Too much expansion of the unit cell : ReTe2

[AD]

Hi,

I have a ReTe2 unit cell with 20 Ang vacuum on top. I would like to analyze monolayer properties by geometry optimization.
My parameters are 0.01 ev/Ang max_force and 0.001 ev/Ang3 max_stress. As per one of the (atk forum) posts, 0.001 is good as a stress value. I use GGA, PBE with DFT-D3 correction. xyz directions are free to move and bravais lattice is unconstrained. Isotropic pressure = 0 is used. I found that after optimization, the unit cell has expanded a lot. The total z direction now stands at 42 Ang. Is it normal given the above optimization conditions? I see the same for some other materials as well. What could lead to such a high expansion? I was assuming that 20 Ang vacuum should be enough to avoid z-direction expansion. I was intially planning to fix z-direction but clearly that would have given me incorrect results (?). If I use a supercell for further study, I would have to deal with a huge cell. Any work around? Does it make sense to fix the z-direction now for the supercell to study interaction with molecules?

Thanks

[lknife]

I think D3-correction is not needed for the calculation of a monolayer. Maybe you can ignore the change of vector z since it can be set as you like.

[Petr Khomyakov]

I would guess that you may want to constrain the Bravais lattice type, provided that you have built the unit cell corresponding to the ReTe2 Bravais lattice of your interest, i.e., assuming that you are not looking for a different phase of this material.

What is definitely better to do for a monolayer geometry optimization is to constrain the Bravais lattice vector in the out-out-plane (z) direction. In this case, the vacuum thickness would not change, and you may avoid nonphysical geometry optimization along the z-direction.

[AD]

Thanks for your response.

@Petr I guess then I should keep it fixed (bravais lattice and z-direction) all throughout even when I am modelling interactions of molecules with ReTe2 (?) since I don't wish to obtain a new phase as a result of interactions. I am interested in total energy calculations.

[Petr Khomyakov]

Assuming that you are using periodic boundary conditions in the out-of-plane direction, you have to make sure that the vacuum thickness is sufficient to avoid a spurious interaction between periodic images of your molecule-substrate structure.

An alternative option is to adopt non-periodic boundary conditions by selecting the Poisson solver type to FFT2D and setting the Left(C) (Right (C)) boundary condition to Dirichlet (Neumann) type. It will not be as fast as using FFT3D, but might be a more accurate approach to your problem. You may give it a try for both options to see if there is any difference in the results obtained.

[AD]

Thanks for the suggestions. I will try them.