Venerdì 7 luglio 2023 - Ore 15:00 - Aula 22
Prof. John Herbert, Ohio State University
  Seminario  

Fragment-based quantum chemistry holds the promise to bring high-level ab initio methods to large systems, by exploiting the “short-sightedness” of electronic matter and harnessing the power of distributed computing. In the race to push such methods to exascale computing resources, however, emphasis on accuracy has been lost and the dirty secret in the field is that most existing fragment-based strategies cannot be considered to be accurate approximations to any well-defined supersystem calculation. This talk will examine sources of error and propose a promising solution: the energy-screened many-body expansion, which provides a stable and accurate route to apply correlated wave function methods to large systems. Proof-of-concept calculations (at DFT levels of theory) demonstrate that this approach does converge to a well-defined result, in thermochemical calculations involving enzyme models with 600+ atoms.