Calculation ends before EDIFFG is reached

Message

ajhatt · #1 Post by **ajhatt** » Wed Jul 11, 2007 5:17 pm

I am doing a spin-polarized calculation with LDA+U for a 10 atom system.

To optimize the atomic postitions, I am using the settings:
EDIFF = 1e-5
EDIFFG = -1e-4
NSW = 200
ISIF = 2
IBRION = 2

However, the calculation terminates after a small number, ~20-50, of iterations while the largest magnitude total force on an atom is larger than 1e-4 eV/Angstrom. For example, it may quit after 24 iterations while the force on a given atom is -0.03 eV/Angstrom. My understanding, based on the manual, is that this calculation should not stop until the absolute value of the largest force is smaller than 1e-4 eV/Angstrom.

Please could you help explain, why might the calculation be stopping before EDIFFG is reached?

Thanks for your help,
AH

#2 Post by **admin** » Tue Jul 31, 2007 8:59 am

did you use selective dynamics? If so, the forces on the atoms which are kept fixed are not considererd when checking the convergence

ajhatt · #3 Post by **ajhatt** » Tue Aug 14, 2007 4:46 pm

Thank you for your response.

No, I didn't use selective dynamics.

ajhatt · #4 Post by **ajhatt** » Mon Sep 03, 2007 9:59 pm

Is there any other information I could give you about my problem that might help suggest an answer?

Also, does anyone else have this problem?