Abelian fiber spaces and their Tate-Shafarevich twists with sections
János Kollár · 2025
Read Abelian fiber spaces and their Tate-Shafarevich twists with sections by János Kollár (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-2504-21705.txt.
Introduction
Janos Kollar, a solo author, works on Abelian fiber spaces and their Tate-Shafarevich twists, in a 2025 mathematics paper. Starting from an Abelian fiber space, the aim is to construct a Tate-Shafarevich twist of it that admits a rational section. Twists are variant forms of a space that can differ in whether they carry such a section, a basic structural feature. The paper carries out that construction.
Catalog record
| CREATOR | János Kollár |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 66 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2504-21705
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Abelian fiber spaces and their Tate-Shafarevich twists with sections János Kollár Abstract . Starting with an Abelian fiber space, the aim is to construct a Tate-Shafarevich twist that has a rational section. Kodaira showed that to any elliptic surface S\to C without multiple fibers one can associate its relative Jacobian surface \operatorname{Jac}(S/C)\to C , which has the same local structure, and also a section. Generalizations to elliptic n -folds and to hyperkähler manifolds are studied in [ Nak91 , DG94 , Gro94 , Gro97 , Nak02b , Nak02a , DCS21 , CDCH+21 , BDCS24 , Sac25 , BKV25 , Kim24 ] ; see ( 14 ). This note considers Abelian fiber spaces in general. Definition 1 (Abelian torsors and fiber spaces) . An Abelian variety over a field k is a smooth, proper, connected algebraic group A_{k} over k . An Abelian torsor X_{k} is a principal homogenous space under an Abelian variety, which can be given by A_{k}=\operatorname{\mathbf{Aut}}^{\circ}(X_{k})\cong\operatorname{\mathbf{Pic}}^{\circ}\bigl(\operatorname{\mathbf{Pic}}^{\circ}(X_{k})\bigr) . A fiber space is a proper morphism \pi:X\to Z between normal, irreducible, algebraic spaces, such that \pi_{*}{\mathcal{O}}_{X}={\mathcal{O}}_{Z} . It is an Abelian fiber space if general fibers are Abelian torsors. Definition 2 (Twisted forms) . A twisted form of a morphism \pi:X\to Z is obtained by the following procedure. Take an open cover \amalg_{i}U_{i}\to Z with U_{ij}:=U_{i}\times_{Z}U_{j} , and set X_{i}:=X\times_{Z}U_{i} and X_{ij}:=X\times_{Z}U_{ij} . Choose isomorphisms (of U_{ij} -schemes) [formula] \phi_{ij}:X_{ij}\cong X_{ji},\quad\mbox{such that}\quad\phi_{jk}\circ\phi_{ij}=\phi_{ik}. Use these to patch the X_{i} together to get a morphism \pi^{\Phi}:X^{\Phi}\to Z . Over {\mathbb{C}} we use the Euclidean topology, in general the étale topology. These are also called étale or Tate-Shafarevich twists. A twisted form of an Abelian fiber space is also an Abelian fiber space. Our aim is to prove that, under some natural conditions, Abelian fiber spaces have a twist with a rational section. For clarity, we state first a weaker form over {\mathbb{C}} . After further definitions, the final version over {\mathbb{C}} is given in Theorem 8 . The general result ( 40 ) over arbitrary base schemes is very similar, but needs more definitions; see ( 38 ) and ( 39 ). Assumption 3 . Let \pi:X\to Z be an Abelian fiber space over {\mathbb{C}} , such that X has log terminal singularities, K_{X} is numerically \pi -trivial, \pi is locally projective (in the Euclidean or étale topology) and of pure relative dimension d . Note that in characteristic 0, any Abelian fiber space is birational to some \pi:X\to Z such that \pi is projective, X has terminal singularities, and K_{X} is numerically \pi -trivial; see ( 38 .3 ′ ) for details. So these 3 are mild assumptions. (Note that terminal implies log terminal.) Having pure relative dimension holds if X is hyperkähler or if \dim Z\leq 2 , but not in general. Theorem 4 (Untwisting of Abelian fiber spaces) . Let \pi:X\to Z be an Abelian fiber space satisfying Assumption 3 . Let \eta\in H^{2d}(X,{\mathbb{Z}}) be a cohomology class such that \eta\cap[\mbox{general fiber}]=1 . Then there is a canonical construction yielding a twisted form [formula] \pi_{\eta}:\operatorname{\mathbb{T}}_{\eta}(X/Z)\to Z\quad\mbox{with a rational section}\quad s_{\eta}:Z\dasharrow\operatorname{\mathbb{T}}_{\eta}(X/Z). The construction commutes with smooth base changes. For fixed X,Z , the different \operatorname{\mathbb{T}}_{\eta}(X/Z) are birational to each other, but for \dim Z\geq 2 one can get infinitely many different ones up to isomorphism, see ( 18 ). The existence of such \eta is the higher dimensional version of having no multiple fibers. It is an almost necessary condition, see Comment 9 for a precise statement. 5 Plan of the proof. We construct \operatorname{\mathbb{T}}_{\eta}(X/Z) locally on Z , and patch the pieces together. We start with the latter. ( 5 .1) Patching. Assume that Z has an open cover Z=\cup_{i}U_{i} , and over each U_{i} we have a rational section s_{i}:U_{i}\dasharrow X_{i}:=\pi^{-1}U_{i} . As in ( 2 .1), the patching is given by isomorphisms \phi_{ij}:X_{ij}\cong X_{ji} such that s_{j}=\phi_{ij}\circ s_{i} on U_{ij} . If the \phi_{ij} are unique, then the cocyle conditions \phi_{jk}\circ\phi_{ij}=\phi_{ik} are automatic. How can we achieve this? Let z\in U_{i}\cap U_{j} be a point such that the fiber X_{z} is smooth. The \phi_{ij} that we aim to construct should restrict to an isomorphism \phi_{ij}|_{z}:\bigl(X_{z},s_{i}(z)\bigr)\cong\bigl(X_{z},s_{j}(z)\bigr) . Since X_{z} is homogeneous, such an isomorphism exists, but it is not unique. However, there is a unique such translation-isomorphism, namely translation by s_{j}(z)-s_{i}(z) . Thus the cocyle conditions are automatic if we use only translation-isomorphisms for the \phi_{ij} . ( 5 .2) Translation-isomorphisms. The notion of translation is more problematic for singular fibers, but in Definitions 6 and 24 we establish the group of translation-isomorphisms \operatorname{Aut}^{\circ}(X/Z) . Sheafifying in the étale topology gives {\mathchanc{Aut}}^{\circ}(X/Z) . ( 5 .3) Changing \eta . These consideration also suggest that the right place for \eta to live is not in H^{2d}(X,{\mathbb{Z}}) , but in H^{0}(Z,R^{2d}\pi_{*}{\mathbb{Z}}_{X}) . Restriction gives a natural map H^{2d}(X,{\mathbb{Z}})\to H^{0}(Z,R^{2d}\pi_{*}{\mathbb{Z}}_{X}). This is an isomorphism if Z is a germ, but otherwise the target can be much larger, so working with H^{0}(Z,R^{2d}\pi_{*}{\mathbb{Z}}_{X}) gives a richer theory. See also Comment 9 . The local problem now becomes the following. ( 5 .4) Local uniqueness. For a germ z\in U\subset Z and \eta\in H^{2d}(X_{U},{\mathbb{Z}}) , we want a canonical way to specify a set of rational sections \operatorname{RatSec}(\eta)=\{s:U\dasharrow X_{U}\} that forms a single \operatorname{Aut}^{\circ}(X_{U}/U) -orbit. We discuss this in more detail in Paragraph 10 . Once this is done, the local patches automatically glue to give the global \operatorname{\mathbb{T}}_{\eta}(X/Z) . ( 5 .5) Arbitrary bases. We replace the analytic germs with the strict Henselizations of the local rings, but otherwise the plan is the same. Definition 6 (Automorphisms) . For a morphism \pi:X\to Z , we let \operatorname{Aut}(X/Z) denote the group of automorphism of X that satisfy \pi=\pi\circ\phi . By base change, it defines a functor V\mapsto\operatorname{Aut}(X\times_{Z}V/V) . If this functor is represented by a group scheme (or algebraic space), we denote it by \operatorname{\mathbf{Aut}}(X/Z) or \operatorname{\mathbf{Aut}}(X/Z)\to Z . By a standard Hilbert scheme argument, representability holds if \pi is proper and flat [ Kol96 , I.1.10] . If \operatorname{\mathbf{Aut}}(X/Z)\to Z is smooth, let \operatorname{\mathbf{Aut}}^{\circ}(X/Z)\subset\operatorname{\mathbf{Aut}}(X/Z) denote the identity component. We would like \operatorname{Aut}^{\circ}(X/Z) to be the global sections of \operatorname{\mathbf{Aut}}^{\circ}(X/Z) . Unfortunately, \operatorname{\mathbf{Aut}}(X/Z) does not exists in general, and even if it does, \operatorname{\mathbf{Aut}}(X/Z) is frequently not even flat over Z . Nonetheless, if \pi:X\to Z is an Abelian fiber space, in ( 24 ) we define \operatorname{Aut}^{\circ}(X/Z) and the corresponding sheaf {\mathchanc{Aut}}^{\circ}(X/Z) . Note on notation. The papers [ AR23 , Kim24 ] define \operatorname{Aut}^{\circ}(X/Z) for hyperkähler manifolds. The various definitions most likely agree when they all apply. Definition 7 (Translation-automorphisms and twists) . Let \pi:X\to Z be an Abelian fiber space. Since \operatorname{Aut}^{\circ}(X/Z) acts by translations on general fibers, we call it the group of translation-automorphisms. It is a commutative group. A twisted form of \pi is translation-twisted if the patching maps \phi_{ij} in ( 2 .1) are translation-automorphisms. The stronger version of Theorem 4 is the following. Theorem 8 . Let \pi:X\to Z be an Abelian fiber space satisfying Assumption 3 . Let \eta\in H^{0}(Z,R^{2d}\pi_{*}{\mathbb{Z}}_{X}) be a cohomology class such that \eta\cap[\mbox{general fiber}]=1 . Then \pi_{\eta}:\operatorname{\mathbb{T}}_{\eta}(X/Z)\to Z is a translation-twisted form of X , with a rational section s_{\eta}:Z\dasharrow\operatorname{\mathbb{T}}_{\eta}(X/Z) . Comment 9 . If X is smooth, then the existence