Group rings of graded rings. Applications (Q795134)

A ring R is graded by a group G if \(R=\oplus R(x)\) is a direct sum of the additive subgroups R(x) indexed by the elements \(x\in G\) and if \(R(x)R(y)\subseteq R(xy).\) For example, R could be a group ring S[G] or more generally a crossed product S*G. In studying R, one wants to relate its graded structure to its ordinary structure and of course to do this as cheaply as possible. Since group rings and crossed products have already been extensively studied, the hope might be to obtain such results directly from known results in these special cases. This was achieved quite dramatically by \textit{M. Cohen} and \textit{S. Montgomery} [Trans. Am. Math. Soc. 282, 237-258 (1984; Zbl 0533.16001)] using a duality machine and their idea was further extended by \textit{Quinn} [''Group-graded rings and duality'', ibid. (to appear)] to handle infinite groups. Using a different construction, \textit{M. van den Bergh} [''On a theorem of Cohen and Montgomery'', Proc. Am. Math. Soc. (to appear)] and later this paper offered another theorem proving technique. Since the applications here are for the most part not new, it seems more appropriate to just briefly discuss the methods. For this we use van den Bergh's point of view which is a little clearer. Given the G-graded ring R, form the ordinary group ring R[G] and then embed R into this ring via the map \(\sim:r=\sum r(x)\to \tilde r=\sum r(x)x.\) Then \(\tilde R\) is isomorphic to R and is embedded as a graded subring of R[G]. Of course R is also contained naturally in R[G] as the identity component. The idea is to play off these two different embeddings of R. A special case here is particularly informative. Thus suppose \(R=S[G]\) is an ordinary group ring. Then \(R[G]=S[G\times G]\) and \(\tilde R=S[D]\subseteq S[G\times G]\) where D is the diagonal subgroup of \(G\times G\). Note that if G is nonabelian, then D is not normal in G. Because of this, the relationship between S[D] and S[\(G\times G]\) is not very well understood. On the other hand, the graded structure of S[D] is in some sense determined by S so the graded structures of S[D] and S[\(G\times G]\) are more tightly related. Reviewer's remarks: Let \(M_ G(R)\) denote the ring of \(| G| \times | G|\) matrices over R. If we follow the embedding \(\sim:R\to R[G]\) by the natural inclusion \(R[G]\to M_ G(R)\) then we actually obtain the Cohen-Montgomery-Quinn embedding of R into \(M_ G(R)\). So in some sense the inclusion used here is not really new. Furthermore I believe that the matrix embedding which contains within it certain smash products and crossed products has much more usable structure. For example, I see no way to use the construction here to prove the result that for G finite, if \(R=R(G)\) is semiprime, then so is R(H) for any subgroup H of G. Similarly the incomparability result Corollary 6.6 is already known to be true more generally without torsion assumptions, but this more general result most likely cannot be proved by these methods. In addition Theorems 3.1 and 3.2 which are only proved here for strong polycyclic-by-finite groups (essentially supersolvable) are easily proved for all polycyclic-by-finite groups from the methods in Quinn's paper. Still it is nice to have new techniques and the Van den Bergh-Năstăsescu embedding, when it does apply, can sometimes yield shorter proofs.