Converging your DFT calculations

[handsomecat]

any CASTEP users or DFT planewave-pseudopotential method users out there?

how do I ensure that the basis set is adequate? I'm performing calculations with a rather
large unit cell ( orthorhombic, 8 atoms/cell) of a compound AB.

Can I justify my choice of energy-cutoff by testing convergence on a smaller cell of AB eg the B2 structure? then using the energy-cutoff that is found adequate for the small cell to do calculations for the orthorhombic cell?

 

[kanato]

The planewave energy cutoff is generally a property of the pseudopotentials you use. Increasing the cutoff adds higher frequency planewaves to the basis, which is generally only necessary where the potential varies rapidly, e.g. near the nuclei. You should check your final value in your full cell just to be sure your total energy is converged, but you can reduce the number of k-points to increase the speed of the calculation.

 

[Modey3]

handsomecat

To answer your question in two parts:

First. You need to make sure that the k-point mesh is adequate for the particular super-cell you use. Smaller super-cells require more k-points for Brillouin-zone integration. The reason being that the Brillouin-zone for a smaller super-cell is larger than a larger super-cell. For large super-cells only one k-point is required for integration since the Brillouin-zone is small. The width of the Brillouin-zone in one direction is \pi/L where L is the size of the super-cell. Remember that the super-cell is your basic repeated unit for this calculation, and hence governs the periodic boundary conditions.

Second. If you have plane-wave energy (Ecut) cut-off convergence for a smaller cell. You will have convergence for a larger cell using that cut-off energy. The reason being as you increase cell size the G-spacing in reciprocal space becomes finer giving you more plane waves.

Hope this helps.

modey3

 

[handsomecat]

Dear Modey3, thank you!

so suppose i wish to do calculations for A_x B_y, then I'd get a sense of what the adequate cutoff is by doing geometry optimization calculations for A and B in their ground states separately?

 

[Modey3]

handsomecat,

Since you are using psudopotentials the core potentials (i.e. potentials inside your cut off radius) for each atom will remain fixed no matter how you alloy your material. Your cut-off needs to just model the most rapidly varying core wave function of the two components, and how rapidly a core wave function oscillates depends on how rapidly the core potential varies. If you have two or more components in your system you need to use the highest cutoff that you determine for either pure A or pure B. Alternatively, just make your alloy and do a cut-off convergence test. Either way you will get the same results. I do a lot of Fe,Ni,Co,C, and N modeling. That is the rule I go by. You just have to make sure that your plane wave basis set is flexible enough to model the actual wave-function and therefore model the actual stationary eigenstates of the ground-state Hamiltonian. Beyond this I can recommend some helpful books that helped me in my studies.

Best Regards

modey3

 

[mayerz]

Hello everyone,

I have some problems with the convergence for the planewave cutoff energy, I tried to find the correct PWcE for a system with 2 atoms, I've determined the minimum k-point mesh with success, but when I compute the Total energy varying the PWcE, it seems that the convergence will be achieved till too high values of PWcE, should I change the pseudopotentials for the atoms?

I use DACAPO,

I'm Mexican by the way, so sorry if I made a mistake while I wrote this.

best,