TheoremComplete

Waldhausen's Additivity Theorem

The additivity theorem is the fundamental structural result for Waldhausen K-theory. It states that the K-theory of the category of cofibration sequences splits as the product of the K-theories of the "fiber" and "cofiber." This result underpins essentially all computations in Waldhausen K-theory.


Statement

Theorem7.3Waldhausen's Additivity Theorem

Let C\mathcal{C} be a Waldhausen category and E(C)=S2(C)E(\mathcal{C}) = S_2(\mathcal{C}) the category of cofibration sequences in C\mathcal{C} (objects are A↣Bβ† CA \rightarrowtail B \twoheadrightarrow C with C=B/AC = B/A). The "source and quotient" functor:

(s,q):E(C)β†’CΓ—C,(A↣Bβ† C)↦(A,C)(s, q): E(\mathcal{C}) \to \mathcal{C} \times \mathcal{C}, \quad (A \rightarrowtail B \twoheadrightarrow C) \mapsto (A, C)

induces a homotopy equivalence on K-theory spaces:

K(E(C))β†’β€…β€Šβ‰ƒβ€…β€ŠK(C)Γ—K(C).K(E(\mathcal{C})) \xrightarrow{\;\simeq\;} K(\mathcal{C}) \times K(\mathcal{C}).

Equivalently, K(s)+K(q)=K(id⁑):K(E(C))β†’K(C)K(s) + K(q) = K(\operatorname{id}): K(E(\mathcal{C})) \to K(\mathcal{C}), meaning:

[B]=[A]+[C]Β inΒ K0(C)forΒ everyΒ A↣Bβ† C.[B] = [A] + [C] \text{ in } K_0(\mathcal{C}) \quad \text{for every } A \rightarrowtail B \twoheadrightarrow C.


Proof

Proof

The proof uses the S-construction and a "swallowing" argument.

Step 1: Setup. We work with the bisimplicial set wSpSqCwS_p S_q \mathcal{C} and its realization. The additivity theorem is equivalent to showing that the two functors

s,q:S2C⇉Cs, q: S_2 \mathcal{C} \rightrightarrows \mathcal{C}

(source and quotient of the cofibration sequence) satisfy ∣wSβˆ™s∣+∣wSβˆ™qβˆ£β‰ƒβˆ£wSβˆ™total⁑∣|wS_\bullet s| + |wS_\bullet q| \simeq |wS_\bullet \operatorname{total}| at the level of K-theory spaces.

Step 2: The "generic cofibration" argument. Consider the functor F:Cβ†’E(C)F: \mathcal{C} \to E(\mathcal{C}) defined by A↦(A↣Aβ† βˆ—)A \mapsto (A \rightarrowtail A \twoheadrightarrow *). Then (s,q)∘F=(id⁑,0)(s, q) \circ F = (\operatorname{id}, 0). Similarly, G:Cβ†’E(C)G: \mathcal{C} \to E(\mathcal{C}) via C↦(βˆ—β†£Cβ† C)C \mapsto (* \rightarrowtail C \twoheadrightarrow C) gives (s,q)∘G=(0,id⁑)(s, q) \circ G = (0, \operatorname{id}).

The functor total⁑:E(C)β†’C\operatorname{total}: E(\mathcal{C}) \to \mathcal{C}, (A↣Bβ† C)↦B(A \rightarrowtail B \twoheadrightarrow C) \mapsto B, satisfies total⁑∘F=id⁑\operatorname{total} \circ F = \operatorname{id} and total⁑∘G=id⁑\operatorname{total} \circ G = \operatorname{id}.

Step 3: Key lemma (Waldhausen's "swallowing" lemma). Define two functors F′,G′:C→E(C)F', G': \mathcal{C} \to E(\mathcal{C}) that are naturally weakly equivalent to FF and GG but have the property that F′F' and G′G' together "generate" E(C)E(\mathcal{C}).

Consider the filtration on SnE(C)S_n E(\mathcal{C}). An object is an nn-step filtration of cofibration sequences:

(βˆ—)=(A0β†’B0β†’C0)↣(A1β†’B1β†’C1)↣⋯↣(Anβ†’Bnβ†’Cn)(*) = (A_0 \to B_0 \to C_0) \rightarrowtail (A_1 \to B_1 \to C_1) \rightarrowtail \cdots \rightarrowtail (A_n \to B_n \to C_n)

This can be decomposed into the filtration on the AA-components and the CC-components independently, using the cofibration axioms and the fact that cofibrations in E(C)E(\mathcal{C}) are cofibrations on each component.

Step 4: Realization argument. The bisimplicial identity wSβˆ™E(C)≃wSβˆ™CΓ—wSβˆ™CwS_\bullet E(\mathcal{C}) \simeq wS_\bullet \mathcal{C} \times wS_\bullet \mathcal{C} follows from the decomposition in Step 3 and the "realization lemma": a map of bisimplicial sets that is a weak equivalence on each row (or column) is a weak equivalence on realizations.

More precisely, the functor (s,q)(s, q) on SnS_n-level gives:

wSnE(C)β†’wSnCΓ—wSnCwS_n E(\mathcal{C}) \to wS_n \mathcal{C} \times wS_n \mathcal{C}

which is a homotopy equivalence for each nn (by induction on nn using the cofibration structure). Taking geometric realization over nn preserves the equivalence.

Step 5: Passage to loop spaces. Since K(C)=Ω∣wSβˆ™C∣K(\mathcal{C}) = \Omega|wS_\bullet \mathcal{C}| and K(E(C))=Ω∣wSβˆ™E(C)∣K(E(\mathcal{C})) = \Omega|wS_\bullet E(\mathcal{C})|, the equivalence on ∣wSβˆ™βˆ£|wS_\bullet| gives K(E(C))≃K(C)Γ—K(C)K(E(\mathcal{C})) \simeq K(\mathcal{C}) \times K(\mathcal{C}). β–‘\square

β– 

Consequences

ExampleEuler characteristic

The additivity theorem implies that K-theory is an "additive invariant": for any exact sequence A↣Bβ† CA \rightarrowtail B \twoheadrightarrow C, we have [B]=[A]+[C][B] = [A] + [C] in K0K_0. This is the algebraic version of the Euler characteristic formula:

Ο‡(B)=Ο‡(A)+Ο‡(C)\chi(B) = \chi(A) + \chi(C)

for short exact sequences of complexes. More generally, for a bounded complex Cβˆ™C^\bullet with homology Hi(Cβˆ™)H^i(C^\bullet):

[Cβˆ™]=βˆ‘i(βˆ’1)i[Ci]=βˆ‘i(βˆ’1)i[Hi(Cβˆ™)]∈K0[C^\bullet] = \sum_i (-1)^i [C^i] = \sum_i (-1)^i [H^i(C^\bullet)] \in K_0

(assuming all terms are in the Waldhausen category).

RemarkFibration theorem

The additivity theorem implies Waldhausen's fibration theorem: if A→B→C\mathcal{A} \to \mathcal{B} \to \mathcal{C} is a "short exact sequence" of Waldhausen categories (meaning the functor from A\mathcal{A} to the "fiber" of B→C\mathcal{B} \to \mathcal{C} is a K-theory equivalence), then there is a homotopy fiber sequence:

K(A)β†’K(B)β†’K(C).K(\mathcal{A}) \to K(\mathcal{B}) \to K(\mathcal{C}).

This gives long exact sequences in K-groups:

β‹―β†’Kn+1(C)β†’βˆ‚Kn(A)β†’Kn(B)β†’Kn(C)β†’β‹―\cdots \to K_{n+1}(\mathcal{C}) \xrightarrow{\partial} K_n(\mathcal{A}) \to K_n(\mathcal{B}) \to K_n(\mathcal{C}) \to \cdots

The Thomason--Trobaugh localization sequence is a special case, applied to Perf⁑Z(X)β†’Perf⁑(X)β†’Perf⁑(U)\operatorname{Perf}_Z(X) \to \operatorname{Perf}(X) \to \operatorname{Perf}(U).

ExampleSplitting and cofinality

The additivity theorem implies:

  1. Eilenberg swindle: If C\mathcal{C} has countable coproducts, then K(C)β‰ƒβˆ—K(\mathcal{C}) \simeq * (contractible), because [A]=[AβŠ•B]βˆ’[B][A] = [A \oplus B] - [B] and [AβŠ•B]=[A]+[B][A \oplus B] = [A] + [B], giving [A]=0[A] = 0 for all AA.

  2. Cofinality theorem: If AβŠ†B\mathcal{A} \subseteq \mathcal{B} is cofinal (every B∈BB \in \mathcal{B} is a summand of some A∈AA \in \mathcal{A}), then Kn(A)β‰…Kn(B)K_n(\mathcal{A}) \cong K_n(\mathcal{B}) for nβ‰₯1n \geq 1 and K0(A)β†ͺK0(B)K_0(\mathcal{A}) \hookrightarrow K_0(\mathcal{B}) with cokernel determined by the "missing" summands.

  3. Agreement principle: For two Waldhausen category structures on the same category that have the same weak equivalences, the K-theories agree. This is because additivity forces the K-theory to depend only on the "homological" structure.