Models of curves and wild ramification (Q2268327)

Let \(K\) be a complete discrete valuation field with ring of integers \(\mathcal{O}_K\) and algebraically closed residue field \(k\), of characteristic \(p\geq 0\). Let \(X/K\) be a smooth proper geometrically connected curve of genus \(g\geq 1\), and let \(\mathcal{X}/\mathcal{O}_K\) be its minimal regular model. When \(g\geq 2\), or \(g=1\) and \(X(K)\neq\emptyset\), there exists a finite Galois extension \(L/K\) minimal with the property that \(X_L/L\) has semi-stable reduction. Let \(\mathcal{X'}/\mathcal{O}_L\) be the minimal regular model of \(X_L/L\). In this paper, it is shown that certain properties of the special fiber \(\mathcal{X}'\) and of the extension \(L/K\) can be inferred from the knowledge of the combinatorial properties of the special fiber of \(\mathcal{X}\). For instance, if \(p=2\) and \(X/K\) is an elliptic curve with potentially multiplicative reduction, then \([L: K]=2\) and \(X/K\) has reduction of type \(I^*_{\nu+4s}\), where \(\nu>0\) is such that \(X_L/L\) has reduction of type \(I_{2\nu}\) and \(s+1\) is the valuation of the different of the extension \(L/K\). Or, in the same vein, if \(g+1<p\leq 2g+1\), \([L: K]=p^r\), and the reduction of the Jacobian \(A/K\) of \(X/K\) is purely additive, then \(A/K\) has potentially good reduction. When \(p=g+1\), the reduction is either potentially good or potentially purely multiplicative; in the latter case, the paper provides a precise description, similar to the case \(g=1\), for the possible types of reduction of a curve \(X/K\) such that \(p\mid [L: K]\) and \(\text{Jac}(X_L)\) has purely multiplicative reduction. These kind of relationships are then considered in a more general setting. Let us call a model \(\mathcal{X}/\mathcal{O}_K\) \textit{good} if the special fiber has smooth irreducible components \(C_i\) of genus \(g(C_i)\), with multiplicities: \(\mathcal{X}_k=\sum_{i=1}^vr_iC_i\), satisfying \(\gcd(r_1,\dots,r_v)=1\), and such that the reduced special fiber \((\mathcal{X}_k)^{red}\) has normal crossings. To such a model, a graph \(G(\mathcal{X})\) is attached, with vertices given by the curves \(C_i\) and any two vertices \(C_i,C_h\) being linked by \((C_i\cdot C_h)\) edges. Then, the author asks if the conditions: (1) \(a_K=\sum_{i=1}^vg(C_i)=0\), (2) \(G(\mathcal{X})\) is a tree, and (3) all chains of \(G(\mathcal{X})\) have weight equal to \(1\), would be sufficient to ensure that \(\text{Jac}(X)/K\) has potentially good reduction. Here, \(a_K\) is the abelian rank of \(\text{Jac}(X)/K\), so that \(a_K=0\) means that the special fiber of the Néron model of the Jacobian is a product of a torus by an unipotent group. Some of the above mentioned theorems give evidence that this question may have a positive answer. The paper contains some more results in this direction.