Worked Examples of an Eigenbasis of a Biquaternion
Introduction
This article computes the eigenvalue decomposition, the eigenbasis, the nilpotents and the spinors of eight biquaternions, one after another. The eight elements are those of Worked Examples of the Complete Polar Decomposition of an N-Regular Biquaternion, so that the spectral data and the polar data can be read side by side. The decomposition and the eigenprojections are those of The Biquaternion Eigenvalue Decomposition, §The Eigenprojectors; the existence of the eigenbasis is proved in The Spectrum of a Biquaternion; its construction and the nilpotents are those of The Eigenbasis of a Spectrally Regular Biquaternion, §The Corners and the Nilpotents and §The Eigenbasis; no definition or proof is repeated.
Each case is read from the four complex coordinates $Q_t,Q_x,Q_y,Q_z$ of $\tilde Q=Q_te_t+Q_x\vec e_x+Q_y\vec e_y+Q_z\vec e_z$. Write
$$ B^2=Q_x^2+Q_y^2+Q_z^2, \qquad B=\bigl(Q_x^2+Q_y^2+Q_z^2\bigr)^{1/2}, $$
and fix one square root $B$. The eigenvalues, the eigenprojections, the nilpotents and the matrices are computed from them by the formulas of §The Method. The eigenbasis is checked by the four identities
$$ \tilde Q\tilde\Pi_+=\lambda_+\tilde\Pi_+, \qquad \tilde Q\vec\Upsilon_-=\lambda_+\vec\Upsilon_-, \qquad \tilde Q\vec\Upsilon_+=\lambda_-\vec\Upsilon_+, \qquad \tilde Q\tilde\Pi_-=\lambda_-\tilde\Pi_-. $$
The eight cases cover five spectrally regular elements with $B\neq0$, two spectrally degenerate elements with $B=0$ and $\vec Q\neq0$, and the scalar $e_t$. In the tables, distinct names the spectrally regular regime, and Jordan and scalar name the two spectrally degenerate regimes. In the eigenbasis a general element is $\tilde Q_2=Q_{2,11}\tilde\Pi_++Q_{2,12}\vec\Upsilon_-+Q_{2,21}\vec\Upsilon_++Q_{2,22}\tilde\Pi_-$, with the matrix $\begin{pmatrix}Q_{2,11}&Q_{2,12}\\Q_{2,21}&Q_{2,22}\end{pmatrix}$; its two columns are the two spinors.
| Regime | Case | Spectral reading |
|---|---|---|
| Distinct, vector part along $\vec e_x$ | $\vec e_x$ | $B=1$, $\lambda_\pm=\pm i$ |
| Distinct, same direction | $\frac{1}{\sqrt2}(e_t+\vec e_x)$ | $B=\frac{1}{\sqrt2}$, $\lambda_\pm=\frac{1\pm i}{\sqrt2}$ |
| Distinct, real eigenvalues | $\sqrt2\,e_t+i\vec e_x$ | $B=i$, $\lambda_+=\sqrt2-1$, $\lambda_-=\sqrt2+1$ |
| Distinct, same direction | $\bigl(1-\frac{i}{\sqrt2}\bigr)e_t+\bigl(1+\frac{i}{\sqrt2}\bigr)\vec e_x$ | $B=1+\frac{i}{\sqrt2}$, $\lambda_+=a(1+i)$, $\lambda_-=b(1-i)$ |
| Distinct, generic direction | $\bigl(1-\frac{i}{2}\bigr)e_t+\frac{1+i}{\sqrt2}\vec e_x+\frac{1}{\sqrt2}\vec e_y+\frac{i}{2}\vec e_z$ | $B=\sqrt{\frac14+i}$, $\lambda_\pm=1-\frac{i}{2}\pm iB$ |
| Scalar | $e_t$ | $B=0$, $\lambda_\pm=1$ |
| Jordan | $e_t+\vec e_y+\frac{i}{\sqrt2}\vec e_x+\frac{i}{\sqrt2}\vec e_z$ | $B=0$, $\lambda_\pm=1$ |
| Jordan | $(1+i)e_t+\vec e_x+i\vec e_y$ | $B=0$, $\lambda_\pm=1+i$ |
Here $a=1-\frac{1}{\sqrt2}$ and $b=1+\frac{1}{\sqrt2}$.
The plan. The method is stated once (§The Method). The eight cases follow (§The Eight Cases), each with its coordinates, its spectrum, its eigenprojections, its nilpotents when they exist, its eigenbasis and its matrix. The summary collects the results in three tables and reads the patterns across the cases (§Summary).
The Method
The Coordinates and the Spectrum
For $\tilde Q=Q_te_t+Q_x\vec e_x+Q_y\vec e_y+Q_z\vec e_z$ the four complex coordinates are $Q_t,Q_x,Q_y,Q_z$, and the vector part is $\vec Q=Q_x\vec e_x+Q_y\vec e_y+Q_z\vec e_z$ with square
$$ \vec Q^2=-(Q_x^2+Q_y^2+Q_z^2)e_t=-B^2e_t. $$
The two eigenvalues are
$$ \lambda_+=Q_t+iB, \qquad \lambda_-=Q_t-iB. $$
Their sum and product are
$$ \lambda_++\lambda_-=2Q_t, \qquad \lambda_+\lambda_-=Q_t^2+B^2=\langle\tilde Q\rangle_{\natural}. $$
The eigenvalues are distinct exactly when $B\neq0$, because
$$ \lambda_+-\lambda_-=2iB. $$
The characteristic polynomial is
$$ p_{\tilde Q}(X)=(X-Q_t)^2+B^2=(X-\lambda_+)(X-\lambda_-), $$
with discriminant $-4B^2$. The element is invertible exactly when $Q_t^2+B^2\neq0$.
The Eigenprojections
Put
$$ \vec U=\frac{\vec Q}{B}, \qquad \vec{Ζ}=-i\vec U=-\frac{i\vec Q}{B}. $$
Then $\vec{Ζ}^2=e_t$, and the two eigenprojections are
$$ \tilde\Pi_+=\frac12(e_t+\vec{Ζ}), \qquad \tilde\Pi_-=\frac12(e_t-\vec{Ζ}). $$
They are idempotent, orthogonal and complete:
$$ \tilde\Pi_+^2=\tilde\Pi_+, \qquad \tilde\Pi_-^2=\tilde\Pi_-, \qquad \tilde\Pi_+\tilde\Pi_-=\tilde\Pi_-\tilde\Pi_+=0, \qquad \tilde\Pi_++\tilde\Pi_-=e_t. $$
For every $\tilde Q$ with $B\neq0$,
$$ \tilde Q\tilde\Pi_+=\lambda_+\tilde\Pi_+, \qquad \tilde Q\tilde\Pi_-=\lambda_-\tilde\Pi_-, $$
and the eigenvalue decomposition is
$$ \tilde Q=\lambda_+\tilde\Pi_++\lambda_-\tilde\Pi_-. $$
The swap $B\mapsto-B$ exchanges $\tilde\Pi_+$ and $\tilde\Pi_-$ and the two eigenvalues. The decomposition is unique up to this exchange.
Check. For $\tilde Q=(1+i)e_t+\vec e_x+i\vec e_y+\vec e_z$, $B=1$, $\vec U=\vec e_x+i\vec e_y+\vec e_z$ and $\vec{Ζ}=-i\vec e_x+\vec e_y-i\vec e_z$, so $\tilde\Pi_\pm=\frac12(e_t\pm\vec{Ζ})$ and $$ (1+i)e_t+\vec e_x+i\vec e_y+\vec e_z=(1+2i)\tilde\Pi_++\tilde\Pi_-. $$ The scalar-vector form confirms it: $\frac{\lambda_++\lambda_-}{2}e_t+\frac{\lambda_+-\lambda_-}{2}\vec{Ζ}=(1+i)e_t+i\vec{Ζ}=\tilde Q$.
The Peirce Corners and the Nilpotents
The Peirce corners relative to $\tilde\Pi_+$ are
$$ \mathbb{B}_{++}=\tilde\Pi_+\mathbb{B}\tilde\Pi_+, \qquad \mathbb{B}_{+-}=\tilde\Pi_+\mathbb{B}\tilde\Pi_-, \qquad \mathbb{B}_{-+}=\tilde\Pi_-\mathbb{B}\tilde\Pi_+, \qquad \mathbb{B}_{--}=\tilde\Pi_-\mathbb{B}\tilde\Pi_-. $$
The two diagonal corners are the complex lines of the idempotents:
$$ \mathbb{B}_{++}=\mathbb{C}\tilde\Pi_+, \qquad \mathbb{B}_{--}=\mathbb{C}\tilde\Pi_-. $$
The two mixed corners are one-dimensional over $\mathbb{C}$. The convention is:
$$ \vec\Upsilon_+\in\mathbb{B}_{-+}=\tilde\Pi_-\mathbb{B}\tilde\Pi_+, \qquad \vec\Upsilon_-\in\mathbb{B}_{+-}=\tilde\Pi_+\mathbb{B}\tilde\Pi_-. $$
The indexing is by the left ideal, not by the corner.
Construction. Choose a pure element $\tilde Q_1$ with
$$ (\tilde Q_1,\tilde Q_1)+\frac{(\tilde Q_1,\vec Q)^2}{B^2}\neq0, \qquad (\tilde Q_1,\vec Q)=\operatorname{Sc}(\tilde Q_1\vec Q). $$
Put
$$ u=\tilde\Pi_-\tilde Q_1\tilde\Pi_+, \qquad v=\tilde\Pi_+\tilde Q_1\tilde\Pi_-. $$
Then
$$ uv=c\,\tilde\Pi_-, \qquad vu=c\,\tilde\Pi_+, \qquad c=(\tilde Q_1,\tilde Q_1)+\frac{(\tilde Q_1,\vec Q)^2}{B^2}, $$
and
$$ u^2=v^2=0. $$
Choose a square root $\alpha$ of $c$ and set
$$ \vec\Upsilon_+=\frac{u}{\alpha}, \qquad \vec\Upsilon_-=\frac{v}{\alpha}. $$
Then
$$ \vec\Upsilon_+\vec\Upsilon_-=\tilde\Pi_-, \qquad \vec\Upsilon_-\vec\Upsilon_+=\tilde\Pi_+, \qquad \vec\Upsilon_+^2=\vec\Upsilon_-^2=0. $$
The normalisation fixes the products, not the individual generators: it leaves the reciprocal rescaling
$$ \vec\Upsilon_+\longmapsto\mu\vec\Upsilon_+, \qquad \vec\Upsilon_-\longmapsto\mu^{-1}\vec\Upsilon_-, \qquad \mu\in\mathbb{C}^{\times}. $$
The Eigenbasis of a Spectrally Regular Biquaternion fixes one representative uniformly, by the symmetric closed form and its axis chart. The examples with $B\neq0$ below choose the auxiliary vector $\tilde Q_1$ case by case, so their representatives differ from that one by the rescaling above. The idempotents, the spectrum and the diagonal form do not depend on the choice.
The shared solution for $\vec Q=Q_x\vec e_x$. For cases 2 to 5 the vector part is along $\vec e_x$, so
$$ \vec{Ζ}=-i\vec e_x, \qquad \tilde\Pi_+=\frac12(e_t-i\vec e_x), \qquad \tilde\Pi_-=\frac12(e_t+i\vec e_x). $$
Choose $\tilde Q_1=\vec e_y$. Then
$$ u=\frac12(\vec e_y+i\vec e_z), \qquad v=\frac12(\vec e_y-i\vec e_z), \qquad uv=-\tilde\Pi_-, \qquad vu=-\tilde\Pi_+. $$
Thus $c=-1$, and the choice $\alpha=i$ gives
$$ \boxed{\; \vec\Upsilon_+=\frac12(\vec e_z-i\vec e_y), \qquad \vec\Upsilon_-=\frac12(-\vec e_z-i\vec e_y). \;} $$
The pair is the same for the four elements, because it depends only on $\vec{Ζ}$, not on the eigenvalues.
The Eigenbasis
The four elements
$$ \tilde\Pi_+, \qquad \vec\Upsilon_-, \qquad \vec\Upsilon_+, \qquad \tilde\Pi_- $$
form a $\mathbb{C}$-basis of $\mathbb{B}$. It is the eigenbasis. Write $L_{\tilde Q}(\tilde Q_2)=\tilde Q\tilde Q_2$ and $R_{\tilde Q}(\tilde Q_2)=\tilde Q_2\tilde Q$ for left and right multiplication. In this basis multiplication by $\tilde Q$ is diagonal:
$$ \tilde Q\tilde\Pi_+=\lambda_+\tilde\Pi_+, \qquad \tilde Q\vec\Upsilon_-=\lambda_+\vec\Upsilon_-, \qquad \tilde Q\vec\Upsilon_+=\lambda_-\vec\Upsilon_+, \qquad \tilde Q\tilde\Pi_-=\lambda_-\tilde\Pi_-. $$
Thus the matrix of left multiplication $L_{\tilde Q}$ in the eigenbasis is
$$ \operatorname{diag}(\lambda_+,\lambda_+,\lambda_-,\lambda_-). $$
Because each eigenvalue occurs twice, the eigenbasis is not unique: any basis of the $\lambda_+$-eigenspace together with any basis of the $\lambda_-$-eigenspace is an eigenbasis. Write $\mathcal{B}(\tilde Q)=(\tilde\Pi_+,\vec\Upsilon_-,\vec\Upsilon_+,\tilde\Pi_-)$ for the Peirce eigenbasis, normalised by $\vec\Upsilon_+\vec\Upsilon_-=\tilde\Pi_-$: the two idempotents are the spectral idempotents of $\tilde Q$, and the two nilpotents are an admissible pair, determined up to the reciprocal rescaling.
The two minimal left ideals are
$$ S_+=\mathbb{B}\tilde\Pi_+=\mathbb{C}\tilde\Pi_+\oplus\mathbb{C}\vec\Upsilon_+, \qquad S_-=\mathbb{B}\tilde\Pi_-=\mathbb{C}\vec\Upsilon_-\oplus\mathbb{C}\tilde\Pi_-, $$
and
$$ \mathbb{B}=S_+\oplus S_-. $$
On $S_+$ the eigenbasis is $(\tilde\Pi_+,\vec\Upsilon_+)$, with eigenvalues $\lambda_+$ and $\lambda_-$; on $S_-$ it is $(\vec\Upsilon_-,\tilde\Pi_-)$, with eigenvalues $\lambda_+$ and $\lambda_-$. Each is an eigenbasis with distinct eigenvalues, because $B\neq0$.
The spaces $S_\pm$ are the spinor spaces, the Peirce carriers and here the minimal left ideals. They are not the eigenspaces of $L_{\tilde Q}$: the $\lambda_+$-eigenspace is the right ideal $\tilde\Pi_+\mathbb{B}=\mathbb{C}\tilde\Pi_+\oplus\mathbb{C}\vec\Upsilon_-$, and the $\lambda_-$-eigenspace is $\tilde\Pi_-\mathbb{B}=\mathbb{C}\vec\Upsilon_+\oplus\mathbb{C}\tilde\Pi_-$. The spaces $S_\pm$ are the eigenspaces of right multiplication $R_{\tilde Q}$. This basis is an eigenbasis of $L_{\tilde Q}$; on each $S_\pm$, $L_{\tilde Q}$ carries both eigenvalues, while $R_{\tilde Q}$ is scalar. The side of the multiplication is the whole difference. The labels first and second are fixed by the convention $\tilde\Pi_+=(e_t+\vec{Ζ})/2$; replacing $B$ by $-B$ exchanges them.
Spinors and the $2\times2$ Model
A spinor of the first Peirce copy is an element of $S_+$,
$$ \zeta_+=Q_{2,11}\tilde\Pi_++Q_{2,21}\vec\Upsilon_+, \qquad Q_{2,11},Q_{2,21}\in\mathbb{C}, $$
with coordinates $(Q_{2,11},Q_{2,21})$, the first column of the matrix. Under multiplication by $\tilde Q$,
$$ \tilde Q\zeta_+=\lambda_+Q_{2,11}\tilde\Pi_++\lambda_-Q_{2,21}\vec\Upsilon_+. $$
A spinor of the second Peirce copy is an element of $S_-$,
$$ \zeta_-=Q_{2,12}\vec\Upsilon_-+Q_{2,22}\tilde\Pi_-, \qquad Q_{2,12},Q_{2,22}\in\mathbb{C}, $$
with coordinates $(Q_{2,12},Q_{2,22})$, the second column, and
$$ \tilde Q\zeta_-=\lambda_+Q_{2,12}\vec\Upsilon_-+\lambda_-Q_{2,22}\tilde\Pi_-. $$
The same diagonal matrix $\operatorname{diag}(\lambda_+,\lambda_-)$ acts on the coordinates of both spinor spaces.
Every element of $\mathbb{B}$ splits as $\tilde Q_2=\tilde Q_2\tilde\Pi_++\tilde Q_2\tilde\Pi_-$, a spinor of the first Peirce copy plus a spinor of the second Peirce copy. A general element is therefore a pair of spinors, not a spinor.
The assignment
$$ \tilde\Pi_+\longmapsto E_{11}, \qquad \vec\Upsilon_-\longmapsto E_{12}, \qquad \vec\Upsilon_+\longmapsto E_{21}, \qquad \tilde\Pi_-\longmapsto E_{22} $$
gives the eigenbasis model $\mathsf M_2^Q$. For a general element
$$ \tilde Q_2=a_{11}\tilde\Pi_++a_{12}\vec\Upsilon_-+a_{21}\vec\Upsilon_++a_{22}\tilde\Pi_-, $$
the matrix is
$$ \mathsf M_2^Q(\tilde Q_2)= \begin{pmatrix} a_{11}&a_{12}\\ a_{21}&a_{22} \end{pmatrix}, $$
and for $\tilde Q$ itself
$$ \mathsf M_2^Q(\tilde Q)= \begin{pmatrix} \lambda_+&0\\ 0&\lambda_- \end{pmatrix}. $$
The first column is $S_+$ and the second column is $S_-$.
The Fixed Matrix Model
The fixed model sends
$$ \mathsf M_2(e_t)=I_2, \qquad \mathsf M_2(\vec e_x)= \begin{pmatrix}0&-i\\-i&0\end{pmatrix}, \qquad \mathsf M_2(\vec e_y)= \begin{pmatrix}0&-1\\1&0\end{pmatrix}, \qquad \mathsf M_2(\vec e_z)= \begin{pmatrix}-i&0\\0&i\end{pmatrix}. $$
For a general element,
$$ \mathsf M_2(\tilde Q_2)= \begin{pmatrix} Q_{2,t}-iQ_{2,z}&-iQ_{2,x}-Q_{2,y}\\ -iQ_{2,x}+Q_{2,y}&Q_{2,t}+iQ_{2,z} \end{pmatrix}, $$
with trace $2Q_{2,t}$ and determinant $Q_{2,t}^2+Q_{2,x}^2+Q_{2,y}^2+Q_{2,z}^2$.
The spectral projectors are
$$ \rho_\pm=\mathsf M_2(\tilde\Pi_\pm) =\frac12\left(I_2\pm\mathsf M_2(\vec{Ζ})\right), $$
and they satisfy
$$ \rho_\pm^2=\rho_\pm, \qquad \rho_+\rho_-=\rho_-\rho_+=0, \qquad \rho_++\rho_-=I_2, \qquad \mathsf M_2(\tilde Q)=\lambda_+\rho_++\lambda_-\rho_-. $$
The Regimes
| Spectral regime | Condition | Spectrum | Eigenbasis |
|---|---|---|---|
| Spectrally regular (distinct) | $B\neq0$ | $\lambda_+\neq\lambda_-$ | the Peirce eigenbasis; unique up to the reciprocal rescaling of $\vec\Upsilon_\pm$ and the exchange from $B\mapsto-B$ |
| Spectrally degenerate (Jordan) | $B=0$, $\vec Q\neq0$ | $\lambda_+= \lambda_-=Q_t$ | none: the eigenspace is two-dimensional |
| Spectrally degenerate (scalar) | $B=0$, $\vec Q=0$ | $\lambda_+= \lambda_-=Q_t$ | every basis |
In the Jordan regime, write $\tilde Q=Q_te_t+\vec N$ with $\vec N=\vec Q$ and $\vec N^2=0$. The minimal polynomial is $(X-Q_t)^2$. When $Q_t\neq0$ the element is unipotent up to the scalar $Q_t$; when $Q_t=0$ it is nilpotent. In both subcases the eigenspace is $\ker L_{\vec N}$, of dimension two. A two-dimensional eigenspace in a four-dimensional algebra is not a basis of eigenvectors, so there is no eigenbasis. In the scalar regime the operator is $Q_t\,\mathrm{id}$, every nonzero vector is an eigenvector, and every basis is an eigenbasis; the construction of the eigenbasis does not apply because $\vec{Ζ}$ is undefined.
The Eight Cases
Case 1: The Identity
Coordinates. For $\tilde Q=e_t$,
$$ Q_t=1, \qquad Q_x=Q_y=Q_z=0, \qquad B=0, \qquad \langle\tilde Q\rangle_{\natural}=1. $$
The scalar regime. The vector part vanishes, so the element is the scalar $1$. The two eigenvalues coincide, $\lambda_\pm=1$, and the left multiplication is the identity:
$$ L_{\tilde Q}=\mathrm{id}. $$
Every nonzero vector is an eigenvector, every basis is an eigenbasis, and the eigenspace is the whole algebra, of dimension four. There is no eigenbasis attached to $\tilde Q$, because the root $\vec{Ζ}=-i\vec Q/B$ is undefined. The standard Peirce pair $\tilde\Pi_\pm=\tfrac12(e_t\pm i\vec e_z)$ is a choice; in it the matrix of $L_{\tilde Q}$ is the identity.
Spinors. There is no distinguished spinor pair: $B=0$, the two eigenvalues coincide, and the eigenspace is two-dimensional, so $\tilde Q$ selects no pair of complementary minimal left ideals. A general element still has two spinor components in the fixed Peirce basis. With $$ \mathsf M_2(\tilde Q_2)=\begin{pmatrix}Q_{2,11}&Q_{2,12}\\Q_{2,21}&Q_{2,22}\end{pmatrix}, $$ the first column $(Q_{2,11},Q_{2,21})$ is the spinor in the first Peirce carrier and the second column $(Q_{2,12},Q_{2,22})$ is the spinor in the second. Their matrices are $$ \mathsf M_2(\zeta_+)=\begin{pmatrix}Q_{2,11}&0\\Q_{2,21}&0\end{pmatrix}, \qquad \mathsf M_2(\zeta_-)=\begin{pmatrix}0&Q_{2,12}\\0&Q_{2,22}\end{pmatrix}. $$ In the special case $$ \mathsf M_2(\tilde Q)=\begin{pmatrix}1&0\\0&1\end{pmatrix}, $$ the two columns are $(1,0)$ and $(0,1)$. Both columns are eigenvectors with eigenvalue $1$: every idempotent is an eigenprojection, so every pair of complementary minimal left ideals is a pair of spinor spaces.
Matrix. In the fixed model,
$$ \mathsf M_2(\tilde Q)=I_2, \qquad \mathrm{tr}\,\mathsf M_2(\tilde Q)=2, \qquad \det=1. $$
The eigenprojections of the matrix are not unique: the spectral projector of the only eigenvalue is $I_2$, and every decomposition of $I_2$ into two complementary idempotents is a choice.
Consequences. $\tilde Q^2=e_t$ and $\tilde Q^{-1}=e_t$. The case is the trivial eigenbasis: the algebra is already diagonal, and the element does not select a basis.
Case 2: A Pure Vector of the Unit Sphere
Coordinates. For $\tilde Q=\vec e_x$,
$$ Q_t=0, \qquad Q_x=1, \qquad Q_y=Q_z=0, \qquad B=1, \qquad \langle\tilde Q\rangle_{\natural}=1. $$
Spectrum. The eigenvalues are
$$ \lambda_+=i, \qquad \lambda_-=-i. $$
Eigenprojections. With $\vec{Ζ}=-i\vec e_x$,
$$ \tilde\Pi_+=\frac12(e_t-i\vec e_x), \qquad \tilde\Pi_-=\frac12(e_t+i\vec e_x), $$
and
$$ \tilde Q\tilde\Pi_+=i\tilde\Pi_+, \qquad \tilde Q\tilde\Pi_-=-i\tilde\Pi_-, \qquad \tilde Q=i\tilde\Pi_+-i\tilde\Pi_-. $$
Nilpotents. The shared solution of §The Method gives
$$ \vec\Upsilon_+=\frac12(\vec e_z-i\vec e_y), \qquad \vec\Upsilon_-=\frac12(-\vec e_z-i\vec e_y). $$
The checks are
$$ \vec\Upsilon_+\vec\Upsilon_-=\tilde\Pi_-, \qquad \vec\Upsilon_-\vec\Upsilon_+=\tilde\Pi_+, \qquad \vec\Upsilon_+^2=\vec\Upsilon_-^2=0. $$
The eigenbasis. The basis is
$$ \tilde\Pi_+, \qquad \vec\Upsilon_-, \qquad \vec\Upsilon_+, \qquad \tilde\Pi_-, $$
with the diagonal action
$$ \tilde Q\tilde\Pi_+=i\tilde\Pi_+, \qquad \tilde Q\vec\Upsilon_-=i\vec\Upsilon_-, \qquad \tilde Q\vec\Upsilon_+=-i\vec\Upsilon_+, \qquad \tilde Q\tilde\Pi_-=-i\tilde\Pi_-. $$
Thus $L_{\tilde Q}$ has the diagonal form
$$ \operatorname{diag}(i,i,-i,-i), $$
and the $2\times2$ model is
$$ \mathsf M_2^Q(\tilde Q)=\operatorname{diag}(i,-i). $$
Spinors. A general element is $\tilde Q_2=Q_{2,11}\tilde\Pi_++Q_{2,12}\vec\Upsilon_-+Q_{2,21}\vec\Upsilon_++Q_{2,22}\tilde\Pi_-$, with matrix $$ \mathsf M_2^Q(\tilde Q_2)=\begin{pmatrix}Q_{2,11}&Q_{2,12}\\Q_{2,21}&Q_{2,22}\end{pmatrix} $$ in the eigenbasis $(\tilde\Pi_+,\vec\Upsilon_-,\vec\Upsilon_+,\tilde\Pi_-)$. Its two columns are the two spinors: $$ \zeta_+=Q_{2,11}\tilde\Pi_++Q_{2,21}\vec\Upsilon_+\longleftrightarrow\begin{pmatrix}Q_{2,11}\\Q_{2,21}\end{pmatrix}, \qquad \zeta_-=Q_{2,12}\vec\Upsilon_-+Q_{2,22}\tilde\Pi_-\longleftrightarrow\begin{pmatrix}Q_{2,12}\\Q_{2,22}\end{pmatrix}, $$ with matrices $$ \mathsf M_2^Q(\zeta_+)=\begin{pmatrix}Q_{2,11}&0\\Q_{2,21}&0\end{pmatrix}, \qquad \mathsf M_2^Q(\zeta_-)=\begin{pmatrix}0&Q_{2,12}\\0&Q_{2,22}\end{pmatrix}. $$ Left multiplication by $\tilde Q$ multiplies both columns by $\operatorname{diag}(\lambda_+,\lambda_-)$: $$ (Q_{2,11},Q_{2,21})\longmapsto(\lambda_+Q_{2,11},\lambda_-Q_{2,21}), \qquad (Q_{2,12},Q_{2,22})\longmapsto(\lambda_+Q_{2,12},\lambda_-Q_{2,22}). $$ In the special case $\tilde Q_2=\vec e_x$ one has $Q_{2,11}=\lambda_+=i$, $Q_{2,21}=0$, $Q_{2,12}=0$, $Q_{2,22}=\lambda_-=-i$, so $$ \mathsf M_2^Q(\tilde Q)=\begin{pmatrix}i&0\\0&-i\end{pmatrix}, \qquad \mathsf M_2^Q(\zeta_+)=\begin{pmatrix}i&0\\0&0\end{pmatrix}, \qquad \mathsf M_2^Q(\zeta_-)=\begin{pmatrix}0&0\\0&-i\end{pmatrix}. $$
Matrix. In the fixed model,
$$ \mathsf M_2(\tilde Q)= \begin{pmatrix} 0&-i\\ -i&0 \end{pmatrix}, \qquad \mathrm{tr}\,\mathsf M_2(\tilde Q)=0, \qquad \det=1. $$
The spectral projectors are
$$ \rho_+=\frac12\begin{pmatrix}1&-1\\-1&1\end{pmatrix}, \qquad \rho_-=\frac12\begin{pmatrix}1&1\\1&1\end{pmatrix}, $$
and indeed
$$ i\rho_+-i\rho_-= \begin{pmatrix}0&-i\\-i&0\end{pmatrix}. $$
Consequences. The element is a square root of $-e_t$:
$$ \tilde Q^2=-e_t, \qquad \tilde Q^{-1}=-\tilde Q=-\vec e_x, \qquad \tilde Q^3=-\vec e_x, \qquad \tilde Q^4=e_t. $$
Case 3: A Sum of the Identity and a Pure Vector
Coordinates. For $\tilde Q=\frac{1}{\sqrt2}(e_t+\vec e_x)$,
$$ Q_t=\frac{1}{\sqrt2}, \qquad Q_x=\frac{1}{\sqrt2}, \qquad Q_y=Q_z=0, \qquad B=\frac{1}{\sqrt2}, \qquad \langle\tilde Q\rangle_{\natural}=1. $$
Spectrum. The eigenvalues are
$$ \lambda_+=\frac{1+i}{\sqrt2}, \qquad \lambda_-=\frac{1-i}{\sqrt2}. $$
Eigenprojections. The vector part is again along $\vec e_x$, so the projectors and the nilpotents are those of case 2:
$$ \tilde\Pi_+=\frac12(e_t-i\vec e_x), \qquad \tilde\Pi_-=\frac12(e_t+i\vec e_x), \qquad \vec\Upsilon_\pm=\frac12(\pm\vec e_z-i\vec e_y). $$
Therefore
$$ \tilde Q=\frac{1+i}{\sqrt2}\tilde\Pi_++\frac{1-i}{\sqrt2}\tilde\Pi_-. $$
The eigenbasis. The basis is again
$$ (\tilde\Pi_+,\vec\Upsilon_-,\vec\Upsilon_+,\tilde\Pi_-), $$
with the diagonal action
$$ \operatorname{diag}\left(\frac{1+i}{\sqrt2},\frac{1+i}{\sqrt2},\frac{1-i}{\sqrt2},\frac{1-i}{\sqrt2}\right). $$
The $2\times2$ model is
$$ \mathsf M_2^Q(\tilde Q)= \operatorname{diag}\left(\frac{1+i}{\sqrt2},\frac{1-i}{\sqrt2}\right). $$
Spinors. A general element is $\tilde Q_2=Q_{2,11}\tilde\Pi_++Q_{2,12}\vec\Upsilon_-+Q_{2,21}\vec\Upsilon_++Q_{2,22}\tilde\Pi_-$, with matrix $$ \mathsf M_2^Q(\tilde Q_2)=\begin{pmatrix}Q_{2,11}&Q_{2,12}\\Q_{2,21}&Q_{2,22}\end{pmatrix} $$ in the eigenbasis $(\tilde\Pi_+,\vec\Upsilon_-,\vec\Upsilon_+,\tilde\Pi_-)$. Its two columns are the two spinors: $$ \zeta_+=Q_{2,11}\tilde\Pi_++Q_{2,21}\vec\Upsilon_+\longleftrightarrow\begin{pmatrix}Q_{2,11}\\Q_{2,21}\end{pmatrix}, \qquad \zeta_-=Q_{2,12}\vec\Upsilon_-+Q_{2,22}\tilde\Pi_-\longleftrightarrow\begin{pmatrix}Q_{2,12}\\Q_{2,22}\end{pmatrix}, $$ with matrices $$ \mathsf M_2^Q(\zeta_+)=\begin{pmatrix}Q_{2,11}&0\\Q_{2,21}&0\end{pmatrix}, \qquad \mathsf M_2^Q(\zeta_-)=\begin{pmatrix}0&Q_{2,12}\\0&Q_{2,22}\end{pmatrix}. $$ Left multiplication by $\tilde Q$ multiplies both columns by $\operatorname{diag}(\lambda_+,\lambda_-)$: $$ (Q_{2,11},Q_{2,21})\longmapsto(\lambda_+Q_{2,11},\lambda_-Q_{2,21}), \qquad (Q_{2,12},Q_{2,22})\longmapsto(\lambda_+Q_{2,12},\lambda_-Q_{2,22}). $$ In the special case $\tilde Q_2=\tfrac{1}{\sqrt2}(e_t+\vec e_x)$ one has $Q_{2,11}=\lambda_+=\tfrac{1+i}{\sqrt2}$, $Q_{2,21}=0$, $Q_{2,12}=0$, $Q_{2,22}=\lambda_-=\tfrac{1-i}{\sqrt2}$, so $$ \mathsf M_2^Q(\tilde Q)=\begin{pmatrix}\tfrac{1+i}{\sqrt2}&0\\0&\tfrac{1-i}{\sqrt2}\end{pmatrix}, \qquad \mathsf M_2^Q(\zeta_+)=\begin{pmatrix}\tfrac{1+i}{\sqrt2}&0\\0&0\end{pmatrix}, \qquad \mathsf M_2^Q(\zeta_-)=\begin{pmatrix}0&0\\0&\tfrac{1-i}{\sqrt2}\end{pmatrix}. $$
Matrix. In the fixed model,
$$ \mathsf M_2(\tilde Q)= \frac{1}{\sqrt2} \begin{pmatrix} 1&-i\\ -i&1 \end{pmatrix}, \qquad \mathrm{tr}\,\mathsf M_2(\tilde Q)=\sqrt2, \qquad \det=1. $$
Its spectral projectors are the same $\rho_\pm$ as in case 2, and
$$ \lambda_+\rho_++\lambda_-\rho_-= \frac{1}{\sqrt2} \begin{pmatrix} 1&-i\\ -i&1 \end{pmatrix}. $$
Consequences. The square is
$$ \tilde Q^2=\vec e_x, $$
the cube is
$$ \tilde Q^3=\frac{\vec e_x-e_t}{\sqrt2}=-\tilde Q^{-1}, $$
and the inverse is
$$ \tilde Q^{-1}=\frac{e_t-\vec e_x}{\sqrt2}. $$
The fourth power is $-e_t$. The element is a square root of $\vec e_x$.
Case 4: A Real Spectrum from an Imaginary Root
Coordinates. For $\tilde Q=\sqrt2\,e_t+i\vec e_x$,
$$ Q_t=\sqrt2, \qquad Q_x=i, \qquad Q_y=Q_z=0, \qquad B=i, \qquad \langle\tilde Q\rangle_{\natural}=1. $$
Spectrum. With the root $B=i$,
$$ \lambda_+=\sqrt2-1, \qquad \lambda_-=\sqrt2+1. $$
The two eigenvalues are real and distinct, although $B$ is imaginary. The other root $-i$ exchanges them.
Eigenprojections. The vector part is along $\vec e_x$, so
$$ \tilde\Pi_+=\frac12(e_t-i\vec e_x), \qquad \tilde\Pi_-=\frac12(e_t+i\vec e_x), $$
and
$$ \tilde Q=(\sqrt2-1)\tilde\Pi_++(\sqrt2+1)\tilde\Pi_-. $$
Nilpotents. The nilpotents are the shared pair
$$ \vec\Upsilon_+=\frac12(\vec e_z-i\vec e_y), \qquad \vec\Upsilon_-=\frac12(-\vec e_z-i\vec e_y). $$
The eigenbasis. The diagonal action is
$$ \operatorname{diag}(\sqrt2-1,\sqrt2-1,\sqrt2+1,\sqrt2+1), $$
and the $2\times2$ model is
$$ \mathsf M_2^Q(\tilde Q)=\operatorname{diag}(\sqrt2-1,\sqrt2+1). $$
Spinors. A general element is $\tilde Q_2=Q_{2,11}\tilde\Pi_++Q_{2,12}\vec\Upsilon_-+Q_{2,21}\vec\Upsilon_++Q_{2,22}\tilde\Pi_-$, with matrix $$ \mathsf M_2^Q(\tilde Q_2)=\begin{pmatrix}Q_{2,11}&Q_{2,12}\\Q_{2,21}&Q_{2,22}\end{pmatrix} $$ in the eigenbasis $(\tilde\Pi_+,\vec\Upsilon_-,\vec\Upsilon_+,\tilde\Pi_-)$. Its two columns are the two spinors: $$ \zeta_+=Q_{2,11}\tilde\Pi_++Q_{2,21}\vec\Upsilon_+\longleftrightarrow\begin{pmatrix}Q_{2,11}\\Q_{2,21}\end{pmatrix}, \qquad \zeta_-=Q_{2,12}\vec\Upsilon_-+Q_{2,22}\tilde\Pi_-\longleftrightarrow\begin{pmatrix}Q_{2,12}\\Q_{2,22}\end{pmatrix}, $$ with matrices $$ \mathsf M_2^Q(\zeta_+)=\begin{pmatrix}Q_{2,11}&0\\Q_{2,21}&0\end{pmatrix}, \qquad \mathsf M_2^Q(\zeta_-)=\begin{pmatrix}0&Q_{2,12}\\0&Q_{2,22}\end{pmatrix}. $$ Left multiplication by $\tilde Q$ multiplies both columns by $\operatorname{diag}(\lambda_+,\lambda_-)$: $$ (Q_{2,11},Q_{2,21})\longmapsto(\lambda_+Q_{2,11},\lambda_-Q_{2,21}), \qquad (Q_{2,12},Q_{2,22})\longmapsto(\lambda_+Q_{2,12},\lambda_-Q_{2,22}). $$ In the special case $\tilde Q_2=\sqrt2\,e_t+i\vec e_x$ one has $Q_{2,11}=\lambda_+=\sqrt2-1$, $Q_{2,21}=0$, $Q_{2,12}=0$, $Q_{2,22}=\lambda_-=\sqrt2+1$, so $$ \mathsf M_2^Q(\tilde Q)=\begin{pmatrix}\sqrt2-1&0\\0&\sqrt2+1\end{pmatrix}, \qquad \mathsf M_2^Q(\zeta_+)=\begin{pmatrix}\sqrt2-1&0\\0&0\end{pmatrix}, \qquad \mathsf M_2^Q(\zeta_-)=\begin{pmatrix}0&0\\0&\sqrt2+1\end{pmatrix}. $$
Matrix. In the fixed model,
$$ \mathsf M_2(\tilde Q)= \begin{pmatrix} \sqrt2&1\\ 1&\sqrt2 \end{pmatrix}, \qquad \mathrm{tr}\,\mathsf M_2(\tilde Q)=2\sqrt2, \qquad \det=1. $$
This is a real symmetric matrix with eigenvectors $(1,-1)$ and $(1,1)$, which are the columns of the projectors up to scale.
Consequences. The square is
$$ \tilde Q^2=3+2\sqrt2\,i\vec e_x, $$
the cube is
$$ \tilde Q^3=5\sqrt2+7i\vec e_x, $$
and the inverse is
$$ \tilde Q^{-1}=\sqrt2-i\vec e_x. $$
Case 5: A Generic Element on the Same Axis
Coordinates. For
$$ \tilde Q=\left(1-\frac{i}{\sqrt2}\right)e_t+\left(1+\frac{i}{\sqrt2}\right)\vec e_x, $$
the coordinates are
$$ Q_t=1-\frac{i}{\sqrt2}, \qquad Q_x=1+\frac{i}{\sqrt2}, \qquad Q_y=Q_z=0, \qquad B=1+\frac{i}{\sqrt2}, \qquad \langle\tilde Q\rangle_{\natural}=1. $$
Spectrum. Put $a=1-\frac{1}{\sqrt2}$ and $b=1+\frac{1}{\sqrt2}$. Then
$$ \lambda_+=a(1+i), \qquad \lambda_-=b(1-i). $$
Numerically,
$$ \lambda_+\approx 0.2928932188+0.2928932188\,i, \qquad \lambda_-\approx 1.7071067812-1.7071067812\,i. $$
Eigenprojections. The vector part is along $\vec e_x$, so
$$ \tilde\Pi_+=\frac12(e_t-i\vec e_x), \qquad \tilde\Pi_-=\frac12(e_t+i\vec e_x), $$
and
$$ \tilde Q=a(1+i)\tilde\Pi_++b(1-i)\tilde\Pi_-. $$
Nilpotents. The shared pair is
$$ \vec\Upsilon_+=\frac12(\vec e_z-i\vec e_y), \qquad \vec\Upsilon_-=\frac12(-\vec e_z-i\vec e_y). $$
The eigenbasis. The diagonal action is
$$ \operatorname{diag}\bigl(a(1+i),a(1+i),b(1-i),b(1-i)\bigr), $$
and the $2\times2$ model is
$$ \mathsf M_2^Q(\tilde Q)=\operatorname{diag}\bigl(a(1+i),b(1-i)\bigr). $$
Spinors. A general element is $\tilde Q_2=Q_{2,11}\tilde\Pi_++Q_{2,12}\vec\Upsilon_-+Q_{2,21}\vec\Upsilon_++Q_{2,22}\tilde\Pi_-$, with matrix $$ \mathsf M_2^Q(\tilde Q_2)=\begin{pmatrix}Q_{2,11}&Q_{2,12}\\Q_{2,21}&Q_{2,22}\end{pmatrix} $$ in the eigenbasis $(\tilde\Pi_+,\vec\Upsilon_-,\vec\Upsilon_+,\tilde\Pi_-)$. Its two columns are the two spinors: $$ \zeta_+=Q_{2,11}\tilde\Pi_++Q_{2,21}\vec\Upsilon_+\longleftrightarrow\begin{pmatrix}Q_{2,11}\\Q_{2,21}\end{pmatrix}, \qquad \zeta_-=Q_{2,12}\vec\Upsilon_-+Q_{2,22}\tilde\Pi_-\longleftrightarrow\begin{pmatrix}Q_{2,12}\\Q_{2,22}\end{pmatrix}, $$ with matrices $$ \mathsf M_2^Q(\zeta_+)=\begin{pmatrix}Q_{2,11}&0\\Q_{2,21}&0\end{pmatrix}, \qquad \mathsf M_2^Q(\zeta_-)=\begin{pmatrix}0&Q_{2,12}\\0&Q_{2,22}\end{pmatrix}. $$ Left multiplication by $\tilde Q$ multiplies both columns by $\operatorname{diag}(\lambda_+,\lambda_-)$: $$ (Q_{2,11},Q_{2,21})\longmapsto(\lambda_+Q_{2,11},\lambda_-Q_{2,21}), \qquad (Q_{2,12},Q_{2,22})\longmapsto(\lambda_+Q_{2,12},\lambda_-Q_{2,22}). $$ In the special case $\tilde Q_2=(1-\tfrac{i}{\sqrt2})e_t+(1+\tfrac{i}{\sqrt2})\vec e_x$ one has $Q_{2,11}=\lambda_+=a(1+i)$, $Q_{2,21}=0$, $Q_{2,12}=0$, $Q_{2,22}=\lambda_-=b(1-i)$, so $$ \mathsf M_2^Q(\tilde Q)=\begin{pmatrix}a(1+i)&0\\0&b(1-i)\end{pmatrix}, \qquad \mathsf M_2^Q(\zeta_+)=\begin{pmatrix}a(1+i)&0\\0&0\end{pmatrix}, \qquad \mathsf M_2^Q(\zeta_-)=\begin{pmatrix}0&0\\0&b(1-i)\end{pmatrix}. $$
Matrix. In the fixed model,
$$ \mathsf M_2(\tilde Q)= \begin{pmatrix} 1-\frac{i}{\sqrt2}&\frac{1}{\sqrt2}-i\\ \frac{1}{\sqrt2}-i&1-\frac{i}{\sqrt2} \end{pmatrix}, \qquad \mathrm{tr}\,\mathsf M_2(\tilde Q)=2-i\sqrt2, \qquad \det=1. $$
Consequences. The square is
$$ \tilde Q^2=3\vec e_x-2\sqrt2\,i\,e_t, $$
and the inverse is
$$ \tilde Q^{-1}=\left(1-\frac{i}{\sqrt2}\right)-\left(1+\frac{i}{\sqrt2}\right)\vec e_x. $$
Case 6: Orthogonal Axes and a Unipotent Element
Coordinates. For
$$ \tilde Q=e_t+\vec e_y+\frac{i}{\sqrt2}\vec e_x+\frac{i}{\sqrt2}\vec e_z, $$
the coordinates are
$$ Q_t=1, \qquad Q_x=\frac{i}{\sqrt2}, \qquad Q_y=1, \qquad Q_z=\frac{i}{\sqrt2}, \qquad B=0, \qquad \langle\tilde Q\rangle_{\natural}=1. $$
The Jordan regime. The vector part is nonzero and null:
$$ Q_x^2+Q_y^2+Q_z^2=-\frac12+1-\frac12=0. $$
Write
$$ \tilde N=\frac{i}{\sqrt2}\vec e_x+\vec e_y+\frac{i}{\sqrt2}\vec e_z. $$
Then
$$ \tilde N^2=-\frac12+1-\frac12=0, \qquad \tilde Q=e_t+\tilde N. $$
The minimal polynomial is $(X-1)^2$, the only eigenvalue is $1$, and the eigenspace is
$$ \ker L_{\tilde N}=\{\tilde Q_2:\tilde N\tilde Q_2=0\}, $$
of dimension two. A two-dimensional eigenspace does not span the four-dimensional algebra, so there is no eigenbasis. The element is unipotent: it is invertible, with inverse
$$ \tilde Q^{-1}=e_t-\tilde N =e_t-\frac{i}{\sqrt2}\vec e_x-\vec e_y-\frac{i}{\sqrt2}\vec e_z. $$
Spinors. There is no distinguished spinor pair: $B=0$, the two eigenvalues coincide, and the eigenspace is two-dimensional, so $\tilde Q$ selects no pair of complementary minimal left ideals. A general element still has two spinor components in the fixed Peirce basis. With $$ \mathsf M_2(\tilde Q_2)=\begin{pmatrix}Q_{2,11}&Q_{2,12}\\Q_{2,21}&Q_{2,22}\end{pmatrix}, $$ the first column $(Q_{2,11},Q_{2,21})$ is the spinor in the first Peirce carrier and the second column $(Q_{2,12},Q_{2,22})$ is the spinor in the second. Their matrices are $$ \mathsf M_2(\zeta_+)=\begin{pmatrix}Q_{2,11}&0\\Q_{2,21}&0\end{pmatrix}, \qquad \mathsf M_2(\zeta_-)=\begin{pmatrix}0&Q_{2,12}\\0&Q_{2,22}\end{pmatrix}. $$ In the special case $$ \mathsf M_2(\tilde Q)=\begin{pmatrix}1+\frac{1}{\sqrt2}&\frac{1}{\sqrt2}-1\\[2pt]1+\frac{1}{\sqrt2}&1-\frac{1}{\sqrt2}\end{pmatrix}, $$ the two columns are $(1+\frac{1}{\sqrt2},\,1+\frac{1}{\sqrt2})$ and $(\frac{1}{\sqrt2}-1,\,1-\frac{1}{\sqrt2})$. The first column is an eigenvector with eigenvalue $1$; the second is not. The eigenspace is one line, so there is one eigenspinor and no second one, and no spinor decomposition of $\tilde Q$.
Matrix. In the fixed model,
$$ \mathsf M_2(\tilde Q)= \begin{pmatrix} 1+\frac{1}{\sqrt2}&\frac{1}{\sqrt2}-1\\ 1+\frac{1}{\sqrt2}&1-\frac{1}{\sqrt2} \end{pmatrix} =I_2+\mathsf N, \qquad \mathsf N= \begin{pmatrix} \frac{1}{\sqrt2}&\frac{1}{\sqrt2}-1\\ 1+\frac{1}{\sqrt2}&-\frac{1}{\sqrt2} \end{pmatrix}, $$
with $\mathsf N^2=0$. The kernel of $\mathsf N$ is the line spanned by $(1,1+\sqrt2)$, and the eigenspace of $\mathsf M_2(\tilde Q)$ is the two-dimensional space of matrices whose two columns lie on that line.
Consequences. The powers are
$$ \tilde Q^n=e_t+n\tilde N, $$
so
$$ \tilde Q^2=e_t+2\tilde N =e_t+i\sqrt2\,\vec e_x+2\vec e_y+i\sqrt2\,\vec e_z. $$
The element is $\natural$-regular and of norm one, although its vector part is null.
Case 7: A Generic Direction
Coordinates. For
$$ \tilde Q=\left(1-\frac{i}{2}\right)e_t+\frac{1+i}{\sqrt2}\vec e_x+\frac{1}{\sqrt2}\vec e_y+\frac{i}{2}\vec e_z, $$
the coordinates are
$$ Q_t=1-\frac{i}{2}, \qquad Q_x=\frac{1+i}{\sqrt2}, \qquad Q_y=\frac{1}{\sqrt2}, \qquad Q_z=\frac{i}{2}. $$
The vector part has square
$$ B^2=\frac{(1+i)^2}{2}+\frac12+\frac{i^2}{4} =i+\frac12-\frac14=\frac14+i, $$
so
$$ B=\sqrt{\frac14+i}\approx 0.8002425902+0.6248105338\,i. $$
Spectrum. The eigenvalues are
$$ \lambda_+=1-\frac{i}{2}+iB\approx 0.3751894662+0.3002425902\,i, $$
$$ \lambda_-=1-\frac{i}{2}-iB\approx 1.6248105338-1.3002425902\,i. $$
Their product is $1$. The element is invertible.
Eigenprojections. With $\vec{Ζ}=-i\vec Q/B$,
$$ \vec{Ζ}=\frac{1-i}{\sqrt2\,B}\vec e_x-\frac{i}{\sqrt2\,B}\vec e_y+\frac{1}{2B}\vec e_z, $$
and $\tilde\Pi_\pm=\frac12(e_t\pm\vec{Ζ})$. Numerically,
$$ \tilde\Pi_+\approx \frac12e_t+(0.0601727-0.4887892\,i)\vec e_x+(-0.2143082-0.2744809\,i)\vec e_y+(0.1940873-0.1515388\,i)\vec e_z, $$
$$ \tilde\Pi_-\approx \frac12e_t+(-0.0601727+0.4887892\,i)\vec e_x+(0.2143082+0.2744809\,i)\vec e_y+(-0.1940873+0.1515388\,i)\vec e_z. $$
The decomposition is
$$ \tilde Q=\lambda_+\tilde\Pi_++\lambda_-\tilde\Pi_-. $$
Nilpotents. Choose the transverse pure element
$$ \tilde Q_1=i\vec e_y-\sqrt2\,\vec e_z, \qquad \tilde Q_1^2=-e_t. $$
The construction gives
$$ u=\tilde\Pi_-\tilde Q_1\tilde\Pi_+, \qquad v=\tilde\Pi_+\tilde Q_1\tilde\Pi_-, \qquad uv=-\tilde\Pi_-, \qquad vu=-\tilde\Pi_+. $$
With $\alpha=i$,
$$ \vec\Upsilon_+=\frac{u}{i} \approx (-0.1940873+0.1515388\,i)\vec e_x+(1.1912523+0.0850970\,i)\vec e_y+(-0.0601727+1.1958960\,i)\vec e_z, $$
$$ \vec\Upsilon_-=\frac{v}{i} \approx (0.1940873-0.1515388\,i)\vec e_x+(-0.1912523-0.0850970\,i)\vec e_y+(0.0601727+0.2183176\,i)\vec e_z. $$
The checks hold for the computed values to machine precision:
$$ \vec\Upsilon_+\vec\Upsilon_-=\tilde\Pi_-, \qquad \vec\Upsilon_-\vec\Upsilon_+=\tilde\Pi_+, \qquad \vec\Upsilon_+^2=\vec\Upsilon_-^2=0. $$
The nilpotents are not simple coordinate expressions, because the direction of $\vec Q$ is not aligned with an axis. They are computed from the transverse-vector construction, which is the general construction.
The eigenbasis. The basis is
$$ (\tilde\Pi_+,\vec\Upsilon_-,\vec\Upsilon_+,\tilde\Pi_-), $$
with the diagonal action
$$ \operatorname{diag}(\lambda_+,\lambda_+,\lambda_-,\lambda_-), $$
and the $2\times2$ model is
$$ \mathsf M_2^Q(\tilde Q)=\operatorname{diag}(\lambda_+,\lambda_-). $$
Spinors. A general element is $\tilde Q_2=Q_{2,11}\tilde\Pi_++Q_{2,12}\vec\Upsilon_-+Q_{2,21}\vec\Upsilon_++Q_{2,22}\tilde\Pi_-$, with matrix $$ \mathsf M_2^Q(\tilde Q_2)=\begin{pmatrix}Q_{2,11}&Q_{2,12}\\Q_{2,21}&Q_{2,22}\end{pmatrix} $$ in the eigenbasis $(\tilde\Pi_+,\vec\Upsilon_-,\vec\Upsilon_+,\tilde\Pi_-)$. Its two columns are the two spinors: $$ \zeta_+=Q_{2,11}\tilde\Pi_++Q_{2,21}\vec\Upsilon_+\longleftrightarrow\begin{pmatrix}Q_{2,11}\\Q_{2,21}\end{pmatrix}, \qquad \zeta_-=Q_{2,12}\vec\Upsilon_-+Q_{2,22}\tilde\Pi_-\longleftrightarrow\begin{pmatrix}Q_{2,12}\\Q_{2,22}\end{pmatrix}, $$ with matrices $$ \mathsf M_2^Q(\zeta_+)=\begin{pmatrix}Q_{2,11}&0\\Q_{2,21}&0\end{pmatrix}, \qquad \mathsf M_2^Q(\zeta_-)=\begin{pmatrix}0&Q_{2,12}\\0&Q_{2,22}\end{pmatrix}. $$ Left multiplication by $\tilde Q$ multiplies both columns by $\operatorname{diag}(\lambda_+,\lambda_-)$: $$ (Q_{2,11},Q_{2,21})\longmapsto(\lambda_+Q_{2,11},\lambda_-Q_{2,21}), \qquad (Q_{2,12},Q_{2,22})\longmapsto(\lambda_+Q_{2,12},\lambda_-Q_{2,22}). $$ In the special case $\tilde Q_2=(1-\tfrac{i}{2})e_t+\tfrac{1+i}{\sqrt2}\vec e_x+\tfrac{1}{\sqrt2}\vec e_y+\tfrac{i}{2}\vec e_z$ one has $Q_{2,11}=\lambda_+=1-\tfrac{i}{2}+iB$, $Q_{2,21}=0$, $Q_{2,12}=0$, $Q_{2,22}=\lambda_-=1-\tfrac{i}{2}-iB$, so $$ \mathsf M_2^Q(\tilde Q)=\begin{pmatrix}1-\tfrac{i}{2}+iB&0\\0&1-\tfrac{i}{2}-iB\end{pmatrix}, \qquad \mathsf M_2^Q(\zeta_+)=\begin{pmatrix}1-\tfrac{i}{2}+iB&0\\0&0\end{pmatrix}, \qquad \mathsf M_2^Q(\zeta_-)=\begin{pmatrix}0&0\\0&1-\tfrac{i}{2}-iB\end{pmatrix}. $$
Matrix. In the fixed model,
$$ \mathsf M_2(\tilde Q)= \begin{pmatrix} \frac32-\frac{i}{2}&-\frac{i}{\sqrt2}\\ \frac{2-i}{\sqrt2}&\frac12-\frac{i}{2} \end{pmatrix}, \qquad \mathrm{tr}\,\mathsf M_2(\tilde Q)=2-i, \qquad \det=1. $$
The spectral projectors are
$$ \rho_+=\frac{1}{2B} \begin{pmatrix} B-\frac{i}{2}&-\frac{1}{\sqrt2}\\ -\frac{1+2i}{\sqrt2}&B+\frac{i}{2} \end{pmatrix}, \qquad \rho_-=I_2-\rho_+, $$
and $\mathsf M_2(\tilde Q)=\lambda_+\rho_++\lambda_-\rho_-$.
Consequences. The square is
$$ \tilde Q^2=\left(\frac12-2i\right)e_t+\frac{3+i}{\sqrt2}\vec e_x+\frac{2-i}{\sqrt2}\vec e_y+\left(\frac12+i\right)\vec e_z, $$
and the inverse is the conjugate, because the norm is one:
$$ \tilde Q^{-1}=\left(1-\frac{i}{2}\right)e_t-\frac{1+i}{\sqrt2}\vec e_x-\frac{1}{\sqrt2}\vec e_y-\frac{i}{2}\vec e_z. $$
Case 8: A Null Vector Part and a Unipotent Unit
Coordinates. For
$$ \tilde Q=(1+i)e_t+\vec e_x+i\vec e_y, $$
the coordinates are
$$ Q_t=1+i, \qquad Q_x=1, \qquad Q_y=i, \qquad Q_z=0, \qquad B=0, \qquad \langle\tilde Q\rangle_{\natural}=2i. $$
The Jordan regime. The vector part is nonzero and null:
$$ Q_x^2+Q_y^2=1+i^2=0. $$
Write
$$ \tilde N=\vec e_x+i\vec e_y. $$
Then
$$ \tilde N^2=-e_t+0+e_t=0, \qquad \tilde Q=(1+i)e_t+\tilde N. $$
The only eigenvalue is $1+i$, the minimal polynomial is $(X-(1+i))^2$, and the eigenspace is $\ker L_{\tilde N}$, of dimension two. There is no eigenbasis. The element is invertible, with inverse
$$ \tilde Q^{-1}=\frac{1-i}{2}e_t+\frac{i}{2}\vec e_x-\frac12\vec e_y. $$
Spinors. There is no distinguished spinor pair: $B=0$, the two eigenvalues coincide, and the eigenspace is two-dimensional, so $\tilde Q$ selects no pair of complementary minimal left ideals. A general element still has two spinor components in the fixed Peirce basis. With $$ \mathsf M_2(\tilde Q_2)=\begin{pmatrix}Q_{2,11}&Q_{2,12}\\Q_{2,21}&Q_{2,22}\end{pmatrix}, $$ the first column $(Q_{2,11},Q_{2,21})$ is the spinor in the first Peirce carrier and the second column $(Q_{2,12},Q_{2,22})$ is the spinor in the second. Their matrices are $$ \mathsf M_2(\zeta_+)=\begin{pmatrix}Q_{2,11}&0\\Q_{2,21}&0\end{pmatrix}, \qquad \mathsf M_2(\zeta_-)=\begin{pmatrix}0&Q_{2,12}\\0&Q_{2,22}\end{pmatrix}. $$ In the special case $$ \mathsf M_2(\tilde Q)=\begin{pmatrix}1+i&-2i\\0&1+i\end{pmatrix}, $$ the two columns are $(1+i,\,0)$ and $(-2i,\,1+i)$. The first column is an eigenvector with eigenvalue $1+i$; the second is not. The eigenspace is one line, so there is one eigenspinor and no second one, and no spinor decomposition of $\tilde Q$.
Matrix. In the fixed model,
$$ \mathsf M_2(\tilde Q)= \begin{pmatrix} 1+i&-2i\\ 0&1+i \end{pmatrix} =(1+i)I_2+\mathsf N, \qquad \mathsf N= \begin{pmatrix} 0&-2i\\ 0&0 \end{pmatrix}, \qquad \mathsf N^2=0. $$
The matrix is a single Jordan block. The kernel of $\mathsf N$ is the line spanned by $(1,0)$, and the eigenspace of $\mathsf M_2(\tilde Q)$ under left multiplication is the two-dimensional space of matrices whose two columns lie on that line, that is, whose second row is zero.
Consequences. The square is
$$ \tilde Q^2=2i\,e_t+2(1+i)\vec e_x+2(i-1)\vec e_y, $$
and the general power is
$$ \tilde Q^n=(1+i)^n\left(e_t+\frac{n}{1+i}\tilde N\right). $$
The vector part is null, but the norm is $2i\neq0$: the element is $\natural$-regular. A null vector part is not a null element.
Summary
Main Results
| Case | $\tilde Q$ | $B$ | $\lambda_+$ | $\lambda_-$ | $\langle\tilde Q\rangle_{\natural}$ | Regime | Eigenbasis |
|---|---|---|---|---|---|---|---|
| $1$ | $e_t$ | $0$ | $1$ | $1$ | $1$ | scalar | every basis |
| $2$ | $\vec e_x$ | $1$ | $i$ | $-i$ | $1$ | distinct | yes |
| $3$ | $\frac{1}{\sqrt2}(e_t+\vec e_x)$ | $\frac{1}{\sqrt2}$ | $\frac{1+i}{\sqrt2}$ | $\frac{1-i}{\sqrt2}$ | $1$ | distinct | yes |
| $4$ | $\sqrt2\,e_t+i\vec e_x$ | $i$ | $\sqrt2-1$ | $\sqrt2+1$ | $1$ | distinct | yes |
| $5$ | $\bigl(1-\frac{i}{\sqrt2}\bigr)e_t+\bigl(1+\frac{i}{\sqrt2}\bigr)\vec e_x$ | $1+\frac{i}{\sqrt2}$ | $a(1+i)$ | $b(1-i)$ | $1$ | distinct | yes |
| $6$ | $e_t+\vec e_y+\frac{i}{\sqrt2}\vec e_x+\frac{i}{\sqrt2}\vec e_z$ | $0$ | $1$ | $1$ | $1$ | Jordan | none |
| $7$ | $\bigl(1-\frac{i}{2}\bigr)e_t+\frac{1+i}{\sqrt2}\vec e_x+\frac{1}{\sqrt2}\vec e_y+\frac{i}{2}\vec e_z$ | $\sqrt{\frac14+i}$ | $1-\frac{i}{2}+iB$ | $1-\frac{i}{2}-iB$ | $1$ | distinct | yes |
| $8$ | $(1+i)e_t+\vec e_x+i\vec e_y$ | $0$ | $1+i$ | $1+i$ | $2i$ | Jordan | none |
The Eigenbasis Table
For the five spectrally regular cases the Peirce eigenbasis is $\mathcal{B}(\tilde Q)=(\tilde\Pi_+,\vec\Upsilon_-,\vec\Upsilon_+,\tilde\Pi_-)$, and the diagonal values are $(\lambda_+,\lambda_+,\lambda_-,\lambda_-)$.
| Case | $\tilde\Pi_+$ | $\vec\Upsilon_\pm$ | $\operatorname{diag}$ |
|---|---|---|---|
| $2$ | $\frac12(e_t-i\vec e_x)$ | $\vec\Upsilon_\pm=\frac12(\pm\vec e_z-i\vec e_y)$ | $(i,i,-i,-i)$ |
| $3$ | $\frac12(e_t-i\vec e_x)$ | $\vec\Upsilon_\pm=\frac12(\pm\vec e_z-i\vec e_y)$ | $\bigl(\frac{1+i}{\sqrt2},\frac{1+i}{\sqrt2},\frac{1-i}{\sqrt2},\frac{1-i}{\sqrt2}\bigr)$ |
| $4$ | $\frac12(e_t-i\vec e_x)$ | $\vec\Upsilon_\pm=\frac12(\pm\vec e_z-i\vec e_y)$ | $(\sqrt2-1,\sqrt2-1,\sqrt2+1,\sqrt2+1)$ |
| $5$ | $\frac12(e_t-i\vec e_x)$ | $\vec\Upsilon_\pm=\frac12(\pm\vec e_z-i\vec e_y)$ | $(a(1+i),a(1+i),b(1-i),b(1-i))$ |
| $7$ | $\frac12(e_t+\vec{Ζ})$ | computed from $\tilde Q_1=i\vec e_y-\sqrt2\vec e_z$ | $(\lambda_+,\lambda_+,\lambda_-,\lambda_-)$ |
For cases 1, 6 and 8 there is no eigenbasis attached to the element: case 1 is scalar, and cases 6 and 8 are Jordan.
The Matrices
| Case | $\mathsf M_2(\tilde Q)$ | $\mathrm{tr}\,\mathsf M_2(\tilde Q)$ | $\det$ |
|---|---|---|---|
| $1$ | $I_2$ | $2$ | $1$ |
| $2$ | $\begin{pmatrix}0&-i\\-i&0\end{pmatrix}$ | $0$ | $1$ |
| $3$ | $\frac{1}{\sqrt2}\begin{pmatrix}1&-i\\-i&1\end{pmatrix}$ | $\sqrt2$ | $1$ |
| $4$ | $\begin{pmatrix}\sqrt2&1\\1&\sqrt2\end{pmatrix}$ | $2\sqrt2$ | $1$ |
| $5$ | $\begin{pmatrix}1-\frac{i}{\sqrt2}&\frac{1}{\sqrt2}-i\\\frac{1}{\sqrt2}-i&1-\frac{i}{\sqrt2}\end{pmatrix}$ | $2-i\sqrt2$ | $1$ |
| $6$ | $\begin{pmatrix}1+\frac{1}{\sqrt2}&\frac{1}{\sqrt2}-1\\1+\frac{1}{\sqrt2}&1-\frac{1}{\sqrt2}\end{pmatrix}$ | $2$ | $1$ |
| $7$ | $\begin{pmatrix}\frac32-\frac{i}{2}&-\frac{i}{\sqrt2}\\\frac{2-i}{\sqrt2}&\frac12-\frac{i}{2}\end{pmatrix}$ | $2-i$ | $1$ |
| $8$ | $\begin{pmatrix}1+i&-2i\\0&1+i\end{pmatrix}$ | $2+2i$ | $2i$ |
What the Cases Show
- Cases 2 to 5 have the same vector direction $\vec e_x$. They therefore have the same root $\vec{Ζ}=-i\vec e_x$, the same eigenprojections $\tilde\Pi_\pm$, and the same nilpotents $\vec\Upsilon_\pm$. Only the two eigenvalues change. The eigenbasis is attached to the direction of the vector part, not to its magnitude or to the scalar part. This is the biquaternion form of the statement that the eigenbasis is attached to the root.
- Case 4 has $B=i$ and real eigenvalues. The root need not be real for the spectrum to be real. The sign of the root exchanges the two eigenvalues and the two projectors.
- Case 6 is the unipotent case with a real scalar part. The vector part is null, $B=0$, and the element is $e_t$ plus a square-zero element. The eigenspace has dimension two, so there is no eigenbasis.
- Case 8 is the unipotent case with a complex scalar part. The vector part is again null, the norm is $2i\neq0$, and the element is $\natural$-regular. A null vector part does not imply an $\natural$-degenerate element.
- None of the eight is $\natural$-degenerate. The $\natural$-degenerate regime with $B\neq0$ and $\langle\tilde Q\rangle_{\natural}=0$ is not represented, and neither is the nilpotent subcase $Q_t=0$, $B=0$. Case 8 is the closest miss: its vector part is null, but its scalar part keeps the norm nonzero.
- Case 1 is the scalar regime. Every basis is an eigenbasis, and the construction of the eigenbasis does not apply.
- Case 7 is the generic direction. The nilpotents are not aligned with the axes, and they are computed from the transverse-vector construction. The symmetric closed form of The Eigenbasis of a Spectrally Regular Biquaternion, §The Transverse-Vector Construction, gives an admissible pair up to the reciprocal rescaling when the direction is off the exceptional cone; the transverse-vector construction is the one used here.
- In the distinct regime the eigenvalues repeat twice in the four-dimensional eigenbasis, and they are distinct on each two-dimensional spinor space. The $2\times2$ model records the two values once, one for each column.
Summary of Notation
| Symbol | Meaning |
|---|---|
| $\mathbb{B}$ | the biquaternion algebra, $\mathbb{B}=\mathbb{C}\otimes_{\mathbb{R}}\mathbb{H}\cong M_2(\mathbb{C})$ |
| $\tilde Q=Q_te_t+Q_x\vec e_x+Q_y\vec e_y+Q_z\vec e_z$ | a biquaternion |
| $B=\bigl(Q_x^2+Q_y^2+Q_z^2\bigr)^{1/2}$ | a fixed square root of the spectral discriminant $B^2$ |
| $\lambda_\pm=Q_t\pm iB$ | the two eigenvalues |
| $\vec{Ζ}=-i\vec Q/B$ | the pure root of $+1$ attached to $\tilde Q$ |
| $\tilde\Pi_\pm=\frac12(e_t\pm\vec{Ζ})$ | the two eigenprojections |
| $\mathbb{B}_{st}=\tilde\Pi_s\mathbb{B}\tilde\Pi_t$ | the Peirce corners |
| $\vec\Upsilon_\pm$ | the nilpotents of the mixed Peirce corners, normalised by $\vec\Upsilon_+\vec\Upsilon_-=\tilde\Pi_-$ |
| $\mathcal{B}(\tilde Q)$ | the Peirce eigenbasis $(\tilde\Pi_+,\vec\Upsilon_-,\vec\Upsilon_+,\tilde\Pi_-)$ |
| $S_\pm=\mathbb{B}\tilde\Pi_\pm$ | the two Peirce carriers, the minimal left ideals, the spinor spaces |
| $\mathsf M_2$ | the fixed $2\times2$ matrix model |
| $\mathsf M_2^Q$ | the eigenbasis matrix model |
| $\rho_\pm=\mathsf M_2(\tilde\Pi_\pm)$ | the spectral projectors |
| $L_{\tilde Q},R_{\tilde Q}$ | left and right multiplication by $\tilde Q$ |
| $N=\langle\tilde Q\rangle_{\natural}=Q_t^2+Q_x^2+Q_y^2+Q_z^2$ | the norm |
Further Reading
- The Biquaternion Eigenvalue Decomposition, §The Eigenprojectors, for the eigenvalue decomposition and the eigenprojections.
- The Spectrum, for the spectral criterion, the two eigenvalues and the existence of the eigenbasis.
- The Eigenbasis of a Spectrally Regular Biquaternion, for the construction of the eigenbasis, the nilpotent pair and the normalisation.
- The N-Regular, Spectrally Regular Sector, for the generic sector $\Sigma_{11}$, the Peirce corners and the spinor spaces.
- The N-Regular, Spectrally Degenerate Sector, for the scalar line and the Jordan part, where the construction stops.
- The Eigenbasis of a Spectrally Regular Biquaternion, for the fixed model, the eigenbasis model, the diagonalising matrix and the $4\times4$ regular matrices.
- Worked Examples of the Complete Polar Decomposition of an N-Regular Biquaternion, for the same eight elements read through the polar word.
- Worked Examples of Polar Decomposition with Eigenbasis of Biquaternions, for the same eight cases read in the eigenbasis, with the nilpotents and the spinors.
- Worked Examples of Eigenbasis of a Split-Complex Number, for the two-dimensional commutative companion.
- Worked Examples of Eigenbasis of a Reduced Biquaternion, for the four-dimensional commutative companion.
- Peirce Decompositions of the Biquaternion Algebra, for the Peirce corners.
- Biquaternion Elementary Functions, for the exponential and the logarithm built on the decomposition.
- Nilpotent Elements and the Jordan–Chevalley Decomposition, for the Jordan and nilpotent regimes.
- Pertti Lounesto, Clifford Algebras and Spinors (Cambridge University Press, 2nd ed. 2001), for the algebraic structure of the biquaternion algebra.
- Conventions in Mathematics, for the notation used throughout.