Two dimensional neighborhoods of elliptic curves: analytic classification in the torsion case (Q2104862)

The problem of classifying neighborhoods of curves in complex surfaces has a long story. Ran the year 1962 when \textit{H. Grauert} [Math. Ann. 146, 331--368 (1962; Zbl 0173.33004)] and \textit{K. Kodaira} [Ann. Math. (2) 75, 146--162 (1962; Zbl 0112.38404)] proved two fundamental results. Let us state them for elliptic curves since the article under reviewed is only concerned by this case. If \(C\subset S\) is an embedding of \(C\) in a complex surface \(S\), \(N_C\) shall denote the normal bundle of \(C\) in \(S\). \begin{itemize} \item[Grauert:] if \(\mathrm{deg}(N_C)<0\), \(C\subset S\) is linearizable (i.e., is isomorphic to a neighborhood of the zero section in \(N_C\)). \item[Kodaira:] if \(\mathrm{deg}(N_C)>0\), \(C\) moves in \(S\) and there is a classification of such neighborhood. \end{itemize} The remaining case \(\mathrm{deg}(N_C)=0\) happens exactly when \(N_C\) is topologically torsion and it was first studied by \textit{V. I. Arnol'd} in the celebrated article [Funct. Anal. Appl. 10, 249--259 (1977; Zbl 0346.58003); translation from Funkts. Anal. Prilozh. 10, No. 4, 1--12 (1976)]. He established that the neighborhood \((S,C)\) is always formally linearizable if \(N_C\) is not torsion and that it is even analytically linearizable if \(N_C\) is \textit{generic} in \(\mathrm{Pic}^0(C)\) (i.e., \(N_C\) is required to satify a Diophantine condition). It was then further investigated by \textit{T. Ueda} [J. Math. Kyoto Univ. 22, 583--607 (1983; Zbl 0519.32019)] where a formal invariant was introduced: the Ueda type is a positive integer that measures to what extent the flat structure on \(N_C\) can be extended on infinitesimal neigborhoods of \(C\) in \(S\). Using this, the authors [Mosc. Math. J. 19, No. 2, 357--392 (2019; Zbl 1460.32015)] were able to give a formal classification of neighborhoods of ellitptic curves. In the case \(N_C=\mathcal{O}_C\) and \(\mathrm{Uedatype}(S,C)=1\) (the only one to be considered here), the formal moduli space is \(\mathbb{C}^2\) and, if \(C\) is written as \(\mathbb{C}^*_z/(z\sim qz)\) (with \(\vert q\vert<1\)) any germ of embedding \((S,C)\) is formally equivalent to a quotient \((\mathbb{C}^*_z\times \mathbb{C}_y,(y=0))/\langle F_{1,\nu,\lambda}\rangle\) where \(F_{1,\nu,\lambda}\) is an explicit germ of a diffeomorphism. Let us now describe the point corresponding to \((0,0)\). An elliptic curve \(C\) has one (essentially unique) unsplit vector bundle \(V\) of rank 2 that is an extension of \(\mathcal{O}_C\) by itself and let us denote by \(S_0:=\mathbb{P}(V)\to C\) the projective bundle associated with \(V\). It has a unique section \(C\) with \(C^2=0\) and, in the variable \((z,\xi=1/y)\), \(S_0\) is the quotient \(\mathbb{C}^*_z\times \mathbb{P}^1_\xi/\langle F_{1,0,0}\rangle\) with \(F_{1,0,0}(z,\xi)=(qz,\xi-1)\) (and \(C\) is the section \(\xi=\infty\)). It is also known that \(S_0\setminus C\simeq \mathbb{C}^*\times \mathbb{C}^*\) (analytically but not algebraically!) and \(S_0\setminus C\) can then be compactified in another way: add a cycle of four projective lines \(L_1,\ldots,\,L_4\) to obtain \(\mathbb{P}^1\times \mathbb{P}^1\). An analytic neighborhood of \(C\) being formally equivalent to (the germe of) \((S_0,C)\) can thus be seen as formal neighborhood of \(L_1\cup\cdots\cup L_4\). At that point, a kind of cocyle when comparing the change of coordinates when passing from \(L_i\) to \(L_{i+1}\) and using sectorial normalization. The classification of such analytic neighborhoods can then be stated in terms of this cocycle (\emph{cf.} Theorem~C). Let us know describe a bit the content of this article. First, it is important to note that the authors deal almost only with the case considered above: \(N_C=\mathcal{O}_C\) and \(\mathrm{Uedatype}(S,C)=1\). Section~2 describes models for anaytic neighborhoods of \(C\) that are formally isomorphic to \((S_0,C)\) and sectorial decomposition is introduced in Section~3. Section~4 gives an overview of the classification: each model gives rise to a sectorial decomposition and there is a natural equivalence relation (on the induced cocycles) that gives back the analytic equvalence of neighborhoods. Sections~5 and 6 contain the proof of the classification but the one of Lemma~4.1 (to be given in Section~9). Foliations and symmetries of these neighborhoods are studied in Sections~7 and 8. The technical part of the article is Section~9 where the sectorial normalization is proven: it requires in particular to solve a functional equation and it involves some pages of hard analysis. The last two sections explain how to formulate and extend the result for higher Ueda type (Section~10) and higher torsion order (Section~11). The analytic classifcation is thus achieved in the torsion case: Theorem~E in Section~10 describes explicitely the moduli space of analytic neighborhoods of elliptic curves being formally equivalent to a fixed one.