# Vacancy with USPEX

I am using USPEX code. I would like to create structures with vacancies. I defined in the file "INPUT.txt" vacancy as Vc, but it stopped and said:

  'Element Vc does not exist. Please go back to check it again.'


Do you know how to define vacancy in USPEX? I did not find any description in USPEX manual.

• What kind of calculation are you attempting? Do you have seeds or just randomly generated structures? – Hitanshu Sachania Oct 22 '20 at 10:12
• My concerning compound is AxMeO2, with x from 0 to 1. I have a seed structure. – Binh Thien Oct 22 '20 at 10:45
• USPEX is a structure predictor so does it accomodate vacancies. Vacancies can lead to a disruption in the lattice periodicity. So I'm not entire sure if its the package you should use when youre intersted in introducing vacancies. – Anoop A Nair Oct 22 '20 at 13:35

This seems like a restricted crystal structure prediction problem since knowing the crystal structure of the system beforehand becomes a requirement. You must specify a space group (or other symmetry groups like the point, layer, etc groups) via the symmetries tag in the INPUT.txt file. Given that, there arise two scenarios:

1. When you know exactly how many vacancies there are.
2. When there can be a variable number of vacancies.

Let's say your material is $$A_2B_3O_{10}$$ and you wish to study $$O$$ vacancies.

Scenario 1: Fixed composition USPEX calculation.

calculationType = 300


Mention less number of $$O$$ atoms in numSpecies in accordance with the number of vacancies the material has. For instance, when there are two $$O$$ vacancies:

%atomType
A B O
%EndAtomType

%numSpecies
2 3 8
%EndNumSpecies


Scenario 2: Variable composition USPEX calculation.

calculationType = 301


Let's assume there could be up to three $$O$$ vacancies:

%atomType
A B O
%EndAtomType

%numSpecies
2 3 7
0 0 1
%EndNumSpecies


All predicted structures would be a combination of $$x\cdot(A_2B_3O_7)$$ and $$y\cdot O$$ i.e. $$A_{2x}B_{3x}O_{7(x+y)}$$. This way you'll get the best structure among all with $$O$$ vacancies ranging from one to three.

I never have used USPEX, but here is a bad solution to the problem even if it doesn't support vacancies. Use a dummy element such as Na that is not in your system. Post-process the structure afterwards to remove it, using something like ASE. This is definitely not ideal, but you might find that it works to generate reasonable structures.

USPEX might use the size information of the atom to fit them (I am unsure), so if it does be sure to pick an element with a similar size or charge etc to the atom where the vacancy would have formed. This ensures it comes up with reasonable starting guesses.

• I read ASE manual and I found that it is not easy to use like USPEX. Also, I found that USPEX needs ASE for running, then maybe USPEX has a similar algorithm. I do not find any example ASE manual directly related to my case. – Binh Thien Oct 22 '20 at 14:07