Tate's conjecture, algebraic cycles and rational \(K\)-theory in characteristic \(p\) (Q1382105)

Let \(k\) be a field of characteristic \(p>0\). Write \({\mathcal M}={\mathcal M}_{\text{num}}(k)\) for the category of pure motives over \(k\) for numerical equivalence, and let \(\pi_M\) denote the Frobenius endomorphism of the object \(M\) of \({\mathcal M}\). For \(k={\mathbb F}_q\) one can state Tate's conjecture in its strong form as follows: For a smooth projective variety \(X\) over \(k\) one has \[ \dim_{\mathbb Q}A^r_{\text{num}}(X)=\text{ ord}_{s=r}\zeta(X,s), \] where \(A^r_{\text{num}}(X)\) is the group of \({\mathbb Q}\)-linear algebraic cycles of codimension \(r\) on \(X\) modulo numerical equivalence. This conjecture can be reformulated into several other forms as in [3]. Then, relying on results from [1] \textit{U. Jannsen} [``Motives, numerical equivalence, and semi-simplicity'', Invent. Math. 107, No. 3, 447-452 (1992; Zbl 0762.14003)], [2] \textit{J. S. Milne} [``Motives over finite fields'', in: Motives, Proc. Symp. Pure Math. 55, Pt. 1, 401-459 (1994; Zbl 0811.14018)] and [3] \textit{J. Tate} [``Conjectures on algebraic cycles in \(\ell\)-adic cohomology'', in: Motives, Proc. Symp. Pure Math. 55, Pt. 1, 71-83 (1994; Zbl 0814.14009)], a remarkable theorem can be proved: For \(k={\mathbb F}_q\) the following statements are equivalent: (i) Tate's conjecture holds for all smooth projective \(X/k\); (ii) For simple motives \(M,M'\) in \({\mathcal M}\), one has \(M\cong M'\iff \pi_M=\pi_{M'}\); (iii) For a simple motive \(M\) in \({\mathcal M}\), \(M\cong 1\iff\pi_M=1\); (iv) For every \(M\) in \({\mathcal M}\), \({\mathbb Q}[\pi_M]\) is the center of \(\text{ End}(M)\). For the implication (iii)\(\Rightarrow\)(i) one decomposes \(X\) into the sum of simple motives and by [2] one can define étale cohomology groups for motives in \({\mathcal M}\) with geometric Frobenius action, and one may reduce to the case where \(X\) is a power of the Lefschetz motive. From now on, assume the validity of Tate's conjecture. Assuming furthermore that numerical and rational equivalence over finite fields agree, the foregoing results lead to the validity of Parshin's conjecture: For all smooth projective varieties \(X/{\mathbb F}_q\) and integers \(a>0\), \(K_a(X)_{\mathbb Q}=0\). The idea of the proof is again by decomposing \(X\) into simple motives and then to show that the Adams eigenspaces vanish. With Parshin's conjecture in hand, and using de Jong's result on alterations and a Gersten-Quillen spectral argument, one shows that, for any field \(k\) of characteristic \(p\), the rational \(K\)-groups \(K_a(k)_{\mathbb Q}\) vanish for \(a>\text{ trdeg} k/{\mathbb F}_p\), and \(K_a(k)_{\mathbb Q}=K_a(k)^{(a)}_{\mathbb Q}=K^M_a(k)_{\mathbb Q}\), where the superscript \(a\) denotes the corresponding Adams eigenspace and \(K^M\) is Milnor's \(K\)-theory. Several corollaries can be obtained, e.g. one shows that for \(k\) of characteristic \(p\) and assuming Parshin's conjecture, motivic cohomology \(H^i_{\mathcal M}(X,{\mathbb Q}(j))\) vanishes unless \(0\leq j\leq\text{ min}\{i,\dim X+\text{ trdeg} k\}\), in particular, motivic cohomology vanishes in negative degrees. As a matter of fact, various corollaries are avatars of the following one: Let \(X\) be a quasi-projective variety of dimension \(d\) over a field \(k\) of characteristic \(p\) with \(\text{trdeg} k/{\mathbb F}_p=r\). Then \[ K'_a(X)_{\mathbb Q}=\bigoplus^{\text{ min}(a+d,r+d)}_{j=a}K'_a(X)^{(j)}_{\mathbb Q}, \] where, as usual, \(K'\) denotes \(K\)-groups associated to coherent sheaves.