ScalingStacks

Chapter 79

Full preserved author HTML · source-object tagging pending · cited publication edition alignment unverified.

Limiting behavior of local Calabi-Yau metrics DUKE-CGTP-03-02

Ilia Zharkov Address: Mathematics Department
Duke University
Durham, NC 27708
USA
Email address: zharkov@math.duke.edu
Abstract.

We use a generalization of the Gibbons-Hawking ansatz to study the behavior of certain non-compact Calabi-Yau manifolds in the large complex structure limit. This analysis provides an intermediate step toward proving the metric collapse conjecture for toric hypersurfaces and complete intersections.

1. Introduction

Since the Strominger-Yau-Zaslow conjecture [SYZ96] was made there has been a considerable interest in geometry of Kähler Ricci-flat half-dimensional torus fibrations. For a nonsingular fibration the metrics which are flat along the fibers were used by Hitchin [Hit97] in relation with mirror symmetry and SYZ conjecture. This, so called, semi-flat case was further studied in [LYZ01] and [Leu00]. The Ricci-flatness condition becomes equivalent to the real Monge-Ampère equation on the base. The mirror symmetry is then provided by the Legendre transform.

This simple case has inspired another, more recent, conjecture of Gross and Wilson [GW00] and Kontsevich and Soibelman [KS01] about existence of an integral Kähler affine structure on the limiting space of the metric collapse. But in order to understand this conjecture for general compact Calabi-Yau manifolds (not just tori) one needs a local description of the metric behavior near the singular fibers.

The first example of such a metric was constructed by Ooguri and Vafa [OV96] in two dimensions via periodic Gibbons-Hawking ansatz and was used later by Gross and Wilson [GW00] to justify the collapse picture for K3. Pedersen and Poon [PP91] have written the (non-linear) differential equations for the GH ansatz in higher dimensions but the solutions they found were not periodic in the remaining number of torus variables, and hence cannot apply to our case of interest. An attempt to apply the generalization of Gibbons-Hawking in the periodic situation was made by Matessi [Mat01], though no explicit solutions were found.

In this paper we do not try to solve Gibbons-Hawking differential equation. Rather the goal is to set up the geometric framework for investigation of limiting behavior of such metrics as the size of tori goes to zero. Unfortunately, the key exponential decay lemma is left unproven. A substantial amount of hard analysis of non-linear elliptic PDE with singularities is required for the proof, and we plan to do it elsewhere.

The conjectural description of the limiting metric space is in agreement with the metric collapse picture of [KS01] and [GW00]. But we suggest a more explicit condition for asymptotics at the singular locus. The coordinates of the Gibbons-Hawking ansatz are related to the affine ones via the partial Legendre transform introduced in the last section of the paper.

Notations.

We use the usual convention to sum over repeated indices. Also we deal with orbifolds on the same footing as with regular complex manifolds. That is, when we say a form or a map is holomorphic, it is meant in this orbifold sense.

Acknowledgments.

I am very grateful to M. Gross for explaining to me the right normalization of the holomorphic volume form and several other key issues. Also I have considerably benefited from conversations with R. Bryant, C. Haase, M. Kontsevich, D. Morrison and M. Stern. Finally, I would like to thank IHES for its hospitality and financial support during the summers of 2001 and 2002 where a significant part of the work has been done.

2. Ricci-flat metrics

2.1. Generalized Gibbons-Hawking ansatz

Suppose TnT^{n}, a nn-dimensional (real) torus, acts freely on an NN-(complex) dimensional Kähler manifold MM by Hamiltonian holomorphic isometries. Then MM can be considered as a principal TnT^{n}-bundle over a real manifold of dimension 2​N−n2N-n. The generalized Gibbons-Hawking ansatz expresses the Kähler and Ricci-flat conditions as differential equations in the nn moment map coordinates and N−nN-n holomorphic coordinates on the Kähler quotient.

Let 𝔱\mathfrak{t} denote the Lie algebra of the torus Lie group TnT^{n}, and let 𝔱ℤ\mathfrak{t}_{\mathbb{Z}} be the natural integral lattice in 𝔱\mathfrak{t}. We will fix a basis in 𝔱ℤ\mathfrak{t}_{\mathbb{Z}}. This defines affine coordinates uiu_{i} on the dual space 𝔱∗≅ℝn\mathfrak{t}^{*}\cong\mathbb{R}^{n}. Let YY be either ℂN−n\mathbb{C}^{N-n} or (ℂ∗)N−n≅ℝN−n×(S1)N−n(\mathbb{C}^{*})^{N-n}\cong\mathbb{R}^{N-n}\times(S^{1})^{N-n}, with the affine complex coordinates ηp=xp+i​yp\eta_{p}=x_{p}+iy_{p}, where ypy_{p} are the phase coordinates on the torus (S1)N−n(S^{1})^{N-n} in the latter case.

Consider a principal TnT^{n}-bundle π:M→B∘\pi:M\to B^{\circ} over an open set B∘B^{\circ} in 𝔱∗×Y\mathfrak{t}^{*}\times Y, with coordinates (ui,ηp,η¯q)(u_{i},\eta_{p},\bar{\eta}_{q}). Denote by [ν][\nu] its integral Chern class as an element in H2​(B,𝔱ℤ)H^{2}(B,\mathfrak{t}_{\mathbb{Z}}).

Theorem 2.1 (cf. [PP91]).

Let Vi​jV^{ij}, respectively Wp​qW^{pq}, be real symmetric, respectively hermitian, positive definite matrices of smooth functions on B∘B^{\circ}, locally given by some potential function Φ\Phi:

(1) Vi​j=∂2Φ∂uj​∂uj,Wp​q=−4​∂2Φ∂ηp​∂η¯q,1≤i,j≤nn+1≤p,q≤N.V^{ij}=\frac{\partial^{2}\Phi}{\partial u_{j}\partial u_{j}},\quad W^{pq}=-4\frac{\partial^{2}\Phi}{\partial\eta_{p}\partial\bar{\eta}_{q}},\qquad 1\leq i,j\leq n\quad n+1\leq p,q\leq N.

Then the following 𝔱\mathfrak{t}-valued 2-form is closed:

(2) Fj=−1​(12​∂Wp​q∂uj​d​ηp∧d​η¯q+∂Vi​j∂ηp​d​ui∧d​ηp−∂Vi​j∂η¯q​d​ui∧d​η¯q).F_{j}=\sqrt{-1}\left(\frac{1}{2}\frac{\partial W^{pq}}{\partial u_{j}}d\eta_{p}\wedge d\bar{\eta}_{q}+\frac{\partial V^{ij}}{\partial\eta_{p}}du_{i}\wedge d\eta_{p}-\frac{\partial V^{ij}}{\partial\bar{\eta}_{q}}du_{i}\wedge d\bar{\eta}_{q}\right).

Suppose, in addition, that detVi​j=detWp​q\det V^{ij}=\det W^{pq} and (F1,…,Fn)(F_{1},\dots,F_{n}) is in the cohomology class 2​π​[ν]2\pi[\nu]. Then there exist a connection on the bundle M→BM\to B with associated 1-forms AiA_{i} and the curvature (F1,…,Fn)(F_{1},\dots,F_{n}) such that MM is a Kähler manifold with Ricci-flat metric given by

(3) h=(V−1)i​j​d​zi⊗d​z¯j+Wp​q​d​ηp⊗d​η¯q,h=(V^{-1})^{ij}dz_{i}\otimes d\bar{z}_{j}+W^{pq}d\eta_{p}\otimes d\bar{\eta}_{q},

where d​zj=Vi​j​d​ui+−1⋅Ajdz_{j}=V^{ij}du_{i}+\sqrt{-1}\cdot A_{j} and d​ηpd\eta_{p} form a basis of holomorphic 1-forms. The holomorphic NN-form and the Kähler form:

(4) Ω=∧j=1kdzj⋀∧p=1ldηp,ω=duj∧Aj+−12⋅Wp​qdηp∧dη¯q\Omega=\wedge_{j=1}^{k}dz_{j}\bigwedge\wedge_{p=1}^{l}d\eta_{p},\qquad\omega=du_{j}\wedge A_{j}+\frac{\sqrt{-1}}{2}\cdot W^{pq}d\eta_{p}\wedge d\bar{\eta}_{q}

are compatible in the sense that Ω∧Ω¯=c​o​n​s​t⋅ωN\Omega\wedge\bar{\Omega}=const\cdot\omega^{N}.

Proof.

First we note that the local potential description of VV and WW by (1) insures that

(5) d​Fj=−12⋅d⁡(∂Wp​q∂uj)∧d​ηp∧d​η¯q+−1⋅d(∂Vi​j∂ηp)∧dui∧dηp−−1⋅d(∂Vi​j∂η¯q)∧dui∧dη¯q=−12​(∂2Wp​q∂ui​∂uj+4​∂2Vi​j∂ηp​∂η¯q)​d​ui∧d​ηp∧d​η¯q=0.dF_{j}=\frac{\sqrt{-1}}{2}\cdot d\left(\frac{\partial W^{pq}}{\partial u_{j}}\right)\wedge d\eta_{p}\wedge d\bar{\eta}_{q}\\ +\sqrt{-1}\cdot d\left(\frac{\partial V^{ij}}{\partial\eta_{p}}\right)\wedge du_{i}\wedge d\eta_{p}-\sqrt{-1}\cdot d\left(\frac{\partial V^{ij}}{\partial\bar{\eta}_{q}}\right)\wedge du_{i}\wedge d\bar{\eta}_{q}\\ =\frac{\sqrt{-1}}{2}\left(\frac{\partial^{2}W^{pq}}{\partial u_{i}\partial u_{j}}+4\frac{\partial^{2}V^{ij}}{\partial\eta_{p}\partial\bar{\eta}_{q}}\right)du_{i}\wedge d\eta_{p}\wedge d\bar{\eta}_{q}=0.

Moreover, if θj\theta_{j} denote the coordinates on the torus fiber such that ∂/∂θj\partial/\partial\theta_{j} are the Hamiltonian vector fields, then the connection 1-forms can be written up to exact forms on B∘B^{\circ} in terms of the local potential Φ\Phi:

(6) Aj=d​θj+−1​(∂2Φ∂uj​∂ηp​d​ηp−∂2Φ∂uj​∂η¯q​d​η¯q).A_{j}=d\theta_{j}+\sqrt{-1}\left(\frac{\partial^{2}\Phi}{\partial u_{j}\partial\eta_{p}}d\eta_{p}-\frac{\partial^{2}\Phi}{\partial u_{j}\partial\bar{\eta}_{q}}d\bar{\eta}_{q}\right).

And one can see explicitly that Fj=d​AjF_{j}=dA_{j}.

The integrability of the complex structure follows from the fact that the differential ideal generated by (1,0)(1,0)-forms is closed:

(7) d⁡(d​zj)=d​Vi​j∧d​ui+−1⋅d​Aj=−∂Vi​j∂uk​d​ui∧d​uk−∂Vi​j∂ηp​d​ui∧d​ηp−∂Vi​j∂η¯q​d​ui∧d​η¯q−(12​∂Wp​q∂uj​d​ηp∧d​η¯q+∂Vi​j∂ηp​d​ui∧d​ηp−∂Vi​j∂η¯q​d​ui∧d​η¯q)=(12​∂Wp​q∂uj​d​η¯q−2​∂Vi​j∂ηp​d​ui)∧d​ηp,d(dz_{j})=dV^{ij}\wedge du_{i}+\sqrt{-1}\cdot dA_{j}\\ =-\frac{\partial V^{ij}}{\partial u_{k}}du_{i}\wedge du_{k}-\frac{\partial V^{ij}}{\partial\eta_{p}}du_{i}\wedge d\eta_{p}-\frac{\partial V^{ij}}{\partial\bar{\eta}_{q}}du_{i}\wedge d\bar{\eta}_{q}\\ -\left(\frac{1}{2}\frac{\partial W^{pq}}{\partial u_{j}}d\eta_{p}\wedge d\bar{\eta}_{q}+\frac{\partial V^{ij}}{\partial\eta_{p}}du_{i}\wedge d\eta_{p}-\frac{\partial V^{ij}}{\partial\bar{\eta}_{q}}du_{i}\wedge d\bar{\eta}_{q}\right)\\ =\left(\frac{1}{2}\frac{\partial W^{pq}}{\partial u_{j}}d\bar{\eta}_{q}-2\frac{\partial V^{ij}}{\partial\eta_{p}}du_{i}\right)\wedge d\eta_{p},

where we have only used ∂Vi​j∂uk=∂Vk​j∂ui\frac{\partial V^{ij}}{\partial u_{k}}=\frac{\partial V^{kj}}{\partial u_{i}}.

It is equally easy to verify the Kähler condition:

(8) d​ω=−d​uj∧d​Aj+−12⋅d​Wp​q∧d​ηp∧d​η¯q=−−1​(12​∂Wp​q∂uj​d​uj∧d​ηp∧d​η¯q)+−12​∂Wp​q∂uj​d​uj∧d​ηp∧d​η¯q=0.d\omega=-du_{j}\wedge dA_{j}+\frac{\sqrt{-1}}{2}\cdot dW^{pq}\wedge d\eta_{p}\wedge d\bar{\eta}_{q}\\ =-\sqrt{-1}\left(\frac{1}{2}\frac{\partial W^{pq}}{\partial u_{j}}du_{j}\wedge d\eta_{p}\wedge d\bar{\eta}_{q}\right)+\frac{\sqrt{-1}}{2}\frac{\partial W^{pq}}{\partial u_{j}}du_{j}\wedge d\eta_{p}\wedge d\bar{\eta}_{q}=0.

Finally, the Ricci-flatness is manifest since det(h)=detV−1⋅detW=1\det(h)=\det V^{-1}\cdot\det W=1 in the complex coordinates d​zj,d​ηpdz_{j},d\eta_{p}. ∎

We are interested in applying the Gibbons-Hawking ansatz to description of the metrics on the toric Calabi-Yau hypersurfaces. Given a torus fibration of such hypersurface near the large complex structure point one may approximate the true Calabi-Yau metric by non-compact solutions of GH equations over different regions of the base. Away from the discriminant locus a semi-flat metric gives a good approximation. We try to argue that as the tori shrink to zero size the metric behavior near singular fibers can also be approximated by certain yy-periodic solutions of GH ansatz. This local description is the main subject of the paper.

2.2. Example: toric orbifold

This is an important toy example which provides the local description of Gibbons-Hawking solutions for more interesting cases. Here for the standard toric orbifold metric one can actually write down an explicit solution to the Gibbons-Hawking equations.

First we set up the notations. Let N≅ℤn+1N\cong\mathbb{Z}^{n+1} be an integral lattice in a real vector space Nℝ=N⊗ℝN_{\mathbb{R}}=N\otimes\mathbb{R}. Denote by N∗⊂Nℝ∗N^{*}\subset N_{\mathbb{R}}^{*} the dual lattice in the dual space.

Let τ\tau be an nn-simplex with vertices (w0,w1,…,wn)(w_{0},w_{1},\dots,w_{n}) in the lattice N≅ℤn+1N\cong\mathbb{Z}^{n+1}, whose affine distance from the origin is 1. That is, there is a vector ρ\rho in the dual lattice N∗N^{*} such that ⟨wi,ρ⟩=1\langle w_{i},\rho\rangle=1, all i=0,…,ni=0,\dots,n. Denote by 𝒯⊂Nℝ\mathcal{T}\subset N_{\mathbb{R}} the cone over τ\tau and by 𝒯∨⊂Nℝ∗\mathcal{T}^{\vee}\subset N_{\mathbb{R}}^{*} the dual cone.

Let X𝒯:=Spec[zm:m∈𝒯∨∩N∗]X_{\mathcal{T}}:=\operatorname{Spec}[z^{m}\ :\ m\in\mathcal{T}^{\vee}\cap N^{*}] be the associated affine toric variety (cf., e.g. [Ful93]). If ℤ⁡⟨w0,…,wn⟩\mathbb{Z}\langle w_{0},\dots,w_{n}\rangle denotes the (finite index) sublattice in NN generated by wiw_{i} and GG is the quotient group N/ℤ⁡⟨w0,…,wn⟩N/\mathbb{Z}\langle w_{0},\dots,w_{n}\rangle, then X𝒯X_{\mathcal{T}} is isomorphic to the orbifold ℂk+1/G\mathbb{C}^{k+1}/G.

The real torus Nℝ/NN_{\mathbb{R}}/N acts on X𝒯X_{\mathcal{T}}. But we will be interested rather in the action of its subtorus Tn:=(Nρ)ℝ/NρT^{n}:=({N_{\rho}})_{\mathbb{R}}/{N_{\rho}}, where Nρ:={v∈N:⟨v,ρ⟩=0}N_{\rho}:=\{v\in N\ :\ \langle v,\rho\rangle=0\}. The nn-dimensional subspace (Nρ)ℝ⊂Nℝ(N_{\rho})_{\mathbb{R}}\subset N_{\mathbb{R}} can be naturally identified with the Lie algebra 𝔱\mathfrak{t} of TnT^{n}. The dual quotient space Nℝ∗/ρN_{\mathbb{R}}^{*}/\rho is identified with 𝔱∗{\mathfrak{t}}^{*}.

Let Qiτ⊂Nℝ∗/ρQ^{\tau}_{i}\subset N_{\mathbb{R}}^{*}/\rho denote the open normal cones to the vertices wiw_{i} of τ\tau. Define a polyhedral complex Π⁡(τ)\Pi(\tau) in Nℝ∗/ρN_{\mathbb{R}}^{*}/\rho to be the union of walls separating the QiτQ^{\tau}_{i}’s:

Π⁡(τ):=⋃i≠jwalli​j,\Pi(\tau):=\bigcup_{i\neq j}\mathrm{wall}_{ij},

with the orientation of each wall determined by the ordering of {i,j}\{i,j\}. Another way to look at Π⁡(τ)\Pi(\tau) is as being the image of the union of (n−1)(n-1)-dimensional cones in 𝒯∨\mathcal{T}^{\vee} under the quotient map Nℝ∗→Nℝ∗/ρN_{\mathbb{R}}^{*}\to N_{\mathbb{R}}^{*}/\rho.

The vector ρ\rho lies in the interior of 𝒯∨\mathcal{T}^{\vee}, hence η=zρ:X𝒯→ℂ\eta=z^{\rho}:X_{\mathcal{T}}\to\mathbb{C} defines a regular function, which vanishes at the divisor in X𝒯X_{\mathcal{T}} corresponding to the boundary of 𝒯∨{\mathcal{T}}^{\vee}. Together with the moment map μ:X𝒯→𝔱∗\mu:X_{\mathcal{T}}\to{\mathfrak{t}}^{*} we have the torus fibration

(μ,η):X𝒯→Nℝ∗/ρ×ℂ,(\mu,\eta):X_{\mathcal{T}}\to N_{\mathbb{R}}^{*}/\rho\times\mathbb{C},

whose restriction to B∘=Nℝ∗/ρ×ℂ∖Π⁡(τ)×{0}B^{\circ}=N_{\mathbb{R}}^{*}/\rho\times\mathbb{C}\setminus\Pi(\tau)\times\{0\} is a principal TnT^{n}-bundle π:M→B∘\pi:M\to B^{\circ}.

To describe the topology of this bundle note that the homology group H2​(Nℝ∗/ρ∖Π⁡(τ),ℤ)H_{2}(N_{\mathbb{R}}^{*}/\rho\setminus\Pi(\tau),\mathbb{Z}) can be naturally identified with Λτ\Lambda_{\tau}, the (finite index) sublattice of NρN_{\rho} generated by the elements wi−wjw_{i}-w_{j} for all pairs of i,ji,j. Then the Chern class of this bundle

[ν]∈H2​(Nℝ∗/ρ∖Π⁡(τ),ℤ)⊗Nρ≅Hom⁡(Λτ,Nρ)[\nu]\in H^{2}(N_{\mathbb{R}}^{*}/\rho\setminus\Pi(\tau),\mathbb{Z})\otimes N_{\rho}\cong\operatorname{Hom}(\Lambda_{\tau},N_{\rho})

is the element given by the natural inclusion ι:Λτ↪Nρ\iota:\Lambda_{\tau}\hookrightarrow N_{\rho}.

The final piece of notation before we describe the standard orbifold metric on X𝒯X_{\mathcal{T}} is the (finite index) sublattice N′⊂NN^{\prime}\subset N generated by wiw_{i}’s. Let (N′)∗⊃N∗(N^{\prime})^{*}\supset N^{*} be its dual lattice. Let m0,…,mnm_{0},\dots,m_{n} be the minimal vectors in (N′)∗(N^{\prime})^{*} along the rays of 𝒯∨{\mathcal{T}}^{\vee}.

In polar coordinates the standard orbifold metric on the algebraic torus (ℂ∗)n+1⊂X𝒯(\mathbb{C}^{*})^{n+1}\subset X_{\mathcal{T}} will be

h=∑i=0n(d​|zmi|+−1​|zmi|​⟨mi,d​θ⟩)⊗(d​|zmi​|−−1|​zmi|​⟨mi,d​θ⟩).h=\sum_{i=0}^{n}(d|z^{m_{i}}|+\sqrt{-1}|z^{m_{i}}|\langle m_{i},d\theta\rangle)\otimes(d|z^{m_{i}}|-\sqrt{-1}|z^{m_{i}}|\langle m_{i},d\theta\rangle).

The functions zmiz^{m_{i}} are defined only on the |G||G|-fold covering space of X𝒯X_{\mathcal{T}}, but |z|mi|z|^{m_{i}} are well defined on X𝒯X_{\mathcal{T}} itself. So are the differential forms ⟨mi,d​θ⟩\langle m_{i},d\theta\rangle.

To write this metric in the Gibbons-Hawking ansatz we choose a basis {ei}\{e_{i}\} of Nρ=𝔱N_{\rho}=\mathfrak{t}. Evaluating the moment map on the basis vectors defines the coordinates ui=μ⁡(ei)u_{i}=\mu(e_{i}) on Nℝ∗/ρ=𝔱∗N_{\mathbb{R}}^{*}/\rho={\mathfrak{t}}^{*}, thus giving an identification Nℝ∗/ρN_{\mathbb{R}}^{*}/\rho with ℝn\mathbb{R}^{n}. The metric on each phase torus T|z|:={z:|z|mi=c​o​n​s​t}≅Nℝ/NT_{|z|}:=\{z\ :\ |z|^{m_{i}}=const\}\cong N_{\mathbb{R}}/N is constant, and, hence, it is given by a quadratic form Q|z|Q_{|z|} on NℝN_{\mathbb{R}}. Let (V−1)i​j(V^{-1})^{ij} be the matrix of restriction of Q|z|Q_{|z|} to (Nρ)ℝ(N_{\rho})_{\mathbb{R}} in the basis {ei}\{e_{i}\}. Then the functions Vi​jV^{ij} and W=detVi​jW=\det V^{ij} give a solution to the GH equations.

Note that the top degree holomorphic form Ωτ\Omega_{\tau} coincide with the push forward under the projection ℂn+1→ℂn+1/G\mathbb{C}^{n+1}\to\mathbb{C}^{n+1}/G of the standard volume form on ℂn+1\mathbb{C}^{n+1}.

As an illustration free of orbifold complications let us write the ansatz for the standard Euclidean metric on ℂn+1\mathbb{C}^{n+1} explicitly. In this case, τ\tau is the standard nn-simplex in N:=ℤn+1N:=\mathbb{Z}^{n+1}, i.e. wiw_{i} form a basis in NN. We will fix the coordinates zi=|zi|​ei​θiz_{i}=|z_{i}|e^{i\theta_{i}} on ℂn+1\mathbb{C}^{n+1}. The action of the torus

Tn={(θ0,…,θn):∑θi=0}T^{n}=\{(\theta_{0},\dots,\theta_{n})\ :\ \sum\theta_{i}=0\}

on ℂn+1\mathbb{C}^{n+1} gives rise to a principal TnT^{n}-bundle over M=ℝn×ℂ∖Π⁡(τ)×{0}M=\mathbb{R}^{n}\times\mathbb{C}\setminus\Pi(\tau)\times\{0\}. Then, in the Gibbons-Hawking coordinates the metric on MM can be written as

h=(V−1)i​j​(Vk​i​d​uk+−1⋅Ai)⊗(Vk​j​d​uk−−1⋅Aj)+W​d​η⊗d​η¯,h=(V^{-1})^{ij}(V^{ki}du_{k}+\sqrt{-1}\cdot A_{i})\otimes(V^{kj}du_{k}-\sqrt{-1}\cdot A_{j})+Wd\eta\otimes d\bar{\eta},

where

ui=|zi|2−|z0|2,i=1,…,n,η=z0z1…zn,\displaystyle u_{i}=|z_{i}|^{2}-|z_{0}|^{2},\ i=1,\dots,n,\qquad\eta=z_{0}z_{1}\dots z_{n},
W−1=|z0​z1​…​zn|2​(1|z0|2+1|z1|2+⋯+1|zn|2),\displaystyle W^{-1}=|z_{0}z_{1}\dots z_{n}|^{2}\left(\frac{1}{|z_{0}|^{2}}+\frac{1}{|z_{1}|^{2}}+\dots+\frac{1}{|z_{n}|^{2}}\right),
(V−1)i​j=|z0|2+δi​j​|zi|2,\displaystyle(V^{-1})^{ij}=|z_{0}|^{2}+\delta^{ij}|z_{i}|^{2},
Aj=d​θj−W​|z0​z1​…​zk^​…​zn|2⋅d⁡(θ0+θ1+⋯+θn).\displaystyle A_{j}=d\theta_{j}-W|z_{0}z_{1}\dots\widehat{z_{k}}\dots z_{n}|^{2}\cdot d(\theta_{0}+\theta_{1}+\dots+\theta_{n}).

The above expressions degenerate whenever two or more of the coordinates ziz_{i} vanish. Thus, the discriminant locus D⊂ℝn×ℂD\subset\mathbb{R}^{n}\times\mathbb{C} is given by u∈Π⁡(τ)u\in\Pi(\tau) and η=0\eta=0. However, when written in the Euclidean coordinates the metric extends from MM to the standard flat metric h=∑i=0nd​zi⊗d​z¯ih=\sum_{i=0}^{n}dz_{i}\otimes d\bar{z}_{i} on ℂn+1\mathbb{C}^{n+1}.

2.3. Non-flat orbifold metrics

We start with a TnT^{n}-torus bundle π:M→Rn×ℂ∖Π⁡(τ)×{0}\pi:M\to R^{n}\times\mathbb{C}\setminus\Pi(\tau)\times\{0\} which has the same topology as the orbifold bundle above. It is convenient to encode the topological information about the bundle by rewriting the equation (5) for the curvature in a distributional form. Let γτj​(u)\gamma^{j}_{\tau}(u) be the NρN_{\rho}-valued 1-current supported on Π⁡(τ)\Pi(\tau) defined by

(9) γτ​(α)=∑i,j(wi−wj)​∫walli​jα,\gamma_{\tau}(\alpha)=\sum\limits_{i,j}(w_{i}-w_{j})\!\int\limits_{\mathrm{wall}_{ij}}\!\alpha,

for an (n−1)(n-1)-form α\alpha. Then adding the distributional equation

(10) −14​π​(∂2W∂ui​∂uj+4​∂2Vi​j∂η​∂η¯)​d​ui∧d​η∧d​η¯=γτj​(u)∧δ⁡(η)\frac{\sqrt{-1}}{4\pi}\left(\frac{\partial^{2}W}{\partial u_{i}\partial u_{j}}+4\frac{\partial^{2}V^{ij}}{\partial\eta\partial\bar{\eta}}\right)du_{i}\wedge d\eta\wedge d\bar{\eta}=\gamma^{j}_{\tau}(u)\wedge\delta(\eta)

to the Gibbons-Hawking ansatz will automatically guarantee that the fibration π:M→Rn×ℂ∖Π⁡(τ)×{0}\pi:M\to R^{n}\times\mathbb{C}\setminus\Pi(\tau)\times\{0\} has the right Chern class. Here δ⁡(η)\delta(\eta) stands for the two-current associated to the origin in ℂ\mathbb{C} (the Dirac delta-function).

A remark on notation: γτ\gamma_{\tau} and δ\delta in (10) really mean the pull back of the corresponding currents to the product ℝn×ℂ\mathbb{R}^{n}\times\mathbb{C}. We will continue to abuse this notation throughout the rest of the paper when there is no confusion possible.

Lemma 2.2.

Suppose we have a Gibbons-Hawking solution on Rn×ℂ∖Π⁡(τ)×{0}R^{n}\times\mathbb{C}\setminus\Pi(\tau)\times\{0\}, that is, a positive definite matrix function locally given by Vi​j=∂2Φ∂uj​∂ujV^{ij}=\frac{\partial^{2}\Phi}{\partial u_{j}\partial u_{j}} such that W:=detVi​j=−4​∂2Φ∂η​∂η¯W:=\det V^{ij}=-4\frac{\partial^{2}\Phi}{\partial\eta\partial\bar{\eta}}, which satisfy the distributional equation (10) in Rn×ℂR^{n}\times\mathbb{C} and, in addition, ∫0∞Vi​j​(u,η,η¯)​d​ui=∞\int\limits_{0}^{\infty}V^{ij}(u,\eta,\bar{\eta})du_{i}=\infty. Then the total space of the torus bundle π:M→Rn×ℂ∖Π⁡(τ)×{0}\pi:M\to R^{n}\times\mathbb{C}\setminus\Pi(\tau)\times\{0\} can be compactified to the fibration π¯:M¯→Rn×ℂ\bar{\pi}:\bar{M}\to R^{n}\times\mathbb{C} such that M¯\bar{M} is biholomorphic (in the orbifold sense) to X𝒯X_{\mathcal{T}} in a manner which respects the map

η:X𝒯→ℂ.\eta:X_{\mathcal{T}}\to\mathbb{C}.

In particular, such solution defines a complete Ricci-flat Kähler metric on the orbifold X𝒯X_{\mathcal{T}} with the standard holomorphic volume form Ωτ\Omega_{\tau}.

Proof.

The discriminant locus D=Π⁡(τ)×{0}D=\Pi(\tau)\times\{0\} is of codimension 3 in Rn×ℂR^{n}\times\mathbb{C} and has a nice simplicial stratification. Using this stratification the topological compactification from MM to M¯=X𝒯\bar{M}=X_{\mathcal{T}} follows by extending the argument of [Gro01, Prop. 2.9] to arbitrary dimensions and including the orbifold singularities. To prove the matching of complex structure we will follow closely [LeB91] where the argument is given for the ℂ2\mathbb{C}^{2} case.

Let ∂∂θj\frac{\partial}{\partial\theta_{j}} be the Hamiltonian vector fields generating the TnT^{n}-action on MM. Consider the commuting vector fields

(11) ξj=−12​(∂∂θj−−1​J​∂∂θj)=12​((V−1)i​j​∂^∂ui−−1​∂∂θj),\xi_{j}=\frac{\sqrt{-1}}{2}\left(\frac{\partial}{\partial\theta_{j}}-\sqrt{-1}J\frac{\partial}{\partial\theta_{j}}\right)=\frac{1}{2}\left((V^{-1})^{ij}\frac{\hat{\partial}}{\partial u_{i}}-\sqrt{-1}\frac{\partial}{\partial\theta_{j}}\right),

where

∂^∂ui=∂∂ui−Ak​(∂∂ui)​∂∂θk,\frac{\hat{\partial}}{\partial u_{i}}=\frac{\partial}{\partial u_{i}}-A_{k}\left(\frac{\partial}{\partial u_{i}}\right)\frac{\partial}{\partial\theta_{k}},

denote the horizontal lifts of the ∂∂ui\frac{\partial}{\partial u_{i}}, and J:T​M→T​MJ:TM\to TM is the complex structure. Since the ∂∂θj\frac{\partial}{\partial\theta_{j}} preserve both the metric and the complex structure, it follows that the ξj\xi_{j} are holomorphic vector fields. Moreover, the their flows are complete because

(12) ∫0∞Vi​j​(u,η,η¯)​d​ui=∞,\int\limits_{0}^{\infty}V^{ij}(u,\eta,\bar{\eta})du_{i}=\infty,

hence, the ξj\xi_{j} generate a holomorphic action of (ℂ∗)n(\mathbb{C}^{*})^{n} on M¯\bar{M}.

The orbit structure of this action is easily seen to be identical with the toric (ℂ∗)n(\mathbb{C}^{*})^{n}-action on the orbifold X𝒯X_{\mathcal{T}}. Namely, for each subspace

La={(u,η):η=a},L_{a}=\{(u,\eta)\ :\ \eta=a\},

the set π−1​(La)\pi^{-1}(L_{a}) is a union of orbits. For a≠0a\neq 0 the π−1​(La)\pi^{-1}(L_{a}) is a single orbit, where as the π−1​(L0)\pi^{-1}(L_{0}) decomposes into (n+1)(n+1) orbits isomorphic to (ℂ∗)n(\mathbb{C}^{*})^{n} and a bunch of smaller dimensional orbits according to the polyhedral decomposition of Nℝ∗/ρN^{*}_{\mathbb{R}}/\rho induced by Π⁡(τ)\Pi(\tau).

Consider the following subsets of Nℝ∗/ρ×ℂN^{*}_{\mathbb{R}}/\rho\times\mathbb{C}:

Li:={(u,η):η≠0}∪{(u,η):u∈Qiτ}.L_{i}:=\{(u,\eta)\ :\ \eta\neq 0\}\cup\{(u,\eta)\ :\ u\in Q^{\tau}_{i}\}.

We can cover the space MM by (n+1)(n+1) open sets 𝒰i:=π−1​(Li)\mathcal{U}_{i}:=\pi^{-1}(L_{i}). Each map η:𝒰i→ℂ\eta:\mathcal{U}_{i}\to\mathbb{C} defines a holomorphic principal (ℂ∗)n(\mathbb{C}^{*})^{n}-bundle over ℂ\mathbb{C}. Since ℂ\mathbb{C} is Stein and contractible we conclude that each 𝒰i\mathcal{U}_{i} is biholomorphic to the product (ℂ∗)n×ℂ(\mathbb{C}^{*})^{n}\times\mathbb{C}.

Now notice that the bundle structures on 𝒰i\mathcal{U}_{i} agree on their intersection π:⋂i𝒰i→ℂ∗\pi:\bigcap_{i}\mathcal{U}_{i}\to\mathbb{C}^{*}. Thus, MM is biholomorphic to the quotient space

∐i((ℂ∗)n×ℂ∗)/∼,\coprod_{i}\left((\mathbb{C}^{*})^{n}\times\mathbb{C}^{*}\right)/\sim,

where the equivalence relation is of the form

(13) (ui,η)i∼(fi​j​(η)​u,η)j,0≤i,j≤n,(u_{i},\eta)_{i}\sim(f_{ij}(\eta)u,\eta)_{j},\quad 0\leq i,j\leq n,

for some cocycle fi​jf_{ij} with values in the holomorphic maps ℂ∗→Aut⁡(ℂ∗)n\mathbb{C}^{*}\to\operatorname{Aut}(\mathbb{C}^{*})^{n}. But any automorphism of the principal homogeneous space (ℂ∗)n(\mathbb{C}^{*})^{n} is given by an nn-tuple of non-zero complex numbers. In other words, each fi​jf_{ij} is an nn-tuple of holomorphic functions ℂ∗→ℂ∗\mathbb{C}^{*}\to\mathbb{C}^{*}. Another choice of trivializations 𝒰i≅(ℂ∗)n×ℂ\mathcal{U}_{i}\cong(\mathbb{C}^{*})^{n}\times\mathbb{C} amounts to modifying the cocycle fi​jf_{ij} by a coboundary. That is, the biholomorphism type of MM is determined by the singularities of fi​j​(η)f_{ij}(\eta) at η=0\eta=0.

Next we notice that if any of the fi​j​(η)f_{ij}(\eta) had an essential singularity at 0, then it would be possible to find a sequence of points in MM converging to several distinct points. Hence MM would not even be Hausdorff. Hence, every singularity has to be removable. But the orders of vanishing of the fi​j​(η)f_{ij}(\eta) can be read off from the residues of the 1-forms d​log⁡fi​jd\log f_{ij}, and this is a topological information given by the Chern class of the original TnT^{n}-bundle. THus, the claimed biholomorphism is established on MM, and it can be extended to M¯\bar{M} by an orbifold version of the Hartog’s theorem.

In order to show that the holomorphic volume form is necessary the standard one we will prove the following (stronger) statement. Given two sets of commuting vector fields ξi\xi_{i} and ζi\zeta_{i} generating the (ℂ∗)(\mathbb{C}^{*})-actions on η:M→ℂ\eta:M\to\mathbb{C} which agree topologically, there is a biholomorphism ϕ:M→M\phi:M\to M such that ϕ∗​ξi=ζi\phi_{*}\xi_{i}=\zeta_{i}. Then, taking XiX_{i} to be the standard action on X𝒯X_{\mathcal{T}} yields the claim about the volume form.

Let {e1,…,en}∈N∗\{e_{1},\dots,e_{n}\}\in N^{*} together with ρ\rho form a basis of N∗N^{*} topologically compatible with ξi\xi_{i} and ζi\zeta_{i}. That is, if {αi,η}\{\alpha_{i},\eta\} and {βi,η}\{\beta_{i},\eta\} are the two sets of coordinates, such that in the respective coordinates ξj=αj​∂∂αj\xi_{j}=\alpha_{j}\frac{\partial}{\partial\alpha_{j}} and ζj=βj​∂∂βj\zeta_{j}=\beta_{j}\frac{\partial}{\partial\beta_{j}} on the big toric orbit {η≠0}≅(ℂ∗)n+1\{\eta\neq 0\}\cong(\mathbb{C}^{*})^{n+1}, then the transition maps between these coordinates and the standard ones {zei,η}\{z^{e_{i}},\eta\} extends from (ℂ∗)n+1(\mathbb{C}^{*})^{n+1} to the whole of MM without zeros/poles. As a consequence, if we write

αj​∂∂αj=Bjk​βk​∂∂βk,\alpha_{j}\frac{\partial}{\partial\alpha_{j}}=B_{j}^{k}\beta_{k}\frac{\partial}{\partial\beta_{k}},

then the matrix BjkB_{j}^{k} is invertible and extends to MM.

The goal is to find a change of coordinates βj=αj​eψj\beta_{j}=\alpha_{j}e^{\psi_{j}} which would induce such a Jacobian matrix. Using the chain rule

αj​∂∂αj=(δjk+αj​∂ψk∂αj)​βk​∂∂βk,\alpha_{j}\frac{\partial}{\partial\alpha_{j}}=\left(\delta^{k}_{j}+\alpha_{j}\frac{\partial\psi_{k}}{\partial\alpha_{j}}\right)\beta_{k}\frac{\partial}{\partial\beta_{k}},

this amounts to solving the differential system

(14) αj​∂ψk∂αj=Bjk−δjk,\alpha_{j}\frac{\partial\psi_{k}}{\partial\alpha_{j}}=B_{j}^{k}-\delta^{k}_{j},

which, in general, is overdetermined.

However, in our case there are several restrictions on BjkB_{j}^{k}. Namely, note that the forms

d​βjβj=Bjk​d​αkαk mod ​d​η\frac{d\beta_{j}}{\beta_{j}}=B^{k}_{j}\frac{d\alpha_{k}}{\alpha_{k}}\quad\text{ mod }d\eta

are closed and have to satisfy the periodicity requirements of the action:

12​π​−1​∫γkd​βjβj=12​π​∫02​πBjk​(α1,…,αk​e−1​θ,…,αn,η)​𝑑θ=δjk,\frac{1}{2\pi\sqrt{-1}}\int_{\gamma_{k}}\frac{d\beta_{j}}{\beta_{j}}=\frac{1}{2\pi}\int_{0}^{2\pi}B^{k}_{j}(\alpha_{1},\dots,\alpha_{k}e^{\sqrt{-1}\theta},\dots,\alpha_{n},\eta)\,d\theta=\delta_{j}^{k},

where γk\gamma_{k} is the generating cycle in the kk-th factor of ℂ∗\mathbb{C}^{*} in (ℂ∗)n(\mathbb{C}^{*})^{n}. Then, by writing BjkB^{k}_{j} in the power series form

(15) Bjk=∑m1​e1+⋯+mn​en+r​ρ∈𝒯∨(bjk)m,r​α1m1​…​αnmn​ηr,mj∈ℤ,B^{k}_{j}=\sum\limits_{m_{1}e_{1}+\dots+m_{n}e_{n}+r\rho\in\mathcal{T}^{\vee}}(b^{k}_{j})_{m,r}\alpha_{1}^{m_{1}}\dots\alpha_{n}^{m_{n}}\eta^{r},\quad m_{j}\in\mathbb{Z},

we conclude that (bjk)m,r=0(b^{k}_{j})_{m,r}=0 for mj=0m_{j}=0, except for (bjk)0,0=δjk(b^{k}_{j})_{0,0}=\delta_{j}^{k}, and (bjk)m,r=0(b^{k}_{j})_{m,r}=0 if there exists an i≠ji\neq j with mi≠0m_{i}\neq 0. Hence, one can simultaneously solve the equations

mj​(ak)m,r=(bjk)m,r,m_{j}(a^{k})_{m,r}=(b^{k}_{j})_{m,r},

for all m,r,k,jm,r,k,j, and write down the power series solution to (14):

ψk=∑m1​e1+⋯+mn​en+r​ρ∈𝒯∨(ak)m,r​α1m1​…​αnmn​ηr,\psi_{k}=\sum\limits_{m_{1}e_{1}+\dots+m_{n}e_{n}+r\rho\in\mathcal{T}^{\vee}}(a^{k})_{m,r}\alpha_{1}^{m_{1}}\dots\alpha_{n}^{m_{n}}\eta^{r},

that converges on the same polydisk in MM as the power series (15) for BjkB^{k}_{j} does. These functions ψk\psi^{k} provide the desired coordinate change. ∎

It is an interesting problem to exhibit the existence (and abundance) of the Gibbons-Hawking solutions. For instance, one can try to deform the standard (flat) orbifold metric. Applying continuity method techniques (cf. [GT83, Ch. 17]) this amounts to inverting a second order linear elliptic differential operator – the linearization of the GH operator at the flat solution. The difficulty is that one of the eigenvalues of its symbol blows off as we approach the discriminant.

In two dimensions (the original GH ansatz [Haw77],[GH78]) the equation becomes the usual Laplace equation. Since there are no positive harmonic functions on ℝ3\mathbb{R}^{3} except constants, any solution has the form

V=ℓ⁡(τ)2​u2+|η|2+a,V=\frac{\ell(\tau)}{2\sqrt{u^{2}+|\eta|^{2}}}+a,

where ℓ⁡(τ)\ell(\tau) is the length of τ\tau, and aa is a positive constant. This defines the famous Taub-NUT metric – the first example of a non-trivial complete Kähler Ricci-flat metric on ℂ2\mathbb{C}^{2} and its quotients by cyclic groups.

2.4. Periodic solutions

The goal here is to set up the Gibbons-Hawking ansatz in such a way that the resulting complex manifold is identifiable with the local model for a Calabi-Yau toric hypersurface. There are no explicit solutions known in dimension higher than 2, unlike the orbifold case. But we will try to make use of the ansatz to get some information about the limiting behavior of solutions in certain degenerations.

We will adapt the notations from the orbifold example. Namely, τ={w0,…,wn}\tau=\{w_{0},\dots,w_{n}\} is a simplex (but now not necessarily of codimension 1) in the lattice N≅ℤn+l+1N\cong\mathbb{Z}^{n+l+1}, which has the distance 1 from the origin. Let σ={v0,…,vl}\sigma=\{v_{0},\dots,v_{l}\} be a simplex in N∗N^{*} such that ⟨σ,τ⟩=1\langle\sigma,\tau\rangle=1. In particular, it means that σ\sigma also has distance 1 from the origin. Let Nσ⊂NN_{\sigma}\subset N and Nτ∗⊂N∗N^{*}_{\tau}\subset N^{*} be the sublattices orthogonal to σ\sigma and τ\tau, respectively. And let

Nℝ∗/σ:=Nℝ∗/⟨v0,…,vl⟩,Nℝ/τ:=Nℝ/⟨w0,…,wn⟩N_{\mathbb{R}}^{*}/\sigma:=N_{\mathbb{R}}^{*}/\langle v_{0},\dots,v_{l}\rangle,\quad N_{\mathbb{R}}/\tau:=N_{\mathbb{R}}/\langle w_{0},\dots,w_{n}\rangle

be the corresponding (dual) quotients. The polyhedral complex Π⁡(τ)\Pi(\tau) provides a polyhedral decomposition of Nℝ∗/σN^{*}_{\mathbb{R}}/\sigma into cells QiτQ_{i}^{\tau} and γτ\gamma_{\tau} is the associated 1-current supported on Π⁡(τ)\Pi(\tau), as before. Also, we define the cone 𝒯:=cone⁡(τ)\mathcal{T}:=\operatorname{cone}(\tau) in NℝN_{\mathbb{R}}, its dual 𝒯∨⊂Nℝ∗\mathcal{T}^{\vee}\subset N^{*}_{\mathbb{R}} and let

X𝒯:=Spec[zm:m∈𝒯∨∩N∗]X_{\mathcal{T}}:=\operatorname{Spec}[z^{m}\ :\ m\in\mathcal{T}^{\vee}\cap N^{*}]

be the associated affine toric variety.

Every vertex vi∈σv_{i}\in\sigma lies in the interior of 𝒯∨\mathcal{T}^{\vee}, and, hence the monomials zviz^{v_{i}} belong to the coordinate ring of X𝒯X_{\mathcal{T}}. Let Zσ,τZ_{\sigma,\tau} denote the closure of the affine hypersurface

{z∈(ℂ∗)k+l+1:∑i=0lzvi=1}\{z\in(\mathbb{C}^{*})^{k+l+1}\ :\ \sum_{i=0}^{l}z^{v_{i}}=1\}

in the toric variety X𝒯X_{\mathcal{T}}.

The real torus Tn:=(Nσ)ℝ/NσT^{n}:=(N_{\sigma})_{\mathbb{R}}/N_{\sigma} acts on X𝒯X_{\mathcal{T}} and leaves the hypersurface Zσ,τZ_{\sigma,\tau} invariant. We assume that this action is a holomorphic isometry and denote by μ:Zσ,τ→Nℝ∗/σ\mu:Z_{\sigma,\tau}\to N_{\mathbb{R}}^{*}/\sigma the corresponding moment map.

The natural inclusion Nτ∗⊂𝒯∨∩N∗N^{*}_{\tau}\subset\mathcal{T}^{\vee}\cap N^{*} gives the projection κ:X𝒯→Tτ\kappa:X_{\mathcal{T}}\to T_{\tau} onto the algebraic torus Tτ:=Spec[zm:m∈Nτ∗]T_{\tau}:=\operatorname{Spec}[z^{m}\ :\ m\in N^{*}_{\tau}]. A choice of ρ∈N∗\rho\in N^{*}, such that ⟨ρ,τ⟩=1\langle\rho,\tau\rangle=1, defines the polynomial

Pσ​(z):=z−ρ​∑i=0lzvi,P_{\sigma}(z):=z^{-\rho}\sum_{i=0}^{l}z^{v_{i}},

which can be thought of as a polynomial in TτT_{\tau}. The zero divisor of PσP_{\sigma} does not depend on the choice of ρ\rho and let Γσ\Gamma_{\sigma} denote the 2-current in TτT_{\tau} associated to it.

We will consider the map (μ,κ):Zσ,τ→Nℝ∗/σ×Tτ(\mu,\kappa):Z_{\sigma,\tau}\to N_{\mathbb{R}}^{*}/\sigma\times T_{\tau} as a torus fibration with the discriminant locus D=Π(τ)×{Pσ(z)=0}D=\Pi(\tau)\times\{P_{\sigma}(z)=0\}. When restricted to a domain B⊂Nℝ∗/σ×TτB\subset N_{\mathbb{R}}^{*}/\sigma\times T_{\tau} it defines a torus fibration Zσ,τ​(B)→BZ_{\sigma,\tau}(B)\to B which is a principal TnT^{n}-bundle over B∘:=B∖DB^{\circ}:=B\setminus D. The Chern class is given, as before, by the inclusion ι:Λτ↪Nσ\iota:\Lambda_{\tau}\hookrightarrow N_{\sigma}.

If the torus action is a holomorphic isometry a Ricci-flat metric on Zσ,τ​(B)Z_{\sigma,\tau}(B) can be written in the Gibbons-Hawking form. Our main goal of this section is to prove the converse. That is if we have a GH solution with the right Chern class then it defines a Ricci-flat metric on Zσ,τ​(B)Z_{\sigma,\tau}(B).

To write everything in coordinates we choose a basis {ei}\{e_{i}\} in NσN_{\sigma} and a basis {mp}\{m_{p}\} in Nτ∗N^{*}_{\tau}. This will defines the coordinates ui:=μ⁡(ei)u_{i}:=\mu(e_{i}) on Nℝ∗/σ≅ℝnN_{\mathbb{R}}^{*}/\sigma\cong\mathbb{R}^{n} and ηp:=log⁡(zmp)=⟨mp,log⁡z⟩\eta_{p}:=\log(z^{m_{p}})=\langle m_{p},\log z\rangle on Tτ≅(ℂ∗)lT_{\tau}\cong(\mathbb{C}^{*})^{l}.

Definition.

Given a domain B{B} in ℝn×(ℂ∗)l\mathbb{R}^{n}\times(\mathbb{C}^{*})^{l} a (σ,τ)(\sigma,\tau)-type solution to the Gibbons-Hawking ansatz in B{B} are two positive definite matrix functions – a real Vi​jV^{ij} and a hermitian Wp​qW^{pq} – on B∘B^{\circ} locally given by a potential:

Vi​j=∂2Φ∂uj​∂uj,Wp​q=−4​∂2Φ∂ηp​∂η¯q,1≤i,j≤n,n+1≤p,q≤n+l,V^{ij}=\frac{\partial^{2}\Phi}{\partial u_{j}\partial u_{j}},\quad W^{pq}=-4\frac{\partial^{2}\Phi}{\partial\eta_{p}\partial\bar{\eta}_{q}},\qquad 1\leq i,j\leq n,\quad n+1\leq p,q\leq n+l,

such that detVi​j=detWp​q\det V^{ij}=\det W^{pq} and the distributional equation

(16) −14​π​(∂2Wp​q∂ui​∂uj+4​∂2Vi​j∂ηp​∂η¯q)​d​ui∧d​ηp∧d​η¯q=γτj​(u)∧Γσ​(η)\frac{\sqrt{-1}}{4\pi}\left(\frac{\partial^{2}W^{pq}}{\partial u_{i}\partial u_{j}}+4\frac{\partial^{2}V^{ij}}{\partial\eta_{p}\partial\bar{\eta}_{q}}\right)du_{i}\wedge d\eta_{p}\wedge d\bar{\eta}_{q}=\gamma^{j}_{\tau}(u)\wedge\Gamma_{\sigma}(\eta)

is satisfied in B{B}.

The topological information about the bundle is again encoded in right hand side of the equation (16).

To state the compatibility with the desired holomorphic volume form we recall (cf., e.g., [Bat93]) that given an affine hypersurface Zf={f(z)=0}⊂(ℂ∗)N+1Z_{f}=\{f(z)=0\}\subset(\mathbb{C}^{*})^{N+1} there is a distinguished top degree holomorphic form ΩC​Y\Omega_{CY} on ZfZ_{f}, which is defined as a Poincaré residue of the meromorphic (N+1)(N+1)-form

d​z0∧⋯∧d​zNf⋅z0​…​zN\frac{dz_{0}\wedge\dots\wedge dz_{N}}{f\cdot z_{0}\dots z_{N}}

on (ℂ∗)N+1(\mathbb{C}^{*})^{N+1} with a single pole along ZfZ_{f}. This form is special in the following sense. If the hypersurface ZfZ_{f} is compactified to a Calabi-Yau hypersurface in a projective toric variety, then ΩC​Y\Omega_{CY} is the restriction of the unique (up to a scalar multiple) non-vanishing holomorphic volume form on the Calabi-Yau manifold (orbifold).

Proposition 2.3.

Given a (σ,τ)(\sigma,\tau)-type Gibbons-Hawking solution on a domain BB, the total space of the torus bundle M→B∘M\to B^{\circ} can be compactified to the fibration M¯→B\bar{M}\to B such that M¯\bar{M} is biholomorphic (in the orbifold sense) to Zσ,τ​(B)Z_{\sigma,\tau}(B) in a manner which respects the fibration

π:Zσ,τ→Tτ.\pi:Z_{\sigma,\tau}\to T_{\tau}.

In particular, such a solution defines a Ricci-flat Kähler (orbifold) metric on Zσ,τ​(B)Z_{\sigma,\tau}(B) with the holomorphic volume form Ω=ΩC​Y\Omega=\Omega_{CY}.

Proof.

The topological compactification again can be drawn from [Gro01, Prop. 2.9]. All we need to show is that a GH solution with asymptotics determined by (16) produces the right complex structure on MM, which would then uniquely extend to M¯\bar{M} by the orbifold version of Hartog’s theorem. But this is a purely local question and it follows directly from Lemma 2.2 dropping the completeness condition (12) that becomes irrelevant.

To see matching of the volume forms let us choose local complex coordinates {η~1,…,η~l}\{\tilde{\eta}_{1},\dots,\tilde{\eta}_{l}\} on (ℂ∗)l(\mathbb{C}^{*})^{l} such that the local equation for {Pσ=0}\{P_{\sigma}=0\} is η~1=0\tilde{\eta}_{1}=0. In these coordinates the top degree holomorphic form on MM will be Ω=Ωτ∧d​η2∧⋯∧ηl\Omega=\Omega_{\tau}\wedge d\eta_{2}\wedge\dots\wedge\eta_{l}, where Ωτ\Omega_{\tau} is the standard orbifold volume form. Then, Ω\Omega is easily seen to coincide with the local expression for the distinguished form ΩC​Y\Omega_{CY} on Zσ,τZ_{\sigma,\tau}. ∎

3. Limiting behavior of solutions

3.1. Exponential decay lemma

We would like to analyze the behavior of Gibbons-Hawking solutions when the tori (both in the fibers and in the base) are shrinking. First, let us introduce a non-linear differential equation of the Monge-Ampère type.

Definition.

We refer to a pair of real positive definite matrix functions Vi​j,Wp​qV^{ij},W^{pq} as a solution to the split Monge-Ampère equation in an open subset R⊂ℝn×ℝlR\subset\mathbb{R}^{n}\times\mathbb{R}^{l} if detVi​j=detWp​q\det V^{ij}=\det W^{pq} and they are locally given by a smooth potential function KK:

(17) Vi​j=∂2K∂si​∂sj,Wp​q=−∂2K∂tp​∂tq,1≤i,j≤n,k+1≤p,q≤n+l.V^{ij}=\frac{\partial^{2}K}{\partial s_{i}\partial s_{j}},\quad W^{pq}=-\frac{\partial^{2}K}{\partial t_{p}\partial t_{q}},\quad 1\leq i,j\leq n,\quad k+1\leq p,q\leq n+l.

To describe the asymptotics at the discriminant we would like to treat the simplex σ⊂N∗\sigma\subset N^{*} on the same footing as τ\tau. Namely, we let Σ⊂Nℝ∗\Sigma\subset N^{*}_{\mathbb{R}} be the cone over σ\sigma, and let Σ∨\Sigma^{\vee} be its dual cone in NℝN_{\mathbb{R}}. The polyhedral complex Π⁡(σ)\Pi(\sigma) provides a polyhedral decomposition of Nℝ/τN_{\mathbb{R}}/\tau into cells QiσQ_{i}^{\sigma}. Denote by γσ\gamma_{\sigma} the Nτ∗N^{*}_{\tau}-valued 1-current defined in the same way as γτ\gamma_{\tau}.

Definition.

Given a domain RR in Nℝ∗/σ×Nℝ/τ≅ℝn×ℝlN^{*}_{\mathbb{R}}/\sigma\times N_{\mathbb{R}}/\tau\cong\mathbb{R}^{n}\times\mathbb{R}^{l} a (σ,τ)(\sigma,\tau)-type singular solution to the split Monge-Ampère equation in RR is a pair of matrix functions Vi​j,Wp​qV^{ij},W^{pq} which are local Monge-Ampère solutions in R∖(Π⁡(τ)×Π⁡(σ))R\setminus(\Pi(\tau)\times\Pi(\sigma)) with asymptotics at the discriminant locus governed by the distributional equation

(18) 12​π​(∂2Wp​q∂si​∂sj+∂2Vi​j∂tp​∂tq)​d​si∧d​tp=γτj​(s)​γσq​(t).\frac{1}{2\pi}\left(\frac{\partial^{2}W^{pq}}{\partial s_{i}\partial s_{j}}+\frac{\partial^{2}V^{ij}}{\partial t_{p}\partial t_{q}}\right)ds_{i}\wedge dt_{p}=\gamma^{j}_{\tau}(s)\gamma^{q}_{\sigma}(t).
Conjecture 3.1 (Exponential decay lemma).

Given a convex domain RR in ℝk×ℝl\mathbb{R}^{k}\times\mathbb{R}^{l} and a (σ,τ)(\sigma,\tau)-type solution V,WV,W of the split Monge-Ampère equation in RR there is a real one-parameter family of (σ,τ)(\sigma,\tau)-solutions Vλ,WλV_{\lambda},W_{\lambda} to the Gibbons-Hawking ansatz in λ​R×(S1)l\lambda R\times(S^{1})^{l} such that

  • •

    The diameter of the circles both in the fiber TnT^{n} and in the torus part (S1)l(S^{1})^{l} of the base away from the discriminant is roughly given by λ−1\lambda^{-1}.

  • •

    The zero Fourier modes of the GH solutions Vλ0​(u,x),Wλ0​(u,x)V_{\lambda}^{0}(u,x),W_{\lambda}^{0}(u,x) as functions of the rescaled variables s,ts,t, where u=λ​s,x=λ​tu=\lambda s,x=\lambda t, will converge (in some properly weighted norm on the function space) to V⁡(s,t),W⁡(s,t)V(s,t),W(s,t) as λ→∞\lambda\to\infty.

  • •

    The higher Fourier modes decay exponentially away from the discriminant Π⁡(τ)×Π⁡(σ)\Pi(\tau)\times\Pi(\sigma) in λ​R\lambda R, uniformly in λ\lambda. That is, if β⁡(u,x)\beta(u,x) denotes the Euclidean distance from the point (u,x)(u,x) to the discriminant, then

    |Vλm​(u,x)|≤C1​e−β⁡(u,x)​|m|,|Wλm​(u,x)|≤C2​e−β⁡(u,x)​|m|,|V_{\lambda}^{m}(u,x)|\leq C_{1}e^{-\beta(u,x)|m|},\quad|W_{\lambda}^{m}(u,x)|\leq C_{2}e^{-\beta(u,x)|m|},

    for some constants C1,C2C_{1},C_{2}, and large enough λ\lambda and β\beta.

We would like to give some easy examples and a rough argument based on those why we believe this conjecture is true. Note, however, that once justified, it will have an important consequence for the metric collapse program for the toric hypersurfaces and complete intersections:

Corollary 3.2.

The metric space (Zσ,τ​(λ−1​R),λ−2​gλ)(Z_{\sigma,\tau}(\lambda^{-1}R),\lambda^{-2}g_{\lambda}), where gλg_{\lambda} is the Riemannian (orbifold) metric from the Gibbons-Hawking ansatz, converges in the Gromov-Hausdorff sense to (R,g∞i​j)(R,g^{ij}_{\infty}), with the limiting metric g∞i​j=Vi​j​d​si​d​sj+Wp​q​d​tp​d​tqg^{ij}_{\infty}=V^{ij}ds_{i}ds_{j}+W^{pq}dt_{p}dt_{q}.

3.2. The semi-flat case

We consider the case when either l=0l=0, or k=0k=0. In both situations the discriminant locus is empty and the total space MM is just the product of the domain RR and the torus TnT^{n}. We can use any solution of the classical real Monge-Ampère equation in RR and extend it to a Gibbons-Hawking solution on MM by setting higher Fourier modes to zero. In the obvious complex structure this will give a Ricci-flat metric on M=M¯M=\bar{M} (cf. [Hit97], [Leu00], [LYZ01]).

3.3. Two dimensional example: local K3 (after [OV96] and [GW00])

This is the periodic version of the original Gibbons-Hawking ansatz [GH78],[Haw77]. We consider the case when k=l=1k=l=1 and both simplices τ\tau and σ\sigma are of length 1, although the construction works for a non-unimodular case as well.

The Gibbons-Hawking equation in this case is equivalent to the Laplace equation for V⁡(u,x,y)V(u,x,y) (=W⁡(u,x,y)=W(u,x,y)) on a domain in the cylinder ℝ×ℝ×S1\mathbb{R}\times\mathbb{R}\times S^{1} with the Dirac δ\delta-function on the right hand side. We can write both the solution V⁡(u,x,y)V(u,x,y) and the δ\delta-function in the Fourier expansion:

V(u,x,y)=∑m∈ℤVme2​π​i​m​y,γτj(u)δPσ(η)=−δ(u,x,y)=−∑m∈ℤδ(u,x)e2​π​i​m​y.V(u,x,y)=\sum_{m\in\mathbb{Z}}V_{m}e^{2\pi imy},\quad\gamma^{j}_{\tau}(u)\delta_{P_{\sigma}}(\eta)=-\delta(u,x,y)=-\sum_{m\in\mathbb{Z}}\delta(u,x)e^{2\pi imy}.

Here the minus sign takes into account the orientation of the circle action when passing from currents to generalized functions.

Being linear, the Gibbons-Hawking equation

∂2Vλ∂u2+∂2Vλ∂x2+∂2Vλ∂y2=−2π⋅δ(u,x,y)\frac{\partial^{2}V_{\lambda}}{\partial u^{2}}+\frac{\partial^{2}V_{\lambda}}{\partial x^{2}}+\frac{\partial^{2}V_{\lambda}}{\partial y^{2}}=-2\pi\cdot\delta(u,x,y)

will decompose into the Helmholtz equations according to the Fourier modes:

∂2Vλm∂u2+∂2Vλm∂x2−(2πm)2Vλm=−2π⋅δ(u,x),m∈ℤ.\frac{\partial^{2}V_{\lambda}^{m}}{\partial u^{2}}+\frac{\partial^{2}V_{\lambda}^{m}}{\partial x^{2}}-(2\pi m)^{2}V_{\lambda}^{m}=-2\pi\cdot\delta(u,x),\quad m\in\mathbb{Z}.

On the other hand, the (σ,τ)(\sigma,\tau)-type split Monge-Ampère equation in the rescaled coordinates s=λ−1​u,t=λ−1​us=\lambda^{-1}u,t=\lambda^{-1}u is the two-dimensional Laplace equation:

∂2V∂s2+∂2V∂t2=−2π⋅δ(s,t),\frac{\partial^{2}V}{\partial s^{2}}+\frac{\partial^{2}V}{\partial t^{2}}=-2\pi\cdot\delta(s,t),

whose fundamental solutions are in the form V⁡(s,t)=−12​log⁡|s2+t2|+h⁡(s,t)V(s,t)=-\frac{1}{2}\log|s^{2}+t^{2}|+h(s,t), for a harmonic function hh. Thus, one can take the zero mode of the corresponding Gibbons-Hawking solution to be Vλ0​(u,x)=V⁡(λ−1​u,λ−1​x)V_{\lambda}^{0}(u,x)=V(\lambda^{-1}u,\lambda^{-1}x), as long as V⁡(s,t)V(s,t) stays positive on RR. As for the higher modes, it is known that a fundamental solution to the Helmholtz equation with m≠0m\neq 0 may be given by the Bessel function

Vλm=K0​(2​π​|m|​r)∼14​|m|​r​e−2​π|m|r​(1+O⁡(r−1)), where ​r2=u2+x2,V_{\lambda}^{m}=K_{0}(2\pi|m|r)\sim\frac{1}{\sqrt{4|m|r}}\,e^{-{2\pi|m|r}}\left(1+O(r^{-1})\right),\text{ where }r^{2}=u^{2}+x^{2},

which decays exponentially as required.

3.4. Higher dimensional case

The full proof of the conjecture in this general case will probably require some very non-trivial application of the continuity method to deform the given split solution, then introduce exponentially small higher modes and do some clever estimates afterwards. Meanwhile, we want to indicate a rough argument why some of the ideas from the K3 example above may still work in general.

To have the Fourier modes of the solutions defined on the same domain, independent of λ\lambda, we can scale the variables by λ\lambda:

si=uiλ,tp=xpλ,yp~=ypλ.s_{i}=\frac{u_{i}}{\lambda},\quad t_{p}=\frac{x_{p}}{\lambda},\quad\tilde{y_{p}}=\frac{y_{p}}{\lambda}.

Then the GH solutions are on R×(S1/λ)lR\times(S^{1}/\lambda)^{l} for all λ\lambda, and their Fourier modes are functions on RR. To keep up with the complex structure the logarithmic map (ℂ∗)l→ℝl(\mathbb{C}^{*})^{l}\to\mathbb{R}^{l} has to scale by λ\lambda as well:

logeλ⁡(z1,…,zn):=1λ​(log⁡|z1|,…,log⁡|zn|).\log_{e^{\lambda}}(z_{1},\dots,z_{n}):=\frac{1}{\lambda}(\log|z_{1}|,\dots,\log|z_{n}|).

We would like to recall a few basic facts from “tropical” geometry (cf., e.g., [Mik01]). Given a polynomial Pσ​(z)P_{\sigma}(z) in (ℂ∗)l(\mathbb{C}^{*})^{l} the amoeba 𝒜σλ\mathcal{A}^{\lambda}_{\sigma} is defined to be the image of the rescaled log map:

𝒜σλ:=logeλ({Pσ=0)}⊂ℝl.\mathcal{A}^{\lambda}_{\sigma}:={\log_{e^{\lambda}}(\{P_{\sigma}=0)\}}\subset\mathbb{R}^{l}.

As λ→∞\lambda\to\infty the amoeba 𝒜σλ\mathcal{A}^{\lambda}_{\sigma} approaches its spine 𝒜σ∞=Π⁡(σ)\mathcal{A}^{\infty}_{\sigma}=\Pi(\sigma). The Ronkin function

Nσ​(x):=1(2​π​−1)l​∫log⁡|z|=xlog⁡|Pσ​(z)|2​d​z1z1∧⋯∧d​zlzlN_{\sigma}(x):=\frac{1}{(2\pi\sqrt{-1})^{l}}\int\limits_{\log|z|=x}\!\log|P_{\sigma}(z)|^{2}\ \frac{dz_{1}}{z_{1}}\wedge\dots\wedge\frac{dz_{l}}{z_{l}}

is defined up to a linear function, which depends on a particular choice of ρ∈N∗\rho\in N^{*} used in the definition of the polynomial PσP_{\sigma}. Denote by

Nσλ​(t):=1λ​Nσ​(λ​t)N_{\sigma}^{\lambda}(t):=\frac{1}{\lambda}N_{\sigma}(\lambda t)

the rescaled Ronkin function. The point is that Nσλ​(t)N^{\lambda}_{\sigma}(t) is a continuous function, linear on each connected component of ℝl∖Aσλ\mathbb{R}^{l}\setminus A^{\lambda}_{\sigma}, with the slopes given by the viv_{i}’s. As λ→∞\lambda\to\infty, it converges to the piece-wise linear function Nσ∞​(t)N_{\sigma}^{\infty}(t) whose corner locus is Π⁡(σ)\Pi(\sigma) with the viv_{i}-slopes over the QiσQ^{\sigma}_{i}.

We would like to analyze the right hand side of the equation (16) written in the Fourier expansion. The factor γτ\gamma_{\tau} carries over to every mode, while

Γσ=−12​π​∂∂¯​log⁡|Pσ|2\Gamma_{\sigma}=\frac{\sqrt{-1}}{2\pi}\partial\bar{\partial}\log|P_{\sigma}|^{2}

decomposes into currents supported on 𝒜σλ\mathcal{A}^{\lambda}_{\sigma}. In particular, since the exterior differentiation commutes with averaging, we conclude that the zero mode of Γσ\Gamma_{\sigma} is given by the Hessian of the Ronkin function:

Γσ0=1(2​π​−1)l​∫log⁡|z|=xΓσ​d​z1z1∧⋯∧d​zlzl=−12​π⋅1(2​π​−1)l∂∂¯∫log⁡|z|=xlog|Pσ|2d​z1z1…d​zlzl=∂2Nσ​(x)∂xp​∂xqdxp∧dyq.\Gamma_{\sigma}^{0}=\frac{1}{(2\pi\sqrt{-1})^{l}}\!\int\limits_{\log|z|=x}\!\Gamma_{\sigma}\ \frac{dz_{1}}{z_{1}}\wedge\dots\wedge\frac{dz_{l}}{z_{l}}=\\ \frac{\sqrt{-1}}{2\pi}\cdot\frac{1}{(2\pi\sqrt{-1})^{l}}\ \partial\bar{\partial}\!\int\limits_{\log|z|=x}\!\log|P_{\sigma}|^{2}\ \frac{dz_{1}}{z_{1}}\dots\frac{dz_{l}}{z_{l}}=\frac{\partial^{2}N_{\sigma}(x)}{\partial x_{p}\partial x_{q}}dx_{p}\wedge dy_{q}.

But substituting x=λ​tx=\lambda t yields

∂2Nσ​(x)∂xp​∂xq​d​xp=∂2Nσλ​(t)∂tp​∂tq​d​tp.\frac{\partial^{2}N_{\sigma}(x)}{\partial x_{p}\partial x_{q}}dx_{p}=\frac{\partial^{2}N^{\lambda}_{\sigma}(t)}{\partial t_{p}\partial t_{q}}dt_{p}.

Hence, as λ→∞\lambda\to\infty, the current Γσ0\Gamma_{\sigma}^{0} converges to γσ​d​yq\gamma_{\sigma}dy_{q}.

As for the higher modes, we note that since y~\tilde{y} is now (2​π​λ−1)(2\pi\lambda^{-1})-periodic, there is a factor of λ2\lambda^{2} in the zero order term of the Helmholtz-type equation for m≠0m\neq 0. By analogy with the Bessel functions we hope that the spectral theory will force the higher modes decay exponentially away from the locus 𝒜σλ×Π⁡(τ)\mathcal{A}^{\lambda}_{\sigma}\times\Pi(\tau) with the exponent now multiplied by an arbitrary large number λ\lambda.

One can go about proving the lemma by starting with the given split Monge-Ampère solution and constructing a family of solutions but with the factor γσ​d​yq\gamma_{\sigma}dy_{q} in the right hand side being replaced by a more regular Γσ0\Gamma_{\sigma}^{0}. This will give a family of semi-flat Gibbons-Hawking solutions. Then one can argue that since the higher modes can be taken exponentially small, they may be considered, in some sense, as perturbation of the semi-flat solution.

4. Local mirror symmetry and Legendre transform

4.1. Linear algebra of Legendre transform and Monge-Ampére equations

The classical fact, implicitly used in [GW00], is that solving the two-dimensional Laplace equation is equivalent to solving the real two-dimensional Monge-Ampère equation. This can be easily generalized to higher dimensions. Namely, the chain rule and elementary linear algebra for a particular coordinate transformation yields the following.

Lemma 4.1.

If K⁡(s,t)K(s,t) is a local solution to the split Monge-Ampère equation (17), then Ψ⁡(y)\Psi(y) is a (local) solution of the classical (real) Monge-Ampère equation

(19) det∂2Ψ∂yi​∂yp=1,1≤i,j≤n,\det\frac{\partial^{2}\Psi}{\partial y_{i}\partial y_{p}}=1,\quad 1\leq i,j\leq n,

where

(20) yi=∂K∂si, 1≤i≤k,yp=tp,k+1≤p≤n,y_{i}=\frac{\partial K}{\partial s_{i}},\ 1\leq i\leq k,\qquad y_{p}=t_{p},\ k+1\leq p\leq n,

and Ψ⁡(y)\Psi(y) is the partial Legendre transform of K⁡(s,t)K(s,t), defined by

(21) ∂Ψ∂yi=si, 1≤i≤k,∂Ψ∂yp=−∂K∂tp,k+1≤p≤n.\frac{\partial\Psi}{\partial y_{i}}=s_{i},\ 1\leq i\leq k,\qquad\frac{\partial\Psi}{\partial y_{p}}=-\frac{\partial K}{\partial t_{p}},\ k+1\leq p\leq n.
Proof.

First, we check the very existence of the partial Legendre transform. Consider the following Jacobian and Hessian matrices:

(22) ∂(yi,yp)∂(s,t)=(∂2K∂sj​∂si∂2K∂tq​∂si0𝟙)=(VB0𝟙),Hess⁡K⁡(s,t)=(VBBt−W).\frac{\partial(y_{i},y_{p})}{\partial(s,t)}=\left(\begin{array}[]{cc}\frac{\partial^{2}K}{\partial s_{j}\partial s_{i}}&\frac{\partial^{2}K}{\partial t_{q}\partial s_{i}}\\ 0&\mathbbm{1}\end{array}\right)=\left(\begin{array}[]{cc}V&B\\ 0&\mathbbm{1}\end{array}\right),\quad\operatorname{Hess}K(s,t)=\left(\begin{array}[]{cc}V&B\\ {}^{t}B&-W\end{array}\right).

Then

(23) ∂(s,t)∂(yi,yp)=(V−1−V−1​B0𝟙),Hess⁡Ψ⁡(y)=(V−1−V−1​Bt(−V−1B)W+Bt​V−1​B).\frac{\partial(s,t)}{\partial(y_{i},y_{p})}=\left(\begin{array}[]{cc}V^{-1}&-V^{-1}B\\ 0&\mathbbm{1}\end{array}\right),\quad\operatorname{Hess}\Psi(y)=\left(\begin{array}[]{cc}V^{-1}&-V^{-1}B\\ {}^{t}(-V^{-1}B)&W+{{}^{t}B}V^{-1}B\end{array}\right).

This shows that if both VV and WW are symmetric and positive definite, then Hess⁡Ψ\operatorname{Hess}\Psi is also a symmetric positive definite matrix, so there exists (locally) a convex function Ψ\Psi – the partial Legendre transform.

On the other hand the inverse of a non-degenerate 2×22\times 2 block matrix with detA≠0\det A\neq 0 and detD≠0\det D\neq 0 is:

(24) (ABCD)−1=((A−B​D−1​C)−1(−A+B​D−1​C)−1​B​D−1(−D+C​A−1​B)−1​C​A−1(D−C​A−1​B)−1),\left(\begin{array}[]{cc}A&B\\ C&D\end{array}\right)^{-1}=\left(\begin{array}[]{cc}(A-BD^{-1}C)^{-1}&(-A+BD^{-1}C)^{-1}BD^{-1}\\ (-D+CA^{-1}B)^{-1}CA^{-1}&(D-CA^{-1}B)^{-1}\end{array}\right),

which implies that

(25) det(ABCD)=1⟺detA−1=det(D−CA−1B).\det\left(\begin{array}[]{cc}A&B\\ C&D\end{array}\right)=1\quad\Longleftrightarrow\quad\det A^{-1}=\det(D-CA^{-1}B).

Applied to Hess⁡Ψ\operatorname{Hess}\Psi the last observation shows that Ψ\Psi is a local Monge-Ampère solution if and only if detV=detW\det V=\det W. ∎

4.2. Monge-Ampère manifolds

Definition.

A Riemannian manifold (Y,g)(Y,g) is called Monge-Ampère if it possesses an integral affine structure and the metric is, locally in affine coordinates, given by a potential Ψ\Psi, that is, gi​j=∂2Ψ∂yi​∂yjg_{ij}=\frac{\partial^{2}\Psi}{\partial y_{i}\partial y_{j}}, that satisfies the real Monge-Ampère equation detgi​j=1\det g_{ij}=1.

Cheng and Yau [CY82] proved that every compact Monge-Ampère manifold is diffeomorphic to a torus and the Monge-Ampère structure is a deformation of the standard flat structure on ℝn/ℤn\mathbb{R}^{n}/\mathbb{Z}^{n}. To enrich this fairly boring class of manifolds we will allow certain singularities along a codimension 2 discriminant locus.

The Monge-Ampère condition in the affine geometry is an analog of the Ricci-flatness condition on a Kähler manifold with similar global rigidity properties. Roughly speaking, one should expect some sort of the Calabi conjecture saying that each topological class of the metric has a unique Monge-Ampère representative (cf. [HZ03] and [KT02]).

Here we are concerned with the local picture in which many Monge-Ampère structures may exist. However, for applications to the metric collapse of toric Calabi-Yau hypersurfaces we will be interested only in a certain special class of the Monge-Ampère structures, called bi-PIKAS in [HZ03]. We will show that the latter always arise from singular split Monge-Ampère solutions of the corresponding type.

The base (ℝn×ℝl)∖(Π⁡(τ)×Π⁡(σ))(\mathbb{R}^{n}\times\mathbb{R}^{l})\setminus(\Pi(\tau)\times\Pi(\sigma)) has an open covering {Uvi,Uwj}\{U_{v_{i}},U_{w_{j}}\} where Uvi:=ℝn×QiσU_{v_{i}}:=\mathbb{R}^{n}\times Q^{\sigma}_{i} and Uwj:=Qjτ×ℝlU_{w_{j}}:=Q^{\tau}_{j}\times\mathbb{R}^{l}. The nerve of this covering is the complete bipartite graph on the vertices viv_{i} of σ\sigma and wjw_{j} of τ\tau. We assume that the affine structure on (ℝn×ℝl)∖(Π⁡(τ)×Π⁡(σ))(\mathbb{R}^{n}\times\mathbb{R}^{l})\setminus(\Pi(\tau)\times\Pi(\sigma)) is polyhedral of type ({Uv},∂Σ∨)(\{U_{v}\},\partial\Sigma^{\vee}). That is, there is a homeomorphism ℝn×ℝl→∂Σ∨\mathbb{R}^{n}\times\mathbb{R}^{l}\to\partial\Sigma^{\vee} which provides affine coordinates on every chart UviU_{v_{i}} by identifying it with the maximal dimensional face of ∂Σ∨\partial\Sigma^{\vee} orthogonal to viv_{i}. Also we assume that the dual affine structure is polyhedral of type ({Uw},∂𝒯∨)(\{U_{w}\},\partial\mathcal{T}^{\vee}). Moreover, the transformation maps between the affine coordinates on UviU_{v_{i}} and UwjU_{w_{j}} are assumed to be the natural projections from the subspaces vi⟂:={y∈Nℝ:⟨vi,y⟩=0}v_{i}^{\perp}:=\{y\in N_{\mathbb{R}}\ :\ \langle v_{i},y\rangle=0\} to the quotients Nℝ/wjN_{\mathbb{R}}/w_{j}. This data constitutes a bi-polyhedral integral Kähler affine structure (bi-PIKAS for short) on (ℝn×ℝl)∖(Π⁡(τ)×Π⁡(σ))(\mathbb{R}^{n}\times\mathbb{R}^{l})\setminus(\Pi(\tau)\times\Pi(\sigma)) of type (∂Σ∨,∂𝒯∨)(\partial\Sigma^{\vee},\partial\mathcal{T}^{\vee}) (cf. [HZ03]). We will abbreviate the type to be (σ,τ)(\sigma,\tau).

It is straight forward to see that the monodromy of the (linear part of the) affine structure along a loop (vi1​wj1​vi2​wj2)(v_{i_{1}}w_{j_{1}}v_{i_{2}}w_{j_{2}}) is given by y↦y+⟨vi2−vi1,y⟩​(wj2−wj1)y\mapsto y+\langle v_{i_{2}}-v_{i_{1}},y\rangle(w_{j_{2}}-w_{j_{1}}) (cf. [HZ02] for details). Since the radiance obstruction class vanishes locally (cf. [GS02]) we can always choose the affine coordinates such that there is no translational part of the monodromy.

Given a convex domain RR in ℝn×ℝl\mathbb{R}^{n}\times\mathbb{R}^{l} which contains the origin, a Monge-Ampère bi-PIKAS on RR of type (σ,τ)(\sigma,\tau) will be a restriction of a Monge-Ampère bi-PIKAS of the same type on ℝn×ℝl\mathbb{R}^{n}\times\mathbb{R}^{l}.

Theorem 4.2.

There is a bijection between the sets of (σ,τ)(\sigma,\tau)-type split Monge-Ampère solutions on a domain RR and (σ,τ)(\sigma,\tau)-type Monge-Ampère bi-PIKAS on RR. The Riemannian metric on R∘:=R∖Π⁡(s​i​g​m​a)×Π⁡(τ)R^{\circ}:=R\setminus\Pi(sigma)\times\Pi(\tau) is given by Vi​j​d​si​d​sj+Wp​q​d​tp​d​tqV^{ij}ds_{i}ds_{j}+W^{pq}dt_{p}dt_{q}, where (Vi​j,Wp​q)(V^{ij},W^{pq}) is the corresponding split MA solution.

Proof.

Let KαK_{\alpha} be local potentials for a given split Monge-Ampère solution (V,W)(V,W). We will use the partial Legendre transform to define new coordinates yi,ypy_{i},y_{p} as in Lemma 4.1. We claim that these are affine coordinates, that is the transition maps are affine linear.

Indeed, by comparing the two local potentials in the overlap Uα∩UβU_{\alpha}\cap U_{\beta} we see that Kα−KβK_{\alpha}-K_{\beta} has to be in the form f⁡(s)​g​(t)f(s)g(t) where both f⁡(s)f(s) and g⁡(t)g(t) are affine functions. Thus yα−yβy_{\alpha}-y_{\beta} are affine functions of tpt_{p}, hence affine functions of ypy_{p}.

Lemma 4.1 also guarantees that Ψα\Psi_{\alpha} are local Monge-Ampère potentials in the affine coordinates yy. The polyhedral property follows from noticing that the charts UvU_{v} are bounded by linear inequalities in tpt_{p}, and hence by (the same) linear inequalities in the new (affine) coordinates ypy_{p}.

Applying the partial Legendre transform to the other half of the variables (s,t)(s,t) gives the dual MA structure on R∘R^{\circ}. Same considerations as above show that the dual affine structure is polyhedral of type ({Uw},∂𝒯∨)(\{U_{w}\},\partial\mathcal{T}^{\vee}).

The final step is to show that the resulting Monge-Ampère bi-PIKAS has the right type, that is, to compute its monodromy around a loop (vi1​wj1​vi2​wj2)(v_{i_{1}}w_{j_{1}}v_{i_{2}}w_{j_{2}}). For this we consider the closed (Nτ∗)ℝ⊗(Nσ)ℝ(N^{*}_{\tau})_{\mathbb{R}}\otimes(N_{\sigma})_{\mathbb{R}}-valued 1-form on R∘R^{\circ}:

βi​q=∂Wp​q∂si​d​tp−∂2Vi​j∂tq​d​sj.\beta^{iq}=\frac{\partial W^{pq}}{\partial s_{i}}dt_{p}-\frac{\partial^{2}V^{ij}}{\partial t_{q}}ds_{j}.

Making use of the distributional equation (18) and applying Stokes’ theorem we can compute the integral of β\beta along the loop (vi1​wj1​vi2​wj2)=:∂S(v_{i_{1}}w_{j_{1}}v_{i_{2}}w_{j_{2}})=:\partial S:

∮β=∬S𝑑β=2​π​∬Sγσ​γτ=2​π​(vi2−vi1)⊗(wj2−wj1).\oint\beta=\iint_{S}d\beta=2\pi\iint_{S}\gamma_{\sigma}\gamma_{\tau}=2\pi(v_{i_{2}}-v_{i_{1}})\otimes(w_{j_{2}}-w_{j_{1}}).

Thus, β\beta has holonomy which depends only on the homotopy class of the path. That is, we can think of β\beta as given by the differential of a multi-valued function

fi​q=∂2K∂si​∂tq.f^{iq}=\frac{\partial^{2}K}{\partial s_{i}\partial t_{q}}.

The affine coordinates yp,n+1≤p≤n+l,y_{p},\ n+1\leq p\leq n+l, are single valued. Hence, there is no monodromy in d​ypdy_{p}. On the other hand, the ambiguity in the remaining differentials

d​yi=∂2K∂si​∂tq​d​tq+∂2K∂si​∂sj​d​sjdy_{i}=\frac{\partial^{2}K}{\partial s_{i}\partial t_{q}}dt_{q}+\frac{\partial^{2}K}{\partial s_{i}\partial s_{j}}ds_{j}

is coming exactly from the multi-valuedness of fi​qf^{iq}. Thus, the monodromy around the loop (vi1​wj1​vi2​wj2)(v_{i_{1}}w_{j_{1}}v_{i_{2}}w_{j_{2}}) is given by 𝟙+2​π​(vi2−vi1)⊗(wj2−wj1)\mathbbm{1}+2\pi(v_{i_{2}}-v_{i_{1}})\otimes(w_{j_{2}}-w_{j_{1}}). The factor of 2​π2\pi can be absorbed into a redefinition of the affine coordinates.

Tracing backwards through the above argument shows the converse statement is also true. Namely, given a Monge-Ampère bi-PIKAS of type (σ,τ)(\sigma,\tau) we can use it together with its dual to define single-valued coordinates (s,t)(s,t) on R∘R^{\circ} which will obviously extend to RR. Moreover, due to the polyhedral properties of these bi-PIKAS the discriminant locus in RR will be exactly given by ∂𝒰v∩∂𝒰w=Π⁡(σ)×Π⁡(τ)\partial\mathcal{U}_{v}\cap\partial\mathcal{U}_{w}=\Pi(\sigma)\times\Pi(\tau), and the prescribed monodromy of the affine structure will guarantee the (σ,τ)(\sigma,\tau)-type asymptotics of the Monge-Ampère potential K⁡(s,t)K(s,t). ∎

Remark.

The discriminant locus D=∂𝒰v∩∂𝒰wD=\partial\mathcal{U}_{v}\cap\partial\mathcal{U}_{w} does not have to be affine linear. Even though DD lies in the polyhedral boundary ∂𝒰v\partial\mathcal{U}_{v}, the boundary ∂𝒰w\partial\mathcal{U}_{w} will be wiggled in the affine coordinates unless the partial Legendre transform is linear.

We would like to finish by mentioning an obvious application of the construction in this section to mirror symmetry. As was noted in [Hit97] the mirror duality in the semi-flat case is provided by the Legendre transform. This statement continues to hold in the neighborhood of the discriminant locus as well. Namely, in the single-valued coordinates (which are not affine) the full Legendre transform takes a singular split Monge-Ampère solution of type (σ,τ)(\sigma,\tau) into that of type (τ,σ)(\tau,\sigma).

References

  • Editorial link: [00YU] · shared chapter 98. [Bat93] V. Batyrev. Variations of the mixed Hodge structure of affine hypersurfaces in algebraic tori. Duke Math. J., 69(2):349–409, 1993.
  • Editorial link: [00YW] · shared chapter 99. [CY82] S.-Y. Cheng and S.-T. Yau. The real Monge-Ampère equation and affine flat structures. In Proceedings of the 1980 Beijing Symposium on Differential Geometry and Differential Equations, Vol. 1 (Beijing, 1980), pages 339–370, Beijing, 1982. Science Press.
  • Editorial link: [00YY] · shared chapter 6. [Ful93] W. Fulton. Introduction to Toric Varieties, volume 131 of Annals of Math. Studies. Princeton University Press, 1993.
  • Editorial link: [00Z0] · shared chapter 100. [GH78] G. W. Gibbons and S. W. Hawking. Gravitational instantons. Phys. Lett. B, 78:430–432, 1978.
  • Editorial link: [00Z2] · shared chapter 7. [GKZ94] I. Gelfand, M. Kapranov, and A. Zelevinsky. Discriminants, Resultants, and Multidimensional Determinants. Mathematics: Theory & Applications. Birkhäuser, 1994.
  • Editorial link: [00Z3] · shared chapter 195. [Gro01] Mark Gross. Topological mirror symmetry. Invent. Math., 144(1):75–137, 2001.
  • Editorial link: [00Z5] · shared chapter 101. [GS02] M. Gross and B. Siebert. Affine manifolds, log structures, and mirror symmetry. Preprint math.AG/0211094, 2002.
  • Editorial link: [00Z7] · shared chapter 102. [GT83] D. Gilbarg and N. S. Trudinger. Elliptic partial differential equations of second order, volume 224 of Grundlehren der Mathematischen Wissenschaften [Fundamental Principles of Mathematical Sciences]. Springer-Verlag, Berlin, second edition, 1983.
  • Editorial link: [00Z8] · shared chapter 197. [GW00] M. Gross and P. M. H. Wilson. Large complex structure limits of K​3{K}3 surfaces. J. Differential Geom., 55(3):475–546, 2000.
  • Editorial link: [00ZA] · shared chapter 103. [Haw77] S. W. Hawking. Gravitational instantons. Phys. Lett. A, 60(2):81–83, 1977.
  • Editorial link: [00ZC] · shared chapter 104. [Hit97] N. Hitchin. The moduli space of special Lagrangian submanifolds. Ann. Scuola Norm. Sup. Pisa Cl. Sci. (4), 25(3-4):503–515 (1998), 1997. Dedicated to Ennio De Giorgi.
  • Editorial link: [00ZD] · shared chapter 229. [HZ02] C. Haase and I. Zharkov. Integral affine structures on spheres and torus fibrations of Calabi-Yau toric hypersurfaces I. Preprint math.AG/0205321, 2002.
  • Editorial link: [00ZE] · shared chapter 230. [HZ03] C. Haase and I. Zharkov. Integral affine structures on spheres and torus fibrations of Calabi-Yau toric hypersurfaces II. Preprint math.AG/0301222, 2003.
  • Editorial link: [00ZF] · shared chapter 209. [KS01] M. Kontsevich and Y. Soibelman. Homological mirror symmetry and torus fibrations. In Symplectic geometry and mirror symmetry (Seoul, 2000), pages 203–263. World Sci. Publishing, River Edge, NJ, 2001.
  • Editorial link: [00ZG] · shared chapter 211. [KT02] M. Kontsevich and Yu. Tschinkel. Non-Archimedean Kähler geometry. In preparation, 2002.
  • Editorial link: [00ZI] · shared chapter 65. [LeB91] C. LeBrun. Complete Ricci-flat Kähler metrics on 𝐂n{\bf C}^{n} need not be flat. In Several complex variables and complex geometry, Part 2 (Santa Cruz, CA, 1989), volume 52 of Proc. Sympos. Pure Math., pages 297–304. Amer. Math. Soc., Providence, RI, 1991.
  • Editorial link: [00ZK] · shared chapter 105. [Leu00] N. C. Leung. Mirror symmetry without corrections. Preprint math.DG/0009235, 2000.
  • Editorial link: [00ZM] · shared chapter 106. [LYZ01] N. C. Leung, S. T. Yau, and E. Zaslow. From special Lagrangian to Hermitian-Yang-Mills via Fourier-Mukai transform. In Winter School on Mirror Symmetry, Vector Bundles and Lagrangian Submanifolds (Cambridge, MA, 1999), volume 23 of AMS/IP Stud. Adv. Math., pages 209–225. Amer. Math. Soc., Providence, RI, 2001.
  • Editorial link: [00ZP] · shared chapter 107. [Mat01] D. Matessi. Constructions of Calabi-Yau metrics and of special Lagrangian submanifolds. Ph.D Thesis, University of Warwick, UK, 2001.
  • Editorial link: [00ZR] · shared chapter 14. [Mik01] G. Mikhalkin. Amoebas of algebraic varieties. Preprint math.AG/0108225, 2001.
  • Editorial link: [00ZT] · shared chapter 72. [OV96] H. Ooguri and C. Vafa. Summing up Dirichlet instantons. Phys. Rev. Lett., 77(16):3296–3298, 1996.
  • Editorial link: [00ZV] · shared chapter 108. [PP91] H. Pedersen and Y. S. Poon. Hamiltonian constructions of Kähler-Einstein metrics and Kähler metrics of constant scalar curvature. Comm. Math. Phys., 136(2):309–326, 1991.
  • Editorial link: [00ZW] · shared chapter 228. [SYZ96] A. Strominger, S.-T. Yau, and E. Zaslow. Mirror symmetry is T{T}-duality. Nuclear Phys. B, 479(1-2):243–259, 1996.

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