On a theorem of Fuller (Q5906836)

Let \(R\) be a semiperfect ring with Jacobson radical \(J\). If \(e\) and \(f\) are primitive idempotents in \(R\), the pair \((eR,Rf)\) is called an \(i\)-pair if \(\text{Soc}(eR_R)\cong fR/fJ\) and \(\text{Soc}(_RRf)\cong Re/Je\). If, moreover, both \(eR_R\) and \(_RRf\) have essential socles (as, for example, if \(R\) is left and right perfect) then \((eR,Rf)\) is called a Nakayama pair. A theorem of \textit{K. R. Fuller} [Pac. J. Math. 29, 115-135 (1969; Zbl 0182.05702)] states that if \(R\) is either left or right Artinian and \(e\) is a primitive idempotent then \(eR_R\) is injective if and only if there exists a primitive idempotent \(f\) in \(R\) such that \((eR,Rf)\) is an \(i\)-pair. This has been improved on by \textit{Y. Baba} and \textit{K. Oshiro} [J. Algebra 154, No. 1, 86-94 (1993; Zbl 0808.16022)] who showed that if \(R\) is a semiprimary ring and \((eR,Rf)\) is an \(i\)-pair then the following are equivalent: (1) \(_{eRe}eR\) is Artinian; (2) \(Rf_{fRf}\) is Artinian; (3) both \(eR_R\) and \(_RRf\) are injective. As a further generalization of this, the main theorem of the paper under review states that if \((eR,Rf)\) is a Nakayama pair then the following four statements are equivalent: (1) both \(_{eRe}eR\) and \(Rf_{fRf}\) are linearly compact; (2) both \(eR_R\) and \(_RRf\) are injective; (3) \(eR_R\) is injective and \(Rf_{fRf}\) is linearly compact; (4) \(_{eRe}eR\) is linearly compact and \(_RRf\) is injective. Examples are given to illustrate and compare the theorem with Baba and Oshiro's. Moreover an example is given showing that Baba and Oshiro's condition (3) cannot be weakened to one-sided injectivity.