Differential Forms and Exterior Calculus

The penultimate page (analysis spine). Differential forms are the objects built to be integrated over oriented $k$-dimensional regions, and the exterior derivative $d$ is the single operator that contains gradient, curl, and divergence. With forms in hand, the three classical integral theorems collapse into one (stokes-general.md). This page builds on the linear algebra of multivariable.md and the tangent-space language of topology-manifolds.md §3.

References: Spivak, Calculus on Manifolds, ch. 4; Lee, Smooth Manifolds, ch. 11–14; Rudin ch. 10.

1. Multilinear Algebra: Alternating Tensors

Fix a vector space $V={\mathbb{R}}^n$ (think tangent space $T_{p}U$). A $k$-covector is an alternating multilinear map $\omega :V^k\to \mathbb{R}$ — multilinear in each slot and changing sign under any transposition of arguments (so it vanishes when two arguments coincide). The space of these is ${\Lambda }^k{V}^{\ast }$, with

$$\dim {\Lambda }^k{V}^{\ast }=\left(\genfrac{}{}{0ex}{}{n}{k}\right).$$

In particular ${\Lambda }^0=\mathbb{R}$ (scalars), ${\Lambda }^1=V^{\ast }$ (covectors, "row vectors"), and ${\Lambda }^n$ is $1$-dimensional (top forms ↔ determinants). ${\Lambda }^k=\{0\}$ for $k>n$ — the reason space runs out at dimension $n$.

1.1 The alternation operator and the dimension count

Alternating tensors are the completely antisymmetric multilinear maps. Any $k$-linear $T:V^k\to \mathbb{R}$ can be projected onto ${\Lambda }^k{V}^{\ast }$ by the alternation (antisymmetrization) operator

$$(\mathrm{Alt}T)(v_1,\dots ,v_k)=\frac{1}{k!}\sum _{\sigma \in S_k}(\mathrm{sgn}\sigma )T(v_{\sigma (1)},\dots ,v_{\sigma (k)}),$$

where $S_k$ is the symmetric group and $\mathrm{sgn}\sigma =\pm 1$ is the sign of the permutation. $\mathrm{Alt}$ is a projection (${\mathrm{Alt}}^2=\mathrm{Alt}$) fixing exactly the alternating tensors. The count $\dim {\Lambda }^k{V}^{\ast }=\left(\genfrac{}{}{0ex}{}{n}{k}\right)$ follows because an alternating map is determined by its values on strictly increasing index tuples $i_1<\cdots <i_k$: permuting indices only flips a sign, and repeating an index forces the value to $0$. There are exactly $\left(\genfrac{}{}{0ex}{}{n}{k}\right)$ such tuples.

2. The Wedge Product

The wedge (exterior) product $\wedge :{\Lambda }^k\times {\Lambda }^{\ell }\to {\Lambda }^{k+\ell }$ is the alternating, associative, bilinear product satisfying

$$\alpha \wedge \beta =(-1{)}^{k\ell }\beta \wedge \alpha (\text{graded commutativity}),$$

so in particular $d{x}^i\wedge d{x}^j=-d{x}^j\wedge d{x}^i$ and $d{x}^i\wedge d{x}^i=0$. With the coordinate $1$-forms $d{x}^1,\dots ,d{x}^n$ (the dual basis), a basis of ${\Lambda }^k$ is

$$d{x}^{i_1}\wedge \cdots \wedge d{x}^{i_k},i_1<\cdots <i_k.$$

The wedge encodes oriented volume: $d{x}^1\wedge \cdots \wedge d{x}^n$ evaluated on vectors $v_1,\dots ,v_n$ returns $\det [v_1\cdots v_n]$, the signed volume of the parallelepiped they span. This is exactly the Jacobian determinant of multiple-integrals.md §4 made into an algebraic object.

2.1 Evaluation as a determinant, and a worked example

On decomposable forms the wedge is computed by a determinant of evaluations: for $1$-forms ${\alpha }^1,\dots ,{\alpha }^k$ and vectors $v_1,\dots ,v_k$,

$$({\alpha }^1\wedge \cdots \wedge {\alpha }^k)(v_1,\dots ,v_k)=\det [{\alpha }^i(v_j){]}_{i,j=1}^k.$$

The determinant's antisymmetry is the alternation of the form. Two immediate consequences: $\alpha \wedge \alpha =0$ for any odd-degree $\alpha$ (a determinant with two equal rows), and a decomposable $k$-form vanishes on $v_1,\dots ,v_k$ iff those vectors are linearly dependent.

Example. In ${\mathbb{R}}^3$, evaluate $dx\wedge dy$ on $u=(u_1,u_2,u_3)$ and $v=(v_1,v_2,v_3)$: $$(dx\wedge dy)(u,v)=\det \left[\begin{array}{cc}dx(u)& dx(v)\\ dy(u)& dy(v)\end{array}\right]>=\det \left[\begin{array}{cc}{u}_1& v_1\\ u_2& v_2\end{array}\right]=u_1{v}_2-u_2{v}_1,$$ the signed area of the shadow of the parallelogram $[u,v]$ on the $xy$-plane. Likewise $dy\wedge dz$ and $dz\wedge dx$ read off the other two coordinate-plane shadows — which is precisely why the three of them assemble a flux $2$-form in the §3 dictionary.

3. Differential Forms

A differential $k$-form on an open $U\subseteq {\mathbb{R}}^n$ (or on a manifold) is a smooth assignment $p\mapsto {\omega }_p\in {\Lambda }^k{T}_p^{\ast }U$. In coordinates,

$$\omega =\sum _{i_1<\cdots <i_k}{a}_{i_1\cdots i_k}(x)d{x}^{i_1}\wedge \cdots \wedge d{x}^{i_k},$$

with smooth coefficient functions. Forms of degree $0$ are functions; the space of all $k$-forms is written ${\Omega }^k(U)$.

The ${\mathbb{R}}^3$ dictionary. Forms repackage the fields of vector-calculus.md:

4. The Exterior Derivative

Exterior derivative. There is a unique linear operator $d:{\Omega }^k\to >{\Omega }^{k+1}$ with: (i) on $0$-forms $df={\sum }_i\frac{\partial f}{\partial x^i}d{x}^i$; (ii) graded Leibniz $d(\alpha \wedge \beta )=d\alpha \wedge \beta +>(-1{)}^{\mathrm{deg}\alpha }\alpha \wedge d\beta$; and (iii) $d\circ d=0$.

On a general form, $d$ acts on the coefficients and wedges in the new $dx$: $d(\sum a_{I}d{x}^I)=\sum d{a}_I\wedge d{x}^I$.

4.1 $d$ is grad, curl, div

Through the §3 dictionary in ${\mathbb{R}}^3$:

$$df\leftrightarrow \nabla f,d{\omega }_{\mathbf{F}}^1\leftrightarrow {\omega }_{\nabla \times \mathbf{F}}^2,d{\omega }_{\mathbf{F}}^2\leftrightarrow (\nabla \cdot \mathbf{F})dx\wedge dy\wedge dz.$$

The three differential operators are the single operator $d$ acting on degrees $0,1,2$. This is the structural statement behind the unified pattern of integral-theorems.md §5.

The correspondence is a direct coordinate computation. For a $1$-form $\omega =Pdx+Qdy+Rdz$, expand each $dP=P_{x}dx+P_{y}dy+P_{z}dz$ (and similarly $dQ,dR$), wedge on the trailing coordinate, and cancel the vanishing $dx\wedge dx$ terms:

$$d\omega =(R_y-Q_z)dy\wedge dz+(P_z-R_x)dz\wedge dx+(Q_x-P_y)dx\wedge dy,$$

whose three coefficients are exactly the components of $\nabla \times \mathbf{F}$. For a $2$-form $\beta =Ady\wedge dz+Bdz\wedge dx+Cdx\wedge dy$, only the "missing" coordinate survives in each $dA,dB,dC$:

$$d\beta =(A_x+B_y+C_z)dx\wedge dy\wedge dz,$$

the divergence $\nabla \cdot \mathbf{F}$ times the volume form. So $d$ literally is curl on $1$-forms and divergence on $2$-forms.

4.2 $d^2=0$ encodes the two identities

The defining property $d^2=0$ is exactly the symmetry of mixed partials (multivariable.md §4). Reading it through the dictionary:

$$d^{2}f=0\leftrightarrow \nabla \times (\nabla f)=0,d^2{\omega }^1=0\leftrightarrow \nabla \cdot (\nabla \times \mathbf{F})=0.$$

The two "fundamental identities" of vector calculus (vector-calculus.md §2.1) are one algebraic fact.

Why $d^2=0$ holds. On a function it is Clairaut's theorem outright:

$$d(df)=d\left(\sum _i\frac{\partial f}{\partial x^i}d{x}^i\right)=\sum _{i,j}\frac{{\partial }^{2}f}{\partial x^j\partial x^i}d{x}^j\wedge d{x}^i=0,$$

because the coefficient ${\partial }^{2}f/\partial x^j\partial x^i$ is symmetric in $i,j$ (multivariable.md §4) while $d{x}^j\wedge d{x}^i$ is antisymmetric, and the contraction of a symmetric with an antisymmetric array is zero. For a general $\omega =\sum a_{I}d{x}^I$, graded Leibniz gives $d\omega =\sum d{a}_I\wedge d{x}^I$, and applying $d$ again reduces to the $0$-form case on each $a_I$ (the $d{x}^I$ factor is constant, killed by $d$). So $d^2=0$ everywhere.

4.3 Interior product and the Lie derivative (Cartan's formula)

Two more operators complete the working toolkit. Given a vector field $X$, the interior product (contraction) ${\iota }_X:{\Omega }^k\to {\Omega }^{k-1}$ plugs $X$ into the first slot,

$$({\iota }_X\omega )(v_1,\dots ,v_{k-1})=\omega (X,v_1,\dots ,v_{k-1}),$$

and is an antiderivation of degree $-1$: ${\iota }_X(\alpha \wedge \beta )=({\iota }_X\alpha )\wedge \beta +(-1{)}^{\mathrm{deg}\alpha }\alpha \wedge ({\iota }_X\beta )$, with ${\iota }_X{\iota }_X=0$. The Lie derivative ${\mathcal{L}}_X\omega$ measures the infinitesimal change of $\omega$ dragged along the flow of $X$. The three operators are tied together by Cartan's magic formula:

$$\boxed{{\mathcal{L}}_X=d{\iota }_X+{\iota }_{X}d}$$

This single identity powers computations in fluid dynamics (transport of circulation), Hamiltonian mechanics (a symplectic form is preserved iff ${\mathcal{L}}_X\omega =0$), and the derivation of conservation laws from symmetries. Because $d^2=0$, it also gives ${\mathcal{L}}_{X}d=d{\mathcal{L}}_X$ — the Lie derivative commutes with the exterior derivative.

5. Pullback, Orientation, and Integration of Forms

5.1 Pullback

A smooth map $\varphi :U\to W$ induces the pullback ${\varphi }^{\ast }:{\Omega }^k(W)\to {\Omega }^k(U)$, substituting $y=\varphi (x)$ and expanding the $dy$'s via $d{y}^j={\sum }_i\frac{\partial {\varphi }^j}{\partial x^i}d{x}^i$. Pullback commutes with $d$ (${\varphi }^{\ast }d\omega =d{\varphi }^{\ast }\omega$) and with $\wedge$. Crucially, pulling back a top form produces the Jacobian determinant:

$${\varphi }^{\ast }(d{y}^1\wedge \cdots \wedge d{y}^n)=\det (D\varphi )d{x}^1\wedge \cdots \wedge d{x}^n.$$

This is why forms are the natural integrands: the change-of-variables factor appears automatically, and with its sign (orientation), not the absolute value of multiple-integrals.md §4.

Worked pullback — polar coordinates. Let $\varphi (r,\theta )=(r\cos \theta ,>r\sin \theta )$, so $x=r\cos \theta$, $y=r\sin \theta$. Pulling back the coordinate $1$-forms, $${\varphi }^{\ast }dx=\cos \theta dr-r\sin \theta d\theta ,>{\varphi }^{\ast }dy=\sin \theta dr+r\cos \theta d\theta .$$ Wedging (and using $dr\wedge dr=d\theta \wedge d\theta =0$, $d\theta \wedge dr=>-dr\wedge d\theta$), $${\varphi }^{\ast }(dx\wedge dy)=({\cos }^2\theta +{\sin }^2\theta )rdr\wedge d\theta =>rdr\wedge d\theta ,$$ recovering the familiar area element $rdrd\theta$ — here the Jacobian factor $r=\det D\varphi$ drops out of the algebra with no separate computation.

5.2 Orientation

An orientation of $U$ (or a manifold) is a consistent choice of "positively oriented" ordered bases — equivalently a nowhere-zero top form. A diffeomorphism is orientation-preserving iff $\det D\varphi >0$. Forms know about orientation intrinsically, which is what lets the integral theorems track signs (integral-theorems.md §1).

5.3 Integration of a $k$-form

A $k$-form integrates over an oriented $k$-dimensional region. Over a single oriented coordinate patch parametrized by $\varphi :D\subseteq {\mathbb{R}}^k\to U$,

$${\int }_{\varphi }\omega ={\int }_D{\varphi }^{\ast }\omega ,$$

the right side being an ordinary multiple integral (multiple-integrals.md). Pullback-invariance under orientation-preserving reparametrization (by §5.1) makes this well-defined. This single definition specializes to line integrals ($k=1$), surface/flux integrals ($k=2$), and volume integrals ($k=n$) of vector-calculus.md.

6. Closed and Exact Forms (de Rham, briefly)

$\omega$ is closed if $d\omega =0$, and exact if $\omega =d\eta$ for some $\eta$. Since $d^2=0$, every exact form is closed. The converse — is every closed form exact? — is a question about the topology of $U$:

$$H_{\mathrm{dR}}^k(U)=\frac{\{\text{closed }k\text{-forms}\}}{\{\text{exact }k\text{-forms}\}}$$

is the de Rham cohomology, measuring "holes." Poincaré lemma: on a contractible (e.g. star-shaped) set, closed $\Rightarrow$ exact, so all cohomology vanishes. This is the precise version of "irrotational $\Rightarrow$ conservative on simply connected domains" (vector-calculus.md §3.1); the counterexample field around the $z$-axis represents a nonzero class in $H_{\mathrm{dR}}^1$. (A full treatment belongs to algebraic topology and is beyond this spine.)

The canonical closed-but-not-exact form. On the punctured plane ${\mathbb{R}}^2\setminus \{0\}$, the angle form $$\omega =\frac{-ydx+xdy}{x^2+y^2}$$ is closed: a direct computation gives $d\omega =0$ (equivalently, it is locally $d(\arg )$). But it is not exact, because its integral around the unit circle is ${\oint }_{S^1}\omega =2\pi \ne 0$, whereas every exact form integrates to $0$ over a closed loop (by Stokes, stokes-general.md). It therefore represents a nonzero class generating $H_{\mathrm{dR}}^1({\mathbb{R}}^2>\setminus \{0\})\cong \mathbb{R}$ — the "hole" at the origin made algebraic. This is the exact analytic content of the winding-number obstruction in vector-calculus.md §3.1.

Constructive Poincaré (the homotopy operator). On a star-shaped $U$ the lemma is proved by an explicit homotopy operator $K:{\Omega }^k\to {\Omega }^{k-1}$ satisfying $dK+Kd=\mathrm{id}$ on positive-degree forms. Applied to a closed $\omega$ ($d\omega =0$) it yields $\omega =d(K\omega )$, so $\eta =K\omega$ is an explicit primitive. This is the same integrate-along-radial-rays construction that builds a potential from a conservative field.

7. The Hodge Star (brief)

With an inner product and orientation, the Hodge star $\star :{\Lambda }^k\to {\Lambda }^{n-k}$ identifies complementary-degree forms (e.g. in ${\mathbb{R}}^3$ it turns the $1$-form ${\omega }_{\mathbf{F}}^1$ into the $2$-form ${\omega }_{\mathbf{F}}^2$, explaining why a vector field has two form-incarnations in §3). It makes the divergence $\delta =\pm \star d\star$ and the Laplacian $\Delta =d\delta +\delta d$ into operations on forms, and is the language in which Maxwell's equations read $dF=0$, $d\star F=\star J$ — see SR/fields/electromagnetism.md.

On ${\mathbb{R}}^3$ with the standard metric and orientation it acts on the basis as

$$\star 1=dx\wedge dy\wedge dz,\star dx=dy\wedge dz,\star dy=dz\wedge dx,\star dz=dx\wedge dy,$$

and $\star$ on the $2$- and $3$-forms is the inverse of this (with $\star (dx\wedge dy\wedge dz)=1$). In general $\star \star =(-1{)}^{k(n-k)}$ on ${\Lambda }^k$ of an $n$-dimensional Euclidean space. Reading the curl formula of §4.1 as $\nabla \times \mathbf{F}\leftrightarrow \star d{\omega }_{\mathbf{F}}^1$ and the divergence as $\nabla \cdot \mathbf{F}\leftrightarrow \star d\star {\omega }_{\mathbf{F}}^1$ makes the ${\mathbb{R}}^3$ dictionary of §3 exact rather than heuristic.

8. Where this page is used

• The exterior derivative $d$ and integration of forms are exactly what the generalized Stokes theorem ${\int }_{M}d\omega ={\int }_{\partial M}\omega$ relates (stokes-general.md).
• The form/operator dictionary recovers Green's, Stokes', and the divergence theorem as special cases (integral-theorems.md).
• Forms on manifolds use the tangent/cotangent structure of topology-manifolds.md §3.

Next: The Generalized Stokes Theorem — the capstone.