The rank two \(p\)-curvature conjecture on generic curves (Q2039549)

Let \(X\) be a smooth connected complex variety and \(V\) a vector bundle on \(X\) equipped with an integrable connection \(\nabla\). There is a finitely generated \(\mathbb{Z}\)-algebra \(R\) such that \(X\) and \((V, \nabla)\) descend to \(R\). For all but finitely many prime numbers \(p\) one may reduce \(X\) and \((V,\nabla)\) mod \(p\) to obtain a vector bundle with connection mod \(p\) and attach to it the \(p\)-curvature operator \(\psi_p\). Its vanishing is equivalent to the existence of a full set of solutions mod \(p\). The Grothendieck-Katz \(p\)-curvature conjecture predicts that if the \(p\)-curvature of \((V,\nabla)\) vanishes for all but finitely many primes \(p\), then \((V,\nabla)\) has a full set of algebraic solutions. This is equivalent to asking that the associated local system has finite monodromy group. In this paper the authors prove the \(p\)-curvature conjecture for connections of rank \(2\) on generic curves of genus \(g \ge 0\) and with \(n \ge 0\) punctures. Here a complex curve \(C\) is called \textit{generic} if it descends to a smooth model over a finitely generated \(\mathbb{Z}\)-algebra \(R\) such that the morphism \(\mathrm{Spec}(R) \to \mathcal{M}_{g,n}\) to the moduli stack of curves is dominant. Their strategy relies on finiteness criteria for the monodromy groups of local systems on topological surfaces. Namely they prove that a semisimple local system of rank 2 on a punctured surface \(\Sigma\) of genus \(g \ge 1\) with finite monodromy along any simple loop on \(\Sigma\) has finite monodromy group, see Theorem 1.2. In the case of genus \(g=0\) the conclusion remains the same but one has to assume that the restriction to any pair of pants on \(\Sigma\) has finite monodromy, see Theorem 1.3. Let \((V,\nabla)\) be a rank \(2\) connection on a generic curve with vanishing \(p\)-curvature for almost all \(p\). To establish their main result Theorem 1.1. in the case of \(g \ge 1\) the authors make use of Theorem 1.3. in [\textit{A. N. Shankar}, Duke Math. J. 167, No. 10, 1951--1980 (2018; Zbl 1453.11085)] which guarantees finiteness of the monodromy of \((V,\nabla)\) along simple loops. For the case \(g=0\) the authors establish Theorem 1.4. which states that any rank \(2\) connection on a curve of genus \(g=0\) with \(n\) generic punctures and vanishing \(p\)-curvature at almost all primes \(p\) has finite monodromy when restricted to any pair of pants. Together with their finiteness criteria they can conclude their main result. To prove Theorem 1.4. the authors use nodal degenerations of the generic curve to a reducible curve containing \(\mathbb{P}^1 \setminus \{0,1,\infty\}\) as irreducible component. Then the connection may be extended to this irreducible component. Katz has proven the \(p\)-curvature conjecture for \(\mathbb{P}^1 \setminus \{0,1,\infty\}\), see [\textit{N. M. Katz}, Invent. Math. 18, 1--118 (1972; Zbl 0278.14004)], and in this way the finiteness of monodromy of \((V,\nabla)\) restricted to the pair of pants is deduced. The proof involves a systematic study of the vanishing of \(p\)-curvature in families and the authors highlight a few results that may be of independent interests, see Theorem 1.5., Proposition 1.6. and Theorem 1.8.