ProbXiv
Explorersign in
Problem archiveProblem record

Statement

Let kk be a field and let H∗H^{*} be a Weil cohomology theory on smooth projective varieties over kk with coefficients in a field FF of characteristic 00: for each smooth projective XX over kk of dimension nn, a finite-dimensional graded FF-algebra H∗(X)=⨁i≥0Hi(X)H^{*}(X)=\bigoplus_{i\ge 0}H^{i}(X) with a trace isomorphism trX ⁣:H2n(X)→ ∼ F\mathrm{tr}_{X}\colon H^{2n}(X)\xrightarrow{\ \sim\ }F inducing a perfect Poincar'e pairing Hi(X)×H2n−i(X)→FH^{i}(X)\times H^{2n-i}(X)\to F, (x,y)↦trX(x∪y)(x,y)\mapsto \mathrm{tr}_{X}(x\cup y), a K"unneth isomorphism H∗(X×Y)≅H∗(X)⊗H∗(Y)H^{*}(X\times Y)\cong H^{*}(X)\otimes H^{*}(Y), and cycle class maps γX ⁣:Zi(X)Q→H2i(X)\gamma_{X}\colon Z^{i}(X)_{\mathbb{Q}}\to H^{2i}(X) (where Zi(X)QZ^{i}(X)_{\mathbb{Q}} denotes codimension-ii algebraic cycles on XX with rational coefficients) compatible with products, pushforwards and pullbacks, and satisfying the weak and hard Lefschetz axioms.

For smooth projective X,YX,Y over kk with dim⁡X=n\dim X=n, a cycle α∈Z∗(X×Y)Q\alpha\in Z^{*}(X\times Y)_{\mathbb{Q}} induces FF-linear maps

H∗(X)⟶H∗(Y),x⟼pY∗(γX×Y(α)∪pX∗x),H^{*}(X)\longrightarrow H^{*}(Y),\qquad x\longmapsto p_{Y*}\bigl(\gamma_{X\times Y}(\alpha)\cup p_{X}^{*}x\bigr),

where pX,pYp_{X},p_{Y} are the two projections. A collection of FF-linear maps Hi(X)→Hj(X)H^{i}(X)\to H^{j}(X) is called algebraic if it arises in this way from a single cycle in Z∗(X×X)QZ^{*}(X\times X)_{\mathbb{Q}}.

Fix a smooth projective X⊂PNX\subset\mathbb{P}^{N} over kk of dimension nn and a smooth hyperplane section W=X∩HW=X\cap \mathcal{H}, and let

L ⁣:Hi(X)→Hi+2(X),L(x)=x∪γX(W)L\colon H^{i}(X)\to H^{i+2}(X),\qquad L(x)=x\cup \gamma_{X}(W)

be the associated Lefschetz operator, so that by hard Lefschetz Ln−i ⁣:Hi(X)→H2n−i(X)L^{n-i}\colon H^{i}(X)\to H^{2n-i}(X) is an isomorphism for all i≤ni\le n. Define

Λ=(Ln−i+2)−1∘L∘L n−i ⁣:Hi(X)→Hi−2(X)(i≤n),\Lambda=\bigl(L^{n-i+2}\bigr)^{-1}\circ L\circ L^{\,n-i}\colon H^{i}(X)\to H^{i-2}(X)\quad (i\le n), Λ=L n−i∘L∘(Ln−i+2)−1 ⁣:H2n−i+2(X)→H2n−i(X)(i≤n).\Lambda=L^{\,n-i}\circ L\circ \bigl(L^{n-i+2}\bigr)^{-1}\colon H^{2n-i+2}(X)\to H^{2n-i}(X)\quad (i\le n).

Let πi∈EndF(H∗(X))\pi^{i}\in\mathrm{End}_{F}\bigl(H^{*}(X)\bigr), 0≤i≤2n0\le i\le 2n, be the projector onto Hi(X)H^{i}(X); equivalently, πi\pi^{i} is the component of the class of the diagonal ΔX⊂X×X\Delta_{X}\subset X\times X in the summand Hi(X)⊗H2n−i(X)H^{i}(X)\otimes H^{2n-i}(X) of H2n(X×X)H^{2n}(X\times X).

Call a cycle z∈Zi(X)Qz\in Z^{i}(X)_{\mathbb{Q}} numerically equivalent to zero if deg⁡(z⋅z′)=0\deg(z\cdot z')=0 for every z′∈Zn−i(X)Qz'\in Z^{n-i}(X)_{\mathbb{Q}}, and homologically equivalent to zero if γX(z)=0\gamma_{X}(z)=0. Let Ai(X)=γX(Zi(X)Q)⊂H2i(X)A^{i}(X)=\gamma_{X}\bigl(Z^{i}(X)_{\mathbb{Q}}\bigr)\subset H^{2i}(X) be the Q\mathbb{Q}-vector space of algebraic classes, and let

Aprimi(X)=Ai(X)∩ker⁡(L n−2i+1 ⁣:H2i(X)→H2n−2i+2(X))(2i≤n).A^{i}_{\mathrm{prim}}(X)=A^{i}(X)\cap \ker\bigl(L^{\,n-2i+1}\colon H^{2i}(X)\to H^{2n-2i+2}(X)\bigr)\qquad (2i\le n).

Then, for every field kk, every such Weil cohomology theory H∗H^{*}, and every smooth projective variety XX over kk of dimension nn with a smooth hyperplane section as above, the following hold.

(B) The operator Λ\Lambda is algebraic: there is a cycle λ∈Zn−1(X×X)Q\lambda\in Z^{n-1}(X\times X)_{\mathbb{Q}} whose induced maps on cohomology coincide with Λ\Lambda in every degree.

(C) For each ii with 0≤i≤2n0\le i\le 2n, the projector πi\pi^{i} is algebraic: there is a cycle πi∈Zn(X×X)Q\pi^{i}\in Z^{n}(X\times X)_{\mathbb{Q}} inducing it.

(D) For every ii, a cycle in Zi(X)QZ^{i}(X)_{\mathbb{Q}} is numerically equivalent to zero if and only if it is homologically equivalent to zero.

(Hodge standard) For every ii with 2i≤n2i\le n, the Q\mathbb{Q}-valued symmetric bilinear form

(x,y)⟼(−1)i trX(L n−2i(x∪y))(x,y)\longmapsto (-1)^{i}\,\mathrm{tr}_{X}\bigl(L^{\,n-2i}(x\cup y)\bigr)

is positive definite on Aprimi(X)A^{i}_{\mathrm{prim}}(X).

Record

Added
Links

Comments

Nothing has been published against this problem yet, and nothing has been said about it.

Sign in with an institutional address to take part in the discussion. Reading every thread stays open to everyone.

Sign in

Solve with an agent

Open the statement in a chat, with the problem and the ground rules already written into the prompt.

This opens a third-party site. Nothing is posted back to ProbXiv and nothing you write there is recorded here — what a model gives you is an attempt, which a person still has to check.