Peirce decompositions, idempotents and rings (Q2201070)

One defines an \(n\times n\) generalized matrix ring in terms of rings \(R_1,\dots, R_n\), \((R_i,R_j)\)-bimodules \(M_{ij}\) and bimodule homomorphisms \(\theta_{ijk}: M_{ik}\otimes M_{kj}\longrightarrow M_{ij}\). In this case, the diagonal matrix units \(E_{11}, \dots, E_{nn}\) form a complete set of orthogonal idempotents. Reciprocally, an arbitrary ring \(R\) with a complete set of \(n\) orthogonal idempotents has a Peirce decomposition that can be arranged as an \(n\times n\) generalized matrix ring isomorphic to \(R\). In a previous paper [\textit{P. N. Ánh} et al., Linear Multilinear Algebra 64, No. 10, 2002--2029 (2016; Zbl 1376.16023)], the authors studied the subclass \(\mathcal{T}_n\) of \(n\times n\) generalized matrix rings for which \(\theta_{iji}=0\) for \(i\neq j\). The triviality of these bimodule homomorphisms motivates the introduction of the so called trivial Peirce idempotents, which are idempotents \(e=e^2\) of \(R\) satisfying \(eR(1-e)Re=0=(1-e)ReR(1-e)\). The set of Peirce trivial idempotents of R is denoted by \(\mathfrak{P}_t(R)\) and with that one define the concept of \(n\)-Peirce rings inductively as follows: A ring \(R\) is \(1\)-Peirce ring if \(\mathfrak{P}_t(R)=\{0, 1\}\), with \(1 \neq 0\). Inductively, for a natural number \(n > 1\), a ring \(R\) is called an \(n\)-Peirce ring if there is an \(e \in \mathfrak{P}_t(R)\) such that \(eRe\) is an \(m\)-Peirce ring for some \(m, 1 \leq m < n\), and \((1-e)R(1-e)\) is an \((n-m)\)-Peirce ring. The above definition is shown to be independent on the choice of elements in \(\mathfrak{P}_t(R)\). Peirce trivial idempotents generalize the notion of semicentral idempotents which occur naturally in the structure of \(2\times 2\) generalized triangular matrix rings. For a natural number \(n\), \(n\)-Peirce rings are generalizations of \(n\)-strongly triangular matrix rings (see [\textit{P. N. Ánh} and \textit{L. van Wyk}, Linear Algebra Appl. 438, No. 11, 4374--4381 (2013; Zbl 1284.16034)]). The paper investigates the \(n\)-Peirce rings, with a coordinate-free treatment. In section 2, it is shown that a structure theory of Peirce rings can be developed in a similar way as that of Bass for semiperfect rings. In section 3, the authors define the so called trivial idempotents relative to certain radicals and sketch the process of how to lift results on semisimple factors to the rings. The present approach offers a way to unify and to expand the classical theory of semiperfect rings and idempotents to much larger classes of rings. Examples and applications are included in section 4.