How to direct the self-assembly of simple templates toward constructs with complex shape and topology is still an open problem. Recent advancements have made it possible to self-assemble various types of knotted constructs, but targeting general multicomponent topologies, that is, links and catenanes, has proved much harder. Here, we study how the yield and complexity of self-assembled links depends on both intrinsic and extrinsic system properties and, particularly, template shape and spatial confinement. We show that slit confinement does not necessarily suppress linking but can rather enhance it significantly thanks to entropic effects. We also found that only a limited set of binary links are recurrent for different template shapes. These privileged topologies include all those experimentally realized so far plus a few additional ones, such as the 7(7)(2) and 7(8)(2) links that, hence, ought to be ideal candidates for broadening the current class of constructs with addressable topology.
