My Community

Hello VASP users and admins!
I am in need of an urgent advice to whether use VCA calculations for my system or not.

I am trying to use VCA for a system that has 6 elements: O Nb W Ta Va Mo
The VCA tag for this is : 1.0 0.2 0.2 0.2 0.2

I use the recommended VASP potentials from PAW_PBE.
I do see electronic convergence in the first step but as soon as the system wants to go through with first ionic relaxation, the energy levels are reported as positive values and the reported force is very high. This keeps happening at every ionic relaxation after that.
What do you think the reason might be?

Could it be because of the 4 elements that are sharing the same spots? I have done VCA runs with 2 elements having partial occupancy and everything went fine.The example on VASP manual is for a system with two types of elements and sharing the same positions in POSCAR.

On a separate note, I have never done a VCA calculation but I think the positions of elements that have VCA value Lower than 1, should be duplicated in POSCAR file. Is that correct?
This is based on what I see from the manual here: wiki/index.php/VCA

Hi,

Thank you for your inquiry regarding your virtual-crystal-approximation (VCA) calculations. To provide you with a more detailed response, it would be beneficial to review the input and main output files. Please share these files as per the forum guidelines.

Here are some key points and suggestions based on the information you provided:

• You mentioned setting VCA = 1.0 0.2 0.2 0.2 0.2 for a system with 6 elements. The VCA tag must consist of a list of 6 floats, each corresponding to the weight of an element. Please ensure the tag matches the number of elements in your system.

• You mentioned that you've never done a VCA calculation before. It might be helpful to start with a simpler example. The Wiki provides an example of simulating Sn doping in a Ge lattice, which you can use as a reference to understand the method better.

• Your understanding of duplicating positions for elements with VCA values less than 1 is correct. Elements with VCA values less than 1 should be duplicated in the POSCAR file to represent their fractional occupancy in the virtual crystal structure.

• As described on VCA tag documentation. There are many caveats to the method. Particularly, the pseudo-orbitals must be very similar for the potentials that are "mixed" and semi-core states that are treated as valence states represent an obstacle. Check if the pseudopotentials that you selected contain semi-electron states and start by substituting each element one by one in your calculation to see when the calculation becomes unstable.

Finally, you may consider constructing a supercell that corresponds to the desired alloy. This is computationally expensive but might be the only way to obtain reliable results for such a complex composition of elements.

Thank you for you reply. I have attached a zip file that includes 3 folders named,"bulk_relax_v80" and "bulk_relax_v87" and "bulk_relax_v88" and "bulk_relax_v88_non_sv_pv".
1) "bulk_relax_v80": (Failed)
This folder has the VCA run for the system that has all the 6 elements Nb W Ta Mo V O with Nb W Ta Mo V having partial occupancy.
The VCA tag for it would be: VCA= 0.2 0.2 0.2 0.2 0.2 1.0
Observation: In the first ionic step, electronic convergence is reached and the force is calculated to be g(F)= 0.116E+03 which is huge in my opinion.
In the second ionic step, the energy level is positive and stays positive during all the electronic steps. And from there algorithm has a hard time relaxing the system.
These are the POTCARs that have been used in this run: (Just as VASP manual recommends)
POTCAR: PAW_PBE Nb_sv 25May2007
POTCAR: PAW_PBE W_sv 04Sep2015
POTCAR: PAW_PBE Ta_pv 07Sep2000
POTCAR: PAW_PBE Mo_sv 02Feb2006
POTCAR: PAW_PBE V_sv 02Aug2007
POTCAR: PAW_PBE O 08Apr2002


2) "bulk_relax_v87": (Successful)
This folder has the VCA run for the system that has all the 6 elements W Mo O with W Mo having partial occupancy.
The VCA tag for it would be: VCA= 0.5 0.5 1.0
Observation: This run has no problem and reaches the threshold for the Force and relaxes fully and all the energy values at all the steps are reported as negative.
These are the POTCARs that have been used in this run: (Just as VASP manual recommends)
POTCAR: PAW_PBE W_sv 04Sep2015
POTCAR: PAW_PBE Mo_sv 02Feb2006
POTCAR: PAW_PBE O 08Apr2002

3) "bulk_relax_v88": (Failed)
This folder has the VCA run for the system that has all the 6 elements W Mo V O with W Mo V having partial occupancy.
The VCA tag for it would be: VCA= 0.3333 0.3333 0.3333 1.0
Observation: This run has similar fate as the "bulk_relax_v80". The Force value is calculated to be g(F)= 0.897E+01. In the second ionic step, the energy level is positive and stays positive during all the electronic steps. And from there algorithm has a hard time relaxing the system.
These are the POTCARs that have been used in this run: (Just as VASP manual recommends)

POTCAR: PAW_PBE W_sv 04Sep2015
POTCAR: PAW_PBE Mo_sv 02Feb2006
POTCAR: PAW_PBE V_sv 02Aug2007
POTCAR: PAW_PBE O 08Apr2002

4) bulk_relax_v88_non_sv_pv": (Failed)
This run is exactly like "bulk_relax_v88" or case number 3 but the POTCARs are different.
These are the POTCARs that have been used in this run: (PAW_PBE POTCAR files that don't have the s_v or p_v extension)

POTCAR: PAW_PBE W 08Apr2002
POTCAR: PAW_PBE Mo 08Apr2002
POTCAR: PAW_PBE V 08Apr2002
POTCAR: PAW_PBE O 08Apr2002



I am trying to understand how I can make this work. It seems that as I include the third element, then everything falls apart and energy values go to positive when the relaxation wants to begin.

Also, I want to know in these cases where ionic relaxation fails, is the energy level that is reported in the first step reliable and the only problem is the the ionic relaxation.

This brings another question and that is why that the first step goes very well but all the other steps have positive energy values. Could this only be an issue related to ionic relaxation?

What changes can I bring to INCAR or POTCAR that could give us more insight to solve this problem?

I appreciate it.