Ordinary Flops
CHIN-LUNG WANG (NCU and NCTS)
(Joint work with H.-W. Lin) September 6, 2004
CONTENTS
1. Ordinary flips/flops and local models 2. Chow motives and Poincar´e pairing
3. Ordinary/quantum product for simple flops 4. Deformations and degenerations
5. Slicing — Mukai flops
1.1 Ordinary (r, r0) Flips
X smooth projective,
ψ : X → ¯X log-extremal small contraction, R = R+[C], the log-extremal ray,
Z ⊂ X and S ⊂ ¯X: ψ exceptional sets, ψ = ψ|¯ Z : Z → S, Zs := ¯ψ−1(s).
ψ is a (r, r0) flipping contraction if
(i) ¯ψ : Z = PS(F ) → S for some rank r + 1 vector bundle F over a smooth base S,
(ii) NZ/X|Zs =∼ OPr(−1)⊕(r0+1).
Fact: Let ¯ψ : Z = PS(F ) → S and V → Z a vector bundle such that V |Zs is trivial ∀s ∈ S.
Then V ∼= ¯ψ∗F0 for some vector bundle F0.
Apply to V = OP
S(F )(1) ⊗ NZ/X, we get NZ/X =∼ OP
S(F )(−1) ⊗ ¯ψ∗F0. Since OP
Z(L⊗F )(−1) = ¯φ∗L ⊗ OP
Z(F )(−1) for L ∈ Pic(Z), on the blow-up φ : Y = BlZX → X,
E = PZ(NZ/X) ∼= PZ( ¯ψ∗F0) = ¯ψ∗PS(F0) = PS(F )×SPS(F0), NE/Y = OPZ(NZ/X)(−1) = ¯φ∗OPS(F )(−1) ⊗ ¯φ0∗OPS(F0)(−1).
Basic diagram: g = ψ ◦ φ : Y → ¯X, ¯g = g|E,
E = PS(F ) ×S PS(F0) ⊂ Y
φ¯
ssffffffffffffffffffffff
φ¯0WWWWWWWW ++W
WW WW WW WW WW W
¯ g ⊂g
²²
Z = PS(F ) ⊂ X
ψ¯
++XX XX XX XX XX XX XX XX XX XX XX XX
X Z0 = PS(F0)
ψ¯0
ssggggggggggggggggggggggg
S ⊂ ¯X
The pair (F, F0) is unique up to a twisting:
(F, F0) ∼ (F ⊗ L, F0 ⊗ L∗) for all L ∈ Pic(S).
Theorem 1 Ordinary (r, r0)-flip f : X 99K X0 exists. Moreover, Y = ¯Γf = X ×X¯ X0 ⊂ X × X0.
Proof. From 0 → TC → TX|C → NC/X → 0 and NC/X =∼ OC(1)⊕(r−1) ⊕ OC(−1)⊕(r0+1),
KX.C = 2g(C) − 2 − ((r − 1) − (r0 + 1)) = r0 − r.
Pick a line CY ∈ ¯φ−1(pt), φ(CY ) = C. Then
KY.CY = (φ∗KX + r0E).CY = (r0 − r) − r0 = −r < 0.
Let H be very ample on X and L a supporting divisor of C. Let c = H.C. For large k,
kφ∗L − (φ∗H + cE)
is big and nef, and vanishes precisely on [CY ].
Thus CY is a KY -negative extremal ray and
0 0
1.2 Analytic Local Models
F → S, F0 → S: holomorphic vector bundles, ψ : Z =¯ PS(F ) → S, ¯ψ0 : Z0 = PS(F0) → S,
E = Z ×S Z0 with projections ¯φ and ¯φ0.
Y = total space of N := ¯φ∗OZ(−1)⊗¯φ0∗OZ0(−1), E = zero section, NE/Y = N .
We have analytic contraction diagram
E
φ¯
}}zzzzzzzz EEEE
φ¯0
""E EE
j //
Y
φ
||yyyyyyyy φ0
""F FF FF FF F
Z
ψC¯CCCCC!!C C
i //
X
ψDDDDDDDZ!!D 0
zzzz
ψ¯0
}}zzzz
i0 //
X0
ψ0
||yyyyyyyyy
S j0 //X
X and X0 are smooth, S = Sing( ¯X).
X = total space of NZ/X = OPS(F )(−1) ⊗ ¯ψ∗F0, X0 = total space of NZ0/X0 = OPS(F0)(−1) ⊗ ¯ψ0∗F .
Again, (F, F0) and (F1, F10) define isomorphic analytic local model if and only if (F1, F10) = (F ⊗ L, F0 ⊗ L∗) for some L ∈ Pic(S).
An ordinary (r, r)-flip is called an ordinary Pr flop or simply a Pr flop.
2.1 Canonical Correspondences
M: category of motives. Objects = smooth varieties, morphisms = correspondences
HomM( ˆX1, ˆX2) = A∗(X1 × X2)
under composition law: for U ∈ A∗(X1 × X2), V ∈ A∗(X2 × X3), pij : X1 × X2 × X3 → Xi × Xj,
V ◦ U = p13∗(p∗12U.p∗23V ),
[U ] : A∗(X1) → A∗(X2); a 7→ p2∗(U.p∗1a).
Induced map on T -valued points Hom( ˆT , ˆXi):
UT : A∗(T × X1) −→ AU ◦ ∗(T × X2).
Identity Principle: Let U, V ∈ Hom( ˆX, ˆX0).
Then U = V if and only if UT = VT for all T . (UX ◦ ∆X = VX ◦ ∆X0 implies U = V .)
Theorem 2 For ordinary flops f : X 99K X0, the graph closure F := ¯Γf induces ˆX ∼= ˆX0 via F∗ ◦ F = ∆X and F ◦ F∗ = ∆X0.
Proof. For any T , idT × f : T × X 99K T × X0 is also an ordinary flop. By the identity principle we only need to show F∗F = id on A∗(X).
FW = p0∗(¯Γf.p∗W ) = φ0∗φ∗W.
φ∗W = ˜W + j∗³c(E).¯φ∗s(W ∩ Z, W )´
dim W, where 0 → NE/Y → φ∗NZ/X → E → 0 and s(W ∩ Z, W ) is the relative Segre class.
Observation: the error term is lying over W ∩Z.
Pr × Pr //E2r+s //
²²
Y 2r+s+1
Pr //Zr+s
²²Ss
Let W ∈ Ak(X). By Chow’s moving lemma we may assume that W intersects Z transversally.
dim W ∩ Z := ` = k + (r + s) − (2r + s + 1) = k − r − 1. Since dim φ−1(W ∩ Z) = ` + r < k, we get φ∗W = ˜W and φ−1(W ) ∩ E = φ−1(W ∩ Z).
Hence FW = W0, the proper transform of W in X0. (W0 may not be transversal to Z0.)
Let B be an irreducible component of W ∩ Z and ¯B = ¯ψ(B) ⊂ S with dimension `B ≤ `.
Notice that W0∩Z0 has irreducible components { ¯ψ0−1( ¯B)}B. Let φ0∗W0 = ˜W + PB EB.
EB ⊂ ¯φ0−1ψ¯−1( ¯B), a Pr × Pr bundle over ¯B.
For the generic point s ∈ ¯B, we thus have dim EB,s ≥ k − `B = r + 1 + (` − `B) > r = 122r.
In particular, EB,s contains positive dimensional fibers of φ and φ0 and φ∗(EB) = 0. So F∗FW = W . The proof is completed. ¤
2.2 The Poincar´e Pairing
Corollary 3 Let f : X 99K X0 be an ordinary flop. If dim α + dim β = dim X, then
(Fα.Fβ) = (α.β).
That is, F is orthogonal with respect to (−.−).
Proof. α.β = φ∗α.φ∗β = (φ0∗Fα + ξ).φ∗β = (φ0∗Fα).φ∗β = Fα.(φ∗0 φ∗β) = Fα.Fβ. ¤ Remark: F−1 = F∗ both in the sense of corre-
3.1 Triple Product for Simple Flops
f : X 99K X0 a simple Pr flop, S = pt, h = hyperplane class of Z = Pr,
h0 = hyperplane class of Z0,
x = [h × Pr], y = [Pr × h0] in E = Pr × Pr.
φ∗[hs] = xsyr − xs+1yr−1 + · · · + (−1)r−sxrys, F[hs] = (−1)r−s[h0s],
φ0∗α0 = φ∗α+(α.hr−i) xi + (−1)i−1yi
x + y , α ∈ Ai(X).
Theorem 4 For simple Pr-flops, α ∈ Ai(X), β ∈ Aj(X), γ ∈ Ak(X) with i ≤ j ≤ k ≤ r, i + j + k = dim X = 2r + 1,
Fα.Fβ.Fγ = α.β.γ + (−1)r(α.hr−i)(β.hr−j)(γ.hr−k).
Example: r = 2, dim X = 5, (i, j, k) = (1, 2, 2):
T α.T β.T γ = α0.β0.γ0 = φ0∗α0.φ0∗γ0.φ∗γ0
= (φ∗α + (α.h)E)(φ∗β + (β.Z)(x − y))(φ∗γ + (γ.Z)(x − y))
= α.β.γ + (β.Z)(γ.Z)φ∗α.(x − y)2
+ (α.h)(γ.Z)φ∗β.E.(x − y) + (α.h)(β.Z)φ∗γ.E.(x − y) + (α.h)(β.Z)(γ.Z)E.(x − y)2
= α.β.γ + (α.h)(β.Z)(γ.Z).
3.2 Quantum Corrections (Outline)
The three point functions hα, β, γi = X
d∈A1(X) hα, β, γi0,3,d
= α.β.γ + X
k∈N hα, β, γi0,3,k[C] qk[C]
+ X
d6=k[C] hα, β, γi0,3,d qd
and (−, −) determine the quantum product.
The difference of α.β.γ is already determined.
Deformations to the normal cone: X = X ×P1, Φ : M → X be the blowing-up along Z × {∞}.
Mt = X for all t 6= ∞ and M∼ ∞ = Y ∪ ˜E where E =˜ PS(NZ/X ⊕ O). Y ∩ ˜E = E = PS(NZ/X) is the infinity part of ˜E. Similarly Φ0 : M0 → X0 = X0×P1 and M∞0 = Y 0∪ ˜E0. Y = Y 0 and E = E0.
When S = pt, ˜E ∼= ˜E0. J. Li’s degeneration formula (A. Li and Y. Ruan) implies the equiv- alence of hα, β, γi0,3,d with d 6= k[C].
For simple P1-flops, the second term gives
X
k(α.k[C])(β.k[C])(γ.k[C])DI0,0,k[C]Eqk[C].
= (α.C)(β.C)(γ.C) q[C]
1 − q[C]
by the multiple cover formula (Voisin).
For simple P2-flops of type (1, 2, 2), hα, β, γi0,3,k[C] = k(α.C)(β.Z)(γ.Z)
×
Z
M0,2(P2,k)c3(k−1)(R1π∗e∗3O(−1)⊕3), with e3 : M0,3(P2, k) → X and π : M0,3(P2, k) → M0,2(P2, k). (Work in progress.)
3.3 Some Explicit Formulae
For Pr-flop with non-trivial base S, α ∈ A∗(Z) has the form α = Pξiψ¯∗ai; ξ = c1(OP(F )(−1)), ai ∈ A∗(S). E = ¯φ∗OP(F )(−1) ⊗ ¯φ0∗QF0.
Fα = XF(ξi). ¯ψ0∗ai = X φ¯0∗(cr(E).¯φ∗ξi). ¯ψ0∗ai.
Fξ1 = (−1)r−1(ξ0 − ¯ψ∗[c1(F ) + c1(F0)]).
Fξ2 = (−1)r−2(ξ02 − ¯ψ∗[(c1 + c01).ξ0 + (c21 + c1c01 − c2 + c02)]).
Fξ3 = (−1)r−3(ξ03 − ¯ψ0∗[(c1 + c01)ξ02 + (c21 + c1c01 − c2 + c02)ξ0 + (c31 − 2c1c2 − c2c01 + c21c01 + c1c02 + c3)]).
4.1 Deformations
Theorem 5 Ordinary flips deform in families:
let f : X 99K X0 be an (r, r0) flip with base S and X → ∆ be a smooth family with X0 = X. Then there is a smooth family X0 → ∆ and a ∆- birational map F : X 99K X0 such that F0 = f . Moreover, F is also an (r, r0) flip, with base S → ∆ an one parameter deformations of S.
Key: the ray [C] is stable in deformations.
Idea. HilbC/X is a G(2, r + 1) bundle over S.
NC/X =∼ O(1)⊕(r−1) ⊕ O(−1)⊕(r0+1) ⊕ Os+1. H1(C,O(k)) = 0 for all k ≥ −1 implies that HilbC/X is smooth at [C] for all C ⊂ Z and the natural map π : HilbC/X → ∆ is a smooth fibration with special fiber HilbC/X. By the stability of Grassmannian bundles we obtain Z → S → ∆. The supporting line bundles L for C on X is the unique extension of the sup- porting line bundle L for C on X. ¤
4.2 Degenerations
Fact. Every three dimensional smooth flop is the limit of composite of P1 flops.
Question: What is the closure of composite of general ordinary flops?
5.1 Generalized Mukai Flops
ψ : (X, Z) → ( ¯X, S) with NZ/X = TZ/S∗ ⊗ ¯ψ∗L, L ∈ Pic(S). Will construct the local model as a section of ordinary flops with F0 = F∗ ⊗ L.
E = PS(F ) ×S PS(F0) ⊂ Y
Φ
ssgggggggggggggggggggggg Φ0
++XX XX XX XX XX XX XX XX XX XX XX
g
²²
Z = PS(F ) ⊂ X
ΨXXXXXXXXXXXXXXX ++X XX
XX XX XX
X Z0 = PS(F0) ⊂ X0
Ψ0
ssffffffffffffffffffffffffff
S ⊂ ¯X
Suppose ∃ bi-linear map F ×S F0 → ηS to a line bundle ηS over S. OP(F )(−1) → ¯ψ∗F pulls back to ¯φ∗OP (−1) → ¯g∗F , hence a linear map
φ¯∗OZ(−1) ⊗E φ¯0∗OZ0(−1) → ¯g∗(F ⊗S F0) → ¯g∗ηS. Y := inverse image of the zero section of ¯g∗ηS in Y. X = Φ(Y ) ⊃ Z, X0 = Φ0(Y ) ⊃ Z0, ¯X = g(Y ) ⊃ S with restriction maps φ, φ0, ψ, ψ0. By tensoring the Euler sequence
0 → OZ(−1) → ¯ψ∗F → Q → 0
with S∗ = OZ(1) and notice that S∗⊗Q ∼= TZ/S, we get by dualization
0 → TZ/S∗ → OZ(−1) ⊗ ¯ψ∗F∗ → OZ → 0.
The inclusion maps Z ,→ X ,→ X leads to 0 → NZ/X → NZ/X → NX/X|Z → 0.
NX/X|Z = O(X)|Z = ¯ψ∗O( ¯X)|S. Denote O( ¯X)|S by L. Recall NZ/X =∼ OP
S(F )(−1) ⊗ ¯ψ∗F0. By tensoring with ¯ψ∗L∗, we get
0 → NZ/X ⊗ ¯ψ∗L∗ → OPS(F )(−1) ⊗ ¯ψ∗(F0 ⊗ L∗) → OZ → 0.
So F0 = F∗⊗L if and only if NZ/X = T∼ Z/S∗ ⊗ ¯ψ∗L.
5.2 Mukai Flops as Limits of Isomorphisms
For Mukai flops, L ∼= OS, F0 = F∗ with duality pairing F ×S F∗ → OS. Consider π : Y → C via
Y → ¯g∗OS = OE =∼ E × C−→π2 C.
We get a fibration with Yt := π−1(t), being smooth for t 6= 0 and Y0 = Y ∪ E. E = Y ∩ E restricts to the degree (1, 1) hypersurface over each fiber along E → S. Let Xt, X0t and ¯Xt be the proper transforms of Yt in X, X0 and ¯X.
For t 6= 0, all maps in the diagram
Yt
}}{{{{{{{{
C!!C CC CC CC
Xt
BÃÃB BB BB
BB X0t
~~||||||||
X¯t
are all isomorphisms. For t = 0 this is the Mukai flop. Thus Mukai flops are limits of iso- morphisms. They preserve all interesting in- variants like diffeomorphism type, Hodge type (Chow motive via [Y ]+[E]) and quantum rings etc. In fact all quantum corrections are zero.