Energy change during optimization

[hfruchtl]

Probably a stupid question from a beginner... I am trying to optimize the geometry of GaAs with a cubic unit cell. I tried different algorithms, optimizing all positions or just the cell volume, and I always get roughly the same final energy, so I assume the result is correct. But if I grep the output for TOTEN, then the final value is not the lowest one. Is this not the value that's minimized?

Thanks in advance,

Herbert

[admin]

it depends on how EDIFFG is set: if >0, the total energy is minimized (please rather look at the 'energy without entropy' term or the sigma--> 0 term than for TOTEN). Please, in any case take care that the convergence criterium for electronic convergence (EDIFF) has to be at least one order of magnitude more accurate than EDIFFG.
if EDIFFG<0 the forces (and stress tensors) are minimized.
Actually, if you are in the equilibrium, both, the forces and the total energy should be at their minimum.
please also take care of the following
-- is each ionic step really converged electronically before the forces are calculated or did the electronic convergence procedure stop after the default number of el. steps (60)
-- if you do automatic geometry optimization (ISIF), please mind that you have to use high accuracy to get reliable results (PREC=accurate, ENCUT = at least 1.3*max(ENMAX) of the POTCARs you use
-- please also check for k-point convergence before you optimize the geometry.