ScalingStacks

Part III [03W0]

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

Part III

We fix field KK satisfying Zero Characteristic Assumption.

Let BB be a compact oriented surface, Bs​i​n​g⊂BB^{sing}\subset B a finite set, and A​f​fK=A​f​fK,YAff_{K}=Aff_{K,Y} a sheaf defining a KK-affine structure on Y:=Bs​m=B∖Bs​i​n​gY:=B^{sm}=B\setminus B^{sing}. We assume that all singularities of the underlying 𝐙{{\bf Z}}-affine structure are standard (see Section 6.4), and local monodromy around each b∈Bs​i​n​gb\in B^{sing} acts on (K×)2(K^{\times})^{2} with a fixed point (see the Fixed Point property at the end of Section 7.1). Main result of Part 3 of the paper can be formulated such as follows.

Theorem 5

There exist a compact KK-analytic surface Xa​nX^{an}, a top degree analytic form Ω=ΩXa​n\Omega=\Omega_{X^{an}} and a continuous proper Stein map π:Xa​n→B\pi:X^{an}\to B such that the set of π\pi-smooth points coincides with YY and the induced KK-affine structure coincides with the one given by A​f​fKAff_{K}.

In other words, the triple (Xa​n,π,Ω)(X^{an},\pi,\Omega) is a solution of the Lifting Problem.

By Stein property it suffices to construct the sheaf 𝒪B=π∗​(𝒪Xa​n){\cal O}_{B}=\pi_{\ast}({\cal O}_{X^{an}}) of KK-algebras on BB. We will see that outside of the finite singular set S={x1,…,x24}S=\{x_{1},\dots,x_{24}\} the sheaf 𝒪B{\cal O}_{B} is locally isomorphic to 𝒪Yc​a​n{\cal O}_{Y}^{can}. In the next section we will describe the local model for the sheaf 𝒪B{\cal O}_{B} near each singular point. It will be glued together with a modification of the canonical sheaf 𝒪Yc​a​n{\cal O}_{Y}^{can}. This modification depends on the data called lines. Appearance of lines is motivated by Homological Mirror Symmetry (see [Ko], [KoSo]) 66 6 The main idea is that XX is a component of the moduli space of certain objects (skyscrapper sheaves) in the derived category Db​(C​o​h​(X))D^{b}(Coh(X)). These objects correspond to U⁡(1)U(1)-local systems on Lagrangian tori in the Fukaya category of the mirror dual symplectic manifold.. Roughly speaking, lines correspond (for mirror dual K3 surface) to collapsing holomorphic discs with boundaries belonging to fibers of the dual torus fibration (see Section 5.1 and [KoSo]). Such “bad” fibers are Lagrangian tori, but they do not correspond to objects of the Fukaya category (A-branes in terminology of physicists). There are infinitely many such fibers and hence infinitely many lines. We will axiomatize this piece of data in Section 9. Subsequently, with each line ll we will associate an automorphism of the restriction of 𝒪Yc​a​n{\cal O}_{Y}^{can} to ll. This will give us the above-mentioned modified canonical sheaf.

8 Model near a singular point

Here we will construct an analytic torus fibration corresponding to standard singularity (see Sections 3.2.4 and 6.4).

Let X⊂𝐀3X\subset{\bf A}^{3} be the algebraic surface given by equation (α​β−1)​γ=1(\alpha\beta-1)\gamma=1 in coordinates (α,β,γ)(\alpha,\beta,\gamma), and Xa​nX^{an} be the corresponding analytic space. We define a continuous map f:Xa​n→𝐑3f:X^{an}\to{{\bf R}}^{3} by the formula f⁡(α,β,γ)=(a,b,c)f(\alpha,\beta,\gamma)=(a,b,c) where a=max⁡(0,log⁡|α|p),b=max⁡(0,log⁡|β|p),c=log⁡|γ|p=−log⁡|α​β−1|pa=\max(0,\log|\alpha|_{p}),b=\max(0,\log|\beta|_{p}),c=\log|\gamma|_{p}=-\log|\alpha\beta-1|_{p}. Here |⋅|p=exp(−valp(⋅))|\cdot|_{p}=\exp(-val_{p}(\cdot)) denotes the multiplicative seminorm corresponding to the point p∈Xa​np\in X^{an} (see Appendix A).

Proposition 4

The map ff is proper. Moreover

a) Image of ff is homeomorphic to 𝐑2{\bf R}^{2}.

b) All points of the image except of (0,0,0)(0,0,0) are ff-smooth.

Proof. Here is the plan of the proof.

  1. 1.

    We define three open domains Ti,i=1,2,3T_{i},\,\,i=1,2,3 in three copies of the standard two-dimensional analytic torus (𝐆ma​n)2({\bf G}_{m}^{an})^{2}, and continuous maps πi:Ti→𝐑2\pi_{i}:T_{i}\to{{\bf R}}^{2} such that all points of the image Ui=πi​(Ti)U_{i}=\pi_{i}(T_{i}) are πi\pi_{i}-smooth (i.e. each πi\pi_{i} is an analytic torus fibration). Domains UiU_{i} cover 𝐑2∖{(0,0)}{\bf R}^{2}\setminus\{(0,0)\}.

  2. 2.

    For each i,1≤i≤3i,1\leq i\leq 3 we construct an open embedding gi:Ti↪Xa​ng_{i}:T_{i}\hookrightarrow X^{an}.

  3. 3.

    We construct an embedding j:𝐑2↪𝐑3j:{{\bf R}}^{2}\hookrightarrow{{\bf R}}^{3} such that each open set UiU_{i} is homeomorphically identified with f​(gi​(Ti))f(g_{i}(T_{i})) and j⁡((,,,))=(0,0,0)j((0,0))=(0,0,0). Moreover, πi\pi_{i}-smooth points are mapped into ff-smooth points.

The Proposition will follow from 1)-3).

Let us describe the constructions and formulas. We start with open sets Ui,1≤i≤3U_{i},1\leq i\leq 3. Let us fix a number 0<ε<10<\varepsilon<1 and define

U1={(x,y)∈𝐑2|x<ε​|y|}U2={(x,y)∈𝐑2|x>0,y<εx}U3={(x,y)∈𝐑2|x>0,y>0}\begin{array}[]{lll}U_{1}&=&\{(x,y)\in{{\bf R}}^{2}|x<\varepsilon|y|\,\}\\ U_{2}&=&\{(x,y)\in{{\bf R}}^{2}|x>0,y<\varepsilon x\,\}\\ U_{3}&=&\{(x,y)\in{{\bf R}}^{2}|x>0,y>0\}\end{array}

Clearly 𝐑2∖{(0,0)}=U1∪U2∪U3{{\bf R}}^{2}\setminus\{(0,0)\}=U_{1}\cup U_{2}\cup U_{3}. We define also a slightly modified domain U2′U_{2}^{\prime} as {(x,y)∈𝐑2|x>0,y<ε1+εx}\{(x,y)\in{{\bf R}}^{2}|x>0,y<\frac{\varepsilon}{1+\varepsilon}x\,\}.

We define Ti:=πc​a​n−1(Ui)⊂(𝐆ma​n)2,i=1,3T_{i}:=\pi_{can}^{-1}(U_{i})\subset({\bf G}_{m}^{an})^{2},i=1,3 and T2:=πc​a​n−1​(U2′)⊂(𝐆ma​n)2T_{2}:=\pi_{can}^{-1}(U_{2}^{\prime})\subset({\bf G}_{m}^{an})^{2}. Then the projections πi:Tl→Ul\pi_{i}:T_{l}\to U_{l} are given by the formulas

πi(ξi,ηi)=πc​a​n(ξi,ηi)=(log|ξi|,log|ηi|),i=1,3,\pi_{i}(\xi_{i},\eta_{i})=\pi_{can}(\xi_{i},\eta_{i})=(\log|\xi_{i}|,\log|\eta_{i}|),\,\,\,i=1,3\,\,,
π2​(ξ2,η2)={(log⁡|ξ2|,log⁡|η2|) if ​|η2|<1(log⁡|ξ2|−log⁡|η2|,log⁡|η2|) if ​|η2|≥1.\pi_{2}(\xi_{2},\eta_{2})=\left\{\begin{array}[]{ll}(\log|\xi_{2}|,\log|\eta_{2}|)&\mbox{ if }|\eta_{2}|<1\\ (\log|\xi_{2}|-\log|\eta_{2}|,\log|\eta_{2}|)&\mbox{ if }|\eta_{2}|\geq 1\end{array}\right.\,\,.

In these formulas (ξi,ηi)(\xi_{i},\eta_{i}) are coordinates on Ti,1≤i≤3T_{i},1\leq i\leq 3.

We define inclusion gi:Ti↪X,1≤i≤3g_{i}:T_{i}\hookrightarrow X,1\leq i\leq 3 by the following formulas:

g1​(ξ1,η1)=(1ξ1,ξ1​(1+η1),1η1)g2​(ξ2,η2)=(1+η2ξ2,ξ2,1η2)g3​(ξ3,η3)=(1+η3ξ3​η3,ξ3​η3,1η3)\begin{array}[]{lll}g_{1}(\xi_{1},\eta_{1})&=&({1\over{\xi_{1}}},\xi_{1}(1+\eta_{1}),{1\over{\eta_{1}}})\\ g_{2}(\xi_{2},\eta_{2})&=&({1+\eta_{2}\over{\xi_{2}}},\xi_{2},{1\over{\eta_{2}}})\\ g_{3}(\xi_{3},\eta_{3})&=&({1+\eta_{3}\over{\xi_{3}\eta_{3}}},\xi_{3}\eta_{3},{1\over{\eta_{3}}})\end{array}

Let us decompose Xa​n=X−∪X0∪X+X^{an}=X_{-}\cup X_{0}\cup X_{+} according to the sign of log⁡|γ|p\log|\gamma|_{p} where p∈Xa​np\in X^{an} is a point. It is easy to see that

f⁡(X−)={(a,b,c)∈𝐑3|c<0,a≥0,b≥0,ab(a+b+c)=0}f⁡(X0)={(a,b,c)∈𝐑3|c=0,a≥0,b≥0,ab=0}f⁡(X+)={(a,b,c)∈𝐑3|c>0,a≥0,b≥0,ab=0}\begin{array}[]{lll}f(X_{-})&=&\{\,(a,b,c)\in{{\bf R}}^{3}\,|\,c<0,a\geq 0,b\geq 0,\,ab(a+b+c)=0\,\}\\ f(X_{0})&=&\{\,(a,b,c)\in{{\bf R}}^{3}\,|\,c=0,a\geq 0,b\geq 0,\,ab=0\,\}\\ f(X_{+})&=&\{\,(a,b,c)\in{{\bf R}}^{3}\,|\,c>0,a\geq 0,b\geq 0,\,ab=0\,\}\end{array}

From this explicit description we see that ff is proper and the image of ff is homeomorphic to 𝐑2{{\bf R}}^{2}.

Let us consider the embedding j:𝐑2→𝐑3j:{{\bf R}}^{2}\to{{\bf R}}^{3} given by formula

j⁡(x,y)={(−x,max⁡(x+y,0),−y) if x≤0( 0,x+max⁡(y,0),−y) if x≥0j(x,y)=\left\{\begin{array}[]{lll}(-x\,,\,\max(x+y,0)\,,\,-y\,)&\mbox{ if }&x\leq 0\\ (\,0\,,\,x+\max(y,0)\,,\,-y\,)&\mbox{ if }&x\geq 0\end{array}\right.

One can easily check that the image of jj coincides with the image of ff, j∘πi=f∘gij\circ\pi_{i}=f\circ g_{i} and f−1​(j⁡(Ui))=gi​(Ti)f^{-1}(j(U_{i}))=g_{i}(T_{i}) for all 1≤i≤31\leq i\leq 3. This concludes the proof of Proposition. ■\blacksquare

We can derive more from explicit formulas given in the proof.

Let us denote by π:Xa​n→𝐑2\pi:X^{an}\to{{\bf R}}^{2} the map j(−1)∘fj^{(-1)}\circ f. It is an analytic torus fibration outside of point (0,0)(0,0). The induced 𝐙{\bf Z}-affine structure on 𝐑2∖{(0,0)}{{\bf R}}^{2}\setminus\{(0,0)\} is in fact the standard singular 𝐙{\bf Z}-affine structure described in Sections 3.2.4 and 6.4, as follows immediately from formulas for projections πi,i=1,2,3\pi_{i},\,i=1,2,3.

Let us introduce another sheaf 𝒪c​a​n{\cal O}^{can} on 𝐑2∖{(0,0)}{{\bf R}}^{2}\setminus\{(0,0)\}. It is defined as (πi)∗​(𝒪Ti)(\pi_{i})_{*}\left({\cal O}_{T_{i}}\right) in each domain UiU_{i}, with identifications

(ξ1,η1)=(ξ2,η2) on ​U1∩U2(ξ1,η1)=(ξ3,η3) on ​U1∩U3(ξ2,η2)=(ξ3​η3,η3) on ​U2∩U3\begin{array}[]{llcl}(\xi_{1},\eta_{1})&=&(\xi_{2},\eta_{2})&\mbox{ on }U_{1}\cap U_{2}\\ (\xi_{1},\eta_{1})&=&(\xi_{3},\eta_{3})&\mbox{ on }U_{1}\cap U_{3}\\ (\xi_{2},\eta_{2})&=&(\xi_{3}\eta_{3},\eta_{3})&\mbox{ on }U_{2}\cap U_{3}\end{array}

Let us consider the direct image sheaf π∗​(𝒪Xa​n)\pi_{\ast}({\cal O}_{X^{an}}). It is easy to see that on the sets U1U_{1} and U2∪U3U_{2}\cup U_{3} this sheaf is canonically isomorphic to 𝒪c​a​n{\cal O}^{can}. The isomorphism is given by the identification of coordinates (ξ1,η1)(\xi_{1},\eta_{1}) on U1U_{1}, and of coordinates (ξ2,η2)(\xi_{2},\eta_{2}) and (ξ3,η3)(\xi_{3},\eta_{3}) on U2∪U3U_{2}\cup U_{3}. Therefore on the intersection U1∩(U2∪U3)U_{1}\cap(U_{2}\cup U_{3}) we identify two copies of the canonical sheaf by certain automorphism φ\varphi of 𝒪c​a​n{\cal O}^{can} which preserves one coordinate (namely, the coordinate η\eta). We will develop the theory of such transformations and their analytic continuations in Section 11. The explicit formulas for φ\varphi is

φ⁡(ξ,η)={(ξ⁡(1+η),η) on U1∩U2(ξ⁡(1+1/η),η) on U1∩U3\varphi(\xi,\eta)=\left\{\begin{array}[]{cll}(\xi(1+\eta),\eta)&\mbox{ on }&U_{1}\cap U_{2}\\ (\xi(1+1/\eta),\eta)&\mbox{ on }&U_{1}\cap U_{3}\end{array}\right.

We would like to say now few words about analytic volume forms. Notice that each Ti⊂(𝐆ma​n)2T_{i}\subset({\bf G}_{m}^{an})^{2} carries a nowhere vanishing top degree analytic form given by the formula Ωi=d​ξi∧d​ηiξi​ηi\Omega_{i}={d\xi_{i}\wedge d\eta_{i}\over{\xi_{i}\eta_{i}}}. Then a straightforward calculation shows that ΩTi\Omega^{T_{i}} is the pullback under gig_{i} of nowhere vanishing on Xa​nX^{an} analytic top degree form

Ω=−γ​d​α∧d​β.\Omega=-\gamma\,d\alpha\wedge d\beta\,\,.

Form Ω\Omega satisfies Constant Norm Assumption, hence it gives a KK-affine structure on 𝐑2∖{(0,0)}{\bf R}^{2}\setminus\{(0,0)\}. On the other hand, the sheaf 𝒪c​a​n{\cal O}^{can} of algebras is also endowed with top-degree form Ωc​a​n\Omega^{can}, equal to d​ξi∧d​ηiξi​ηi{d\xi_{i}\wedge d\eta_{i}\over{\xi_{i}\eta_{i}}} in local coordinates.

Lemma 3

The KK-affine structure on 𝐑2∖{(0,0)}{\bf R}^{2}\setminus\{(0,0)\} associated with Ω\Omega coincides with the one associated with Ωc​a​n\Omega^{can}.

Proof: Using definitions from Section 7.2 one sees immediately that the statement of the Lemma follows from the equality pΩc​a​n​(1+η)=1p_{\Omega^{can}}(1+\eta)=1, which is straightoforward: pΩc​a​n​(1+η)=e​x​p​(R​e​s​(Ωc​a​n​log⁡(1+η)))=1∈𝒪K×p_{\Omega^{can}}(1+\eta)=exp\left(Res(\Omega^{can}\,\log(1+\eta))\right)=1\in{\cal O}_{K}^{\times}. ■\,\blacksquare

In all the definitions and formulas in this section on can shift domains Ui,,i=1,2,3U_{i},_{,}i=1,2,3 by vector (x0,0)∈𝐑2(x_{0},0)\in{\bf R}^{2} for arbitrary x0∈𝐑x_{0}\in{\bf R}, thus giving a map Xa​n→𝐑2X^{an}\to{\bf R}^{2} with singularity at the point (x0,0)(x_{0},0).

Finally, we denote π∗​(𝒪Xa​n)\pi_{\ast}({\cal O}_{X^{an}}) by 𝒪𝐑2m​o​d​e​l{\cal O}^{model}_{{\bf R}^{2}}. This will be our model for the sheaf 𝒪B{\cal O}_{B} near each point of the singular set Bs​i​n​gB^{sing}.

9 Lines on surfaces

In this section we are going to describe axiomatically the notion of collection of lines on a surface.

9.1 Data

a)

A compact oriented surface BB, a finite subset Bs​i​n​g⊂BB^{sing}\subset B.

b)

A 𝐙{\bf Z}-affine structure on Y=Bs​m=B∖Bs​i​n​gY=B^{sm}=B\setminus B^{sing} with the standard singularities near each b∈Bs​i​n​gb\in B^{sing}.

c)

A set ℒ{\cal L} of lines. With each line l∈ℒl\in{\cal L} there is an associated continuous map fl:(0,+∞)→Yf_{l}:(0,+\infty)\to Y. We assume that ℒ{\cal L} is decomposed into a disjoint union of two subsets ℒ=ℒi​n⊔ℒc​o​m{\cal L}={\cal L}_{in}\sqcup{\cal L}_{com}. Lines belonging to ℒi​n{\cal L}_{in} are called initial, while those in ℒc​o​m{\cal L}_{com} are called composite. We assume that for any l∈ℒl\in{\cal L} there exists a continuous extension fl:[0,+∞)→Bf_{l}:[0,+\infty)\to B such that fl​(0)∈Bs​i​n​gf_{l}(0)\in B^{sing} if l∈ℒi​nl\in{\cal L}_{in} and fl​(0)∈Y=Bs​mf_{l}(0)\in Y=B^{sm} if l∈ℒc​o​ml\in{\cal L}_{com}.

d)

A collection of covariantly constant nowhere vanishing integer-valued 11-forms αl∈Γ⁡((0,+∞),fl∗​((T∗)𝐙),l∈ℒCLOSE\alpha_{l}\in\Gamma((0,+\infty),f_{l}^{\ast}((T^{\ast})^{\bf Z}),l\in{\cal L}. We assume that for l∈ℒi​nl\in{\cal L}_{in} in the standard coordinates (x,y)(x,y) near singular point fl​(0)f_{l}(0) we have: fl​(t)=(0,t)f_{l}(t)=(0,t) or fl​(t)=(0,−t)f_{l}(t)=(0,-t) for all sufficiently small t>0t>0, and αl​(t)=±fl∗​(d​y)\alpha_{l}(t)=\pm f_{l}^{\ast}(dy).

e)

A map ℒ→ℒ×ℒ,l↦(pl​e​f​t​(l),pr​i​g​h​t​(l)){\cal L}\to{\cal L}\times{\cal L},\,\,\,l\mapsto(p_{left}(l),p_{right}(l)) (the letter pp stands for “parent”: one can think about these lines as “generating ll in a collision”).

Notice that since the form d​ydy is invariant with respect to the monodromy, the condition in d) is coordinate-independent. The covector αl​(t)\alpha_{l}(t) will be called a direction covector of ll at time tt. It gives rise to a half-plane

Pl,t(0)={v∈Tfl​(t)​Y|⟨αl​(t),v⟩>0}.P_{l,t}^{(0)}=\{v\in T_{f_{l}(t)}Y|\langle\alpha_{l}(t),v\rangle>0\}\,\,.

9.2 Axioms

To every l∈ℒi​nl\in{\cal L}_{in} we assign a pair (fl​(0),s​g​n​(αl​(0)))∈Bs​i​n​g×{±1}\left(f_{l}(0),sgn\left(\alpha_{l}(0)\right)\right)\in B^{sing}\times\{\pm 1\}, where s​g​n​(αl​(0))sgn(\alpha_{l}(0)) is a choice of sign in ±fl∗​(d​y)\pm f_{l}^{\ast}(dy) (see data d) in the previous subsection). In this way we obtain a map r:ℒi​n→Bs​i​n​g×{±1}r:{\cal L}_{in}\to B^{sing}\times\{\pm 1\}.

Axiom 1.

Map rr is one-to-one.

Let U⊂YU\subset Y be a simply-connected domain, and line ll intersects UU. Let I⊂𝐑+I\subset{\bf R}_{+} be an interval such that fl​(I)⊂Uf_{l}(I)\subset U. Then there exists a covariantly constant closed non-zero 11-form βU\beta_{U} in UU (with constant integer coefficients), such that fl∗​(βU)=αlf_{l}^{\ast}(\beta_{U})=\alpha_{l}, when both sides are restricted to II.

Axiom 2.

For any t1,t2∈It_{1},t_{2}\in I one has

∫fl​(t1)fl​(t2)βU=t2−t1.\int_{f_{l}(t_{1})}^{f_{l}(t_{2})}\beta_{U}=t_{2}-t_{1}\,\,.

Let l1,l2∈ℒl_{1},l_{2}\in{\cal L}, t1,t2>0t_{1},t_{2}>0 satisfy the condition fl1​(t1)=fl2​(t2)=x∈Yf_{l_{1}}(t_{1})=f_{l_{2}}(t_{2})=x\in Y. In this case we say that lines l1l_{1} and l2l_{2} have a collision at xx at the times t1t_{1} and t2t_{2} respectively.

Axiom 3.

Under the above assumptions there are only two possibilities:

3a) either l1=l2l_{1}=l_{2} and t1=t2t_{1}=t_{2}, or

3b) covector αl1​(t1)\alpha_{l_{1}}(t_{1}) is not proportional to αl2​(t2)\alpha_{l_{2}}(t_{2}). Then we may assume that αl1​(t1)∧αl2​(t2)>0\alpha_{l_{1}}(t_{1})\wedge\alpha_{l_{2}}(t_{2})>0. Under these conditions we require that for any coprime positive integers n1,n2n_{1},n_{2} there exists a unique line l∈ℒl\in{\cal L} such that l1=pl​e​f​t​(l),l2=pr​i​g​h​t​(l)l_{1}=p_{left(l)},l_{2}=p_{right}(l), fl​(0)=xf_{l}(0)=x and αl​(0)=n1​αl1​(t1)+n2​αl2​(t2)\alpha_{l}(0)=n_{1}\alpha_{l_{1}}(t_{1})+n_{2}\alpha_{l_{2}}(t_{2}).

In other words, l1l_{1} and l2l_{2} are “parents of ll”, and the direction covector of ll at the intersection point is a primitive integral linear combination of those for l1l_{1} and l2l_{2} (see Figure 4).

Refer to caption

Figure 4: Line ll and its two parents l1,l2l_{1},l_{2}. Dashed half-planes are domains in tangent planes where 11-forms α\alpha take positive values.
Axiom 4.

For every line l∈ℒc​o​ml\in{\cal L}_{com} there exist l1l_{1} and l2l_{2} such that they satisfy the condition 3b).

Axiom 5.

For any x∈Yx\in Y there are no more than two pairs (l,t)∈ℒ×(0,+∞)(l,t)\in{\cal L}\times(0,+\infty) such that x=fl​(t)x=f_{l}(t). In other words, there are no more than two lines intersecting at a point in YY.

Let l1,l2,t1,t2,xl_{1},l_{2},t_{1},t_{2},x mean the same as in the Axiom 3, and assume that αl1​(t1)∧αl2​(t2)>0\alpha_{l_{1}}(t_{1})\wedge\alpha_{l_{2}}(t_{2})>0. Let us consider the set ℒ(x){\cal L}_{(x)} of germs of all l∈ℒc​o​ml\in{\cal L}_{com} starting at xx (i.e. such that fl​(0)=xf_{l}(0)=x).

Axiom 6.

For any finite subset ℒ′⊂ℒ(x){\cal L}^{\prime}\subset{\cal L}_{(x)} there is an orientation preserving homeomorphism of a neighborhood of xx onto a neighborhood of (0,0)∈𝐑2(0,0)\in{\bf R}^{2} such that:

6a) Germs of oriented curves which are images of l1l_{1} and l2l_{2} get transformed into the germs at (0,0)(0,0) of coordinate axes (x,0)(x,0) and (0,y)(0,y) respectively.

6b) Germ of the image of l∈ℒ′l\in{\cal L}^{\prime} gets transformed into the germ of the ray {(n1​t,n2​t)|t>0}\{(n_{1}t,n_{2}t)\,|\,t>0\} where αl​(0)=n1​αl1​(t1)+n2​αl2​(t2)\alpha_{l}(0)=n_{1}\alpha_{l_{1}}(t_{1})+n_{2}\alpha_{l_{2}}(t_{2}).

Figure 5 illustrates this axiom.

Refer to caption

Figure 5: Two intersecting lines and some of new lines obtained as a result of collision. All lines are straightened by a homeomorphism of 𝐑2{\bf R}^{2}.
Axiom 7.

Let pip_{i} denotes either pl​e​f​tp_{left} or pr​i​g​h​tp_{right}. Then for any l∈ℒl\in{\cal L} there exists N≥1N\geq 1 such that if the line p1​(p2​(…​pN​(l)​…)CLOSEp_{1}(p_{2}(\dots p_{N}(l)\dots) is well-defined then it belongs to ℒi​n{\cal L}_{in}.

This axiom says that any composed line l∈ℒc​o​ml\in{\cal L}_{com} appears as a result of finitely many collisions. The tree of ancestors of a given line form a tree embedded in BB, see Figure 6.

Refer to caption

Figure 6: Tree of ancestors of line ll starting from 33 singular points s1,s2,s3∈Bs​i​n​gs_{1},s_{2},s_{3}\in B^{sing}.

9.3 Example: gradient lines

Here we offer a construction of the set of lines satisfying the above axioms.

Let us use the standard 𝐑2{\bf R}^{2} as a model around each b∈Bs​i​n​gb\in B^{sing} in order to fix a structure of smooth manifold on the whole surface BB. Let Y~\widetilde{Y} denotes the covering of YY such that the fiber over y∈Yy\in Y is (Ty∗​Y)𝐙∖{0}(T_{y}^{\ast}Y)^{{\bf Z}}\setminus\{0\}.

Let us fix a generic smooth metric on BB. By the pull-back it gives a metric on Y~\widetilde{Y}. Notice that there is a canonical closed 11-form β\beta on Y~\widetilde{Y} such that β|(y,μ)=μ\beta_{|(y,\mu)}=\mu, where y∈Y,μ∈(Ty∗​Y)𝐙y\in Y,\,\,\mu\in(T_{y}^{\ast}Y)^{{\bf Z}}. Using the metric we obtain dual to β\beta gradient vector field vv on Y~\widetilde{Y}.

For any s∈Bs​i​n​gs\in B^{sing} and a choice of 1-form α⁡(0)=±d​y\alpha(0)=\pm dy in local coordinates, we take the unique integral line of vv starting at (s,α⁡(0))(s,\alpha(0)). Set ℒi​n{\cal L}_{in} will be the set of all lines obtained in this way. Each line l∈ℒi​nl\in{\cal L}_{in} carries a covariantly constant closed 11-form αl\alpha_{l}. Using Axiom 2 as a definition, we obtain a canonical parametrization of each line by the time parameter tt. Since the metric is generic, a line cannot return to a point in Bs​i​n​gB^{sing}.

Then we proceed inductively. If two already constructed lines l1,l2∈ℒl_{1},l_{2}\in{\cal L} meet at x∈Yx\in Y we produce a new integral line ll of vv with the direction covector satisfying the condition 3b) for any pair of coprime positive integers n1,n2n_{1},n_{2}. In this way we construct a set of lines ℒ{\cal L} satisfying all the axioms. The only non-trivial thing to check is that for each line values of the parameter tt are in one-to-one correspondence with the interval (0,+∞)(0,+\infty). In order to see this we observe that the length of each line is infinite. Indeed, an integral curve of vv cannot have a limiting point in YY (since the flow generated by vv is smooth, and the lengths of tangent vectors are bounded from below because of the integrality of 11-forms).

We conclude that there exists a set ℒ{\cal L} of lines satisfying Axioms 1-7.

10 Groups and symplectomorphisms

In this section we are going to discuss the sheaf of groups of symplectomorphisms S​y​m​p:=S​y​m​p​(𝒪Yc​a​n)Symp:=Symp({\cal O}_{Y}^{can}) of the sheaf 𝒪Yc​a​n{\cal O}_{Y}^{can}. Let U⊂YU\subset Y be an open convex subset. By definition, a symplectomorphism of 𝒪Yc​a​n​(U){\cal O}_{Y}^{can}(U) is an automorphism of KK-algebra 𝒪Yc​a​n​(U){\cal O}_{Y}^{can}(U) preserving projection to YY and the canonical symplectic form Ω=d​ξ∧d​ηξ​η\Omega={d\xi\wedge d\eta\over{\xi\eta}} (the latter is understood as an element of the algebra of Kähler differential forms). To each line ll we will assign a symplectomorphism of the restriction of 𝒪Yc​a​n{\cal O}_{Y}^{can} to ll, so that the assignment will be compatible with the collision of lines. Then we are going to modify the sheaf 𝒪Yc​a​n{\cal O}_{Y}^{can} using symplectomorphisms, associated with lines and obtain the sheaf 𝒪Ym​o​d​i​f{\cal O}_{Y}^{modif}. This sheaf will be glued with the sheaf 𝒪𝐑2m​o​d​e​l{\cal O}^{model}_{{\bf R}^{2}} near each point of Bs​i​n​gB^{sing}.

10.1 Pro-nilpotent Lie algebra

Here it will be convenient to work in local coordinates (x,y)=(log⁡|ξ|,log⁡|η|)(x,y)=(\log|\xi|,\log|\eta|) on YY.

Let (x0,y0)∈𝐑2(x_{0},y_{0})\in{\bf R}^{2} be a point, α1,α2∈(𝐙2)∗\alpha_{1},\alpha_{2}\in({\bf Z}^{2})^{\ast} be 11-covectors such that α1∧α2>0\alpha_{1}\wedge\alpha_{2}>0. Denote by V=V(x0,y0),α1,α2V=V_{(x_{0},y_{0}),\alpha_{1},\alpha_{2}} the closed angle

{(x,y)∈𝐑2|⟨αi,(x,y)−(x0,y0)⟩≥0,i=1,2}.\{(x,y)\in{\bf R}^{2}|\langle\alpha_{i},(x,y)-(x_{0},y_{0})\rangle\geq 0,i=1,2\,\}\,\,.

Let 𝒪⁡(V){\cal O}(V) be a KK-algebra consisting of series f=∑n,m∈𝐙cn,m​ξn​ηmf=\sum_{n,m\in{\bf Z}}c_{n,m}\xi^{n}\eta^{m}, such that cn,m∈Kc_{n,m}\in K and for all (x,y)∈V(x,y)\in V we have:

  1. 1.

    if cn,m≠0c_{n,m}\neq 0 then ⟨(n,m),(x,y)−(x0,y0)⟩≤0\langle(n,m),(x,y)-(x_{0},y_{0})\rangle\leq 0, where we identified (n,m)∈𝐙2(n,m)\in{\bf Z}^{2} with a covector in (Tp∗​Y)𝐙(T_{p}^{\ast}Y)^{{\bf Z}};

  2. 2.

    log⁡|cn,m|+n​x+m​y→−∞\log|c_{n,m}|+nx+my\to-\infty as long as |n|+|m|→+∞|n|+|m|\to+\infty.

Algebra 𝒪⁡(V){\cal O}(V) is a Poisson algebra with respect to the bracket {ξ,η}=ξ​η\{\xi,\eta\}=\xi\eta.

For an integer covector μ=a​d​x+b​d​y∈(𝐙2)∗\mu=adx+bdy\in({\bf Z}^{2})^{*} we denote by RμR_{\mu} the monomial ξa​ηb\xi^{a}\eta^{b}.

Let us consider a pro-nilpotent Lie algebra 𝐠:=𝐠α1,α2,V⊂𝒪⁡(V){\bf g}:={\bf g}_{\alpha_{1},\alpha_{2},V}\subset{\cal O}(V) consisting of series

f=∑n1,n2≥0,n1+n2>0cn1,n2​Rα1−n1​Rα2−n2f=\sum_{n_{1},n_{2}\geq 0,n_{1}+n_{2}>0}c_{n_{1},n_{2}}R_{\alpha_{1}}^{-n_{1}}R_{\alpha_{2}}^{-n_{2}}

satisfying the condition

log⁡|cn,m|−n1​⟨α1,(x,y)⟩−n2​⟨α2,(x,y)⟩≤0​∀(x,y)∈V.\log|c_{n,m}|-n_{1}\langle\alpha_{1},(x,y)\rangle-n_{2}\langle\alpha_{2},(x,y)\rangle\leq 0\,\,\,\,\forall\,(x,y)\in V\,\,.

The latter condition is equivalent to log⁡|cn,m|−⟨n1​α1+n2​α2,(x0,y0)⟩≤0\log|c_{n,m}|-\langle n_{1}\alpha_{1}+n_{2}\alpha_{2},(x_{0},y_{0})\rangle\leq 0.

Lie algebra 𝐠{\bf g} admits a filtration by Lie subalgebras 𝐠≥k,k∈𝐙,k≥1{\bf g}^{\geq k},k\in{\bf Z},k\geq 1, 𝐠=𝐠≥1{\bf g}={\bf g}^{\geq 1}, such that 𝐠≥k{\bf g}^{\geq k} consists of the above series which satisfy the condition n1+n2≥kn_{1}+n_{2}\geq k. Clearly [𝐠≥k1,𝐠≥k2]⊂𝐠≥k1+k2[{\bf g}^{\geq k_{1}},{\bf g}^{\geq k_{2}}]\subset{\bf g}^{\geq k_{1}+k_{2}}, and 𝐠=lim←k→+∞⁡𝐠/𝐠≥k{\bf g}=\varprojlim_{k\to+\infty}{\bf g}/{\bf g}^{\geq k}.

Thus, 𝐠{\bf g} is a topological complete pro-nilpotent Lie algebra over KK. We denote by GG the corresponding pro-nilpotent Lie group exp⁡(𝐠)\exp({\bf g}). It inherits the filtration by normal subgroups G≥kG^{\geq k} obtained from the corresponding Lie algebras.

10.2 Lie groups GλG_{\lambda}

For each λ∈[0,+∞]𝐐:=𝐐≥0∪∞\lambda\in[0,+\infty]_{{\bf Q}}:={\bf Q}_{\geq 0}\cup\infty we define a Lie subalgebra

𝐠λ={∑n1,n2cm,nRα1−n1Rα2−n2∈𝐠|cn1,n2∈K,n2n1=λ}.{\bf g}_{\lambda}=\left\{\sum_{n_{1},n_{2}}c_{m,n}R_{\alpha_{1}}^{-n_{1}}R_{\alpha_{2}}^{-n_{2}}\in{\bf g}\,|\,\,c_{n_{1},n_{2}}\in K,\,\,{n_{2}\over{n_{1}}}=\lambda\,\right\}\,\,.

Each 𝐠λ{\bf g}_{\lambda} is an abelian Lie algebra. It carries the induced filtration by Lie algebras 𝐠λ≥k=𝐠λ∩𝐠≥k{\bf g}_{\lambda}^{\geq k}={\bf g}_{\lambda}\cap{\bf g}^{\geq k}. Denote by Gλ=exp⁡(𝐠λ)G_{\lambda}=\exp({\bf g}_{\lambda}) the corresponding pro-nilpotent group.

Lemma 4

For any given k≥1k\geq 1 there exist finitely many λ1<λ2<⋯<λNk\lambda_{1}<\lambda_{2}<\dots<\lambda_{N_{k}} such that 𝐠λ/𝐠λ≥k=0{\bf g}_{\lambda}/{\bf g}^{\geq k}_{\lambda}=0 for λ≠λi,1≤i≤Nk\lambda\neq\lambda_{i},1\leq i\leq N_{k}.

Proof. Indeed, for the monomial Rα1−n1​Rα2−n2∈𝐠λR_{\alpha_{1}}^{-n_{1}}R_{\alpha_{2}}^{-n_{2}}\in{\bf g}_{\lambda} which maps non-trivially to the quotient 𝐠λ/𝐠λ≥k{\bf g}_{\lambda}/{\bf g}^{\geq k}_{\lambda} we have: n1+n2≤k,n1/n2=λn_{1}+n_{2}\leq k,n_{1}/n_{2}=\lambda, where n1,n2n_{1},n_{2} are non-negative integers. There are finitely many such non-negative integers n1n_{1} and n2n_{2}. ■\blacksquare

It follows from the Lemma that we have a natural isomorphism of vector spaces ∏λ∈[0,+∞]𝐐𝐠λ/𝐠λ≥k→𝐠/𝐠≥k\prod_{\lambda\in[0,+\infty]_{{\bf Q}}}{\bf g}_{\lambda}/{\bf g}_{\lambda}^{\geq k}\to{\bf g}/{\bf g}^{\geq k}, hence the map

(fλ)λ∈[0,+∞]𝐐↦∑λfλ=∑i=1Nifλi, where fλ∈𝐠λ/𝐠λ≥k∀λ∈[0,+∞]𝐐(f_{\lambda})_{\lambda\in[0,+\infty]_{{\bf Q}}}\mapsto\sum_{\lambda}f_{\lambda}=\sum_{i=1}^{N_{i}}f_{\lambda_{i}},\,\,\,\,\mbox{ where }f_{\lambda}\in{\bf g}_{\lambda}/{\bf g}_{\lambda}^{\geq k}\,\,\,\forall\lambda\in[0,+\infty]_{{\bf Q}}

is well-defined and gives rise (after taking the projective limit as k→+∞k\to+\infty) to the isomorphism 𝐠≃∏λ∈[0,+∞]𝐐𝐠λ{\bf g}\simeq\prod_{\lambda\in[0,+\infty]_{{\bf Q}}}{\bf g}_{\lambda}.

In a similar way we define the map ∏→:∏λ∈[0,+∞]𝐐Gλ→G\prod_{\to}:\prod_{\lambda\in[0,+\infty]_{{\bf Q}}}G_{\lambda}\to G, the product is taken with respect to the natural order on 𝐐{\bf Q}. Namely, for any k≥1k\geq 1 we define

∏→(k):∏i=1NkGλi/Gλi≥k→G/G≥k,(g1,…,gNk)↦g1​…​gNk, for ​gi∈Gλi/Gλi≥k{\textstyle\prod_{\to}^{(k)}}:\prod_{i=1}^{N_{k}}G_{\lambda_{i}}/G_{\lambda_{i}}^{\geq k}\to G/G^{\geq k}\,,\,(g_{1},\dots,g_{N_{k}})\mapsto g_{1}\dots g_{N_{k}},\,\mbox{ for }g_{i}\in G_{\lambda_{i}}/G_{\lambda_{i}}^{\geq k}

and then set ∏→:=lim←k∏→(k)\prod_{\to}:=\varprojlim_{k}\prod_{\to}^{(k)}\,.

Theorem 6

Map ∏→\prod_{\to} is a bijection of sets.

Proof. Let k≥1k\geq 1 be an integer. We claim that ∏→(k)\prod_{\to}^{(k)} is a bijection of sets (this implies the proposition by taking the projective limit as k→+∞k\to+\infty). We will prove the bijection by induction in kk . Case k=1k=1 is obvious because all the groups under considerations are trivial.

We would like to prove that ∏→(k+1)\prod_{\to}^{(k+1)} is a bijection assuming that ∏→(k)\prod_{\to}^{(k)} is a bijection. Let hh be an element of G/G≥k+1G/G^{\geq k+1} and h¯\overline{h} its image in G/G≥kG/G^{\geq k}. By the induction assumption there exist unique h¯i∈Gλi/Gλi≥k+1,1≤i≤Nk+1\overline{h}_{i}\in G_{\lambda_{i}}/G_{\lambda_{i}}^{\geq k+1},1\leq i\leq N_{k+1} such that h¯1​…​h¯Nk+1=h¯\overline{h}_{1}\dots\overline{h}_{N_{k+1}}=\overline{h}. Let hi,1≤i≤Nk+1h_{i},1\leq i\leq N_{k+1} be any liftings of h¯i\overline{h}_{i} to Gi/Gi≥kG_{i}/G_{i}^{\geq k}. Then h1​…​hNk+1=h(modG≥k)h_{1}\dots h_{N_{k+1}}=h\pmod{G^{\geq k}}, hence c:=h1​…​hNk+1​h−1c:=h_{1}\dots h_{N_{k+1}}h^{-1} belongs to G≥k/G≥k+1⊂C​e​n​t​e​r​(G/G≥k+1)G^{\geq k}/G^{\geq k+1}\subset Center(G/G^{\geq k+1}). The last inclusion holds because [𝐠,𝐠≥k]=[𝐠≥1,𝐠≥k]⊂𝐠≥k+1[{\bf g},{\bf g}^{\geq k}]=[{\bf g}^{\geq 1},{\bf g}^{\geq k}]\subset{\bf g}^{\geq k+1}.

Next we observe that the isomorphism of abelian Lie algebras

⨁1≤i≤Nk+1𝐠λi≥k/𝐠λi≥k+1≃𝐠≥k/𝐠≥k+1\bigoplus_{1\leq i\leq{N_{k+1}}}{\bf g}_{\lambda_{i}}^{\geq k}/{\bf g}_{\lambda_{i}}^{\geq k+1}\simeq{\bf g}^{\geq k}/{\bf g}^{\geq k+1}

implies an isomorphism of the corresponding abelian groups

∏1≤i≤Nk+1Gi≥k/Gi≥k+1≃G≥k/G≥k+1.\prod_{1\leq i\leq{N_{k+1}}}G_{i}^{\geq k}/G_{i}^{\geq k+1}\simeq G^{\geq k}/G^{\geq k+1}\,\,.

Hence we can write uniquely c=c1​…​cNk+1c=c_{1}\dots c_{N_{k+1}}, where ci∈G≥k/G≥k+1⊂C​e​n​t​e​r​(G/G≥k+1)c_{i}\in G^{\geq k}/G^{\geq k+1}\subset Center(G/G^{\geq k+1}). It follows that ∏→(k+1)((hi​ci−1))=h\prod_{\to}^{(k+1)}\left((h_{i}c_{i}^{-1})\right)=h. Also it is now clear that this decomposition of hh is unique. This concludes the proof. ■\blacksquare

10.3 Function o​r​dlord_{l}

For l∈ℒl\in{\cal L} we will define an order function

o​r​dl∈Γ⁡((0,+∞),fl∗​(A​f​f𝐙,Y))ord_{l}\in\Gamma((0,+\infty),f_{l}^{\ast}(Aff_{{\bf Z},Y}))

(its meaning will become clear later) by the following inductive procedure:

  1. 1.

    Let l∈ℒi​nl\in{\cal L}_{in} and t>0t>0 be sufficiently small. Then in the standard affine coordinates near s=fl​(0)s=f_{l}(0) one has αl=±fl∗​(d​y)\alpha_{l}=\pm f_{l}^{\ast}(dy). We define o​r​dl=±fl∗​(y)ord_{l}=\pm f_{l}^{\ast}(y). Then d⁡(o​r​dl)=αld(ord_{l})=\alpha_{l}, and we can extend uniquely o​r​dlord_{l} for all t∈(0,+∞)t\in(0,+\infty).

  2. 2.

    Let l∈ℒc​o​ml\in{\cal L}_{com} and l1,l2l_{1},l_{2} be parents of ll. In the notation of Axiom 3 we have fl1​(t1)=fl2​(t2)=fl​(0)f_{l_{1}}(t_{1})=f_{l_{2}}(t_{2})=f_{l}(0) and αl​(0)=n1​αl1​(t1)+n2​αl2​(t2)\alpha_{l}(0)=n_{1}\alpha_{l_{1}}(t_{1})+n_{2}\alpha_{l_{2}}(t_{2}). Then we define o​r​dl​(0):=n1​o​r​dl1​(t1)+n2​o​r​dl2​(t2)ord_{l}(0):=n_{1}ord_{l_{1}}(t_{1})+n_{2}ord_{l_{2}}(t_{2}). Again, using the condition d⁡(o​r​dl)=αld(ord_{l})=\alpha_{l} and the knowledge of o​r​dl​(0)ord_{l}(0) we can extend o​r​dlord_{l} for t>0t>0.

Notice that o​r​dl​(t)ord_{l}(t) can be thought of as affine function on the tangent space Tfl​(t)​YT_{f_{l}(t)}Y (in the induced integral affine structure). In particular, we have a half-plane Pl,t⊂Tfl​(t)​YP_{l,t}\subset T_{f_{l}(t)}Y defined by the inequality o​r​dl​(t)>0ord_{l}(t)>0. The family of half-planes Pl,tP_{l,t} is covariantly constant with respect to ∇a​f​f\nabla^{aff}.

Each half-plane Pl,tP_{l,t} contains 0∈Tfl​(t)​Y0\in T_{f_{l}(t)}Y strictly in its interior. Recall that at the end of Section 9.1 we defined another half-plane Pl,t(0)⊂Tfl​(t)​YP_{l,t}^{(0)}\subset T_{f_{l}(t)}Y. It is easy to see that Pl,t(0)P_{l,t}^{(0)} is the half-plane parallel to Pl,tP_{l,t} such that 0∈Tfl​(t)​Y0\in T_{f_{l}(t)}Y is on the boundary of Pl,t(0)P_{l,t}^{(0)}.

10.4 Symplectomorphisms assigned to lines

In this section we are going to assign to each line l∈ℒl\in{\cal L} a symplectomorphism

φl∈Γ⁡((0,+∞),fl∗​(S​y​m​p))\varphi_{l}\in\Gamma\left((0,+\infty),f_{l}^{*}\left(Symp\right)\right)

giving for each t>0t>0 a transformation φl​(t):𝒪Y,fl​(t)c​a​n→𝒪Y,fl​(t)c​a​n\varphi_{l}(t):{\cal O}^{can}_{Y,f_{l}(t)}\to{\cal O}^{can}_{Y,f_{l}(t)}. This symplectomorphism in local coordinates will belong to the subgroup GλG_{\lambda} where λ\lambda is the slope of αl​(t)\alpha_{l}(t). More precisely, we demand that φl​(t)\varphi_{l}(t) is of the form

φl​(t)=exp⁡{Fl,t​(ξ−a​η−b),⋅},\varphi_{l}(t)=\exp\{F_{l,t}(\xi^{-a}\eta^{-b}),\cdot\}\,\,,

where αl​(t)=a​d​x+b​d​y\alpha_{l}(t)=adx+bdy, operation {⋅,⋅}\{\cdot,\cdot\} is the Poisson bracket on 𝒪Y,fl​(t)c​a​n{\cal O}^{can}_{Y,f_{l}(t)} and Fl,t​(z)∈z​K​[[z]]F_{l,t}(z)\in zK[[z]] is an analytic function of one variable satisfying the following condition. Let us consider the pullback (by the exponential map) of the function Fl,t​(ξ−a​η−b)F_{l,t}(\xi^{-a}\eta^{-b}) to a section of the sheaf 𝒪c​a​n{\cal O}^{can} on vector space Tfl​(t)​Y≃𝐑2T_{f_{l}(t)}Y\simeq{\bf R}^{2} considered as a manifold with 𝐙{\bf Z}-affine structure. Then this pullback should admit an analytic continuation from 0∈Tfl​(t)0\in T_{f_{l}(t)} to the half-plane Pl,tP_{l,t}, and obey there the bound

|Fl,t​(ξ−a​η−b)|≤exp⁡(−o​r​dl​(t)).|F_{l,t}(\xi^{-a}\eta^{-b})|\leq\exp(-ord_{l}(t))\,\,.

Let us explain the construction of φl​(t)\varphi_{l}(t), leaving the justification for the next sections.

Symplectomorphisms φl\varphi_{l} are constructed by an inductive procedure. Let l=l+∈ℒi​nl=l_{+}\in{\cal L}_{in} be (in standard affine coordinates) a line in the half-plane y>0y>0 emerging from (0,0)(0,0) (there is another such line l−l_{-} in the half-plane y<0y<0). Assume that tt is sufficiently small. Then we define φl​(t)∈S​y​m​pfl​(t)\varphi_{l}(t)\in Symp_{f_{l}(t)} on topological generators ξ,η\xi,\eta by the formula (as in Section 8)

φl​(t)​(ξ,η)=(ξ⁡(1+1/η),η).\varphi_{l}(t)(\xi,\eta)=(\xi(1+1/\eta),\eta)\,\,.

Notice that φl​(t)=exp⁡{F⁡(η−1),⋅}\varphi_{l}(t)=\exp\{F(\eta^{-1}),\cdot\}, where F⁡(z)=∑n>0(−1)n​zn/n2F(z)=\sum_{n>0}(-1)^{n}z^{n}/n^{2} is convergent for |z|<1|z|<1.

In order to extend φl​(t)\varphi_{l}(t) to the interval (0,t0)(0,t_{0}), where t0t_{0} is not small, we cover the corresponding segment of ll by open charts. Notice that change of affine coordinates transforms η\eta into a monomial multiplied by a constant from K×K^{\times}. Therefore η\eta extends analytically in a unique way to a global section over (0,+∞)(0,+\infty) of the sheaf fl∗​((𝒪c​a​n)×)f_{l}^{\ast}(({\cal O}^{can})^{\times}). Moreover, the norm |η||\eta| strictly decreases as tt increases, and remains strictly smaller than 11. Hence F⁡(η)F(\eta) can be canonically extended for all t>0t>0.

Each symplectomorphism φl​(t)\varphi_{l}(t) is defined by a series which converges in the half-plane Pl,tP_{l,t}. Using the exponential map associated with the affine structure as well as estimates of o​r​dl​(t)ord_{l}(t), we can extend analytically φl​(t)\varphi_{l}(t) into a neighborhood of fl​(t)f_{l}(t).

Let us now assume that l1l_{1} and l2l_{2} collide at p=fl1​(t1)=fl2​(t2)p=f_{l_{1}}(t_{1})=f_{l_{2}}(t_{2}), generating the line l∈ℒc​o​ml\in{\cal L}_{com}. Then φl​(0)\varphi_{l}(0) is defined with the help of factorization theorem in the group GG. More precisely, we set αi:=αli(ti),i=1,2\alpha_{i}:=\alpha_{l_{i}}(t_{i}),\,\,i=1,2 and the angle VV to be the intersection of half-planes Pl1,t1∩Pl2,t2P_{l_{1},t_{1}}\cap P_{l_{2},t_{2}}. By construction elements g0:=φl1​(t1)g_{0}:=\varphi_{l_{1}}(t_{1}) and g+∞:=φl2​(t2)g_{+\infty}:=\varphi_{l_{2}}(t_{2}) belong respectively to G0G_{0} and G+∞G_{+\infty}. Then we can use the factorization Theorem 6 and write down the formula

g+∞​g0=∏→((gλ)λ∈[0,+∞]𝐐)=g0​…​g1/2​…​g1​…​g+∞,g_{+\infty}g_{0}={\textstyle\prod_{\to}}\left((g_{\lambda})_{\lambda\in[0,+\infty]_{\bf Q}}\right)=g_{0}\dots g_{1/2}\dots g_{1}\dots g_{+\infty}\,\,,

where gλ∈Gλg_{\lambda}\in G_{\lambda} and the product on the right is in the increasing order. There is no clash of notations because it is easy to see that the boundary factors in the decomposition from above are indeed equal to g0g_{0} and g+∞g_{+\infty}. Each term gλg_{\lambda} with 0<λ=n1/n2<+∞0<\lambda=n_{1}/n_{2}<+\infty corresponds to the newborn line ll with the direction covector n1​αl1​(t1)+n2​αl2​(t2)n_{1}\alpha_{l_{1}}(t_{1})+n_{2}\alpha_{l_{2}}(t_{2}). Then we set φl​(0):=gλ\varphi_{l}(0):=g_{\lambda}. This transformation is defined by a series which is convergent in a neighborhood of pp, and using the analytic continuation as above, we obtain φl​(t)\varphi_{l}(t) for t>0t>0. The decomposition identity can be rewritten as

g0​…​g1/2​…​g1​…​g+∞​g0−1​g+∞−1=i​dg_{0}\dots g_{1/2}\dots g_{1}\dots g_{+\infty}g_{0}^{-1}g_{+\infty}^{-1}=id

where each factor corresponds to half-lines at the collision point (see Figure 5), and the meaning of the identity is that the infinite composition of symplectomorphisms in the natural cyclic order on half-lines, is trivial.

11 Modification of the sheaf 𝒪c​a​n{\cal O}^{can}

11.1 Pieces of lines and convergence regions

Definition 13

A neighborhood UU of a point x∈Yx\in Y is convex if there exists an open convex U1∈Tx​Y,0∈U1U_{1}\in T_{x}Y,0\in U_{1} which is isomorphic to UU by means of the exponential map expx:Tx​Y→Y\exp_{x}:T_{x}Y\to Y associated with the affine structure on YY.

For x∈Yx\in Y let U⊂U′U\subset U^{\prime} be convex neighborhoods of xx such that UU is relatively compact in U′U^{\prime}. Let l∈ℒl\in{\cal L}. Then there is a natural embedding fl−1​(U)→fl−1​(U′)f_{l}^{-1}(U)\to f_{l}^{-1}(U^{\prime}).

Definition 14

A piece of ll defined by the pair (U,U′)(U,U^{\prime}) is an element of the image of the set of connected components π0​(fl−1​(U))\pi_{0}(f_{l}^{-1}(U)) into π0​(fl−1​(U′))\pi_{0}(f_{l}^{-1}(U^{\prime})) under the above embedding.

In plain words a piece LL of ll is an equivalence class of a connected interval of l∩Ul\cap U. Two connected intervals are equivalent if they are contained in a larger connected interval of l∩U′l\cap U^{\prime}. The sole purpose of the introduction of the notion of a piece is to avoid some pathology. Namely, for any pair (U,U′)(U,U^{\prime}) as above, any l∈ℒl\in{\cal L} and any T∈𝐑>0T\in{\bf R}_{>0}, there is only a finite number of pieces of ll in (U,U′)(U,U^{\prime}) which have points with time parameter t∈(0,T)t\in(0,T).

Let LL be a piece of ll defined by a pair (U,U′)(U,U^{\prime}). Then one can define an affine function o​r​dL∈A​f​f𝐙,Y​(U′)ord_{L}\in Aff_{{\bf Z},Y}(U^{\prime}) in the following way. Let t>0t>0 be such that fl​(t)f_{l}(t) belongs to LL. Since U′U^{\prime} is convex, there is a unique continuation of o​r​dl​(t)ord_{l}(t) to U′U^{\prime}. This is an affine function which does not depend on the choice of tt. We will denote it by o​r​dLord_{L}.

For any germ of a symplectomorphism φ∈S​y​m​pp\varphi\in Symp_{p} at a point p∈Yp\in Y we define its convergence region as the maximal convex subset Ω⁡(φ)⊂Tp​Y\Omega(\varphi)\subset T_{p}Y such that the pullback expp∗⁡(φ)\exp_{p}^{\ast}(\varphi) extends to Ω⁡(φ)\Omega(\varphi). Since the definition of φl\varphi_{l} (and hence its convergence region) is covariant with respect to the affine connection we have the following result:

Proposition 5

Let p=fl​(t)p=f_{l}(t) belongs to a line ll. Then the convergence region of φl​(t)\varphi_{l}(t) at pp contains an open half-plane Pl,tP_{l,t}.

It is clear that one can define convergence regions for symplectomorphisms associated with pieces of lines, and a similar property holds for them.

11.2 Main assumptions, and an apology

Let us suppose that our collection of lines satisfies the following assumptions:

Assumption A1

There is a smooth metric g=gBg=g_{B} and a collection of balls D⁡(s,rs)D(s,r_{s}) with centers at s∈Bs​i​n​gs\in B^{sing} such that each ball D⁡(s,rs)D(s,r_{s}) contains exactly two lines l±∈ℒi​nl_{\pm}\in{\cal L}_{in} outcoming of ss.

Assumption A2

There exists ε>0\varepsilon>0 such that for any p=fl(t)∈Y′:=B∖∪s∈Bs​i​n​gD(s,rs)p=f_{l}(t)\in Y^{\prime}:=B\setminus\cup_{s\in B^{sing}}D(s,r_{s}) the distance in Tp​YT_{p}Y between 0∈Tp​Y0\in T_{p}Y and the boundary of Pl,tP_{l,t} is greater or equal to ε\varepsilon.

We are going to show that such a collection does exist in Section 11.

Assumptions A1 and A2 are very artificial, they do not hold in physical picture which is the main motivation for the construction. It is quite possible that they can be weakened or even omitted. The main purpose of introducing them here is the possibility to define the sheaf of analytic functions by simple gluing. In complex geometry it is similar to the gluing of closed Riemann surfaces with boundaries by the mean of real-analytic identifications of the boundaries. It is well-known that one can replace real-analytic maps by smooth ones (or even by quasi-symmetric continuous maps). Maybe the rest of this section is unnecessary, and unpleasant technical arguments in Section 11.5 can be avoided.

11.3 Infinite product and its convergence

Denote by Wℒ:=∪l∈ℒfl([0,+∞))W_{\cal L}:=\cup_{l\in{\cal L}}f_{l}([0,+\infty)) the set of all points of all lines. It has measure zero. Let pp be a point of YY. We consider two convex neighborhoods U⊂U′U\subset U^{\prime} of pp such that UU is relatively compact in U′U^{\prime}.

For any two points x,yx,y belonging to U∖WℒU\setminus W_{\cal L}, and a path γ\gamma joining xx and yy in UU, we would like to define an infinite ordered product ix,yγi_{x,y}^{\gamma} of transformations φL±1\varphi_{L}^{\pm 1}, where factors correspond to the intersection points of γ\gamma with all possible pieces LL relative to (U,U′)(U,U^{\prime}). Factors in the infinite product are ordered according to the time parameter of γ\gamma, the sign corresponds to the mutual position of orientations of γ\gamma and a piece LL at the intersection point.

In order to give a precise meaning to the infinite product the neighborhood UU of pp should be sufficiently small. Then we will have an analytic continuation of symplectomorphisms φL\varphi_{L} to UU, and the convergence of the infinite product. We are also going to prove that the product is independent of the choice of path γ\gamma. In order to achieve these goals it suffices to assume:

C1

for any l,tl,t such fl​(t)∈Uf_{l}(t)\in U the set (expfl​(t))−1​(U)(\exp_{f_{l}(t)})^{-1}(U) is contained in Pl,tP_{l,t}\,;

C2

for any C∈𝐑C\in{\bf R} there is only a finite number of pieces LL of lines in UU such that infx∈Uo​r​dL​(x)<C\inf_{x\in U}ord_{L}(x)<C.

Theorem 7

Assume two above conditions. Then the product defining ix,yγi_{x,y}^{\gamma} converges at every point of UU and in fact gives an element of S​y​m​p​(U)Symp(U). Moreover, the product does not depend on the choice of path γ\gamma, and for any x,y,z∈U∖Wℒx,y,z\in U\setminus W_{\cal L} satisfies the relation ix,y​iy,z=ix,zi_{x,y}i_{y,z}=i_{x,z}.

Proof. Condition C1 implies that all transformations φL\varphi_{L} admit an analytic continuation to UU. Let us introduce a decreasing filtration by positive real numbers S​y​m​p≥r​(U),r∈𝐑,r≥0Symp^{\geq r}(U),\,\,r\in{\bf R},r\geq 0 on group S​y​m​p​(U)Symp(U) by the formula

{g∈Symp(U)|log|ξ′/ξ−1|,log|η′/η−1|<−r where (ξ′,η′)=g((ξ,η))}\left\{g\in Symp(U)\,|\,\,\log|\xi^{\prime}/\xi-1|,\log|\eta^{\prime}/\eta-1|<-r\,\mbox{ where }(\xi^{\prime},\eta^{\prime})=g((\xi,\eta))\,\right\}

This is a complete filtration, and condition C2 implies that in any quotient S​y​m​p​(U)/S​y​m​p≥r​(U)Symp(U)/Symp^{\geq r}(U) only a finite number of elements φL\varphi_{L} are non-trivial. Therefore we can define the product in the quotient group.

In order to prove independence of γ\gamma, we consider the quotient group S​y​m​p​(U)/S​y​m​p≥r​(U)Symp(U)/Symp^{\geq r}(U), and the finite 11-dimensional CW-complex (graph) consisting of finitely many pieces LL, such that φL≠1\varphi_{L}\neq 1 in the quotient. For each vertex vv of the graph there is a natural cyclic order on the edges incident to vv. The product φv=∏LφL±1\varphi_{v}=\prod_{L}\varphi_{L}^{\pm 1} taken in the cyclic order over the set of edges incident to vv is equal to i​did (this follows from the construction of φl\varphi_{l} via factorizations). Since UU is simply-connected, we conclude that the image of ix,yγi_{x,y}^{\gamma} in S​y​m​p​(U)/S​y​m​p≥r​(U)Symp(U)/Symp^{\geq r}(U) does not depend on γ\gamma. Using completeness of the filtration we see that ix,y:=ix,yγi_{x,y}:=i_{x,y}^{\gamma} does not depend on γ\gamma. Proof of the identity ix,y​iy,z=ix,zi_{x,y}i_{y,z}=i_{x,z} is similar. ■\blacksquare

Theorem 8

Assumptions A1 and A2 imply that for any p∈Yp\in Y there exist neighborhood UU (and also U′U^{\prime}) satisfying conditions C1 and C2.

Proof. Assumption A1 implies that the result near any singular point s∈Bs​i​n​gs\in B^{sing}\,, as there are only two lines near ss. If we are far from Bs​i​n​gB^{sing} then obviously A2 implies C1.

In order to check C2 we prove the following lemma

Lemma 5

Under Assumptions A1 and A2, for any C>0C>0 the set

{(l,t)|o​r​dl​(t)​(fl​(t))<C}⊂ℒ×(0,+∞)\{(l,t)|\,\,ord_{l}(t)(f_{l}(t))<C\,\}\subset{\cal L}\times(0,+\infty)

consists of a finite number of intervals.

Proof: We proceed by induction in “complexity of the line”. Let δ∈𝐑>0\delta\in{\bf R}_{>0} be the infimum of o​r​dl​(t)​(fl​(t))ord_{l}(t)(f_{l}(t)) where l∈ℒi​nl\in{\cal L}_{in} has a collision at time tt. This number is strictly positive because the number of initial lines is finite, and by A1 there is no collisions at small times. Observe that the value of o​r​dl​(0)ord_{l}(0) at the beginning of any composite line ll is greater or equal to the sum o​r​dl1​(t1)+o​r​dl2​(t2)ord_{l_{1}}(t_{1})+ord_{l_{2}}(t_{2}). Therefore the inequality in the lemma implies that the number of collisions is bounded from above by C/δC/\delta. Also we have an upper bound on integer coefficients (n1,n2)(n_{1},n_{2}) in each collision (see Axiom OPEN𝟑​𝐛)\bf 3b) in Section 9.2). Let us observe that the length of each edge of the ansector tree of ll is also bounded from above by A​o​r​dlA\,ord_{l}, for some absolute constant A>0A>0. Hence we have only finitely many possibilities for intersections. ■\blacksquare

For point p∈Yp\in Y which is far from Bs​i​n​gB^{sing} we chose as UU a neighborhood of radius ϵ′≪ϵ\epsilon^{\prime}\ll\epsilon where ϵ>0\epsilon>0 is constant from Assumption A2. Then for any point of a line fl​(t)∈Uf_{l}(t)\in U we will have the inclusion

U⊂expfl​(t)⁡(12​Pl,t).U\subset\exp_{f_{l}(t)}\left(\frac{1}{2}P_{l,t}\right)\,\,.

This implies that o​r​dLord_{L} in UU for the corresponding piece LL is bounded below by

12​o​r​dl​(t)​(fl​(t)).\frac{1}{2}ord_{l}(t)(f_{l}(t))\,\,.

Since (by the last lemma) there exists only a finite number of pieces LL intersecting such UU, we obtain convergence condition C2. ■\blacksquare

11.4 Construction of the modified sheaf 𝒪Bm​o​d​i​f{\cal O}^{modif}_{B}

For any point p∈Yp\in Y and a neighborhood UU satisfying conditions C1 and C2 we define the sheaf 𝒪Um​o​d​i​f{\cal O}^{modif}_{U} as the result of the identification of copies of the sheaf (𝒪Yc​a​n)|U\left({\cal O}^{can}_{Y}\right)_{|U} labeled by points x∈U∖Wℒx\in U\setminus W_{\cal L}, by isomorphisms ix,yi_{x,y}. It follows from formulas in Section 8 that near singular points one can identify canonically this sheaf with the restriction of the sheaf 𝒪𝐑2m​o​d​e​l{\cal O}^{model}_{{\bf R}^{2}} to a punctured neighborhood of (0,0)∈𝐑2(0,0)\in{\bf R}^{2}.

Proposition 6

For the modified sheaf 𝒪m​o​d​i​f{\cal O}^{modif} one has a canonical nowhere vanishing section Ω\Omega of the associated sheaf of KK-analytic 22-forms.

The KK-affine structure A​f​fK,YΩAff^{\Omega}_{K,Y} on YY associated with Ω\Omega coincides with the initial one A​f​fK,YAff_{K,Y}.

Proof. Existence of Ω\Omega follows from the fact that all modifications associated with lines are symplectomorphisms. In order to finish the proof it suffices to check that the modification associated with a line does not change the KK-affine structure on YY. In local coordinates we may assume that Ω=d​ξξ∧d​ηη\Omega={d\xi\over\xi}\wedge{d\eta\over\eta} and the modification is of the form φ⁡(ξ,η)=(ξ​f​(η−1),η)\varphi(\xi,\eta)=(\xi f(\eta^{-1}),\eta), where f⁡(z)=1+∑n≥1cn​zn∈K⁡[[z]]f(z)=1+\sum_{n\geq 1}c_{n}z^{n}\in K[[z]] is convergent in an appropriate domain. We need to check that the automorphism φ\varphi acts trivially on the quotient sheaf π∗​(𝒪X×)/ker⁡pΩ\pi_{\ast}({\cal O}_{X}^{\times})/\ker\,p_{\Omega} (see Section 7.2 for the notation). This check reduces to the calculation of

pΩ​(ξ​f​(η)ξ)=exp⁡(OPENR​e​s​(Ω​log⁡(ξ​f​(η)/ξ)))R​e​s​(Ω)).p_{\Omega}\left({\xi f(\eta)\over\xi}\right)=\exp\left({Res(\Omega\log(\xi f(\eta)/\xi)))\over Res(\Omega)}\right)\,\,.

The latter is equal to exp⁡(R​e​s​(Ω​log⁡(f⁡(η))))=1\exp\left(Res(\Omega\log(f(\eta)))\right)=1 because log⁡(f⁡(η−1))\log(f(\eta^{-1})) belongs to η−1​K​[[η−1]]\eta^{-1}K[[\eta^{-1}]] and therefore has no constant term. ■\blacksquare

Thus, we have a solution of the Lifting Problem under Assumptions A1 and A2.

11.5 Construction of the collection of lines

We would like to show that there exists a smooth metric gg and a collection of lines satisfying the Assumptions A1 and A2.

Let g0g_{0} be an arbitrary smooth metric, flat near singular points. We define germs of lines l∈ℒi​nl\in{\cal L}_{in} in such a way that for each s∈Bs​i​n​gs\in B^{sing} in local coordinates these lines are given by {(0,y)|y>0}\{(0,y)|y>0\} and {(0,y)|y<0}\{(0,y)|y<0\}. The metric gg will coincide with g0g_{0} in a sufficiently small neighborhood U=∪s∈Bs​i​n​gD(s,rs)U=\cup_{s\in B^{sing}}D(s,r_{s}) of the singular set. Hence Assumption A1 will be satisfied.

In order to construct the whole family of lines we introduce a 33-dimensional manifold ℳ{\cal M} consisting of pairs (x,P)(x,P) where x∈B∖U¯1x\in B\setminus\overline{U}_{1} and PP is a half-plane in Tx​BT_{x}B whose boundary contains zero. Here U1:=∪s∈Bs​i​n​gD(s,2rs)U_{1}:=\cup_{s\in B^{sing}}D(s,2r_{s}) is a larger neighborhood of Bs​i​n​gB^{sing}.

We would like to construct a smooth section v:(x,P)↦v(x,P)∈Tx​Bv:(x,P)\mapsto v_{(x,P)}\in T_{x}B of the pull-back to ℳ{\cal M} of the tangent bundle T​BTB satisfying the following conditions:

  1. 1.

    for any (x,P)∈ℳ(x,P)\in{\cal M} one has v(x,P)∈i​n​t​(P)v_{(x,P)}\in int(P)\,\,;

  2. 2.

    for any x∈B∖U¯1x\in B\setminus\overline{U}_{1} the map (x,P)↦𝐑>0×⋅v(x,P)(x,P)\mapsto{{\bf R}}_{>0}^{\times}\cdot v_{(x,P)} is an orientation-preserving diffeomorphism

    S1≃(Tx​B∗∖{0})/𝐑>0×→S1≃(Tx​B∖{0})/𝐑>0×;S^{1}\simeq(T_{x}B^{\ast}\setminus\{0\})/{{\bf R}}_{>0}^{\times}\to S^{1}\simeq(T_{x}B\setminus\{0\})/{{\bf R}}_{>0}^{\times}\,\,\,;
  3. 3.

    for every l∈ℒi​nl\in{\cal L}_{in} there exists a smooth extension of the piece of ll in U1U_{1} to a larger piece intersecting ∂U1\partial U_{1} such that

    f˙l​(t)∈𝐑>0×⋅v(fl​(t),Pl,t),\dot{f}_{l}(t)\in{{\bf R}}_{>0}^{\times}\cdot v_{(f_{l}(t),P_{l,t})}\,\,,

    for such t>0t>0 that fl​(t)∈B∖U¯1f_{l}(t)\in B\setminus\overline{U}_{1}\,\,;

Let us associate with the section vv a nowhere vanishing vector field v^\hat{v} on T∗​(B∖U¯1)∖(Z​e​r​o​S​e​c​t​i​o​n)T^{\ast}(B\setminus\overline{U}_{1})\setminus(Zero\,\,\,Section) in the following way:

  • •

    For each (x,α)∈Tx∗​B(x,\alpha)\in T_{x}^{\ast}B the vector v^​(x,α)\hat{v}(x,\alpha) is tangent to the horizontal distribution associated with the flat connection ∇\nabla (the one which defines the affine structure on B∖Bs​i​n​gB\setminus B^{sing}).

  • •

    Projection of v^​(x,α)\hat{v}(x,\alpha) to BB coincides with v(x,Pα)v_{(x,P_{\alpha})}, where Pα={γ|(α,γ)>0}P_{\alpha}=\{\gamma|(\alpha,\gamma)>0\}.

Clearly these conditions determines v^\hat{v} uniquely. Now we formulate last condition:

  1. 4.

    there exits rs′>2​rsr_{s}^{\prime}>2r_{s} such that for almost all (in the sense of Baire category) initial values (x0,P0)∈ℳ(x_{0},P_{0})\in{\cal M} the integral curve of v^\hat{v} starting at (x0,P0)(x_{0},P_{0}) reaches the pullback of B∖∪s∈Bs​i​n​gD(s,rs′)B\setminus\cup_{s\in B^{sing}}D(s,r_{s}^{\prime}) in finite time.

Using the vector field v^\hat{v} we will construct (under certain genericity assumptions) a set ℒ{\cal L} of lines satisfying Assumption A1. Namely, the data consisting of a line ll and an integer-valued 11-form αl\alpha_{l} (see Section 9) will be an integral line of v^\hat{v}.

We are going to construct lines by induction by the number of collisions. Lines l∈ℒi​nl\in{\cal L}_{in} will be constructed using condition 3. The genericity assumption mentioned after the condition 4 is the assumption that no more than two lines collide and that initial values for newborn lines will be sufficiently generic. Conditions 1 and 4 plus genericity imply that one can parametrize any line l∈ℒl\in{\cal L} by the new “time” t>0t>0 such that the Axiom 2 is satisfied. Axiom 6 follows from the condition 2. Other axioms and the Assumption A1 will be satisfied automatically.

Now we would like to discuss Assumption A2.

Proposition 7

Suppose that the metric gg and field vv described above are such that for any (x,P)∈ℳ(x,P)\in{\cal M} there exists C>0C>0 such that

(∇v(x,P)g)​(nP,nP)≤C​g​(nP,v(x,P)),(\nabla_{v_{(x,P)}}\,g)(n_{P},n_{P})\leq C\,g(n_{P},v_{(x,P)})\,\,,

where nPn_{P} is the normal unit vector to PP directed inside and ∇v(x,P)g\nabla_{v_{(x,P)}}\,g is the covariant derivative of the metric gg considered as a symmetric tensor on the cotangent bundle.

Then the Assumption A2 is satisfied.

Proof. In order to satisfy Assumption A2 it suffices to find such ε>0\varepsilon>0 that for any x∈B∖U¯1x\in B\setminus\overline{U}_{1} and any half-plane Px⊂Tx​B,  0∈i​n​t​(Px)P_{x}\subset T_{x}B,\,\,0\in int(P_{x}) with the distance d​i​s​tgx​(0,∂Px)=εdist_{g_{x}}(0,\partial P_{x})=\varepsilon, and another half-plane Px′⊂Tx​BP_{x}^{\prime}\subset T_{x}B parallel to PxP_{x} such that 0∈∂Px′0\in\partial P_{x}^{\prime}, one has the following property: if Px+δ​t​v(x,Px′)P_{x+\delta tv_{(x,P_{x}^{\prime})}} is the half-plane obtained from PxP_{x} by a small covariant (with respect to the affine connection ∇a​f​f\nabla^{aff}) shift δ​t\delta t in the direction of v(x,Px′)v_{(x,P_{x}^{\prime})}, then

d​i​s​tgx+δ​t​v(x,Px′)​(0,Px+δ​t​v(x,Px′))≥d​i​s​tgx​(0,∂Px).dist_{g_{x+\delta tv_{(x,P_{x}^{\prime})}}}(0,P_{x+\delta tv_{(x,P_{x}^{\prime})}})\geq dist_{g_{x}}(0,\partial P_{x})\,\,.

Here gxg_{x} etc. denotes the induced flat metric on the tangent space Tx​BT_{x}B. This property guarantees that the condition d​i​s​tgx​(0,∂Pl,t)≥εdist_{g_{x}}(0,\partial P_{l,t})\geq\varepsilon will propagate along the line. For a new line obtained as a result of collision of l1l_{1} and l2l_{2} at the times t1t_{1} and t2t_{2} respectively one has

d​i​s​tgx​(0,∂Pl,0)≥min⁡{d​i​s​tgx​(0,∂Pl1,t1),d​i​s​tgx​(0,∂Pl2,t2)}dist_{g_{x}}(0,\partial P_{l,0})\geq\min\{dist_{g_{x}}(0,\partial P_{l_{1},t_{1}}),dist_{g_{x}}(0,\partial P_{l_{2},t_{2}})\}

since ∂Pl,0\partial P_{l,0} contains the intersection point ∂Pl1,t1∩∂Pl2,t2\partial P_{l_{1},t_{1}}\cap\partial P_{l_{2},t_{2}}, see Figure 7.

Refer to caption

Figure 7: Three half-planes containing zero.

One can easily see that the infinitesimal inequality from above is equivalent to

δ​t​gx​(v(x,Px′),nPx′)+ε/2​(gx+δ​t​v(x,Px′)−gx)​(nPx′,nPx′)≥0\delta tg_{x}(v_{(x,P_{x}^{\prime})},n_{P_{x}^{\prime}})+\varepsilon/2(g_{x+\delta tv_{(x,P_{x}^{\prime})}}-g_{x})(n_{P_{x}^{\prime}},n_{P_{x}^{\prime}})\geq 0

(the change of the distance consists of two summands: one corresponds to the shift along δ​t​v(x,Px)\delta tv_{(x,P_{x})} with the fixed metric, and the other one corresponds to the change of the metric). Taking the limit δ​t→0\delta t\to 0 we arrive to the inequality for the covariant derivative of the metric with C=2/εC=2/\varepsilon. ■\blacksquare

Now our goal is to construct the field of directions vv and the metric gg satisfying the conditions 1–4 and the inequality from the last Propostion. This will conclude the construction of the set ℒ{\cal L} of lines satisfying the Assumptions A1 and A2.

Since ∂U1\partial U_{1} is a boundary of the convex set, we can locally model it by the graph of function y=f⁡(x)y=f(x) such that f′′​(x)>0f^{\prime\prime}(x)>0, f′​(x0)=0f^{\prime}(x_{0})=0. We may assume that P=P0P=P_{0} is the upper half-plane. Then we take

v((x,y),P)=∂/∂y+(f⁡(x)−f⁡(x0))/f′​(x)f⁡(x)−f⁡(x0)+f⁡(x)−y∂/∂x.v_{\left((x,y),P\right)}=\partial/\partial y+{(f(x)-f(x_{0}))/f^{\prime}(x)\over{f(x)-f(x_{0})+f(x)-y}}\,\,\partial/\partial x\,\,.

We extend this local model of vv near ∂U1\partial U_{1} to B∖U¯1B\setminus\overline{U}_{1} in such a way that conditions 1 and 2 are satisfied. It is clear that we can satisfy conditions 3,4 as well by taking a small perturbation of vv. On Figure 8 there is a picture of the field (x,y)↦v((x,y),P0)(x,y)\mapsto v_{\left((x,y),P_{0}\right)}.

Refer to caption

Figure 8: Vector field near ∂U1\partial U_{1} for P= the upper half-planeP=\mbox{ the upper half-plane}.

For an arbitrary choice of the metric gg we have g⁡(nP,vz,P)>0g(n_{P},v_{z,P})>0 for all (z,P)∈ℳ(z,P)\in{\cal M}. The problem with inequality

(∇v(z,P)g)​(nP,nP)≤C​g​(nP,v(z,P))(\nabla_{v_{(z,P)}}\,g)(n_{P},n_{P})\leq Cg(n_{P},v_{(z,P)})

arises only as the point zz approaches ∂U1\partial U_{1}. Indeed, in this case the vector v(z,P)v_{(z,P)} can be very close to the tangent vector to ∂Pz⊂Tz​B\partial P_{z}\subset T_{z}B.

Lemma 6

With the above choice of vv assume that the metric satisfies for any z∈∂U1z\in\partial U_{1} the condition

(∇ezg)​(nz,nz)=0,(\nabla_{e_{z}}\,g)(n_{z},n_{z})=0\,\,,

where ez∈Tz​Be_{z}\in T_{z}B is the unit tangent vector to ∂U1\partial U_{1} and nzn_{z} is the normal vector to ∂U1\partial U_{1} (all scalar products and lengths are taken with respect to the metric gg).

Then there exists C>0C>0 such that

(∇v(z,P)g)​(nP,nP)≤C​g​(nP,v(z,P))(\nabla_{v_{(z,P)}}\,g)(n_{P},n_{P})\leq Cg(n_{P},v_{(z,P)})

for all (z,P)∈ℳ(z,P)\in{\cal M}.

Proof. We need to check that the ratio

(∇v(z,P)g)​(nP,nP)g⁡(nP,v(z,P)){(\nabla_{v_{(z,P)}}\,g)(n_{P},n_{P})}\over{g(n_{P},v_{(z,P)})}

is bounded for (z,P)∈ℳ(z,P)\in{\cal M}.

It suffices to prove the Lemma assuming that U1U_{1} is the parabolic domain {(x,y)∈𝐑2|y>x2}\{(x,y)\in{{\bf R}}^{2}|y>x^{2}\} and PP is the upper half-plane. The vector field v(z,P)v_{(z,P)} is given for z=(x,y)z=(x,y) by the formulas

v(z,P)=∂/∂y+x4​x2−2​y∂/∂x.v_{(z,P)}=\partial/\partial y+{x\over{4x^{2}-2y}}\,\partial/\partial x\,\,.

The denominator is equal to g⁡(nP,v(z,P))=⟨d​y,v(z,P)⟩⋅g⁡(∂/∂y,∂/∂y)=g⁡(∂/∂y,∂/∂y)=exp⁡(O⁡(1))g(n_{P},v_{(z,P)})=\langle dy,v_{(z,P)}\rangle\cdot\sqrt{g(\partial/\partial y,\partial/\partial y)}=\sqrt{g(\partial/\partial y,\partial/\partial y)}=\exp(O(1)) near (0,0)(0,0).

The numerator is equal to

x4​x2−2​y​f1​(x,y)+f2​(x,y),{x\over{4x^{2}-2y}}f_{1}(x,y)+f_{2}(x,y)\,\,,

where f1​(x,y)=(∇∂/∂xg)​(nP,nP)f_{1}(x,y)=(\nabla_{\partial/\partial x}\,g)(n_{P},n_{P}) and f2​(x,y)=(∇∂/∂yg)​(nP,nP)f_{2}(x,y)=(\nabla_{\partial/\partial y}\,g)(n_{P},n_{P}) are two C∞C^{\infty}-functions.

By assumption of the Lemma we have f1​(0,0)=0f_{1}(0,0)=0. Therefore |f1​(x,y)|≤c​o​n​s​t​max⁡{|x|,|w|}|f_{1}(x,y)|\leq const\,\max\{|x|,|w|\} where w=x2−yw=x^{2}-y is a convenient local coordinate near the point (0,0)(0,0). Notice also that f2​(x,y)=O​(1)f_{2}(x,y)=O(1).

Now we can estimate first summand of the numerator assuming that |x||x| and |w||w| are sufficiently small. As we have seen, it is bounded by

I:=xx2+w​O​(max⁡{|x|,|w|}CLOSE.I:={x\over{x^{2}+w}}O(\max\{|x|,|w|\}\,\,.

There are three cases which we need to consider.

a) If 0<w<x20<w<x^{2} then I=xx2​O​(|x|)=O⁡(1)I={x\over{x^{2}}}O(|x|)=O(1).

b) if x2≤w<xx^{2}\leq w<x then I=xw​O​(|x|)=O⁡(1)I={x\over w}O(|x|)=O(1).

c) If x≤w≤1x\leq w\leq 1 the I=xw​O​(|w|)=O⁡(1)I={x\over w}O(|w|)=O(1).

We see that the numerator is bounded. This concludes the proof of Lemma. ■\blacksquare

Finally, we have the following result.

Lemma 7

There exists metric gg satisfying the conditions of Lemma 6.

Proof: First of all, the condition on gg from Lemma 6 is the condition on a loop g|TzBg_{|T_{z}B} of scalar products on 2-dimensional spaces, here z∈∂U1≃S1z\in\partial U_{1}\simeq S^{1}. We can write g=exp⁡(ψ)​g0g=\exp(\psi)g_{0} where det(g0)=1\det(g_{0})=1 and ψ\psi is a smooth function. Then we have

∇ez(exp⁡ψ​g0)=exp⁡(ψ)​∇ezg0+exp⁡(ψ)​∂ez(ψ)​g0.\nabla_{e_{z}}(\exp{\psi}g_{0})=\exp({\psi})\nabla_{e_{z}}g_{0}+\exp(\psi)\partial_{e_{z}}(\psi)\,g_{0}\,\,.

The equation of Lemma 6 gives ∂ezψ=−(∇ezg0)(nz,nz)/g0(nz,nz)\partial_{e_{z}}\psi=-(\nabla_{e_{z}}g_{0})(n_{z},n_{z})/g_{0}(n_{z},n_{z}). The RHS of this expression is known as long as we know g0g_{0}. Hence we can say that d​ψ=βg0d\psi=\beta_{g_{0}}, where βg0\beta_{g_{0}} is a 1-form depending on the restriction (g0)|∂U1(g_{0})_{|\partial U_{1}}. We see that it suffices to find such g0g_{0} that ∫∂U1βg0=0\int_{\partial U_{1}}\beta_{g_{0}}=0 (then ψ\psi and hence gg does exist).

Let us consider the functional I⁡(g0)=∫S1βg0I(g_{0})=\int_{S^{1}}\beta_{g_{0}}. We can interpret a metric g0g_{0} as a point in the Lobachevsky plane ℋ=S​L​(2,𝐑)/S​O​(2){\cal H}=SL(2,{\bf R})/SO(2). More precisely, let us consider the space SS of pairs (g0,P)(g_{0},P) where g0g_{0} is a positive quadratic form on 𝐑2{\bf R}^{2} such that det(g0)=1\det(g_{0})=1 and PP is a half-plane in 𝐑2{\bf R}^{2} (the meaning of PP is the inward oriented tangent half-plane to ∂U1\partial U_{1} at point z∈∂U1z\in\partial U_{1}). This space is naturally diffeomorphic to S∗​(𝐑2)×ℋS^{\ast}({\bf R}^{2})\times{\cal H}. The latter manifold can be identified in S​L​(2,𝐑)SL(2,{\bf R})-equivariant way with the manifold consisting of pairs (x,y)(x,y), where x∈ℋx\in{\cal H} and yy belongs to the absolute. Hence (g0)|∂U1(g_{0})_{|\partial U_{1}} is (locally) a non-parametrized path in SS (it would be a global path, if the bundle over S1S^{1} given by the all metrics on S1S^{1} with the determinant 11 was trivial).

Next we observe that the variation δ​I​(g0)=∫Nω\delta I(g_{0})=\int_{N}\omega, where NN is a 22-dimensional surface bounded by the paths defined by g0g_{0} and g0+δ​g0g_{0}+\delta g_{0}, and ω\omega is a canonical S​L​(2,𝐑)SL(2,{\bf R})-invariant 22-form on SS. One can show that even by a small variation of the path defined by g0g_{0} we can make I⁡(g0)I(g_{0}) an arbitrary real number. In particular, we can find g0g_{0} such that I⁡(g0)=0I(g_{0})=0. This concludes the proof of Lemma 6. ■\blacksquare

Summarizing, we have constructed a set of lines satisfying the Assumptions A1 and A2. This concludes the proof of Theorem 5. Thus we have obtained a solution of the Lifting Problem, which is a KK-analytic K3 surface.

11.6 Independence and uniqueness

It is natural to ask how the above construction of the KK-analytic K3 surface (Xa​n,Ω)(X^{an},\Omega) depends on the choice of the set ℒ{\cal L} of lines. We know that the “periods” of Ω\Omega (they are encoded in the initial KK-affine structure) do not depend on ℒ{\cal L} (see Sections 7.3, 10.4). In the light of Torelli theorem (see Appendix B) it is natural to formulate the following conjecture.

Conjecture 11

The isomorphism class of the pair (Xa​n,Ω)(X^{an},\Omega) does not depend on the choice of the set ℒ{\cal L} of lines.

More precisely, the change of ℒ{\cal L} corresponds to the change of the projection π:=πℒ:Xa​n→B\pi:=\pi_{\cal L}:X^{an}\to B (see Section 7.3).

Remark 4

For B=S2B=S^{2} and Bs​i​n​g={x1,…,x24}B^{sing}=\{x_{1},\dots,x_{24}\} with the standard singular 𝐙{\bf Z}-affine structure we have constructed a KK-analytic K3 surface depending on 2020 parameters in K×K^{\times}. More precisely, we have a 2020-dimensional KK-analytic space of conjugacy classes of representations

π1​(S2∖Bs​i​n​g)→S​L​(2,𝐙)⋉(K×)2\pi_{1}(S^{2}\setminus B^{sing})\to SL(2,{{\bf Z}})\ltimes(K^{\times})^{2}

such that the monodromy around each singular point is conjugate to the pair (A,(1,1))(A,(1,1)) where A∈S​L​(2,𝐙)A\in SL(2,{{\bf Z}}) is equal to

(1101).\left(\begin{array}[]{cc}1&1\\ 0&1\end{array}\right)\,\,.

(compare with Section 3.3).

11.7 Remark on the case of positive and mixed characteristic

Our construction of (Xa​n,Ω)(X^{an},\Omega) works even without the assumption c​h​a​r​k=0char\,k=0 where kk is the residue field of KK. This can be explained from the point of view of factorization theorem (see Section 10.4). It turns out that symplectomorphisms which appear in the infinite product in the RHS of the factorization theorem are infinite series whose coefficients are integer polynomials in the coefficients of the “parent” symplectomorphisms.

For example, let f0​(z)=1+∑n≥1cn​znf_{0}(z)=1+\sum_{n\geq 1}c_{n}z^{n} and f∞​(z)=1+∑n≥1dn​znf_{\infty}(z)=1+\sum_{n\geq 1}d_{n}z^{n} be two power series convergent when |z|<1|z|<1. Let us consider two symplectomorphisms: F0​(ξ,η)=(ξ,η​f0​(ξ−1))F_{0}(\xi,\eta)=(\xi,\eta f_{0}(\xi^{-1})) and F∞​(ξ,η)=(ξ​f∞​(η−1),η)F_{\infty}(\xi,\eta)=(\xi f_{\infty}(\eta^{-1}),\eta) and decompose F∞∘F0F_{\infty}\circ F_{0} into the infinite ordered product ∏→(Fλ)\prod_{\to}(F_{\lambda}). Here

Fp/q​(ξ,η)=(ξ​fp/q​(ξ−p​η−q)q,η​fp/q​(ξ−p​η−q)−p)F_{p/q}(\xi,\eta)=(\xi f_{p/q}(\xi^{-p}\eta^{-q})^{q},\eta f_{p/q}(\xi^{-p}\eta^{-q})^{-p})

where fp/q​(z)=1+∑n≥1cnp/q​znf_{p/q}(z)=1+\sum_{n\geq 1}c_{n}^{p/q}z^{n}. Then one can check that for any coprime p,q∈𝐙>0p,q\in{\bf Z}_{>0} and any n≥1n\geq 1 one has

cnp/q∈𝐙⁡[c1,c2,…,d1,d2,…].c_{n}^{p/q}\in{{\bf Z}}[c_{1},c_{2},\dots,d_{1},d_{2},\dots]\,\,.

This implies that our construction works when one replaces KK by arbitrary commutative ring RR endowed with a complete non-trivial valuation val:R→(−∞,+∞]val:R\to(-\infty,+\infty].

11.8 Further generalizations

First of all, one can introduce a small parameter ℏ∈K,|ℏ|<1\hbar\in K,|\hbar|<1 of noncommutativity in the picture, coordinates ξ,η\xi,\eta will not commute but instead satisfy the relation

η​ξ=ξ​η​exp⁡(ℏ).\eta\xi=\xi\eta\exp(\hbar)\,\,.

For such a noncommutative analytic torus one can still define sheaf 𝒪ℏc​a​n{\cal O}^{can}_{\hbar} on 𝐑2{\bf R}^{2} by the “same” formula as in the commutative case:

𝒪ℏc​a​n​(U)={∑n,m∈𝐙cn,m​ξn​ηm|∀(x,y)∈U​supn,m(log⁡|cn,m|+n​x+m​y)<∞}{\cal O}^{can}_{\hbar}(U)=\left\{\sum_{n,m\in{\bf Z}}c_{n,m}\xi^{n}\eta^{m}\,|\,\forall(x,y)\in U\,\,\,\sup_{n,m}\left(\log|c_{n,m}|+nx+my\right)<\infty\right\}

where U⊂𝐑2U\subset{\bf R}^{2} is connected. Also one can construct a non-commutative deformation of the model sheaf near the singular point. All arguments with the groups work as well. In this way we will obtain a kind of quantized K3 surface over a non-archimeden field.

Secondly, we believe that one can generalize our construction to higher dimensions. Instead of lines there will be codimension one walls which should be flat hypersurfaces with respect to 𝐙{\bf Z}-affine structure and carry foliations by parallel lines. Generically on the intersection of two such foliated hypersurfaces one can “separate” variables into the product of a purely 2-dimensional situation studied in the present paper, and n−2n-2 dummy variables. Presumably everywhere except a countable union of codimension 2 subsets one can use 2-dimensional factorization and define gluing volume preserving maps. One can hope that by a kind of Hartogs principle the sheaf will have a canonical extension to the whole space BB.

Appendix A Analytic geometry

In this section we collect several facts and definitions about rigid analytic spaces and Clemens polytopes. Some of them are well-known, the rest is borrowed from [KoT].

We always work over a complete non-archimedean local field KK. The field KK carries a valuation map v​a​lK:=v​a​l:K→𝐑∪{+∞}val_{K}:=val:K\to{{\bf R}}\cup\{+\infty\} such that v​a​l​(0)=+∞,v​a​l​(1)=0,v​a​l​(x​y)=v​a​l​(x)+v​a​l​(y),v​a​l​(x+y)≥min⁡(v​a​l​(x),v​a​l​(y))val(0)=+\infty,val(1)=0,val(xy)=val(x)+val(y),val(x+y)\geq\min(val(x),val(y)).

We will assume that the valuation is non-trivial. The ring

𝒪K=v​a​lK−1​(𝐑≥0∪{+∞}){\cal O}_{K}=val_{K}^{-1}({{\bf R}}_{\geq 0}\cup\{+\infty\})

is called the ring of integers of KK. The residue field is defined as k=𝒪K/mKk={\cal O}_{K}/m_{K}, where mK=v​a​lK−1​(𝐑>0∪{+∞})m_{K}=val_{K}^{-1}({{\bf R}}_{>0}\cup\{+\infty\}) is the maximal ideal in 𝒪K{\cal O}_{K}.

Our main example is the field K=𝐂⁡((t))K={{\bf C}}((t)) of Laurent series in one variable. In this case v​a​lK​(∑n≥n0cn​tn)=n0val_{K}(\sum_{n\geq n_{0}}c_{n}t^{n})=n_{0}, as long as cn0≠0c_{n_{0}}\neq 0.

A.1 Berkovich spectrum

We refer the reader to [Be1] for the general definition of an analytic space and more details. In this Appendix we restrict ourselves to analytic spaces associated with algebraic varieties (although we use the general definition in the paper as well).

Let R=R/KR=R/K be a commutative unital finitely generated KK-algebra. The underlying set of the Berkovich spectrum S​p​e​ca​n​(R):=S​p​e​ca​n​(R/K)Spec^{an}(R):=Spec^{an}(R/K) can be defined in two ways. First one uses valuations (or, equivalently, multiplicative seminorms).

Definition 15

(Valuations) A point xx of S​p​e​ca​n​(R/K)Spec^{an}(R/K) is an additive valuation

v​a​lx:R→𝐑∪{+∞}val_{x}:R\to{\bf R}\cup\{+\infty\}

extending v​a​l:=v​a​lKval:=val_{K}, i.e. it is a map satisfying the conditions

  • •

    v​a​lx​(r+r′)≤max⁡(v​a​lx​(r),v​a​lx​(r′))val_{x}(r+r^{\prime})\leq\max(val_{x}(r),val_{x}(r^{\prime}));

  • •

    v​a​lx​(r​r′)=v​a​lx​(r)+v​a​lx​(r′)val_{x}(rr^{\prime})=val_{x}(r)+val_{x}(r^{\prime});

  • •

    v​a​lx​(λ)=v​a​lK​(λ)val_{x}(\lambda)=val_{K}(\lambda)

for all r,r′∈Rr,r^{\prime}\in R and all λ∈K\lambda\in K.

Having a valuation and a real number q0∈(0,1)q_{0}\in(0,1) one can define the multiplicative seminorm |a|=q0v​a​lK​(a),a∈R|a|=q_{0}^{val_{K}(a)},a\in R. In particular, in the previous definition one can take seminorms |⋅|x|\cdot|_{x} instead of valuations v​a​lx​(⋅)val_{x}(\cdot). The reader has noticed that in the main body of the paper, for R=KR=K we often took |a|=e−v​a​l​(a)|a|=e^{-val(a)}. It is easy to translate the definition of Berkovich spectrum to the language of multiplicative seminorms. We use it freely in the paper.

The second way to define Xa​nX^{an} uses evaluations (characters).

Definition 16

(Evaluation maps) A point xx of S​p​e​ca​n​(R/K)Spec^{an}(R/K) is an equivalence class of homomorphisms of KK-algebras

e​v​a​lx:R→Kx,eval_{x}\,:\,R\to K_{x}\,\,,

where Kx⊃KK_{x}\supset K is a complete field equipped with a non-archimedean valuation, which extends the valuation v​a​lKval_{K}, and such that KxK_{x} is generated by the closure of the image of e​v​a​lxeval_{x}.

The field KxK_{x} is determined by x∈Xa​nx\in X^{an} in a canonical way. We define for r∈Rr\in R and x∈Xa​nx\in X^{an} the “value” r⁡(x)∈Kxr(x)\in K_{x} as the image e​v​a​lx​(r)eval_{x}(r).

In order to pass from the first description of S​p​e​ca​n​(R/K)Spec^{an}(R/K) to the second, starting with a valuation v​a​lxval_{x} one defines the field KxK_{x} as the completion of the field of fractions of R/IxR/I_{x}, where Ix=(v​a​lx)−1​({+∞})I_{x}=(val_{x})^{-1}(\{+\infty\}).

Definition 17

The topology on S​p​e​ca​n​(R/K)Spec^{an}(R/K) is the weakest topology such that for all r∈Rr\in R the map

S​p​e​ca​n​(R/K)→𝐑∪{+∞},x↦v​a​lx​(r)\begin{array}[]{ccc}Spec^{an}(R/K)&\to&{\bf R}\cup\{+\infty\},\\ x&\mapsto&val_{x}(r)\end{array}

is continuous.

An element f∈Rf\in R defines a function f:S​p​e​ca​n​(R)→Kxf:Spec^{an}(R)\to K_{x}, where KxK_{x} is the non-archimedean valuation field, which is the completion of the field of fractions of the domain R/ker⁡(v​a​lx)R/\ker(val_{x}). Since each KxK_{x} carries a seminorm, we obtain a function |f|:S​p​e​ca​n→𝐑≥0,x↦|f⁡(x)||f|:Spec^{an}\to{{\bf R}}_{\geq 0},x\mapsto|f(x)|.

A fundamental system of neighborhoods U=Ux⊂S​p​e​ca​n​(R)U=U_{x}\subset Spec^{an}(R) of a point xx is parametrized by the following data: a finite collections of functions

(fi)i∈I,(gj)j∈J∈R(f_{i})_{i\in I},\,\,(g_{j})_{j\in J}\,\,\,\in R

and numbers

βi+,βi−,γj∈𝐑>0\beta^{+}_{i},\beta^{-}_{i},\gamma_{j}\in{\bf R}_{>0}

such that βi−<|fi​(x)|<βi+,|gj​(x)|=0\beta_{i}^{-}<|{f_{i}(x)}|<\beta_{i}^{+},\,\,|{g_{j}(x)}|=0, The corresponding neighborhood consists of points x′x^{\prime} such that βi−<|fi​(x′)|<βi+,|gj​(x′)|<γj\beta_{i}^{-}<|{f_{i}(x^{\prime})}|<\beta_{i}^{+},\,\,|{g_{j}(x^{\prime})}|<\gamma_{j} for all i∈I,j∈Ji\in I,j\in J and x′∈Ux^{\prime}\in U.

Let us assume that elements (fi)i∈I,(gj)j∈J(f_{i})_{i\in I},\,(g_{j})_{j\in J} generate RR, i.e.

R=K⁡[(fi)i∈I,(gj)j∈J]/IR=K[(f_{i})_{i\in I},\,(g_{j})_{j\in J}]/I

where II is an ideal. Let us consider the algebra of series

s=∑nI∈𝐙I,mJ∈𝐍JcI,J​fInI​gJmJs=\sum_{n_{I}\in{\bf Z}^{I},\,m_{J}\in{\bf N}^{J}}c_{I,J}f_{I}^{n_{I}}g_{J}^{m_{J}}

with constants cI,J∈Kc_{I,J}\in K, absolutely convergent when variables (fi)i∈I,(gj)j∈J(f_{i})_{i\in I},\,(g_{j})_{j\in J} satisfy the above inequalities. The quotient of this algebra by the topological closure of ideal II is the algebra 𝒪S​p​e​ca​n​(R/K)​(U){\cal O}_{Spec^{an}(R/K)}(U).

As in the case of schemes we can glue S​p​e​ca​n​(R/K)Spec^{an}(R/K) into ringed spaces called analytic spaces (or rigid analytic spaces). Moreover we get a functor

(S​c​h​e​m​e​s/K)→(K−analytic​spaces)X↦(Xa​n,𝒪Xa​n).\begin{array}[]{ccc}(Schemes/K)&\to&(K-{\rm analytic}\,\,{\rm spaces})\\ X&\mapsto&(X^{an},{\cal O}_{X^{an}}).\end{array}
Proposition 8

The space Xa​nX^{an}

a) is a locally compact Hausdorff space as long as XX is separated;

b) has the homotopy type of a finite C​WCW-complex;

c) is contractible if XX has good reduction with irreducible special fiber.

Example 1

Let X=𝐀1=S​p​e​c​(K⁡[x])X={\bf A}^{1}=Spec(K[x]) be the affine line. The analytic space Xa​nX^{an} contains, among others, points of the following types:

  • •

    X⁡(K)↪X⁡(K¯)/G​a​l​(K¯/K)↪Xa​nX(K)\hookrightarrow X(\overline{K})/Gal(\overline{K}/K)\hookrightarrow X^{an};

  • •

    for r∈𝐑≥0r\in{{\bf R}}_{\geq 0} define

    |∑j=0dcj​zj|r:=maxj⁡(|cj|​rj).|\sum_{j=0}^{d}c_{j}z^{j}|_{r}:=\max_{j}(|{c_{j}}|r^{j})\,\,.

    This gives an embedding 𝐑≥0↪Xa​n{\bf R}_{\geq 0}\hookrightarrow X^{an}.

We see that Xa​nX^{an} contains, in a sense, both pp-adic and real points.

Define the cone over Xa​nX^{an} as

CXa​n​(𝐑):=Xa​n×𝐑>0.C_{X^{an}}({\bf R}):=X^{an}\times{\bf R}_{>0}\,\,.

We interpret a point 𝐱=(x,λ){\bf x}=(x,\lambda) of CXa​n​(𝐑)C_{X^{an}}({\bf R}) as a KxK_{x}-point of XX, where Kx⊃KK_{x}\supset K is a complete field with the 𝐑{\bf R}-valued valuation

v​a​l𝐱:=λ​v​a​lx,val_{\bf x}:=\lambda\,val_{x}\,\,,

whose restriction to KK is proportional to v​a​lKval_{K}. The set of points 𝐱∈CXa​n​(𝐑){\bf x}\in C_{X^{an}}({\bf R}) such that the valuation v​a​l𝐱val_{\bf x} is 𝐙{\bf Z}-valued is denoted by CXa​n​(𝐙)C_{X^{an}}({\bf Z}).

A.2 Algebraic torus and the logarithmic map

Here we will describe explicitly the main example for our paper. Let X=𝐆mn=S​p​e​c​(K⁡[zi±1]),1≤i≤nX={\bf G}_{m}^{n}=Spec(K[z_{i}^{\pm 1}]),1\leq i\leq n be an algebraic torus. and Xa​n=(𝐆ma​n)nX^{an}=({\bf G}_{m}^{an})^{n} the corresponding analytic space.

Firstly, we define an embedding ic​a​n:𝐑n↪Xa​ni_{can}:{\bf R}^{n}\hookrightarrow X^{an}. For real vector (xi)1≤i≤n∈𝐑b(x_{i})_{1\leq i\leq n}\in{\bf R}^{b} the corresponding point p:=ic​a​n​(x1,…,xn)∈Xa​np:=i_{can}(x_{1},\dots,x_{n})\in X^{an} will be described in terms of valuations.

For every Laurent polynomial f=∑I∈𝐙ncI​zI,cI∈Kf=\sum_{I\in{{\bf Z}}^{n}}c_{I}z^{I},\,\,c_{I}\in K we set

v​a​lp​(f):=minI∈𝐙n⁡(v​a​l​(cI)−∑i=1nxi​Ii).val_{p}(f):=\min_{I\in{\bf Z}^{n}}\left(val(c_{I})-\sum_{i=1}^{n}x_{i}I_{i}\right)\,\,.

Secondly, we define a projection πc​a​n:Xa​n→𝐑n\pi_{can}:X^{an}\to{\bf R}^{n} by formula

πc​a​n​(y)=(−v​a​ly​(z1),…,−v​a​ly​(zn))=(log⁡|z1|y,…,log⁡|zn|y).\pi_{can}(y)=\left(-val_{y}(z_{1}),\dots,-val_{y}(z_{n})\right)=\left(\log|z_{1}|_{y},\dots,\log|z_{n}|_{y}\right)\,\,.

The fiber over a point (x1,…,xn)∈𝐑n(x_{1},\dots,x_{n})\in{{\bf R}}^{n} can be identified with the set of such seminorms |⋅|y|\cdot|_{y} that |zi|y=exp⁡(xi),1≤i≤n|z_{i}|_{y}=\exp(x_{i}),1\leq i\leq n. We see πc​a​n\pi_{can} is a kind of torus fibration77 7 This is the origin of the term “analytic torus fibration” introduced in Section 4.1.. Moreover, πc​a​n∘ic​a​n=i​d𝐑n\pi_{can}\circ i_{can}=id_{\,{\bf R}^{n}}.

For any open connected U∈(𝐑)nU\in({{\bf R}})^{n} the KK-algebra of analytic functions on πc​a​n−1​(U)\pi_{can}^{-1}(U) consists of series f=∑I∈𝐙ncI​zIf=\sum_{I\in{{\bf Z}}^{n}}c_{I}z^{I} with coefficients cI∈Kc_{I}\in K such that for any p=(x1,…,xn)∈Up=(x_{1},\dots,x_{n})\in U we have log⁡|cI|+∑i=1nxi​Ii→+∞\log|c_{I}|+\sum_{i=1}^{n}x_{i}I_{i}\to+\infty when |l|→+∞|l|\to+\infty. It is easy to see that πc​a​n−1​(U)=πc​a​n−1​(C​o​n​v​(U))\pi_{can}^{-1}(U)=\pi_{can}^{-1}(Conv(U)) where C​o​n​v​(U)Conv(U) is the convex hull of UU.

The sheaf (πc​a​n)∗​(𝒪Xa​n):=𝒪𝐑nc​a​n(\pi_{can})_{\ast}({\cal O}_{X^{an}}):={\cal O}^{can}_{{{\bf R}}^{n}} (canonical sheaf) plays an important role in the paper (see Sections 4.1, 7.3, 8).

A.3 Clemens polytopes

Let XX be a smooth proper scheme over the non-archimedean field KK. We assume that KK carries a discrete valuation v​a​lval such that v​a​l​(K×)=𝐙val(K^{\times})={{\bf Z}}.

Definition 18

A model of XX is a scheme of finite type 𝒳/𝒪K{\cal X}/{{\cal O}}_{K} flat and proper over 𝒪K{{\cal O}}_{K}, together with an isomorphism 𝒳×S​p​e​c​(K)S​p​e​c​(𝒪K)≃X{\cal X}\times_{Spec(K)}Spec({{\cal O}}_{K})\simeq X. Denote the special fiber of 𝒳{\cal X} by

𝒳0:=𝒳×S​p​e​c​(K)S​p​e​c​(k).{\cal X}^{0}:={\cal X}\times_{Spec(K)}Spec(k)\,\,.

A model has no nontrivial automorphisms. Thus, the stack of equivalence classes of models is in fact a set, which we denote by M​o​dXMod_{X}. It carries a natural partial order. Namely, we say that 𝒳1≥𝒳2{\cal X}_{1}\geq{\cal X}_{2} if there exists a map 𝒳1→𝒳2{\cal X}_{1}\to{\cal X}_{2} over S​p​e​c​(𝒪K)Spec({{\cal O}}_{K}). Such a map is automatically unique.

Definition 19

A model 𝒳{\cal X} has normal crossings if the scheme 𝒳{\cal X} is regular and the reduced subscheme 𝒳r​e​d0{\cal X}^{0}_{red} is a divisor with normal crossings.

By the resolution of singularities, in the case c​h​a​r​k=0char\,k=0 we know that every model is dominated by a model with normal crossings.

Definition 20

A model 𝒳{\cal X} has simple normal crossings (snc model for short) if

  • •

    it has normal crossings;

  • •

    all irreducible components of 𝒳r​e​d0{\cal X}^{0}_{red} are smooth and

  • •

    all intersections of irreducible components of 𝒳r​e​d0{\cal X}^{0}_{red} are either empty or irreducible.

The set of equivalence classes of snc models will be denoted by M​o​dXs​n​cMod_{X}^{snc}. It is a filtered partially ordered set. The order is given by dominating maps of models which give the identity automorphism on the generic fiber.

It is easy to show that starting with any model with normal crossings and applying blow-ups centered at certain self-intersection loci of the special fiber we can get a snc model. In what follows we use snc models only. This choice is dictated by convenience and not by necessity. Working with snc models has the advantage that all definitions and calculations can be made very transparent. The reader can consult [Be2] for the approach in the general case, without the use of the resolution of singularities.

Let 𝒳{\cal X} be an snc model and I=I𝒳I=I_{{\cal X}} the set of irreducible components of 𝒳r​e​d0{\cal X}^{0}_{red}. Denote by Di⊂𝒳D_{i}\subset{\cal X} the divisor corresponding to i∈Ii\in I. For any finite non-empty subset J⊂IJ\subset I put

DJ:=⋂j∈JDj.D_{J}:=\bigcap_{j\in J}D_{j}\,\,.

By the snc property the set DJD_{J} is either empty or is a smooth connected proper variety over kk of dimension dim(DJ)=(n−|J|+1)\dim(D_{J})=(n-|J|+1). For a divisor Di⊂𝒳0D_{i}\subset{\cal X}^{0} we denote by di∈𝐙>0d_{i}\in{\bf Z}_{>0} the order of vanishing of uu at DiD_{i}, where u∈Ku\in K is an uniformizing element, v​a​lK​(u)=1val_{K}(u)=1. Equivalently, did_{i} is the multiplicity of DiD_{i} in 𝒳0{\cal X}^{0}.

Definition 21

The Clemens polytope S𝒳S_{\cal X} is the finite simplicial subcomplex of the simplex ΔI\Delta^{I} such that ΔJ\Delta^{J} is a face of S𝒳S_{\cal X} iff DJ≠∅D_{J}\neq\emptyset.

Clearly, S𝒳S_{\cal X} is a nonempty connected CW-complex. We will also consider the cone over S𝒳S_{\cal X}:

C𝒳(𝐑):={∑i∈Iai⟨Di⟩|ai∈𝐑≥0,⋂i:ai>0Di≠∅}∖{0}⊂𝐑I.C_{\cal X}({\bf R}):=\left\{\sum_{i\in I}a_{i}\langle D_{i}\rangle|\,a_{i}\in{{\bf R}}_{\geq 0},\,\,\,\bigcap_{i:\,a_{i}>0}D_{i}\neq\emptyset\right\}\setminus\{0\}\subset{\bf R}^{I}\,\,.

Analogously, we can define C𝒳​(𝐙),C_{\cal X}({\bf Z}),\,.

We identify S𝒳S_{\cal X} with the following subset of C𝒳​(𝐑)C_{\cal X}({\bf R}):

{∑i∈Iai​⟨Di⟩∈C𝒳​(𝐑)|∑iai​di=1}.\left\{\sum_{i\in I}a_{i}\langle D_{i}\rangle\in C_{\cal X}({\bf R})|\,\sum_{i}a_{i}d_{i}=1\right\}\,\,.

Obviously, we can also describe S𝒳S_{\cal X} as a quotient of C𝒳​(𝐑)C_{\cal X}({\bf R}):

S𝒳=C𝒳​(𝐑)/𝐑+×.S_{\cal X}=C_{\cal X}({\bf R})/{\bf R}^{\times}_{+}\,\,.

A.4 Simple blow-ups

Let 𝒳{\cal X} be an snc model, J⊂I𝒳J\subset I_{\cal X} a non-empty subset and Y⊂DJY\subset D_{J} a smooth irreducible variety of dimension less or equal than nn. Let us assume that YY intersects transversally (in DJD_{J}) all subvarieties DJ′D_{J^{\prime}} of DJD_{J} (for J′⊃JJ^{\prime}\supset J), and that all intersections Y∩DJY\cap D_{J} are either empty or irreducible. It is obvious that the blow-up 𝒳′:=B​lY​(𝒳){\cal X}^{\prime}:=Bl_{Y}({\cal X}) of 𝒳{\cal X} with the center at YY is again a snc model.

Definition 22

For a pair of snc models 𝒳′≥𝒳{\cal X}^{\prime}\geq{\cal X} as above we say that 𝒳′{\cal X}^{\prime} is obtained from 𝒳{\cal X} by a simple blow-up. If Y=DJY=D_{J} we say that we have a simple blow-up of the first type. Otherwise (when dim(Y)<dim(DJ)\dim(Y)<\dim(D_{J})), we have a simple blow-up of the second type.

Let us describe the behavior of S𝒳S_{\cal X} under simple blow-ups. To the set of vertices we add a new vertex corresponding to the divisor Y~\widetilde{Y} obtained from YY:

I𝒳′=I𝒳⊔{n​e​w},Dn​e​w:=Y~.I_{{\cal X}^{\prime}}=I_{\cal X}\sqcup\{new\},\,\,\,D_{new}:=\widetilde{Y}\,\,.

The degree of the new divisor is (for both the first and the second type)

dn​e​w:=∑i∈Jdj.d_{new}:=\sum_{i\in J}d_{j}\,\,.

For blow-ups of the first type we have automatically #​J>1\#J>1. Here is the list of faces of S𝒳′S_{\cal{X}^{\prime}}:

1) I′I^{\prime} for I′∈F​a​c​e​s​(S𝒳),I′⊄JI^{\prime}\in Faces(S_{\cal X}),\,I^{\prime}\not\subset J;

2) I′⊔{n​e​w}I^{\prime}\sqcup\{new\} for I′∈F​a​c​e​s​(S𝒳),I′≠J,I′∪J∈F​a​c​e​s​(S𝒳)I^{\prime}\in Faces(S_{\cal X}),\,I^{\prime}\neq J,\,I^{\prime}\cup J\in Faces(S_{\cal X});

3) the vertex {n​e​w}\{new\}.
For blow-ups of the second type the list of faces of S𝒳′S_{{\cal X}^{\prime}} is

1) I′I^{\prime} for I′∈F​a​c​e​s​(S𝒳)I^{\prime}\in Faces(S_{\cal X});

2) I′⊔{n​e​w}I^{\prime}\sqcup\{new\} for I′∈F​a​c​e​s​(S𝒳),I′⊃J,Y∩DI′≠∅I^{\prime}\in Faces(S_{\cal X}),\,I^{\prime}\supset J,\,Y\cap D_{I^{\prime}}\neq\emptyset;

3) the vertex {n​e​w}\{new\}.

On can deduce from results [AKMW] the following

Theorem 9

(Weak factorization) Assume that c​h​a​r​k=0char\,k=0. Then for any two snc models 𝒳,𝒳′{\cal X},\,{\cal X}^{\prime} there exists a finite alternating sequence of simple blow-ups

𝒳<𝒳1>𝒳2<⋯<𝒳2​m+1>𝒳′.{\cal X}<{\cal X}_{1}>{\cal X}_{2}<\dots<{\cal X}_{2m+1}>{\cal X}^{\prime}\,\,.
Corollary 3

Simple homotopy type of S𝒳S_{\cal X} does not depend on the choice of a snc model 𝒳{\cal X}.

A.5 Clemens cones and valuations

Let 𝒳{\cal X} be a snc model of XX. We define a map

i𝒳:C𝒳​(𝐑)→CXa​n​(𝐑)i_{\cal X}:C_{\cal X}({\bf R})\to C_{X^{an}}({\bf R})

such as follows. For J={j1,…,jk}⊂I𝒳J=\{j_{1},\dots,j_{k}\}\subset I_{\cal X} such that DJ≠∅D_{J}\neq\emptyset let us consider a point x∈C𝒳​(𝐑)x\in C_{\cal X}({\bf R})

x=∑i=1kai​⟨Dji⟩,ai∈𝐑>0​∀i∈{1,…,k}x=\sum_{i=1}^{k}a_{i}\langle D_{j_{i}}\rangle,\,\,\,a_{i}\in{{\bf R}}_{>0}\,\,\forall i\in\{1,\dots,k\}

and an affine Zariski open subset U⊂𝒳U\subset{\cal X} containing the generic point of DJD_{J}. One can embed 𝒪⁡(U){\cal O}(U) into the algebra of formal series KJ​[[z1,…,zk]]K_{J}[[z_{1},\dots,z_{k}]] where KJK_{J} is the field of rational functions on DJD_{J} and zi=0z_{i}=0 are equations of divisors Dji,i=1,…,kD_{j_{i}},\,\,i=1,\dots,k\,. We define a valuation vxv_{x} of 𝒪⁡(U){\cal O}(U) by the formula

vx​(∑n1,…,nk≥0cn1,…,nk​∏i=1kzini)=inf{∑ai​ni|cn1,…,nk≠0}.v_{x}\left(\sum_{n_{1},\dots,n_{k}\geq 0}c_{n_{1},\dots,n_{k}}\prod_{i=1}^{k}z_{i}^{n_{i}}\right)=\inf\left\{\sum a_{i}n_{i}\,|\,c_{n_{1},\dots,n_{k}}\neq 0\right\}\,\,.

We define i𝒳​(x)i_{\cal X}(x) to be the image of the point vx∈S​p​e​ca​n​(𝒪⁡(U)/K)v_{x}\in Spec^{an}({\cal O}(U)/K) in Xa​nX^{an}. It is easy to check that the element i𝒳​(x)i_{\cal X}(x) does not depend on the choice of the open subset UU.

The following proposition is obvious:

Proposition 9

The map i𝒳𝐑i_{\cal X}^{\bf R} is an embedding.

We will denote also by i𝒳i_{\cal X} the induced embedding S𝒳↪Xa​nS_{\cal X}\hookrightarrow X^{an}.

A.6 Clemens cones and paths

For a model 𝒳{\cal X} we can interpret elements of CXa​n​(𝐙)C_{X^{an}}({\bf Z}) as paths in 𝒳{\cal X}, i.e. equivalence classes of maps

ϕ:S​p​e​c​(𝒪L)→𝒳,\phi:Spec({{\cal O}}_{L})\to{\cal X},

where 𝒪L{{\cal O}}_{L} is the ring of integers in a field LL with discrete valuation in 𝐙{\bf Z}, such that the image of ϕ\phi does not lie in 𝒳{\cal X}. We define the map

p𝒳𝐙:CXa​n​(𝐙)→C𝒳​(𝐙)p_{\cal X}^{\bf Z}\,:\,C_{X^{an}}({\bf Z})\to C_{\cal X}({\bf Z})

as

p𝒳𝐙​([ϕ]):=∑iai​⟨Di⟩,p_{\cal X}^{\bf Z}([\phi]):=\sum_{i}a_{i}\langle D_{i}\rangle,

where ai∈𝐙≥0a_{i}\in{\bf Z}_{\geq 0} is the multiplicity of the intersection of the path ϕ\phi with the divisor Di,i∈I𝒳D_{i},\,\,i\in I_{\cal X}.

The following proposition can be derived from [Be1].

Proposition 10

The map p𝒳𝐙p_{\cal X}^{\bf Z} extends uniquely to a continuous 𝐑+×{\bf R}_{+}^{\times}-equivariant map p𝒳𝐑:CXa​n​(𝐑)→C𝒳​(𝐑)p_{\cal X}^{\bf R}\,:\,C_{X^{an}}({\bf R})\to C_{\cal X}({\bf R}). The map p𝒳𝐑p_{\cal X}^{\bf R} is a surjection.

We denote by p𝒳:Xa​n→S𝒳p_{\cal X}:X^{an}\to S_{\cal X} the map induced by p𝒳𝐑p_{\cal X}^{\bf R}.

Let f:𝒳′→𝒳f:{\cal X}^{\prime}\to{\cal X} be a dominating map of models. Let us denote by mi,i′∈𝐙≥0m_{i,i^{\prime}}\in{{\bf Z}}_{\geq 0} the multiplicity of a divisor Di′,i′∈I𝒳′D_{i^{\prime}},i^{\prime}\in I_{{\cal X}^{\prime}} in the proper pull-back of Di,i∈I𝒳D_{i},i\in I_{\cal X}. We define p𝒳′,𝒳𝐙:CXa​n​(𝐙)→C𝒳​(𝐙)p_{{\cal X}^{\prime},{\cal X}}^{{\bf Z}}:C_{X^{an}}({\bf Z})\to C_{\cal X}({\bf Z}) by the formulas ∑i′ai′​⟨Di′⟩↦∑imi,i′​ai′​⟨Di⟩\sum_{i^{\prime}}a_{i^{\prime}}\langle D_{i^{\prime}}\rangle\mapsto\sum_{i}m_{i,i^{\prime}}a_{i^{\prime}}\langle D_{i}\rangle. Let p𝒳′,𝒳:SXa​n​(𝐑)→S𝒳​(𝐑)p_{{\cal X}^{\prime},{\cal X}}:S_{X^{an}}({\bf R})\to S_{\cal X}({\bf R}) be the corresponding by map of Clemens polytopes.

Then we have the following result, which is easy to prove.

Lemma 8

For any dominating map of models 𝒳′→𝒳{\cal X}^{\prime}\to{\cal X} we have

p𝒳𝐙=p𝒳′,𝒳𝐙∘p𝒳′𝐙.p_{{\cal X}}^{\bf Z}=p_{{\cal X}^{\prime},{\cal X}}^{\bf Z}\circ p_{{\cal X}^{\prime}}^{{\bf Z}}\,\,.
Corollary 4

For dominating maps 𝒳′′≥𝒳′≥𝒳{\cal X}^{\prime\prime}\geq{\cal X}^{\prime}\geq{\cal X} we have

p𝒳′′,𝒳=p𝒳′,𝒳∘p𝒳′′,𝒳′.p_{{\cal X}^{\prime\prime},{\cal X}}=p_{{\cal X}^{\prime},{\cal X}}\circ p_{{\cal X}^{\prime\prime},{\cal X}^{\prime}}\,\,.
Theorem 10

For any algebraic XX the analytic space Xa​nX^{an} is a projective limit over the partially ordered set of snc models 𝒳{\cal X} of Clemens polytopes S𝒳S_{\cal X}. The connecting maps are p𝒳′,𝒳p_{{\cal X}^{\prime},{\cal X}}.

With any meromorphic at t=0t=0 family of smooth complex projective varieties Xt,   0<|t|<ϵX_{t},\,\,\,0<|t|<\epsilon one can associate a variety XX over the field 𝐂⁡((t)){\bf C}((t)). It is easy to see that for any snc model 𝒳\cal X one can canonically complete the family XtX_{t} by adding S𝒳S_{\cal X} as the fiber over t=0t=0. The total space is not a complex manifold by just a Hausdorff locally compact space which maps properly to the dick {t∈𝐂||t|<ϵ}\{t\in{\bf C}\,|\,|t|<\epsilon\}. Passing to the projective limit we see that one can compactify the family XtX_{t} at t=0t=0 by Xa​nX^{an}.

Appendix B Torelli theorem for K3 surfaces

Here we recall the classification theory of complex K3 sufaces (see [PSS] and its extension to non-algebraic case in [LP]). Let XX be a complex K3 surface, i.e. smooth connected complex manifold with dim𝐂X=2\dim_{{\bf C}}X=2 which admits a nowhere vanishing holomorphic 22-form Ω\Omega, and such that H1​(X,𝐙)=0H^{1}(X,{{\bf Z}})=0.

It is known that the group H2​(X,𝐙)H^{2}(X,{{\bf Z}}) endowed with the Poincare pairing (⋅,⋅)(\cdot,\cdot) is isomorphic to the lattice

ΛK​3=(0110)⊕(0110)⊕(0110)⊕(−E8)⊕(−E8)\Lambda_{K3}=\left(\begin{array}[]{cc}0&1\\ 1&0\end{array}\right)\oplus\left(\begin{array}[]{cc}0&1\\ 1&0\end{array}\right)\oplus\left(\begin{array}[]{cc}0&1\\ 1&0\end{array}\right)\oplus\left(-E_{8}\right)\oplus\left(-E_{8}\right)

of signature (3,19)(3,19).

Complex 11-dimensional vector space H2,0​(X)=𝐂⋅[Ω]⊂H2​(X,𝐙)⊗𝐂H^{2,0}(X)={{\bf C}}\cdot[\Omega]\subset H^{2}(X,{{\bf Z}})\otimes{{\bf C}} satisfies the condition (v,v)=0,(v,v¯)>0(v,v)=0,(v,\overline{v})>0 for any non-zero vector vv. Finally, it is known that XX admits a Kähler metric, and Kähler cone 𝒦X⊂H2​(X,𝐑){\cal K}_{X}\subset H^{2}(X,{{\bf R}}) of all Kähler metrics on XX is an open subset of CX:={[ω]∈H2(X,𝐙)⊗𝐑|([ω],[Ω])=0,([ω],[ω])>0}C_{X}:=\{[\omega]\in H^{2}(X,{{\bf Z}})\otimes{{\bf R}}|([\omega],[\Omega])=0,([\omega],[\omega])>0\}. In fact 𝒦X{\cal K}_{X} is a connected component of the set CX∖∪v∈H2​(X,𝐙),(v,v)=−2,(v,[Ω])=0HvC_{X}\setminus\cup_{v\in H^{2}(X,{{\bf Z}}),(v,v)=-2,(v,[\Omega])=0}H_{v}, where HvH_{v} is the hyperplane orthogonal to vv.

Axiomatizing these data we arrive to the following definition.

Definition 23

K3 period data is a quadruple (Λ,(⋅,⋅),H2,0,𝒦)(\Lambda,(\cdot,\cdot),H^{2,0},{\cal K}) consisting of a free abelian group Λ\Lambda, a symmetric pairing (⋅,⋅):Λ×Λ→𝐙(\cdot,\cdot):\Lambda\times\Lambda\to{{\bf Z}}, a 11-dimensional complex vector subspace H2,0⊂Λ⊗𝐂H^{2,0}\subset\Lambda\otimes{{\bf C}} and a set 𝒦⊂Λ⊗𝐑{\cal K}\subset\Lambda\otimes{{\bf R}} satisfying the following conditions:

  1. 1.

    r​k​Λ=22rk\,\Lambda=22;

  2. 2.

    (Λ,(⋅,⋅))(\Lambda,(\cdot,\cdot)) is isomorphic to ΛK​3\Lambda_{K3};

  3. 3.

    for any v∈H2,0∖{0}v\in H^{2,0}\setminus\{0\} one has (v,v)=0(v,v)=0 and (v,v¯)>0(v,\overline{v})>0;

  4. 4.

    the set 𝒦{\cal K} is a connected component of C∖∪v∈Λ,(v,v)=−2,(v,H2,0)=0HvC\setminus\cup_{v\in\Lambda,(v,v)=-2,(v,H^{2,0})=0}H_{v}\,, where C={w∈Λ⊗𝐑|(w,H2,0)=0,(w,w)>0}C=\{w\in\Lambda\otimes{{\bf R}}|(w,H^{2,0})=0,(w,w)>0\} and HvH_{v} is the hyperplane orthogonal to vv.

The K3 period data form a groupoid. On the other hand, K3 surfaces also form a groupoid (morphisms are isomorphisms of K3 surfaces). Then classical global Torelli theorem can be formulated in the following way.

Theorem 11

Groupoid of K3 surfaces is equivalent to the groupoid of K3 period data.

In particular the automorphism group of a K3 surface is isomorphic to the automorphism group of its period data.

More generally one can speak about holomorphic families of K3 surfaces over complex analytic spaces. For a K3 surface over an analytic space MM the period data consist of a local system of integral lattices (Λ,(⋅,⋅))(\Lambda,(\cdot,\cdot)) pointwise isomorphic to ΛK​3\Lambda_{K3}, a holomorphic line subbundle H2,0H^{2,0} of Λ⊗𝐙𝒪M\Lambda\otimes_{{\bf Z}}{\cal O}_{M} which is isotropic with respect to the symmetric pairing (⋅,⋅)(\cdot,\cdot), and satisfies pointwise the condition (v,v¯)>0,v∈Hx2,0∖{0},x∈Mr​e​d(v,\overline{v})>0,v\in H^{2,0}_{x}\setminus\{0\},x\in M^{red}, and an open subset of the total space of the bundle over Mr​e​dM^{red} with the fibers Λx⊗𝐑∩(H2,0)⟂\Lambda_{x}\otimes{{\bf R}}\cap(H^{2,0})^{\perp} ((H2,0)⟂(H^{2,0})^{\perp} is the orthogonal complement) satisfying pointwise the condition 4) from the definition of K3 period data. Then Torelli theorem holds for families as well.

References

[AKMW], D. Abramovich, K. Karu, K. Matsuki, J. Wlodarczyk, Torification and factorization of birational maps, J. Amer. Math. Soc. 15 (2002), no. 3, 531–572, and preprint math.AG/9904135.

[Ar] V. Arnold, Mathematical methods of classical mechanics, Springer, 1997.

[Au] M. Audin, Spinning tops. A course on integrable systems, Cambridge Studies Adv. Math., vol. 51, Cambridge University Press, 1996.

[Be1] V. Berkovich, Spectral theory and analytic geometry over non-archimedean fields, AMS Mathematical Surveys and Monographs, n. 33, 1990.

[Be2] V. Berkovich, Smooth p-adic analytic spaces are locally contractible, Inv.Math., 137, 1–84 (1999).

[Be3] V. Berkovich, Smooth p-adic analytic spaces are locally contractible, II, to appear.

[GS] M. Gross, B. Siebert, Mirror Symmetry via Logarithmic degeneration data, I, preprint math.AG/0309070.

[GW] M. Gross, P. M. H. Wilson, Large complex structure limits of K3 surfaces, J. Differential Geom. 55 (2000), no. 3, 475–546, and preprint math.DG/0008018.

[HZh] C. Haase, I. Zharkov, Integral affine structures on spheres and torus fibrations of Calabi-Yau toric hypersurfaces I, preprint math.AG/0205321.

[KN] S. Kobayashi, K. Nomizu, Foundations of differential geometry, vol. 1, John Wiley and Sons, 1963.

[Ko] M. Kontsevich, Homological algebra of mirror symmetry, Proc. ICM Zürich, vol.1, 1994, and preprint math.AG/9411018.

[KoSo] M. Kontsevich, Y. Soibelman, Homological mirror symmetry and torus fibrations, in Symplectic geometry and mirror symmetry (Seoul, 2000), World Sci. Publishing, River Edge, NJ, 2001, 203–263, and math.SG/0011041.

[KoT] M. Kontsevich, Yu. Tschinkel, Non-archimedean Kähler geometry, in preparation.

[LeS] N. C. Leung, M. Symington, Almost toric symplectic four-manifolds, preprint math.SG/0312165.

[LYZ] J. Loftin, S.-T. Yau, R. Zaslow, Affine manifolds, SYZ geometry and the “Y” vertex, preprint math.DG/0405061.

[LP] E. Looijenga, C. Peters, Torelli theorems for Kähler K3 surfaces, Comp. Math. 42 (1981), 145–186.

[Mi] G. Mikhalkin, Amoebas of algebraic varieties and tropical geometry, preprint math.AG/0403015.

[Mor] D. R. Morrison, Mathematical aspects of mirror symmetry, in Complex algebraic geometry (Park City, UT, 1993), IAS/Park City Math. Ser., 3, Amer. Math. Soc., Providence, RI, 1997, 265–327, and preprint alg-geom/9609021.

[PSS] I. I. Pjateckiĭ-S̆apiro, I. R. S̆afarevic̆, A Torelli theorem for algebraic surfaces of type K3, Math. USSR Izvestija 5 (1971), No. 3, 547–588.

[SYZ] A. Strominger, S.-T. Yau, E. Zaslow, Mirror symmetry is T-duality, Nucl. Phys. B479 (1996), 243-259.

[Tyu] A. Tyurin, On Bohr-Sommerfeld bases, Izv. Math. 64 (2000), no. 5, 1033–1064, and preprint math.AG/9909084.

[Zu] N. Zung, Symplectic topology of integrable Hamiltonian systems, II: Topological classification , Compositio Math., 138:2 (2003), 125-156.

Addresses:

M.K.: IHES, 35 route de Chartres, F-91440, France

maxim@ihes.fr

Y.S.: Department of Mathematics, KSU, Manhattan, KS 66506, USA

soibel@math.ksu.edu

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