The Self-Dual and Anti-Self-Dual Split: Spin 1 from the Bivector Plane
Introduction
The biquaternion algebra $\mathbb{B} = \mathbb{C}\otimes_\mathbb{R}\mathbb{H}$ carries four-dimensional real slices, whose elements are the four-vectors of physics when a slice is read as spacetime, and it carries a unique simple module, the spinor module, whose elements are the spinors. The two objects have different transformation characters under the Lorentz group, and the difference is the reason that the framework treats half-integer spin and integer spin by different means. A spinor is an element of a module over the algebra; a four-vector is an element of the algebra itself, and it transforms by conjugation rather than by multiplication. Spin one has to be found in the second setting.
This article isolates the algebraic structure that makes integer spin possible at all: the decomposition of the algebra's antisymmetric tensors into a self-dual and an anti-self-dual part. The decomposition is a standard feature of the Lorentz group — after complexification, the six-dimensional space of two-forms splits into the two three-dimensional complex representations $(1,0)$ and $(0,1)$ — and in the biquaternion algebra it acquires a compact form. The vector part of the algebra is complex three-dimensional, and on it the Hodge dual of the field strength is nothing but minus left multiplication by the scalar imaginary,
$$ \star\tilde{F} = -i\,\tilde{F}. $$
After complexification, the two halves are the two eigenspaces of the Hodge dual, and each is, on restriction to the rotation group, a spin-one object. This is what the phrase "spin one from the bivector plane" means precisely: spin one is not an irreducible module of the algebra, it is a three-dimensional complex representation carried by the vector part of the algebra, equivalently by the antisymmetric tensors built from the four-vectors of a slice. Which slice is read is a matter of which physics is wanted; the split is a fact about the bivectors and not about the slicing.
The article is organised as follows. The first section recalls why integer spin requires a representation that the spinor module does not provide: the four-vector representation of the Lorentz group is $(\tfrac12,\tfrac12)$, which on restriction to rotations contains spin one and spin zero, while the antisymmetric tensor representation is $(1,0)\oplus(0,1)$, which is pure spin one. The second identifies the carrier inside the algebra: the six-dimensional real vector part $\mathbb{B}_{V}$, which is at once the real field-strength space and the Lorentz Lie algebra. The third defines the Hodge dual on it and proves the identity above, which is linear in the field and therefore holds for a superposition of fields and not only for a single Fourier mode. The fourth constructs the projectors and the split, the fifth reads off the spin content of each half and writes the triplet in the algebra's letters, and the sixth connects the split to duality rotations and to the two helicities of the massless field. A closing section states where the split sits in the algebra's module category and where it does not, which is the boundary that the final article of this subcategory draws in general.
The conventions are those of the companion articles. The algebra is $\mathbb{B} = \mathbb{C}\otimes_\mathbb{R}\mathbb{H}$, with basis $e_t = 1, \vec e_x, \vec e_y, \vec e_z$ satisfying $\vec e_\eta^2 = -e_t$ and $\vec e_x \vec e_y = \vec e_z$, and central scalar imaginary $i$. The gradient is $\tilde\nabla_\Phi = e_t\partial/\partial R_t + \vec e_x\partial/\partial R_x + \vec e_y\partial/\partial R_y + \vec e_z\partial/\partial R_z$, with conjugate $(\tilde\nabla_\Phi)^{\natural} = e_t\partial/\partial R_t - \vec e_x\partial/\partial R_x - \vec e_y\partial/\partial R_y - \vec e_z\partial/\partial R_z$, so that $\tilde\nabla_\Phi(\tilde\nabla_\Phi)^{\natural}= \Box_\Phi = \Delta_\Phi - \partial^2/\partial {r^{\prime}_t}^2$. The potential and field strength are
$$ \tilde{A}_{\Phi} = i\,a'_t\,e_t + \vec A, \qquad \tilde{F} = (\tilde\nabla_\Phi)^{\natural}\tilde{A}_{\Phi} - \mathrm{Sc}\!\left((\tilde\nabla_\Phi)^{\natural}\tilde{A}_{\Phi}\right) = i\sqrt{\epsilon}\,\vec E - \sqrt{\mu}\,\vec H, $$
with $\vec H$ the magnetic field and $\vec B = \mu\vec H$ the magnetic induction, and $c = 1/\sqrt{\epsilon\mu}$ the speed of light in the medium. The normalization of the potential is the one for which the two expressions for $\tilde{F}$ in the display agree, a constant multiple of $\tilde{A}_{\Phi}$ being absorbed in it; the field strength itself is fixed unambiguously by the tensor formula $F_{\mu\nu} = \partial/\partial R_\mu A_\nu - \partial/\partial R_\nu A_\mu$, as the companion field-strength article records. The Lorentz rotor acts by conjugation, $\tilde{Q}\mapsto\tilde{U}_{\mathrm{lor}}\tilde{Q}\tilde{U}_{\mathrm{lor}}^{*}$, and the material interval is $\langle\tilde Q\rangle_\natural$ with $Q_t=i q'_t$.
The companion articles, and what each supplies, are the following.
- The Field-Strength Biquaternion and Its Invariants, for the field-strength biquaternion, the Hodge dual and the field invariants.
- Maxwell's Equations in the Biquaternionic Formulation, for the potential biquaternion, the field strength and the gauge structure of the massless case.
- The Spinor Module in Biquaternionic Form and Its Lorentz Action, for the defining module, its conjugate and the Lorentz action on the defining module.
- Integer-Spin Quantization and the Adjoint Action on the Material Sector in Biquaternionic Form, for the explicit construction of the spin-one triplet from the adjoint action: the generators, the commutators, the Casimir and the weight basis.
- The Photon in Biquaternionic Form, for the massless helicity states and the circular polarization basis.
Integer Spin and the Representation It Needs
The finite-dimensional irreducible representations of the Lorentz group are labelled by a pair $(s,s_2)$ of half-integers, with complex dimension $(2s+1)(2s_2+1)$, and the restriction to the rotation subgroup is by the Clebsch–Gordan rule $s\otimes s_2$. Two entries of the list are relevant before any field equation is written.
The four-vector representation is $(\tfrac12,\tfrac12)$, of complex dimension four. Its restriction to the rotation subgroup is
$$ (\tfrac12,\tfrac12)\big|_{SU(2)} \;\cong\; \tfrac12\otimes\tfrac12 \;\cong\; 1\oplus 0, $$
of dimension $3\oplus1$. A four-vector therefore contains a three-dimensional spin-one piece — the spatial components — and a one-dimensional spin-zero piece, the time component. This is the standard statement that a vector field describes spin one together with a scalar admixture, and it is the reason the four-vector alone does not isolate integer spin.
The antisymmetric tensor representation is the antisymmetric square of the four-vector,
$$ \Lambda^2\big((\tfrac12,\tfrac12)\big) \;\cong\; (1,0)\oplus(0,1), $$
of complex dimension $3+3 = 6$, the complexification of the real six-dimensional two-form space. Its two summands are conjugate to one another and, since $(1,0)|_{SU(2)} = 1$ and $(0,1)|_{SU(2)} = 1$, each summand is a pure spin-one multiplet: neither contains a spin-zero piece. The two summands are the self-dual and the anti-self-dual parts of the two-form, distinguished by the eigenvalue of the Hodge dual. This is the representation in which integer spin appears without a scalar admixture, and it is the representation the material sector supplies once its four-vectors are assembled into field strengths.
The dimension count is worth recording because it is the whole content of the split. A real two-form has six real components, which is the same as a complex three-vector; the complexification of the two-form space doubles it to six complex components, and the Hodge dual, whose square is $-1$, separates those six complex components into two three-dimensional complex eigenspaces. The two eigenspaces are the two summands $(1,0)$ and $(0,1)$, and the real two-form is the real form on which the two halves are conjugate.
The Carrier Inside the Algebra
The vector part is the bivector space
An element of $\mathbb{B}$ is written in components as $\tilde{Q} = Q_te_t + Q_x\vec e_x + Q_y\vec e_y + Q_z\vec e_z$ with $Q_\mu \in \mathbb{C}$. The vector part is the space of elements with vanishing scalar part,
$$ \mathbb{B}_{V} = \{\,Q_x\vec e_x + Q_y\vec e_y + Q_z\vec e_z\,\}, $$
a complex vector space of dimension three, equivalently a real vector space of dimension six with basis $\{\vec e_x,\vec e_y,\vec e_z,i\vec e_x,i\vec e_y,i\vec e_z\}$. The same six real basis elements are the six bivectors of the Clifford algebra $\mathrm{Cl}_{1,3}$: the three spatial rotation generators are the images of $\vec e_\eta$, and the three boost generators are the images of $i\vec e_\eta$. The real vector part is therefore at once
- the space of real antisymmetric tensors, that is, of bivectors and of field strengths, through the identification of a complex three-vector with the pair $(\vec E,\vec B)$; and
- the Lorentz Lie algebra $\mathfrak{so}(1,3)$ of rotations and boosts, with the commutator bracket.
That the two are the same six-dimensional real space is not a coincidence of dimension: both are the antisymmetric square of the four-vector space, and in the $i r'_t$ convention the four-vector space is the material sector. The material sector $\mathbb{B}_{\Phi}$ is the four-dimensional real subspace spanned by $ie_t, \vec e_x, \vec e_y, \vec e_z$; its antisymmetric square is the six-dimensional space spanned by the products of distinct basis elements, which is exactly the real vector part of $\mathbb{B}$.
The field strength as a vector of the algebra
The field-strength biquaternion of the companion article is a pure vector,
$$ \tilde{F} = i\sqrt{\epsilon}\,\vec E - \sqrt{\mu}\,\vec H \;\in\; \mathbb{B}_{V}, $$
with imaginary electric part and real magnetic part, and it is obtained from the four-potential by differentiation. This object carries all six real components of $(\vec E,\vec B)$, and the standard complex combination
$$ \vec N = \vec E + ic\,\vec B, \qquad \tilde{F} = i\sqrt{\epsilon}\,\vec N, $$
is the Riemann–Silberstein vector. The identification of $\vec N$ with a complex three-vector is precisely the identification of a real two-form with an element of the complex vector part, and it is the identification that makes the self-dual split transparent: the complex structure that appears in $\vec N$ is the algebra's scalar imaginary, and it is the same complex structure that defines the Hodge dual, as the next section shows.
The Hodge Dual on the Vector Part
Definition
The Hodge dual of a field is defined on the fields themselves, by its action on the pair $(\vec E,\vec B)$,
$$ \star(\vec E,\vec B) = \left(c\vec B, -\frac{\vec E}{c}\right), \qquad \star^2 = -1 , $$
which is the $\theta_{\mathrm{dua}} = \pi/2$ case of the duality rotation used in the companion article on the field-strength biquaternion and its invariants, at the real duality angle $\theta_{\mathrm{dua}}$. On the Riemann–Silberstein vector the dual is multiplication by $-i$,
$$ \star\vec N = -i\vec N, $$
because $\star(\vec E+ic\vec B) = c\vec B - i\vec E = -i(\vec E+ic\vec B)$. Since $\tilde{F} = i\sqrt{\epsilon}\,\vec N$, the same statement holds on the field-strength biquaternion up to the overall constant:
$$ \boxed{\;\star\tilde{F} = -i\,\tilde{F}\;}. $$
Direct check
The identity is worth checking without the Riemann–Silberstein shortcut, because the factors of $\epsilon$ and $\mu$ are the place where a sign can hide. Write $\vec H = \vec B/\mu$. The dual acts on the magnetic field as
$$ \star\vec H = \frac{1}{\mu}\,\star\vec B = -\frac{\vec E}{c\mu}, $$
so that
$$ \star\tilde{F} = i\sqrt{\epsilon}\,\star\vec E - \sqrt{\mu}\,\star\vec H = i\sqrt{\epsilon}\,c\vec B + \frac{\sqrt{\mu}}{c\mu}\,\vec E = \frac{i}{\sqrt{\mu}}\,\vec B + \sqrt{\epsilon}\,\vec E, $$
using $\sqrt{\epsilon}\,c = \sqrt{\epsilon}/\sqrt{\epsilon\mu} = 1/\sqrt{\mu}$ and $\sqrt{\mu}/(c\mu) = \sqrt{\mu}\sqrt{\epsilon\mu}/\mu = \sqrt{\epsilon}$. On the other hand
$$ -i\tilde{F} = -i\left(i\sqrt{\epsilon}\,\vec E - \sqrt{\mu}\,\vec H\right) = \sqrt{\epsilon}\,\vec E + i\sqrt{\mu}\,\vec H = \sqrt{\epsilon}\,\vec E + \frac{i}{\sqrt{\mu}}\,\vec B, $$
which is the same. The identity $\star = -i$ therefore holds for every field, with no assumption on the field's time dependence or on the wavevector.
The identity is linear in the field, so it is enough to test it once and for all on a general superposition. With $\epsilon$ and $\mu$ at fixed numerical values and with $\vec E$ and $\vec B$ taken as a sum of two unrelated uniform fields — equivalently two Fourier components with different wavevectors and independent amplitudes — the residual of $\star\tilde{F} + i\tilde{F}$, computed componentwise from the definitions above, is at the level of $10^{-16}$ in units where the field components are of order one, that is, at the level of the floating-point round-off of the arithmetic. A single plane wave would not test the linearity of the identification, and it is the linearity that lets the dual be read as an operator on the whole vector part.
The dual as an operator on the algebra
Because $\star$ is real-linear and $\star^2 = -1$, it is a complex structure on the six-dimensional real vector part. The identity above says that this complex structure is, up to sign, the algebra's own: left multiplication by $i$ is central, squares to $-1$, and preserves the vector part, and on the field strength
$$ \star\tilde{F} = -i\tilde{F} = \tilde{F}(-i), $$
the scalar imaginary being central, so that left and right multiplication agree. The statement is intrinsic: no Clifford generator, no gamma matrix (no basis operator $T_{\mu\nu}$), and no choice of representation enters. The Hodge dual of the framework's field strength is the scalar imaginary acting on the vector part of the algebra.
The Self-Dual and Anti-Self-Dual Split
Projectors
Because $\star^2 = -1$, the dual is a complex structure on the real vector part. On its complexification, described in the paragraph after the next, it has the eigenvalues $\pm i$, and the projectors onto the two eigenspaces are
$$ \tilde\Pi_{sd,asd} = \tfrac12\left(1 \pm i\,\star\right), \qquad \star \tilde\Pi_{sd,asd} = \mp i\,\tilde\Pi_{sd,asd}. $$
They are idempotent and orthogonal,
$$ \tilde\Pi_{sd}^2 = \tilde\Pi_{sd}, \qquad \tilde\Pi_{asd}^2 = \tilde\Pi_{asd}, \qquad \tilde\Pi_{sd}\tilde\Pi_{asd} = \tilde\Pi_{asd}\tilde\Pi_{sd} = 0, \qquad \tilde\Pi_{sd} + \tilde\Pi_{asd} = 1, $$
and they decompose the complexified vector part into two three-dimensional complex spaces. Following the companion article's convention, the $\tilde\Pi_{sd}$ space, on which $\star = -i$, is the self-dual half, and the $\tilde\Pi_{asd}$ space, on which $\star = +i$, is the anti-self-dual half. Each is a copy of the representation $(1,0)$, respectively $(0,1)$, of the Lorentz group.
The subtlety is that the real vector part is already complex, with the scalar imaginary supplying its complex structure. On $\mathbb{B}_{V}$ itself the dual is therefore the scalar $-i$, every element carries that one eigenvalue, and the split is invisible. The complexification that the projectors require is a second, external complexification: one allows coefficients in a new unit, the $i$ of the projectors, which is distinct from the algebra's $i$ in $\star\tilde{F}=-i\tilde{F}$, so that the complexified vector part is $\mathbb{C}\otimes_\mathbb{R}\mathbb{B}_{V}$, of complex dimension six. On it the two eigenvalues are $\pm i$, each with eigenspace of complex dimension three, and these are the two halves above. The real field strength is a real form of this six-dimensional space: it is an element of the vector part, and the biquaternion that the framework assigns to it carries the self-dual coordinate. Concretely, with the two combinations
$$ \vec N = \vec E+ic\,\vec B, \qquad \vec N^* = \vec E-ic\,\vec B, $$
the field-strength biquaternion and its coefficient conjugate are
$$ \tilde{F} = i\sqrt{\epsilon}\,\vec N, \qquad \bar{\tilde{F}} = -i\sqrt{\epsilon}\,\vec N^*, $$
and the dual acts on the two coordinates with opposite signs, being $+i$ on $\vec N^*$ where it is $-i$ on $\vec N$:
$$ \vec N(\star F) = -i\,\vec N(F), \qquad \vec N^*(\star F) = +i\,\vec N^*(F) . $$
Because the dual commutes with coefficient conjugation, $\star(\bar{\tilde{F}}) = \overline{\star\tilde{F}}$, the two relations are statements about the two coordinates and not about $\tilde{F}$ and $\bar{\tilde{F}}$ separately. The pair $(\vec N,\vec N^*)$ is a real form of the pair of halves, and for a real field the two are complex conjugates of one another, so that $\tilde{F}$ together with its coefficient conjugate already carries all six real components of $(\vec E,\vec B)$.
The split of a general two-form
For a general complexified two-form, written as the sum of a self-dual and an anti-self-dual piece,
$$ \boxed{\;\tilde{F}_{\mathbb{C}} = \tilde{F}_{+} + \tilde{F}_{-}, \qquad \tilde{F}_{+} = \tilde\Pi_{sd}\tilde{F}_{\mathbb{C}} = \tfrac12\left(\tilde{F}_{\mathbb{C}} + i\,\star\tilde{F}_{\mathbb{C}}\right), \qquad \tilde{F}_{-} = \tfrac12\left(\tilde{F}_{\mathbb{C}} - i\,\star\tilde{F}_{\mathbb{C}}\right)\;}, $$
the two pieces are independent and transform separately under the Lorentz group, in $(1,0)$ and in $(0,1)$. This is the decomposition of the antisymmetric tensor into its self-dual and anti-self-dual parts, and in the biquaternion formulation it is the decomposition of a complexified vector of the algebra into the $+i$ and $-i$ eigencomponents of the Hodge dual.
The split is Lorentz covariant, and the reason is short. The Lorentz action on the field strength is complex-linear on the Riemann–Silberstein vector: writing a boost as a complex $3\times3$ matrix $M$ acting on $\vec N$, the same boost acts on $i\vec N$ as $iM$, so the action commutes with multiplication by $i$ and consequently $\star$ commutes with the boost and maps each half to itself. The two halves are therefore invariant subspaces of the Lorentz action, which is what it means for them to be the two three-dimensional complex representations.
The Spin Content of the Two Halves
Restriction to rotations
The rotation subgroup $SU(2)\subset SL(2,\mathbb{C})$ acts on the Lorentz representations by the restriction of the pair $(s,s_2)$ to the diagonal, and the two halves behave as
$$ (1,0)\big|_{SU(2)} = 1, \qquad (0,1)\big|_{SU(2)} = 1, $$
each a three-dimensional, irreducible, spin-one multiplet. This is verifiable from the weights alone. The representation $(1,0)$ has $S_{\mathrm{tot},3}$ weights $\{+1,0,-1\}$, one each, and the quadratic Casimir takes the value $s(s+1) = 2$; the representation $(0,1)$ has the same weights. Both are spin one.
Pure spin one without an admixture
The contrast with the four-vector is the point of the split. The four-vector representation restricted to rotations has weights
$$ (\tfrac12,\tfrac12)\big|_{SU(2)}: \quad \{+1,\,0,\,0,\,-1\}, $$
that is, one state of weight $+1$, one of weight $-1$, and two of weight zero. A single spin-one multiplet would have one state of each of the three weights; the extra weight-zero state is a spin-zero multiplet, and the decomposition is $1\oplus0$. The two-form representation, by contrast, restricts to two copies of the spin-one weight set with no extra zero,
$$ (1,0)\oplus(0,1)\big|_{SU(2)}: \quad \{+1,0,-1\}\oplus\{+1,0,-1\}, $$
so that the antisymmetric tensor is pure integer spin. This is the algebraic content of the statement that the vector potential carries a scalar admixture while the field strength does not, and it is the reason the self-dual split is the natural home of spin one in the framework: it separates the two pure spin-one multiplets from one another, and it leaves no scalar behind.
The triplet in the algebra's letters
The count is realised concretely, and the realisation is the one constructed in the maths article From the Spin to the Carriers and the Closures, §The Anti-Hermitian Space.
The rotation algebra inside $\mathbb{B}$ is the real span of the three pure quaternions, $\mathfrak{g}=\mathrm{span}_{\mathbb{R}}\{\vec e_x,\vec e_y,\vec e_z\}$, with bracket $[\vec e_\mu,\vec e_\nu]=2\epsilon_{\mu\nu\lambda}\vec e_\lambda$. Its Hermitian normalisation is $\tilde S_k=\tfrac12 i\vec e_k$, with $[\tilde S_i,\tilde S_j]=i\epsilon_{ijk}\tilde S_k$ and $\tilde S_k^{*}=\tilde S_k$. The factor $i$ is the only difference between the two normalisations.
The carrier is the vector part itself, of complex dimension three, with the adjoint action $\tilde{Q}\mapsto[\tilde S_k,\tilde{Q}]$. A weight basis is
$$ v_+=\vec e_x+i\vec e_y,\qquad v_0=\vec e_z,\qquad v_-=\vec e_x-i\vec e_y, $$
with
$$ [\tilde S_3,v_\pm]=\pm v_\pm,\qquad [\tilde S_3,v_0]=0,\qquad [\tilde S_+,v_+]=0,\qquad [\tilde S_-,v_+]=-2v_0,\qquad [\tilde S_-,v_0]=v_-. $$
The weights are $+1,0,-1$, one state each, so the maximal weight is $1$, and the Casimir is
$$ \sum_{k=1}^{3}\mathrm{ad}_{\tilde S_k}^{\,2} = 2\ \ \text{on } \mathbb{B}_{V}, \qquad \sum_{k=1}^{3}\mathrm{ad}_{\tilde S_k}^{\,2} = 0\ \ \text{on } \mathbb{B}_{C}, $$
which is $s(s+1)$ at $s=1$ and at $s=0$; here $\mathbb{B}_{C}=\{Q_te_t\}$ is the centre. In the notation of the maths article, where $\mathcal{V}_n$ is the irreducible module of dimension $n+1$, the adjoint module is $\mathbb{B}\cong\mathcal{V}_0\oplus\mathcal{V}_2$.
Three properties of the construction are used below. The highest-weight vector is nilpotent, $v_+^2=(-e_t+i\vec e_z)+(-i\vec e_z+e_t)=0$, so the state of maximal weight multiplies to zero; the vanishing is a statement about the product and not about a length, and $v_+$ has a positive trace-form norm. The generators act by the commutator and not by left multiplication, so the triplet is a carrier and not a module. And the weights are integers where the spinor's are half-integers, so that a rotation by $2\pi$ returns the triplet to itself and the spinor to minus itself.
The generator of a physical rotation of the field strength is the real $D_k=\tfrac12\mathrm{ad}_{\vec e_k}$, not the Hermitian $\tilde S_k$. On $v_+$ it has the eigenvalue $-i$, and multiplication by $i$ recovers the integer weight $+1$. The same factor $i$ therefore makes the generators self-adjoint for the trace form and turns their anti-Hermitian spectrum into the integer weights.
Each half of the two-form is a copy of this triplet. With the axis taken along the propagation direction, the three weights are the three helicity states, and the later section reads them off on a plane wave.
The same two halves as symmetric spinor products
The representation theory of the spinor module supplies a second description of the same split, which will be used again in the final article. The symmetric square of the defining module $S$ is
$$ \mathrm{sym}^2(S) \cong (1,0), \qquad \dim_{\mathbb{C}}\mathrm{sym}^2(S) = 3, $$
and the symmetric square of the conjugate module is $(0,1)$. The self-dual half of the field strength is therefore a symmetric two-spinor, and the anti-self-dual half is a symmetric two-spinor of the conjugate type. The carrier of the split is the pair of symmetric squares, which are representations of the Lorentz group but not modules over $\mathbb{B}$ — a point taken up in the closing section.
Duality, Helicity, and the Two Halves
Duality rotation
The duality rotation of the field-strength article is
$$ \vec N \mapsto \exp(-i\theta_{\mathrm{dua}})\,\vec N, $$
and at $\theta_{\mathrm{dua}} = \pi/2$ it is the classical electric–magnetic duality. On the biquaternion side the rotation is multiplication of the field strength by a central phase, $\tilde{F}\mapsto \exp(-i\theta_{\mathrm{dua}})\tilde{F}$, and it acts on the two halves by opposite phases: because $\star = -i$ on the self-dual coordinate and $+i$ on the anti-self-dual one, the two acquire the phases $\exp(-i\theta_{\mathrm{dua}})$ and $\exp(+i\theta_{\mathrm{dua}})$ respectively, so each half is carried into itself while the relative phase between them runs through the circle. The invariants of the field are the two scalars of the companion articles, $I_1 = \vec E^2 - c^2\vec B^2$ and $I_2 = \vec E\cdot\vec B$, which are the real and imaginary parts of $\vec N\cdot\vec N = I_1 + 2ic\,I_2$; under the duality rotation they mix, $I_1\mapsto I_1\cos2\theta_{\mathrm{dua}}+2cI_2\sin2\theta_{\mathrm{dua}}$ and $2cI_2\mapsto 2cI_2\cos2\theta_{\mathrm{dua}}-I_1\sin2\theta_{\mathrm{dua}}$, so that the modulus $I_1^2 + 4c^2I_2^2$ is invariant under duality while each of $I_1$ and $I_2$ separately is not.
Helicity
For a plane wave of definite wavevector $\vec u_\Phi$, Faraday's law gives $\vec B = \vec u_\Phi\times\vec E/c$ and the transverse amplitude splits into circular components satisfying $\vec u_\Phi\times\varepsilon_{\mathrm{fop},\pm} = \mp i\varepsilon_{\mathrm{fop},\pm}$. Then
$$ \boxed{\;\vec E + ic\vec B = 2E_+\varepsilon_{\mathrm{fop},+}, \qquad \vec E - ic\vec B = 2E_-\varepsilon_{\mathrm{fop},-}\;}, $$
so that the self-dual combination $\vec E+ic\vec B$ contains only the helicity $+1$ amplitude and the anti-self-dual combination $\vec E-ic\vec B$ only the helicity $-1$ amplitude. The two halves of the two-form are therefore the two helicities of the radiation field: on each of two independent transverse directions, an amplitude proportional to $\varepsilon_{\mathrm{fop},+}$ gives a vanishing anti-self-dual combination and an amplitude proportional to $\varepsilon_{\mathrm{fop},-}$ the reverse, and no transverse amplitude gives both. The third helicity eigenvalue, zero, belongs to the propagation direction and is not a transverse polarization; for the massless field the longitudinal mode is removed by the gauge freedom, whose structure is the subject of the companion articles on the Maxwell field. The next article in this subcategory, The Third Polarization and the Self-Dual Split of the Massive Field Strength, takes up the massive case, where the gauge freedom is absent, the longitudinal polarization becomes physical, and the third state has to be counted as a genuine degree of freedom.
Where the Split Lives, and Where It Does Not
The split has been constructed on the vector part of the algebra, and the closing observation is that this is the correct location and not an accident. Three statements hold.
First, the carrier of the split is not a module over $\mathbb{B}$. The algebra $\mathbb{B}\cong M_2(\mathbb{C})$ is simple, its only simple module is the two-dimensional spinor module $S$, and every finite-dimensional module over it is a direct sum of copies of $S$. The vector part is not such a sum: it is invariant under multiplication by central scalars and under the adjoint action, but not under left multiplication by a general element of $\mathbb{B}$. The self-dual half is the symmetric square of the defining module, of complex dimension three, and no module over $M_2(\mathbb{C})$ has odd complex dimension, since each is $S^{\oplus k}$ of dimension $2k$. The split therefore lives in the tensor category generated by the defining module and its conjugate, not in the module category of the algebra.
Second, the four-vector representation from which the two-form was built is itself carried by the algebra by conjugation, not by multiplication, and the same remark applies to it: $\mathbb{B}_{\Phi}$ is a real form of the $(\tfrac12,\tfrac12)$ representation, and it is not a module over $\mathbb{B}$ either. The integer-spin representations of the framework are all of this kind.
Third, the adjoint action of the algebra on itself exhibits the spin-one algebra directly. The inner derivations $D_\eta = \tfrac12\mathrm{ad}_{\vec e_\eta}$, with $\mathrm{ad}_{\vec e_\eta}(\tilde{Q}) = [\vec e_\eta,\tilde{Q}]$, are the adjoint action of a basis of the rotation algebra, which sits inside the scalar-free part, and they preserve the two real triples $\mathrm{span}_\mathbb{R}\{\vec e_x,\vec e_y,\vec e_z\}$ and $\mathrm{span}_\mathbb{R}\{i\vec e_x,i\vec e_y,i\vec e_z\}$, acting on each as an irreducible three-dimensional angular momentum: for instance $D_1(\vec e_y)=\vec e_z$ and $D_1(\vec e_z)=-\vec e_y$, and the same relations hold with $\vec e_\eta$ replaced by $i\vec e_\eta$ throughout. The two triples are the field's own. The magnetic three-vector lies in $\mathrm{span}_\mathbb{R}\{\vec e_x,\vec e_y,\vec e_z\}$ and the electric three-vector in $\mathrm{span}_\mathbb{R}\{i\vec e_x,i\vec e_y,i\vec e_z\}$, so the invariance is the familiar statement that a rotation turns $\vec E$ and $\vec B$ each into itself and never mixes them. The six-dimensional real vector part, read as a Lie algebra rather than as a field-strength space, therefore splits under its own rotation subalgebra into the compact rotation triple and the hyperbolic boost triple, each a real three-dimensional irreducible; the proposition and the remark in §Real Forms and the Rotation Subgroup of the maths article give the statement and its proof, including the observation that the Hermitian generators exchange the two triples while the real ones preserve them. These are the rotation and boost generators, not the self-dual and anti-self-dual halves: the chiral halves are the complex eigenspaces of the Hodge dual, they mix the two triples, and they appear only after the external complexification that the projectors require. The structural reason the same space carries both descriptions is that the field-strength space and the Lorentz Lie algebra are both the antisymmetric square of the material sector.
The general statement that these observations point to — that the algebra's own module category contains only spin one-half, with the scalar carried by the centre and the vector by the adjoint, and that spin above one arrives through tensor products of the defining module — is the subject of the final article of this subcategory, which uses the explicit spin-one case constructed here and the spin-$\tfrac32$ case constructed in The Rarita–Schwinger Equation: Spin 3/2 in Biquaternionic Form.
The reality obstruction
A real field strength is never an eigenform of the dual, and the reason is a count. With $\vec E$ and $\vec B$ real, the condition $\vec E\pm ic\vec B=0$ separates into $\vec E=0$ and $\vec B=0$: six real equations on six real components, whose only solution is the zero field. Nullness is a different condition and does not help: $\vec N\cdot\vec N=0$ is one complex equation, met by a real field when $\vec E\cdot\vec B=0$ and $\lvert\vec E\rvert=c\lvert\vec B\rvert$, and it neither implies duality purity nor follows from it. Nothing is lost by this, because a real field is always the real part of a duality-pure complex field — $\vec N=\vec E+ic\vec B$ is that field — and that is the step by which the programme's complex solutions return to real physics.
The consequence for the split is a statement about what has been decomposed. The projectors $\tilde\Pi_{sd,asd}$ act on the complexified vector part, so the two halves are the halves of a complex field. A real field strength lies in neither half, and the two halves appear separately only after complexification, as the two circular polarizations of a plane wave. The split is therefore a decomposition and never a condition that a real field could be asked to satisfy.
Any scheme that imposes self-duality as an equation must consequently take its field outside the material sector $\mathbb{B}_{\Phi}$. The external algebrodynamical programme does exactly this: its field is $\mathbb{C}$-valued throughout, the duality-purity condition $\vec N=0$ — the anti-self-dual sign in this article's naming — is imposed on that complex field, and the ordinary real field is read off from it by $\Im\vec H=\Re\vec E$, $\Im\vec E=-\Re\vec H$. The Coulomb solution of The Algebrodynamical Programme shows the pattern ($\vec E$ real, $\vec H$ purely imaginary) and The Complex Potential of a Self-Dual Maxwell Field states the general relation with the potential that generates it. No result of that programme transfers to the corpus's real Maxwell field without this step, and the step is a change of carrier, not a gauge choice.
Summary
The material sector $\mathbb{B}_{\Phi}$ is the home of the four-vectors, and the antisymmetric tensors built from them are the carriers of integer spin. The four-vector representation of the Lorentz group is $(\tfrac12,\tfrac12)$, whose restriction to rotations is $1\oplus0$: it carries spin one together with a scalar admixture. The two-form representation is $(1,0)\oplus(0,1)$, whose restriction to rotations is two pure spin-one multiplets with no scalar. The self-dual and anti-self-dual parts of the two-form are these two multiplets.
In the biquaternion algebra the two-form space is the six-dimensional real vector part $\mathbb{B}_{V}$, which is simultaneously the space of real field strengths and the Lorentz Lie algebra. The Hodge dual acts on the field-strength biquaternion as
$$ \star\tilde{F} = -i\,\tilde{F}, $$
that is, as minus left multiplication by the scalar imaginary; equivalently $\star = -i$ on the Riemann–Silberstein vector $\vec N = \vec E+ic\vec B$. Since $\star^2 = -1$, its projectors $\tilde\Pi_{sd,asd} = \tfrac12(1\pm i\star)$ decompose the complexified vector part into two three-dimensional complex spaces, the self-dual and anti-self-dual halves, which are the representations $(1,0)$ and $(0,1)$ and are preserved by every Lorentz transformation because the boost acts complex-linearly on $\vec N$. A real field strength is a real form of this space: its field-strength biquaternion $\tilde{F} = i\sqrt{\epsilon}\vec N$ carries the self-dual coordinate $\vec N = \vec E+ic\vec B$, on which $\star = -i$, and the anti-self-dual coordinate is the combination $\vec N^* = \vec E-ic\vec B$ carried by $\bar{\tilde{F}}$, on which $\star = +i$.
The split is pure integer spin. The self-dual half is the symmetric square of the defining module, $\mathrm{sym}^2(S)\cong(1,0)$, of complex dimension three; the anti-self-dual half is $\mathrm{sym}^2(\bar{S})\cong(0,1)$. In the algebra's letters the triplet is the adjoint module: the vector part $\mathbb{B}_{V}$ with the weights $+1,0,-1$ of $\mathrm{ad}_{\tilde S_3}$, Casimir $2$, and nilpotent highest-weight vector $\vec e_x+i\vec e_y$. The scalar of the adjoint is the centre, with Casimir $0$, and the two together are $\mathcal{V}_0\oplus\mathcal{V}_2$. Under duality rotation the two halves acquire opposite phases, and for a plane wave they are exactly the two helicities: $\vec E+ic\vec B$ carries helicity $+1$ and $\vec E-ic\vec B$ helicity $-1$. It is a split of the complexified field and not of a physical one: a real field strength is an eigenform of neither half, and is recovered from one of them by taking a real part. The carrier of all of this is the vector part of the algebra, which is not a module over $\mathbb{B}$ but a representation carried by conjugation and by tensor products of the defining module; that distinction is the boundary of the algebra's native spin content, and it is drawn in the final article of the subcategory.
Summary of Notation
| Symbol | Meaning |
|---|---|
| $\mathbb{B} = \mathbb{C}\otimes_\mathbb{R}\mathbb{H}$ | Biquaternion algebra, isomorphic to $M_2(\mathbb{C})$ |
| $e_t=1,\vec e_x,\vec e_y,\vec e_z$ | Quaternion basis, $\vec e_\eta^2=-e_t$, $\vec e_x \vec e_y=\vec e_z$ |
| $i$ | Central scalar imaginary, $i^2=-1$ |
| $\mathbb{B}_{\Phi}$, $\mathbb{B}_{\Psi}$ | Material (anti-Hermitian) and informational (Hermitian) sectors |
| $\mathbb{B}_{V}$ | Complex three-dimensional vector part, real dimension six |
| $\mathbb{B}_{C}=\{Q_te_t\}$ | Centre, the complex scalar subspace |
| $\tilde S_k=\tfrac12 i\vec e_k$ | Hermitian rotation generators, $[\tilde S_i,\tilde S_j]=i\epsilon_{ijk}\tilde S_k$, $\tilde S_k^{*}=\tilde S_k$ |
| $D_\eta=\tfrac12\mathrm{ad}_{\vec e_\eta}$ | Rotation generators of the adjoint action, the real form of $\tilde S_k$ |
| $v_+=\vec e_x+i\vec e_y$, $v_0=\vec e_z$, $v_-=\vec e_x-i\vec e_y$ | Weight basis of the adjoint, weights $+1,0,-1$ |
| $\mathcal{V}_n$ | Irreducible module of dimension $n+1$, spin $n/2$ |
| $\tilde\nabla_\Phi, (\tilde\nabla_\Phi)^{\natural}$ | Biquaternionic gradient and its quaternion conjugate; $\Box_\Phi = \tilde\nabla_\Phi(\tilde\nabla_\Phi)^{\natural}= \Delta_\Phi - \partial^2/\partial {r^{\prime}_t}^2$ |
| $\tilde{A}_{\Phi} = i\,a'_t\,e_t + \vec A$ | Four-potential biquaternion |
| $\tilde{F} = (\tilde\nabla_\Phi)^{\natural}\tilde{A}_{\Phi} - \mathrm{Sc}((\tilde\nabla_\Phi)^{\natural}\tilde{A}_{\Phi})$ | Field-strength biquaternion, pure vector |
| $\vec E, \vec H, \vec B=\mu\vec H$ | Electric field, magnetic field, magnetic induction |
| $\epsilon,\mu$, $c=1/\sqrt{\epsilon\mu}$ | Medium constants and speed of light in the medium |
| $\vec N = \vec E+ic\vec B$ | Riemann–Silberstein vector, $\tilde{F}=i\sqrt{\epsilon}\,\vec N$ |
| $I_1 = \vec E^2-c^2\vec B^2$, $I_2 = \vec E\cdot\vec B$ | The two field invariants, real and imaginary parts of $\vec N\cdot\vec N = I_1+2icI_2$ |
| $\star$ | Hodge dual, $\star(\vec E,\vec B)=(c\vec B,-\vec E/c)$, $\star^2=-1$, $\star\tilde{F}=-i\tilde{F}$ |
| $\tilde\Pi_{sd,asd} = \tfrac12(1\pm i\star)$ | Projectors onto the self-dual ($\star=-i$) and anti-self-dual ($\star=+i$) halves |
| $\vec E\pm ic\vec B=0$ | Duality-purity; unavailable to a real field, where it forces $\vec E=\vec B=0$ |
| $(1,0)$, $(0,1)$ | The two three-dimensional complex Lorentz representations |
| $S$ | Defining (spinor) module, $\mathrm{sym}^2(S)\cong(1,0)$ |
| $\tilde{U}_{\mathrm{lor}}\in SL(2,\mathbb{C})$ | Lorentz rotor, $\tilde{U}_{\mathrm{lor}}\tilde{U}_{\mathrm{lor}}^{\natural}=e_t$ |
Further Reading
- From the Spin to the Carriers and the Closures, for the adjoint module, the spins zero and one, the quadratic element, the nilpotent highest-weight vector, and the two real forms of the carrier in the compact and hyperbolic triples.
- The Tensor Square of $\mathbb{C}^2$, for $\mathrm{sym}^2\mathbb{C}^2$ as the spin-one representation of dimension three and $\Lambda^2\mathbb{C}^2$ as the scalar.
- Biquaternion Rotations and Lorentz Transformations, for the $(m,n)$ classification, the adjoint representation $(1,0)\oplus(0,1)$, and the split of the trace-free part into compact rotation and hyperbolic boost directions.
- The Volume Element, Duality and the Hodge Star, for $\star A=-iA$ in the biquaternion algebra and the split of the bivectors in the two cases.
- The Antisymmetric Sector, the Pfaffian and the Klein Quadric of $M_4(\mathbb{C})$, for $\Lambda^2\mathbb{C}^4$ as the Lorentz Lie algebra and its self-dual and anti-self-dual three-spaces.
- Alexandre Proca, "Sur la théorie ondulatoire des électrons positifs et négatifs", Journal de Physique et le Radium 7 (1936) 347–353, for the original massive vector field equation and the third polarization it carries.
- Roger Penrose and Wolfgang Rindler, Spinors and Space-Time (Cambridge, 1984), for the self-dual and anti-self-dual decomposition of the field tensor and the two three-dimensional complex representations of the Lorentz group.
- Iwo Białynicki-Birula and Zofia Białynicka-Birula, "The role of the Riemann–Silberstein vector in classical and quantum theories of electromagnetism", Journal of Physics A 46 (2013) 053001, for the complex-vector description of the field and its duality structure.
- Steven Weinberg, The Quantum Theory of Fields, Vol. I: Foundations (Cambridge, 1995), for the $(s,s_2)$ classification of the finite-dimensional Lorentz representations and the reduction of the four-vector representation.
- J. D. Jackson, Classical Electrodynamics (Wiley, 1999), for the transformation of the fields, the field invariants, and the relation between duality and the circular polarizations.
- Lev Landau and Evgeny Lifshitz, The Classical Theory of Fields (Pergamon, 1975), for the invariant classification of the electromagnetic field and the reality of the self-dual combination for a real field.
- Pertti Lounesto, Clifford Algebras and Spinors (Cambridge, 2001), for bivectors, the Lie algebra of the Lorentz group, and the biquaternion realization of the even Clifford algebra.
- Chris Doran and Anthony Lasenby, Geometric Algebra for Physicists (Cambridge, 2003), for the spacetime-algebra treatment of duality rotations and the field strength as a bivector.