Dear all,
I read in some references that in order to include a charged defect in VASP calculation, we need to add electrons (holes) into the defect cell until the additional charge relax into conduction (valence) band. I have two questions:
Any favorable suggestions are greatly appreciated!
Thanks in advance!
Cheng
Reference: http://dx.doi.org/10.1016/j.jnucmat.2010.10.088
To study charged defect using the supercell method as included in many DFT packages, we can add or remove some electrons from the neutral supercell trought the parameter nelec (nombre of electron in the system). we should note here that adding electrons means a negative state for the defect while removing electrons simulate the case of the positive charged defects. in addition, the code will automatically add a backgroung jellium to make the supercell in neutral charge and the jellium charge is distributed in the voulme but it will be localized near the defects. the rest of the work will be done by the code to deduce the eignvalue as well as the defect related levels.
Dear Cheng,
I guess that VASP is some kind of Density Functional Theory Package. I have no experience with these calculations, but I have a word of warning, although it might be irrelevant (the effects are being taken care off). The calculations are only as good as is the Hamiltonian used as the starting point. Charged defects in semiconductors (and insulators for that matter) are often made of mobile electrical charges (electrons and/or holes) strongly coupled to the lattice. This electron-lattice interaction is particularly important when discussing situations with strong localisation. If it is not properly treated, some strange results (well, wrong) can result. My question to you therefore could be, can your package calculate a single polaron self-trapping states ? If yes, then I think you are all right, if not, I would be worried.
With bet regards
Petr
Calculation of charged defects is a topic which has gotten much interest in recent years. Unfortunately there are a lot of problems which pop up when you want to calculate such systems using ab-initio methods. Often the charge needs to be localized (Note that Gaussian basis sets will do this for the wrong reason, eg: J. Comput. Chem. 34(5), 422-427 (2013),) which can be hard for standard DFT with a plane wave basis-set (such as VASP). Depending on the materials you are working with, it may be interesting to check if DFT+U methods are used (required) for one or more of the composing elements. If so, using them can give rise to the required localisation. When I was studying a MOF which contained TiO-clusters including +U parameters for Ti resulted in the experimentally observed localisation, which was nicely visualized by plotting the atomic charges calculated using the Hirshfeld-I atoms-in-molecules partitioning scheme (Hirshfeld-i: J. Comput. Chem. 34(5), 405-417 (2013), TiO-containing MOF, in the SI of Angew. Chem. Int. Ed. 54(47), 13912-13917 (2015)). This just to give a picture of the localisation issue. Luckily VASP is rather good at handling such systems, once you know how to get the required localisation.
For practically calculating the defect formation energy in case of charged defects, there is a really useful paper by Lany and Zunger on this:
Phys. Rev. B 78, 235104 (2008)
You may also want to have a look at the review paper by Kresse of the First-Principles calculation of point defects in solids:
Rev. Mod. Phys. 86, 253-305 (2014)
These two papers will help you out a lot, mainly with regard to the title question. The first question would be answered with setting NELECT as you state, but in addition, you may want/need to play with NUPDOWN, MAGMOM and the DFT+U related parameters. For your second question regarding the evaluation of the localisation, there I think Hirshfeld-I charges will be of use, as they will give you a quantitative picture.
Danny
As far as I know VASP does not have an implementations of Lany Zunger corrections. One can also use the FNV corrections, which is a more versatile approach since it is extendable to low-dimensional systems as well. FNV has been recently implemented in the CoFFEE code:
Article CoFFEE: Corrections For Formation Energy and Eigenvalues for...
Code: http://www.physics.iisc.ernet.in/~mjain/pages/software.html
- Badges
- Science topic
- Similar topics
- Physics
- Condensed Matter Physics