Hey SK M

I'd use the 'relax' method for the individual components. Using 'vc-relax' could indeed change the vacuum spacing and distort the environment, which would affect binding energy calculations. For a fair comparison, use only the 'relax' mode.

Rgds

Gabriel Vinicius

Hi SK M

You can indeed use vc-relax in Quantum ESPRESSO as well. The newer versions of QE have an option 'cell_dofree' under &CELL block to keep a certain dimension or shape of the system unchanged. For example, if your vacuum is along the z direction, you can define cell_dofree = c. You can also combine multiple options like cell_dofree = ibrav + c, to keep the z dimensions of the unit cell unchanged and also the lattice shape.

I hope it helps.

Yuvam

Gabriel Vinicius and Yuvam Bhateja thank you for your answers.

Yuvam Bhateja if we fix the way you mentioned as freeze z-direction, cell shape also unchanged, then isn't it same as using 'relax' option?

SK M

Well not really, if I want to relax my lateral cell dimensions while keeping say my orthorhombic unit cell shape, it is useful and entirely different from the normal 'relax' calculation that does not change cell structure. Anyway, there is a lot of freedom with how you use this input.

Gabriel Vinicius and Yuvam Bhateja by the way, does one need to do the optimisation in the sequence? as in first 'SCF'-->'Relax' -->'vc-relax'? is this a must? kindly give some details, please.

Hey SK M

I wasn't aware of the feature Yuvam Bhateja mentioned earlier. I usually stick to the basics in QE - I mainly use VASP instead. As for your latest question, I don't think 'SCF'-->'Relax' -->'vc-relax' is a must. If I were unsure, I'd take the following empirical approach:

(assuming the question is whether or not to relax the cell)

1. Try 'SCF'-->'Relax'-->'vc-relax'

2. Try 'SCF'-->'Relax'

Then, compare the results to see which one makes more physical sense for the system you're modeling, or if there’s any significant change and make your decisions accordingly.

Best,

@Yuvam bhatija, Gabriel Vinicius thank you for the reply.

by the way, after SCF step, how to update the output coordinates into the input files of 'relax' and then 'vcrelax'? i did many years before ( a little of QE) and do not have previous data to check the sequence of steps. I also mostly use VASP.

SK M

If you are doing SCF first and then relax/vc-relax, you will keep the same coordinates for the input files of relax/vc-relax that you used for the scf input. On the other way where the coordinates and the cell dimensions change through relax/vc-relax calculations, you can simply copy the updated coordinates and cell dimensions from the relax/vc-relax output files for any subsequent calculations.

Personally, for a bulk or periodic system I prefer to do relax/vc-relax first and the scf, as electronic properties will change if your system geometry is under stress. But, it entirely depends on what you intend to do.

Yuvam Bhateja based on your discussion and what I understand, the SCF calculation is for finding the ground state energy of the system. In such case under the following scenarios i express my opinion, please comment that:

1. If I do SCF first and then do relax and then vc-relax: Its meaning less to do SCF first, as relax and VC-relax would change the geometry and thus the ground state energy. Will there be any situation we need to know the SCF energy first and then do relax and vc-relax?

2. If i do relax and take revised coordinates as input to vc-relax: This seems logical to me to do geometry optimisation steps correctly. comment on this please.