Accuracy of DFTB for high-T carbon chemistry:
Summary
• DFTB reproduces B3LYP/6-31G(d) energetics very well
• DFTB
reproduces B3LYP/6-31G(d) geometries excellently (PES is qualitatively and quantitatively
correct)
• DFTB
compares much better w/DFT than conventional semiempirical methods at similar computational
cost
• High-level DFT may not always give the
qualitatively correct answer
in unusual cases, especially when ring strain is high
• Fortunately, cases where DFT fails are rare
and related to
extremely unlikely geometries
Approximate DFT/MD* possible over ~ ns
timescale