GAUGE FREEDOM AND POTENTIALS

Back in §03 we turned the magnetic field into a curl: B = ∇×A. One equation, one definition, one clean replacement of three numbers per point with three others. And we noticed, almost in passing, that this replacement was not unique. Add the gradient of any scalar function to A and the curl is unchanged, because the curl of a gradient is zero. We filed that under "a quirk of magnetostatics" and moved on.

Now the magnetic world has a time-varying partner. Faraday's law (§04.1) says a changing flux drives an EMF; in potential language that means E isn't just −∇V anymore, it has to pick up a term from ∂A/∂t. The two potentials now dance together. And the quirk — that scalar function you can slide under the rug without anyone noticing — comes along for the ride. Only now it can slide under both potentials at once.

This topic is about that freedom. What it is, why it's there, how to pin it down, and — the preview — why physics in the twentieth century turned out to care about it more than anyone in the nineteenth could have guessed.

In electrodynamics the two fields come from the two potentials by a pair of rules. For B we reuse the old definition. For E we add a time-derivative term:

Read those in plain language. B is the curl of A — how much A swirls around a point. E has two sources: the steepness of the scalar V (how fast V falls as you walk forward), and the rate at which A is changing with time (how fast the vector field itself is reshuffling). Together, (V, A) carry four numbers per point and describe the whole electromagnetic field.

Now pick any smooth scalar function f(x, y, z, t) — literally any. Define a new pair of potentials by

The symbol ∇f means "the vector whose components are the rates at which f changes in each direction" — how steep the terrain f is. ∂f/∂t is how fast f is changing at a fixed point as time passes. Plug the primed pair back into EQ.01 and watch what happens. B′ = ∇×A′ = ∇×A + ∇×(∇f) = B + 0, because the curl of a gradient is identically zero. And E′ = −∇V′ − ∂A′/∂t = −∇V + ∇(∂f/∂t) − ∂A/∂t − ∂(∇f)/∂t = E. The two correction terms cancel exactly — they're the same mixed partial derivative, written in opposite signs.

So E and B — the fields that actually push on charges — are blind to the choice of f. The potentials shift; the fields don't. Every measurement anyone ever made with a voltmeter, a compass, a Hall probe, or an oscilloscope is stubbornly insensitive to which f you picked.

Underdetermined theories aren't usually something to celebrate. But the kind of underdetermination the potentials have is the friendly kind: it's a freedom, not an ambiguity. You get to use it.

A concrete example. The potentials of an arbitrary time-varying charge-and-current distribution normally satisfy a pair of messy coupled wave equations — wave equations with cross-terms linking V and A, so you can't solve one without the other. Apply a clever gauge transformation, chosen so that the cross-terms vanish, and suddenly V and A each solve a plain wave equation independently. We went from "two tangled partial differential equations" to "two identical decoupled ones". No physics changed — we still describe the exact same field — but the bookkeeping collapsed from hard to easy.

This is how gauge freedom is used in practice across every corner of electrodynamics, quantum mechanics, and field theory: as a knob you turn to simplify whatever equation is in front of you.

The top panel is V(x) — a line that tilts and slides as you change the sliders. The bottom panel is A(x) — a row of arrows that grow, flip, shrink. The number in the HUD is the physical field E. It never moves. That's gauge invariance with your own hands on it: the potentials are a house of cards; the field is the floor underneath them.

Two specific choices of gauge show up everywhere. The first — and the one physicists quietly default to when special relativity is in the room — is the Lorenz gauge:

Read ∇·A as "how much A is flowing outward from a point" — the divergence, the same operator as in Gauss's law. The whole condition says: whatever net outward flow the vector potential has at a point, it must be precisely cancelled by the rate at which V is changing there, divided by c². A balance equation between the two potentials, tying them together.

Why this particular balance? Because under it the messy coupled wave equations for V and A fall apart into two decoupled wave equations with the same d'Alembertian operator — the relativistically natural one — on each side. Sources (ρ, J) in; potentials (V, A) out; one equation each; light speed baked in.

The gauge is named after the Danish physicist Ludvig Lorenz, not after Hendrik Antoon Lorentz of the Lorentz force and Lorentz transformation. Their names differ by one consonant and they lived at the same time, which has spawned more than a century of textbook typos. Ludvig got there first by a couple of decades; the "z" variant is the one that belongs on this gauge.

The second standard choice is the Coulomb gauge:

Flat. Simple. A is divergence-free everywhere, full stop — it loops but it doesn't spread. Under this constraint V drops the time-derivative partner in its equation and obeys the plain Poisson equation of electrostatics, as Poisson wrote it in 1812:

So if you know where the charge is right now, you know V right now — no waiting, no retardation, no wave equation. That looks as if the theory is faster than light. It isn't: the catch is that A in this gauge picks up the difference and propagates the causal, delayed part of the field. The measurable combination −∇V − ∂A/∂t still travels at c. Coulomb gauge is the toolkit of choice for bound-state problems, atomic physics, and anywhere the sources are essentially stationary — because half the equation becomes trivial.

Here's the test. Pick a pair of potentials in the Coulomb gauge. Compute the force on a charge at some point: F = q(E + v×B). Write down the answer. Now pick a different pair — same physical field, shifted by ∇f and ∂f/∂t into, say, the Lorenz gauge. Recompute F. Identical.

The left panel carries a heavy scalar potential (Coulomb-style: V does all the work). The right panel has a lighter V and a set of amber A-arrows (Lorenz-style: the vector potential carries part of the load). The white E-field arrows — the thing a charge would actually feel — are in exactly the same places, pointing the same way, with the same lengths. What differs is the accounting; what stays is the physics. Any local, classical measurement — a force on a test charge, a current in a wire, a voltage between two terminals — is a gauge-invariant quantity and cannot tell the two panels apart.

If gauge freedom is invisible, why is it one of the deepest ideas in modern physics? Because in the twentieth century it turned out that gauge invariance isn't a peculiarity of electromagnetism — it's the template the universe uses to generate fundamental forces. Demand that a quantum field theory be invariant under a local phase change (a position-dependent gauge transformation) and a force-carrying boson must appear to keep the bookkeeping consistent. Demand it for the electron's wavefunction and the photon drops out of the maths. Demand it for broader symmetry groups and you get the W and Z bosons (weak force) and the gluons (strong force).

Every fundamental force in the Standard Model is a gauge theory. We'll take this apart in §12.1; the thing to carry away now is that the quiet "the potentials are redundant by ∇f" of this topic is the small end of a wedge whose large end is the entire modern picture of fundamental interactions.