You are on page 1of 4

M ath. Res. Lett. 16 (2009), no.

00, 10001–10004

c International Press 2009

SMASH-NILPOTENT CYCLES ON ABELIAN 3-FOLDS Bruno Kahn and Ronnie Sebastian
Abstract. We show that homologically trivial algebraic cycles on a 3-dimensional abelian variety are smash-nilpotent.

Introduction Let X be a smooth projective variety over a ﬁeld k . An algebraic cycle Z on X (with rational coeﬃcients) is smash-nilpotent if there exists n > 0 such that Z n is rationally equivalent to 0 on X n . Smash-nilpotent cycles have the following properties: (1) The sum of two smash-nilpotent cycles is smash-nilpotent. (2) The subgroup of smash-nilpotent cycles forms an ideal under the intersection product as (x · y ) × (x · y ) · · · × (x · y ) = (x × x × · · · × x) · (y × y × · · · × y ). (3) On an abelian variety, the subgroup of smash-nilpotent cycles forms an ideal under the Pontryagin product as (x ∗ y ) × (x ∗ y ) × · · · × (x ∗ y ) = (x × x × · · · × x) ∗ (y × y × · · · × y ) where ∗ denotes the Pontryagin product. Voevodsky [11, Cor. 3.3] and Voisin [12, Lemma 2.3] proved that any cycle algebraically equivalent to 0 is smash-nilpotent. On the other hand, because of cohomology, any smash-nilpotent cycle is numerically equivalent to 0; Voevodsky conjectured that the converse is true [11, Conj. 4.2]. This conjecture is open in general. The main result of this note is: Theorem 1. Let A be an abelian variety of dimension ≤ 3. Any homologically trivial cycle on A is smash-nilpotent. In characteristic 0 we can improve “homologivally trivial” to “numerically trivial”, thanks to Lieberman’s theorem [7]. Nori’s results in [8] give an example of a group of smash-nilpotent cycles which is not ﬁnitely generated modulo algebraic equivalence. The proof of Theorem 1 actually gives the uniform bound 21 for the degree of smash-nilpotence on this group, see Remark 2. See Proposition 2 for partial results in dimension 4. 1. Beauville’s decomposition, motivically For any smooth projective variety X and any integer n ≥ 0, we write as in [1] n CHQ (X ) = CH n (X ) ⊗ Q, where CH n (X ) is the Chow group of cycles of codimension n on X modulo rational equivalence. Let A be an abelian variety of dimension g . For m ∈ Z, we denote by m the endomorphism of multiplication by m on A, viewed as an algebraic correspondence. In [1], Beauville introduces an eigenspace decomposition of the rational Chow groups
2000 Mathematics Subject Classiﬁcation. 14C15, 14K05. 10001

h(A)) (where L is the Lefschetz motive) we have n CH n (A)[r] := {x ∈ CHQ (A) | m ∗ x = mr x ∀m ∈ Z} = Hom(Ln . View β as a morphism n L → h(A) in the category of Chow motives. This shows that β factors through a morphism ˜ : Ln → hodd (A) β with hodd (A) = i odd hi (A). n Proof. Th.2]). where the πi are orthogonal idempotents and πi is characterised by πi m ∗ = mi πi for any m ∈ Z. For g = 3.10 p.10002 Bruno Kahn and Ronnie Sebastian of A for the actions of the operators m . For g ≤ 2. 5. −πi β = πi −1 ∗ β = (−1)i πi β hence πi β = 0 for i even. S 2g+1 (h1 (A)) = h2g+1 (A) = 0 by [9. the Griﬃths group of A is 0 and there is nothing to prove. the g endomorphism 1A ∈ End(h(A)) = CHQ (A × A) is given by the class of the diagonal ∆A . hr (A)). Prop. 188]. say. 2. hi (A) = (A. Remark 1. Here is an equivalent deﬁnition: in the category of Chow motives with rational coeﬃcients. . Thus. Any skew cycle on an abelian variety is smash-nilpotent. using the Fourier transform. we have n CH n (A)[r] = CH2 n−r (A). prop. the Griﬃths group of A is a quotient of CH 2 (A)[3] [1. Theorem].1]. 186]. under the isomorphism n CHQ (A) = Hom(Ln . Skew cycles on abelian varieties Let β ∈ Assume that −1 ∗ β = −β : we say that β is skew. We shall use his notation in §3. Prop. This implies that β is homologically equivalent to 0.1 p. and a direct summand of an odd tensor power of an oddly ﬁnite-dimensional motive is oddly ﬁnite dimensional by [6. We have the canonical Chow-K¨ unneth decomposition of Deninger-Murre 2g 1A = i=0 πi [4. (Indeed. This yields a canonical Chow-K¨ unneth decomposition of the Chow motive h(A) of A: 2g h(A) = i=0 hi (A). In Beauville’s notation. for all i: ∗ CHQ (A).-I.) Hence the conclusion follows from [6. πi ) (see [10. 5. We may assume β homogeneous. Kimura. β ∈ CHQ (A). 6. thus Theorem 1 follows from the more general Proposition 1. for any genus. 6]. Th. But Ln is evenly ﬁnite-dimensional and hodd (A) is oddly ﬁnite-dimensional in the sense of S. Then. 3. This applies notably to the Ceresa cycle [3].

Both authors would like to thank V. [1. 3a]. h3 (A)) verify z ⊗21 = 0. namely. with the notation of §1. 90 (1968). Math. Beauville Sur l’anneau de Chow d’une vari´ et´ e ab´ elienne. J. Some of these arguments also follow from a paper of Bloch and Srinivas [2]. 1.8]. 2 In the case of CH0 (A). 273 (1986). 117 (1983). If z ∈ Hom(Ln . Ann. Acknowledgements The second author would like to thank Prof. Sympos. Proc. g } and s < 0. Sci.SMASH-NILPOTENT CYCLES ON ABELIAN 3-FOLDS 10003 Remark 2. 422(1991). for i = 3 and if A is a 3-fold. J. the sum of the odd Betti numbers of A. 285–291.P. 5. Indian Acad. Math. 331 (2005). 99 (1989). Ceresa C is not algebraically equivalent to C − in its Jacobian. h4 (A)). Thus. 105 (1983). from where he learnt about the conjecture. g − 2. Murre for his course on Motives at TIFR in January 2008. 2 CH0 (A) = Hom(L2 . J. Srinivas for discussions leading to this result. References [1] A. in a sense. The second author is being funded by CSIR. Note that. 366–374. Soc. 4]. Jannsen Motivic sheaves and ﬁltrations on Chow groups. . J. Math. Murre Motivic decomposition of abelian shecmes and the Fourier transform. 55 (1). Amer. [6] S. hi (A)) for some odd i. Math. h4 (A)). Let A be an abelian variety and let A know. Bloch. Amer. of Math. [7] D. Kimura Chow groups are ﬁnite dimensional.-I. Math. 245–302. 647–651. Sci. using these results and Proposition 1 one can conclude smash nilpotence 2 3 for homologically trivial cycles which are not in CH0 (A) or CH2 (A). Remark 3. we ﬁnd that all z ∈ Hom(L. 201–219. Pure Math. We Proof. (1) CHs p g (2) CHp (A) and CHs (A) consist of cycles algebraically equivalent to 0 for all values of p and all values of s > 0. from [1]. g − 1. Prop. ˆ denote its dual abelian variety. the problem is whether there are any homologically trivial cycles: in view of the above expression. the following: p (A) = 0 for p ∈ {0. hodd (A)) verify z ⊗N +1 = 0 for a ﬁxed N . 191–196. The 4-dimensional case Proposition 2. Math. Deninger. I. are smashnilpotent. in Motives. 3.1 shows in fact that all z ∈ Hom(Ln . [4] C. If g = 4. Nori Cycles on the generic abelian threefold.P. [8] M. Prop. 1994. [3] G.. 1235–1253. Kimura’s proposition 6. V. [5. For g = 4. J. [1. this is conjecturally not the case. Prop. [5] U. [2] S. Math. homologically trivial cycles on A. then we may take for N the i-th Betti number of A. 173–201. 3 CH2 (A) = Hom(L3 . Ann. except perhaps those 2 3 which occur in parts CH0 (A) or CH2 (A) of the Beauville decomposition. reine angew. Srinivas Remarks on Correspondences and Algebraic Cycles. Amer. Ann. cf. Lieberman Numerical and homological equivalence of algebraic cycles on Hodge manifolds. Proc.

Voisin Remarks on zero-cycles of self-products of varieties. 75013 Paris. Sympos. Homi Bhabha Road. 179. Lect. 1994. Math. Appl. Amer.. 1996. [12] C. Math. Res.fr Tata Institute of Fundamental Research.tifr. Funct.res. [10] A. 163–187. rue du Chevaleret. Shermenev The motive of an abelian variety. Soc. Anal. 47–53. Internat.in . [11] V. Mumbai 400 005. India E-mail address : ronnie@math.10004 Bruno Kahn and Ronnie Sebastian [9] A. 265–285. in Moduli of vector bundles. Voevodsky A nilpotence theorem for cycles algebraically equivalent to 0. Pure Math. Math.jussieu. 55 (1). Notices 1995. 187–198. in Motives. Scholl Classical motives. ´matiques de Jussieu. 175–179. 8 (1974). Dekker. France Institut de Mathe E-mail address : kahn@math. M. Notes in Pure Appl. Proc.