ScalingStacks

7. Proof of the main theorem [055I]

Original official author HTML, exact retained edition. Historical TeX conversion verdicts remain unchanged. Cited-edition alignment and mathematical self-containment are not assessed.

Complete original source context · Original author HTML

7. Proof of the main theorem

The goal of this Section is to prove Theorem 1.1. We shall work with the special family of Calabi-Yau varieties 𝒳\mathcal{X} defined in the Introduction. In Section 7.1 we show how to modify the family 𝒳\mathcal{X} to a new family 𝒳^\widehat{\mathcal{X}} such that the new central fiber consists of a chain of three components, with the middle component given by the compactification of the space 𝒩0\mathcal{N}^{0} defined in Section 4.2. Notice in Section 4 a family of neck metrics are constructed on an exhausting family of domains in 𝒩0\mathcal{N}^{0}. In Section 7.2 we review general facts about the Tian-Yau metrics on the complement of a smooth anti-canonical divisor in a Fano manifold. These give Ricci-flat Kähler metrics on the other two components of the central fiber in 𝒳^\widehat{\mathcal{X}}. In Section 7.3 we explain how to graft the above neck metrics and Tian-Yau metrics on the central fiber of 𝒳^\widehat{\mathcal{X}} to the nearby smooth fibers, and obtain approximately Calabi-Yau metrics in a suitable sense. In Section 7.4 we finish the proof of Theorem 1.1. The arguments are very similar to those in Section 6.3 and 6.4, so we will not provide full details.

7.1. Algebro-geometric aspect

7.1.1. Poincaré residue

We first recall some general facts about Poincaré residues. Given a smooth divisor ZZ in a complex manifold MM of dimension mm, the Poincaré residue map

(7.1) Res:H0​(M,KM⊗[Z])→H0​(Z,KZ)\Res:H^{0}(M,K_{M}\otimes[Z])\rightarrow H^{0}(Z,K_{Z})

can be defined as follows. Given a holomorphic mm form Ω\Omega on MM with a simple pole along ZZ, locally if we choose a defining function hh of ZZ, then h​Ωh\Omega is a holomorphic mm form, and we can write

(7.2) h​Ω=d​h∧Ω~h\Omega=dh\wedge\tilde{\Omega}

for some locally defined holomorphic m−1m-1 form Ω~\tilde{\Omega}. The Poincaré residue of Ω\Omega along ZZ is given by

(7.3) Res⁡(Ω)≡Ω~|Z\Res(\Omega)\equiv\tilde{\Omega}|_{Z}

It is straightforward to check that this does not depend on the choice of hh and Ω~\tilde{\Omega}, and gives rise to a well-defined holomorphic volume form ΩZ\Omega_{Z} globally on ZZ.

If we choose local holomorphic coordinates z1,⋯,zmz_{1},\cdots,z_{m} on MM, then we may write

(7.4) Ω=ghdz1∧⋯dzm.\Omega=\frac{g}{h}dz_{1}\wedge\cdots dz_{m}.

At a point on ZZ where ∂h∂z1≠0\frac{\partial h}{\partial z_{1}}\neq 0, we have then by definition

(7.5) Res⁡(Ω)=g∂h∂z1​d​z2∧⋯∧d​zm.\Res(\Omega)=\frac{g}{\frac{\partial h}{\partial z_{1}}}dz_{2}\wedge\cdots\wedge dz_{m}.

From the local expression one can see that if ZZ is an anti-canonical divisor in MM, and we pick a holomorphic volume form ΩM\Omega_{M} on M∖ZM\setminus Z with a simple pole along ZZ, and then Res⁡(ΩM)\Res(\Omega_{M}) gives a holomorphic volume form ΩZ\Omega_{Z} on ZZ.

A special case is when we have a globally defined holomorphic function h:M→ℂh:M\rightarrow\mathbb{C}, and we are given a holomorphic volume form Ω\Omega on MM, then for each w∈ℂw\in\mathbb{C}, we can apply the above construction to the meromorphic form (h−w)−1​Ω(h-w)^{-1}\Omega. In this way we obtain a nowhere vanishing section Ω′\Omega^{\prime} of the relative canonical bundle KM⊗(h∗​Kℂ)−1K_{M}\otimes(h^{*}K_{\mathbb{C}})^{-1}, on the set where hh is a submersion, and it satisfies the equation

(7.6) d​h∧Ω′=Ω.dh\wedge\Omega^{\prime}=\Omega.

We may also view Ω′\Omega^{\prime} as a holomorphic varying family of holomorphic volume forms on the fibers of hh.

7.1.2. A model partial resolution of singularities

Let 𝒮\mathcal{S} be a two dimensional Ak−1​(k≥2)A_{k-1}(k\geq 2) singularity, which is a hypersurface in ℂ3\mathbb{C}^{3} with defining equation

(7.7) z1​z2+z3k=0.z_{1}z_{2}+z_{3}^{k}=0.

Given two positive integers a1≥a2a_{1}\geq a_{2} with a1+a2=ka_{1}+a_{2}=k, we can define a partial resolution of 𝒮\mathcal{S} as follows. Let 𝒮¯\overline{\mathcal{S}} be the subvariety in the product space ℂ3×ℂ​ℙ2\mathbb{C}^{3}\times\mathbb{C}\mathbb{P}^{2} cut out by the following system of equations

(7.8) {z3a1​u1=z1​u3;z3a2​u2=z2​u3;u1​u2+u32=0;z3a1−a2​u1​z2=u2​z1;z3a2​u3+u1​z2=0.\begin{cases}z_{3}^{a_{1}}u_{1}=z_{1}u_{3};\\ z_{3}^{a_{2}}u_{2}=z_{2}u_{3};\\ u_{1}u_{2}+u_{3}^{2}=0;\\ z_{3}^{a_{1}-a_{2}}u_{1}z_{2}=u_{2}z_{1};\\ z_{3}^{a_{2}}u_{3}+u_{1}z_{2}=0.\end{cases}

where [u1:u2:u3][u_{1}:u_{2}:u_{3}] denotes homogeneous coordinates on ℂ​ℙ2\mathbb{C}\mathbb{P}^{2}. Alternatively, 𝒮¯\overline{\mathcal{S}} can also be described as the closure in ℂ3×ℂ​ℙ2\mathbb{C}^{3}\times\mathbb{C}\mathbb{P}^{2} of the graph of the rational map 𝒮→ℂℙ2;(z1,z2,z3)↦[z1z3a1:z2z3a2:1]\mathcal{S}\rightarrow\mathbb{C}\mathbb{P}^{2};(z_{1},z_{2},z_{3})\mapsto[\frac{z_{1}}{z_{3}^{a_{1}}}:\frac{z_{2}}{z_{3}^{a_{2}}}:1]. On the affine chart {ui≠0}\{u_{i}\neq 0\} we shall denote by vj=uj/ui​(j≠i)v_{j}=u_{j}/u_{i}(j\neq i) the affine coordinates.

Lemma 7.1.

𝒮¯\overline{\mathcal{S}} has at most two possible singularities, which are of type Aa1−1A_{a_{1}-1} and Aa2−1A_{a_{2}-1} respectively, and the projection map 𝒮¯→𝒮\overline{\mathcal{S}}\rightarrow\mathcal{S} is a partial resolution, with exceptional divisor isomorphic to ℂ​ℙ1\mathbb{C}\mathbb{P}^{1}.

Proof.

We first show that the system of equations implies z1​z2+z3k=0z_{1}z_{2}+z_{3}^{k}=0, so that 𝒮¯\overline{\mathcal{S}} does project to 𝒮\mathcal{S}. To see this, we notice the first three equations imply

(7.9) u32​(z1​z2+z3k)=0.u_{3}^{2}(z_{1}z_{2}+z_{3}^{k})=0.

If u3≠0u_{3}\neq 0, then we get z1​z2+z3k=0z_{1}z_{2}+z_{3}^{k}=0. If u3=0u_{3}=0, then by the third equation we get that either u1≠0,u2=0u_{1}\neq 0,u_{2}=0 or u1=0,u2≠0u_{1}=0,u_{2}\neq 0. In the first case using the remaining equations we get z3=z2=0z_{3}=z_{2}=0. In the second case we get z3=z1=0z_{3}=z_{1}=0. In both cases the equation z1​z2+z3k=0z_{1}z_{2}+z_{3}^{k}=0 is indeed satisfied.

Now we study singularities of 𝒮¯\overline{\mathcal{S}}. In the affine chart {u1≠0}\{u_{1}\neq 0\}, we get

(7.10) {v2+v32=0;z2+z3a2​v3=0,\begin{cases}v_{2}+v_{3}^{2}=0;\\ z_{2}+z_{3}^{a_{2}}v_{3}=0,\end{cases}

so we reduce the defining equations to a single equation in the z1,z3,v3z_{1},z_{3},v_{3} variable given by

(7.11) z3a1=z1​v3.z_{3}^{a_{1}}=z_{1}v_{3}.

This has exactly one Aa1−1A_{a_{1}-1} singularity at {z1=z3=v3=0}\{z_{1}=z_{3}=v_{3}=0\}. Similarly, on the affine chart {u2≠0}\{u_{2}\neq 0\} we reduce the equations to

(7.12) z3a2=z2​v3.z_{3}^{a_{2}}=z_{2}v_{3}.

This has exactly one Aa2−1A_{a_{2}-1} singularity at {z2=z3=v3=0}\{z_{2}=z_{3}=v_{3}=0\}. On the affine chart {u3≠0}\{u_{3}\neq 0\}, we reduce the equations to

(7.13) v1​v2+1=0.v_{1}v_{2}+1=0.

which is smooth.

It is then easy to verify that the projection map 𝒮¯→𝒮\overline{\mathcal{S}}\rightarrow\mathcal{S} is an isomorphism outside the point {z1=z2=z3=0}\{z_{1}=z_{2}=z_{3}=0\}, and if z1=z2=z3=0z_{1}=z_{2}=z_{3}=0, we get the equation

(7.14) u1​u2+u32=0,u_{1}u_{2}+u_{3}^{2}=0,

which gives a conic in ℂ​ℙ2\mathbb{C}\mathbb{P}^{2}. ∎

From another point of view, we can view 𝒮\mathcal{S} and 𝒮¯\overline{\mathcal{S}} as families of algebraic curves by projecting to the z3z_{3} variable. For 𝒮\mathcal{S} this is simply the standard nodal degeneration of conics in ℂ2\mathbb{C}^{2}, modified by a base change. The family corresponding to 𝒮¯\overline{\mathcal{S}} is isomorphic to 𝒮\mathcal{S} over any general fiber {z3≠0}\{z_{3}\neq 0\}, and the special fiber of 𝒮¯\overline{\mathcal{S}} is now given by a chain consisting of three components, two of which are given by the proper transforms of the two lines {z1=0}\{z_{1}=0\} and {z2=0}\{z_{2}=0\} in ℂ2\mathbb{C}^{2}, and the middle component is the conic {u1u2+u32=0}\{u_{1}u_{2}+u_{3}^{2}=0\} in ℂ​ℙ2\mathbb{C}\mathbb{P}^{2}. In the special case when a1=a2=1a_{1}=a_{2}=1, 𝒮¯\overline{\mathcal{S}} is smooth and the projection map is precisely the minimal resolution of singularity.

It is well-known that 𝒮\mathcal{S} has a canonical singularity, meaning that the canonical line bundle K𝒮K_{\mathcal{S}} is trivial. An explicit holomorphic volume form Ω𝒮\Omega_{\mathcal{S}} can be written by applying the Poincaré residue to the standard meromorphic 1z1​z2+z3k​d​z1∧d​z2∧d​z3\frac{1}{z_{1}z_{2}+z_{3}^{k}}dz_{1}\wedge dz_{2}\wedge dz_{3} on ℂ3\mathbb{C}^{3}. In the chart {z1≠0}\{z_{1}\neq 0\}, it is given by

(7.15) Ω𝒮=d​z2∧d​z3z2.\Omega_{\mathcal{S}}=\frac{dz_{2}\wedge dz_{3}}{z_{2}}.

Notice 𝒮\mathcal{S} is isomorphic to the quotient ℂ2/ℤk\mathbb{C}^{2}/\mathbb{Z}_{k}, and Ω𝒮\Omega_{\mathcal{S}} pulls-back to a multiple of the standard holomorphic volume form on ℂ2\mathbb{C}^{2}.

Viewing 𝒮\mathcal{S} as fibered over z3∈ℂz_{3}\in\mathbb{C}, we further get a relative holomorphic volume form

(7.16) Ω′=−d​z2z2=d​z1z1.\Omega^{\prime}=-\frac{dz_{2}}{z_{2}}=\frac{dz_{1}}{z_{1}}.

One can see Ω′\Omega^{\prime} is smooth away from the singularity {z1=z2=z3=0}\{z_{1}=z_{2}=z_{3}=0\}, and on each component of the singular fiber it is a meromorphic 1-form with a simple pole along the singularity.

The partial resolution 𝒮¯\overline{\mathcal{S}} is a crepant resolution, i.e. the canonical line bundle K𝒮¯K_{\overline{\mathcal{S}}} is also trivial. Indeed the pull-back Ω𝒮¯\Omega_{\overline{\mathcal{S}}} of Ω𝒮\Omega_{\mathcal{S}} is nowhere vanishing on 𝒮¯\overline{\mathcal{S}}, and by applying the Poincaré residue to the function z3z_{3}, we then get a meromorphic 1-form on each component of the special fiber. On the conic {u1u2+u32=0}\{u_{1}u_{2}+u_{3}^{2}=0\} the meromorphic 1-form is given by v1−1​d​v1=−v2−1​d​v2v_{1}^{-1}dv_{1}=-v_{2}^{-1}dv_{2}. The upshot is that we still get a meromorphic section of the relative canonical bundle, which is smooth away from the two singularities {u1=u3=z1=z2=z3=0}\{u_{1}=u_{3}=z_{1}=z_{2}=z_{3}=0\} and {u2=u3=z1=z2=z3=0}\{u_{2}=u_{3}=z_{1}=z_{2}=z_{3}=0\} of S¯\overline{S}.

7.1.3. A modification of the degenerating family

We now recall the set-up in the introduction. Let n≥2n\geq 2 be an integer. Let f1,f2,ff_{1},f_{2},f be homogeneous polynomials of degree d1≥d2,d1+d2=n+2d_{1}\geq d_{2},d_{1}+d_{2}=n+2 respectively, and let 𝒳⊂ℂ​ℙn+1×Δ\mathcal{X}\subset\mathbb{C}\mathbb{P}^{n+1}\times\Delta be a family of Calabi-Yau hypersurfaces in ℂ​ℙn+1\mathbb{C}\mathbb{P}^{n+1} defined by the equation Ft​(x)=0F_{t}(x)=0, where

(7.17) Ft​(x)≡f1​(x)​f2​(x)+t​f​(x)F_{t}(x)\equiv f_{1}(x)f_{2}(x)+tf(x)

and tt is the complex parameter on the unit disc Δ⊂ℂ\Delta\subset\mathbb{C}. Let p:𝒳→Δp:\mathcal{X}\rightarrow\Delta be the projection map and we denote X^t=p−1​(t)\widehat{X}_{t}=p^{-1}(t).

We further assume f1,f2,ff_{1},f_{2},f are sufficiently general so that the following hold:

  1. (i)

    X0=Y1∪Y2X_{0}=Y_{1}\cup Y_{2}, where Y1={f1=0}Y_{1}=\{f_{1}=0\} and Y2={f2=0}Y_{2}=\{f_{2}=0\} are smooth;

  2. (ii)

    X^t\widehat{X}_{t} is smooth for t≠0t\neq 0.;

  3. (iii)

    D={f1=f2=0}D=\{f_{1}=f_{2}=0\} is a smooth complete intersection;

  4. (iv)

    H={f1=f2=f=0}H=\{f_{1}=f_{2}=f=0\} is a smooth complete intersection in ℂ​ℙn+1\mathbb{C}\mathbb{P}^{n+1}.

The total space 𝒳\mathcal{X} is singular along HH and transverse to H×{0}H\times\{0\} the singularities are locally modeled on a two dimensional ordinary double point. For our purpose we need to perform certain birational transformations to 𝒳\mathcal{X} keeping the general fibers unchanged.

We first do a base change t↦tn+2t\mapsto t^{n+2}, and work on the new family, which we still denote by 𝒳\mathcal{X}. Then 𝒳\mathcal{X} now has singularities along D×{0}D\times\{0\}, transversal to which generically it is a two dimensional Ad−1A_{d-1} singularity, which becomes worse along H×{0}H\times\{0\}. This is usually referred to as a compounded Du Val (cDV) singularity .

Now we apply the family version of the above model partial resolution to 𝒳\mathcal{X}. Let 𝒳^\widehat{\mathcal{X}} be the subvariety in the projective bundle ℙ⁡(𝒪⁡(d2)⊕𝒪⁡(d1)⊕ℂ)\mathbb{P}(\mathcal{O}(d_{2})\oplus\mathcal{O}(d_{1})\oplus\mathbb{C}) over ℂ​ℙn+1×Δ\mathbb{C}\mathbb{P}^{n+1}\times\Delta cut out by the equations

(7.18) {td1​s1=s3​f2​(x);td2​s2=s3​f1​(x);s1⊗s2+s32​f​(x)=0;td1−d2​s1⊗f1​(x)=f2​(x)⊗s2;td2​s3​f​(x)+s1⊗f1​(x)=0.\begin{cases}t^{d_{1}}s_{1}=s_{3}f_{2}(x);\\ t^{d_{2}}s_{2}=s_{3}f_{1}(x);\\ s_{1}\otimes s_{2}+s_{3}^{2}f(x)=0;\\ t^{d_{1}-d_{2}}s_{1}\otimes f_{1}(x)=f_{2}(x)\otimes s_{2};\\ t^{d_{2}}s_{3}f(x)+s_{1}\otimes f_{1}(x)=0.\end{cases}

where naturally we view fi∈H0​(ℂ​ℙn+1,𝒪⁡(di))f_{i}\in H^{0}(\mathbb{C}\mathbb{P}^{n+1},\mathcal{O}(d_{i})), f∈H0​(ℂ​ℙn+1,𝒪⁡(n+2))f\in H^{0}(\mathbb{C}\mathbb{P}^{n+1},\mathcal{O}(n+2)), and [s1:s2:s3][s_{1}:s_{2}:s_{3}] denotes a point in the fiber of the projective bundle over the point (x,t)∈ℂ​ℙn+1×Δ(x,t)\in\mathbb{C}\mathbb{P}^{n+1}\times\Delta.

For our discussion in the rest of this section we shall always take [x0:x1:⋯:xn+1][x_{0}:x_{1}:\cdots:x_{n+1}] to be the homogeneous coordinates of a point xx on ℂ​ℙn+1\mathbb{C}\mathbb{P}^{n+1}. On the affine chart {xi≠0}\{x_{i}\neq 0\} of ℂ​ℙn+1\mathbb{C}\mathbb{P}^{n+1} we denote by u={uj=xj/xi,j≠i}u=\{u_{j}=x_{j}/x_{i},j\neq i\} the affine coordinates, and we view xix_{i} as a local trivialization of 𝒪⁡(1)\mathcal{O}(1). Then on this chart we can view any holomorphic sections of powers of 𝒪⁡(1)\mathcal{O}(1) as local holomorphic functions. In particular, for a homogeneous function R⁡(x)R(x), we denote by R⁡(u)R(u) the corresponding inhomogeneous function. On the affine trivialization of the projective bundle {si≠0}\{s_{i}\neq 0\}, we denote by {ζj=sj/si,j≠i}\{\zeta_{j}=s_{j}/s_{i},j\neq i\} the affine coordinates on the fibers.

We define

(7.19) D1\displaystyle D_{1} ≡{f1(x)=f2(x)=t=0,s2=s3=0},\displaystyle\equiv\{f_{1}(x)=f_{2}(x)=t=0,s_{2}=s_{3}=0\},
(7.20) D2\displaystyle D_{2} ≡{f1(x)=f2(x)=t=0,s1=s3=0}.\displaystyle\equiv\{f_{1}(x)=f_{2}(x)=t=0,s_{1}=s_{3}=0\}.
Lemma 7.2.

𝒳^\widehat{\mathcal{X}} is smooth away from the union D1∪D2D_{1}\cup D_{2}, and transverse to each DiD_{i} the singularity is a two dimensional Adi−1A_{d_{i}-1} singularity.

Proof.

We know 𝒳^\widehat{\mathcal{X}} is isomorphic to 𝒳\mathcal{X} away from D×{0}D\times\{0\}, so it suffices to consider around a point (x,0)(x,0) where f1​(x)=f2​(x)=0f_{1}(x)=f_{2}(x)=0. Locally in an affine chart {s1≠0}\{s_{1}\neq 0\}, 𝒳^\widehat{\mathcal{X}} is then cut out by the equations

(7.21) {f2​(u)​ζ3=td1;f1​(u)​ζ3=td2​ζ2;ζ2+ζ32​f​(u)=0;f2​(u)​ζ2=td1−d2​f1​(u);td2​ζ3​f​(u)+f1​(u)=0.\begin{cases}f_{2}(u)\zeta_{3}=t^{d_{1}};\\ f_{1}(u)\zeta_{3}=t^{d_{2}}\zeta_{2};\\ \zeta_{2}+\zeta_{3}^{2}f(u)=0;\\ f_{2}(u)\zeta_{2}=t^{d_{1}-d_{2}}f_{1}(u);\\ t^{d_{2}}\zeta_{3}f(u)+f_{1}(u)=0.\end{cases}

These can be reduced to two equations on the coordinates uu, tt and ζ3\zeta_{3}, given by

(7.22) {f2​(u)​ζ3−td1=0td2​ζ3​f​(u)+f1​(u)=0.\begin{cases}f_{2}(u)\zeta_{3}-t^{d_{1}}=0\\ t^{d_{2}}\zeta_{3}f(u)+f_{1}(u)=0.\end{cases}

By our assumption (iii) locally we may use v1=f1​(u)v_{1}=f_{1}(u) and v2=f2​(u)v_{2}=f_{2}(u) to replace u1,u2u_{1},u_{2} (say) as local holomorphic coordinates on a neighborhood of xx in ℂ​ℙn+1\mathbb{C}\mathbb{P}^{n+1}. Then it is easy to see the corresponding subvariety is smooth if ζ3≠0\zeta_{3}\neq 0, and has transversal Ad1−1A_{d_{1}-1} singularities along D1D_{1}. So this gives the local description of 𝒳^\widehat{\mathcal{X}} in a neighborhood of D1D_{1}. Similarly on {s2≠0}\{s_{2}\neq 0\} we also know the space is smooth except with transversal Ad2−1A_{d_{2}-1} singularities along D2D_{2}.

On {s3≠0}\{s_{3}\neq 0\}, we use u,t,ζ1,ζ2u,t,\zeta_{1},\zeta_{2} as coordinates, and we get the constraint equations

(7.23) {ζ1​ζ2+f⁡(u)=0,f2​(u)−td1​ζ1=0,f1​(u)−td2​ζ2=0.\begin{cases}\zeta_{1}\zeta_{2}+f(u)=0,\\ f_{2}(u)-t^{d_{1}}\zeta_{1}=0,\\ f_{1}(u)-t^{d_{2}}\zeta_{2}=0.\end{cases}

We only need to consider the points where ζ1=ζ2=t=0\zeta_{1}=\zeta_{2}=t=0, so in particular we also have f⁡(u)=0f(u)=0. At such a point, the differentials of these three equations are (∇f​(u),∇f2​(u),∇f1​(u))(\nabla f(u),\nabla f_{2}(u),\nabla f_{1}(u)). This is non-zero by our assumption (iv). ∎

One can see that the new central fiber X^0\hat{X}_{0} consists of a chain of three smooth components intersecting transversally, given by the proper transforms Y^1,Y^2\hat{Y}_{1},\hat{Y}_{2} of Y1,Y2Y_{1},Y_{2} respectively and the submanifold 𝒩\mathcal{N} in the projective bundle ℙ⁡(L1⊕L2⊕ℂ)\mathbb{P}(L_{1}\oplus L_{2}\oplus\mathbb{C}) over DD cut out by the equation s1​s2=s32​f​(x)s_{1}s_{2}=s_{3}^{2}f(x) (so that 𝒩\mathcal{N} is a quadric bundle over DD, and singular fibers are over HH). Notice 𝒩\mathcal{N} itself is a smooth manifold.

∙\bullet∙\bullet∙\bullet∙\bullet∙\bulletX^t\widehat{X}_{t}Y^1\hat{Y}_{1}Y^2\hat{Y}_{2}D1D_{1}D2D_{2}𝒩\mathcal{N}H×{t}H\times\{t\}X^0=Y^1∪D1𝒩∪D2Y^2\widehat{X}_{0}=\hat{Y}_{1}\cup_{D_{1}}\mathcal{N}\cup_{D_{2}}\hat{Y}_{2}
Figure 7.1. The modified family 𝒳^\widehat{\mathcal{X}}

We then have

(7.24) D1=Y^1∩𝒩,D2=Y^2∩𝒩.D_{1}=\hat{Y}_{1}\cap\mathcal{N},\ \ D_{2}=\hat{Y}_{2}\cap\mathcal{N}.

It is straightforward to see that the normal bundle of DiD_{i} in 𝒩\mathcal{N} is Li−1L_{i}^{-1}.

Next we consider holomorphic volume forms. Viewing 𝒳\mathcal{X} as an anti-canonical divisor in ℂ​ℙn+1×Δ\mathbb{C}\mathbb{P}^{n+1}\times\Delta, then away from D×{0}D\times\{0\}, 𝒳\mathcal{X} is smooth and we then obtain a holomorphic volume form Γ\Gamma. In the affine chart {x0≠0}×Δ⊂ℂℙn+1×Δ\{x_{0}\neq 0\}\times\Delta\subset\mathbb{C}\mathbb{P}^{n+1}\times\Delta, the meromorphic volume form is given by

(7.25) 1Ft​(u)​d​t∧d​u1∧⋯∧d​un+1.\frac{1}{F_{t}(u)}dt\wedge du_{1}\wedge\cdots\wedge du_{n+1}.

So the Poincaré residue on 𝒳\mathcal{X} is

(7.26) Γ=−1(n+2)​tn+1​f​(u)du1∧⋯dun+1.\Gamma=-\frac{1}{(n+2)t^{n+1}f(u)}du_{1}\wedge\cdots du_{n+1}.

It is easy to check using the equation and the genericity assumptions that Γ\Gamma is indeed holomorphic on 𝒳∖D×{0}\mathcal{X}\setminus D\times\{0\}.

Now applying the above discussion to the global function tt on 𝒳\mathcal{X}, then we get a holomorphic family of holomorphic volume forms Γt\Gamma_{t} on each X^t\widehat{X}_{t}. Differentiating the equation Ft​(u)=f1​(u)​f2​(u)+tn+2​f​(u)=0F_{t}(u)=f_{1}(u)f_{2}(u)+t^{n+2}f(u)=0, we get

(7.27) (n+2)​tn+1​f​(u)​d​t+du​Ft=0.(n+2)t^{n+1}f(u)dt+d_{u}F_{t}=0.

In the above affine chart, on the set where ∂Ft∂u1≠0\frac{\partial F_{t}}{\partial u_{1}}\neq 0, we have

(7.28) Γt=1∂Ft​(u)∂u1du2∧⋯dun+1.\Gamma_{t}=\frac{1}{\frac{\partial F_{t}(u)}{\partial u_{1}}}du_{2}\wedge\cdots du_{n+1}.

This is indeed well-defined on X^t\widehat{X}_{t} for t≠0t\neq 0 and also on X0∖DX_{0}\setminus D. On each component YiY_{i} of X0X_{0}, it has a simple pole along DD. Notice Γt\Gamma_{t} is also the natural holomorphic volume form on X^t\widehat{X}_{t} when we apply the Poincaré residue to the divisor X^t\widehat{X}_{t} in ℂ​ℙn+1\mathbb{C}\mathbb{P}^{n+1}.

Now we pass to the resolution 𝒳^\widehat{\mathcal{X}}. Abusing notation we still denote by Γ\Gamma its pull-back.

Lemma 7.3.

Γ\Gamma extends to a global holomorphic volume form on 𝒳^∖(D1∪D2)\widehat{\mathcal{X}}\setminus(D_{1}\cup D_{2}).

Proof.

We only need to consider around a point (x,t,s)(x,t,s) on the exceptional set 𝒩\mathcal{N}, so (x,t)∈D×{0}(x,t)\in D\times\{0\}. Without loss of generality may assume x0≠0x_{0}\neq 0. Since DD is a complete intersection by assumption (iii), we may use v1=f1​(u)v_{1}=f_{1}(u) and v2=f2​(u)v_{2}=f_{2}(u) to replace u1,u2u_{1},u_{2} (say) as local holomorphic coordinates on a neighborhood of xx in ℂ​ℙn+1\mathbb{C}\mathbb{P}^{n+1}. So we can write

(7.29) Γ=−J−1(n+2)​tn+1​f​(u)​d​v1∧d​v2∧d​u3∧⋯∧d​un+1,\Gamma=-\frac{J^{-1}}{(n+2)t^{n+1}f(u)}dv_{1}\wedge dv_{2}\wedge du_{3}\cdots\wedge du_{n+1},

where JJ is the Jacobian given by

(7.30) J=∂f1∂u1​∂f2∂u2−∂f1∂u2​∂f2∂u1.J=\frac{\partial f_{1}}{\partial u_{1}}\frac{\partial f_{2}}{\partial u_{2}}-\frac{\partial f_{1}}{\partial u_{2}}\frac{\partial f_{2}}{\partial u_{1}}.

Suppose first we work on the affine chart {s1≠0}\{s_{1}\neq 0\}. Then we get the local equations for 𝒳^\widehat{\mathcal{X}} given by (7.22). Since we are away from D1D_{1}, we must have ζ3≠0\zeta_{3}\neq 0. Then we can use ζ3,t,u3,⋯,un+1\zeta_{3},t,u_{3},\cdots,u_{n+1} as local holomorphic coordinates on 𝒳^\widehat{\mathcal{X}}. We have

(7.31) d​v1=−td2​f​d​ζ3−d2​td2−1​ζ3​f​d​t−tn+2​ζ3​d​f,dv_{1}=-t^{d_{2}}fd\zeta_{3}-d_{2}t^{d_{2}-1}\zeta_{3}fdt-t^{n+2}\zeta_{3}df,
(7.32) d​v2=d1​td1−1​ζ3−1​d​t−ζ3−2​td1​d​ζ3dv_{2}=d_{1}t^{d_{1}-1}\zeta_{3}^{-1}dt-\zeta_{3}^{-2}t^{d_{1}}d\zeta_{3}

and

(7.33) d​f=∂f∂v1​d​v1+∂f∂v2​d​v2+∑j≥3∂f∂uj​d​uj.df=\frac{\partial f}{\partial v_{1}}dv_{1}+\frac{\partial f}{\partial v_{2}}dv_{2}+\sum_{j\geq 3}\frac{\partial f}{\partial u_{j}}du_{j}.

So we get

(1+td2​ζ3​∂f∂v1)​d​v1\displaystyle(1+t^{d_{2}}\zeta_{3}\frac{\partial f}{\partial v_{1}})dv_{1}
(7.34) =\displaystyle= (−td2​f+tn+2​ζ3−1​∂f∂v2)​d​ζ3−(d2​td2−1​ζ3​f+d1​tn+1​∂f∂v2)​d​t\displaystyle(-t^{d_{2}}f+t^{n+2}\zeta_{3}^{-1}\frac{\partial f}{\partial v_{2}})d\zeta_{3}-(d_{2}t^{d_{2}-1}\zeta_{3}f+d_{1}t^{n+1}\frac{\partial f}{\partial v_{2}})dt mod(d​u3,⋯,d​un+1).\displaystyle\mod(du_{3},\cdots,du_{n+1}).

Hence we get

(7.35) Γ=ζ3−1(1+td2​ζ3​∂f∂v1)​J−1​d​ζ3∧d​t∧d​u3∧⋯∧d​un+1.\Gamma=\frac{\zeta_{3}^{-1}}{(1+t^{d_{2}}\zeta_{3}\frac{\partial f}{\partial v_{1}})}J^{-1}d\zeta_{3}\wedge dt\wedge du_{3}\wedge\cdots\wedge du_{n+1}.

Near t=0t=0 we see Γ\Gamma is smooth around such a point. Similarly we can deal with the chart {s2≠0}\{s_{2}\neq 0\}.

Now on {s3≠0}\{s_{3}\neq 0\}, we only need to consider a point on DD where f=0f=0, then by our assumption (iv) we may use v3=fv_{3}=f as a local holomorphic coordinate to replace u3u_{3} for instance. Then we can write

(7.36) Γ=−1(n+2)​tn+1​f​K−1​d​v1∧d​v2∧d​v3∧d​u4∧⋯∧d​un+1,\Gamma=-\frac{1}{(n+2)t^{n+1}f}K^{-1}dv_{1}\wedge dv_{2}\wedge dv_{3}\wedge du_{4}\cdots\wedge du_{n+1},

where KK is the Jacobian for the change of coordinates. We have

(7.37) d​v3=−(ζ1​d​ζ2+ζ2​d​ζ1),dv_{3}=-(\zeta_{1}d\zeta_{2}+\zeta_{2}d\zeta_{1}),
(7.38) d​v1=td2​d​ζ2+d2​ζ2​td2−1​d​t,dv_{1}=t^{d_{2}}d\zeta_{2}+d_{2}\zeta_{2}t^{d_{2}-1}dt,
(7.39) d​v2=td1​d​ζ1+d1​ζ1​td1−1​d​t.dv_{2}=t^{d_{1}}d\zeta_{1}+d_{1}\zeta_{1}t^{d_{1}-1}dt.

Then we get

(7.40) Γ=K−1​d​t∧d​ζ1∧d​ζ2∧d​u4∧⋯∧d​un+1,\Gamma=K^{-1}dt\wedge d\zeta_{1}\wedge d\zeta_{2}\wedge du_{4}\cdots\wedge du_{n+1},

which is smooth. ∎

Now we can apply the previous Poincaré residue to the function tt on 𝒳^\widehat{\mathcal{X}}. Since the exceptional set of the resolution lies over D×{0}D\times\{0\}, we still get Γt\Gamma_{t} for t≠0t\neq 0. On the central fiber X^0\hat{X}_{0}, we still get Γ0\Gamma_{0} on Y^1∖D1\hat{Y}_{1}\setminus D_{1} and Y^2∖D2\hat{Y}_{2}\setminus D_{2}. Over 𝒩∖(D1∪D2)\mathcal{N}\setminus(D_{1}\cup D_{2}), using (7.35) and (7.40) we get the corresponding Poincaré residue

(7.41) Γ𝒩=J−1​d​ζ1ζ1∧d​u3∧⋯∧d​un+1=−J−1​d​ζ2ζ2∧d​u3∧⋯∧d​un+1.\Gamma_{\mathcal{N}}=J^{-1}\frac{d\zeta_{1}}{\zeta_{1}}\wedge du_{3}\wedge\cdots\wedge du_{n+1}=-J^{-1}\frac{d\zeta_{2}}{\zeta_{2}}\wedge du_{3}\wedge\cdots\wedge du_{n+1}.

Notice by applying Poincaré residue twice to the complete intersection D={f1=f2=0}D=\{f_{1}=f_{2}=0\}, we obtain a holomorphic volume form ΩD\Omega_{D} on DD, which in the above local coordinates can be written as

(7.42) ΩD=J−1​d​u3∧⋯∧d​un+1.\Omega_{D}=J^{-1}du_{3}\wedge\cdots\wedge du_{n+1}.

So we get

(7.43) Γ𝒩=d​ζ1ζ1∧ΩD.\Gamma_{\mathcal{N}}=\frac{d\zeta_{1}}{\zeta_{1}}\wedge\Omega_{D}.

This means that up to multiplying by −−1-\sqrt{-1}, Γ𝒩\Gamma_{\mathcal{N}} agrees with the natural holomorphic volume form Ω0\Omega_{0} on 𝒩0\mathcal{N}_{0} defined in Section 4.2, under the identification k−=d2,k+=−d1k_{-}=d_{2},k_{+}=-d_{1}.

7.2. Tian-Yau metrics

In this subsection we briefly review the complete Ricci-flat Kähler metrics, constructed in [TY90] on the complement of a smooth anti-canonical divisor in a Fano manifold. We will state without proof some facts on the asymptotics of these metrics. Interested readers are referred to [HSVZ18], Section 3 for details.

Let YY be an nn dimensional Fano manifold, DD a smooth anti-canonical divisor in YY, and denote Z=Y∖DZ=Y\setminus D. By adjunction formula DD itself is Calabi-Yau, and we can find a Ricci-flat Kähler metric ωD∈2​π​c1​(LD)\omega_{D}\in 2\pi c_{1}(L_{D}), where LDL_{D} is the restriction of KY−1K_{Y}^{-1} to DD. Fixing a defining section SS of DD, we can view S−1S^{-1} as a holomorphic nn-form ΩZ\Omega_{Z} on ZZ with a simple pole along DD. Rescaling suitably we may assume the Poincaré residue of ΩZ\Omega_{Z} gives a holomorphic volume form ΩD\Omega_{D} on DD satisfying the normalization condition (4.1).

As before we can fix the hermitian metric on LDL_{D} whose curvature form is −−1​ωD-\sqrt{-1}\omega_{D} and we also fix a smooth extension to YY with strictly positive curvature. Then

(7.44) ωZ≡nn+1​−1​∂∂¯​(−log⁡|S|2)n+1n\omega_{Z}\equiv\frac{n}{n+1}\sqrt{-1}\partial\bar{\partial}(-{\log|S|^{2}})^{\frac{n+1}{n}}

defines a Kähler form on a neighborhood of infinity in ZZ. The Tian-Yau metric ωT​Y\omega_{TY} on ZZ is then obtained by solving a Monge-Ampère equation with reference metric ωZ\omega_{Z}. Let 𝒞\mathcal{C} be the Calabi model space constructed using (D,LD,ωD)(D,L_{D},\omega_{D}), as in Section 2.2.

Proposition 7.4 ([TY90], see also [HSVZ18]).

There is a smooth function ϕ\phi on ZZ such that ωT​Y≡ωZ+−1​∂∂¯​ϕ\omega_{TY}\equiv\omega_{Z}+\sqrt{-1}\partial\bar{\partial}\phi is a complete Ricci-flat Kähler metric on ZZ solving the Monge-Ampère equation

(7.45) ωT​Yn=1n⋅2n−1​(−1)n2​ΩZ∧Ω¯Z.\omega_{TY}^{n}=\frac{1}{n\cdot 2^{n-1}}(\sqrt{-1})^{n^{2}}\Omega_{Z}\wedge\overline{\Omega}_{Z}.

Moreover, there is a diffeomorphism Φ:𝒞∖K′→Y∖K\Phi:\mathcal{C}\setminus K^{\prime}\rightarrow Y\setminus K, where K⊂ZK\subset Z is compact and K′={|ξ|≥12}K^{\prime}=\{|\xi|\geq\frac{1}{2}\} and constant δZ>0\delta_{Z}>0, such that the following asymptotics hold uniformly for all zz large

  1. (1)
    (7.46) |∇gZkϕ|gZ=O⁡(e−δZ​(−log⁡|S|2)1/2)​for all​k≥0.|\nabla_{g_{Z}}^{k}\phi|_{g_{Z}}=O(e^{-\delta_{Z}(-\log|S|^{2})^{1/2}})\ \text{for all}\ k\geq 0.
  2. (2)
    (7.47) |∇g𝒞k(Φ∗​JZ−J𝒞)|g𝒞=O⁡(e−(12−ϵ)​zn)​for all​k≥0,ϵ>0.|\nabla_{g_{\mathcal{C}}}^{k}(\Phi^{*}J_{Z}-J_{\mathcal{C}})|_{g_{\mathcal{C}}}=O(e^{-(\frac{1}{2}-\epsilon)z^{n}})\ \text{for all}\ k\geq 0,\epsilon>0.
  3. (3)
    (7.48) |∇g𝒞k(Φ∗​ΩZ−Ω𝒞)|g𝒞=O⁡(e−(12−ϵ)​zn)​for all​k≥0,ϵ>0.|\nabla_{g_{\mathcal{C}}}^{k}(\Phi^{*}\Omega_{Z}-\Omega_{\mathcal{C}})|_{g_{\mathcal{C}}}=O(e^{-(\frac{1}{2}-\epsilon)z^{n}})\ \text{for all}\ k\geq 0,\epsilon>0.
  4. (4)
    (7.49) |∇g𝒞k(Φ∗​ωT​Y−ω𝒞)|g𝒞=O⁡(e−δZ​zn/2).|\nabla_{g_{\mathcal{C}}}^{k}(\Phi^{*}\omega_{TY}-\omega_{\mathcal{C}})|_{g_{\mathcal{C}}}=O(e^{-{\delta_{Z}}z^{n/2}}).
  5. (5)

    There is a constant C>0C>0 such that

    (7.50) C−1​z≤Φ∗​((−log⁡|S|2)1n)≤C​z.C^{-1}z\leq\Phi^{*}((-\log|S|^{2})^{\frac{1}{n}})\leq Cz.

In particular, the space (Z,ωT​Y)(Z,\omega_{TY}) is δZ\delta_{Z}-asymptotically Calabi in the sense of Definition 5.1. For later purposes we also need a simple observation regarding the asymptotics of ωT​Y\omega_{TY}. Fix a local holomorphic chart {U,w1,⋯,wn}\{U,w_{1},\cdots,w_{n}\} centered at a point p∈Dp\in D, i.e. wi​(p)=0w_{i}(p)=0 for all ii, and such that SS is locally defined by w1=0w_{1}=0. Define a cylindrical type Kähler metric as follows

(7.51) ωc​y​l≡∑j≥2−1​d​wj∧d​w¯j+−1​|w1|−2​d​w1∧d​w¯1.\omega_{cyl}\equiv\sum_{j\geq 2}\sqrt{-1}dw_{j}\wedge d\bar{w}_{j}+\sqrt{-1}|w_{1}|^{-2}dw_{1}\wedge d\bar{w}_{1}.

By a straightforward computation we get

Lemma 7.5.

On U∖DU\setminus D, there is a constant C>0C>0 such that

(7.52) C−1​(−log⁡|S|2)1n−1​ωc​y​l≤ωT​Y≤C​(−log⁡|S|2)1n​ωc​y​l,C^{-1}(-\log|S|^{2})^{\frac{1}{n}-1}\omega_{cyl}\leq\omega_{TY}\leq C(-\log|S|^{2})^{\frac{1}{n}}\omega_{cyl},

and for all k≥1k\geq 1, there are constants Ck,mk>0C_{k},m_{k}>0 such that

(7.53) |∇ωc​y​lkωT​Y|ωc​y​l≤Ck​(−log⁡|S|2)mk.|\nabla^{k}_{\omega_{cyl}}\omega_{TY}|_{\omega_{cyl}}\leq C_{k}(-\log|S|^{2})^{m_{k}}.

Using this Lemma, later when we do estimates for quantities using the Tian-Yau metric, we can do computations using the cylindrical metric which becomes much simpler, and in the end we only get an error which is of polynomial order in −log⁡|S|2-\log|S|^{2}.

7.3. Construction of approximately Calabi-Yau metrics

We shall work in the set-up of Section 7.1. Let us recall some notation from previous discussion. The algebro-geometric setup is

  • •

    We have the family of Calabi-Yau varieties p:𝒳^→Δp:\widehat{\mathcal{X}}\rightarrow\Delta in ℂ​ℙn+1×Δ\mathbb{C}\mathbb{P}^{n+1}\times\Delta. Let us denote by X^t\widehat{X}_{t} the fiber p−1​(t)p^{-1}(t). By construction, for t≠0t\neq 0 we know X^t\widehat{X}_{t} can be identified with Xtn+2X_{t^{n+2}} in the original family.

  • •

    The central fiber X^0\widehat{X}_{0} is given by the union of three smooth components: Y^1\hat{Y}_{1}, Y^2\hat{Y}_{2} and 𝒩\mathcal{N}, with Y^j∩𝒩=Dj\hat{Y}_{j}\cap\mathcal{N}=D_{j} both canonically isomorphic to DD.

  • •

    Under the identification k−=d2k_{-}=d_{2}, k+=−d1k_{+}=-d_{1}, and L=𝒪⁡(1)|DL=\mathcal{O}(1)|_{D}, 𝒩∖(D1∪D2)\mathcal{N}\setminus(D_{1}\cup D_{2}) is naturally identified with the space 𝒩0\mathcal{N}^{0} defined in Section 4.2.

  • •

    The normal bundle of DjD_{j} in Y^j\hat{Y}_{j} is Lj=𝒪⁡(d3−j)|DL_{j}=\mathcal{O}(d_{3-j})|_{D} and in 𝒩\mathcal{N} is Lj−1L_{j}^{-1}.

  • •

    There is a relative holomorphic volume form Γt​(t∈Δ)\Gamma_{t}(t\in\Delta) defined on 𝒳^∖{D1∪D2}\widehat{\mathcal{X}}\setminus\{D_{1}\cup D_{2}\}. We denote

    (7.54) {Γ0,1≡Γ0|Z1Γ0,2≡Γ0|Z2,\begin{cases}\Gamma_{0,1}\equiv\Gamma_{0}|_{Z_{1}}\\ \Gamma_{0,2}\equiv\Gamma_{0}|_{Z_{2}},\end{cases}

    where Zj≡Y^j∖DjZ_{j}\equiv\hat{Y}_{j}\setminus D_{j}, and we know

    (7.55) Γ0|𝒩0=−−1​Ω0,\Gamma_{0}|_{\mathcal{N}^{0}}=-\sqrt{-1}\Omega_{0},

    where Ω0\Omega_{0} is the holomorphic volume form on 𝒩0\mathcal{N}^{0} defined in (4.89).

The corresponding metric ingredients are

  • •

    We have the Calabi-Yau metric ωD∈2​π​c1​(L)\omega_{D}\in 2\pi c_{1}(L) on DD, where L=𝒪⁡(1)|DL=\mathcal{O}(1)|_{D}. We fix a hermitian metric on LL with curvature −−1​ωD-\sqrt{-1}\omega_{D}. We also extend this hermitian metric to the whole ℂ​ℙn+1\mathbb{C}\mathbb{P}^{n+1} such that its curvature form defines a smooth Kähler metric ωℂ​ℙn+1\omega_{\mathbb{C}\mathbb{P}^{n+1}}. This then induces hermitian metrics on 𝒪⁡(l)\mathcal{O}(l) for all ll, and also on the pull-back of 𝒪⁡(l)\mathcal{O}(l) to the projective bundle ℙ⁡(𝒪⁡(d2)⊕𝒪⁡(d1)⊕ℂ)\mathbb{P}(\mathcal{O}(d_{2})\oplus\mathcal{O}(d_{1})\oplus\mathbb{C}). Later when ss is a holomorphic section of some 𝒪⁡(l)\mathcal{O}(l), |s||s| will always mean the norm of ss with respect to this fixed hermitian metric.

  • •

    We have the Tian-Yau metrics ωT​Y,j\omega_{TY,j} on ZjZ_{j} for j=1,2j=1,2, by applying the construction in Section 7.2 to the line bundle Lj→DjL_{j}\rightarrow D_{j} and the Calabi-Yau metric ωDj=d3−j⋅ωD\omega_{D_{j}}=d_{3-j}\cdot\omega_{D}. So ωT​Y,j\omega_{TY,j} is asymptotic to

    (7.56) ωZj=nn+1​−1​∂∂¯​(−log⁡|f3−j|2)n+1n,\omega_{Z_{j}}=\frac{n}{n+1}\sqrt{-1}\partial\bar{\partial}(-\log|f_{3-j}|^{2})^{\frac{n+1}{n}},

    and

    (7.57) ωT​Y,jn=(−1)n2n⋅2n−1​d3−jn−1​Γ0,j∧Γ¯0,j,\omega_{TY,j}^{n}=\frac{(\sqrt{-1})^{n^{2}}}{n\cdot 2^{n-1}}d_{3-j}^{n-1}\Gamma_{0,j}\wedge\bar{\Gamma}_{0,j},

    where the coefficient d3−jn−1d_{3-j}^{n-1} arises from the fact we are using ωDj\omega_{D_{j}} instead of ωD\omega_{D} in the construction.

  • •

    The family of incomplete C2,αC^{2,\alpha} approximately Calabi-Yau metrics (ωT,ΩT)(\omega_{T},\Omega_{T}) on ℳT\mathcal{M}_{T}, and (ℳT,ΩT)(\mathcal{M}_{T},\Omega_{T}) is embedded in (𝒩0,Ω0)(\mathcal{N}^{0},\Omega_{0}) as in Section 4.2, with k−=d2k_{-}=d_{2} and k+=−d1k_{+}=-d_{1}.

Remark 7.5.1.

In the case b1​(D)≠0b_{1}(D)\neq 0, by Remark 4.8.2 to ensure ℳT\mathcal{M}_{T} is holomorphically embedded in 𝒩0\mathcal{N}^{0}, we need an appropriate choice of the connection 1-form in the construction of the Kähler metrics ωT\omega_{T}. It is not difficult to see this is always achievable.

Our goal in this subsection is to construct for each tt small a C1,αC^{1,\alpha} Kähler metric ω⁡(t)\omega(t) on X^t\widehat{X}_{t} which is approximately Calabi-Yau in a suitable weighted sense. In the next subsection we shall prove these metrics can be perturbed to genuine Calabi-Yau metrics for tt small.

7.3.1. Matching between the parameters tt and TT

The relationship between the parameters tt and TT can be determined by studying the matching between the Tian-Yau ends and the neck.

In our setting we need to first normalize the Tian-Yau metrics ωT​Y,i\omega_{TY,i} on ZiZ_{i} (as defined in Section 7.2). We define

(7.58) ω~T​Y,j=2−1n​n1n​d3−j−n−1n​ωT​Y,j.\tilde{\omega}_{TY,j}=2^{\frac{-1}{n}}n^{\frac{1}{n}}d_{3-j}^{-\frac{n-1}{n}}\omega_{TY,j}.

Then we have

(7.59) ω~T​Y,j=(−1)n22n​Γ0,j∧Γ¯0,j.\tilde{\omega}_{TY,j}=\frac{(\sqrt{-1})^{n^{2}}}{2^{n}}\Gamma_{0,j}\wedge\bar{\Gamma}_{0,j}.

By definition we can write

(7.60) ω~T​Y,j≡d​dc​ϕj=2​−1​∂∂¯​ϕj,\tilde{\omega}_{TY,j}\equiv dd^{c}\phi_{j}=2\sqrt{-1}\partial\bar{\partial}\phi_{j},

where

(7.61) ϕj=ηj+ψj,\phi_{j}=\eta_{j}+\psi_{j},

with

(7.62) ηj=1n+1⋅k3−j1−nn​nn+1n​(−log⁡|f3−j|)n+1n,\eta_{j}=\frac{1}{n+1}\cdot k_{3-j}^{\frac{1-n}{n}}n^{\frac{n+1}{n}}(-\log|f_{3-j}|)^{\frac{n+1}{n}},

and

(7.63) |∇kψ1|=O⁡(e−δ0​(−log⁡|f3−j|2)1/2),|\nabla^{k}\psi_{1}|=O(e^{-\delta_{0}(-\log|f_{3-j}|^{2})^{1/2}}),

for all k≥0k\geq 0, where the derivatives and norms are taken with respect to the Tian-Yau metric itself (which is equivalent to taking with respect to the metric ωZj\omega_{Z_{j}}).

Now on the neck ℳT\mathcal{M}_{T} we have the asymptotics of the Kähler potential given in Section 4.2. By the discussion there we identify ℳT\mathcal{M}_{T} with an open set in 𝒩0\mathcal{N}^{0}, and we can write

(7.64) Tn−2n​ωT=d​dc​ϕT,T^{\frac{n-2}{n}}\omega_{T}=dd^{c}\phi_{T},

with

(7.65) ϕT={ϕ−≡φ−+ψ−,z<0;ϕ+≡φ++ψ+,z>0,\phi_{T}=\begin{cases}\phi_{-}\equiv\varphi_{-}+\psi_{-},\ \ \ \ z<0;\\ \phi_{+}\equiv\varphi_{+}+\psi_{+},\ \ \ \ z>0,\end{cases}

where

(7.66) {φ−=1n+1​nn+1n​k−−n−1n​(A−−log⁡|s1/s3|);φ+=1n+1​nn+1n​(−k+)−n−1n​(A+−log⁡|s2/s3|),\begin{cases}\varphi_{-}=\frac{1}{n+1}n^{\frac{n+1}{n}}k_{-}^{-\frac{n-1}{n}}(A_{-}-\log|s_{1}/s_{3}|);\\ \varphi_{+}=\frac{1}{n+1}n^{\frac{n+1}{n}}(-k_{+})^{-\frac{n-1}{n}}(A_{+}-\log|s_{2}/s_{3}|),\end{cases}

and for |z|≥1|z|\geq 1 we have

(7.67) |ψ±|=ϵ⁡(z)+ϵT.|\psi_{\pm}|=\epsilon(z)+\epsilon_{T}.

Now on ℳT\mathcal{M}_{T} for |t||t| small

(7.68) td1​s1=s3​f2​(x)t^{d_{1}}s_{1}=s_{3}f_{2}(x)

which gives

(7.69) −d1​log|t|−log⁡|s1||s3|=log⁡|s3||s1|=−log⁡|f2|.-d_{1}\log|t|-\log\frac{|s_{1}|}{|s_{3}|}=\log\frac{|s_{3}|}{|s_{1}|}=-\log|f_{2}|.

So if we want to graft the metrics on the three components of X^0\widehat{X}_{0} to nearby X^t\widehat{X}_{t}, then we need

(7.70) d1​log⁡|t|=−A−.d_{1}\log|t|=-A_{-}.

Similarly at the positive end we need

(7.71) d2​log⁡|t|=−A+.d_{2}\log|t|=-A_{+}.

This suggests that we should choose

(7.72) |t|=e−1d1​A−=e−1d2​A+.|t|=e^{-\frac{1}{d_{1}}A_{-}}=e^{-\frac{1}{d_{2}}A_{+}}.

Given |t||t| small we can find TT big so that (7.72) holds. It is not necessary that TT is uniquely determined by tt, but we shall always fix a particular choice for each tt throughout this section so that (7.72) holds. With this choice it is easy to see that

(7.73) C−1​e−1d1​d2​n​T2≤|t|≤C​e−1d1​d2​n​T2.C^{-1}e^{-\frac{1}{d_{1}d_{2}n}T^{2}}\leq|t|\leq Ce^{-\frac{1}{d_{1}d_{2}n}T^{2}}.

7.3.2. Fixing the constants in the definition of weighted spaces

From now on, we will fix weight parameters in the definition of weight spaces, which allows us to prove the uniform injectivity estimate in Proposition 7.15 and apply the implicit function theorem to complete the proof the main theorem in Section 7.4. The parameters δ\delta, μ\mu, ν\nu are fixed as follows (similar to the specification of the parameters in Section 6.1):

  1. (GP1)

    (Fix ν\nu) The parameter ν∈ℝ\nu\in\mathbb{R} is chosen such that

    (7.74) ν∈(−1,0).\displaystyle\nu\in(-1,0).
  2. (GP2)

    (Fix α\alpha) The Hölder constant α∈(0,1)\alpha\in(0,1) is chosen such that

    (7.75) ν+α<0.\displaystyle\nu+\alpha<0.
  3. (GP3)

    (Fix δ\delta) The constant δ>0\delta>0 is chosen such that

    (7.76) 0<δ<δG≡1n⋅(|k−|+|k+|)⋅min⁡{δe,δZ1,δZ2,ϵZ1,ϵZ2,λD},0<\delta<\delta_{G}\equiv\frac{1}{n\cdot(|k_{-}|+|k_{+}|)}\cdot\min\{\delta_{e},\delta_{Z_{1}},\delta_{Z_{2}},\epsilon_{Z_{1}},\epsilon_{Z_{2}},\sqrt{\lambda_{D}}\},

    where λD\sqrt{\lambda_{D}} is in Lemma 6.7 (Liouville theorem on QQ), δe>0\delta_{e}>0 is in Proposition 4.23, δZ1,δZ2\delta_{Z_{1}},\delta_{Z_{2}} are the constants in Proposition 7.4 applied to Z1,Z2Z_{1},Z_{2}, and ϵZ1,ϵZ2\epsilon_{Z_{1}},\epsilon_{Z_{2}} are the constants in Theorem 5.2 applied to Z1,Z2Z_{1},Z_{2}.

  4. (GP4)

    (Fix μ\mu) The parameter μ>0\mu>0 is chosen as

    (7.77) μ=(1−1n)​(ν+2+α).\mu=(1-\frac{1}{n})(\nu+2+\alpha).

As a comparison, on the neck region ℳT\mathcal{M}_{T}, the corresponding choice of parameters are given in (6.10), (6.11), (6.12) and (6.13).

7.3.3. Construction of ω⁡(t)\omega(t)

We will divide a neighborhood of X^0\widehat{X}_{0} into various regions (c.f. Figure 7.2)

  • •

    Region 𝐈\bf{I} is given by 2​|s3|≥max⁡(|s1|,|s2|)2|s_{3}|\geq\max(|s_{1}|,|s_{2}|);

  • •

    Region 𝐈𝐈−\bf{II}_{-} is given by s1≠0s_{1}\neq 0, and |s3|≤2​|s1|,−log⁡|f2|≥−d12​log⁡|t||s_{3}|\leq 2|s_{1}|,-\log|f_{2}|\geq-\frac{d_{1}}{2}\log|t|;

  • •

    Region 𝐈𝐈𝐈−\bf{III}_{-} is given by s1≠0s_{1}\neq 0, and |f2|≤1/2|f_{2}|\leq 1/2, −log⁡|f2|≤−d12​log⁡|t|+1-\log|f_{2}|\leq-\frac{d_{1}}{2}\log|t|+1;

  • •

    Region 𝐈𝐕−\bf{IV}_{-} is given by s1≠0s_{1}\neq 0 and |f2|≥1/4|f_{2}|\geq 1/4;

  • •

    Region 𝐈𝐈+\bf{II}_{+} is given by s2≠0s_{2}\neq 0, and |s3|≤2​|s2|,−log⁡|f1|≥−d22​log⁡|t||s_{3}|\leq 2|s_{2}|,-\log|f_{1}|\geq-\frac{d_{2}}{2}\log|t|;

  • •

    Region 𝐈𝐈𝐈+\bf{III}_{+} is given by s2≠0s_{2}\neq 0 and |f1|≤1/2|f_{1}|\leq 1/2, −log⁡|f1|≥−d22​log⁡|t|+1-\log|f_{1}|\geq-\frac{d_{2}}{2}\log|t|+1;

  • •

    Region 𝐈𝐕+\bf{IV}_{+} is given by s2≠0s_{2}\neq 0 and |f1|≥1/4|f_{1}|\geq 1/4

∙\bullet∙\bulletY1Y_{1}Y2Y_{2}X^t\widehat{X}_{t}D1D_{1}D2D_{2}𝒩\mathcal{N}Region 𝐈𝐕−\bf{IV}_{-}Region 𝐈𝐈𝐈−\bf{III}_{-}Region 𝐈𝐈−\bf{II}_{-}Region 𝐈\bf{I}Region 𝐈𝐕+\bf{IV}_{+}Region 𝐈𝐈𝐈+\bf{III}_{+}Region 𝐈𝐈+\bf{II}_{+}
Figure 7.2. Division of a neighborhood of X^0\widehat{X}_{0}

For all |t||t| sufficiently small, then we also get a division of X^t\widehat{X}_{t} into 7 regions. Notice we have non-empty intersections between these regions and we shall need a cut-off (gluing) on the overlap.

For the convenience of later analysis, we now fix a finite cover 𝒰={Uβ1,Uγ2,U𝒩,U−,U+}\mathcal{U}=\{U_{\beta}^{1},U_{\gamma}^{2},U_{\mathcal{N}},U_{-},U_{+}\} of a neighborhood of X^0\widehat{X}_{0} in 𝒳^\widehat{\mathcal{X}} obtained as follows.

We first cover a neighborhood of D1D_{1}. Given any point in (x,t,[s1:s2:s3])∈D1(x,t,[s_{1}:s_{2}:s_{3}])\in D_{1}, we have t=f1​(x)=f2​(x)=s2=s3=0,s1≠0t=f_{1}(x)=f_{2}(x)=s_{2}=s_{3}=0,s_{1}\neq 0. On the open subset {xj≠0}\{x_{j}\neq 0\} in ℂ​ℙn+1\mathbb{C}\mathbb{P}^{n+1}, we can view σ=xj\sigma=x_{j} as a trivialization of 𝒪⁡(1)\mathcal{O}(1). Without loss of generality we may assume j=0j=0. Then we get affine coordinates {ui=xi/x0(i=1,⋯,n+1)}\{u_{i}=x_{i}/x_{0}(i=1,\cdots,n+1)\}, and we can v1=f1​(u)v_{1}=f_{1}(u) and v2=f2​(u)v_{2}=f_{2}(u) as local holomorphic functions on ℂ​ℙn+1\mathbb{C}\mathbb{P}^{n+1}. Further without loss of generality we can assume {v1,v2,ui=xi/x0​(i=3,⋯)}\{v_{1},v_{2},u_{i}=x_{i}/x_{0}(i=3,\cdots)\} yield local holomorphic coordinates in a neighborhood of xx in ℂ​ℙn+1\mathbb{C}\mathbb{P}^{n+1}. Correspondingly we can pull-back these to local holomorphic functions on the projective bundle ℙ⁡(𝒪⁡(d2)⊕𝒪⁡(d1)⊕ℂ)\mathbb{P}(\mathcal{O}(d_{2})\oplus\mathcal{O}(d_{1})\oplus\mathbb{C}). As before we also introduce local holomorphic functions ζ3=s3/s1,ζ2=s2/s1\zeta_{3}=s_{3}/s_{1},\zeta_{2}=s_{2}/s_{1} on the projective bundle, and the space 𝒳^\widehat{\mathcal{X}} is then defined by the equations as in (7.21), which essentially reduces to one relation v2​ζ3=td1v_{2}\zeta_{3}=t^{d_{1}} in the three variables v2,ζ3,tv_{2},\zeta_{3},t. We denote by Uβ1U_{\beta}^{1} an open subset in 𝒳^\widehat{\mathcal{X}} defined by the inequalities |ζ3|<3​|σ|d2|\zeta_{3}|<3|\sigma|^{d_{2}}, |v2|<3​|σ|d2|v_{2}|<3|\sigma|^{d_{2}}, and |ui|<C⁡(i=3,⋯)|u_{i}|<C(i=3,\cdots) for some fixed C>0C>0. Call such an open set Uβ1U_{\beta}^{1}, and denote the trivializing section σ\sigma by σβ1\sigma_{\beta}^{1}. For |t||t| small, Uβ,t1≡Uβ1∩X^tU^{1}_{\beta,t}\equiv U^{1}_{\beta}\cap\widehat{X}_{t} is then defined by the equation v2​ζ3=td1v_{2}\zeta_{3}=t^{d_{1}}.

We have the natural projection maps

(7.78) πβ1:Uβ,t1→Uβ,0∩Y1;(x,t,v2,ζ3)↦(v2,0),\pi_{\beta}^{1}:U^{1}_{\beta,t}\rightarrow U_{\beta,0}\cap Y_{1};(x,t,v_{2},\zeta_{3})\mapsto(v_{2},0),
(7.79) πβ𝒩:Uβ,t1→Uβ,0∩𝒩;(x,t,v2,ζ3)↦(0,ζ3),\pi_{\beta}^{\mathcal{N}}:U^{1}_{\beta,t}\rightarrow U_{\beta,0}\cap\mathcal{N};(x,t,v_{2},\zeta_{3})\mapsto(0,\zeta_{3}),
(7.80) πβ1,D:Uβ,t1→D;(x,t,v2,ζ3)→x.\pi^{1,D}_{\beta}:U^{1}_{\beta,t}\rightarrow D;(x,t,v_{2},\zeta_{3})\rightarrow x.

Then the union of images πβ1,D​(Uβ,t1)\pi^{1,D}_{\beta}(U^{1}_{\beta,t}) form an open cover of DD. By compactness we can choose and then fix finitely many of them which also cover DD, and we put these Uβ1U^{1}_{\beta}’s in 𝒰\mathcal{U}. Then we obtain also a cover of a neighborhood of D1D_{1} in Y1Y_{1} by {Uβ,0∩Y1}\{U_{\beta,0}\cap Y_{1}\} and a cover of a neighborhood of D1D_{1} in 𝒩\mathcal{N} by {Uβ,0∩𝒩}\{U_{\beta,0}\cap\mathcal{N}\} so that on each element in the cover we have holomorphic coordinates. Without loss of generality we may assume these cover the neighborhood defined by |f2|≤3|f_{2}|\leq 3 and |s3/s1|≤3|s_{3}/s_{1}|\leq 3. So in particular they contain Regions 𝐈𝐈−\bf{II}_{-} and 𝐈𝐈𝐈−\bf{III}_{-}.

We can do the same with D2D_{2}, and add the corresponding elements Uγ2U_{\gamma}^{2} to 𝒰\mathcal{U}. Now away from D1∪D2D_{1}\cup D_{2} we may find a trivialization of the fibration 𝒳^→Δ\widehat{\mathcal{X}}\rightarrow\Delta. So we can obtain three open subsets of 𝒳^\widehat{\mathcal{X}}, each of which has a differentiable trivialization over Δ\Delta. Call these U𝒩U_{\mathcal{N}}, U−U_{-}, U+U_{+}. Adding these to 𝒰\mathcal{U} we then obtain an open cover of a neighborhood of X^0\widehat{X}_{0}. Over each of the three subsets we also have the projection map π−,π+\pi_{-},\pi_{+}, and π𝒩\pi_{{\mathcal{N}}} from them into X^0∖(D1∪D2)\widehat{X}_{0}\setminus(D_{1}\cup D_{2}). We may assume that Region 𝐈\bf{I} is contained in U𝒩U_{\mathcal{N}}, Region 𝐈𝐕±\bf{IV}_{\pm} is contained in U±U_{\pm}.

We shall fix a partition of unity χβ1\chi_{\beta}^{1} of DD subordinate to the cover πβ1,D​(Uβ,t1)\pi_{\beta}^{1,D}(U^{1}_{\beta,t}), and χγ2{\chi_{\gamma}^{2}} of DD subordinate to the cover πγ1,D​(Uγ,t2)\pi_{\gamma}^{1,D}(U^{2}_{\gamma,t}). We view these naturally as functions on the corresponding Uβ,t1U^{1}_{\beta,t} and Uγ,t2U^{2}_{\gamma,t}, though not compactly supported (along the fiber direction).

Below we define the approximately Calabi-Yau metric ω⁡(t)\omega(t) on (X^t,Γt)(\widehat{X}_{t},\Gamma_{t}) for each region above, and we also define the weight function ρt​(𝒙)\rho_{t}(\bm{x}) simultaneously and measure the error of the Calabi-Yau equation in the weighted sense.

(7.81) Errt≡(−1)n2​2−n​Γt∧Γ¯tω​(t)n/n!−1.\mathrm{Err}_{t}\equiv\frac{(\sqrt{-1})^{n^{2}}2^{-n}\Gamma_{t}\wedge\bar{\Gamma}_{t}}{\omega(t)^{n}/n!}-1.

Obviously Errt=0\mathrm{Err}_{t}=0 if and only if ω⁡(t)\omega(t) is Calabi-Yau. Also in the meantime we discuss the gluing in the intersection of neighboring regions.

Region 𝐈\bf{I}. In this region we define

(7.82) ω⁡(t)=T2−nn​(T​ωℂ​ℙn+1|X^t+d​dc​ϕt,𝒩),\omega(t)=T^{\frac{2-n}{n}}(T\omega_{\mathbb{C}\mathbb{P}^{n+1}}|_{\widehat{X}_{t}}+dd^{c}\phi_{t,\mathcal{N}}),

where

(7.83) ϕt,𝒩=π𝒩∗​ϕT\phi_{t,\mathcal{N}}=\pi_{\mathcal{N}}^{*}\phi_{T}

Using the fixed diffeomorphism π𝒩\pi_{\mathcal{N}} we may view the Kähler structures (ω⁡(t),Ω⁡(t)=Γt)(\omega(t),\Omega(t)=\Gamma_{t}) on X^t\widehat{X}_{t} as a perturbation of the Kähler structure (ωT,Ω0)(\omega_{T},\Omega_{0}) on 𝒩0\mathcal{N}^{0}.

Notice by Corollary 4.11.1 it is not difficult to see that 𝐈∩𝒩\bf{I}\cap\mathcal{N} is contained in the union 𝐈𝟏∪𝐈𝟐\bf{I}_{1}\cup\bf{I}_{2} (as defined in Section 4.4). So we can define

(7.84) z⁡(𝒙)≡z⁡(π𝒩​(𝒙)),Lt​(𝒙)≡Lt​(π𝒩​(𝒙)),𝔯⁡(𝒙)≡𝔯⁡(π𝒩​(𝒙))z(\bm{x})\equiv z(\pi_{\mathcal{N}}(\bm{x})),\ \ L_{t}(\bm{x})\equiv L_{t}(\pi_{\mathcal{N}}(\bm{x})),\ \ \mathfrak{r}(\bm{x})\equiv\mathfrak{r}(\pi_{\mathcal{N}}(\bm{x}))

and then use (4.273) to define the weight function ρt​(𝒙)\rho_{t}(\bm{x}). We can then apply Proposition 4.24 to conclude that

(7.85) {|Ω⁡(t)−Ω0|Cδ,ν,μ1,α=ϵ¯T2;|ω⁡(t)−ωt|Cδ,ν,μ1,α=ϵ¯T2.\begin{cases}|\Omega(t)-\Omega_{0}|_{C^{1,\alpha}_{\delta,\nu,\mu}}=\underline{\epsilon}_{T^{2}};\\ |\omega(t)-\omega_{t}|_{C^{1,\alpha}_{\delta,\nu,\mu}}=\underline{\epsilon}_{T^{2}}.\end{cases}

Then by Proposition 4.23 we get an error estimate

(7.86) ‖Errt‖Cδ,ν+2,μ0,α​(𝐈∩X^t)=O⁡(Tν+α).\|\mathrm{Err}_{t}\|_{C^{0,\alpha}_{\delta,\nu+2,\mu}({\bf{I}}\cap\widehat{X}_{t})}=O(T^{\nu+\alpha}).

At the two ends of 𝐈∩X^t{\bf{I}}\cap{\widehat{X}_{t}}, we can write down the metric ω⁡(t)\omega(t) in potential form. In the negative end we have f2≠0f_{2}\neq 0, so we can write

(7.87) ωℂ​ℙn+1|X^t=−1d2​d​dc​log⁡|f2|,\omega_{\mathbb{C}\mathbb{P}^{n+1}}|_{\widehat{X}_{t}}=-\frac{1}{d_{2}}dd^{c}\log|f_{2}|,

and

(7.88) ω⁡(t)=T−n−2n​d​dc​ϕt,𝐈−,\omega(t)=T^{-\frac{n-2}{n}}dd^{c}\phi_{t,\bf{I}_{-}},

where

(7.89) ϕt,𝐈−=−Td2​log⁡|f2|+ϕt,𝒩.\phi_{t,\bf{I}_{-}}=-\frac{T}{d_{2}}\log|f_{2}|+\phi_{t,\mathcal{N}}.

Similarly at the positive end we have

(7.90) ω⁡(t)=T−n−2n​d​dc​ϕt,𝐈+,\omega(t)=T^{-\frac{n-2}{n}}dd^{c}\phi_{t,\bf{I}_{+}},

where

(7.91) ϕt,𝐈+=−Td1​log⁡|f1|+ϕt,𝒩.\phi_{t,\bf{I}_{+}}=-\frac{T}{d_{1}}\log|f_{1}|+\phi_{t,\mathcal{N}}.

Region 𝐈𝐕±\bf{IV}_{\pm}. We only consider the region 𝐈𝐕−\bf{IV}_{-}, and the other region is similar. We define

(7.92) ω⁡(t)=d​dc​(ϕ1∘π−).\omega(t)=dd^{c}(\phi_{1}\circ\pi_{-}).

Then for |t||t| small we can view (X^t∩𝐈𝐕−,ω⁡(t))(\widehat{X}_{t}\cap{\bf{IV}_{-}},\omega(t)) as a perturbation of the Tian-Yau metric ω~T​Y,1\tilde{\omega}_{TY,1}. It is easy to see that in the intersection X^t∩𝐈𝐕−\widehat{X}_{t}\cap\bf{IV}_{-}, for all k≥0k\geq 0 we have

(7.93) {|∇ω~T​Y,1k(ω⁡(t)−ω~T​Y,1)|ω~T​Y,1=ϵ¯T2;|∇ω~T​Y,1k((Γ⁡(t)−Γ0,1))|ω~T​Y,1=ϵ¯T2.\begin{cases}|\nabla_{\tilde{\omega}_{TY,1}}^{k}(\omega(t)-\tilde{\omega}_{TY,1})|_{\tilde{\omega}_{TY,1}}=\underline{\epsilon}_{T^{2}};\\ |\nabla_{\tilde{\omega}_{TY,1}}^{k}((\Gamma(t)-\Gamma_{0,1}))|_{\tilde{\omega}_{TY,1}}=\underline{\epsilon}_{T^{2}}.\end{cases}

To define the weight we let

(7.94) Lt​(𝒙)≡Tn−2n​(n​k−)1n​(−log⁡4)1n,L_{t}(\bm{x})\equiv T^{\frac{n-2}{n}}(nk_{-})^{\frac{1}{n}}(-\log 4)^{\frac{1}{n}},
(7.95) Ut​(𝒙)=T−T1−n2​Lt​(𝒙)n2,U_{t}(\bm{x})=T-T^{1-\frac{n}{2}}L_{t}(\bm{x})^{\frac{n}{2}},

and then define ρt​(𝒙)\rho_{t}(\bm{x}) as in (4.273).Then we obtain that

(7.96) ‖Errt‖Cδ,ν+2,μ0,α​(𝐈𝐕−∩X^t)=ϵ¯T2.\|\mathrm{Err}_{t}\|_{C^{0,\alpha}_{\delta,\nu+2,\mu}({\bf{IV}_{-}}\cap\widehat{X}_{t})}=\underline{\epsilon}_{T^{2}}.

We also have by assumption the asymptotics at the end

(7.97) ϕ1∘π−=η1∘π−+ψ1∘π−.\phi_{1}\circ\pi_{-}=\eta_{1}\circ\pi_{-}+\psi_{1}\circ\pi_{-}.

Region 𝐈𝐈±\bf{II}_{\pm}. Again we only consider the region 𝐈𝐈−\bf{II}_{-}. We define

(7.98) ω⁡(t)=T2−nn​d​dc​ϕt,𝐈𝐈−,\omega(t)=T^{\frac{2-n}{n}}dd^{c}\phi_{t,\bf{II}_{-}},

where

(7.99) ϕt,𝐈𝐈−=∑χβ1⋅ϕ−∘πβ𝒩.\phi_{t,\bf{II}_{-}}=\sum{\chi_{\beta}^{1}}\cdot\phi_{-}\circ\pi_{\beta}^{\mathcal{N}}.

We need the following Lemma.

Lemma 7.6.

We have the following

  1. (1)

    On πβ1​(Uβ,t1∩Uβ′,t1)\pi^{1}_{\beta}(U^{1}_{\beta,t}\cap U^{1}_{\beta^{\prime},t}), we write πβ′1∘(πβ1)−1​(𝒙)=𝒙′\pi^{1}_{\beta^{\prime}}\circ(\pi^{1}_{\beta})^{-1}(\bm{x})=\bm{x}^{\prime}. Suppose 𝒙\bm{x} and 𝒙′\bm{x}^{\prime} have coordinates given by (v2,0,ui)(v_{2},0,u_{i}) and (v2′′,0,ui′′)(v_{2}^{\prime\prime},0,u_{i}^{\prime\prime}) in the chart Uβ,01∩Y1U^{1}_{\beta,0}\cap Y_{1}. Then we have

    (7.100) {v2′′=v2⋅(1+v1​F2)ui′′=ui+v1​Gi,\begin{cases}v_{2}^{\prime\prime}=v_{2}\cdot(1+v_{1}F_{2})\\ u_{i}^{\prime\prime}=u_{i}+v_{1}G_{i},\end{cases}

    where F2F_{2} and GiG_{i} are smooth functions in v1,v2,uiv_{1},v_{2},u_{i}, and v1v_{1} is implicitly determined by v2,uiv_{2},u_{i} and tt by the equation (7.22)

  2. (2)

    On πβ𝒩​(Uβ,t1∩Uβ′,t1)\pi^{\mathcal{N}}_{\beta}(U^{1}_{\beta,t}\cap U^{1}_{\beta^{\prime},t}), we write πβ′𝒩∘(πβ𝒩)−1​(q)=q′\pi^{\mathcal{N}}_{\beta^{\prime}}\circ(\pi^{\mathcal{N}}_{\beta})^{-1}(q)=q^{\prime}. Suppose qq and q′q^{\prime} have coordinates given by (0,ζ3,ui)(0,\zeta_{3},u_{i}) and (0,ζ3′′,ui′′)(0,\zeta_{3}^{\prime\prime},u_{i}^{\prime\prime}) in the chart Uβ,01∩𝒩U^{1}_{\beta,0}\cap\mathcal{N}. Then we have

    (7.101) {ζ3′′=ζ3⋅(1+v2​F~3)ui′′=ui+v2​G~i,\begin{cases}\zeta_{3}^{\prime\prime}=\zeta_{3}\cdot(1+v_{2}\tilde{F}_{3})\\ u_{i}^{\prime\prime}=u_{i}+v_{2}\tilde{G}_{i},\end{cases}

    where F~2\tilde{F}_{2} and G~i\tilde{G}_{i} are smooth functions in v1,v2,uiv_{1},v_{2},u_{i}, and v1v_{1} is implicitly determined by ζ3,ui\zeta_{3},u_{i} and tt by the equation (7.22).

Proof.

This involves only local discussion. By construction we get overlapping local holomorphic charts on ℂ​ℙn+1\mathbb{C}\mathbb{P}^{n+1} given by {v1,v2,ui​(i≥3)}\{v_{1},v_{2},u_{i}(i\geq 3)\} and {v1′,v2′,ui′​(i≥3)}\{v_{1}^{\prime},v_{2}^{\prime},u_{i}^{\prime}(i\geq 3)\}. Given a point in this overlap with coordinates (v1,v2,ui)(v_{1},v_{2},u_{i}) and (v1′,v2′,ui′)(v_{1}^{\prime},v_{2}^{\prime},u_{i}^{\prime}) in these two coordinate charts respectively, then we have

(7.102) {v1′=v1⋅Q1​(v1,v2,ui)v2′=v2⋅Q2​(v1,v2,ui)ui′=Ri′​(v1,v2,ui).\begin{cases}v_{1}^{\prime}=v_{1}\cdot Q_{1}(v_{1},v_{2},u_{i})\\ v_{2}^{\prime}=v_{2}\cdot Q_{2}(v_{1},v_{2},u_{i})\\ u_{i}^{\prime}=R_{i}^{\prime}(v_{1},v_{2},u_{i}).\end{cases}

where Q1,Q2Q_{1},Q_{2} are smooth and non-vanishing along DD. More precisely, we have

(7.103) Qi=(σβ′1/σβ1)di.Q_{i}=(\sigma_{\beta^{\prime}}^{1}/\sigma_{\beta}^{1})^{d_{i}}.

Correspondingly we obtain the transition maps on Uβ1∩Uβ′1U^{1}_{\beta}\cap U^{1}_{\beta^{\prime}} given by

(7.104) {v2′=v2⋅Q2​(v1,v2,ui);ζ3′=ζ3⋅Q2−1​(v1,v2,ui);ui′=Ri′​(v1,v2,ui);\begin{cases}v_{2}^{\prime}=v_{2}\cdot Q_{2}(v_{1},v_{2},u_{i});\\ \zeta_{3}^{\prime}=\zeta_{3}\cdot Q_{2}^{-1}(v_{1},v_{2},u_{i});\\ u_{i}^{\prime}=R_{i}^{\prime}(v_{1},v_{2},u_{i});\end{cases}

where using (7.22) we can write v1v_{1} implicitly as a function of v2,ζ3′v_{2},\zeta_{3}^{\prime} and uiu_{i}. In particular, we obtain the transition function of Y1∩Uβ1∩Uβ′1Y_{1}\cap U_{\beta}^{1}\cap U_{\beta^{\prime}}^{1} given by

(7.105) {v2′=v2⋅Q2​(0,v2,ui);ui′=Ri′​(0,v2,ui),\begin{cases}v_{2}^{\prime}=v_{2}\cdot Q_{2}(0,v_{2},u_{i});\\ u_{i}^{\prime}=R_{i}^{\prime}(0,v_{2},u_{i}),\end{cases}

and 𝒩∩Uβ1∩Uβ′1\mathcal{N}\cap U_{\beta}^{1}\cap U_{\beta^{\prime}}^{1} given by

(7.106) {ζ3′=ζ3⋅Q2−1​(v1,0,ui);ui′=Ri′​(v1,0,ui).\begin{cases}\zeta_{3}^{\prime}=\zeta_{3}\cdot Q_{2}^{-1}(v_{1},0,u_{i});\\ u_{i}^{\prime}=R_{i}^{\prime}(v_{1},0,u_{i}).\end{cases}

Then the conclusion follows by a direct calculation. ∎

Proposition 7.7.

In the Region 𝐈𝐈−∩Uβ,t1{\bf{II}_{-}}\cap U_{\beta,t}^{1}, we have for all k≥0k\geq 0

(7.107) |∇k(ϕt,𝐈𝐈−∘(πβ𝒩)−1−ϕ−)|=ϵ¯T2,|\nabla^{k}(\phi_{t,\bf{II}_{-}}\circ(\pi_{\beta}^{\mathcal{N}})^{-1}-\phi_{-})|=\underline{\epsilon}_{T^{2}},

where derivative and norm are taken with respect to the metric ωT\omega_{T}.

Proof.

We may write

(7.108) ϕt,𝐈𝐈−∘(πβ𝒩)−1(𝒙)−ϕ−(𝒙)=∑β′:q∈U1,β′χβ′1(𝒙)(ϕ−∘πβ′𝒩∘(πβ𝒩)−1(𝒙)−ϕ−(𝒙)).\phi_{t,\bf{II}_{-}}\circ(\pi_{\beta}^{\mathcal{N}})^{-1}(\bm{x})-\phi_{-}(\bm{x})=\sum_{\beta^{\prime}:q\in U_{1,\beta^{\prime}}}\chi_{\beta^{\prime}}^{1}(\bm{x})(\phi_{-}\circ\pi_{\beta^{\prime}}^{\mathcal{N}}\circ(\pi_{\beta}^{\mathcal{N}})^{-1}(\bm{x})-\phi_{-}(\bm{x})).

Write

(7.109) πβ′𝒩∘(πβ𝒩)−1​(𝒙)=𝒙′′=(ζ3′′,ui′′).\pi_{\beta^{\prime}}^{\mathcal{N}}\circ(\pi_{\beta}^{\mathcal{N}})^{-1}(\bm{x})=\bm{x}^{\prime\prime}=(\zeta_{3}^{\prime\prime},u_{i}^{\prime\prime}).

Then we write

(7.110) ϕ−​(𝒙′′)−ϕ−​(𝒙)=∫01⟨∇ϕ−​(t​𝒙′′+(1−t)​𝒙),𝒙′′−𝒙⟩​𝑑t.\phi_{-}(\bm{x}^{\prime\prime})-\phi_{-}(\bm{x})=\int_{0}^{1}\langle\nabla\phi_{-}(t\bm{x}^{\prime\prime}+(1-t)\bm{x}),\bm{x}^{\prime\prime}-\bm{x}\rangle dt.

Claim: For any k≥0k\geq 0, there is a Ck>0C_{k}>0 such at for all 𝒙∈𝐈𝐈−\bm{x}\in\bf{II}_{-},

(7.111) |∇ωTk​ϕ−​(𝒙)|≤eCk​T.|\nabla^{k}_{\omega_{T}}\phi_{-}(\bm{x})|\leq e^{C_{k}T}.

To see this we notice by definition ϕ−\phi_{-} satisfies the equation

(7.112) ΔωT​ϕ−=TrωT⁡ωT=n.\Delta_{\omega_{T}}\phi_{-}=\Tr_{\omega_{T}}\omega_{T}=n.

Then we apply the local weighted Schauder estimate Proposition 4.22, (2). Notice by Corollary 4.11.1 Item (2), given c∈(0,1/2)c\in(0,1/2) we have for all 𝒙∈𝐈𝐈−\bm{x}\in\bf{II}_{-},

(7.113) r⁡(𝒙)≥c​T−1​log⁡T.r(\bm{x})\geq cT^{-1}\log T.

Hence for all 𝒙\bm{x}, every point 𝒚\bm{y} in the regularity ball B𝔰⁡(x)​(𝒙)B_{\mathfrak{s}(x)}(\bm{x}) satisfies

(7.114) r⁡(𝒚)≥c2​T−1​log⁡T.r(\bm{y})\geq\frac{c}{2}T^{-1}\log T.

So we can apply the Item (2) in Proposition 4.22, and it suffices to show a bound on the C0C^{0} norm of ϕ−\phi_{-}. By (7.66) it suffices to bound log⁡|ζ3|\log|\zeta_{3}|. By our definition for 𝒙∈𝐈𝐈−\bm{x}\in\bf{II}_{-} we have

(7.115) log⁡|ζ3​(𝒙)|≤C−log⁡r−≤C.\log|\zeta_{3}(\bm{x})|\leq C-\log r_{-}\leq C.

Also since z≥−Tz\geq-T, by Proposition 4.11,

(7.116) log⁡|ζ3​(𝒙)|≥C−log⁡r−≥−C​T2.\log|\zeta_{3}(\bm{x})|\geq C-\log r_{-}\geq-CT^{2}.

So we get

(7.117) |ϕ−​(𝒙)|≤C​Tm,|\phi_{-}(\bm{x})|\leq CT^{m},

for some m>0m>0. This then proves the Claim.

Now it suffices to bound the norm of the vector field 𝒙′′−𝒙\bm{x}^{\prime\prime}-\bm{x} and its convariant derivatives. To this end we divide into two cases.

Case 1: z≤−1z\leq-1. Notice by Lemma 4.9 comparing with the cylindrical metric, we obtain the norm of the tangent vectors |𝒙′′−𝒙|≤|v2|​Tm|\bm{x}^{\prime\prime}-\bm{x}|\leq|v_{2}|T^{m} for some m>0m>0. On the other hand we have |v2|≤C​|f2|≤ϵ¯T2|v_{2}|\leq C|f_{2}|\leq\underline{\epsilon}_{T^{2}}. So we obtain

(7.118) |ϕ−​(𝒙′′)−ϕ−​(𝒙)|=ϵ¯T2.|\phi_{-}(\bm{x}^{\prime\prime})-\phi_{-}(\bm{x})|=\underline{\epsilon}_{T^{2}}.

The higher order derivatives follows similarly by differentiating (7.110) and Lemma 4.9, using the fact that all derivatives of the vector field 𝒙′′−𝒙\bm{x}^{\prime\prime}-\bm{x} in the cylindrical metric is bounded by C​|v2|C|v_{2}|.

Case 2. z≥−1z\geq-1. Then we instead compare the metric ωT\omega_{T} with the standard metric

(7.119) ωs​t​d≡∑j=1n−1−1​d​wj∧d​w¯j+−1​d​ζ3∧d​ζ¯3.\omega_{std}\equiv\sum_{j=1}^{n-1}\sqrt{-1}dw_{j}\wedge d\bar{w}_{j}+\sqrt{-1}d\zeta_{3}\wedge d\bar{\zeta}_{3}.

As in the proof of Proposition 4.24 we first notice

(7.120) ΔωT​wj=ΔωT​ζ3=ΔωT​ζ3−1=0.\Delta_{\omega_{T}}w_{j}=\Delta_{\omega_{T}}\zeta_{3}=\Delta_{\omega_{T}}\zeta_{3}^{-1}=0.

By assumption we have |ζ3|≤C|\zeta_{3}|\leq C in this case, and also by Corollary 4.11.1, Item (3) we get |ζ3−1|≤C​eC​T|\zeta_{3}^{-1}|\leq Ce^{CT}. Then we again apply Schauder estimates Proposition 4.22, Item (2), to get

(7.121) |∇kwj|≤C​eCk​T,|∇kζ3|≤C​eCk​T.|\nabla^{k}w_{j}|\leq Ce^{C_{k}T},|\nabla^{k}\zeta_{3}|\leq Ce^{C_{k}T}.

Hence we get for all k≥0k\geq 0.

(7.122) |∇ωTkωs​t​d|ωT≤C​eCk​T.|\nabla^{k}_{\omega_{T}}\omega_{std}|_{\omega_{T}}\leq Ce^{C_{k}T}.

Now to get a lower bound we use the fact that

(7.123) ωTn≤C​ΩT∧Ω¯T≤C​|ζ3|−2​ωs​t​dn.\omega_{T}^{n}\leq C\Omega_{T}\wedge\bar{\Omega}_{T}\leq C|\zeta_{3}|^{-2}\omega_{std}^{n}.

So we get that

(7.124) ωs​t​d≥C​e−C​T​ωT.\omega_{std}\geq Ce^{-CT}\omega_{T}.

Now we again can first estimate the norm of 𝒙′′−𝒙\bm{x}^{\prime\prime}-\bm{x} and its derivatives using the standard metric, and use the above information to conclude. ∎

Now we define the weight function ρt\rho_{t}. We first define

(7.125) Lt​(𝒙)=∑βχβ​(𝒙)⋅L⁡(πβ𝒩​(𝒙)),𝔯⁡(𝒙)≡e∑βχβ​(𝒙)⋅log⁡𝔯⁡(πβ𝒩​(𝒙)).L_{t}(\bm{x})=\sum_{\beta}\chi_{\beta}(\bm{x})\cdot L(\pi_{\beta}^{\mathcal{N}}(\bm{x})),\ \ \ \ \mathfrak{r}(\bm{x})\equiv e^{\sum_{\beta}\chi_{\beta}(\bm{x})\cdot\log\mathfrak{r}(\pi_{\beta}^{\mathcal{N}}(\bm{x}))}.

Then we define the weight function ρt​(𝒙)\rho_{t}(\bm{x}) as (4.273). Notice we have that on 𝐈𝐈−∩X^t∩Uβ1{\bf{II}_{-}}\cap\widehat{X}_{t}\cap U_{\beta}^{1},

(7.126) Lt​(𝒙)=Tn2−1​(n​k−)1n​(A−−log⁡|r−​(πβ𝒩​(𝒙))|+ϵ⁡(z))1nL_{t}(\bm{x})=T^{\frac{n}{2}-1}(nk_{-})^{\frac{1}{n}}(A_{-}-\log|r_{-}(\pi_{\beta}^{\mathcal{N}}(\bm{x}))|+\epsilon(z))^{\frac{1}{n}}

From this we get that

(7.127) |ω⁡(t)−(πβ𝒩)∗​ωT|Cδ,ν,μ1,α​(𝐈𝐈−∩X^t)=ϵ¯T2.|\omega(t)-(\pi_{\beta}^{\mathcal{N}})^{*}\omega_{T}|_{C^{1,\alpha}_{\delta,\nu,\mu}({\bf{II}_{-}}\cap\widehat{X}_{t})}=\underline{\epsilon}_{T^{2}}.

Now we understand the holomorphic volume form. Using (7.35) we get that

(7.128) Γt=(1+H)​(πβ𝒩)∗​Ω0,\Gamma_{t}=(1+H)(\pi_{\beta}^{\mathcal{N}})^{*}\Omega_{0},

where HH is a holomorphic function in ζ3,v2,w2,⋯,wn−1\zeta_{3},v_{2},w_{2},\cdots,w_{n-1}, and its derivatives is of order ϵ¯T2\underline{\epsilon}_{T^{2}} in these coordinates. Then we again apply weighted Schauder estimates to get that

(7.129) |H|Cδ,ν,μ0,α​(𝐈𝐈−∩X^t)=ϵ¯T2.|H|_{C^{0,\alpha}_{\delta,\nu,\mu}({\bf{II}_{-}}\cap\widehat{X}_{t})}=\underline{\epsilon}_{T^{2}}.

So by Proposition 4.23 we obtain

(7.130) ‖Errt‖Cδ,ν+2,μ0,α​(𝐈𝐈−∩X^t)=O⁡(Tν+α).\|\mathrm{Err}_{t}\|_{C^{0,\alpha}_{\delta,\nu+2,\mu}({\bf{II}_{-}}\cap\widehat{X}_{t})}=O(T^{\nu+\alpha}).

Notice 𝐈𝐈−∩X^t{\bf{II}_{-}}\cap\widehat{X}_{t} has two ends. Along one end it is close to the negative end of Region 𝐈\bf{I}.

Proposition 7.8.

On the intersection 𝐈𝐈−∩𝐈∩X^t{\bf{II}_{-}}\cap{\bf{I}}\cap\widehat{X}_{t} we have for all k≥0k\geq 0

(7.131) |∇k(ϕt,𝐈𝐈−−ϕt,𝐈−−d1​log⁡t)|=ϵ¯T2,|\nabla^{k}(\phi_{t,\bf{II}_{-}}-\phi_{t,\bf{I}_{-}}-d_{1}\log t)|=\underline{\epsilon}_{T^{2}},

where the derivative and norm are taken with respect to ω⁡(t)\omega(t).

Proof.

We work in Uβ1U_{\beta}^{1} for a fixed β\beta. We have

(7.132) ϕt,𝐈−​(𝒙)=Td2​log⁡|f2​(𝒙)|+π𝒩∗​ϕt​(𝒙)\phi_{t,\bf{I}_{-}}(\bm{x})=\frac{T}{d_{2}}\log|f_{2}(\bm{x})|+\pi_{\mathcal{N}}^{*}\phi_{t}(\bm{x})

and

(7.133) (πβ𝒩)∗​ϕ−​(𝒙)=ϕt​(𝒚)−Td2​log⁡r−​(𝒚)(\pi_{\beta}^{\mathcal{N}})^{*}\phi_{-}(\bm{x})=\phi_{t}(\bm{y})-\frac{T}{d_{2}}\log r_{-}(\bm{y})

where 𝒚=πβ𝒩​(𝒙)\bm{y}=\pi_{\beta}^{\mathcal{N}}(\bm{x}). By definition it is easy to see that 𝒚−𝒙\bm{y}-\bm{x} is of order ϵ¯T2\underline{\epsilon}_{T^{2}} in the coordinates in v2,ζ3,w2,⋯,wn−1v_{2},\zeta_{3},w_{2},\cdots,w_{n-1}. By our choice of TT in terms of tt we have

(7.134) −log⁡|r−​(𝒚)|=d1​log|t|−log⁡|f2​(𝒚)|.-\log|r_{-}(\bm{y})|=d_{1}\log|t|-\log|f_{2}(\bm{y})|.

Then by Lemma 7.6, and use weighed Schauder estimates as above we get the conclusion.

∎

By Proposition 7.8, we can easily glue the the potentials in Region 𝐈−\bf{I}_{-} and 𝐈𝐈−\bf{II}_{-}, using a simple cut-off function of the form

(7.135) ϕ⁡(t)≡χ⁡(r−​(𝒙))⋅ϕt,𝐈−+(1−χ⁡(r−​(𝒙)))⋅ϕt,𝐈𝐈−\phi(t)\equiv\chi(r_{-}(\bm{x}))\cdot\phi_{t,\bf{I}_{-}}+(1-\chi(r_{-}(\bm{x})))\cdot\phi_{t,\bf{II}_{-}}

where χ\chi is a cut-off function in ss satisfying

(7.136) χ⁡(s)={1,s≤3/40,s≥5/4.\chi(s)=\begin{cases}1,s\leq 3/4\\ 0,s\geq 5/4.\end{cases}

Along the other end, Region 𝐈𝐈−\bf{II}_{-} is close to the region 𝐈𝐈𝐈−\bf{III}_{-}.

Proposition 7.9.

On the intersection 𝐈𝐈−∩𝐈𝐈𝐈−∩X^t{\bf{II}_{-}}\cap{\bf{III}_{-}}\cap\widehat{X}_{t}, we have for all k≥0k\geq 0

(7.137) |∇k(ϕt,𝐈𝐈−−η1)|=O⁡(e−δe​T),|\nabla^{k}(\phi_{t,\bf{II}_{-}}-\eta_{1})|=O(e^{-\delta_{e}T}),

where the derivative and norm are taken with respect to ω⁡(t)\omega(t), and δe\delta_{e} is defined as in Proposition 4.23.

Proof.

The proof is similar to the previous Proposition. One works in a fixed Uβ1U_{\beta}^{1}, and then we use the asymptotics of ϕ−\phi_{-} (c.f. (7.64)) and the relation between tt and TT (c.f. (7.72)). We omit the details.

∎

Region 𝐈𝐈𝐈±\bf{III}_{\pm}. Again we only consider the Region 𝐈𝐈𝐈−\bf{III}_{-}. The discussion here is very similar to the case of Region 𝐈𝐈−\bf{II}_{-} so we will be sketchy. We define

(7.138) ω⁡(t)=d​dc​ϕt,𝐈𝐈𝐈−,\omega(t)=dd^{c}\phi_{t,\bf{III}_{-}},

where

(7.139) ϕt,𝐈𝐈𝐈−​(𝒙)=∑χβ1​(𝒙)⋅ϕ1∘πβ1​(𝒙).\phi_{t,\bf{III}_{-}}(\bm{x})=\sum\chi_{\beta}^{1}(\bm{x})\cdot\phi_{1}\circ\pi_{\beta}^{1}(\bm{x}).
Proposition 7.10.

In the intersection 𝐈𝐈𝐈−∩Uβ,t1{\bf{III}_{-}}\cap{U_{\beta,t}^{1}}, we have for all k≥0k\geq 0

(7.140) |∇k(ϕt,𝐈𝐈𝐈−∘(πβ1)−1−ϕ1)|=ϵ¯T2,|\nabla^{k}(\phi_{t,\bf{III}_{-}}\circ(\pi_{\beta}^{1})^{-1}-\phi_{1})|=\underline{\epsilon}_{T^{2}},

where derivative is taken with respect to the metric ωT​Y,1\omega_{TY,1}.

The proof is very similar to the proof of Proposition 7.7, except one compares with the cylindrical metric and uses Lemma 7.5. We omit the details.

To define the weight, we also define the function LtL_{t} by setting

(7.141) Lt​(𝒙)=Tn2−1​(n​k−)1n​(−log⁡|f2​(𝒙)|)1nL_{t}(\bm{x})=T^{\frac{n}{2}-1}(nk_{-})^{\frac{1}{n}}(-\log|f_{2}(\bm{x})|)^{\frac{1}{n}}

and correspondingly the weight ρt\rho_{t} using (4.273).

Similar to the case of Region 𝐈𝐈−\bf{II}_{-} we have the holomorphic volume form

(7.142) Γt=(1+H)​(πβ1)∗​Ω0\Gamma_{t}=(1+H)(\pi_{\beta}^{1})^{*}\Omega_{0}

where HH is a holomorphic function in v2,ζ3,w2,⋯,wn−1v_{2},\zeta_{3},w_{2},\cdots,w_{n-1} and is of order ϵ¯T2\underline{\epsilon}_{T^{2}} in these coordinates. We get

(7.143) ‖Errt‖Cδ,ν+2,μ0,α​(𝐈𝐈𝐈−∩X^t)=ϵ¯T2.\|\mathrm{Err}_{t}\|_{C^{0,\alpha}_{\delta,\nu+2,\mu}({\bf{III}_{-}}\cap\widehat{X}_{t})}=\underline{\epsilon}_{T^{2}}.

Region 𝐈𝐈𝐈−\bf{III}_{-} has two ends. One end intersects Region 𝐈𝐕−\bf{IV}-.

Proposition 7.11.

On 𝐈𝐈𝐈−∩𝐈𝐕−\bf{III}_{-}\cap\bf{IV}_{-}, we have for all k≥0k\geq 0

(7.144) |∇k(ϕt,𝐈𝐈𝐈−−ϕ1)|=ϵ¯T2,|\nabla^{k}(\phi_{t,\bf{III}_{-}}-\phi_{1})|=\underline{\epsilon}_{T^{2}},

where the derivative and norm are taken with respect to ω⁡(t)\omega(t).

This is fairly easy to see, by working in a fixed Uβ1U^{1}_{\beta}.

The other end is close to the Region 𝐈𝐈−\bf{II}_{-}.

Proposition 7.12.

On 𝐈𝐈𝐈−∩𝐈𝐈−\bf{III}_{-}\cap\bf{II}_{-} we have for all k≥0k\geq 0

(7.145) |∇k(ϕt,𝐈𝐈𝐈−−η1)|=O⁡(e−δZ1​T),|\nabla^{k}(\phi_{t,\bf{III}_{-}}-\eta_{1})|=O(e^{-\delta_{Z_{1}}T}),

where the derivative and norm are taken with respect to ω⁡(t)\omega(t), and δZ1\delta_{Z_{1}} is the constant in Proposition 7.4 applied to Z1Z_{1}.

To see this we only need to work in a fixed Uβ1U_{\beta}^{1} and use the asymptotics of the Tian-Yau metric ω~T​Y,1\tilde{\omega}_{TY,1}.

Now by Proposition 7.9 and 7.12, we can choose a cut-off function to glue together ϕt,𝐈𝐈−\phi_{t,\bf{II}_{-}} and ϕt,𝐈𝐈𝐈−\phi_{t,\bf{III}_{-}}. Similarly we may also glue the corresponding weight function ρt​(𝒙)\rho_{t}(\bm{x}). Here we need to use (7.126), the fact that

(7.146) −log⁡|f2|=−log⁡|s1||s3|−d1​log⁡|t|,-\log|f_{2}|=-\log\frac{|s_{1}|}{|s_{3}|}-d_{1}\log|t|,

and the relation between |t||t| and TT (7.72).

We also choose a cut-off function to glue together ϕt,𝐈𝐈𝐈−\phi_{t,\bf{III}_{-}} and ϕt,𝐈𝐕−\phi_{t,\bf{IV}_{-}} and also the corresponding weight function ρt​(𝒙)\rho_{t}(\bm{x}).

Similarly we can define the metrics ω⁡(t)\omega(t) on 𝐈𝐈+,𝐈𝐈𝐈+,𝐈𝐕+\bf{II}_{+},\bf{III}_{+},\bf{IV}_{+} and glue together in the intersections and also glue the weight functions.

To sum up, we have constructed a family of C1,αC^{1,\alpha} Kähler metrics ω⁡(t)\omega(t) on X^t\widehat{X}_{t} for |t||t| small such that in the above defined weighted norm

(7.147) ‖Errt‖Cδ,ν+2,μ0,α​(X^t)=O⁡(Tν+α).\|\mathrm{Err}_{t}\|_{C^{0,\alpha}_{\delta,\nu+2,\mu}(\widehat{X}_{t})}=O(T^{\nu+\alpha}).
Remark 7.12.1.

It follows from the construction that ω⁡(t)\omega(t) in the cohomology class T2n​2​π​c1​(𝒪⁡(1)|X^t)T^{\frac{2}{n}}2\pi c_{1}(\mathcal{O}(1)|_{\widehat{X}_{t}}). Hence we get the volume

(7.148) ∫X^tω​(t)nn!=C​T2∼(−log⁡|t|)−1.\int_{\widehat{X}_{t}}\frac{\omega(t)^{n}}{n!}=CT^{2}\sim(-\log|t|)^{-1}.

The above error estimate in particular gives

(7.149) ∫X^tΓt∧Γ¯t∼T2∼(−log⁡|t|)−1.\int_{\widehat{X}_{t}}\Gamma_{t}\wedge\bar{\Gamma}_{t}\sim T^{2}\sim(-\log|t|)^{-1}.

For our analysis in the next subsection we define the normalized holomorphic volume form as

(7.150) Ω⁡(t)≡(2n​∫X^tω​(t)n(−1)n2​∫X^tΓt∧Γ¯t)12⋅Γt.\Omega(t)\equiv(\frac{2^{n}\int_{\widehat{X}_{t}}\omega(t)^{n}}{(\sqrt{-1})^{n^{2}}\int_{\widehat{X}_{t}}\Gamma_{t}\wedge\bar{\Gamma}_{t}})^{\frac{1}{2}}\cdot\Gamma_{t}.

Abusing notation we define Errt\mathrm{Err}_{t} by

(7.151) (−1)n22n​Ω​(t)∧Ω¯​(t)=(1+Errt)​ω​(t)nn!,\frac{(\sqrt{-1})^{n^{2}}}{2^{n}}\Omega(t)\wedge\bar{\Omega}(t)=(1+\mathrm{Err}_{t})\frac{\omega(t)^{n}}{n!},

where

(7.152) ∫X^tErrt​ω​(t)n=0\int_{\widehat{X}_{t}}\mathrm{Err}_{t}\omega(t)^{n}=0

and

(7.153) ‖Errt‖Cδ,ν+2,μ0,α​(X^t)=O⁡(Tν+α).\|\mathrm{Err}_{t}\|_{C^{0,\alpha}_{\delta,\nu+2,\mu}(\widehat{X}_{t})}=O(T^{\nu+\alpha}).

7.4. Global weighted analysis on X^t\widehat{X}_{t} and the proof of the main theorem

Now we are in a position to set up the whole package to implement the global weighted analysis on the glued manifold.

To begin with, let (ω⁡(t),Ω⁡(t))(\omega(t),\Omega(t)) be the C1,αC^{1,\alpha}-Kähler structure constructed in Section 7.3. on X^t\widehat{X}_{t}. So we define the linear spaces

𝔄\displaystyle\mathfrak{A} ≡{−1​∂∂¯​ϕ∈Ω1,1​(X^t)|ϕ∈C2,α​(X^t)},\displaystyle\equiv\Big\{\sqrt{-1}\partial\bar{\partial}\phi\in\Omega^{1,1}(\widehat{X}_{t})\Big|\phi\in C^{2,\alpha}(\widehat{X}_{t})\Big\},
(7.154) 𝔅\displaystyle\mathfrak{B} ≡{f∈C0,α​(X^t)|∫X^tf⋅ω​(t)nn!=0},\displaystyle\equiv\Big\{f\in C^{0,\alpha}(\widehat{X}_{t})\Big|\int_{\widehat{X}_{t}}f\cdot\frac{\omega(t)^{n}}{n!}=0\Big\},

which are equipped with the weighted norms

(7.155) ‖−1​∂∂¯​ϕ‖𝔄\displaystyle\|\sqrt{-1}\partial\bar{\partial}\phi\|_{\mathfrak{A}} ≡‖−1​∂∂¯​ϕ‖Cδ,ν+2,μ0,α​(X^t),−1​∂∂¯​ϕ∈𝔄,\displaystyle\equiv\|\sqrt{-1}\partial\bar{\partial}\phi\|_{C_{\delta,\nu+2,\mu}^{0,\alpha}(\widehat{X}_{t})},\quad\sqrt{-1}\partial\bar{\partial}\phi\in\mathfrak{A},
(7.156) ‖f‖𝔅\displaystyle\|f\|_{\mathfrak{B}} ≡‖−1​∂∂¯​ϕ‖Cδ,ν+2,μ0,α​(X^t),f∈𝔅,\displaystyle\equiv\|\sqrt{-1}\partial\bar{\partial}\phi\|_{C_{\delta,\nu+2,\mu}^{0,\alpha}(\widehat{X}_{t})},\quad f\in\mathfrak{B},

such that both 𝔄\mathfrak{A} and 𝔅\mathfrak{B} are Banach spaces. As in Section 7.3.2, the parameters δ,ν,μ\delta,\nu,\mu are chosen as

(7.157) 0<δ\displaystyle 0<\delta <δG,\displaystyle<\delta_{G},
(7.158) −1<ν\displaystyle-1<\nu <0.\displaystyle<0.

Morevoer, α∈(0,1)\alpha\in(0,1) is sufficiently small such that

(7.159) ν+α<0\nu+\alpha<0

and μ=(1−1n)​(ν+2+α)\mu=(1-\frac{1}{n})(\nu+2+\alpha).

For |t|≪1|t|\ll 1, starting with the Kähler structure (ω⁡(t),Ω⁡(t))(\omega(t),\Omega(t)), we will solve the nonlinear equation

(7.160) 1n!​(ω⁡(t)+−1​∂∂¯​ϕ)n=(−1)n2​2−n⋅Ω⁡(t)∧Ω¯​(t).\frac{1}{n!}(\omega(t)+\sqrt{-1}\partial\bar{\partial}\phi)^{n}=(\sqrt{-1})^{n^{2}}2^{-n}\cdot\Omega(t)\wedge\bar{\Omega}(t).

Let ℱ:𝔄→𝔅\mathscr{F}:\mathfrak{A}\rightarrow\mathfrak{B} be defined by

(7.161) ℱ⁡(−1​∂∂¯​ϕ)⋅ω​(t)n≡(ω⁡(t)+−1​∂∂¯​ϕ)n−ω​(t)n​(1−Errt).\mathscr{F}(\sqrt{-1}\partial\bar{\partial}\phi)\cdot\omega(t)^{n}\equiv(\omega(t)+\sqrt{-1}\partial\bar{\partial}\phi)^{n}-\omega(t)^{n}(1-\mathrm{Err}_{t}).

Then (7.160) is equivalent to

(7.162) ℱ⁡(−1​∂∂¯​ϕ)=0.\mathscr{F}(\sqrt{-1}\partial\bar{\partial}\phi)=0.

Now we write

(7.163) ℱ⁡(v)−ℱ⁡(0)=ℒ⁡(v)+𝒩⁡(v),\mathscr{F}(v)-\mathscr{F}(0)=\mathscr{L}(v)+\mathscr{N}(v),

for any v∈𝔄v\in\mathfrak{A}, where

(7.164) ℒ⁡(−1​∂∂¯​ϕ)=Δ​ϕ,\mathscr{L}(\sqrt{-1}\partial\bar{\partial}\phi)=\Delta\phi,

is the linearization of ℱ\mathscr{F} and

(7.165) 𝒩⁡(−1​∂∂¯​ϕ)⋅ω​(t)n\displaystyle\mathscr{N}(\sqrt{-1}\partial\bar{\partial}\phi)\cdot\omega(t)^{n} =(ω⁡(t)+−1​∂∂¯​ϕ)n−ω​(t)n−n​ω​(t)n−1∧−1​∂∂¯​ϕ.\displaystyle=(\omega(t)+\sqrt{-1}\partial\bar{\partial}\phi)^{n}-\omega(t)^{n}-n\omega(t)^{n-1}\wedge\sqrt{-1}\partial\bar{\partial}\phi.

The proof of the following is identical to Proposition 6.4.

Proposition 7.13 (Nonlinear error estimate).

There exists a constant CN>0C_{N}>0 independent of 0<|t|≪10<|t|\ll 1 such that for all

(7.166) ϱ∈(0,12)\varrho\in(0,\frac{1}{2})

and

(7.167) −1​∂∂¯​ϕ2∈Bϱ​(𝟎)¯⊂𝔄,−1​∂∂¯​ϕ2∈Bϱ​(𝟎)¯⊂𝔄,\sqrt{-1}\partial\bar{\partial}\phi_{2}\in\overline{B_{\varrho}(\bm{0})}\subset\mathfrak{A},\quad\sqrt{-1}\partial\bar{\partial}\phi_{2}\in\overline{B_{\varrho}(\bm{0})}\subset\mathfrak{A},

we have the pointwise estimate

(7.168) ‖𝒩⁡(−1​∂∂¯​ϕ1)−𝒩⁡(−1​∂∂¯​ϕ2)‖𝔅≤CN⋅ϱ⋅‖−1​∂∂¯​(ϕ1−ϕ2)‖𝔄.\displaystyle\|\mathscr{N}(\sqrt{-1}\partial\bar{\partial}\phi_{1})-\mathscr{N}(\sqrt{-1}\partial\bar{\partial}\phi_{2})\|_{\mathfrak{B}}\leq C_{N}\cdot\varrho\cdot\|\sqrt{-1}\partial\bar{\partial}(\phi_{1}-\phi_{2})\|_{\mathfrak{A}}.

The global version of the weighted Schauder estimate Proposition 4.22 takes the following form. Note that the weighted Schauder estimate on the neck is given by Proposition 6.9.

Proposition 7.14 (Weighted Schauder estimate, the global version).

For every α∈(0,1)\alpha\in(0,1), there exists a uniform constant C>0C>0 (independent of |t|≪1|t|\ll 1) such that for every u∈𝔄u\in\mathfrak{A},

(7.169) ‖u‖Cδ,ν,μ2,α​(X^t)≤C⁡(‖Δ​u‖Cδ,ν+2,μ0,α​(X^t)+‖u‖Cδ,ν,μ0​(X^t)).\displaystyle\|u\|_{C_{\delta,\nu,\mu}^{2,\alpha}(\widehat{X}_{t})}\leq C\Big(\|\Delta u\|_{C_{\delta,\nu+2,\mu}^{0,\alpha}(\widehat{X}_{t})}+\|u\|_{C_{\delta,\nu,\mu}^{0}(\widehat{X}_{t})}\Big).

The proof is similar to the proof of Proposition 4.22. From the construction of the metric ω⁡(t)\omega(t) in Section 7.3 the rescaled limit geometries will be the same as in the case of the neck ℳT\mathcal{M}_{T} studied in Section 4.3, except two possible incomplete Calabi model space limits replaced by the two Tian-Yau metrics on the ends. We omit the details.

Proposition 7.15 (Global injectivity estimates).

For all parameters α∈(0,1)\alpha\in(0,1), δ>0\delta>0, μ,ν∈ℝ\mu,\nu\in\mathbb{R} satisfying

(7.170) 0<δ<δG,−1<ν<0,ν+α<0,μ=(1−1n)​(ν+2+α),\displaystyle 0<\delta<\delta_{G},\quad-1<\nu<0,\quad\nu+\alpha<0,\quad\mu=(1-\frac{1}{n})(\nu+2+\alpha),

there exists a uniform constant C>0C>0 (independent of tt) such that for every u∈C2,α​(X^t)u\in C^{2,\alpha}(\widehat{X}_{t}),

(7.171) ‖∇u‖Cδ,ν+1,μ0​(X^t)+‖∇2u‖Cδ,ν+2,μ0​(X^t)≤C⋅‖Δ​u‖Cδ,ν+2,μ0,α​(X^t),\displaystyle\|\nabla u\|_{C_{\delta,\nu+1,\mu}^{0}(\widehat{X}_{t})}+\|\nabla^{2}u\|_{C_{\delta,\nu+2,\mu}^{0}(\widehat{X}_{t})}\leq C\cdot\|\Delta u\|_{C_{\delta,\nu+2,\mu}^{0,\alpha}(\widehat{X}_{t})},
(7.172) [u]Cδ,ν,μ2,α​(X^t)≤C⋅‖Δ​u‖Cδ,ν+2,μ0,α​(X^t).\displaystyle[u]_{C_{\delta,\nu,\mu}^{2,\alpha}(\widehat{X}_{t})}\leq C\cdot\|\Delta u\|_{C_{\delta,\nu+2,\mu}^{0,\alpha}(\widehat{X}_{t})}.

The proof is very similar to the proof of Proposition 6.8, by using a contradiction argument and applying various Liouville theorems. We omit the details and only mention two different points. The first point is that from our construction of ω⁡(t)\omega(t) on X^t\widehat{X}_{t}, if we rescale around points in Region 𝐈𝐕±\bf{IV}_{\pm}, then we will get the Tian-Yau spaces (instead of the incomplete Calabi model spaces) as limits, and we need to use Theorem 5.2. The second point is that the other rescaled limits will be exactly the same as considered in the proof of Proposition 6.8, and this follows from the fact that by construction our metric ω⁡(t)\omega(t) away from the region 𝐈𝐕±\bf{IV}_{\pm} is essentially a small perturbation of the neck region (ℳT,ωT)(\mathcal{M}_{T},\omega_{T}).

Now given Proposition 7.15 as before it is straightforward to see that for |t|>0|t|>0 sufficiently small, there is a ϕ⁡(t)∈𝔄\phi(t)\in\mathfrak{A} solving the Calabi-Yau equation (7.160). By uniqueness of Calabi-Yau metrics, we know T−2n​(ω⁡(t)+−1​∂∂¯​ϕ​(t))T^{-\frac{2}{n}}(\omega(t)+\sqrt{-1}\partial\bar{\partial}\phi(t)) must agree with the Calabi-Yau metric ωC​Y,tn+2\omega_{CY,t^{n+2}} on Xtn+2≃X^tX_{t^{n+2}}\simeq\widehat{X}_{t} in the Introduction. The geometric statements in Theorem 1.1 then follow from similar arguments as in Section 6.4. We omit the details.

Original mathematics by the credited authors. Source-backed reader collection; mathematical self-containment is not assessed.