Dear Sahar, I had similar experience for my system. The directions of magnetization were changing during the scf cycles. What i think is that if the gain in energy as a result of changing the direction outweighs the energy cost of not respecting the penalty, the directions will get changed. Also, the constraint of the magnetization induces a very strong external magnetic field. You can try specifying angles without using the flag "constrained_magnetization".
The radius of the sphere is chosen to be a little smaller (i.e., divided by 2*1.2) than the minimum distance (between two atoms in your system). See PW/make_pointlists.f90 Hope it helps, Mighfar Imam JNCASR, Bangalore, India. > Hi pwscf users, > I am trying to do non-collinear calculation and > add a penalty functional > to the Hamiltonian in order to constrain > direction of magnetic moment of > some atoms in my desired direction. So I added > these tags to my usual > input file: > noncolin=.TRUE. > constrained_magnetization='atomic' > angle1(i)= desired amount in degree i=1,ntyp > angle2(i)= desired amount in degree i=1,ntyp > lambda=desired amount > as the result, I expect that the code will keep > the direction of magnetic > moments in desired direction and in every step > just changes the magnitude > of magnetization until the convergence achieves, > but I see the magnetic > moment of every atom is changing in both > magnitude and direction in every > step. > 1) Am I missing something in my input file? or > the whole procedure is > different from what I expected? > 2) How starting_magnetization will be related to > this constraining > process? > 3) How does code choose the radius of the sphere > around every atom for > calculating the charge and magnetic moment of > every atom? > I appreciate any help in finding the answers of > these questions. > ______________________________ > Sahar Mirshamsi > Quantum Theory Project (QTP) and Physics > Department, > University of Florida > _______________________________________________ > Pw_forum mailing list > Pw_forum at pwscf.org > http://www.democritos.it/mailman/listinfo/pw_forum
