On the \(p\)-torsion of the Tate-Shafarevich group of abelian varieties over higher dimensional bases over finite fields (Q2086417)

Earlier work of the author gives some relative boundedness for the \(\ell\)-primary part of \(\Sha(\mathscr{A}/X)=H^1_{\text{ét}}(X,\mathscr{A})\) of an abelian scheme \(\mathscr{A}\) of good reduction over a proper base scheme \(X/k\) of dimension greater than \(1\). The case of greatest interest here is when \(k\) is a finite field of characteristic \(p\), and the results are extended to the \(p\)-primary part. The main results are restatements of results from [the author, Manuscr. Math. 150, No. 1--2, 211--245 (2016; Zbl 1342.19005)] with \(\ell\) replaced by \(p\) (and a slight strengthening in one case: \(\Sha(\mathscr{A}/X)[p^\infty]\) is cofinitely generated; isogeny preserves finiteness of \(\Sha(\mathscr{A}/X)[p^\infty]\); under some conditions, alterations of \(X\) also preserve finiteness of \(\Sha(\mathscr{A}/X)[\ell^\infty]\) as long as the degree is coprime to \(p\) or \(\ell\neq p\). (The improvement from the earlier result for \(\ell\neq p\) is that the degree is not then required to be coprime to \(\ell\).) \par These all follow from a technical result that may be useful in other contexts: if \(X\) is a proper integral normal variety over a finite field and \(G\) is a finite flat commutative group scheme over \(X\) then \(H^1_{\mathrm{fppf}}(X,G)\) is finite. This is proved by reducing to the case where \(G\) is simple, and counting \(G\)-torsors. Then the main results follow by arguments similar but not always identical to those used for the \(\ell\)-primary part in [loc. cit.].