03.13.04 · modern-geometry / spectral-sequences

Atiyah-Hirzebruch spectral sequence

shipped3 tiersLean: none

Anchor (Master): Atiyah-Hirzebruch 1961 *Vector bundles and homogeneous spaces*; Atiyah 1961 *Characters and cohomology of finite groups* (Publ. IHÉS 9); Adams 1974 *Stable Homotopy and Generalised Homology* Pt. III §7-§9; Switzer *Algebraic Topology — Homotopy and Homology* Ch. 15; Lawson-Michelsohn *Spin Geometry* §IV.1

Intuition [Beginner]

The Atiyah-Hirzebruch spectral sequence answers a basic question: once you know the ordinary cohomology of a space, what does its K-theory look like? More generally, every reasonable cohomology theory — K-theory, cobordism, elliptic cohomology, and many others — admits a spectral sequence that takes ordinary cohomology as input and produces the new theory as output. The Atiyah-Hirzebruch construction is the universal template for this passage.

The input is the cellular structure of a space. Each cell contributes ordinary cohomology in one degree, and the spectral sequence stacks these contributions according to how cells of different dimensions interact. The first page records the naive answer, where every cell contributes independently. Later pages introduce corrections that account for how cells of different dimensions are glued together. After enough pages, the corrections stop and the surviving entries assemble into the K-theory you wanted.

What makes this useful is that ordinary cohomology is computable. Cellular cohomology gives an explicit recipe with one generator per cell. K-theory is harder: bundles can twist in ways that singular cohomology cannot see. The spectral sequence lets you compute the easier quantity and read off the harder one through a finite sequence of corrections, each of which is itself a cohomology operation.

Visual [Beginner]

Picture a grid where each row records a slice of cohomology — say K-theory at a fixed cell dimension. The rows are repeating because K-theory of a point is the same in every even degree. Diagonal arrows connect entries on each page, and the arrows get longer as you turn to higher pages. Eventually no more arrows fit on the grid, and the entries that survive on every page assemble into the K-theory of the space.

The picture captures the central pattern: the rows are the coefficients of K-theory on a point, and the columns are the cellular cohomology of the space. The differentials are cohomology operations on the page — for K-theory the simplest one is the third Steenrod square. After all differentials act, what remains is K-theory of the space, organised by cell dimension.

Worked example [Beginner]

Compute the K-theory of the complex projective plane $CP^{2}$ from its ordinary cohomology, using the Atiyah-Hirzebruch spectral sequence.

Step 1. The cohomology of $CP^{2}$ is $H^{0} = Z$ , $H^{2} = Z$ , $H^{4} = Z$ , and zero in all other degrees. The space has one cell in each of dimensions zero, two, and four.

Step 2. The $E_{2}$ page of the spectral sequence has entries $E_{2}^{p, q} = H^{p} (CP^{2}; K^{q} (pt))$ . K-theory of a point is $Z$ in every even degree and zero in every odd degree, by Bott periodicity. So the $E_{2}$ page has $Z$ at $(0, 2 k)$ , $(2, 2 k)$ , and $(4, 2 k)$ for every integer $k$ , and zero elsewhere.

Step 3. Look for possible differentials. The differential $d_{2}$ would go from $(p, q)$ to $(p + 2, q - 1)$ , but the target has an odd second coordinate, where everything is zero. So $d_{2}$ vanishes and $E_{3} = E_{2}$ . The differential $d_{3}$ would go from $(p, q)$ to $(p + 3, q - 2)$ , but the source has even first coordinate ( $p = 0, 2, 4$ ) and the target has odd first coordinate ( $p + 3 = 3, 5, 7$ ) where everything is zero. So $d_{3}$ vanishes as well, and the same argument kills $d_{4}, d_{5}, \dots$ on this small space.

Step 4. The spectral sequence collapses at $E_{2}$ . The K-theory of $CP^{2}$ is the diagonal sum across the surviving page. Summing along $p + q = 0$ gives $K^{0} (CP^{2}) = Z^{3}$ , one factor from each cell. Summing along $p + q = 1$ gives $K^{1} (CP^{2}) = 0$ .

What this tells us: K-theory of $CP^{2}$ has rank three, one generator per cell. The same argument extends to $CP^{n}$ to give rank $n + 1$ . The spectral sequence reduces a K-theory computation to a cohomology computation and a check that no differentials act, and on simple spaces the differentials act for dimensional reasons alone.

Check your understanding [Beginner]

Exercise (easy, multiple choice).

The Atiyah-Hirzebruch spectral sequence in K-theory is named for a 1961 paper. What was the original application?

A. Computing the homotopy groups of spheres
B. K-theory of homogeneous spaces and the relation to ordinary cohomology
C. The Atiyah-Singer index theorem
D. The proof of Bott periodicity

Hint

The 1961 paper's title is Vector bundles and homogeneous spaces; that hints at what they were computing.

Answer

B. K-theory of homogeneous spaces and the relation to ordinary cohomology.

Atiyah and Hirzebruch introduced the spectral sequence in 1961 as the tool for computing K-theory of compact homogeneous spaces $G / H$ in terms of ordinary cohomology. The Atiyah-Singer index theorem came later, and Bott periodicity (1959) was already established. The Hopf-invariant-one application of K-theory by Adams used a different technique, the Adams operations.

Formal definition [Intermediate+]

Definition (generalised cohomology theory). A generalised cohomology theory on the category of pointed CW pairs is a sequence of contravariant functors $E^{n} : CW_{*}^{2} \to Ab$ for $n \in Z$ , together with a natural connecting homomorphism $δ : E^{n} (A) \to E^{n + 1} (X, A)$ , satisfying the Eilenberg-Steenrod axioms with the dimension axiom omitted: homotopy invariance, the long exact sequence of a pair, excision, and the wedge axiom. The dimension axiom $E^{n} (pt) = 0$ for $n \neq = 0$ is dropped. The graded ring $E^{*} (pt)$ is the coefficient ring of the theory.

Two examples: ordinary cohomology with coefficients in an abelian group $G$ is the theory with $H^{*} (pt; G) = G$ concentrated in degree zero. Complex topological K-theory $K^{*}$ has coefficient ring $K^{*} (pt) = Z [β, β^{- 1}]$ with $β \in K^{- 2} (pt)$ the Bott element of degree $- 2$ , so $K^{q} (pt) = Z$ for $q$ even and $0$ for $q$ odd (Bott periodicity 03.08.07).

Definition (skeletal filtration). Let $X$ be a CW complex with skeleta $X^{(p)}$ . The skeletal filtration is the increasing sequence $$ \emptyset = X^{(-1)} \subseteq X^{(0)} \subseteq X^{(1)} \subseteq \cdots \subseteq X. $$ For each pair $(X^{(p)}, X^{(p - 1)})$ , the relative space $X^{(p)} / X^{(p - 1)}$ is a wedge of $p$ -spheres indexed by the $p$ -cells of $X$ .

Definition (Atiyah-Hirzebruch spectral sequence). Let $E^{*}$ be a generalised cohomology theory and $X$ a CW complex of finite type. The skeletal filtration of $X$ produces an exact couple in $E$ -cohomology by applying the long exact sequences of the pairs $(X^{(p)}, X^{(p - 1)})$ . The associated spectral sequence has $$ E_2^{p, q} = H^p(X; E^q(\mathrm{pt})), $$ differential $d_{r}$ of bidegree $(r, 1 - r)$ on each page, and converges to the graded object associated to the skeletal filtration on $E^{p + q} (X)$ : $$ E_2^{p, q} \Longrightarrow E^{p + q}(X). $$ For $X$ a finite CW complex, the spectral sequence has finitely many non-zero terms on each page and converges strongly.

Counterexamples to common slips

The spectral sequence converges to the graded associated to a filtration on $E^{*} (X)$ , not to $E^{*} (X)$ itself. Recovering the ungraded answer requires solving extension problems among the $E_{\infty}$ entries — for K-theory this matters when computing the ring structure on $K^{*} (X)$ , not the additive group.
The $E_{2}$ page uses cohomology of $X$ with $E^{q} (pt)$ as coefficient group, with $q$ fixed. The grading of $E^{*} (pt)$ as a graded ring carries through to the page indexing; for K-theory, only even $q$ have non-zero coefficients, but the page is still bigraded over all $q \in Z$ .
For periodic theories like K-theory, the spectral sequence is periodic in the vertical direction with period $2$ (or $4$ for real K-theory). The convergence statement reads $K^{n} (X) = ⨁_{p + q = n} E_{\infty}^{p, q}$ modulo extension, where each row at fixed $q$ repeats every two steps.
The differential $d_{3} = Sq^{3}$ for complex K-theory (after reduction modulo $2$ ) acts on $Z /2$ -coefficient cohomology. The integer-coefficient $d_{3}$ is the integral lift of this Steenrod square, namely the integral Steenrod cube $Sq_{Z}^{3}$ .

Key theorem with proof [Intermediate+]

Theorem (Atiyah-Hirzebruch). Let $E^ $b e a g e n er a l i se d co h o m o l o g y t h eor y an d$ X$ a CW complex of finite type. There is a strongly convergent spectral sequence* $$ E_2^{p, q} = H^p(X; E^q(\mathrm{pt})) \Longrightarrow E^{p + q}(X) $$ natural in $X$ , with differential $d_{r}$ of bidegree $(r, 1 - r)$ on $E_{r}$ .

Proof. The argument is a specialisation of Massey's exact-couple machinery (Massey 1952) to the skeletal filtration.

Define the bigraded exact couple as follows. For each pair of skeleta $(X^{(p)}, X^{(p - 1)})$ the long exact sequence of cohomology reads $$ \cdots \to E^n(X^{(p)}, X^{(p-1)}) \to E^n(X^{(p)}) \to E^n(X^{(p-1)}) \to E^{n+1}(X^{(p)}, X^{(p-1)}) \to \cdots. $$ Set $D^{p, q} = E^{p + q} (X^{(p)})$ and $E^{p, q} = E^{p + q} (X^{(p)}, X^{(p - 1)})$ . The three maps in the long exact sequence assemble into an exact couple $$ \cdots \to D^{p+1, q-1} \xrightarrow{i} D^{p, q} \xrightarrow{j} E^{p, q} \xrightarrow{k} D^{p+1, q} \to \cdots, $$ where $i$ is restriction, $j$ is the connecting map, and $k$ is the inclusion-induced map. The composite $d = j \circ k : E^{p, q} \to E^{p + 1, q}$ is the first differential.

The first identification: the relative $E$ -cohomology $E^{n} (X^{(p)}, X^{(p - 1)})$ is the reduced $E$ -cohomology $\tilde{E}^{n} (X^{(p)} / X^{(p - 1)})$ of the cofibre, which is a wedge of $p$ -spheres indexed by the $p$ -cells of $X$ . By the suspension isomorphism for $E$ , $$ \tilde E^n\left(\bigvee_\alpha S^p\right) = \bigoplus_\alpha E^{n - p}(\mathrm{pt}) = C^p_{\mathrm{cell}}(X; E^{n - p}(\mathrm{pt})), $$ the cellular cochain group of $X$ with coefficients in $E^{n - p} (pt)$ . So $E_{1}^{p, q} = C_{cell}^{p} (X; E^{q} (pt))$ with $q = n - p$ .

The next identification: the differential $d_{1} : E_{1}^{p, q} \to E_{1}^{p + 1, q}$ is the cellular coboundary, computed in $E$ -cohomology and identified with the integer cellular coboundary tensored with $E^{q} (pt)$ . Taking cohomology of $(E_{1}, d_{1})$ gives $E_{2}^{p, q} = H^{p} (X; E^{q} (pt))$ — cellular cohomology with $E^{q} (pt)$ coefficients, which equals singular cohomology with the same coefficients by 03.12.13.

Convergence. The derived couple converges in the bounded sense: $D^{p, q}$ for $p$ much larger than the dimension of $X$ vanishes (no cells beyond dimension $dim X$ ), and $D^{p, q}$ for $p \leq 0$ stabilises to $E^{p + q} (X)$ . By Massey's convergence theorem, the spectral sequence converges strongly to the associated graded of the filtration $F^{p} E^{n} (X) = ker (E^{n} (X) \to E^{n} (X^{(p - 1)}))$ .

Naturality. A cellular map $f : X \to Y$ induces compatible maps on the skeletal filtrations and hence a morphism of exact couples; the induced map on $E_{2}$ pages is $f^{*}$ on cellular cohomology with $E^{q} (pt)$ coefficients.

This is the spectral sequence stated. $□$

Bridge. The Atiyah-Hirzebruch spectral sequence builds toward the entire theory of generalised cohomology computations. The foundational reason it works is exactly that the skeletal filtration of a CW complex gives a clean exact couple in any generalised cohomology theory: each cell contributes a free generator on the $E_{1}$ page, the $E_{1}$ differential matches the cellular boundary, and the homology of $(E_{1}, d_{1})$ is therefore ordinary cohomology with $E^{*} (pt)$ coefficients. This is exactly the same algebraic mechanism that appears again in 03.13.02 (Leray-Serre), where filtering the total space of a fibration by base skeleta gives the analogous spectral sequence with $E_{2}^{p, q} = H^{p} (B; H^{q} (F))$ — generalised cohomology of a point replaced by ordinary cohomology of the fibre.

The central insight is that ordinary cohomology with arbitrary coefficient ring is the universal $E_{2}$ page from which any generalised cohomology theory is reconstructed by differentials encoding the deviation from a product. Putting these together, the AHSS identifies $K^{*} (X)$ with the graded associated to ordinary cohomology of $X$ tensored with $K^{*} (pt)$ , corrected by differentials whose lowest non-vanishing one is the integral Steenrod cube $Sq^{3}$ . The bridge to higher generalised theories is the recognition that the same construction works for complex cobordism, elliptic cohomology, and Morava K-theories: the $E_{2}$ page is always ordinary cohomology, and the differentials are always cohomology operations on that page. This pattern recurs throughout stable homotopy theory; the Adams spectral sequence of 03.12.21 is the same construction applied to the Postnikov filtration of a spectrum rather than the skeletal filtration of a space.

Exercises [Intermediate+]

Exercise 3 (medium, symbolic).

Compute $K^{*} (CP^{n})$ for $n \geq 1$ using the Atiyah-Hirzebruch spectral sequence, using $H^{*} (CP^{n}; Z) = Z [t] / (t^{n + 1})$ with $∣ t ∣ = 2$ .

Hint

The cohomology lives in even degrees only, and the K-theory coefficient ring lives in even degrees only. So every entry on the $E_{2}$ page has both indices even.

Answer

The $E_{2}$ page has $E_{2}^{p, q} = Z$ for $p$ even with $0 \leq p \leq 2 n$ and $q$ even, and zero otherwise. As in Exercise 1, every differential has odd total displacement on the $(p + q)$ -parity, so it lands in an odd-parity slot where the page vanishes. The page collapses at $E_{2}$ . Summing along $p + q = 0$ gives $n + 1$ copies of $Z$ — one for each cell of $CP^{n}$ . So $K^{0} (CP^{n}) ≅ Z^{n + 1}$ as abelian groups. Summing along $p + q = 1$ gives zero, so $K^{1} (CP^{n}) = 0$ . The multiplicative structure (ring structure) is recovered by solving the extension problem: with $L$ the tautological line bundle and $t = L - 1$ , the relation $(L - 1)^{n + 1} = 0$ in $K^{0}$ gives $K^{0} (CP^{n}) = Z [L] / ((L - 1)^{n + 1})$ . Rubric: full credit for both the additive computation and the ring-structure identification.

Exercise 4 (medium, short-answer).

Show that the Atiyah-Hirzebruch spectral sequence for K-theory of a CW complex of dimension at most $4$ collapses at $E_{2}$ over $Z$ .

Hint

The differential $d_{r}$ has bidegree $(r, 1 - r)$ . On a complex of dimension $4$ , only finitely many differentials can act before the source or target leaves the grid. Identify which differentials could carry a non-zero map and rule them out.

Answer

The $E_{2}$ page has entries only in columns $0 \leq p \leq 4$ . The differential $d_{r}$ sends $(p, q)$ to $(p + r, q + 1 - r)$ , which is non-zero in cohomology only if $0 \leq p + r \leq 4$ and $q + 1 - r$ is even. For K-theory the rows alternate between $Z$ at even $q$ and zero at odd $q$ . The total parity $(p + q)$ changes by $1$ across a differential, so an even-parity source maps to an odd-parity target — but odd-parity entries vanish. Every differential is zero. The page collapses, and the K-theory is the direct sum of cellular cohomology groups: $K^{n} (X) = ⨁_{p + q = n} H^{p} (X; K^{q} (pt))$ . Rubric: full credit for the parity argument plus the dimension cap.

Exercise 5 (medium, symbolic).

Identify the differential $d_{3}$ in the Atiyah-Hirzebruch spectral sequence for mod- $2$ K-theory $K^{*} (-; Z /2)$ , and compute it on the cellular cohomology $H^{p} (RP^{4}; Z /2)$ .

Hint

The differential $d_{3}$ in mod- $2$ K-theory equals $Sq^{3}$ , the third Steenrod square. The mod- $2$ cohomology of $RP^{4}$ is $Z /2 [x] / (x^{5})$ with $∣ x ∣ = 1$ , on which $Sq^{3}$ acts by Cartan's formula on monomials.

Answer

By Atiyah 1961 ^{[Atiyah 1961]}, the differential $d_{3}$ on the $E_{3}$ page of the AHSS for mod- $2$ K-theory equals the third Steenrod square $Sq^{3}$ . On $H^{*} (RP^{4}; Z /2) = Z /2 [x] / (x^{5})$ the Cartan formula gives $Sq^{3} (x^{p}) = (3 p) x^{p + 3}$ . Computing: $Sq^{3} (x) = 0$ (binomial $(3 1) = 0$ ); $Sq^{3} (x^{2}) = 0$ ; $Sq^{3} (x^{3}) = x^{6} = 0$ in $H^{*} (RP^{4}; Z /2)$ since $x^{5} = 0$ ; $Sq^{3} (x^{4}) = 0$ for the same reason. So $d_{3} = 0$ on every cellular cohomology class of $RP^{4}$ mod $2$ , and the spectral sequence collapses at $E_{4} = E_{3} = E_{2}$ for this space. Rubric: full credit for the identification $d_{3} = Sq^{3}$ plus the Cartan-formula evaluation.

Exercise 6 (medium, short-answer).

Prove that the AHSS for K-theory of any CW complex $X$ of finite type collapses at $E_{2}$ rationally, using the Chern character $ch : K^{*} (X) \otimes Q \to H^{even} (X; Q)$ .

Hint

The Chern character is a rational isomorphism. If every differential $d_{r}$ on the rationalised spectral sequence were non-zero, the dimension of the limit would drop below the rank of the $E_{2}$ page; the Chern character forbids this.

Answer

The Chern character is a ring homomorphism $ch : K^{*} (X) \otimes Q \to H^{even} (X; Q)$ and Atiyah-Hirzebruch ^{[Atiyah-Hirzebruch 1961]} proved it is an isomorphism on finite CW complexes. The total rank of $K^{*} (X) \otimes Q$ equals the total rank of $H^{even} (X; Q)$ . On the other side, the rationalised $E_{2}$ page of the AHSS is $E_{2}^{p, q} \otimes Q = H^{p} (X; Q) \otimes K^{q} (pt; Q)$ , whose total rank is the rank of $H^{even} (X; Q)$ times the contribution from each row (which by Bott periodicity is one rank per row). The $E_{\infty}$ page has the rank of $K^{*} (X) \otimes Q$ , which by the Chern character equals the rank of the $E_{2}$ page. So no rationalised differential can drop rank; every $d_{r} \otimes Q$ is zero; the page collapses rationally. Rubric: full credit for the rank-count argument backed by the Chern-character isomorphism.

Exercise 7 (hard, short-answer).

Exhibit a CW complex $X$ for which the AHSS for K-theory has a non-zero differential $d_{3}$ on integral K-theory, and identify the surviving page.

Hint

The simplest example is a $2$ -cell complex built by attaching a $4$ -cell to $S^{2}$ via a degree- $2$ map between $π_{3} (S^{2}) = Z$ and an appropriate target. The torsion in $π_{3} (S^{2})$ is the source of the Steenrod-cube obstruction.

Answer

Take $X = S^{2} \cup_{2 η} D^{4}$ , where $η : S^{3} \to S^{2}$ is the Hopf map and $2 η : S^{3} \to S^{2}$ is the doubled Hopf class (an element of order in $π_{3} (S^{2}) = Z$ ). The cohomology is $H^{0} = H^{2} = H^{4} = Z$ , with $Sq_{Z}^{2}$ on the generator of $H^{2}$ giving twice the generator of $H^{4}$ . (Concretely, the Steenrod cube $Sq_{Z}^{3}$ on $H^{0}$ vanishes here; the relevant differential is the integral $Sq^{3}$ on $H^{1}$ , which also vanishes by absence of a $1$ -cell. The non-zero $d_{3}$ comes from a different position.) The cleanest non-vanishing example is $X = Σ RP^{2} \land Σ RP^{2}$ where the integral $d_{3}$ realises the integer Steenrod cube on a $Z /2$ -class, producing a non-zero map between $E_{2}^{p, q}$ and $E_{2}^{p + 3, q - 2}$ and reducing the rank of $E_{4}$ below that of $E_{2}$ . The integer K-theory of $X$ is then the surviving $E_{\infty}$ page, with rank strictly less than the rank of integer cohomology. Rubric: full credit for any specific construction with a non-zero integral $d_{3} = Sq_{Z}^{3}$ .

Exercise 8 (hard, short-answer).

For a Lie group $G$ acting freely on a contractible space with quotient $B G$ , identify the AHSS for $K^{*} (B G)$ with the Atiyah-Segal completion theorem at the $E_{2}$ page level.

Hint

The Atiyah-Segal completion theorem identifies $K^{*} (B G)$ with the completion of the representation ring $R (G)$ at the augmentation ideal. The $E_{2}$ page of the AHSS is $H^{*} (B G; K^{*} (pt))$ , which for finite $G$ involves the cohomology of $B G$ with integer coefficients in every even degree.

Answer

For $G$ a compact Lie group, $B G$ is a CW complex of infinite dimension, and the AHSS reads $E_{2}^{p, q} = H^{p} (B G; K^{q} (pt)) \Rightarrow K^{p + q} (B G)$ . For $G$ finite, $H^{*} (B G; Z)$ is concentrated in even degrees outside dimension $0$ and is the group cohomology of $G$ . The Atiyah-Segal completion theorem (Atiyah-Segal 1969 Equivariant K-theory and completion) identifies $K^{0} (B G)$ with the completion of the representation ring $R (G)$ at the augmentation ideal $I$ . At the $E_{2}$ page, the additive structure $H^{*} (B G; Z)$ corresponds to the associated graded of the $I$ -adic filtration of $R (G)_{I}^{\land}$ : a class in $H^{2 k} (B G; Z)$ lifts (via the AHSS edge map plus a Chern-character correction) to an element of $I^{k} / I^{k + 1} \cdot R (G)_{I}^{\land}$ . The Chern character realises this correspondence rationally; integrally the Bockstein and the Steenrod-cube differentials match the higher-order obstructions to deforming a representation through the $I$ -adic filtration. Rubric: full credit for the $E_{2}$ -to-completion correspondence and one explicit example like $G = Z / p$ with $R (Z / p) = Z [ζ_{p}]$ matching $H^{*} (B Z / p; Z)$ .

Lean formalization [Intermediate+]

Mathlib has neither generalised cohomology theories nor topological K-theory packaged, so the Atiyah-Hirzebruch spectral sequence cannot yet be stated in Mathlib syntax. The intended formalisation would read schematically:

import Mathlib.AlgebraicTopology.CWComplex
import Mathlib.CategoryTheory.Filtered
import Mathlib.Algebra.Homology.SpectralSequence

variable {E : GeneralisedCohomologyTheory}

/-- The Atiyah-Hirzebruch spectral sequence of a CW complex in a
    generalised cohomology theory. -/
noncomputable def atiyahHirzebruchSS
    (X : CWComplex) (E : GeneralisedCohomologyTheory) :
    SpectralSequence ℤ × ℤ AddCommGrp :=
  skeletalFiltration X |>.toExactCouple E |>.spectralSequence

theorem atiyahHirzebruch_E2 (X : CWComplex) (E : GeneralisedCohomologyTheory)
    (p q : ℤ) :
    (atiyahHirzebruchSS X E).page 2 (p, q) ≅
      cohomology p X (E.coefficients q) :=
  sorry  -- identify the E_1 page with cellular cochains, take H of (E_1, d_1)

theorem atiyahHirzebruch_converges (X : CWComplex) [FiniteType X]
    (E : GeneralisedCohomologyTheory) (n : ℤ) :
    (atiyahHirzebruchSS X E).Converges (E.cohomology X n) :=
  sorry  -- bounded-below filtration of a finite-type complex

theorem atiyahHirzebruch_d3_eq_Sq3 (X : CWComplex) :
    (atiyahHirzebruchSS X K_mod2).differential 3 =
      steenrodSquare 3 :=
  sorry  -- Atiyah 1961 originator identification

The proof gap is substantive. Mathlib needs a structured generalised cohomology theory (with suspension isomorphism axioms), an exact-couple library that produces a spectral sequence from a filtration in an additive category, a packaged convergence theorem for bounded filtrations, and the identification of the $E_{2}$ page with cellular cohomology of the coefficient ring. The Steenrod-cube identification $d_{3} = Sq^{3}$ requires the Steenrod-square library (also not yet packaged) plus the explicit computation of Atiyah 1961.

Advanced results [Master]

Theorem (K-theory specialisation; Atiyah-Hirzebruch 1961). Let $X$ be a finite CW complex. The Atiyah-Hirzebruch spectral sequence for complex topological K-theory takes the form $$ E_2^{p, q} = H^p(X; K^q(\mathrm{pt})) = \begin{cases} H^p(X; \mathbb{Z}) & q \text{ even} \ 0 & q \text{ odd} \end{cases} $$ and converges to the graded associated to the skeletal filtration on $K^(X)$.*

The coefficient identification uses Bott periodicity 03.08.07: $K^{*} (pt) = Z [β, β^{- 1}]$ with $β \in K^{- 2} (pt)$ the Bott element. The page is two-periodic in $q$ , and the differentials are determined by their action on a single period.

Theorem (rational collapse via Chern character). Let $X$ be a finite CW complex. The Chern character $$ \mathrm{ch}: K^*(X) \otimes \mathbb{Q} \xrightarrow{\sim} H^{\mathrm{even}}(X; \mathbb{Q}) $$ is an isomorphism of $Z /2$ -graded rings. Consequently, the Atiyah-Hirzebruch spectral sequence for complex K-theory collapses at $E_{2}$ after tensoring with $Q$ .

The Chern character is constructed using the splitting principle 03.13.03 and the exponential power series: for a line bundle $L$ with $c_{1} (L) = x$ , set $ch (L) = e^{x} = 1 + x + x^{2} /2 + \dots$ . The construction extends multiplicatively to sums of line bundles by the splitting principle. The isomorphism statement reads in two parts: rationalising K-theory and rationalising even cohomology produces $Q$ -vector spaces of the same dimension, and the Chern character is a ring isomorphism between them.

Theorem (Steenrod-cube differential; Atiyah 1961). Let $X$ be a CW complex. The differential $d_{3}$ on the $E_{3}$ page of the Atiyah-Hirzebruch spectral sequence for mod- $2$ K-theory equals the third Steenrod square $$ d_3 = \mathrm{Sq}^3: H^p(X; \mathbb{Z}/2) \to H^{p + 3}(X; \mathbb{Z}/2) $$ on each row. Integrally, $d_{3}$ equals the integral Steenrod cube $Sq_{Z}^{3}$ , the lift of $Sq^{3}$ to integer cohomology that exists when $Sq^{2} \circ Sq^{1}$ vanishes.

Atiyah's 1961 paper Characters and cohomology of finite groups (Publ. IHÉS 9) gave the first explicit computation of $d_{3}$ on a specific class of spaces, identifying it as the Steenrod-cube cohomology operation. The identification proceeds by computing the differential explicitly on the K-theory of a cyclic group of prime order, where the page $E_{3}$ has a single non-zero column and the differential is forced to match a known Steenrod operation by naturality.

Theorem (higher differentials). On the $E_{5}$ page of the AHSS for K-theory, the differential $d_{5}$ is a higher cohomology operation determined by a secondary Massey product of Steenrod operations. The differentials $d_{2 k + 1}$ for $k \geq 1$ are detected by tertiary and higher operations; $d_{2 k}$ vanishes for parity reasons.

The full structure of the AHSS differentials is part of the calculation of cohomology operations on $K$ -theory. The first non-vanishing differential is $d_{3} = Sq^{3}$ ; the next is $d_{5}$ , which is a quaternary operation related to the Massey product $⟨ Sq^{2}, Sq^{2}, Sq^{2} ⟩$ ; higher differentials involve higher-order operations and become harder to identify explicitly.

Theorem (collapse criteria). The Atiyah-Hirzebruch spectral sequence for complex K-theory of a finite CW complex $X$ collapses at $E_{2}$ in the following cases:

(i) $X$ has cells only in even dimensions.

(ii) $X$ has dimension at most $4$ .

(iii) $X$ has all integer cohomology torsion-free in degrees where $Sq^{3}$ could act with non-zero image.

(iv) After tensoring with $Q$ on any finite CW complex.

(v) After tensoring with $C$ on any finite CW complex.

Case (i) is the parity argument of Exercise 4 applied to even-cell complexes; case (ii) extends the same argument by dimension bound. Case (iii) is the cohomology-operation criterion: $d_{3} = Sq_{Z}^{3}$ vanishes on torsion-free integer classes when the source has no $Sq^{1}$ -cohomology contribution. Case (iv) is the Chern-character argument. Case (v) holds because tensoring with $C$ also kills the differentials, by extension of scalars from the rational case.

Theorem (Atiyah-Hirzebruch for KO-theory). Let $X$ be a finite CW complex. The Atiyah-Hirzebruch spectral sequence for real topological K-theory $KO^ $ha s$ E_2^{p, q} = H^p(X; KO^q(\mathrm{pt}))$ with the eight-periodic coefficient ring* $$ KO^*(\mathrm{pt}) = \mathbb{Z}[\eta, \omega, \mu, \mu^{-1}] / (\eta^3, 2\eta, \omega \eta, \omega^2 - 4\mu), $$ and the differentials $d_{2} = Sq^{2} \cdot η$ at the lowest non-vanishing level, with $d_{3} = Sq^{3}$ at higher levels.

The real version of the AHSS is computationally richer: the coefficient ring has torsion ( $η$ of order $2$ ) starting in degree $- 1$ , so the $E_{2}$ page has $Z /2$ -coefficient rows in addition to integer rows. The differential $d_{2}$ becomes non-zero on the $η$ -rows, producing the lowest-degree obstruction to a real bundle lifting to complex K-theory. This is the basis of the analysis of Stiefel-Whitney and Pontryagin classes in 03.06.03 from the spectral-sequence viewpoint.

Theorem (originator computation: $K^{*} (CP^{n})$ ). The Atiyah-Hirzebruch spectral sequence for complex K-theory of $CP^{n}$ collapses at $E_{2}$ for parity reasons. The resulting K-theory is $$ K^0(\mathbb{CP}^n) \cong \mathbb{Z}[L] / ((L - 1)^{n + 1}), \qquad K^1(\mathbb{CP}^n) = 0, $$ where $L$ is the class of the tautological line bundle and $L - 1$ generates the augmentation ideal.

This is the originator computation of Atiyah-Hirzebruch 1961: the AHSS reduces a K-theory question to a cohomology computation plus the identification of the multiplicative structure via the Chern character.

Synthesis. The Atiyah-Hirzebruch spectral sequence is the foundational reason that ordinary cohomology determines generalised cohomology up to a finite sequence of corrections, each a stable cohomology operation. The central insight is that the skeletal filtration of a CW complex produces a clean exact couple in any generalised cohomology theory, and the resulting spectral sequence has $E_{2}$ page identified with ordinary cohomology with coefficients in the generalised theory of a point. Putting these together, the AHSS reads $K^{*} (X)$ , $Ω^{*} (X)$ , $Ell^{*} (X)$ , and many other generalised cohomology groups as $H^{*} (X; -)$ with the appropriate coefficient ring, corrected by differentials that are cohomology operations on the page. The bridge to the broader theory is the recognition that the Atiyah-Hirzebruch construction is the universal way to compute generalised cohomology from ordinary cohomology: every page above $E_{2}$ is determined by a stable cohomology operation on the previous page, and the operations encountered along the way are exactly the classical Steenrod operations and their refinements.

This same algebraic mechanism appears again in 03.13.02 (Leray-Serre) where the skeletal filtration is replaced by the base filtration of a fibration and the coefficient ring of $E^{*} (pt)$ is replaced by the cohomology of the fibre. Putting these together, the AHSS and the Leray-Serre spectral sequence are two specialisations of a single construction: the spectral sequence of a filtered cohomology theory, where the filtration is either by cell skeleta or by fibration base, and the $E_{2}$ page is determined by the filtration-quotient cohomology. This pattern recurs throughout stable homotopy theory; the bridge is the recognition that every spectral sequence converging to a generalised cohomology theory is determined by the filtration and the page-zero coefficients, and the differentials encode the homotopy-theoretic structure of how the filtration is glued. The foundational reason this works is the same exact-couple machinery of Massey 1952 that produces every classical spectral sequence — Leray's, Serre's, the Adams, the Eilenberg-Moore, and now Atiyah-Hirzebruch's — from a single algebraic input.

Full proof set [Master]

Proposition (parity collapse on even-dimensional complexes). Let $X$ be a CW complex with cells only in even dimensions. Then the Atiyah-Hirzebruch spectral sequence for complex K-theory collapses at $E_{2}$ .

Proof. The cellular cohomology $H^{p} (X; Z)$ vanishes for odd $p$ (no odd cells). The K-theory coefficient ring $K^{q} (pt) = Z$ for even $q$ and zero for odd $q$ . Hence $E_{2}^{p, q}$ is non-zero only when both $p$ and $q$ are even, i.e., when $p + q$ is even.

A differential $d_{r}$ on $E_{r}$ has bidegree $(r, 1 - r)$ , sending $(p, q)$ to $(p + r, q + 1 - r)$ . The total degree changes by $1$ : from $p + q$ even to $p + q + 1$ odd. The target sits in an odd-total-degree slot. Since both rows and columns are non-zero only in even total degree, the target is zero. Every differential vanishes, and $E_{\infty} = E_{2}$ . $□$

Proposition (rational collapse). Let $X$ be a finite CW complex. The Atiyah-Hirzebruch spectral sequence for complex K-theory of $X$ collapses at $E_{2}$ after tensoring with $Q$ .

Proof. The Chern character is the natural transformation $ch : K^{*} (-) \to H^{even} (-; Q)$ defined on line bundles by $ch (L) = e^{c_{1} (L)}$ and extended multiplicatively. The Chern character is a ring homomorphism, and Atiyah-Hirzebruch 1961 ^{[Atiyah-Hirzebruch 1961]} proved it is an isomorphism on finite CW complexes after tensoring K-theory with $Q$ .

Apply $- \otimes Q$ to the AHSS. The rationalised $E_{2}$ page is $$ E_2^{p, q} \otimes \mathbb{Q} = H^p(X; \mathbb{Q}) \otimes K^q(\mathrm{pt}; \mathbb{Q}) = \begin{cases} H^p(X; \mathbb{Q}) & q \text{ even} \ 0 & q \text{ odd}. \end{cases} $$ The total $Q$ -dimension of the rationalised $E_{2}$ page (counted with multiplicity in $q$ modulo the periodic identification) equals the total $Q$ -dimension of $H^{even} (X; Q)$ , which equals the total $Q$ -dimension of $K^{*} (X) \otimes Q$ by the Chern character isomorphism. The rationalised $E_{\infty}$ page equals the graded associated to the filtration on $K^{*} (X) \otimes Q$ , which has the same total $Q$ -dimension. Hence no rationalised differential drops dimension, so every $d_{r} \otimes Q$ is zero. The page collapses at $E_{2}$ . $□$

Proposition (the $d_{3}$ differential equals $Sq^{3}$ in mod- $2$ K-theory). Let $X$ be a CW complex. The differential $d_{3}$ on the $E_{3}$ page of the Atiyah-Hirzebruch spectral sequence for mod- $2$ K-theory $K^(-; \mathbb{Z}/2)$ equals the third Steenrod square* $$ d_3 = \mathrm{Sq}^3 : H^p(X; \mathbb{Z}/2) \to H^{p + 3}(X; \mathbb{Z}/2). $$

Proof sketch. The strategy is naturality plus a specific computation. The differentials of the AHSS are stable cohomology operations on the $E_{2}$ page: by naturality of the spectral sequence in the CW complex $X$ , the differential $d_{r} : H^{p} (X; Z /2) \to H^{p + r} (X; Z /2)$ commutes with all continuous maps and so is a stable operation of bidegree $(r, 1 - r)$ acting on $Z /2$ -cohomology.

The classification of stable mod- $2$ cohomology operations is the Steenrod algebra: every stable operation $H^{p} \to H^{p + k}$ is a polynomial in the Steenrod squares $Sq^{i}$ (Serre, Cartan, Adem). In bidegree $(3, - 2)$ — meaning operations $H^{p} \to H^{p + 3}$ that are part of a degree- $1$ change in total grading consistent with the spectral sequence convention — the candidates are $Sq^{3}$ , $Sq^{2} Sq^{1}$ , and $Sq^{1} Sq^{2}$ . The Adem relation $Sq^{2} Sq^{1} = Sq^{3} + Sq^{1} Sq^{2}$ reduces this to two independent operations.

The specific computation done by Atiyah in 1961 ^{[Atiyah 1961]} is on the classifying space $B Z /2 = RP^{\infty}$ . The mod- $2$ K-theory of $RP^{\infty}$ is computable explicitly via the Atiyah-Segal completion theorem (representation ring of $Z /2$ completed at the augmentation ideal). On this space, $d_{3}$ acts on $H^{*} (RP^{\infty}; Z /2) = Z /2 [x]$ with $∣ x ∣ = 1$ . Atiyah computed $d_{3} (x) = 0$ , $d_{3} (x^{2}) = 0$ , $d_{3} (x^{3}) = x^{6}$ , and so on by $d_{3} (x^{p}) = (3 p) x^{p + 3}$ — the formula for $Sq^{3}$ . The operation $d_{3} = Sq^{3}$ is uniquely identified by its action on $RP^{\infty}$ (because $RP^{\infty}$ is a universal model for mod- $2$ cohomology classes via the Eilenberg-MacLane representability theorem $H^{n} (X; Z /2) = [X, K (Z /2, n)]$ ).

The integral lift of this identification — namely $d_{3}$ on integer K-theory equals the integral Steenrod cube $Sq_{Z}^{3}$ on the integer cohomology of $X$ — follows by considering the integer-reduction map $K^{*} (-) \to K^{*} (-; Z /2)$ and the corresponding map $H^{*} (-; Z) \to H^{*} (-; Z /2)$ , plus the fact that $Sq^{3} \cdot 2 = 0$ in the Steenrod algebra. $□$

Proposition ( $K^{*} (CP^{n})$ via AHSS). The Atiyah-Hirzebruch spectral sequence for complex K-theory of $CP^{n}$ collapses at $E_{2}$ . The ring structure is $K^{0} (CP^{n}) = Z [L] / ((L - 1)^{n + 1})$ and $K^{1} (CP^{n}) = 0$ .

Proof. The cohomology of $CP^{n}$ is $H^{*} (CP^{n}; Z) = Z [t] / (t^{n + 1})$ with $∣ t ∣ = 2$ , generated by the Chern class $t = c_{1} (L^{*})$ of the dual tautological line bundle. All cohomology is in even degrees, so by the previous proposition the AHSS collapses at $E_{2}$ .

The additive structure: $K^{0} (CP^{n}) = ⨁_{k = 0}^{n} H^{2 k} (CP^{n}; Z) = Z^{n + 1}$ . The grading is by the skeletal filtration: the $k$ -th summand is the image of $K^{0} (CP^{n}) \to K^{0} (CP^{k}) / K^{0} (CP^{k - 1})$ .

The ring structure: the tautological line bundle $L$ on $CP^{n}$ generates $K^{0} (CP^{n})$ as a ring, with relation $(L - 1)^{n + 1} = 0$ . The relation comes from the Koszul resolution of $O_{CP^{n}}$ as a module over $L^{*} \cdot O_{CP^{n}}$ , or equivalently from the splitting principle applied to the product bundle $C^{n + 1} = L \oplus L^{⊥}$ . The Chern character verification: $ch (L) = e^{t} = 1 + t + t^{2} /2 + \dots$ , and the ring isomorphism $Z [L] / ((L - 1)^{n + 1}) \otimes Q \to Q [t] / (t^{n + 1})$ matches the rationalised K-theory with the rationalised cohomology, confirming the relation. The reduction modulo torsion-free structure forces the integral relation $(L - 1)^{n + 1} = 0$ . $□$

Proposition (collapse on dimension $\leq 4$ ). Let $X$ be a CW complex of dimension at most $4$ . Then the integer Atiyah-Hirzebruch spectral sequence for complex K-theory collapses at $E_{2}$ .

Proof. The $E_{2}$ page has non-zero entries only at $(p, q)$ with $0 \leq p \leq 4$ and $q$ even. The first candidate non-zero differential is $d_{3}$ , sending $(p, q)$ to $(p + 3, q - 2)$ . For the target to be in range, $p + 3 \leq 4$ , so $p \leq 1$ . With $q$ even and $q - 2$ also even, $d_{3}$ acts on rows. The only position with $p \leq 1$ and a non-zero target is $p = 0$ or $p = 1$ , with target $p = 3$ or $p = 4$ . For $p = 0$ : $d_{3} : H^{0} (X; Z) \to H^{3} (X; Z)$ . By the integral Steenrod cube identification (Atiyah 1961, integral lift), $d_{3} = Sq_{Z}^{3}$ acts on $H^{0} (X; Z)$ , which has only the constant cocycle; $Sq^{3}$ on a constant is zero. So $d_{3} = 0$ from $p = 0$ .

For $p = 1$ : $d_{3} : H^{1} (X; Z) \to H^{4} (X; Z)$ . The integral Steenrod cube on a degree- $1$ class is zero modulo a higher-order obstruction; in fact $Sq_{Z}^{3}$ on $H^{1}$ vanishes for parity reasons (since $Sq^{3} = Sq^{1} Sq^{2}$ and $Sq^{2}$ on $H^{1}$ would land in $H^{3}$ , but the integral lift of $Sq^{2}$ on degree- $1$ classes is zero). So $d_{3} = 0$ on dimension $\leq 4$ .

Subsequent differentials $d_{5}, d_{7}, \dots$ have target columns $\geq 5$ , which are out of range on a dimension- $\leq 4$ complex. So all higher differentials vanish. The page collapses at $E_{2}$ . $□$

Proposition (Chern character compatibility). The Chern character $\mathrm{ch}: K^(X) \otimes \mathbb{Q} \to H^{\mathrm{even}}(X; \mathbb{Q})$ is the edge map of the Atiyah-Hirzebruch spectral sequence rationally.*

Proof. The edge map of the AHSS on the $E_{\infty}^{0, q}$ corner is the projection $K^{q} (X) \to E_{\infty}^{0, q} = H^{0} (X; K^{q} (pt))$ when this position survives. For complex K-theory, $E_{\infty}^{0, q} \otimes Q = H^{0} (X; Q)$ for $q$ even, embedded in $H^{even} (X; Q)$ . The construction of the AHSS shows that this edge map is compatible with the Chern character on the rationalised K-theory, since both arise from the natural transformation $K^{*} (X) \to K^{*} (X^{(0)}) = H^{*} (X^{(0)}; K^{*} (pt))$ on the $0$ -skeleton, factored through the filtration. Putting the levels together over all $q$ and using rational collapse, the Chern character realises the abutment of the spectral sequence as a $Z /2$ -graded ring isomorphism. $□$

Connections [Master]

Spectral sequences 03.13.01. The Atiyah-Hirzebruch construction is a direct application of the general spectral-sequence machinery — exact couples (Massey 1952) applied to the skeletal filtration of a CW complex in any generalised cohomology theory. The construction inherits all the structure: bidegree- $(r, 1 - r)$ differentials, convergence to the graded associated to a filtration, multiplicative structure when the underlying theory has a product. Without the spectral-sequence framework of 03.13.01, the AHSS could not be stated cleanly; with it, the AHSS is the canonical example of the framework applied to a non-additive cohomology theory.
K-theory definition 03.08.01. The AHSS is the primary computational tool for K-theory of a CW complex. Topological K-theory $K^{*} (X)$ is defined as the Grothendieck group of complex vector bundles on $X$ , which is computable directly only for simple spaces. The AHSS converts the K-theory computation into a cohomology computation plus a finite sequence of cohomology-operation differentials. Every concrete K-theory calculation in the literature — Atiyah's K-theory of homogeneous spaces, the Atiyah-Segal completion theorem, the Atiyah-Singer index theorem — uses the AHSS at some point.
Adams operations 03.08.02. Adams operations $ψ^{k}$ act on K-theory and on the Atiyah-Hirzebruch spectral sequence. Under the Chern character, $ψ^{k}$ corresponds to multiplication by $k^{p}$ on $H^{2 p} (X; Q)$ . The AHSS therefore provides the framework for computing $ψ^{k}$ on integer K-theory: identify the cohomology that maps to a given K-class via the spectral sequence, apply the rational Adams formula, and lift back through the spectral-sequence differentials. The Hopf-invariant-one application of Adams operations uses exactly this circle of ideas.
Thom isomorphism in K-theory 03.08.03. The K-theoretic Thom isomorphism is computable via the AHSS applied to the Thom space of a bundle. On the $E_{2}$ page, the cohomological Thom isomorphism appears; the spectral sequence converges to the K-theoretic Thom isomorphism. The compatibility between these two Thom classes is exactly the statement that the AHSS edge map sends the cohomological Thom class to the K-theoretic Thom class modulo higher-order corrections — the Todd-genus correction.
Bott periodicity 03.08.07. Bott periodicity is the input to the AHSS at the coefficient-ring level: $K^{*} (pt) = Z [β, β^{- 1}]$ with the Bott element $β \in K^{- 2} (pt)$ . The $E_{2}$ page of the AHSS for K-theory inherits two-periodicity from Bott periodicity. The vertical structure of the page — repeating $Z$ rows in even degrees, zero rows in odd degrees — is exactly the Bott structure, and the spectral sequence's convergence to two-periodic K-theory recovers Bott as a consistency check.
Leray-Serre spectral sequence 03.13.02. The AHSS and the Leray-Serre spectral sequence are two specialisations of a common framework: the spectral sequence of a filtered cohomology theory. The AHSS filters by cell skeleta in a generalised cohomology theory; Leray-Serre filters by base skeleta in a fibration with ordinary cohomology. Both have $E_{2} = H^{p} (filtration base; F^{q})$ for an appropriate coefficient $F^{q}$ (filtration quotient cohomology), and both converge to a graded associated to the total cohomology. The AHSS is the universal-coefficient-style specialisation; Leray-Serre is the fibration specialisation.
Cellular homology 03.12.13. The $E_{1}$ page of the AHSS is cellular cochains in the generalised theory: $E_{1}^{p, q} = C_{cell}^{p} (X; E^{q} (pt))$ . The $d_{1}$ differential is the cellular coboundary tensored with the coefficient ring. Cellular homology is therefore the foundational input to the AHSS, and the agreement of cellular and singular cohomology (proved in 03.12.13) is what allows the $E_{2}$ page to be identified with singular cohomology in any chosen coefficient ring.
Spectrum 03.12.04. A generalised cohomology theory is represented by a spectrum. The AHSS extends from K-theory to any spectrum-represented cohomology theory: complex cobordism $M U^{*}$ , real cobordism $M O^{*}$ , elliptic cohomology $Ell^{*}$ , Morava K-theories $K (n)^{*}$ , and the sphere spectrum itself. The construction does not change — only the coefficient ring $E^{*} (pt)$ does. The AHSS is therefore the universal computational framework for spectrum-represented cohomology.
Equivariant K-theory and the representation ring 03.08.10. The AHSS computation of $K^{*} (B G)$ for a compact Lie group $G$ is the concrete instance underlying the Atiyah-Segal completion theorem $K^{*} (B G) ≅ R (G)_{I}$ . The $E_{2}$ page $H^{*} (B G; Z)$ encodes characteristic-class data via the Borel construction; convergence to the $I$ -adic completion of the representation ring records how irreducible characters of $G$ lift through the filtration. Atiyah's 1961 computation of $K^{*} (B Z / p)$ via the AHSS — identifying the answer as $Z_{p} [[y]]$ with $y = t - 1$ and the $d_{3} = Sq^{3}$ differential as the obstruction — was the first concrete instance of the completion theorem and provided the prototype calculation. More generally, the equivariant AHSS converging to $K_{G}^{*} (X)$ has $E_{2}$ page $H_{G}^{*} (X; R (G))$ (equivariant cohomology with $R (G)$ -coefficients), with differentials given by equivariant cohomology operations; passing to the Borel construction $X \times_{G} E G$ then recovers the ordinary AHSS plus the completion theorem. Connection type: organising spectral sequence — the AHSS is the page-by-page bookkeeping device through which the abstract representation-ring structure of $R (G)$ refines into ordinary K-theory of $B G$ .

Historical & philosophical context [Master]

Atiyah and Hirzebruch introduced their spectral sequence in 1961 at the Tucson symposium on differential geometry, in the paper Vector bundles and homogeneous spaces (Proc. Sympos. Pure Math. 3) ^{[Atiyah-Hirzebruch 1961]}. The setting was specific: compute the topological K-theory of compact homogeneous spaces $G / H$ from their ordinary cohomology, with a view to applications in the topology of Lie groups and the structure of vector bundles on flag varieties. The spectral sequence was constructed as the page-by-page approximation of the K-theory of the CW complex $G / H$ by its ordinary cohomology with K-theory-of-a-point coefficients, with the differentials encoding the obstructions to a cellular K-class extending across higher cells.

The key technical insight was that the construction is entirely formal: the only inputs are the skeletal filtration of a CW complex and the long exact sequence axiom for a generalised cohomology theory. Atiyah and Hirzebruch realised that the same construction extends to any cohomology theory satisfying the Eilenberg-Steenrod axioms minus the dimension axiom, opening the door to a unified treatment of computations in K-theory, cobordism, and what later came to be called stable homotopy theory.

Atiyah's 1961 follow-up paper Characters and cohomology of finite groups (Publ. IHÉS 9) ^{[Atiyah 1961]} computed the K-theory of classifying spaces $B G$ for finite groups, and in doing so identified the lowest non-vanishing differential of the AHSS as the third Steenrod square $Sq^{3}$ . This identification — that the differentials of the AHSS are cohomology operations on the page — was the first concrete instance of a more general principle, later codified by Adams in his lectures on stable homotopy theory (Adams 1969-74) ^{[Adams 1974]}: the differentials of any spectral sequence converging to a generalised cohomology theory of a space are stable cohomology operations, classified by the stable homotopy category and explicitly computable via the Steenrod algebra.

The exact-couple machinery underlying the AHSS was developed by Massey in his 1952 paper Exact couples in algebraic topology (Ann. of Math. 56) ^{[Massey 1952]}, building on the spectral-sequence framework of Cartan-Eilenberg Homological Algebra (Princeton 1956) ^{[Cartan-Eilenberg 1956]}. Massey's exact couples gave the algebraic backbone that Atiyah-Hirzebruch then specialised to the skeletal filtration; the same machinery underlies Leray's spectral sequence (1946), the Serre spectral sequence (1951), and the Adams spectral sequence (1958). The AHSS was the application of this machinery that integrated K-theory into the broader framework of generalised cohomology theories, completing the picture started by Eilenberg-Steenrod and continued by Whitehead, Adams, and Boardman.

Bibliography [Master]

@inproceedings{AtiyahHirzebruch1961,
  author    = {Atiyah, Michael F. and Hirzebruch, Friedrich},
  title     = {Vector bundles and homogeneous spaces},
  booktitle = {Differential Geometry (Proc. Sympos. Pure Math., Vol. III, Tucson, 1960)},
  pages     = {7--38},
  publisher = {American Mathematical Society},
  year      = {1961}
}

@article{Atiyah1961,
  author  = {Atiyah, Michael F.},
  title   = {Characters and cohomology of finite groups},
  journal = {Publ. Math. Inst. Hautes \'Etudes Sci.},
  volume  = {9},
  year    = {1961},
  pages   = {23--64}
}

@book{Adams1974,
  author    = {Adams, J. Frank},
  title     = {Stable Homotopy and Generalised Homology},
  series    = {Chicago Lectures in Mathematics},
  publisher = {University of Chicago Press},
  year      = {1974}
}

@article{Massey1952,
  author  = {Massey, William S.},
  title   = {Exact couples in algebraic topology, {I, II}},
  journal = {Ann. of Math. (2)},
  volume  = {56},
  year    = {1952},
  pages   = {363--396}
}

@book{CartanEilenberg1956,
  author    = {Cartan, Henri and Eilenberg, Samuel},
  title     = {Homological Algebra},
  publisher = {Princeton University Press},
  year      = {1956}
}

@book{McCleary2001,
  author    = {McCleary, John},
  title     = {A User's Guide to Spectral Sequences},
  edition   = {2nd},
  series    = {Cambridge Studies in Advanced Mathematics},
  volume    = {58},
  publisher = {Cambridge University Press},
  year      = {2001}
}

@book{Switzer1975,
  author    = {Switzer, Robert M.},
  title     = {Algebraic Topology: Homotopy and Homology},
  series    = {Grundlehren der mathematischen Wissenschaften},
  volume    = {212},
  publisher = {Springer-Verlag},
  year      = {1975}
}

@book{LawsonMichelsohn1989,
  author    = {Lawson, H. Blaine and Michelsohn, Marie-Louise},
  title     = {Spin Geometry},
  publisher = {Princeton University Press},
  year      = {1989}
}

@book{Karoubi1978,
  author    = {Karoubi, Max},
  title     = {K-Theory: An Introduction},
  series    = {Grundlehren der mathematischen Wissenschaften},
  volume    = {226},
  publisher = {Springer-Verlag},
  year      = {1978}
}

@article{AtiyahSegal1969,
  author  = {Atiyah, Michael F. and Segal, Graeme B.},
  title   = {Equivariant {$K$}-theory and completion},
  journal = {J. Differential Geom.},
  volume  = {3},
  year    = {1969},
  pages   = {1--18}
}

Prerequisites

03.13.01
03.12.10
03.12.13
03.08.01
03.08.07
03.12.04

Tier anchors

beginner: the cellular skeleton viewed page by page in any cohomology theory — informal opening of McCleary §10 or the Atiyah-Hirzebruch 1961 BAMS announcement
intermediate: Atiyah-Hirzebruch 1961 *Vector bundles and homogeneous spaces* (Proc. Sympos. Pure Math. 3); McCleary §10; Hatcher *Spectral Sequences* §1.5; Adams *Stable Homotopy and Generalised Homology* Part III
master: Atiyah-Hirzebruch 1961 *Vector bundles and homogeneous spaces*; Atiyah 1961 *Characters and cohomology of finite groups* (Publ. IHÉS 9); Adams 1974 *Stable Homotopy and Generalised Homology* Pt. III §7-§9; Switzer *Algebraic Topology — Homotopy and Homology* Ch. 15; Lawson-Michelsohn *Spin Geometry* §IV.1

References

TODO_REF
Atiyah-Hirzebruch — Vector bundles and homogeneous spaces · Proc. Sympos. Pure Math. 3 (Differential Geometry), Amer. Math. Soc. (1961), 7-38 — originator paper for the AHSS in K-theory
TODO_REF
Atiyah — Characters and cohomology of finite groups · Publ. Math. IHÉS 9 (1961), 23-64 — first explicit identification of $d_3 = \mathrm{Sq}^3$ as the lowest non-vanishing differential in K-theory
TODO_REF
Adams — Stable Homotopy and Generalised Homology · Chicago Lectures in Mathematics 1974; Part III §7-§9, the AHSS for an arbitrary generalised cohomology theory on a CW complex via the skeletal filtration and exact couple
TODO_REF
Cartan-Eilenberg — Homological Algebra · Princeton 1956, §XV exact-couple and filtered-complex machinery underlying the spectral sequence
TODO_REF
Massey — Exact couples in algebraic topology · Ann. of Math. 56 (1952), 363-396 — foundational exact-couple framework specialised by the AHSS
TODO_REF
McCleary — A User's Guide to Spectral Sequences · Cambridge Studies in Advanced Mathematics 58 (2nd ed., 2001); §10 the AHSS for generalised cohomology theories with worked computation for $K^*(\mathbb{C}P^n)$
TODO_REF
Switzer — Algebraic Topology: Homotopy and Homology · Springer Grundlehren 212 (1975); Ch. 15 generalised cohomology theories and the AHSS
TODO_REF
Hatcher — Spectral Sequences (online supplement) · §1.5 Atiyah-Hirzebruch spectral sequence, Steenrod-square differential, collapse arguments
TODO_REF
Lawson-Michelsohn — Spin Geometry · Princeton 1989; §IV.1 the AHSS in K-theory and KO-theory, applications to the index theorem
TODO_REF
Karoubi — K-Theory: An Introduction · Springer Grundlehren 226 (1978); §IV.6 the Chern character isomorphism over $\mathbb{Q}$, equivalent to AHSS collapse at $E_2$ on the rational page

Reviewer

TBD

Estimated time

beginner: 20m
intermediate: 55m
master: 90m