Pathological MMP singularities as α p \alpha_{p} -quotients
Quentin Posva · 2025
Read Pathological MMP singularities as α p \alpha_{p} -quotients by Quentin Posva (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-2501-01179.txt.
Introduction
Quentin Posva constructs pathological singularity examples in algebraic geometry in this 2025 paper. The setting is the minimal model program in positive characteristic. Singularities there can behave worse than in characteristic zero, and sharp examples are valuable. Posva builds them as quotients by a special group action, called an alpha p action, in every positive characteristic.
Catalog record
| CREATOR | Quentin Posva |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 85 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2501-01179
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Pathological MMP singularities as α p \alpha_{p} -quotients Quentin Posva Abstract Abstract. We construct pathological examples of MMP singularities in every positive characteristic using quotients by \alpha_{p} -actions. In particular, we obtain non- S_{3} terminal singularities, as well as locally stable (respectively stable) families whose general fibers are smooth (respectively klt, Cohen–Macaulay and F -injective) and whose special fibers are non- S_{2} . The dimensions of these examples are bounded below by a linear function of the characteristic. 1. Introduction Singularities of the Minimal Model Program (MMP) have poor cohomological properties in positive characteristic, unlike in characteristic 0 . Striking examples are canonical and terminal singularities that are not Cohen–Macaulay [ Kov18 , Ber19 , Yas19 , Tot19 , Tot24 ] . In this note, we construct several new examples of non- S_{3} singularities of the MMP in every positive characteristic p>0 . The examples constructed in the aforementioned works are of two types: their stem either from a cone construction on a variety failing Kodaira vanishing, or from a wild \mathbb{Z}/p\mathbb{Z} -quotient. Our examples are \alpha_{p} -quotients: to understand their singularities we use the perspective of 1 -foliations and the tools developed in [ Pos23 ] . Our first series of examples gives the following: Theorem 1 ( Theorem 3.0.2 ) . Let k be an algebraically closed field of characteristic p>0 . Then there exists an isolated singularity (\mathbb{0}\in Y) over k that is klt (resp. canonical, resp. terminal) of dimension \max\{p,3\} (resp. \max\{2p-2,3\} , resp. 2p-1 ) and non- S_{3} . In particular, we obtain a three-dimensional terminal singularity in characteristic p=2 that is not S_{3} : the other knowns examples of this dimension were all constructed by Totaro in positive and mixed characteristics 2,3 and 5 [ Tot19 , Tot24 ] . For an arbitrary characteristic p>2 , Totaro also constructed terminal non- S_{3} cone singularities of dimension 2p+2 [ Tot19 ] . In this regard, we achieve a small improvement on the asymptotic dimensional lower bound. The lower bound of Theorem 1 on the dimension is not optimal for canonical singularities in characteristic p>3 : we shall find examples of smaller dimensions, although with non-isolated singularities, in Section 4.1 and Theorem 4.2.3 . We use the same method to produce another series of pathological examples. In characteristic 0 , fibers of locally stable one-parameter families (in the sense of Section 2.1 ) are S_{2} [ Kol13 , §7.3] . This is true for one-parameter families of surfaces with residue characteristics different from 2 , 3 and 5 [ ABP23 ] , but recent examples of Kollár [ Kol23a ] show that this property can fail for families of threefolds in every positive characteristic. Using \alpha_{p} -quotients, we construct new examples of this kind. First we obtain: Theorem 2 ( Theorem 4.1.3 ) . Let k be an algebraically closed field of characteristic p>0 . Then there exists a locally stable family Y\to\mathbb{A}^{1} over k of relative dimension \max\{p,3\} such that Y_{t} is smooth for t\neq 0 and Y_{0} is reduced but non- S_{2} . Then we compactify the families of Theorem 2 to obtain pathological stable families (in the sense of Section 2.1 ): Theorem 3 ( Theorem 4.1.7 ) . Let k be an algebraically closed field of characteristic p>0 . Then there exists a projective flat family of pairs (\mathcal{Y},\mathcal{B})\to\mathbb{A}^{1} over k of relative dimension \max\{p,3\} such that: (a) \mathcal{Y} is normal and (\mathcal{Y},\mathcal{B})\to\mathbb{A}^{1} is stable, (b) \mathcal{Y}_{t} has only \mu_{p} -quotient singularities (or is regular if p=2 ) for t\neq 0 , (c) \mathcal{Y}_{0} is reduced but non- S_{2} . Moreover, if C is a normal curve and if C\to\mathbb{A}^{1} is a finite flat morphism, then \mathcal{Y}_{C} is normal and (\mathcal{Y}_{C},\mathcal{B}_{C})\to C is stable. Our method produces boundary divisors \mathcal{B}\neq 0 whose coefficents are quite small, but the number of irreducible components of \mathcal{B} and the volume of K_{\mathcal{Y}_{t}}+\mathcal{B}_{t} are up to our liking: see Section 4.1 and the proof of Theorem A.0.11 . The examples given by Theorem 3 differ from Kollár’s ones [ Kol23a ] in two important ways: their dimension is constrained by the characteristic (while Kollár constructs pathological families of threefold pairs in every characteristic), but their general fibers are klt, Cohen–Macaulay and F -injective by [ Pos23 , Theorems 1 and 2] (while the general fibres in Kollár’s examples satisfy none of these properties). We discuss the implications of these examples for moduli theory in Section 1.1 below. Let me sketch the construction of these examples. We start with a 1 -foliation \mathcal{F} of rank 1 on an affine space \mathbb{A}^{n} , and we wish to understand the singularities of \mathbb{A}^{n}/\mathcal{F} . There is a simple criterion guaranteeing that \mathbb{A}^{n}/\mathcal{F} is only S_{2} (see [ Pos23 , Remark 2.5.4] ), so we only need to study its singularities in the sense of the MMP. If F is a divisor over \mathbb{A}^{n}/\mathcal{F} , then by [ Pos23 , §4.2] (see Theorem 2.2.8 ) we can compute the discrepancy a(F;\mathbb{A}^{n}/\mathcal{F}) in terms of a(E;\mathbb{A}^{n}) and a(E;\mathcal{F}) where E is the closure of F in the function field of \mathbb{A}^{n} . The problem is that we cannot estimate a(E;\mathcal{F}) easily. So we blow-up \mathbb{A}^{n} until the singularities of \mathcal{F} become log canonical. This leads to a commutative diagram {\lx@inpgf@ignorespaces X} {\lx@inpgf@ignorespaces Z=X/b^{*}\mathcal{F}} {\lx@inpgf@ignorespaces\mathbb{A}^{n}} {\lx@inpgf@ignorespaces\mathbb{A}^{n}/\mathcal{F}} \scriptstyle{\lx@inpgf@ignorespaces b} \scriptstyle{\lx@inpgf@ignorespaces\mathfrak{b}} where the horizontal arrows are the quotient morphisms, and both morphisms b and \mathfrak{b} are birational. Writing K_{Z}-\sum_{i}a_{i}F_{i}=\mathfrak{b}^{*}K_{\mathbb{A}^{n}/\mathcal{F}} where the F_{i} ’s are the \mathfrak{b} -exceptional divisors, we reduce to computing the discrepancies over the sub-pair (Z,-\sum_{i}a_{i}F_{i}) . The advantage is now that b^{*}\mathcal{F} is log canonical, so the tools given by Theorem 2.2.8 are much more efficient. The appearance of the characteristic in the dimensional lower bound comes from the adjunction formula of Section 2.2 . For Theorem 2 and Theorem 3 we use a relative version of this construction. Proving that the central fiber of the quotient is not S_{2} is more involved (as the criterion used before does not apply): we use the tools developed in [ Pos23 , §5] (see Theorem 2.2.10 ). In our examples, the 1 -foliation \mathcal{F} is generated by (some variation of) the quadratic derivation \partial=\sum_{i=1}^{n}x_{i}^{2}\partial_{x_{i}} . It satisfies \partial^{[p]}=0 in every characteristic p>0 , and thus defines an \alpha_{p} -action on \mathbb{A}^{n} such that \mathbb{A}^{n}/\alpha_{p}=\mathbb{A}^{n}/\mathcal{F} [ Pos23 , Proposition 2.3.7] . If n=2=p then \mathbb{A}^{2}/\mathcal{F} is the RDP D^{0}_{4} . The quadratic derivation has the nice property that a single blow-up simplifies its singularities into lc ones, making the necessary computations straightforward. In principle, other choices of 1 -foliations \mathcal{F} can lead to similar examples, possibly with better dimensional bounds. The difficult part is the computation of the blow-up sequence b\colon X\to\mathbb{A}^{n} needed to simplify the singularities of \mathcal{F} . The results of [ Pos24b ] suggests that such a sequence usually exists, but the number of necessary blow-ups might be quite large. To illustrate this, in Section 4.2 we consider the derivation y^{3}\partial_{x}+x\partial_{y}+t\partial_{z} on \mathbb{A}^{3}_{x,y,z}\times\mathbb{A}^{1}_{t} : this choice was motivated by the fact that the quotient of \mathbb{A}^{2} by y^{3}\partial_{x}+x\partial_{y} gives the RDP E^{0}_{6} in characteristic 3 . The quotient gives pathological families of affine threefolds in characteristics 3 and 5 : Theorem