Cap product

Intuition [Beginner]

The cap product is the rule that lets a measurement on a space (a cochain) eat part of a shape (a chain) and leave a smaller shape behind. A cochain assigns a number to each small simplex of a given dimension; a chain is a formal sum of simplices. Pairing a $k$-dimensional cochain with an $n$-dimensional chain feeds the cochain the front $k$-dimensional face of each simplex, multiplies by the resulting number, and keeps the back $(n-k)$-dimensional face. The leftover is an $(n-k)$-dimensional chain.

Why bother? Two reasons. First, the cap product turns the homology of a space into a module over the cohomology ring of the same space — so you can multiply a cohomology class by a homology class and get a homology class. Second, capping with a special class called the fundamental class of a closed oriented manifold turns cohomology into homology in a way that pairs degree $k$ with degree $n-k$. This pairing is Poincaré duality, the structural theorem behind every intersection-theory computation on manifolds.

The picture to keep in mind: a cochain is like a stack of horizontal slices, a chain is a vertical stick of cells, and capping is what happens when the slices cut the stick — the part above the cut is consumed by evaluation, the part below the cut survives as a smaller chain.

Visual [Beginner]

A schematic of an $n$-simplex split into a front $k$-face and a back $(n-k)$-face along the vertex $v_k$. An arrow shows the cochain $\varphi$ being evaluated on the front face, producing a number; the back face survives as the $(n-k)$-chain coefficient.

The picture captures the front-face / back-face split that defines the cap product. The cochain consumes the front $k$ vertices, the back $n-k$ vertices index the remaining chain, and the number returned by the cochain is the coefficient of that surviving chain.

Worked example [Beginner]

Compute the cap product of a $1$-cochain with a $2$-chain on the standard $2$-simplex.

Step 1. Take the standard $2$-simplex ${\Delta }^2$ with ordered vertices $v_0,v_1,v_2$. The $2$-chain is the simplex itself, written $\sigma =[v_0,v_1,v_2]$. The $1$-cochain $\varphi$ is defined by $\varphi ([v_0,v_1])=3$, $\varphi ([v_1,v_2])=5$, $\varphi ([v_0,v_2])=2$.

Step 2. Apply the front-face / back-face split with $k=1$, $n=2$. The front $1$-face of $\sigma$ is $\sigma {|}_{[v_0,v_1]}=[v_0,v_1]$. The back $1$-face is $\sigma {|}_{[v_1,v_2]}=[v_1,v_2]$.

Step 3. Evaluate the cochain on the front face: $\varphi ([v_0,v_1])=3$. The cap product is the number $3$ multiplied by the back face: $\varphi ⌢\sigma =3\cdot [v_1,v_2]$.

Step 4. The result is a $1$-chain — the edge $[v_1,v_2]$ taken with coefficient $3$. Capping a $1$-cochain with a $2$-chain has produced a $1$-chain, matching the dimension count $n-k=2-1=1$.

What this tells us: the cap product is concrete and combinatorial. It does not require any abstract machinery to compute on a single simplex; the front-face / back-face split is the entire definition. Linear extension to formal sums of simplices and the resulting descent to homology are the only pieces that need extra work.

Check your understanding [Beginner]

Formal definition [Intermediate+]

Let $X$ be a topological space. The singular chain complex $C_{\bullet }(X)$ has $C_n(X)$ the free abelian group on continuous maps $\sigma :{\Delta }^n\to X$, with the standard alternating-face boundary $\partial$. The singular cochain complex $C^{\bullet }(X)$ has $C^k(X)=\mathrm{Hom}(C_k(X),\mathbb{Z})$, with coboundary $\delta \varphi (\sigma )=\varphi (\partial \sigma )$.

For a singular $n$-simplex $\sigma :{\Delta }^n\to X$ with $k\le n$, write $[v_0,\dots ,v_n]$ for the ordered vertex set of the standard simplex and let $\sigma {|}_{[v_0,\dots ,v_k]}:{\Delta }^k\to X$ denote the front $k$-face, the restriction of $\sigma$ to the affine $k$-subsimplex spanned by the first $k+1$ vertices. Let $\sigma {|}_{[v_k,\dots ,v_n]}$ denote the back $(n-k)$-face, the restriction to the affine $(n-k)$-subsimplex spanned by the last $n-k+1$ vertices. The two faces overlap in the single vertex $v_k$.

The cap product of a singular $k$-cochain $\varphi \in C^k(X)$ with a singular $n$-simplex $\sigma$ is

$$\varphi ⌢\sigma =\varphi \left(\sigma {|}_{[v_0,\dots ,v_k]}\right)\cdot \sigma {|}_{[v_k,\dots ,v_n]}\in C_{n-k}(X).$$

The cochain is evaluated on the front face producing an integer; that integer is the coefficient of the back face viewed as an $(n-k)$-chain. Linear extension over $\sigma$ and bilinearity in $\varphi$ produce a chain-level pairing

$$⌢:C^k(X)\otimes C_n(X)\to C_{n-k}(X).$$

Leibniz rule. For $\varphi \in C^k(X)$ and a singular $n$-chain $c\in C_n(X)$,

$$\partial (\varphi ⌢c)=(-1{)}^k(\delta \varphi ⌢c-\varphi ⌢\partial c).$$

The signed identity is verified by direct computation on a single simplex via the alternating-face formula. Both cocycles paired with cycles produce cycles, and modifications by coboundaries or boundaries shift the result by a boundary. The cap therefore descends to a well-defined pairing on (co)homology

$$⌢:H^k(X)\otimes H_n(X)\to H_{n-k}(X).$$

Module structure. The cap product makes $H_{\ast }(X)$ a graded module over the cohomology ring $H^{\ast }(X)$ with cup product. Bilinearity is immediate; associativity is the cup-cap compatibility identity stated in the next section.

Naturality (projection formula). For a continuous map $f:X\to Y$, a cohomology class $\varphi \in H^k(Y)$, and a homology class $\alpha \in H_n(X)$,

$$f_{\ast }(f^{\ast }\varphi ⌢\alpha )=\varphi ⌢f_{\ast }\alpha \in H_{n-k}(Y).$$

The pullback of $\varphi$ along $f$ caps with $\alpha$ in $X$; pushing the result forward agrees with capping the original $\varphi$ against the pushforward of $\alpha$. This is the cap-product analogue of the projection formula in algebraic geometry and is what makes the cap a natural transformation between functors of pairs.

Counterexamples to common slips

• The front-face / back-face split is asymmetric: the cochain consumes the front $k$-face, not the back. Reversing the convention produces a different pairing related to the original by a sign and an opposite-orientation reordering. Hatcher and Bredon use the front-face convention; Spanier sometimes writes the back-face convention. Pinning the convention is required before any sign computation.
• The Leibniz sign $(-1{)}^k$ tracks the cochain degree, not the chain degree. A common slip is to write $(-1{)}^n$ or $(-1{)}^{n-k}$; the correct sign is $(-1{)}^k$ and is what makes the cup-cap compatibility hold without further sign adjustment.
• The cap is not graded-commutative. Cup is graded-commutative on cohomology; cap is a module action and has no commutation rule beyond the cup-cap identity.

Key theorem with proof [Intermediate+]

Theorem (cup-cap compatibility; Hatcher §3.3). Let $X$ be a topological space, $\varphi \in C^k(X)$, $\psi \in C^{\ell }(X)$, and $\sigma \in C_n(X)$ a singular $n$-simplex with $k+\ell \le n$. Then

$$\varphi ⌢(\psi ⌢\sigma )=(\psi ⌣\varphi )⌢\sigma .$$

The identity descends to (co)homology and exhibits $H_(X)$asagradedmoduleoverthecohomologyring$H^(X)$.

Proof. Apply the front-face / back-face split twice. The inner cap is

$$\psi ⌢\sigma =\psi \left(\sigma {|}_{[v_0,\dots ,v_{\ell }]}\right)\cdot \sigma {|}_{[v_{\ell },\dots ,v_n]}.$$

Set $\tau =\sigma {|}_{[v_{\ell },\dots ,v_n]}$, a singular $(n-\ell )$-simplex with ordered vertices $v_{\ell },v_{\ell +1},\dots ,v_n$ (relabel as $w_0,w_1,\dots ,w_{n-\ell }$ if needed). The outer cap of $\varphi$ with $\psi ⌢\sigma$ is

$$\varphi ⌢(\psi ⌢\sigma )=\psi \left(\sigma {|}_{[v_0,\dots ,v_{\ell }]}\right)\cdot \varphi \left(\tau {|}_{[w_0,\dots ,w_k]}\right)\cdot \tau {|}_{[w_k,\dots ,w_{n-\ell }]}.$$

Translating back to the original vertex labels of $\sigma$, the first front-face is $[v_0,\dots ,v_{\ell }]$, the second front-face is $[v_{\ell },\dots ,v_{\ell +k}]$, and the back-face that survives is $[v_{\ell +k},\dots ,v_n]$. The two evaluations consume the first $\ell +1$ vertices and the next $k$ vertices respectively, and the back $(n-\ell -k)$-face survives.

The right-hand side. The cup product on cochains is

$$(\psi ⌣\varphi )(\sigma {|}_{[v_0,\dots ,v_{k+\ell }]})=\psi \left(\sigma {|}_{[v_0,\dots ,v_{\ell }]}\right)\cdot \varphi \left(\sigma {|}_{[v_{\ell },\dots ,v_{\ell +k}]}\right).$$

Capping $\psi ⌣\varphi$ with $\sigma$ takes the front $(k+\ell )$-face for the cochain evaluation and keeps the back $(n-k-\ell )$-face,

$$(\psi ⌣\varphi )⌢\sigma =(\psi ⌣\varphi )\left(\sigma {|}_{[v_0,\dots ,v_{k+\ell }]}\right)\cdot \sigma {|}_{[v_{k+\ell },\dots ,v_n]}.$$

Substituting the cup-product expansion,

$$(\psi ⌣\varphi )⌢\sigma =\psi \left(\sigma {|}_{[v_0,\dots ,v_{\ell }]}\right)\cdot \varphi \left(\sigma {|}_{[v_{\ell },\dots ,v_{\ell +k}]}\right)\cdot \sigma {|}_{[v_{\ell +k},\dots ,v_n]}.$$

The two expressions agree term by term: the same $\psi$-evaluation, the same $\varphi$-evaluation, the same surviving back face. The chain-level identity holds.

Descent to homology uses the Leibniz rule $\partial (\varphi ⌢c)=(-1{)}^k(\delta \varphi ⌢c-\varphi ⌢\partial c)$. For cocycles $\varphi ,\psi$ and a cycle $c$, both sides of the identity pair cocycles with a cycle and produce a cycle; both shifts by a coboundary or a boundary modify the result by a boundary; the descent to $H^k(X)\otimes H^{\ell }(X)\otimes H_n(X)\to H_{n-k-\ell }(X)$ produces a well-defined identity on classes. The associativity of the resulting action of the cup-product ring on $H_{\ast }$ is exactly this identity. $\square$

Bridge. The cap product builds toward the entire intersection-theory machinery on manifolds, with Poincaré duality as the load-bearing application. The foundational reason it works is the front-face / back-face split combined with the Leibniz rule: the chain-level pairing is constrained enough to descend to homology, and the cup-cap compatibility makes the descent multiplicative. This is exactly the same descent principle that appears again in 03.12.13 (cellular homology), where the cellular cap product is defined via a diagonal approximation on the cellular chain complex and inherits its Leibniz rule from the singular case via the cellular-to-singular comparison theorem. The central insight is that the cap product is dual to the cup product in a precise sense: cup pairs cohomology with cohomology to produce cohomology; cap pairs cohomology with homology to produce homology; the two together generalise the evaluation pairing $H^k(X)\otimes H_k(X)\to \mathbb{Z}$ that already appears in 03.12.11 (singular homology) as the Kronecker pairing. Putting these together, the cap is what turns the homology of a space into a module over its cohomology ring, and the bridge is the recognition that capping with a fundamental class identifies cohomology with homology when a fundamental class exists. The bridge appears again in 03.12.16 (Poincaré duality), where the cap with $[M]$ is the duality isomorphism, and is the structural reason the duality is natural in continuous maps via the projection formula.

Exercises [Intermediate+]

Lean formalization [Intermediate+]

Mathlib has Algebra.Homology.HomologicalComplex for chain-complex homology and AlgebraicTopology.SingularSet for the simplicial-set side, but no named cap product on singular chains. The intended formalisation would read schematically:

[object Promise]

The proof gap is substantive. Mathlib needs the cap product as a chain map, the Leibniz identity, the descent to homology, and the cup-cap compatibility. Each piece is formalisable from existing infrastructure but has not been packaged. The Poincaré duality consequence requires a separate fundamental-class API, which depends on orientation and is itself a Mathlib gap.

Advanced results [Master]

Theorem (Leibniz rule, full statement). Let $X$ be a topological space, $\varphi \in C^k(X)$, and $c\in C_n(X)$ a singular $n$-chain with $k\le n$. Then

$$\partial (\varphi ⌢c)=(-1{)}^k(\delta \varphi ⌢c-\varphi ⌢\partial c)\in C_{n-k-1}(X).$$

Consequently the cap descends to a pairing $⌢:H^k(X)\otimes H_n(X)\to H_{n-k}(X)$, and the descent is bilinear and natural in $X$.

The Leibniz rule is the chain-level identity that makes the cap a homological invariant. Its proof is a computation on a single simplex via the alternating-face formula for $\partial$ and $\delta$; the sign $(-1{)}^k$ tracks the cochain degree and is exactly what is needed to align with the cup-product Leibniz rule on the cohomology side.

Theorem (cup-cap compatibility). For $\varphi \in H^k(X)$, $\psi \in H^{\ell }(X)$, $\alpha \in H_n(X)$,

$$\varphi ⌢(\psi ⌢\alpha )=(\psi ⌣\varphi )⌢\alpha \in H_{n-k-\ell }(X).$$

The identity exhibits $H_(X)$asagradedmoduleoverthecohomologyring$H^(X) = \bigoplus_k H^k(X)$ with the cup product.

Theorem (projection formula / naturality). For a continuous map $f:X\to Y$, $\varphi \in H^k(Y)$, $\alpha \in H_n(X)$,

$$f_{\ast }(f^{\ast }\varphi ⌢\alpha )=\varphi ⌢f_{\ast }\alpha \in H_{n-k}(Y).$$

The cap is a natural transformation between bifunctors $\mathbf{Top}\to \mathbf{Ab}$, contravariant in cohomology and covariant in homology.

Theorem (Poincaré duality via cap). Let $M$ be a closed oriented $n$-manifold with fundamental class $[M]\in H_n(M;\mathbb{Z})$. The map

$$D=-⌢[M]:H^k(M;\mathbb{Z})\to H_{n-k}(M;\mathbb{Z})$$

is an isomorphism for every $k$. For non-compact oriented manifolds the corresponding statement uses compactly supported cohomology and Borel-Moore homology with the appropriate fundamental class.

The duality theorem identifies the cohomology and homology of a closed oriented manifold via a single class — the fundamental class — and the cap pairing. The classical proof uses cellular cap on a CW decomposition of $M$, descent through the chain-level Poincaré-Lefschetz argument, and a Mayer-Vietoris induction.

Theorem (cellular cap product). Let $X$ be a CW complex with cellular chain complex $C_{\bullet }^{\mathrm{cell}}(X)$ and cellular cochain complex $C_{\mathrm{cell}}^{\bullet }(X)$. There is a cellular cap product

$${⌢}^{\mathrm{cell}}:C_{\mathrm{cell}}^k(X)\otimes C_n^{\mathrm{cell}}(X)\to C_{n-k}^{\mathrm{cell}}(X)$$

compatible with the singular cap via the cellular-to-singular comparison map. The cellular cap is constructed via a diagonal approximation on the cellular chain complex, computed cell-by-cell using the attaching maps.

The cellular cap is what makes Poincaré duality computable in practice. On a CW manifold with a single top cell, the cellular cap with the fundamental class is a near-identity map on the dual cell decomposition; the duality isomorphism becomes an explicit pairing of cells with their dual cells. This is the modern formulation of the classical Poincaré-Lefschetz duality argument.

Theorem (cap in spectral sequences). For a Serre fibration $F\to E\to B$ with simply-connected $B$, the Leray-Serre spectral sequence carries a multiplicative structure. The cap product $H^(E) \otimes H_*(E) \to H_*(E)$inducesacaponthespectralsequencepages:thecuponthe$E_2$page$H^(B; H^(F)) \otimes H^(B; H^(F))$capswiththehomologySerrespectralsequenceconvergingto$H_*(E)$, and the resulting graded module structure is compatible with the cap on the abutment.*

The spectral-sequence cap is the source of many computations in fibre-bundle homology, including the cohomology of classifying spaces $BG$ for compact Lie groups via the fibration $G\to EG\to BG$.

Synthesis. The cap product is the foundational reason cohomology and homology pair into a module structure rather than sitting as parallel but unrelated functors. The central insight is that the cap is dual to the cup in a precise sense: cup is the diagonal-induced product on cohomology, cap is the diagonal-induced module action of cohomology on homology, and the two together generalise the Kronecker evaluation pairing $H^k\otimes H_k\to \mathbb{Z}$ that already appears at the chain level. This is exactly the same diagonal-approximation principle that appears again in 03.12.13 (cellular homology) when constructing the cellular cap, and the bridge is that diagonal approximations are unique up to acyclic-models chain homotopy. Putting these together, the cap exhibits the homology of any space as a graded module over its cohomology ring, and the Poincaré duality theorem is the statement that for closed oriented manifolds this module is free of rank one with generator the fundamental class.

This is exactly the same structural pattern that appears again in 03.12.16 (Poincaré duality), where capping with $[M]$ identifies cohomology with homology in complementary degrees, and the bridge is the recognition that the cap is the natural pairing through which any cohomology operation acts on homology classes — including the Steenrod operations on $H^{\ast }(X;{\mathbb{F}}_p)$ and the secondary cohomology operations of higher degree. The cap product generalises to spectra: the smash-product pairing of a spectrum with its dual recovers the cap product on the underlying graded abelian groups, and the projection formula becomes a statement about smash products and duality in the stable homotopy category. Putting these together, the cap is what links cohomology theory with homology theory at every level of generality, and the identification of homology with cohomology under Poincaré duality identifies intersection-theoretic data with cohomological data on closed oriented manifolds.

Full proof set [Master]

Theorem (Leibniz rule), proof. Compute $\partial (\varphi ⌢\sigma )$ on a single $n$-simplex $\sigma$ with $\varphi \in C^k(X)$. By definition $\varphi ⌢\sigma =\varphi (\sigma {|}_{[v_0,\dots ,v_k]})\cdot \sigma {|}_{[v_k,\dots ,v_n]}$, an $(n-k)$-chain with integer coefficient $\varphi (\sigma {|}_{[v_0,\dots ,v_k]})$. Apply $\partial$:

$$\partial (\varphi ⌢\sigma )=\varphi (\sigma {|}_{[v_0,\dots ,v_k]})\cdot \partial \left(\sigma {|}_{[v_k,\dots ,v_n]}\right)=\varphi (\sigma {|}_{[v_0,\dots ,v_k]})\cdot \sum _{i=k}^n(-1{)}^{i-k}\sigma {|}_{[v_k,\dots ,{\hat{v}}_i,\dots ,v_n]}.$$

The right side: $\delta \varphi \in C^{k+1}(X)$ has $\delta \varphi (\tau )=\varphi (\partial \tau )$. The cap is

$$\delta \varphi ⌢\sigma =\delta \varphi (\sigma {|}_{[v_0,\dots ,v_{k+1}]})\cdot \sigma {|}_{[v_{k+1},\dots ,v_n]}=\varphi (\partial \sigma {|}_{[v_0,\dots ,v_{k+1}]})\cdot \sigma {|}_{[v_{k+1},\dots ,v_n]}.$$

Expand $\partial \sigma {|}_{[v_0,\dots ,v_{k+1}]}={\sum }_{j=0}^{k+1}(-1{)}^j\sigma {|}_{[v_0,\dots ,{\hat{v}}_j,\dots ,v_{k+1}]}$. The $j=k+1$ term is $(-1{)}^{k+1}\sigma {|}_{[v_0,\dots ,v_k]}$ and contributes $(-1{)}^{k+1}\varphi (\sigma {|}_{[v_0,\dots ,v_k]})\cdot \sigma {|}_{[v_{k+1},\dots ,v_n]}$ to $\delta \varphi ⌢\sigma$. The remaining $j\le k$ terms reorganise via the boundary of $\sigma$ on the front face.

The other piece $\varphi ⌢\partial \sigma$ expands as ${\sum }_{i=0}^n(-1{)}^i\varphi ⌢\sigma {|}_{[v_0,\dots ,{\hat{v}}_i,\dots ,v_n]}$. Splitting the sum at $i=k$ and $i=k+1$ partitions the terms into those that match the $j\le k$ part of $\delta \varphi ⌢\sigma$ and those that match $\partial$ acting on the back face. The signed combination $(-1{)}^k(\delta \varphi ⌢\sigma -\varphi ⌢\partial \sigma )$ collects the front-face boundary with sign $+1$ and the back-face boundary with sign $-1$; matching against $\partial (\varphi ⌢\sigma )=\varphi (\sigma {|}_{[v_0,\dots ,v_k]})\cdot \partial \sigma {|}_{[v_k,\dots ,v_n]}$ produces the identity term-by-term. The full computation is in Hatcher §3.3, with the explicit signed match presented as an alternating-face bookkeeping argument ^[pending]. $\square$

Theorem (cup-cap compatibility), proof. Given in the Key Theorem section. The two-step front-face / back-face expansion produces the same surviving back face on both sides, with both cochain evaluations consuming the matching subset of front vertices. The chain-level identity descends to (co)homology via the Leibniz rule for both caps. $\square$

Theorem (projection formula), proof. On a single simplex $\sigma :{\Delta }^n\to X$, the chain-level pullback $f^{\ast }\varphi (\tau )=\varphi (f\circ \tau )$ gives

$$f^{\ast }\varphi ⌢\sigma =\varphi (f\circ \sigma {|}_{[v_0,\dots ,v_k]})\cdot \sigma {|}_{[v_k,\dots ,v_n]}.$$

Applying $f_{\ast }$ pushes the back face to $f\circ \sigma {|}_{[v_k,\dots ,v_n]}=(f\circ \sigma ){|}_{[v_k,\dots ,v_n]}$, while the integer coefficient $\varphi (f\circ \sigma {|}_{[v_0,\dots ,v_k]})=\varphi ((f\circ \sigma ){|}_{[v_0,\dots ,v_k]})$ is exactly the value of $\varphi$ on the front face of $f\circ \sigma$. The result is $\varphi ⌢(f\circ \sigma )=\varphi ⌢f_{\ast }\sigma$. Linear extension and descent to homology via the Leibniz rule produce the homology-level identity. $\square$

Theorem (Poincaré duality via cap), stated without proof in full generality — see Hatcher §3.3 ^[pending]. The proof for closed oriented manifolds: cellular CW structure on $M$, cellular cap with $[M]$ on the chain level, identification of the dual cell decomposition with the cellular chains of the dual CW structure, Mayer-Vietoris induction along the open-cover decomposition $M=U\cup V$ with $U,V$ unions of open cells. The non-compact and twisted-coefficient versions use compactly supported cohomology and the orientation local system; the full development is in Hatcher §3.3 and Bredon §VI.

Theorem (cellular cap product), proof sketch. The cellular chain complex $C_{\bullet }^{\mathrm{cell}}(X)=H_{\bullet }(X^{(\bullet )},X^{(\bullet -1)})$ admits a diagonal approximation ${\Delta }^{\mathrm{cell}}:C_{\bullet }^{\mathrm{cell}}(X)\to C_{\bullet }^{\mathrm{cell}}(X)\otimes C_{\bullet }^{\mathrm{cell}}(X)$ constructed cell-by-cell from the attaching maps. The acyclic-models theorem of Eilenberg-Mac Lane, applied to the category of CW complexes, produces a unique-up-to-chain-homotopy diagonal extending the identity on the $0$-skeleton. The cellular cap $\varphi {⌢}^{\mathrm{cell}}c=(\varphi \otimes \mathrm{id})\circ {\Delta }^{\mathrm{cell}}(c)$ inherits the Leibniz rule from the singular cap via the comparison map $C_{\bullet }^{\mathrm{cell}}\to C_{\bullet }^{\mathrm{sing}}$, which is a chain-homotopy equivalence by the cellular-singular agreement theorem. $\square$

Theorem (cap in spectral sequences), stated without proof — see McCleary A User's Guide to Spectral Sequences §5 ^[pending]. The multiplicative structure on the Leray-Serre spectral sequence carries the cup product on each page; the homology spectral sequence carries the corresponding cap action of the cohomology spectral sequence; convergence at $E_{\infty }$ matches the cup and cap on the abutment by a standard filtration argument. The full development requires the multiplicative-spectral-sequence machinery and is treated in McCleary, Hatcher §4.D, and Spanier §IX.

Connections [Master]

• Singular homology 03.12.11. The cap product is defined on singular chains and cochains, pairing $C^k(X)\otimes C_n(X)\to C_{n-k}(X)$ via the front-face / back-face split. The descent to homology relies on the boundary $\partial$ and the Leibniz rule, both of which live at the level of singular chains. The chain-level definition specialises the diagonal map $X\to X\times X$ via the Alexander-Whitney formula, and the resulting bilinear pairing is the chain-level engine that produces the $H^{\ast }$-module structure on $H_{\ast }$.

• Cellular homology 03.12.13. The cellular cap product is constructed via a diagonal approximation on the cellular chain complex, computed cell-by-cell using the attaching maps. The cellular and singular caps agree via the cellular-singular comparison theorem, which is itself a chain-homotopy equivalence proved by acyclic models. This is the version of the cap that makes Poincaré duality computable on a CW manifold: the dual cell decomposition pairs with the original under the cellular cap with the fundamental class.

• Poincaré duality 03.12.16. The cap product with the fundamental class $[M]$ of a closed oriented $n$-manifold gives the Poincaré duality isomorphism $D:H^k(M)\to H_{n-k}(M)$. This is the load-bearing application of the cap product and the structural reason cohomology and homology of a closed oriented manifold are linked by a single rank-one module structure. Naturality of $D$ in continuous maps is the projection formula; module structure is the cup-cap compatibility.

• Singular cohomology 03.04.13. Cup product on cohomology and cap product on homology are companion operations, both arising from the diagonal map $X\to X\times X$ via the Alexander-Whitney formula. The cup gives $H^{\ast }(X)$ a graded ring structure; the cap gives $H_{\ast }(X)$ a graded module structure over that ring; the cup-cap compatibility makes the module action associative.

• Eilenberg-Steenrod axioms 03.12.15. The cap product is part of the multiplicative structure of an ordinary homology theory. The Eilenberg-Steenrod axiomatic framework treats cup and cap together, with the seven axioms supplemented by the multiplicative axioms of associativity, commutativity (for cup), and the projection formula (for cap). Generalised homology theories — $K$-theory, bordism, stable homotopy — carry their own cup and cap products, with the cap as the module action of the cohomology ring on the homology module.

• Spectra and stable homotopy 03.12.04. The cap product generalises to spectra: the smash-product pairing of a spectrum $E$ with its dual $DE$ recovers the cap product on the underlying graded abelian groups, and the projection formula becomes a statement about smash products and Spanier-Whitehead duality in the stable homotopy category. Generalised Poincaré duality for closed oriented manifolds in $E$-theory follows from Atiyah duality ${\Sigma }^{\infty }{M}_{+}\simeq {\Sigma }^{\infty }{M}^{TM}\wedge S^{-n}$.

Historical & philosophical context [Master]

The cup-product framework on cohomology was introduced by Eduard Čech in 1936 (Multiplications on a complex, Ann. Math. 37, 681-697) ^[pending] and refined by Hassler Whitney in 1938 (On products in a complex, Ann. Math. 39, 397-432) ^[pending]. Whitney's paper introduced the modern combinatorial cup product on simplicial cochains and identified the dual operation on chains — what is now called the cap product — as the structurally companion construction. The cap was defined by Whitney in the simplicial setting; the corresponding construction on singular chains required the development of singular homology theory, which was completed by Eilenberg in 1944 (Singular homology theory, Ann. Math. 45, 407-447) ^[pending].

Solomon Lefschetz, in his 1942 monograph Algebraic Topology (AMS Colloquium Publications XXVII) ^[pending], gave the first systematic treatment of cap products in the general singular setting and used them to formalise intersection theory on manifolds. Lefschetz's intersection-theoretic interpretation — capping a cohomology class with the fundamental class of a manifold and reading the result as the intersection of representing submanifolds — is the geometric content of Poincaré duality, made algebraic by the cap product. Lefschetz himself had introduced the duality theorem in 1926 (Intersections and transformations of complexes and manifolds, Trans. AMS 28, 1-49) for combinatorial manifolds; the cap-product formulation was the modern singular-theoretic reformulation.

The axiomatic treatment of cap products appears in Eilenberg and Steenrod's 1952 Foundations of Algebraic Topology (Princeton) ^[pending], §V. The multiplicative structure axioms supplement the seven Eilenberg-Steenrod axioms and characterise cup and cap on ordinary homology theories with coefficients in a commutative ring. The Eilenberg-Mac Lane acyclic-models theorem (Eilenberg-Mac Lane 1953, Acyclic models, Amer. J. Math. 75, 189-199) ^[pending] supplies the existence and uniqueness of diagonal approximations on the singular and cellular chain complexes, which is the modern chain-level foundation of both cup and cap. The generalisation of cap product to extraordinary cohomology theories appears in Adams's 1974 monograph Stable Homotopy and Generalised Homology (University of Chicago Press) ^[pending], where the Spanier-Whitehead duality framework places cap product at the centre of every duality theorem in stable homotopy theory.

Bibliography [Master]

[object Promise]