Cancellation and direct summands in dimension 1 (Q1178921): Difference between revisions

In integral representation theory and in the theory of classical orders, an important role is played by theorems on cancellation in direct sums (e.g., Bass, Drozd, Jacobinski) and on the existence of direct summands (e.g., Serre, Roiter, Jacobinski). The sharpest forms of these theorems were developed for finitely generated modules and/or lattices under assumptions such as the coefficient ring being an order over a Dedekind domain in a semisimple Artinian ring, or an order contained in a maximal order. Here the authors extend these classical theorems to a more general and very natural setting, namely arbitrary finitely generated modules over a module-finite algebra \(\Lambda\) over a commutative Noetherian ring \(R\) of Krull dimension 1. Nilpotent ideals are allowed, and maximal orders are not required. The main cancellation theorem of the paper gives a sufficient condition (too technical to state here) for cancellation to hold within the genus of a \(\Lambda\)-module \(M\) that is, for \(\Lambda\)-modules that become isomorphic to \(M\) on localization at each maximal ideal of \(R\). Corollaries include \textit{Yu. A. Drozd}'s cancellation theorem [Izv. Akad. Nauk SSSR, Ser. Mat. 33, 1080--1088 (1969; Zbl 0208.04502)], as well as the first author's generalization [J. Algebra 18, 169--177 (1984; Zbl 0533.13006)] of \textit{H. Jacobinski}'s cancellation theorem [Acta Math. 121, 1--29 (1968; Zbl 0167.04503)]. A stability theorem is also obtained: if cancellation holds in the genera of \(\Lambda\)-modules \(M\) and \(M'\), then it also holds in the genus of \(M\oplus M'\). Turning to direct summands, the following generalization of \textit{A. V. Roiter}'s direct summand theorem [Izv. Akad. Nauk SSSR, Ser. Mat. 30, 1315--1324 (1966; Zbl 0232.20006)] is proved: if \(X\), \(M\), \(N\) are \(\Lambda\)-modules such that (a) for each maximal ideal \(\mathfrak m\) of \(R\), the localization \(X_{\mathfrak m}\) is isomorphic to a direct summand of \(M_{\mathfrak m}\), and (b) for each minimal prime ideal \(\mathfrak p\) of \(R\), each indecomposable direct summand of \(X_{\mathfrak p}\) is isomorphic to a direct summand of \(N_{\mathfrak p}\), then \(X\) is isomorphic to a direct summand of \(M\oplus N\). An extension of the first author's generalization [J. Number Theory 26, 149--165 (1987; Zbl 0621.16007)] of Jacobinski's direct summand theorem [op. cit.] is obtained as a corollary. Namely, if \(X\) and \(M\) are \(\Lambda\)-modules satisfying (a) above and (c) for each minimal prime ideal \(\mathfrak p\) of \(R\), each indecomposable direct summand of \(X_{\mathfrak p}\) occurs with greater multiplicity as a direct summand of \(M_{\mathfrak p}\), then \(X\) is isomorphic to a direct summand of \(M\).