De Rham cohomology is a method for extracting global topological information from the differential calculus of a smooth manifold. Its basic objects are differential forms, which generalize functions, line-integral expressions, and volume elements. A form is called closed when its exterior derivative vanishes, and exact when it is itself an exterior derivative. De Rham cohomology measures the difference between these conditions: every exact form is closed, but a closed form need not possess a globally defined primitive.
The central result, de Rham’s theorem, identifies these spaces of differential forms with singular cohomology with real coefficients. Consequently, calculations involving derivatives and integrals can detect features such as independent loops, higher-dimensional cycles, and orientability. The theory connects differential geometry with algebraic topology, while also supplying tools for differential equations, complex geometry, characteristic classes, and mathematical physics.
The de Rham complex and its cohomology
Let be a smooth manifold of dimension , conventionally assumed Hausdorff and second countable. Write for the real vector space of smooth differential -forms on . In degree zero, . In local coordinates , a -form has the expression
where the coefficients are smooth functions and the wedge product is alternating. In particular, exchanging two adjacent coordinate differentials changes the sign.
The exterior derivative is a linear operator
For a function , it is the differential
On higher-degree forms, it differentiates the coefficient functions and satisfies the graded product rule
Its decisive identity is , which follows locally from equality of mixed partial derivatives and the alternating nature of the wedge product.
These operators form the de Rham cochain complex:
The spaces of closed and exact -forms are
Since , the inclusion holds. The -th de Rham cohomology group is the quotient vector space
Despite the traditional word “group,” it naturally carries the richer structure of a real vector space.
A class consists of all closed forms differing from by an exact form. Thus precisely when for a globally defined form . No metric, orientation, or embedding is required for this definition. Forms of degree greater than vanish, so for . In degree zero there are no nonzero exact forms, and closed functions are locally constant. Consequently, when is connected; more generally, it is the space of real-valued functions on the set of connected components.
Local exactness and global obstructions
The Poincaré lemma states that every closed differential form of positive degree on a star-shaped open subset of Euclidean space is exact. Since sufficiently small coordinate neighborhoods can be chosen to correspond to balls, every closed positive-degree form on a smooth manifold is locally exact. De Rham cohomology therefore detects a global obstruction rather than a local failure of differentiation.
A concrete mechanism behind the lemma is contraction toward a point. If is star-shaped about the origin, define . Integration in the parameter produces an operator satisfying
where is the constant map to the origin. For , the pullback vanishes on -forms. Thus a closed form satisfies . For a 1-form , a corresponding primitive is
Closedness is the compatibility condition ensuring that .
The punctured plane shows why this local conclusion need not extend globally. On , consider
Direct calculation gives . Locally, is the differential of an angle coordinate. Nevertheless, no globally defined smooth real-valued angle function exists. For the positively oriented unit circle,
An exact form has integral zero around every closed smooth curve, so this nonzero period proves that is not exact.
The same distinction appears in vector calculus. A curl-free vector field may have local scalar potentials without having one global potential. On a connected, simply connected smooth manifold, every closed 1-form is exact: integration from a fixed base point defines a potential independent of the chosen path. Simple connectedness does not, however, force all higher-degree cohomology groups to vanish. The 2-sphere is simply connected but has nonzero second de Rham cohomology.
A partition of unity permits smooth local constructions to be assembled globally, but does not automatically assemble local primitives into a primitive. If locally, then
The extra term records the incompatibility of the local choices. On overlaps, the differences are locally constant; their pattern encodes the obstruction that the cohomology class measures.
Integration and de Rham’s theorem
The bridge to topology is integration over chains. A smooth singular -simplex in is a suitably smooth map , where is the standard simplex. A differential -form defines a cochain by
Extending linearly assigns a number to every smooth singular -chain. The Stokes theorem gives
so integration intertwines exterior differentiation with the singular cochain differential.
Two consequences explain why this construction descends to cohomology. If is closed, its integral over a boundary is zero. If changes by an exact form, its integral over a cycle is unchanged. Integration therefore defines a pairing
where homology records cycles modulo boundaries. Integrals of this kind are called periods.
De Rham’s theorem states that integration induces a natural isomorphism
with singular cohomology on the right. Smooth singular chains compute the same homology as ordinary continuous singular chains on a smooth manifold, so the smoothness needed for integration does not change the resulting topological invariant. An important equivalent consequence is that a closed form is exact if and only if all its periods over smooth cycles vanish. In fact, over the real numbers the pairing identifies de Rham cohomology with the algebraic dual of real homology.
Georges de Rham established the comparison in his 1931 doctoral work. It gave a rigorous general formulation of connections between multiple integrals and topology developed in the work of Henri Poincaré and Élie Cartan. Later proofs recast the theorem using local exactness, partitions of unity, and sheaf theory. These approaches emphasize that smooth differential forms provide a flexible local resolution of the constant real coefficients underlying topological cohomology.
The coefficient field matters. Real cohomology does not detect torsion in integral homology. For example, real projective 2-space has first integral homology , but its first de Rham cohomology is zero. Its positive-degree real cohomology vanishes entirely. Thus differential forms detect continuous real-valued period data, not every integral topological distinction. Conversely, their real coefficients make the theory particularly compatible with integration, differential equations, and linear analysis.
Basic examples and the cohomology ring
For a point, the only nonzero group is . The same is true for every contractible smooth manifold, including Euclidean space . Contractibility is a global condition here: local resemblance to Euclidean space does not imply that a manifold has trivial higher cohomology.
For the circle, one has
A normalized representative of the degree-one generator is the restriction of
whose integral around the circle is one. The 1-form commonly written is globally defined on the circle even though is not a globally defined real-valued function there. This notation is useful but must not be mistaken for an assertion that the form is globally exact.
For , the sphere has
An oriented volume form normalized to total integral one represents its top-degree generator. The punctured space , for , retracts onto and therefore has the same cohomology. Its nontrivial positive-degree class can be interpreted through flux around the missing point.
The wedge product descends to cohomology:
This is well-defined because wedging a closed form with an exact form gives an exact form, up to the sign dictated by degree. The resulting algebra
is graded-commutative:
for classes of degrees and . Under de Rham’s isomorphism, this multiplication agrees with the cup product in singular cohomology.
The -torus illustrates this algebra explicitly. The translation-invariant forms descend to the quotient, and
Hence its degree- dimension is . These dimensions are the Betti numbers. For a closed oriented surface of genus , the Betti numbers are , and multiplication in degree one records the intersection pairing of loops. More generally, Betti numbers alone do not determine the cohomology ring: complex projective 2-space and have the same real cohomology dimensions, but the square of a degree-two generator is nonzero only in the former.
Homotopy invariance and computational methods
A smooth map pulls differential forms backward:
Because pullback commutes with , it induces
The direction reverses: a map between manifolds produces a map from the cohomology of its target to that of its source. Pullback also preserves wedge products, so de Rham cohomology is a contravariant construction taking smooth manifolds to graded-commutative real algebras.
Its dependence is actually topological rather than specifically differential. If two smooth maps are joined by a smooth homotopy, integration along the homotopy parameter gives an operator satisfying
Their induced maps on cohomology are therefore equal. Together with de Rham’s theorem and smooth approximation, this gives invariance under ordinary homotopy equivalence. A deformation retraction can consequently simplify a calculation dramatically: an annulus has the cohomology of a circle, and a Möbius strip also has the cohomology of its central circle, despite their different orientability.
For a cover by open subsets, the Mayer–Vietoris sequence relates the cohomology of the pieces:
It arises from a short exact sequence of complexes of forms. The map from forms on and to forms on takes the difference of their restrictions; a partition of unity establishes the required surjectivity. Covering a sphere by two contractible enlarged hemispheres reduces its positive-degree cohomology to that of an equatorial sphere one dimension lower.
For products, the Künneth theorem gives, under standard finite-type hypotheses,
The class corresponding to is represented by
Over a field there are no torsion correction terms of the kind appearing in integral versions. Applying this formula repeatedly computes the torus from the circle.
These methods often avoid solving differential equations for representatives. A finite good cover, whose nonempty finite intersections are contractible, reduces the computation to combinatorial overlap information. Alternatively, an explicit basis of cycles and a collection of closed forms can be compared by the period matrix . An invertible period matrix proves linear independence of the represented classes and, when the dimension is known, identifies a basis.
Orientation, compact support, and duality
Integration of top-degree forms requires an orientation, but de Rham cohomology itself does not. If is connected, compact, oriented, and without boundary, with dimension , integration induces an isomorphism
Stokes’ theorem ensures that exact -forms integrate to zero. The converse assertion—that every top form of integral zero is exact—is a substantial global consequence of the theory.
More generally, Poincaré duality states that the pairing
is nondegenerate for a closed oriented manifold. Here “closed manifold” means compact and without boundary. These cohomology groups are finite-dimensional, and each is the dual vector space of the other. In particular,
For a connected closed nonorientable manifold, ordinary top-degree real cohomology is zero; orientation-twisted coefficients provide the appropriate version of duality.
Noncompact manifolds require compact supports on one side of the integration pairing. Let be the forms whose supports are compact. Since differentiation preserves compact support, they define
For , ordinary cohomology is concentrated in degree zero, whereas compactly supported cohomology is concentrated in degree , where integration identifies it with . A compactly supported top form can therefore be exact as an ordinary form while having no compactly supported primitive.
On an oriented manifold without boundary, integration pairs with . The general duality statement identifies compactly supported cohomology with ordinary homology in complementary degree. When the relevant groups are finite-dimensional, the integration pairing is perfect. Without finiteness assumptions, identifying both factors with each other’s full algebraic dual requires care; infinite-dimensional vector spaces are not generally isomorphic to their double duals.
For a compact oriented manifold with boundary, the corresponding result is Poincaré–Lefschetz duality. Relative de Rham cohomology can be computed using forms whose pullback to the boundary vanishes, and it pairs with ordinary cohomology in complementary degree. Duality also constrains the Euler characteristic:
For a closed oriented odd-dimensional manifold, the complementary terms cancel, giving .
Hodge theory and canonical representatives
The definition of de Rham cohomology makes no choice of representative inside a class. On a compact manifold, a Riemannian metric supplies a distinguished one. This is the central connection with Hodge theory, which turns topological questions about classes into analytic questions about elliptic differential operators.
Assume first that is oriented, compact, and without boundary. Its metric defines the Hodge star
and the inner product
The formal adjoint of is characterized by
The Hodge Laplacian is
It extends the metric Laplace operator from functions to differential forms, with the nonnegative sign convention.
A form is harmonic if . The identity
shows that harmonicity is equivalent to and . The Hodge theorem asserts that every de Rham class contains exactly one harmonic representative:
Equivalently, smooth forms admit the orthogonal decomposition
Ellipticity and compactness make the harmonic space finite-dimensional, providing an analytic proof of finite-dimensionality of de Rham cohomology.
The harmonic representative also minimizes the norm among all closed representatives of its class. If is harmonic, then
because harmonic forms are orthogonal to exact forms. The selected representative changes when the metric changes, but the dimension of the harmonic space remains the topological number . On a flat torus, the constant-coefficient forms give an explicit harmonic basis.
W. V. D. Hodge developed this harmonic description in the 1930s and presented it systematically in his 1941 book on harmonic integrals. The compactness and boundary assumptions are essential to the simplest formulation. On manifolds with boundary, absolute or relative boundary conditions lead to different cohomology theories. On noncompact manifolds, square-integrable harmonic forms generally compute an theory rather than ordinary de Rham cohomology. For example, the constant function represents the nonzero class in but is not square-integrable under the Euclidean metric.
Characteristic classes and symplectic geometry
De Rham cohomology provides a natural home for geometric quantities constructed from curvature. A vector bundle admits connections that differentiate its sections, and the curvature of a connection is an endomorphism-valued 2-form. Although the curvature changes with the connection, certain polynomial expressions in it determine cohomology classes independent of that choice.
This construction is formalized by Chern–Weil theory. If is the curvature and is an invariant homogeneous polynomial of degree on the relevant matrix Lie algebra, then
is a closed -form. The Bianchi identity is the differential mechanism behind closedness, while invariance of removes dependence on local bundle frames. For two connections , a transgression formula gives
for a suitable -form . Thus the resulting characteristic class depends on the bundle, not the auxiliary connection.
For a Hermitian complex vector bundle with unitary connection, the total Chern form is expressed by
Its homogeneous components represent the real images of the integral Chern classes. On a complex line bundle, the first Chern class is represented by
These formulas show both the power and the limitation of differential forms: curvature represents the real part of the characteristic-class information, while torsion classes can remain invisible.
A related curvature formula is the Gauss–Bonnet theorem. On a closed oriented Riemannian surface,
where is Gaussian curvature. In higher even dimensions, an Euler form built from curvature represents the Euler class of the tangent bundle, and its integral equals the Euler characteristic. Local geometric curvature therefore produces a global topological integer through a de Rham class.
In symplectic geometry, the defining structure is itself a closed nondegenerate 2-form . On a closed symplectic manifold of dimension , its class cannot vanish. If , then closedness gives
so Stokes’ theorem would force . But is a positive volume form for the symplectic orientation. More generally, . This immediately rules out symplectic structures on for , since those spheres have zero second real cohomology. Noncompact symplectic manifolds behave differently: the standard symplectic form on is exact.
Potentials, conservation laws, and discretization
The distinction between closed and exact forms appears directly in field theories. On a spacetime manifold, the electromagnetic field can be written as a 2-form . The homogeneous Maxwell equations take the coordinate-independent form
A globally defined electromagnetic potential is a 1-form satisfying . Such a potential exists exactly when in second de Rham cohomology. Local potentials always exist on sufficiently small neighborhoods, but nonzero flux periods can obstruct a single global one.
For example, a closed 2-form on a region surrounding an excluded point in three-dimensional space can have nonzero integral over an enclosing sphere. The sphere represents a nontrivial homology class in that punctured region. The form is locally exact, but cannot be globally exact because
for any global . In bundle formulations of electromagnetism, local potentials are connected by gauge transformations on overlaps. Appropriate normalization of the curvature then relates its periods to an integral first Chern class. This integrality is additional information, not a consequence of real de Rham cohomology alone.
Closed forms also encode conservation. A current on an -dimensional oriented spacetime can be represented by an -form , with the local conservation equation . If two oriented hypersurfaces differ by the boundary of a region, and any additional boundary contribution vanishes, Stokes’ theorem gives equal integrals of over them. The conserved quantity is consequently a period. Cohomology distinguishes the local differential statement from the global classes on which conserved integrals may depend.
In three-dimensional vector calculus, after choosing a Euclidean metric and orientation, the operators in the de Rham complex correspond to gradient, curl, and divergence:
The identities “curl of a gradient is zero” and “divergence of a curl is zero” are both manifestations of . Cohomology describes precisely why their converses need not hold globally.
These structures guide compatible numerical methods, including finite element exterior calculus within the finite element method. Suitable finite-dimensional spaces of forms are chosen so that differentiation maps each degree into the next, producing a discrete subcomplex. Commuting projection operators preserve the relation between continuous and discrete differentiation. This helps retain conservation identities and correctly represent the topological null spaces of differential operators. Harmonic components associated with nonzero cohomology can otherwise be mistaken for numerical artifacts or omitted from a solver. The required topological dimensions are determined by the domain rather than by mesh refinement or material coefficients.
Complex, sheaf-theoretic, and algebraic extensions
De Rham cohomology can be reformulated through sheaf cohomology. Let denote the sheaf of locally constant real-valued functions, and let be the sheaf of smooth -forms. The Poincaré lemma makes
an exact sequence of sheaves. Partitions of unity make the sheaves of smooth forms fine and hence acyclic for global sections. Taking global sections therefore computes the sheaf cohomology of . This explains why local exactness can produce nonzero global cohomology: the failure occurs when passing from local data to global sections.
On a complex manifold, complex-valued differential forms split into types , and exterior differentiation decomposes as
The identities imply
The cohomology of defines Dolbeault cohomology. The resulting double complex connects holomorphic geometry with complexified de Rham cohomology. On a compact Kähler manifold, Hodge theory yields
This direct-sum description is not valid in this form for arbitrary compact complex manifolds; a spectral sequence records the more general relationship.
In algebraic geometry, algebraic de Rham cohomology replaces smooth forms with algebraic differential forms. For a smooth algebraic variety over a characteristic-zero field, it is defined through the hypercohomology of
For affine varieties, the complex of global algebraic forms suffices; for nonaffine varieties, simply taking global forms can miss cohomological information. Grothendieck’s algebraic de Rham comparison theorem identifies this theory, for smooth complex algebraic varieties, with the singular cohomology of the associated analytic space with complex coefficients.
Characteristic zero is crucial to the familiar local behavior. Over a field of characteristic , the nonconstant polynomial satisfies
Thus even degree-zero closed algebraic functions need not be locally constant in the classical sense, and a direct analogue of the smooth Poincaré lemma fails. The Cartier isomorphism describes the characteristic- de Rham complex through the Frobenius morphism, while crystalline cohomology supplies a related theory with finer arithmetic information. These extensions preserve the organizing idea of studying differential complexes, but their coefficient systems and comparison theorems reflect the geometry of the underlying category.