This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

A model theoretic proof for o-minimal coherence theorem

Yayi Fu
Abstract

Bakker, Brunebarbe, Tsimerman showed in [BBT22] that the definable structure sheaf 𝒪n\mathcal{O}_{\mathbb{C}^{n}} of n\mathbb{C}^{n} is a coherent 𝒪n\mathcal{O}_{\mathbb{C}^{n}}-module as a sheaf on the site n¯\underline{\mathbb{C}^{n}}, where the coverings are finite coverings by definable open sets. In general, let 𝒦\mathcal{K} be an algebraically closed field of characteristic zero. We give another proof of the coherence of 𝒪𝒦n\mathcal{O}_{\mathcal{K}^{n}} as a sheaf of 𝒪𝒦n\mathcal{O}_{\mathcal{K}^{n}}-modules on the site 𝒦n¯\underline{\mathcal{K}^{n}} using spectral topology on the type space Sn(𝒦)S_{n}(\mathcal{K}). (Here Sn(𝒦)S_{n}(\mathcal{K}) means S2n()S_{2n}(\mathcal{R}) for some real closed field \mathcal{R}.) It also gives an example of how the intuition that sheaves on the type space are the same as sheaves on the site with finite coverings (see [EJP06, Proposition 3.2]) can be applied.

1 Introduction

Let 𝒪n\mathcal{O}_{\mathbb{C}^{n}} denote the sheaf of rings where 𝒪n(U)\mathcal{O}_{\mathbb{C}^{n}}(U) is the ring of holomorphic functions defined on UU, for each UnU\subseteq\mathbb{C}^{n} open. It’s also an 𝒪n\mathcal{O}_{\mathbb{C}^{n}}-module.
In complex analysis, it is well-known that

Fact 1.1.

[Oka50] (Oka) For any positive integer nn, 𝒪n\mathcal{O}_{\mathbb{C}^{n}} is a coherent 𝒪n\mathcal{O}_{\mathbb{C}^{n}}-module. i.e. 𝒪n\mathcal{O}_{\mathbb{C}^{n}} satisfies that

  1. 1.

    𝒪n\mathcal{O}_{\mathbb{C}^{n}} locally finite.

  2. 2.

    Every relation sheaf of 𝒪n\mathcal{O}_{\mathbb{C}^{n}}is locally finite.

This result is generalized in [PS08] to the case of any algebraically closed field 𝒦\mathcal{K} of characteristic 0.

Fact 1.2.

(Peterzil, Starchenko) For any positive integer nn, 𝒪𝒦n\mathcal{O}_{\mathcal{K}^{n}} is a coherent 𝒪𝒦n\mathcal{O}_{\mathcal{K}^{n}}-module.

In fact 1.1, 1.2, a sheaf means the usual sheaf in e.g. [Har13, Chapter II]. In [BBT22], coherence theorem is proved on the site n¯\underline{\mathbb{C}^{n}} where the coverings are finite coverings by definable open sets:

Fact 1.3.

(Bakker, Brunebarbe, Tsimerman) The definable structure sheaf 𝒪n\mathcal{O}_{\mathbb{C}^{n}} of n\mathbb{C}^{n} is a coherent 𝒪n\mathcal{O}_{\mathbb{C}^{n}}-module (as a sheaf on the site n¯\underline{\mathbb{C}^{n}}).

(The sheaves on a site in [BBT22] are different from the usual sheaves defined in [Har13]. We will explain more in later sections.)
In this paper, we use a method different from the one used in [BBT22] to prove the coherence of 𝒪𝒦n\mathcal{O}_{\mathcal{K}^{n}} as a sheaf on the site 𝒦n¯\underline{\mathcal{K}^{n}}, where 𝒦\mathcal{K} is an algebraically closed field of characteristic 0.

Theorem 1.4.

The definable structure sheaf 𝒪𝒦n\mathcal{O}_{\mathcal{K}^{n}} of 𝒦n\mathcal{K}^{n} is a coherent 𝒪𝒦n\mathcal{O}_{\mathcal{K}^{n}}-module as a sheaf on the site 𝒦n¯\underline{\mathcal{K}^{n}}.

Motivation of this proof comes from [EJP06, Proposition 3.2] which says we can consider a sheaf on the site 𝒦n¯\underline{\mathcal{K}^{n}} the same as a usual sheaf in [Har13, Chapter II] on the type space Sn(𝒦)S_{n}(\mathcal{K}) with spectral topology.
Section 2 gives definitions of sites, presheaves and sheaves on a site, spectral topology, coherence, tubular neighborhoods. Section 3 gives the proof of theorem 1.4. Section 4 shows that we can prove theorem 1.4 using an isomorphism of categories similar to that in [EJP06, Proposition 3.2].

Acknowledgements.

The author is grateful to her advisor Sergei Starchenko for the suggestion of using spectral topology and compactness to give a more model-theoretic proof and the suggestion of using tubular neighborhoods to prove lemma 2.2.

2 Preliminaries

2.1 Basic notions

Setting.

(The same setting as in [PS01].) Let 𝒦\mathcal{K} be an algebraically closed field of characteristic zero. Then 𝒦=(1)\mathcal{K}=\mathcal{R}(\sqrt{-1}) for some real closed subfield \mathcal{R}. Such \mathcal{R} is not unique. We fix one such \mathcal{R} and fix an o-minimal expansion of the chosen real closed field. The topology on \mathcal{R} is generated by the definable open intervals. The topology on 𝒦\mathcal{K} is identified with that on 2\mathcal{R}^{2}. When we say definable, we mean definable in the o-minimal structure \mathcal{R} with parameters in \mathcal{R}.

Definition 2.1.

[PS03, Definition 2.1.] For U𝒦U\subseteq\mathcal{K} a definable open set and F:U𝒦F:U\rightarrow\mathcal{K} a definable function, z0Uz_{0}\in U, we say that FF is 𝒦\mathcal{K}-differentiable at z0z_{0} if the limit as zz tends to z0z_{0} in 𝒦\mathcal{K} of (f(z)f(z0))/(zz0)(f(z)-f(z_{0}))/(z-z_{0}) exists in 𝒦\mathcal{K} (all operations taken in 𝒦\mathcal{K}, while the limit is taken in the topology induced on 𝒦\mathcal{K} by 2\mathcal{R}^{2}).

Definition 2.2.

[PS03, Definition 2.8.] Let V𝒦nV\subseteq\mathcal{K}^{n} be a definable open set, F:V𝒦F:V\rightarrow\mathcal{K} a definable map. FF is called 𝒦\mathcal{K}-differentiable on VV if it is continuous on VV and for every (z1,,zn)V(z_{1},...,z_{n})\in V and i=1,,ni=1,...,n, the function F(z1,,zi1,,zi+1,,zn)F(z_{1},...,z_{i-1},-,z_{i+1},...,z_{n}) is 𝒦\mathcal{K}-differentiable in the ii-th variable at ziz_{i} (in other words, FF is continuous on VV and 𝒦\mathcal{K}-differentiable in each variable separately).

2.2 Spectral topology

Definition 2.3.

[EJP06, Definition 2.2.] Let XmX\subseteq\mathcal{R}^{m} be a definable set (with parameters in \mathcal{R}). The o-minimal spectrum X~\tilde{X} of XX is the set of complete mm-types Sm()S_{m}(\mathcal{R}) of the first order theory Th()Th_{\mathcal{R}}(\mathcal{R}) which imply a formula defining XX. This is equipped with the topology generated by the basic open sets of the form U~={αX~:Uα}\tilde{U}=\{\alpha\in\tilde{X}:U\in\alpha\}, where UU is a definable, relatively open subset of XX, and UαU\in\alpha means the formula defining UU is in α\alpha. We call this topology on XX the spectral topology.

2.3 Sheaves on the type space

Let Sn(𝒦)S_{n}(\mathcal{K}) denote S2n()S_{2n}(\mathcal{R}). We use this unconventional notation to emphasize that we are considering functions on 𝒦n\mathcal{K}^{n}.
Given a definable open set U𝒦nU\subseteq\mathcal{K}^{n}, let 𝒪𝒦n(U~)\mathcal{O}_{\mathcal{K}^{n}}(\tilde{U}) be the ring of 𝒦\mathcal{K}-differentiable functions defined on UU. It’s easy to see that this defines a sheaf on the type space Sn(𝒦)S_{n}(\mathcal{K}) with spectral topology. (We mean the usual notion of sheaves.)
Let 𝒪𝒦n\mathcal{O}_{\mathcal{K}^{n}} denote the sheaf of rings where 𝒪𝒦n(U~)\mathcal{O}_{\mathcal{K}^{n}}(\tilde{U}) is the ring of 𝒦\mathcal{K}-differentiable functions defined on UU, for each U𝒦nU\subseteq\mathcal{K}^{n} definable open.

Definition 2.4.

Given pSn(𝒦)p\in S_{n}(\mathcal{K}), let 𝒪p\mathcal{O}_{p} denote the set of germs for functions

{f:U𝒦:U\{f:U\rightarrow\mathcal{K}:U is some open definable set such that pU~p\in\tilde{U}
and ff is 𝒦\mathcal{K}-holomophic on UU }\}.

Given a definable set A𝒦nA\subseteq\mathcal{K}^{n}, let p(A)𝒪p\mathcal{I}_{p}(A)\subseteq\mathcal{O}_{p} denote the set of germs for functions

{f:U𝒦:U\{f:U\rightarrow\mathcal{K}:U is some open definable set such that pU~p\in\tilde{U},
ff is 𝒦\mathcal{K}-holomophic on UU and xAU\forall x\in A\cap U, f(x)=0f(x)=0 }\}.

Let g1,,gt𝒪pg_{1},...,g_{t}\in\mathcal{O}_{p}. Let Rp(g1,,gt)R_{p}(g_{1},...,g_{t}) denote the set

{(f1,,ft)𝒪pt:f1g1++ftgt=0}\{(f_{1},...,f_{t})\in\mathcal{O}^{t}_{p}:f_{1}g_{1}+...+f_{t}g_{t}=0\}.

Notice that for a sheaf on the type space Sn(𝒦)S_{n}(\mathcal{K}), we mean a usual sheaf as defined in [Har13, Chapter 2]. In the next subsection, we define a different notion of sheaves, the sheaves on a site, where coverings are finite.

2.4 o-minimal site

We translate definitions about sites in [Sta18] into o-minimal context: (For the formal definitions related to sites and how the usual notion of sheaves defined in [Har13, Chapter 2] are defined in the context of sites, see [Sta18, Part 1, Chapter 7].)

Definition 2.5.

[Sta18, Part 1, Chapter 7, Definition 6.2] Let X𝒦nX\subseteq\mathcal{K}^{n} be a definable set. The o-minimal site X¯\underline{X} on XX consists of definable (relative) open subsets of XX, together with Cov(X):={(U,{Ui}i=1k):U,U1,,UkXCov(X):=\{(U,\{U_{i}\}_{i=1}^{k}):U,U_{1},...,U_{k}\subseteq X definable open, {Ui}i=1k\{U_{i}\}_{i=1}^{k} a finite covering of UU }\}. (This means, in the formal definition of a site, the objects of the category are definable open subsets of XX; the morphisms of the category are inclusions; the coverings are finite coverings by definable open sets.)

Definition 2.6.

[Sta18, Part 1, Chapter 6, Section 5] A presheaf of abelian groups (resp. rings) on an o-minimal site X¯\underline{X} is defined the same as usual:
Let XX be a topological space. A presheaf \mathcal{F} of abelian groups (resp. rings) on an o-minimal site X¯\underline{X} consists of the following data:

  1. (a)

    a collection of non empty abelian groups (resp. rings) (U)\mathcal{F}(U) associated with every definable open set UXU\subseteq X,

  2. (b)

    a collection of morphisms of abelian groups (resp. rings) ρU,V:(V)(U)\rho_{U,V}:\mathcal{F}(V)\rightarrow\mathcal{F}(U) defined whenever UVU\subseteq V and satisfying the transitivity property,

  3. (c)

    ρU,VρV,W=ρU,W\rho_{U,V}\circ\rho_{V,W}=\rho_{U,W} for UVWU\subseteq V\subseteq W, ρU,U=IdU\rho_{U,U}=Id_{U} for every UU.

Definition 2.7.

[Sta18, Part 1, Chapter 6, Definition 6.1] Let XX be a topological space. Let 𝒪\mathcal{O} be a presheaf of rings on the o-minimal site X¯\underline{X}. A presheaf of 𝒪\mathcal{O}-modules \mathcal{F} on an o-minimal site X¯\underline{X} is a presheaf \mathcal{F} of abelian groups with the following additional data:

  1. (a)

    For every definable open set UXU\subseteq X, (U)\mathcal{F}(U) is a non empty 𝒪(U)\mathcal{O}(U)-module;

  2. (b)

    for every definable open UXU\subseteq X the 𝒪(U)\mathcal{O}(U)-module structure of (U)\mathcal{F}(U) is compatible with restriction mappings (of \mathcal{F} and 𝒪\mathcal{O}). i.e. for definable open UVXU\subseteq V\subseteq X, r𝒪(V)r\in\mathcal{O}(V), x(V)x\in\mathcal{F}(V), ρU,V(r)τU,V(x)=τU,V(rx)\rho_{U,V}(r)\tau_{U,V}(x)=\tau_{U,V}(rx).

Definition 2.8.

[Sta18, Part 1, Chapter 6, Definition 7.1.] Let X¯\underline{X} be an o-minimal site, and let \mathcal{F} be a presheaf of abelian groups (resp. rings, 𝒪\mathcal{O}-modules) on X¯\underline{X}. We say \mathcal{F} is a sheaf if for every definable open UXU\subseteq X and every definable open finite covering {Ui}i=1k\{U_{i}\}_{i=1}^{k} of UU,

  1. (i)

    if (si)i=1k(s_{i})_{i=1}^{k} satisfies si(Ui)s_{i}\in\mathcal{F}(U_{i}) for each ii and si|UiUj=sj|UiUjs_{i}|_{U_{i}\cap U_{j}}=s_{j}|_{U_{i}\cap U_{j}} for each pair i,ji,j, then there is a unique sUs\in U such that s|Ui=sis|_{U_{i}}=s_{i} for each ii;

  2. (ii)

    for s,t(U)s,t\in\mathcal{F}(U), if s|Ui=t|Uis|_{U_{i}}=t|_{U_{i}} for each ii then s=ts=t.

Definition 2.9.

[Sta18, Part 1, Chapter 7. Definition 11.1.] Let X¯\underline{X} be an o-minimal site, and let φ:𝒢\varphi:\mathcal{F}\rightarrow\mathcal{G} be a map of sheaves of modules. (i.e. φ\varphi is a morphism of \mathcal{F} and 𝒢\mathcal{G} considered as presheaves. A presheaf morphism is, as usual, a map compatible with the restiction maps.)

  1. (1)

    We say that φ\varphi is injective if for every definable open UXU\subseteq X the map φ:(U)𝒢(U)\varphi:\mathcal{F}(U)\rightarrow\mathcal{G}(U) is injective.

  2. (2)

    We say that φ\varphi is surjective if for every definable open UXU\subseteq X and every section s𝒢(U)s\in\mathcal{G}(U) there exists a finite covering {Ui}i=1k\{U_{i}\}_{i=1}^{k} of UU such that for each ii, UiU_{i} is definable open and the restriction s|Uis|_{U_{i}} is in the image of φ:(Ui)𝒢(Ui)\varphi:\mathcal{F}(U_{i})\rightarrow\mathcal{G}(U_{i}).

Definition 2.10.

([BBT22, Definition 2.13]) Let X¯\underline{X} be an o-minimal site. Given an 𝒪X\mathcal{O}_{X}-module MM, we say that MM is of finite type (as an 𝒪X\mathcal{O}_{X}-module) if there exists a finite definable open (relative to XX) cover XiX_{i} of XX and surjections 𝒪XinMXi\mathcal{O}^{n}_{X_{i}}\twoheadrightarrow M_{X_{i}} for some positive integer nn on each of those open sets. We say that MM is coherent (as an 𝒪X\mathcal{O}_{X}-module) if it is of finite type, and given any definable open UXU\subseteq X and any 𝒪U\mathcal{O}_{U} -module homomorphism φ:𝒪UnMU\varphi:\mathcal{O}^{n}_{U}\rightarrow M_{U} , the kernel of φ\varphi is of finite type.

Remark.

By definition 2.9 and definition 2.10, given a definable open UU and an 𝒪U\mathcal{O}_{U}-module MM, to show that MM is of finite type, it suffices to show that there exist a finite family of definable open sets U1,,UkU_{1},...,U_{k} covering UU and sheaf morphisms φi:𝒪UiMUi\varphi_{i}:\mathcal{O}_{U_{i}}\rightarrow M_{U_{i}}, i=1,,ki=1,...,k such that for any fixed ii, for any definable open VUiV\subseteq U_{i} and every section sM(V)s\in M(V), there exist a finite family of definable open sets V1,,VlV_{1},...,V_{l} covering VV and for each j{1,,l}j\in\{1,...,l\}, tj𝒪Vjt_{j}\in\mathcal{O}_{V_{j}} such that φ(Vj)(tj)=s|Vj\varphi(V_{j})(t_{j})=s|_{V_{j}}.

2.5 Motivation

Let X𝒦nX\subseteq\mathcal{K}^{n} be a definable set.

Definition 2.11.

[EJP06, Definition 2.2] For the o-minimal spectrum X~\tilde{X} of XX, since it is a topological space, we use the classical notation Sh(X~)Sh(\tilde{X}) to denote the category of sheaves of abelian groups on X~\tilde{X}. Since the topology on the o-minimal spectrum X~\tilde{X} of XX is generated by the constructible open subsets, i.e. sets of the form U~\tilde{U} with UU an open definable subset of XX, a sheaf on X~\tilde{X} is determined by its values on the sets U~\tilde{U} with UU an open definable subset of XX. (We may also consider X~\tilde{X} as a site where the objects of the category are the U~\tilde{U}’s where each UU is some definable open subset of XX; the morphisms of the category are inclusions; the coverings are any coverings by the U~\tilde{U}’s.)

Definition 2.12.

[EJP06, Definition 3.1.] We denote by Shdtop(X)Sh_{dtop}(X) the category of sheaves of abelian groups on XX with respect to the o-minimal site on XX.
Thus, for a definable set XX, we define the functor of the categories of sheaves of abelian groups Shdtop(X)Sh(X~)Sh_{dtop}(X)\rightarrow Sh(\tilde{X}) which sends FShdtop(X)F\in Sh_{dtop}(X) into F~\tilde{F} where, for UU an open definable subset of XX, we define F~(U~)={s~:sF(U)}F(U)\tilde{F}(\tilde{U})=\{\tilde{s}:s\in F(U)\}\simeq F(U), and Sh(X~)Shdtop(X)Sh(\tilde{X})\rightarrow Sh_{dtop}(X) which sends F~\tilde{F} into FF where, for UU an open definable subset of XX, we define F(U)={s:s~F~(U~)}F~(U~)F(U)=\{s:\tilde{s}\in\tilde{F}(\tilde{U})\}\simeq\tilde{F}(\tilde{U}).

The following fact is the motivation for our proof in section 3. It says that a sheaf on the site 𝒦n¯\underline{\mathcal{K}^{n}} is the same as a usual sheaf in [Har13, Chapter II] on the type space Sn(𝒦)S_{n}(\mathcal{K}) with spectral topology.

Fact 2.1.

[EJP06, Proposition 3.2] Sh(X~)Sh(\tilde{X}) and Shdtop(X)Sh_{dtop}(X) are isomorphic.

2.6 Tubular neighborhood

[PS01, Theorem 2.56] roughly says that given a 𝒦\mathcal{K}-holomorphic function ff and p𝒦np\in\mathcal{K}^{n}, the number of zeroes is fixed locally around pp. In this section, we prove the following lemma, which says that for all pSn(𝒦)p\in S_{n}(\mathcal{K}), the number of zeroes is fixed locally. This lemma will be used in the proof of the type version of Weierstrass division theorem.
Let π:𝒦n𝒦n1\pi:\mathcal{K}^{n}\rightarrow\mathcal{K}^{n-1}, πn:𝒦n𝒦\pi_{n}:\mathcal{K}^{n}\rightarrow\mathcal{K} denote the projection onto the first (n1)(n-1) coordinates and the projection onto the nn-th coordinate resp.
For notational convenience, given pSn(𝒦)p\in S_{n}(\mathcal{K}) and a definable set UU, when we say “pUp\in U”, we actually mean “pU~p\in\tilde{U}”. Similarly, when we say “an open neighborhood UU of pp”, we actually mean “an open neighborhood U~\tilde{U} of pp”.

Lemma 2.2.

Let pSn(𝒦)p\in S_{n}(\mathcal{K}). Fix f𝒪pf\in\mathcal{O}_{p} and an open neighborhood UU of pp on which ff is defined and is 𝒦\mathcal{K}-differentiable. Suppose for all yπ(U)y\in\pi(U), there are finitely many zeroes of f(y,)f(y,-) in Uy:={x𝒦:(y,x)U}U_{y}:=\{x\in\mathcal{K}:(y,x)\in U\}, counting multiplicity.
Then there exist ii\in\mathbb{N} and VUV\subseteq U a definable open neighborhood of pp such that for any yπ(V)y\in\pi(V), there are exactly ii zeroes of f(y,)f(y,-) in VyV_{y} counting multiplicity.

We need some basic definitions and facts about o-minimal structures.

Definition 2.13.

[VdD98, Chapter 3] Call a set Ym+1Y\subseteq\mathcal{R}^{m+1} is finite over m\mathcal{R}^{m} if for each xmx\in\mathcal{R}^{m} the fiber Yx:={r:(x,r)Y}Y_{x}:=\{r\in\mathcal{R}:(x,r)\in Y\} is finite; call YY uniformly finite over m\mathcal{R}^{m} if there is NN\in\mathbb{N} such that |Yx|N|Y_{x}|\leq N for all xmx\in\mathcal{R}^{m}.

Fact 2.3.

[VdD98, Chapter 3, Lemma (2.13)] (UNIFORM FINITENESS PROPERTY). Suppose the definable subset YY of m+1\mathcal{R}^{m+1} is finite over m\mathcal{R}^{m}. Then YY is uniformly finite over m\mathcal{R}^{m}.

Fact 2.4.

[PS01, Theorem 2.56.] Let WnW\subseteq\mathcal{R}^{n}, U𝒦U\subseteq\mathcal{K} be definable open sets, F:U×W𝒦F:U\times W\rightarrow\mathcal{K} a definable continuous function such that for every wWw\in W, F(,w)F(-,w) is a 𝒦\mathcal{K}- differentiable function on UU. Take (z0,w0)U×W(z_{0},w_{0})\in U\times W and suppose that z0z_{0} is a zero of order mm of F(,w0)F(-,w_{0}).
Then for every definable neighborhood VV of z0z_{0} there are definable open neighborhoods U1VU_{1}\subseteq V of z0z_{0} and W1WW_{1}\subseteq W of w0w_{0} such that F(,w)F(-,w) has exactly mm zeroes in U1U_{1} (counted with multiplicity) for every wW1w\in W_{1}.

We may assume in fact 2.4 that U1,W1U_{1},W_{1} are open balls: Let U1U,W1WU_{1}\subseteq U,W_{1}\subseteq W be definable neighborhoods of z0z_{0}, w0w_{0} resp. such that for all wW1w\in W_{1}, F(,w)F(-,w) has exactly mm zeroes in U1U_{1}. Let U2U1U_{2}\subseteq U_{1} and W2W1W_{2}\subseteq W_{1} be definable open balls of z0z_{0}, w0w_{0} resp. Then for all wW2w\in W_{2}, F(,w)F(-,w) has m\leq m zeroes in U2U_{2}. By fact 2.4, there exist U3U2U_{3}\subseteq U_{2}, W3W2W_{3}\subseteq W_{2} definable open neighborhoods of z0z_{0}, w0w_{0} resp. such that F(,w)F(-,w) has exactly mm zeroes in U3U_{3} (counted with multiplicity) for every wW3w\in W_{3}. Let W4W3W_{4}\subseteq W_{3} be a definable open ball around w0w_{0}. Then for all wW4w\in W_{4}, F(,w)F(-,w) has m\geq m zeroes in U2U_{2}. Hence U2U_{2} and W4W_{4} are open balls satisfying the conclusion of fact 2.4.

Definition 2.14.

[Cos00, Section 6.2] A CkC^{k} cylindrical definable cell decomposition of n\mathcal{R}^{n} is a cdcd satisfying extra smoothness conditions which imply, in particular, that each cell is a CkC^{k} submanifold of n\mathcal{R}^{n}.

  • A CkC^{k} cdcd of \mathcal{R} is any cdcd of \mathcal{R} (i.e. a finite subdivision of \mathcal{R}).

  • If n>1n>1, a CkC^{k} cdcd of n\mathcal{R}^{n} is given by a CkC^{k} cdcd of n1\mathcal{R}^{n-1} and, for each cell DD of n1\mathcal{R}^{n-1}, definable functions of class CkC^{k} ζD,1<<ζD,l(D):D\zeta_{D,1}<...<\zeta_{D,l(D)}:D\rightarrow\mathcal{R}. The cells of n\mathcal{R}^{n} are, of course, the graphs of the ζD,i\zeta_{D,i} and the bands delimited by these graphs.

Fact 2.5.

[Cos00, Theorem 6.6] (CkC^{k} Cell Decomposition: CkC^{k} CDCDnCDCD_{n}) Given finitely many definable subsets X1,,XlX_{1},...,X_{l} of n\mathcal{R}^{n}, there is a CkC^{k} cdcd of n\mathcal{R}^{n} adapted to X1,,XlX_{1},...,X_{l} (i.e. each XiX_{i} is a union of cells).

Fact 2.6.

[Cos00, Theorem 6.7] (Piecewise CkC^{k} PCnkPC^{k}_{n}) Given a definable function f:Anf:A\rightarrow\mathcal{R}^{n}, where AA is a definable subset of n\mathcal{R}^{n}, there is a finite partition of AA into definable CkC^{k} submanifolds 𝒞1,,𝒞l\mathcal{C}_{1},...,\mathcal{C}_{l}, such that each restriction f|𝒞if|\mathcal{C}_{i} is CkC^{k}.

Definition 2.15.

[Cos00, Chapter 6] Let MnM\subseteq\mathcal{R}^{n} be a definable CkC^{k} submanifold (we always assume 1k<1\leq k<\infty). The tangent bundle TMTM is the set of (x,v)M×n(x,v)\in M\times\mathcal{R}^{n} such that vv is a tangent vector to MM at xx. The normal bundle NMNM is the set of (x,v)(x,v) in M×nM\times\mathcal{R}^{n} such that vv is orthogonal to TxMT_{x}M. This is a Ck1C^{k-1} submanifold of n×n\mathcal{R}^{n}\times\mathcal{R}^{n}, and it is definable since TMTM is definable.

Let φ\varphi be the function φ:N𝒟2n2\varphi:N\mathcal{D}\rightarrow\mathcal{R}^{2n-2} where φ(x,v)=x+v\varphi(x,v)=x+v.

Fact 2.7.

[Cos00, Theorem 6.11] (Definable Tubular Neighborhood) Let MM be a definable CkC^{k} submanifold of n\mathcal{R}^{n}. There exists a definable open neighborhood UU of the zero-section M×{0}M\times\{0\} in the normal bundle NMNM such that the restriction φ|U\varphi|U is a Ck1C^{k-1} diffeomorphism onto an open neighborhood Ω\Omega of MM in n\mathcal{R}^{n}. Moreover, we can take UU of the form

U={(x,v)NM:v<ϵ(x)},U=\{(x,v)\in NM:\|v\|<\epsilon(x)\},

where ϵ\epsilon is a positive definable CkC^{k} function on MM.

Fact 2.8.

[Cos00, Lemma 6.12] Let MM be a definable CkC^{k} submanifold of n\mathcal{R}^{n}, closed in n\mathcal{R}^{n}. Let ψ:M\psi:M\rightarrow\mathcal{R} be a positive definable function, which is locally bounded from below by positive constants (for every xx in MM, there exist c>0c>0 and a neighborhood VV of xx in MM such that ψ>c\psi>c on VV). Then there exists a positive definable CkC^{k} function ϵ:M\epsilon:M\rightarrow\mathcal{R} such that ϵ<ψ\epsilon<\psi on MM.

Fact 2.9.

[Cos00, Lemma 6.15] The definable tubular neighborhood in [Cos00, Theorem 6.11] holds if MM is closed in n\mathcal{R}^{n}, or definably CkC^{k} diffeomorphic to a closed submanifold in some m\mathcal{R}^{m} (e.g. if MM is a cell of a CkC^{k} cdcd).

Now we prove lemma 2.2 by imitating the proof of fact 2.9.

Proof.

Since the set Z(f):={z:f(z)=0}Z(f):=\{z:f(z)=0\} is a closed set, if pZ(f)p\notin Z(f), then UZ(f)U\setminus Z(f) is a definable open set satisfying the lemma. Hence, may assume pZ(f)p\in Z(f).
Let pSn(𝒦)p\in S_{n}(\mathcal{K}). Fix f𝒪pf\in\mathcal{O}_{p} and an open neighborhood UU of pp on which ff is defined and is 𝒦\mathcal{K}-differentiable.
By fact 2.3, since for all yπ(U)y\in\pi(U), there are finitely many zeroes of f(y,)f(y,-) in UyU_{y}, there is mm such that for all yπ(U)y\in\pi(U), there are m\leq m many zeroes of f(y,)f(y,-) in UyU_{y}.
Let MiM_{i} be the set

{xU:πn(x) is a zero of order i for the function f(π(x),)}.\{x\in U:\pi_{n}(x)\text{ is a zero of order $i$ for the function }f(\pi(x),-)\}.

Then pMip\in M_{i} for some ii. Fix such ii and let M=MiM=M_{i}. Let π(M)\pi(M) denote {y𝒦n1:z\{y\in\mathcal{K}^{n-1}:\exists z (y,z)M}(y,z)\in M\}.
Define F1,,Fm:π(M)𝒦F_{1},...,F_{m}:\pi(M)\rightarrow\mathcal{K} as follows:

F1(u)= the least v such that (u,v)M.F_{1}(u)=\text{ the least }v\text{ such that }(u,v)\in M.

Suppose F1,,FjF_{1},...,F_{j} are defined. Let

Fj+1(u)= the least v such that (u,v)M and v{F1(u),,Fj(u)}F_{j+1}(u)=\text{ the least }v\text{ such that }(u,v)\in M\text{ and }v\notin\{F_{1}(u),...,F_{j}(u)\}

if such vv exists; otherwise, let

Fj+1(u)=Fj(u).F_{j+1}(u)=F_{j}(u).

Then M=(π(M)×𝒦)M=j=1𝑚{(u,Fj(u)):uπ(M)}M=(\pi(M)\times\mathcal{K})\cap M=\underset{j=1}{\overset{m}{\bigcup}}\{(u,F_{j}(u)):u\in\pi(M)\}.
Take jj such that p{(u,Fj(u)):uπ(M)}p\in\{(u,F_{j}(u)):u\in\pi(M)\}.
Let F:π(M)MF:\pi(M)\rightarrow M be the definable function FjF_{j}. Then for each yπ(M)y\in\pi(M), F(y)F(y) is a zero of order ii for the function f(y,)f(y,-) and pgrpah(F)p\in grpah(F). Then there is a CkC^{k}-cell 𝒞π(M)\mathcal{C}\subseteq\pi(M) such that FF is continuous on 𝒞\mathcal{C} and p(𝒞×𝒦)Mp\in(\mathcal{C}\times\mathcal{K})\cap M. ([Cos00] the theorem for cell decomposition and the theorem for piecewise continuous)
Recall that sup\sup and inf\inf exist for definable sets by [VdD98, Chapter 1, Lemma (3.3) (i)]. Also recall fact 2.4, which roughly says that the number of zeroes remains the same in a small neighborhood around a fixed zero. Then we can define functions α\alpha, β\beta as follows:
Define α:𝒞>0\alpha:\mathcal{C}\rightarrow\mathcal{R}^{>0} by

α(x)=12sup{r: there is s>0 such that for all yB(x,r)\alpha(x)=\frac{1}{2}\sup\{r\in\mathcal{R}:\text{ there is $s>0$ such that for all $y\in B(x,r)$}
 there are exactly i zeroes in B(F(x),s) for f(y,)}.\text{ there are exactly $i$ zeroes in $B(F(x),s)$ for $f(y,-)\}$}.

Define β:𝒞>0\beta:\mathcal{C}\rightarrow\mathcal{R}^{>0} by

β(x)=sup{r: for all yB(x,α(x)) there are exactly i zeroes \beta(x)=\sup\{r\in\mathcal{R}:\text{ for all $y\in B(x,\alpha(x))$ there are exactly $i$ zeroes }
in B(F(x),r)B(F(x),r) for f(y,)}f(y,-)\}.

Then there is a CkC^{k}-cell 𝒟𝒞\mathcal{D}\subseteq\mathcal{C} such that α,β\alpha,\beta are continuous on 𝒟\mathcal{D}. Note: β(x)\beta(x) satisfies that for all yB(x,α(x))y\in B(x,\alpha(x)) there are exactly ii zeroes in B(F(x),β(x))B(F(x),\beta(x)) for f(y,)}f(y,-)\} because B(F(x),β(x))=r<β(x)B(F(x),r)B(F(x),\beta(x))=\underset{r<\beta(x)}{\bigcup}B(F(x),r).
As in [Cos00], let φ\varphi be the function φ:N𝒟2n2\varphi:N\mathcal{D}\rightarrow\mathcal{R}^{2n-2} where φ(x,v)=x+v\varphi(x,v)=x+v and let ZZ be the subset of (x,v)(x,v) in N𝒟N\mathcal{D} such that

d(x,v)φ:T(x,v)(N𝒟)2n2d_{(x,v)}\varphi:T_{(x,v)}(N\mathcal{D})\rightarrow\mathcal{R}^{2n-2}

is not an isomorphism. Define θ:𝒟>0\theta:\mathcal{D}\rightarrow\mathcal{R}^{>0} such that

  1. 1.

    θ(x)min{1,dist((x,0),Z),α(x)}\theta(x)\leq\min\{1,dist((x,0),Z),\alpha(x)\}

  2. 2.

    θ(x)inf{r:(y,w)N𝒟\theta(x)\leq\inf\{r\in\mathcal{R}:\exists(y,w)\in N\mathcal{D} vNx𝒟\exists v\in N_{x}\mathcal{D} wv=r\|w\|\leq\|v\|=r and y+w=x+vy+w=x+v }\}.

By continuity of α\alpha and by the proof of fact 2.9, θ\theta is locally bounded below. By fact 2.8, there is a definable continuous ϵ:𝒟>0\epsilon:\mathcal{D}\rightarrow\mathcal{R}^{>0} such that ϵ<θ\epsilon<\theta. (We are not using fact 2.6 to get a cell 𝒟𝒟\mathcal{D}^{\prime}\subseteq\mathcal{D} on which θ\theta is continuous since N𝒟N\mathcal{D}^{\prime} might be very different from N𝒟N\mathcal{D}.) Define a set

V={(y,z)𝒦n1×𝒦:y=u+v for some unique u𝒟,vNu𝒟 and V=\{(y,z)\in\mathcal{K}^{n-1}\times\mathcal{K}:y=u+v\text{ for some unique }u\in\mathcal{D},v\in N_{u}\mathcal{D}\text{ and }
v<ϵ(u),zB(F(u),β(u))}\|v\|<\epsilon(u),z\in B(F(u),\beta(u))\}

By the proof of fact 2.9, π(V)\pi(V) is definable open and φ|π(V)\varphi|\pi(V) is a diffeomorphism.

Claim 2.10.

VV is a open neighborhood of pp such that for any yπ(V)y\in\pi(V), there are exactly ii zeroes of f(y,)f(y,-) in VyV_{y}.

Proof.

pVp\in V since pgraph(F)Vp\in graph(F)\subseteq V.
Fix yπ(V)y\in\pi(V). Let u𝒟u\in\mathcal{D}, vNu𝒟v\in N_{u}\mathcal{D} be the unique elements such that y=u+vy=u+v. Then zVyz\in V_{y} iff zB(F(u),β(u)))z\in B(F(u),\beta(u))). By the choice of β\beta, ϵ\epsilon, there are exactly ii zeroes of f(y,)f(y,-) in B(F(u),β(u))B(F(u),\beta(u)).
We now show that VV is open. Let (y,z)V(y,z)\in V. Write y=u+vy=u+v where u𝒟u\in\mathcal{D} and vNu𝒟v\in N_{u}\mathcal{D} are unique. By continuity of α,β,F\alpha,\beta,F, there is s>0s>0 such that for all u𝒟u^{\prime}\in\mathcal{D} with uu<s\|u^{\prime}-u\|<s,

  • α(u)>12(α(u)v)+v\alpha(u^{\prime})>\frac{1}{2}(\alpha(u)-\|v\|)+\|v\|,

  • ϵ(u)>12(ϵ(u)v)+v\epsilon(u^{\prime})>\frac{1}{2}(\epsilon(u)-\|v\|)+\|v\|,

  • β(u)>zF(u)+12(β(u)zF(u))\beta(u^{\prime})>\|z-F(u)\|+\frac{1}{2}(\beta(u)-\|z-F(u)\|) and

  • F(u)F(u)14(β(u)zF(u))\|F(u)-F(u^{\prime})\|\leq\frac{1}{4}(\beta(u)-\|z-F(u)\|).

Since by the proof of fact 2.9, φ|π(V)\varphi|\pi(V) is a diffeomorphism, there is r>0r>0 such that for all y𝒦n1y^{\prime}\in\mathcal{K}^{n-1} with yy<r\|y^{\prime}-y\|<r, y=u+vy^{\prime}=u^{\prime}+v^{\prime} where u𝒟u^{\prime}\in\mathcal{D}, vNu𝒟v^{\prime}\in N_{u^{\prime}}\mathcal{D} are unique and uu<s\|u^{\prime}-u\|<s, vv<min{14(ϵ(u)v),14(α(u)v)}\|v^{\prime}-v\|<\min\{\frac{1}{4}(\epsilon(u)-\|v\|),\frac{1}{4}(\alpha(u)-\|v\|)\}. Then for (y,z)B(y,r)×B(z,14(β(u)zF(u)))(y^{\prime},z^{\prime})\in B(y,r)\times B(z,\frac{1}{4}(\beta(u)-\|z-F(u)\|)), write yy^{\prime} as y=u+vy^{\prime}=u^{\prime}+v^{\prime} with unique u𝒟u^{\prime}\in\mathcal{D} and vNu𝒟v^{\prime}\in N_{u}\mathcal{D}. We have

vvv+vmin{14(α(u)v)+v,14(ϵ(u)v)+v}\|v^{\prime}\|\leq\|v^{\prime}-v\|+\|v\|\leq\min\{\frac{1}{4}(\alpha(u)-\|v\|)+\|v\|,\frac{1}{4}(\epsilon(u)-\|v\|)+\|v\|\}
<min{ϵ(u),α(u)}<\min\{\epsilon(u^{\prime}),\alpha(u^{\prime})\}

and

zF(u)zz+zF(u)+F(u)F(u)\|z^{\prime}-F(u^{\prime})\|\leq\|z^{\prime}-z\|+\|z-F(u)\|+\|F(u)-F(u^{\prime})\|\leq
14(β(u)zF(u))+zF(u)+14(β(u)zF(u))\frac{1}{4}(\beta(u)-\|z-F(u)\|)+\|z-F(u)\|+\frac{1}{4}(\beta(u)-\|z-F(u)\|)\leq
zF(u)+12(β(u)zF(u))<β(u)\|z-F(u)\|+\frac{1}{2}(\beta(u)-\|z-F(u)\|)<\beta(u^{\prime})

Hence B(y,r)×B(z,14(β(u)zF(u)))VB(y,r)\times B(z,\frac{1}{4}(\beta(u)-\|z-F(u)\|))\subseteq V and VV is open. ∎

3 Proof

Outline of the proof: Given pSn(𝒦)p\in S_{n}(\mathcal{K}), use lemma 2.2 to get a neighborhood of pp with fixed number of zeroes. Then follow the proof of [PS03, Theorem 2.23] to get the type version of Weierstrass division theorem. Then the rest is just the same as in [PS08].

Theorem 3.1.

(type version of [PS03, Theorem 2.23.])
Let pSn(𝒦)p\in S_{n}(\mathcal{K}), UU a definable open neighborhood of pp. Let f(z1,,zn1,y),g(z,y)𝒪p,nf(z_{1},...,z_{n-1},y),g(z,y)\in\mathcal{O}_{p,n} be defined and 𝒦\mathcal{K}-differentiable on UU. Suppose for all yπ(U)y\in\pi(U), there are finitely many zeroes of f(y,)f(y,-) in Uy:={x𝒦:(y,x)U}U_{y}:=\{x\in\mathcal{K}:(y,x)\in U\}, counting multiplicity.
Then there is kk\in\mathbb{N}, a definable open set VUV\subseteq U with pVp\in V and unique q(z,y)𝒪V,nq(z,y)\in\mathcal{O}_{V,n}, R0(z),,Rk1(z)𝒪V,n1R_{0}(z),...,R_{k-1}(z)\in\mathcal{O}_{V,n-1} such that

g(z,y)=q(z,y)f(z,y)+Rk1(z)yk1++R1(z)y+R0(z) on V.g(z,y)=q(z,y)f(z,y)+R_{k-1}(z)y^{k-1}+...+R_{1}(z)y+R_{0}(z)\text{ on $V$}.
Proof.

Fix g𝒪p,ng\in\mathcal{O}_{p,n}. Suppose f,gf,g are defined and 𝒦\mathcal{K}- differentiable on a definable open set UUU^{\prime}\subseteq U. By lemma 2.2, there exist kk\in\mathbb{N} and VUV\subseteq U^{\prime} a definable open neighborhood of pp such that for any yπ(V)y\in\pi(V), there are exactly kk zeroes of f(y,)f(y,-) in VyV_{y} (counting multiplicity). Then the rest of the proof is the same as in [PS03, Theorem 2.23.].∎

Theorem 3.2.

(type version of [PS08, Theorem 11.2.]) Assume that U𝒦nU\subseteq\mathcal{K}^{n} is a definable open set and AUA\subseteq U an irreducible 𝒦\mathcal{K}-analytic subset of UU of dimension dd. Assume also:

  1. (i)

    The projection π\pi of AA on the first dd coordinates is definably proper over its image, and π(A)\pi(A) is open in 𝒦d\mathcal{K}^{d}.

  2. (ii)

    There is a definable set S𝒦dS\subseteq\mathcal{K}^{d}, of \mathcal{R}-dimension 2d2\leq 2d-2 and a natural number mm, such that π|A\pi|A is mm-to-11 outside the set Aπ1(S)A\cap\pi^{-1}(S), π\pi is a local homeomorphism outside of the set π1(S)\pi^{-1}(S), and Aπ1(S)A\setminus\pi^{-1}(S) is dense in AA.

  3. (iii)

    The coordinate function zzd+1z\mapsto z_{d+1} is injective on Aπ1(x)A\cap\pi^{-1}(x^{\prime}) for every \mathcal{R}-generic xπ(A)x^{\prime}\in\pi(A). Namely, for all z,wπ1(x)z,w\in\pi^{-1}(x^{\prime}), if zd+1=wd+1z_{d+1}=w_{d+1} then z=wz=w.

Then, there is a definable open set UUU^{\prime}\subseteq U containing AA, a natural number ss and 𝒦\mathcal{K}-holomorphic functions G1,,Gr,D:U𝒦G_{1},...,G_{r},D:U^{\prime}\rightarrow\mathcal{K}, such that for every pSn(𝒦)p\in S_{n}(\mathcal{K}) with pAp\in A and f𝒪pf\in\mathcal{O}_{p}, if g1,,gr,δg_{1},...,g_{r},\delta are the germs at pp of G1,,Gr,DG_{1},...,G_{r},D, resp, then:

f(A)pf1,,fr𝒪p(δsf=f1g1++frgr)f\in\mathcal{I}(A)_{p}\iff\exists f_{1},...,f_{r}\in\mathcal{O}_{p}(\delta^{s}f=f_{1}g_{1}+...+f_{r}g_{r}) (1)
Proof.

Define Pd+1,,PnP_{d+1},...,P_{n} and {D(z)ziRi(z,zd+1):i=d+2,,n}\{D(z^{\prime})z_{i}-R_{i}(z^{\prime},z_{d+1}):i=d+2,...,n\} satisfying [PS08, Claim 11.4., Claim 11.5., Claim 11.6.] as in the proof of [PS08, Theorem 11.2.]. These are 𝒦\mathcal{K}-holomorphic functions defined on the open set π(A)×𝒦ndA\pi(A)\times\mathcal{K}^{n-d}\supseteq A. Hence pπ(A)×𝒦ndp\in\pi(A)\times\mathcal{K}^{n-d} and we can consider the germs pd+1,,pnp_{d+1},...,p_{n} for Pd+1,,PnP_{d+1},...,P_{n} respectively, the germ δ\delta for D(z)D(z^{\prime}), and ri(z,zd+1)r_{i}(z^{\prime},z_{d+1}) for Ri(z,zd+1)R_{i}(z^{\prime},z_{d+1}) in the ring 𝒪p\mathcal{O}_{p}.
Let JpJ_{p} be the ideal of 𝒪p\mathcal{O}_{p} generated by the germs of Pd+1,,PnP_{d+1},...,P_{n} and D(z)ziRi(z,zd+1)D(z^{\prime})z_{i}-R_{i}(z^{\prime},z_{d+1}), i=d+2,,ni=d+2,...,n at pp. Let s=(m1)(n(d+1))s=(m-1)(n-(d+1)). As in [PS08, Theorem 11.2.], for all f𝒪pf\in\mathcal{O}_{p}, δsfJp\delta^{s}f\in J_{p}.
Now we show that for all f𝒪pf\in\mathcal{O}_{p}, if δsfJp\delta^{s}f\in J_{p} then f(A)pf\in\mathcal{I}(A)_{p}. We follow the proof in [PS08, Theorem 11.2.]

Claim 3.3.

(type version of [PS08, Claim 11.7.]) For every h(z,zd+1)𝒪ph(z^{\prime},z_{d+1})\in\mathcal{O}_{p}, if h(A)ph\in\mathcal{I}(A)_{p} then hh is divisible by pd+1(z,zd+1)p_{d+1}(z^{\prime},z_{d+1}).

Proof.

Let hh be defined on some definable open set WW with pWp\in W. Since Aπ(A)×𝒦ndA\subseteq\pi(A)\times\mathcal{K}^{n-d} and pAp\in A, may assume π(W)π(A)\pi(W)\subseteq\pi(A). zW\forall z\in W, if pd+1(z)=0p_{d+1}(z)=0 then h(z)=0h(z)=0: Suppose zWz\in W and pd+1(z)=0p_{d+1}(z)=0. Write zz as z=(π(z),zd+1,,zn)z=(\pi(z),z_{d+1},...,z_{n}). If π(z)S\pi(z)\notin S, then zd+1=ϕi,d+1(π(z))z_{d+1}=\phi_{i,d+1}(\pi(z)) for some i{1,,m}i\in\{1,...,m\}. (The ϕi=(ϕi,d+1,,ϕi,n)\phi_{i}=(\phi_{i,d+1},...,\phi_{i,n})’s are defined as in [PS08, Claim 11.4.].) Since (π(z),ϕi(π(z)))A(\pi(z),\phi_{i}(\pi(z)))\in A and h(A)ph\in\mathcal{I}(A)_{p}, h(z)=h(π(z),ϕi(π(z)))=0h(z)=h(\pi(z),\phi_{i}(\pi(z)))=0. Note that we didn’t say z=(π(z),ϕi(π(z))z=(\pi(z),\phi_{i}(\pi(z)), but the value of hh is determined by the first d+1d+1 coordinates. If π(z)S\pi(z)\in S, since Wπ1(S)W\setminus\pi^{-1}(S) is dense in WW and zeroes of pd+1p_{d+1} are zeroes of hh, by fact 2.4, every zero of pd+1p_{d+1} in Wπ1(S)W\cap\pi^{-1}(S) is also a zero of hh, of at least the same multiplicity. (i.e. If zz is a zero of pd+1p_{d+1} of multiplicity mm, then zz is a zero of hh of multiplicity m\geq m.)
Also by fact 2.4, for all zπ(W)z^{\prime}\in\pi(W),

 |{zW:π(z)=z and pd+1(z)=0}|m.\text{ $|\{z\in W:\pi(z)=z^{\prime}$ and $p_{d+1}(z)=0\}|\leq m$}.

Define the function uu on WZ(pd+1)W\setminus Z(p_{d+1}), where Z(pd+1)Z(p_{d+1}) means the zero set of pd+1p_{d+1}, by

u(z)=h(z)pd+1(z).u(z)=\dfrac{h(z)}{p_{d+1}(z)}.

Define a function u¯\overline{u} on WW by

u¯(z)=u(z) if zZ(pd+1);\overline{u}(z)=u(z)\text{ if $z\notin Z(p_{d+1})$};
u¯(z)=limwzu(w) if zZ(pd+1).\overline{u}(z)=\lim_{w\rightarrow z}u(w)\text{ if $z\in Z(p_{d+1})$}.

By [PS01, Lemma 2.42], for any zπ(W)z^{\prime}\in\pi(W), the function u¯z(y):=limyyh(z,y)pd+1(z,y)\overline{u}_{z^{\prime}}(y):=\lim_{y^{\prime}\rightarrow y}\dfrac{h(z^{\prime},y)}{p_{d+1}(z^{\prime},y)} is well-defined and 𝒦\mathcal{K}-holomorphic, since it has isolated singularities only. By [PS03, Theorem 2.7], u¯\overline{u} is continuous on WW. By [PS03, Theorem 2.14], since dimZ(pd+1)2n1dim_{\mathcal{R}}Z(p_{d+1})\leq 2n-1, u¯\overline{u} is 𝒦\mathcal{K}-holomorphic on WW. Since h=u¯pd+1h=\overline{u}\cdot p_{d+1} on Wπ1(S)W\setminus\pi^{-1}(S), which is dense in WW, h=u¯pd+1h=\overline{u}\cdot p_{d+1} on WW. ∎

Claim 3.4.

(type version of [PS08, Claim 11.8.]) For every g(z)𝒪pg(z)\in\mathcal{O}_{p}, there is h(z,ud+1,,un)𝒪d,p[u¯]h(z^{\prime},u_{d+1},...,u_{n})\in\mathcal{O}_{d,p}[\bar{u}] of degree less than mm in each variable ud+1,,unu_{d+1},...,u_{n} such that g(z)g(z) is equivalent to h(z,zd+1,,zn)h(z^{\prime},z_{d+1},...,z_{n}) modulo JpJ_{p}.

Proof.

Given g(z)𝒪pg(z)\in\mathcal{O}_{p} defined on some definable open U𝒦nU\subseteq\mathcal{K}^{n} with pUp\in U, may assume π(U)π(A)\pi(U)\subseteq\pi(A) as in Claim 3.3. Since zπ(U)\forall z^{\prime}\in\pi(U), pn(z,w)p_{n}(z^{\prime},w) has mm zeroes, by theorem 3.1, there exist kk\in\mathbb{N} with kmk\leq m, a definable open set VUV\subseteq U with pVp\in V and unique

q(z,ud+1,,un)𝒪V,n,q(z^{\prime},u_{d+1},...,u_{n})\in\mathcal{O}_{V,n},
R0(z,ud+1,,un1),,Rk1(z,ud+1,,un1)𝒪V,n1R_{0}(z^{\prime},u_{d+1},...,u_{n-1}),...,R_{k-1}(z^{\prime},u_{d+1},...,u_{n-1})\in\mathcal{O}_{V,n-1}

such that

g(z,ud+1,,un)=q(z,ud+1,,un)pn(z,un)+g(z^{\prime},u_{d+1},...,u_{n})=q(z^{\prime},u_{d+1},...,u_{n})p_{n}(z^{\prime},u_{n})+
Rk1(z,ud+1,,un1)unk1++R1(z,ud+1,,un1)un+R0(z,ud+1,,un1)R_{k-1}(z^{\prime},u_{d+1},...,u_{n-1})u_{n}^{k-1}+...+R_{1}(z^{\prime},u_{d+1},...,u_{n-1})u_{n}+R_{0}(z^{\prime},u_{d+1},...,u_{n-1})

on VV. Apply theorem 3.1. to R0,,Rk1R_{0},...,R_{k-1} by dividing pn1(z,un1)p_{n-1}(z^{\prime},u_{n-1}). Repeat this as in [PS08, Claim 11.8.] and we get the conclusion. ∎

Claim 3.5.

(type version of [PS08, Claim 11.9.]) For every g(z)𝒪pg(z)\in\mathcal{O}_{p}, there is q(z,u)𝒪d,p[u]q(z^{\prime},u)\in\mathcal{O}_{d,p}[u] such that δsg(z)\delta^{s}g(z) is equivalent modulo JpJ_{p} to q(z,zd+1)q(z^{\prime},z_{d+1}).

Proof.

The same as in [PS08, Claim 11.9.]. ∎

We get Theorem 3.2. as in [PS08, Theorem 11.2.] using Claim 3.5 and Claim 3.3.

Theorem 3.6.

(type version of [PS08, Theorem 11.3.]) Assume that AA is a 𝒦\mathcal{K}-analytic subset of U𝒦nU\subseteq\mathcal{K}^{n} and assume that G1,,GtG_{1},...,G_{t} are 𝒦\mathcal{K}-holomorphic maps from AA into 𝒦N\mathcal{K}^{N}. Then we can write AA as a union of finitely many relatively open sets A1,,AmA_{1},...,A_{m} such that on each AiA_{i} the following holds:
There are finitely many tuples of 𝒦\mathcal{K}-holomorphic functions on AiA_{i},

{(Hj,1,,Hj,t):j=1,,k},k=k(i),\{(H_{j,1},...,H_{j,t}):j=1,...,k\},k=k(i),

with the property that for every pSn(𝒦)p\in S_{n}(\mathcal{K}) with pAip\in A_{i}, the module Rp(g1,,gt)R_{p}(g_{1},...,g_{t}) equals its submodule generated by {(hj,1,,hj,t):j=1,,k}\{(h_{j,1},...,h_{j,t}):j=1,...,k\} over 𝒪p\mathcal{O}_{p} (where gig_{i} and hi,jh_{i,j} are the germs of GiG_{i} and Hi,jH_{i,j} at pp, resp).

Proof.

Induction on nn. When n=0n=0, (G1,..,Gt)(G_{1},..,G_{t}) can be considered as a vector in 𝒦t\mathcal{K}^{t} and {(ϕ1,,ϕt):U𝒦t:ϕ1G1++ϕtGt=0}\{(\phi_{1},...,\phi_{t}):U\rightarrow\mathcal{K}^{t}:\phi_{1}G_{1}+...+\phi_{t}G_{t}=0\} is a vector subspace of 𝒦t\mathcal{K}^{t}.
Assume true for n1n-1 and prove for nn. Consider first N=1N=1. As in [PS08, Claim 1 in Theorem 11.3.], may assume G1,,GtG_{1},...,G_{t} are Weierstrass polynomials ω1(z,zn),.,ωt(z,zn):V×𝒦=π(U)×𝒦𝒦\omega_{1}(z^{\prime},z_{n}),....,\omega_{t}(z^{\prime},z_{n}):V\times\mathcal{K}=\pi(U)\times\mathcal{K}\rightarrow\mathcal{K}, where π\pi is the projection onto the first n1n-1 coordinates.
We reduce to the case where (ϕ1,,ϕt)(\phi_{1},...,\phi_{t}) are polynomials as in [PS08, Claim 2 in Theorem 11.3.]:

Claim 3.7.

(type version of [PS08, Claim 2 in Theorem 11.3.]) For pSn(𝒦)p\in S_{n}(\mathcal{K}) with pUp\in U^{\prime}, UUU^{\prime}\subseteq U definable and open, let ϕ=(ϕ1,,ϕs)RU(ω)\phi=(\phi_{1},...,\phi_{s})\in R_{U^{\prime}}(\omega).
Then there are tuples {ψi=(ψ1i,,ψsi):i=1,,t}\{\psi^{i}=(\psi^{i}_{1},...,\psi^{i}_{s}):i=1,...,t\}, where each ψji\psi^{i}_{j} is a polynomial in znz_{n} of degree m\leq m over 𝒪V′′\mathcal{O}_{V^{\prime\prime}} (V′′=π(U′′)V^{\prime\prime}=\pi(U^{\prime\prime}), where U′′UU^{\prime\prime}\subseteq U^{\prime} is definable open and pU′′p\in U^{\prime\prime}), such that ϕ\phi is in the module generated by ψ1,,ψt\psi^{1},...,\psi^{t} over 𝒪U′′\mathcal{O}_{U^{\prime\prime}}.

Proof.

Let V=π(U)V=\pi(U). By lemma 2.2, there exist kk\in\mathbb{N} and U′′UU^{\prime\prime}\subseteq U^{\prime} a definable open neighborhood of pp such that for any zπ(U′′)z\in\pi(U^{\prime\prime}), there are exactly kk zeroes of ω1(z,)\omega_{1}(z,-) in Uy′′U^{\prime\prime}_{y} (counting multiplicity). Let h1,,hkh_{1},...,h_{k} be definable functions such that for all zπ(U′′)z\in\pi(U^{\prime\prime}), {h1(z),,hk(z)}\{h_{1}(z),...,h_{k}(z)\} list all zeroes of ω1(z,)\omega_{1}(z,-) in Uy′′U_{y}^{\prime\prime}. Take

ω1(z,y)=i=1𝑘(yhi(z))\omega_{1}^{\prime}(z^{\prime},y)=\underset{i=1}{\overset{k}{\prod}}(y-h_{i}(z^{\prime})).

ω1\omega_{1}^{\prime} is holomorphic by the same argument as in [PS03, Theorem 2.20]. By theorem 3.1, there exist definable unique qi(z,zn)𝒪U′′q_{i}(z^{\prime},z_{n})\in\mathcal{O}_{U^{\prime\prime}}, polynomials ri(z,zn)𝒪U′′r_{i}(z^{\prime},z_{n})\in\mathcal{O}_{U^{\prime\prime}} of degree <k<k such that ϕi=qiω1+ri\phi_{i}=q_{i}\omega_{1}^{\prime}+r_{i} on U′′U^{\prime\prime}. Since (z,y)U′′\forall(z^{\prime},y)\in U^{\prime\prime}, ω1(z,y)=0\omega_{1}^{\prime}(z^{\prime},y)=0 implies ω1(z,y)=0\omega_{1}(z^{\prime},y)=0 and for any such zero (z,y)U′′(z^{\prime},y)\in U^{\prime\prime}, the degree of the zero yy in ω1(z)\omega_{1}^{\prime}(z^{\prime}-) is the same as the degree of that zero in ω1(z)\omega_{1}(z^{\prime}-), by the same construction as in Claim 3.3, there is u𝒪U′′u^{\prime}\in\mathcal{O}_{U^{\prime\prime}} such that ω1=uω1\omega_{1}=u^{\prime}\omega_{1}^{\prime}. Moreover, (z,y)U′′\forall(z^{\prime},y)\in U^{\prime\prime}, u0u^{\prime}\neq 0. Define

ψ2=(ω2,ω1,0,,0),,\psi^{2}=(-\omega_{2},\omega_{1},0,...,0),...,
ψs=(ωs,0,,0,ω1),\psi^{s}=(-\omega_{s},0,...,0,\omega_{1}),
ψ=ϕ1+q2ω2++qsωs,\psi=\phi_{1}+q_{2}\omega_{2}+...+q_{s}\omega_{s},
P=(r2ω2++rsωs),P=-(r_{2}\omega_{2}+...+r_{s}\omega_{s}),
Q=Q= a 𝒦\mathcal{K}-holomorphic polynomial in variable znz_{n} over π(U′′)\pi(U^{\prime\prime}) extending uψu^{\prime}\psi,
Qi=Q_{i}= a 𝒦\mathcal{K}-holomorphic polynomial in variable znz_{n} extending uriu^{\prime}r_{i} for i{2,,s}i\in\{2,...,s\},

and

ψ1=(Q,Q2,,Qs).\text{$\psi_{1}=(Q,Q_{2},...,Q_{s})$}.

As in [Whi72, Chapter 8, Theorem 8B],

ϕ=i=2𝑠qiψi+(1/u)(Q,Q2,Qs)\phi=\underset{i=2}{\overset{s}{\sum}}q_{i}\psi_{i}+(1/u^{\prime})(Q,Q_{2}...,Q_{s}) in U′′U^{\prime\prime}.

(For the same reason as in [Whi72, Chapter 8, Theorem 8B], it’s important to use ω1\omega_{1}^{\prime} instead of ω1\omega_{1} because there might be zeros of ω1\omega_{1} outside the domain of ϕ\phi, i.e. outside UU^{\prime}. When there exist zeros of ω1\omega_{1} outside the domain of ϕ\phi, we cannot use the equation ψω1=P\psi\omega_{1}=P to conclude that zeros of ω1\omega_{1} are also zeros of PP, and hence cannot extend ψ\psi as P/ω1P/\omega_{1}.). ∎

The rest is the same argument as in [PS08, Theorem 11.3.]

Theorem 3.8.

(type version of [PS08, Theorem 11.1.]) Let MM be a 𝒦\mathcal{K}-manifold and AMA\subseteq M a 𝒦\mathcal{K}-analytic subset of MM. Then there are finitely many open sets V1,,VkV_{1},...,V_{k} whose union covers MM and for each i=1,,ki=1,...,k there are finitely many 𝒦\mathcal{K}-holomorphic functions fi,1,,fi,mif_{i,1},...,f_{i,m_{i}} in Vi(A)\mathcal{I}_{V_{i}}(A), such that for every pSn(𝒦)p\in S_{n}(\mathcal{K}) with pVip\in V_{i} the functions fi,1,,fi,mif_{i,1},...,f_{i,m_{i}} generate the ideal (A)p\mathcal{I}(A)_{p} in 𝒪p\mathcal{O}_{p}.
Moreover, the ViV_{i}’s and the fi,jf_{i,j}’s are all definable over the same parameters defining MM and AA.

Proof.

The same as in [PS08] using Theorem 3.2 and Theorem 3.6. ∎

Theorem 3.9.

(another proof of [BBT22, Theorem 2.21]) The definable structure sheaf 𝒪𝒦n\mathcal{O}_{\mathcal{K}^{n}} of 𝒦n\mathcal{K}^{n} is a coherent 𝒪𝒦n\mathcal{O}_{\mathcal{K}^{n}}-module as a sheaf on the site 𝒦n¯\underline{\mathcal{K}^{n}}.

Proof.

𝒪𝒦n\mathcal{O}_{\mathcal{K}^{n}} is a generated by 11 as an 𝒪𝒦n\mathcal{O}_{\mathcal{K}^{n}}-module.
It suffices to show that given any definable open U𝒦nU\subseteq\mathcal{K}^{n} and any 𝒪U\mathcal{O}_{U}-module homomorphism φ:𝒪Um𝒪U\varphi:\mathcal{O}^{m}_{U}\rightarrow\mathcal{O}_{U}, the kernel of φ\varphi is of finite type. i.e. we want a finite definable cover UiU_{i} of UU and surjections 𝒪Uin\mathcal{O}^{n}_{U_{i}}\twoheadrightarrow (kerφ)Ui(ker\varphi)_{{}_{U_{i}}} for some positive integer nn on each of those open sets. Let G1,,GmG_{1},...,G_{m} be definable 𝒦\mathcal{K}-holomophic functions from UU to 𝒦\mathcal{K} such that φ(ej)=Gj\varphi(e_{j})=G_{j} for all eje_{j} in the canonical basis {e1,,em}\{e_{1},...,e_{m}\} of 𝒪Um\mathcal{O}^{m}_{U}. By theorem 3.6, UU is a union of finitely many definable open sets U1,,UlU_{1},...,U_{l} such that on each UiU_{i} the following holds:
There are finitely many tuples of 𝒦\mathcal{K}-holomorphic functions on UiU_{i}, {(Hj,1,,Hj,m):j=1,,k}\{(H_{j,1},...,H_{j,m}):j=1,...,k\}, k=k(i)k=k(i), with the property that for every pSn(𝒦)p\in S_{n}(\mathcal{K}) with pUip\in U_{i}, the module Rp(g1,,gm)R_{p}(g_{1},...,g_{m}) equals its submodule generated by {(hj,1,,hj,m):j=1,,k}\{(h_{j,1},...,h_{j,m}):j=1,...,k\} over 𝒪p\mathcal{O}_{p} (where gig_{i} and hi,jh_{i,j} are the germs of GiG_{i} and Hi,jH_{i,j} at pp, resp).
Hence, given UiU_{i} as above and a definable open VUiV\subseteq U_{i}, fix any sker(φ)Ui(V)s\in ker(\varphi)_{U_{i}}(V). For each pSn(𝒦)p\in S_{n}(\mathcal{K}) with pVp\in V, since spker(φ)p=Rp(g1,,gm)s_{p}\in ker(\varphi)_{p}=R_{p}(g_{1},...,g_{m}), there is a definable open VpV_{p} with pVpVp\in V_{p}\subseteq V such that s|Vps|_{V_{p}} is generated by {(Hj,1|Vp,,Hj,m|Vp):j=1,,k}\{(H_{j,1}|_{V_{p}},...,H_{j,m}|_{V_{p}}):j=1,...,k\}, k=k(i)k=k(i) on VpV_{p}. By compactness of Sn(𝒦)S_{n}(\mathcal{K}), there exist finitely many p1,,psp_{1},...,p_{s} such that V=Vp1VpsV=V_{p_{1}}\cup...\cup V_{p_{s}} and on each of these VpαV_{p_{\alpha}}’s, s|Vpαs|_{V_{p_{\alpha}}} is generated by {(Hj,1|Vpα,,Hj,t|Vpα):j=1,,k}\{(H_{j,1}|_{V_{p_{\alpha}}},...,H_{j,t}|_{V_{p_{\alpha}}}):j=1,...,k\}, k=k(i)k=k(i). By definition 2.9, the morphism 𝒪Uik(i)ker(φ)Ui\mathcal{O}_{U_{i}}^{k(i)}\rightarrow ker(\varphi)_{U_{i}} given by mapping the canonical basis to {(Hj,1|Ui,,Hj,t|Ui):j=1,,k}\{(H_{j,1}|_{U_{i}},...,H_{j,t}|_{U_{i}}):j=1,...,k\} is surjective. Hence ker(φ)ker(\varphi) is of finite type and 𝒪𝒦n\mathcal{O}_{\mathcal{K}^{n}} is a coherent 𝒪𝒦n\mathcal{O}_{\mathcal{K}^{n}}-module.

4 Remark

Let X𝒦nX\subseteq\mathcal{K}^{n} be definable open. Let Sh𝒪X(X~)Sh^{\mathcal{O}_{X}}(\tilde{X}) be the category of usual sheaves 𝒪X\mathcal{O}_{X}-modules on X~\tilde{X}. Let Shdtop𝒪X(X)Sh_{dtop}^{\mathcal{O}_{X}}(X) be the category of sheaves of 𝒪X\mathcal{O}_{X}-modules on X¯\underline{X} as an o-minimal site. We show that Sh𝒪X(X~)Sh^{\mathcal{O}_{X}}(\tilde{X}) and Shdtop𝒪X(X)Sh_{dtop}^{\mathcal{O}_{X}}(X) are isomorphic categories, and the surjective maps are exactly the epimorphisms in both categories. Hence, from a category-theoretic perspective, theorem 3.8 immediately implies theorem 3.9.

4.1 Sheafification

Definition 4.1.

[Sta18, Part 1, Chapter 7, Section 10] Let X𝒦nX\subseteq\mathcal{K}^{n} be a definable open set. Let \mathcal{F} be a presheaf of 𝒪X\mathcal{O}_{X}-modules, and let U={Ui}i=1kU=\{U_{i}\}_{i=1}^{k} be a covering of UU. Let us use the notation (U)\mathcal{F}(U) to indicate the equalizer

H0(U,)={(si)i{1,,k}i(Ui):si|UiUj=sj|UiUji,j{1,,k}}.H^{0}(U,\mathcal{F})=\{(s_{i})_{i\in\{1,...,k\}}\in\prod_{i}\mathcal{F}(U_{i}):s_{i}|_{U_{i}\cap U_{j}}=s_{j}|_{U_{i}\cap U_{j}}\forall i,j\in\{1,...,k\}\}.
Definition 4.2.

[Sta18, Part 1, Chapter 7, Section 10] Let X𝒦nX\subseteq\mathcal{K}^{n} be a definable open set. For UXU\subseteq X be definable open, let JUJ_{U} be the set of all finite definable open coverings of UU. Define \leq on JUJ_{U} by UVU\leq V if VV is a refinement of UU. (JU,)(J_{U},\leq) is a directed set. For V={Vj}j=1lV=\{V_{j}\}_{j=1}^{l} a refinement of U={Ui}i=1kU=\{U_{i}\}_{i=1}^{k}, fix a function α:[l][k]\alpha:[l]\rightarrow[k] such that VjUα(j)V_{j}\subseteq U_{\alpha(j)}. Define μU,V:H0(U,)H0(V,)\mu_{U,V}:H^{0}(U,\mathcal{F})\rightarrow H^{0}(V,\mathcal{F}) by μU,V((si:i=1,k))=(sα(j)|Vj:j=1,,l)\mu_{U,V}((s_{i}:i=1,...k))=(s_{\alpha(j)}|_{V_{j}}:j=1,...,l). (In fact, μU,V\mu_{U,V} is independent of the choice of α\alpha: if VjUiV_{j}\subseteq U_{i} and VjUiV_{j}\subseteq U_{i^{\prime}}, then VjUiUiV_{j}\subseteq U_{i}\cap U_{i^{\prime}} and si|Vj=si|UiUi|Vj=si|UiUi|Vj=si|Vjs_{i}|_{V_{j}}=s_{i}|_{U_{i}\cap U_{i^{\prime}}}|_{V_{j}}=s_{i^{\prime}}|_{U_{i}\cap U_{i^{\prime}}}|_{V_{j}}=s_{i^{\prime}}|_{V_{j}}.) (H0(U,),μU,VH^{0}(U,\mathcal{F}),\mu_{U,V}) is a directed system. Define the presheaf +\mathcal{F}^{+} by

+(U)=UJUH0(U,)/\mathcal{F}^{+}(U)=\coprod_{U\in J_{U}}H^{0}(U,\mathcal{F})/\sim

where for sH0(U,)s\in H^{0}(U,\mathcal{F}) and sH0(V,)s^{\prime}\in H^{0}(V,\mathcal{F}) we have ssμU,𝒲(s)=μV,𝒲(s)s\sim s^{\prime}\iff\mu_{U,\mathcal{W}}(s)=\mu_{V,\mathcal{W}}(s^{\prime}) for some 𝒲U,V\mathcal{W}\geq U,V.
For a presheaf \mathcal{F}, define the canonical map τ:+\tau:\mathcal{F}\rightarrow\mathcal{F}^{+} by (U)+(U):s(s)/\mathcal{F}(U)\rightarrow\mathcal{F}^{+}(U):s\mapsto(s)/\sim.

Fact 4.1.

[Sta18, Part 1, Chapter 7, Section 10, Theorem 10.10]

  1. (1)

    The presheaf +\mathcal{F}^{+} is separated.

  2. (2)

    If \mathcal{F} is separated, then +\mathcal{F}^{+} is a sheaf and the map of presheaves +\mathcal{F}\rightarrow\mathcal{F}^{+} is injective.

  3. (3)

    If \mathcal{F} is a sheaf, then +\mathcal{F}\rightarrow\mathcal{F}^{+} is an isomorphism.

  4. (4)

    The presheaf ++\mathcal{F}^{++} is always a sheaf.

Definition 4.3.

[Sta18, Part 1, Chapter 7, Section 10, Definition 10.11] The sheaf #:=++\mathcal{F}^{\#}:=\mathcal{F}^{++} together with the canonical map τ#=τ+τ:+#\tau^{\#}=\tau^{+}\circ\tau:\mathcal{F}\rightarrow\mathcal{F}^{+}\rightarrow\mathcal{F}^{\#} is called the sheaf associated to \mathcal{F}.

4.2 Epimorphism

Following proofs in [Sta18], we show that surjective maps and epimorphisms coincide in the categories Shdtop𝒪X(X)Sh_{dtop}^{\mathcal{O}_{X}}(X) and Sh𝒪X(X~)Sh^{\mathcal{O}_{X}}(\tilde{X})

Lemma 4.2.

[Sta18, Part 1, Chapter 7, Lemma 11.2.] The surjective maps defined in definition 2.9 are exactly the epimorphisms of the category Shdtop𝒪X(X)Sh_{dtop}^{\mathcal{O}_{X}}(X).

Proof.

Let φ:𝒢\varphi:\mathcal{F}\rightarrow\mathcal{G} be an epimorphism between \mathcal{F}, 𝒢\mathcal{G} which are 𝒪X\mathcal{O}_{X}-modules on the o-minimal site X¯\underline{X}. Consider the presheaf \mathcal{H} defined by (U)=𝒢(U)𝒢(U)/S(U)\mathcal{H}(U)=\mathcal{G}(U)\oplus\mathcal{G}(U)/S(U) where S(U)S(U) is the 𝒪(U)\mathcal{O}(U)-submodule {(y,z)𝒢(U)𝒢(U):x(U)\{(y,z)\in\mathcal{G}(U)\oplus\mathcal{G}(U):\exists x\in\mathcal{F}(U) φU(x)=y,z=y}\varphi_{U}(x)=y,z=-y\} for UXU\subseteq X definable open (i.e. the pushout).
As in [Sta18, Part 1, Chapter 7, Section 3, Lemma 3.2.], consider the presheaf morphisms i1,i2:𝒢i_{1},i_{2}:\mathcal{G}\rightarrow\mathcal{H} defined by i1U(x)=(x,0)/S(U)i_{1_{U}}(x)=(x,0)/S(U) and i2U(x)=(0,x)/S(U)i_{2_{U}}(x)=(0,x)/S(U). Let i1=τ+τi1:𝒢#i_{1}^{\prime}=\tau^{+}\circ\tau\circ i_{1}:\mathcal{G}\rightarrow\mathcal{H}^{\#}, i2=τ+τi2:𝒢#i_{2}^{\prime}=\tau^{+}\circ\tau\circ i_{2}:\mathcal{G}\rightarrow\mathcal{H}^{\#}. Then i1i_{1}^{\prime}, i2i_{2}^{\prime} are morphisms in Shdtop𝒪(X)(X)Sh_{dtop}^{\mathcal{O}(X)}(X). Since i1φ=i2φi_{1}\circ\varphi=i_{2}\circ\varphi as presheaf morphisms by definition, i1φ=i2φi_{1}^{\prime}\circ\varphi=i_{2}^{\prime}\circ\varphi as sheaf morphisms. Since φ\varphi is an epimorphism, i1=i2i_{1}^{\prime}=i_{2}^{\prime}. Fix UXU\subseteq X definable open and y𝒢(U)y\in\mathcal{G}(U). Since i1(y)=i2(y)i_{1}^{\prime}(y)=i_{2}^{\prime}(y) and τ+\tau^{+} is injective, by fact 4.1 (1), (2), τ(i1(y))=τ(i2(y))\tau(i_{1}(y))=\tau(i_{2}(y)). By definition 4.2, there exists a finite definable open covering {Ui}i=1k\{U_{i}\}_{i=1}^{k} of UU such that (y|Ui,0|Ui)/S(Ui)=(0|Ui,y|Ui)/S(Ui)(y|_{U_{i}},0|_{U_{i}})/S(U_{i})=(0|_{U_{i}},y|_{U_{i}})/S(U_{i}) for all i=1,,ki=1,...,k. By definition of \mathcal{H}, for each i{1,,k}i\in\{1,...,k\}, there exists xi(Ui)x_{i}\in\mathcal{F}(U_{i}) such that φUi(xi)=y|Ui\varphi_{U_{i}}(x_{i})=y|_{U_{i}}. Hence φ\varphi is a surjective morphism.
The other direction is just checking definitions.

We have a similar result for Sh𝒪X(X~)Sh^{\mathcal{O}_{X}}(\tilde{X}), the category of (classical) sheaves on X~\tilde{X} as a topological space.

Lemma 4.3.

The surjective maps (i.e. surjective at the stalks) are exactly the epimorphisms of the category Sh𝒪X(X~)Sh^{\mathcal{O}_{X}}(\tilde{X}).

Proof.

The same as in the proof lemma 4.2 using the usual sheafification of sheaves. ∎

Proposition 4.4.

Sh𝒪X(X~)Sh^{\mathcal{O}_{X}}(\tilde{X}) and Shdtop𝒪X(X)Sh_{dtop}^{\mathcal{O}_{X}}(X) are isomorphic categories.

Proof.

The same proof as in [EJP06, Proposition 3.2]

Another proof of theorem 3.9: Let ι:Sh𝒪X(X~)Shdtop𝒪X(X)\iota:Sh^{\mathcal{O}_{X}}(\tilde{X})\rightarrow Sh_{dtop}^{\mathcal{O}_{X}}(X) be an isomorphism. Let U𝒦nU\subseteq\mathcal{K}^{n} be definable open and φ:𝒪Um𝒪U\varphi:\mathcal{O}^{m}_{U}\rightarrow\mathcal{O}_{U} a 𝒪U\mathcal{O}_{U}-module homomorphism. By theorem 3.6, there exists a finite definable open covering {Ui}i=1k\{U_{i}\}_{i=1}^{k} of UU such that for some ll\in\mathbb{N} and for each i{1,,k}i\in\{1,...,k\}, there exists ψi:𝒪U~ilker(ι1(φ))U~i\psi_{i}:\mathcal{O}_{\tilde{U}_{i}}^{l}\twoheadrightarrow ker(\iota^{-1}(\varphi))_{\tilde{U}_{i}}. Since surjective morphisms are epimorphisms in Sh𝒪X(X~)Sh^{\mathcal{O}_{X}}(\tilde{X}), ι(ψi):𝒪Uilker(φ)Ui\iota(\psi_{i}):\mathcal{O}_{U_{i}}^{l}\rightarrow ker(\varphi)_{U_{i}} is an epimorphism and hence a surjective morphism by lemma 4.2.

References

  • [BBT22] Benjamin Bakker, Yohan Brunebarbe, and Jacob Tsimerman. o-minimal gaga and a conjecture of griffiths. Inventiones mathematicae, pages 1–66, 2022.
  • [Cos00] Michel Coste. An introduction to o-minimal geometry. Istituti editoriali e poligrafici internazionali Pisa, 2000.
  • [EJP06] Mário Jorge Edmundo, Gareth O Jones, and Nicholas J Peatfield. Sheaf cohomology in o-minimal structures. Journal of Mathematical Logic, pages 1–20, 2006.
  • [Har13] Robin Hartshorne. Algebraic geometry, volume 52. Springer Science & Business Media, 2013.
  • [Oka50] Kiyoshi Oka. Sur les fonctions analytiques de plusieurs variables. vii. sur quelques notions arithmétiques. Bulletin de la Société mathématique de France, 78:1–27, 1950.
  • [PS01] Ya’acov Peterzil and Sergei Starchenko. Expansions of algebraically closed fields in o-minimal structures. Selecta Mathematica, 7(3):409–445, 2001.
  • [PS03] Ya’acov Peterzil and Sergei Starchenko. Expansions of algebraically closed fields ii: functions of several variables. Journal of Mathematical Logic, 3(01):1–35, 2003.
  • [PS08] Ya’acov Peterzil and Sergei Starchenko. Complex analytic geometry in a nonstandard setting. LONDON MATHEMATICAL SOCIETY LECTURE NOTE SERIES, 349:117, 2008.
  • [Sta18] The Stacks Project Authors. Stacks Project. https://stacks.math.columbia.edu, 2018.
  • [VdD98] Lou Van den Dries. Tame topology and o-minimal structures, volume 248. Cambridge university press, 1998.
  • [Whi72] Hassler Whitney. Complex analytic varieties, volume 131. Addison-Wesley Reading, 1972.