Differential Forms and the Exterior Algebra

Give a geometer a manifold with no ruler, no protractor, and no way to compare vectors at different points, and one calculus still survives intact: the calculus of differential forms. Strip away the metric and the connection — everything that lets you measure length, angle, or parallel transport — and you can still wedge forms together, differentiate them, pull them back through smooth maps, and integrate them over curves, surfaces, and volumes. This is the skeleton of differential geometry, the part that is pure topology and smooth structure, standing before any physics of distance is poured in. The reason forms occupy this privileged place is a single algebraic fact: total antisymmetry. A quantity that flips sign whenever you swap two of its slots behaves like a determinant, and the determinant is exactly the Jacobian factor that the change-of-variables formula demands for a coordinate-independent integral. So the totally antisymmetric covariant tensors — and only they — are the legitimate integrands of geometry. A $p$ -form is the object you integrate over a $p$ -dimensional surface, and it does so without ever asking what coordinates you chose. That is why a $p$ -form has only $\binom{n}{p}$ components instead of $n^p$ : antisymmetry throws away everything that would spoil the determinant law. Two operations then close the system. The wedge product $\wedge$ builds higher-degree forms and is graded-anticommutative; the exterior derivative $\mathrm{d}$ raises degree by one, is nilpotent ( $\mathrm{d}^2=0$ ), and needs no connection. Tie them to the boundary operator through Stokes' theorem and the whole edifice — de Rham cohomology, $F=\mathrm{d}A$ , Cartan's structure equations, conserved charges in general relativity — falls out of these few lines.

Key takeaways

A differential $p$ -form is a totally antisymmetric covariant tensor field; the forms of all degrees assemble into the exterior algebra $\Lambda^\bullet$ under the wedge product, with $\dim\Lambda^p=\binom{n}{p}$ .
The wedge is graded-anticommutative, $\alpha\wedge\beta=(-1)^{pq}\beta\wedge\alpha$ , so any odd-degree form wedged with itself vanishes — this antisymmetry is what makes forms behave like signed volumes.
The exterior derivative $\mathrm{d}$ is the unique nilpotent antiderivation that extends the differential of a function; $\mathrm{d}^2=0$ holds because mixed partial derivatives commute while $\mathrm{d}$ antisymmetrizes them.
Forms, $\wedge$ , $\mathrm{d}$ , pullback, and integration are all metric-free — they live on a bare smooth manifold; the metric re-enters only through the Hodge star $\star$ .
Stokes' theorem $\int_M\mathrm{d}\omega=\int_{\partial M}\omega$ makes $\mathrm{d}$ dual to $\partial$ , so $\mathrm{d}^2=0$ is the algebraic shadow of "the boundary of a boundary is empty".
In four-dimensional spacetime a 2-form has exactly $\binom{4}{2}=6$ components — the electric and magnetic fields packaged as $F=\mathrm{d}A$ , with the source-free Maxwell equations reduced to $\mathrm{d}F=0$ .
Key idea: a $p$ -form is a machine that eats $p$ tangent vectors and returns a number totally antisymmetrically; antisymmetry makes that number a signed volume, signed volumes are what you integrate, and $\wedge$ , $\mathrm{d}$ , and Stokes are the algebra, calculus, and integral theorem of those integrands.

The core insight

A $p$ -form is a machine that eats $p$ tangent vectors and returns a number, totally antisymmetrically. Antisymmetry is exactly the property that makes the answer behave like a signed volume — and signed volumes are what you integrate. Wedge builds higher forms, $\mathrm{d}$ differentiates them with $\mathrm{d}^2=0$ , and Stokes' theorem ties $\mathrm{d}$ to the boundary operator.

Where this lives today

The form calculus is the working vocabulary of every corner of modern gravitational physics. When numerical relativists and mathematical physicists compute the curvature of a new black-hole or cosmological metric, they almost never grind through Christoffel symbols by hand — they set up an orthonormal coframe $\{e^a\}$ and turn Cartan's structure equations, $\mathrm{d}e^a+\omega^a{}_b\wedge e^b=0$ and $R^a{}_b=\mathrm{d}\omega^a{}_b+\omega^a{}_c\wedge\omega^c{}_b$ , into a few lines of wedge-product algebra, cutting a page of tensor bookkeeping to a paragraph. The same language runs through the frontier: in supergravity and string theory the fundamental fields are higher $p$ -form gauge potentials, with $F=\mathrm{d}A$ generalizing to $p$ -form field strengths whose fluxes $\int F$ are quantized topological charges. Closed-but-not-exact forms detect the magnetic monopole (a nontrivial $F$ on a sphere surrounding it) and the Aharonov–Bohm phase (a holonomy $\oint A$ with $\mathrm{d}A=0$ ), turning de Rham cohomology into measurable physics. And the ADM and Komar expressions for the mass and angular momentum of an isolated gravitating system are surface integrals of two-forms — conserved charges that, thanks to $\mathrm{d}^2=0$ and Stokes, can be evaluated on any enclosing sphere at once.

The physical picture

Forget components for a moment and picture what a form does. A one-form is not an arrow; it is a stack of parallel level surfaces — the constant-value sheets of some scalar. To evaluate it on a vector you count how many sheets the arrow pierces. A denser stack is a "bigger" one-form, and a vector lying inside a single sheet pierces nothing and returns zero. This is the honest geometric meaning of the gradient $\mathrm{d}f$ : the surfaces are the level sets of $f$ , and $\mathrm{d}f(V)$ is the rate at which $f$ climbs along $V$ , read off as a pierce-count. Two-forms and higher repeat the idea one dimension up: a two-form is a stack of tubes, and it returns the flux of a parallelogram threading through them.

Figure

A one-form ω is a stack of level surfaces, and a vector V is evaluated by counting the surfaces it pierces — here ⟨ω, V⟩ = 3. A two-form eats two vectors and returns the signed area of the parallelogram they span, so ω(u, v) = −ω(v, u): swapping the two edges flips the sign, exactly as a determinant does

Now the exterior derivative. Think of $\mathrm{d}$ as the operator that turns a form into its "boundary content one degree up". Stokes' theorem $\int_M\mathrm{d}\omega=\int_{\partial M}\omega$ is the precise statement: integrating $\mathrm{d}\omega$ over a region equals integrating $\omega$ over the region's edge. Read backwards, this says $\mathrm{d}$ is the algebraic adjoint of the geometric boundary operator $\partial$ . And because the boundary of a boundary is empty — the rim of a filled disk is a circle, but a circle has no endpoints — the adjoint statement must be $\mathrm{d}^2=0$ . That one identity is the seed of de Rham cohomology: forms that are closed ( $\mathrm{d}\omega=0$ ) but not exact ( $\omega\ne\mathrm{d}\eta$ ) detect holes in the manifold that no local computation can see. The final structural fact is dimensional. On an $n$ -manifold there is a top degree: $\Lambda^p=0$ for $p>n$ , because you cannot find $p>n$ linearly independent one-forms at a point to wedge together — the pigeonhole principle of antisymmetry. The number of independent components climbs and then falls symmetrically as $\binom{n}{p}$ , and the whole exterior algebra has dimension $2^n$ . The figure below shows this for the physically decisive case $n=4$ .

Figure

The de Rham complex on a 4-manifold: d raises degree by one and dim Λᵖ = binomial(4, p) runs 1, 4, 6, 4, 1, summing to 2⁴ = 16 for the whole exterior algebra, while d² = 0 makes the arrows a complex. The Maxwell field strength F sits in Λ² with exactly 6 independent components — the 3 electric plus the 3 magnetic

The core equations

A $p$ -form is written in a coordinate basis by antisymmetrizing the wedge of $p$ basis one-forms. The $1/p!$ prefactor corrects for the fact that the antisymmetric sum already visits each ordering of the indices $p!$ times, so it prevents double counting. The component array is totally antisymmetric by construction.

\omega=\frac{1}{p!}\,\omega_{\mu_1\cdots\mu_p}\,\mathrm{d}x^{\mu_1}\wedge\cdots\wedge\mathrm{d}x^{\mu_p},\qquad \omega_{\mu_1\cdots\mu_p}=\omega_{[\mu_1\cdots\mu_p]} \qquad (1)

where: $\omega_{\mu_1\cdots\mu_p}$ — the totally antisymmetric component array; $[\cdots]$ — antisymmetrization over the enclosed indices with weight $1/p!$ ; $\mathrm{d}x^\mu$ — the coordinate basis one-forms; $p!$ — cancels the overcounting of the $p!$ index orderings. Physical reading: the form is a bookkeeping device for signed volumes, and only the antisymmetric part of any component array survives. Antisymmetry drastically shrinks the number of independent components. Choosing $p$ distinct indices out of $n$ is a binomial count, and summing over all degrees recovers a full power of two — this is why the exterior algebra is finite-dimensional at each point even though the tensor algebra is not.

\binom{n}{p}=\dim\Lambda^p M,\qquad \sum_{p=0}^n \binom{n}{p}=2^n=\dim\Lambda^\bullet M \qquad (2)

where: $n$ — the manifold dimension; $p$ — the form degree, $0\le p\le n$ ; $\binom{n}{p}$ — the count of independent components. Physical reading: on 4D spacetime the degrees carry $1,4,6,4,1$ components; the $6$ at $p=2$ is exactly the number of electromagnetic field components. The wedge product multiplies forms by antisymmetrizing their tensor product. The combinatorial prefactor keeps the normalization consistent with equation (1), so that wedging basis forms reproduces the expected antisymmetrized products.

(\alpha\wedge\beta)_{\mu_1\cdots\mu_{p+q}}=\frac{(p+q)!}{p!\,q!}\,\alpha_{[\mu_1\cdots\mu_p}\,\beta_{\mu_{p+1}\cdots\mu_{p+q}]} \qquad (3)

where: $\alpha$ — a $p$ -form; $\beta$ — a $q$ -form; $\alpha\wedge\beta$ — the resulting $(p+q)$ -form. Physical reading: wedging is how you build a flux 2-form from two gradient 1-forms, or the volume form from $n$ independent one-forms. The two defining algebraic properties of the wedge are graded-anticommutativity and associativity. The sign $(-1)^{pq}$ records how many index transpositions it takes to slide one factor past the other. For an odd-degree form this forces $\alpha\wedge\alpha=0$ .

\alpha\wedge\beta=(-1)^{pq}\,\beta\wedge\alpha,\qquad (\alpha\wedge\beta)\wedge\gamma=\alpha\wedge(\beta\wedge\gamma) \qquad (4)

where: $p,q$ — the degrees of $\alpha,\beta$ ; $(-1)^{pq}$ — the sign from commuting one factor through the other. Physical reading: a one-form wedged with itself is zero, but a two-form need not be — $F\wedge F$ is the nonzero Pontryagin (instanton) density. The exterior derivative raises degree by one and is built purely from partial derivatives, antisymmetrized. No connection appears, so $\mathrm{d}$ is defined on any smooth manifold before a metric is chosen.

(\mathrm{d}\omega)_{\mu_0\mu_1\cdots\mu_p}=(p+1)\,\partial_{[\mu_0}\omega_{\mu_1\cdots\mu_p]} \qquad (5)

where: $\mathrm{d}\omega$ — the $(p+1)$ -form derivative of $\omega$ ; $\partial_{[\mu_0}$ — the partial derivative antisymmetrized against the form indices; $(p+1)$ — the combinatorial weight matching equation (1). Physical reading: on a function this is the gradient; on a 1-form it is the curl; on an $(n-1)$ -form it is the divergence. Two identities pin the exterior derivative down completely: it squares to zero, and it obeys a graded Leibniz rule where the sign depends on the degree of the factor $\mathrm{d}$ passes through. Together with $\mathrm{d}f=\partial_\mu f\,\mathrm{d}x^\mu$ on functions, these properties make $\mathrm{d}$ unique.

\mathrm{d}^2=0,\qquad \mathrm{d}(\alpha\wedge\beta)=\mathrm{d}\alpha\wedge\beta+(-1)^{p}\,\alpha\wedge\mathrm{d}\beta \qquad (6)

where: $\mathrm{d}^2$ — the composite $\mathrm{d}\circ\mathrm{d}$ ; $(-1)^p$ — the sign from $\mathrm{d}$ (degree $+1$ ) crossing the $p$ -form $\alpha$ . Physical reading: $\mathrm{d}^2=0$ delivers both $\nabla\times\nabla f=0$ and $\nabla\cdot(\nabla\times\mathbf A)=0$ in one stroke. Stokes' theorem is the master integral identity. It says the integral of $\mathrm{d}\omega$ over an oriented region equals the integral of $\omega$ over the boundary, with the induced orientation. Green's theorem, the divergence theorem, and the classical curl theorem are all special cases.

\int_M \mathrm{d}\omega=\int_{\partial M}\omega \qquad (7)

where: $M$ — an oriented $p$ -dimensional manifold-with-boundary; $\partial M$ — its boundary with induced orientation; $\omega$ — a $(p-1)$ -form, compactly supported if $M$ is noncompact. Physical reading: this is where "the field on the boundary determines the charge inside" becomes a theorem. Only now does the metric enter, through the Hodge star, which pairs a $p$ -form with an $(n-p)$ -form using the volume element. It is the single place where signature and length reappear in the form calculus.

(\star\omega)_{\mu_1\cdots\mu_{n-p}}=\frac{1}{p!}\,\sqrt{|g|}\,\epsilon_{\mu_1\cdots\mu_{n-p}}{}^{\nu_1\cdots\nu_p}\,\omega_{\nu_1\cdots\nu_p} \qquad (8)

where: $\star$ — the Hodge dual $\Lambda^p\to\Lambda^{n-p}$ ; $\epsilon$ — the Levi-Civita symbol; $\sqrt{|g|}$ — the square root of the absolute metric determinant. Physical reading: $\star$ turns the field 2-form $F$ into its dual $\star F$ , and the metric hides entirely inside it — which is why $\mathrm{d}F=0$ is topology and $\mathrm{d}\star F=\star J$ is dynamics.

Derivation ACoordinate independence of the exterior derivative

The exterior derivative is defined by the innocuous-looking formula $(\mathrm{d}\omega)=(p+1)\,\partial_{[\mu_0}\omega_{\mu_1\cdots\mu_p]}$ , yet a plain partial derivative of a tensor is not a tensor — under a coordinate change it picks up inhomogeneous terms full of second derivatives of the transition map. The goal here is to show that the antisymmetrization in the definition kills exactly those bad terms, so that $\mathrm{d}\omega$ transforms as a genuine $(p+1)$ -form. This is what licenses using $\mathrm{d}$ on any bare manifold with no connection at all, and it is worth seeing in detail because the same antisymmetrization argument recurs everywhere in the theory.

Step 1. Start from the covariant transformation of the components under a change $y^a\mapsto x^\mu(y)$ : $\omega'_{a_1\cdots a_p}=\frac{\partial x^{\mu_1}}{\partial y^{a_1}}\cdots\frac{\partial x^{\mu_p}}{\partial y^{a_p}}\,\omega_{\mu_1\cdots\mu_p}$ . Why: this is just the statement that $\omega$ is a covariant tensor field; everything else follows from differentiating it.

Step 2. Differentiate with $\partial_{a_0}$ . The product rule gives one "good" term where the derivative lands on $\omega$ , carrying the full Jacobian string, plus $p$ "bad" terms where it lands on a Jacobian factor, each producing a second derivative $\partial^2 x^{\mu_i}/\partial y^{a_0}\partial y^{a_i}$ . Why: the bad terms are precisely the obstruction to $\partial_a\omega'$ being tensorial — they must be shown to cancel.

Step 3. Note that each second derivative $\partial^2 x^{\mu_i}/\partial y^{a_0}\partial y^{a_i}$ is symmetric under exchanging the two lower indices $a_0\leftrightarrow a_i$ , by equality of mixed partials for a smooth coordinate map. Why: symmetry in the derivative indices is exactly the property that antisymmetrization annihilates.

Step 4. Antisymmetrize the whole expression over $[a_0 a_1\cdots a_p]$ . Every bad term is symmetric in a pair of those indices while being antisymmetrized over that pair, so it vanishes identically, leaving only the good term. Why: a quantity symmetric in two indices, antisymmetrized over the same two, is zero — the workhorse identity of the entire subject.

Step 5. What remains is $(\mathrm{d}\omega')_{a_0\cdots a_p}=\frac{\partial x^{\mu_0}}{\partial y^{a_0}}\cdots\frac{\partial x^{\mu_p}}{\partial y^{a_p}}\,(\mathrm{d}\omega)_{\mu_0\cdots\mu_p}$ , the clean covariant transformation of a $(p+1)$ -form. Why: this proves $\mathrm{d}$ maps forms to forms with no metric or connection, the fact that makes it universal.

Go deeper · Derivation B · 1 minWhy the exterior derivative squares to zero

Derivation BWhy the exterior derivative squares to zero

The nilpotency $\mathrm{d}^2=0$ is the single most consequential identity attached to differential forms: it is what makes the sequence of spaces $\Lambda^0\to\Lambda^1\to\cdots\to\Lambda^n$ a complex, what defines closed and exact forms, and what powers de Rham cohomology, the Bianchi identity, and charge conservation. The goal is to prove it directly from components and then to read off its geometric meaning through Stokes' theorem, so that " $\mathrm{d}^2=0$ " and "the boundary of a boundary is empty" are seen to be the same statement wearing two costumes.

Step 1. Apply $\mathrm{d}$ once: $(\mathrm{d}\omega)_{\mu_0\cdots\mu_p}=(p+1)\,\partial_{[\mu_0}\omega_{\mu_1\cdots\mu_p]}$ , a $(p+1)$ -form built from a single partial derivative antisymmetrized with the form indices. Why: to iterate we need the first derivative in explicit component form.

Step 2. Apply $\mathrm{d}$ again. The result is proportional to $\partial_{[\nu}\partial_{\mu_0}\omega_{\mu_1\cdots\mu_p]}$ — two partial derivatives, both swept into the same antisymmetrization together with all the original indices. Why: the outer $\mathrm{d}$ contributes a second derivative index $\nu$ that joins the bracket, so both derivative indices are now antisymmetrized.

Step 3. Invoke equality of mixed partials, $\partial_\nu\partial_{\mu_0}\omega=\partial_{\mu_0}\partial_\nu\omega$ for smooth $\omega$ : the pair of derivative indices is symmetric. Why: smoothness guarantees the second derivative is symmetric in its two slots, and that pair sits inside the antisymmetrizer.

Step 4. Antisymmetrizing a symmetric pair gives zero, so $\mathrm{d}(\mathrm{d}\omega)=0$ for every form $\omega$ — no exceptions, no conditions beyond smoothness. Why: same annihilation identity as the previous derivation, applied to the derivative indices rather than the transition-map indices.

Step 5. Read it geometrically through Stokes: $\int_M\mathrm{d}(\mathrm{d}\eta)=\int_{\partial M}\mathrm{d}\eta=\int_{\partial(\partial M)}\eta$ . Since $\partial(\partial M)=\varnothing$ , the integral vanishes for all $\eta$ , which is the integral form of $\mathrm{d}^2=0$ . Why: this exhibits $\mathrm{d}$ as adjoint to $\partial$ , making nilpotency the dual of "a boundary has no boundary".

Go deeper · Derivation C · 1 minMaxwell's equations as two form identities

Derivation CMaxwell's equations as two form identities

Classical electromagnetism, four vector equations in the textbooks, collapses to two lines in the language of forms — and the collapse is not cosmetic. It cleanly separates the part of Maxwell's theory that is pure topology, requiring no metric, from the part that carries all the dynamics and all the dependence on spacetime signature. The goal is to package the potential as a one-form, define the field strength as its exterior derivative, and show that $\mathrm{d}F=0$ and $\mathrm{d}\star F=\star J$ reproduce the four equations while exposing exactly where the metric lives.

Step 1. Package the electromagnetic potential as a one-form $A=A_\mu\,\mathrm{d}x^\mu$ and define the field strength as the two-form $F=\mathrm{d}A$ , so $F_{\mu\nu}=\partial_\mu A_\nu-\partial_\nu A_\mu$ . Why: $F$ is manifestly antisymmetric and manifestly closed by construction, which is the whole trick.

Step 2. The homogeneous pair is automatic: $\mathrm{d}F=\mathrm{d}(\mathrm{d}A)=0$ by $\mathrm{d}^2=0$ . In components this is $\partial_{[\lambda}F_{\mu\nu]}=0$ , which unpacks to $\nabla\cdot\mathbf B=0$ and Faraday's law. Why: these two Maxwell equations are a topological identity — they hold with no metric and no sources, purely because $F$ is exact.

Step 3. The sourced pair introduces the Hodge dual: $\mathrm{d}\star F=\star J$ , where $J=J_\mu\,\mathrm{d}x^\mu$ is the current one-form. This single equation unpacks to Gauss's law and the Ampère–Maxwell law. Why: the star is the only place the metric and the Lorentzian signature enter, so all the "dynamics" is quarantined into one operator.

Step 4. Charge conservation is now a free corollary: apply $\mathrm{d}$ to the sourced equation to get $\mathrm{d}\star J=\mathrm{d}(\mathrm{d}\star F)=0$ , i.e. $\nabla_\mu J^\mu=0$ . Why: continuity is not an extra assumption but a consequence of $\mathrm{d}^2=0$ acting on the sourced equation — conservation is built into the geometry.

Limiting cases

Forms specialize to a whole zoo of familiar objects as the degree and dimension are dialed:

Case	Result	Meaning
Degree zero	$\Lambda^0 M=C^\infty(M),\ \ \mathrm{d}f=\partial_\mu f\,\mathrm{d}x^\mu$	A 0-form is a smooth function; its exterior derivative is the gradient one-form.
Vector calculus in $\mathbb{R}^3$	$\mathrm{d}:\Lambda^0\!\to\!\Lambda^1\!\to\!\Lambda^2\!\to\!\Lambda^3\ \leftrightarrow\ \mathrm{grad},\mathrm{curl},\mathrm{div}$	With the Euclidean $\star$ , $\mathrm{d}^2=0$ encodes both $\mathrm{curl\,grad}=0$ and $\mathrm{div\,curl}=0$ .
Top degree	$\omega\in\Lambda^n M:\ \omega=f\,\mathrm{d}x^1\wedge\cdots\wedge\mathrm{d}x^n$	The unique kind of object that integrates over the whole manifold; $\mathrm{d}\omega=0$ automatically since degree $n+1>n$ .
Closed vs exact	$\mathrm{d}\omega=0$ (closed); $\omega=\mathrm{d}\eta$ (exact)	Exact $\Rightarrow$ closed by $\mathrm{d}^2=0$ ; the converse fails globally, and the failure is measured by $H^p_{\mathrm{dR}}(M)$ .
Flat contractible patch	Poincaré lemma: closed $\Rightarrow$ exact	On a star-shaped region every closed form is exact, so cohomology is trivial — holes are a global phenomenon.

Common trapThe exterior derivative needs Christoffel symbols to be a real tensor

This is the reflex to resist. Because $\partial_\mu$ of a general tensor is not a tensor, students instinctically reach for $\nabla_\mu=\partial_\mu+\Gamma$ to "fix" $\mathrm{d}\omega$ — but for forms the fix is unnecessary and, in fact, changes nothing. The connection terms in $\nabla_{[\mu_0}\omega_{\mu_1\cdots\mu_p]}$ are each symmetric in a pair of lower indices (torsion-free $\Gamma^\lambda_{\mu\nu}=\Gamma^\lambda_{\nu\mu}$ ) that is simultaneously being antisymmetrized, so they cancel exactly: $\nabla_{[\mu}\omega_{\nu\cdots]}=\partial_{[\mu}\omega_{\nu\cdots]}$ . Never decorate $\mathrm{d}\omega$ with Christoffel symbols. The converse warning matters too: $\mathrm{d}$ only makes sense on totally antisymmetric tensors — do not try to apply it to a symmetric tensor like the metric or a general $(0,2)$ -tensor, where the antisymmetrization would simply discard most of the object.

Worked examples

1 · Counting components and building the Maxwell 2-form (numeric)

On a 4-dimensional spacetime, tabulate how many independent components a $p$ -form has for each degree, check the total against $2^n$ , and use the count at $p=2$ to identify the electromagnetic field strength.

Step 1. Evaluate $\binom{4}{p}$ for $p=0,1,2,3,4$ . Directly: $\binom{4}{0}=1$ , $\binom{4}{1}=4$ , $\binom{4}{2}=\frac{4\cdot3}{2}=6$ , $\binom{4}{3}=4$ , $\binom{4}{4}=1$ . The counts are the symmetric row 1, 4, 6, 4, 1 — Pascal's row for $n=4$ .

Step 2. Sum them: $1+4+6+4+1=\mathbf{16}=2^4$ , confirming $\dim\Lambda^\bullet=2^n$ . The symmetry $\binom{4}{p}=\binom{4}{4-p}$ is the combinatorial fingerprint of Hodge duality, which pairs $\Lambda^p$ with $\Lambda^{4-p}$ of equal dimension.

Step 3. Focus on $p=2$ : a 2-form in 4D has exactly 6 independent components. That is precisely the count of a 2-form $F_{\mu\nu}=-F_{\nu\mu}$ : the three mixed time-space components $F_{0i}=E_i$ (electric field) and the three purely spatial components $F_{ij}=\epsilon_{ijk}B_k$ (magnetic field). So $\mathbf{3+3=6}$ : the field strength is a single geometric object, not two vectors that happen to travel together.

Step 4. Sanity-check the top degree: a 4-form has $\binom{4}{4}=1$ component, the volume element $\sqrt{|g|}\,\mathrm{d}x^0\wedge\mathrm{d}x^1\wedge\mathrm{d}x^2\wedge\mathrm{d}x^3$ , and a 5-form is impossible ( $5>4$ ), so $\mathrm{d}$ of any 4-form vanishes automatically. The whole exterior algebra of spacetime has 16 dimensions, and Maxwell's F occupies the 6-dimensional middle.

2 · The exterior derivative is the curl, and $\mathrm{d}^2=0$ by hand (numeric)

Take the rigid-rotation velocity field of the plane, written as the one-form $\alpha=-y\,\mathrm{d}x+x\,\mathrm{d}y$ on $\mathbb{R}^3$ . Compute $\mathrm{d}\alpha$ , read off the curl, then verify $\mathrm{d}^2=0$ on the function $f=xyz$ .

Step 1. Apply $\mathrm{d}$ using the graded Leibniz rule and $\mathrm{d}(\mathrm{d}x)=0$ : $\mathrm{d}\alpha=\mathrm{d}(-y)\wedge\mathrm{d}x+\mathrm{d}(x)\wedge\mathrm{d}y=(-\mathrm{d}y)\wedge\mathrm{d}x+\mathrm{d}x\wedge\mathrm{d}y$ . Only the derivative of the coefficient survives, wedged onto the basis form.

Step 2. Use antisymmetry $\mathrm{d}y\wedge\mathrm{d}x=-\mathrm{d}x\wedge\mathrm{d}y$ : the first term becomes $+\mathrm{d}x\wedge\mathrm{d}y$ , so $\mathrm{d}\alpha=\mathrm{d}x\wedge\mathrm{d}y+\mathrm{d}x\wedge\mathrm{d}y=\mathbf{2\,\mathrm{d}x\wedge\mathrm{d}y}$ . The coefficient $2$ is exactly the $z$ -component of $\nabla\times(-y,x,0)=(0,0,2)$ — the exterior derivative is the curl, and a rigid rotation has vorticity $2\Omega$ for unit angular speed.

Step 3. Now verify nilpotency directly. Take $f=xyz$ , so $\mathrm{d}f=yz\,\mathrm{d}x+xz\,\mathrm{d}y+xy\,\mathrm{d}z$ . Apply $\mathrm{d}$ again and collect one representative pair: the $\mathrm{d}x\wedge\mathrm{d}y$ coefficient is $\partial_x(xz)-\partial_y(yz)$ ... more carefully, from $yz\,\mathrm{d}x$ we get $\partial_y(yz)\,\mathrm{d}y\wedge\mathrm{d}x=z\,\mathrm{d}y\wedge\mathrm{d}x$ , and from $xz\,\mathrm{d}y$ we get $\partial_x(xz)\,\mathrm{d}x\wedge\mathrm{d}y=z\,\mathrm{d}x\wedge\mathrm{d}y$ .

Step 4. These cancel: $z\,\mathrm{d}y\wedge\mathrm{d}x+z\,\mathrm{d}x\wedge\mathrm{d}y=-z\,\mathrm{d}x\wedge\mathrm{d}y+z\,\mathrm{d}x\wedge\mathrm{d}y=0$ , and the same happens for every index pair. So $\mathrm{d}(\mathrm{d}f)=0$ — the mixed partials $\partial_x\partial_y f=\partial_y\partial_x f$ cancel under antisymmetrization, which is precisely $\nabla\times\nabla f=0$ .

Physics of the portrait

The exterior calculus did not arrive with the physicists who now use it daily; it was built by a geometer solving a problem in differential equations, decades before general relativity needed it.

Origins · 1 minÉlie Cartan and how this was discovered

ÉC

Élie Cartan

1869–1951 · founder of the exterior calculus and moving-frame methods in geometry

Élie Cartan is the architect of the machinery in this article. Born in 1869 to a village blacksmith in the French Alps, he rose through the École Normale Supérieure to become the deepest geometer of his generation. Between 1899 and 1902 he created the calculus of exterior differential forms, introducing both the wedge product (building on Grassmann's half-forgotten exterior algebra) and the exterior derivative $\mathrm{d}$ , precisely to give a coordinate-free formulation of Pfaffian systems of differential equations. What began as a tool for integrability conditions became, in his hands, the language of an entire subject: his method of moving frames recasts curvature as a matrix of two-forms through the structure equations $\mathrm{d}e^a+\omega^a{}_b\wedge e^b=0$ and $R^a{}_b=\mathrm{d}\omega^a{}_b+\omega^a{}_c\wedge\omega^c{}_b$ , which remain the fastest route to the Riemann tensor in general relativity. Cartan also classified the simple Lie groups, discovered spinors in 1913 (two decades before Dirac needed them), and developed the theory of symmetric spaces. For years his work was thought too abstract; it took de Rham's 1931 theorem, tying his forms to topology, and the physics of gauge fields to reveal that Cartan had quietly written down the grammar of modern geometry. When you compute $F=\mathrm{d}A$ or read $\mathrm{d}^2=0$ as a Bianchi identity, you are speaking Cartan's language.

History

1844

— Grassmann's Ausdehnungslehre introduces the alternating exterior product

1889

— Volterra and Poincaré establish that closed forms are locally exact

1899

— Élie Cartan founds the calculus of exterior differential forms

1931

— de Rham proves that cohomology is computable from differential forms

1963

— Flanders' textbook brings the exterior calculus to working physicists

Going further

References. Misner, Thorne & Wheeler, Gravitation (1973), Ch. 4 "Electromagnetism and Differential Forms" · Wald, General Relativity (1984), Ch. 2.7 and Appendix B · Carroll, Spacetime and Geometry (2004), Ch. 2.9–2.10 · Schutz, A First Course in General Relativity (2009), Ch. 4 · Spivak, Calculus on Manifolds (1965), Ch. 4–5 · Flanders, Differential Forms with Applications to the Physical Sciences (1963). The component conventions for the wedge, the exterior derivative, and the Hodge star are standard throughout, as are the dimension counts $\binom{n}{p}$ and $2^n$ (elementary combinatorics) and the Cartan structure equations. Key consistency points: the $\partial^2 x$ terms cancel in the coordinate-independence derivation by symmetry of mixed partials; the graded-anticommutativity sign is $(-1)^{pq}$ ; $\mathrm{d}$ and $\star$ reduce to grad, curl, and div in $\mathbb{R}^3$ ; $\mathrm{d}F=0$ and $\mathrm{d}\star F=\star J$ reproduce all four Maxwell equations; and the worked-example results are $\mathrm{d}\alpha=2\,\mathrm{d}x\wedge\mathrm{d}y$ and $\mathrm{d}(\mathrm{d}f)=0$ . No nonstandard numerical values appear.

Builds on: Cotangent Space and One-Forms · General Tensors and the Tensor Product · Tensor Algebra: Contraction, Symmetrization and Antisymmetrization · Tangent Vectors and the Tangent Space · Smooth Manifolds and Coordinate Charts. Leads to: Tensor Densities and Weights · The Metric Tensor and the Line Element · Metric Signature and Lorentzian Geometry.

Ask the tutor: Ask me to walk through building the curvature two-forms of the Schwarzschild metric with Cartan's structure equations, or to show how $\mathrm{d}\theta$ on the punctured plane makes de Rham cohomology tangible.

What is the difference between a differential form and a general tensor

A general covariant tensor of rank $p$ has arbitrary index symmetry and $n^p$ components. A $p$ -form is the special case that is totally antisymmetric — swapping any two indices flips the sign — which cuts the count to $\binom{n}{p}$ . That restriction is not a loss but the whole point: only antisymmetric components transform with the Jacobian determinant, which is what a coordinate-independent integrand needs. So forms are the tensors you can integrate, and the operations $\wedge$ and $\mathrm{d}$ keep you inside the antisymmetric world. A symmetric tensor like the metric is emphatically not a form and cannot be fed to $\mathrm{d}$ .

Why is there a factor 1/p! in front of a p-form

The factor prevents double counting. When you write $\omega=\frac{1}{p!}\omega_{\mu_1\cdots\mu_p}\,\mathrm{d}x^{\mu_1}\wedge\cdots\wedge\mathrm{d}x^{\mu_p}$ , the sum runs over all values of the indices, not just increasing ones. Because both $\omega_{\mu_1\cdots\mu_p}$ and the wedge of basis forms are totally antisymmetric, every distinct set of $p$ indices appears $p!$ times — once for each ordering — and all those copies are equal. Dividing by $p!$ collapses them to a single contribution. Equivalently, you can sum only over strictly increasing indices and drop the factor; the two conventions agree. The same combinatorics explains the $(p+q)!/(p!q!)$ in the wedge and the $(p+1)$ in $\mathrm{d}$ .

How does d unify grad, curl, and div

On $\mathbb{R}^3$ the exterior derivative acts along the chain $\Lambda^0\to\Lambda^1\to\Lambda^2\to\Lambda^3$ . Starting from a function (0-form), $\mathrm{d}f=\partial_i f\,\mathrm{d}x^i$ is the gradient. On a 1-form $\alpha=A_i\,\mathrm{d}x^i$ , $\mathrm{d}\alpha$ is a 2-form whose components are $\partial_i A_j-\partial_j A_i$ — the curl, once you use the Euclidean Hodge star to turn the 2-form back into a vector. On a 2-form, $\mathrm{d}$ gives a 3-form whose single component is the divergence. The two ways of composing these, $\mathrm{d}^2=0$ , are exactly $\nabla\times\nabla f=0$ and $\nabla\cdot(\nabla\times\mathbf A)=0$ . One operator, three classical derivatives, two vanishing identities.

What does it mean for a form to be closed but not exact

A form is closed if $\mathrm{d}\omega=0$ and exact if $\omega=\mathrm{d}\eta$ for some $\eta$ . Since $\mathrm{d}^2=0$ , every exact form is automatically closed. The interesting question is the converse, and it fails whenever the manifold has holes. On a plane with the origin removed, the angle one-form $\mathrm{d}\theta=(x\,\mathrm{d}y-y\,\mathrm{d}x)/(x^2+y^2)$ is closed everywhere it is defined, yet its integral around the origin is $2\pi\ne0$ , so it cannot be $\mathrm{d}$ of any globally defined function. The space of closed-modulo-exact forms is the de Rham cohomology $H^p_{\mathrm{dR}}(M)$ , a topological invariant you can compute with calculus.

Where exactly does the metric enter the theory of forms

Almost nowhere — and that is the headline. The wedge product, the exterior derivative, pullback, the notions of closed and exact, Stokes' theorem, and de Rham cohomology are all defined on a bare smooth manifold with no metric and no connection. The metric makes its single appearance through the Hodge star $\star$ , which needs the volume element $\sqrt{|g|}$ and the ability to raise indices to map a $p$ -form to an $(n-p)$ -form. This clean split is why Maxwell's equations factor into a metric-free topological half, $\mathrm{d}F=0$ , and a metric-dependent dynamical half, $\mathrm{d}\star F=\star J$ . All the geometry of length and signature is quarantined inside that one star.

Why can I not have a p-form with p greater than the dimension n

Because you would need $p$ linearly independent one-forms to wedge together, and at any point of an $n$ -manifold the cotangent space is only $n$ -dimensional. By the pigeonhole principle, any $n+1$ one-forms are linearly dependent, so at least two factors in the wedge repeat — and a wedge with a repeated one-form vanishes by antisymmetry, since $\mathrm{d}x^\mu\wedge\mathrm{d}x^\mu=0$ . Hence $\Lambda^p M=0$ for all $p>n$ . A concrete consequence: on 4D spacetime the top nonzero degree is $p=4$ (the volume form), every 5-form is identically zero, and $\mathrm{d}$ of any 4-form vanishes automatically because its target degree exceeds $n$ .

Differential Forms and the Exterior Algebra

Where this lives today

The physical picture

The core equations

Limiting cases

Worked examples

1 · Counting components and building the Maxwell 2-form (numeric)

2 · The exterior derivative is the curl, and $\mathrm{d}^2=0$ by hand (numeric)

Physics of the portrait

Going further

Before this

Where this leads

Differential Forms and the Exterior Algebra

Where this lives today

The physical picture

The core equations

Limiting cases

Worked examples

1 · Counting components and building the Maxwell 2-form (numeric)

2 · The exterior derivative is the curl, and d2=0\mathrm{d}^2=0d2=0 by hand (numeric)

Physics of the portrait

Going further

Before this

Where this leads

2 · The exterior derivative is the curl, and $\mathrm{d}^2=0$ by hand (numeric)