For geometry relaxation ISIF 3 or 4 is best. I prefer ISIF 4 in which volume remains fixed and shape varies by varying different volume you can observe corresponding energy. By plotting E-V curve you have to look at which volume, energy is minimum and corresponding pressure should be also zero or equivalent to zero that is your optimized structure.

Sometime it does not happen in that case you have to look other parameters like ENCUT  and KPOINTS. You can increase ENCUT upto 1.5 times maximum value of ENCUT in POTCAR among all elements in your system. If it also does not work then you have to look other ISIF value (3) in which you can vary volume also.

So you have to optimize your system by different ISIF value. You can compare your optimize system with experimental values for pure unit cell. I am not getting comparison of supercell. For lattice constants you have to compare only unit cell.

Hi Shilendra,

Thank you for the comprehensive reply. Do mean that in optimizing the structure, I should aim for zero external pressure in the system? And am I correct that I can find this external pressure line in OUTCAR file?

I am confused with the first statement in the second paragraph. What do you mean by "it does not happen"? Do you mean that if pressure does not go to zero?

You also mentioned about increasing the ENCUT 1.5 times the maximum value of ENCUT in all the elements in the system. May I ask why 1.5 times? In the manual, it was recommended to use something like 1.3 times to avoid problems with Pulay stress, (although I do not fully understand what Pulay Stress is.)

Also, I have another question. If one is to obtain the binding energy of an adsorbate, one will need to get the energy of system, the adsorbate, and the substrate. After which, one should subtract the adsorbate energy and substrate energy from the system energy to get the adsorption energy. For this adsorption energy to be meaningful, important parameters should be the same for all three such as ENCUT, K Points, etc. I am assuming that supercell size should also be the same among the  three. So when getting the energy of the three (substrate, adsorbate, and system), one should not allow the supercell shape and volume to change right? In this case, would the best ISIF setting be ISIF=2 when getting the energies individually for the three? Or would it be better to just use IBRION=-1 and not allow the system components to relax after getting the optimized geometry?

Thank you very much.

http://cms.mpi.univie.ac.at/vasp/guide/node112.html

http://cms.mpi.univie.ac.at/vasp/guide/node161.html#pullay

Yes, you should do zero external pressure. VASP calculate pressure after each ionic iteration for example if you have kept NSW=5 then it will give 5 time pressure and 5th one is the final pressure in the system for that particular volume. If you change volume then pressure will change. If you are increasing volume then pressure will decrease. You can find out pressure by typing command grep pressure OUTCAR then it will show pressure.

IN VASP manual it has written take ENCUT 1.3 times but I am taking in some cases when system is large then people use 1.5 time also for good convergence.

INCAR parameter depends on your system if your adsorbate, substrate and system having same geometry then you can put same parameter.

For pulley stress there is nice discussion on researchgate you can search in Q.A. section.

I see. Thank you very much for your answers.