Bull. Aust. Math. Soc. 82 (2010), 99–112 doi:10.1017/S0004972710000213 MODULES WITH ABELIAN ENDOMORPHISM RINGS GRIGORE CALUG˘ AREANU˘ and PHILL SCHULTZ ˛ (Received 15 October 2009) Abstract The results of Szele and Szendrei [‘On Abelian groups with commutative endomorphism rings’, Acta Math. Acad. Sci. Hungar. 2 (1951), 309–324] characterizing abelian groups with commutative endomorphism rings are generalized to modules whose endomorphism rings have various restrictions on their idempotents. Such properties include central or commuting idempotents, and one-sided ideals being two-sided. Related properties include direct summands having unique complements, or being fully invariant. 2000 Mathematics subject classification: primary 16D70; secondary 16D99. Keywords and phrases: commutative endomorphism rings, central or commuting idempotents, duo rings, unique complements, fully invariant submodules. 1. Introduction Throughout this paper, unless specifically noted, the word ring denotes a unital associative ring, group denotes an abelian group, and module denotes a unital right R-module. It is a long-standing problem in abelian group theory to describe those groups G whose endomorphism ring End.G/ is commutative and those commutative rings which are isomorphic to the endomorphism ring of an abelian group. Krylov et al. TKMT03U present an authoritative survey of the history of the problems. The known results are of two sorts: if G is characterized by torsion, then a complete classification is known for both parts of the problem TKMT03, Corollary 19.3U; if G is torsion-free, then no classification is feasible; instead, there are realization theorems stating that large classes of commutative rings can be realized as End.G/ for some group G TKMT03, Section 29U. Naturally, mixed groups lie between these two extremes. The purpose of this paper is to generalize the problem by considering modules whose endomorphism ring is ‘almost commutative’. For example, we describe classes of modules whose endomorphism rings have commuting idempotents, or in which one-sided ideals are two-sided. We also consider modules for which direct summands have unique complements or are fully invariant. c 2010 Australian Mathematical Publishing Association Inc. 0004-9727/2010 $16.00 99 Downloaded from https://www.cambridge.org/core. IP address: 170.106.35.93, on 30 Sep 2021 at 20:21:42, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0004972710000213 100 G. Calug˘ areanu˘ and P. Schultz [2] Let us first establish some notation. Elements of a ring or module are denoted by lower-case Latin letters, and homomorphisms by lower-case Greek letters. Functions are written on the left of their arguments. We denote a submodule N of M by N ≤ M and a direct summand N of M by N ≤⊕ M. In particular, if R is a ring, then I ≤ R means that I is a right ideal of R. We denote the (not necessarily abelian) group of units of R by U D U.R/ and the centre of a ring R by Z D Z.R/. Let R be a ring and let E D E.R/ denote the set of idempotents in R. If e 2 E, then e is called proper if e 6D 0 or 1. There is a natural partial order, denoted ≤, on E defined by e ≤ f if ef D f e D e, and nonzero minimal elements of E with respect to this order are called primitive idempotents. A ring R is called abelian if E.R/ ⊆ Z.R/, that is, if all idempotents are central. If e 2 E, then eN D 1 − e 2 E is called the complementary idempotent and eeN D eeN D 0. In general, e and f 2 E are called orthogonal if ef D f e D 0. If e 2 E is proper, then R D eR ⊕eR N is a proper direct decomposition of R considered as right R-module. In particular, e is primitive if and only if eR is a minimal right ideal summand. Similar remarks hold of course for left modules Re and ReN. If e 2 E is central, then so is eN and the decomposition is into a direct sum of ideals. For right R-modules M and N; Hom.M; N/ D HomR.M; N/ denotes the additive group of R-homomorphisms of M into N and End.M/ D EndR.M/ the ring of R-endomorphisms of M. The following facts concerning idempotents " 2 End.M/ are well known TF73, Section 106U. If " is a proper idempotent then " determines a nontrivial direct decomposition M D im."/ ⊕ ker."/. Conversely, if M D N ⊕ K is a decomposition of M into nonzero summands, then there is a unique idempotent " 2 End.M/ such that "jN is the identity map and "jK is the zero map. We say that N ⊕ K is the decomposition corresponding to " and that " is the idempotent corresponding to the decomposition M D N ⊕ K . The paper is organized as follows. In Section 2, we describe the rings in which direct summand right ideals are uniquely complemented and relate this property to subcommutativity and certain one-sided ideals being two-sided. In Section 3 we show that these properties are equivalent to R being abelian. In the rest of the paper, the results of Sections 2 and 3 are applied to endomorphism rings of modules. For example, in Section 4 we show how unique complementation of summands of End.M/ implies unique complementation of summands of M. We also find a module theoretic generalization of the main result of Szele and Szendrei TSzSz51U on abelian groups with commutative endomorphism rings. Section 5 is devoted to characterizations of the fully invariant submodules and direct summands of a module. 2. Uniquely complemented summands of rings Suppose that R D I ⊕ J is a nontrivial decomposition of the ring R into right ideals. Then the idempotent endomorphisms considered above can be realized by multiplication by elements of E, the set of idempotents of R. In this section, Downloaded from https://www.cambridge.org/core. IP address: 170.106.35.93, on 30 Sep 2021 at 20:21:42, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0004972710000213 [3] Modules with abelian endomorphism rings 101 we examine the connection between the centrality of e 2 E and uniqueness of complements of the direct summand eR. We begin with some known facts and some elementary results. PROPOSITION 2.1 (Folklore). Let RR D I ⊕ J be a right ideal direct decomposition of R. There exists e 2 E such that I D eR and J D eR. Much is known about idempotents e and f for which eR D f R. For example, here is a selection from Lam TLam95, Section 21U. PROPOSITION 2.2. Let e and f 2 E. The following are equivalent: (1) eR D f R; (2) ef D f and f e D e; (3) there exists r 2 R such that f D e C ere; (4) there exists u 2 U such that f D eu; (5) Re D R fN . The direct complements of a right ideal direct summand may be characterized as follows. LEMMA 2.3. Let R D eR ⊕ gR. Then gR D .e − ere/R for a suitable r 2 R. Conversely, for every r 2 R, .e − ere/R is a direct complement for eR. PROOF. According to Proposition 2.1, there is an idempotent f such that eR D f R and gR D f R. By Proposition 2.2, f D e C ere and g D f C f s f for suitable r; s 2 R. Thus g D f .1 C f s f / and u D 1 C f s f is invertible in R with inverse 1 − f s f . Hence g D .e − ere/u so that gR D .e − ere/R. As for the converse, again by Proposition 2.1, eR ⊕ .e − ere/R D R if there is an idempotent f 2 R such that eR D f R and .e − ere/R D f R. Clearly f D e C ere is suitable for this. 2 Since eR always has the direct complement eR, we say that eR has a unique complement if R D eR ⊕ gR implies that eR D gR. PROPOSITION 2.4. eR has a unique complement if and only if eRe D 0. PROOF. Since by Lemma 2.3 all the direct complements of eR are of the form .e − ere/R; eRe D 0 implies that eR has a unique complement. Conversely, if for all r 2 R;.e − ere/R D eR, it follows from Proposition 2.2(2) that .e − ere/e D e and so ere D 0 for all r 2 R. 2 Similarly, eR is the unique direct complement of eR if and only if eRe D 0. We now consider the question of when not just the complement of eR but even its idempotent generator e is unique. PROPOSITION 2.5. Let R D eR ⊕ f R with e; f 2 E. Then f D e if and only if ef D f e. Downloaded from https://www.cambridge.org/core. IP address: 170.106.35.93, on 30 Sep 2021 at 20:21:42, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0004972710000213 102 G. Calug˘ areanu˘ and P. Schultz [4] PROOF. If f D e then ef D e.1 − e/ D .1 − e/e D f e.