Orbifold modifications of complex analytic spaces
János Kollár and Wenhao Ou · 2025
Read Orbifold modifications of complex analytic spaces by János Kollár and Wenhao Ou (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2512-20708.txt.
Introduction
János Kollár and Wenhao Ou prove a modification theorem for singular complex spaces. Orbifolds are spaces that look locally like quotients by finite groups, and many complex analytic spaces are almost orbifolds except at their worst points. The authors show that every compact complex analytic space admits a bimeromorphic orbifold modification. The new space agrees with the original exactly on the region where the original was already a locally trivial orbifold. Only the singular locus is genuinely changed.
Catalog record
| CREATOR | János Kollár and Wenhao Ou |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 27 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2512-20708
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Orbifold modifications of complex analytic spaces János Kollár and Wenhao Ou Abstract . We show that a compact, complex analytic space X X has a bimeromorphic orbifold modification that is an isomorphism over the locally trivial orbifold locus of X X . August 10, 2026 A complex analytic space X is an orbifold if each point has an open neighborhood that is biholomorphic to an open neighborhood of 0\in{\mathbb{C}}^{n}/G for some finite subgroup G\subset\mathrm{GL}_{n}({\mathbb{C}}) . We can always assume that G contains no pseudo-reflections, and then such a G is unique. Any orbifold X is locally trivial along a dense, Zariski open subset of \operatorname{Sing}(X) , that is, locally biholomorphic to an open subset of {\mathbb{C}}^{r}/G\times{\mathbb{C}}^{n-r} , where {\mathbb{C}}^{r}/G has an isolated singularity at the origin. Our aim is to prove the following theorem. Theorem 1 . Let X be a compact, complex space and S^{\circ}\subset\operatorname{Sing}(X) a Zariski open subset such that X is a locally trivial orbifold along S^{\circ} . Then there is a projective, bimeromorphic modification \pi_{Y}:Y\to X , such that Y is an orbifold, and \pi_{Y} is an isomorphism over both S^{\circ} and the smooth locus of X . Complement 2 . The \pi_{Y}:Y\to X we construct has a \pi_{Y} -ample, exceptional, anti-effective divisor, and it is functorial for biholomorphisms ( 11 ). The case when X is projective is due to Chenyang Xu, without assuming local triviality of the orbifold structure along S^{\circ} . The proof is written up in [ LT19 , Sec.3] . The method appears to use projectivity in an essential way, and the complex analytic version without the local triviality assumption remains open. Applications of Theorem 1 to Bogomolov-Gieseker–type inequalities on complex spaces are treated in [ Ou25 ] . The proof relies on some general observations on neighborhoods of subspaces and on modifications. Both of these may be of independent interest. The arguments work more generally if X is countable at infinity, and the closure of S^{\circ} is compact. The following example illustrates some of the difficulties, and also the method to overcome them. Example 3 . This singularity X:=(xy=sz^{2}) is not {\mathbb{Q}} -factorial; D:=(x=s=0) is a divisor that is not {\mathbb{Q}} -Cartier. Thus X is not a quotient singularity. Blowing up the origin, in the chart x_{1}=x/s,y_{1}=y/s,z_{1}=z/s the equation becomes x_{1}y_{1}=sz_{1}^{2} . This is the same singularity as before, so it seems that it may not have an orbifold modification, without blowing up the s -axis. However, adjoining u=\sqrt{x},v=\sqrt{y},t=\sqrt{s} we get a Galois extension \bar{X}\to X with group G={\mathbb{Z}}/2\times{\mathbb{Z}}/2 , and equation \bar{X}=(uv=tz) . This can be resolved by blowing up the origin to get \tilde{X}\to\bar{X}\to X . Note that \tilde{X} is smooth and the G -action on \bar{X} lifts to a G -action on \tilde{X} . Thus Y:=\tilde{X}/G has only quotient singularities, \pi:Y\to X is projective, birational, and an isomorphism along the s -axis (except at the origin). Notation 4 . All complex spaces are assumed reduced and countable at infinity. Note that even if we start with a normal space X , some of the intermediate blow-ups in our construction can be non-normal (but reduced). This does not cause any problems. Let X be a complex space and D a divisor on X . Then (X,D) has simple normal crossing at x\in X if X is smooth at x , there are local coordinates x_{1},\dots,x_{n} such that D=(x_{1}\cdots x_{m}=0) near x for some m\leq n , and the (x_{i}=0) lie on different irreducible components of D . This is usually abbreviated as snc. {\mathbb{D}} denotes a complex disc, that is allowed to shrink as needed. The punctured disc is denoted by {\mathbb{D}}^{\circ} . We usually work with the Euclidean topology on X , and make clear when the Zariski topology is used. X^{\circ} always denotes a dense, Zariski open subset of a complex space X . 5 Outline of the proof. Let X be a reduced, complex space. Write \operatorname{Sing}(X)=S\cup R , where S,R are both unions of irreducible components of \operatorname{Sing}(X) , and S compact. The proof of Theorem 1 has 5 steps. In the first 2 we make no assumptions on the singularities of X along S . 5 .1 (Well prepared neighborhoods) Let B\subset S be the set of points where X is not equisingular along S ( 8 ). If p\in S\setminus B , then a transversal slice of X through p is topologically a cone in some neighborhood of radius \epsilon(p) . We use the Łojasiewicz inequality to bound \epsilon(p) from below as p goes to B . Once this is done, it is easy to see that, after a sequence of blow-ups whose centers lie over B , we may assume that S is smooth, there is a divisor D\subset X , a Euclidean open S\subset N\subset X , and a real analytic retraction \rho:N\to S , whose restriction \rho^{\circ}:(N\setminus D)\to(S\setminus D) is a topological fiber bundle. (There may be different fibers over different connected components of S\setminus D .) This is done in Section 1 . 5 .2 (Separating S and R ) We construct a sequence of blow-ups \pi:X^{r}\to X such that all blow-up centers lie over R , \pi induces a log resolution of X\setminus S , and one can write \operatorname{Ex}(\pi)=E^{r}+F^{r} such that \pi(E^{r})\subset S\cap R\quad\mbox{and}\quad(X^{r},E^{r}+F^{r})\quad\mbox{is snc along}\quad F^{r}. Note that S^{r}\cup E^{r}=\pi^{-1}(S) can be very singular. By contrast, F^{r} , which is the closure of \pi^{-1}(R\setminus S) , is snc. Although S^{r}\cup E^{r} and F^{r} are not disjoint, they only meet along the snc locus. Thus we separated the effects of S from the effects of R . We will obtain Y as a modification Y\to X^{r} that is an isomorphism over X^{r}\setminus E^{r} This is done in Section 2 . We will actually need a version that also keeps track of the divisor D ; this is small technical issue. Assume from now on that X has an isolated quotient singularity {\mathbb{C}}^{r}/G transversally along S\setminus D . 5 .3 (Local charts, first attempt) Let (V,V\cap D)\subset S be a local chart biholomorphic to \bigl({\mathbb{D}}^{n-r}_{\mathbf{s}},(s_{1}\cdots s_{m}=0)\bigr) for some m\leq n-r . Set V^{\circ}:=V\setminus D . Let U:=\rho^{-1}(V) be its preimage in N . Note that U^{\circ}:=U\setminus(D\cup S)=\rho^{-1}(V^{\circ})\setminus S is a topological ({\mathbb{C}}^{r}/G)\setminus\{0\} bundle over V^{\circ} . Its fundamental group is an extension 1\to G\to\pi_{1}(U^{\circ})\to{\mathbb{Z}}^{m}\to 0. Let H=H(U) be a quotient of \pi_{1}(U^{\circ}) such that the induced map G\to H is an injection. Let \bar{U}\to U be the corresponding finite, normal cover. \bar{U} is clearly smooth over U\setminus(D\cup S) , but it is also smooth over S\setminus D since G\to H is an injection. However, \bar{U} can be very singular along the preimage of D . Let us now take an H -equivariant resolution \tilde{U}\to\bar{U} . Then Y_{U}:=\tilde{U}/H\to U is a projective modification that has only quotient singularities. It is easy to see that if the quotients \pi_{1}(U^{\circ})\twoheadrightarrow H are chosen carefully, then the Y_{U} can be patched together to get the required resolution over the neighborhood N\supset S . If S is a connected component of \operatorname{Sing}X , then the rest of \operatorname{Sing}X can be resolved separately, and we are done. Otherwise \operatorname{Sing}X has an irreducible component R_{i} that intersects both N and X\setminus N . Then