The Activation-Relaxation Technique (ARTn) is an efficient technique for finding the minima and saddle points of multidimensional functions such as the potential energy surface of atomic systems in chemistry. In this work we detail and illustrate significant improvements made to the algorithm, regarding both preprocessing and the activation process itself. As showcased, these advances significantly reduce ARTn computational costs, especially when applied with ab initio description. With these modifications, ARTn establishes itself as a very efficient method for exploring the energy landscape and chemical reactions associated with complex mechanisms.
