A NOTE ON THE PICARD NUMBER OF SINGULAR FANO 3-FOLDS 3 1 GLORIADELLA NOCE 0 2 n Abstract. UsingaconstructionduetoC.Casagrandeandfurtherdevelopedbytheauthorin a [DN12], we prove that the Picard number of a non-smooth Fano 3-fold with isolated factorial J canonical singularities, is at most 6. 1 1 ] G Introduction A LetX beaFano3-fold. IfX issmooth, weknow fromtheclassifiction resultsin[MM81], that . h its Picard number ρX is at most 10. Moreover, if ρX ≥ 6, then X is isomorphic to a product t S ×P1, where S is a smooth Del Pezzo surface. a m If X is singular, bounds for ρ are known only in particular cases. If X is toric and has X [ canonical singularities, then ρX ≤ 5 ([Bat82] and [WW82]). If X has Gorenstein terminal 1 singularities, then ρX ≤ 10, because X has a smoothing which preserves ρX (see [Nam97, Thorem 11] and [JR11, Theorem 1.4]). If, instead, X has Gorenstein canonical singularities, it v 8 does not admit, in general, a smooth deformation (see [Pro05, Example 1.4] for an example). In 5 this setting, the following holds. 5 2 Theorem 0.1. [DN12, Theorem 1.3] Let X be a 3-dimensional Q-factorial Gorenstein Fano . 1 variety with isolated canonical singularities. Then ρX ≤ 10. 0 3 The proof of this theorem uses a construction introduced by C. Casagrande in [Cas12], and 1 relies on the result of [BCHM10] that Fano varieties are Mori dream spaces (see [HK00] for the v: definition). i In this paper, using the same construction, we show that the bound given by Theorem 0.1 X can be improved if X is actually singular and its singularities are also factorial. Our result is r a the following. Theorem 0.2. Let X be a non-smooth factorial Fano 3-fold with isolated canonical singularities. Then ρ ≤ 6. X In the first section of this paper, we recall some preliminary results from [DN12]; the second section contains the proof of Theorem 0.2 and an observation concerning the case ρ = 6. X Notation and terminology We work over the field of complex number. Let X be a normal variety. We call X Fano if −K has a multiple which is an ample Cartier X divisor. We denote by X the non-singular locus of X. We say that X is Q-factorial if every reg Weil divisor is Q-Cartier, i.e. it admits a multiple which is Cartier. We call X factorial if all its local rings are UFD; by [Har77, II, Proposition 6.11], this implies that every Weil divisor of This work has been partially supported by PRIN 2009 “Moduli, strutture geometriche e loro applicazioni”. 1 X is Cartier. We refer the reader to [KM98] for the definition and properties of terminal and canonical singularities. If X has canonical singularities, it is called Gorenstein if its canonical divisor K is a Cartier divisor. X WedenotewithN (X)(resp. N1(X))thevectorspaceofone-cycles(resp. Q-Cartierdivisors) 1 with real coefficients, modulo the relation of numerical equivalence. The dimension of these two real vector spaces is, by definition, the Picard number of X, and is denoted by ρ . We denote X by [C] (resp. [D]) the numerical equivalence class of a one-cycle (resp. a Q-Cartier divisor). Given [D] ∈ N1(X), we set D⊥ := {γ ∈ N (X)|D ·γ = 0}, where · denotes the intersection 1 product. We define NE(X) ⊂ N (X) as the convex cone generated by classes of effective curves 1 and NE(X) is its closure. An extremal ray R of X is a one-dimensional face of NE(X). We denote by Locus(R) the subset of X given by the union of curves whose class belongs to R. A contraction of X is a projective surjective morphism with connected fibers ϕ : X → Y onto a projective normal variety Y. It induces a linear map ϕ : N (X) → N (Y) given by the ∗ 1 1 push-forward of one-cycles. We set NE(ϕ) := NE(X)∩ker(ϕ ). We say that ϕ is K -negative ∗ X if K ·γ < 0 for every γ ∈ NE(ϕ). X Theexceptional locus ofϕisthelocuswhereϕisnotanisomorphism;wedenoteitbyExc(ϕ). We say that ϕ is of fiber type if dim(X) > dim(Y), otherwise ϕ is birational. We say that ϕ is elementary if dim(ker(ϕ )) = 1. In this case NE(ϕ) is an extremal ray of NE(X); we say that ∗ ϕ (or NE(ϕ)) is divisorial if Exc(ϕ) is a prime divisor of X and it is small if its codimension is greater than 1. An elementary contraction from a 3-fold X is called of type (2,1) if ϕ is K -negative, bira- X tional, dim(Exc(ϕ)) = 2 and dim(ϕ(Exc(ϕ))) = 1. IfD ⊂ X isaWeildivisorandi: D → X istheinclusionmap,wesetN (D,X) := i N (D) ⊆ 1 ∗ 1 N (X). 1 1. Preliminaries Inthefollowing statement, we collect someresults from [DN12]. For thereader’s convenience, we recall here the main steps of their proof. We refer the reader to [DN12, Theorem 2.2] for the properties of contractions of type (2,1) defined on mildly singular 3-folds. Lemma 1.1. [DN12,Theorem1.2anditsproof-Remark5.2] Let X be a Q-factorial Gorenstein Fano 3-fold with isolated canonical singularities. Suppose ρ ≥ 6. Then there exist morphisms X ψ :X → P1 and ξ :X → S, where S is a normal surface with ρ = ρ −1, and the morphism S X π := (ξ,ψ) : X → S ×P1 is finite. Moreover there exist extremal rays R ,...,R (m ≥ 3) in NE(X) such that: 0 m • each R is of type (2,1); i • NE(ψ) = R +···+R ; 0 m • for i =0,...,m, set E = LocusR and Q =NE(ξ). Then i i m ψ(Ei)= P1, N1(Ei,X) = RRi⊕RQ and Q ⊆ \Ei⊥; i=0 2 • ψ factors as X →σ X˜ → P1, where σ is birational, X˜ is a Fano 3-fold with canonical isolated singularities, NE(σ) = R + ··· + R , with m ≥ s ∈ {ρ − 2,ρ − 3} and 1 s X X σ(E ),...,σ(E )⊂ X˜ are pairwise disjoint. 1 s Proof. By [DN12, Remark 5.2], the assumption ρ ≥ 6 implies that all the assumptions of X [DN12, Theorem 1.2] are satisfied, from which the existence of the finite morphism π. The properties of its projections ψ and ξ follow by their construction, that we briefly recall. All the details can be found in the proof of [DN12, Theorem 1.2]. By [DN12, Proposition 3.5], there exists an extremal ray R ⊂ NE(X) of type (2,1). Set 0 E = Locus(R ); we have dimN (E ,X) = 2. As in [DN12, Lemma 3.1], we may find a Mori 0 0 1 0 program (1.1) X = X 9σ90K X 99K ··· 99K X σ9k9−K1 X −→ϕ Y 0 1 k−1 k whereX ,...,X areQ-factorial3-foldswithcanonicalsingularitiesand,foreachi =0,...,k−1, 1 k there esists a K -negative extremal ray Q ⊂ NE(X ) such that σ is either its contraction, if Xi i i i Q is divisorial, or its flip, if it is small. Moreover, if (E ) ⊂ X is the transform of E and i 0 i i 0 (E ) := E , then (E ) ·Q > 0. Finally, ϕ is a fiber type contraction to a Q-factorial normal 0 0 0 0 i i variety Y. Let us set {i ,...,i }:= {i ∈ {0,...,k−1}|codimN (D ,X ) = codimN (D ,X )−1}. 1 s 1 i+1 i+1 1 i i Then, by [DN12, Lemma 3.3], s ∈ {ρ −2,ρ −3} (in particular s ≥ 3); moreover, for every X X j ∈ {1,...,s}, Q is a divisorial ray, σ is a birational contraction of type (2,1) and, if E ⊂ X ij ij j is the transform of the exceptional divisors of the contraction σ as above, then E ,...,E are ij 1 s pairwise disjoint. Since s ≥ 3, [DN12, Proposition 3.5] assures that, for each j = 1,...,s, there exists an extremal ray R ⊂ NE(X) of type (2,1) such that E = Locus(R ). The divisor −K +E + j j j X 1 ···+E comes out to be nef, and its associated contraction σ :X → X˜ verifies s ker(σ ) = RR +···+RR and Exc(σ) = E ∪···∪E . ∗ 1 s 1 s It is thus possible to look at σ a part of a Mori program as in (1.1), and to find a fiber type contraction ϕ : X˜ → Y giving rise to a morphism ψ := ϕ◦σ : X → Y as in the statement. In particular, we have NE(ψ) = R +···+R , where m ≥ s and R ,...,R are extremal rays 0 m s+1 m of type (2,1). We notice that, since dim(X) = 3, we have Y ∼=P1 by [DN12, Remark 4.2]. The second projection ξ arises as the contraction associated to a certain nef divisor defined as a combination of the prime divisors E ,...,E constructed above (recall that E = LocusR 0 m i i for i = 0,...,m). It is an elementary contraction and the one-dimensional subspace generated by NE(ξ) belongs to N (E ,X) for every i = 0,...,m. (cid:3) 1 i 2. Theorem 0.2 Proof of Theorem 0.2. Let us prove that, if ρ ≥ 7, then the morphism π : X → S ×P1 given X by Lemma 1.1 is an isomorphism. This will give a contradiction with our assumptions on the singularities of X, since S ×P1 is smooth or has one-dimensional singular locus. We are in the setting of Lemma 1.1; let us keep its notations. By [AW97, Corollary 1.9 and Theorem 4.1(2)], the general fiber of ξ is a smooth rational curve, and the other fibers have 3 at most two irreducible components (that might coincide) whose whose reduced structures are isomorphic to P1. Our assumptions imply that S is factorial: if C ⊂ S is a Weil divisor, its counterimage D := ξ−1(C) ⊂ X is a Cartier divisor, because X is factorial. Moreover D · Q = 0 (where Q = NE(ξ)), because D is disjoint from the general fiber of ξ. Then D = ξ∗(C′) for a certain Cartier divisor C′ on S. But then C = C′ is Cartier. Fix i = 0,...,m; let ϕ : X → Y be the contraction of R and set G := ϕ (E ) ⊆ Y , i i i i i i i T := ξ(E ) ⊆ S: i i E i ϕi|~~}E}}i}}}}} AAAAAξAA|EA i G T . i i Notice that T ⊂ S is a curve. Indeed, by Lemma 1.1, E ·Q = 0, which implies that T ⊂ S is i i i a (Cartier) divisor and E = ξ∗(T ). i i Let f be the general fiber of ϕ . Since f is a smooth rational curve which dominates T , T i i i i i is a (possibly singular) rational curve. The same conclusion holds for G , which is dominated i by any smooth curve contained in a fiber of ξ over T . i We have −1= E ·f = ξ∗(T )·f = T2·deg(ξ ), i i i i i |fi from which −T2 = deg(ξ ) = 1. Then the general fiber g of ξ over T is a smooth rational i |fi i curve. Indeed, g has no embedded points, and if, by contradiction, the 1-cycle associated to g is of the type C +C , then g would intersect f in at least two (distinct or coincident) points. 1 2 i This is impossible because g is general and deg(ξ )= 1. |fi Then E is smooth along the general fibers of both ϕ and ξ; we deduce that E is smooth in i i i codimension one. Moreover E is a Cohen-Macaulay variety, because X is factorial. Then, by i Serre’s criterion, E is normal. Then the finite morphism (ξ ,ϕ ): E → T ×G , which has i |Ei i|Ei i i i degree one, factors through the normalization of the target: there is a commutative diagram E τ // P1×P1 i HHHHHHHHH## (cid:15)(cid:15) ν T ×G . i i Sinceτ isfiniteofdegreeone,byZariskiMainTheorem,itisanisomorphism. ThusE ∼= P1×P1, i and ξ : E → T ∼= P1 and ϕ : E → G ∼= P1 are the projections. In particular, since both |Ei i i i|Ei i i E and T are Cartier divisors, they are contained in the smooth loci of, respectively, X and S. i i We have (K −ξ∗(K ))·f = (K −ξ∗ (K ))·f = (ϕ∗ (K ))·f = 0. X S i Ei |Ei Ti i i|Ei Gi i LetF beageneralfiberofψ :X → P1. ThenF isasmoothDelPezzo surfaceand,byLemma 1.1, N (F) ⊆ R[f ]; thus K −ξ∗(K ) is numerically trivial in F. Moreover ζ := ξ :F → S 1 P i X S |F is a finite morphism of degree d := deg(π) and (2.1) K = (K ) = (ξ∗(K )) = ζ∗K ; F X |F S |F S 4 in particular ζ is unramified in the open subset ξ−1(S ), which contains E ∩ F for every reg i i = 0,...,m. SetF˜ := σ(F) ⊂ X˜, whereσ : X → X˜ is thebirational contraction given by Lemma1.1; then F˜ is again a smooth Del Pezzo surface and σ :F → F˜ is a contraction. For every i= 1,...,s, |F theintersection E ∩F istheunionof ddisjoint curves numerically equivalent tof ; inparticular i i σ realizes F astheblow-upofF˜ alongs·ddistinctpoints(wheres = ρ −ρ ). Then,recalling |F X X˜ that s ≥ ρ −3 and ρ ≥ 7, we get X X 9 ≥ ρ = ρ +s·d ≥ 1+4d, F F˜ and then d≤ 2. Moreover, if d= 2, then ρ = 9 and, by 2.1, F 1 = K2 = ζ∗(K )·K = 2(K )2, F S F S which is impossible becauseS is factorial and thus K2 is integral. Hence d= deg(ζ)= deg(π) = S 1 and the statement is proved. (cid:3) The case ρ = 6 is more complicated to analyze. Indeed, though Lemma 1.1 still holds in X that case, we are not able to conclude that π is an isomorphism and that, as a consequence, X is smooth. Proposition 2.1. Let X be a factorial Fano 3-fold with isolated canonical singularities and with ρ = 6. If X is not smooth, there exists a finite morphism of degree 2 X π :X → S ×P1, where S is a singular DelPezzo surface with factorial canonical singularities, ρ = 5, (K )2 = 1. S S Moreover the ramification locus of π contains a surface R which dominates S. Proof. We argue as in the proof of Theorem 0.2 and we use the same notations. Since X is not smooth, the degree of π must be 2. Exactly as in the above case, we have (2.2) K =(K ) = (ξ∗(K )) = (ξ∗K ) = ζ∗K , F X |F S |F S |F S and (2.3) ρ = 10−(K )2 = 10−2(K )2, F F S so that ρ needs to be even. Since ρ = 6, we have s ∈ {3,4}, and then F X 9 ≥ ρ =ρ +2s. F F˜ Thus the only possibility is that ρ = 2 and ρ = 8. By (2.3), we get (K )2 = 1. F˜ F S Let us call R the ramification divisor (possibly trivial) of π. Let C be the general fiber of ξ. Then C ∼= P1 and ψ : P1 → P1 is finite of degree 2. By Hurwitz’s formula we have R·C = 2, |C and hence R is not trivial and it dominates S. (cid:3) Acknowledgments. This paper is part of my PhD thesis; I am deeply grateful to my advisor Cinzia Casagrande for her constant guidance. 5 References [AW97] M.AndreattaandJ.A.Wi´sniewski,Aviewoncontractions ofhigherdimensionalvarieties,Algebraic Geometry-SantaCruz 1995, Proc. Symp.Pure Math., vol. 62, 1997, pp.153–183. [Bat82] V.V. Batyrev,Toric Fano threefolds (English traslation), Math. USSR-Izv.19 (1982), 13–25. [BCHM10] C. Birkar, P. Cascini, C. D. Hacon, and J. McKernan, Existence of minimal models for varieties of log general type, J. Amer.Math. Soc. 23 (2010), 405–468. [Cas12] C. Casagrande, On the Picard number of divisors in Fano manifolds, Ann.Sci. E´c. Norm. Sup´er.45 (2012), 363–403. [DN12] G.Della Noce, On the Picard number of singular Fano varieties, preprint arXiv:1202.1154 (2012), to appear in Internat.Math. Res. Notices. [Har77] R.Hartshorne, Algebraic Geometry, Springer, 1977. [HK00] Y.Hu and S. Keel, Mori dream spaces and GIT, Michigan Math. J. 48 (2000), 331–348. [JR11] P.JahnkeandI.Radloff,TerminalFanothreefoldsandtheirsmoothing,Math.Zeitschrift269(2011), 1129–1136. [KM98] J. Koll´ar and S. Mori, Birational Geometry of Algebraic Varieties, Cambridge Univ.Press, 1998. [MM81] S.MoriandS.Mukai,Classification on Fano 3-folds with b2≥2,Manuscr.Mah.36(1981),147–162, Erratum: 110 (2003), 407. [Nam97] Y.Namikawa, Smoothing Fano 3-folds, J. Alg. Geom. 6 (1997), 307–324. [Pro05] Y. G. Prokhorov, On the degree of Fano threefolds with canonical Gorenstein singularities (English translation), Sb.Math. 196 (2005), 77–114. [WW82] K. Watanabe and M. Watanabe, The classification of Fano 3-folds with torus embeddings, Tokio J. Math. 5 (1982), 37–48. Dipartimento di Matematica, Universita` di Pavia, via Ferrata, 1 27100 Pavia - Italy E-mail address: [email protected] 6