Cohomology Rings

⚡ Key Takeaways

The cohomology ring of a topological space $X$ is the graded ring $H^*(X;R) = \bigoplus_{n\geq 0} H^n(X;R)$ obtained by equipping cohomology with the cup product, a graded-commutative multiplication. This ring structure is a far stronger invariant than the cohomology groups alone, distinguishing spaces with identical Betti numbers. It applies throughout algebraic topology, algebraic geometry, and mathematical physics. The cup product was introduced in the 1930s–40s by Alexander, Kolmogorov, Čech, and Whitney.

💡 The Core Insight

Cohomology is contravariant and carries a natural diagonal-induced product that homology lacks. The diagonal map $\Delta: X \to X \times X$ , $x \mapsto (x,x)$ , pulls back the external (cross) product on $H^*(X\times X)$ to an internal product on $H^*(X)$ . Because the diagonal is intrinsic to any space, every space's cohomology is automatically a ring — and ring homomorphisms induced by maps make $H^*(-;R)$ a contravariant functor into graded rings.

DefinitionSingular Cochains and Cup Product

Let $X$ be a topological space and $R$ a commutative ring. The singular cochain group is $C^n(X;R) = \mathrm{Hom}(C_n(X), R)$ , where $C_n(X)$ is the free abelian group on singular $n$ -simplices $\sigma : \Delta^n \to X$ . For cochains $\varphi \in C^p$ , $\psi \in C^q$ , the cup product $\varphi \smile \psi \in C^{p+q}$ is defined on a singular $(p+q)$ -simplex $\sigma$ by

1

\quad (\varphi \smile \psi)(\sigma) = \varphi\big(\sigma|_{[v_0,\dots,v_p]}\big)\cdot \psi\big(\sigma|_{[v_p,\dots,v_{p+q}]}\big)

Derivation hint: split the standard simplex along the vertex $v_p$ into a "front $p$ -face" and a "back $q$ -face"; the product on the right is multiplication in $R$ . This is the Alexander–Whitney formula.

DefinitionThe Cohomology Ring

The coboundary $\delta : C^n \to C^{n+1}$ is the transpose of the simplicial boundary, $(\delta\varphi)(\sigma) = \varphi(\partial\sigma)$ . The cup product satisfies the Leibniz rule

2

\quad \delta(\varphi \smile \psi) = (\delta\varphi)\smile\psi + (-1)^{|\varphi|}\,\varphi\smile(\delta\psi)

Derivation hint: expand both sides using the Alexander–Whitney formula and the alternating-sum definition of $\partial$ ; the cross terms cancel telescopically along the splitting vertex. Consequently the cup product of cocycles is a cocycle, and cup with a coboundary is a coboundary, so $\smile$ descends to a well-defined product on cohomology classes. The cohomology ring is

3

\quad H^*(X;R) = \bigoplus_{n=0}^{\infty} H^n(X;R), \qquad [\varphi]\smile[\psi] = [\varphi\smile\psi]

Derivation hint: well-definedness follows directly from [2] — replacing $\varphi$ by $\varphi + \delta\eta$ changes the product by a coboundary, hence the same cohomology class.

TheoremGraded-Commutativity

For $\alpha \in H^p(X;R)$ and $\beta \in H^q(X;R)$ ,

4

\quad \alpha \smile \beta = (-1)^{pq}\,\beta \smile \alpha

so $H^*(X;R)$ is a graded-commutative (anticommutative) ring with unit $1 \in H^0(X;R)$ represented by the cochain assigning $1\in R$ to every $0$ -simplex.

Proof

The cochain-level cup product is not commutative — only commutative up to chain homotopy. One constructs explicit homotopy operators interpolating between front-back splittings. 1. Define for each $n$ a "reversal" chain map $\rho_n: C_n(X)\to C_n(X)$ sending a simplex $[v_0,\dots,v_n]$ to $(-1)^{n(n+1)/2}[v_n,\dots,v_0]$ . One checks $\rho$ is chain-homotopic to the identity via a standard prism/subdivision operator $P$ . 2. Compute that $\rho^*(\varphi\smile\psi)$ relates to $(\rho^*\psi)\smile(\rho^*\varphi)$ with the sign $(-1)^{pq}$ , by reversing vertex orders and tracking the sign $(-1)^{n(n+1)/2}$ with $n = p+q$ . 3. Since $\rho \simeq \mathrm{id}$ , passing to cohomology kills the homotopy term $P$ , and $\rho^* = \mathrm{id}$ on $H^*$ . The surviving identity is precisely $\alpha\smile\beta = (-1)^{pq}\beta\smile\alpha$ . The sign bookkeeping is the heart: $(-1)^{p(p+1)/2}(-1)^{q(q+1)/2}(-1)^{(p+q)(p+q+1)/2} = (-1)^{pq}$ . $\blacksquare$

∎

mathematician · J. W. Alexander & Hassler Whitney | 1888–1971, 1907–1989

James Waddell Alexander II proposed an early product on cohomology in 1936; independently Andrei Kolmogorov announced similar ideas the same year at the Moscow Topology Conference, where a famous error (the product failed associativity in the original form) was discovered and corrected. Eduard Čech and Hassler Whitney refined the construction; the explicit Alexander–Whitney formula [1] — splitting a simplex at a vertex into front and back faces — became the standard cochain-level definition. Whitney also coined the term "cup product" (from the symbol $\smile$ , a cup), with the dual "cap product" $\frown$ describing the homology module structure.

🧠 Intuition

Think of a cohomology class $\alpha \in H^p(X)$ as a "measuring device" that pairs with $p$ -dimensional cycles. The cup product $\alpha \smile \beta$ is a combined measurement of dimension $p+q$ . The cleanest geometric picture is Poincaré duality for a closed oriented $n$ -manifold $M$ : a class in $H^p(M)$ corresponds to an $(n-p)$ -dimensional submanifold (its Poincaré dual). Under this dictionary, cup product becomes geometric intersection: if $\alpha$ is dual to submanifold $A$ and $\beta$ dual to $B$ , then $\alpha\smile\beta$ is dual to $A\cap B$ (after perturbing $A, B$ into transverse position). The dimension count $\dim(A\cap B) = (n-p)+(n-q)-n = n-(p+q)$ exactly matches the degree $p+q$ of $\alpha\smile\beta$ . The graded-commutative sign $(-1)^{pq}$ is the sign incurred when reordering the transverse intersection. So the cohomology ring is, morally, the algebra of how submanifolds intersect.

Figure

Cup product via transverse intersection of Poincaré-dual submanifolds on a torus

```svg

```

ExampleThe 2-Torus

T^2 = S^1 \times S^1

Given: Compute $H^*(T^2;\mathbb{Z})$ as a ring.

Steps: 1. By the Künneth theorem, $H^*(T^2) \cong H^*(S^1)\otimes H^*(S^1)$ as rings. For $S^1$ : $H^*(S^1;\mathbb{Z}) = \mathbb{Z}[\alpha]/(\alpha^2)$ with $|\alpha|=1$ (an exterior algebra on one generator). 2. Hence with generators $a,b$ in degree 1 (one from each factor):

5

\quad H^*(T^2;\mathbb{Z}) \cong \Lambda_{\mathbb{Z}}[a,b] = \mathbb{Z}[a,b]/(a^2,\,b^2,\,ab+ba)

Derivation hint: Künneth for cohomology of finite-type spaces over $\mathbb{Z}$ gives a ring isomorphism with the tensor product of cohomology rings; the sign in $ab+ba$ comes from [4] with $p=q=1$ . 3. The additive ranks are $1,2,1$ in degrees $0,1,2$ . The key relation $a^2 = 0$ , $b^2 = 0$ but $a\smile b = -\,b\smile a \neq 0$ generates $H^2$ .

Answer: $\boxed{H^*(T^2;\mathbb{Z}) = \Lambda_{\mathbb{Z}}[a,b],\ |a|=|b|=1,\ ab = -ba \text{ generates } H^2 \cong \mathbb{Z}}$

ExampleReal Projective Space

\mathbb{RP}^n

with

\mathbb{Z}/2

Coefficients

Given: Compute $H^*(\mathbb{RP}^n;\mathbb{Z}/2)$ as a ring.

Steps: 1. The $\mathbb{Z}/2$ -cohomology groups are $H^k(\mathbb{RP}^n;\mathbb{Z}/2) = \mathbb{Z}/2$ for $0\le k\le n$ and $0$ otherwise. Let $\omega$ generate $H^1$ . 2. One shows by induction (e.g., via the CW structure and the Gysin sequence of the double cover $S^n \to \mathbb{RP}^n$ ) that $\omega^k$ generates $H^k$ for each $k\le n$ :

6

\quad H^*(\mathbb{RP}^n;\mathbb{Z}/2) \cong (\mathbb{Z}/2)[\omega]/(\omega^{n+1}),\qquad |\omega|=1

Derivation hint: the cellular cochain complex has zero differentials mod 2; the cup-square $\omega^2 = \omega\smile\omega$ is computed nonzero on $\mathbb{RP}^2$ directly from [1] on the standard CW/simplicial structure, then powers propagate. 3. Note that mod 2 there is no sign obstruction: $\omega^2 = \omega\smile\omega$ is honestly $\omega^2$ (since $-1=1$ ). This is a truncated polynomial ring.

Answer: $\boxed{H^*(\mathbb{RP}^n;\mathbb{Z}/2) = (\mathbb{Z}/2)[\omega]/(\omega^{n+1})}$ — a polynomial ring truncated at degree $n+1$ . For comparison, $\mathbb{CP}^n$ gives $H^*(\mathbb{CP}^n;\mathbb{Z}) = \mathbb{Z}[x]/(x^{n+1})$ with $|x|=2$ .

⚠️ Common mistake

The cohomology ring is not determined by the cohomology groups. Two spaces can have isomorphic graded cohomology groups in every degree yet non-isomorphic cohomology rings. The classic case: $\mathbb{RP}^2$ versus $S^1 \vee S^2$ over $\mathbb{Z}/2$ . Both have $H^0=H^1=H^2=\mathbb{Z}/2$ , but in $\mathbb{RP}^2$ the generator $\omega$ satisfies $\omega^2 \neq 0$ , while in $S^1\vee S^2$ the degree-1 generator squares to zero (its dual circle has self-intersection $0$ , and there is no 2-cell glued along it). Thus the ring structure is genuinely extra information.

Do not assume the cup product is strictly commutative. It is graded-commutative — the sign $(-1)^{pq}$ is essential. Odd-degree classes anticommute and square to zero over $\mathbb{Q}$ or $\mathbb{Z}$ (since $\alpha^2 = -\alpha^2 \Rightarrow 2\alpha^2 = 0$ ), but over $\mathbb{Z}/2$ they need not.

Counterexample

Cohomology cannot always distinguish a product from a twisted bundle additively, but the ring can. Consider the Klein bottle $K$ versus the torus $T^2$ . With $\mathbb{Z}/2$ coefficients both have Poincaré polynomial $1 + 2t + t^2$ . Yet their integral cohomology rings differ: $H^2(T^2;\mathbb{Z})=\mathbb{Z}$ (orientable) while $H^2(K;\mathbb{Z})=\mathbb{Z}/2$ (non-orientable), and even the $\mathbb{Z}/2$ ring structures differ in how generators multiply. The product structure detects orientability and twisting that the dimensions hide. Conversely, the connected sum $\mathbb{CP}^2 \# \overline{\mathbb{CP}}^2$ and $S^2\times S^2$ have the same Betti numbers $(1,0,2,0,1)$ but distinct intersection forms (cup products on $H^2$ ): $\mathrm{diag}(1,-1)$ versus $\begin{pmatrix}0&1\\1&0\end{pmatrix}$ — these are not isomorphic over $\mathbb{Z}$ , so the spaces are not homotopy equivalent.

connection

The cohomology ring sits at the crossroads of many structures:

<a class="wiki-link" href="/wiki/poincare-duality">Poincaré Duality</a> turns the cup product into the intersection form; for closed oriented $4k$ -manifolds the middle-dimensional pairing gives a symmetric unimodular form, a deep invariant (Donaldson, Freedman).
Künneth Theorem computes $H^*(X\times Y)$ as a tensor product of rings, making the external/cross product the engine behind the internal cup product via the diagonal.
Steenrod Operations ( $Sq^i$ , $P^i$ ) refine the ring with natural cohomology operations; they are built from the failure of cup product to be commutative on the cochain level (the cup- $i$ products).
Cap Product makes homology a module over the cohomology ring, encoding the duality pairing.
Spectral Sequences (Serre, Eilenberg–Moore) are multiplicative: they converge to the cohomology ring with all products, not just groups.
In Algebraic Geometry, the Chow ring and de Rham cohomology ring with the wedge product play analogous roles; the cup product corresponds to $\wedge$ on differential forms via the de Rham isomorphism.

🔬 Where this appears

⏳ History

- 1895–1904 — Henri Poincaré founds homology and states duality; the dual product structure is implicit but not yet formalized.

- 1925–1930 — Emmy Noether reframes homology in terms of abelian groups; Leopold Vietoris and others develop cohomology precursors.

- 1935 — Andrey Kolmogorov and James Alexander independently define cohomology and propose a product. Kolmogorov's first definition was non-associative.

- 1936 (Moscow Topology Conference) — Alexander and Kolmogorov present; the associativity flaw is identified and the corrected cup product emerges from the discussions involving Čech.

- 1938 — Hassler Whitney systematizes the cup product notation and the front-face/back-face (Alexander–Whitney) formula; introduces "cup" and "cap."

- 1944 — Samuel Eilenberg gives a rigorous singular-cohomology foundation; the ring structure becomes standard.

- 1947 — Norman Steenrod introduces the Steenrod squares

Sq^i

via cup-

i

products, deepening the algebra of the cochain-level product's non-commutativity.

- 1950 — Heinz Hopf and Armand Borel classify the cohomology rings of Lie groups and

H

-spaces (Hopf algebras); Borel computes

H^*(BG)

.

- 1982–2003 — Simon Donaldson and Michael Freedman use the cup-product intersection form to revolutionize 4-manifold topology; exotic

\mathbb{R}^4

's emerge.

open-problem

The cohomology ring is a complete invariant in some categories but not in general. Two simply connected spaces can share isomorphic cohomology rings (and even isomorphic rational homotopy types of the ring) yet fail to be homotopy equivalent because of higher-order operations: Massey products and Steenrod operations carry information beyond the ring. A landmark subtlety is formality: a space is formal if its rational homotopy type is determined by its cohomology ring (its singular cochain DGA is quasi-isomorphic to its cohomology with zero differential). Compact Kähler manifolds are formal (Deligne–Griffiths–Morgan–Sullivan, 1975), but general symplectic manifolds need not be — the first non-formal closed symplectic manifolds were constructed to settle this. Detecting non-formality requires Massey products, the first secondary operations refining the cup product.

Open territory. Determining which graded-commutative rings arise as cohomology rings of closed manifolds (the realization problem), and classifying spaces up to homotopy from ring-plus-secondary-operations data, remain active. The interplay of the cohomology ring with the Steenrod algebra action (the structure of an unstable algebra) constrains realizability in ways not fully understood, e.g., Hovey–Palmieri–Strickland program and the still-incomplete classification of polynomial cohomology rings (related to the resolved Adams–Wilkerson and Steenrod realization questions).

Formal Statement

Building Intuition

Worked Examples

Common Misconceptions

Connections

Real-World Applications

Historical Development

Subtleties & Limits

See Also