On the compositum of wildly ramified extensions (Q2439313)

This paper studies Abhyankar's inertia conjecture, which is the following : Let \(k\) be an algebraic closed field of characteristic \(p>0\), \(G\) a quasi-\(p\) group and \(I\) a subgroup of \(G\) such that \(I=P\rtimes {\mathbb Z}/n{\mathbb Z}\) with \(P\) a \(p\)-group whose conjugates generate \(G\) and \((n,p)=1\). Then there exists a \(G\)-Galois cover \(X\rightarrow {\mathbb P}_k^1\) branched only at \(\infty\) such that the inertia group at a point of \(X\) lying above \(\infty\) is \(I\). When the inertia conjecture holds for \((G,I)\), the pair \((G,I)\) is called realizable. This conjecture is still open, and the author gives several results concerning this by computing the ramification filtration on the compositum of two wildly ramified extensions of local fields and using the existence of Harbater-Katz-Gabber covers and author's previous results. In Section 2, the ramification theory for Galois extensions of local fields and Artin-Schreier-Witt theory are recalled, and several lemmas needed in the subsequent sections are provided. In Section 3, ramification filtrations of the Galois groups of Galois extensions of local fields of characteristic \(p>0\) with perfect residue field are considered. Especially, upper jumps and lower jumps for the compositum of two Galois extensions, including those of two \(p\)-cyclic and two \(p^2\)-cyclic extensions, are computed. In the final section, several results related to Abhyankar's inertia conjecture including the following are obtained : {\parindent = 0.5cm \begin{itemize}\item[(1)] For a quasi-\(p\) group \(G\), if the inertia conjecture holds for every \(p\)-subgroup of \(G\) and there is no epimorphism from \(G\) to \({\mathbb Z}/p{\mathbb Z}\), then the conjecture is true for every \(p\)-subgroup of \(G\times P\) for any \(p\)-group \(P\)~(Corollary 4.6). \item[(2)] If \((G,P)\) is realizable for a \(p\)-subgroup \(P\) of \(G\), \(Q\) is an index \(p\)-subgroup of \(P\times {\mathbb Z}/p{\mathbb Z}\) such that the projection of \(Q\) on \(P\) is surjective and conjugates of \(Q\) in \(G\times {\mathbb Z}/p{\mathbb Z}\) generate \(G\times {\mathbb Z}/p{\mathbb Z}\), then \((G\times {\mathbb Z}/p{\mathbb Z},Q)\) is realizable. Moreover, the lower jumps of \(Q\) are explicitly given by the upper jumps of \(P\)~(Theorem 4.8). So if the inertia conjecture holds for every \(p\)-subgroup of \(G\), then it does for every \(p\)-subgroup of \(G\times ({\mathbb Z}/p{\mathbb Z})^n\) for any \(n\geq 0\)~(Corollary 4.9). \item[(3)] If there is no epimorphism from \(G\) to \({\mathbb Z}/p{\mathbb Z}\) and \((G,P)\) is realizable for a \(p\)-subgroup \(P\) of \(G\) such that there is an epimorphism from \(P\) to \({\mathbb Z}/p^r{\mathbb Z}\), then \((G\times {\mathbb Z}/p^r{\mathbb Z},Q)\) is realizable, where \(Q\) is an index \(p\) subgroup of \(P\times_{{\mathbb Z}/p^{r-1}{\mathbb Z}}{\mathbb Z}/p^r{\mathbb Z}\) with two projections from \(Q\) on \(P\) and \({\mathbb Z}/p^r{\mathbb Z}\) surjective. The lower jumps of \(Q\) are also explicitly given by the upper jumps of \(P\)~(Theorem 4.10). \end{itemize}}