A Notion To Ponder

Landau's version (§18). $A_i \to A_i + \partial_i f$ leaves $F_{ik}$, and all physics, unchanged.

Two geometric readings (both used; distinguish them!). 1. Passive: a change of Local Trivialization / local section $s \to s\cdot g$, $g: U \to G$. Nothing moves; the description changes. This is Landau's version: $A \to g^{-1}Ag + g^{-1}dg$, for $U(1)$ with $g = e^{i\theta}$: $A \to A + i\,d\theta$. 2. Active: a bundle automorphism $\Phi: P \to P$ covering $\mathrm{id}_M$ and commuting with the $G$-action. The group of these is the gauge group $\mathscr{G}$ (infinite-dimensional).

The conceptual upgrade. In Landau, gauge freedom looks like a defect (non-uniqueness of $A$). Geometrically it is inevitable: on a nontrivial bundle no global trivialization exists, so descriptions are patchwise and must be glued by gauge transformations — same logic as coordinate charts on a curved manifold. "Gauge symmetry" is not a symmetry of nature but a redundancy of description; what's real is the connection itself and its gauge-invariants (holonomies, $F$, Chern numbers).

Matter co-transforms. A charge-$n$ section transforms $\psi \to e^{in\theta}\psi$ — and the Covariant Derivative is engineered so $\nabla\psi$ co-transforms too. That's the whole design.

Related: Local Trivialization · Connection 1-Form · Principal Bundle · Minimal Coupling · Module 09 - Electromagnetism as a U(1) Gauge Theory

Sources: Landau §18; Woit (gauge chapters).