ASF

Analysis on Manifolds IV: Integration

Part IV of a five-part lecture series on differential forms and the generalised Stokes theorem.

April 13, 2026|
A
A
SymPy Verification Environment

This lecture includes interactive SymPy cells that verify key results symbolically. SymPy is a Python library for symbolic mathematics. Loading it fetches a WebAssembly Python runtime (approx. 15 MB, cached after first load). You can also load it on demand from any code cell below.

0. Exterior derivative and integration

Parts I through III built the analytic and algebraic machinery that makes integration on manifolds possible. Part I established the calculus of : the inverse and implicit function theorems, the definition of the derivative as a linear map, and the theory of smooth maps. Part II built the algebra of a single vector space: the dual, alternating tensors, the wedge product, and the pullback of a linear map. Part III glued the algebra onto the smooth structure and produced differential -forms, the exterior derivative , and the identity . Part III closed by defining closed and exact forms and previewing de Rham cohomology.

Part IV uses all of that to integrate. A -form can be integrated over an oriented -dimensional region: the alternating sign encodes orientation and the pullback formula built into the form's definition handles changes of coordinates. Integration requires no new constructions beyond Parts I-III.

This post has two acts. Act I (Sections 1-6) sets up integration: what it means for a form to be a derivative or antiderivative (Section 1), what orientation means for manifolds (Section 2), what the boundary of a manifold is (Section 3), how to patch local definitions globally with partitions of unity (Section 4), the definition of the integral of a -form (Section 5), and the Riemannian volume form that lets one integrate ordinary functions (Section 6). Act II (Sections 7-10) covers periods and the winding number (Section 7), the Mayer-Vietoris long exact sequence (Section 8), cohomology computations for a range of surfaces (Section 9), and Poincaré duality (Section 10). Section 11 previews Part V.

Notation. Part I worked entirely in , where points and vectors are the same object. That identification breaks down on a general manifold. A point lives in the manifold; a tangent vector lives in the separate vector space attached at . Throughout Part IV we write points in roman italic () and tangent vectors in boldface (). The distinction collapses only when .

We follow Bott-Tu [1], Lee [2], Tu [3] and Bredon [4].


Integration (Act I)

1. Forms as derivatives and antiderivatives

Forms are the objects to integrate because they already encode the infinitesimal change-of-variable formula. Hence the exterior derivative is the correct notion of derivative for forms, and the antiderivatives (primitives) of forms are exactly the exact forms.

1.1. Ratio

Let be smooth. In Part III we defined by , where is the standard (constant) 1-form on . The symbol is a differential 1-form; it pairs with a tangent vector to give .

The classical derivative is the coefficient of in the 1-form . That is:

The equality is a literal equality of 1-forms, rearranged: dividing both sides by the 1-form (a non-vanishing section of the cotangent bundle on ) gives the smooth function . In dimension one the cotangent bundle is a line bundle and the ratio of two 1-forms is a well-defined smooth function at each point. This is the origin of Leibniz notation.

On with coordinates , for :

The coefficient of is the partial derivative . The 1-form packages all partial derivatives into a single coordinate-free object (its value on a tangent vector is the directional derivative ). The ratio interpretation generalises: .

1.2. Antiderivatives as exact forms

Definition (Primitive).

A -form is a primitive (or antiderivative) of if . Equivalently, is exact.

On , a 1-form is exact iff has a smooth antiderivative (i.e. ), in which case . This is the classical notion. Primitives of forms are forms of one degree lower. When a primitive exists it is unique up to adding a closed form.

On , a 1-form is exact iff for some smooth , which requires for all (the curl-free condition). The condition is necessary because implies for any exact form. The Poincaré lemma (Part III, §10.3) shows this is also sufficient locally. Globally the failure is measured by .

Example (Angular form).

On , define the 1-form

A direct computation gives (closed). In polar coordinates where , , one finds . Locally is a smooth antiderivative. Globally jumps by as one circles the origin and no single-valued smooth function is a primitive on all of . The form is closed with no global primitive: it generates .

1.3. Fundamental theorem as baby Stokes

The exterior derivative assigns to a smooth function the 1-form . Integrating this 1-form over the interval should recover at the endpoints. That is the content of the fundamental theorem of calculus.

Theorem (Fundamental theorem of calculus).

Let be smooth. Then

Proof

Define the accumulation function by

Step 1 (). For and small ,

Since is continuous, for every there exists such that whenever . For ,

Hence for all , with the same argument applying one-sidedly at the endpoints.

Step 2 ( constant). Since on and is connected, the mean value theorem implies is constant.

Step 3 (evaluation). At , , so the constant equals . Therefore for all , and setting gives

Since is the exterior derivative of the 0-form , and since the interval is a 1-manifold with boundary (two points with the induced orientation, positive at and negative at ), the theorem reads

Equation [eq:baby-stokes] is the generalised Stokes theorem in dimension 1: the integral of the exterior derivative of a form over a manifold equals the integral of the form over the boundary. Every classical integral theorem (Green's theorem in the plane, the divergence theorem in space, the classical Stokes theorem for surfaces) is the same identity with replaced by a higher-dimensional manifold with boundary. Part V proves the full theorem.

2. Orientation

Integration requires a choice of sign: which way is "positive"? This is orientation. On a vector space, orientation amounts to choosing a preferred basis up to positive-determinant changes. On a manifold, orientation amounts to making this choice consistently across all tangent spaces.

2.1. Oriented vector spaces

Definition (Orientation of a vector space).

Let be a real vector space of dimension . Two ordered bases and are equivalent if the transition matrix defined by has positive determinant. This is an equivalence relation with exactly two classes; an orientation of is a choice of one class.

An orientation of is equivalently a choice of nonzero element of up to positive scalar: a nonzero -form is positively oriented if for a positively oriented basis .

Example (Orientation of ).

has a standard orientation: the standard basis is positive. The standard -form is .

2.2. Oriented manifolds

Definition (Orientation of a manifold).

A smooth -manifold is orientable if there exists a nowhere-vanishing -form . An orientation is a choice of such an up to positive smooth function; two choices are equivalent if their ratio is everywhere positive. An oriented manifold is a pair .

Proposition (Atlas characterisation).

A smooth -manifold is orientable if and only if it has an atlas whose transition maps have everywhere positive Jacobian determinant. Such an atlas is called oriented.

Proof

Given a non-vanishing -form , choose charts such that in local coordinates with (take connected chart domains and replace by its composition with if needed). On an overlap . So the transition map has Jacobian determinant .

Conversely, given an oriented atlas, define a local -form in each chart, and patch these with a partition of unity subordinate to the cover . Since the Jacobians are positive, each is a positive multiple of every other on overlaps. Hence vanishes nowhere.

Two oriented charts on a manifold with a positively oriented transition map between them
Two oriented charts and on a manifold . The transition map has positive Jacobian on the overlap (shown in green). Reversing the orientation of one chart (dashed frame) produces a negative Jacobian transition.
Example (Orientability of standard spaces).
  • : orientable, with the standard orientation .
  • : orientable for all . The outward unit normal to the embedding gives a volume form via .
  • : orientable, with volume form (where are the angle coordinates).
  • : orientable iff is odd.
  • Möbius band: not orientable. As one traverses the core circle the fibre orientation is reversed and no consistent choice exists.
  • Klein bottle: not orientable.

2.3. Orientation-reversing maps

A diffeomorphism between oriented manifolds is orientation-preserving if its Jacobian in oriented charts has positive determinant everywhere, and orientation-reversing if the determinant is negative everywhere. The reflection defined by is orientation-reversing (). The antipodal map on is orientation-preserving iff is odd.

3. Manifolds with boundary

Stokes' theorem relates integrals over a manifold to integrals over its boundary. This section defines precisely.

3.1. Definition

Definition (Manifold with boundary).

A smooth -manifold with boundary is a second-countable Hausdorff topological space together with a maximal smooth atlas in which each chart maps to either an open subset of or an open subset of the upper half-space .

The boundary is the set of points that map to the hyperplane under some (hence every) chart whose domain contains . The interior is .

Proposition (Boundary as a manifold).

If is a smooth -manifold with boundary, then is a smooth -manifold without boundary (i.e., ).

Proof

Each boundary chart restricts to a homeomorphism from to the open set in (identifying with ). The transition maps of are the restrictions of those of to the boundary hyperplane; they remain smooth (since the defining maps are smooth up to the boundary). The chart images lie in the open subsets of and no second boundary stratum arises.

Example (Standard examples).
  • The closed interval : boundary , both points.
  • The closed disc : boundary .
  • A compact surface (genus- surface) with circular holes cut out: boundary is disjoint copies of .
  • itself: boundary .
Example (Möbius band boundary).

The Möbius band is a smooth 2-manifold with boundary. Its boundary is a single circle that goes around the band twice. The twisted identification joins the two free edges of the rectangle used to construct end to end into this one circle. The untwisted identification gives the cylinder with two boundary circles.

3.2. Induced orientation on the boundary

If is an oriented -manifold with boundary, the boundary inherits a canonical orientation.

Definition (Induced (outward-normal-first) orientation).

Let be an oriented -manifold with boundary. At each , let denote an outward-pointing tangent vector (one whose image in any half-space chart points in the direction of decreasing ). The induced orientation on is defined by: a basis of is positively oriented iff is positively oriented in .

This is the "outward normal first" convention. For the interval : the orientation of is the positive direction; the outward normal at points right (positive), at points left (negative). The induced orientation assigns to and to . Hence . This recovers the sign in the fundamental theorem.

Induced boundary orientation on a 2D region: outward normal first convention
The induced orientation on the boundary of an oriented 2-manifold . The blue arrows show the orientation of (counterclockwise in the plane); the gold arrows on show the induced orientation (also counterclockwise when viewed from outside). The outward-normal-first rule gives the counterclockwise direction.

4. Partitions of unity

A differential form is a global object, but we can only compute with it locally in charts. Partitions of unity break a global computation into local pieces and reassemble them.

4.1. Smooth bump functions

Smooth functions on (unlike holomorphic functions on ) can be non-zero and compactly supported.

Lemma (Bump function).

For any there exists a smooth function with on the closed ball and . Such a function is called a smooth bump function.

Proof

Define for and for ; this is smooth on . Set ; then is smooth, for , and for . Fix . Then is a smooth radially symmetric bump with support and equal to 1 on .

4.2. Existence of partitions of unity

Theorem (Existence of partitions of unity).

Let be a smooth manifold and an open cover of . There exists a smooth partition of unity subordinate to : a collection of smooth functions such that:

  1. for all .
  2. The collection is locally finite (every has a neighbourhood meeting only finitely many supports).
  3. for all (the sum is finite at each point by local finiteness).
Proof sketch

Since is second-countable (part of the definition of a smooth manifold), every open cover has a countable locally finite refinement (by paracompactness, which follows from second-countability). Take a countable locally finite refinement by chart balls and shrink it to open sets with that still cover . Choose smooth supported in and positive on . The sum is finite near each point by local finiteness and positive everywhere because the cover . So is smooth. Pick for each an index with and set .

Three smooth bump functions forming a partition of unity on a line
Three smooth bump functions (blue, gold, green) on , each supported in one of three overlapping intervals, summing to 1 at every point. The partition of unity decomposes any global integral into a sum of local integrals.
Remark (Smooth partitions of unity).

The analytic fact that smooth bump functions exist distinguishes smooth manifolds from real-analytic or complex-analytic manifolds. On a connected complex manifold there are no compactly supported holomorphic functions (other than zero) and the same patching argument fails. Smooth partitions of unity are why many global results (integration, Riemannian metrics, connections) are easier to construct in the smooth category than in the holomorphic one.

5. Integration of differential forms

5.1. Integration in a single chart

Let be an open set and an -form with compact support. Write . Define

where the right-hand side is the standard Lebesgue (or Riemann) integral in .

This definition depends on the orientation: reversing the order of two wedge factors negates and negates the integral.

5.2. Change-of-variables formula

Proposition (Pullback and integration).

Let be an orientation-preserving diffeomorphism of open subsets of , and . Then

If is orientation-reversing, the equality becomes .

Proof

Write in coordinates on , with . Then

where we used the fact that (Part II, wedge product and determinants). Therefore

where in the middle step we used the classical change-of-variables theorem and the fact that (orientation-preserving). The orientation-reversing case introduces a factor of since .

The pullback formula built into the definition of differential forms (Part II) is the change-of-variables formula for integrals.

5.3. Global definition via partition of unity

Definition (Integral of an n-form).

Let be a compact oriented -manifold (possibly with boundary) and . Choose an atlas compatible with the orientation and a subordinate partition of unity . Define

Each term is an integral of a compactly supported -form on an open subset of , as in §5.1.

Proposition (Well-definedness).

The integral is independent of the choice of oriented atlas and subordinate partition of unity.

Proof

Let be a second oriented atlas with subordinate partition of unity . Since , linearity of each chart integral gives , computed in the chart . The form is supported in . Since the transition map has positive Jacobian, the change-of-variables proposition above gives the same value in the chart . Summing over and then over turns the double sum into the integral defined by the second system.

Proposition (Properties of integration).

Let be a compact oriented -manifold and .

  1. Linearity: for .
  2. Orientation reversal: if denotes with the opposite orientation, then .
  3. Positivity: if is a volume form compatible with the orientation, then .
  4. Pullback: if is an orientation-preserving diffeomorphism, then .
  5. Disjoint union: if with having measure zero, then .

5.4. Integration over submanifolds

More generally, one integrates a -form over an oriented -dimensional submanifold :

where is the inclusion and is the pullback (restriction) to . This is the integral of a -form over a -manifold, reduced to the case above.

Example (Line integrals).

A 1-form on integrates over an oriented curve as

This is the classical line integral. The form is the object that, when given the velocity tangent vector , returns the integrand at time .

Example (Surface integrals).

A 2-form on integrates over an oriented surface with parametrisation as the classical flux integral . The wedge product encodes the cross product; the orientation determines the sign.

6. Riemannian volume form

A differential form is the natural object to integrate. In applications one often wants to integrate an ordinary function (a density, a charge distribution, a probability). A Riemannian metric determines a canonical volume -form that converts functions into -forms.

6.1. Metric volume form

Definition (Riemannian volume form).

Let be an oriented Riemannian -manifold. In any positively oriented local chart with coordinates , the Riemannian volume form is

where is the metric tensor in local coordinates.

Proposition (Well-definedness).

The form is independent of the choice of positively oriented chart.

Proof

Under a change of positively oriented coordinates with positive Jacobian , the metric transforms as . Hence . Thus . Meanwhile . The product gives . Hence is the same -form in both charts.

The integral of a function against the Riemannian measure is

computed via a partition of unity as in §5.3.

Example (Volume of ).

The standard round metric on in spherical coordinates , where , , and is the polar angle from the north pole, gives metric components , , . This chart is positively oriented for the outward normal, since . Thus on and

Example (Flat torus ).

The flat metric has . Hence and .


de Rham Cohomology (Act II)

7. Period integrals and closed forms

Part III defined closed and exact forms and hinted that their quotient is a topological invariant. Integration makes this precise: closed forms integrate to periods, real numbers that depend only on the cohomology class of the form. Periods link the analytic definition of de Rham cohomology to its topological content.

7.1. Periods over cycles

Definition (Period).

Let be a closed form and a compact oriented -dimensional submanifold of without boundary (a -cycle). The period of over is

Proposition (Periods depend only on cohomology).

If and are cohomologous closed forms, then for every cycle . Similarly, if and differ by a boundary (i.e., for some -chain ), then .

Proof

For the first claim: , since (Z has no boundary). For the second claim: , since is closed. (Both applications use the Stokes theorem, proved in Part V; we use them here as motivation.)

Periods define a bilinear pairing

The de Rham theorem (Part III, §10.2) implies this pairing is non-degenerate: cohomology classes are completely determined by their periods against all cycles.

7.2. Winding number as a period

The angular 1-form on ,

is closed (as computed in Part III). Its period over the unit circle (parametrised as ) is

For an arbitrary closed loop (with ), the winding number around the origin is

Since is not exact, the period can be nonzero. The winding number measures how many times wraps around the origin; it is always an integer because changes by integer multiples of around a closed loop.

The interactive figure below lets you drag four control points to deform a closed Catmull-Rom loop around the origin. The winding number is computed in real time as the period .

Remark (Winding number and ).

The class generates and the winding number [eq:winding-num] is the matching coordinate: any closed 1-form on is of the form for some and , and is the period over the unit circle divided by . The integer values reflect that (equation [eq:angular-form]) generates the -lattice inside corresponding to singular cohomology .

8. Mayer-Vietoris sequence

Computing de Rham cohomology directly from the definition (quotient of closed forms by exact forms) is usually intractable. The Mayer-Vietoris sequence is the main inductive tool: it expresses in terms of , and when .

8.1. Restriction maps

The direct sum of two vector spaces is their Cartesian product made into a vector space by componentwise addition and scalar multiplication. As an example, if and , elements of are column vectors with an upper block from and a lower block from :

A linear map is captured by a block row . For infinite-dimensional spaces such as and , the direct sum has the same componentwise structure: a pair with , .

Let with open. The inclusions and and the inclusions , induce restriction maps on forms:

where restricts to each open set, and takes the difference of the two restrictions to the overlap. This sequence is exact (a partition of unity argument proves surjectivity of ).

8.2. Long exact sequence

Theorem (Mayer-Vietoris long exact sequence).

Let with open. There is a long exact sequence

The connecting homomorphism is defined as follows: given a closed form representing a class, write for forms and (which exist by the short exact sequence above). Then on . Hence they glue to a global closed -form on (with ). Set .

Proof

The short exact sequence [eq:mv-ses] of cochain complexes (each map commuting with ) induces a long exact sequence in cohomology by the standard algebraic snake lemma. The connecting homomorphism is exactly the one described above. Exactness of the short sequence at each degree uses a partition of unity subordinate to for the surjectivity of .

8.3. Computation for

Decompose as the union of two open arcs:

where and are the north and south poles. Each of and is diffeomorphic to (contractible), so and for . The intersection has two connected components (two open arcs), so .

Applying [eq:mv-les] in low degrees:

Substituting the known groups: . The map is (restriction of constants to each arc). Its kernel is (the image of the map from ) and its image has rank 1 (the diagonal). Thus has dimension 1, and by exactness . Hence .

Result: , , for .

The Mayer-Vietoris decomposition of (and the dependence of on the gap size) is illustrated in the interactive figure below.

Remark (Two components of the intersection).

has two components when the gaps at and are small. The two components of are forced by the topology of : every decomposition of into two contractible open sets produces a disconnected intersection. The rank-1 cokernel of then forces .

8.4. Computation for

Decompose as and , both diffeomorphic to (via stereographic projection). The intersection has the homotopy type of a circle, so and .

The Mayer-Vietoris sequence in degree 1:

Since and are contractible, . The sequence becomes . Hence .

Result: , .

Mayer-Vietoris decomposition of S^2 into two contractible caps with an equatorial strip as intersection
Mayer-Vietoris decomposition of . Blue cap and gold cap are each contractible. Their intersection (purple equatorial band) is homotopy equivalent to . The long exact sequence forces .

8.5. Computation for

The torus can be handled by Mayer-Vietoris applied to where and are open annular strips. Alternatively, the Künneth formula for de Rham cohomology gives the cohomology of a product directly.

The binary direct sum generalises to an iterated direct sum indexed by a condition on degrees. Writing means: take one copy of for each pair of non-negative integers with , and sum them all. For instance, with and spaces indexed by :

Each summand contributes a block: if and , the full direct sum has dimension . This is the graded analogue of how polynomial multiplication combines terms of matching total degree.

Theorem (Künneth formula for de Rham cohomology).

For smooth manifolds and with having a finite good cover (for instance compact),

Applying this to with :

The generators of are the two angular 1-forms and (one for each circle factor). The generator of is the area form .

Mayer-Vietoris decomposition of T^2 = S^1 x S^1 into two open annular strips
Mayer-Vietoris decomposition of . The two open sets (blue) and (gold) are open annular strips, each homotopy equivalent to . Their intersection consists of two disjoint annuli, each homotopy equivalent to . The MV sequence recovers and .

9. Computing cohomology

Mayer-Vietoris and Künneth compute the cohomology of a wide range of spaces. Examples follow.

9.1. Spheres

Theorem (Cohomology of spheres).
Proof by induction via Mayer-Vietoris

Base case (): proved in §8.3 above.

Inductive step: Assume the result for . Decompose with and , so (contractible) and (an equatorial sphere, of dimension ). The Mayer-Vietoris sequence in degree reads:

So is an isomorphism for .

By induction and for . Hence and for . The and cases are handled separately by the degree-0 part of MV (connectedness) and the degree-1 part (which gives for since and the difference map has rank 1, kernel , cokernel 0, so ).

9.2. Tori

By the Künneth formula applied inductively to ,

The generators of are the wedge products for . The full cohomology ring is the exterior algebra:

This is the Part II algebra at the level of global cohomology. The Betti numbers are the binomial coefficients: .

The interactive figure shows the harmonic generators on in degrees 0, 1 and 2:

9.3. Real projective spaces

Theorem (Cohomology of real projective spaces).

With real coefficients:

Proof sketch

The quotient is a double cover with deck group , where is the antipodal map. On , the antipodal map acts by on the generator of (since is orientation-preserving iff is odd) and by on . Forms on correspond to -invariant forms on . Averaging over shows that identifies with the -invariant part of . The generator of is -invariant iff , i.e. is odd. For the group is already zero.

In particular with real coefficients (the torsion in integral cohomology disappears upon tensoring with ).

9.4. Möbius band

The Möbius band deformation-retracts onto its core circle (the centre line). Therefore :

The generator of is the pullback of along the retraction. With real coefficients de Rham cohomology sees only the homotopy type: and the orientable annulus share the same cohomology groups. They are distinct as smooth manifolds (one orientable, one not).

9.5. Klein bottle

The Klein bottle is a compact non-orientable surface without boundary. It can be constructed from two Möbius bands by gluing their boundaries. A Mayer-Vietoris argument gives:

The vanishing of reflects non-orientability: a connected compact -manifold without boundary has iff it is orientable, and otherwise. The torsion in integral homology (equivalently ) disappears with real coefficients. Only remains in degree 1.

9.6. Euler characteristic

Definition (Euler characteristic).

The Euler characteristic of a compact manifold is

Examples:

Manifold
1012
1210
1001
Klein bottle1100
Genus- surface11

The Euler characteristic satisfies (from Künneth) and (from Mayer-Vietoris).

10. Poincaré duality

For a compact oriented manifold there is a symmetry between cohomology in complementary degrees: and are dual to each other, paired by integration.

10.1. Poincaré pairing

Theorem (Poincaré duality).

Let be a compact oriented smooth -manifold without boundary. The bilinear pairing

is non-degenerate. Consequently , and since all groups are finite-dimensional, .

Proof sketch

That the pairing is well-defined (independent of representative) follows from Stokes' theorem (Part V): if , then . The computation uses (closedness) and on a closed manifold (Stokes).

Non-degeneracy is proved using Hodge theory: each cohomology class has a unique harmonic representative (a form in the kernel of the Laplacian ), and the Hodge star (Part II) maps harmonic -forms to harmonic -forms. For a harmonic , taking gives .

Corollary (Betti number symmetry).

If is a compact oriented -manifold without boundary, then for all .

The examples above all show it: ; ; the Euler characteristic of a compact oriented odd-dimensional manifold without boundary is 0 (since and gives cancellation for odd).

Schematic of the Poincaré pairing: k-forms wedged with (n-k)-forms integrate to a real number over M
Schematic of the Poincaré pairing on a compact oriented 2-manifold . A 1-form (blue) and a 1-form (gold) are wedged to produce a 2-form , which is then integrated over to give a real number. Non-degeneracy means the map is an isomorphism .

10.2. Examples

Example (Poincaré duality for ).

. The pairing [eq:poincare-pair] for , reduces to (product of a function and a 2-form). Non-degeneracy: take and to get .

Example (Poincaré duality for ).

with generators and . The pairing [eq:poincare-pair] gives and (since ). In the basis , the pairing matrix is (skew-symmetric, as expected for a 2-manifold: since ). This matrix has determinant 1 and is therefore non-degenerate.

Remark (Non-orientable manifolds).

Poincaré duality with real coefficients fails for non-orientable manifolds. For : but , so . The correct statement for non-orientable manifolds requires twisted (local system) coefficients. This post does not treat them.

11. Preview of Stokes' theorem

Part V proves the generalised Stokes theorem

valid for any compact oriented -manifold with boundary and any -form . Equation [eq:baby-stokes] is the case of [eq:stokes-gen] with . Specialised further:

  • , a smooth function: the fundamental theorem (Theorem [ftc]).
  • , a region in , a 1-form: Green's theorem.
  • , a region in , a 2-form: the divergence theorem.
  • , an oriented surface in , a 1-form: the classical Stokes theorem for surface integrals.

Part V will prove the theorem, recover all classical cases, and use it to prove that de Rham cohomology is a topological invariant. The series ends with the identification of the period pairing as the natural map

and the statement that this pairing induces the de Rham isomorphism between the analytic and topological descriptions of the same cohomology.

References

  1. R. Bott and L.W. Tu, Differential Forms in Algebraic Topology, Springer, 1982. The primary reference for de Rham cohomology, Mayer-Vietoris, and Poincaré duality. DOI
  2. J.M. Lee, Introduction to Smooth Manifolds, 2nd ed., Springer, 2012. Chapters 15-17 (integration, Stokes). DOI
  3. L.W. Tu, An Introduction to Manifolds, 2nd ed., Springer, 2011. Chapters 6-7 (integration and de Rham theory). DOI
  4. G.E. Bredon, Topology and Geometry, Springer, 1993. Chapter V (cohomology). DOI
  5. M. Spivak, Calculus on Manifolds, W.A. Benjamin, 1965. The classic short treatment of integration on manifolds. DOI
  6. W. Rudin, Principles of Mathematical Analysis, 3rd ed., McGraw-Hill, 1976. Chapter 10 (classical change-of-variables).