Component 1: Which functional should I use?
Given the sheer amount of options available, a key issue in modern
KS-DFT research is the choice of an approximate xc functional.
This is not an easy task, as the accuracy and reliability of the results
is dependent on the functional chosen to describe the chemical system.
If one chooses a functional that is not suitable for a given chemical
application, the results will not be reliable (in accordance with the
old computer science adage: garbage in, garbage out). This does not mean
that DFT fails as a whole, but rather that the specific approximation is
inadequate for the chemical problem under investigation. In this
context, we may think of the functional (and not of DFT) as a tool
needed to perform a specific task. If the tool is not adequate, we
should choose a different tool, or a different task. If we need to put a
nail in the wall, we will pick a hammer, regardless of the number of
screwdrivers we have at our disposal. As Kathrin Hopmann recently
advised: “Do not use DFT for mechanisms it cannot
handle”.70 Many topics at the center of the last
decades of DFT research, such as dispersion
corrections,71–76 broadly applicable meta -GGA
functionals,36,37,40,43,44 and functionals for strong
correlation,36,42,69,77–81 have been introduced and
studied to provide better and more reliable approximations. Several of
these new advances have been facilitated by the introduction of large
databases of accurate data of broad chemical
interest.25,26,36,43,69,82 The use of databases
outside the theoretical chemistry community is encouraged as they are
the primary source for suggestions when choosing the xcfunctional for a specific application. It is good practice to compare
the specific reaction under investigation with a similar one in a
computational database before choosing a functional. In addition,
databases can also give an estimate of the expected accuracy of the
results over a broad spectrum of chemical properties. The reactions we
will analyze in this tutorial come from three important collections of
chemical data, namely MGCDB84,26GMTKN55,25 and Minnesota
2015.36,43,69
Table 1. Structure of the laboratory notebook and corresponding take-home message.