Spectral Sequences Introduction

Core Idea

A spectral sequence is a doubly-indexed family of abelian groups (or objects in an abelian category) with differentials that exhibit increasingly refined information, converging to a target homology group. Spectral sequences systematize the computation of derived functors and homological invariants through a sequence of approximations, and are among the most powerful computational tools in algebraic topology and homological algebra.

How It's Best Learned

Study the spectral sequence associated to a filtered complex and the Leray spectral sequence for sheaf cohomology. Understand the E₁, E₂, E∞ pages and how differentials at each stage carry information. Practice computing spectral sequences in concrete examples and understanding convergence criteria.

Common Misconceptions

Spectral sequences do not directly yield the answer; they require careful analysis of differentials and extension problems. Not every spectral sequence converges, and convergence properties can be subtle. The notion of 'convergence' itself requires precise formulation.

Explainer

From chain complexes and long exact sequences, you know that a short exact sequence of chain complexes 0 → A• → B• → C• → 0 produces a long exact sequence in homology, connecting H_*(A), H_*(B), and H_*(C) through connecting homomorphisms. This is powerful when you have an exact sequence. A spectral sequence generalizes this: when your chain complex has a filtration — a nested chain of subcomplexes F₀ ⊂ F₁ ⊂ ··· ⊂ Fₙ = C• — the spectral sequence systematically extracts the homology of C• from the homology of the layers Fₚ/Fₚ₋₁, one approximation at a time.

The construction begins with the E¹ page: E¹_{p,q} = H_{p+q}(Fₚ/Fₚ₋₁), the homology of the graded pieces (quotient complexes). These are the "first approximation" — they see only what happens within each layer of the filtration. At the E¹ page, there are differentials d₁: E¹_{p,q} → E¹_{p−1,q} that encode boundary maps between adjacent layers. Taking homology with respect to d₁ produces the E² page, which sees interactions across adjacent layers. At each page Eʳ, differentials dᵣ go from Eʳ_{p,q} to Eʳ_{p−r,q+r−1} (step r to the left and r−1 up), and the next page is their homology. As r increases, the differentials reach further across the bigraded diagram.

The sequence converges when all differentials above some page r₀ are zero — the pages stabilize. The E∞ page gives the associated graded of the filtered homology: E∞_{p,q} ≅ Fₚ H_{p+q}(C•)/Fₚ₋₁ H_{p+q}(C•). In favorable cases (e.g., when the filtration is finite and bounded), this determines H_*(C•) up to extension problems. The mental model is a sequence of approximations converging on the true answer: each page strips away noise and reveals finer structure, until only the essential homology information remains.

A landmark application is the Leray-Serre spectral sequence for a fibration F → E → B: the E² page is E²_{p,q} = H_p(B; H_q(F)) (cohomology of the base with coefficients in the fiber's cohomology), and the spectral sequence converges to H_*(E). This computes the cohomology of E — often a complicated space — from the cohomology of the simpler spaces B and F. For a product E = B × F, all differentials above E² vanish (the sequence collapses), and H_*(B × F) = H_*(B) ⊗ H_*(F) (Künneth formula). For a non-trivial fibration, non-zero higher differentials encode the twisting, and reading them correctly is the analytical work. The hard technical skill with spectral sequences is not building them but interpreting them: surviving the differentials, tracking which elements are killed at which page, and then solving the extension problems that arise when assembling E∞ back into H_*(C•) — since knowing the associated graded does not uniquely determine the group when extensions are non-trivial.