Spinors as Peirce Coordinates

Introduction

A spinor is not a column of a matrix, and it is not every irreducible Peirce carrier. It is a coordinate vector on a subspace that one idempotent singles out inside an algebra, when that subspace carries the highest-weight-$\tfrac12$ weight of the frame root. The idempotent fixes the subspace, its companion corner supplies the extra coordinate directions, and the diagonal corner supplies the coefficients. This article states that definition in a general finite-dimensional algebra, proves when the subspace is irreducible, identifies the weight condition that makes it a spinor space, and separates the algebraic object from the pictures that a matrix model provides. It closes by fixing the two names built on the condition: the elementary spinor, the two-dimensional object itself, and the Dirac spinor, the pair of elementary spinors that the chirality operator of a declared frame separates.

The subspace is the Peirce carrier

$$ C(e)=Ae=eAe\oplus fAe, \qquad f=1_A-e, $$

of an idempotent $e$. The diagonal corner $D=eAe$ is the algebra of coefficients; the companion corner $fAe$ is square-zero and its elements are the companion nilpotents. If $(\nu_1,\dots,\nu_s)$ generates $fAe$ over $D$, every element of $C(e)$ is written

$$ \zeta=d_0+\nu_1d_1+\dots+\nu_sd_s, \qquad d_i\in D, $$

and the spinor is the list $(d_0;d_1,\dots,d_s)$. The number of entries is $1+s$.

Four results organise the article. First, a spinor is an element of a Peirce carrier that also satisfies a weight condition: the half-root acts with weights $\tfrac12$ and $-\tfrac12$, each once. Second, semisimplicity of the algebra is exactly what makes the carrier irreducible: in a semisimple algebra a primitive idempotent gives an irreducible carrier, while the upper triangular algebra $T_2$ has a primitive idempotent with a reducible carrier. Third, the weight condition is restrictive: in $M_N(k)$ with the standard frame it holds only for $N=2$, so the column of $M_N$ is a spinor space only in the biquaternion case; for $N>2$ the column is the spin-$(N-1)/2$ carrier of the Peirce chain. Fourth, the column of a matrix model is one picture of the carrier, not its definition, and the number of coordinates $1+s$ is not the spin.

Setting. Throughout, $k$ is a field, $A$ is a finite-dimensional unital associative $k$-algebra with $1_A\neq0$, and $e\in A$ is an idempotent with $f=1_A-e$. The corners and the companion count are those of Peirce Coordinates and Companion Nilpotents: $A_{st}=e_sAe_t$, $D=A_{++}=eAe$, and $s$ is the least number of generators of $A_{-+}=fAe$ over $D$. The primitive idempotents are those of Primitive Idempotents and Minimal Left Ideals; the module-theoretic comparison, used only for the theorem below, is Modules over an Algebra and Simple and Semisimple Modules. No form, no involution, no matrix model and no module theory is used in the definition.

The Carrier

Definition. The Peirce carrier of $e$ is $C(e)=Ae=\{\,x\in A:xe=x\,\}$. Its diagonal part is $D=eAe$ and its companion part is $fAe$.

Proposition. $C(e)=D\oplus fAe$, and left multiplication by any $a\in A$ maps $C(e)$ into itself.

Proof. This is Peirce Coordinates and Companion Nilpotents, §The Carrier and Its Coordinates: the Peirce expansion of $x\in Ae$ has only the corners with right index $+$, and $(ax)e=a(xe)=ax$.

Definition. A companion frame of $e$ is a family $(\nu_1,\dots,\nu_s)$ of elements of $fAe$ that generates $fAe$ over $D$ with $s$ minimal. The coordinate vector of $\zeta\in C(e)$ is the list $(d_0;d_1,\dots,d_s)$ of the expansion above; it is unique when the companion family is a $D$-basis.

Proposition. If $D$ is a division ring, then $fAe$ is free over $D$, the companion frame is a basis, and $C(e)\cong D^{1+s}$.

Proof. Every module over a division ring is free. The expansion is unique by the direct sum $D\oplus fAe$.

Example (the biquaternion). For $A=\mathbb{B}$, $e=\tilde\Pi_+$, $f=\tilde\Pi_-$, one has $D=\mathbb{C}\tilde\Pi_+\cong\mathbb{C}$, $s=1$, and

$$ C(\tilde\Pi_+)=\mathbb{C}\tilde\Pi_+\oplus\mathbb{C}\vec\Upsilon_y\cong\mathbb{C}^2 . $$

The two coordinates are the coefficient of the idempotent and the coefficient of the companion nilpotent. The other idempotent gives $C(\tilde\Pi_-)=\mathbb{C}\vec\Upsilon_x\oplus\mathbb{C}\tilde\Pi_-$, a second carrier with the same structure. The two are different subspaces of the algebra and the same two-dimensional complex space up to isomorphism.

The Definition of a Spinor

Definition. Let $e$ be an idempotent with $h=2e-1_A$, and let $C(e)=Ae$ be its Peirce carrier. A spinor of the frame $e$ is an element of $C(e)$ when the half-root $\tfrac12h$ acts on $C(e)$ by left multiplication with the weight string

$$ \tfrac12,\ -\tfrac12, $$

each weight occurring once over $D=eAe$; that is, $C(e)=L_+\oplus L_-$ with $hx=\pm x$ for $x\in L_\pm$. A spinor frame is the pair of $e$ and a companion frame $(\nu_1,\dots,\nu_s)$; in that frame the spinor is its coordinate vector $(d_0;d_1,\dots,d_s)\in D^{1+s}$.

When the weight condition fails, $C(e)$ is a Peirce carrier and its coordinate vector is a Peirce coordinate, not a spinor. The definition has four ingredients: an idempotent, the subspace it fixes on the right, the companion corner that completes it, and the weight condition on the half-root. The last condition supplies the highest-weight-$\tfrac12$ structure; irreducibility alone does not.

Theorem (the matrix case). Let $A=M_N(k)$ and $e=E_{11}$. Then $C(e)$ is a spinor space exactly when $N=2$.

Proof. The carrier is the first column, $C(e)=\operatorname{span}\{E_{11},\dots,E_{N1}\}$, and $h=2E_{11}-I_N$ acts by $hE_{11}=E_{11}$ and $hE_{k1}=-E_{k1}$ for $k\ge2$. The half-root therefore has the weight $+\tfrac12$ once and $-\tfrac12$ with multiplicity $N-1$. The weight string is $(+\tfrac12,-\tfrac12)$ exactly when $N=2$.

Corollary. In a semisimple algebra with a primitive idempotent of a simple component $M_N(D)$, the carrier is a spinor space exactly when $N=2$. The biquaternion is the case $N=2$, $D=\mathbb{C}$.

Remark (why the whole carrier, not one corner). The diagonal corner $D$ alone is not the spinor space. In $M_2(k)$ with $e=E_{11}$, $D=kE_{11}$ is a single coefficient, and the spinor also needs the companion direction $E_{21}$; without it there is no second coordinate. The companion corner alone is not the spinor space either: its elements are nilpotents, and a nilpotent is a direction, not a state. The spinor is an element of their sum, with one diagonal coordinate and $s$ companion coordinates.

Remark (coordinates over $D$, not over $k$). The entries of the coordinate vector are elements of $D=eAe$, not scalars. When $D$ is a division ring, this is a vector of $1+s$ entries over a division ring, and $1+s$ is its dimension over $D$. When $D$ is a matrix algebra, an entry is itself a matrix; the coordinate vector is a block vector, and $1+s$ counts blocks, not scalars.

The Elementary Spinor and the Dirac Spinor

The condition of the previous section isolates one object. Two names are built on it, and the second is the first taken twice.

Definition (elementary spinor). An elementary spinor is an element of a complex two-dimensional space on which the left multiplications act irreducibly with weights $+\tfrac12$ and $-\tfrac12$, so that a rotation through $\theta$ about $\vec u$ multiplies it by $e^{-i\theta\vec u\cdot\vec S}$.

Definition (Dirac spinor). A Dirac spinor is an element $\tilde\Psi=\tilde\Psi_L+\tilde\Psi_R$ of $\mathbb{B}$, the two parts being its two components in the decomposition of $\mathbb{B}$ into minimal left ideals,

$$ \mathbb{B}=\mathbb{B}\tilde\Pi_+\oplus\mathbb{B}\tilde\Pi_-,\qquad \tilde\Pi_\pm=\tfrac12(e_t\pm i\vec u), $$

for a declared unit imaginary direction $\vec u$, and each part is an elementary spinor of complex dimension two.

Proposition. The two parts are closed under left multiplication, and each is an elementary spinor.

Proof. For $x\in\mathbb{B}\tilde\Pi_+$ and $a\in\mathbb{B}$, the product $ax=(a)(x\tilde\Pi_+)=(ax)\tilde\Pi_+$ lies in $\mathbb{B}\tilde\Pi_+$, because $\tilde\Pi_+^2=\tilde\Pi_+$; the same for $\tilde\Pi_-$. So the summands are left-invariant and the decomposition is a decomposition into left submodules. Each summand is $\mathbb{C}^2$ over $\mathbb{C}$, and on it the half-root $\tfrac12iL_{\vec u}$ has the weights $+\tfrac12$ and $-\tfrac12$, each once, so each is a carrier of the elementary type. The two summands are the two columns of the algebra, $\mathbb{C}E_{11}\oplus\mathbb{C}E_{21}$ and $\mathbb{C}E_{12}\oplus\mathbb{C}E_{22}$ for $\vec u=\vec e_z$.

Remark (the multiplicity, and why the pair is not a repetition). As a left module $\mathbb{B}\cong\mathbb{C}^2\otimes\mathbb{C}^2$, and left multiplication acts on the first factor alone. The first factor is the spin; the second, of dimension two, is a multiplicity. The idempotent multiplies on the right, so it acts on the second factor only, and its two summands are the two lines of that factor tensored with the whole spin space. The Dirac spinor is therefore a carrier of spin $\tfrac12$ with multiplicity two, and the pair of elementary spinors is the two lines of the multiplicity, not two copies of one state. This is what the definition buys and what a definition by "a pair of spinors" alone would miss.

Remark (the declared direction). The direction $\vec u$ inside $\tilde\Pi_\pm$ is the only datum of the second definition, and it is declared, not derived. A change of $\vec u$ changes the decomposition; it changes neither the dimension, nor the spin, nor the number of parts. The definition of the elementary spinor names no direction at all, so the two definitions differ in exactly one respect: the second names the direction that selects its two parts.

Remark (what the two definitions do not use). No Clifford relation, no Pauli array, no gamma matrix, no gamma basis operator, no volume element and no second rotation triple. The two parts are the two summands of a decomposition into left ideals, and the summands are isomorphic copies of one module: the index that separates them is a copy index, which is also the multiplicity of the remark above, and it is not a chirality (Operator Chirality on the Biquaternion Algebra, §What Operator Chirality Is Not). A triple cannot separate them, because every left multiplier commutes with every right multiplier and therefore acts on the two summands in the same way. What the operator chirality of that article names is the side on which an operator multiplies, read from the operator and not from this decomposition; the two values of a basis operator are its two eigenspaces on a carrier, and those are not these two copies.

Remark ($N=4$). The same two definitions cover the four-dimensional case. There the Peirce carrier is of dimension four and of spin $\tfrac32$ under the chain, while the volume element splits it $2+2$ into two half-spin modules of spin $\tfrac12$ (Spinors as Coordinates in $M_4(\mathbb{C})$, §The Clifford reading). The elementary spinor is the half-spin module of dimension two; the Dirac spinor is the module that splits into the two halves.

Irreducibility

Proposition. If $C(e)$ is irreducible under left multiplication, then $e$ is primitive and $D=eAe$ is a division ring.

Proof. If $e=e_1+e_2$ with $e_1,e_2$ nonzero orthogonal idempotents, then $Ae=Ae_1\oplus Ae_2$ is a nontrivial decomposition into left-invariant subspaces, against irreducibility. An $A$-endomorphism $\varphi$ of $Ae$ satisfies $\varphi(x)=\varphi(xe)=x\varphi(e)$, with $\varphi(e)\in Ae$ and $e\varphi(e)=\varphi(e)e=\varphi(e)$, so $\varphi$ is the right multiplication by an element of $D=eAe$; conversely every such right multiplication is $A$-linear. Hence $\operatorname{End}_A(C(e))\cong D_{\mathrm{opp}}$, and the endomorphism ring of an irreducible module is a division ring by Schur's lemma. So $D$ is a division ring.

Theorem. Let $A$ be semisimple and $e$ primitive. Then $C(e)$ is irreducible under left multiplication.

Proof. In a semisimple algebra the primitive idempotent $e$ has $D=eAe$ a division ring and $Ae$ a simple left module; this is the equivalence proved in Primitive Idempotents and Minimal Left Ideals, §The equivalence, and the simplicity is exactly the absence of a proper nonzero left-invariant subspace.

Example (the theorem fails without semisimplicity). Let $A=T_2(k)$ be the upper triangular algebra and $e=E_{22}$. Then $e$ is primitive and $D=eAe=kE_{22}$ is a division ring, but

$$ C(e)=kE_{12}\oplus kE_{22}, $$

and $kE_{12}$ is a proper left-invariant subspace: $E_{11}E_{12}=E_{12}$, $E_{22}E_{12}=0$ and $E_{12}^2=0$. So $C(e)$ is reducible. Primitiveness and a division corner do not suffice; semisimplicity is the missing hypothesis.

Example (the dual numbers). In $A=k[\varepsilon]$, the identity $e=1_A$ is primitive, $D=A$ is not a division ring, and $C(e)=A$ contains the invariant subspace $k\varepsilon$. The carrier is reducible. The companion corner is zero, so the reducibility is entirely in the diagonal part.

Corollary. In a semisimple algebra, the idempotents whose carriers are irreducible are exactly the primitive idempotents. The spinor carriers are the further subset selected by the weight condition: the primitive idempotents of a simple component $M_N(D)$ with $N=2$.

What the Definition Is Not

It is not a corner. The space $A_{++}$ is the coefficient algebra, and a general element of it is a matrix of labels. The spinor uses $A_{++}$ for its first coordinate and $A_{-+}$ for the rest; neither corner alone is the carrier.

It is not a column. A column appears when the algebra is written as $M_N(k)$ and the carrier is read on the first column of the matrix. That picture requires a basis of $A$ and the identification $A\cong M_N(k)$; it is not available in a general algebra. The column is the image of $C(e)$ under one model, not the object.

It is not defined by minimality. The carrier is defined by the idempotent before any minimality is mentioned. The closure of $C(e)$ under left multiplication is a consequence of the product rule, and its minimality is a theorem about semisimple algebras, proved in Primitive Idempotents and Minimal Left Ideals. The order is: idempotent, carrier, coordinates, weight. Irreducibility is a theorem about the carrier; it is not the spinor condition, and the irreducible carrier of $M_3$ with $e=E_{11}$ is the counterexample.

The number of coordinates is not a weight. A carrier of dimension $N$ over $k$ is not a weight-$N$ object. A weight needs a grading element and an action on the carrier; the idempotent alone supplies only the $\mathbb{Z}_2$ grading, not a weight string. A weight string of the form $\lambda,\lambda-2,\dots,-\lambda$ needs an $\mathfrak{sl}_2$ triple, and the triple is extra data.

Examples

algebra idempotent $D$ $s$ carrier irreducible? spinor?
$\mathbb{B}=M_2(\mathbb{C})$ $\tilde\Pi_+$ $\mathbb{C}$ $1$ $\mathbb{C}^2$ yes yes
$M_2(k)$ $E_{11}$ $k$ $1$ $k^2$ yes yes
$M_3(k)$ $E_{11}$ $k$ $2$ $k^3$ yes no
$M_N(k)$ $E_{11}$ $k$ $N-1$ $k^N$ yes no, $N>2$
$M_N(k)$ $E_{11}+E_{22}$, $N\ge3$ $M_2(k)$ $\lceil N/2\rceil-1$ two-column block no no
$\mathbb{H}$ $1_A$ $\mathbb{H}$ $0$ $\mathbb{H}$ yes no
$T_2(k)$ $E_{11}$ $k$ $0$ $k$ yes no
$T_2(k)$ $E_{22}$ $k$ $1$ $k^2$ no no
$k[\varepsilon]$ $1_A$ $k[\varepsilon]$ $0$ $k[\varepsilon]$ no no

The matrix algebra with a primitive idempotent is the model case: $C(e)$ is the first column, of dimension $N$ over $k$, and the companion frame is $E_{21},\dots,E_{N1}$. The biquaternion is the only matrix case in which the column is a spinor space: $N=2$, the column has two entries, and the half-root has weights $\pm\tfrac12$. For $N>2$ the column is the carrier of the spin-$(N-1)/2$ Peirce chain, so its $N$ entries are Peirce coordinates, not spinor coordinates. The quaternions are the case $s=0$: the carrier is the algebra itself, of dimension one over $\mathbb{H}$, and the half-root has the single weight $+\tfrac12$, so the carrier is not a spinor space.

The Clifford case. For the complex Clifford algebra the structure depends on the parity of $N$. For $N=2m$ even,

$$ \mathrm{Cl}_{2m}(\mathbb{C})\cong M_{2^m}(\mathbb{C}), $$

with one simple component and a primitive idempotent of $D\cong\mathbb{C}$, $s=2^m-1$, and carrier dimension $2^m$. For $N=2m+1$ odd,

$$ \mathrm{Cl}_{2m+1}(\mathbb{C})\cong M_{2^m}(\mathbb{C})\oplus M_{2^m}(\mathbb{C}), $$

with two simple components, the two chiralities; a primitive idempotent of either component has $D\cong\mathbb{C}$, $s=2^m-1$, and carrier dimension $2^m$. In both cases the number of companion nilpotents and the number of spinor coordinates are the same datum:

$N$ Clifford algebra $m$ $s$ spinor dimension
$2$ $M_2(\mathbb{C})$ $1$ $1$ $2$
$3$ $M_2(\mathbb{C})\oplus M_2(\mathbb{C})$ $1$ $1$ $2$
$4$ $M_4(\mathbb{C})$ $2$ $3$ $4$
$5$ $M_4(\mathbb{C})\oplus M_4(\mathbb{C})$ $2$ $3$ $4$
$6$ $M_8(\mathbb{C})$ $3$ $7$ $8$
$8$ $M_{16}(\mathbb{C})$ $4$ $15$ $16$

The Clifford reading. In this table "spinor dimension" is the dimension of the Clifford module carried by the Peirce ideal, not a claim that the half-root alone supplies the weights $\pm\tfrac12$. The Clifford spin structure is the action of the bivectors. At $N=2$ the Peirce root and the Clifford vector agree, and the two-dimensional ideal is the highest-weight-$\tfrac12$ module. At $N=4$ the four-dimensional ideal is the Dirac module; under the volume element it splits into two two-dimensional half-spin modules of spin $\tfrac12$. The Peirce carrier is the module; the bivector action is what makes it a spinor representation.

Summary

A spinor is an element of the Peirce carrier $C(e)=Ae=eAe\oplus fAe$ of an idempotent $e$ when the half-root $\tfrac12h$, with $h=2e-1_A$, acts on $C(e)$ by left multiplication with the weights $+\tfrac12$ and $-\tfrac12$, each once over $D$. In coordinates over $D=eAe$ it is the vector $(d_0;d_1,\dots,d_s)$ with one diagonal entry and one entry for each of the $s$ companion directions. The definition uses an idempotent, a subspace, a square-zero corner and the weight condition; it does not use a matrix model or a column.

Two names are built on it. An elementary spinor is an element of a complex two-dimensional carrier on which the left multiplications act irreducibly with weights $+\tfrac12$ and $-\tfrac12$, so that a rotation through $\theta$ about $\vec u$ multiplies it by $e^{-i\theta\vec u\cdot\vec S}$. A Dirac spinor is an element $\tilde\Psi=\tilde\Psi_L+\tilde\Psi_R$ of $\mathbb{B}$ whose two parts are its two components in the decomposition into minimal left ideals, with $\tilde\Pi_\pm=\tfrac12(e_t\pm i\vec u)$ for a declared direction $\vec u$, and each part is an elementary spinor of complex dimension two. The two parts are the two lines of the multiplicity in $\mathbb{B}\cong\mathbb{C}^2\otimes\mathbb{C}^2$; the idempotent multiplies on the right, so no triple of left multipliers can tell the two parts apart, and the side of the multiplication is what does. The elementary spinor names no direction; the Dirac spinor names the direction that selects its two parts.

In a semisimple algebra the primitive idempotents are exactly the idempotents with irreducible carriers, and $D$ is a division ring, so the carrier is $D^{1+s}$. Without semisimplicity the equivalence fails: the upper triangular algebra has a primitive idempotent with a reducible carrier. Irreducibility is not the spinor condition; the weight string is. In $M_N(k)$ with a primitive idempotent, $s=N-1$ and the carrier has $N$ coordinates. The weight condition holds only for $N=2$, so the biquaternion is the only matrix case in which those coordinates are spinor coordinates; for $N>2$ the carrier is the spin-$(N-1)/2$ module and the coordinates are Peirce coordinates.

Summary of Notation

Symbol Meaning
$k$ A field
$A$ A finite-dimensional unital associative $k$-algebra
$e$ An idempotent, $e^2=e$
$f$ The complement $1_A-e$
$h=2e-1_A$ The frame root
$D=eAe$ The diagonal corner, the coefficient algebra
$fAe$ The companion corner, square-zero
$s$ The least number of generators of $fAe$ over $D$
$C(e)=Ae$ The Peirce carrier
$(\nu_1,\dots,\nu_s)$ A companion frame
$(d_0;d_1,\dots,d_s)$ The coordinates of a spinor
$\tfrac12h$ with weights $\pm\tfrac12$ The condition that makes $C(e)$ a spinor space
elementary spinor an element of a two-dimensional carrier satisfying that condition
$\tilde\Pi_\pm=\tfrac12(e_t\pm i\vec u)$ the idempotents of the declared direction $\vec u$, the two parts of a Dirac spinor
Dirac spinor an element $\tilde\Psi=\tilde\Psi_L+\tilde\Psi_R$ of $\mathbb{B}$, its two parts the two minimal left ideals

Further Reading

  • Richard S. Pierce, Associative Algebras (Springer, 1982), for the Peirce decomposition and the corner algebras.
  • Yurij A. Drozd and Vladimir V. Kirichenko, Finite Dimensional Algebras (Springer, 1994), for primitive idempotents and the local corners.
  • Frank W. Anderson and Kent R. Fuller, Rings and Categories of Modules (Springer, 2nd ed. 1992), for the module-theoretic reading of the carrier.
  • The corpus article Peirce Coordinates and Companion Nilpotents, for the coordinate theory.