1) The initial dH jump occurs because it is sort of a linear mixing of 1 since it reads the DM, creates H, then calculates the new DM and the new H. So I would consider this a feature when mixing H. When you re-read the H it should be somewhat different.
Yes, you can save the SCF DM's, but you can't read it in yet... :(
=> I think not. I think it is better to have a *very* small linear mixing for the initial iscf=1 step, then you can bump up the weight in subsequent SCF iterations.
1) The initial dH jump occurs because it is sort of a linear mixing of 1 since it reads the DM, creates H, then calculates the new DM and the new H. So I would consider this a feature when mixing H. When you re-read the H it should be somewhat different.
Yes, you can save the SCF DM's, but you can't read it in yet... :(
=> I think not. I think it is better to have a *very* small linear mixing for the initial iscf=1 step, then you can bump up the weight in subsequent SCF iterations.
2) Ok.