Common splitting fields of symbol algebras (Q6155372)

Let \(F\) be a field of nonzero characteristic \(p\), and let Br\((F)\) be the Brauer group of \(F\). Suppose that the \(p\)-component Br\((F)_p\) of Br\((F)\) is nontrivial, and for each \(m \in \mathbb{N}\), denote by \(_{p^{m}}Br(F)\) the maximal subgroup of Br\((F)\) of period \(p^m\). It is known (due to Teichmüller and Albert, see Chapter VII of: [\textit{A. A. Albert}, Structure of algebras. Providence, RI: American Mathematical Society (AMS) (1939; Zbl 0023.19901)], or Chapter 9 of: [\textit{P. Gille} and \textit{T. Szamuely}, Central simple algebras and Galois cohomology. 2nd revised and updated edition. Cambridge: Cambridge University Press (2017; Zbl 1373.19001)]) that the group \(_{p^{m}}Br(F)\) is generated by the Brauer equivalence classes of symbol (cyclic) \(F\)-algebras of degree \(p^m\). Also, Albert has proved that any central simple \(F\)-algebra of degree \(p^m\) is Brauer equivalent to a symbol algebra of degree \(p^t\), for some \(t \ge m\). This has been done by finding a common degree \(p^{m+t}\) simple purely inseparable splitting field, for any two symbol algebras of degrees \(p^m\) and \(p^t\), respectively, which implies their tensor product is a symbol algebra of degree \(p^{m+t}\). The paper under review generalizes the latter result by finding a common splitting field of degree \(p^{(m_{0}+ \dots +m_{s})}\), for any finite number of symbol \(F\)-algebras of degrees \(p^{m_{0}}, \dots , p^{m_{s}}\), respectively. The authors deduce this from one of the main results of the paper, which states that any finite number of symbol algebras of the same degree sharing a simple purely inseparable maximal subfield also share a cyclic maximal subfield. The stated result has been proved in an earlier paper by the first-named author, for a pair of symbol algebras of degree \(p\) (see [\textit{A. Chapman}, Bull. Belg. Math. Soc. - Simon Stevin 22, No. 4, 683--688 (2015; Zbl 1367.16015)]). The rest of the paper under review considers the problem of finding the symbol length in \(H_{p^{m}}^{n+1}(F)\) of a class \(A\) in \(H_{p^{m}}^{n+1}(F)\) which has a prescribed symbol length \(r \ge 1\) when embedded in \(H_{p^{t}}^{n+1}(F)\), for some integer \(t \ge m\). It provides an explicit upper bound of this symbol length, for any class of this kind, in each of the following cases: (i) \(n = 2\), \(t = m + 1\), \(r = 1\); (ii) \(p = 2\), \(t = m + 1\), \(r = 1\); (iii) \(n = 1\); (iv) \(n = 1\), \(r = 1\); (v) \(p = 2\), \(n = 1\), \(r \le 2\), \(t = m + 1\); (vi) \(p = 3\), \(n = 1\), \(r \le 2\), \(t = m + 1\). In view of the isomorphisms \(H_{p^{m}}^2(F) \cong_{p^{m}}Br(F)\), for all \(m \in \mathbb{N}\) (see [\textit{O. Izhboldin}, Geom. Topol. Monogr. 3, 19--29 (2000; Zbl 1008.11052)]), they specify the above-noted theorems of Teichmüller and Albert. Analogous results to those in cases (v) and (vi) are obtained, under the assumptions that \(E\) is a field with char\((E)\) different from \(2\) and \(3\), and \(E\) contains a primitive \(p^{m+1}\)-th root of unity, for \(p \in \{2, 3\}\). The concluding results of this paper are proved applying one of Tignol's theorems and a method developed by Matzri (see [\textit{J. P. Tignol}, Proc. Am. Math. Soc. 89, 587--588 (1983; Zbl 0558.16008)], and [\textit{E. Matzri}, J. Algebra 418, 1--7 (2014; Zbl 1312.16014)]).