ABINIT Discussion Forums

Hello, this is my first post in the forum.

I'm relatively new doing PAW calculations. I have a question regarding crystal symmetries and the atomic basis: is the value of ecut dependent on them? Specifically for convergence of Etotal.

What I mean, do crystals with more complex symmetry (than, say, cubic symmetry), or those with more atoms in the basis require higher values for ecut? (Or for that matter, higher pawecutdg?) I mean, is a value of ecut higher than 30 hartree too high, say for example the ferroelectric LiNbO3 crystal?

Any advice would be welcome, and thank you in advance.

Welcome to the forum! Ecut depends on the PAW datasets, not on the complexity of the unit cell. Generally speaking, light atoms will need higher ecut's than heavy atoms, and also, the lightest atom in the sample will determine ecut for the whole sample (that is, you could have a bunch of heavy atoms and only a single light atom, and you will still have to use a large ecut due to the single light atom). So something like LiNbO3 will need a high ecut because of Li and O, not because of cell size. You could do some kind of As-Se glass alloy with a huge supercell, and need only a very small ecut because of the properties of As and Se. Of course, there would still be many bands and it would be a big calculation, but not because of ecut!

Hope this helps--

Thank you for your reply! And yes, it does help. I prepared some data sets for Li, O, and Nb (*not* with a Bloechl but rather a Vanderbilt scheme for pseudization) for a calculation on LiNbO3, and when I realized I needed and Ecut higher than 40 hartree to achieve convergence on the order of 1 part in 10^6 in the total cell energy I tought I might be doing something wrong. I guess I'll have to live with it hehe

Again, thank you very much.

40 Ha is very high. For O and Li you should be able to get to 20 Ha, though it is hard to get much lower. My general experience is that "custom rrkj" works best for main group atoms, and "custom polynom2 7 10 vanderbiltortho besselshape" works best for transition metals.

Thank you once again. I will certainly try your suggestion, since I did think it was a bit high at 40