ConceptComplete

K-Theory of Schemes

Algebraic K-theory extends naturally from rings to schemes, where it captures deep geometric and arithmetic information. For a scheme XX, the K-groups Kn(X)K_n(X) are defined using vector bundles (or perfect complexes), while the G-theory groups Gn(X)G_n(X) use coherent sheaves.


Definitions

Definition5.1K-theory of a scheme

For a scheme XX, define:

  1. Kn(X)K_n(X) (K-theory): the Quillen K-groups of the exact category V(X)\mathcal{V}(X) of locally free sheaves (vector bundles) of finite rank on XX: Kn(X)=Kn(V(X)).K_n(X) = K_n(\mathcal{V}(X)).

  2. Gn(X)G_n(X) (G-theory): the Quillen K-groups of the abelian category Coh⁑(X)\operatorname{Coh}(X) of coherent sheaves on XX (when XX is Noetherian): Gn(X)=Kn(Coh⁑(X)).G_n(X) = K_n(\operatorname{Coh}(X)).

There is a natural Cartan map Kn(X)β†’Gn(X)K_n(X) \to G_n(X) induced by the inclusion V(X)β†ͺCoh⁑(X)\mathcal{V}(X) \hookrightarrow \operatorname{Coh}(X). By the resolution theorem, this is an isomorphism when XX is regular.

RemarkFunctoriality

The two theories have complementary functoriality:

K-theory Kn(X)K_n(X):

  • Contravariant for arbitrary morphisms f:Xβ†’Yf: X \to Y: pullback fβˆ—:Kn(Y)β†’Kn(X)f^*: K_n(Y) \to K_n(X) via E↦fβˆ—E\mathcal{E} \mapsto f^*\mathcal{E}.
  • Makes K0(X)K_0(X) a commutative ring via tensor product: [E]β‹…[F]=[EβŠ—F][\mathcal{E}] \cdot [\mathcal{F}] = [\mathcal{E} \otimes \mathcal{F}].

G-theory Gn(X)G_n(X):

  • Contravariant for flat morphisms: fβˆ—:Gn(Y)β†’Gn(X)f^*: G_n(Y) \to G_n(X) when ff is flat.
  • Covariant for proper morphisms: fβˆ—:Gn(X)β†’Gn(Y)f_*: G_n(X) \to G_n(Y) via Fβ†¦βˆ‘(βˆ’1)i[Rifβˆ—F]\mathcal{F} \mapsto \sum (-1)^i [R^i f_* \mathcal{F}].
  • G0(X)G_0(X) is a module over K0(X)K_0(X) via [E]β‹…[F]=[EβŠ—F][\mathcal{E}] \cdot [\mathcal{F}] = [\mathcal{E} \otimes \mathcal{F}].

Kβ‚€ of smooth varieties

ExampleKβ‚€ of a smooth curve

For a smooth projective curve CC over a field kk:

K0(C)β‰…ZβŠ•Pic⁑(C)K_0(C) \cong \mathbb{Z} \oplus \operatorname{Pic}(C)

where Z\mathbb{Z} is the rank and Pic⁑(C)\operatorname{Pic}(C) is the Picard group. Every vector bundle E\mathcal{E} on CC satisfies [E]=rk⁑(E)[O]+[det⁑E]βˆ’[O][\mathcal{E}] = \operatorname{rk}(\mathcal{E})[\mathcal{O}] + [\det \mathcal{E}] - [\mathcal{O}] in K0(C)K_0(C) modulo higher filtration.

For C=P1C = \mathbb{P}^1: Pic⁑(P1)β‰…Z\operatorname{Pic}(\mathbb{P}^1) \cong \mathbb{Z} generated by O(1)\mathcal{O}(1), so K0(P1)β‰…Z2K_0(\mathbb{P}^1) \cong \mathbb{Z}^2 with basis [O],[O(1)][\mathcal{O}], [\mathcal{O}(1)].

For an elliptic curve EE: Pic⁑0(E)β‰…E(k)\operatorname{Pic}^0(E) \cong E(k) (the group of rational points), so K0(E)β‰…ZβŠ•E(k)βŠ•ZK_0(E) \cong \mathbb{Z} \oplus E(k) \oplus \mathbb{Z} where the first Z\mathbb{Z} is rank and the last Z\mathbb{Z} is the degree component of Pic⁑\operatorname{Pic}.

ExampleKβ‚€ of smooth surfaces

For a smooth projective surface SS over an algebraically closed field:

K0(S)βŠ—Qβ‰…CHβˆ—(S)βŠ—Q=⨁i=02CHi(S)βŠ—QK_0(S) \otimes \mathbb{Q} \cong CH^*(S) \otimes \mathbb{Q} = \bigoplus_{i=0}^{2} CH^i(S) \otimes \mathbb{Q}

via the Chern character. The integral structure is more subtle. The filtration FiK0(S)F^i K_0(S) by codimension of support gives:

gr⁑0K0(S)=Z(rank)\operatorname{gr}^0 K_0(S) = \mathbb{Z} \quad (\text{rank}) gr⁑1K0(S)=Pic⁑(S)(first Chern class)\operatorname{gr}^1 K_0(S) = \operatorname{Pic}(S) \quad (\text{first Chern class}) gr⁑2K0(S)=CH2(S)(second Chern class)\operatorname{gr}^2 K_0(S) = CH^2(S) \quad (\text{second Chern class})

For S=P2S = \mathbb{P}^2: K0(P2)β‰…Z3K_0(\mathbb{P}^2) \cong \mathbb{Z}^3 with basis [O],[O(1)],[O(2)][\mathcal{O}], [\mathcal{O}(1)], [\mathcal{O}(2)] and ring structure K0(P2)β‰…Z[h]/(h3)K_0(\mathbb{P}^2) \cong \mathbb{Z}[h]/(h^3) where h=[O(1)]βˆ’1h = [\mathcal{O}(1)] - 1.


Thomason--Trobaugh K-theory

Definition5.2K-theory of perfect complexes

For a general (possibly singular or non-Noetherian) scheme XX, the Thomason--Trobaugh K-theory is defined using the Waldhausen construction applied to the category of perfect complexes Perf⁑(X)\operatorname{Perf}(X):

Kn(X)=Kn(Perf⁑(X))K_n(X) = K_n(\operatorname{Perf}(X))

where Perf⁑(X)\operatorname{Perf}(X) is the full subcategory of the derived category D(OX-Mod⁑)D(\mathcal{O}_X\text{-}\operatorname{Mod}) consisting of complexes locally quasi-isomorphic to bounded complexes of locally free sheaves of finite rank.

For a Noetherian regular scheme, Perf⁑(X)=Db(Coh⁑(X))\operatorname{Perf}(X) = D^b(\operatorname{Coh}(X)) and this agrees with the Quillen definition. For singular schemes, Perf⁑(X)⊊Db(Coh⁑(X))\operatorname{Perf}(X) \subsetneq D^b(\operatorname{Coh}(X)).

ExampleK-theory of singular schemes

For the singular scheme X=Spec⁑k[x,y]/(xy)X = \operatorname{Spec} k[x, y]/(xy) (the node):

Using Thomason--Trobaugh K-theory:

  • K0(X)β‰…Z2K_0(X) \cong \mathbb{Z}^2 (one for each irreducible component, since every perfect complex splits at the generic points).
  • K1(X)β‰…kΓ—Γ—kΓ—K_1(X) \cong k^{\times} \times k^{\times} (units from each branch).
  • Kβˆ’1(X)β‰…ZK_{-1}(X) \cong \mathbb{Z} (detecting the singularity via Bass's negative K-theory).

The negative K-group Kβˆ’1(X)K_{-1}(X) vanishes for regular schemes and provides a measure of the severity of the singularity.


The Brown--Gersten--Quillen spectral sequence

Definition5.3BGQ spectral sequence

For a regular Noetherian scheme XX of finite Krull dimension, there is a convergent spectral sequence:

E1p,q=⨁x∈X(p)Kβˆ’pβˆ’q(k(x))β€…β€ŠβŸΉβ€…β€ŠKβˆ’pβˆ’q(X)E_1^{p,q} = \bigoplus_{x \in X^{(p)}} K_{-p-q}(k(x)) \implies K_{-p-q}(X)

where X(p)X^{(p)} denotes the set of points of codimension pp and k(x)k(x) is the residue field at xx.

The E1E_1-page is the Gersten complex:

⨁x∈X(0)Kn(k(x))→⨁x∈X(1)Knβˆ’1(k(x))→⋯→⨁x∈X(n)K0(k(x))\bigoplus_{x \in X^{(0)}} K_n(k(x)) \to \bigoplus_{x \in X^{(1)}} K_{n-1}(k(x)) \to \cdots \to \bigoplus_{x \in X^{(n)}} K_0(k(x))

and the E2E_2-terms are the Zariski cohomology of the K-theory sheaf:

E2p,q=HZarp(X,Kβˆ’q).E_2^{p,q} = H^p_{\text{Zar}}(X, \mathcal{K}_{-q}).

By Gersten's conjecture (proved for smooth varieties over a field), the Gersten complex is exact, so E2p,q=Hp(X,Kβˆ’q)E_2^{p,q} = H^p(X, \mathcal{K}_{-q}) where Kn\mathcal{K}_n is the Zariski sheaf associated to the presheaf U↦Kn(U)U \mapsto K_n(U).

RemarkRelation to Chow groups

From the BGQ spectral sequence, E2p,βˆ’p=HZarp(X,Kp)=CHp(X)E_2^{p,-p} = H^p_{\text{Zar}}(X, \mathcal{K}_p) = CH^p(X) (Chow groups). The edge map gives the cycle class map:

CHp(X)=E2p,βˆ’pβ†’E∞p,βˆ’pβ†ͺgr⁑pK0(X)CH^p(X) = E_2^{p,-p} \to E_\infty^{p,-p} \hookrightarrow \operatorname{gr}^p K_0(X)

which is Grothendieck's Chern character in the appropriate degree. After tensoring with Q\mathbb{Q}, the spectral sequence degenerates at E2E_2 and gives:

Kn(X)βŠ—Q≅⨁pHZarp(X,Kn+p)βŠ—QK_n(X) \otimes \mathbb{Q} \cong \bigoplus_{p} H^p_{\text{Zar}}(X, \mathcal{K}_{n+p}) \otimes \mathbb{Q}

relating K-groups to motivic cohomology.