Component 3: What is special about B3LYP/6‑31G*?
B3LYP31–34 is undoubtedly the most used and cited
functional,121 and it is the de facto standard
for calculations involving organic molecules and even for teaching
purposes.122–126 B3LYP is the acronym for B88
exchange31 with Lee, Yang and Parr
correlation32 in the 3-parameter scheme of
Becke.33 B3LYP was proposed by Frisch et al. in 1994
for a study involving circular dichroism spectra of different
molecules,34 and it is based on a variation of the
B3PW91 functional of Becke.31,33,127,128 In B3LYP, the
PW91 correlation functional127,128 is replaced by the
LYP functional, but the remaining parts are unchanged (B88 exchange and
the 3-parameter scheme). The 6‑31G* basis set is a double-ζ basis set of
the Pople family,94–102 and it includes a very
manageable amount of basis functions. This method rose to popularity in
the mid 1990s, but since the mid 2000s many failures have been
documented in the specialized
literature.83,85,129–135 Unfortunately though,
B3LYP/6‑31G* has become almost a synonym for DFT calculation itself in
much of the broader audience (non-specialistic)
literature.136–140 The educational literature is not
exempted, as this method is applied without any formal justification,
nor critical evaluation of the results even for reactions that are known
to be problematic.122–126,139,141 We specifically
designed experiment 7 to give first-hand experience on the limitations
of B3LYP/6-31G*. Selected results are reported in Figure 2 ,
while full results are given in the SI. Instead of looking for systems
where B3LYP/6‑31G* is successful, we encourage practitioners—as others
did before us131—to move away from it and to seek
better methods for potentially problematic cases. As a guide, all other
hybrid GGA functionals are computationally as expensive as B3LYP (they
are on the same rung of Jacob’s Ladder), while modern basis sets have
similar number of basis functions as 6-31G*
(def2-SV(P)107 has even less).