Covariant Form of Maxwell's Equations
Statement
The four Maxwell equations, together with the Lorentz force and charge conservation, are equivalent to two four-tensor equations: the inhomogeneous pair ∂μFμν = μ0Jν, and the homogeneous pair ∂μ(⋆F)μν = 0 (equivalently ∂[αFβγ] = 0). Here Fμν is the electromagnetic field tensor built from E and B, Jν = (cρ, J) is the four-current, and ∂μ = (∂ct, ∇). Both equations transform as tensors under the Lorentz group, so their validity in one inertial frame guarantees validity in every inertial frame.
Why it matters
Maxwell's equations were the first field theory whose form already secretly respected special relativity, years before Einstein made the symmetry explicit. Recasting the four vector equations as two tensor equations makes Lorentz invariance manifest: a single equation that is a tensor identity is automatically true in all frames, so the mixing of E and B under boosts is no longer a surprise to be checked but a structural consequence.
The covariant form is also the launch point for the Lagrangian formulation (ℒ = −FμνFμν/4μ0 − JμAμ), for gauge theory, and for the extension to curved spacetime by minimal coupling ∂ → ∇. Nearly all of modern field theory inherits this template.
Assumptions
Derivation
Result
Reading. The inhomogeneous equation ties the field to its sources: its ν = 0 component is Gauss's law and its ν = i components are the Ampère–Maxwell law, with the four-current Jν = (cρ, J) on the right. The homogeneous equation is source-free and follows purely from F being built from a potential: its all-spatial component is ∇·B = 0 and its mixed components are Faraday's law. Eight scalar Maxwell equations collapse into two manifestly Lorentz-covariant tensor statements. Contracting the inhomogeneous equation with ∂ν and using antisymmetry of F gives 0 = μ0∂νJν, so charge conservation is built in, not imposed.
Units check. [∂μFμν]: F0i = E/c has units (V m−1)/(m s−1) = V s m−2 = T; ∂μ ~ m−1, so the left side is T m−1. Right side: μ0Jν with [μ0] = T m A−1 and [J0] = cρ = (m s−1)(A s m−3) = A m−2, giving (T m A−1)(A m−2) = T m−1. Both sides T m−1. ✓
Limiting cases
- Statics (∂t → 0): the time components of ∂μFμν drop out, leaving ∇·E = ρ/ε0 and ∇×B = μ0J — electro- and magnetostatics decouple.
- Vacuum (Jν = 0): both equations are homogeneous; taking ∂ν of the first and using the second gives □Aμ = 0 in Lorenz gauge — light as a free-field solution.
- Non-relativistic source, v ≪ c: Ji = ρvi ≪ J0 = cρ; the current four-vector is dominated by its charge-density component and boosts act nearly as Galilean, recovering pre-relativistic magnetostatics as a small correction.
- Single boost of a static charge: Fμν with only E in the rest frame develops B in the boosted frame via the tensor transformation — the covariant equations reproduce the moving-charge magnetic field automatically.
Breaks when
- Curved spacetime / strong gravity. Partial derivatives are not tensorial on a curved manifold; ∂μFμν = μ0Jν must be promoted to ∇μFμν = μ0Jν with Christoffel connection terms (equivalently ∂μ(√−g Fμν) = μ0√−g Jν). The flat-space form is then only locally valid in a freely-falling frame.
- Media with polarization/magnetization. In matter one must split into free and bound sources; the covariant equation for the free current uses a separate tensor Gμν built from (D, H), and constitutive relations Gμν = f(Fαβ) are needed. The single vacuum equation no longer suffices.
- Quantum / strong-field regime. At field strengths approaching the Schwinger limit Ec ≈ 1.3×1018 V/m, vacuum polarization makes Maxwell's equations nonlinear (Euler–Heisenberg corrections); the linear tensor equation is only the leading term.
- Magnetic monopoles present. If ∇·B ≠ 0, the homogeneous equation gains a magnetic four-current, ∂μ(⋆F)μν = μ0Jmν, and F can no longer be written globally as ∂μAν − ∂νAμ.
Failure modes
- Index-position slip: writing ∂μFμν (both down) instead of ∂μFμν. The contraction requires one up and one down index; forgetting to raise with η loses the relative sign between E and B terms.
- Dropping the factor of c in F0i: using F0i = Ei rather than Ei/c (SI). The displacement-current term then comes out with the wrong dimensional coefficient.
- Assuming the homogeneous equation needs a source constant: setting ∂μ(⋆F)μν = μ0(something). It is identically zero in standard electromagnetism — the Bianchi identity, not a dynamical law.
- Confusing Fμν with (⋆F)μν: swapping E ↔ cB under the dual. This turns Gauss's law into no-monopoles and silently exchanges the two equations.
- Signature error: mixing (+,−,−,−) definitions of F with (−,+,+,+) index raising, producing spurious sign flips in Faraday's law.
- Treating Jν as independent of conservation: imposing a charge/current pair that violates ∂νJν = 0, which makes ∂μFμν = μ0Jν internally inconsistent.
Discussion
The deepest lesson is that Lorentz invariance is not an extra property Maxwell's equations happen to have — it is the reason they take the form they do. Once you accept that the sources form a four-vector Jν and the fields an antisymmetric tensor Fμν, the only first-order, gauge-invariant, Lorentz-covariant equation linking them is ∂μFμν ∝ Jν. Electromagnetism is, in a real sense, the simplest relativistic field theory of a spin-1 field. The apparent asymmetry between the E-source (charge) and the absent B-source (monopole) is exactly the asymmetry between the inhomogeneous and homogeneous tensor equations.
The homogeneous equation deserves special emphasis: it carries no physical constants and no source because it is a mathematical identity following from F = dA. In the language of differential forms, F = dA and the homogeneous pair is dF = d(dA) = 0, while the inhomogeneous pair is d⋆F = μ0⋆J. This compresses all of electromagnetism into two lines and makes the topological content — that monopoles would obstruct the global existence of A — visible.
Charge conservation emerges as a consistency condition rather than a separate postulate. Because Fμν is antisymmetric and partial derivatives commute, ∂ν∂μFμν = 0 identically; applying ∂ν to the field equation therefore forces ∂νJν = 0. In Noether's framework this same conservation law is the consequence of the U(1) gauge symmetry of the action, and gauging that symmetry is precisely what generates the coupling JμAμ. The covariant Maxwell equations are thus the prototype every Yang–Mills theory generalizes: replace the abelian Fμν with a non-abelian field strength and ∂ with a gauge-covariant D, and the same two-equation structure describes the strong and electroweak interactions.
Common misconceptions. Students often think the covariant form contains more physics than the vector Maxwell equations — it does not; it is exactly the same content, reorganized so that Lorentz symmetry is manifest. Another frequent error is believing that "covariant" means the equations are unchanged numerically between frames; rather, each side transforms as a tensor, so the equation is preserved even though the individual components of F and J change. Finally, the homogeneous equation is not "Maxwell's other two equations dressed up" in the sense of new dynamics — it is a kinematic identity that holds the moment a vector potential exists.
Worked examples
Reading. No choice of consistent electromagnetic field can source a current that fails to conserve charge. The antisymmetry of Fμν alone guarantees it.
Units check. [∂tρ] = (A s m−3)/s = A m−3; [∇·J] = (A m−2)/m = A m−3. ✓
Reading. A purely electric field in S becomes an electric-plus-magnetic field in S′. The magnetic field is exactly what an observer in S′ attributes to the now-moving line charge (a current I′ = λ′v), and it is generated for free by the tensor transformation of Fμν — no separate magnetostatics calculation needed.
Units check. [B′] = [v/c2][E] = (m s−1)/(m2 s−2) · (V m−1) = (s m−1)(V m−1) = V s m−2 = T. ✓
Problems
- Starting from ∂μFμν = μ0Jν, write out the ν = 2 (y-component) equation explicitly in terms of E and B components and identify which Maxwell law it represents.
Solution
For ν = 2: ∂μFμ2 = ∂0F02 + ∂1F12 + ∂3F32. Using F02 = Ey/c, F12 = −Bz, F32 = −Bx (from Fij = −εijkBk): = (1/c2)∂tEy − ∂xBz + ∂zBx = (1/c2)∂tEy − (∇×B)y. Setting equal to μ0Jy and rearranging: (∇×B)y = μ0Jy + μ0ε0∂tEy — the y-component of the Ampère–Maxwell law. - Show explicitly that ∂νJν = 0 follows from ∂μFμν = μ0Jν, and state which property of Fμν is essential.
Solution
Apply ∂ν to both sides: ∂ν∂μFμν = μ0∂νJν. The operator ∂ν∂μ is symmetric under μ ↔ ν (partial derivatives commute), while Fμν = −Fνμ is antisymmetric. The full contraction of a symmetric object with an antisymmetric one vanishes identically: ∂ν∂μFμν = 0. Hence μ0∂νJν = 0, i.e. ∂tρ + ∇·J = 0. The essential property is the antisymmetry of the field tensor. - Verify that the number of independent scalar equations in the two covariant equations (4 + 4 = 8, before accounting for the built-in conservation identity) matches the count in the four vector Maxwell equations.
Solution
Vector form: ∇·E (1) + ∇·B (1) + ∇×E = −∂tB (3) + ∇×B = μ0J + ... (3) = 8 scalar equations. Covariant form: ∂μFμν = μ0Jν has ν = 0,1,2,3 → 4 equations (Gauss + 3 Ampère–Maxwell). ∂[αFβγ] = 0 has 4 independent index-triples (choosing 3 of 4 indices) → 4 equations (no-monopole + 3 Faraday). Total 8, matching. Note the inhomogeneous set has one redundancy via ∂νJν = 0, consistent with the vector form's implied continuity. - A proton beam carries current density J = 5.0×104 A/m² and charge density ρ = 1.0×10−4 C/m³ along x. Compute the four-current Jν and its Lorentz invariant JνJν. Interpret the sign.
Solution
Jν = (cρ, Jx, 0, 0) = (3.0×108 · 1.0×10−4, 5.0×104, 0, 0) = (3.0×104, 5.0×104, 0, 0) A/m². Invariant with η = (+,−,−,−): JνJν = (cρ)2 − |J|2 = (3.0×104)2 − (5.0×104)2 = 9.0×108 − 25×108 = −1.6×109 (A/m²)². The invariant is negative, meaning the four-current is spacelike here — physically the drift current dominates cρ, which happens when carriers move relativistically or the net charge density is small compared to the current. (For a single-species non-relativistic beam one expects cρ > J, i.e. timelike; a spacelike result signals a nearly neutral current-carrying medium.) - In the Lorenz gauge ∂μAμ = 0, show that ∂μFμν = μ0Jν reduces to a wave equation for Aν, and give the vacuum dispersion relation.
Solution
Insert Fμν = ∂μAν − ∂νAμ: ∂μFμν = ∂μ∂μAν − ∂ν(∂μAμ). In Lorenz gauge the second term vanishes, leaving □Aν = μ0Jν, where □ = ∂μ∂μ = (1/c2)∂t2 − ∇2. In vacuum (Jν = 0): □Aν = 0. Plane-wave ansatz Aν ∝ exp(−ikμxμ) gives kμkμ = 0, i.e. ω2/c2 = |k|2, or ω = c|k|. Electromagnetic waves propagate at c with a massless (lightlike) four-wavevector — the quantum photon has zero rest mass.