“Slim” Benchmark Sets for Faster Method Development

The construction of large benchmark sets has accelerated
advancement
of quantum chemistry methods, especially in density functional theory
and lower-cost methods. However, these large benchmark sets can be
unsuitable for cutting-edge method development, because research codes
developed for fundamentally new approaches are often inefficient and
may consequently struggle to handle large molecules. Here, we introduce Slim benchmark sets that are designed to ‘summarize’
the statistics of larger (in number and size of molecules) counterparts,
but have the advantage that molecules are restricted in size (to 5,
16, and 20 atoms) and may therefore be treated by inefficient implementations.
Remarkably, our 16 and 20 atom Slim sets effectively
summarize reactions involving much larger numbers of atoms. They thereby
allow data-driven methodologies to be exploited in the early stages
of cutting-edge method development.