De Rham Cohomology

De Rham cohomology is a topological invariant of a smooth manifold built from differential forms:

H_{d R}^{k} (M) = ker d / im d

. Key result: the de Rham theorem identifies these analytically-defined groups with singular cohomology over

R

, so calculus on

M

detects its global topology (holes). Applies to: smooth manifolds, with rich generalizations to sheaf theory, Hodge theory, and physics. Who/when: Georges de Rham (1931), building on Poincaré, Cartan, and Stokes.

De Rham cohomology measures the failure of closed differential forms to be exact on a smooth manifold $M$ . A $k$ -form $ω$ is closed if $d ω = 0$ and exact if $ω = d η$ ; since $d^{2} = 0$ , every exact form is closed, and the quotient $H_{d R}^{k} (M) = Z^{k} / B^{k}$ records the nontrivial closed-but-not-exact forms. This matters because it converts a purely differential-geometric/analytic question into a topological invariant — the dimensions of these vector spaces (Betti numbers) count the "holes" of $M$ . The surprising fact, the de Rham theorem, is that this construction from analysis exactly reproduces the singular cohomology with real coefficients, a combinatorial-topological object. Thus integration of forms over cycles provides a perfect pairing between geometry and topology.

The Core Insight

The exterior derivative $d$ satisfies $d \circ d = 0$ , turning the spaces of differential forms into a cochain complex (the de Rham complex). Cohomology of this complex is a priori a delicate analytic quantity, yet it depends only on the smooth homotopy type of $M$ . Stokes' theorem, $\int_{M} d ω = \int_{\partial M} ω$ , is the bridge: it makes integration a pairing between closed forms and cycles that descends to a pairing on (co)homology. The whole subject is the systematic exploitation of $d^{2} = 0$ plus Stokes.

Formal Statement

The de Rham Complex

Let $M$ be a smooth $n$ -dimensional manifold. Let $Ω^{k} (M) = Γ (Λ^{k} T^{*} M)$ denote the $R$ -vector space of smooth differential $k$ -forms, with $Ω^{k} (M) = 0$ for $k < 0$ or $k > n$ . The exterior derivative is the unique family of $R$ -linear maps

d_{k} : Ω^{k} (M) ⟶ Ω^{k + 1} (M), d_{k + 1} \circ d_{k} = 0,

⟳existence/uniqueness from the axioms

d (df) = 0

, the Leibniz rule

d (α \land β) = d α \land β + (- 1)^{d e g α} α \land d β

, and agreement with the differential on functions.

In local coordinates $(x^{1}, \dots, x^{n})$ , writing $ω = \sum_{I} f_{I} d x^{I}$ with multi-index $I = (i_{1} < \dots < i_{k})$ and $d x^{I} = d x^{i_{1}} \land \dots \land d x^{i_{k}}$ ,

d ω = I \sum j = 1 \sum n \frac{\partial f _{I}}{\partial x ^{j}} d x^{j} \land d x^{I} .

⟳apply the Leibniz rule to each term and use

d (d x^{i}) = 0

; mixed partials cancel in

d^{2}

by symmetry of

\partial_{j} \partial_{k}

against antisymmetry of

d x^{j} \land d x^{k}

A form $ω$ is closed if $d ω = 0$ ; the space of closed $k$ -forms is $Z^{k} (M) = ker d_{k}$ . A form is exact if $ω = d η$ ; the space of exact $k$ -forms is $B^{k} (M) = im d_{k - 1}$ . Since $d^{2} = 0$ , $B^{k} \subseteq Z^{k}$ .

de Rham Cohomology Groups

The $k$ -th de Rham cohomology group of $M$ is the quotient $R$ -vector space

H_{d R}^{k} (M) = \frac{Z ^{k} ( M )}{B ^{k} ( M )} = \frac{ker ( d _{k} : Ω ^{k} \to Ω ^{k + 1} )}{im ( d _{k - 1} : Ω ^{k - 1} \to Ω ^{k} )} .

⟳the quotient is well-defined precisely because

im d_{k - 1} \subseteq ker d_{k}

, which is

d^{2} = 0

Its dimension is the $k$ -th Betti number $b_{k} (M) = dim_{R} H_{d R}^{k} (M)$ (finite when $M$ is compact, or more generally of finite type). The Euler characteristic is the alternating sum

χ (M) = k = 0 \sum n (- 1)^{k} b_{k} (M) .

⟳Euler–Poincaré principle — the alternating sum of dimensions of a finite cochain complex equals the alternating sum of its cohomology dimensions.

de Rham Theorem

For a smooth manifold $M$ , the integration pairing induces a natural isomorphism of graded $R$ -algebras between de Rham cohomology and singular cohomology with real coefficients:

I : H_{d R}^{k} (M) ≅ H_{sing}^{k} (M; R), I [ω] (σ) = \int_{σ} ω,

⟳define a cochain map from forms to smooth singular cochains by

ω \mapsto (σ \mapsto \int_{σ} ω)

; Stokes' theorem shows it is a cochain map; prove it induces an isomorphism on a basis of opens (cubes) and globalize via Mayer–Vietoris and the five lemma.

The algebra structure on $H_{d R}^{∙}$ is induced by the wedge product; on the singular side it is the cup product. In particular the Betti numbers $b_{k}$ are smooth homotopy invariants and equal the ranks of the integral singular homology groups.

Poincaré Lemma

If $U \subseteq R^{n}$ is open and star-shaped (or more generally contractible), then

H_{d R}^{k} (U) = {R, 0, k = 0, k \geq 1.

⟳construct a chain homotopy

K

with

d K + K d = id

on positive degrees, by integrating the radial homotopy

h_{t} (x) = t x

against the form; explicit

K ω = \int_{0}^{1} ι_{X} (h_{t}^{*} ω) \frac{d t}{t}

with

X

the radial Euler vector field.

This is the local triviality that makes de Rham cohomology a measure of global topology only.

Why It Holds / Derivation

Homotopy Invariance via the Chain Homotopy Operator

We sketch the central computational engine: smoothly homotopic maps induce equal maps on cohomology.

Step 1 — Setup. Let

f_{0}, f_{1} : M \to N

be smoothly homotopic via

H : M \times [0, 1] \to N

H (\cdot, 0) = f_{0}

H (\cdot, 1) = f_{1}

. Each

f_{i}^{*} : Ω^{k} (N) \to Ω^{k} (M)

is a cochain map (since pullback commutes with

d

), hence induces

f_{i}^{*} : H_{d R}^{k} (N) \to H_{d R}^{k} (M)

. Step 2 — The integration-along-the-fiber operator. On

M \times [0, 1]

, with

t

the coordinate on

[0, 1]

, every form decomposes as

α + d t \land β

where

α, β

contain no

d t

. Define

(K Θ)_{x} = \int_{0}^{1} β (x, t) d t, Θ = α + d t \land β .

⟳

K

integrates out the interval direction; it lowers form degree on

M

by one because it removes the

d t

factor and consumes one slot.

Step 3 — The homotopy identity. A direct computation using

d_{M \times I} = d_{M} + d t \land \partial_{t}

gives, for the inclusions

j_{t} : M ↪ M \times {t}

j_{1}^{*} - j_{0}^{*} = d_{M} K + K d_{M \times I} .

⟳split

d Θ

into

M

-part and

d t

-part, apply the fundamental theorem of calculus to the

\partial_{t}

term; the boundary values

t = 1, 0

produce

j_{1}^{*} - j_{0}^{*}

Step 4 — Conclusion. Pull back by

H

and compose:

f_{1}^{*} - f_{0}^{*} = H^{*} (j_{1}^{*} - j_{0}^{*}) = d (K H^{*}) + (K H^{*}) d

. On closed forms the right side is exact, so

f_{1}^{*} = f_{0}^{*}

H_{d R}^{k}

■

Corollary. Contractible spaces have the cohomology of a point (the Poincaré lemma), and homotopy-equivalent manifolds have isomorphic de Rham cohomology.

Mayer–Vietoris Sequence

The computational workhorse. Let $M = U \cup V$ with $U, V$ open.

Step 1. Restriction maps assemble into a short exact sequence of de Rham complexes

0 \to Ω^{∙} (M) (i_{U}^{*}, i_{V}^{*}) Ω^{∙} (U) \oplus Ω^{∙} (V) j_{U}^{*} - j_{V}^{*} Ω^{∙} (U \cap V) \to 0.

⟳injectivity is obvious; surjectivity uses a partition of unity

{ρ_{U}, ρ_{V}}

subordinate to

{U, V}

to split a form on

U \cap V

into pieces extendable to

U

and

V

Step 2. A short exact sequence of cochain complexes yields a long exact sequence in cohomology:

\dots \to H_{d R}^{k} (M) \to H_{d R}^{k} (U) \oplus H_{d R}^{k} (V) \to H_{d R}^{k} (U \cap V) δ H_{d R}^{k + 1} (M) \to \dots

⟳the connecting homomorphism

δ

comes from the snake lemma; concretely lift a closed form on

U \cap V

, apply

d

, and recognize the result as coming from

M

Iterating Mayer–Vietoris from contractible pieces (a good cover) computes $H_{d R}^{∙}$ of any manifold. $■$

Georges de Rham | 1903–1990

Swiss mathematician at the universities of Lausanne and Geneva. In his 1931 thesis Sur l'analysis situs des variétés à $n$ dimensions, de Rham proved the theorem now bearing his name: that the periods of closed forms over cycles determine, and are determined by, the topology of the manifold — confirming a conjecture implicit in the work of Poincaré and Élie Cartan. He introduced currents (a unification of forms and chains via distribution theory), developed the theory of harmonic forms with Kodaira, and shaped the modern view in which analysis, geometry, and topology are inseparable. His characterization of which period-systems are realizable (the de Rham theorem proper) closed a chapter Poincaré had opened in the 1890s.

Building Intuition

Think of a closed form as a "flow with no local sources or sinks" and an exact form as one that is "globally a gradient/potential." On $R^{2}$ , the field $\frac{- y d x + x d y}{x ^{2} + y ^{2}}$ has zero curl (closed) yet is not the differential of a single-valued function — its integral around the origin is $2 π$ , not $0$ . That nonzero period is the hole at the origin talking back to you. De Rham cohomology systematizes exactly this: $H^{k}$ counts the independent ways a closed form can fail to be exact, and each such failure is "detected" by integrating over a $k$ -cycle that wraps a $k$ -dimensional hole. Degree $0$ counts connected components ( $b_{0}$ = number of pieces, since a closed $0$ -form is a locally constant function). The dimensions $b_{k}$ are rigid topological data, even though forms themselves can be wiggled freely.

The cohomology hole on the punctured plane: a closed 1-form with nonzero period detects the missing point.

<svg viewBox="0 0 460 320" xmlns="http://www.w3.org/2000/svg">
<rect width="460" height="320" fill="#0a0e17"/>
<text x="230" y="26" fill="#e2e8f0" font-size="15" text-anchor="middle">H¹(ℝ²∖{0}) ≅ ℝ, generated by [dθ]</text>
<!-- axes -->
<line x1="60" y1="180" x2="400" y2="180" stroke="#94a3b8" stroke-width="1"/>
<line x1="230" y1="60" x2="230" y2="300" stroke="#94a3b8" stroke-width="1"/>
<!-- puncture -->
<circle cx="230" cy="180" r="4" fill="#0a0e17" stroke="#e2e8f0" stroke-width="1.5"/>
<text x="238" y="196" fill="#e2e8f0" font-size="12">removed point 0</text>
<!-- loop (key curve) -->
<circle cx="230" cy="180" r="80" fill="none" stroke="#d4a853" stroke-width="2.5"/>
<!-- arrow on loop -->
<polygon points="312,176 300,170 300,182" fill="#d4a853"/>
<text x="318" y="150" fill="#d4a853" font-size="12">cycle γ</text>
<!-- vector field hints -->
<g stroke="#94a3b8" stroke-width="1.2">
<line x1="230" y1="110" x2="248" y2="118"/><polygon points="248,118 240,113 244,123" fill="#94a3b8"/>
<line x1="300" y1="180" x2="296" y2="200" /><polygon points="296,200 290,192 302,194" fill="#94a3b8"/>
<line x1="230" y1="250" x2="212" y2="242"/><polygon points="212,242 220,247 216,237" fill="#94a3b8"/>
<line x1="160" y1="180" x2="164" y2="160"/><polygon points="164,160 170,168 158,166" fill="#94a3b8"/>
</g>
<text x="60" y="300" fill="#e2e8f0" font-size="12">∮_γ dθ = 2π ≠ 0  ⇒  dθ closed but not exact</text>
</svg>

Worked Examples

The Circle $S^1$

Given: the

1

-dimensional manifold

S^{1}

, parametrized by angle

θ \in [0, 2 π)

. Compute

H_{d R}^{k} (S^{1})

. Step 1 — Degree 0. Closed

0

-forms are locally constant functions;

S^{1}

is connected, so

Z^{0} = R

B^{0} = 0

. Thus

b_{0} = 1

. Step 2 — Degree 1. Every

1

-form on

S^{1}

f (θ) d θ

, automatically closed (top degree). Exact forms are

d g = g^{'} (θ) d θ

. The form

f d θ

is exact iff

f

has a

2 π

-periodic antiderivative, i.e. iff

\int_{0}^{2 π} f d θ = 0

. Step 3 — The isomorphism. The map

[f d θ] \mapsto \int_{0}^{2 π} f d θ

is a linear isomorphism

H_{d R}^{1} (S^{1}) \to R

, with

d θ

itself representing the generator (integral

2 π

). Result:

H_{d R}^{0} (S^{1}) = R, H_{d R}^{1} (S^{1}) = R, b_{0} = b_{1} = 1, χ (S^{1}) = 1 - 1 = 0.

The 2-Sphere $S^2$ via Mayer–Vietoris

Given:

S^{2} = U \cup V

U, V

open disks (each contractible,

≅ R^{2}

) overlapping in a band

U \cap V ≃ S^{1}

. Compute

H_{d R}^{k} (S^{2})

. Step 1 — Inputs.

H^{∙} (U) = H^{∙} (V) = (R, 0, 0)

in degrees

0, 1, 2

;

H^{∙} (U \cap V) = H^{∙} (S^{1}) = (R, R, 0)

. Step 2 — The sequence (low degrees).

0 \to H^{0} (S^{2}) \to R^{2} \to R \to H^{1} (S^{2}) \to 0 \to R δ H^{2} (S^{2}) \to 0.

S^{2}

is connected:

H^{0} (S^{2}) = R

. The map

R^{2} \to R

(restriction difference) sends

(a, b) \mapsto a - b

, rank

1

, so its image is all of

R

, forcing

H^{1} (S^{2}) = 0

. Step 3 — Top degree. Exactness gives

H^{2} (S^{2}) ≅ R

via the connecting map

δ

from

H^{1} (S^{1}) = R

. Result:

H^{0} (S^{2}) = R, H^{1} (S^{2}) = 0, H^{2} (S^{2}) = R; b_{0} = 1, b_{1} = 0, b_{2} = 1, χ (S^{2}) = 2.

This matches

χ = 2

from the Euler formula

V - E + F

for a polyhedron.

Common Misconceptions

"Closed implies exact." False in general — that is the entire point of cohomology. It holds locally (Poincaré lemma) but can fail globally. Conflating local and global exactness is the single most common error. Likewise, " $d ω = 0$ means $ω$ is constant" is wrong for

k \geq 1

; closedness is a differential condition, not a constancy condition. "De Rham cohomology sees torsion." It does not. Because coefficients are

R

(a field of characteristic

0

), de Rham cohomology equals singular cohomology with real coefficients, which kills all torsion. For example,

H^{1} (RP^{2}; Z) = Z /2

but

H_{d R}^{1} (RP^{2}) = 0

. "Betti numbers determine the manifold." No — they are coarse invariants. Lens spaces and other examples can share Betti numbers yet be non-homeomorphic.

The Punctured Plane and the "Angle Form"

On $M = R^{2} ∖ {0}$ , set

ω = \frac{- y d x + x d y}{x ^{2} + y ^{2}} = d θ (locally) .

⟳in polar coordinates

ω = d θ

; but

θ

is only defined modulo

2 π

, so it is not a globally single-valued function on

M

Then $d ω = 0$ (direct computation: $\partial_{x} \frac{x}{x ^{2} + y ^{2}} = \partial_{y} \frac{- y}{x ^{2} + y ^{2}}$ ), so $ω$ is closed. Yet $\oint_{S^{1}} ω = 2 π \neq = 0$ , so $ω$ cannot be exact (an exact form integrates to $0$ over any cycle by Stokes). This single form generates $H_{d R}^{1} (R^{2} ∖ {0}) ≅ R$ , demonstrating that removing one point creates a nontrivial degree-1 class. It is the canonical refutation of "closed $\Rightarrow$ exact."

Connections

De Rham cohomology sits at a crossroads of several theories:

Singular Cohomology — the de Rham theorem makes $H_{d R}^{∙} (M) ≅ H^{∙} (M; R)$ as rings, linking analysis to algebraic topology.
Hodge Theory — on a compact oriented Riemannian manifold, each de Rham class has a unique harmonic representative ( $Δ ω = 0$ ), giving $H_{d R}^{k} ≅ H^{k}$ and, on Kähler manifolds, the Hodge decomposition $H_{C}^{k} = ⨁_{p + q = k} H^{p, q}$ .
Sheaf Cohomology — the de Rham complex is a fine resolution of the constant sheaf $\underline{R}$ , so $H_{d R}^{k} (M) ≅ H^{k} (M; \underline{R})$ ; this is the conceptual proof of the de Rham theorem.
Poincaré Duality — on a compact oriented $n$ -manifold, the wedge-and-integrate pairing $H_{d R}^{k} \times H_{d R}^{n - k} \to R$ is perfect, so $b_{k} = b_{n - k}$ .
Chern–Weil Theory — characteristic classes of vector bundles are constructed as de Rham classes of curvature polynomials.
Čech Cohomology — agrees with de Rham via the Čech–de Rham double complex.

Real-World Applications

Electromagnetism (Maxwell's equations). With the field strength $2$ -form $F$ , the homogeneous equations are $d F = 0$ ; whether $F = d A$ globally (existence of a vector potential) is a $H_{d R}^{2}$ question — the obstruction underlies the Aharonov–Bohm effect and Dirac's monopole quantization.
Gauge theory & instantons. Topological charges of Yang–Mills fields are integrals of curvature characteristic forms (Chern–Weil Theory), classifying field configurations by de Rham/cohomology data.
Condensed matter / topological insulators. Berry curvature is a closed $2$ -form; its integral (Chern number) is quantized and explains the integer quantum Hall conductance.
Fluid dynamics & topology of flows. Helicity and the linking of vortex lines are expressed via closed forms; conservation laws follow from cohomological invariance.
Robotics & control. Holonomy and the non-existence of global "phase" functions in configuration spaces with holes are detected by $H_{d R}^{1}$ .
Numerical analysis (FEEC). Finite Element Exterior Calculus discretizes the de Rham complex to build stable, structure-preserving solvers for PDEs in electromagnetics and elasticity.
Data analysis. Persistent cohomology (a discrete cousin) uses the same closed/exact philosophy to detect circular coordinates and holes in point-cloud data.

Historical Development

1750s–1850s — Euler's polyhedron formula $V - E + F = 2$ and later Riemann's notion of connectivity foreshadow topological invariants computable from "holes."
1895 — Henri Poincaré, Analysis Situs, introduces homology and Betti numbers and studies periods of differentials, implicitly conjecturing the form–cycle duality.
1899–1900 — Élie Cartan develops the exterior calculus of differential forms and the exterior derivative $d$ , providing the language.
1922 — Stokes' theorem in its general $n$ -dimensional form is fully articulated within Cartan's calculus, supplying the pairing $\int_{\partial} = \int d$ .
1931 — Georges de Rham's thesis proves the de Rham theorem: closed forms modulo exact forms are isomorphic to (real) homology via periods.
1935–1943 — W. V. D. Hodge develops harmonic forms; de Rham–Kodaira put the analysis (elliptic theory of the Laplacian) on firm footing, yielding the Hodge theorem.
1945–1950 — Cartan, Leray, Weil reinterpret de Rham cohomology via sheaves and the Čech–de Rham bridge; Weil's note clarifies the comparison.
1958 — Bott and others develop the modern Mayer–Vietoris/good-cover computational toolkit (later canonized in Bott–Tu, 1982).
1960s onward — Generalizations: algebraic de Rham cohomology (Grothendieck), crystalline cohomology, and infinite-dimensional analogues extend the philosophy across mathematics.

Subtleties & Limits

Where de Rham cohomology is delicate or incomplete:

Loss of torsion. Real coefficients discard $Z$ -torsion, so de Rham cohomology cannot distinguish spaces that differ only torsionally. Recovering integral information requires integrality of periods (the de Rham theorem's "period lattice") or other theories.
Non-smooth and singular spaces. The classical theory requires smoothness. Extending it to singular varieties or stratified spaces motivates intersection cohomology, L²-cohomology, and analytic compactification questions that remain technically subtle.
Algebraic comparison. Grothendieck's algebraic de Rham cohomology over fields of characteristic $0$ matches the classical theory (Grothendieck's comparison theorem), but in positive characteristic the naive de Rham complex misbehaves — fixing this led to crystalline cohomology and the still-active $p$ -adic Hodge theory program (period rings, comparison isomorphisms).
Infinite dimensions. For infinite-dimensional manifolds (loop spaces, mapping spaces), exactness of the Poincaré lemma and finiteness of Betti numbers can fail; convenient calculus and homotopical methods are needed.
Effective computation. While Mayer–Vietoris is conceptually complete, producing explicit harmonic representatives or period matrices for general manifolds is computationally hard and an active area in numerical geometry.