[Wien] Iterative diagonalization in SCF cycles

[msoumeli at physics.auth.gr]

At first I tried full diagonalization and it failed. After changing  
the convergence criteria it still crashed, so I tried iterative  
diagonalization. The command I used was run_lapw -it, which is  
iterative diagonalization after the first cycle. My system reached  
convergence with this command and my concern is if this command is as  
accurate as full diagonalization.

In another case I am studying the file .scf1 containes an atom with  
E(top)=-200. When this value appears the scf crashes. Which parameter  
affects E(top)? How can I have an easier convergence? And what can  
someone do if an scf fails?

Thank you.