Algebraic K-theory
Daniel Ramras
Typed by Virgil Chan
29 October 2004 – 11 November 2004 2 Contents
Preface 5
1 Projective Modules7 1.1 Equivalent Definitions...... 7 1.2 Topological K-theory...... 7 1.3 K0 for Schemes...... 8
2 Higher K-theory9 2.1 Simplicial Sets...... 9 2.1.1 Canonical Example–Singular Sets...... 10 2.1.2 Formal Definition...... 11 2.1.3 Geometric Realisation of a Simplicial Set...... 11 2.2 Classifying Spaces of a Small Category...... 12 2.3 Definition of Higher K-theory...... 14 2.3.1 The Q-construction...... 14 2.3.2 Exact Categories and the Q-construction...... 16 2.3.3 Universal Property of the Q-construction...... 18 2.3.4 An Explicit Map α : K0(R) → π1 Q P ...... 23 2.3.5 Another Approach to π1 Q M ...... 26
3 Bloch’s Formula for Chow Groups 31 3.1 Product...... 35
4 Theorem A and D´evissage 39
3 4 CONTENTS Preface
These are notes from a series of three lectures I gave at Stanford in Fall 2004. They cover ∼ Swan’s Theorem, the Q-construction, the isomorphism π1 Q P (R) = K0(R), Bloch’s Formula, and D´evissage. Thanks to Virgil Chan for turning my handwritten notes into a readable document.
5 6 CONTENTS Chapter 1
Projective Modules
1.1 Equivalent Definitions
Let R be a (commutative) ring with identity 1 = 1R; M be a finitely generated R-module. We say M is projective if it is a direct summand of a free module. We denote P (R) the category of projective R-modules with module homomorphisms, then the Grothendieck group K0(R) is defined as follows: Let F be the free abelain group with a generator hP i for each isomorphism class of projective modules over R, and let R be the subgroup of F generated by hP ⊕ Qi − hP i − hQi . (Note: P ⊕ Q ∈ P (R) whenever P , Q ∈ P (R)). We define
K0(R) := F/R and write [P ] for hP i R. In other words, we formally add inverse to the monoid M of isomorphism classes. K0(R) is universal in the sense that any monoid map M → A with A an abelian group factors through K0(R) uniquely. Remark [P ] = [Q] if and only if P is stably isomorphic to Q, that is, there exists n such n ∼ n that P ⊕ R = Q ⊕ R . So K0(R) is the ”group of modules up to stable isomorphism”. ∼ Example 1.1.1. If R is a field, then K0(R) = Z, generated by [R].
1.2 Topological K-theory
Let X be a compact Hausdorff space. We form a group K0(X) just as before, replacing projective modules by vector bundles over X. (We can use complex, real, or quaternion, but 0 0 n ∼ 0 n let us use C.) Again, in K (X) we have [V ] = [V ] if and only if V ⊕ ε = V ⊕ ε for some n n n, where ε is the n-dimensional trivial bundle X × C → X.
7 8 CHAPTER 1. PROJECTIVE MODULES
0 ∼ Theorem 1.2.1 ([Swa62]). K (X) = K0(C(X)), where C(X) is the ring of continuous functions X → C.
Sketch of Proof. We need to show that vector bundles correspond to projective C(X)-modules:
{V → X} ←→ Γ(V ) := {continuous sections of V } . If V is a vector bundle, then there exists W such that V ⊕ W = εn, so Γ(V ) ⊕ Γ(W ) = Γ(εn) = C(X)n. i.e., Γ is projective. If M is a projective, finitely generated C(X)-module, then M ⊕ P ∼= C(X)n for some C(X)-module P and some integer n. So there exists f : C(X)n → C(X)n with f 2 = f. (f(C(X)) = M.) Now f corresponds to a map g : εn → εn with g2 = g, and we can check that im (g) is a vector bundle with sections Γ(im (g)) = M. (To get g, note Γ(ξ)/{s ∈ Γ | s(X) = 0} is fibre over X. So f induces map on each fibre, which is a bundle map.)
1.3 K0 for Schemes
Let X be a N¨otherianscheme. We define K0(X) to be the Grothendieck group of the category of coherent sheaves on X (sheaves which are locally finitely generated modules). This is the group with generators [F] for each isomorphism classes of coherent sheaves on X, and relations
[F] = [F 0] + [F 00] for each short exact sequence
0 → F 0 → F → F 00 → 0. Note we are not assume such sequences split in the category of coherent sheaves, by in K-theory, [F] = [F 0] + [F 00] = [F 0 ⊕ F 00] for each short exact sequence
0 → F 0 → F → F 00 → 0. This is the case originally studied by Grothendieck in his proof of the Riemann-Roch Theo- rem. Chapter 2
Higher K-theory
For a commutative ring R with identity, we define
ab K1(R) := GL (∞,R) ab = colim GL (n, R) , n
K2(R) := group of non-trivial relations among elementary matrices.
λ Recall an elementary matrix eij ∈ GL (n, Λ) has 1’s on the main diagonal, λ in the (i, j)- entry, and 0’s else where. The groups K0, K1, and K2 are related by exact sequences. For example, if Λ is a Dedekind domain with fraction field F , there is a localisation sequence:
M M K2(F ) K1(Λ/p) K1(Λ) K1(F ) K0(Λ/p) K0(Λ) K0(F ) 0 p p
Quillen extended this to the left by defining higher K-groups (even to odd other K2-terms one needs a completely new way of thinking about K0, K1, K2).
2.1 Simplicial Sets
A simplicial set X• provides a combinatorial description of a topological space. It is a sequence of sets X0,X1,X2, ···, and structure maps
di : Xn → Xn−1,
si : Xn → Xn+1 for each i with 0 ≤ i ≤ n, satisfying the following relations
9 10 CHAPTER 2. HIGHER K-THEORY
didj = dj−1di if i < j, disj = sj−1di if i < j, djsj = dj+1sj = id, (2.1.1) disj = sjdi−1 if i > j + 1, sisj = sj+1si if i ≤ j.
We think of Xi as the collection of i-simplices of X•, with face di and degeneracy si operators. For example, given an “edge” e ∈ X1, its “endpoints” are the “vertices” d0(e), d1(e) ∈ X0. si’s correspond to repeating a vertex in a simplex. For instance, if v is a vertex, then (v, v) is a degenerate 1-simplex. More formally, if σ ∈ Xn is a simplex with an ordered set of vertices σ = [v0, v1, ··· , vn], then
di(σ) = [v0, v1, ··· , vˆi, ··· , vn] ∈ Xn−1, and
si(σ) = [v0, ··· , vi, vi, ··· , vn] ∈ Xn+1. From this readers can verify Equation (2.1.1) easily.
2.1.1 Canonical Example–Singular Sets Let X be any topological space, we define the singular set of X to be the simplicial set
n Sn(X) := {f : ∆ → X | f is continuous} , ( n )
n n+1 X where ∆ := (x0, ··· , xn) ∈ R 0 ≤ xi ≤ 1, xi = 1 is the standard n-simplex. To i=0 define face and degeneracy operators, we first define
n−1 n δi : ∆ → ∆ (inclusion as i-th face) , n+1 n σi : ∆ → ∆ (collapse onto i-th face) by
δi(x0, ··· , xn−1) = (x0, ··· , xi−1, 0, ··· , xn−1)
σi(x0, ··· , xn+1) = (x0, ··· , xi + xi+1, ··· , xn+1). Then define
di(f) := f ◦ δi,
si(f) := f ◦ σi. 2.1. SIMPLICIAL SETS 11
2.1.2 Formal Definition
op op A simplicial set is a functor X• : ∆ → Sets, where ∆ is the category of finite totally ordered sets and order-preserving morphisms. So if [n] := {0 < 1 < ··· < n}, then Xn := X•([n]) is the set of n-simplices, and the face and degeneracy faces are induced by the maps
δi :[n] → [n + 1] σi :[n] → [n − 1] j if j < i, j if j ≤ i, δ (j) = (skip i) σ (j) = (repeat i) i j + 1 if j ≥ i i j − 1 if j > i
Example 2.1.1 (The Simplicial Interval). The simplicial interval I• := ∆[1] is the simplicial set
I0 = {0, 1} ,
I1 = {[0, 1], [0, 0], [1, 1]} ,
I2 = {[0, 0, 1], [0, 1, 1]} ,
with
d0([0, 1]) = 1,
d1([0, 1]) = 0
and et cetera. For a formal definition of I•, see Example 2.2.2.
2.1.3 Geometric Realisation of a Simplicial Set
Given a simpicial set X•, we want to create a topological space ! G n |X•| := Xn × ∆ /∼ . n≥0 The relation “∼” identifies faces and collapses degeneracies:
n (τ, δi(v)) ∼ (di(τ), v) for τ ∈ Xn+1 and v ∈ ∆ , n (τ, σi(v)) ∼ (si(τ), v) for τ ∈ Xn−1 and v ∈ ∆ . 12 CHAPTER 2. HIGHER K-THEORY
Theorem 2.1.2 (Milnor).
(i)[Mil57, Theorem 1] |X•| is a CW-complex with one n-cell for each non-degenerate simplex τ ∈ Xn.
(ii)[Mil57, Theorem 2] |X• × Y•| and |X•| × |Y•| are weakly equivalent, and are home- omorphic when |X•| or |Y•| is locally finite.
Here, X• × Y• has n-simplices Xn × Yn and si, di operate coordinate-wise.
Remark 2.1.3.1. Earlier definitions of simplicial sets ignored degeneracies and hence we did not get |X• × Y•| = |X•| × |Y•|. Note that (σ, τ) ∈ Xn × Yn is degenerate only when σ and τ are both degenerate.
Remark 2.1.3.2. A map f : X• → Y• (i.e., fn : Xn → Yn commutes with di, si) defines a continuous map |f| : |X•| → |Y•| by f(xn, t) = (f(xn), t).
2.2 Classifying Spaces of a Small Category
Let C be a small category (i.e., Obj (C), MorC (X,Y ) are all sets). We think of C as a concrete combinatorial object, not as an abstraction of some entire field of math. For us, categories will serve as a mean of describing complicated spaces algebraically/combinatorially.
Definition 2.2.1 (Nerve of a Category). Let C be a small category. We construct a simplicial set BC : ∆op → Sets, called the nerve of C, by
(BC)n = BC([n]) := {A0 → · · · → An} .
For Φ ∈ Mor∆op ([n], [m]), its image (BC)(Φ) is defined as
ϕ0 ϕn−1 ϕΦ(0) ϕΦ(n−1) A0 −→· · · −−−→ An 7→ AΦ(0) −−−→· · · −−−−→ AΦ(n)
Face operators are given by composing morphisms, and degeneracy operators insert identity morphisms.
For instance,
ϕ0 ϕ1 ϕ2 ϕ0 ϕ2◦ϕ1 d2 A0 −→ A1 −→ A2 −→ A3 = A0 −→ A1 −−−→ A3 ,
ϕ0 ϕ1 ϕ0 id ϕ1 s1 A0 −→ A1 −→ A2 = A0 −→ A1 −→ A1 −→ A2 2.2. CLASSIFYING SPACES OF A SMALL CATEGORY 13
Example 2.2.2 (Formal Definition of Simplicial Interval). Let C be the category with two objects 0, 1, and
MorC (0, 1) := {0 → 1, id0, id1} . Then |BC| = ∆[1]. What is BC then? We have
BC0 = {0, 1} ,
BC1 = {0 → 1, degenerate simplices} ,
BCi = {degenerate simplices} for i > 1. G So |BC| = {0, 1} × ∆0 {0 → 1} × ∆1 /∼ is the following graph
0 1 • •
which is just the standard 1-simplex.
Proposition 2.2.3. A functor F : C → D induces a continuous map |F | : |BC| → |BD|. A natural transformation η : F → G of functors F,G : C → D induces a homotopy from |F | to |G|. Notice we assume that all functors here are covariant.
Proof. A functor F : C → D induces a simplicial map BC → BD, i.e., maps BCi onto BDi for each i which commute with all face and degeneracy maps. These maps are obvious:
F (A0 → · · · → An) = F (A0) → · · · → F (An). Now a natural transformation η : F → G is really a collection of commuting diagrams:
F (ϕ) F (A) F (B)
ηA ηB
G(A) G(B), G(ϕ)
ϕ one for each map A −→ B in C. Another description of η is that by considering the category I × C (I is the category in Example 2.2.2), we define
η˜ : I × C → D by 14 CHAPTER 2. HIGHER K-THEORY
η˜(0,A) := F (A) η˜(1,A) := G(A), ϕ η˜(id0,A −→ B) := F (ϕ), ϕ η˜(id1,A −→ B) := G(ϕ), ϕ η˜(0 → 1,A −→ B) := G(ϕ) ◦ ηA =η ˜B ◦ F (ϕ)
So we get a functorη ˜ : I × C → D, hence a map
|I × C| D
|I| × |C|
by Milnor’s Theorem 2.1.2, and I = [0, 1], so this is exactly a homotopy from |F | to |G|.
2.3 Definition of Higher K-theory
In order to define the higher K-groups of a ring, we will work with the category P (R) of
projective R-modules. Unfortunately, one cannot just consider the space BP (R) . Instead, we will need to modify the morphisms to construct a category Q P (R). Then the homotopy groups