ABINIT Discussion Forums

Hi folks!

Using ABINIT, I was trying to find the optimized geometry of the CrO2Cl2 molecule. I found some old literature values (somewhere from the 70s) that I used as a starting point. After convergence testing (i.e. energy cutoff and cell size), I first tried using the BFGS scheme (ionmov 2) for the optimization. But after the results did give any clues (moving of positions mostly happened in one direction only), I started the same calculations, only this time with MD+Verlet scheme (ionmov 6), but here, no actual ion movement occurs.

I have uploaded the input files and the corresponding results so you can have a look at them. (Dataset 1 was the literature given geometry, 2-4 are modifications of this by changing the bond angles. Actually, some forces should occur driving the constituents apart into the positions of dataset 1 assuming this is the minimal energy configuration.)

Is it possibly necessary to state the degrees of freedom or disable symmetry calculations?

Thank you very much for your help!

Best regards,
Oliver

P.S.: bfgs.out was too large to attach (~450KB), but I can mail it if someone wants to have a look at it

P.S.:
It seems adding

nsym 1
chkprim 0

to the input file solved the problem. The coordinates are now moving freely.

Just wanted to give some update,
Oliver

ok, well if it's working great - nsym 1 probably made some forces small but /= 0 and that started the motion. Note that ionmov 6 is MD, not relaxation, since there is no friction term. Is there a specific reason you're not happy with BFGS? It is usually the most efficient.

Matthieu

Dear Matthieu,

thank you for the hint. I just saw, that ionmov 1 should be relaxation with some friction. Nope, the only reason to go away from BFGS was, that I was not sure, whether my problems were related to it. After adding the options I mentioned before, the Broyden scheme also works as I expected.

Thank you and
best regards

Oliver