K_0: Turning Modules into a Group

From a monoid to a group

Here is the recurring problem K-theory solves. You have a collection of objects you can add (direct sum) but not subtract, and you want a group. For a commutative ring R, take the set of isomorphism classes of finitely generated projective modules. Under direct sum this is a commutative monoid: associative, with the zero module as identity, but no inverses — you cannot un-add a module. The Grothendieck group K_0(R) is the universal group you get by formally adding inverses, exactly the way Z is built from the natural numbers under addition.

Concretely, elements of K_0(R) are formal differences [P] − [Q] of projective modules, with the rule that [P ⊕ Q] = [P] + [Q]. Two such differences are equal when they agree after adding a common module — the same trick that makes 3 − 5 and 1 − 3 name the same integer.

The universal property, precisely

The construction is governed by a clean universal property. Let (M, +) be a commutative monoid. Then there is an abelian group K(M) and a monoid map M → K(M) such that every monoid map M → A into an abelian group factors uniquely through it. This makes K(M) unique up to isomorphism, so we never have to fuss over which construction we chose — a recurring comfort in this subject.

1. Form pairs (a, b) ∈ M × M, thinking of (a, b) as the formal difference a − b.
2. Declare (a, b) ∼ (c, d) when a + d + e = b + c + e for some e ∈ M (the extra e absorbs non-cancellable elements).
3. Add componentwise; the quotient by ∼ is K(M), with −(a, b) = (b, a).

Example: R = field k (or any PID, e.g. Z).
Every finitely generated projective module over a field is free,
so it is just k^n, determined by its rank n >= 0.

Monoid of iso-classes:  {0, k, k^2, k^3, ...}  =  (N, +)  via n <-> k^n.
[k^m] + [k^n] = [k^(m+n)]   matches   m + n.

Grothendieck group of (N, +) is (Z, +).
The map  [P] |-> rank(P)  is an isomorphism

        K_0(k)  ~=  Z ,   and likewise   K_0(Z) ~= Z.

The class [k^n] goes to n; the formal difference [k^a] - [k^b] goes to a - b.

Stably free, and the first surprise

In K_0 we can only ever detect modules up to adding free pieces. A projective P is stably free if P ⊕ R^m ≅ R^n for some m, n. Such a P contributes [P] = n − m to K_0 — the same class a genuinely free module would. So K_0 is blind to the difference between stably free and free; that distinction is precisely what the next invariant, K_1, will start to see.

For most rings you meet first — fields, Z, any PID — K_0 is just Z, and the reduced part K̃_0 = K_0 / Z vanishes. The invariant only comes alive when projectives can fail to be free. The cleanest example is the ring of integers of a number field: there K̃_0(O_K) is the ideal class group, and we will return to that beautiful coincidence in Guide 5.