Stokes' theorem — statement, curl-circulation connection: Visual walkthrough

We will need only two ideas from before you started reading, and I will rebuild both:

A vector field — an arrow attached to every point of the plane/space (think: water-current arrows).
A tiny bit of movement — how much a flow pushes you along as you take one small step.

Everything else is earned below.

Step 1 — What "flow along a path" even means

WHAT. Picture a river seen from above. At each point the water moves in some direction with some speed — that is the arrow $F$ at that point.

WHY. Before we can talk about "swirl" or "circulation," we must agree on how much the flow helps you as you walk. If you step a tiny amount $d r$ (a short arrow showing your step), the helpful part of the flow is the projection of $F$ onto your step: $F \cdot d r = flow’s x P step’s x d x + flow’s y Q step’s y d y .$ The dot $\cdot$ multiplies matching components and adds them. If the flow points with your step, this is positive (a push); against, negative (a drag); sideways, zero.

PICTURE. The blue arrows are the flow. The orange step arrow $d r$ shows the direction you walk; the green shadow is the part of the flow that actually helps you.

Figure — Stokes' theorem — statement, curl-circulation connection

Step 2 — Circulation around one tiny rectangle

WHAT. Take the smallest loop we can reason about exactly: a little rectangle with corners at $(x_{0}, y_{0})$ , width $Δ x$ , height $Δ y$ . Walk it counter-clockwise: right along the bottom, up the right side, left along the top, down the left side.

WHY. A curved boundary is hard, but a rectangle has four straight sides where the flow barely changes. If we can compute circulation for one tiny rectangle, we can later tile any surface out of them.

PICTURE. Follow the four orange arrows in order. Notice the bottom and top are travelled in opposite $x$ -directions, and likewise the two vertical sides.

The lap splits into four straight pieces: $\oint F \cdot d r = y = y_{0}, \to \int P d x_{bottom} + x = x_{0} + Δ x, ↑ \int Q d y_{right} + y = y_{0} + Δ y, \leftarrow \int P d x_{top} + x = x_{0}, ↓ \int Q d y_{left} .$ Each label says: which side, which coordinate is fixed, and which way the arrow points.

Step 3 — Pairing opposite sides makes derivatives appear

WHAT. Pair the horizontal sides (bottom and top) and the vertical sides (right and left).

WHY. The top is walked leftward, so its contribution is minus the rightward version. Bottom and top run over the same $x$ -range but at different heights $y_{0}$ and $y_{0} + Δ y$ . Their difference is exactly "how much $P$ changed when we raised $y$ " — and rate of change is what a partial derivative measures.

Horizontal pair (bottom minus top, because top runs backward): $\int P d x_{y_{0}} + \int P d x_{y_{0} + Δ y} \approx - how P changes with height \frac{\partial P}{\partial y} height Δ y width Δ x .$

Vertical pair (right minus left): $\int Q d y_{x_{0} + Δ x} + \int Q d y_{x_{0}} \approx + how Q changes across \frac{\partial Q}{\partial x} width Δ x height Δ y .$

PICTURE. The color-coded arrows show the two pairs; the small $Δ$ -labels sit right on the sides they measure.

Adding the two pairs: $\oint_{tiny rect} F \cdot d r \approx (one number: the swirl \frac{\partial Q}{\partial x} - \frac{\partial P}{\partial y}) area Δ A Δ x Δ y .$

Step 4 — Naming the swirl: the curl

WHAT. We just proved: (tiny circulation) $\approx$ (curl) $\times$ (area).

WHY. This is the whole engine. It converts a loop quantity (circulation) into an area quantity (curl × area). Loops we can chain; areas we can add up — that's what Step 6 exploits.

PICTURE. Two little paddle-wheels. Where $\partial Q / \partial x$ and $- \partial P / \partial y$ agree in sign, the wheel spins fast (big curl); where they fight, it barely turns.

Step 5 — Lifting to 3D: the curl becomes a vector

WHAT. In space $F = (P, Q, R)$ and a tiny loop can lie in any plane, tilted by a unit normal $\hat{n}$ (an arrow of length 1 sticking straight out of the loop).

WHY. A loop lying flat catches only the swirl about the vertical axis; a tilted loop catches swirl about its axis. So swirl needs a direction — it is a vector, not a number. The tiny-circulation law becomes $\oint_{tiny loop} F \cdot d r \approx swirl vector (\nabla \times F) \cdot loop’s axis \hat{n} loop’s area Δ A .$ The dot picks out the part of the swirl aligned with the loop's axis — exactly the part that circulation feels.

WHY the right-hand rule? "Counter-clockwise" only means something once you say which side you're looking from. Point your right thumb along $\hat{n}$ ; your curled fingers show the positive lap direction. Flip $\hat{n}$ and the lap direction flips too — this is why the sign of the two sides always stays matched. (See Orientation & the Right-Hand Rule.)

PICTURE. The same physical spin, seen from opposite normals: one $\hat{n}$ gives CCW, the reversed $\hat{n}$ gives CW.

Step 6 — Tiling: interior edges cancel, only the rim survives

WHAT. Cover the surface $S$ with a fine grid of tiny loops, all walked the same way (CCW as seen from $\hat{n}$ ). Add up every tiny circulation.

WHY the magic works. Any inner edge is shared by two neighbouring tiles. One tile walks it left-to-right; the neighbour walks the same edge right-to-left. Same flow, opposite direction → the two contributions are exact negatives and cancel. An edge on the outer boundary $C$ has no neighbour, so nothing cancels it. Sum of tiny loops = one lap around $C$ .

PICTURE. Zoom on two tiles: the shared middle edge has two opposing orange arrows (they annihilate); the outer edges have a single surviving arrow.

$only the rim left \oint_{C} F \cdot d r = \sum_{tiles} \oint_{tile} F \cdot d r \approx \sum_{tiles} (\nabla \times F) \cdot \hat{n} Δ A .$

Step 7 — The limit: sum becomes a surface integral

WHAT. Shrink every tile ( $Δ A \to 0$ ). The " $\approx$ " from the Taylor step becomes exact, and the sum becomes an integral.

WHY the limit is legal. The error in each tile is higher-order (proportional to $Δ A^{3/2}$ or smaller). There are about $1/Δ A$ tiles, so total error $\sim Δ A^{1/2} \to 0$ . The leading terms survive; the errors vanish.

$\oint_{C} F \cdot d r = \iint_{S} (\nabla \times F) \cdot d S$

Read left to right: circulation around the rim = total curl-flux through the skin.

Step 8 — Edge and degenerate cases (never let the reader fall through)

Intuition Every case checked

Curl is zero everywhere on $S$ . Then the RHS is $0$ , so circulation is $0$ for this loop. If it's zero for every loop, $F$ is conservative. Stokes gives the "why."
$S$ is flat and horizontal ( $\hat{n} = \hat{k}$ ). Only $(\nabla \times F)_{z}$ counts and Stokes collapses to Green's Theorem.
Two surfaces, same rim. RHS only depends on $S$ through its boundary $C$ , so any surface with boundary $C$ gives the same flux — swap to the easiest one.
Degenerate surface (a single point or a line). Its boundary is empty or retraces itself; both sides are $0$ . Consistent.
Singularity inside $S$ (curl undefined at a point). The tiling argument breaks there — you must exclude that point or pick another surface.
Reversed orientation. Flip $\hat{n}$ : RHS flips sign; the right-hand rule flips the lap direction, so LHS flips too. Equality still holds.

The one-picture summary

Everything on this page in a single frame: chop the skin into tiny swirls (curl × area), watch interior arrows cancel, and only the rim-lap (circulation) survives.

Recall Feynman: tell it back in plain words

Drop a wire loop into a shallow, swirling tray of water. I want to know how fast water races around the edge of the loop. Instead of following the whole edge, I chop the inside into a grid of tiny squares and measure how much water spins around each little square — that's the "curl." When two squares touch, one shoves water clockwise along their shared wall and the other shoves it counter-clockwise: perfect cancellation. Every inside wall is cancelled by its neighbour. Only the outside walls — the actual rim of the loop — have nobody to cancel them. So adding up all the tiny spins inside gives me exactly the flow around the rim. That sentence, made exact, is Stokes' theorem: curl-flux through the skin equals circulation around the rim.

Recall

Tiny-loop law (2D swirl) ::: $\oint_{tiny} F \cdot d r \approx (\partial_{x} Q - \partial_{y} P) Δ A$ Why do interior edges vanish when tiling? ::: Each interior edge is walked once in each direction by two neighbours, so the two contributions cancel exactly. Which tool measures "how $P$ changes as $y$ moves"? ::: The partial derivative $\partial P / \partial y$ . What does $\hat{n}$ do in the 3D tiny-loop law? ::: It selects, via the dot product, the part of the curl vector aligned with the loop's axis. Flat horizontal $S$ reduces Stokes to which theorem? ::: Green's Theorem.