Ramification and cleanliness (Q765680): Difference between revisions

The goal of this paper is to set up a framework to understand and describe the ramification of \(\ell\)-adic sheaves in positive characteristic. A degenerate but very crucial situation was already treated by the second author in [J. Inst. Math. Jussieu 8, No. 4, 769--829 (2009; Zbl 1177.14044)]. Let \(k\) be a perfect field of characteristic \(p>0\). Let \(X\) be a proper smooth variety over \(k\) and \(D\) a simple normal crossing divisor on \(X\). Put \(U = X - D\). Let \(\mathcal F\) be a lisse \(\overline {\mathbb{Q}}_{\ell}\)-sheaf over \(U\). The goal of ramification theory is (1) to describe the (wild) ramification of \(\mathcal F\) along \(D\) (by developing a theory analogous to that of (log-)characteristic cycles for algebraic \(D\)-modules), and (2) to give a Riemann-Roch type formula for \(\mathcal F\), that is, to compute the Euler-Poincaré characteristic of \(\mathcal F\) in terms of its invariants from ramification provided by (a). For \(R\) an effective rational divisor supported on \(D\), the authors construct certain ``weighted blowups'' \((X \ast_k X)^{(R)}\) of \(X \times X\) along the divisors \(D \times X\) and \(X \times D\). This space contains \(U \times U\) as a subspace. Let \(j^{(R)}\) denote the natural inclusion. The authors study the behavior of the sheaf \(\mathcal G = j^{(R)}_*\Hom(\mathrm{pr}_1^*\mathcal F, \mathrm{pr}_2^* \mathcal F)\) along the complement of \(U \times U\) and define the notion of \(\mathcal F\) having ramification bounded by \(R+\). This agrees with the definition using the logarithmic ramification filtration for the Galois group of the completion of \(k(X)\) along each generic point of \(D\), as defined by the same authors in [Am. J. Math. 124, No. 5, 879--920 (2002; Zbl 1084.11064)] and [Doc. Math., J. DMV Extra Vol., 5--72 (2003; Zbl 1127.11349)]. The author then introduce the key notion of cleanness of \(\mathcal F\), which, roughly speaking, says that the ramification along \(D\) is completely controlled by the ramification given by the restriction of \(\mathcal F\) to the localization at the generic points of \(D\). In technical terms, this means that, after decomposing \(\mathcal F\) locally according to the ramification filtration, the support of the Fourier-Deligne transform of \(\mathcal G\) (for each pure slope direct summand) does not meet the zero section of the corresponding vector bundle introduced in the process of weighted blowups. When the cleanness condition is satisfied, the authors define certain cycles in the log-cotangent space of \(X\) using the aforementioned support; this should be viewed as analogous to the definition of (log-)charactersitic cycles for algebraic \(D\)-modules over \(\mathbb C\). The authors then conjecture that the intersection number of this cycle with the zero section of the log-cotangent space gives the Euler-Poincaré characteristic of \(\mathcal F\). In a special case, this conjecture is known by the work of the second author in [J. Inst. Math. Jussieu 8, No. 4, 769--829 (2009; Zbl 1177.14044)]. This paper, although well written, is quite technical. We recommend the reader to first familiarize himself/herself with earlier works in this area, e.g. [Am. J. Math. 124, No. 5, 879--920 (2002; Zbl 1084.11064)] and [J. Inst. Math. Jussieu 8, No. 4, 769--829 (2009; Zbl 1177.14044)], before diving into the technical constructions.