Towards A + B Theory in Conifold Transitions for Calabi–Yau Threefolds
Chin-Lung Wang National Taiwan University
(Joint work with Yuan-Pin Lee and Hui-Wen Lin)
Second Annual Meeting of ICCM Taipei, Taiwan
December 27, 2018
Contents
1 Calabi–Yau 3-folds 2 A model and B model 3 Mirror, flops, and transitions
4 An observation from ordinary k-fold singularity 5 Statements for conifold transitions
6 Linked GW invariants for A model 7 Linked B model via VMHS
I. Calabi–Yau 3-folds
I A projective manifold X/C is Calabi–Yau if π1(X)is finite and KX=0 (or c1(X) =0).
I Yau’s solution to the Calabi conjecture=⇒for any cpt K¨ahler X with c1(X)R=0,∃finite cover eX→X:
Xe =A×B×C.
A∼=Cg/Λ (flat), B is hyperk¨ahler (SU(m)), and C is CY (SU(n)). Also π1(B) =π1(C) =0,C is projective.
I The first new case appears in dim=3. We have
h1(O) =h2(O) =0. WLOG we assume that π1(X) =0.
I Question:classification of CY 3-folds?
I What is the global structure (symmetries?) ofMCY3?
I Examples.Adjunction formula for hypersurfaces X⊂Y:
KX= (KY+X)|X =0⇐⇒X is anti-canonical in Fano Y.
I X= (n+1) ⊂Pn. E.g. the Fermat hypersurfaces xn0+1+ · · · +xnn+1 =0
is a CY(n−1)-fold. E.g. quintic 3-folds.
I X= (~d1, . . . ,~d
k) ⊂∏mi=1Pni with~d
j = (dji)mi=1and
∑
k j=1dji=ni+1, 1≤i≤m.
This is a CICY of dimension D=∑ ni−m.
I Let NDbe the numbers of them. ThenN3=7890.
I Toric CY.A lattice polytope4 ⊂MR, M∼=Zn+1is reflexive if 0∈int4and its polar (dual) polytope
4◦ := {w∈N :=M∨ | hw, vi ≥ −1, ∀v∈ 4}
is also a lattice polytope, in NR.
I Number of them [Kruezer–Skarke, 2000]:
N1=16, N2 =4319, N3 =473800776, . . . .
I For a reflexive pair(4,4◦), the toric variety P4 :=Proj(M
k≥0Ck4∩M) is Fano withH0(K−P1
4) =Lv∈4∩MC tv; similarly for P4◦.
I For a general section f , Xf := {f =0}is a CY n-fold.
II. Classical A model and B model
I The Hodge numbers of a CY 3-fold X are 1
0 0
0 h22 0
1 h21 h12 1
0 h11 0
0 0
1
I h11 =h1(X,ΩX) =h2parametrizes K¨ahler classes.
I A(X) =QH(X)is the g=0 Gromov–Witten theory in ω=B+iH ∈KXC=H2(X,R) ⊕√
−1Amp(X).
I h21 =h1(X, TX)parametrizes complex deformations.
I B(X) = (H3,∇GM)is the VHS on the complex moduliMX
under the Gauss–Manin connection with lattice H3(X,Z).
I A model. For a CY 3-fold X, let β∈ H2(X,Z),
M(X, β) = {h : C→X stable|C is nodal, pa(C) =0, h∗[C] =β}/∼. I Virtual dim=0: the essential genus 0 GW invariants are
nXβ = h−iXβ = Z
[M(X,β)]virt1∈Q.
I Toric example. Let Xf ⊂P4with f ∈H0(K−P41).
I A(Xf)is determined byC×-localizationdata [LLY, G 1999]:
IX(q•, z−1) =
∑
β∈H2(Xf,Z)
qβ ∏Km=1−1.β(K−1P4+mz)
∏ρ∈Σ1∏Dm=1ρ.β(Dρ+mz) .
I Σ is the normal fan of P4and Dρis the torus invariant divisor corresponding to the one-edge ρ∈Σ1,qβ =e2πi(β.ω).
I B model. For the CY family π :X →S :=MX, H3 :=R3π∗C⊗OS→S,
Fp =π∗ΩpX /S,Hpq=Fp∩Fq,Ω∈ Γ(S, F3). Then
∇GMFp,→Fp−1⊗Ω1S, h ∇GM∂/∂x
jΩihj=211 = H21.
I Periods. Let δm ∈H3(X)be a basis with dual δ∗m∈ H3(X). For η ∈Γ(S,H3), since∇GMδm∗ =0, we have
∇GM∂/∂x
jη=
∑
m
δm∗ ∂
∂xj Z
δm
η, j∈ [1, h21]. GM⇐⇒Picard–Fuchs equations of period integralsR
δmΩ.
I Toric example: B(Xf)is determined by the GKZ* system:
(1) symmetry operators;
(2) for`a relation ofmi ∈ 4 ∩Mwith∑`i=0,
` :=
∏
`i>0∂i`i−∏
`i<0∂−`i i.III. Mirror, flops, and transitions
I Mirror symmetry.
I Topological MS:(Y, Y◦)is a mirror pair of CY 3-folds if h21(Y) =h11(Y◦), h11(Y) =h21(Y◦).
I Classical MS, or A↔B MS:B(Y) ∼=A(Y◦),A(Y) ∼=B(Y◦).
I Toric Example: Consider 2 families of CY 3-folds Xf ⊂P4, Xg◦ ⊂P4◦.
I Topological MS holds [Batyrev ’94].
I A↔B MS holds for “many cases”.
I Observation: Σ1=rays from 0 to Vert(4◦).
I [HLY 1998]∃max-deg-point (⇒mirror transform).
I Flops.A D-flop between CY 3-folds is a birational diagram
` ⊂Y f //
ψ
""
Y0 ⊃ `0
ψ0
{{X
where ψ is D-negative (log-extremal) and ψ0is D0-positive.
I [Koll´ar, Kawamata 1988] Birational CY 3-folds are connected by flops.3D flops are classified.
I [Koll´ar–Mori 1992] Birational CY 3-folds Y and Y0 have MX∼=MX0 =⇒B(Y) ∼=B(Y0)
since flops can be performed in flat families.
I [Li–Ruan 2000]A(Y) ∼=A(Y0)under qβ 7→qf∗β(` 7→ −`0).
I Transitions.
I Geometric transition X%Y (or Y&X) of CY 3-folds:
Y
ψ
KY=ψ∗KX,
X //X NF3∞ =0.
I X%Y is a conifold transition if Xsinghas only ODPs (X, pi):= {x21+x22+x23+x24 =0}.
I Q1[Reid 1987] CanALL CY 3-foldsbe connected through (possibly non-projective) conifold transitions?
I Q2[W 2009] Does(A(X), B(X))determines(A(Y), B(Y)) and vice versa? NoticeA(X) <A(Y)andB(X) >B(Y).
IV. An observation from ordinary k-fold singularity
A + B Model in Quantum Geometry
Oct. 12, 2009 NTU Math
Dragon
LOCAL EXAMPLES: Consider the dim k hyper-surface X 0 C k+1 : x 0 k + x 1 k + … + x k k = 0
with p = 0 X 0 being an ordinary k-fold singularity. The blow-up f: X = Bl p (X 0 ) → X 0 is crepant with exceptional divisor
E = (k) P k , N E/X = O(-1)| E .
The local structure of E X, namely the germ (E, X) is equivalent to P k “cut out” by the rank 2 vector bundle:
V k = O(k)⊕O(-1) → P k .
X 0 can be smoothed into a flat family M → with general smooth
fiber X’ = M t . The semi-stable reduction : W → is used to compare
X and X’ since W t = X’ and W 0 = X∪E’ for some Fano E’.
Quantum Transition from A to B:
The Gromov-Witten extremal function f(a) = dN <a> dL q dL attached to the extremal ray L NE(X) can be calculated, using the quantum Serre duality principle, by the bundle
V k + = O(k)⊕O(1) → P k .
This is in turn reduced to O(k) → P k-1 , the Calabi-Yau CY k ! Where is the Picard-Fuchs operator P k for f(a)?
Since dim CY k = k – 2, we must have deg P = k – 2. But dim X’ =
k. It must be the case that there is a “sub-VHS of R k * C of weight
k – 2” which starts at Ω H n-1,1 = H 1 (X’, T). Let Γ be the vanishing
cycle along , then P k is the Picard-Fuchs op for
ΓΩ .
V. Statements for conifold transitions
Let X%Y be a projective conifold transition of CY 3-folds through X with k ODPs p1, . . . , pk, π :X →∆, ψ : Y→X:
Ci⊂Y
ψ
NCi/Y=OP1(−1)⊕2
NSi/X =T∗S3 Si⊂X π // pi ∈X
Let µ :=h2,1(X) −h2,1(Y) >0 and ρ :=h1,1(Y) −h1,1(X) >0.
χ(X) −kχ(S3) =χ(Y) −kχ(S2) =⇒µ+ρ=k.
Hence there arenon-trivial relationsbetween the “vanishing cycles”:
A= (aij) ∈Mk×µ,
∑
ki=1aij[Ci] =0, B= (bij) ∈Mk×ρ,∑
ki=1bij[Si] =0.Let 0→VZ,→H3(X,Z) →H3(X,¯ Z) →0 and V :=CZ⊗C.
Theorem (Basic exact sequence)
We have an exact sequence ofweight two pure Hodge structures:
0→H2(Y)/H2(X)−→B Ck A−→t V→0.
Since ψ : Y→X deforms in families, this identifiesMYas a codimenison µ boundary strata inMXand locallyMX ∼=∆µ×MY. Write V=ChΓ1, . . . ,Γµiin terms of a basisΓj’s. Then the α-periods
rj= Z
Γj
Ω, 1≤j≤µ
form the degeneration coordinates around[X]. The discriminant loci of MXis described by acentral hyperplane arrangementDB=Ski=1Di: Proposition (Friedman 1986)
Let wi=ai1r1+ · · · +aiµrµ, then the divisor Di:= {wi=0} ⊂MXis the loci where the sphere Sishrinks to an ODP pi.
I The β-periods in transversal directions are given by a function u:
up=∂pu= Z
βp
Ω
I The Bryant–Griffiths–Yukawa couplings extend over DBand
upmn:=∂3pmnu=O(1) +
∑
k i=11 2π√
−1
aipaimain wi =
Z
∂p∂m∂nΩ∧Ω
for 1≤p, m, n≤µ. It is holomorphic outside this index range.
I Let y=∑ki=1yiei∈Ck, with ei’s being the dual basis on(Ck)∨. Thetrivial logarithmic connectiononCk⊕ (Ck)∨−→Ckis
∇k=d+1 z
∑
ki=1dyi
yi ⊗ (ei⊗e∗i).
Theorem (Local invariance: Exc(A) +Exc(B) =trivial)
(1) ∇krestricts to the logarithmic part of∇GMon V∗.
(2) ∇krestricts to the logarithmic part of∇Dubrovinon H2(Y)/H2(X).
Theorem (LinkedA +B theory)
Let[X]be a nearby point of[X]inMX,
(1) A(X)is a sub-theory ofA(Y)(i.e. quantum sub-ring).
(2) B(Y)is a sub-theory ofB(X)(sub-moduli, invariant sub-VHS).
(3) A(Y)can be reconstructed from a “refinedA theory” on
X◦:=X\[k
i=1Si
“linked” by the vanishing 3-spheres inB(X).
(4) B(X)can be reconstructed from the variations of MHS on H3(Y◦),
Y◦ :=Y\[k
i=1Ci,
“linked” by the exceptional curves inA(Y).
For (3) and (4),effective methodsare under developed.
VI. Linked GW invariants for A model
I It is easy to see thatA(X) ⊂A(Y): for 06=β∈H2(X), by degeneration formula in GW theory [Li] we can show
nXβ =
∑
γ7→βnYγ.I To determineA(Y)fromA(X) +B(X), it is equivalent to find a definition of each individual term nYγ with the same βin terms of a refined data in X.
Lemma
H2(X◦) ∼=H2(Y◦) ∼=H2(Y). In particular, for a map h : C→X◦, γ:=h∗[C] ∈H2(Y)is well-defined.
I This γ is calleda linking data(β, L). It encodes the link between h(C)(2D) and Si’s (3D) inside a 6D space X.
I If the stable maps h : C→Xdo not touchSiS3i, then the linked GW invariants nX(β,L)is defined and nX(β,L) =nYγ.
I In general this is not true. However, it istrue in the virtual sense, which is all we need:
Proposition
For Xtwith t∈A1\ {0}small in the degenerating family π:X →A1arising from the semi-stable reduction, we have a decomposition of the virtual class[M(Xt, β)]virtinto a finite disjoint union of cycles
[M(Xt, β)]virt =
ä
γ∈H2(X◦)[M(Xt, γ)]virt,where[M(Y, γ)]virt ∼ [M(Xt, γ)]virt ∈Avdim M(Xt, β) is a cycle class corresponding to the linking data γ of Xt.
VII. Linked B model via VMHS
I B(Y)is a sub-theory ofB(X)by viewingMY,→MXas a boundary strata ofMX.
I We will show thatB(Y), together with the knowledge of extremal curves Z :=SiCi⊂Y determinesB(X).
I Proposition
There is a short exact sequence ofmixed Hodge structures
0→V→H3(X) →H3(U) →0, (1) where H3(X)is equipped with thelimiting MHS of Schmid,
V∼=H1,1∞ H3(X),
and H3(U)is equipped with thecanonical mixed Hodge structure of Deligne. In particular, F3H3(X) ∼=F3H3(U), F2H3(X) ∼=F2H3(U).
I We have on X, U∼=X◦ :=X\p, where p=S{pi}:
· · ·H1p(ΘX¯) →H1(ΘX¯) →H1(U, TU) →H2p(ΘX¯) → · · ·.
I [Schlessinge] pi is a hypersurface singularity=⇒
depthOpi =3=⇒Hp1(ΘX¯) =0 and H2p(ΘX¯) ∼= Lki=1Cpi: 0→H1(ΘX¯) →H1(U, TU) →H2p(ΘX¯) → · · ·.
I Comparing with the local to global spectral sequence
0→H1(ΘX¯)→λ Ext1(ΩX¯,OX¯) →H0(E xt1(ΩX¯,OX¯))→κ H2(ΘX¯),
I ⇒Def(X¯) ∼=H1(U, TU). Similarly, for Y⊃Z=SCiwe get Def(Y) =H1(TY) ⊂H1(U, TU) ∼=Def(X¯),
and thenMY ,→MX¯ (unobstructedness theorem).
I WriteI :=I as the ideal sheaf ofM ⊂M .
I Since H2(U, TU) 6=0, the deformation of U could be obstructed. Nevertheless, the first-order deformation of U exists and is parameterized by H1(U, TU) ⊃Def(Y).
I Therefore, we have the following smooth family π: U→ Z1:=ZMX(I2) ⊃MY, whereZ1 =ZMX¯(I2)stands for the nonreduced subscheme ofMXas the first jet extension ofMYinMX.
I [Katz]∇GMfor π : U→ Z1is defined by the lattice
H3(U,Z) ⊂H3(U,C).It underlies VMHSinstead of VHS.
I The proposition implies
WiH3(U) =0, i≤2; W3 ⊂W4 with GrW3 H3(U) ∼=H3(Y)and GrW4 H3(U) ∼=V∗.
I The Hodge filtration of the local system F0=H3(U,C): F• = {F3 ⊂F2⊂F1⊂F0}
satisfiesGriffiths’ transversality.
I Since KU ∼=OU, F3is a line bundle overZ1spanned by Ω∈Ω3U/Z
1. Near[Y] ∈ Z1,
F2is then spanned byΩ and v(Ω) where v runs through a basis of H1(U, TU).
I Notice that v(Ω) ∈W3precisely when v∈H1(Y, TY).
I Proposition⇒F3⊂F2on H3(U)overZ1lifts uniquely to
˜F3 ⊂ ˜F2on H3(X)overZ1with
˜F3 ∼=F3, ˜F2 ∼=F2.
I The complete lifting ˜F•is then determined since
˜F1= (˜F3)⊥
bythe first Hodge–Riemann relationon H3(X).
I Now ˜F•overZ1uniquely determines a horizontal map Z1→D.ˇ
I Since it has maximal tangent dimension h1(U, TU) =h1(X, TX), it determinesthe maximal horizontal slice
ψ:M →Dˇ withM ∼=MXnearMY.
I LetΓ be the monodromy group generated by the local monodromy T(i)=exp N(i)around the divisor
Di:= {wi=
∑
jµ=1aijrj =0}.I Under the coordinates t= (r, s), the period map φ:MX= MX¯\[k
i=1Di →D/Γ is then given (by an extension of Schmid’s NOT) as
φ(r, s) =exp
∑
k i=1log wi 2π√
−1N(i)
!
ψ(r, s),
I Since N(i)is determined by the Picard–Lefschetz formula, the period map φ is completely determined by A and Ci’s.
I Hence the refinedB model on Y\Z=U determines theB