Artinian and Noetherian serial rings (Q1608126)

The aim of the paper is to give, from a ``uniform point of view'', new proofs for well-known theorems and some new results on Artinian and on Noetherian serial (= generalized uniserial) rings. Concerning the Noetherian case a short proof for Kirichenko's theorem is given, that a Noetherian, non-Artinian indecomposable serial ring \(R\) is a factor ring of a matrix ring \(H_n(D)\), where \(D\) is a Noetherian uniserial domain [\textit{V. V. Kirichenko}, Mat. Sb., N. Ser. 99(141), 559-581 (1976; Zbl 0328.16022)]. As to the Artinian case, the demanding aim is: (1) The reduction of serial rings to serial quasi-Frobenius rings. (2) The explicit construction of ``almost all'' indecomposable serial quasi-Frobenius rings by means of (local) uniserial rings. Here, the term ``almost all'' is misleading. Behind it is the exclusion of all weakly symmetric serial rings; namely, if \(n\) is the number of indecomposable projectives \(P_1,\dots,P_n\) and \(d=d(R)\) the Loewy-length, then \(d\equiv 1(n)\), i.e. \(P_i/\text{rad }P_i\cong\text{soc}(P_i)\) is excluded. Moreover, the number of different rings in case \(d(R)=\kappa n+1\) can be much larger than, for example, in case \(d(R)=\kappa n\) or \(\kappa n+2\). The essential error concerning (2), however, is that the group-ring-construction (5.4) does not meet (2), as claimed; in particular, the Corollary 5.6 is wrong, too. A simple counterexample is the following: Let \(K\) be a field, \(q\neq\text{id}\in\text{aut }K\) and \(A=Ke_1\oplus Ke_2\oplus Ku_{21}\oplus Ku_{12}\) a 4-dimensional \(K\)-vector space with a multiplication \[ \begin{alignedat}{3} e_ie_j&=\delta_{ij}e_i, &e_ju_{ji}&=u_{ji}= u_{ji}e_i, &u_{ji}u_{\kappa\ell}&=0,\\ e_i\alpha &=\alpha e_i, &u_{21}\alpha &=\alpha u_{21}, &u_{12}\alpha &=g(\alpha)u_{12}\text{ for }\alpha\in K.\end{alignedat} \] Then \(A\) is a serial quasi-Frobenius ring with \(n=2\) and \(d(A)=2\not\equiv 1(2)\). However, because of \(g(\alpha)\neq\alpha\) for some \(\alpha\), \(A\) is not of the type (5.4) in the paper. Instead of (5.4) the family \(A_n(P,g,b)\) of serial rings constructed by the reviewer, where \(P\) is a uniserial ring, \(g\in\text{aut }P\) and \(b\in\text{rad }P\), describes all (basic) indecomposable serial quasi-Frobenius rings \(R\) with \(d(R)\not\equiv 1(n)\). (See [\textit{H. Kupisch}, Arch. Math. 26, 23-35 (1975; Zbl 0304.16006) and Arch. Math. 17, 20-35 (1966; Zbl 0135.07503)], where a full classification of all Artinian serial rings in terms of uniserial rings has been given.) Besides, the ``blow-up'' matrix-construction which is central in the reduction (1), was already used in [J. Algebra 33, 68-74 (1975; Zbl 0296.16013)] by the reviewer.