The Generalized Stokes Theorem

The capstone of the analysis spine. Everything converges here: the real numbers gave us limits, limits gave derivatives and the Riemann integral, the FTC linked them, the multivariable theory gave the total derivative and multiple integrals, and differential forms gave the single operator $d$. The reward is one theorem that contains the Fundamental Theorem of Calculus, Green's theorem, Stokes' theorem, and the divergence theorem as special cases:

$$\boxed{{\int }_{M}d\omega ={\int }_{\partial M}\omega }$$

References: Spivak, Calculus on Manifolds, ch. 5; Lee, Introduction to Smooth Manifolds, ch. 16; Rudin ch. 10.

1. Manifolds with Boundary and Orientation

The objects we integrate over are oriented smooth $k$-manifolds with boundary (the definition is in topology-manifolds.md §2.5): spaces locally modelled on the half-space ${\mathbb{H}}^k=\{x^1\ge 0\}$, whose boundary $\partial M$ is a $(k-1)$-manifold.

• An orientation is a consistent choice of positively oriented charts — equivalently a nowhere-vanishing top form (differential-forms.md §5.2).
• The orientation of $M$ induces one on $\partial M$ by the outward-normal- first convention: an ordered basis $(v_2,\dots ,v_k)$ of $T_p\partial M$ is positive iff $(\nu ,v_2,\dots ,v_k)$ is positive in $T_{p}M$, where $\nu$ points out of $M$. This single rule reproduces all the orientation conventions of integral-theorems.md §1: CCW boundary of a plane region, right-hand rule for a surface, outward normal for a solid.

2. Integration of Forms over a Manifold

A $k$-form $\omega$ on $M$ is integrated by partition of unity: choose a locally finite cover by oriented charts $\{(U_{\alpha },{\varphi }_{\alpha })\}$ and smooth functions ${\rho }_{\alpha }\ge 0$ supported in $U_{\alpha }$ with ${\sum }_{\alpha }{\rho }_{\alpha }\equiv 1$, then

$${\int }_M\omega =\sum _{\alpha }{\int }_{{\varphi }_{\alpha }(U_{\alpha })}({\varphi }_{\alpha }^{-1}{)}^{\ast }({\rho }_{\alpha }\omega ).$$

Each summand is an ordinary multiple integral (multiple-integrals.md), and the result is independent of all choices because pullback transforms forms by the signed Jacobian (differential-forms.md §5.1) and the charts are orientation-compatible.

Why bump functions exist. Partitions of unity require smooth functions that are $1$ on a set and $0$ outside a slightly larger one. These exist precisely because $C^{\infty }\ne C^{\omega }$ — the non-analytic bump $e^{-1/x^2}$ of differentiation.md §5 is the seed. This is the technical debt from the one-variable theory being repaid at the summit.

3. The Theorem

Generalized Stokes Theorem. Let $M$ be a compact, oriented, smooth $k$-manifold with boundary $\partial M$ (given the induced orientation), and let $\omega$ be a smooth $(k-1)$-form on $M$. Then $${\int }_{M}d\omega ={\int }_{\partial M}\omega .$$ (If $\partial M=\varnothing$, the right side is $0$: the integral of an exact form over a closed manifold vanishes.)

3.1 Proof outline

The proof is a clean three-step reduction:

1. Partition of unity. Using §2 and the linearity of both sides, reduce to the case where $\omega$ is supported inside a single chart. The interior chart contributions and boundary chart contributions can be handled separately.
2. Local model. Transport to the half-space ${\mathbb{H}}^k$ via the chart. Write $\omega ={\sum }_j{a}_{j}d{x}^1\wedge \cdots \widehat{d{x}^j}\cdots \wedge d{x}^k$ (the hat omits $d{x}^j$). Then $d\omega ={\sum }_j(-1{)}^{j-1}\frac{\partial a_j}{\partial x^j}d{x}^1\wedge \cdots \wedge d{x}^k$.
3. Fundamental Theorem of Calculus. Integrating each term over ${\mathbb{H}}^k$ and applying Fubini (multiple-integrals.md §3), the one-variable FTC (riemann-integral.md §5) collapses each $\int \frac{\partial a_j}{\partial x^j}d{x}^j$ to a boundary evaluation. Only the $x^1$-term survives the boundary $\{x^1=0\}$ of ${\mathbb{H}}^k$ (compact support kills the rest), and it equals exactly ${\int }_{\partial M}\omega$ with the induced orientation. $\square$

So the entire theorem is the FTC, applied one coordinate at a time and stitched together by partitions of unity. The whole spine exists to make this sentence rigorous.

4. Recovering the Classical Theorems

Each classical theorem is ${\int }_{M}d\omega ={\int }_{\partial M}\omega$ for a specific $k$ and dictionary entry (differential-forms.md §3):

In each row, $d\omega$ is grad/curl/div of the corresponding field (§4.1 of differential-forms.md), and the induced boundary orientation (§1) is the classical sign convention. The four cornerstone theorems of calculus are one theorem.

5. Consequences and Outlook

• Stokes depends only on the boundary. If $\partial M_1=\partial M_2$ and $\omega$ is closed, ${\int }_{M_1}\omega ={\int }_{M_2}\omega$ — the integral of a closed form depends only on the homology class. This is the bridge from forms to de Rham cohomology (differential-forms.md §6).
• Conservation laws. In physics, $d\omega =0$ (a continuity equation) integrated via Stokes gives a conserved charge; Gauss's and Ampère's laws are the ${\mathbb{R}}^3$ instances, and the covariant Maxwell equations $dF=0$, $d\star F=\star J$ are the spacetime version (SR/fields/electromagnetism.md).
• Onward. The same theorem on Riemannian manifolds (with the Hodge star) yields Green's identities, Hodge theory, and the analytical backbone of gauge theory — the mathematics underlying the physics sections.

6. The Spine, End to End

$$\underset{\text{real-numbers}}{\underbrace{\text{completeness of }\mathbb{R}}}\to \underset{\text{sequences, continuity}}{\underbrace{\text{limits}}}\to \underset{\text{differentiation, riemann}}{\underbrace{\text{derivative}+\text{integral}}}\stackrel{\text{FTC}}{\to }\underset{\text{metric, multivariable, multiple}}{\underbrace{\text{total derivative}+\text{multiple integrals}}}\to \underset{\text{forms}}{\underbrace{d+\int \omega }}\stackrel{{\int }_{M}d\omega ={\int }_{\partial M}\omega }{\to }\text{Stokes.}$$

Every page was a prerequisite for the box at the top of this one.

Back to: Analysis spine index · Mathematics overview.