In Gaussian program package, while performing geometry optimization, when we choose opt=tight, convergence threshold for force set to 1.5*10^-5 Hartree/Bohr, for displacement it is 6.0*10^-5 Angstrom but what is the convergence threshold for energy? I am having this question because I have seen different energy threshold values in different cases.
Your suggestions would be appreciated.
Thank you!
Please have a look, for example, at
http://www.lct.jussieu.fr/manuels/Gaussian98/nl001_tn.htm
Is says: "... using tighter convergence criteria makes almost no difference in terms of energy ... The energy is converged to less then 1 microHartree..."
Thanks a lot Basiuk!
This is clear that opt=tight does not make much difference in energy. But when we see the output file for a Gaussian input (say regarding geometry optimization), we see the convergence threshold values for force and displacement, in the same way I want to know what is the convergence threshold for Energy. From the file sent by you, what I am getting is 1*10^-6 Hartree is the convergence threshold for energy. Am I getting it correctly or not?
If you expand the energy as a Taylor series in the geometry, what you get is
E(x) = E_0 - f * x + 1/2 x^T H x
where f = -nabla E is the force and H is the Hessian.
If "convergence threshold for force set to 1.5*10^-5 Hartree/Bohr, for displacement it is 6.0*10^-5 Angstrom" then you can see that the energy should be converged to roughly |f| |delta x| = 1.5e-5 Ha/a0 * 6e-5 Å which is about 2e-9 Hartree.
The convergence of the electronic wave function will, of course, also affect this, so for accurate geometries the SCF has to be very tight: the 2e-9 Hartree energy uncertainty between geometries should be much larger than the uncertainty of the total energy at a single geometry.
https://www.researchgate.net/deref/http%3A%2F%2Fwww.lct.jussieu.fr%2Fmanuels%2FGaussian98%2Fnl001_tn.htm
following
- Badges
- Science topic