ConceptComplete

The Grothendieck Group K₀

The Grothendieck group K0K_0 is the most fundamental invariant in algebraic K-theory. Given a ring RR, the group K0(R)K_0(R) encodes information about finitely generated projective modules over RR, modulo a relation that identifies modules related by direct sum decomposition.


Construction of K₀

Definition1.1Grothendieck group of a commutative monoid

Let (M,+,0)(M, +, 0) be a commutative monoid. The Grothendieck group (or group completion) of MM is an abelian group K(M)K(M) together with a monoid homomorphism ι:M→K(M)\iota: M \to K(M) satisfying the universal property: for every monoid homomorphism f:M→Af: M \to A with AA an abelian group, there exists a unique group homomorphism fˉ:K(M)→A\bar{f}: K(M) \to A such that fˉ∘ι=f\bar{f} \circ \iota = f.

Concretely, K(M)K(M) can be constructed as the quotient of the free abelian group Z[M]\mathbb{Z}[M] by the subgroup generated by elements [m+n]−[m]−[n][m + n] - [m] - [n] for all m,n∈Mm, n \in M. Equivalently, K(M)=(M×M)/∼K(M) = (M \times M) / \sim where (a,b)∼(c,d)(a, b) \sim (c, d) if there exists e∈Me \in M with a+d+e=b+c+ea + d + e = b + c + e.

Definition1.2K₀ of a ring

Let RR be a ring (associative, unital). The group K0(R)K_0(R) is the Grothendieck group of the commutative monoid (Proj⁡(R),⊕)(\operatorname{Proj}(R), \oplus), where Proj⁡(R)\operatorname{Proj}(R) denotes the set of isomorphism classes of finitely generated projective RR-modules under direct sum.

Explicitly, elements of K0(R)K_0(R) are formal differences [P]−[Q][P] - [Q] of finitely generated projective modules, with the relation [P]−[Q]=[P′]−[Q′][P] - [Q] = [P'] - [Q'] if and only if there exists a finitely generated projective module SS such that

P⊕Q′⊕S≅P′⊕Q⊕S.P \oplus Q' \oplus S \cong P' \oplus Q \oplus S.

ExampleK₀ of a field

If FF is a field, every finitely generated projective FF-module is free, so Proj⁡(F)≅N\operatorname{Proj}(F) \cong \mathbb{N} via [Fn]↦n[F^n] \mapsto n. The Grothendieck group is

K0(F)≅Z,K_0(F) \cong \mathbb{Z},

generated by [F][F]. The rank homomorphism rk⁡:K0(F)→Z\operatorname{rk}: K_0(F) \to \mathbb{Z} sending [P]−[Q]↦dim⁡P−dim⁡Q[P] - [Q] \mapsto \dim P - \dim Q is an isomorphism.

ExampleK₀ of a PID

If RR is a principal ideal domain (PID), then every finitely generated projective module is free (since submodules of free modules over PIDs are free). Thus K0(R)≅ZK_0(R) \cong \mathbb{Z}, generated by [R][R]. This applies to Z\mathbb{Z}, k[x]k[x], and all DVRs.


The rank and reduced K₀

Definition1.3Rank homomorphism and reduced K₀

For a ring RR, the rank homomorphism is the map

rk⁡:K0(R)→Z,[P]↦rk⁡(P)\operatorname{rk}: K_0(R) \to \mathbb{Z}, \quad [P] \mapsto \operatorname{rk}(P)

where the rank is computed at any prime ideal. The reduced Grothendieck group is

K~0(R)=ker⁡(rk⁡:K0(R)→Z).\widetilde{K}_0(R) = \ker(\operatorname{rk}: K_0(R) \to \mathbb{Z}).

Elements of K~0(R)\widetilde{K}_0(R) measure the failure of finitely generated projective modules to be stably free. If K~0(R)=0\widetilde{K}_0(R) = 0, then every finitely generated projective module is stably free (i.e., P⊕Rm≅RnP \oplus R^m \cong R^n for some m,nm, n).

ExampleK₀ of a Dedekind domain

For a Dedekind domain RR with fraction field FF, the structure theorem gives K0(R)≅Z⊕Cl⁡(R)K_0(R) \cong \mathbb{Z} \oplus \operatorname{Cl}(R), where Cl⁡(R)\operatorname{Cl}(R) is the ideal class group. The isomorphism sends [P][P] to (rk⁡(P),[det⁡P])(\operatorname{rk}(P), [\det P]) where det⁡P=⋀rk⁡PP\det P = \bigwedge^{\operatorname{rk} P} P.

For R=Z[−5]R = \mathbb{Z}[\sqrt{-5}], we have Cl⁡(R)≅Z/2Z\operatorname{Cl}(R) \cong \mathbb{Z}/2\mathbb{Z}, so K0(Z[−5])≅Z⊕Z/2ZK_0(\mathbb{Z}[\sqrt{-5}]) \cong \mathbb{Z} \oplus \mathbb{Z}/2\mathbb{Z}. The nontrivial element of K~0\widetilde{K}_0 is represented by the ideal (2,1+−5)(2, 1 + \sqrt{-5}).


Functorial properties

RemarkFunctoriality of K₀

K0K_0 is a covariant functor from rings to abelian groups. A ring homomorphism f:R→Sf: R \to S induces

f∗:K0(R)→K0(S),[P]↦[P⊗RS]f_*: K_0(R) \to K_0(S), \quad [P] \mapsto [P \otimes_R S]

via extension of scalars. This is well-defined because base change preserves finitely generated projective modules.

For commutative rings, K0(R)K_0(R) is a commutative ring with multiplication [P]⋅[Q]=[P⊗RQ][P] \cdot [Q] = [P \otimes_R Q] and identity [R][R]. The rank map rk⁡:K0(R)→Z\operatorname{rk}: K_0(R) \to \mathbb{Z} is then a ring homomorphism, and K~0(R)\widetilde{K}_0(R) is an ideal.

If RR is a commutative Noetherian ring, there is also the group G0(R)=K0(mod⁡-R)G_0(R) = K_0(\operatorname{mod}\text{-}R), the Grothendieck group of finitely generated (not necessarily projective) modules. The natural map K0(R)→G0(R)K_0(R) \to G_0(R) is an isomorphism when RR is regular (the resolution theorem).

ExampleMorita invariance

For the matrix ring Mn(R)M_n(R), Morita equivalence gives K0(Mn(R))≅K0(R)K_0(M_n(R)) \cong K_0(R) for all n≥1n \geq 1. The equivalence sends a projective Mn(R)M_n(R)-module PP to the projective RR-module ePeP where e=e11e = e_{11} is the matrix unit. This is a special case of the general principle: K0K_0 is an invariant of the Morita equivalence class of a ring.