I am doing a graphene flake system and I am using Quantum ESPRESSO. And I must do some set of simulation variation. I was asked to find a stable configuration of graphene flake system (probably unterminated), then to rather add or subtract the atom one by one, probably. So the first question is how do I find stable configuration? Is that can be done by looking at the lower total energy? For example, if 13 atoms config is lower in energy than 15 atoms for certain config, is 13 atom more stable?
Other than that, I was asked to vary the cutoff radii of the pseudopotential, that is the inner cutoff radius (rcut) and outer (rcutus). I have done some calculation with ld1.x and I get a rough idea on how varying them. But here the problem comes out, how do I know the pseudopotential I used with certain radius cutoff is already optimized? Is it by comparing which total energy is lower at the final configuration in relax calculation? (just like I mentioned above?)
Would any kind of pseudopotential expert plis tell me what to do, in simple terms, so I would not get lost? I feel like I finally got an exit way, but I do not know what to do.
Thanks a lot for your attention and please tell me if I need to attach any file.
I can answer the first question, at least. The stable phase is the phase with the lowest energy per atom, rather than the lowest total energy. A phase with more atoms is very likely to have a lower total energy but that doesn't mean it's stable. For example, if phase 1 has 5 atoms and a total energy of -10 eV and phase 2 has 10 atoms and a total energy of -15 eV, in reality a piece of phase 2 will decompose into two pieces of phase 1.
I am not an expert on psudopotentials, but I don't think it's safe to assume that the pseudopotential which gives the lowest energy is most realistic. In your situation, I would assume that that a very small cutoff is the most accurate (it makes the fewest assumptions), and then increase it to see if a larger cutoff gives a good speed gain while still giving results very close to the small cutoff. As a check, you might see how well each cutoff matches the experimental lattice constant (bond length) of graphene. If someone who knows pseudopotentials better answers your question, feel free to disregard this advice.
Hi Steven Hartman
I know from the paper of Kresse-Joubert, "From USPP to PAW method" in 1999 (forgive me for the abbreviation) that I probably should take the upper limit of cutoff radii in 1.2*half bond length of C-C dimer, since I am simulating graphene flake.The paper said it this way, though, "The partial core radius is set to r_pc=r_rad/1.2", where r_pc is partial core radius while r_rad is the grid radius. It makes sense though when you think about it. I have tried to decrease the inner cutoff radii, but it leads to disaster.
The pseudo file I used is called C.pbe-n-kjpaw_psl.1.0.0.UPF. I was trying to get the comparison with atom O and N, which has smaller atomic radius. I reckon that the inner cutoff radii was not meant to be varied. By this in mind, I can say that I can vary the outer cutoff radii. It is said the cutoff radii are determined by the maximum of the radial wfc. I am using Quantum Espresso, the ld1.x probably could determine this so I can make the prediction on how I vary this, somehow, I am still uncertain.
In short, I have an idea, but trying it costs time and computational memory which is sucks in my place. I would be very grateful if you can probably point out some ideas or perhaps, it is much better if you can refer me to an expert of this, which is likely very considerate.
Thanks a lot