Dear Deepak,

Generally speaking, the plane-wave cut-off energy is determined by the elements (i.e. pseudopotentials) you have in the simulation, and the k-point spacing is determined by the bonding and electronic properties of the simulation system itself. This means that if you have a perfect supercell you can use the same cut-off energy as for the primitive cell, and the same k-point spacing as the primitive cell (which won't lead to the same k-point grid, because the supercell is larger).

*However* if you change any of the elements (or pseudopotentials) then you need to check the cut-off energy again, and if you change the bonding and electronic properties (e.g. have a surface, interface, vacancy, different elements) then you need to check the k-point convergence again.

Checking the cut-off energy again is fairly straightforward: because it only depends on the elements, you can determine an appropriate cut-off energy with any simulation using those elements, it doesn't have to be the actual supercell. In fact you can often manage doing pure elemental materials for the cut-off energy convergence, and then choosing whichever is the highest cut-off energy needed by the different elements.

Checking the k-point sampling is more computationally demanding, because it actually requires calculations on the full simulation cell.

Hope that helps,

Phil Hasnip

(CASTEP developer)

Dear Deepak,

Generally speaking, the plane-wave cut-off energy is determined by the elements (i.e. pseudopotentials) you have in the simulation, and the k-point spacing is determined by the bonding and electronic properties of the simulation system itself. This means that if you have a perfect supercell you can use the same cut-off energy as for the primitive cell, and the same k-point spacing as the primitive cell (which won't lead to the same k-point grid, because the supercell is larger).

*However* if you change any of the elements (or pseudopotentials) then you need to check the cut-off energy again, and if you change the bonding and electronic properties (e.g. have a surface, interface, vacancy, different elements) then you need to check the k-point convergence again.

Checking the cut-off energy again is fairly straightforward: because it only depends on the elements, you can determine an appropriate cut-off energy with any simulation using those elements, it doesn't have to be the actual supercell. In fact you can often manage doing pure elemental materials for the cut-off energy convergence, and then choosing whichever is the highest cut-off energy needed by the different elements.

Checking the k-point sampling is more computationally demanding, because it actually requires calculations on the full simulation cell.

Hope that helps,

Phil Hasnip

(CASTEP developer)

Dear Philip James Hasnip

That is very helpful thank you, I will try out your strategy now.

Regards

-----

Deepak

Dear Deepak

As Philip said, the cut-off energy depends on the pseudopotential, so you just need to find optimum cut-off for that pseudopotential in a simpler crystal or for a dimer or just in an elemental calculation.

Julen has provided a simple script for finding this optimum cut-off energy:

http://larrucea.eu/checking-optimum-cutoff-qe/

check the pseudopotentials, check the cutoff inside and take the large one cutoff, after that make a convergence test varying the cutoff value

 for every dopant or foreign atom you must check into the correspondign pseudo to the sure abour the convergence, if the cutoff is lower it will be fine, if not them always take the largest one, even you can increase it adding some amount 

Hi dears,

Would you please answer my question,

Is it necessary to converge mesh cut off and K-points parameters for whole atoms in a super-cell, or convergence for unit cell is adequate?

Dear Milad,

I'm not quite sure what you mean by "whole atoms", but I think you're asking whether you have to do all the convergence tests for a supercell, or whether you can do this in a unit cell. This question has mostly been addressed in earlier posts, but I'll try to summarise here:

The plane-wave cut-off energy (is that what you mean by "mesh cut off"?) can be determined from any simulation containing the same pseudopotentials (see earlier posts for a longer description). The unit cell would be fine.

For a perfect supercell, you can use the same k-point spacing as the unit cell (spacing, not mesh) - but this is not very common, because if the supercell is perfect then it would be much quicker to simulate the unit cell instead, and not bother with the supercell at all!

If the supercell is not a perfect replication of the unit cell, e.g. you've introduced a point defect or a grain boundary, then you will need to re-check your k-point sampling for the supercell itself. This is because the k-points are sampling your electronic structure, so anything which changes your electronic structure could change the number of k-points you require. However you could still start with the k-point spacing determined from the unit cell, and then increase the grid from there, so it isn't as costly as doing a full k-point convergence calculation.

Many codes have tools to automate some or all of the convergence process, e.g.

https://github.com/CCP-NC/castepconv (for CASTEP)

http://larrucea.eu/checking-optimum-cutoff-qe/ (for QE)

Hope that helps,

Phil Hasnip

Deer Deepak:

it's recommonded that keeping all the parameters constant during the whole calculation as possible, so you can choose a ecut that suitable for all the composition elements first , this should be carefully tested.