I am a beginner using VASP to do material modeling. I have made a heterostructure model written in the form of a POSCAR file. But the supercell is too big (it contains 60 atoms) so that the computational speed is slow (unreasonably slow) even I put it into the cluster. How can I make an effective INCAR setting to do the calculation in structure relaxation and self-consistent calculation process?
-
1$\begingroup$ Can you supply an INCAR and some cluster details (CPU count etc) in your question? A representative POSCAR could also be useful (you can change the system if need be, just be sure it is vaguely similar). $\endgroup$– Tristan MaxsonCommented Oct 6, 2020 at 16:09
-
1$\begingroup$ As with most DFT calculations, symmetries will help reduce the computation load. With heterostructures, this can get tricky because symmetries of a monolayer may be broken when you include the other layer in the system. Have you looked at your VASP output to see if any symmetries were recognized? Apart from this, the general ways of speeding up the calculation is to reduce plane wave cutoff (ENCUT) and reducing density of k-mesh. $\endgroup$– Xivi76Commented Oct 6, 2020 at 21:38
1 Answer
Two things are important for the computational speed for VASP users. The first thing is ENCUT, which is controlled by the INCAR file. For a quick calculation, you can pick up this value just by referring to the ENMAX values of your POTCAR file. The other thing is the k sampling, which is decided by KPOINTS file. For your system with 60 atoms, I guess your $a$ and $b$ are very large and hence you can use lesser points for your simulations. For convenience, you can utilize the VASPKIT tool to automatically generate KPOINTS file with low/medium KP-resolved values for a quick calculation.
-
$\begingroup$ +1. By the way we have a VASP chat room now, can you come by and write "hello" at least once, so that we know you're interested in VASP? chat.stackexchange.com/rooms/109983/vasp $\endgroup$ Commented Oct 26, 2020 at 19:37