Visual walkthrough — Noether's theorem — symmetry ↔ conservation law

---

Step 0 — The picture we are trying to explain

Before any algebra, let us agree on what the words even mean, using a picture.

WHAT. A particle moves. It traces a curve. We ask: is there a number that stays frozen along that curve?

WHY. Physics is full of "conserved quantities" — energy, momentum, spin. We want to know where they come from. The claim is: they come from symmetry, which is just "you can nudge the setup a little and nothing important changes."

PICTURE. On the left, a bead slides on a smooth horizontal rail. Slide the whole rail sideways — the physics is identical (the bead doesn't know where the rail is, only its shape). That "doesn't matter where" is a symmetry. The frozen number it hides is the bead's momentum.

We need one more idea before Step 1: the machine that decides which paths are physical.

---

Step 1 — The scorecard for paths: the action

WHAT. Nature does not try every path. It picks the one path that makes a certain total score, the action, as flat-as-possible (a minimum or a saddle). We import this from Lagrangian Mechanics.

WHY this tool and not "$F=ma$"? Because a scorecard for whole paths is exactly the setting where "nudge the path and the score doesn't change" makes sense. Symmetry is a statement about nudges; the action is the thing nudges act on. Forces alone give us no natural place to talk about "nudge the coordinate."

PICTURE. Every candidate path gets a number $S$. Plot $S$ against "which path." The true path sits at the bottom of the valley — where the ground is flat. Flatness is the whole game.

The "bottom of the valley" condition, written in symbols, is the Euler-Lagrange Equation — that is Step 2.

---

Step 2 — The physical-path law (Euler–Lagrange)

WHAT. "The score is flat at the true path" turns into one equation the path must obey.

WHY. We will need this equation in the derivation to convert an ugly term into a tidy time-derivative. Think of it as the tool that says "on a real path, force = rate of change of momentum," but written in the action's language.

PICTURE. Take the true path (solid) and a slightly wiggled neighbour (dashed) that agrees at both endpoints. Flatness means the score of the wiggle equals the score of the true path to first order — the two scores differ by nothing linear in the wiggle.

---

Step 3 — What a symmetry is, precisely

WHAT. Nudge every coordinate by a tiny, shape-controlled amount: $q_i\to q_i^{\prime }=q_i+ϵ\delta q_i.$

WHY this form? A continuous symmetry needs a continuous dial $ϵ$. We split the nudge into "how much" ($ϵ$, a small number) times "which direction/shape" ($\delta q_i$, a pattern that can depend on the $q$'s). This separation is what lets us later take a derivative in $ϵ$.

PICTURE. At each point of the path, an arrow $\delta q_i$ points "where this nudge would push me." A translation gives all-parallel arrows; a rotation gives arrows curling around a centre. Same path, two different nudge-fields.

---

Step 4 — Change the score under the nudge (chain rule)

WHAT. Compute how $L$ changes, $\delta L$, when we apply the nudge.

WHY the chain rule and nothing fancier? $L$ depends on $q_i$ and ${\dot{q}}_i$. If both wiggle a little, the only honest first-order accounting of the change is "how sensitive $L$ is to each input, times how much that input moved." That is the chain rule.

PICTURE. $L$ is a hill over the $(q,\dot{q})$ plane. Moving by $(\delta q,\delta \dot{q})$ changes height by (slope-in-$q$)×(step-in-$q$) + (slope-in-$\dot{q}$)×(step-in-$\dot{q}$).

---

Step 5 — Trade the position-slope for a momentum-rate (use EL)

WHAT. Replace the awkward first term $\frac{\partial L}{\partial q_i}$ using Step 2's law.

WHY. We are hunting for a single time-derivative $\frac{d}{dt}(\text{something})$, because that is what produces a conserved quantity. The term $\partial L/\partial q_i$ blocks us; Euler–Lagrange lets us swap it for $\frac{d}{dt}{p}_i$, which fits the pattern.

PICTURE. A little swap-card: the "force" slot $\partial L/\partial q_i$ is exchanged for the "rate of momentum" card $\frac{d}{dt}{p}_i$ — legal only on a physical path.

---

Step 6 — Collapse into one time-derivative → the conserved charge

WHAT. Recognize the product rule and fold the two terms into a single $\frac{d}{dt}(\cdot )$.

WHY. The identity $\frac{d}{dt}(uv)=\dot{u}v+u\dot{v}$ turns our two-term mess into one derivative. If the whole thing equals $\delta L=0$ (a symmetry!), then "derivative $=0$" means "the inside is constant." That constant is the prize.

PICTURE. Two puzzle pieces $\dot{u}v$ and $u\dot{v}$ snap together into the single block $\frac{d}{dt}(uv)$. Under it, a flat line labelled "constant."

---

Step 7 — The boundary case: when $L$ does change

WHAT. Sometimes the score isn't perfectly invariant: instead $\delta L=ϵ\frac{dF}{dt}$ for some function $F$. The action is still invariant, because $\int \frac{dF}{dt}dt$ only reads off endpoint values, and the nudge is defined to vanish at the endpoints.

WHY this is not cheating. A total time-derivative integrates to "value at $t_2$ minus value at $t_1$" — pure boundary. Since we hold endpoints fixed while nudging, it contributes nothing to the variation of the action.

PICTURE. A ramp of "extra score" that only depends on the endpoints; the interior of the path is untouched, so it cannot affect which path is chosen.

---

Worked check — free particle, rotation ↔ angular momentum

---

The one-picture summary

Everything above is one arrow diagram: a nudge that leaves the score flat forces a time-derivative to vanish, and the thing inside that derivative is the conserved charge.

Recall Feynman retelling — the whole walkthrough in plain words

Imagine grading every possible way a marble could roll, giving each way a single score. Nature always picks the flattest-scoring way. Now gently shove the whole setup — slide it, spin it, wait a while. If the score doesn't budge, you've found a "doesn't-matter." I then wrote down how much the score changes when I shove (that's the chain rule), swapped one clumsy piece using the rule that real paths obey (Euler–Lagrange), and suddenly the two leftover pieces snapped together into "the rate of change of one single thing." If shoving doesn't change the score, that rate of change is zero — so that one single thing never moves. Sliding freezes momentum; spinning freezes angular momentum; waiting freezes energy (as long as the rules themselves don't tick with the clock). Every "doesn't-matter" hides a number that refuses to change.

Recall

A cyclic coordinate is one that $L$ ... ::: does not depend on; then its conjugate momentum is conserved. The Noether charge formula (no boundary term) is ::: $Q={\sum }_i{p}_i\delta q_i=\text{const}$. With a boundary term $\delta L=dF/dt$, the charge becomes ::: $Q={\sum }_i{p}_i\delta q_i-F$. Time-translation shape $\delta q_i$ equals ::: ${\dot{q}}_i$, giving the energy function $h={\sum }_i{p}_i{\dot{q}}_i-L$. Why do discrete symmetries give no Noether charge? ::: There is no continuous parameter $ϵ$ to differentiate with respect to.