
2 A. DI LORENZO AND G. INCHIOSTRO
(namely, the stack parametrizing line bundles with section). In our example, we argue that one can assume
the chain of destabilizing stacky P1’s to have length at most 1, to map to the closed point of [A1/Gm], and
we explicitly describe the line bundle on the family of curves. We expect a similarly accessible description
to be available in general.
Where are we headed. As it is, our definition of twisted maps does not lead to an algebraic stack,
essentially because we are imposing no numerical stability conditions. For instance, when the target stack
is BGm, we can ask for the line bundles to satisfy Caporaso’s basic inequality [Cap94, 0.3], a numerical
condition involving the degrees of the line bundles on the irreducible components of the curve. This is the
key ingredient for making the moduli problem an algebraic stack with a projective good moduli space.
However, our definition of twisted maps can be regarded as a first step towards the construction of a stack
of twisted stable maps to a quotient stack having a projective good moduli space. The next step in our
project is to tackle the case M= [X/Gn
m], leveraging our understanding of the behaviour of degenerations
of twisted maps. In particular, we aim to show that, after imposing some numerical conditions, we actually
get an algebraic stack which is Θ-reductive and has unpunctured inertia: applying the results of Alper,
Halpern-Leistner and Heinloth [AHLH18, Theorem A], we would get a good moduli space of twisted stable
maps to [X/Gn
m].
Another interesting direction to study is when M=BGfor a linear algebraic group G; in this case
our functor of twisted maps can be regarded as the functor of G-torsors over families of curves, and our
main theorem shows how to extend these families of G-torsors over curves. Hopefully, after imposing some
numerical stability conditions, one can extract a proper algebraic stack out of our functor; we are currently
investigating this path too.
Organization of the paper. In Section 2we introduce the main object of interest, namely twisted maps to
an algebraic stack, and we present some basic properties of them. For example, we show that our definition
recovers the one of Abramovich and Vistoli if Mis Deligne-Mumford. In Section 3we show how to extend
maps to codimension one points of the special fiber of the family of curves (Theorem 3.9). In Section 4we
introduce the main tools to extend maps to codimension two points: we prove two technical propositions
(Proposition 4.1 and Proposition 4.5) concerning extensions of line bundles over stacky surfaces, that will
be essential for proving the main results of this section. Section 5is divided into two parts. In the first part,
we prove our main result (Theorem 1.1), and in the second part (namely Section 5.1) we prove some results
on the birational geometry of stacky surfaces, which are of independent interest: we study when one can
contract certain (stacky) P1s on a surface. Our results can be understood as a generalization of the results
of Artin in [Art62] to stacky surfaces. Finally in Section 6we run a specific example: the case of twisted
maps to M= [A1/Gm].
Conventions. In what follows, unless otherwise stated, every stack and morphism is assumed to be essen-
tially of finite type over a field kof characteristic zero. If Cis a one dimensional Deligne-Mumford stack
with coarse moduli space C → C, and Lis a line bundle on C, there is a minimal natural number nsuch
that L⊗ndescends to a line bundle Lon C. The degree of Lwill be deg(L)
n.
Acknowledgements. We are thankful to Dan Abramovich, Jarod Alper, Dori Bejleri, Samir Canning,
Andres Fernandez Herrero, Daniel Halpern-Leistner, Ming Hao Quek, Mauro Porta, Zinovy Reichstein,
Minseon Shin and Angelo Vistoli for helpful conversations.