Closed foxtran closed 3 weeks ago
An extensive list of functionals, I believe the damping parameters are accurate with two digits.
I believe the damping parameters are accurate with two digits.
There is a problem that it is not enough.
Taking parameters from #92 (s6=0.418, a2=5.65), the dispersion energy for example.xyz is -1.5172415429713E-02
. With two digits (s6=0.42, a2=5.7), the dispersion energy is -1.4655518783476E-02
. The difference is ~3.5 % and that is huge.
The output:
build % app/s-dftd3 ../app/example.xyz --bj-param 0.418 0.0 0.0 5.65
-----------------------------------
s i m p l e D F T - D 3 v1.1.1
-----------------------------------
Rational (Becke-Johnson) damping: D3(BJ)
---------------------
s6 0.4180
s8 0.0000
s9 0.0000
a1 0.0000
a2 5.6500
alp 14.0000
--------------------
Dispersion energy: -1.5172415429713E-02 Eh
[Info] Dispersion energy written to .EDISP
build % app/s-dftd3 ../app/example.xyz --bj-param 0.42 0.0 0.0 5.7
-----------------------------------
s i m p l e D F T - D 3 v1.1.1
-----------------------------------
Rational (Becke-Johnson) damping: D3(BJ)
---------------------
s6 0.4200
s8 0.0000
s9 0.0000
a1 0.0000
a2 5.7000
alp 14.0000
--------------------
Dispersion energy: -1.4655518783476E-02 Eh
DSD-BLYP is already implemented.
A huge list of parameters for a set of functionals was presented in 10.1002/jcc.23391. Unfortunately, only 2-digit precision is. See the SI of this paper.