A Brief History of Biquaternions in Physics

Introduction

The biquaternion algebra $\mathbb{B} = \mathbb{C}\otimes_\mathbb{R}\mathbb{H}$ is the complexification of Hamilton's quaternions. In its complexified form it is older than the special theory of relativity and older by three quarters of a century than the quantum mechanics of spin; it is also the algebra in which the Pauli matrices, the two-component spinors, and the double cover of the Lorentz group are written. This article is a history of how $\mathbb{B}$ entered physics: how it was found, how it was used in electromagnetism, how it was published as a language for relativity and then displaced by vectors and tensors, and how it returned in the middle of the twentieth century.

Two qualifications belong at the start.

The first is scope. Most nineteenth-century physics that used "quaternions" used the real quaternions $\mathbb{H}$, not the biquaternions. The quaternion formulation of Maxwell's equations, the quaternionists' campaign, and the vector revolt all concern real quaternions and their vector parts. The biquaternion thread is narrower: it is the complexified algebra, and its principal physical use before 1950 is the formulation of the Lorentz group and of special relativity, first by Noether and Klein and then by Conway and Silberstein. This article follows that narrower thread, keeping the real-quaternion context in view because the thread cannot be understood without it. A history that silently substituted the real quaternions for the biquaternions would be signing for more than the record contains.

The second is what a history can show. The algebra $\mathbb{B}$ contains $M_2(\mathbb{C})$, and therefore contains, as a matter of algebra, the objects that quantum mechanics later used. That is a structural fact. It is not the same as a historical fact — that anyone intended a quantum-mechanical application, or was on the verge of one. The word "near-miss" that this subject attracts is a retrospective description; where it is used below it is marked as such. Dates and attributions follow the standard references, and the few items that could not be pinned down are flagged in the text and recorded in the companion note.

Hamilton's Quaternions and Their Complexification

William Rowan Hamilton discovered the quaternions on 16 October 1843, while walking along the Royal Canal in Dublin. The multiplication rules he carved into Broom Bridge (then Brougham Bridge) were

$$ i^2 = j^2 = k^2 = ijk = -1 . $$

These are Hamilton's own letters: here $i, j, k$ are the three quaternion units. The article's notation, in which those units are written $e_1, e_2, e_3$ and $i$ denotes the scalar imaginary, is set out below.

Hamilton had spent years trying to multiply triples of numbers to represent points in space; the breakthrough was to give up triples for quadruples. He named the quadruple a quaternion and spent the rest of his life on it. The Lectures on Quaternions (1853) is the comprehensive treatise; the Elements of Quaternions appeared posthumously in 1866, edited by his son William Edwin Hamilton, with a later edition by Charles Jasper Joly (1899–1901). Olinde Rodrigues had published formulas equivalent in substance to quaternion multiplication in the composition of rotations in 1840, without the algebra; the composition law is accordingly shared, while the algebra is Hamilton's.

Why the coefficients had to become complex. Hamilton was studying the quaternion algebra and wanted to solve equations with it, and when he worked through the cases by degree he found that the algebra stops short.

Linear equations: solved, and within $\mathbb{H}$ itself. If $ax = b$ with $a \neq 0$, then $x = a^{-1}b$, and if $xa = b$ then $x = ba^{-1}$, each unique. The side on which $a$ stands matters — $e_1^{-1}e_2 = -e_3$ while $e_2e_1^{-1} = +e_3$ — but every nonzero quaternion is invertible, so a linear equation asked on a definite side always has its answer in $\mathbb{H}$.
Quadratic equations: no longer routine. There is no quadratic formula, and the solutions, when they are found, sometimes require complex coefficients — that is, complex quaternions rather than quaternions.
Higher degree: worse. The root sets cease to be finite. The cubic $x^{3} = 1$ has the root $1$ and a whole sphere of further roots $x = -\tfrac{1}{2} + \tfrac{\sqrt{3}}{2}\mathbf{u}$, one for each unit pure quaternion $\mathbf{u}$, so a degree-$n$ equation over $\mathbb{H}$ cannot be reduced to $n$ linear factors.

The quadratic case fails at the first step of the classroom method. Completing the square requires

$$ \left(x + \frac{b}{2}\right)^{2} = x^{2} + b\,x + \frac{1}{2}\left(bx - xb\right) + \frac{b^{2}}{4}, $$

and the linear term disappears only if $x$ commutes with $b$, which is not the general case. With the commutator left standing there is no reduction to $y^{2} = \text{const}$, and hence no discriminant to compute with.

A basic example. The shortest way to see the obstruction is a quadratic with ordinary real quaternion coefficients:

$$ x^{2} + 2e_2\,x - 3 = 0 . $$

Write $x = x_0 + x_1 e_1 + x_2 e_2 + x_3 e_3$ and separate the four components of the equation. The $e_1$- and $e_3$-components give $x_3 = -x_0 x_1$ and $x_1 = x_0 x_3$, hence $x_1 = x_3 = 0$; the $e_2$-component gives $x_0\,(x_2 + 1) = 0$, which leaves two branches.

First branch, $x_2 = -1$. The scalar component becomes $x_0^{2} = 2$, so $x_0 = \pm\sqrt{2}$ and

$$ x = \pm\sqrt{2} - e_2 , $$

two solutions in $\mathbb{H}$.

Second branch, $x_0 = 0$. The scalar component becomes

$$ x_2^{2} + 2x_2 + 3 = 0, \qquad \Delta = 4 - 12 = -8 < 0 , $$

which no real $x_2$ satisfies. The resolution is $x_2 = -1 \pm \sqrt{2}\,i$, so

$$ x = \left(-1 \pm \sqrt{2}\,i\right) e_2 , $$

two complex quaternions, not quaternions.

Substitution confirms all four. A quadratic with real quaternion coefficients has produced four solutions, two in $\mathbb{H}$ and two outside it — and the second branch is where the coefficients leave the algebra. It leaves because a single scalar equation, $x_2^{2} + 2x_2 + 3 = 0$, has a negative discriminant, and solving it calls for a scalar square root of $-1$: a number that commutes with everything, so that the ordinary algebra of equations still applies. The quaternion units are roots of $-1$ in abundance, but they are not central, and $\mathbb{H}$ contains no central root of $-1$ at all, since the only quaternions commuting with every element are the real numbers. Nothing in the reduction is repaired by a quaternion unit; the repair is to let the coefficients themselves be complex.

That the complexification is forced by this branch, rather than by a shortage of roots, is worth recording, since the point is often overstated. By the Eilenberg–Niven theorem a one-sided quaternion polynomial of positive degree always has a quaternion root, and the first branch above duly supplies two. What fails in $\mathbb{H}$ is the splitting: the quadratic has solutions galore but no factorisation into linear factors, and no scalar discriminant to separate them.

This is the precise content of the remark that the quaternions were "not enough": they are enough to solve the equation in one branch, and not enough to solve it as an equation.

The quaternions form a four-dimensional real division algebra. Making them complex produces the biquaternions: the four coefficients are allowed to be complex,

$$ \tilde{Q} = Q_0 e_0 + Q_1 e_1 + Q_2 e_2 + Q_3 e_3, \qquad Q_\mu \in \mathbb{C}, $$

with the quaternion units $e_0 = 1, e_1, e_2, e_3$ multiplying as in $\mathbb{H}$ and commuting with every complex scalar. Hamilton himself took this step and named the result: the biquaternion is the complex quaternion, and its fullest treatment in his own work is in the Lectures on Quaternions (1853). His motive was the algebraic one set out above: in the treatment of general quadratic equations with quaternion coefficients, the resolution called for complex coefficients. The word is his, and so is the surrounding vocabulary — he introduced bivector, biconjugate, bitensor, and biversor alongside it. His notation for the unit part of a quaternion was $Uq$ — the versor, from the Latin versari, to turn — and his word for the norm $|q|$ was the tensor, so that every quaternion is the product of its tensor and its versor, $q = Tq\,Uq$. Because the quaternion units were written $i, j, k$, Hamilton used the letter $h$ — not $i$ — for the square root of $-1$ in the coefficient field, to keep the two imaginary units apart, and Arthur W. Conway later followed the same convention. The companion articles write the quaternion units as $e_1, e_2, e_3$ and reserve $i$ for the scalar imaginary, which removes the clash that $h$ was introduced to avoid. The translation is

$$ e_0 \leftrightarrow 1, \qquad e_1 \leftrightarrow i, \qquad e_2 \leftrightarrow j, \qquad e_3 \leftrightarrow k, \qquad i \leftrightarrow h . $$

The biquaternion algebra thus arrived with the quaternions themselves, within the first decade of their discovery. Nothing in its later physics was new algebra; what was new, repeatedly, was the recognition of where the algebra applied.

The bivectors in print, and the null square. The complexification was not only a convenience for equations. In a note read to the Royal Irish Academy on 28 February 1853 — On the Geometrical Interpretation of some Results obtained by calculation with Biquaternions — Hamilton worked with the bivectors

$$ \rho + \sqrt{-1}\,\rho' , $$

where $\rho$ and $\rho'$ are real vectors and the root of $-1$ is the ordinary commutative one; in the letters of the calculus, with his $h$ for that root, the examples he takes are the null-square bivectors

$$ i + hj, \qquad j + hk . $$

The name records the computation he gives for the second, $(j + hk)^2 = j^2 + h(jk + kj) + h^2 k^2 = -1 + 0 + (-1)(-1) = 0$: the cross terms cancel because $jk + kj = 0$, and the term $h^2k^2 = +1$ cancels the term $j^2 = -1$. The square vanishes, so every power from the second on vanishes with it and the binomial series terminates at first order; this yields what the note reports as a simplification useful in the geometrical investigation that follows,

$$ (1 + j + hk)^{x} = 1 + x\,(j + hk), $$

for any numerical exponent $x$. These bivectors are the zero divisors of the algebra, and the note is where they first appear in print — it refers to the Lectures on Quaternions as "as yet unpublished" — while the simplification is the origin of the nilpotent regime of Biquaternion Elementary Functions. That the zero divisors entered through geometry rather than through algebra is the point of the communication: he reports a remark he had made to the Academy in November 1844, that the geometrically unreal tangents to a sphere drawn from an internal point — no real tangent from an interior point exists, the discriminant of the intersection being negative — are represented in his calculus by bivectors of this kind, and have positive squares. A square that comes out positive is the bivector's doing and not a real vector's, since a real vector obeys $\rho^2 = -|\rho|^2 < 0$ always, while a bivector whose two real parts are perpendicular has the real square $|\rho'|^2 - |\rho|^2$, which is zero when they have equal length — the null-square case above — and positive when the imaginary part is the longer. The note closes with the theorem the algebra had been used to obtain, on the alternate points generated by a continued fraction and the two circles on which they lie; that belongs to the geometry of the period and is not carried further here.

A terminological caution is necessary, because the word "biquaternion" has not been used uniformly. The modern convention distinguishes the three tensor products of $\mathbb{H}$ with the three two-dimensional commutative real algebras: the biquaternions $\mathbb{C}\otimes_\mathbb{R}\mathbb{H}$, the split-biquaternions $\mathbb{D}\otimes_\mathbb{R}\mathbb{H}$ (with $\mathbb{D}$ the split-complex numbers), and the dual quaternions $\mathbb{N}\otimes_\mathbb{R}\mathbb{H}$ (with $\mathbb{N}$ the dual numbers). William Kingdon Clifford observed that Hamilton's biquaternions were quaternions with complex coefficients, and then used the word himself in a different sense in his "Preliminary Sketch of Biquaternions" (Proceedings of the London Mathematical Society, vol. 4, pp. 381–395; the volume is dated variously 1871 and 1873): his biquaternions are the split-biquaternions, the quaternions with split-complex coefficients, $\mathbb{D}\otimes_\mathbb{R}\mathbb{H}$, and he recorded the change of meaning in a note on the word. The dual quaternions $\mathbb{N}\otimes_\mathbb{R}\mathbb{H}$ enter a later line, Alexander McAulay's and Eduard Study's, and are the algebra of the rigid motions of space. In older texts the words are not always used in this sense, and a reader meeting "biquaternion" in a nineteenth-century source should check which algebra is meant. This article follows the modern convention: the biquaternions are $\mathbb{C}\otimes_\mathbb{R}\mathbb{H}$.

The Algebra, in the Notation of the Companion Articles

The algebra and its subspaces are established in the companion articles and are recalled here only in the form the history needs. The notation is inherited, not redefined.

$\mathbb{B}$ is four-dimensional over $\mathbb{C}$ and eight-dimensional over $\mathbb{R}$. Its four natural conjugations have as fixed-point sets the distinguished real subspaces used throughout the corpus: the complex subspace $\mathbb{C}_{\mathbb{B}}$ (the complex scalars, the center of $\mathbb{B}$), the quaternion subspace $\mathbb{H}_{\mathbb{B}}$ (the real quaternions), the Hermitian subspace $\mathbb{M}_{+}$ (real scalar part, imaginary vector part, the informational sector), and the anti-Hermitian subspace $\mathbb{M}_{-}$ (imaginary scalar part, real vector part, the material sector). The natural quadratic form is the biquaternion norm

$$ N(\tilde{Q}) = \tilde{Q}\tilde{Q}^{\natural} = \sum_{\mu=0}^{3} Q_\mu^2 , $$

whose indefinite signature on $\mathbb{M}_{-}$ arises algebraically from $i^2 = -1$ in the imaginary time coefficient $ict$. The null elements $N(\tilde{Q}) = 0$, $\tilde{Q} \neq 0$, are the zero divisors; on $\mathbb{M}_{-}$ they are the light cone.

The single structural fact that organizes the whole subsequent history is the isomorphism

$$ \mathbb{B} \;\cong\; M_2(\mathbb{C}) , $$

realized by sending the quaternion units to $-i$ times the Pauli matrices, $e_k \mapsto -i\sigma_k$. Equivalently, $\mathbb{B}$ is the real Clifford algebra $Cl_{3,0}(\mathbb{R})$ — the Pauli algebra — and the even subalgebra of the spacetime algebra $Cl_{1,3}$. Under this isomorphism the biquaternion norm is the determinant, $N(\tilde{Q}) = \det\Phi(\tilde{Q})$, and the unit-norm elements

$$ SL(2,\mathbb{C}) = \{\tilde{\Lambda} \in \mathbb{B} : \tilde{\Lambda}\tilde{\Lambda}^{\natural} = e_0\} $$

are the double cover of the proper orthochronous Lorentz group $SO^{+}(1,3)$, acting on $\mathbb{M}_{-}$ by the rotor conjugation

$$ \tilde{Q} \;\longmapsto\; \tilde{\Lambda}\,\tilde{Q}\,\tilde{\Lambda}^{*} . $$

These are the facts that make the history intelligible. The algebra of Hamilton's complex quaternions is the algebra of the Pauli matrices; the group of unit biquaternions is the Lorentz double cover; the idempotents $\tilde\Pi_\pm = \tfrac{1}{2}(e_0 \pm i\hat{\mu})$ in $\mathbb{M}_{+}$ are the pure states of a two-state system, with $\mathrm{Tr}(\tilde{P}\tilde{H}) = 2\,\mathrm{Sc}(\tilde{P}\tilde{H})$ as the Born rule written in the algebra (see the companion article The Hermitian Subspace $\mathbb{M}_+$ as the Informational Sector). None of these identifications was known to Hamilton.

Clifford's Rotors: Position, Lines, and Screws

Hamilton's vectors have magnitude and direction but no position: the vector $AB$ is the same vector as $CD$ whenever the two are equal, parallel, and in the same sense, so a vector may be moved anywhere parallel to itself. The translation of a rigid body is the type case, and a couple can be represented by a vector. Many quantities of mechanics are not of this kind. The rotational velocity of a rigid body is about a definite axis, and equal rotations about two parallel axes are not equivalent; a force has a definite line of action, and equal forces along parallel lines differ by a couple. The calculus follows the difference: the force-polygon and the addition of vectors suffice for the motion of a particle and for the composition of couples, while the statics and kinematics of solids call for the link-polygon and the theory of the involution of lines in space, that is, of the linear complex.

The name. William Kingdon Clifford took this difference as the reason to go past the quaternion, and his paper of 1873 is where the word "rotor" enters the vocabulary. He proposes rotor — short for rotator — for a quantity having magnitude, direction, and position, of which the simplest type is a velocity of rotation about an axis; a rotor is drawn as a length on its axis, and where a vector may be moved anywise parallel to itself, a rotor may be moved only along its own line, so that two rotors are equal only when they lie on the same straight line with the same length and the same sense. Addition is where the analogy with vectors breaks: the sum of vectors is always a vector, but the sum of rotors is a rotor only in special cases, and in general the composition of two forces whose lines of action do not intersect, or of two rotation-velocities whose axes do not intersect, gives the more complex objects — a system of forces, and the most general velocity of a rigid body. For these Clifford credits the theory of screws to Ball: a system of forces reduces to a single force together with a couple perpendicular to it, a wrench about a screw whose axis is the central axis and whose pitch is the ratio of the couple to the force, and the velocity of a rigid body is represented in the same way, as a twist-velocity about a screw.

The ladder of quantities. Clifford arranges the whole scale in a table, and the table is the paper's answer to why a fourth kind of unit is wanted at all. The rungs run from the scalar on a line, through the point in a plane and the direction in space, to the axis and the screw, and each rung carries a quantity, a typical example, and a kind of ratio.

Geometrical form	Quantity	Example	Ratio
Scalar on a line	Vector on a line	Addition or subtraction	Signed ratio
Point in a plane	Vector in a plane	Complex quantity	Complex ratio
Direction in space	Vector in space	Translation, couple	Quaternion
Axis	Rotor	Rotation-velocity, force	Twist
Screw	Motor	Twist-velocity, system of forces	Biquaternion

The last column is the point of the table. On a line the ratio of two quantities is a signed number; in a plane it is a complex number; in space the ratio of two vectors is a quaternion; the ratio of two rotors is a twist; and the ratio of two motors is a biquaternion. Geometry and algebra grow together, rung by rung, from the real numbers to the biquaternions, and the last rung is where a fourth kind of unit becomes unavoidable.

The setting: elliptic geometry. The paper works not in Euclidean space, which Clifford with Klein calls the parabolic geometry, but in elliptic geometry, and the choice is load-bearing. The metric geometry is Cayley's: a quadric surface, the absolute, is fixed, and distance is read from it as an angle, points conjugate in regard to it being a quadrant apart. Every line has a polar line and every point a polar plane; two lines equidistant along their whole length are right parallel or left parallel according to the handedness of the twist that carries one to the other; and a twist-velocity of a rigid body has two polar axes. In elliptic space a translation along an axis is the same thing as a rotation about the polar axis, so position and the infinitesimal motion are both carried by rotations and no separate translation parameter is needed. Clifford notes that the formulation in free vectors and unsituated rotors is the limit of this one, in which the difficulty he has just met — that the ratio of two motors is a symbol whose two parts are separately meaningful but not jointly so — does not arise. Euclidean space, with its free vectors and its degenerate absolute, is that parabolic limit.

The algebra. Three mutually perpendicular unit rotors through the origin — the corpus's $e_1$, $e_2$, $e_3$ — satisfy $e_1^{2} = e_2^{2} = e_3^{2} = e_1e_2e_3 = -e_0$ by the ordinary rules of quaternions, and the same equations hold when the rotors do not pass through the origin; each symbol carries a second meaning, as the operation of turning a rotor that meets its axis at right angles through a right angle about that axis. A rotor through the origin is $e_1x + e_2y + e_3z$, and the ratio of two such rotors is a quaternion. The general motor is the sum of a rotor through the origin and one polar to a rotor through the origin, and the paper writes it $\alpha + \omega\beta$, where $\omega$ denotes the rectangular twist of pitch 1 and $\omega\alpha$ is the rotor polar to $\alpha$, of the same magnitude, obtained from it by a left-handed twist. The new symbol is not a second copy of the scalar imaginary. Continuing the twist through two right angles restores the rotor to its original position and not to its reverse, so

$$ \omega^{2} = +1 , $$

and $\omega$ commutes with the quaternion units. It is the split-complex unit, written $j$ in the corpus's notation, and with $j^{2} = +1$ the two combinations

$$ \Pi_1 = \tfrac{1}{2}(1 + j), \qquad \Pi_2 = \tfrac{1}{2}(1 - j) $$

are idempotent, with $\Pi_1\Pi_2 = 0$; multiplication by $\Pi_1$ picks out the right vector part of a motor and multiplication by $\Pi_2$ the left vector part, so that every motor is the sum of a right and a left vector. The paper calls $\Pi_1$ and $\Pi_2$ selective symbols and compares them to the $\mathrm{S}$ and $\mathrm{V}$ of quaternions.

The algebra is the split-biquaternions. The paper's "biquaternion" is therefore the quaternions extended by one central element of square $+1$: the split-biquaternion algebra $\mathbb{D}\otimes_\mathbb{R}\mathbb{H}$, isomorphic to $\mathbb{H}\oplus\mathbb{H}$, with the idempotents $\Pi_1$ and $\Pi_2$ exhibiting the splitting. It is not Hamilton's biquaternion, the quaternions with complex coefficients, $\mathbb{C}\otimes_\mathbb{R}\mathbb{H}$ — the algebra this article and the corpus call the biquaternions — and Clifford's own note on the point records the collision: Hamilton's biquaternion is a quaternion with complex coefficients, and since the word was "no longer wanted in its old meaning" he had "made bold to use it in a new one". The naming collision this article warns about is thus the paper's own and deliberate, and a reader meeting a nineteenth-century "biquaternion" has three algebras to choose from. The algebra of the paper has been known since as the elliptic biquaternions, a name that records the geometry; the later literature calls it the Clifford biquaternion or the dyquaternion, or writes it $\mathbb{H}\oplus\mathbb{H}$. The quaternions with dual-number coefficients, $\mathbb{N}\otimes_\mathbb{R}\mathbb{H}$ with $\varepsilon^{2} = 0$, are the parabolic case and belong to a later line, McAulay's and Study's; they are the algebra of the rigid motions of Euclidean space, and the screw motions, the double cover of $SE(3)$ and the motor calculus built on them are the subject of the companion article Projective Geometric Algebra and Dual Quaternions.

The geometric constructions. The last section of the paper turns the algebra back on the geometry, in five constructions, all in the elliptic case and all expressed with the position-rotor, the rotor drawn from the origin to a point. Relative to a quadrantal tetrahedron $1234$ with the origin at the vertex $4$, the equation of the absolute being $\sum_k x_k^{2} = 0$, the position-rotor of the point $x$ is

$$ \rho = e_1\frac{x_1}{x_4} + e_2\frac{x_2}{x_4} + e_3\frac{x_3}{x_4} , $$

and its tensor, the length of the rotor, is the tangent of the angular distance from the origin to the point. The five constructions are the position-rotor just given; the equation of a straight line, built on the perpendicular drawn from the origin to the line; the rotor along a straight line whose equation is given; the rotor joining two points whose position-rotors are $\alpha$ and $\beta$, namely $mR = \beta - \alpha + \omega\,V\alpha\beta$, with $V$ the vector product; and the rotor parallel to a given line through a given point whose position-rotor is $\sigma$. The list is the paper's demonstration that the algebra computes with lines and with screws, and not only with directions.

The corpus uses "rotor" throughout for the unit-norm element of $\mathbb{B}$ that implements a Lorentz rotation, $\tilde{Q}\mapsto\tilde{\Lambda}\tilde{Q}\tilde{\Lambda}^{*}$. That is a later and a different sense of the same word: Clifford's rotor is a position-carrying line element, a quantity of the geometry of lines, while the corpus's rotor is an element of the algebra acting on the material sector. The two objects share a name and little else, and the coinage is recorded here because the name runs through the rest of the corpus.

Quaternions in Electromagnetism, and the Vector Revolt

The first substantial physics to use quaternions was electromagnetism. James Clerk Maxwell's A Treatise on Electricity and Magnetism (1873) employed quaternion notation for some of its vector operations, and in the decades after his death the quaternion calculus became a standard language for the field: topics now written with vectors — kinematics in space and Maxwell's equations among them — were described in quaternion terms. Peter Guthrie Tait, Maxwell's friend and the leading quaternionist after Hamilton's death, published an Elementary Treatise on Quaternions in 1867 (written with Hamilton's advice and published after Hamilton's death) and an Introduction to Quaternions with Philip Kelland in 1873. The quaternion product was attractive because the dot and cross products are cut from a single operation, and the quaternion operator $\nabla$ carried gradient, divergence, and curl together.

The algebra's hold on physics did not last. From the mid-1880s, vector analysis — developed by Josiah Willard Gibbs and Oliver Heaviside — displaced the quaternions in physics and engineering. Heaviside's Electromagnetic Theory (1893) taught the field in vector form, and Gibbs's lectures, published as Vector Analysis by Edwin Bidwell Wilson (1901), gave the notation its textbook. The reasons were practical rather than algebraic: the vector calculus is notationally cleaner, it separates the dot and cross products instead of carrying the full quaternion product, and it suited the needs of electrical engineering and the emerging field theories. The transition was contested — the quaternionists, organized in a Quaternion Association (Alexander Macfarlane was its secretary from 1896, its president in 1909, and edited its Bibliography of Quaternions in 1904), argued the case for years — but the outcome was not close. Michael J. Crowe's A History of Vector Analysis (1967) is the standard account.

This part of the story concerns real quaternions and their vector parts almost entirely. The biquaternions played little role in the electromagnetic debates; their physics appearance came later and was about a different subject.

The Mechanical Origin of the Field Equations

The section above tells the story of a notation. A second and older strand runs from mechanics into the content of the field equations, and the corpus's account of where the equations come from is incomplete without it.

James Clerk Maxwell's first formulation of electromagnetism, the memoir On Physical Lines of Force (1861), was built on an explicit mechanical model. The field was carried by a lattice of incompressible "molecular vortices", and magnetism was their rotation; Maxwell read Faraday's law as the couple exerted on the vortices by the tangential (electric) forces, and he introduced the displacement current as the elastic reaction of the vortices to being strained, the step that gave him transverse propagation. Hermann von Helmholtz, working on organ pipes, had written the equations of vorticity in fluids in 1858, and the correspondence between fluid vorticity and the magnetism of currents was one of the routes through which the field equations were first understood. Lord Kelvin's dictum states the standard of the period: the test of whether a point of physics is understood is whether a mechanical model of it can be made.

The point of recording this is chronological. The corpus's history traces the algebra's road into physics — Conway and Silberstein's relativity, Lanczos's reformulation of the Dirac equation — and that is a twentieth-century road. The mechanical route is earlier, and it ran in the opposite direction: mechanics produced the field equations rather than interpreting them after the fact. Two cautions belong with it. The model is not a derivation of the equations from mechanics, and the correspondence was not maintained as the theory matured; and the original texts are hard to read for a reason the corpus understands well, namely that vector notation did not yet exist, so the equations were written in components and in forms that no longer match how they are taught. The analogy itself, its technical content and its limits, is recorded in Maxwell's Equations in the Biquaternionic Formulation.

The Group Before the Physics

The Lorentz group had a history of more than half a century before it took the name of Lorentz, and that history is geometry, not electromagnetism. Transformations equivalent to the Lorentz transformations in various dimensions were studied through the nineteenth century in the theory of quadratic forms, in hyperbolic geometry, in Möbius geometry and in sphere geometry. The indefinite quadratic form whose invariance defines the group is the same object as the Cayley–Klein metric of the hyperboloid model, so the group appears as the group of motions of hyperbolic space; the same group acts on the sphere by Möbius transformations; and it appears again, under the name of Edmond Laguerre, as the group of Laguerre transformations, the stabiliser of the family of planes inside the Lie sphere group of Möbius and Lie Sphere Geometry. The identification is one statement in three geometries: in the relevant dimensions, and up to the usual finite covers, the group of motions of hyperbolic space, the Möbius group $\mathrm{PSL}(2,\mathbb{C})$, and the Laguerre group are the Lorentz group.

The physical literature reached the identification late. Harry Bateman (1910–12), in the work with Ebenezer Cunningham on the transformations that leave the spherical wave equation invariant, alluded to the identity between the Laguerre inversion and the Lorentz transformation; the general isomorphism between the Laguerre group and the Lorentz group was pointed out by Élie Cartan (1912, and again in his Encyclopédie article of 1915) and by Henri Poincaré (1912). Gustav Herglotz (1909–10), following Klein and Fricke on the Cayley absolute, classified the one-parameter Lorentz transformations as loxodromic, hyperbolic, parabolic and elliptic — the four types of Möbius transformation, which is the same classification read in hyperbolic geometry.

The group also had a notation of its own before it had a name, in the hyperbolic versor. Hamilton's versor is a unit quaternion $\cos\theta + r\sin\theta$ with $r^2 = -1$, the exponential of a rotation generator; the same expression with the sign reversed, $\cosh t + r\sinh t$ with $r^2 = +1$, is the exponential of a boost generator and needs an algebra in which some element squares to $+1$. The element appeared first in James Cockle's tessarines (1848) — the commutative algebra $\mathbb{C}\otimes_{\mathbb{R}}\mathbb{C}$ that the corpus records as the tessarine or reduced biquaternion algebra in List of Algebras by Dimension — was used with quaternion multiplication by Homersham Cox (1882), and became the hyperbolic quaternions of Alexander Macfarlane (1891), an algebra whose difficulties, by Macfarlane's own account, put the subject on the road to the biquaternions after 1900; in a review of 1899 he drew the distinction the construction rests on, that a root of a quadratic may be "versor in nature or scalar in nature", circular in the former case and hyperbolic in the latter. The parameter of the family $\cosh t + r\sinh t$ was identified with the change of frame by Alfred Robb (1911), whose Optical Geometry of Motion names the rapidity; with the arrival of special relativity the one-parameter group of a hyperbolic versor came to be called a Lorentz boost. The vocabulary of Hamilton and Macfarlane was later subsumed by Lie's: for every $r$ with $r^2 = \pm1$ the map $t\mapsto\exp(tr)$ is a one-parameter subgroup, ordinary or hyperbolic.

The point of the section is a comparison of two prehistories. The article's refrain about the algebra has a parallel on the group side: the algebra is Hamilton's of the 1840s, and the group is Klein's, Lie's and Laguerre's of the 1870s and 1880s. In both cases the mathematics came first and the physical reading came second, and in neither case is the outcome a near-miss to be regretted: both were published, and both lost a competition of notations to the tensor and vector calculus.

The Biquaternion Formulation of Relativity

Special relativity is where the biquaternions came closest to becoming a standard physical language, and the closeness is structural before it is historical.

The structural fact is recorded above: the unit-norm biquaternions are $SL(2,\mathbb{C})$, the double cover of the Lorentz group, and they act on the four-dimensional real space $\mathbb{M}_{-}$ by rotor conjugation. A four-vector of relativistic physics is an element of $\mathbb{M}_{-}$; a Lorentz transformation is a unit biquaternion; the invariance of the interval is the invariance of the biquaternion norm under the action. The biquaternion algebra therefore contained the kinematics of special relativity as soon as the physics was written down — not because anyone looked for it there, but because the group of unit biquaternions had sat inside Hamilton's algebra since the 1840s.

The German route came first, and it is the one usually left out. The biquaternion Lorentz transformation was written down before Conway's paper and before Silberstein's, in the school around Felix Klein. Klein had described Cayley's four-dimensional quaternion multiplication as a Drehstreckung — an orthogonal substitution leaving a quadratic form invariant up to a factor — and had said that Minkowski's relativity is essentially the consequent application of such substitutions, without carrying the remark further. The working out was Fritz Noether's, in an appendix to Klein and Sommerfeld's Über die Theorie des Kreisels (1910): Noether wrote the hyperbolic rotations with biquaternions whose imaginary unit $\omega$ obeys $\omega^2 = -c^2$, tied that unit to the speed of light, and called the construction the principal ingredient of a rational representation of the group of Lorentz transformations. Citing Noether, Klein himself published quaternion substitutions forming that group in August 1910 and again in March 1911. The biquaternion Lorentz transformation was therefore in print in the Klein–Sommerfeld circle in 1910, a year before the priority exchange between Conway (February 1911) and Silberstein (November 1911) recorded below. This does not change the conclusion drawn in this section — the formulation was published, and it lost a competition of notations — but it corrects the attribution of the first publication, and it shows that the algebra entered relativity from geometry and group theory rather than from electromagnetism.

The historical record then shows that this formulation was published, not merely available. Arthur W. Conway applied biquaternion algebra to special relativity, publishing an article in 1911 and in 1912 asserting priority over Ludwik Silberstein, who had independently applied biquaternions to relativity; Conway's claim was later backed by George Temple, and in 1915 Conway published a 43-page tract, Relativity, in Edinburgh. Silberstein's treatment appeared as "Quaternionic form of relativity" (Philosophical Magazine 23 (1912) 790–809), with a second memoir in 1913, and it became a textbook: The Theory of Relativity (Macmillan, 1914; second edition, 1924), which used biquaternions throughout. Silberstein had also introduced, in 1907, the complex three-vector $\mathbf{E} + ic\mathbf{B}$ of the electromagnetic field — now called the Riemann–Silberstein vector. The corpus's Maxwell article writes the same object, up to a normalisation, as $i\sqrt{\epsilon}\,\mathbf{E}-\sqrt{\mu}\,\mathbf{H}$. (The attribution of the name's "Riemann" to Bernhard Riemann I have not verified; the 1907 introduction is Silberstein's. This is flagged in the companion.)

Around the same time, Alexander Macfarlane developed his "space analysis," an adaptation of quaternions to physics; the historical record notes that his first publication on it preceded Minkowski's presentation of spacetime by seventeen years. Macfarlane's hyperbolic-quaternion program was not the biquaternion formulation, but it belongs to the same moment in which quaternion methods were being stretched toward the geometry of spacetime.

The displacement, when it came, was by Minkowski's four-dimensional tensor formulation (1908) and the four-vector calculus that grew from it. The tensor formalism was explicitly covariant, it generalized without alteration to curved spacetime, and it separated vectors, covectors, and tensors in a way the quaternion product did not. It is worth stating plainly, because "near-miss" suggests otherwise: the biquaternion relativity of Conway and Silberstein was a published, textbook-level formulation that lost a competition of notations, not a possibility that was never realized.

The Later Recognition: Biquaternions and the Quantum Algebra

The connection of $\mathbb{B}$ to quantum mechanics is the sharpest structural fact in this history and the weakest historical one.

The structural fact is the isomorphism $\mathbb{B} \cong M_2(\mathbb{C})$: the biquaternion algebra is the algebra generated by the Pauli matrices, the algebra of operators on a two-state system. In the language of the companion articles, the Hermitian subspace $\mathbb{M}_{+}$ carries the observables, its idempotents the pure states, and the trace formula is the Born rule. Nothing in the algebra of a qubit lies outside $\mathbb{B}$.

The historical chronology, however, runs the other way, and by a wide margin. Élie Cartan introduced spinors in geometry in 1913, without physics. Wolfgang Pauli introduced the three spin matrices for the electron in 1927 ("Zur Quantenmechanik des magnetischen Elektrons," Zeitschrift für Physik 43 (1927) 601–623), and Paul Dirac gave the relativistic electron equation in 1928. In 1929 the Dirac equation was reformulated in tensor-analytic and covariant terms by Cornelius Lanczos ("Die tensoranalytischen Beziehungen der Diracschen Gleichung" and its companion, Zeitschrift für Physik 57 (1929) 447–473, 474–483, 484–493); Lanczos's doctoral dissertation — handwritten in 1919, The Functional Theoretical Relationships of the Maxwell Aether Equations — had already rewritten Maxwell's equations in quaternion terms and applied a relativistic variational principle to them. In the same year, Hermann Weyl and Bartel van der Waerden brought the two-component spinor into physics. Gsponer and Hurni's commentary on that dissertation (1998) records a further first in its third chapter: the quaternion generalisation of the Cauchy–Riemann regularity conditions and of Cauchy's integral formula, the latter now called the Cauchy–Lanczos–Fueter formula, which Moisil and Fueter rediscovered in 1931, more than twelve years later. The corpus records that priority as the commentators' claim, resting on the dissertation itself; what is structural here is that Lanczos's electrodynamics is a function theory of the homogeneous Maxwell equation, which the article on radiation from accelerated charges takes up. The algebra these developments needed had existed since Hamilton; the developments did not derive from it, and the standard history of quantum mechanics does not run through the biquaternion literature.

Even the later quaternion work is a separate line: Conway returned to quaternions for quantum mechanics only at the end of his life, in work published in 1948 and cited in a 1950 thesis by Joachim Lambek, and the recognition that the biquaternion algebra and the Pauli algebra are the same became standard in the geometric-algebra literature of the 1960s. The honest statement is two-part: the algebra contains the quantum structures, and no one in the biquaternion tradition is documented as having used it to anticipate them. Only the second kind of connection would justify the word "anticipation."

Separating $i$ from $\hbar$. One reading of the algebra's quantum content is a claim about what the algebra clarifies rather than about priority. In the Schrödinger equation the imaginary unit and the quantum of action appear as one combined factor, $i\hbar\,\partial_t$; in the Lanczos biquaternion equation $i$ sits in the scalar part of the four-gradient and $\hbar$ multiplies the mass, so the two are separated. Gsponer and Hurni read the separation as an identification: $i$ belongs to time — Hamilton's own intuition about imaginary numbers — and $\hbar$ belongs to mass, that is, to the particle aspect of a wave, a lump of energy localized in space. On the same reading Maxwell's theory needs no $\hbar$ at all and can be read as a quantum theory without it. The corpus records both as the commentators' reading; what the algebra demonstrably does is exhibit the two ingredients of the quantum formalism as independent places in one equation, rather than as one product.

Why the Biquaternions Were Displaced

The biquaternions were not abandoned because the algebra failed. They were displaced by notations better suited to the practice of physics, for separable reasons.

Vector analysis. The vector calculus of Gibbs and Heaviside was easier to teach, easier to compute with, and adequate for the field theories of the period. It took the parts of the quaternion product that physics used — the dot and cross products — and discarded the rest, and it won on usability.

Tensor calculus and differential geometry. Minkowski's four-tensor formulation of relativity (1908) and the tensor calculus of Ricci and Levi-Civita made covariance manifest, distinguished vectors from covectors and one-forms, and generalized directly to the curved spacetime of general relativity. The quaternion product has no corresponding tensor apparatus; when physics moved to differential geometry, the quaternion notation had no place to stand.

The $ict$ convention and its later abandonment. The complex structure that the biquaternion formulation makes explicit appeared in mainstream practice as the $ict$ convention — the imaginary time coordinate that makes the Minkowski interval look Euclidean. The companion article Why Complexify Spacetime? records why that convention was given up: it is tied to a global inertial frame, it sits awkwardly with spinors and with curved spacetime, and in quantum field theory the Wick rotation is better understood as a change of causal structure than as a change of coordinates. The abandonment of $ict$ was not an abandonment of the complex structure, but it removed the one place where the complexification was visible in ordinary practice.

A failure internal to the algebra: order reversal. Gsponer and Hurni add a diagnosis of a different kind, one that is internal to the algebra rather than external to it. The quaternion product of two quadruplets $[a;\vec a][b;\vec b]=[ab+p\,\vec a\cdot\vec b\,;\,a\vec b+b\vec a+q\,\vec a\times\vec b]$ carries two constants tied by associativity in the form $p+q^{2}=0$: the sign of $p$ fixes the signature of the metric, and the sign of $q$ fixes the handedness of the vector product. Reversing the order of the factors reverses $q$, so there is an involution — order reversal, $(\;)^{\sim}$ — tied to space inversion; unlike the choice of $p$, the sign of $q$ has no overall invariant criterion. A quantity whose meaning does not depend on that sign is called order-reversal covariant (ordinal covariant), and the requirement is specific to quaternions and Clifford numbers. Their claim is that early quaternion physics often met with problems by failing to impose it — treating fundamental quantities as order-reversal invariant when they are only covariant — and that this is part of why the algebra was abandoned. Their example is the compact Conway–Silberstein Maxwell system $\tilde\nabla\tilde A=\tilde B$, $\tilde\nabla\tilde B=-4\pi\tilde J$, which carries the supplementary conditions $\mathrm{Sc}(\tilde\nabla\tilde A)=0$, $\tilde A=\tilde A^{\sim}$, $\tilde J=\tilde J^{\sim}$; the gauge- and order-reversal-invariant pair $\mathrm{Vec}(\tilde\nabla\tilde A)=\tilde B$, $\tfrac12(\tilde\nabla\tilde B+\tilde B^{\sim}\tilde\nabla)=-4\pi\tilde J$ needs none of them (the first is the source's $\tilde\nabla\wedge\tilde A$, the vector part). The diagnosis is a claim about the practice of the algebra, and it is recorded here as the commentators'.

Community and curriculum. The quaternionists' program was a school with a doctrine; the vectorists' program was a tool set with users. Such a contest is settled by adoption, not by proof, and adoption followed the textbooks and the engineering curriculum.

The through-line is that the biquaternions lost on notation and practice, not on content — which is exactly what makes their return possible.

The Return

The biquaternions and the closely related Clifford algebras returned to physics in the second half of the twentieth century, through several independent routes.

Feza Gürsey's work on conformal-invariant spinor and quaternion equations in the 1950s reopened the quaternion formulation of the Dirac field ("On a conform-invariant spinor wave equation," Il Nuovo Cimento 3 (1956)). In 1962, David Finkelstein, Josef Maria Jauch, Samuel Schiminovich, and David Speiser gave a systematic "Foundations of Quaternion Quantum Mechanics" (Journal of Mathematical Physics 3 (1962) 207–220), the first sustained attempt to build quantum theory on a quaternionic Hilbert space. In 1966 David Hestenes's Space-Time Algebra developed the Clifford algebra $Cl_{1,3}$ as the natural language of relativistic physics; the real Pauli algebra $Cl_{3,0}(\mathbb{R})$ is the biquaternion algebra in real form, and is used in physics under the name "algebra of physical space." Stephen Adler's Quaternionic Quantum Mechanics and Quantum Fields (Oxford, 1995) collected the mid-century program and examined its physical consequences.

The wave-equation line, Lanczos to Petiau. A second route into the algebra ran through the relativistic wave equations themselves. Lanczos's 1929 papers, cited above for the tensor-analytic form of the Dirac equation, write the spin-$\tfrac12$ equation as the coupled system $\tilde{\nabla}\tilde{A}=m\tilde{B}$, $\tilde{\nabla}\tilde{B}=m\tilde{A}$, and the doubling of the solution that the system carries was later read as isospin: Gürsey's London thesis (1950) reviewed the quaternionic theory of Lanczos and his contemporaries, and his papers of 1957–58 gave the neutron as the second idempotent superposition of the pair — equal mass, different charge — with the two-parameter group K. Nishijima wrote down in 1957. Einstein and Mayer (1933) generalised the construction by promoting the mass to a biquaternion and treating the wave equations of different spins as semivectors of one system; Petiau (1958, 1965) closed the system by giving the mass field an equation of its own, at the price of nonlinearity, and found its double-periodic elliptic solutions. This is the line that the corpus's articles on the Proca equation, the Petiau system and the proton–neutron doublet develop.

The Lanczos revival, and its register. The line was reopened at the end of the twentieth century by Adrian Gsponer and Jean-Pierre Hurni, whose work on Lanczos's electrodynamics the corpus draws on in several articles. Three of their items are distinct and should be kept so. (i) A commentary on Lanczos's doctoral dissertation, arXiv:math-ph/0402012v2, which shows the 1919 Lagrangian to be equivalent to the standard Lagrangian of classical electrodynamics and dates the quaternion Cauchy–Riemann conditions to 1919 — the priority over Moisil and Fueter that the chronology above records. (ii) Their 2005 Foundations of Physics paper on Lanczos's derivation of the usual action integral. (iii) A four-page contribution to the Cornelius Lanczos centenary conference (1994, arXiv:hep-ph/0112317), which is the sharpest claim in the whole line: it presents the Lanczos quaternionic equation as a replacement for Dirac's equation, describes particles of spin $0$, $\tfrac12$, $1$ and $\tfrac32$ in one framework, and reports two classes of solutions — point-like partons in two families, quarks and leptons, with the charges and baryon number coming out of the equations and the neutrino and u-quark masses zero by eigenvalue equations, and extended hadrons with a non-local strong interaction. Its electroweak sector is claimed identical to the Standard Model's with the same two free parameters and no Higgs mechanism.

The register of the third item matters as much as its content, and the corpus fixes it here so that no later reader mistakes it for a result. It is four pages, it is an abstract-level report, it exhibits no derivation, and its title is a question. The corpus therefore files it in its lowest evidential tier — claims reported, derivations not available in this source — and treats it as a pointer to a programme with a history rather than as evidence. The two tests it would have to pass are the ones the corpus applies to any mass-generation claim: relations for the $W$ and $Z$ masses, and the pattern of fermion masses and mixings, not merely the existence of eigenvalues. Neither is exhibited. The item is nonetheless worth recording, because the Lanczos line's own logic runs from a quaternionic reformulation of the Dirac equation to a claimed particle classification, and a history of the algebra that stopped before the claim would hide the direction the line was pointing.

A citation hazard. The corpus holds two copies of the 2004 commentary (item (i) above) that carry the same arXiv identifier, math-ph/0402012v2, and differ only in an internal date — one dated 6 June 2005, one dated 29 October 2018. They are one paper, not two, and should be cited by the arXiv identifier so that the duplication is visible.

Two names from Dublin, and three later writers. The objects the corpus calls the Hermitian subspace and the null cone have older names: in J. L. Synge's Dublin memoir on quaternions and the Lorentz group (1972) a bireal biquaternion — fixed by the biquaternion conjugation, the corpus's $\mathbb{M}_+$ — is a minquat, and a singular biquaternion — a zero divisor, hence an idempotent such as the $\sigma$ of the corpus's spinor construction — is a nullquat; the same memoir treats the Conway–Dirac–Eddington matrices. In the same decades Paul Weiss (1941), a student of Max Born, used the quaternion spinor decomposition of the four-velocity, $\tilde{U}=\tilde{B}\tilde{B}^{+}$, to write explicit formulas for the four-acceleration and the retarded null separation, and derived the Lorentz–Dirac equation from the condition that the energy–momentum flow through a hypersurface around the charge be stationary — plausibly the first significant use of spinors in classical physics; C. W. Kilmister (1955) set out the linear quaternion functions that tensor analysis requires; and the review volume of Gsponer and Hurni from which much of this section is drawn returned to Lanczos's dissertation in the 1990s.

A separate route is twistor theory: Roger Penrose's twistors live on complexified Minkowski space, with real Minkowski space as a slice, so the complexification that the biquaternion formulation makes algebraic is, in twistor theory, the fundamental arena. Modern biquaternionic treatments of the classical field equations — Silberstein's complex-vector Maxwell theory, the algebrodynamics of Vladimir Kassandrov and his collaborators, and the review work of Adrian Gsponer and Jean-Pierre Hurni — continue to use the algebra directly.

The programme's own rationale for the algebra. Gsponer and Hurni's review is the fullest statement by the programme's authors of why the algebra should be used at all, and its conclusion is methodological rather than physical: the physics is in the mathematical structure, not in the formalism. Their grounds are practical. Biquaternions are whole symbols — one object compounding between one and eight real numbers that belong to a single, or a few related, tensor quantities — so formulas are generally shorter than their matrix, tensor, or higher-rank Clifford counterparts, and at least one level of tensor index is dispensed with. They are expressive: elements of the simplest non-trivial Clifford algebra, they give final-form expressions directly usable in symbolic or numerical calculation. They are suggestive: they often indicate how to generalize a result, or how to relate results that look independent. And they are unifying: they allow complex numbers to be used consistently in classical and quantum physics, give expressions that are similar in Galilean and Lorentzian relativity, and formulate the physics of the Standard Model compactly. The caution the same authors attach to the list is the one this article has already met in the order-reversal paragraph: because the symbols are whole, an expression has a definite tensor character only if it is built from elementary symbols that have one, and non-commutativity and order-reversal covariance must be handled explicitly rather than assumed. The corpus records the rationale as the commentators', and keeps it apart from the derivation question — the four grounds are reasons the algebra is usable, not evidence that it is nature's. The companion article The Standard Model under the Biquaternion Framework records the same authors' caution on the other side, that reformulating standard physics in the algebra produces few genuinely new results.

The framework developed in this corpus belongs to this return. The corpus reads the two subspaces $\mathbb{M}_{-}$ and $\mathbb{M}_{+}$ as a material sector and an informational sector, with the rotor conjugation as the coupling between them. That reading is a proposal: it is enabled by the algebra, and the algebra is old, but the historical record does not contain the proposal and nothing in the history vindicates it. The history establishes the existence and richness of the algebra, not the truth of the interpretation.

What the History Does and Does Not Show

The record shows that the algebra was Hamilton's and was named by him; that its structural identifications with the Pauli algebra, the Lorentz double cover, and the qubit operator algebra are theorems about $\mathbb{B}$; that the geometry of lines and screws was built on the quaternion units in 1873, in elliptic space, and gave the word "rotor" to the vocabulary; that the group it generates was studied as a transformation group in hyperbolic, Möbius and Laguerre geometry before it had any physics; that biquaternions were applied to that group by Noether and Klein in 1910 and to special relativity by Conway and Silberstein in 1911, and then displaced by tensor and vector notation; and that the algebra returned in the mid-twentieth century through quaternionic quantum mechanics, geometric algebra, and twistor theory.

The record does not show that the biquaternion algebra is nature's, that any near-miss would have changed the physics, or that the biquaternion tradition anticipated quantum mechanics. The needed structures were in the algebra; no documented path runs from Hamilton's or Conway's work to Pauli's or Dirac's, and the connection is a structural identity recognized in retrospect. Nor does the record support the corpus's material/informational reading, which is a proposal placed on an old algebra.

The distinction on which this turns is between an algebra containing a structure and a person using it. Containment is a theorem of the 1840s; use is an event of the 1910s or the 1920s. "Near-miss" conflates the two, and this article uses it only where the documented record — the published relativity of Conway and Silberstein — justifies it.

Summary

The biquaternions are the complex quaternions, $\mathbb{B} = \mathbb{C}\otimes_\mathbb{R}\mathbb{H}$, introduced and named by Hamilton and treated at length in his Lectures on Quaternions (1853). The motive for the complexification was algebraic: in $\mathbb{H}$ a quadratic equation has no quadratic formula, its solution set need not be finite, and its resolution can call for a scalar square root of $-1$ — that is, for complex coefficients. As an algebra the biquaternions are $M_2(\mathbb{C})$, equivalently the Pauli algebra $Cl_{3,0}(\mathbb{R})$ and the even subalgebra of the spacetime algebra $Cl_{1,3}$. Their unit-norm elements are $SL(2,\mathbb{C})$, the double cover of the Lorentz group, acting on the material sector $\mathbb{M}_{-}$ by rotor conjugation; the full algebra $\mathbb{B}\cong M_2(\mathbb{C})$ is the operator algebra of a two-state system, of which $\mathbb{M}_{+}$ is the Hermitian subspace.

In physics, real quaternions first entered through Maxwell's electromagnetism and Tait's advocacy, and were displaced from the mid-1880s by the vector analysis of Gibbs and Heaviside. Biquaternions were applied to the Lorentz group by Noether and Klein in 1910 and to special relativity by Conway and by Silberstein in 1911, most fully in Silberstein's The Theory of Relativity (1914), and were displaced by Minkowski's tensor formalism. The group itself, under the names of hyperbolic motions, Möbius transformations and Laguerre transformations, had been studied in geometry since the 1870s. The group also carried a one-parameter notation before it had a name, in the hyperbolic versors of Cockle, Cox, Macfarlane and Robb, whose parameter is the rapidity of a Lorentz boost. The algebra returned in the mid-twentieth century through quaternionic quantum mechanics, geometric algebra, twistor theory, and modern biquaternionic field theory, and through a direct transcription of the relativistic wave equations themselves — Lanczos's coupled system and its doubling, read by Gürsey as isospin; Einstein and Mayer's biquaternionic mass and semivectors; Petiau's closure of the mass field; and Weiss's quaternion derivation of the Lorentz–Dirac equation.

The history shows that the algebra is old, rich, and structurally sufficient for relativity and for the qubit. It does not show that the algebra is nature's, nor that the biquaternion tradition anticipated the physics that later used it. The gap between containing a structure and having used it is the honest content of this history.

The word "rotor" is Clifford's, coined in 1873 for a quantity having position — the line element of the geometry of screws — and his biquaternions, the ratios of two such line elements, are the quaternions with split-complex coefficients, the algebra of the lines and screws of elliptic space; the complex quaternions this article calls the biquaternions are Hamilton's.

Summary of Notation

Symbol	Meaning
$\mathbb{B} = \mathbb{C}\otimes_\mathbb{R}\mathbb{H}$	Biquaternion algebra (complex quaternions)
$e_0 = 1, e_1, e_2, e_3$	Quaternion basis, $e_k^2 = -e_0$
$i$	Scalar imaginary, $i^2 = -1$ (Hamilton's $h$)
${}^{\natural}, \bar{\cdot}, {}^{*}, {}^{\flat}$	The four conjugations
$\mathbb{C}_{\mathbb{B}}$	Complex subspace (complex scalars; center of $\mathbb{B}$)
$\mathbb{H}_{\mathbb{B}}$	Quaternion subspace (real quaternions; norm $(4,0)$)
$\mathbb{M}_{+}$	Hermitian subspace (informational sector; norm $(1,3)$)
$\mathbb{M}_{-}$	Anti-Hermitian subspace (material sector; norm $(3,1)$)
$N(\tilde{Q}) = \tilde{Q}\tilde{Q}^{\natural} = \sum_\mu Q_\mu^2$	Biquaternion norm (zero divisors where $N=0$)
$\mathbb{B} \cong M_2(\mathbb{C}) \cong Cl_{3,0}(\mathbb{R})$	Pauli algebra; $\Phi(e_k) = -i\sigma_k$
$SL(2,\mathbb{C}) = \{\tilde{\Lambda} : \tilde{\Lambda}\tilde{\Lambda}^{\natural} = e_0\}$	Unit-norm biquaternions; Lorentz double cover
$\tilde{Q} \mapsto \tilde{\Lambda}\tilde{Q}\tilde{\Lambda}^{*}$	Rotor conjugation on $\mathbb{M}_{-}$
$\tilde\Pi_\pm = \tfrac{1}{2}(e_0 \pm i\hat{\mu})$	Idempotent (pure state of $\mathbb{M}_{+}$)
$\mathrm{Tr}(\tilde{P}\tilde{H}) = 2\,\mathrm{Sc}(\tilde{P}\tilde{H})$	Trace formula (Born rule)
$\mathbf{E} + ic\mathbf{B}$	Riemann–Silberstein vector (Silberstein, 1907)
$j$	Split-complex unit, $j^{2} = +1$ (Clifford's twist $\omega$)
$\Pi_1 = \tfrac{1}{2}(1 + j), \ \Pi_2 = \tfrac{1}{2}(1 - j)$	Idempotents; right and left vector parts of a motor
$\mathbb{H}$	Real quaternion algebra (Hamilton 1843)
$Uq = q/\lvert q\rvert$, $Tq = \lvert q\rvert$	Versor and tensor of a quaternion (Hamilton)
$ax = b$, $x = a^{-1}b$	Linear equation in $\mathbb{H}$; the side on which $a$ stands matters
$x^{2} + bx + c = 0$	Quaternion quadratic: no completing the square, no discriminant
$x^{2} + 2e_2x - 3 = 0$	Worked quadratic: roots $\pm\sqrt{2} - e_2$ in $\mathbb{H}$, and $(-1 \pm \sqrt{2}\,i)e_2$ in $\mathbb{B}$