Differential Forms

What a differential form is

A k-form is a smooth, point-dependent rule that takes k tangent vectors and returns a real number — multilinearly and antisymmetrically. The output flips sign if you swap any two inputs.

Concretely, on R³ with coordinates (x, y, z):

• 0-form — a smooth function f(x, y, z).
• 1-form — α = P dx + Q dy + R dz; pairs with a tangent vector to produce a number. Examples: df = ∂f/∂x dx + ∂f/∂y dy + ∂f/∂z dz.
• 2-form — β = A dy∧dz + B dz∧dx + C dx∧dy; pairs with two tangent vectors to produce a signed area-like number.
• 3-form — γ = f dx∧dy∧dz; pairs with three vectors to give a signed volume.

On an n-manifold there are no k-forms for k > n — you cannot antisymmetrise more vectors than the dimension allows.

Wedge product — antisymmetric multiplication

The wedge ∧ multiplies forms while remembering orientation:

dx ∧ dy = −dy ∧ dx        (so dx ∧ dx = 0)
α ∧ β = (−1)^(deg α · deg β) β ∧ α

A 1-form wedge a 1-form gives a 2-form. A 1-form wedge a 2-form gives a 3-form. The wedge is the multiplication of the exterior algebra Λ*T*M.

Geometric meaning: (dx∧dy)(v, w) = v_x·w_y − v_y·w_x — the signed area of the parallelogram spanned by v and w projected to the xy-plane. Antisymmetry is exactly the parallelogram-area sign rule.

Exterior derivative d

The exterior derivative d sends k-forms to (k+1)-forms:

d(f)        = ∂f/∂x dx + ∂f/∂y dy + ∂f/∂z dz                     (gradient as 1-form)
d(P dx + Q dy)
            = (∂Q/∂x − ∂P/∂y) dx ∧ dy                              (2D curl)
d(A dy∧dz + B dz∧dx + C dx∧dy)
            = (∂A/∂x + ∂B/∂y + ∂C/∂z) dx∧dy∧dz                    (divergence)

So d unifies grad, curl, and div. The single rule that picks the right one is "increase the degree by 1; respect antisymmetry; use the product rule on wedges."

The killer property: d² = 0. Applying d twice to anything gives zero, because mixed partials commute. In vector-calculus language, d² = 0 simultaneously says curl(grad f) = 0 and div(curl F) = 0.

Stokes' theorem — one identity, four classical theorems

For an oriented k-manifold M with boundary ∂M and a (k−1)-form ω:

∫_∂M ω = ∫_M dω

Dimension	Classical name	Form
1D	Fundamental theorem of calculus	f(b) − f(a) = ∫_a^b f'(x) dx
2D (planar)	Green's theorem	∮ P dx + Q dy = ∬ (∂Q/∂x − ∂P/∂y) dA
2D (in R³)	Classical Stokes	∮ F·dr = ∬ (∇×F)·dS
3D	Divergence theorem	∯ F·dS = ∭ (∇·F) dV
n-D	General Stokes (Cartan)	∫_∂M ω = ∫_M dω
arbitrary	Manifolds with corners	same identity, smooth boundary pieces

Historical context

Hermann Grassmann published Die lineare Ausdehnungslehre in 1844 — the first systematic exterior algebra. Few read it; the second edition (1862) helped but Grassmann died in 1877 mostly unrecognised in mathematics. Henri Poincaré used 1-forms extensively for Hamiltonian mechanics in the 1890s.

Élie Cartan rebuilt the theory from 1899 onward. His 1922 monograph on integral invariants codified the exterior derivative and the modular Stokes identity. Cartan applied forms to Lie groups, general relativity, and what would become gauge theory; his student de Rham (1931) proved that the algebraic invariants of forms — closed mod exact — reproduce the singular cohomology of the underlying manifold.

By 1953 Chern's work and Atiyah–Singer (1963) had made forms the standard language of modern geometry. Today they are the integrands of every gauge field theory (Yang–Mills, Maxwell, gravity in tetrad formalism) and the foundation of the integrals in the AdS/CFT correspondence.

Worked example — a 1-form and its derivative

Let ω = x dy − y dx on R². Compute dω:

dω = dx ∧ dy − dy ∧ dx
   = dx ∧ dy + dx ∧ dy            (using dy ∧ dx = − dx ∧ dy)
   = 2 dx ∧ dy

Integrate dω over the unit disk D — ∬_D 2 dA = 2π. By Stokes, ∫_∂D ω = 2π too. Parametrise ∂D by (cos t, sin t): ω = cos t · cos t dt − sin t · (−sin t) dt = (cos² t + sin² t) dt = dt — integrating from 0 to 2π yields 2π. Stokes checks out.

Why differential forms matter

• Gauge theory. Yang–Mills connections are Lie-algebra-valued 1-forms; the field strength F = dA + A∧A is a 2-form. The action ∫ tr(F∧*F) defines QCD and the electroweak theory.
• General relativity (Cartan formalism). The metric is encoded by an orthonormal tetrad e^a — a set of 1-forms — and the connection by a 1-form ω^{ab}. The curvature 2-form Ω = dω + ω∧ω plugs into Einstein's equation in coordinate-free form.
• De Rham cohomology. Closed forms (dω = 0) modulo exact (ω = dη) reproduce topological invariants. The Mayer–Vietoris sequence, Poincaré duality, and the index theorem live here.
• Symplectic geometry. Classical mechanics phase space is a symplectic manifold (M, ω) with ω a closed non-degenerate 2-form. Hamilton's equations are dH = ι_X ω. Geometric quantisation begins here.
• Maxwell's equations in 2 lines. With F a 2-form on spacetime, the source equations are dF = 0 and d*F = J — replacing four vector-calculus equations.
• Numerical PDEs. Finite-element exterior calculus (FEEC, Arnold–Falk–Winther 2006) uses forms to build provably stable mixed methods for electromagnetics and elasticity.

Closed, exact, and topology

A form is closed if dω = 0; exact if ω = dη for some η. Every exact form is closed (because d² = 0), but the converse fails on topologically interesting spaces.

Classic example — on R²\{0}, the 1-form ω = (x dy − y dx)/(x² + y²) is closed but not exact. Its integral around the unit circle is 2π, not 0 — which Stokes would force if ω = dη. The obstruction is the puncture at the origin: H¹(R²\{0}) = R, generated by ω.

This is how forms see topology: closed-mod-exact = de Rham cohomology = singular cohomology of the manifold. The dimension of H^k(M) counts k-dimensional "holes" detected analytically.

JavaScript — 2-form on a parallelogram

// (dx ∧ dy)(v, w) = v.x * w.y - v.y * w.x — signed parallelogram area
function wedgeDxDy(v, w) {
  return v.x * w.y - v.y * w.x;
}

const v = { x: 2, y: 1 };
const w = { x: 1, y: 3 };

console.log(wedgeDxDy(v, w));  //  5  (oriented area, v then w)
console.log(wedgeDxDy(w, v));  // -5  (swap inputs, sign flips)
console.log(wedgeDxDy(v, v));  //  0  (degenerate parallelogram)

// Exterior derivative of f(x, y) = x*y as a 1-form pulled back to a curve
// df = y dx + x dy
function df(p) { return { dx: p.y, dy: p.x }; }
// Verify d²f = 0:  d(y dx + x dy) = (1 - 1) dx ∧ dy = 0
console.log('d^2 f =', 1 - 1);  // 0

// Stokes' theorem check: ω = -y dx + x dy around unit circle = 2π = area · 2
function lineIntegralOmega(N) {
  let total = 0;
  for (let i = 0; i < N; i++) {
    const t = 2 * Math.PI * i / N;
    const dt = 2 * Math.PI / N;
    const x = Math.cos(t), y = Math.sin(t);
    const dx = -Math.sin(t) * dt, dy = Math.cos(t) * dt;
    total += -y * dx + x * dy;
  }
  return total;
}
console.log(lineIntegralOmega(10000).toFixed(4));  // 6.2832 ≈ 2π
// dω = 2 dx∧dy, integrated over unit disk = 2π ✓

Variants and generalisations

• Hodge star *. On an oriented Riemannian n-manifold, * sends k-forms to (n−k)-forms. Lets you define the codifferential δ = ±*d* and the Hodge Laplacian Δ = dδ + δd. The Hodge decomposition theorem (1934) writes any form uniquely as exact + co-exact + harmonic.
• Vector-valued forms. Sections of Λ^k T*M ⊗ E for a vector bundle E. Yang–Mills connections live here; the covariant exterior derivative replaces d.
• Currents. De Rham's "generalised forms" dual to smooth forms — they include things like the delta function δ_p as a 0-current. Foundation of geometric measure theory (Federer 1969).
• Complex (p,q)-forms. On a complex manifold, d splits as ∂ + ∂̄ with ∂² = ∂̄² = ∂∂̄ + ∂̄∂ = 0. Dolbeault cohomology refines de Rham.
• L-infinity / supergeometry. Forms valued in graded super-Lie algebras describe BRST cohomology and the cohomology of moduli spaces.
• Discrete exterior calculus (DEC). Forms on simplicial complexes — k-forms become k-cochains. Crane, Hirani, Desbrun popularised it for computational geometry.

Common misconceptions

• "Forms are just vectors with a hat on." 1-forms live in the cotangent space and pair with vectors; they transform covariantly under coordinate change while vectors transform contravariantly. Equating them requires a metric (musical isomorphism).
• "dx is an infinitesimal." In nonstandard analysis it can be; in modern differential geometry dx is a perfectly defined 1-form — the dual basis element to ∂/∂x. The pre-Cartan "infinitesimal" reading survives only in informal calculations.
• "d is just the gradient." Only on 0-forms (functions). On 1-forms d is the curl; on 2-forms in R³ it is the divergence. The point of d is the unification.
• "d² = 0 is trivial." It is algebraically forced by Clairaut's theorem, but it is a deep fact: it is exactly what makes de Rham cohomology a well-defined invariant. Without d² = 0, no cohomology — and topology becomes invisible to forms.
• "You can integrate a function over a surface." Only after choosing a volume form, which requires orientation and (usually) a metric. A k-form is the coordinate-free object you actually integrate over a k-manifold; functions alone do not have invariant integrals.
• "Stokes' theorem requires R³." The classical version does. Cartan's version works on any oriented smooth n-manifold with boundary, including non-orientable settings via twisted forms. It is the deepest single identity in differential calculus.