A remark on the generalized smashing conjecture (Q1340233): Difference between revisions

Recall briefly the generalized smashing conjecture: Let \(\mathcal S\) be a triangulated category in the sense of Verdier, with coproducts. An object \(X\) of \(\mathcal S\) is small if \(\text{Hom} (X,-)\) commutes with arbitrary coproducts. A localizing full subcategory of \(\mathcal S\) is a triangulated subcategory closed under forming coproducts with respect to \(\mathcal S\). The category \(\mathcal S\) is supposed to be the smallest localizing subcategory containing its small objects. A localizing subcategory \({\mathcal R} \subset {\mathcal S}\) is smashing if the inclusion admits a right adjoint commuting with arbitrary coproducts. For instance, if \(\mathcal R\) is generated by small objects, then \(\mathcal R\) is smashing. The generalized smashing conjecture states the converse: ``Every smashing subcategory is generated by small objects.'' In this paper, the author disproves this conjecture with \({\mathcal R} \subset {\mathcal D}A\), where \({\mathcal D} A\) is the unbounded derived category of the category of right unitary modules over a ring \(A\), and \(\mathcal R\) is the localizing subcategory generated by an ideal \(I\) contained in the Jacobson ideal of \(A\) and verifying \(\text{Tor}^A_i (A/I, A/I) = 0\), \(i > 0\). For that, the author proves first that the left derived functor of the tensor product is a right adjoint of the inclusion, \({\mathcal R} \subset {\mathcal D} A\), and then shows that any small object of \({\mathcal D} A\) belonging to \(\mathcal R\) is 0. An explicit example of this situation is given.