\(\mathbb Z_3 \times\mathbb Z_3\) crossed products. (Q406343): Difference between revisions

Let \(F\) be a field containing the roots of unity of order \(e\). Merkurjev and Suslin proved that any central simple algebra \(A\) of exponent \(e\) over \(F\) is similar, in the Brauer group, to a tensor product of cyclic algebras of degree \(e\). However, the ``symbol length'' of \(A\), namely the number of cyclic algebras in such a presentation, is not known. The author has recently established an upper bound depending on the essential dimension, [see \textit{E. Matzri}, ``Symbol length in the Brauer group of a field'', Trans. Am. Math. Soc. (to appear; \url{arxiv:1402.0332})]. The goal of this paper is to bound the symbol length of algebras of degree \(9\). Recall that an algebra is a crossed product if it has a maximal subfield Galois over \(F\). The main results are as follows. (1) The symbol length of a crossed product of degree \(9\) is at most \(4\). This is based on an explicit presentation of the algebra as a product of \(3\) cyclic algebras after a quadratic extension, and taking the corestriction down to the base field. Incidentally this result proves that the essential \(3\)-dimension for division algebras of degree \(9\) is \(\text{edc}_3(A)\leq 6\). (2) The symbol length of a crossed product of degree \(9\) and exponent \(3\) is at most \(31\). The bound has gone up, since now only cyclic algebras of degree \(3\) are allowed. The proof here uses the above presentation, a slot lemma for algebras of degree \(3\), and a theorem of Tignol that the symbol length of cyclic algebras of degree \(9\) and exponent \(3\) is at most \(3\). (3) The symbol length for any algebra of degree \(9\) is at most \(8m\), where \(m=4480\) is the index of the \(3\)-Sylow subgroup of \(S_9\). And in exponent \(3\), the symbol length is at most \(62m=277760\). This is proved by a prime-to-\(3\) scalar extension to obtain a crossed product, and taking the corestriction back to \(F\).