← All posts

What is Stokes’ theorem, really?

Every calculus student meets Stokes’ theorem three separate times and is told three separate stories. First it is the fundamental theorem of calculus, dressed up along a curve: the change in a function equals the accumulation of its derivative. Then it reappears as Green’s theorem, trading a line integral around a region for a double integral over its interior. Finally it shows up in its “real” costume — flux through a surface equals circulation around its boundary. These are presented as three theorems that happen to share a name. They are not. They are one theorem, and the reason they look different is that we keep forgetting to say what “boundary” and “derivative” really mean.

The one-line version

Let $\omega$ be a differential form of degree $k$ on a smooth, oriented $(k+1)$-manifold-with-boundary $M$. Then

$$\int_M d\omega \;=\; \int_{\partial M} \omega.$$

That is the entire theorem. Everything else — the gradient theorem, Green’s theorem, the classical curl-form of Stokes’ theorem, and the divergence theorem — is this one line, read for a specific choice of $k$ and a specific shape of $M$. The exterior derivative $d$ is doing the work that “gradient,” “curl,” and “divergence” each did separately in vector calculus; it is the same operation each time, just applied to forms of different degree.

Recovering the familiar cases

Take $k = 0$, so $\omega = f$ is just a scalar function and $M$ is a curve $\gamma$ from $a$ to $b$. Then $d\omega = \nabla f \cdot d\mathbf{r}$, the boundary $\partial M$ is the two-point set $\{a, b\}$ with orientation, and the theorem reads

$$\int_\gamma \nabla f \cdot d\mathbf{r} \;=\; f(b) - f(a).$$

That is the gradient theorem — itself the fundamental theorem of calculus once you parametrize $\gamma$ by arc length. Push $k$ up to $1$ with $M$ a planar region: the same line reads out as Green’s theorem. Push it up again with $M$ a surface in $\mathbb{R}^3$: out comes the classical Stokes’ theorem relating circulation to curl. Let $M$ be a solid region instead: out comes the divergence theorem. Nothing about the underlying argument changes between these cases — only the dimension of the stage.

The right way to think about it: a manifold is glued together out of little oriented cubes, an integral is a sum over those cubes, and $d$ tracks how much a form’s value changes from one face of a cube to the opposite face. Sum that change over every cube in $M$ and almost everything cancels — each internal face is shared by two cubes with opposite orientation — except the faces sitting on $\partial M$, which have nothing to cancel against. What survives the cancellation is the boundary integral.

Why this is worth internalizing

  1. It explains why the vector calculus theorems all require an orientation, and why getting that orientation backwards flips a sign rather than producing nonsense — orientation is baked into what “$\partial M$” means as an oriented boundary.
  2. It explains why $\operatorname{curl}(\nabla f) = 0$ and $\operatorname{div}(\operatorname{curl}\mathbf{F}) = 0$ are the same fact stated twice: both are instances of $d^2 = 0$, the statement that applying the exterior derivative twice always yields zero.
  3. It generalizes immediately to manifolds that don’t sit inside $\mathbb{R}^n$ at all — the theorem doesn’t know or care whether $M$ is embedded in Euclidean space, which is exactly what you need once you start doing analysis on manifolds in earnest.

The takeaway

“Boundary of the whole equals accumulation of the parts” is one idea wearing three costumes. Once $d$ and $\partial$ are defined properly — as dual operations, one on forms and one on chains — the theorem stops being a family of formulas to memorize and becomes something closer to a definition you could have written down yourself.