The Spin-Half Bispinors and N Norm Regularity in $M_2(\mathbb{C})$

Introduction

The question. The spin-half carrier is the plane of multiples of a rank-one projection and its closures are the minimal left ideals. What is the determinant on these spaces, and is any element of a single ideal N-regular?

Why it is the same statement. This is the in-depth matrix reading of The Spin-Half Bispinors and N Norm Regularity. There the norm is $Q_t^2+Q_x^2+Q_y^2+Q_z^2$ and the null plane is $\mathbb{C}\tilde\Pi_+$; here the norm is $\det A$, the plane is $\mathbb{C}P$ and the ideal is $M_2P$. The statements are the same statements, and only the writing is new.

What the article proves. In this order. That $\det P=0$ and $\det(Y\!P)=0$ for every $Y$, so the plane and both ideals are null. That the ideals are maximal isotropic, of complex dimension two, half of four. That on the bispinor the determinant is the pairing, $\det(X_++X_-)=2B(X_+,X_-)$. That the pairing is nondegenerate and that the polar form is $\sum_\mu x_\mu y_\mu$ in the coefficient basis. The comparison with the other two carriers.

What the article does not do. It does not treat the discriminant on these spaces, which is The Spin-Half Bispinors and Spectral Regularity in $M_2(\mathbb{C})$; it does not treat the four sectors in full, which is The Spin-Half Bispinors and the Sectors of N and Spectral Regularity in $M_2(\mathbb{C})$.

Conventions. $\mathsf M_2(\tilde Q)=Q_t\mathrm I_2-i(Q_x\sigma_1+Q_y\sigma_2+Q_z\sigma_3)$, so $Q_t=\tfrac12\operatorname{tr}A$ and $N(\tilde Q)=\det A$. The coefficient basis is $(\mathrm I_2,E_x,E_y,E_z)$ with $E_k=-i\sigma_k$ and coefficient tuple $x=(x_0,x_1,x_2,x_3)$. The polar form is $B(X,Y)=\tfrac12(\det(X+Y)-\det X-\det Y)$. The Hermitian form is $\langle X,Y\rangle=\tfrac12\operatorname{tr}(X^{\dagger}Y)$, a different object.

Boundaries. The spaces are Introduction to the Spin-Half Bispinors Spaces in $M_2(\mathbb{C})$. The Dirac pairing is Spinors and Dirac Spinors in the Biquaternion Algebra in $M_2(\mathbb{C})$. The norm in general is N-Regularity and N-Degeneracy in the Biquaternion Algebra. The $n=2$ spine is The One-Sided Operators at $n=2$ and The Two-Sided Operators and the Segre Cone at $n=2$, and the display is $M_2(\mathbb{C})$ Pauli Representation : Operators.

Letters. $P$ a rank-one projection, $P_{\perp}=\mathrm I_2-P$, $S_\pm=M_2P^{(\perp)}$, $x$ a coefficient tuple, $B$ the polar form, $\langle\cdot,\cdot\rangle$ the Hermitian form.

The Norm of the Projection

The formula. In coefficients $P=\tfrac12(\mathrm I_2+n_1E_1+n_2E_2+n_3E_3)$ because $\sigma_k=iE_k$, so $$ \det P=\frac14\left(1-\sum_k n_k^2\right)=0, $$ the cancellation being $|n|=1$. Verified exactly.

The two squares. The scalar coordinate contributes $\tfrac14$ and the vector coordinates contribute $-\tfrac14|n|^2$. The cancellation is the unit length of the axis written as an identity.

An idempotent of determinant zero. From $P^2=P$, multiplicativity gives $(\det P)^2=\det P$, so the determinant of an idempotent is $0$ or $1$. The rank-one projections are the case $0$.

The Plane and the Ideals Are Null

The plane. $\det(\lambda P)=\lambda^2\det P=0$ for every $\lambda$: the whole plane lies in the singular locus.

The ideal. For every $Y$, $$ \det(YP)=\det Y\det P=0 . $$ So both ideals $M_2P$ and $M_2P_{\perp}$ lie in the singular locus, and no element of a single ideal is invertible.

Verified. On a hundred random elements of the plane and of the ideal the determinant vanishes to $8\cdot10^{-16}$.

The closure does not repair the nullity. The ideal repairs the invariance of the plane but stays inside the singular locus; the two notions are independent.

The Isotropic Dimension

The definition. A subspace is isotropic when the determinant vanishes on it and totally isotropic when the polar form vanishes on it in both arguments.

The ideals are totally isotropic. $B(XP,YP)=\tfrac12(\det((X+Y)P)-0-0)=0$, verified for all pairs in the basis. Each ideal is isotropic of complex dimension two.

The maximality. For a nondegenerate quadratic form on a space of dimension $N$ an isotropic subspace has dimension at most $N/2$. Here $N=4$, so the bound is $2$, and the ideals attain it: they are maximal isotropic.

The plane is isotropic and not maximal. It has complex dimension one, and it is totally isotropic as well, since $B(\lambda P,\mu P)=\lambda\mu\det P=0$. The dimension, and not the total isotropy, separates it from the ideals.

The Bispinor Read

The pairing. For $X=X_++X_-$ with $X_\pm=XP^{(\perp)}$, $$ \det X=2B(X_+,X_-), $$ because the two pure terms vanish. Verified on a hundred random pairs to $10^{-15}$.

The criterion. $X$ is invertible exactly when its two parts pair non-trivially.

The nondegeneracy. In the coefficient basis the Gram matrix of $B$ is the identity, verified: $B$ has rank four and no radical. So the pairing $S_+\times S_-\to\mathbb{C}$ is nondegenerate, and the two ideals are dual to each other under $B$.

The coefficient formula. $B(X,Y)=\sum_{\mu=0}^{3}x_\mu y_\mu$ and $\det X=\sum_\mu x_\mu^2$, both verified on a hundred random pairs to $2.4\cdot10^{-15}$ and $4.5\cdot10^{-15}$. These are the source's $N=Q_t^2+Q_x^2+Q_y^2+Q_z^2$ and its polar form, in the matrix coefficients.

The form is complex-bilinear, not Hermitian. $B(P,P)=0$ while $\langle P,P\rangle=\tfrac12$: the two forms have the same Gram matrix on the basis and differ off the real span, because $B(iX,iY)=-B(X,Y)$ and $\langle iX,iY\rangle=\langle X,Y\rangle$.

The two ideals are dual. The determinant restricted to the bispinor is the duality pairing of $S_+$ with $S_-$; in a basis of one and its dual basis of the other the pairing is the identity matrix, up to the factor two. Neither ideal carries a norm of its own.

The Pairing and the Axis

The pairing moves with the axis. Both ideals move, so the pairing does; every statement holds at every axis, and the axis is a parameter.

The pairing is symmetric. $B(X_+,X_-)=B(X_-,X_+)$, so the order of the two parts does not matter.

The Three Carriers Compared

carrier $\det$ on it singular locus in it
centre $\mathbb{C}\mathrm I_2$ $\lambda^2$ $\{0\}$
plane and ideals $0$ the whole space
vector part $\mathfrak{su}(2)$ $\lvert v\rvert^2$ $\{0\}$

Where the carrier sits. The plane away from the origin lies in $\Sigma_{01}$: singular and spectrally regular. The centre lies in $\Sigma_{10}$ and the vector part in $\Sigma_{11}$.

Worked Example

The axis $e_3$. $P=\operatorname{diag}(1,0)$, $P_{\perp}=\operatorname{diag}(0,1)$, and $S_+$ is the set of matrices with second column zero.

An element of the ideal. $X=\begin{pmatrix}2&0\\3&0\end{pmatrix}$: $\det=0$, rank one, non-zero, singular.

A bispinor. $X=\begin{pmatrix}2&0\\3&1\end{pmatrix}=X_++X_-$ with $X_+=\begin{pmatrix}2&0\\3&0\end{pmatrix}$ and $X_-=\begin{pmatrix}0&0\\0&1\end{pmatrix}$; $\det=2$ and $2B(X_+,X_-)=2$, verified.

Where $n=2$ Differs

The isotropic dimension. At $n=2$ the ideals are maximal isotropic of complex dimension two, half of four. At order $n$ the maximal isotropic subspaces of the determinant on $M_n(\mathbb{C})$ have dimension $\lfloor n^2/2\rfloor$, and the minimal left ideals have dimension $n$, which is smaller for $n\ge3$.

The pairing. At $n=2$ the bispinor is a sum of two ideals and the determinant is their pairing. At order $n$ the two Peirce halves are blocks of sizes $1$ and $n-1$ for a rank-one projection, and the determinant of their sum is the product of the two block determinants, so the pairing is replaced by a product of a scalar and an $(n-1)$-minor.

What does not differ. The nullity of the ideal, $\det(Y\!P)=0$, holds at every order, because the rank of $YP$ is at most one.

Summary

The determinant of a rank-one projection vanishes, and multiplicativity gives $\det(Y\!P)=0$ for every $Y$: the plane and both minimal left ideals are null, and no element of a single ideal is invertible. The ideals are totally isotropic of complex dimension two, which is maximal for a quadratic form on four dimensions.

On the bispinor the determinant becomes the pairing, $\det(X_++X_-)=2B(X_+,X_-)$, and the pairing is nondegenerate: in the coefficient basis $B$ has the identity Gram matrix, and $B(X,Y)=\sum x_\mu y_\mu$. The Hermitian form $\langle X,Y\rangle=\tfrac12\operatorname{tr}(X^{\dagger}Y)$ is a different form, complex-linear in both arguments versus conjugate-linear in one.

Summary of Notation

Symbol Meaning
$P,\ P_{\perp}$ a rank-one projection and its complement
$S_\pm=M_2P^{(\perp)}$ the two minimal left ideals
$x=(x_0,x_1,x_2,x_3)$ the coefficient tuple in $(\mathrm I_2,E_x,E_y,E_z)$
$B(X,Y)$ the polar form of the determinant
$\langle X,Y\rangle=\tfrac12\operatorname{tr}(X^{\dagger}Y)$ the Hermitian form

Further Reading