Cancellation of finite-dimensional Noetherian modules (Q2144390)

The present paper is devoted to establishing the following module cancellation theorem (cancellation in the sense that \(K\oplus M\cong K\oplus L\Rightarrow M\cong L\)) that generalizes -- and unifies the proofs of -- some previous cancellation results in the literature in several papers, namely Theorem 9.3. of \textit{H. Bass} [Publ. Math., Inst. Hautes Étud. Sci. 22, 489--544 (1964; Zbl 0248.18025)], Theorem 2 of \textit{A. Dress} [Bull. Am. Math. Soc. 75, 984--986 (1969; Zbl 0201.04204)] and Theorem 3.14 of \textit{A. De Stefani} et al. [Proc. Am. Math. Soc. 146, No. 4, 1417--1430 (2018; Zbl 1427.13009)]: \textbf{Theorem 1.2 of the paper under review}. Let \(K,L,M,\) and \(N\) be right modules over a ring \(S\) that is algebra over a commutative ring \(R\), and let \(E:=\mathrm{End}_S(N)\) denote the \(R\)-algebra of all \(S\)-linear endomorphisms of \(N\). Assume the following: \begin{itemize} \item[(1)] \(X:=j-\mathrm{Spec}(R)\cap\mathrm{Supp}_R(N)\) is a Noetherian space (here \(j-\mathrm{Spec}(R)\) is the Zariski subspace of \(\mathrm{Spec}(R)\) composed of the prime ideals of \(R\) that are intersections of maximal ideals of \(R\)). \item[(2)] \(N\) is a finitely presented \(S\)-module, and \(E\) is a module finite \(R\)-algebra. \item[(3)] There is a finitely presented left \(E\)-submodule \(F\) of \(\mathrm{Hom}_S(M,N)\) such that, for every \(\mathfrak{p}\in X\), there is a split surjection in \(F_{\mathfrak{p}}^{\bigoplus (1+\dim_X(\mathfrak{p}))}\subseteq\mathrm{Hom}_{S_{\mathfrak{p}}}(M_{\mathfrak{p}},N_{\mathfrak{p}}^{\bigoplus (1+\dim_X(\mathfrak{p}))})\). \end{itemize} Then \(N\oplus L\cong N\oplus M\Rightarrow L\cong M\). More generally, if \(K\) is a direct summand of a direct sum of finitely many copies of \(N\) over \(S\), then \(K\oplus L\cong K\oplus M\Rightarrow L\cong M\). The introduction of the present paper includes an informative -- probably comprehensive -- explanation of the cancellations results as well as counterexamples to the cancellation property that exist in the literature. The proof of the Theorem 1.2 above involves many computations and preparatory lemmas, but one important fact used in the proof is the fact that under the notation and conditions of Theorem 1.2 \(E_{\mathfrak{p}}\cong \mathrm{End}_{S_{\mathfrak{p}}}(N_\mathfrak{p})\) is a ring of \textit{stable rank \(1\)} and \(E_{\mathfrak{p}}/\mathrm{Jac}(E_{\mathfrak{p}})\) is a \textit{unit-regular} ring for every \(\mathfrak{p}\in X\). Here a \textit{unit regular ring} is a ring for which every element \(a\) can be presented as \(a=ava\) for some unit element \(v\). The definition of having \textit{stable rank \(m\)} is more lengthy, however, for an imprecise explanation it amounts to obtaining from \(n>m\)-elements of \(E\) generating the improper ideal \(E\), certain \((n-1)\)-elements of a specified form generating again \(E\).