# In VASP, which option is better for monolayer relaxation: ISIF=3 or ISIF=4?

Since I have started working on monolayers I was using the ISIF=3 tag for relaxation, this was after recompiling VASP with the modified "constr_cell_relax.F" file. There was no problem in relaxing pure materials, but with the doping of 4x4x1 supercells I don't get convergence in relaxation and the calculations are taking too too much time. On the other hand my friend is using the tag ISIF=4 and he is getting always converged results.

Now I am a little bit confused with which tag should be used, since I think that because of vacuum along the c axis there should be no relaxation along it, so ISIF=4 tag shouldn't be used. I want to know your opinions, especially if you have worked on relaxing doped monolayers.

• The convergence problems may be related to ALGO and IBRION, rather than ISIF.
– Jack
Feb 17, 2021 at 4:31
• Take a look at this post: mattermodeling.stackexchange.com/questions/2022/…. By the way, the ISIF=3 tag with recompiled VASP [ modified constr_cell_relax.F] is suitable for monolayer relaxation. Also this post: mattermodeling.stackexchange.com/questions/2032/…
– Jack
Feb 17, 2021 at 10:46
• @NikeDattani Thanks for your suggestions. Later I will post an answer.
– Jack
Sep 5, 2021 at 23:41
• @Jack any luck coming up with an answer. Also, I was able to edit your 3 comments into just 1.
– Tyberius
Nov 19, 2021 at 2:17
• @ChiKou were you able to figure out any of this? Is there any update you can give us? Jan 13 at 0:53

When dealing with monolayers, I have found that using the ISIF=4 tag is better and is less time-consuming. As mentioned in Vaspwiki, it allows the relaxation of both lattice constants and atomic coordinates with a fixed volume. However, what is important here is uncommenting the FCELL(3,3)=SAVE line in constr_cell_relax.F file and recompiling VASP. This trick prevents the relaxation along the z-direction, which is negligible in monolayers.