Density function theory (DFT) has been a popular method for electronic structure calculations in computational chemistry. B3LYP is frequently used as a DFT functional for organic molecules because DFT calculations are less computationally demanding. Standard DFT functional including B3LYP do not consider weak interactions such as van der Waals forces, and hydrogen bonding in the long-range region. The description of the B3LYP fitting of long-range gradual increasing correlation potential is incorrect. Weak interactions (long-range electron correlations) such as hydrogen bonding and π-π interaction effects dominated by dispersion are impossible with standard DFT. Dispersion correction should be considered for accurate thermochemical properties it acts intramolecularly between atoms or functional groups (aromatic rings and dimers etc.) that are not bonded directly.
It contributes to the internal energies of large molecules. Gaussian 16 includes dispersion correction (DFT-D3) and also damping function BJ “Becke-Johnson” to D3. DFT-D with damped, atom-pair-wise potential is added to standard Kohn-Sham DFT computed results. BJ damping scheme is basically empirical and appears to be correct, especially at short distances (i.e. vanishing forces), and provides a good estimation of dispersion energy. The dispersion can be introduced in the Gaussian root section with the keyword EmpiricalDispersion equals the desired dispersion term D3 or D3BJ as shown below.
#p DFT_method/basis_set Opt=(MaxCycles=1000) IOp(3/124=40) EmpiricalDispersion=GD3BJ Integral=UltraFine SCF=QC
The dispersion energy (in Hartree units) is reported in the output section as
Output—–> R6Disp: Grimme-D3(BJ) Dispersion energy=-0.0494906761 Hartrees.
This contribution will be added to the final energy, as the dispersion term is an add-on, it doesn’t affect the wave function directly. Dispersion correction contributes mainly to the force acting on the atoms, so SCF energy with the dispersion correction term could be different from the energy without the dispersion term and so does the subsequent molecular properties. Dispersion correction used for geometry optimization must be included for the evaluation of other molecular properties.
Most likely, the frequency calculation with large basis sets may not converge properly for large complex molecular systems. This problem can be overcome by optimizing the system with small basis sets, ex. 6-31G, and the system once converged with small basis sets should use that geometry as an initial guess for large basis sets.
#p DFT_method/6-31G Opt=(MaxCycles=1000) IOp(3/124=40) EmpiricalDispersion=GD3BJ Integral=UltraFine SCF=QC
Converged geometry of 6-31G as an initial guess for 6-311++G(d, p) or higher basis set as shown below.
#p DFT_method/6-311++G(d, p) Opt=ReadFC Freq Geom=AllCheck EmpiricalDispersion=GD3BJ Integral=UltraFine Guess=Read
If frequency calculation shows large negative frequencies, then the optimization is very close to the stationary point but not at the energy minimum. Negative frequencies can be avoided by selecting the problematic frequency in Gauss view/results/vibrations and scale by manual displacement towards +1. Then save the structure, and optimize it again to get the desired optimized structure. One can also start from the unoptimized (but very close to the stationary point) structure as a starting point and optimize again by using the following keywords in the root section.
#p DFT_method/basis_set Opt=ReadFC Freq Geom=AllCheck EmpiricalDispersion=GD3BJ Integral=UltraFine Guess=Read
Please note that the dispersion correction term must be included in weakly bound complexes and complexes involving dimers.
Worth reading
- Heiner et al. J. Chem. Theory Comput. 2015, 11, 7, 3163–3170.
- Jonny Proppe et al. J. Chem. Theory Comput. 2019, 15, 11, 6046–6060.
- Stefan Grimme et al. Int. j. Comput. Chem. 2011, 32: 1456-1465.
- Kaining Zhang et al. Phys. Chem. Chem. Phys., 2020,22, 13248-13260.
- Seiji Tsuzuki et al. Phys. Chem. Chem. Phys., 2020,22, 22508-22519.
