Higher \(l\)-adic Abel-Jacobi mappings and filtrations on Chow groups (Q1896008)

In recent years, several people (Beilinson, Bloch, Murre, Nori, S. Saito, H. Saito) have made conjectures asserting the existence of certain filtrations on the Chow groups of any smooth projective variety. These conjectures would imply that the Chow groups have much more structure than has been hitherto uncovered. In this paper we define and study a filtration arising from the cycle map into \(l\)-adic étale cohomology. It is somewhat arithmetic in nature because it is necessary to first define it over a field which is finitely generated over its prime subfield. For varieties over larger fields, it may be defined by passing to the limit over finitely generated subfields. Others have studied this filtration, in particular Beauville (unpublished) and \textit{U. Jannsen} [in: Motives. Proc. Summer Res. Conf., Univ. Washington 1991, Proc. Symp. Pure Math. 55, part 1, 245-302 (1994; Zbl 0811.14004)]. With their permission, I have included some of their results in this paper. Each of the filtrations defined so far has its advantages and disadvantages. The filtration in the paper under review has the advantage that it may be ``computed'' in some cases, and the powerful tools of arithmetic algebraic geometry may be brought to bear on it. However, it has two basic disadvantages. First, it is not at all obvious that it is independent of \(l\). Second, suppose \(X\) is defined over a field \(k\) which is finitely generated over its prime subfield, and \(l\) is a prime number different from the characteristic of \(k\). \textit{U. Jannsen} [Math. Ann. 280, No. 2, 207-245 (1988; Zbl 0649.14011)] defined a cycle map into continuous étale cohomology (see \S 1): \[ c_n:CH^n(X) \otimes\mathbb{Q} \to H^{2n} (X, \mathbb{Q}_t(n)). \] The following is a more general version of a conjecture first made by Soulé over finite fields and by Jannsen and the author over number fields. Conjecture. \(c_n\) is injective. This is known when \(n=1\) (a consequence of the finite generation of \(CH^1(X))\), but except for special classes of varieties, we have little information for \(n\geq 2\). The Hochschild-Serre spectral sequence \[ H^r\biggl( k,H^q\bigl(\overline X,\mathbb{Q}_l(n) \bigr) \biggr) \Rightarrow H^{r+s}\bigl(X, \mathbb{Q}_l(n)\bigr) \] defines a filtration on \(H^{2n}(X,\mathbb{Q}_l(n))\), which we pull back to \(CH^n(X)\otimes\mathbb{Q}\) via the cycle map \(c_n\). We call this the \(l\)-adic filtration on \(CH^n(X)\otimes\mathbb{Q}\). By definition, the kernel of the cycle map is contained in all the steps of the filtration. For this reason, until we have more information on this kernel, we can only get results about the filtration modulo the kernel. We define another filtration \({\mathcal F}^\bullet CH^n(X)\otimes\mathbb{Q}\) on the image of the cycle map by taking the intersection with the filtration obtained from the Hochschild-Serre spectral sequence. It is really this filtration about which we can say something. In \S 2 of this paper we refine the cycle map \(c_n\) to get higher \(l\)-adic Abel-Jacobi maps from certain parts of the Chow group to Galois cohomology groups. For example, suppose that \(n=\dim X\), and let \(A_0(X)\) denote the group of zero-cycles of degree zero modulo rational equivalence. Let \(\text{Alb}(X)\) be the group of \(k\)-points of the Albanese variety of \(X\) and \(T(X)\) the kernel of the Albanese map: \(A_0(X)\to\text{Alb}(X)\). Then we define a map: \[ d^2_n:T(X) \otimes\mathbb{Q} \to H^2\biggl(k,H^{2n-2} \bigl(\overline X, \mathbb{Q}_l(n)\bigr)\biggr). \] For the results of this paper we will need to assume that if \(X\) is a smooth projective variety over a local field \(k\) with ring of integers \({\mathcal O}\), then \(X\) has a regular proper model over \({\mathcal O}\). Assuming this, the main result of this paper is the following theorem. Theorem 0.1. Let \(k\) be a function field in one variable over a finite field of characteristic \(p\neq 1\), and \(X\) a smooth projective variety of dimension \(n\) over \(k\). Then the map \(d^2_n\) is zero. Since a global field is of cohomological dimension 2 for \(\mathbb{Q}_l\)-cohomology, we have the following. Corollary. For \(X\) as above of dimension \(n\), we have \({\mathcal F}^iCH^n(X)\otimes\mathbb{Q}=0\) for all \(i\geq 2\). Remark: Beilinson and Bloch made the conjecture that if \(k\) is a global field and \(X\) is a smooth projective variety over \(k\), then the Albanese kernel \(T(X)\) is a torsion group. Thus the theorem provides ``evidence'' for this conjecture for \(k\) a global field of positive characteristic by showing that the cycle classes of elements of \(T(X)\) in \(H^{2n}(X,\mathbb{Q}_l(n))\) are trivial.