ASF

Maxwell's Equations and Gauge Theory

March 22, 2026|
A
A

A compass needle aligns with a magnetic field threading through empty air. A lightning bolt drives a current through the atmosphere via the same electric force that pulls charged particles in a wire. Radio waves cross a continent at the speed of light. That speed is the constant that emerges from the equations governing both electricity and magnetism. All of this follows from the theory James Clerk Maxwell published in 1865, which Oliver Heaviside recast in the 1880s as four vector equations.

The classical formulation in vector calculus is explicit but hides structure. The equations have a symmetry (gauge invariance) which suggests that the fields E and B are derived objects. The vector potential A underlies them and the geometry of a principal fibre bundle over spacetime underlies A. The same geometric structure reappears with a non-abelian gauge group in the quantum chromodynamics of quarks and the electroweak theory of the Standard Model. In 2000 the Clay Mathematics Institute named the question of whether the non-abelian (Yang–Mills) version of this structure always has a positive mass gap as one of the seven Millennium Prize Problems. The million-dollar prize is unclaimed.

This post develops four parallel languages for Maxwell's equations: the classical vector calculus form, the coordinate-free language of differential forms on Minkowski spacetime, the spacetime algebra (geometric algebra), which encodes all four equations in a single line, and the principal fibre bundle construction, which presents gauge theory as differential geometry. The physical content is identical at each level and only the visible structure changes.

1. Classical Laws

The four Maxwell equations rest on four independent experimental observations, each discovered in a 46-year window between 1785 and 1831. Coulomb measured the force between static charges. Biot and Savart quantified the magnetic field produced by a wire conducting a current. Faraday discovered that a moving magnet induces a voltage in a nearby coil. Ampère related magnetic effects to the currents that produce them. In modern form his law relates closed line integrals of to enclosed currents. In 1861–62 Maxwell added one term to Ampère's law to repair an inconsistency with charge conservation. In 1865 he showed that the equations predict electromagnetic waves travelling at the speed of light.

Two pieces of notation come first.

Charge is an intrinsic scalar property of matter, analogous to mass but signed: it can be positive, negative or zero. Two particles with the same sign of charge repel each other and two with opposite signs attract. Coulomb quantified this force in 1785. That measurement justifies introducing charge as a concept. Classical theory gives no deeper account of why matter has charge: charge is a label that organises the observed forces. Field theory adds structure to the label: charge is the source term that couples matter to the electromagnetic field. Without charge Gauss's law reduces to the homogeneous equation and the field need not be nonzero anywhere. Charge is what makes the field respond to the presence of matter.

Charge density extends this to continuous distributions. When charge is spread through a volume (as in a plasma or the conduction band of a metal) a charge cannot be assigned to each geometric point. The question is how much charge occupies an infinitesimally small region near . Let be a ball of radius centred at . Then

where is the total charge inside that ball at time and . This is the same construction as mass density in continuum mechanics: charge per unit volume, measured in coulombs per cubic metre. The total charge in any region is recovered by integration:

Point charges and the Dirac measure. An electron has all of its charge at a single spatial point to every precision ever measured. A density function for this situation must be zero everywhere except at the electron's location, yet integrate over all of space to the total charge . No function in the ordinary (Lebesgue) sense can achieve this: a function that is zero almost everywhere has integral zero, regardless of the value at any isolated point.

The -dimensional Dirac delta resolves this within the class of distributions (continuous linear functionals on smooth test functions). The three-dimensional Dirac delta centred at the origin is defined by its action on any smooth compactly supported test function :

It is the evaluation functional at the origin. Equivalently is the distributional limit of the family of normalised Gaussians

Drag the slider right to push . The Gaussian concentrates to an infinitely tall spike while its area stays fixed at 1. The Dirac delta is what the family converges to in the distributional sense.

as . For any fixed smooth , one has as . The limit is a spike of unit mass collapsed to a single point. The -dimensional version is the same construction in and satisfies .

The charge density of a point charge at the origin is then . Substituting into Gauss's law and integrating over any sphere enclosing the origin recovers Coulomb's inverse-square field exactly. A point particle moving along a worldline generalises this to , with current density . In field theory the interaction between a charged particle and the electromagnetic potential is the spacetime integral . For a point particle, the delta function collapses this volume integral to a line integral along the worldline. That line integral is the geometric object whose ambiguity encodes gauge freedom and whose measurability produces the Aharonov–Bohm phase in Section 8.

In a conductor or plasma is a smooth function spread continuously through space. The point-supported delta and the smooth distribution are both instances of the same distributional framework and differ only in spatial concentration.

Quantum mechanics. The treatment above assigns charge density via Dirac deltas for point particles or smooth functions for continuous distributions, and treats the electron as occupying a definite location or a definite extended region. Quantum mechanics alters this. An electron occupies a quantum state described by a wavefunction , and is the probability density for finding the electron near . The charge density it contributes is accordingly (negative because the electron has charge ). This is the expectation value of the charge density operator in the state . In an atom is an orbital and is a smooth function spread over the orbital volume. It sources the electric field through Gauss's law exactly as a classical continuous distribution does. Section 10 develops this connection: how quantum charge densities enter Maxwell's equations, how the resulting electrostatic potential feeds back into the Schrödinger equation and how the coupled system is solved computationally.

The closed surface integral adds up the outward flux of a vector field through every infinitesimal patch of a closed surface (one with no boundary, like a sphere or a box). It measures the net outflow of from the enclosed region. The divergence theorem relates this to a volume integral: . It converts the integral form of each law into the differential (pointwise) form below.

Each law is stated as a proposition and derived from the corresponding experiment below.

Proposition (Gauss's law for electricity).

Let be the charge density. The electric field satisfies

Derivation. Coulomb's law gives the field of a point charge at the origin as . The flux through a sphere of radius is

By the divergence theorem, . Replacing by a continuous distribution and equating the integrands gives pointwise.

Proposition (Gauss's law for magnetism).

For any magnetic field produced by currents,

Derivation. The Biot–Savart law gives , which can be written as [eq:b-potential] with . Then by the identity , valid for any smooth vector field.

Proposition (Faraday's law of induction).

The electric and magnetic fields satisfy

Derivation. Faraday observed in 1831 that a changing magnetic flux through a loop drives an EMF:

For a stationary loop, . By Stokes' theorem, . Since this holds for every , the integrands must match: .

Proposition (Ampère–Maxwell law).

For current density and electric field ,

Derivation. Ampère's part. The Biot–Savart law for a long straight wire gives at distance . Integrating around a circular Amperian loop: . By Stokes, this becomes in the static case.

Maxwell's correction. Taking the divergence: , but charge conservation requires and this is nonzero during transients. By Gauss's law, . Adding the displacement current to the right side restores consistency: by continuity.

Corollary (Electromagnetic wave speed).

In vacuum (, ), the fields and satisfy the wave equation with propagation speed

equal to the measured speed of light.

Proof. Apply to [eq:faraday-eq] (Faraday's law): using [eq:ampere-eq] (vacuum Ampère–Maxwell law). The identity (with in vacuum) gives . Maxwell identified m/s with the known speed of light and concluded that light is an electromagnetic wave. The same derivation for gives an identical equation.

Remark (Natural units convention).

Where it reduces clutter without ambiguity we set and (natural units). The speed-of-light factors can be reinstated by dimensional analysis: every time appears where a spatial derivative would sit, supply a factor of .

2. Vector Potential and Gauge Freedom

The equation says that the magnetic field has no divergence. A standard result from vector calculus (the Poincaré lemma in disguise) says that any divergence-free vector field on a contractible domain (for instance all of ) can be written as the curl of another vector field. Simple connectivity does not suffice: is divergence-free on and has no vector potential there. This defines the magnetic vector potential.

Definition (Electromagnetic potentials).

Because , we write

for a vector field called the magnetic vector potential. Substituting into Faraday's law gives . Hence the combination is curl-free and can be written as the gradient of a scalar:

The scalar is the electric scalar potential.

The fields E and B are determined by the pair and different potentials can produce the same fields.

Theorem (Gauge invariance).

For any smooth scalar function , the transformation

leaves the physical fields E and B unchanged: and .

Proof. Direct substitution: since any gradient is curl-free. For E: since partial derivatives commute.

The freedom in choosing is called gauge freedom. Two gauge choices are common:

Definition (Lorenz and Coulomb gauges).

The Lorenz gauge imposes the covariant condition , which reduces the potential equations to decoupled wave equations: and , where is the d'Alembertian.

The Coulomb gauge imposes and makes the scalar potential satisfy Poisson's equation instantaneously, without retardation. This gauge is often used in quantum mechanics and in the canonical quantisation of the electromagnetic field.

The potentials were originally treated as mathematical auxiliaries with no independent physical meaning. Section 8 shows otherwise: the Aharonov–Bohm effect demonstrates that the vector potential has measurable physical consequences in regions where .

3. Differential Forms and the Faraday Tensor

Vector calculus works in flat three-dimensional space but extends poorly to curved spacetime and to general dimensions. Differential forms are coordinate-free and extend naturally. All four Maxwell equations reduce to two statements.

A -form on a smooth manifold is a field of antisymmetric multilinear maps on the tangent space, taking tangent vectors to a number. These spaces are organised into the de Rham complex: a sequence of vector spaces connected by the exterior derivative , which maps -forms to -forms, subject to the identity . This algebraic fact encodes conservation laws, Bianchi identities and the existence of gauge potentials.

Definition (De Rham complex).

On a smooth -manifold , the de Rham complex is the sequence

where is the space of smooth -forms on and each arrow is the exterior derivative. The defining property says that applying the exterior derivative twice gives zero. Hence every exact form (one of the form ) is closed (satisfies ).

Three dimensions. On the sequence specialises to

and the identity recovers two classical identities simultaneously: (the curl of any gradient is zero) and (the divergence of any curl is zero). Both are instances of read at different grades of the complex.

A -form is closed if , and exact if for some -form . Every exact form is closed, but whether every closed form is exact depends on the global topology of . The -th de Rham cohomology group measures the gap:

A nonzero class in corresponds to a closed form that is not globally exact and signals a -dimensional topological hole in .

The Poincaré lemma provides the converse on contractible domains: on any contractible open set , every closed form is exact. In three dimensions this says that on a contractible region any curl-free vector field is a gradient and any divergence-free field is a curl. For the first statement simple connectivity suffices. On non-simply-connected domains (such as ), closed forms need not be exact, and this failure is what allows the Aharonov–Bohm effect (Section 8).

The de Rham complex is the continuum structure that discrete exterior calculus (DEC) discretises: smooth manifolds become simplicial complexes, smooth -forms become -cochains, and the exterior derivative becomes a combinatorial coboundary operator that inherits exactly from the boundary identity . Section 10 develops this construction and uses it to formulate Maxwell's equations on a discrete mesh, where and gauge invariance hold exactly at the discrete level. For Maxwell's equations the relevant complex runs over four-dimensional Minkowski spacetime up to .

Definition (Minkowski spacetime and conventions).

Minkowski spacetime has coordinates with metric signature . The metric tensor is . Lowering and raising indices uses and its inverse .

The Levi-Civita symbol with defines the Hodge star: for a -form with components , the Hodge dual is the -form with components proportional to .

Definition (Faraday 2-form).

The electromagnetic field is encoded in the Faraday 2-form , with components

In terms of the electric and magnetic field components:

Explicitly in coordinates:

The current density defines a 1-form . With these signs gives Faraday's law with the correct sign. The opposite overall sign for both and is also common and leaves the equations below unchanged.

Theorem (Maxwell's equations as two differential form equations).

The four Maxwell equations are equivalent to

The first equation, , encodes Faraday's law and the no-monopole condition . The second, , encodes Gauss's law and Ampère's law.

Proof. We work in coordinates with . With and as in the Definition the field 2-form reads

Equation . Because is a 3-form in four dimensions it has exactly four independent components. The general formula for the antisymmetrised derivative of a 2-form is

Component . This is the only purely spatial triple. Only the magnetic 2-forms contribute (the electric terms involve and vanish on a purely spatial triple):

Setting this to zero gives the no-monopole condition .

Component . Only , and contribute because they are the only field components whose basis 2-forms share two of the three indices :

Setting this to zero gives , the -component of Faraday's law . The remaining two components follow by cycling :

which are the - and -components of .

Equation . The Hodge star in acts on basis 2-forms by with as in the Definition (equivalently ). Applying this to and checking signs with :

The Hodge star has swapped the electric and magnetic sectors. Now apply to :

Component . Only the spatial 2-forms in contribute a purely spatial 3-form:

The source 3-form (the Hodge dual of the current 1-form ) has in this convention. Setting gives Gauss's law .

Component . The magnetic 2-forms , and the spatial form contribute:

Setting equal to gives , the -component of Ampère's law . The and components give the remaining components by the same calculation.

The eight component equations, four from each equation in [eq:maxwell-forms-eq], are Maxwell's four equations in classical form.

Remark (Gauge potential from the Poincaré lemma).

Since the first equation in [eq:maxwell-forms-eq] says F is a closed 2-form. On a contractible domain the Poincaré lemma guarantees a 1-form A such that . This is the gauge potential in differential form language. In terms of the potentials of Section 2 it is . The gauge transformation leaves unchanged since .

The forms language also makes the charge conservation law automatic: applying to and using gives . In coordinates this is the continuity equation .

4. Spacetime Algebra and the Equation

The differential forms approach breaks Maxwell's equations into two statements. Geometric algebra (Clifford algebra applied to spacetime) collapses them into one. The algebraic structure also makes the spinor formulation of relativistic quantum mechanics and the Yang–Mills generalisation (Section 7) more transparent.

Definition (Spacetime algebra Cl(1,3)).

The spacetime algebra (STA) is the real Clifford algebra generated by four basis vectors satisfying

so that and for . The 16-dimensional algebra has a basis organised by grade: grade 0 (scalars, 1 element), grade 1 (vectors, 4 elements), grade 2 (bivectors, 6 elements), grade 3 (trivectors, 4 elements), grade 4 (pseudoscalar, 1 element). The pseudoscalar is with .

The spacetime gradient is , where .

Definition (Electromagnetic field as a bivector).

In the STA the electromagnetic field is a grade-2 multivector (bivector):

where and are relative bivectors (bivectors with one timelike leg), and is the pseudoscalar. In natural units :

since and cyclically.

Theorem (Maxwell's equation in the STA).

All four Maxwell equations are equivalent to the single multivector equation

where is the 4-current. The vector part encodes Gauss and Ampère. The trivector part (when is equated to the grade-1 object J) encodes Faraday and the no-monopole condition.

Proof. Write with . Then and . The Clifford product of a grade-1 vector with a grade-2 bivector decomposes by grade:

where (the dot part) decreases grade by 1 and (the wedge part) increases grade by 1. Since is a bivector is the sum of a grade-1 vector and a grade-3 trivector.

Grade-1 part and the inhomogeneous equations. Expanding in the canonical blade basis (computed explicitly via the Clifford multiplication rules for and ):

The coefficient is ; equating to gives Gauss's law. The coefficient is ; equating to gives Ampère's law .

Grade-3 part and the homogeneous equations. Expanding in the trivector basis:

The coefficients are, up to sign, the components of ; each set to zero gives a component of Faraday's law. The purely spatial trivector coefficient is

which set to zero gives . Since is a grade-1 object, equating forces the grade-3 part to vanish identically and recovers all four homogeneous equations at once. The grade-1 equation then yields the four inhomogeneous equations of [eq:maxwell-forms-eq].

Remark (Yang–Mills in geometric algebra).

For a non-abelian gauge field (Section 7), the STA equation becomes

where the bracket is the commutator in the Lie algebra, and is the non-abelian curvature. The additional commutator term reflects the self-interaction of non-abelian gauge bosons: gluons in QCD have colour charge and interact with each other, unlike photons. The STA formulation shows that this self-interaction is an algebraic consequence of non-commutativity.

5. Principal Fibre Bundles

The gauge potential A of electromagnetism is a 1-form on spacetime with values in a Lie algebra. Geometrically it is a connection on a principal fibre bundle. The physical picture is a family of gauge-equivalent copies of the structure group, one over each point of spacetime.

Think of a helix wrapping around a cylinder. The cylinder is the base space (the spacetime you navigate), the circle at each height is the fibre (the gauge group), and the helix tells you how to move consistently from one fibre to the next. That consistent prescription is a connection.

Definition (Principal G-bundle).

A principal -bundle over a smooth manifold is a smooth manifold together with:

  1. a smooth surjection (the bundle projection),
  2. a smooth free right action , written , that preserves fibres: for all , ,
  3. local triviality: each has an open neighbourhood and a diffeomorphism intertwining the right action with right multiplication in G.

The fibre over is . A local section is a smooth map with .

Example (Hopf bundle S³ → S²).

The 3-sphere is the set of pairs with . The unit circle acts on by . The quotient is the 2-sphere , and the map given by the Hopf map

is a principal -bundle. This bundle is nontrivial because it has no global section: no smooth choice of one representative from each fibre exists over all of . The first Chern class detects this topologically.

Definition (Electromagnetic U(1) bundle).

The gauge group of electromagnetism is . Over Minkowski spacetime the electromagnetic field lives on a principal -bundle . Since is contractible, this bundle is trivial and a global section exists. Hence A is a globally defined 1-form. Over topologically nontrivial spacetimes (for instance outside an infinitely long solenoid, where the effective base space is with fundamental group ), the bundle is still trivial, since . The connection is flat there but its holonomy around the solenoid need not be 1. That holonomy is the Aharonov–Bohm effect.

6. Connections and Curvature

A connection on a principal bundle lifts a path in the base manifold to a path in the total space in a consistent -equivariant way. This defines parallel transport, and the failure of parallel transport to close around loops is curvature.

A connection splits the tangent space of at each point into a vertical part (tangent to the fibre, "staying at the same base point") and a horizontal part (transverse to the fibre, "moving in the base direction"). The connection is a smooth G-equivariant choice of horizontal subspace.

Definition (Connection 1-form).

A connection on a principal -bundle is a -valued 1-form satisfying:

  1. for all , where is the fundamental vector field generated by X,
  2. (G-equivariance under right multiplication).

Given a local section , the local gauge potential is the pullback

This is the gauge field A of physics.

Theorem (Gauge transformation of the local potential).

If is a different local section (a gauge transformation by a smooth map ), then the local potentials are related by

For the abelian case with , this reduces to

which, after identifying , recovers [eq:gauge-transform-eq] from Section 2.

Definition (Curvature 2-form).

The curvature of a connection is the -valued 2-form

where the bracket is the Lie bracket in . Locally via a section the curvature pulls back to

For the abelian case , the Lie algebra is commutative. Hence and

This is the Faraday 2-form of Section 3. The homogeneous equation of [eq:maxwell-forms-eq] is automatic. The inhomogeneous equation is the Yang–Mills equation (Section 7) on this abelian bundle.

Remark (Bianchi identity from geometry).

The Bianchi identity (for the abelian case) is a geometric identity. It says , which follows from . In the non-abelian case the identity becomes the covariant Bianchi identity . It reflects the structure equations of the Lie algebra.

7. Yang–Mills Theory

In 1954 Chen-Ning Yang and Robert Mills asked what happens if the gauge group is non-abelian. The same geometric structure (principal bundle, connection, curvature) applies, but because the group is non-commutative the gauge bosons themselves have charge and interact. Photons are electrically neutral and do not interact with each other. In a non-abelian gauge theory the gauge bosons have charges under their own gauge group and self-interact.

Definition (Yang–Mills action).

For a principal -bundle with connection A and curvature , the Yang–Mills action is

where is the Killing form on and is the Hodge star of the spacetime metric. For with , this is proportional to , the Maxwell action.

Theorem (Yang–Mills equations).

The Euler-Lagrange equations of coupled to a matter current , under variation of A, are

where is the covariant exterior derivative. Together with the Bianchi identity , these are the Yang–Mills equations. For they reduce to [eq:maxwell-forms-eq] (Maxwell's equations).

Example (Standard Model gauge group).

The Standard Model of particle physics uses the gauge group :

  • The factor gives the hypercharge gauge field, which mixes with the gauge boson to produce the photon (electromagnetism) and the Z boson after spontaneous symmetry breaking.
  • The factor governs the weak nuclear force, with gauge bosons and (after the mixing above) .
  • The factor governs quantum chromodynamics (QCD), with 8 gluon gauge bosons that have colour charge. Because is non-abelian gluons interact with each other and produce confinement: quarks cannot be extracted individually from hadrons.

Each of these three factors is a Yang–Mills theory on the same principal bundle framework.

Remark (Millennium Prize problem on the Yang–Mills mass gap).

Pure Yang–Mills theory on with compact simple gauge group (for instance or ) is expected on physical and numerical grounds to have a mass gap: the lowest energy excitation above the vacuum has energy at least . This would explain why the strong force is short-range although gluons are classically massless. The weak force is short-range for a different reason: the Higgs mechanism gives the and bosons mass. A rigorous mathematical proof (constructing a quantum Yang–Mills theory on and proving a positive mass gap) has not been found. The Clay Mathematics Institute lists this as one of the seven Millennium Prize Problems, each with a $1,000,000 prize.[3]

8. Aharonov–Bohm Effect

Consider a solenoid: a tightly wound coil of wire conducting current. Inside the solenoid there is a strong magnetic field. Outside, by symmetry and Ampère's law, the field is nearly zero. Classical electromagnetism predicts that electrons travelling only in the field-free region outside the solenoid should be unaffected by whatever happens inside.

In 1959 Yakir Aharonov and David Bohm predicted that this classical intuition fails. Even though outside, the vector potential is nonzero (it must be, since is nonzero inside). The vector potential has a directly measurable effect in quantum mechanics: it shifts the phase of the electron wavefunction along any path. If two electron paths encircle the solenoid from opposite sides and recombine, the phase difference between the paths depends on the total enclosed magnetic flux.

Theorem (Aharonov–Bohm phase shift).

An electron wavefunction acquires a phase

along a path in spacetime, where is the electron charge and A is the electromagnetic gauge potential. For two paths (upper) and (lower) that together encircle the solenoid, the relative phase difference is

where is the total magnetic flux through the solenoid, is the magnetic flux quantum, and is any surface bounded by . The second equality uses Stokes' theorem. The effect is nonzero whenever even where B is zero on both paths.

The prediction was confirmed experimentally by Chambers (1960) and by Tonomura and collaborators (1986), who used electron holography with a toroidal magnet enclosed in a superconducting shield so that on the electron paths to high precision.

Drag the flux slider to shift the interference fringe pattern. Even though outside the solenoid, the phase difference [eq:ab-phase-eq] grows linearly with flux.

In the language of fibre bundles the AB phase is the holonomy of the connection A around the closed loop formed by the two paths. Holonomy measures how much parallel transport around a loop fails to return to the starting point in the fibre. For an abelian bundle holonomy is a phase in .

Remark (AB effect as holonomy).

The AB phase around a loop is

This can differ from 1 even when the curvature along the entire path (since outside the solenoid). This is possible because the loop is not contractible in the base space: has fundamental group . The winding number of the loop around the excluded origin counts how many times the path encircles the solenoid. A flat bundle (zero curvature) can still have nontrivial holonomy on a topologically nontrivial base.

Hence A is more than a gauge artefact. The local fields E and B do not capture the full physical content of electromagnetism. The full content is the connection on the principal bundle and its holonomy over topologically nontrivial loops.

9. Dirac Monopoles and Chern Classes

No magnetic monopole has been observed. Dirac showed in 1931 that if a single monopole exists anywhere in the universe electric charge must be quantised in discrete units: every particle's charge is an integer multiple of a fundamental unit.

The argument uses the same two-patch construction that classifies line bundles over spheres, known as the Wu–Yang construction.

Definition (Dirac monopole setup).

A magnetic monopole of strength at the origin creates a radial magnetic field

satisfying . On any sphere of radius surrounding the origin,

The problem: we need to write for a potential A defined on the sphere, but Stokes' theorem would then give (a sphere has no boundary), contradicting the nonzero flux. A globally defined A on cannot exist when the flux is nonzero.

Theorem (Wu–Yang construction and Dirac quantisation).

Split into two overlapping patches: a northern hemisphere (including the north pole) and a southern hemisphere (including the south pole). The patches overlap on an equatorial strip. Define gauge potentials on and on that correctly reproduce the monopole field on each patch. On the equatorial overlap they must differ by a gauge transformation:

For the wavefunction of an electron (charge ) to be single-valued as it traverses the equator once, the gauge transformation parameter must satisfy

Combined with , this gives the Dirac quantisation condition:

If any magnetic charge exists, electric charge must be an integer multiple of . This explains charge quantisation.

Definition (First Chern class).

The two-patch construction defines a principal -bundle over , classified by the transition function on the equatorial overlap. Since the equator is a circle , the transition function is a map classified by its winding number .

This integer is the first Chern class of the bundle. In terms of the curvature:

For the trivial bundle (no monopole), . For the Hopf bundle of Section 5, . The bundle of a Dirac monopole with Dirac charge has .

Remark (Topological quantisation).

The integrality of is a theorem. For any principal -bundle over any compact 2-manifold the integral of the curvature form over the manifold is an integer multiple of . This is a special case of the Chern-Weil theorem, which represents the characteristic classes of a bundle (topological invariants) by polynomials in its curvature form.

Charges are quantised because the topology of gauge bundles is quantised. The integer that counts the winding of the transition function also counts the units of magnetic charge.

Remark (Chern-Simons construction of the secondary class).

The Chern-Weil integrand is closed on the base manifold and therefore locally exact. The differential primitive is the Chern-Simons 3-form

introduced by Chern and Simons in 1974 [7]. On a closed 3-manifold the integral is a secondary characteristic class of the pair , distinct from the primary Chern classes (which vanish in odd dimension). For the abelian Hopf bundle on the standard connection gives . This matches for the bundle. My cartan-geo::chern_simons module implements this calculation (abelian plus non-abelian ) via tensor-product Gauss-Legendre quadrature and recovers the Hopf invariant to numerical tolerance.

10. Orbitals and Computational Electrostatics

The four languages developed in the preceding sections describe the electromagnetic field given its sources. In classical physics those sources are point charges and continuous charge distributions, handled via Dirac deltas and smooth densities as in Section 1. In quantum mechanics even a single particle has an extended charge distribution: the source term in Maxwell's equations is determined by the particle's quantum state. This section develops the semiclassical coupling, in which the charge distribution is quantum-mechanical and the electromagnetic field is classical.

Definition (Quantum charge and current density).

For an electron in a normalised quantum state , the semiclassical charge density and current density sourcing Maxwell's equations are

where is the electron charge, is the electron mass, and is the magnetic vector potential. The two terms in are the paramagnetic current (driven by the phase gradient of ) and the diamagnetic current (driven by the vector potential itself). The continuity equation is satisfied identically as a consequence of the time-dependent Schrödinger equation. Quantum mechanics is thus consistent with charge conservation.

Example (Charge densities of hydrogen orbitals).

For hydrogen, the Schrödinger equation with Coulomb potential has normalised eigenstates , where is the Bohr radius. The charge densities of the lowest orbitals are:

Both are smooth, extended functions. The 1s density is spherically symmetric with characteristic scale . The density has the dumbbell profile of the orbital's angular part. From far away (), both densities integrate to over all space and the electron looks like a point charge at the nucleus. This is consistent with Coulomb's law.

Proposition (Hartree potential from the Poisson equation).

The electrostatic potential sourced by an electron in state satisfies Gauss's law in the form of Poisson's equation:

The solution is the Hartree potential,

This is Coulomb's law with the point charge replaced by the quantum charge density. For the 1s orbital, the integral evaluates in closed form via the expansion of in Legendre polynomials and gives a potential that behaves as for and interpolates smoothly to a finite value at the nucleus.

Derivation. The charge density is the source in Gauss's law [eq:gauss-e]. In the electrostatic case . Hence . Thus . The integral formula is the Green's function solution of Poisson's equation on with source .

Self-consistent fields. In a many-electron atom or molecule with electrons occupying orbitals , the total charge density is and the Hartree potential sourced by all electrons is

Each orbital satisfies a Schrödinger equation with this potential:

The orbitals determine . The density sources through Poisson's equation and in turn enters the Schrödinger equation for the orbitals. This circularity is the self-consistent field (SCF) problem: the solution must be found iteratively. Adding exchange corrections (the antisymmetry of the many-electron wavefunction) gives the Hartree–Fock equations. Replacing the many-body wavefunction with the electron density as the primary variable gives density functional theory (DFT). DFT is exact in principle by the Hohenberg–Kohn theorem and tractable in practice through approximate exchange-correlation functionals.

Minimal coupling and the vector potential. So far only the scalar potential has appeared. The full coupling between a particle of charge (for the electron ) and the electromagnetic field is captured by minimal coupling: replace the canonical momentum with the kinetic momentum everywhere in the Hamiltonian:

Expanding and using the Coulomb gauge (so that and commute):

The term is the paramagnetic coupling, responsible for orbital magnetic moments and NMR chemical shifts. The term is the diamagnetic coupling. The Aharonov–Bohm phase of Section 8 is the unitary transformation implementing minimal coupling for a wavepacket travelling along : substituting into the free Schrödinger equation produces the minimally coupled Hamiltonian above. Gauge invariance of the Schrödinger equation under together with matches the gauge transformation of Section 2.

Computational pipeline. Solving the coupled Schrödinger–Poisson system for a molecule or solid proceeds iteratively:

  1. Initialise trial orbitals (typically superposed atomic orbitals).
  2. Build the electron density .
  3. Solve Poisson's equation for the Hartree potential. In plane-wave codes this is a pointwise division in Fourier space: . Thus . In real-space codes it is solved by multigrid or finite element methods on a numerical grid.
  4. Build the Kohn–Sham or Hartree–Fock Hamiltonian with updated .
  5. Diagonalise: solve .
  6. Check convergence: if , accept; otherwise return to step 2.

Step 3 is the point where Maxwell's equations enter the quantum calculation: the charge density generated by quantum mechanics is fed into the classical electrostatic equation, and the resulting potential modifies the quantum dynamics. The loop between step 2 and step 5 is a discrete version of the self-consistent field condition, The iteration converges when the density mixing map is a contraction (Banach fixed-point theorem). Damped mixing is used to make it one, and in general convergence is not guaranteed.

Discrete exterior calculus and cochain complexes. The computational methods above (finite element grids, plane-wave Fourier solves) are instances of a common algebraic framework: discrete exterior calculus (DEC). The central idea is to replace the smooth de Rham complex of Section 3 with an exact combinatorial analogue defined on a mesh.

Definition (Cochain complex on a simplicial mesh).

Let be a simplicial complex (a mesh of nodes, edges, triangles, and tetrahedra) triangulating a domain in . A -cochain is a linear map from the -cells of to : it assigns a real number to each edge (), each face (), each volume (), or each node (). The space of -cochains is .

The coboundary operator is defined by duality with the boundary operator :

for any -cell , where is its oriented boundary (a signed sum of -cells). Since (the boundary of a boundary is empty), follows. This gives the discrete cochain complex:

In coordinates, is the node-to-edge incidence matrix, is the edge-to-face incidence matrix, and is the face-to-volume incidence matrix. These are sparse integer matrices that encode the combinatorial topology of the mesh.

The de Rham complex of Section 3 is the continuum limit of this construction: as the mesh is refined, smooth -forms integrate over -cells to give cochains, and the coboundary operator converges to the exterior derivative. Whitney forms (piecewise linear -forms associated to each -simplex) are the interpolating basis functions that transfer this correspondence to the finite element setting. Their span over a simplicial complex forms a subcomplex of the de Rham complex, and this subcomplex is the foundation of finite element exterior calculus (FEEC).

Definition (Discrete Hodge star and the primal–dual mesh).

The Hodge star of Section 3 maps -forms to -forms using the metric. Discretely, it maps primal -cochains to dual -cochains, where the dual mesh is constructed by placing a dual node at the circumcentre of each primal simplex and connecting dual nodes across shared faces.

The discrete Hodge star is a diagonal matrix whose entries are ratios of dual and primal volumes:

where is the volume of a primal -cell and is the volume of its dual -cell. The discrete Laplacian on -cochains is then

with . is a sparse positive semi-definite matrix, self-adjoint in the -weighted inner product (so is symmetric). On 0-cochains, is the discrete scalar Laplacian : the equation , with the charge in each dual cell, is the cochain form of Poisson's equation from step 3 of the SCF pipeline.

Theorem (Maxwell's equations as cochain equations).

Place the electromagnetic degrees of freedom on the primal and dual meshes as follows. The scalar potential lives on primal 0-cochains; the vector potential on primal 1-cochains; the magnetic field on primal 2-cochains; and the electric field on primal 1-cochains. With this placement, [eq:maxwell-forms-eq] from Section 3 become exact algebraic identities:

Both hold at the level of integer arithmetic on the mesh because . The source equations (Gauss and Ampère–Maxwell) become

The sign appears because the adjoint of the discrete gradient is minus the divergence.

These are the DEC discretisation of from Section 3.

The Yee finite-difference time-domain (FDTD) grid, the standard computational tool for solving Maxwell's equations in 3D, is a special case of this structure: it is the DEC complex on a Cartesian hexahedral mesh, where E-field components sit on primal edges, B-field components sit on primal faces, and the staggering in space and time ensures that is satisfied exactly at every grid point and every time step.

Three structural properties follow from the cochain complex that are difficult to achieve by ad hoc discretisations. First, is preserved exactly by the time-stepping scheme because it is an algebraic consequence of . Second, the discrete gauge freedom of Section 2 is preserved: replacing leaves unchanged because . Third, the discrete Hodge decomposition gives an exact splitting of any 1-cochain into a gradient part, a curl part and a harmonic part, which permits gauge-fixed solvers (discrete Lorenz gauge or Coulomb gauge) that are consistent by construction.

For the SCF pipeline DEC replaces the ad hoc Poisson solve in step 3 with an algebraic system: the charge density is a dual 3-cochain (the charge in the dual cell around each node), the potential is a 0-cochain (a number per node), and the discrete Poisson operator is the sparse matrix assembled once from the mesh geometry. Solving by conjugate gradient (exploiting the symmetry and sparsity of this matrix) is the core linear algebra step shared by real-space DFT codes such as GPAW and Octopus.

11. Synthesis

We have presented the same physical theory in four languages.

LanguageEquationReveals
Vector calculus etc.Forces on charges, wave propagation, practical computation
Differential formsCoordinate freedom, relativistic covariance, Stokes' theorem, charge conservation
Spacetime algebraSingle compact equation, spinor structure, algebraic unification with gravity
Principal bundlesGauge invariance as geometry, topological invariants, non-abelian extension

The fibre bundle picture is the most fundamental. The electric and magnetic fields E and B are components of the curvature 2-form of a connection on a principal bundle. The connection itself depends on a choice of local section (gauge choice), and only gauge-invariant quantities (such as the curvature and holonomies along loops) are physical. The primary object is the bundle together with the gauge-equivalence class of connections.

Three consequences follow immediately from this geometric viewpoint:

Gauge invariance as geometry. The freedom to choose a gauge is the freedom to choose a local section of the bundle. Changing the section transforms A by . The curvature transforms by and is invariant in the abelian case. This is the geometric origin of gauge invariance. In vector calculus it appears as an arbitrary-seeming symmetry of the potentials.

Aharonov–Bohm effect as holonomy. The phase acquired by a charged particle travelling around a loop is the holonomy of the connection around that loop. It can differ from 1 even when the curvature is zero ( on the path) because the base space with the solenoid excluded is not simply connected. The vector potential A contains the holonomy information that F does not.

Charge quantisation as topology. The first Chern class of a bundle is an integer. If magnetic monopoles exist, the Dirac quantisation condition forces electric charge to be an integer multiple of a fundamental unit. The integrality follows because bundles are classified by integers.

The same framework, extended to non-abelian gauge groups and , gives the full Standard Model of particle physics. The electroweak force is a Yang–Mills theory for coupled to a Higgs field, and the strong force is one for . Both share the principal bundle geometry developed here. The unsolved mass gap problem for Yang–Mills theory with compact simple gauge group is a question about the quantum theory of connections: when the classical theory is quantised, do its energy excitations have a positive lower bound? The classical geometry is understood. What remains is the quantum.

References

  1. Y. Aharonov and D. Bohm, "Significance of electromagnetic potentials in the quantum theory," Phys. Rev. 115 (1959), 485–491. DOI
  2. A. Tonomura et al., "Evidence for Aharonov–Bohm effect with magnetic field completely shielded from electron wave," Phys. Rev. Lett. 56 (1986), 792–795. DOI
  3. A. Jaffe and E. Witten, "Quantum Yang–Mills Theory," in The Millennium Prize Problems, Clay Mathematics Institute, 2006. PDF
  4. P. A. M. Dirac, "Quantised singularities in the electromagnetic field," Proc. R. Soc. London A 133 (1931), 60–72. DOI
  5. T. T. Wu and C. N. Yang, "Concept of nonintegrable phase factors and global formulation of gauge fields," Phys. Rev. D 12 (1975), 3845–3857. DOI
  6. J. C. Maxwell, "A dynamical theory of the electromagnetic field," Phil. Trans. R. Soc. London 155 (1865), 459–512. DOI
  7. S.-S. Chern and J. Simons, "Characteristic forms and geometric invariants," Annals of Mathematics 99 (1974), 48–69. DOI