Atomic quasi-injective modules (Q2523589)

This note is a slight addition to a paper of \textit{R. E. Johnson} and the author [J. Lond. Math. Soc. 36, 260--268 (1961; Zbl 0103.02203)]. Let \(_RM\) be a quasi-injective module with zero singular submodule. Let \(K = \Hom_R(M, M)\). By the above paper, \(K = \Hom_R(\overline M, \overline M)\), where \(\overline M\) is the injective hull of \(M\). An atom of \(M\) is an indecomposable direct summand. Let \(C\) be the sum of the atoms of \(M\). Then \(C\) is a direct sum of atoms, \(C\) is quasi-injective and so a \(K\)-submodule of \(M\), and \(C_K\) is completely reducible. Moreover, the singular submodule of \(C_K\) is zero, and if \(_RM\) is an essential extension of \(_RC\), \(M_K\) is an essential extension of \(C_K\). In the Johnson and Wong paper (loc. cit.), they prove that for an atom \(N\), for each \(h\in \Hom_K(N, N)\) and \(\{x_1\dots x_n\}\in N\), there exist \(r, r'\in R\) such that \(rhx_i = r'x_i\), \(1\leq i\leq n\). This result is shown to hold for the module \(C\). Moreover, if \(M\) is faithful, then \(\Hom_K(C, C)\) is a left quotient ring of \(R\) if and only if any proper \(K\)-submodule of \(C\) has non-zero annihilator in \(R\).