Read afm wave function ? Why does abinit destroy AFM order ?

Moderators: torrent, jzwanzig

Post Reply
Posts: 1
Joined: Thu May 06, 2010 10:36 am

Read afm wave function ? Why does abinit destroy AFM order ?

Post by dadams » Thu May 06, 2010 11:15 am


I would like to do following: generate an abo_WFK file with an antiferromagnetic (AFM) order but without any AFM-symmetry elements (no symafm elements).

The AFM order is in some cases to obtain spontaneously therefore I generate first a afmo_WFK file where the AFM order is imposed (nsppol 1, nspden). However, I converge the calculation well (tolvrs 10d-8). Afterwards I would like to start from that file without imposing the AFM order (nsppol 2, irdwfk 1).

I observe following: abinit reads afmo_WFK file correctly (e.g. the occupations at the first step are identical to the afm run to a precision of 10d-6). But abinit "does not realize" that the wavefunction is already converged and thus performs optimizations steps which strongly perturb the system and destroy the AFM order (here at least the high spin state). Furthermore a state with a much higher energy is
found. Why?

If I impose in the second run the occupation matrix of the AFM, I can recover the AFM state. This indicates that the AFM is at least a local minimum of the magnetic states. It should not be destroyed in a follow-up run!

Usually it works well to converge the WFK and read it in a run with less symmetry elements. But for the transition from AFM to magnetic it does not. Is there a bug in the code?

With my best regards, Donat

Post Reply