The maximal regular ideal of self-injective and continuous rings splits off (Q761529)

The main result states that the maximal regular ideal M(R) splits off as a ring direct summand of R when R is a (right and left) continuous ring. In particular, any continuous ring, e.g., any (right and left) self- injective ring R is a direct product \(R_ 1\times R_ 2\), where \(R_ 1\) is von Neumann regular, and \(R_ 2\) has no regular ideals except 0. Actually R can be a semicontinuous top-regular ring for this. (See below for definitions.) The results extend splitting theorems of M. Hall jun., B. Brown and N. H. McCoy, Van Huynh and the author, and reduce the study of these rings to regular rings \(R_ 1\) and bound rings \(R_ 2\) by the results of \textit{B. Brown} and \textit{N. H. McCoy} [Proc. Am. Math. Soc. 1, 165-171 (1950; Zbl 0036.297)]. When M(R) is a maximal regular right ideal, equivalently M(R) contains all regular right ideals, then the splitting theorem holds for a right continuous or right self-injective ring R. A necessary and sufficient condition in order that M(R) splits off in an arbitrary ring R is that the canonical image of M(R) in the residue ring \(\bar R\) of R modulo radical splits as a right or left \(\bar R-\)module. This always happens whenever \(\bar R\) is semisimple, the principal results of the author [Arch. Math. 12, 179-181 (1961; Zbl 0104.031)]. In the last section we show for a ring R generated by units that M(R) is the intersection of all maximal regular right (left) ideals. We also prove more general splitting theorems for ideals which are maximal with respect to being semiprime rings, or semiprimitive rings, under the assumption that every one-sided ideal is essential in one generated by an idempotent. Rings with this property, which is one half of the defining property of continuous rings, are called semi-continuous rings in the text, and are shown to be equivalent to CS rings defined elsewhere. For these rings, then, any maximal semiprime (or semiprimitive) ideal splits off, and thereby becomes the unique maximal semiprime (or semiprimitive) ideal. Under the further assumption of top- regularity, the maximal semiprimitive ideal coincides with M(R). \textit{D. Van Huynh} [Arch. Math. 33, 232-234 (1979; Zbl 0425.16030)] proved that M(R) splits off for a linearly compact ring R. Since neither self-injective nor semi-continuous rings need be linearly compact, our results are not contained in Van Huynhs, but, conversely, our result does not appear to contain his. (Are linearly compact rings semi-continuous, equivalently CS?). However, the author's earlier result does contain Van Huynh's, since \(\bar R\) is semisimple Artinian in a linearly compact ring R, according to \textit{F. L. Sandomierski} [Ring Theory, Proc. Conf. Ring theory Park City, Utah 1971, 333-346 (1972; Zbl 0234.16013)]. A result of \textit{C. Faith} and \textit{S. S. Page} [FPF Ring Theory: Faithful Modules and Generators of Mod-R (1984; Zbl 0554.16007) (Chapter 5)] states that M(R) splits off in (right and left) FPF rings, generalizing the result of the author [Lect. Notes Math. 700, 151-203 (1979; Zbl 0413.13003), Theorem 10] for commutative FPF rings. While the latter result fits into the framework of this paper (commutative FPF rings are CS rings by [the author, Lect. Notes Pure Appl. Math. 72, 71- 105 (1982; Zbl 0546.13006), p. 83, Cor. 3.10], the non-commutative result does not, at least not obviously. Neither, for that matter, does the earlier theorem of the author. (Misstated by Van Huynh since the theorem states that M(R) splits off when \(\bar R\) satisfies the minimum condition on principal left ideals [of \(\bar R\) rather than R]. Thus, e.g., neither is the author's earlier paper contained in Van Huynh!) The most general criterion for the splitting off of M(R) thus far is the one stated in the third paragraph of the introduction, namely \(\overline{M(R)}\) splits off in \(\bar R\) qua right or left ideal.