Maxwell's Equations in the Biquaternionic Formulation
Introduction
Maxwell's equations are the clearest illustration of why the biquaternionic formulation is natural. In their biquaternionic form, they collapse into a single equation relating a biquaternionic field strength to a biquaternionic source. With the $i r'_t$ structure built into the Minkowski subspace, the biquaternionic structure becomes manifest, and the Lorentz invariance of electromagnetism appears as a rotor conjugation in the Minkowski subspace.
This article develops the biquaternionic formulation of Maxwell's equations in a material medium characterized by permittivity $\epsilon$ and permeability $\mu$, then takes the vacuum limit $\epsilon = \epsilon_0$, $\mu = \mu_0$. The presentation proceeds in the natural order: first the standard Maxwell equations in three-vector form, then the potential biquaternion $\tilde{A}_{\Phi}$, then the biquaternionic gradient $\tilde\nabla_\Phi$, then the field-strength biquaternion $\tilde{F}$, and finally the single biquaternionic equation that replaces the four standard Maxwell equations. It then works through the retarded Green's function, the three-vector and stationary forms, the gauge structure, the hydrodynamic correspondence, the relation between the two formulations and the matrix route that sits inside it, the author's variant with its added scalar, the biquaternionic energy–momentum and the energy law it satisfies, the Lorentz transformation of the potential, the vacuum limit and the question of what is fundamental, the field bivector against the field operator, the vacuum equation read as a Cauchy–Riemann condition, the internal freedom that equation leaves, and the complexified extension; it closes with the frame basis operators that give the gamma names.
Throughout this article, the symbol $c$ denotes the speed of light in the medium: $c = 1/\sqrt{\epsilon\mu}$. In vacuum, $c$ reduces to the constant vacuum speed of light $c_0 = 1/\sqrt{\epsilon_0\mu_0}$. The symbol $v$ (and $\vec v$) is reserved for particle and frame velocities; the fluid velocity is $\vec v_f$ and the boost velocity is $\vec v_b$. This convention keeps the notation consistent with relativistic mechanics, where $c$ is the speed of light and $v$ is a velocity.
The algebra, its units and its central imaginary are those of Algebra and Element Conventions, §The Base and Its Units: the scalar $i$ is central, and a product of two factors of $i$ is a sign, so every display below is reduced with $i^2 = -1$ rather than left with a doubled $i$. The body is written in elements and in the products of the algebra; the one exception is the reported matrix display of A fourth route below, which is the object of that subsection and of nothing else. The gamma names appear only in The Legacy Pauli Gamma Formulation, where they name frame basis operators and where the same rule fixes their signs.
Maxwell's Equations in a Material Medium
In a linear, isotropic, non-dispersive medium characterized by permittivity $\epsilon$ and permeability $\mu$, Maxwell's equations in three-vector form are
$$ \mathrm{rot}\,\vec E = -\frac{\partial \vec B}{\partial r'_t}, $$
$$ \mathrm{rot}\,\vec H = \frac{\partial \vec D}{\partial r'_t} + \vec J, $$
$$ \mathrm{div}\,\vec D = \rho, $$
$$ \mathrm{div}\,\vec B = 0, $$
with the constitutive relations
$$ \vec D = \epsilon\,\vec E, \qquad \vec B = \mu\,\vec H. $$
Here $\vec E$ is the electric field, $\vec H$ the magnetic field, $\vec D$ the electric displacement, $\vec B$ the magnetic induction, $\rho$ the free charge density, and $\vec J$ the free current density. The constants $\epsilon$ and $\mu$ are properties of the medium. In vacuum they take the values $\epsilon_0$ and $\mu_0$.
The speed of electromagnetic waves in the medium is
$$ c = \frac{1}{\sqrt{\epsilon \mu}}. $$
In vacuum this becomes
$$ c_0 = \frac{1}{\sqrt{\epsilon_0 \mu_0}}. $$
The local speed of light $c$ is the factor that controls propagation in the medium, and it enters the equations in the same way for all media. In vacuum, $c = c_0$.
The source fields $\rho$ and $\vec J$ are not independent: they satisfy the charge conservation law
$$ \mathrm{div}\,\vec J + c\,\frac{\partial \rho}{\partial r'_t} = 0, $$
which follows from taking the divergence of the Ampère–Maxwell law and substituting Gauss's law. This constraint is the integrability condition for the Maxwell system, and it will reappear below as the constraint on the biquaternionic source $\sqrt{\mu}\,\tilde{J}_{\Phi}$.
The Potential Biquaternion
The electromagnetic field is described by a potential biquaternion
$$ \tilde{A}_{\Phi} = A_0 + \vec A, \qquad A_0 = i\,a'_t, $$
where $a'_t=\xi_{\mathrm{scp}}/c$ is the temporal parameter, the scalar potential in the unit of $\vec A$, and $\vec A$ is the vector potential. The coefficient of $e_t$ is $A_t=i\,a'_t$, the imaginary case of the general $A_t=a_t+i\,a'_t$. The tilde signals that $\tilde{A}_{\Phi}$ is an element of the biquaternion algebra $\mathbb{B}$, not a four-vector.
The scalar part $A_0 = i\,a'_t$ is purely imaginary, matching the complex time coordinate $i r'_t$. The vector part $\vec A$ is the ordinary vector potential. The factor of $i$ in $A_0$ is the same factor that appears in the complex time coordinate.
The coordinate representation of $\tilde{A}_{\Phi}$ as a four-vector potential is
$$ \tilde{A}_{\Phi} = \sum_{\mu\in\{t,x,y,z\}} A_\mu e_\mu \;\longleftrightarrow\; A_\mu = (A_0, A_1, A_2, A_3), $$
with the four $e_\mu$ the coefficient basis of Algebra and Element Conventions, §The Base and Its Units, and
$$ A_0 = i\,a'_t, \qquad (A_1, A_2, A_3) = \vec A. $$
The four-vector representation is a coordinate representation of the biquaternion, not the fundamental object.
The Biquaternionic Gradient
The operator this article differentiates with is the biquaternionic gradient of Operator and Calculus Conventions, §The Gradient: an element of four terms, one per coordinate, whose quaternion conjugate is the same element with the three spatial signs reversed. Both are the conventions' objects, and neither is defined here.
Two facts about the gradient carry the rest of the article. On the material slice, where $Q_t=i q'_t$, the time derivative carries one factor of $i$,
$$ \frac{\partial}{\partial R_t} = -i\frac{\partial}{\partial r'_t}, $$
the substitution fixed in §Conventions for Derivation of the same article; that factor is the one that combines with $A_0 = i\,a'_t$ below, where the product of the two factors of $i$ is a sign. And the product of the gradient with its conjugate is the d'Alembertian times the unit,
$$ (\tilde\nabla_\Phi)^{\natural}\tilde\nabla_\Phi = \tilde\nabla_\Phi(\tilde\nabla_\Phi)^{\natural}= \Box_\Phi\,e_t , $$
$\Box_\Phi$ being the quaternionic norm of the gradient, the scalar operator of §The Quaternionic Bilinear Product, Its Form and Its Norm, which also gives its form on the material slice. The identity is the conventions' and is not derived here; the unit $e_t$ is dropped where $\Box_\Phi$ acts on a scalar, which is everywhere below.
The Field-Strength Biquaternion
The electromagnetic field is described by a field-strength biquaternion
$$ \tilde{F} = \vec F, \qquad \vec F = i\sqrt{\epsilon}\,\vec E - \sqrt{\mu}\,\vec H. $$
Here $\vec F$ is a complex three-vector combining the electric and magnetic fields, with an imaginary electric part and a real magnetic part: the electric field carries a time index, and in the $i r'_t$ convention it is the time direction that supplies the factor $i$. The tilde signals that $\tilde{F}$ is a biquaternion with vanishing scalar part. The factors $\sqrt{\epsilon}$ and $\sqrt{\mu}$ are the natural normalization factors that make the biquaternionic product symmetric between electric and magnetic contributions.
The field strength is the antisymmetric part of the gradient of the potential. In the biquaternion algebra it is the vector part of $(\tilde\nabla_\Phi)^{\natural}\tilde{A}_{\Phi}$, divided by $\sqrt\mu$:
$$ \tilde{F} = \frac{1}{\sqrt{\mu}}\Bigl((\tilde\nabla_\Phi)^{\natural}\tilde{A}_{\Phi} - \mathrm{Sc}\!\left((\tilde\nabla_\Phi)^{\natural}\tilde{A}_{\Phi}\right)\Bigr). $$
The computation is three steps, and the normalisation constant is fixed in the third.
Step 1: the product. With $\tilde A_{\Phi} = A_0 + \vec A$ and $(\tilde\nabla_\Phi)^{\natural} = e_t\partial/\partial R_t - \vec e_\eta\partial/\partial R_\eta$, the four products give
$$ \mathrm{Sc}\bigl((\tilde\nabla_\Phi)^{\natural}\tilde{A}_{\Phi}\bigr)=\frac{\partial A_0}{\partial R_t}+\mathrm{div}\,\vec A, $$
$$ \mathrm{Vect}\bigl((\tilde\nabla_\Phi)^{\natural}\tilde{A}_{\Phi}\bigr)=\frac{\partial \vec A}{\partial R_t}-\vec\nabla_\Phi A_0-\mathrm{rot}\,\vec A . $$
The scalar part comes from $e_te_t=e_t$ and from the minus in $(\tilde\nabla_\Phi)^{\natural}$ against $\vec e_\eta\vec e_\eta=-e_t$; the rotation is the antisymmetric part of $\vec e_\eta\vec e_j$.
Step 2: the fields. In the $i r'_t$ convention $\partial/\partial R_t = -i\,\partial/\partial r'_t$, and the two derivative terms become
$$ \frac{\partial \vec A}{\partial R_t}-\vec\nabla_\Phi A_0 =-i\frac{\partial\vec A}{\partial r'_t}-i\nabla a'_t =-\frac{i}{c}\Bigl(c\,\frac{\partial\vec A}{\partial r'_t}+\vec\nabla_\Phi\xi_{\mathrm{scp}}\Bigr) =\frac{i}{c}\,\vec E, $$
using $A_0 = i\,a'_t$, $a'_t=\xi_{\mathrm{scp}}/c$, $i^2=-1$ and $\vec E=-\vec\nabla_\Phi\xi_{\mathrm{scp}}-c\,\partial\vec A/\partial r'_t$, the time derivative being $c\,\partial/\partial r'_t$ because $r'_t$ is a length. With $\vec B=\mathrm{rot}\,\vec A$,
$$ \mathrm{Vect}\bigl((\tilde\nabla_\Phi)^{\natural}\tilde{A}_{\Phi}\bigr) =\frac{i}{c}\vec E-\vec B . $$
Step 3: the normalisation. Multiplying by $\sqrt\mu$, and using $\sqrt{\mu}/c=\mu\sqrt{\epsilon}$ and $\vec B=\mu\vec H$,
$$ \sqrt{\mu}\,\mathrm{Vect}\bigl((\tilde\nabla_\Phi)^{\natural}\tilde{A}_{\Phi}\bigr) =i\mu\sqrt{\epsilon}\,\vec E-\mu\sqrt{\mu}\,\vec H =\mu\Bigl(i\sqrt{\epsilon}\,\vec E-\sqrt{\mu}\,\vec H\Bigr) =\mu\,\tilde{F}, $$
so the relation is exact with one constant,
$$ \boxed{\;\mathrm{Vect}\bigl((\tilde\nabla_\Phi)^{\natural}\tilde{A}_{\Phi}\bigr)=\sqrt{\mu}\,\tilde{F}, \qquad \tilde{F} = \frac{1}{\sqrt{\mu}}\Bigl((\tilde\nabla_\Phi)^{\natural}\tilde{A}_{\Phi} - \mathrm{Sc}\bigl((\tilde\nabla_\Phi)^{\natural}\tilde{A}_{\Phi}\bigr)\Bigr)\;}. $$
In Heaviside–Lorentz units, where the two medium factors are absorbed in the definition of $\tilde F$, the factor $\sqrt\mu$ is absent and the relation reads $\tilde{F}=(\tilde\nabla_\Phi)^{\natural}\tilde{A}_{\Phi}-\mathrm{Sc}((\tilde\nabla_\Phi)^{\natural}\tilde{A}_{\Phi})$, the form quoted in the companion articles; in a medium whose relative permeability differs from one the factor is required, and dropping it would make the two sides differ in dimension. No sign convention is left open: the choice $A_0=+i\,a'_t$ fixes the imaginary electric part, and the opposite choice would reverse it.
The same object in tensor language. The four-potential has components $A_\mu$, and the field tensor is
$$ F_{\mu\nu} = \partial/\partial R_\mu A_\nu - \partial/\partial R_\nu A_\mu , $$
antisymmetric, $F_{\mu\nu}=-F_{\nu\mu}$, with $F_{0\eta}=-(i/c)(c\,\partial A_\eta/\partial r'_t+\partial \xi_{\mathrm{scp}}/\partial r_\eta)=iE_\eta/c$ and $F_{j\eta}=\epsilon_{j\eta l}B_l$. The six independent components are the six of $\vec F$: three in the imaginary electric part and three in the real magnetic part.
The complex combination $\vec F = i\sqrt{\epsilon}\,\vec E - \sqrt{\mu}\,\vec H$ has a long history. It was introduced by Ludwik Silberstein in 1907, in his work on the electromagnetic field as a complex three-vector, and is now known as the Riemann–Silberstein vector $\vec N = \vec E + ic\vec B$. The biquaternionic field strength is that vector with one constant phase and one normalisation,
$$ \tilde F = i\sqrt{\epsilon}\,\vec N = i\sqrt{\epsilon}\,(\vec E + ic\vec B), $$
which reproduces $\tilde F$ because $c\sqrt{\epsilon}=1/\sqrt{\mu}$ turns the magnetic term $i^2c\sqrt\epsilon\vec B=-\sqrt\mu\vec H$. The two formulations therefore differ by the constant factor $i\sqrt{\epsilon}$ and not by their content. The biquaternionic formulation is the natural algebraic home of this object: the complex vector $\vec N$ carries the same six components, the biquaternionic field strength is a pure vector, and the operations of the electromagnetic field theory ($\mathrm{rot}$, $\mathrm{div}$, and the wave operator) become biquaternion multiplication and differentiation. The historical construction of Silberstein and its modern biquaternionic formulation are therefore two expressions of the same structure.
The Biquaternionic Source
The source of the electromagnetic field is the four-current
$$ \tilde J_{\Phi} = J_0e_t + \vec J = ic\rho\,e_t + \vec J, $$
whose scalar part carries the charge density and whose vector part carries the current density. The field equation carries the medium through one normalisation constant,
$$ \tilde\nabla_\Phi\tilde F = -\sqrt{\mu}\,\tilde J_{\Phi} . $$
The factor $\sqrt{\mu}$ is the whole difference between the four-current and the combination $i\rho/\sqrt{\epsilon}\,e_t+\sqrt{\mu}\,\vec J$ that the component equations require, since $\sqrt{\mu}\,ic = i/\sqrt{\epsilon}$ by $c=1/\sqrt{\epsilon\mu}$.
The four-current is not arbitrary: the charge conservation law $\mathrm{div}\,\vec J + c\,\partial/\partial r'_t \rho = 0$ becomes, in biquaternionic form,
$$ \mathrm{Sc}\!\left((\tilde\nabla_\Phi)^{\natural}\tilde J_{\Phi}\right) = 0, $$
i.e., the scalar part of $(\tilde\nabla_\Phi)^{\natural}\tilde J_{\Phi}$ vanishes. In components this is the statement $c\,\partial\rho/\partial r'_t + \mathrm{div}\,\vec J = 0$, because
$$ \mathrm{Sc}\bigl((\tilde\nabla_\Phi)^{\natural}\tilde J_{\Phi}\bigr)=\frac{\partial J_0}{\partial R_t}+\mathrm{div}\,\vec J =-i\frac{\partial}{\partial r'_t}\bigl(ic\rho\bigr)+\mathrm{div}\,\vec J =c\,\frac{\partial\rho}{\partial r'_t}+\mathrm{div}\,\vec J, $$
where the two factors of $i$ give $i^2=-1$, leaving the $c$ of the time component of $\tilde J_{\Phi}$ in place. This is the integrability condition for the biquaternionic Maxwell equation $\tilde\nabla_\Phi \tilde{F} = -\sqrt{\mu}\,\tilde J_{\Phi}$, and it is the biquaternionic expression of the conservation of electric charge. That the scalar part is the relevant one follows by applying $(\tilde\nabla_\Phi)^{\natural}$ to the field equation:
$$ (\tilde\nabla_\Phi)^{\natural}\tilde\nabla_\Phi\tilde F=\Box_\Phi\tilde F=-\sqrt{\mu}\,(\tilde\nabla_\Phi)^{\natural}\tilde J_{\Phi} . $$
The left-hand side is a pure vector, since $\Box_\Phi$ is a scalar operator and $\tilde F$ is a pure vector; its scalar part vanishes identically, so $-\sqrt{\mu}\,\mathrm{Sc}((\tilde\nabla_\Phi)^{\natural}\tilde J_{\Phi})$ must vanish for a solution to exist. When the source is written as $\sqrt{\mu}\tilde J_{\Phi}=-\tilde\nabla_\Phi\tilde F$ the condition holds without further assumption.
Maxwell's Equations in Biquaternionic Form
With these definitions, Maxwell's equations take the single compact form
$$ \boxed{\;\tilde\nabla_\Phi \tilde{F} = -\sqrt{\mu}\,\tilde{J}_{\Phi}\;}. $$
One equation replaces the four standard Maxwell equations. No explicit $\epsilon$ appears: the medium is carried by the definition of $\tilde{F}$ and by the single factor $\sqrt{\mu}$.
The counting is visible in the characteristic polynomial. In the Fourier variables $(\vec\zeta,\omega)$ the vector part $\partial/\partial R_t\,\vec F + \mathrm{rot}\,\vec F$ acts on the three components as the factor $-\omega/c$ together with the map $\vec F\mapsto i\,\vec\zeta\times\vec F$, whose eigenvalues are $0$ and $\pm|\vec\zeta|$; the determinant of the symbol is therefore $-({\omega}/{c})\bigl(({\omega}/{c})^2 - |\vec\zeta|^2\bigr)$. The characteristic set is therefore $\omega = 0$ together with the light cone $|\omega| = c|\vec\zeta|$. The four standard Maxwell equations have the same characteristic set with every zero doubled; the single complex equation carries the characteristics of the four with half the multiplicity, which is the symbol-level form of "one equation replaces four".
Taking the scalar and vector parts of $\tilde\nabla_\Phi \tilde{F} = -\sqrt{\mu}\,\tilde{J}_{\Phi}$ separately, we recover the Hamiltonian form of Maxwell's equations. The scalar part is
$$ -\mathrm{div}\,\vec F = -R_0, $$
and the vector part is
$$ \partial/\partial R_t \vec F + \mathrm{rot}\,\vec F = -\vec R. $$
These are equivalent to the four standard Maxwell equations. The integrability condition $\sqrt{\mu}\,\mathrm{Sc}((\tilde\nabla_\Phi)^{\natural}\tilde{J}_{\Phi}) = 0$ is automatically satisfied when the four-current is expressed in terms of the potentials via $\sqrt{\mu}\,\tilde{J}_{\Phi} = -\tilde\nabla_\Phi \tilde{F}$, and it is the consistency condition that must be imposed when the source is specified independently.
The Retarded Green's Function
The biquaternionic Maxwell equation $\tilde\nabla_\Phi \tilde{F} = -\sqrt{\mu}\,\tilde{J}_{\Phi}$ is a linear first-order equation for the field strength $\tilde{F}$ in terms of the four-current $\tilde{J}_{\Phi}$. Its solution can be written as a convolution with a Green's function $\tilde{G}_{\Phi,\tilde\nabla_\Phi}(\tilde Q)$, which is the solution of the equation with a point source:
$$ \tilde\nabla_\Phi \tilde{G}_{\Phi,\tilde\nabla_\Phi}(\tilde Q) = -\delta(\tilde{Q})\,e_t, $$
with the sign convention of The Biquaternion D'Alembertian and Its Green's Functions, $\Box_\Phi G_{\Box,\Phi}=-\delta$.
The physically relevant Green's function is the retarded one, which vanishes for $r'_t < 0$ and has support on the future light cone. It is the first-order kernel of the companion article The Biquaternion D'Alembertian and Its Green's Functions,
$$ \begin{aligned} \tilde{G}_{\Phi,\tilde\nabla_\Phi} = (\tilde\nabla_\Phi)^{\natural}\,G_{\Box,\Phi} &=\frac{1}{4\pi R c}\,\delta'\!\left(r'_t - R\right)\bigl(-i\,e_t + \vec u_R\bigr)\\ &\quad+\frac{1}{4\pi R^2}\,\delta\!\left(r'_t - R\right)\vec u_R, \end{aligned} $$
where $G_{\Box,\Phi}(R,r'_t) = (4\pi R)^{-1}\delta(r'_t - R)$ is the scalar retarded kernel of the wave operator, $\delta'$ is the derivative of the delta with respect to its argument, $R = \sqrt{r_x^2+r_y^2+r_z^2}$, and $\vec u_R = (r_x \vec e_x + r_y \vec e_y + r_z \vec e_z)/R$ is the unit radial biquaternion. It satisfies $\tilde\nabla_\Phi\tilde{G}_{\Phi,\tilde\nabla_\Phi} = (\tilde\nabla_\Phi(\tilde\nabla_\Phi)^{\natural})G_{\Box_\Phi} = \Box_\Phi G_{\Box_\Phi}\,e_t = -\delta(\tilde{Q})\,e_t$, the unit $e_t$ being the one carried by the right-hand side of the defining equation, and it is supported on the future light cone. The kernel is not a multiple of the delta alone: the $\delta'$ term is a double layer on the cone and the $\delta$ term a single layer, and it is the $\delta'$ term — absent from any schematic multiple-of-$\delta$ display — that the wave-front and jump conditions of Shock Electromagnetic Waves require. Its distributional layer calculus is that of Distributions on Surfaces, Layers, and Jump Conditions.
The solution of the inhomogeneous equation is then the retarded convolution
$$ \boxed{\;\tilde{F}(\tilde{Q}) = +\sqrt{\mu}\int \tilde{G}_{\Phi,\tilde\nabla_\Phi}(\tilde{Q} - \tilde{R}_{\Phi,2})\,\tilde{J}_{\Phi}(\tilde{R}_{\Phi,2})\,d^4R_2\;}, $$
where the integral is over the past light cone of $\tilde{Q}$. This is the biquaternionic form of the retarded solution of Maxwell's equations, and it is the physically correct solution for radiation problems: the field at a point depends only on the sources in its past light cone, not on the future sources.
The scalar part of that formula is where the integrability condition acts, and it is worth writing out. Since $\tilde{G}_{\Phi,\tilde\nabla_\Phi} = (\tilde\nabla_\Phi)^{\natural}G_{\Box_\Phi}$, the derivatives in $(\tilde\nabla_\Phi)^{\natural}$ can be moved from the kernel onto the source without changing any sign — the change of variable contributes one minus per derivative and the integration by parts the other — and the minus on the vector part of $(\tilde\nabla_\Phi)^{\natural}$ meets the minus in $\vec v = -\vec\nabla_\Phi G_{\Box,\Phi}$, so the two terms of the scalar part agree. The result is
$$ \mathrm{Sc}\bigl(\tilde F\bigr) =\sqrt{\mu}\int G_{\Box_\Phi}(\tilde Q-\tilde{R}_{\Phi,2})\,\mathrm{Sc}\bigl((\tilde\nabla_\Phi)^{\natural}\tilde J_{\Phi}\bigr)(\tilde{R}_{\Phi,2})\,d^4R_2 . $$
So the boxed solution is a pure vector exactly when $\mathrm{Sc}((\tilde\nabla_\Phi)^{\natural}\tilde J_{\Phi}) = 0$, and when that condition fails the field it returns carries a scalar whose value is the retarded average of the violation. The constraint is therefore not an extra condition imposed beside the formula: it is the statement that the formula's output lies in the space in which $\tilde F$ is defined.
The retarded Green's function is distinct from the Cauchy kernel $\tilde{Q}^{\natural}/\|\tilde{Q}\|^4$ of the elliptic theory (companion articles on biquaternion integration and biquaternion analysis on subspaces). The difference is which reading of one operator is inverted. The Cauchy kernel is the fundamental solution of $\Box_\Phi=\partial^2/\partial R_t^2+\Delta_\Phi$ read with the four real coordinates $Q_\mu$ of the $\mathbb{B}$ frame, where all four coefficients have the same sign; the retarded kernel is the fundamental solution of the same $\Box_\Phi$ read on $\mathbb{B}_{\Phi}$, where $Q_t=i q'_t$ turns it into $\Delta_\Phi-\partial^2/\partial {r'_t}^2$. There is one d'Alembertian, not two: the substitution is a change of coordinate, and it changes the signs. The two kernels are related by the Wick rotation $r'_t \to -i\tau$, which converts the hyperbolic kernel into the elliptic one. The elliptic kernel is the natural object in the Euclidean (imaginary-time) formulation; the retarded kernel is the natural object in the Lorentzian (real-time) formulation.
The Three-Vector Form
The field strength has a complex three-vector form. With
$$ \vec F = i\sqrt{\epsilon}\,\vec E - \sqrt{\mu}\,\vec H, \qquad \tilde F = \vec F, $$
and $\partial/\partial R_t = -i\partial/\partial r'_t$, the biquaternionic Maxwell equation $\tilde\nabla_\Phi\tilde F = -\sqrt{\mu}\,\tilde J_{\Phi}$ is the pair
$$ \partial/\partial R_t\,\vec F + \mathrm{rot}\,\vec F = -\vec R, \qquad -\mathrm{div}\,\vec F = -R_0, $$
with $\vec R = \sqrt{\mu}\,\vec J$ and $R_0 = i\rho/\sqrt{\epsilon}$. This is the Riemann–Silberstein form of the same equation, written in the corpus's own field strength; it introduces no object beyond $\tilde F$. The energy density and the flux are read from the same element, $\tilde F\tilde F^{*} = 2W e_t + \tfrac{2i}{c}\vec S$, as in The Field-Strength Biquaternion and Its Invariants. The point of the form is symmetry: once magnetic charge is admitted, the equation is invariant under the interchange of the electric and magnetic halves, which the four standard Maxwell equations possess only with magnetic sources present.
At fixed frequency $\partial/\partial R_t = -k_\Phi$ with $k_\Phi = \omega/c$, and the equation becomes the complex-vector Helmholtz equation
$$ \mathrm{rot}\,\vec F - k_\Phi\,\vec F = -\vec R . $$
The static limit, $\partial/\partial R_t = 0$, is the pair $\mathrm{rot}\,\vec F = -\vec R$, $\mathrm{div}\,\vec F = R_0$. Both are the biquaternionic equation at the corresponding frequency, not a separate formulation.
The Stationary Limit
When the sources are time-independent, $\partial/\partial r'_t \rho = 0$ and $\partial/\partial r'_t \vec J = 0$, the biquaternionic Maxwell equation reduces to a static equation. The field strength $\tilde{F}$ becomes time-independent, and the equation $\tilde\nabla_\Phi \tilde{F} = -\sqrt{\mu}\,\tilde{J}_{\Phi}$ separates into the two standard static equations:
$$ \mathrm{div}\,\vec D = \rho, \qquad \mathrm{rot}\,\vec H = \vec J, $$
where $\vec D = \epsilon \vec E$ and $\vec H = \vec B/\mu$ are the usual static fields. In biquaternionic form, the static equation is
$$ \vec\nabla_\Phi \tilde{F} = -\sqrt{\mu}\,\tilde{J}_{\Phi}, \qquad \vec\nabla_\Phi = \vec e_x \partial/\partial R_x + \vec e_y \partial/\partial R_y + \vec e_z \partial/\partial R_z, $$
where $\vec\nabla_\Phi$ is the vector part of the biquaternionic gradient, the three spatial terms of $\tilde\nabla_\Phi$ alone. The equation is purely spatial, and the time coordinate plays no role.
The static solution is given by the gradient of the Newton kernel, not by the Newton kernel itself, because the static equation is first order. Write $\vec r = r_x\,\vec e_x+r_y\,\vec e_y+r_z\,\vec e_z$ for the field point and $\vec y$ for the source point. Then
$$ \tilde{F}(\vec r) =\frac{1}{4\pi}\int\frac{(\vec r-\vec y)\,R_0(\vec y)}{\|\vec r-\vec y\|^3}\,d^3y -\frac{1}{4\pi}\int\frac{\vec R(\vec y)\times(\vec r-\vec y)}{\|\vec r-\vec y\|^3}\,d^3y, $$
with $R_0 = \sqrt{\mu}\,ic\rho = i\rho/\sqrt{\epsilon} = \mathrm{Sc}(\sqrt{\mu}\,\tilde{J}_{\Phi})$, $\vec R = \sqrt{\mu}\,\vec J = \mathrm{Vect}(\sqrt{\mu}\,\tilde{J}_{\Phi})$, and $\|\vec r-\vec y\|$ the ordinary Euclidean distance in $\mathbb{R}^3$. The first term is the biquaternionic form of the Coulomb law and the second that of the Biot–Savart law. Its imaginary part is $\sqrt{\epsilon}\,\vec E$ and its real part is $-\sqrt{\mu}\,\vec H$.
For a point charge $ϙ$ at the origin, the four-current is $\tilde J_{\Phi} = i c ϙ\,\delta(\vec r)\,e_t$, so the source is $\sqrt{\mu}\,\tilde{J}_{\Phi} = i ϙ/\sqrt{\epsilon}\,\delta(\vec r)\,e_t$, and the solution is
$$ \tilde{F}(\vec r) = \frac{i ϙ\,\vec u_x}{4\pi\sqrt{\epsilon}\,\|\vec r\|^2}, \qquad \vec u_x = \frac{\vec r}{\|\vec r\|}, $$
which is the Coulomb field of a point charge at the origin, falling as $1/\|\vec r\|^2$. In the same way, a steady current loop produces a magnetic dipole field, and the biquaternionic formulation reproduces the standard results of magnetostatics.
Gauge Structure
The potential biquaternion $\tilde{A}_{\Phi}$ is not uniquely determined by the field-strength biquaternion $\tilde{F}$. Two potentials $\tilde{A}_{\Phi}$ and $\tilde{A}_{\Phi,\mathrm g}$ give the same $\tilde{F}$ if they differ by a gauge transformation
$$ \tilde{A}_{\Phi,\mathrm g} = \tilde{A}_{\Phi} - \tilde\nabla_\Phi \Gamma, $$
where $\Gamma$ is an arbitrary scalar function. This is the biquaternionic form of the standard gauge transformation $A_\mu \to A_\mu - \partial/\partial R_\mu \Gamma$.
Under this transformation, the field-strength biquaternion is invariant:
$$ \tilde{F}_{\mathrm g} = \frac{1}{\sqrt{\mu}}\Bigl((\tilde\nabla_\Phi)^{\natural}\tilde{A}_{\Phi,\mathrm g} - \mathrm{Sc}\!\left((\tilde\nabla_\Phi)^{\natural}\tilde{A}_{\Phi,\mathrm g}\right)\Bigr) = \frac{1}{\sqrt{\mu}}\Bigl((\tilde\nabla_\Phi)^{\natural}\tilde{A}_{\Phi} - \mathrm{Sc}\!\left((\tilde\nabla_\Phi)^{\natural}\tilde{A}_{\Phi}\right)\Bigr) = \tilde{F}, $$
because the gradient of a gradient is central under $(\tilde\nabla_\Phi)^{\natural}$ — the term $(\tilde\nabla_\Phi)^{\natural}\tilde\nabla_\Phi\Gamma=\Box_\Phi\Gamma$ is a scalar and drops out of the vector part.
The scalar part of $(\tilde\nabla_\Phi)^{\natural}\tilde{A}_{\Phi}$ is not invariant. Write
$$ S = \mathrm{Sc}\!\left((\tilde\nabla_\Phi)^{\natural}\tilde{A}_{\Phi}\right). $$
This is the biquaternionic scalar field, whose components are $\partial/\partial R_t A_0 + \mathrm{div}\,\vec A$. Under a gauge transformation, it transforms as
$$ S_2 = S - \Box_\Phi \Gamma. $$
The scalar field $S$ is therefore a gauge degree of freedom. It is not a physical field; it can be changed at will by a gauge transformation. The physical content of the theory is entirely in the vector part of $(\tilde\nabla_\Phi)^{\natural}\tilde{A}_{\Phi}$, which is the field-strength biquaternion $\tilde{F}$ and is invariant under the gauge transformation.
The Lorenz gauge is the condition
$$ S = 0, $$
which means
$$ S=\frac{\partial A_0}{\partial R_t} + \mathrm{div}\,\vec A =-i\frac{\partial}{\partial r'_t}\Bigl(i\,a'_t\Bigr)+\mathrm{div}\,\vec A =\frac{1}{c}\frac{\partial\xi_{\mathrm{scp}}}{\partial r'_t}+\mathrm{div}\,\vec A=0, $$
where the two factors of $i$ give $i^2=-1$, $a'_t=\xi_{\mathrm{scp}}/c$ restores the potential, and the sign of the second term follows. This is the biquaternionic form of the standard Lorenz condition $\sum_\mu \partial A_\mu/\partial R_\mu = 0$, written in the familiar three-vector shape. It is a choice of gauge, not a physical condition. (The electro-gravimagnetic programme of The Electro-Gravimagnetic Field and the Magnetic-Charge–Mass Hypothesis uses the same condition in a different role, imposing it as a constraint of its representation rather than choosing it as a gauge; see that article, where the difference of logical status is recorded and not resolved.) In the Lorenz gauge, the potential satisfies the wave equation
$$ \Box_\Phi \tilde{A}_{\Phi} = -\mu \tilde J_{\Phi}, $$
with
$$ \tilde J_{\Phi} = ic\rho\,e_t + \vec J. $$
This is the biquaternionic wave equation for the potential. The factor of $\mu$ is explicit, because the potential equation in the tensor formulation is $\Box_\Phi A_\mu = -\mu J_\mu$ with $J_\mu = (ic\rho, \vec J)$. Note that the wave equation has a solution if and only if the four-current satisfies the conservation law $\mathrm{Sc}((\tilde\nabla_\Phi)^{\natural}\tilde J_{\Phi}) = 0$, which is equivalent to the charge conservation constraint on $\rho$ and $\vec J$.
The scalar field $S$ has no independent physical meaning. It is a gauge artifact, and the Lorenz gauge is the physical choice that sets it to zero. In the future, if a physical interpretation of $S$ is found, it would have to arise from a coupling to a sector of the theory that is not gauge-invariant. For the classical electromagnetic field, $S$ is not physical.
The gauge function as a velocity potential. There is an interpretation of the gauge freedom, older than the relativistic one and independent of it, in which the gauge function is a velocity potential. Under $\tilde{A}_{\Phi} \to \tilde{A}_{\Phi} - \tilde\nabla_\Phi\Gamma$ only the gradient of $\Gamma$ matters, exactly as in the hydrodynamic statement that only the gradient of a potential matters for a velocity and its curl. On the hydrodynamic dictionary set out below, the vector potential is the image of the fluid velocity, the field strength is the image of its vorticity, and the gauge transformation is the freedom to add the gradient of a scalar to the velocity — a change that leaves the vorticity and hence the Lamb vector untouched. This is a formal correspondence and not an explanation of gauge invariance, which the algebraic account above settles on its own; it is recorded because the corpus's own gauge discussion stops at "a choice of gauge, not a physical condition", and the hydrodynamic picture supplies a concrete image of why only a gradient is free.
The Hydrodynamic Form of the Equations
A second interpretation of the same equations, older than the relativistic one and independent of it, comes from fluid mechanics. It is a formal correspondence with a specific provenance, not an alternative derivation, and its limits are stated with it.
The bridge is the Lamb vector. For an incompressible fluid of velocity $\vec v_f$, vorticity $\vec\omega = \vec\nabla_\Phi\times\vec v_f$, pressure $p$ and mass density $\rho_f$ — the fluid density, not the charge density $\rho$ of the electromagnetic sections — the convective acceleration obeys the identity
$$ (\vec v_f\cdot\nabla)\vec v_f = \vec\omega\times\vec v_f + \vec\nabla_\Phi\frac{|\vec v_f|^2}{2}, $$
so that the Euler equation $\partial/\partial r'_t\vec v_f + (\vec v_f\cdot\nabla)\vec v_f = -\vec\nabla_\Phi p/\rho_f$ takes the Lamb form
$$ \partial/\partial r'_t\vec v_f = -\,\vec\nabla_\Phi\!\left(\frac{p}{\rho_f} + \frac{|\vec v_f|^2}{2}\right) - \vec\ell, \qquad \vec\ell = \vec\omega\times\vec v_f. $$
The quantity $\vec\ell$ is the Lamb vector. Comparison with the potential relations of the electromagnetic theory, $\vec B = \vec\nabla_\Phi\times\vec A$ and $\vec E = -\vec\nabla_\Phi\xi_{\mathrm{scp}} - \partial/\partial r'_t\vec A$, suggests the pairing
$$ \vec v_f\;\leftrightarrow\;\vec A, \qquad \vec\omega\;\leftrightarrow\;\vec B, \qquad \vec\ell\;\leftrightarrow\;\vec E, \qquad \frac{p}{\rho_f}+\frac{|\vec v_f|^2}{2}\;\leftrightarrow\;\xi_{\mathrm{scp}} . $$
With these identifications the Lamb form of the Euler equation is the definition of the electric field from the potentials, $\vec E = -\vec\nabla_\Phi\xi_{\mathrm{scp}} - \partial/\partial r'_t\vec A$, and its curl is the vorticity equation
$$ \partial/\partial r'_t\vec\omega = -\,\vec\nabla_\Phi\times\vec\ell, $$
which is Faraday's law $\partial/\partial r'_t\vec B = -\vec\nabla_\Phi\times\vec E$ under the dictionary. Both statements are exact given the pairing. The counterpart of Gauss's law is the divergence of the Lamb form,
$$ \vec\nabla_\Phi\cdot\vec\ell = -\,\vec\nabla_\Phi^2\!\left(\frac{p}{\rho_f} + \frac{|\vec v_f|^2}{2}\right), $$
which holds for a divergence-free velocity. The hydrodynamic image of the charge density is therefore a Laplacian of the pressure-and-kinetic term rather than an independent quantity, which is where the analogy's electric sector is thinnest: the fluid has no carrier of "charge" that the flow does not already determine.
The correspondence is attributed to H. Marmanis. It is stated here in three-vector language, as the source states it, and it is not a biquaternion formula; the biquaternionic content of these equations is the one the rest of this article records.
Limits. The correspondence is partial, and the corpus does not take it as an explanation of electromagnetism. The full Euler equation is nonlinear while the field equations are not, and the source paper's correspondence is stated for the linearised equations. The divergence law for the magnetic field has no fluid counterpart, so the analogy has no magnetic source sector. The constitutive relations differ, so permittivity and permeability have no single fluid meaning. And the two theories do not share a symmetry group: Maxwell's equations are Lorentz invariant while the fluid equations are Galilean invariant. What the analogy reflects is the part of the symmetry structure that the two actually share. The field–potential relations are kinematics: they hold in every frame and need no symmetry argument at all. Of Maxwell's four equations, the homogeneous pair survives unchanged in the magnetic Galilean limit — the limit obtained by sending $c\to\infty$ at fixed vacuum permeability, which is the regime of the quasi-stationary approximation — while it is Gauss's law that acquires the correction of order $1/c^2$. Those two equations are therefore the ones a Galilean theory can carry, and the one it cannot is exactly the one whose fluid image is a Laplacian rather than an independent density. The other limit, $c\to\infty$ at fixed vacuum permittivity with $\vec B\to c^2\vec B$, is the electric one and is the one used in electrohydrodynamics. The incompressible fluid limit corresponds to the magnetic one, with the vacuum permittivity playing the role of an acoustic compressibility and the vacuum permeability that of a density. The relativistic account of the field remains the standard one. The force-side counterpart of the same correspondence is recorded in The Lorentz Force in Biquaternion Form.
Relationship Between the Two Formulations
The two formulations are related as follows:
- The field-strength biquaternion $\tilde{F}$ is obtained from the potential biquaternion $\tilde{A}_{\Phi}$ by differentiation: $\tilde{F} = \frac{1}{\sqrt{\mu}}\bigl((\tilde\nabla_\Phi)^{\natural}\tilde{A}_{\Phi} - \mathrm{Sc}((\tilde\nabla_\Phi)^{\natural}\tilde{A}_{\Phi})\bigr)$.
- The first-order equation $\tilde\nabla_\Phi \tilde{F} = -\sqrt{\mu}\,\tilde{J}_{\Phi}$ is equivalent to the second-order equation $\Box_\Phi \tilde{A}_{\Phi} = -\mu \tilde J_{\Phi}$ in the Lorenz gauge.
- The source is the normalised four-current, $\sqrt{\mu}\,\tilde J_{\Phi} = i\rho/\sqrt{\epsilon} + \sqrt{\mu}\vec J$, with $\tilde J_{\Phi} = ic\rho\,e_t + \vec J$.
The first-order formulation is the most compact: one equation for the field strength. The second-order formulation is the most familiar: the wave equation for the potential. Both are valid, and both are biquaternionic.
A third route, and where it loses equivalence. The two formulations above are equivalent to each other, and both are equivalent to the Maxwell system. The classical route to a single equation is a third thing, and it is not equivalent. Eliminating one field — solving the Ampère–Maxwell law for $\vec H$ and substituting it into $\mathrm{rot}\,\vec E=-\partial/\partial r'_t\vec B$ — yields a second-order wave equation for $\vec E$ alone, and that equation admits solutions the Maxwell system does not. Any $\vec E=\vec\nabla_\Phi f$ with $\Box_\Phi f=0$ solves it, since $\Box_\Phi\nabla f=\vec\nabla_\Phi\Box_\Phi f$ vanishes, while $\mathrm{div}\,\vec E=\Delta_\Phi f$ need not vanish; for $f=\exp(ik_\Phi(z-r'_{t}))$ the divergence is $-k_\Phi^2f$, so the elimination has enlarged the solution set by a longitudinal sector that Maxwell forbids. The constraints $\mathrm{div}\,\vec D=\rho$ and $\mathrm{div}\,\vec B=0$ are not implied by the wave equation and have to be reimposed by hand, and the boundary data do not transfer either: a Dirichlet or Neumann datum on $\vec E$ is natural data for the wave equation and not for the first-order system, while the natural Maxwell datum — the tangential components of $\vec E$ and $\vec H$ — is not a datum the wave equation is posed with. The reduction is convenient rather than faithful: it is what one wants for propagation, and it is not a statement of the theory.
The biquaternionic formulation pays no such price, because it eliminates nothing. The single equation $\tilde\nabla_\Phi\tilde{F}=-\sqrt{\mu}\,\tilde{J}_{\Phi}$ is first order in one field-strength biquaternion, and the constraints are not separate conditions added afterwards: $\mathrm{div}\,\vec F=R_0$ is the scalar part of that same equation. This is the property Kravchenko calls the diagonalization of the Maxwell system, and it can be reproduced in vector calculus by combining the two fields rather than eliminating one of them into the other; Maxwell's Equations in Chiral Media — The Quaternionic Reformulation displays the combinations and the first-order equations they satisfy. The two routes are distinct claims and both are wanted: the quaternionic form is preferred because it carries a function theory, and the point here is only that the first-order biquaternionic equation already has the equivalence the wave equation lacks.
A fourth route: the matrix form, and what makes it work. The arrays printed in this subsection are the object of the subsection: it reports an external display and compares that display with the element form of this article, so the arrays are shown and nothing of this article is computed with them. Daniel Henry Gottlieb writes the same system as a single matrix equation, in which the field is a $4 \times 4$ matrix and the divergence is a column of operators acting on it. In his units and conventions ($c = 1$, Heaviside–Lorentz, metric inertia $(-,+,+,+)$, $r'_t$ the time coordinate) the real anti-self-adjoint operator of the Lorentz law and its dual are
$$ F = \begin{pmatrix} 0 & \vec E^{\mathsf T} \ \vec E & -[\vec B]_\times \end{pmatrix}, \qquad F^* = \begin{pmatrix} 0 & -\vec B^{\mathsf T} \ -\vec B & -[\vec E]_\times \end{pmatrix}, $$
with $[\vec a]_\times$ the matrix of $\vec w \mapsto \vec a \times \vec w$; the paper writes the lower-right block of $F$ as $\times\vec B$, whose action is $\vec w\mapsto\vec w\times\vec B = -[\vec B]_\times\vec w$, which is the same object. The complexification $cF := F - iF^*$ is the matrix of a single complex three-vector,
$$ cF = \begin{pmatrix} 0 & \vec N^{\mathsf T} \ \vec N & i[\vec N]_\times \end{pmatrix}, \qquad \vec N = \vec E + i\vec B, $$
where $\vec N$ is the Riemann–Silberstein vector of The Field-Strength Biquaternion above, written with $c=1$ as the paper's units require; the letter $\vec A$ is reserved in this article for the vector potential and is not used here.
and with the operator column $(-\partial/\partial r'_t, \vec\nabla_\Phi)^{\mathsf T}$ read entry by entry against the matrix — each operator differentiating the matrix entry beside it — the four equations become the three statements
$$ F\begin{pmatrix} -\partial/\partial r'_t \ \vec\nabla_\Phi \end{pmatrix} = \begin{pmatrix} \rho \ \vec J \end{pmatrix}, \qquad F^*\begin{pmatrix} -\partial/\partial r'_t \ \vec\nabla_\Phi \end{pmatrix} = \begin{pmatrix} -\mathrm{div}\,\vec B \ \partial/\partial r'_t\vec B + \mathrm{rot}\,\vec E \end{pmatrix}, \qquad cF\begin{pmatrix} -\partial/\partial r'_t \ \vec\nabla_\Phi \end{pmatrix} = \begin{pmatrix} \rho \ \vec J \end{pmatrix}. $$
The first holds identically, because it is the two definitions $\rho = \mathrm{div}\,\vec E$ and $\vec J = \mathrm{rot}\,\vec B - \partial/\partial r'_t\vec E$; the second vanishes exactly when the two homogeneous laws hold; so the third, the paper's one equation, holds exactly when the Maxwell system holds. Recomputing the three on fields built from potentials, $\vec B = \mathrm{rot}\,\vec A$ and $\vec E = -\partial/\partial r'_t\vec A - \vec\nabla_\Phi\xi_{\mathrm{scp}}$, gives agreement to $6 \times 10^{-17}$ for the first and the third and $1 \times 10^{-17}$ for the vanishing of the second, and on a field that violates the homogeneous laws the discrepancy in the third is exactly $i\left(\mathrm{div}\,\vec B,\; -(\partial/\partial r'_t\vec B + \mathrm{rot}\,\vec E)\right)$, so the equivalence is exact rather than numerical coincidence. The paper also records the version of the same equation for the four fields $\vec E, \vec B, \vec D, \vec H$, in which $\vec D + i\vec B$ occupies the vector slots: the shape of the matrix is unchanged, but the paper notes that the square of that matrix is no longer a multiple of the identity — as it is for the form above, where $cF^2 = \langle\vec N,\vec N\rangle I$ — and that this property singles out the first form among all matrices of the second. Both statements were checked: $cF^2 = \langle\vec N,\vec N\rangle I$ exactly, and the second form's square misses a multiple of the identity by at least $14$ over $100$ random pairs of vectors.
The dual form, and why it holds. The equation has a dual in which the matrix carries the operators and the vector carries the field. With
$$ c\nabla := \begin{pmatrix} 0 & \vec\nabla_\Phi^{\mathsf T} \ \vec\nabla_\Phi & i[\vec\nabla_\Phi]_\times \end{pmatrix}, $$
the system reads $(\partial/\partial r'_t I - c\nabla)\left(0,\; -\overline{\vec E + i\vec B}\right)^{\mathsf T} = \left(\rho,\; \vec J\right)^{\mathsf T}$, again exactly equivalent, and the two forms are one construction seen from two sides: in the first the matrix is the field and the column the operators, in the second the matrix is the operators and the column the field. Recomputing this form on the same fields gives agreement to machine precision, with one caveat of transcription: the conjugate bar in the column is required. Without it the scalar row still gives $\rho$ and the vector rows are wrong, which is how the formula extracts from the source and is why it is not stated here without the bar.
The paper writes two further equations of the same shape, in which the four-potential $(\xi_{\mathrm{scp}}, \vec A)$ replaces the field in the column and the operator matrix acquires the potential in its entries; they are the potential-level counterparts of the two field-level equations, and they are the paper's route to the wave equation and to the covariant gauge condition $\partial/\partial r'_t\xi_{\mathrm{scp}} + \mathrm{div}\,\vec A = 0$. As printed, both carry a defect in the vector part, and this is recorded rather than reproduced: the first gives $\partial/\partial r'_t\vec A - i\,\mathrm{rot}\,\vec A$ where the definition $\vec E = -\partial/\partial r'_t\vec A - \vec\nabla_\Phi\xi_{\mathrm{scp}}$ requires the missing term $-\vec\nabla_\Phi\xi_{\mathrm{scp}}$, so it holds only in a gauge with $\vec\nabla_\Phi\xi_{\mathrm{scp}} = 0$, and the second gives $\partial/\partial r'_t\vec A + \vec\nabla_\Phi\xi_{\mathrm{scp}} + i\,\mathrm{rot}\,\vec A$ where $-\vec E - i\vec B = \partial/\partial r'_t\vec A + \vec\nabla_\Phi\xi_{\mathrm{scp}} - i\,\mathrm{rot}\,\vec A$ is required, a sign on the magnetic term. Recomputing both against the definitions, the discrepancies are exactly $|\vec\nabla_\Phi\xi_{\mathrm{scp}}|$ and $|2\vec B|$ respectively, term for term; the scalar rows and the operator identity $c\nabla^2 = \Delta_\Phi I$ on which the reduction rests are unaffected, since the scalar row involves only $\partial/\partial r'_t\xi_{\mathrm{scp}} + \mathrm{div}\,\vec A$ and neither defect touches it. The duality between the two potential equations is therefore a statement about the two operator matrices, not a statement that the printed columns are already correct, and the mechanism below is what carries it.
The reason the dual pairs agree is an algebraic identity, and it is the source of everything else in this route. The matrices of the complex three-vectors satisfy, for every pair of vectors,
$$ cF(\vec N_1)\,cF(\vec N_2) + cF(\vec N_2)\,cF(\vec N_1) = 2\langle \vec N_1, \vec N_2 \rangle I, \qquad \langle \vec N, \vec N \rangle = \vec E^2 - \vec B^2 + 2i\,\vec E\cdot\vec B, $$
in the general plain bilinear form — no conjugation, so that $\langle \vec N, \vec N \rangle$ is the two field invariants in one — and they commute with their own complex conjugates,
$$ cF(\vec N_1)\,\overline{cF(\vec N_2)} = \overline{cF(\vec N_2)}\,cF(\vec N_1), $$
for all $\vec N_1, \vec N_2$. The commutation is the whole mechanism: it is why the matrix of the field times the column of operators equals the matrix of the operators times the column of the field, and it is what the source itself names as the reason, together with the algebra it names as the origin of the commutation — the left and the right regular representations of the biquaternions. In the operator case the same identity becomes
$$ c\nabla^2 = \Delta_\Phi I, \qquad (\partial/\partial r'_t I - c\nabla)(\partial/\partial r'_t I + c\nabla) = (\partial^2/\partial {r'_{t}}^2 - \Delta_\Phi) I, $$
both recomputed at the finite-difference floor, and the second is the paper's decoupling into the wave equation, written as a product of two first-order factors: it is the same factorization as $\Box_\Phi = \tilde\nabla_\Phi(\tilde\nabla_\Phi)^{\natural}$ of the biquaternionic gradient, with the operators reorganized and the signs adapted.
What the matrix route says about the algebra. The correspondence with the algebra is exact, and it is a change of basis of The Element and Operator Representations: those $4 \times 4$ matrices are the biquaternion algebra in the basis $e_t, x, y, z$ with $x = i\vec e_x$, $y = i\vec e_y$, $z = i\vec e_z$, whose vector units square to $+e_t$; in that basis the left regular matrix of an element is $a I + cF$ with $cF$ the matrix above, the right regular matrix is exactly its transpose with no sign matrix, and $c\nabla$ is the same construction with the components of $\vec\nabla_\Phi$ in place of the field. The consequence for the question this section is about is not a confirmation but a reversal of direction: the one-equation compactness, with full equivalence and no elimination, is available in a purely matrix setting, so it settles nothing about the algebra; what the algebra supplies here is the explanation of the matrix identities, since the commutation that makes the dual pair agree is the commutation of the left and the right regular representations, and the field matrix is the left regular matrix in a basis. An author working in matrices and counting equations is led back to the algebra as the reason his identities are true. That is a stronger statement for the corpus than another confirmation of compactness, and it is consistent with the position taken in the section The Question of What Is Fundamental below: the algebra supplies the home in which relations become structural, not a claim that no other language can write the equation once.
Two further items of the paper belong to other articles. The three matrices $cF$ of the coordinate units $\vec e_x,\vec e_y,\vec e_z$ just defined — $cF$ evaluated at $(1,0,0)$, $(0,1,0)$, $(0,0,1)$ — anticommute pairwise, square to $+I$, and satisfy $cF(\vec e_x)cF(\vec e_y) = i\,cF(\vec e_z)$, while the sixteen matrices $I$, $cF(\vec e_\mu)$, their conjugates and their products form a Hermitian basis of $M_4(\mathbb{C})$ of which the four $\{I, cF(\vec e_\mu)\}$ span the biquaternions — same matrices, in the same basis, as the sixteen of the second realization recorded in The Element and Operator Representations, where they are listed and where the transposition identity these matrices satisfy is stated. The tensor $\frac{1}{2}cF\,\overline{cF}$ is real, traceless and carries the energy density, the Poynting vector and the Maxwell stress, and its divergence is the four-force density; that balance, and the matrix proof of it from the Leibniz rule, are recorded beside the corpus's own derivation in Exercise: The Electromagnetic Energy–Momentum Tensor. The exponential of the real anti-Hermitian $F$ factors as $\exp(F) = \exp(cF/2)\exp(\overline{cF}/2)$ and is a proper Lorentz transformation, recomputed to $1.4 \times 10^{-14}$ over $100$ random fields; the corpus has no separate article for this exponential as an element of the field theory, and it is recorded here only as a property of $cF$ rather than adopted as a route to the Lorentz group, which the two-sided action of The Element and Operator Representations already supplies.
The Author's Reading: the Objection, the Modification, and Closure
The biquaternionic form is the corpus's own statement of Maxwell's equations, derived above from the four three-vector equations. The author who supplied the shock-inclusive Cauchy theory reads the classical system differently — as defective — and the corpus records the interpretation here, attributed, because it is the stated motive for the modification that defines her programme.
The objection. In Alexeyeva's account the classical system has three faults: the two vector equations are not connected to the two scalar equations; the resulting system is "of a mixed hyperbolic–elliptic type", which she holds to contradict the wave nature of electromagnetic propagation; and it does not describe the longitudinal electromagnetic waves she takes to be observed. The first two are precise and reproducible. The vector pair's characteristic equation is
$$ \nu_\tau^2\left(\frac{\nu_\tau^2}{c^2} - \|\vec\nu\|^2\right)^2 = 0, $$
so the vector system is hyperbolic with the light cone for its characteristics, while the scalar pair $\epsilon\,\mathrm{div}\,\vec E = \rho_E$, $\mathrm{div}\,\vec H = 0$ is purely spatial and elliptic; its Coulomb potential solves $\Delta_\Phi \xi_{\mathrm{scp}} = \rho_E$, whose solution decays at infinity, which the author reads as incompatible with propagation. The corpus's own interpretation is the standard one: the vector equations are the evolution system and the scalar equations are constraints on its initial data, and the two are not meant to be joined by further dynamics. That the constraint sector is elliptic is a feature of the $3+1$ split, not a defect of the four-dimensional equation. The disagreement is one of interpretation, and the corpus records it only because the modification is built on it.
The modification. Joining the vector and scalar equations into one connected system is what the scalar $\alpha$-field does. A complex scalar $\alpha=\alpha_1+i\alpha_2$ is added to the field strength, so that the scalar and vector parts of the single biquaternionic equation, which without it are separate statements about separate components, become coupled: each component of $\alpha$ enters one of the vector equations through a gradient and one of the scalar equations through a time derivative. In the $\vec E,\vec H$ normalisation of the Hamiltonian form the coupled first-order system is
$$ \mathrm{rot}\,\vec H - \epsilon\,\partial/\partial r'_t\vec E + c^2\,\mathrm{grad}\,\alpha_1 = \vec j_E, \qquad \mathrm{rot}\,\vec E + \mu\,\partial/\partial r'_t\vec H - c^2\,\mathrm{grad}\,\alpha_2 = \vec j_H, $$
$$ \epsilon\,\mathrm{div}\,\vec E + \partial/\partial r'_t\alpha_1 = \rho_E, \qquad -\mu\,\mathrm{div}\,\vec H - \partial/\partial r'_t\alpha_2 = \rho_H, $$
in which each of the eight equations carries the field. The normalisation of $\alpha$ is the author's and is not fixed by the algebra alone, because $\tilde F$ carries $\sqrt\epsilon$ in its electric half and $\sqrt\mu$ in its magnetic half; the system is quoted from the 2016 paper and is not re-derived here.
The one sign that is not free. The two $\alpha_2$ entries — the gradient $-c^2\,\mathrm{grad}\,\alpha_2$ in the second equation and the time derivative in the fourth — are the two halves of one construction, and they must carry the same relative sign. As first printed, with $+\partial/\partial r'_t\alpha_2$ in the fourth equation, they do not. The test is the principal symbol $P(\vec\zeta,\omega)$ of the eight-equation system in the Fourier variables $(\vec\zeta,\omega)$, whose determinant is
$$ \det P = -\frac{1}{\epsilon\mu}\left(\epsilon\mu\omega^2+\lVert\vec\zeta\rVert^2\right)\left(\epsilon\mu\omega^2-\lVert\vec\zeta\rVert^2\right)^3 . $$
The light cone is present, and with it a branch $\omega=\pm i\lVert\vec\zeta\rVert/\sqrt{\epsilon\mu}$ at imaginary frequency — exponential growth and decay in time, not propagation. With the sign as displayed above, the same computation gives
$$ \det P = \frac{1}{\epsilon\mu}\left(\epsilon\mu\omega^2-\lVert\vec\zeta\rVert^2\right)^4 , $$
the light cone alone, with every zero of multiplicity four. Both determinants were computed symbolically, and both were confirmed numerically on the singular values of the symbol.
The static limit cannot test the difference, because a time derivative vanishes there. That is why the printed system and the static system of The Electro-Gravimagnetic Field and the Magnetic-Charge–Mass Hypothesis both looked right while the time-dependent one was not hyperbolic; the corrected sign is the one that carries the claim of hyperbolicity from the static limit to the full system.
The author calls it the Maxwell–Dirac system because its differential operator coincides with the differential part of the Dirac operator; the corpus's The Biquaternion d'Alembertian and Its Green's Functions records the related shifted-gradient form of the same name, and The Electro-Gravimagnetic Field and the Magnetic-Charge–Mass Hypothesis supplies the physical interpretation of $\alpha$ as the attraction–resistance field, its front conditions, and the static limit in which it gives Poisson's equation. The corpus records the corrected system as a first-order hyperbolic completion of the Maxwell pair and takes no position on whether the modification is needed.
Closure. The author's own assessment is that the biquaternionic form is mathematically and physically self-consistent but unclosed: the field equations do not determine their own charges and currents, and closure requires the material equations for $\rho$ and $\vec j$, supplied in the companion papers. This is the corpus's own recurring theme — the four-current $\tilde J_{\Phi}$ must be supplied, not derived — restated in the author's terms, and it is the reason the programme's later papers turn to a law of motion for the charge–current field.
The three-vector form in Maxwell's Equations in a Material Medium above is the corpus's derivation; the modified system above and the objection that motivates it are recorded from the author's 2016 paper, and the corpus's own equation is not altered by them.
A counter-claim, from a formulation arranged differently. One external formulation claims the opposite outcome, and it belongs beside the modification. E. P. J. de Haas (Biquaternion Formulation of Relativistic Tensor Dynamics, arXiv:1401.4470v1, 2013) states that his biquaternion tensor calculus "lacks the extra terms that usually arise in biquaternionic electrodynamics", and he gives the condition in the body: the electromagnetic force biquaternion matches the standard field only if the Lorenz gauge holds, that is the divergence of the potential vanishes — he writes $F_0 = 0$ for it, which in the corpus's convention is $\sum_\mu\partial A_\mu/\partial R_\mu = 0$; off that gauge the usual biquaternion expressions for the Lorentz force and for the two inhomogeneous Maxwell equations carry extra terms which his arrangement does not, the terms arising from rearranging the tensor indices into biquaternion slots, an operation he calls external to his system. So the contrast with the $\alpha$-modification is exact and worth stating plainly: the corpus's own Maxwell equations, and the corpus's Gauge Structure section, are reached with no added scalar, and the gauge is fixed by construction; the modification adds a scalar to the field strength because the reduction is held to fail without it; and the counter-claim says the extra terms are an artefact of the mapping and the gauge rather than of the algebra. Both external positions are recorded, attributed, and neither alters the corpus's equation. The full accounting is in Relativistic Mechanics in Biquaternionic Form and Curved Spacetime and the Biquaternion Framework.
The Biquaternionic Energy–Momentum
The energy and momentum of the electromagnetic field are encoded in a biquaternionic energy–momentum, the halved Hermitian form of the field strength:
$$ \boxed{\;\tilde\Sigma_{\Psi} = \frac{1}{2}\tilde{F}\tilde{F}^{*} = W + \frac{i}{c}\vec{S}\;}, $$
where $W$ is the energy density and $\vec S$ is the energy flow density (the Poynting vector), and the star is the Hermitian conjugate of the algebra, the map of Algebra and Element Conventions, §The Conjugations, which conjugates each coefficient and negates the vector part. The star is not optional here: the complex conjugate alone would give $-\bigl(W+\tfrac{i}{c}\vec S\bigr)$, since the electric and the magnetic halves of $\tilde F$ carry opposite phases. For the electromagnetic field in a medium, these are
$$ W = \frac{1}{2}\left(\epsilon\,\vec E\cdot\vec E + \mu\,\vec H\cdot\vec H\right), \qquad \vec S = \vec E\times\vec H. $$
The scalar part of $\tilde\Sigma_{\Psi}$ is the energy density, real; the vector part is $(i/c)\vec S$, purely imaginary. The object is therefore an element of the Hermitian subspace $\mathbb{B}_{\Psi}$, as every Hermitian form is, and the factor $i$ on the Poynting part is required: in the $i r'_t$ convention a temporal component carries an $i$ relative to a spatial one, exactly as the scalar part of the source $\sqrt{\mu}\,\tilde{J}_{\Phi}$ above is $i\rho/\sqrt{\epsilon}$ rather than $\rho/\sqrt{\epsilon}$. The form $\tilde{F}\tilde{F}^{*}$ is computed in The Field-Strength Biquaternion and Its Invariants; the factor $\tfrac{1}{2}$ is fixed by the requirement that the scalar part be the energy density. The energy–momentum tensor $T_{\mu\nu}$ is the source of the gravitational field in any theory that couples gravity to electromagnetism; $\tilde\Sigma_{\Psi}$ carries its energy density and its energy flux.
The energy law is the scalar part of the biquaternionic equation
$$ \tilde\nabla_\Phi \tilde\Sigma_{\Psi} = -\tilde{\mathcal F}, $$
where $\tilde{\mathcal F}$ is the biquaternionic power–force density, whose scalar part is fixed by the work done on the charges. Only the scalar part of this equation is used below. Writing out the two parts of the left-hand side, with $\partial/\partial R_t=-i\,\partial/\partial r'_t$,
$$ \mathrm{Sc}\bigl(\tilde\nabla_\Phi\tilde\Sigma_{\Psi}\bigr)=\frac{\partial W}{\partial R_t}-\frac{i}{c}\,\mathrm{div}\,\vec S, \qquad \mathrm{Vect}\bigl(\tilde\nabla_\Phi\tilde\Sigma_{\Psi}\bigr)=\frac{1}{c}\frac{\partial\vec S}{\partial r'_t}+\vec\nabla_\Phi W+\frac{i}{c}\,\mathrm{rot}\,\vec S . $$
The scalar part is the Poynting theorem: with $\mathrm{Sc}(\tilde{\mathcal F})=-\frac{i}{c}\vec J\cdot\vec E$ the equation reads
$$ \frac{\partial W}{\partial r'_t}+\mathrm{div}\,\vec S+\vec J\cdot\vec E=0 . $$
The vector part is not the momentum balance. That balance pairs $\partial/\partial r'_t\,\vec S/c^2$ and $\vec\nabla_\Phi W$ with the divergence of the Maxwell stress, a symmetric rank-two tensor, and a four-component biquaternion cannot carry it; Exercise: The Electromagnetic Energy–Momentum Tensor constructs the tensor and shows where the four-component object stops. The conservation law used in this article is therefore the energy law, the scalar part of the biquaternionic equation.
Energy Conservation and the Cauchy Problem
The biquaternionic energy–momentum $\tilde\Sigma_{\Psi} = \tfrac{1}{2}\tilde{F}\tilde{F}^{*} = W + (i/c)\vec S$ satisfies the energy law $\mathrm{Sc}(\tilde\nabla_\Phi\tilde\Sigma_{\Psi}) = -\mathrm{Sc}(\tilde{\mathcal F})$, where $\tilde{\mathcal F}$ is the biquaternionic power–force density. Integrating the scalar part over a spacetime region $D$ with smooth boundary $\partial D$, and using the divergence theorem, gives
$$ \int_{\partial D} \mathrm{Sc}\bigl(\tilde U_{\mathrm n} \tilde\Sigma_{\Psi}\bigr)\,dS = -\int_D \mathrm{Sc}\bigl(\tilde{\mathcal F}\bigr)\,dV, $$
where $\tilde U_{\mathrm n} = i n_t e_t + \vec n$, $n_t^2 + \lvert\vec n\rvert^2 = 1$, is the outward unit normal on $\partial D$. This is the biquaternionic form of the Poynting theorem: the flux of energy through the boundary equals the work done by the sources inside.
The energy law implies the uniqueness of solutions of the Cauchy problem. Suppose two solutions $\tilde{F}_1$ and $\tilde{F}_2$ satisfy the same initial data at $r'_t = 0$ and the same four-current $\tilde J_{\Phi}$. Their difference $\tilde{F} = \tilde{F}_1 - \tilde{F}_2$ satisfies the homogeneous equation $\tilde\nabla_\Phi\tilde{F} = 0$ with zero initial data. The energy integral
$$ E(r'_t) = \int_{\mathbb{R}^3} W(r_x\,\vec e_x+r_y\,\vec e_y+r_z\,\vec e_z, r'_t)\,d^3 x = \frac{1}{2}\int_{\mathbb{R}^3} \|\tilde{F}(r_x\,\vec e_x+r_y\,\vec e_y+r_z\,\vec e_z, r'_t)\|^2\,d^3 x $$
is positive-definite and, by the conservation law, time-independent. Since $E(0) = 0$, it follows that $E(r'_t) = 0$ for all $r'_t$, and hence $\tilde{F} = 0$ everywhere. So the Cauchy problem for the biquaternionic Maxwell equation has at most one solution with finite energy. This is the biquaternionic version of the standard uniqueness theorem for Maxwell's equations.
The energy method also provides a stability statement: small perturbations of the initial data lead to small perturbations of the solution, in the $L^2$ norm. The biquaternionic formulation preserves the standard well-posedness of the Maxwell Cauchy problem.
The argument closes under weaker hypotheses than differentiability. In Alexeyeva's treatment the classical solution is continuous and differentiable everywhere except on finitely many wave fronts, on which it satisfies the jump condition. The energy identity then holds with the surface terms included, because the jump of the energy integrand on each front vanishes by the shock article's energy-jump relation, whose electric-energy component is $[W]_{F_t} = c^{-1}(\vec u,[\vec P]_{F_t})$. The energy integral is still positive-definite and time-independent, so the classical solution — shock fronts included — is unique. This is stronger than the finite-energy uniqueness above: the plain energy method presumes enough smoothness to apply the divergence theorem, whereas the distributional statement covers the discontinuous solutions that the shock article needs.
The Lorentz Transformation of the Potential
The transformation of the four-potential under a Lorentz boost is the rotor conjugation
$$ \boxed{\;\tilde{A}_{\Phi,2} = \tilde U\,\tilde{A}_{\Phi}\,\tilde U^{*}\;}, $$
where $\tilde U$ is the boost biquaternion. For a pure boost with velocity $\vec v_b$, the boost biquaternion is
$$ \tilde U = \cosh\frac{\theta_{\mathrm{rpd}}}{2} + i\sinh\frac{\theta_{\mathrm{rpd}}}{2}\,\vec u = \exp\!\left(\frac{\tilde{Ξ}}{2}\,i\vec u\right), $$
with the rapidity $\theta_{\mathrm{rpd}}$ related to the velocity by $\tanh\theta_{\mathrm{rpd}} = v_b/c$, and $\vec u = \vec v_b/v_b$ the unit vector in the direction of the boost. The Lorentz factor — a number, not a frame basis operator, and unrelated to the $\gamma$ names of The Legacy Pauli Gamma Formulation below — is
$$ \gamma = \frac{1}{\sqrt{1 - v_b^2/c^2}} = \cosh\theta_{\mathrm{rpd}} . $$
The boost rotor is Hermitian, $\tilde U^{*}=\tilde U$, while its inverse is its quaternion conjugate, $\tilde U^{-1}=\tilde U^{\natural}$; the star in the conjugation law is therefore not an inversion, and the two-sided product is what carries the four-vector. The full development of the boost biquaternion, its relation to the four-velocity, and its action on the four-potential, is the subject of the companion article on the Lorentz transformation. Here we recall the resulting component formulas for the transformation of the scalar and vector potentials:
$$ \xi_{\mathrm{scp},2} = \gamma\left(\xi_{\mathrm{scp}} - \vec v_b\cdot\vec A\right), $$
$$ \vec A_2 = \vec A + \frac{\gamma - 1}{v_b^2}(\vec v_b\cdot\vec A)\vec v_b - \gamma\frac{\xi_{\mathrm{scp}}}{c^2}\vec v_b. $$
These are the standard Lorentz transformation formulas for the four-potential, obtained from the rotor conjugation $\tilde{A}_{\Phi,2} = \tilde U\tilde{A}_{\Phi}\tilde U^{*}$. The same transformation law applies to the four-current $\tilde J_{\Phi}$ and to the biquaternionic energy–momentum $\tilde\Sigma_{\Psi}$. This confirms that the biquaternionic formulation is Lorentz-covariant, and it shows how the transformation laws look in biquaternionic form.
A one-sided alternative, and why it fails. One sometimes meets the one-sided formula $\tilde{A}_{\Phi,2} = (\gamma/c)\tilde{U}\tilde{A}_{\Phi}$ with $\tilde{U} = c + i\vec v_b$, in which the rotor multiplies on the left only. It does not reproduce the standard component formulas for a general boost, and it does not match the rotor-conjugation formulation established in the companion articles. The two-sided conjugation given above is the correct transformation.
The Vacuum Limit
In vacuum, $\epsilon = \epsilon_0$ and $\mu = \mu_0$, and the speed of light becomes
$$ c_0 = \frac{1}{\sqrt{\epsilon_0 \mu_0}}. $$
All the equations above carry over with the substitution $c \to c_0$, $\epsilon \to \epsilon_0$, $\mu \to \mu_0$. In particular, the field-strength biquaternion becomes
$$ \tilde{F} = i\sqrt{\epsilon_0}\,\vec E - \sqrt{\mu_0}\,\vec H, $$
and the potential biquaternion becomes
$$ \tilde{A}_{\Phi} = i\,a'_t + \vec A. $$
The complex time coordinate becomes $i r'_t$, recovering the familiar $i r'_t$ form with the vacuum speed of light. The structure of the equations is unchanged. This is the content of the statement that the vacuum is a special case of a medium: the same equations describe both, with different values of $\epsilon$ and $\mu$.
The Question of What Is Fundamental
In the standard view, $c_0$ is a fundamental constant of spacetime, and $\epsilon_0$ and $\mu_0$ are properties of the vacuum that are related to $c_0$ by
$$ c_0 = \frac{1}{\sqrt{\epsilon_0 \mu_0}}. $$
In the material-medium view, $c$ is a derived quantity: it is determined by the properties $\epsilon$ and $\mu$ of the medium. In a fluid, the speed of sound is determined by the density and compressibility. By analogy, one may ask whether the speed of light in vacuum is determined by the electromagnetic properties $\epsilon_0$ and $\mu_0$ of the vacuum, rather than the other way around.
This is a structural question, not a settled one. In the SI as revised in 2019, $c_0$ alone is defined exactly; $\mu_0$ is measured, with a relative uncertainty of order $10^{-10}$ fixed by that of the fine-structure constant, and $\epsilon_0 = 1/(\mu_0c_0^2)$ is derived from it. Before 2019 the pair was the other way round: $\mu_0$ was the defined constant and $\epsilon_0$ the derived one. But the choice of which constants are defined and which are derived is a matter of convention, not physics. The physical question is whether the vacuum has electromagnetic properties in the same sense that a material medium does, or whether $\epsilon_0$ and $\mu_0$ are merely conversion factors between unit systems.
The Field Bivector and the Field Operator
The corpus separates an element — a number — from an operator — a number with a slot to be filled. The groups Focus on the Standard Element Representations and Focus on the Polar Decomposition develop the numbers and the group Focus on Operator Representations develops the operators, and the separation is not book-keeping: it decides which objects may be added to one another and which must be applied to an argument.
The Maxwell equation of this article is written with an element. The field strength $\tilde{F}=i\sqrt{\epsilon}\,\vec E-\sqrt{\mu}\,\vec H$ is a bivector, a number, and the equation $\tilde\nabla_\Phi\tilde{F}=-\sqrt{\mu}\,\tilde{J}_{\Phi}$ multiplies numbers. The electromagnetic field tensor is the associated operator: an object that takes a four-vector argument and returns a number, formed from the bivector and its order reverse. Written as an action on an argument, the operator is
$$ F(X) = \tfrac12\left(X\tilde{F} + \tilde{F}^{\sim}X\right), $$
where $\tilde{F}^{\sim}$ is the order reverse of the bivector, the first term multiplying the argument on its left and the second on its right. Gsponer and Hurni write the same expression with a slot, $F(\;)=\tfrac12([\;]\tilde{F}+\tilde{F}^{\sim}[\;])$, where $[\;]$ marks the slot; the source attributes the distinction between the bivector $\tilde{F}$ and the operator $F(\;)$ to Kilmister, who drew it in 1955, and treats the two as objects that must not be confused.
The distinction matters because the field's derived quantities come with slots. The source writes the electromagnetic energy–momentum tensor as $4\pi T(\;)=\tfrac12\tilde{F}^{+}[\;]\tilde{F}$ and the Lorentz force density as a value of the tensor operator at the gradient, again with a slot. The tensor operator is the same construction with the argument standing between the two factors, $T(X)=\tfrac{1}{8\pi}\tilde{F}^{+}X\tilde{F}$. In the corpus's vocabulary the point is the element/operator one: the field-strength number is the subject of this article and of the companions on the field strength, while the operator built from it belongs to The Element and Operator Representations, Two-Sided Operator and the Polar Decomposition of an N-Regular, Spectrally Regular Biquaternion and Bilinear Operators on a Hermitian Module with Hermitian Adjoint. The corpus writes the Maxwell equation with the element and does not profit from the operator form; the source's warning is that the two must be told apart before either is used.
One silent assumption. The compact form also assumes that the potential is its own order reverse, $\tilde{A}_{\Phi}=\tilde{A}_{\Phi}^{\sim}$ — that is, that it is ordinal invariant. Splitting the equation into the parts that are even and odd under reversal, the reversal-even part is the field equation carrying the source and the reversal-odd part is the vanishing of the odd source; both belong to the companion article The Proca Equation: Massive Spin 1 in Biquaternionic Form. A reversal-odd potential is the case the same construction reads as a different field rather than as an electromagnetic one.
The single-term form and spatial reversal. Gsponer and Hurni record a caveat about the writing of the equation rather than about the field. The widely used quaternion form of Maxwell's second set carries a single term — the gradient of the field strength proportional to the current — and the source contrasts it with the two-term form in which the field equation is the reversal-symmetric half of a pair. The two agree only when the reversed term vanishes, and the source identifies that term with the source-free relation recorded above; the single-term form is therefore covariant under spatial reversal exactly when the potential and the current are reversion invariant, which is the ordinal-invariance assumption of the paragraph above, and not otherwise.
In the corpus's conventions the separation is explicit, and it was checked. For a pure-vector field strength $\tilde{F}=\sum_\mu F_\mu \vec e_\mu$ the two order-parities of $\tilde\nabla_\Phi\tilde{F}$ are
$$ \tfrac12\left(\tilde\nabla_\Phi\tilde{F}+\tilde{F}^{\sim}\tilde\nabla_\Phi\right)=\partial/\partial R_t\,\vec F-\mathrm{div}\,\vec F\,e_t, \qquad \tfrac12\left(\tilde\nabla_\Phi\tilde{F}-\tilde{F}^{\sim}\tilde\nabla_\Phi\right)=\mathrm{rot}\,\vec F, $$
where the second product is read with the derivative acting on the reversed field strength and the unit standing on the right, $\tilde{F}^{\sim}\tilde\nabla_\Phi=\sum_\mu(\partial/\partial R_\mu\tilde{F}^{\sim})\,e_\mu$. The first line carries the time derivative and the divergence, a biquaternion whose scalar part is $-\mathrm{div}\,\vec F$ and whose vector part is $\partial/\partial R_t\,\vec F$; the second line is the spatial curl. The one-term equation $\tilde\nabla_\Phi\tilde{F}=-\sqrt{\mu}\,\tilde{J}_{\Phi}$ therefore fixes the sum of the two halves and neither half separately, and the divergence half is not identically zero: on a test field of mixed polynomial structure both its scalar and its vector part are nonzero. So the compact equation is a statement about the sum; its spatial-reversal covariance rests on the two halves coinciding, and the source reads that coincidence as the source-free condition rather than as a property the writing inherits from the equation.
Maxwell's Vacuum Equation as a Cauchy–Riemann Condition
The vacuum equation $\tilde\nabla_\Phi\tilde{F}=0$ has an interpretation that is not a reformulation. Lanczos observed, in his 1919 thesis, that it is the direct four-dimensional generalisation of the Cauchy–Riemann analyticity condition: the pair of two-dimensional Cauchy–Riemann equations is replaced by the single biquaternion equation, and classical electrodynamics becomes a biquaternionic field theory of regular functions, in which the point singularities are read as electrons. The field at a point is then given by the generalisation of Cauchy's integral formula, an integral over a three-dimensional hypersurface $\Sigma$ surrounding the point,
$$ \tilde{F}(X) = \frac{-1}{2\pi^2}\int_\Sigma \frac{R}{|R|^4}\, d^3\Sigma\, \tilde{F}(Y), \qquad R = Y-X, \quad |R|^2 = R\bar{R}, $$
in the source's notation. The function theory behind it — Fueter's for real quaternions (1932), later extended to biquaternions and to algebras of higher dimension — is the corpus's subject in Fueter Theory for Biquaternions, Biquaternion Regular Functions, Biquaternion Analysis and Biquaternion Integration, and the elliptic Cauchy kernel $\tilde{Q}^{\natural}/\|\tilde{Q}\|^4$ of the retarded-Green's-function section above is the kernel this interpretation is built on.
The two kernel normalisations. The factor $-1/2\pi^2$ here is the normalisation of the Cauchy kernel, and it does not conflict with the $1/4\pi^2$ of the fundamental solution of the d'Alembertian recorded in Biquaternion Integration, where $\Delta_\Phi\|\tilde{Q}\|^{-2} = -4\pi^2\delta_0$. The two kernels are separated by one application of the Cauchy–Riemann operator, which lowers the degree by one: applied to the degree-$(-2)$ kernel $\|\tilde{Q}\|^{-2}$ it produces a degree-$(-3)$ kernel proportional to $\tilde{Q}\|\tilde{Q}\|^{-4}$, and the proportionality carries the factor $2$ that converts the $4\pi^2$ of the second-order kernel into the $2\pi^2$ of the first-order one. The Cauchy normalisation is fixed by the surface integral of the kernel, $\int_{S^3_r} n\,G\,dS = 1$ with $G = \tilde{Q}^{\natural}/(2\pi^2\|\tilde{Q}\|^4)$ and $n = \tilde{Q}/\|\tilde{Q}\|$, which has been checked directly. The two constants therefore belong to two kernels of two different orders, and there is no inconsistency between them.
What the interpretation adds to the equation is a regularity content for the vacuum half: the homogeneous Maxwell pair is the statement that the field is a regular biquaternionic function of the position, so the corpus's analysis machinery — the Cauchy formula, residues, contour deformation — is available for the source-free equations. The source records the interpretation as Lanczos's and as an interpretation of the classical theory; it changes no prediction, and it is the framework's most direct contact with the analytic side of the corpus.
The same equation with the time written out. The compact equation can be put in a form in which the time derivative stands alone, $i\partial/\partial r'_t\tilde{F} + D\tilde{F} = 0$ with $D$ the spatial part, and in that form it is the equation K. Imaeda formulated in 1976 as a new formulation of classical electrodynamics. The two writings differ only in how the four-gradient is split and in the inertia convention of the biquaternion units; both say that the vacuum field is annihilated by the four-dimensional Cauchy–Riemann operator. The form is worth recording because it is the one on which the separation of the time is performed — the fixed-frequency reduction of the next paragraph is exactly that separation applied to it — and because it is the massless member of the family of shifted equations that the dictionary of The Dirac Equation in Biquaternionic Form produces. Imaeda's paper carries the quaternionic vacuum Maxwell equation that the corpus here uses under the names of Conway, Silberstein and Lanczos, and it is recorded in the Further Reading below.
The same equation at fixed frequency. The interpretation above is four-dimensional. In the time-harmonic regime it degenerates to a three-dimensional one, and the object is the same. In the convention $\exp(-i\theta_{\mathrm{pwt}})$ of Exercise: Plane-Wave Propagation in a Medium, the temporal derivative $\partial/\partial R_t$ acts as $-\omega/c$ on a monochromatic field, so the vacuum equation $\tilde\nabla_\Phi\tilde{F}=0$ becomes
Throughout, $\theta_{\mathrm{pwt}}:=\omega r'_t/c$ is the plane-wave time angle: $\omega$ is real for a monochromatic wave, so the factor $\exp(-i\theta_{\mathrm{pwt}})$ is a phase of modulus one and it is the time part of the plane-wave angle of the full wave form.
$$ \Bigl(D_3-\frac{\omega}{c}\Bigr)\tilde{F}=0, \qquad D_3=\vec e_x\partial/\partial R_x+\vec e_y\partial/\partial R_y+\vec e_z\partial/\partial R_z, \qquad k_\Phi=\frac{\omega}{c}=\omega\sqrt{\epsilon\mu}, $$
whose scalar part is the constraint $\mathrm{div}\,\vec F=0$. The field-strength biquaternion is therefore left regular for the shifted three-dimensional Cauchy–Riemann operator $D_3-k_\Phi$ of Quaternion Regular Functions: four-dimensional regularity becomes Helmholtz–Dirac regularity in three dimensions, with the shift supplied by the frequency. The companion combination $\vec G=i\sqrt{\epsilon}\vec E+\sqrt{\mu}\vec H$ satisfies the same equation, $(D_3-k_\Phi)\vec G=0$, because a solution of $\tilde\nabla_\Phi\tilde F=0$ has $\mathrm{rot}\,\vec E=k_\Phi\vec E$ and $\mathrm{rot}\,\vec H=k_\Phi\vec H$ separately, and the plus in $\vec G$ leaves both eigenvalues at $+k_\Phi$. The shift $+k_\Phi$ is carried by the other handedness: solutions of $(D_3+k_\Phi)\tilde F=0$ have $\mathrm{rot}\,\vec F=-k_\Phi\vec F$, the Beltrami field of opposite chirality, and $+k_\Phi$ is also the shift that the equation takes if the temporal factor is $e^{+i\theta_{\mathrm{pwt}}}$ instead of $e^{-i\theta_{\mathrm{pwt}}}$. The two halves of the field therefore decouple into two first-order equations whose shifts coincide at $k_\Phi$ in vacuum and separate in a chiral medium, where they become the two wavenumbers $k_\Phi/(1\mp k_\Phi\beta)$ of Maxwell's Equations in Chiral Media — the diagonalization again, read from the four-dimensional side.
The Internal Freedom of the Vacuum Field
The vacuum equation $\tilde\nabla_\Phi\tilde{F}=0$ is Lorentz covariant: a change of frame acts on $\tilde{F}$ on the left, and left multiplication by the rotor is what the covariance means. Lanczos recorded, in the fourth chapter of the 1919 thesis, a second freedom, on the right, which the same equation possesses and which is not a change of frame. His equation (4.8) states that with $\tilde{F}$ the function $\tilde{F}\tilde{q}_0$ is also a solution, for any constant quaternion $\tilde{q}_0$; the unit norm in the display below is the normalisation that keeps the invariants of the field, not a condition the equation imposes:
$$ \tilde\nabla_\Phi(\tilde{F}\tilde{q}_0) = (\tilde\nabla_\Phi\tilde{F})\,\tilde{q}_0 = 0, \qquad \langle\tilde{q}_0\rangle_{\natural} = 1 . $$
The proof is one line — the constant passes through the operator — and it has been checked directly: with $\tilde{q}_0$ a unit quaternion and $\tilde{F}$ a smooth biquaternion field, the residual of the identity is at the level of the finite-difference error. The constants that leave the field's own normalisation untouched are the unit quaternions, a three-parameter group, $\mathrm{SU}(2)\cong\mathbb{H}^\times/\mathbb{R}_{>0}$; the general constant adds a real rescaling of the field on top of that group. In Lanczos's words the transformation of the versor $\tilde{F}$ is "not uniquely fixed at all" by the field equations: a rotation of the frame, with the two unit-quaternion characteristics $\tilde{p},\tilde{q}$, transforms the field as
$$ \tilde{F}_2 = \tilde{p}\,\tilde{F}\,\tilde{q}\,\tilde{q}_0 , $$
so the equations imply the left ($P$) transformation and leave the right ($Q$) transformation free up to the arbitrary constant factor $\tilde{q}_0$. Equivalently: the homogeneous equations fix only the left action, and the solution space is a right module over the constants.
The Lorentz character of the field is not fixed by the equation. The choice of $\tilde{q}_0$ decides which Lorentz object the field is, and the homogeneous Maxwell equations do not decide it. Lanczos distinguishes three cases. For the electromagnetic six-vector — the bivector field of the corpus's $\tilde{F}$ — the constant is the one that makes $\tilde{F}_2$ transform as a six-vector; for a four-vector the constant is the other element, and Lanczos proposes that the field then be the gradient of a scalar potential, which he associates with gravitation; and the third case, which he did not take, is the one in which the field transforms as the simplest of the three, the massless spin-1/2 field. Gsponer and Hurni note that Lanczos had thus reached, in 1919, the freedom that contains the massless spin-1/2 field, nine years before Dirac, and that he returned to exactly this point after 1928, when he used the quaternion formalism to display the spacetime covariance of the spin-1 and spin-1/2 equations.
The corpus records the invariance and the three interpretations as the source's and the commentators'. What is structural is the statement of the freedom: the vacuum Maxwell equation does not determine the spin content of its own solutions; that is fixed by the constant $\tilde{q}_0$, an extra datum the equation leaves free. The three-parameter group of the freedom is the $\mathrm{SU}(2)$ that Why the Biquaternion Algebra Reaches SU(2) but Not SU(3) identifies as the algebra's intrinsic compact group, reached here from the vacuum equation rather than from a gauge construction. The corpus does not adopt Lanczos's assignment of the four-vector case to gravitation; it records that the freedom exists, that its group is $\mathrm{SU}(2)$, and that the electromagnetic interpretation is the one member of the three that the rest of the corpus uses.
Beyond the Minkowski Subspace: Complexified Spacetime
Up to this point the potential has been taken in the anti-Hermitian subspace $\mathbb{B}_{\Phi}$ (the material sector): with $A_0 = i\,a'_t$ and $\vec A$ real, the time coefficient is purely imaginary and the three spatial coefficients are real. The field strength is in neither subspace, and its two halves are the two physical halves of $\tilde F = i\sqrt{\epsilon}\,\vec E - \sqrt{\mu}\,\vec H$.
The electric half $i\sqrt{\epsilon}\,\vec E$ is the Hermitian part of the field and the magnetic half $-\sqrt{\mu}\,\vec H$ the anti-Hermitian part, which is the sector split of Algebra and Element Conventions, §The Hermitian and Anti-Hermitian Subspaces, read on $\tilde F$. In the full biquaternion algebra $\mathbb{B}$, the coefficients $A_\mu$ are arbitrary complex numbers, and the potential and field-strength biquaternions become fully complexified
$$ \tilde{A}_{\Phi} = A_0 + \vec A, \qquad \tilde{F} = \vec F, $$
with complex coefficients that are not restricted to either subspace. Restricting to the material slice reproduces the classical electromagnetic field; the remaining components belong to the additional directions of the complexified spacetime.
The biquaternionic formulation extends naturally to this setting. The gradient $\tilde\nabla_\Phi$ becomes an operator on the complexified coordinates, and the factorization $\tilde\nabla_\Phi(\tilde\nabla_\Phi)^{\natural}= \Box_\Phi\,e_t$ continues to hold, with $\Box_\Phi$ now the complexified d'Alembertian. The single equation $\tilde\nabla_\Phi \tilde{F} = -\sqrt{\mu}\,\tilde{J}_{\Phi}$ remains valid, but the fields and sources are now fully complex. The integrability condition $\mathrm{Sc}((\tilde\nabla_\Phi)^{\natural}\sqrt{\mu}\,\tilde{J}_{\Phi}) = 0$ remains the conservation law for the complexified source.
This is the natural generalization of the $i r'_t$ structure. The electromagnetic field is the material slice of a complex field, just as the material sector $\mathbb{B}_{\Phi}$ is a real slice of the complexified biquaternion algebra. The structure of Maxwell's equations is preserved, but the arena is larger.
The Legacy Pauli Gamma Formulation
The Pauli and gamma formulation is legacy: the reasons are explained in the article The AXF Matrices: Axis-Free, Integer Entries, §Why the Pauli and the Gamma Displays Are Refused. The generators carry no frame label: $\gamma_\mu$ is one set of matrices, and the frame is named on the coordinate.
The body writes its operators as basis operators $L_aR_b$, a left multiplication followed by a right one: $L_a$ multiplies on the left by $a$ and $R_b$ on the right by $b$. The gamma names are the names of the frame basis operators. Two basis operators multiply by multiplying the left factors and the right factors, the right factors in reverse order,
$$ L_aR_b\,L_cR_d=L_{ac}R_{db}, $$
because $R_bR_d=R_{db}$. The gammas are a frame of $\mathrm{Cl}_{1,3}(\mathbb{C})=M_4(\mathbb{C})$, not a different algebra:
$$ \gamma_0=iL_{\mathrm{pau},e_t}R_{\mathrm{pau},\vec e_x},\quad \gamma_1=iL_{\mathrm{pau},\vec e_x}R_{\mathrm{pau},\vec e_y},\quad \gamma_2=iL_{\mathrm{pau},\vec e_y}R_{\mathrm{pau},\vec e_y},\quad \gamma_3=iL_{\mathrm{pau},\vec e_z}R_{\mathrm{pau},\vec e_y},\quad \gamma_5=-iL_{\mathrm{pau},e_t}R_{\mathrm{pau},\vec e_z}. $$
The same in the axis-free display:
$$ \gamma_0=iL_{\mathrm{axf},e_t}R_{\mathrm{axf},\vec e_x},\quad \gamma_1=iL_{\mathrm{axf},\vec e_x}R_{\mathrm{axf},\vec e_y},\quad \gamma_2=iL_{\mathrm{axf},\vec e_y}R_{\mathrm{axf},\vec e_y},\quad \gamma_3=iL_{\mathrm{axf},\vec e_z}R_{\mathrm{axf},\vec e_y},\quad \gamma_5=-iL_{\mathrm{axf},e_t}R_{\mathrm{axf},\vec e_z}. $$
A single product is a pure tensor; the products form the seven-dimensional cone of rank at most one, and a general element is a sum of at most four.
Every generator carries exactly one factor of $i$, so $(iL_aR_b)^2=-L_{a^2}R_{b^2}$ and all the signs are fixed by $i^2=-1$ together with $\vec e_\eta^2=-e_t$:
$$ \gamma_0^2=+I_4,\qquad \gamma_1^2=\gamma_2^2=\gamma_3^2=-I_4,\qquad \gamma_5^2=+I_4, $$
with $\{\gamma_\mu,\gamma_\nu\}=2\varepsilon_\mu\delta_{\mu\nu}I_4$, $\varepsilon=(1,-1,-1,-1)$, and $\{\gamma_5,\gamma_\mu\}=0$. The four frame basis operators multiply to $-i\gamma_5$, that is, $\gamma_5$ is $i$ times their product,
$$ \gamma_0\gamma_1\gamma_2\gamma_3=-i\gamma_5, \qquad \gamma_5=i\gamma_0\gamma_1\gamma_2\gamma_3 . $$
The single $i$ in a generator is also what makes a doubled $i$ collapse: multiplying a generator by $i$ returns a plain basis operator,
$$ i\gamma_\mu=-L_{\mathrm{pau},e_\mu}R_{\mathrm{pau},\vec e_{\nu(\mu)}}, $$
In the axis-free display:
$$ i\gamma_\mu=-L_{\mathrm{axf},e_\mu}R_{\mathrm{axf},\vec e_{\nu(\mu)}}, $$
with $\vec e_{\nu(t)}=\vec e_x$ and $\vec e_{\nu(\eta)}=\vec e_y$, together with $i\gamma_5=+L_{\mathrm{pau},e_t}R_{\mathrm{pau},\vec e_z}$, the sign there being the opposite one because $\gamma_5$ was defined with the opposite sign of $i$. Every product of two imaginary units is therefore a sign, and no display in this article carries a doubled $i$.
In these names the field operator of The Field Bivector and the Field Operator above is $\tfrac12(R_{\tilde F}+L_{\tilde F^{\sim}})$; since the reverse of a bivector is minus itself, $\tilde F^{\sim}=-\tilde F$, and the operator is half the commutator with the field strength,
$$ F(\;)=\tfrac12\bigl(R_{\tilde F}-L_{\tilde F}\bigr). $$
$\gamma_5$ is the reference-frame basis operator of the chirality operator. The central grading $\chi=i\tilde{Ζ}_{\mathrm{ad}}$ — the difference of the two central idempotents of the complexified algebra, so $\chi^2=e_t$ — is a different object, represented by $\gamma_5$ on the Dirac module.
The two-sided structure, and the two Dirac equations. The basis operators come in two families, $L$ and $R$, and that is what separates the corpus's two Dirac equations of Biquaternion Representation of Spin One-Half and the Relativistic One-Electron Atom. The one-sided equation acts with the left family alone, $L_{\tilde\nabla_\Phi}\tilde{Ξ}_{R}=k_m\tilde{Ξ}_{L}$ and $L_{(\tilde\nabla_\Phi)^{\natural}}\tilde{Ξ}_{L}=k_m\tilde{Ξ}_{R}$, and it splits the field into its two chirality halves. The two-sided equation carries the mass through the right family, $iL_{\mathrm{pau},e_t}\partial/\partial r'_t\,F+\sum_\eta L_{\mathrm{pau},\vec e_\eta}a\partial/\partial R_\eta\,F-mi\,R_{\mathrm{pau},\vec e_z}\overline{F}=0$; its chirality sits in that right multiplication, and its mass term is antilinear, the complex conjugation $\overline{F}$ being what the right basis operator acts on. The two share the frame and not the operator. The field operator of The Field Bivector and the Field Operator above is the third object of the same kind, $\tfrac12(R_{\tilde F}-L_{\tilde F})$, one basis operator from each family, antisymmetrised. In this article the field equation is written with the element; the operators of The Field Bivector and the Field Operator above are written as actions on an argument, and the gamma names are attached to the frame basis operators here and nowhere else. Here $\hbar/(mc)$ is the mass length.
The dictionary is complex-linear; over $\mathbb{R}$ the products span $\mathrm{Cl}_{1,3}(\mathbb{R})\cong M_2(\mathbb{H})$ and the basis operators span $\mathrm{Cl}_{3,1}(\mathbb{R})\cong M_4(\mathbb{R})$.
The display. Every array printed in this section is given twice: in the Peirce display first, labelled $\mathrm{pau}$, whose arrays are the ones the literature prints, and in the axis-free display after it, labelled $\mathrm{axf}$, which is the display of the body and of the corpus. The two are carried into one another by $L_{\mathrm{axf},e_\mu}=C_{\Pi}L_{\mathrm{pau},e_\mu}C_{\Pi}^{-1}$, and a conjugation by $C_{\Pi}$ changes no product, no anticommutator, no trace and no determinant, so every identity of the section holds in both. An array is a statement about one display; a basis operator named in the running text keeps the Peirce label, and its axis-free array is the conjugate. The label names the array and not the operator.
Summary
Maxwell's equations in a linear, isotropic, non-dispersive medium collapse into the single biquaternionic equation $\tilde\nabla_\Phi\tilde{F} = -\sqrt{\mu}\,\tilde{J}_{\Phi}$, where $\tilde\nabla_\Phi$ is the biquaternionic gradient of the conventions, $\tilde{F} = i\sqrt{\epsilon}\,\vec E - \sqrt{\mu}\,\vec H$ is the field-strength biquaternion (the Riemann–Silberstein vector of Silberstein's 1907 construction, up to an overall factor), and $\sqrt{\mu}\,\tilde J_{\Phi} = i\rho/\sqrt{\epsilon} + \sqrt{\mu}\,\vec J$ is the source, the normalised four-current, with $\tilde J_{\Phi} = ic\rho\,e_t+\vec J$. One equation replaces the four standard Maxwell equations, and no explicit $\epsilon$ appears, because the medium is carried by the definition of $\tilde{F}$ and by the single factor $\sqrt{\mu}$. The scalar and vector parts of the single equation reproduce the Hamiltonian form, $\mathrm{div}\,\vec F = R_0$ and $\partial/\partial R_t\,\vec F + \mathrm{rot}\,\vec F = -\vec R$. The d'Alembertian is the scalar $\Box_\Phi$, the quaternionic norm of the gradient, with $\tilde\nabla_\Phi(\tilde\nabla_\Phi)^{\natural}= \Box_\Phi\,e_t$, and the conservation of electric charge is exactly the integrability condition $\mathrm{Sc}((\tilde\nabla_\Phi)^{\natural}\tilde{J}_{\Phi}) = 0$. The single equation is first order and eliminates no field, and that is the property the classical reduction to a second-order wave equation lacks: the reduced equation admits solutions, among them longitudinal ones, that violate the divergence constraints. At fixed frequency the same equation reduces to the shifted Cauchy–Riemann condition $(D_3-k_\Phi)\tilde{F}=0$ together with its companion, and in a chiral medium the two shifts separate into the two wavenumbers and the two circular polarizations. The vacuum equation also carries an internal freedom beyond its Lorentz covariance: right multiplication by any constant unit quaternion leaves it invariant, a three-parameter $\mathrm{SU}(2)$ that the equation does not fix and that Lanczos read as the freedom selecting the Lorentz character of the field — the electromagnetic six-vector, a four-vector, or the massless spin-1/2 that he did not take.
The retarded Green's function supplies the causal solution, the field at a point depending only on the sources in its past light cone; it is the fundamental solution of the hyperbolic wave operator and is related by the Wick rotation to the elliptic Cauchy kernel $\tilde{Q}^{\natural}/\|\tilde{Q}\|^4$ used in biquaternion analysis. In the stationary limit the equation becomes purely spatial and the solution is the gradient of the Newton kernel, whose first term is Coulomb's law for a point charge and whose second is the Biot–Savart law. The potential $\tilde{A}_{\Phi}$ is subject to the gauge transformation $\tilde{A}_{\Phi,\mathrm g} = \tilde{A}_{\Phi} - \tilde\nabla_\Phi\Gamma$, under which $\tilde{F}$ is invariant while the scalar $S = \mathrm{Sc}((\tilde\nabla_\Phi)^{\natural}\tilde{A}_{\Phi})$ shifts by $S_2 = S - \Box_\Phi\Gamma$; the Lorenz gauge $S = 0$ reduces the first-order equation to the wave equation $\Box_\Phi\tilde{A}_{\Phi} = -\mu\tilde J_{\Phi}$.
The energy and momentum of the field are carried by the halved Hermitian form $\tilde\Sigma_{\Psi} = \tfrac{1}{2}\tilde{F}\tilde{F}^{*} = W + (i/c)\vec S$, an element of the Hermitian subspace $\mathbb{B}_{\Psi}$ whose scalar part is the energy density $W$ and whose vector part is $(i/c)\vec S$, carrying the Poynting flux. Its scalar part obeys the energy law $\mathrm{Sc}(\tilde\nabla_\Phi\tilde\Sigma_{\Psi}) = -\mathrm{Sc}(\tilde{\mathcal F})$, which integrates to the biquaternionic Poynting theorem and gives uniqueness and stability of the Cauchy problem with finite energy; its vector part is not the momentum balance, which requires the rank-two Maxwell stress. The four-potential transforms by rotor conjugation, $\tilde{A}_{\Phi,2} = \tilde U\tilde{A}_{\Phi}\tilde U^{*}$, which reproduces the standard boost formulas for $\xi_{\mathrm{scp}}$ and $\vec A$ and exhibits the Lorentz covariance of the formulation. The vacuum limit is the substitution $\epsilon \to \epsilon_0$, $\mu \to \mu_0$, $c \to c_0$, leaving the structure of the equations unchanged; whether $c_0$ or the pair $(\epsilon_0,\mu_0)$ is fundamental is a structural question, not a settled one. Finally, the formulation extends to fully complexified coefficients, in which the electromagnetic field is the real projection of a complex field on the larger arena of complexified spacetime.
Summary of Notation
| Symbol | Meaning |
|---|---|
| $\mathbb{B}$ | Biquaternion algebra |
| $e_t = 1, \vec e_x, \vec e_y, \vec e_z$ | Quaternion basis |
| $i$ | Scalar imaginary, $i^2 = -1$, commutes with $\vec e_\mu$ |
| $\mathbb{B}_{\Phi}$ | Anti-Hermitian subspace (material sector) |
| $\mathbb{B}_{\Psi}$ | Hermitian subspace (informational sector) |
| $\tilde{A}_{\Phi}$ | Potential biquaternion |
| $\tilde{F}$ | Field-strength biquaternion |
| $G_{\Box,\Phi}$, $\tilde G_{\Phi,\tilde\nabla_\Phi}$ | Scalar retarded kernel $(4\pi R)^{-1}\delta(r'_t-R)$ and the first-order kernel $(\tilde\nabla_\Phi)^{\natural}G_{\Box_\Phi}$ |
| $\delta'$ | Derivative of the delta with respect to its argument |
| $\vec N = \vec E + ic\vec B$ | Riemann–Silberstein vector; $\tilde F = i\sqrt{\epsilon}\,\vec N$ |
| $L_a$, $R_b$ | Left and right multiplication, by $a$ on the left and by $b$ on the right |
| $\vec r$ | Field point $r_x\,\vec e_x+r_y\,\vec e_y+r_z\,\vec e_z$ |
| $\rho_f$ | Mass density of the fluid (not the charge density $\rho$) |
| $\tilde J_{\Phi} = ic\rho\,e_t+\vec J$ | Four-current; the source enters as $\sqrt{\mu}\,\tilde J_{\Phi}$ |
| $\tilde Q^{*}$ | Hermitian conjugate: complex conjugation followed by the quaternion conjugate |
| $\tilde\Sigma_{\Psi} = \tfrac{1}{2}\tilde{F}\tilde{F}^{*}$ | Energy–momentum biquaternion, in $\mathbb{B}_{\Psi}$ |
| $\tilde U$ | Boost biquaternion (unit-norm biquaternion) |
| $\tilde{q}_0$ | Constant unit quaternion; the internal (right-multiplication) freedom of the vacuum equation, in $\mathrm{SU}(2)$ |
| $\tilde\nabla_\Phi$ | Biquaternionic gradient |
| $(\tilde\nabla_\Phi)^{\natural}$ | Quaternion conjugate gradient |
| $\Box_\Phi$ | d'Alembertian, a scalar, the quaternionic norm of the gradient; $\tilde\nabla_\Phi(\tilde\nabla_\Phi)^{\natural}=\Box_\Phi\,e_t$ |
| $\epsilon, \mu$ | Permittivity and permeability of the medium |
| $\epsilon_0, \mu_0$ | Permittivity and permeability of vacuum |
| $c = 1/\sqrt{\epsilon\mu}$ | Speed of light in the medium |
| $c_0 = 1/\sqrt{\epsilon_0\mu_0}$ | Speed of light in vacuum |
| $D_3 = \vec e_x\partial/\partial R_x+\vec e_y\partial/\partial R_y+\vec e_z\partial/\partial R_z$ | Three-dimensional Cauchy–Riemann operator (time-harmonic interpretation) |
| $P(\vec\zeta,\omega)$ | Principal symbol of the eight-equation system of The Author's Reading, whose determinant gives the characteristic set |
| $F, F^*, cF$ | Real field matrices of the matrix formulation and their complexification, $cF := F - iF^*$; the star here is the paper's dual matrix, not the Hermitian conjugate $\tilde Q^{*}$ of the algebra |
| $c\nabla$ | The operator matrix of the same formulation |
| $k_\Phi = \omega\sqrt{\epsilon\mu}=\omega/c$ | Wave number in the medium at fixed frequency; the shift of $D_3-k_\Phi$, that is the eigenvalue of $\mathrm{rot}$ on the field |
| $\vec v$ | Particle or frame velocity |
| $\vec v_f$ | Fluid velocity |
| $\vec v_b$ | Boost velocity |
| $\langle\tilde{Q}_1,\tilde{Q}_2\rangle_{\natural}$ | the general quaternionic bilinear form, $\mathrm{Sc}(\tilde{Q}_1^{\natural}\tilde{Q}_2)$; on the diagonal $\langle\tilde{Q}_2,\tilde{Q}_2\rangle_{\natural}=\langle\tilde{Q}_2\rangle_{\natural}$ |
| $\langle\tilde{Q}_1,\tilde{Q}_2\rangle_{*}$ | the general plain sesquilinear form, $\mathrm{Sc}(\tilde{Q}_1\tilde{Q}_2^{*})$; on the diagonal $\langle\tilde{Q}_2,\tilde{Q}_2\rangle_{*}=\sum_\mu\lvert Q_{2\mu}\rvert^{2}$ |
| $\langle\tilde{Q}_1,\tilde{Q}_2\rangle$ | the general plain bilinear form, the scalar part of the general plain bilinear product, $\mathrm{Sc}(\tilde{Q}_1\tilde{Q}_2)$ |
Further Reading
- The AXF Matrices: Axis-Free, Integer Entries, for the reasons the AXF matrices replace the Pauli and gamma matrices.
- Hermann Minkowski, "Space and Time" (1908), reprinted in The Principle of Relativity (Dover).
- Albert Einstein, The Meaning of Relativity (Princeton, 1922), for the $i r'_t$ formulation of special relativity.
- James Clerk Maxwell, A Treatise on Electricity and Magnetism (Dover, 1954), for the original formulation.
- G. Rousseaux and É. Guyon, "À propos d'une analogie entre la mécanique des fluides et l'électromagnétisme", Bulletin de l'Union des Physiciens 96 (2002), no. 841 (2), 125–134, for the fluid–electromagnetism correspondence of The Hydrodynamic Form of the Equations — the Lamb vector, the dictionary the text attributes to Marmanis, the two Galilean limits and their fluid interpretations — and for the historical claim, recorded in A Brief History of Biquaternions in Physics, that the mechanical model of the field was the origin of the equations rather than an interpretation of them. An expository article in a teachers' journal, in French; cited for the analogy's content and its historical claims, not as primary literature.
- Lev Landau and Evgeny Lifshitz, The Classical Theory of Fields (Pergamon, 1975), for the four-dimensional formulation.
- L. A. Alexeyeva, "Hamiltonian Form of the Maxwell Equations and Its Generalized Solutions", Differential Equations 39(6) (2003) 807–816 (arXiv:0705.3153 is the Russian original), for the Hamiltonian form and the shock-inclusive Cauchy theory.
- L. A. Alexeyeva, "Maxwell Equations, Their Hamiltonian and Biquaternionic Forms and Properties of Their Solutions", Mathematical Journal 16(2) (2016) 90–103, ISSN 1682-0525, for the objection to the classical system, the $\alpha$-coupled Maxwell–Dirac system, the unclosed diagnosis, and the operator factorization.
- A. Waser, "Application of Bi-Quaternions in Physics" (2000, updated 2007), for the biquaternionic energy–momentum and the Lorentz transformation of the potential.
- A. W. Conway, "On the applications of quaternions to some recent developments of electrical theory", Proceedings of the Royal Irish Academy 29 (1911) 1–9, and L. Silberstein, "Quaternionic form of relativity", Philosophical Magazine 23 (1912) 790–809, for the compact biquaternionic Maxwell equation $\vec\nabla_\Phi\tilde{F}=-4\pi J$ used here.
- C. W. Kilmister, "The application of certain linear quaternion functions to tensor analysis", Proceedings of the Royal Irish Academy 57 (1955) 37–99, for the distinction between the electromagnetic field bivector and the field operator (tensor).
- D. H. Gottlieb, "Maxwell's equations" (1 August 2004, 12 pp.), for the matrix formulation compared in A fourth route above: the single complex matrix equation $cF(-\partial/\partial r'_t,\vec\nabla_\Phi)^{\mathsf T} = (\rho,\vec J)^{\mathsf T}$, the dual form with the operator matrix $c\nabla$, the operator identities $c\nabla^2 = \Delta_\Phi I$ and $(\partial/\partial r'_t I - c\nabla)(\partial/\partial r'_t I + c\nabla) = (\partial^2/\partial {r'_{t}}^2 - \Delta_\Phi)I$, and the commutation of the field matrices with their conjugates. The paper's equations (13) and (14) are not reproduced here: their vector parts carry the term and sign defects recorded in A fourth route. The paper's $4 \times 4$ matrices are those of the second realization of The Element and Operator Representations, where the transposition identity they satisfy is stated.
- C. Lanczos, "Die funktionentheoretischen Beziehungen der Maxwellschen Aethergleichungen — Ein Beitrag zur Relativitäts- und Elektronentheorie" (Budapest, 1919; reprinted in the Collected Published Papers, Vol. VI, A-1–A-82; English typescript arXiv:physics/0408079), for the interpretation of the vacuum Maxwell equation as the four-dimensional Cauchy–Riemann condition and the generalised Cauchy formula.
- K. Imaeda, "A new formulation of classical electrodynamics", Il Nuovo Cimento B 32 (1976) 138–162, for the quaternionic vacuum Maxwell equation $i\partial/\partial r'_tF+DF=0$, the form the compact equation takes when the time derivative is left explicit; the source of the time-harmonic dictionary of The Dirac Equation in Biquaternionic Form cites it in the massless case.
- R. Fueter, "Analytische Funktionen einer Quaternionenvariablen", Commentarii Mathematici Helvetici 4 (1932) 9–20, for the analytic function theory of regular quaternion functions.
- A. Gsponer and J.-P. Hurni, "The physical heritage of Sir W. R. Hamilton", arXiv:math-ph/0201058, §§6–7, for the operator/bivector distinction, the ordinal-invariance assumption, and the analytic interpretation of the vacuum equation.
- V. V. Kravchenko, "Quaternionic Diagonalization of Maxwell's Equations" (expository article, National Polytechnic Institute, Mexico City), for the objection to the wave-equation reduction, the diagonalization of the time-harmonic Maxwell system into first-order quaternionic equations, and the extendability problem.
- E. P. J. de Haas, "Biquaternion Formulation of Relativistic Tensor Dynamics," arXiv:1401.4470v1 [physics.gen-ph] (2013), for the counter-claim recorded after The Author's Reading: a biquaternion tensor calculus that reports no extra terms relative to the standard relativistic language, the condition being the Lorenz gauge. Cited for the negative claim and for the tensor-index rearrangement it identifies as the source of the extra terms.
- C. Castro and M. Pavšič, The Extended Relativity Theory in Clifford Spaces (review, 8 July 2004), for the extension of Maxwell theory to an arena whose coordinates are Clifford-valued polyvectors, where the potential is itself a polyvector, $A = \xi_{\mathrm{scp}} + A_{(1)} + A_{(2)} + \dots$ with $A_{(p)}$ its grade-$p$ part, and the field strength $F = dA$ carries a symmetric and an antisymmetric part of every rank, so that an extended charge couples to antisymmetric tensor fields of arbitrary rank. The formalism is standard — $p$-form gauge fields and the coupling of extended objects to them are ordinary exterior calculus — while the interpretation of its scalar and pseudoscalar components as a dilaton and an axion is the review's suggestion ("one could attempt to identify"), and the arena is not this article's $\mathbb{B}$. It is listed because it is the widest generalization of the single equation recorded here.