I am working on transition metal cluster. I want to do geometry optimization, but the problem is during freq calculations. During the geometry optimization, when i do by turing off the symmetry, it does not go well. I mean scf does not converged. But when i turn on the symmetry, it does, and i am also able to get vibrational frequency in the default smearing, charge and spin mixing factor. So my idea here is,

1) At first i will do geometry optimization either turning symmetry off, or by using harris function from LDA.

2) when a get the optimized structure, i would like to run it with GGA using symmetry on.

By doing this i am able to get some of my simulation done. And since, i am working on cluster, i do not know the symmetry.

What do you think with this technique of geomtery optimization.

I have also tried using DIIS_Restart off, no luck in some of the simulations.

As far as i read the tutorial, it says that using harris functional, we get good accuray in the geometry of the input file, but the energy value and difference of the energy can't be trusted, e.g. binding energy of gas on these cluster.

What do you all think on this?