ScalingStacks

3. Affine manifolds with singularities [02Z9]

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

3. Affine manifolds with singularities

To deal with singular fibres, we define

Definition 3.1.

A (tropical, integral) affine manifold with singularities is a (C0)(C^{0}) manifold BB with an open subset B0⊆BB_{0}\subseteq B which carries a (tropical, integral) affine structure, and such that Γ:=B∖B0\Gamma:=B\setminus B_{0} is a locally finite union of locally closed submanifolds of codimension ≥2\geq 2.

Here we will give a relatively simple construction of such affine manifolds with singularities; a broader class of examples is given in [22]; see also [39] and [40].

Let Δ\Delta be a reflexive polytope in Mℝ=M⊗ℤℝM_{\mathbb{R}}=M\otimes_{\mathbb{Z}}\mathbb{R}, where M=ℤnM=\mathbb{Z}^{n}. This means that Δ\Delta is a lattice polytope with a unique interior integral point 0∈Δ0\in\Delta, and the polar dual polytope

∇:={n∈Nℝ|⟨m,n⟩≥−1 for all m∈Δ}\nabla:=\{n\in N_{\mathbb{R}}|\hbox{$\langle m,n\rangle\geq-1$ for all $m\in\Delta$}\}

is also a lattice polytope.

Let B=∂ΔB=\partial\Delta, and let 𝒫\mathscr{P} be a decomposition of BB into lattice polytopes, i.e., 𝒫\mathscr{P} is a set of lattice polytopes contained in BB such that (1) B=⋃σ∈𝒫σB=\bigcup_{\sigma\in\mathscr{P}}\sigma; (2) σ1,σ2∈𝒫\sigma_{1},\sigma_{2}\in\mathscr{P} implies σ1∩σ2\sigma_{1}\cap\sigma_{2} lies in 𝒫\mathscr{P} and is a face of both σ1\sigma_{1} and σ2\sigma_{2}; (3) if σ∈𝒫\sigma\in\mathscr{P}, any face of σ\sigma lies in 𝒫\mathscr{P}.

We now define a structure of integral affine manifold with singularities on BB, with discriminant locus Γ⊆B\Gamma\subseteq B defined as follows. Let Bar⁡(𝒫)\operatorname{Bar}(\mathscr{P}) denote the first barycentric subdivision of 𝒫\mathscr{P} and let Γ⊆B\Gamma\subseteq B be the union of all simplices of Bar⁡(𝒫)\operatorname{Bar}(\mathscr{P}) not containing a vertex of 𝒫\mathscr{P} (a zero-dimensional cell) or intersecting the interior of a maximal cell of 𝒫\mathscr{P}. Setting B0:=B∖ΓB_{0}:=B\setminus\Gamma, we define an affine structure on B0B_{0} as follows. B0B_{0} has an open cover

{Wσ|σ∈𝒫 maximal}∪{Wv|v∈𝒫 a vertex}\{W_{\sigma}|\hbox{$\sigma\in\mathscr{P}$ maximal}\}\cup\{W_{v}|\hbox{$v\in\mathscr{P}$ a vertex}\}

where Wσ=Int⁡(σ)W_{\sigma}=\operatorname{Int}(\sigma), the interior of σ\sigma, and

Wv=⋃τ∈Bar⁡(𝒫)v∈τInt⁡(τ)W_{v}=\bigcup_{\tau\in\operatorname{Bar}(\mathscr{P})\atop v\in\tau}\operatorname{Int}(\tau)

is the (open) star of vv in Bar⁡(𝒫)\operatorname{Bar}(\mathscr{P}). We define an affine chart

ψσ:Wσ→𝔸n−1⊆Nℝ\psi_{\sigma}:W_{\sigma}\rightarrow\mathbb{A}^{n-1}\subseteq N_{\mathbb{R}}

given by the inclusion of WσW_{\sigma} in 𝔸n−1\mathbb{A}^{n-1}, the affine hyperplane containing σ\sigma. Also, take ψv:Wv→Mℝ/ℝ​v\psi_{v}:W_{v}\rightarrow M_{\mathbb{R}}/\mathbb{R}v to be the projection. One checks easily that for v∈σv\in\sigma, ψσ∘ψv−1\psi_{\sigma}\circ\psi_{v}^{-1} is integral affine linear (integrality follows from reflexivity of Δ\Delta!) so BB is an integral affine manifold with singularities.

Example 3.2.

Let Δ⊆ℝ4\Delta\subseteq\mathbb{R}^{4} be the convex hull of the points

(−1,−1,−1,−1),\displaystyle(-1,-1,-1,-1),
(4,−1,−1,−1),\displaystyle(4,-1,-1,-1),
(−1,4,−1,−1),\displaystyle(-1,4,-1,-1),
(−1,−1,4,−1),\displaystyle(-1,-1,4,-1),
(−1,−1,−1,4).\displaystyle(-1,-1,-1,4).

Choose a triangulation 𝒫\mathscr{P} of B=∂ΔB=\partial\Delta into standard simplices; this can be done in a regular way so that the restriction of 𝒫\mathscr{P} to each two-dimensional face of Δ\Delta is as given by the light lines in Figure 1. This gives a discriminant locus Γ\Gamma depicted by the dark lines in the figure; the line segments coming out of the boundary of the two-face are meant to illustrate the pieces of discriminant locus contained in adjacent two-faces. The discriminant locus there is not contained in the plane of the two-face. In particular, the discriminant locus is not planar at the vertices of Γ\Gamma on the edges of Ξ\Xi with respect to the affine structure we define. Note Γ\Gamma is a trivalent graph, with two types of trivalent vertices, the non-planar ones just mentioned and the planar vertices contained in the interior of two-faces.

Refer to caption
Figure 1.

For an affine manifold, the monodromy of the local system Λ\Lambda is an important feature of the affine structure. In this example, it is very useful to analyze this monodromy around loops about the discriminant locus. If vv is a vertex of Γ\Gamma contained in the interior of a two-face of Δ\Delta, one can consider loops based near vv in B0B_{0} around the three line segments of Γ\Gamma adjacent to vv. It is an enjoyable exercise to calculate that these monodromy matrices take the form, in a suitable basis,

T1=(100110001),T2=(100010101),T3=(100−110−101).T_{1}=\begin{pmatrix}1&0&0\\ 1&1&0\\ 0&0&1\end{pmatrix},T_{2}=\begin{pmatrix}1&0&0\\ 0&1&0\\ 1&0&1\end{pmatrix},T_{3}=\begin{pmatrix}1&0&0\\ -1&1&0\\ -1&0&1\end{pmatrix}.

They are computed by studying the composition of transition maps between charts that a loop passes through. These matrices can be viewed as specifying the obstruction to extending the affine structure across a neighbourhood of vv in Γ\Gamma. Of course, the monodromy of Λˇ\check{\Lambda} is the transpose inverse of these matrices. Similarly, if vv is a vertex of Γ\Gamma contained in an edge of Δ\Delta, then the monodromy will take the form

T1=(1−10010001),T2=(10−1010001),T3=(111010001).T_{1}=\begin{pmatrix}1&-1&0\\ 0&1&0\\ 0&0&1\end{pmatrix},T_{2}=\begin{pmatrix}1&0&-1\\ 0&1&0\\ 0&0&1\end{pmatrix},T_{3}=\begin{pmatrix}1&1&1\\ 0&1&0\\ 0&0&1\end{pmatrix}.

So we see that the monodromy of the two types of vertices are interchanged between Λ\Lambda and Λˇ\check{\Lambda}.

One main result of [20] is

Theorem 3.3.

If BB is a three-dimensional tropical affine manifold with singularities such that Γ\Gamma is trivalent and the monodromy of Λ\Lambda at each vertex is one of the above two types, then f0:X⁡(B0)→B0f_{0}:X(B_{0})\rightarrow B_{0} can be compactified to a topological fibration f:X⁡(B)→Bf:X(B)\rightarrow B. Dually, fˇ0:Xˇ​(B0)→B0\check{f}_{0}:\check{X}(B_{0})\rightarrow B_{0} can be compactified to a topological fibration fˇ:Xˇ​(B)→B\check{f}:\check{X}(B)\rightarrow B. Both X⁡(B)X(B) and Xˇ​(B)\check{X}(B) are topological manifolds.

We won’t give any details here of how this is carried out, but it is not particularly difficult, as long as one restricts to the category of topological (not C∞C^{\infty}) manifolds. However, it is interesting to look at the singular fibres we need to add.

If b∈Γb\in\Gamma is a point which is not a vertex of Γ\Gamma, then f−1​(b)f^{-1}(b) is homeomorphic to I1×S1I_{1}\times S^{1}, where I1I_{1} denotes a Kodaira type I1I_{1} elliptic curve, i.e., a pinched torus.

If vv is a vertex of Γ\Gamma, with monodromy of the first type, then f−1(v)=S1×S1×S1/∼f^{-1}(v)=S^{1}\times S^{1}\times S^{1}/\sim, with (a,b,c)∼(a′,b′,c′)(a,b,c)\sim(a^{\prime},b^{\prime},c^{\prime}) if (a,b,c)=(a′,b′,c′)(a,b,c)=(a^{\prime},b^{\prime},c^{\prime}) or a=a′=1a=a^{\prime}=1, where S1S^{1} is identified with the unit circle in ℂ\mathbb{C}. This is the three-dimensional analogue of a pinched torus, and χ​(f−1​(v))=+1\chi(f^{-1}(v))=+1. We call this a positive fibre.

If vv is a vertex of Γ\Gamma, with monodromy of the second type, then f−1​(v)f^{-1}(v) can be described as S1×S1×S1/∼S^{1}\times S^{1}\times S^{1}/\sim, with (a,b,c)∼(a′,b′,c′)(a,b,c)\sim(a^{\prime},b^{\prime},c^{\prime}) if (a,b,c)=(a′,b′,c′)(a,b,c)=(a^{\prime},b^{\prime},c^{\prime}) or a=a′=1a=a^{\prime}=1, b=b′b=b^{\prime}, or a=a′,b=b′=1a=a^{\prime},b=b^{\prime}=1. The singular locus of this fibre is a figure eight, and χ​(f−1​(v))=−1\chi(f^{-1}(v))=-1. We call this a negative fibre.

So we see a very concrete local consequence of SYZ duality: in the compactifications X⁡(B)X(B) and Xˇ​(B)\check{X}(B), the positive and negative fibres are interchanged. Of course, this results in the observation that the Euler characteristic changes sign under mirror symmetry for Calabi-Yau threefolds.

Example 3.4.

Continuing with Example 3.2, it was proved in [20] that Xˇ​(B)\check{X}(B) is homeomorphic to the quintic and X⁡(B)X(B) is homeomorphic to the mirror quintic. Modulo a paper [29] whose appearance has been long-delayed because of other, more pressing, projects, the results of [22] imply that the SYZ conjecture holds for all complete intersections in toric varieties at a topological level.

W.-D. Ruan in [69] gave a description of Lagrangian torus fibrations for hypersurfaces in toric varieties using a symplectic flow argument, and his construction should coincide with a symplectic compactification of the symplectic manifolds Xˇ​(B0)\check{X}(B_{0}). In the three-dimensional case, such a symplectic compactification has been constructed by Ricardo Castaño-Bernard and Diego Matessi [8]. If this compactification is applied to the affine manifolds with singularities described here, the resulting symplectic manifolds should be symplectomorphic to the corresponding toric hypersurface, but this has not yet been shown.

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