TheoremComplete

Waldhausen's Approximation Theorem

The approximation theorem provides conditions under which a functor between Waldhausen categories induces an equivalence on K-theory. It is the Waldhausen analogue of Quillen's Theorem A and is used ubiquitously to compare different models of K-theory.


Statement

Theorem7.4Waldhausen's Approximation Theorem

Let F:ABF: \mathcal{A} \to \mathcal{B} be an exact functor between Waldhausen categories (preserving zero objects, cofibrations, and weak equivalences). Suppose:

  1. App 1 (Approximation): For every morphism FABFA \to B in B\mathcal{B}, there exists a cofibration AAA \rightarrowtail A' in A\mathcal{A} and a weak equivalence FABFA' \xrightarrow{\sim} B making the obvious triangle commute up to homotopy.

  2. App 2 (Weak equivalence detection): A morphism f:AAf: A \to A' in A\mathcal{A} is a weak equivalence if and only if Ff:FAFAFf: FA \to FA' is a weak equivalence in B\mathcal{B}.

Then FF induces a homotopy equivalence:

K(F):K(A)    K(B).K(F): K(\mathcal{A}) \xrightarrow{\;\simeq\;} K(\mathcal{B}).


Applications

ExampleChain complexes vs derived category

The inclusion Chb(P(R))Chb(mod-R)perf\operatorname{Ch}^b(\mathcal{P}(R)) \hookrightarrow \operatorname{Ch}^b(\operatorname{mod}\text{-}R)^{\text{perf}} (bounded chain complexes of projectives into perfect complexes of f.g. modules) satisfies the approximation conditions:

  • App 1: Every perfect complex is quasi-isomorphic to a bounded complex of projectives (take a projective resolution).
  • App 2: A chain map between complexes of projectives is a quasi-isomorphism iff it is a quasi-isomorphism of the underlying complexes.

Conclusion: K(Chb(P(R)))K(Perf(R))K(\operatorname{Ch}^b(\mathcal{P}(R))) \simeq K(\operatorname{Perf}(R)). This shows that Quillen's K-theory (via projective modules) agrees with Thomason--Trobaugh's (via perfect complexes) for rings.

ExampleFinitely dominated vs finite complexes

Let A\mathcal{A} = finite CW complexes and B\mathcal{B} = finitely dominated CW complexes (spaces homotopy equivalent to a retract of a finite complex). Both are Waldhausen categories with homotopy equivalences as weak equivalences.

The inclusion AB\mathcal{A} \hookrightarrow \mathcal{B} does not satisfy App 1 (not every finitely dominated space has the homotopy type of a finite complex -- the Wall finiteness obstruction in K~0(Z[π1])\widetilde{K}_0(\mathbb{Z}[\pi_1]) is the obstruction).

However, for the S-construction with a different model: S(homotopy finite)S(finitely dominated)S_\bullet(\text{homotopy finite}) \hookrightarrow S_\bullet(\text{finitely dominated}) can be analyzed using the cofinality theorem, giving:

K(B)/K(A)K~0(Z[π1])×(higher terms).K(\mathcal{B}) / K(\mathcal{A}) \simeq \widetilde{K}_0(\mathbb{Z}[\pi_1]) \times (\text{higher terms}).


Proof sketch

Proof

The proof proceeds by analyzing the effect of FF on the SS_\bullet-construction level by level.

Step 1: Level-wise analysis. We must show wSnF:wSnAwSnBwS_n F: wS_n \mathcal{A} \to wS_n \mathcal{B} is a homotopy equivalence for each nn, where wSnCwS_n \mathcal{C} is the nerve of the category of weak equivalences in SnCS_n \mathcal{C}.

Step 2: Apply Quillen's Theorem A. For each nn, show that for any object B=(B0Bn)SnB\vec{B} = (B_0 \rightarrowtail \cdots \rightarrowtail B_n) \in S_n \mathcal{B}, the over-category (BwSnF)(\vec{B} \downarrow wS_n F) has contractible nerve.

The over-category consists of pairs (A,α)(\vec{A}, \alpha) where ASnA\vec{A} \in S_n \mathcal{A} and α:BF(A)\alpha: \vec{B} \xrightarrow{\sim} F(\vec{A}) is a level-wise weak equivalence.

Step 3: Contractibility via App 1. Given any two objects (A,α)(\vec{A}, \alpha) and (A,α)(\vec{A}', \alpha') in the over-category, use App 1 to "approximate" them: construct a cofibration AA\vec{A} \rightarrowtail \vec{A}'' in SnAS_n \mathcal{A} with F(A)BF(\vec{A}'') \xrightarrow{\sim} \vec{B} compatible with both α\alpha and α\alpha'.

The key is that App 1 applies inductively to the filtration levels A0A1AnA_0 \rightarrowtail A_1 \rightarrowtail \cdots \rightarrowtail A_n: at each step, we extend the cofibration one level, using the approximation property and the fact that cofibrations are stable under pushout.

Step 4: App 2 ensures uniqueness. The condition App 2 ensures that the maps constructed in Step 3 are unique up to weak equivalence in A\mathcal{A}. This makes the over-category filtered (any two objects have a common "approximation"), and filtered categories have contractible nerves.

Step 5: Geometric realization. Since wSnF|wS_n F| is a homotopy equivalence for each nn, and realization of a map of simplicial spaces that is a level-wise equivalence is an equivalence, we conclude wSF|wS_\bullet F| is a homotopy equivalence, giving K(F):K(A)K(B)K(F): K(\mathcal{A}) \simeq K(\mathcal{B}). \square


Generalizations

RemarkDG-category and stable infinity-category versions

The approximation theorem has been generalized to more modern settings:

  1. DG-categories: For a DG-functor F:ABF: \mathcal{A} \to \mathcal{B} between pretriangulated DG-categories, if FF induces an equivalence on homotopy categories [A][B][\mathcal{A}] \xrightarrow{\sim} [\mathcal{B}], then K(F)K(F) is an equivalence. This is Waldhausen's approximation theorem in the DG setting.

  2. Stable \infty-categories: In Blumberg--Gepner--Tabuada's framework, K-theory is a functor from stable \infty-categories to spectra, and it sends equivalences to equivalences. The approximation theorem is automatic: if F:CDF: \mathcal{C} \to \mathcal{D} is an equivalence of stable \infty-categories, then K(C)K(D)K(\mathcal{C}) \simeq K(\mathcal{D}).

  3. Invariance under derived equivalence: For rings, if Db(mod-R)Db(mod-S)D^b(\operatorname{mod}\text{-}R) \simeq D^b(\operatorname{mod}\text{-}S) as triangulated categories, then Kn(R)Kn(S)K_n(R) \cong K_n(S) for all nn. This is because the derived equivalence lifts to a Waldhausen equivalence (or DG equivalence) and the approximation theorem applies.