}$ carries the space $\mathbb{H}$ itself, and he will be right in a sense that costs the whole construction: the spinor is then the element, the algebra acts on it faithfully and irreducibly, and the "spinor space" is four-dimensional and is the algebra. What is lost is the properness, and with the properness the whole of the Peirce calculus: there is no corner, no off-diagonal space and no pair of conjugate carriers.
The Chirality Projector Is Invertible
Proposition (the projector of the complexified block is, over $\mathbb{R}$, an automorphism). Let $\tilde u$ be a unit vector and $\tilde e_+=\tfrac12(e_t+\vec u)$ the chirality projector of the biquaternion block. Over $\mathbb{R}$ the element $\tilde e_+$ is invertible, of norm $\tfrac12$, and the left multiplication $L_{\tilde e_+}$ is an isomorphism of $\mathbb{H}$.
Proof. $N(\tilde e_+)=\tfrac14(1+1)=\tfrac12\neq0$, so $\tilde e_+$ is invertible and its left multiplication is invertible with inverse $L_{\tilde e_+^{-1}}$. So the map that cuts a chirality in the complexified block is, over $\mathbb{R}$, an automorphism of the module, with no kernel and no cokernel. Verified numerically: the inverse is $1-\vec u$, of norm two.
Remark (what the failure means). The chirality projector fails over $\mathbb{R}$ for the same reason the idempotent fails: an idempotent is a projector, and an invertible element is a projector only when it is the identity. The two notions coincide only after complexification, where $\tfrac12(1+\vec u)$ satisfies $\tilde e_+^2=\tilde e_+$ because $\vec u^2=-e_t$ is combined with the complex scalar; over $\mathbb{R}$ the same formula is invertible.
The Carrier the Matrix Model Lends
Theorem (the column carrier). In the standard model $\mathbb{H}\subset M_2(\mathbb{C})$, the column space $\mathbb{C}^2$ is a minimal left ideal of $M_2(\mathbb{C})$, hence a carrier of the complexified algebra, and it is a simple left $\mathbb{H}$-module of real dimension four; as an $\mathbb{H}$-module it is isomorphic to $\mathbb{H}$ itself.
Proof. The columns of $M_2(\mathbb{C})$ are a minimal left ideal, since the matrices of rank one with a fixed column span it and the algebra acts transitively on the non-zero vectors of $\mathbb{C}^2$ with the matrices of determinant one. The $\mathbb{H}$-action is the restriction of the matrix action; the module is simple because it is the regular module of the algebra identified with its image in the matrices, and its real dimension is four. Verified by sending the basis of $\mathbb{H}$ to the four matrices and acting on the two columns.
Remark (the price of the carrier). The carrier of $\mathbb{H}$ is not a part of $\mathbb{H}$: it is a part of the matrix algebra, and the algebra reaches it only through the model. The carrier exists because the complexification enlarges the algebra to the point where it has proper ideals, and the enlargement is exactly the operation the quaternion algebra forbids.
The Spinor as a Column
Definition. A spinor of the algebra is a column $$ \psi=\begin{pmatrix}\psi_1\\ \psi_2\end{pmatrix}\in\mathbb{C}^2, $$ with the action of $\tilde q=z_0+z_1\vec e_x+z_2\vec e_y+z_3\vec e_z$ given by the left multiplication of its matrix.
Proposition (the two coordinates). The two entries of the column are the coordinates of the spinor in the two real vectors $e_1,e_2$ of the model, the action of the algebra is complex-linear and faithful, and it is irreducible.
Proof. The matrix of $\tilde q$ acts on the two columns with the two entries of each row, which are the coefficients of the two real vectors; the action is injective because the matrices of the basis of $\mathbb{H}$ are independent, and it is irreducible because every non-zero column is cyclic, the matrices of determinant one acting transitively on the non-zero vectors. Verified on the four basis vectors.
Remark (the dimension count). The column has complex dimension two and real dimension four, which is the dimension of the algebra: a spinor and a quaternion have the same size, and this is the numerical form of the statement that the module is the regular module.
The Reality Condition
Theorem (the quaternionic structure). The column space carries the anti-linear map $$ j(\psi)=J\bar\psi,\qquad J=\begin{pmatrix}0&1\\-1&0\end{pmatrix}, $$ which satisfies $j^2=-\mathrm{id}$ and commutes with the action of $\mathbb{H}$: $j(\tilde q\psi)=\tilde q\,j(\psi)$.
Proof. The matrix $J$ is real, so $j^2(\psi)=J\overline{J\bar\psi}=J^2\psi=-\psi$; and for the matrix $X$ of a quaternion the identity $J\bar X=XJ$ holds on the four basis matrices, so $j(X\psi)=J\bar X\bar\psi=XJ\bar\psi=Xj(\psi)$. Verified on the basis.
Corollary (there is no real spinor). The fixed points of $j$ are zero: an anti-linear map of square $-1$ has no non-zero fixed vector, since $j\psi=\psi$ gives $-\psi=j^2\psi=\psi$. So the spinor of the algebra is quaternionic and has no real part in the sense of a fixed subspace.
Proof. The computation of the square and the definition of a fixed point. Verified on the basis.
Remark (the reality condition of the two algebras). The reality condition that cuts $\mathbb{H}$ inside the biquaternion block is an anti-linear involution of square $+1$, whose fixed set is the algebra; on the module the corresponding structure has square $-1$, so it cuts nothing. The real form of the algebra is a real structure on the coefficients, and the real structure it induces on the spinors is symplectic: the spinor space of the algebra is a quaternionic line, not a real plane.
The Two Carriers of the Biquaternion Block
Theorem (the pair of carriers). Let $\tilde e_\pm=\tfrac12(e_t\pm\vec u)$ be the two idempotents of the complexified block attached to a root $\vec u$ of $-e_t$. Then $S_\pm(\vec u)=B\tilde e_\pm$ are minimal left ideals of $B$, of complex dimension two, and $B=S_+(\vec u)\oplus S_-(\vec u)$.
Proof. The idempotents satisfy $\tilde e_\pm^2=\tilde e_\pm$, $\tilde e_++\tilde e_-=e_t$ and $\tilde e_+\tilde e_-=0$, so the images are left ideals summing to $B$ and intersecting in zero; each is minimal because $B$ is the full matrix algebra of $2\times2$ matrices over $\mathbb{C}$ and the images are the two coordinate lines. Verified on the matrix model.
Theorem (which one survives). The quaternionic structure exchanges the two carriers, $j(S_+(\vec u))=S_-(\vec u)$ and $j(S_-(\vec u))=S_+(\vec u)$, and neither is stable under the action of $\mathbb{H}$; the sum $S_+(\vec u)\oplus S_-(\vec u)=B$ is stable, and it contains the regular module $\mathbb{H}$ as the fixed set of the reality condition.
Proof. In the model with $\vec u=\vec e_z$ the two carriers are the two coordinate lines of $\mathbb{C}^2$, and $j$ acts on the first basis vector by $j(e_1)=Je_1=e_2$, so it exchanges them; for a general root the pair is the image of the standard pair under a conjugation by a versor, which commutes with $j$, so the exchange is preserved. A subspace stable under both $\mathbb{H}$ and $j$ is a quaternionic subspace, and a complex line is not one, so no single carrier descends. Verified on the standard pair and on the pair of a rotated root.
Remark (the verdict of the descent). Two carriers over $\mathbb{C}$, none over $\mathbb{R}$, and the sum in place of the pair. The descent from the biquaternion block to the quaternion algebra merges the two spinor spaces into one quaternionic line, and the merging is exactly the identification of the two chiralities.
The Carrier of the Clifford Algebra
Remark (the algebra is its own spinor space). The Clifford algebra of the three-dimensional Euclidean space is the direct sum of two copies of the quaternion algebra, and its minimal left ideals are the two summands, each of which is $\mathbb{H}$ itself. So the spinor space of that Clifford algebra is a copy of the algebra, and the absence of a proper carrier in $\mathbb{H}$ is the same statement read on the Clifford side: the spinor is the quaternion. This is the algebraic form of the identification of the versors with the spin group of the three-dimensional space, and it is the reason the quaternion algebra is the spinor algebra of the lowest dimension in which a quaternionic spinor appears at all.
Remark (what the complexification changes). Over $\mathbb{C}$ the Clifford algebra becomes a full matrix algebra, its minimal left ideals become the columns, and the carrier appears: the columns are the two-dimensional complex space of the model. The whole difference between the two cases is the difference between a division algebra and a full matrix algebra, which is the difference between one simple module and a space of proper carriers.
A Worked Check
Example (the projectors and their norms). Take $\vec u=\vec e_z$, so that $\tilde e_\pm=\tfrac12(e_t\pm\vec e_z)$. Then $N(\tilde e_+)=\tfrac14(1+1)=\tfrac12$, the inverse is $1-\tilde e_z$ of norm two, and the product $L_{\tilde e_+}L_{1-\vec e_z}$ is the identity; the same holds for $\tilde e_-$. Both are invertible, both are non-idempotent over $\mathbb{R}$, and neither has a kernel. Verified numerically on the basis.
Example (the exchange of the two carriers). In the model the two idempotents become the two coordinate projectors, so $S_+(\vec e_z)$ is the first coordinate line and $S_-(\vec e_z)$ the second. The map $j$ sends the first basis vector to the second, $j(e_1)=Je_1=e_2$, and the second to the negative of the first, so it exchanges the two carriers and acts on the pair as the matrix $J$. The two lines are therefore conjugate under the quaternionic structure, and the quaternionic line they span is the module. Verified on the standard pair and on the pair attached to a rotated root.
The Weyl Reading
Remark (the reading). A Weyl spinor is an element of one of the two carriers, that is a spinor of definite chirality; the chirality projectors are the two idempotents $\tilde e_\pm$. Over $\mathbb{R}$ the projectors are invertible, so the left multiplications that should cut the two chiralities are automorphisms of the module: they do not cut anything, and the two "chiralities" of the algebra are the same module seen twice. There is no Weyl spinor of the quaternion algebra, and the reason is that the chirality projector is invertible in the algebra.
The Majorana Reading
Remark (the reading). A Majorana spinor is a spinor fixed by an anti-linear involution, that is a real spinor inside a complex one. The module of the algebra carries the anti-linear map $j$, but its square is $-1$, so it has no fixed spinor; the fixed spinors appear only for an anti-linear map of square $+1$, which the quaternionic line does not carry. So the quaternionic column is the analogue of a pair of conjugate Weyl spinors rather than of a Majorana spinor: the conjugation pairs a spinor with another, and the pair is what the real form sees.
Remark (the two readings compared). The Weyl reading asks for a proper direct summand and fails because the algebra has no proper left ideal; the Majorana reading asks for a real form of the carrier and fails because the quaternionic structure is symplectic. Both readings ask the module to break into something smaller than a quaternionic line, and the division property is precisely the statement that it does not.
The Verdict on the Carriers
Remark (one carrier, and it is the algebra). The algebra has one simple module, of real dimension four, and that module is the algebra; the carriers of the complexified block are two lines inside it, exchanged by the quaternionic structure. So the count of the carriers is one over $\mathbb{R}$ and two over $\mathbb{C}$, and the two are not the halves of the one: they are two complex lines whose sum is the one quaternionic line.
Remark (what the absence buys). The absence of a proper carrier is not a defect of the algebra; it is the division property read on the modules, and it is what makes the algebra the spinor algebra of the three-dimensional space rather than of the four-dimensional one. The complexified block needs the two carriers because its multiplication degenerates on the null cone; the quaternion algebra has no null cone of zero divisors, and it needs none. The number of the carriers is the number of the simple modules of the complexification, and the descent from two to one is the descent from the light cone to the sphere.
Summary
A carrier is a minimal left ideal, hence a simple submodule of the regular module, and it exists exactly when the algebra has a non-trivial idempotent.
$\mathbb{H}$ has none: its only left ideals are $0$ and the algebra, the regular module is simple, and the only carrier is the algebra itself, which has no proper part.
The chirality projector $\tfrac12(e_t+\vec u)$ of the complexified block is invertible over $\mathbb{R}$, of norm $\tfrac12$, so it is an automorphism of the module and cuts nothing.
The matrix model lends the algebra the carrier $\mathbb{C}^2$, a minimal left ideal of $M_2(\mathbb{C})$ and a simple module of real dimension four, which is the regular module; the spinor is a column and the quaternionic structure $j(\psi)=J\bar\psi$ satisfies $j^2=-1$ and commutes with the action up to the conjugation.
The two carriers $S_\pm(\vec u)$ of the biquaternion block are the two coordinate lines; the quaternionic structure exchanges them, no single one descends, and their sum is the quaternionic line that survives.
The Weyl reading fails because the chirality projector is invertible, and the Majorana reading fails because the anti-linear structure of the module is symplectic; the quaternionic column is the analogue of a conjugate pair of Weyl spinors.
Summary of Notation
| Symbol | Meaning |
|---|---|
| $S$ | a left ideal, and a carrier when minimal |
| $\tilde e_\pm=\tfrac12(e_t\pm\vec u)$ | the chirality projectors of the complexified block |
| $S_\pm(\vec u)$ | the two carriers $B\tilde e_\pm$ |
| $\mathbb{C}^2$ | the column carrier of the matrix model |
| $j(\psi)=J\bar\psi$ | the quaternionic structure, $j^2=-1$ |
| $B=\mathbb{C}\otimes\mathbb{H}$ | the biquaternion algebra |
Further Reading
- The Category of Modules over the Quaternion Algebra for the module structure that forbids the carrier.
- The Peirce Decomposition of the Quaternion Algebra and Its Collapse for the idempotents and their absence.
- The M2C-H Standard Representation for the matrix model and the columns.
- Spinors and Dirac Spinors in the Biquaternion Algebra for the carriers of the complexified block and the Dirac reading.