The Vector Parts of the Products over the Biquaternion Algebra
Introduction
The twelve products of Symmetric and Antisymmetric Decomposition of the Four General Products split every value into a scalar part and a vector part. The scalar parts are the four forms of The Four Forms and the Four Algebraic Norms over the Biquaternion Algebra, together with the zero form of the four antisymmetric products, and they are studied as forms. The vector parts are the other half, and this article treats them as what they are: vector-valued maps, one per product, with a table, a diagonal, an image, a symmetry and a behaviour on the remarkable subspaces.
The vector part is not a scalar form. It is a map into the three-dimensional complex vector subspace $\mathbb{B}_{V}$, and the scalar theory applies to its three contractions with the basis vectors. The article states this distinction first, then reads the twelve vector parts one by one, then develops the properties that are common to them. The two results that organise the table are the two-bit rule for the four general products and the three coincidences: there are twelve products and therefore twelve vector parts, but three identities make some of them equal, so only nine distinct maps occur on the complex space.
The article assumes the four general products of The Four General Products and Their Relations over the Biquaternion Algebra, the twelve products of Symmetric and Antisymmetric Decomposition of the Four General Products, and the four forms of The Four Forms and the Four Algebraic Norms over the Biquaternion Algebra. The element, the basis and the three conjugations are those of Introduction to the Algebra of Biquaternions; the remarkable subspaces are those of Subspaces and Decompositions; the two switches, the coefficient conjugation and the quaternion conjugation, are those of Introduction to the Operators of the Biquaternion Algebra. The diagonal of the plain product is also computed in The Square of an Element in the Biquaternion Algebra, the diagonal of the general quaternionic sesquilinear product in The Square of the General Quaternionic Sesquilinear Product and the Two Halves, and the diagonal of the block in The Vector Part of the Square and the Jacobi Failure of the Antisymmetric Plain Sesqualgebra.
Throughout, $\tilde Q_1=Q_{1t}e_t+\vec Q_1$ and $\tilde Q_2=Q_{2t}e_t+\vec Q_2$, with $Q_{1t},Q_{2t}\in\mathbb{C}$ and $\vec Q_1,\vec Q_2\in\mathbb{B}_{V}$. The coefficientwise conjugation is $\overline{\cdot}$, the quaternion conjugation is ${}^{\natural}$, the Hermitian conjugation is ${}^{*}=\overline{\cdot}\circ{}^{\natural}$, the complex bilinear dot product is $(\vec Q_1,\vec Q_2)=\sum_kQ_{1k}Q_{2k}$, and the complex bilinear cross product is $\vec Q_1\times\vec Q_2$. The vector subspace is $\mathbb{B}_{V}$, the centre is $\mathbb{B}_{C}$, the real quaternion subspace is $\mathbb{B}_{H}$, and the Hermitian and anti-Hermitian subspaces are $\mathbb{B}_{\Psi}$ and $\mathbb{B}_{\Phi}$. The products are written with their operators, as in the decomposition article: $\tilde Q_1\tilde Q_2$, $\tilde Q_1^{\natural}\tilde Q_2$, $\tilde Q_1\tilde Q_2^{*}$, $\tilde Q_1^{\natural}\tilde Q_2^{*}$ for the four general products, $\tilde Q_1\bullet\tilde Q_2$, $\tilde Q_1\bullet_{\natural}\tilde Q_2$, $\tilde Q_1\bullet_{*}\tilde Q_2$, $\tilde Q_1\bullet_{\natural*}\tilde Q_2$ for their symmetric parts, and $\tilde Q_1\wedge\tilde Q_2$, $\tilde Q_1\wedge_{\natural}\tilde Q_2$, $\tilde Q_1\wedge_{*}\tilde Q_2$, $\tilde Q_1\wedge_{\natural*}\tilde Q_2$ for their antisymmetric parts.
The Vector Part of a Product
Definition (the vector part). For a biquaternion $\tilde Q=Q_te_t+\vec Q$, the scalar part and the vector part are
$$ \mathrm{Sc}(\tilde Q)=Q_t,\qquad \mathrm{Vec}(\tilde Q)=\vec Q , $$
so that $\tilde Q=\mathrm{Sc}(\tilde Q)e_t+\mathrm{Vec}(\tilde Q)$. They are the two projections of the scalar–vector split, and both are $\mathbb{C}$-linear:
$$ \mathrm{Sc}(\tilde Q)=\tfrac12\bigl(\tilde Q+\tilde Q^{\natural}\bigr),\qquad \mathrm{Vec}(\tilde Q)=\tfrac12\bigl(\tilde Q-\tilde Q^{\natural}\bigr). $$
Definition (the vector part of a product). For a product $f$ of the twelve, the vector part of $f$ is the map
$$ V_f(\tilde Q_1,\tilde Q_2)=\mathrm{Vec}\bigl(f(\tilde Q_1,\tilde Q_2)\bigr). $$
The product is reconstructed from its two parts,
$$ f(\tilde Q_1,\tilde Q_2)=\mathrm{Sc}\bigl(f(\tilde Q_1,\tilde Q_2)\bigr)e_t+V_f(\tilde Q_1,\tilde Q_2), $$
and the scalar part is one of the four forms, while $V_f$ is the new object. The map $V_f$ is $\mathbb{C}$-bilinear when $f$ is one of the six products of the two bilinear families, and sesquilinear over $(\mathbb{C},\bar{\cdot})$ when $f$ is one of the six products of the two sesquilinear families. Its values lie in $\mathbb{B}_{V}$.
The vector part is not a form. A form of the corpus is scalar-valued: it is the scalar part of a product read in two arguments, bilinear in both slots for the two bilinear families and sesquilinear in the second slot for the two sesquilinear families. The vector part is vector-valued, so it has no scalar Gram matrix and no inertia of its own. What replaces the scalar form is a family of three scalar forms, one for each vector of the subspace: for $w\in\mathbb{B}_{V}$ put
$$ \varphi_{f,w}(\tilde Q_1,\tilde Q_2)=\bigl(w,V_f(\tilde Q_1,\tilde Q_2)\bigr). $$
For $f$ bilinear, $\varphi_{f,w}$ is a complex bilinear form; for $f$ sesquilinear, it is sesquilinear over $(\mathbb{C},\bar{\cdot})$. The three coordinate forms with $w=\vec e_x,\vec e_y,\vec e_z$ determine $V_f$, and $V_f$ determines them. This is the precise sense in which a vector part is a package of three scalar forms. The four forms of the corpus are a different object; they are not a member of this package of three.
The natural quadratic invariants. The scalar forms of $V_f$ are not the only scalars it carries. The complex bilinear square $(V_f,V_f)$ and the real square $(V_f,\overline{V_f})$ are the two natural scalar quantities attached to the vector part; the second is the Euclidean square of the vector. They are quartic in the arguments for a general product, and they are the natural invariants when no vector $w$ is chosen.
The Rule for the Four General Products
The four general products obey one rule. Put $\varepsilon=0$ for the two products whose second slot is plain, $\tilde Q_1\tilde Q_2$ and $\tilde Q_1^{\natural}\tilde Q_2$, and $\varepsilon=1$ for the two whose second slot is the Hermitian conjugation, $\tilde Q_1\tilde Q_2^{*}$ and $\tilde Q_1^{\natural}\tilde Q_2^{*}$. Write
$$ \vec Q_2^{\varepsilon}= \begin{cases} \vec Q_2, & \varepsilon=0,\\ \overline{\vec Q_2}, & \varepsilon=1, \end{cases} \qquad Q_{2t}^{\varepsilon}= \begin{cases} Q_{2t}, & \varepsilon=0,\\ \overline{Q_{2t}}, & \varepsilon=1. \end{cases} $$
Let $\alpha$ be the sign of the vector part of the second factor and $\beta$ the sign of the vector part of the first factor:
| product | first slot | second slot | $\varepsilon$ | $\alpha$ | $\beta$ | $\alpha\beta$ |
|---|---|---|---|---|---|---|
| $\tilde Q_1\tilde Q_2$ | plain | plain | $0$ | $+$ | $+$ | $+$ |
| $\tilde Q_1^{\natural}\tilde Q_2$ | quaternion | plain | $0$ | $+$ | $-$ | $-$ |
| $\tilde Q_1\tilde Q_2^{*}$ | plain | star | $1$ | $-$ | $+$ | $-$ |
| $\tilde Q_1^{\natural}\tilde Q_2^{*}$ | quaternion | star | $1$ | $-$ | $-$ | $+$ |
Theorem (the vector part of a general product). For the four general products,
$$ \boxed{\;V_f(\tilde Q_1,\tilde Q_2)=\alpha\,Q_{1t}\vec Q_2^{\varepsilon}+\beta\,Q_{2t}^{\varepsilon}\vec Q_1+\alpha\beta\,\vec Q_1\times\vec Q_2^{\varepsilon}\;} $$
and the scalar part is
$$ \mathrm{Sc}\bigl(f(\tilde Q_1,\tilde Q_2)\bigr)=Q_{1t}Q_{2t}^{\varepsilon}-\alpha\beta\,\bigl(\vec Q_1,\vec Q_2^{\varepsilon}\bigr). $$
Proof. The first factor is $\tilde Q_1^{\natural}$ for $\beta=-1$ and $\tilde Q_1$ for $\beta=+1$, so it is $Q_{1t}+\beta\vec Q_1$. The second factor is $\tilde Q_2$ for $\varepsilon=0$ and $\tilde Q_2^{*}=Q_{2t}^{\varepsilon}-\vec Q_2^{\varepsilon}$ for $\varepsilon=1$; in both cases it is $Q_{2t}^{\varepsilon}+\alpha\vec Q_2^{\varepsilon}$ with $\alpha=+1$ for $\varepsilon=0$ and $\alpha=-1$ for $\varepsilon=1$. The quaternion product of $Q_{1t}+\beta\vec Q_1$ and $Q_{2t}^{\varepsilon}+\alpha\vec Q_2^{\varepsilon}$ has scalar part $Q_{1t}Q_{2t}^{\varepsilon}-\alpha\beta(\vec Q_1,\vec Q_2^{\varepsilon})$ and vector part $\alpha Q_{1t}\vec Q_2^{\varepsilon}+\beta Q_{2t}^{\varepsilon}\vec Q_1+\alpha\beta\,\vec Q_1\times\vec Q_2^{\varepsilon}$, which is the display.
Remark (the two bits). The sign $\alpha$ is read on the second slot: it is $+$ for the plain and $-$ for the star. The sign $\beta$ is read on the first slot: it is $+$ for the plain and $-$ for the quaternion conjugation. The cross term carries the product $\alpha\beta$. The first mixed term carries $\alpha$, the second carries $\beta$, and the two are the two slot choices of The Four General Products and Their Relations over the Biquaternion Algebra read on the vector part rather than on the scalar form. The four general vector parts are therefore the four sign choices of two independent bits, not a list of four unrelated formulas.
The Table of the Twelve Vector Parts
The table lists the twelve products with their operators, their vector parts, their diagonals, their symmetries and their images. The diagonal is $V_f(\tilde Q,\tilde Q)$ at the single element $\tilde Q=Q_te_t+\vec Q$; the symmetry is the relation between $V_f(\tilde Q_1,\tilde Q_2)$ and $V_f(\tilde Q_2,\tilde Q_1)$, read with the coefficientwise conjugation in the sesquilinear families.
| product | operator | vector part | diagonal | symmetry | image |
|---|---|---|---|---|---|
| the general plain bilinear product | $\tilde Q_1\tilde Q_2$ | $Q_{1t}\vec Q_2+Q_{2t}\vec Q_1+\vec Q_1\times\vec Q_2$ | $2Q_t\vec Q$ | neither | $\mathbb{B}_{V}$ |
| the general quaternionic bilinear product | $\tilde Q_1^{\natural}\tilde Q_2$ | $Q_{1t}\vec Q_2-Q_{2t}\vec Q_1-\vec Q_1\times\vec Q_2$ | $0$ | alternating | $\mathbb{B}_{V}$ |
| the general plain sesquilinear product | $\tilde Q_1\tilde Q_2^{*}$ | $-Q_{1t}\overline{\vec Q_2}+\overline{Q_{2t}}\vec Q_1-\vec Q_1\times\overline{\vec Q_2}$ | $-Q_t\overline{\vec Q}+\overline{Q_t}\vec Q-\vec Q\times\overline{\vec Q}$ | conjugate-alternating | $\mathbb{B}_{V}$ |
| the general quaternionic sesquilinear product | $\tilde Q_1^{\natural}\tilde Q_2^{*}$ | $-Q_{1t}\overline{\vec Q_2}-\overline{Q_{2t}}\vec Q_1+\vec Q_1\times\overline{\vec Q_2}$ | $-Q_t\overline{\vec Q}-\overline{Q_t}\vec Q+\vec Q\times\overline{\vec Q}$ | neither | $\mathbb{B}_{V}$ |
| the symmetric plain bilinear product | $\tilde Q_1\bullet\tilde Q_2$ | $Q_{1t}\vec Q_2+Q_{2t}\vec Q_1$ | $2Q_t\vec Q$ | symmetric | $\mathbb{B}_{V}$ |
| the symmetric quaternionic bilinear product | $\tilde Q_1\bullet_{\natural}\tilde Q_2$ | $0$ | $0$ | zero | $\{0\}$ |
| the symmetric plain sesquilinear product | $\tilde Q_1\bullet_{*}\tilde Q_2$ | $0$ | $0$ | zero | $\{0\}$ |
| the symmetric quaternionic sesquilinear product | $\tilde Q_1\bullet_{\natural*}\tilde Q_2$ | $-Q_{1t}\overline{\vec Q_2}-\overline{Q_{2t}}\vec Q_1$ | $-Q_t\overline{\vec Q}-\overline{Q_t}\vec Q$ | conjugate-symmetric | $\mathbb{B}_{V}$ |
| the antisymmetric plain bilinear product | $\tilde Q_1\wedge\tilde Q_2$ | $\vec Q_1\times\vec Q_2$ | $0$ | alternating | $\mathbb{B}_{V}$ |
| the antisymmetric quaternionic bilinear product | $\tilde Q_1\wedge_{\natural}\tilde Q_2$ | $Q_{1t}\vec Q_2-Q_{2t}\vec Q_1-\vec Q_1\times\vec Q_2$ | $0$ | alternating | $\mathbb{B}_{V}$ |
| the antisymmetric plain sesquilinear product | $\tilde Q_1\wedge_{*}\tilde Q_2$ | $-Q_{1t}\overline{\vec Q_2}+\overline{Q_{2t}}\vec Q_1-\vec Q_1\times\overline{\vec Q_2}$ | $-Q_t\overline{\vec Q}+\overline{Q_t}\vec Q-\vec Q\times\overline{\vec Q}$ | conjugate-alternating | $\mathbb{B}_{V}$ |
| the antisymmetric quaternionic sesquilinear product | $\tilde Q_1\wedge_{\natural*}\tilde Q_2$ | $\vec Q_1\times\overline{\vec Q_2}$ | $\vec Q\times\overline{\vec Q}$ | conjugate-alternating | $\mathbb{B}_{V}$ |
The vector part column is $V_f$; the diagonal column is $V_f$ at the two equal arguments; the symmetry column is the class of the map under the exchange of the arguments, with the coefficientwise conjugation in the sesquilinear families; and the image column is the image of $V_f$ on the whole space. Ten rows have image the whole vector subspace, and two rows, $\tilde Q_1\bullet_{\natural}\tilde Q_2$ and $\tilde Q_1\bullet_{*}\tilde Q_2$, are the zero map. The table has repetitions: the rows $\tilde Q_1^{\natural}\tilde Q_2$ and $\tilde Q_1\wedge_{\natural}\tilde Q_2$ are identical, the rows $\tilde Q_1\tilde Q_2^{*}$ and $\tilde Q_1\wedge_{*}\tilde Q_2$ are identical, and the two zero rows are identical. The next section proves that these are the only repetitions on the complex space.
The Twelve Vector Parts and Their Coincidences
Theorem (the vector part loses the central symmetric parts). The twelve vector parts satisfy
$$ \boxed{\;V(\tilde Q_1^{\natural}\tilde Q_2)=V(\tilde Q_1\wedge_{\natural}\tilde Q_2),\qquad V(\tilde Q_1\tilde Q_2^{*})=V(\tilde Q_1\wedge_{*}\tilde Q_2),\qquad V(\tilde Q_1\bullet_{\natural}\tilde Q_2)=V(\tilde Q_1\bullet_{*}\tilde Q_2)=0\;} $$
and these three identities are the only coincidences among the twelve on the complex space. There are twelve vector parts, one to each product; as maps they fall into nine distinct classes. The count of nine is a count of distinct maps, not a replacement for the twelve: every product keeps its own vector part, including the two zero maps.
Proof. The symmetric part of the general quaternionic bilinear product is the quaternion form read as a product, $\tilde Q_1\bullet_{\natural}\tilde Q_2(\tilde Q_1,\tilde Q_2)=\langle\tilde Q_1,\tilde Q_2\rangle_{\natural}e_t$, which is central; its vector part is zero. Hence $V(\tilde Q_1^{\natural}\tilde Q_2)=V(\tilde Q_1\bullet_{\natural}\tilde Q_2)+V(\tilde Q_1\wedge_{\natural}\tilde Q_2)=V(\tilde Q_1\wedge_{\natural}\tilde Q_2)$. The symmetric part of the general plain sesquilinear product is the Hermitian form read as a product, $\tilde Q_1\bullet_{*}\tilde Q_2(\tilde Q_1,\tilde Q_2)=\langle\tilde Q_1,\tilde Q_2\rangle_{*}e_t$, which is central; hence $V(\tilde Q_1\tilde Q_2^{*})=V(\tilde Q_1\wedge_{*}\tilde Q_2)$. The two central products themselves have zero vector part. The nine classes are separated by the three pairs in the next table.
Theorem (the separation). The nine values in the table are pairwise distinct, and the three pairs $(e_t,\vec e_x)$, $(\vec e_x,\vec e_y)$ and $(\vec e_x,i\vec e_y)$ separate them.
| distinct vector part | products | $(e_t,\vec e_x)$ | $(\vec e_x,\vec e_y)$ | $(\vec e_x,i\vec e_y)$ |
|---|---|---|---|---|
| $\tilde Q_1\tilde Q_2$ | $\tilde Q_1\tilde Q_2$ | $\vec e_x$ | $\vec e_z$ | $i\vec e_z$ |
| $\tilde Q_1^{\natural}\tilde Q_2=\tilde Q_1\wedge_{\natural}\tilde Q_2$ | $\tilde Q_1^{\natural}\tilde Q_2$, $\tilde Q_1\wedge_{\natural}\tilde Q_2$ | $\vec e_x$ | $-\vec e_z$ | $-i\vec e_z$ |
| $\tilde Q_1\tilde Q_2^{*}=\tilde Q_1\wedge_{*}\tilde Q_2$ | $\tilde Q_1\tilde Q_2^{*}$, $\tilde Q_1\wedge_{*}\tilde Q_2$ | $-\vec e_x$ | $-\vec e_z$ | $i\vec e_z$ |
| $\tilde Q_1^{\natural}\tilde Q_2^{*}$ | $\tilde Q_1^{\natural}\tilde Q_2^{*}$ | $-\vec e_x$ | $\vec e_z$ | $-i\vec e_z$ |
| $\tilde Q_1\bullet\tilde Q_2$ | $\tilde Q_1\bullet\tilde Q_2$ | $\vec e_x$ | $0$ | $0$ |
| $\tilde Q_1\bullet_{\natural*}\tilde Q_2$ | $\tilde Q_1\bullet_{\natural*}\tilde Q_2$ | $-\vec e_x$ | $0$ | $0$ |
| $\tilde Q_1\wedge\tilde Q_2$ | $\tilde Q_1\wedge\tilde Q_2$ | $0$ | $\vec e_z$ | $i\vec e_z$ |
| $\tilde Q_1\wedge_{\natural*}\tilde Q_2$ | $\tilde Q_1\wedge_{\natural*}\tilde Q_2$ | $0$ | $\vec e_z$ | $-i\vec e_z$ |
| $0$ | $\tilde Q_1\bullet_{\natural}\tilde Q_2$, $\tilde Q_1\bullet_{*}\tilde Q_2$ | $0$ | $0$ | $0$ |
Each row is different from every other, so the three pairs separate the nine classes. The table also shows how close the classes are: $\tilde Q_1\tilde Q_2$ and $\tilde Q_1^{\natural}\tilde Q_2$ differ only in the sign of the cross term and of the second mixed term; $\tilde Q_1\tilde Q_2^{*}$ and $\tilde Q_1^{\natural}\tilde Q_2^{*}$ differ in the sign of the second mixed term and of the cross term; $\tilde Q_1\bullet\tilde Q_2$ and $\tilde Q_1\bullet_{\natural*}\tilde Q_2$ differ in the sign and in the coefficientwise conjugation of the two mixed terms; and $\tilde Q_1\wedge\tilde Q_2$ and $\tilde Q_1\wedge_{\natural*}\tilde Q_2$ differ only at the complex pair.
Corollary (the vector part does not determine the product). The map $f\mapsto V_f$ has non-singleton fibres $\{\tilde Q_1^{\natural}\tilde Q_2,\tilde Q_1\wedge_{\natural}\tilde Q_2\}$, $\{\tilde Q_1\tilde Q_2^{*},\tilde Q_1\wedge_{*}\tilde Q_2\}$ and $\{\tilde Q_1\bullet_{\natural}\tilde Q_2,\tilde Q_1\bullet_{*}\tilde Q_2\}$, and singleton fibres on the other six products. In each fibre the two products have different scalar parts: $N$ against $0$ for $\{\tilde Q_1^{\natural}\tilde Q_2,\tilde Q_1\wedge_{\natural}\tilde Q_2\}$, $H$ against $0$ for $\{\tilde Q_1\tilde Q_2^{*},\tilde Q_1\wedge_{*}\tilde Q_2\}$, and $N$ against $H$ for $\{\tilde Q_1\bullet_{\natural}\tilde Q_2,\tilde Q_1\bullet_{*}\tilde Q_2\}$. So the pair $(\mathrm{Sc}\circ f,V_f)$ determines $f$, while $V_f$ alone does not. The vector part is one half of the product; the scalar part is the other, and the two are needed together.
The Four General Vector Parts
The General Plain Bilinear Product
The vector part is
$$ V(\tilde Q_1\tilde Q_2)=Q_{1t}\vec Q_2+Q_{2t}\vec Q_1+\vec Q_1\times\vec Q_2. $$
It is $\mathbb{C}$-bilinear and neither symmetric nor alternating: its symmetric part is $V(\tilde Q_1\bullet\tilde Q_2)=Q_{1t}\vec Q_2+Q_{2t}\vec Q_1$ and its alternating part is $V(\tilde Q_1\wedge\tilde Q_2)=\vec Q_1\times\vec Q_2$, so $V(\tilde Q_1\tilde Q_2)=V(\tilde Q_1\bullet\tilde Q_2)+V(\tilde Q_1\wedge\tilde Q_2)$. Its diagonal is $2Q_t\vec Q$, which vanishes exactly on the union of the centre and the vector subspace; it is the vector part of the square computed in The Square of an Element in the Biquaternion Algebra. Its image is all of $\mathbb{B}_{V}$, since at $\tilde Q_1=e_t$ and $\tilde Q_2=\vec w$ it is $\vec w$. On pure vectors it is the cross product $\vec Q_1\times\vec Q_2$. Its scalar part is the plain form $P$. When one argument is central, only the mixed term of the other survives, and the vector part is a complex multiple of the other vector part.
The General Quaternionic Bilinear Product
The vector part is
$$ V(\tilde Q_1^{\natural}\tilde Q_2)=Q_{1t}\vec Q_2-Q_{2t}\vec Q_1-\vec Q_1\times\vec Q_2. $$
It is $\mathbb{C}$-bilinear and alternating, $V(\tilde Q_2^{\natural}\tilde Q_1)=-V(\tilde Q_1^{\natural}\tilde Q_2)$, and its diagonal is zero. It equals $V(\tilde Q_1\wedge_{\natural}\tilde Q_2)$ because the symmetric part $\tilde Q_1\bullet_{\natural}\tilde Q_2$ is central. Its image is all of $\mathbb{B}_{V}$. On pure vectors it is $-\vec Q_1\times\vec Q_2$. It is the reflection of the plain vector part in $Q_{1t}\vec Q_2$,
$$ V(\tilde Q_1^{\natural}\tilde Q_2)=2Q_{1t}\vec Q_2-V(\tilde Q_1\tilde Q_2),\qquad V(\tilde Q_1\tilde Q_2)+V(\tilde Q_1^{\natural}\tilde Q_2)=2Q_{1t}\vec Q_2, $$
which is the vector form of the splitting identity of The Four General Products and Their Relations over the Biquaternion Algebra. Its scalar part is the quaternionic form $N$. The square of the product is central, $N(\tilde Q)e_t$, and the vanishing of the diagonal is that statement. It is the only general product whose vector part is alternating.
The General Plain Sesquilinear Product
The vector part is
$$ V(\tilde Q_1\tilde Q_2^{*})=-Q_{1t}\overline{\vec Q_2}+\overline{Q_{2t}}\vec Q_1-\vec Q_1\times\overline{\vec Q_2}. $$
It is sesquilinear over $(\mathbb{C},\bar{\cdot})$ and conjugate-alternating,
$$ V(\tilde Q_1\tilde Q_2^{*})=-\overline{V(\tilde Q_2\tilde Q_1^{*})}. $$
Its diagonal is the vector part of the Hermitian square $\tilde Q\tilde Q^{*}$, hence a purely imaginary vector; it vanishes exactly when $\tilde Q\tilde Q^{*}$ is central, that is, when the Hermitian square has no vector part. It equals $V(\tilde Q_1\wedge_{*}\tilde Q_2)$ because $\tilde Q_1\bullet_{*}\tilde Q_2$ is central. Its image is all of $\mathbb{B}_{V}$. On pure vectors it is $-\vec Q_1\times\overline{\vec Q_2}$. It is the plain vector part read at the conjugated second argument, which is the vector form of the slot identity. Its scalar part is the Hermitian form $H$.
The General Quaternionic Sesquilinear Product
The vector part is
$$ V(\tilde Q_1^{\natural}\tilde Q_2^{*})=-Q_{1t}\overline{\vec Q_2}-\overline{Q_{2t}}\vec Q_1+\vec Q_1\times\overline{\vec Q_2}. $$
It is sesquilinear and neither conjugate-symmetric nor conjugate-alternating. It splits into its conjugate-symmetric and conjugate-alternating halves,
$$ V(\tilde Q_1^{\natural}\tilde Q_2^{*})=V(\tilde Q_1\bullet_{\natural*}\tilde Q_2)+V(\tilde Q_1\wedge_{\natural*}\tilde Q_2), \qquad V(\tilde Q_1\bullet_{\natural*}\tilde Q_2)=-Q_{1t}\overline{\vec Q_2}-\overline{Q_{2t}}\vec Q_1, \qquad V(\tilde Q_1\wedge_{\natural*}\tilde Q_2)=\vec Q_1\times\overline{\vec Q_2}. $$
Its diagonal is $-Q_t\overline{\vec Q}-\overline{Q_t}\vec Q+\vec Q\times\overline{\vec Q}$; its real part is the real vector $-2(a\vec p+b\vec q)$ and its imaginary part is $-2i\,\vec p\times\vec q$ for $Q_t=a+ib$ and $\vec Q=\vec p+i\vec q$, so the two parts are orthogonal. Its image is all of $\mathbb{B}_{V}$. On pure vectors it is $\vec Q_1\times\overline{\vec Q_2}$. It is the quaternionic vector part read on $(\tilde Q_1,\tilde Q_2^{*})$,
$$ V(\tilde Q_1^{\natural}\tilde Q_2^{*})=V(\tilde Q_1^{\natural}\tilde Q_2^{*}), $$
and it satisfies $V(\tilde Q_1\tilde Q_2^{*})+V(\tilde Q_1^{\natural}\tilde Q_2^{*})=-2Q_{1t}\overline{\vec Q_2}$. Its scalar part is the Krein form $\mathrm{Kr}$. It is the only general product whose vector part has both a conjugate-symmetric and a conjugate-alternating piece.
The Four Symmetric Vector Parts
The Symmetric Plain Bilinear Product
The vector part is
$$ V(\tilde Q_1\bullet\tilde Q_2)=Q_{1t}\vec Q_2+Q_{2t}\vec Q_1. $$
It is symmetric, $V(\tilde Q_1\bullet\tilde Q_2)=V(\tilde Q_2\bullet\tilde Q_1)$, and it is the symmetrisation of the plain vector part,
$$ V(\tilde Q_1\bullet\tilde Q_2)=\tfrac12\bigl(V(\tilde Q_1\tilde Q_2)+V(\tilde Q_2\tilde Q_1)\bigr). $$
Its diagonal is $2Q_t\vec Q$, the same as that of $\tilde Q_1\tilde Q_2$; it vanishes on the union of the centre and the vector subspace. Its image is all of $\mathbb{B}_{V}$. On pure vectors it vanishes identically. Its scalar part is the plain form $P$. The zero set is $Q_{1t}\vec Q_2=-Q_{2t}\vec Q_1$; off the centre and the vector subspace it forces the two vector parts to be proportional.
The Symmetric Quaternionic Bilinear Product
The vector part is zero:
$$ V(\tilde Q_1\bullet_{\natural}\tilde Q_2)=0. $$
The product is central-valued, $\tilde Q_1\bullet_{\natural}\tilde Q_2(\tilde Q_1,\tilde Q_2)=\langle\tilde Q_1,\tilde Q_2\rangle_{\natural}e_t$, the quaternion form read as a product. The vector part has zero diagonal, zero image and zero values on pure vectors. It is one of the two products of the twelve whose vector part vanishes identically.
The Symmetric Plain Sesquilinear Product
The vector part is zero:
$$ V(\tilde Q_1\bullet_{*}\tilde Q_2)=0. $$
The product is central-valued, $\tilde Q_1\bullet_{*}\tilde Q_2(\tilde Q_1,\tilde Q_2)=\langle\tilde Q_1,\tilde Q_2\rangle_{*}e_t$, the Hermitian form read as a product. The vector part has zero diagonal, zero image and zero values on pure vectors. It is the second of the two products whose vector part vanishes identically.
The Symmetric Quaternionic Sesquilinear Product
The vector part is
$$ V(\tilde Q_1\bullet_{\natural*}\tilde Q_2)=-Q_{1t}\overline{\vec Q_2}-\overline{Q_{2t}}\vec Q_1. $$
It is conjugate-symmetric,
$$ V(\tilde Q_1\bullet_{\natural*}\tilde Q_2)=\overline{V(\tilde Q_2\bullet_{\natural*}\tilde Q_1)}, $$
and it is the conjugate-symmetric half of the general quaternionic sesquilinear vector part,
$$ V(\tilde Q_1\bullet_{\natural*}\tilde Q_2)=\tfrac12\bigl(V(\tilde Q_1^{\natural}\tilde Q_2^{*})+\overline{V(\tilde Q_2^{\natural}\tilde Q_1^{*})}\bigr). $$
Its diagonal is $-Q_t\overline{\vec Q}-\overline{Q_t}\vec Q$, a real vector. It vanishes exactly when $Q_t=0$ or $\vec Q=iQ_t\vec r$ with $\vec r$ a real vector. Its image is all of $\mathbb{B}_{V}$, and its diagonal image is the real vector subspace $\mathrm{Vect}_{\mathbb{R}}(\mathbb{B})$. On pure vectors it vanishes. Its scalar part is the Krein form $\mathrm{Kr}$; the values of the product itself lie in no remarkable subspace of Subspaces and Decompositions.
The Four Antisymmetric Vector Parts
The Antisymmetric Plain Bilinear Product
The vector part is
$$ V(\tilde Q_1\wedge\tilde Q_2)=\vec Q_1\times\vec Q_2. $$
It is alternating and satisfies the Jacobi identity; it is the Lie bracket of the vector subspace, which is isomorphic to $\mathfrak{sl}(2,\mathbb{C})$, and the article that develops it is The Lie Algebra of the Antisymmetric Plain Algebra. Its diagonal is zero, its image is all of $\mathbb{B}_{V}$, and on pure vectors it is the cross product. Its scalar part is zero. Its quadratic invariant is the Lagrange identity
$$ \bigl(\vec Q_1\times\vec Q_2,\vec Q_1\times\vec Q_2\bigr) =\bigl(\vec Q_1,\vec Q_1\bigr)\bigl(\vec Q_2,\vec Q_2\bigr)-\bigl(\vec Q_1,\vec Q_2\bigr)^2, $$
and it vanishes exactly when the two vector parts are linearly dependent.
The Antisymmetric Quaternionic Bilinear Product
The vector part is
$$ V(\tilde Q_1\wedge_{\natural}\tilde Q_2)=Q_{1t}\vec Q_2-Q_{2t}\vec Q_1-\vec Q_1\times\vec Q_2. $$
It is alternating, and it equals $V(\tilde Q_1^{\natural}\tilde Q_2)$ because the symmetric part $\tilde Q_1\bullet_{\natural}\tilde Q_2$ is central. Its diagonal is zero, its image is all of $\mathbb{B}_{V}$, and on pure vectors it is $-\vec Q_1\times\vec Q_2$. Its scalar part is zero. It fails the Jacobi identity at $(e_t,\vec e_x,\vec e_y)$ with cyclic sum $-\vec e_z$, as The Jacobi Failure and the Associator Defect of the Antisymmetric Quaternionic Algebra records. It is the antisymmetric part of the quaternionic bilinear product.
The Antisymmetric Plain Sesquilinear Product
The vector part is
$$ V(\tilde Q_1\wedge_{*}\tilde Q_2)=-Q_{1t}\overline{\vec Q_2}+\overline{Q_{2t}}\vec Q_1-\vec Q_1\times\overline{\vec Q_2}. $$
It is conjugate-alternating, and it equals $V(\tilde Q_1\tilde Q_2^{*})$ because the symmetric part $\tilde Q_1\bullet_{*}\tilde Q_2$ is central. Its diagonal is the vector part of $\tilde Q\tilde Q^{*}$, a purely imaginary vector; its image is all of $\mathbb{B}_{V}$; on pure vectors it is $-\vec Q_1\times\overline{\vec Q_2}$. Its scalar part is zero. It fails the Jacobi identity at $(e_t,\vec e_x,\vec e_y)$ with cyclic sum $\vec e_z$ and at $(\vec e_x,\vec e_x,i\vec e_y)$ with $-2i\vec e_y$, as The Vector Part of the Square and the Jacobi Failure of the Antisymmetric Plain Sesqualgebra records. It is the antisymmetric part of the general plain sesquilinear product.
The Antisymmetric Quaternionic Sesquilinear Product
The vector part is
$$ V(\tilde Q_1\wedge_{\natural*}\tilde Q_2)=\vec Q_1\times\overline{\vec Q_2}. $$
It is conjugate-alternating. Its diagonal is
$$ \vec Q\times\overline{\vec Q}=-2i\,\vec p\times\vec q, \qquad \vec Q=\vec p+i\vec q, $$
a purely imaginary vector, and it vanishes exactly when $\vec Q$ is a complex multiple of a real vector. Its image is all of $\mathbb{B}_{V}$, and on pure vectors it is $\vec Q_1\times\overline{\vec Q_2}$. Its scalar part is zero. It fails the Jacobi identity at $(\vec e_x,\vec e_x,i\vec e_y)$ with $-2i\vec e_y$, as The Conjugate Cross Product and the Jacobi Failure of the Antisymmetric Quaternionic Sesqualgebra records. On the real subspace it equals $V(\tilde Q_1\wedge\tilde Q_2)$. It is the antisymmetric part of the general quaternionic sesquilinear product.
The Diagonal and the Square
Definition. The diagonal of a vector part is
$$ D_f(\tilde Q)=V_f(\tilde Q,\tilde Q)=\mathrm{Vec}\bigl(f(\tilde Q,\tilde Q)\bigr). $$
It is the vector part of the square of the element under $f$. The table groups the twelve diagonals by their value.
| products | diagonal | image |
|---|---|---|
| $\tilde Q_1\tilde Q_2$, $\tilde Q_1\bullet\tilde Q_2$ | $2Q_t\vec Q$ | $\mathbb{B}_{V}$ |
| $\tilde Q_1^{\natural}\tilde Q_2$, $\tilde Q_1\bullet_{\natural}\tilde Q_2$, $\tilde Q_1\bullet_{*}\tilde Q_2$, $\tilde Q_1\wedge\tilde Q_2$, $\tilde Q_1\wedge_{\natural}\tilde Q_2$ | $0$ | $\{0\}$ |
| $\tilde Q_1\tilde Q_2^{*}$, $\tilde Q_1\wedge_{*}\tilde Q_2$ | $-Q_t\overline{\vec Q}+\overline{Q_t}\vec Q-\vec Q\times\overline{\vec Q}$ | $i\,\mathrm{Vect}_{\mathbb{R}}(\mathbb{B})$ |
| $\tilde Q_1^{\natural}\tilde Q_2^{*}$ | $-Q_t\overline{\vec Q}-\overline{Q_t}\vec Q+\vec Q\times\overline{\vec Q}$ | $\{r+is:\ r,s\in\mathrm{Vect}_{\mathbb{R}}(\mathbb{B}),\ r\cdot s=0\}$ |
| $\tilde Q_1\bullet_{\natural*}\tilde Q_2$ | $-Q_t\overline{\vec Q}-\overline{Q_t}\vec Q$ | $\mathrm{Vect}_{\mathbb{R}}(\mathbb{B})$ |
| $\tilde Q_1\wedge_{\natural*}\tilde Q_2$ | $\vec Q\times\overline{\vec Q}$ | $i\,\mathrm{Vect}_{\mathbb{R}}(\mathbb{B})$ |
Proposition (homogeneity). The diagonal of a $\mathbb{C}$-bilinear product is homogeneous of degree two over $\mathbb{C}$, $D_f(\lambda\tilde Q)=\lambda^2D_f(\tilde Q)$, and the diagonal of a sesquilinear product is homogeneous of degree two in the real sense, $D_f(\lambda\tilde Q)=\lvert\lambda\rvert^2D_f(\tilde Q)$.
Proof. For a bilinear product, $f(\lambda\tilde Q,\lambda\tilde Q)=\lambda^2f(\tilde Q,\tilde Q)$; for a sesquilinear product, $f(\lambda\tilde Q,\lambda\tilde Q)=\lambda\overline{\lambda}f(\tilde Q,\tilde Q)=\lvert\lambda\rvert^2f(\tilde Q,\tilde Q)$. The vector part carries the scalar.
Proposition (the diagonal of the quaternionic sesquilinear product and orthogonality). Write $Q_t=a+ib$ and $\vec Q=\vec p+i\vec q$ with $a,b\in\mathbb{R}$ and $\vec p,\vec q\in\mathrm{Vect}_{\mathbb{R}}(\mathbb{B})$. Then
$$ V(\tilde Q^{\natural}\tilde Q^{*})=-2(a\vec p+b\vec q)-2i\,\vec p\times\vec q, $$
and its real and imaginary parts are orthogonal. Conversely, every pair of orthogonal real vectors is the real and imaginary parts of a value of $V(\tilde Q^{\natural}\tilde Q^{*})$.
Proof. The mixed terms give $-Q_t\overline{\vec Q}-\overline{Q_t}\vec Q=-2(a\vec p+b\vec q)$, which is real, and the cross term gives $\vec Q\times\overline{\vec Q}=-2i\,\vec p\times\vec q$, which is purely imaginary. The real part lies in the plane spanned by $\vec p,\vec q$, while the imaginary part is perpendicular to that plane, so the two are orthogonal. For the converse, let $r,s$ be orthogonal real vectors; take $\vec p=r/\lvert r\rvert$, $\vec q=-(\lvert s\rvert/2)(\hat s\times\hat p)$, $a=-\lvert r\rvert/2$ and $b=0$, and the display returns $r+is$; the degenerate cases follow by continuity and by the zero vector.
The zero sets of the diagonals. The diagonal of $\tilde Q_1\tilde Q_2$ or $\tilde Q_1\bullet\tilde Q_2$ vanishes exactly on the centre and on the vector subspace, $Q_t=0$ or $\vec Q=0$. The diagonals of $\tilde Q_1^{\natural}\tilde Q_2$, $\tilde Q_1\bullet_{\natural}\tilde Q_2$, $\tilde Q_1\bullet_{*}\tilde Q_2$, $\tilde Q_1\wedge\tilde Q_2$ and $\tilde Q_1\wedge_{\natural}\tilde Q_2$ vanish identically. The diagonal of $\tilde Q_1\bullet_{\natural*}\tilde Q_2$ vanishes exactly when $Q_t=0$ or $\vec Q=iQ_t\vec r$ with $\vec r$ real. The diagonal of $\tilde Q_1\wedge_{\natural*}\tilde Q_2$ vanishes exactly when $\vec Q$ is a complex multiple of a real vector. The diagonal of $\tilde Q_1\tilde Q_2^{*}$ or $\tilde Q_1\wedge_{*}\tilde Q_2$ vanishes exactly when $\tilde Q\tilde Q^{*}$ is central, that is, when the Hermitian square has no vector part. The diagonal of $\tilde Q_1^{\natural}\tilde Q_2^{*}$ vanishes exactly when the two equations $a\vec p+b\vec q=0$ and $\vec p\times\vec q=0$ hold.
Remark (the diagonal is not a form). The diagonal is a quadratic map, not a bilinear or sesquilinear form. This is the same distinction as for the vector part itself, read at the two equal arguments. The diagonal of $\tilde Q_1\tilde Q_2^{*}$ is the object studied in The Vector Part of the Square and the Jacobi Failure of the Antisymmetric Plain Sesqualgebra, and the diagonal of $\tilde Q_1^{\natural}\tilde Q_2^{*}$ is the vector part of the square studied in The Square of the General Quaternionic Sesquilinear Product and the Two Halves.
The Behaviour on the Subspaces
The centre. If both arguments are central, all twelve vector parts vanish. More generally, if one argument is central, every vector part collapses to a scalar multiple of the other vector part, and the multiplier is $Q_{1t}$ or $\overline{Q_{1t}}$ for the first argument and $Q_{2t}$ or $\overline{Q_{2t}}$ for the second, with the sign of the slot:
| central argument | the surviving vector parts |
|---|---|
| $\tilde Q_1\in\mathbb{B}_{C}$ | $V(\tilde Q_1\tilde Q_2)=V(\tilde Q_1^{\natural}\tilde Q_2)=V(\tilde Q_1\bullet\tilde Q_2)=V(\tilde Q_1\wedge_{\natural}\tilde Q_2)=Q_{1t}\vec Q_2$; $V(\tilde Q_1\tilde Q_2^{*})=V(\tilde Q_1^{\natural}\tilde Q_2^{*})=V(\tilde Q_1\bullet_{\natural*}\tilde Q_2)=V(\tilde Q_1\wedge_{*}\tilde Q_2)=-Q_{1t}\overline{\vec Q_2}$; the other four vanish |
| $\tilde Q_2\in\mathbb{B}_{C}$ | $V(\tilde Q_1\tilde Q_2)=V(\tilde Q_1\bullet\tilde Q_2)=Q_{2t}\vec Q_1$; $V(\tilde Q_1^{\natural}\tilde Q_2)=V(\tilde Q_1\wedge_{\natural}\tilde Q_2)=-Q_{2t}\vec Q_1$; $V(\tilde Q_1\tilde Q_2^{*})=V(\tilde Q_1\wedge_{*}\tilde Q_2)=\overline{Q_{2t}}\vec Q_1$; $V(\tilde Q_1^{\natural}\tilde Q_2^{*})=V(\tilde Q_1\bullet_{\natural*}\tilde Q_2)=-\overline{Q_{2t}}\vec Q_1$; the other four vanish |
The vector subspace. On pure vectors, $\tilde Q_1=\vec Q_1$ and $\tilde Q_2=\vec Q_2$, the mixed terms drop and only the cross term survives. The twelve restrictions take five distinct values:
| products | vector part on $\mathbb{B}_{V}$ |
|---|---|
| $\tilde Q_1\tilde Q_2$, $\tilde Q_1\wedge\tilde Q_2$ | $\vec Q_1\times\vec Q_2$ |
| $\tilde Q_1^{\natural}\tilde Q_2$, $\tilde Q_1\wedge_{\natural}\tilde Q_2$ | $-\vec Q_1\times\vec Q_2$ |
| $\tilde Q_1\tilde Q_2^{*}$, $\tilde Q_1\wedge_{*}\tilde Q_2$ | $-\vec Q_1\times\overline{\vec Q_2}$ |
| $\tilde Q_1^{\natural}\tilde Q_2^{*}$, $\tilde Q_1\wedge_{\natural*}\tilde Q_2$ | $\vec Q_1\times\overline{\vec Q_2}$ |
| $\tilde Q_1\bullet\tilde Q_2$, $\tilde Q_1\bullet_{\natural}\tilde Q_2$, $\tilde Q_1\bullet_{*}\tilde Q_2$, $\tilde Q_1\bullet_{\natural*}\tilde Q_2$ | $0$ |
Each general vector part coincides with its antisymmetric part on the vector subspace, and each symmetric part vanishes there. On the vector subspace the vector part is therefore the cross product alone.
The real subspace. On the real quaternion subspace $\mathbb{B}_{H}$, the coefficientwise conjugation is the identity and the twelve restrictions take eight distinct values. The collisions are
$$ V(\tilde Q_1^{\natural}\tilde Q_2)=V(\tilde Q_1\wedge_{\natural}\tilde Q_2),\qquad V(\tilde Q_1\tilde Q_2^{*})=V(\tilde Q_1\wedge_{*}\tilde Q_2),\qquad V(\tilde Q_1\wedge\tilde Q_2)=V(\tilde Q_1\wedge_{\natural*}\tilde Q_2),\qquad V(\tilde Q_1\bullet_{\natural}\tilde Q_2)=V(\tilde Q_1\bullet_{*}\tilde Q_2)=0, $$
and the eight representatives are
| product | vector part on $\mathbb{B}_{H}$ |
|---|---|
| $\tilde Q_1\tilde Q_2$ | $Q_{1t}\vec Q_2+Q_{2t}\vec Q_1+\vec Q_1\times\vec Q_2$ |
| $\tilde Q_1^{\natural}\tilde Q_2$ | $Q_{1t}\vec Q_2-Q_{2t}\vec Q_1-\vec Q_1\times\vec Q_2$ |
| $\tilde Q_1\tilde Q_2^{*}$ | $-Q_{1t}\vec Q_2+Q_{2t}\vec Q_1-\vec Q_1\times\vec Q_2$ |
| $\tilde Q_1^{\natural}\tilde Q_2^{*}$ | $-Q_{1t}\vec Q_2-Q_{2t}\vec Q_1+\vec Q_1\times\vec Q_2$ |
| $\tilde Q_1\bullet\tilde Q_2$ | $Q_{1t}\vec Q_2+Q_{2t}\vec Q_1$ |
| $\tilde Q_1\bullet_{\natural*}\tilde Q_2$ | $-Q_{1t}\vec Q_2-Q_{2t}\vec Q_1$ |
| $\tilde Q_1\wedge\tilde Q_2$ | $\vec Q_1\times\vec Q_2$ |
| $0$ | $0$ |
The Hermitian subspace. On $\mathbb{B}_{\Psi}$, write $\tilde Q_1=a+i\vec p$ and $\tilde Q_2=b+i\vec q$ with $a,b\in\mathbb{R}$ and $\vec p,\vec q\in\mathrm{Vect}_{\mathbb{R}}(\mathbb{B})$. The twelve restrictions take seven distinct values:
| class | vector part on $\mathbb{B}_{\Psi}$ |
|---|---|
| $\tilde Q_1\tilde Q_2=\tilde Q_1\tilde Q_2^{*}=\tilde Q_1\wedge_{*}\tilde Q_2$ | $-\vec p\times\vec q+i(a\vec q+b\vec p)$ |
| $\tilde Q_1^{\natural}\tilde Q_2=\tilde Q_1^{\natural}\tilde Q_2^{*}=\tilde Q_1\wedge_{\natural}\tilde Q_2$ | $\vec p\times\vec q+i(a\vec q-b\vec p)$ |
| $\tilde Q_1\bullet\tilde Q_2$ | $i(a\vec q+b\vec p)$ |
| $\tilde Q_1\bullet_{\natural*}\tilde Q_2$ | $i(a\vec q-b\vec p)$ |
| $\tilde Q_1\wedge\tilde Q_2$ | $-\vec p\times\vec q$ |
| $\tilde Q_1\wedge_{\natural*}\tilde Q_2$ | $\vec p\times\vec q$ |
| $\tilde Q_1\bullet_{\natural}\tilde Q_2=\tilde Q_1\bullet_{*}\tilde Q_2$ | $0$ |
The identities behind the table are $V(\tilde Q_1\tilde Q_2^{*})=V(\tilde Q_1\tilde Q_2)$ and $V(\tilde Q_1^{\natural}\tilde Q_2^{*})=V(\tilde Q_1^{\natural}\tilde Q_2)$ on $\mathbb{B}_{\Psi}$, because $\tilde Q^{*}=\tilde Q$ there, together with $V(\tilde Q_1\bullet_{*}\tilde Q_2)=V(\tilde Q_1\tilde Q_2^{*})-V(\tilde Q_1\wedge_{*}\tilde Q_2)=0$. On the anti-Hermitian subspace $\mathbb{B}_{\Phi}$ the same argument with $\tilde Q^{*}=-\tilde Q$ gives $V(\tilde Q_1\tilde Q_2^{*})=-V(\tilde Q_1\tilde Q_2)$ and $V(\tilde Q_1^{\natural}\tilde Q_2^{*})=-V(\tilde Q_1^{\natural}\tilde Q_2)$.
The Symmetries of the Vector Parts
The exchange of the two arguments classifies the vector parts by the same four symmetries that classify the parts of the products. The following table gives the class, the identity and the products.
| class | identity | products |
|---|---|---|
| symmetric | $V(\tilde Q_1,\tilde Q_2)=V(\tilde Q_2,\tilde Q_1)$ | $\tilde Q_1\bullet\tilde Q_2$ |
| alternating | $V(\tilde Q_1,\tilde Q_2)=-V(\tilde Q_2,\tilde Q_1)$ | $\tilde Q_1^{\natural}\tilde Q_2$, $\tilde Q_1\wedge\tilde Q_2$, $\tilde Q_1\wedge_{\natural}\tilde Q_2$ |
| conjugate-symmetric | $V(\tilde Q_1,\tilde Q_2)=\overline{V(\tilde Q_2,\tilde Q_1)}$ | $\tilde Q_1\bullet_{\natural*}\tilde Q_2$ |
| conjugate-alternating | $V(\tilde Q_1,\tilde Q_2)=-\overline{V(\tilde Q_2,\tilde Q_1)}$ | $\tilde Q_1\tilde Q_2^{*}$, $\tilde Q_1\wedge_{*}\tilde Q_2$, $\tilde Q_1\wedge_{\natural*}\tilde Q_2$ |
| neither | — | $\tilde Q_1\tilde Q_2$, $\tilde Q_1^{\natural}\tilde Q_2^{*}$ |
| zero | $V=0$ | $\tilde Q_1\bullet_{\natural}\tilde Q_2$, $\tilde Q_1\bullet_{*}\tilde Q_2$ |
Theorem (the class of the vector part). The vector part of a product has the symmetry of the part of the product: it is symmetric for the symmetric part of a bilinear product, alternating for the antisymmetric part of a bilinear product, conjugate-symmetric for the symmetric part of a sesquilinear product, and conjugate-alternating for the antisymmetric part of a sesquilinear product. The vector parts of the four general products are not all of one class: the plain bilinear is neither, the quaternionic bilinear is alternating, the plain sesquilinear is conjugate-alternating, and the quaternionic sesquilinear is neither.
Proof. The four definitions of the parts give the four identities term by term. For the bilinear pair, the exchange is the swap and the vector part of the swapped value is the swapped vector part; for the sesquilinear pair, the exchange is the swap followed by the coefficientwise conjugation, and the vector part of the conjugated value is the conjugated vector part because the conjugation is real-linear on the vector subspace. The general rows follow from the decompositions $V(\tilde Q_1\tilde Q_2)=V(\tilde Q_1\bullet\tilde Q_2)+V(\tilde Q_1\wedge\tilde Q_2)$ and $V(\tilde Q_1^{\natural}\tilde Q_2^{*})=V(\tilde Q_1\bullet_{\natural*}\tilde Q_2)+V(\tilde Q_1\wedge_{\natural*}\tilde Q_2)$.
Remark (the vector part can be more symmetric than the product). The general quaternionic bilinear product is not alternating — its scalar part is symmetric — but its vector part is alternating. The general plain sesquilinear product is not conjugate-alternating as a product — its scalar part is conjugate-symmetric — but its vector part is. The symmetry of the vector part is the symmetry of the vector half, not of the whole value; the scalar half carries the other symmetry, and the two halves together are neither.
The Component Forms
Definition (the component forms). For $f$ of the twelve and $w\in\mathbb{B}_{V}$, the component form is
$$ \varphi_{f,w}(\tilde Q_1,\tilde Q_2)=\bigl(w,V_f(\tilde Q_1,\tilde Q_2)\bigr). $$
For $f$ bilinear it is a complex bilinear form, with Gram matrix $G_{ij}=\varphi_{f,w}(e_i,e_j)$ in the basis $(e_t,\vec e_x,\vec e_y,\vec e_z)$ and $\varphi(\sum c_ie_i,\sum d_je_j)=\sum c_iG_{ij}d_j$. For $f$ sesquilinear it is sesquilinear over $(\mathbb{C},\bar{\cdot})$, with the same entries and $\varphi(\sum c_ie_i,\sum d_je_j)=\sum c_iG_{ij}\overline{d_j}$.
Theorem (the component forms of the twelve). The three coordinate forms of each vector part have the following classes, Gram matrices and ranks.
| products | class of $\varphi$ | Gram matrix | rank |
|---|---|---|---|
| $\tilde Q_1\tilde Q_2$ | bilinear | neither; symmetric part the component form of $\tilde Q_1\bullet\tilde Q_2$, alternating part the component form of $\tilde Q_1\wedge\tilde Q_2$ | $4$ |
| $\tilde Q_1^{\natural}\tilde Q_2$ | bilinear | alternating | $4$ |
| $\tilde Q_1\bullet\tilde Q_2$ | bilinear | symmetric | $2$ |
| $\tilde Q_1\bullet_{\natural}\tilde Q_2$ | zero | zero | $0$ |
| $\tilde Q_1\wedge\tilde Q_2$ | bilinear | alternating | $2$ |
| $\tilde Q_1\tilde Q_2^{*}$, $\tilde Q_1\wedge_{*}\tilde Q_2$ | sesquilinear | conjugate-alternating; anti-Hermitian | $4$ |
| $\tilde Q_1\bullet_{*}\tilde Q_2$ | zero | zero | $0$ |
| $\tilde Q_1\bullet_{\natural*}\tilde Q_2$ | sesquilinear | conjugate-symmetric; Hermitian | $2$ |
| $\tilde Q_1\wedge_{\natural*}\tilde Q_2$ | sesquilinear | conjugate-alternating; anti-Hermitian | $2$ |
| $\tilde Q_1^{\natural}\tilde Q_2^{*}$ | sesquilinear | sum of a Hermitian rank $2$ and an anti-Hermitian rank $2$ | $4$ |
Proof. The ranks and the symmetry types are read from the matrices. For $\tilde Q_1\tilde Q_2$ and $w=\vec e_x$ the matrix is
$$ G_x= \begin{pmatrix} 0&1&0&0\\ 1&0&0&0\\ 0&0&0&1\\ 0&0&-1&0 \end{pmatrix}, $$
of rank $4$; its symmetric part has rank $2$ and its alternating part rank $2$, and they are the Gram matrices of the component forms of $V(\tilde Q_1\bullet\tilde Q_2)$ and $V(\tilde Q_1\wedge\tilde Q_2)$. The matrices for $w=\vec e_y$ and $w=\vec e_z$ are obtained from $G_x$ by the cyclic relabelling $x\to y$, $y\to z$, $z\to x$ of the basis vectors. For $\tilde Q_1^{\natural}\tilde Q_2$ and $w=\vec e_x$ the matrix is
$$ N_x= \begin{pmatrix} 0&1&0&0\\ -1&0&0&0\\ 0&0&0&-1\\ 0&0&1&0 \end{pmatrix}, $$
anti-Hermitian of rank $4$; the other two are obtained by the same cyclic relabelling. The zero rows are immediate. For the sesquilinear rows, the Gram matrix of a conjugate-symmetric form is Hermitian and that of a conjugate-alternating form is anti-Hermitian, by the two definitions read in the entries; the ranks are computed from the same matrices. The $\tilde Q_1^{\natural}\tilde Q_2^{*}$ entry is the sum of the $\tilde Q_1\bullet_{\natural*}\tilde Q_2$ and $\tilde Q_1\wedge_{\natural*}\tilde Q_2$ entries; its Hermitian part is the $\tilde Q_1\bullet_{\natural*}\tilde Q_2$ matrix, its anti-Hermitian part is the $\tilde Q_1\wedge_{\natural*}\tilde Q_2$ matrix, and its rank is $4$.
Remark (the diagonal of a component form). At the two equal arguments,
$$ \varphi_{f,w}(\tilde Q,\tilde Q)=\bigl(w,D_f(\tilde Q)\bigr). $$
For a conjugate-symmetric component form this is real, and for a conjugate-alternating one it is purely imaginary; this is the scalar-form reading of the real and imaginary characters of the diagonals of $\tilde Q_1\bullet_{\natural*}\tilde Q_2$ and of $\tilde Q_1\wedge_{\natural*}\tilde Q_2$.
The Relations among the Twelve
The twelve vector parts are linked by the following identities, in which $\tau$ is the swap of the two arguments and the bar is the coefficientwise conjugation.
| identity | content |
|---|---|
| $V(\tilde Q_1^{\natural}\tilde Q_2)=V(\tilde Q_1\wedge_{\natural}\tilde Q_2)$, $V(\tilde Q_1\tilde Q_2^{*})=V(\tilde Q_1\wedge_{*}\tilde Q_2)$, $V(\tilde Q_1\bullet_{\natural}\tilde Q_2)=V(\tilde Q_1\bullet_{*}\tilde Q_2)=0$ | the three coincidences |
| $V(\tilde Q_1\tilde Q_2)=V(\tilde Q_1\bullet\tilde Q_2)+V(\tilde Q_1\wedge\tilde Q_2)$, $V(\tilde Q_1^{\natural}\tilde Q_2^{*})=V(\tilde Q_1\bullet_{\natural*}\tilde Q_2)+V(\tilde Q_1\wedge_{\natural*}\tilde Q_2)$ | the two reconstructions |
| $V(\tilde Q_1\tilde Q_2)+V(\tilde Q_1^{\natural}\tilde Q_2)=2Q_{1t}\vec Q_2$, $V(\tilde Q_1\tilde Q_2^{*})+V(\tilde Q_1^{\natural}\tilde Q_2^{*})=-2Q_{1t}\overline{\vec Q_2}$ | the two sums |
| $V(\tilde Q_1^{\natural}\tilde Q_2)=2Q_{1t}\vec Q_2-V(\tilde Q_1\tilde Q_2)$, $V(\tilde Q_1^{\natural}\tilde Q_2^{*})=-2Q_{1t}\overline{\vec Q_2}-V(\tilde Q_1\tilde Q_2^{*})$ | the two reflections |
| $V(\tilde Q_1\tilde Q_2^{*})=V(\tilde Q_1\tilde Q_2^{*})$, $V(\tilde Q_1^{\natural}\tilde Q_2^{*})=V(\tilde Q_1^{\natural}\tilde Q_2^{*})$ | the two slot identities |
| $V(\tilde Q_1\bullet\tilde Q_2)=\tfrac12(V(\tilde Q_1\tilde Q_2)+V(\tilde Q_1\tilde Q_2)\circ\tau)$, $V(\tilde Q_1\wedge\tilde Q_2)=\tfrac12(V(\tilde Q_1\tilde Q_2)-V(\tilde Q_1\tilde Q_2)\circ\tau)$ | the two bilinear halves |
| $V(\tilde Q_1\bullet_{\natural*}\tilde Q_2)=\tfrac12(V(\tilde Q_1^{\natural}\tilde Q_2^{*})+\overline{V(\tilde Q_1^{\natural}\tilde Q_2^{*})\circ\tau})$, $V(\tilde Q_1\wedge_{\natural*}\tilde Q_2)=\tfrac12(V(\tilde Q_1^{\natural}\tilde Q_2^{*})-\overline{V(\tilde Q_1^{\natural}\tilde Q_2^{*})\circ\tau})$ | the two sesquilinear halves |
The first row collects the coincidences, the second and third the reconstructions and the two sums, and the rest the reflections, the slot identities and the two half formulas. The scalar parts obey the companion identities of The Four General Products and Their Relations over the Biquaternion Algebra, and the two sets together reconstruct the twelve products.
Summary
The vector parts of the twelve products are vector-valued maps into the complex vector subspace, not scalar forms. The four general products obey one two-bit rule: the first slot fixes the sign of the second mixed term, the second slot fixes the sign of the first mixed term, and the cross term carries the product of the two signs. There are twelve vector parts, one to each product. Three coincidences make some of them equal: the two central symmetric products have zero vector part, and the quaternionic bilinear and the plain sesquilinear products share their vector part with their antisymmetric parts. The twelve maps therefore fall into nine distinct classes, separated by three pairs of basis elements, and the vector part alone does not determine the product; the pair of the scalar part and the vector part does.
Each product was read with its own properties: its diagonal, its image, its value on pure vectors, its symmetry and its role in the algebra. The diagonals give the vector parts of the squares: $2Q_t\vec Q$ for the plain bilinear pair; zero for the quaternionic bilinear product, the two central symmetric products and the two bilinear antisymmetric products; the Hermitian square $\mathrm{Vec}(\tilde Q\tilde Q^{*})$ for the plain sesquilinear pair; a real vector for $\tilde Q_1\bullet_{\natural*}\tilde Q_2$; a purely imaginary vector for $\tilde Q_1\wedge_{\natural*}\tilde Q_2$; and a real-plus-imaginary pair with orthogonal parts for $\tilde Q_1^{\natural}\tilde Q_2^{*}$. On the centre all vector parts vanish; on the vector subspace only the cross term survives and the twelve restrictions take five distinct values; on the real subspace they take eight; on the Hermitian subspace they take seven. The component forms are three scalar forms to a vector part: rank $4$ with a rank-$2$ symmetric and a rank-$2$ alternating part for the plain bilinear product, alternating rank $4$ for the quaternionic bilinear product, Hermitian or anti-Hermitian of rank $2$ or $4$ in the sesquilinear families.
Summary of Notation
| Symbol | Meaning |
|---|---|
| $\mathrm{Vec}(\tilde Q)=\vec Q$ | the vector part of an element |
| $\mathbb{B}_{V}$ | the complex vector subspace, the image of the vector parts |
| $\mathrm{Vect}_{\mathbb{R}}(\mathbb{B})$ | the real vector subspace, the real vectors |
| $V_f(\tilde Q_1,\tilde Q_2)=\mathrm{Vec}(f(\tilde Q_1,\tilde Q_2))$ | the vector part of the product $f$ |
| $D_f(\tilde Q)=V_f(\tilde Q,\tilde Q)$ | the diagonal, the vector part of the square |
| $\varphi_{f,w}(\tilde Q_1,\tilde Q_2)=(w,V_f(\tilde Q_1,\tilde Q_2))$ | the component form in the direction $w$ |
| $\varepsilon$, $\alpha$, $\beta$ | the second-slot bit, the second-slot sign and the first-slot sign of the general rule |
| $\vec Q_2^{\varepsilon}$, $Q_{2t}^{\varepsilon}$ | the second vector part and scalar read with or without the coefficientwise conjugation |
| $\tilde Q_1\tilde Q_2$, $\tilde Q_1^{\natural}\tilde Q_2$, $\tilde Q_1\tilde Q_2^{*}$, $\tilde Q_1^{\natural}\tilde Q_2^{*}$ | the four general products |
| $\tilde Q_1\bullet\tilde Q_2$, $\tilde Q_1\bullet_{\natural}\tilde Q_2$, $\tilde Q_1\bullet_{*}\tilde Q_2$, $\tilde Q_1\bullet_{\natural*}\tilde Q_2$ | the four symmetric parts |
| $\tilde Q_1\wedge\tilde Q_2$, $\tilde Q_1\wedge_{\natural}\tilde Q_2$, $\tilde Q_1\wedge_{*}\tilde Q_2$, $\tilde Q_1\wedge_{\natural*}\tilde Q_2$ | the four antisymmetric parts |
| $\mathbb{B}_{C}$, $\mathbb{B}_{H}$, $\mathbb{B}_{\Psi}$, $\mathbb{B}_{\Phi}$ | the centre, the real quaternion subspace and the Hermitian and anti-Hermitian subspaces |
Further Reading
- Symmetric and Antisymmetric Decomposition of the Four General Products, for the twelve products, their codes and their scalar and vector parts.
- The Four General Products and Their Relations over the Biquaternion Algebra, for the four general products and the splitting identities used here.
- The Four Forms and the Four Algebraic Norms over the Biquaternion Algebra, for the scalar parts and the four degenerate sets.
- The Biquaternion Algebra and Subspaces and Decompositions, for the basis, the conjugations and the remarkable subspaces.
- Introduction to the Operators of the Biquaternion Algebra, for the coefficient and quaternion conjugations, the two switches behind the two-bit rule.
- The Square of an Element, for the diagonals.