Spinors as Peirce Coordinates over the Reals
Introduction
The off-diagonal Peirce coordinates of a real matrix are maps between the lines of a frame, and at $n=2$ those coordinates are a pair of real numbers that the transpose exchanges the way a spinor is exchanged with its conjugate.
The article defines the Peirce coordinates of an element, reads the off-diagonal part as a family of real maps, treats the case $n=2$ in full, shows that the transpose exchanges the two chiralities and that the trace form pairs them, counts the dimensions at $n=2,3,4$, compares the real case with the complex one, and identifies the pure spinors with the rank-one matrices. It assumes The Peirce Basis and the Corners over $\mathbb{R}$ for the frame and its corners, and it defers the symmetric reading of the coordinates to The Peirce Decomposition and the Two Halves over $\mathbb{R}$.
The Peirce Coordinates of an Element
Every real matrix is the sum of its corner components, and the components are its Peirce coordinates. For a frame $P_{1},\dots,P_{n}$ with subspaces $V_{k}=\operatorname{im}P_{k}$ of ranks $r_{k}$,
$$ X=\sum_{k,l=1}^{n}X_{kl},\qquad X_{kl}=P_{k}XP_{l}\in A_{kl}=\operatorname{Hom}(V_{l},V_{k}). $$
The coordinates are unique, because the corners are independent. The sum of the diagonal coordinates $X_{kk}$ is the part of $X$ that preserves each $V_{k}$, and the sum of the off-diagonal ones is the part that moves between them.
The diagonal coordinates are square matrices and the off-diagonal ones are rectangular. In a basis adapted to the frame, $X_{kk}$ is an $r_{k}\times r_{k}$ array and $X_{kl}$ is an $r_{k}\times r_{l}$ array. The off-diagonal coordinates are the maps $V_{l}\to V_{k}$, and that is what a spinor coordinate is here: a map between two lines of the frame, not a column of an ambient array.
The coordinates multiply by the corner rule.
$$ (XY)_{kl}=\sum_{m=1}^{n}X_{km}Y_{ml}, $$
so the algebra reads as an $n\times n$ array of blocks with the matrix product, as in The Peirce Basis in $n$ Dimensions for the complex case.
The Off-Diagonal Coordinates as Maps
The off-diagonal part of an element is a collection of real maps, one for each ordered pair of distinct lines.
$$ X_{\text{off}}=\sum_{k\ne l}X_{kl},\qquad X_{kl}:V_{l}\longrightarrow V_{k}. $$
For a complete frame each $V_{k}$ is a line, each $X_{kl}$ is a scalar multiple of a chosen unit, and the off-diagonal part is an $n\times n$ array with zero diagonal. That array is the spinor data of the element.
The diagonal part is not spinor data. It carries the endomorphisms of the lines, of dimension $\sum_{k}r_{k}^{2}$, and the spinorial content of the frame is the off-diagonal part, of real dimension
$$ \sum_{k\ne l}r_{k}r_{l}=n^{2}-\sum_{k}r_{k}^{2}. $$
At a complete frame this is $n^{2}-n=n(n-1)$: two real dimensions for each unordered pair of lines.
The diagonal corners form a subalgebra and the off-diagonal ones a bimodule over it. A spinor is an element of that bimodule, and the frame is what makes the bimodule visible.
The Case $n=2$
At $n=2$ the off-diagonal corner is one-dimensional and the off-diagonal part is a pair of real numbers.
$$ E=\begin{pmatrix}1&0\\0&0\end{pmatrix},\qquad X=aE_{12}+bE_{21}=\begin{pmatrix}0&a\\ b&0\end{pmatrix},\qquad (a,b)\in\mathbb{R}^{2}. $$
The pair $(a,b)$ is the spinor of $X$ in the frame: one real number for the map $V_{2}\to V_{1}$ and one for the map $V_{1}\to V_{2}$.
The spinor determines the element and the element determines the spinor. The diagonal part of $X$ is separate data, so the off-diagonal corner $A_{12}\oplus A_{21}$ is isomorphic to $\mathbb{R}^{2}$ as a real vector space, and the two numbers $a$ and $b$ are its coordinates.
The square of the element is the product of the two components.
$$ X^{2}=ab\,\mathrm{I}, $$
so the element is nilpotent exactly when $ab=0$, that is, when one of the two spinor components vanishes. For $a,b$ both nonzero the element has eigenvalues $\pm\sqrt{ab}$, which are real exactly when $ab>0$ and a conjugate pair when $ab<0$. The sign of the product is a real invariant of the spinor.
The transpose swaps the two components.
$$ X^{\mathsf T}=bE_{12}+aE_{21}, $$
so the transpose acts on the spinor by $(a,b)\mapsto(b,a)$. The symmetric elements are those with $a=b$ and the antisymmetric ones are those with $a=-b$; the symmetric part of $X$ is $\tfrac12(a+b)(E_{12}+E_{21})$ and the antisymmetric part is $\tfrac12(a-b)(E_{12}-E_{21})$. The two halves of the spinor space are the diagonal and the antidiagonal of the pair.
The pairing between the two chiralities is the trace form. With $X=aE_{12}+bE_{21}$ and $Y=a'E_{12}+b'E_{21}$,
$$ \langle X,Y\rangle=\operatorname{tr}(XY)=ab'+ba', $$
because $\operatorname{tr}(E_{12}E_{21})=1$ and $\operatorname{tr}(E_{12}E_{12})=\operatorname{tr}(E_{21}E_{21})=0$. The form is symmetric, nondegenerate, and its matrix in the basis $E_{12},E_{21}$ is $\begin{pmatrix}0&1\\ 1&0\end{pmatrix}$, of inertia $(1,1)$. It is the perfect pairing between the two one-dimensional spaces $A_{12}$ and $A_{21}$.
The pairing is invariant and indefinite. The space $A_{12}\oplus A_{21}$ has the invariant form $\operatorname{tr}(XY)$, of inertia $(1,1)$, and the positive definite form $\operatorname{tr}(X^{\mathsf T}Y)$ is not invariant. So a real spinor pair comes with a symmetric indefinite pairing.
The Two Chiralities
The two off-diagonal corners are the two chiralities of the frame. Write
$$
S^{+}=\bigoplus_{k An element of $S^{+}$ raises the index, an element of $S^{-}$ lowers it, and the off-diagonal part of every element splits as $X_{\text{off}}=X^{+}+X^{-}$ with $X^{+}\in S^{+}$ and $X^{-}\in S^{-}$. The transpose exchanges the two chiralities. For a symmetric frame, $A_{kl}^{\mathsf T}=A_{lk}$, so $$
(S^{+})^{\mathsf T}=S^{-},\qquad (S^{-})^{\mathsf T}=S^{+}.
$$ The transpose is $\mathbb{R}$-linear, so the exchange is an isomorphism of real vector spaces. The fixed points of the exchange are the symmetric off-diagonal elements and the anti-fixed points are the antisymmetric ones. The trace form pairs $S^{+}$ with $S^{-}$ and vanishes on each chirality separately. For $X\in A_{kl}$ and $Y\in A_{pq}$ with $k\ne l$ and $p\ne q$, the trace $\operatorname{tr}(XY)$ is nonzero only when $l=p$ and $q=k$, that is, only when $Y$ lies in the corner $A_{lk}$ opposite to $X$. Hence $$
\langle S^{+},S^{+}\rangle=0,\qquad \langle S^{-},S^{-}\rangle=0,\qquad \langle S^{+},S^{-}\rangle\ \text{nondegenerate}.
$$ This is the real form of the statement that the two spinor spaces are dual to each other. There is one real spinor space, not two. The two chiralities are subspaces of the one real algebra $M_{n}(\mathbb{R})$, and the transpose moves one to the other. Nothing is added by naming them separately: a real spinor is an element of the off-diagonal part, and its two chiral components are read off from the corners. The off-diagonal part has dimension $n(n-1)$ at a complete frame. At $n=2$ each chirality is a line, at $n=3$ each is a plane, and at $n=4$ each has six dimensions. The two chiralities always have equal dimension, $\tfrac12 n(n-1)$, and the two together with the $n$ diagonal coordinates rebuild the algebra: $$
n+\frac{n(n-1)}{2}+\frac{n(n-1)}{2}=n^{2}.
$$ At a coarse frame the counts change with the ranks. The chirality $S^{+}$ has dimension $\sum_{k Each entry is the dimension of a real space. Over $\mathbb{C}$ the corner $A_{kl}$ has complex dimension $r_{k}r_{l}$, the same number, so the passage to $\mathbb{C}$ doubles the real dimension of every corner and leaves the table unchanged. Over $\mathbb{C}$ the two chiralities are exchanged by the conjugate transpose. For a Hermitian frame, with $P_{k}^{*}=P_{k}$, the conjugate transpose sends $A_{kl}$ to $A_{lk}$, and it is conjugate linear. The two spinor spaces are then complex conjugate spaces, and the pairing between them is Hermitian, not symmetric. Over $\mathbb{R}$ the exchanging map is linear and the pairing is symmetric. The transpose sends $A_{kl}$ to $A_{lk}$ and satisfies $\operatorname{tr}(XY)=\operatorname{tr}(YX)$ with no conjugation. The bilinear form on $A_{12}\oplus A_{21}$ at $n=2$ is $\operatorname{tr}(XY)=ab'+ba'$, with inertia $(1,1)$, and its matrix is symmetric with integer entries. The difference is the difference between a real and a Hermitian form. A Hermitian form is conjugate linear in one argument and positive definite for the standard frame; the real form $\operatorname{tr}(XY)$ is bilinear and indefinite. So the real spinor of this article is a pair of real components with an indefinite pairing, and the complex spinor of Spinors as Peirce Coordinates in $n$ Dimensions is a pair of conjugate components with a Hermitian pairing. A rank-one matrix has a spinor whose two chiral components factor through one vector on each side. Let $X=uv^{\mathsf T}$ with $u,v\in\mathbb{R}^{n}$ nonzero. Then $$
X_{kl}=P_{k}uv^{\mathsf T}P_{l}=(P_{k}u)(P_{l}^{\mathsf T}v)^{\mathsf T},
$$ so each corner component of $X$ is the rank-one map $(P_{l}^{\mathsf T}v)\mapsto P_{k}u$ followed by a scalar, and the whole off-diagonal part is determined by the two vectors $P_{k}u$ and $P_{l}^{\mathsf T}v$ together with the frame. Conversely, a spinor whose components factor through one vector on each side is rank one. If $X^{+}$ is built from a single vector $u$ and $X^{-}$ from a single vector $v$ in the sense of the display, then $X=uv^{\mathsf T}$ has rank one. So the pure spinors are the rank-one matrices, read in the frame. The rank of a random rank-one matrix is one. Ten random real $3\times 3$ matrices of the form $uv^{\mathsf T}$, with $u$ and $v$ with independent Gaussian entries, all have rank exactly $1$, as computed with numpy. A rank-one matrix is nilpotent exactly when $v^{\mathsf T}u=0$, which is the spinor condition at $n=2$ written in general. At $n=2$ the pure spinors are the pairs with exactly one nonzero component. With $X=aE_{12}+bE_{21}$ the rank is $2$ when $ab\ne0$ and $1$ when $ab=0$; the trace is zero in both cases. So the pure spinors of the frame are the two coordinate axes of $\mathbb{R}^{2}$, and the general spinor is a sum of two pure ones. This is the real form of the statement that a spinor is a sum of decomposable pieces. For a real frame with subspaces $V_{k}$ of ranks $r_{k}$, every matrix splits into Peirce coordinates $X=\sum_{k,l}X_{kl}$ with $X_{kl}:V_{l}\to V_{k}$ of dimension $r_{k}r_{l}$. The off-diagonal coordinates are the spinor data: at a complete frame they form two blocks of dimension $n(n-1)/2$ each, one raising the index and one lowering it, and at $n=2$ they are a single pair $(a,b)\in\mathbb{R}^{2}$ with $X=aE_{12}+bE_{21}$ and $X^{2}=ab\,\mathrm{I}$. The transpose exchanges the two blocks, so the symmetric and antisymmetric parts of the pair are its diagonal and antidiagonal; the trace form pairs the blocks perfectly, with $\operatorname{tr}(E_{12}E_{21})=1$ and matrix $\begin{pmatrix}0&1\\ 1&0\end{pmatrix}$ of inertia $(1,1)$ on the pair at $n=2$. The off-diagonal dimension is $2$ at $n=2$, $6$ at $n=3$ and $12$ at $n=4$, and the diagonal part of dimension $n$ completes the algebra. Over $\mathbb{C}$ with a Hermitian frame the two blocks are exchanged by a conjugate-linear map, while over $\mathbb{R}$ they are exchanged by the linear transpose and paired by a symmetric indefinite form. The pure spinors are the rank-one matrices, and a random rank-one real matrix has rank one.The Dimension Counts
$n$
$S^{+}$
$S^{-}$
off-diagonal total
$\dim M_{n}(\mathbb{R})$
$2$
$1$
$1$
$2$
$4$
$3$
$3$
$3$
$6$
$9$
$4$
$6$
$6$
$12$
$16$
The Complex Case
The Pure Spinors Are the Rank-One Elements
Summary
Summary of Notation
Symbol
Meaning
$P_{k}$, $V_{k}=\operatorname{im}P_{k}$
The frame and the line of its $k$-th idempotent
$r_{k}=\operatorname{rk}P_{k}$
The rank, with $\sum_{k}r_{k}=n$
$X_{kl}=P_{k}XP_{l}$
The Peirce coordinate in the corner $A_{kl}$
$S^{+}=\bigoplus_{k The raising chirality
$S^{-}=\bigoplus_{k The lowering chirality
$(a,b)$
The spinor of $aE_{12}+bE_{21}$ at $n=2$
$\langle X,Y\rangle=\operatorname{tr}(XY)$
The pairing between the chiralities
$uv^{\mathsf T}$
A pure spinor, a rank-one matrix
Further Reading