[gpaw-users] Lattice parameter with GGA or GGA+U ?

[jun yan]

Hi, 

     I am trying to get the lattice parameter for VO, which has a very simple structure of atoms = bulk('VO', 'rocksalt', a=4.46)
The following is what I got with PBE using spinpol=True.  The OV_0.90000 means scaling lattice parameter a=4.46 with 0.9 . 

OV_0.90000.txt:Total Magnetic Moment: 1.956392
OV_0.95000.txt:Total Magnetic Moment: 2.448952
OV_1.00000.txt:Total Magnetic Moment: 2.600094
OV_1.05000.txt:Total Magnetic Moment: 2.638621

OV_0.90000.txt:Free Energy:    -16.24786
OV_0.95000.txt:Free Energy:    -16.24359
OV_1.00000.txt:Free Energy:    -16.20235
OV_1.05000.txt:Free Energy:    -16.11876

so as the lattice parameter becomes small, the total energy decreases with decreasing magnetic moment. But this is wrong. The expected lattice parameter will be around a=4.46, (scaling factor 1.0) and with magnetic moment 3.0. 

My question is should I use GGA+U to get lattice parameters ?  If so, should I use the same U for all the scaling factors ? I expect that if the cell becomes smaller, the d-electrons interactions change and thus requires a different U…  and even if I use the same U, would different U gives different lattice parameter ?

 I appreciate if someone with more experiences would like to shine some light. 

All the best, 
Jun