ConceptComplete

Lambda-Ring Structure and Adams Operations

The K-theory ring K0(X)K_0(X) of a scheme carries a rich additional structure: the lambda-ring structure coming from exterior powers. The associated Adams operations provide eigenspace decompositions and connect K-theory to motivic cohomology via the gamma filtration.


Lambda-rings

Definition5.6Lambda-ring

A lambda-ring is a commutative ring RR equipped with operations λk:RR\lambda^k: R \to R for k0k \geq 0 satisfying:

  1. λ0(x)=1\lambda^0(x) = 1 for all xRx \in R.
  2. λ1(x)=x\lambda^1(x) = x for all xRx \in R.
  3. λk(x+y)=i=0kλi(x)λki(y)\lambda^k(x + y) = \sum_{i=0}^k \lambda^i(x) \lambda^{k-i}(y) (additivity via exterior power splitting).
  4. λk(xy)=Pk(λ1(x),,λk(x);λ1(y),,λk(y))\lambda^k(xy) = P_k(\lambda^1(x), \ldots, \lambda^k(x); \lambda^1(y), \ldots, \lambda^k(y)) where PkP_k is a universal polynomial.
  5. λk(λl(x))=Pk,l(λ1(x),,λkl(x))\lambda^k(\lambda^l(x)) = P_{k,l}(\lambda^1(x), \ldots, \lambda^{kl}(x)) where Pk,lP_{k,l} is another universal polynomial (the plethysm).

The lambda operations on K0(X)K_0(X) are defined by λk([E])=[ΛkE]\lambda^k([\mathcal{E}]) = [\Lambda^k \mathcal{E}] (the kk-th exterior power).

ExampleLambda operations on K₀(ℙⁿ)

On Pn\mathbb{P}^n with K0(Pn)=Z[h]/(hn+1)K_0(\mathbb{P}^n) = \mathbb{Z}[h]/(h^{n+1}) where h=[O(1)]1h = [\mathcal{O}(1)] - 1:

The lambda operations on the generator [O(1)][\mathcal{O}(1)] are trivial (λk([O(1)])=0\lambda^k([\mathcal{O}(1)]) = 0 for k2k \geq 2 since O(1)\mathcal{O}(1) has rank 1). For hh:

λt(h)=λt([O(1)]1)=λt([O(1)])λt(1)=1+t[O(1)]1+t=1+t(1+h)1+t\lambda_t(h) = \lambda_t([\mathcal{O}(1)] - 1) = \frac{\lambda_t([\mathcal{O}(1)])}{\lambda_t(1)} = \frac{1 + t[\mathcal{O}(1)]}{1 + t} = \frac{1 + t(1+h)}{1+t}

where λt(x)=k0λk(x)tk\lambda_t(x) = \sum_{k \geq 0} \lambda^k(x) t^k is the generating function. Since λt(1)=1+t\lambda_t(1) = 1 + t (the class [R][R] has λk([R])=0\lambda^k([R]) = 0 for k2k \geq 2).

Expanding: λt(h)=1+th1+t=1+tht2h+t3h\lambda_t(h) = 1 + \frac{th}{1+t} = 1 + th - t^2h + t^3h - \cdots, so λk(h)=(1)k1h\lambda^k(h) = (-1)^{k-1}h for k1k \geq 1.


Adams operations

Definition5.7Adams operations

The Adams operations ψk:K0(X)K0(X)\psi^k: K_0(X) \to K_0(X) for k1k \geq 1 are ring homomorphisms defined by Newton's identity from the lambda operations:

ψk(x)=(1)k+1kλk(x)+i=1k1(1)i+1λi(x)ψki(x).\psi^k(x) = (-1)^{k+1} k \lambda^k(x) + \sum_{i=1}^{k-1} (-1)^{i+1} \lambda^i(x) \psi^{k-i}(x).

Equivalently, they are characterized by the generating function identity:

ψt(x):=k1ψk(x)tk1=ddtlogλt(x)1\psi_t(x) := \sum_{k \geq 1} \psi^k(x) t^{k-1} = \frac{d}{dt} \log \lambda_{-t}(x)^{-1}

or the property that for a line bundle L\mathcal{L}: ψk([L])=[Lk]\psi^k([\mathcal{L}]) = [\mathcal{L}^{\otimes k}].

Key properties:

  • ψk\psi^k is a ring homomorphism: ψk(xy)=ψk(x)ψk(y)\psi^k(xy) = \psi^k(x)\psi^k(y).
  • ψkψl=ψkl\psi^k \circ \psi^l = \psi^{kl}.
  • ψp(x)xp(modp)\psi^p(x) \equiv x^p \pmod{p} for prime pp (Frobenius-like).
  • For a vector bundle of rank rr: ψk([E])=[Symk(E)](lower terms)\psi^k([\mathcal{E}]) = [\operatorname{Sym}^k(\mathcal{E})] - (\text{lower terms}).
ExampleAdams operations on vector bundles

For a vector bundle E\mathcal{E} with Chern roots α1,,αr\alpha_1, \ldots, \alpha_r:

ψk([E])=i=1r[Lik]\psi^k([\mathcal{E}]) = \sum_{i=1}^r [\mathcal{L}_i^{\otimes k}]

where Li\mathcal{L}_i are the "formal line bundles" with c1(Li)=αic_1(\mathcal{L}_i) = \alpha_i. Under the Chern character:

ch(ψk(E))=i=1rekαi\operatorname{ch}(\psi^k(\mathcal{E})) = \sum_{i=1}^r e^{k\alpha_i}

So on CHp(X)QK0(X)QCH^p(X) \otimes \mathbb{Q} \subset K_0(X) \otimes \mathbb{Q} (via the Chern character), ψk\psi^k acts as multiplication by kpk^p.

For E=TP2\mathcal{E} = T_{\mathbb{P}^2} (tangent bundle of P2\mathbb{P}^2, rank 2, c1=3Hc_1 = 3H, c2=3H2c_2 = 3H^2):

  • ψ2([E])=2ch0+22ch1+24ch2=\psi^2([\mathcal{E}]) = 2\operatorname{ch}_0 + 2 \cdot 2 \cdot \operatorname{ch}_1 + 2 \cdot 4 \cdot \operatorname{ch}_2 = ... actually the computation uses ψ2=(λ1)22λ2=[E]22[Λ2E]=[EE]2[detE]\psi^2 = (\lambda^1)^2 - 2\lambda^2 = [\mathcal{E}]^2 - 2[\Lambda^2\mathcal{E}] = [\mathcal{E} \otimes \mathcal{E}] - 2[\det \mathcal{E}].

The gamma filtration

Definition5.8Gamma filtration

The gamma filtration on K0(X)K_0(X) is defined by the gamma operations γk(x)=λk(x+(k1)1)\gamma^k(x) = \lambda^k(x + (k-1) \cdot 1) for xK~0(X)=ker(rk)x \in \widetilde{K}_0(X) = \ker(\operatorname{rk}):

FγnK0(X)=subgroup generated by products γi1(x1)γir(xr) with i1++irnF^n_\gamma K_0(X) = \text{subgroup generated by products } \gamma^{i_1}(x_1) \cdots \gamma^{i_r}(x_r) \text{ with } i_1 + \cdots + i_r \geq n

where xjK~0(X)x_j \in \widetilde{K}_0(X).

The graded pieces are:

grγnK0(X)=Fγn/Fγn+1\operatorname{gr}^n_\gamma K_0(X) = F^n_\gamma / F^{n+1}_\gamma

The Chern character identifies:

grγnK0(X)QCHn(X)Q\operatorname{gr}^n_\gamma K_0(X) \otimes \mathbb{Q} \cong CH^n(X) \otimes \mathbb{Q}

and the Adams operations act on grγn\operatorname{gr}^n_\gamma by multiplication by knk^n.

ExampleGamma filtration and torsion

The gamma filtration can detect torsion phenomena invisible to the Chern character:

For X=B(Z/p)X = B(\mathbb{Z}/p) (the classifying space of a cyclic group, in an algebraic setting via equivariant K-theory):

K0(B(Z/p))R(Z/p)=Z[ζp]/Φp(ζp)Z[t]/(1+t++tp1)K_0(B(\mathbb{Z}/p)) \cong R(\mathbb{Z}/p) = \mathbb{Z}[\zeta_p] / \Phi_p(\zeta_p) \cong \mathbb{Z}[t]/(1 + t + \cdots + t^{p-1})

The gamma filtration gives grγnK0Z/p\operatorname{gr}^n_\gamma K_0 \cong \mathbb{Z}/p for 1np21 \leq n \leq p-2 and grn=0\operatorname{gr}^n = 0 for np1n \geq p-1. The successive quotients are all pp-torsion, reflecting the pp-torsion in the Chow groups/motivic cohomology of B(Z/p)B(\mathbb{Z}/p).

This example illustrates that the gamma filtration captures arithmetic information (torsion primes) that the rational Chern character misses.


Relation to motivic cohomology

RemarkAdams eigenspaces and motivic decomposition

The Adams operations provide a motivic decomposition of K-theory. For a smooth variety XX over a field:

Kn(X)Q=p0Kn(X)(p)K_n(X) \otimes \mathbb{Q} = \bigoplus_{p \geq 0} K_n(X)^{(p)}

where Kn(X)(p)={xKn(X)Q:ψk(x)=kpx for all k}K_n(X)^{(p)} = \{x \in K_n(X) \otimes \mathbb{Q} : \psi^k(x) = k^p x \text{ for all } k\} is the weight-pp eigenspace.

The Beilinson--Soule conjecture predicts Kn(X)(p)=0K_n(X)^{(p)} = 0 for p<n/2p < n/2 (the "vanishing conjecture"). For n=0n = 0, this says K0(X)(p)=0K_0(X)^{(p)} = 0 for p<0p < 0, which is trivially true. For n=1n = 1, this says K1(X)(0)=0K_1(X)^{(0)} = 0, which is true for connected XX.

The eigenspaces are identified with motivic cohomology:

Kn(X)(p)Hmot2pn(X,Q(p))K_n(X)^{(p)} \cong H^{2p-n}_{\text{mot}}(X, \mathbb{Q}(p))

This is the rational version of the Atiyah--Hirzebruch spectral sequence in motivic cohomology:

E2p,q=Hmotpq(X,Z(q))    Kpq(X)E_2^{p,q} = H^{p-q}_{\text{mot}}(X, \mathbb{Z}(-q)) \implies K_{-p-q}(X)

which degenerates rationally by the existence of Adams operations.