Hi
i have a nanoparticles with ligands structure with about 600 atoms in the system.
while geometry / energy optimisation, initially the energy is reducing and suddenly increased and after few steps reduced again. i am continuing the previous step with CONTCAR to POSCAR option. So, far 2 parts completed and the energy trend is plotted and attached, here.
Doing Gamma point only optimisation with VASP_gam .
Can anyone suggest what could have gone wrong?
Thanks
You could have a system with more than one quasi-stable electronic states. During the course of the geometry optimization, the system could be fluctuating to another electronic state that is not the ground state. Do you notice significant changes in the geometry in the ionic steps where the energy changes rapidly? One thing you might try is specifying the blocked Davidson solver to be used instead of RMM-DIIS. The advantage of the Davidson solver is it tends to be more numerically stable... the disadvantage is it normally doubles the computation time relative to RMM-DIIS.
The calculation appears to be jumping between two different electronic states as noted by Stephen Giles. As the structure changes, the orbitals are moving around in energy. If the HOMO & LUMO flip order, or the HOMO and HOMO-1 flip order, their occupancy will change, placing the system in a different electronic state. This problem is especially common in small band gap materials. Increasing the Fermi smearing can alleviate the problem. Increasing the number of k-points can as well, but that comes with increased computational expense, which is likely an issue with a 600+ atom system. If the gap is very small, you might try running as a metal. I would: 1) increase the Fermi smearing. 2) Try a different optimizer as suggested by Giles. 3) Try running as a metal if the bg is quite small. 4) Increase the number of k-points.
One curious thing, I've not seen a case where the program takes so many geometry steps between state changes. Hopefully that means it is close to stable and a small tweak will get the calculation to stay in the correct state.
Thank you both for detailed answers.
Sorry Stephen Giles , I did not notice your reply earlier, not sure how I missed it seeing.
its quite some time I had that issue, and not sure how did I resolved it, but I thing I changed the IALGO=38, instead of 48. This is also your suggestion. I need to check your Check point as you mentioned ..."...Do you notice significant changes in the geometry in the ionic steps where the energy changes rapidly?" This is an interesting check I think. I need to go back and see the data, if I still have the data. but I don't understand the point "more than one quasi-stable" electronic states. I am curious on to know this aspect. (--For the benefit of all readers, I would say here-----) because of the reason my system has two nanoparticles (NPs) separated by a distance but linked to ligands in between, and I am trying to optimise to get total energy.
as Karl Sohlberg also mentioned similar opinion, and my energy profile sketch drawn w.r.t distance between NPs is attached. I am curious to know, does the profile reflects your opinions on more than one electronic states? if so, I wish to understand more on this. I also tried change of ISMEAR value to fermi smearing as Karl Sohlberg mentioned, and finally resolved to my satisfaction. frankly not sure which combination of tags solved the issue, I generally meddle with ISMEAR, IALGO, and Mixing parameters (AMIX, BMIX, IMIX). also as he mentioned to run it as a metal option. How could we do that?
await your reply to know more on the issue and for others as well.
Thanks. :)
- Badges
- Science method
- Similar topics
- Methodology
- Crystallography
- X-ray Diffraction