Contramodules over pro-perfect topological rings (Q2121543)

Let \(\mathfrak{R}\) be a separated and complete topological ring, with a base of neighbourhoods of zero consisting of open right ideals, and let \(X\) be a set. Define the following subset of the infinite direct product \(\mathfrak{R}^X\): \[ \mathfrak{R}[[X]] = \left\{\sum_{x\in X} r_x x : \text{for all open right ideals } \mathfrak{I}\subseteq \mathfrak{R}, r_x \in\mathfrak{I} \text{ for cofinitely many } x\in X\right\}. \] The map \(\mathbb{T}_{\mathfrak{R}}\) sending \(X\mapsto \mathfrak{R}[[X]]\) is functorial -- an endofunctor on the category of sets. In fact, \(\mathbb{T}_{\mathfrak{R}}\) is naturally a monad on the category of sets, with monad unit maps \(\varepsilon_X: X\to \mathfrak{R}[[X]]\) sending \(x_0\in X\) to the formal linear combination \(x_0\in\mathfrak{R}[[X]]\), and monad multiplication maps \(\phi_X: \mathfrak{R}[[\mathfrak{R}[[X]]]]\to \mathfrak{R}[[X]]\) given by ``multiplying out the brackets'' in \(\mathfrak{R}\). A \textit{left \(\mathfrak{R}\)-contramodule} is an algebra (or perhaps module) \(\mathfrak{C}\) over the monad \(\mathbb{T}_{\mathfrak{R}}\): that is, a set \(\mathfrak{C}\) given a \textit{left contraction map} \(\pi_\mathfrak{C}: \mathfrak{R}[[\mathfrak{C}]]\to \mathfrak{C}\) such that the following diagrams commute: \[ \begin{tikzcd} &\mathfrak{R}[[\mathfrak{C}]]\ar[dr, "{\pi_\mathfrak{C}}"] & &\mathfrak{R}[[\mathfrak{R}[[\mathfrak{C}]]]]\ar[r, "-{\phi_\mathfrak{C}}"] \ar[d, "{\mathfrak{R}[[\pi_\mathfrak{C}]]}" '] &\mathfrak{R}[[\mathfrak{C}]]\ar[d, "{\pi_\mathfrak{C}}"] \\ \mathfrak{C}\ar[ur, "{\varepsilon_\mathfrak{C}}"]\ar[rr, "{\mathrm{id}_\mathfrak{C}}" '] & &\mathfrak{C}, &\mathfrak{R}[[\mathfrak{C}]]\ar[r, "-{\pi_\mathfrak{C}}" '] &\mathfrak{C} \end{tikzcd} \] The category of left \(\mathfrak{R}\)-contramodules is abelian and has enough projectives. Left \(\mathfrak{R}\)-contramodules are in particular left \(\mathfrak{R}\)-modules. However, the usual notion of ``tensor product'' is badly behaved in this context. If \(\mathcal{N}\) is a left \(A\)-module (for some ring \(A\)) and a \textit{discrete} right \(\mathfrak{R}\)-module, and \(V\) is any left \(A\)-module, the abelian group \(\mathrm{Hom}_A(\mathcal{N},V)\) is naturally a left \(\mathfrak{R}\)-contramodule; the appropriate left adjoint to \(\mathrm{Hom}_A(\mathcal{N}, -)\) in this context is the \textit{contratensor product} \(\mathcal{N}\odot_\mathfrak{R} -\). The functor \(-\odot_{\mathfrak{R}} \mathfrak{C}\), taking discrete right \(\mathfrak{R}\)-modules to abelian groups, is right exact; \(\mathfrak{C}\) is called \textit{flat} if this functor is exact. The paper asks under what conditions all flat left \(\mathfrak{R}\)-contramodules have projective covers (a contramodule analogue of results of Bass and Yekutieli). In the above context, the following four conditions are already known to be equivalent to each other: \begin{itemize} \item[(i)] all flat left \(\mathfrak{R}\)-contramodules have projective covers, \item[(ii)] all left \(\mathfrak{R}\)-contramodules have projective covers, \item[(iii)] all flat left \(\mathfrak{R}\)-contramodules are projective, \item[(iv)] \(\mathfrak{R}\) has a two-sided ideal \(\mathfrak{H}\) (satisfying certain topological nilpotency and closure properties) such that \(\mathfrak{R}/\mathfrak{H}\) is a product of simple artinian discrete rings. \end{itemize} The following two conditions are also known to be equivalent to each other: \begin{itemize} \item[(i\({}^\flat\))] all Bass flat left \(\mathfrak{R}\)-contramodules have projective covers, \item[(iii\({}^\flat\))] all Bass flat left \(\mathfrak{R}\)-contramodules are projective. \end{itemize} (The \textit{Bass flat left \(\mathfrak{R}\)-contramodule} corresponding to a sequence of elements \(a_1, a_2, \dots \in \mathfrak{R}\) is the direct limit of the sequence \[ \mathfrak{R}\xrightarrow{a_1} \mathfrak{R}\xrightarrow{a_2} \mathfrak{R}\xrightarrow{a_3} \dots, \] where the maps are right multiplication.) The main result of the paper is as follows. Suppose, further to the above, that \(\mathfrak{R}\) has a base of neighbourhoods of zero formed by open \textit{two-sided} ideals, and that one of the following three additional conditions holds: \begin{itemize} \item[(a)] \(\mathfrak{R}\) is commutative, or \item[(b)] \(\mathfrak{R}\) has a countable base of neighbourhoods of zero, or \item[(c)] \(\mathfrak{R}\) has only a finite number of semisimple artinian discrete quotient rings. \end{itemize} Then conditions (i--iv), (i\({}^\flat\)) and (iii\({}^\flat\)) are all equivalent, and are also equivalent to each of the following conditions: \begin{itemize} \item[(v)] all descending chains of cyclic discrete right \(\mathfrak{R}\)-modules terminate, \item[(vi)] all the discrete quotient rings of \(\mathfrak{R}\) are left perfect. \end{itemize} In fact, if \(\mathfrak{R}\) has a base of neighbourhoods of zero formed by open \textit{right} ideals, the author shows that if \(\mathfrak{R}\) satisfies \begin{itemize} \item[(d)] \(\mathfrak{R}\) has a two-sided ideal \(\mathfrak{H}\) (satisfying certain topological nilpotency and closure properties) such that \(\mathfrak{R}/\mathfrak{H}\) is a product of topological rings all satisfying one of (a), (b) or (c), \end{itemize} then conditions (i--vi), (i\({}^\flat\)) and (iii\({}^\flat\)) are all equivalent; and conversely, if all of these conditions hold, then \(\mathfrak{R}\) must satisfy condition (d).