Dear friends,
I have tried to get self Consistent band structures of layered material by using VASP, PBE , but I got the underestimated values.
I have repeated the calculations with LDA but am getting the same result.
Can anybody tell me what kind of issues will effect the band structure? Like strain, lattice constant etc.
And which functional will give proper band structure calculations?
Thanks
anji
The essential exchange and correlation effects are crucial to the proper band structure. As these quantities can't be calculated exactly, we use approximations. The Up to till, the approximation which is very useful in calculating the band structure and band gaps closer to the experimental perspective, is Modified Becke-Johnson Ex-Cor approximation. It has been cited more than thousand times since the first appearance in 2009. The original paper for this approximation is "F. Tran, P. Blaha, Phys. Rev. Lett. 102 (2009) 226401".
DFT is a ground-state approach, as such the bandgap is by definition wrong (underestimated) As such, a functional in general is not the solution.
LDA and PBE will in general give you bands with the shape that are to be found in the 'real band structure", however their gaps are too small. Solutions are: using DFT+U (if you know what the size of your band gap should be, you tune U untill you get a match.(this approach cost you not significantly more than normal DFT). You could move to a hybrid functional such as HSE06 which is rather good for solids (though the calculation time goes up by a factor of 100 more or less :s ). Alternately you could check out the GW approach. Of course there are also semi-empirical approaches such as MBJ. The most rigorous (least cheating) approach to me seems GW. Hybrid-functionals and DFT+U both are fitted, (as is MBJ, albeit even worse)
I agree with Danny in general.
These 'beyond' DFT approaches can be used for different systems:
For systems with localized states, such as 3d, 4f, the DFT+U works well though you have to work out the U values for the localized orbitals first.
For s, p semiconductors/insulators, both hybrid functional and GW methods work well. One difference is that a hybrid functional requires some parameters while GW can be regarded as parameter-free. The weak point is the both hybrid functional and GW are very computer- and time- costly.
If this is a layered materials, VdW corrections should be important and you can can get a change in the cell parameter of 10-20%, enough to change the electronic properties.. you can include in Vasp VdW selfconsistently or by using a Grimme like approach... in any case, I would consider this before any DFT+U or anything like that... geometry can be far... you can see those effects in known layered materials such as MoS2 or WS2..
Let me be bit more specific. If you are interested in structure, include VdW... then you can fix that structure and do more advance calculations if the point is to get good optical spectra, exciting effects, etc... for such you can use GW, Bethe-Saltpeter or as it was stated, HSE functionals.. in any case, all of these methods are much more expensive than normal DFT. Of course, you can be more "accurate" if you reoptimize with HSE, but as I said before, it is more expensive.
Hello,
When I draw the band structure by mbj-vasp I don t find the band gap any more, while PBE gives a band gap of 0.9 eV for ZnO. is there a special way to define the k point patch for non-scf band structure calculations?. I note that the dos seems to be ok by mbj-vasp
Bests regard!
- Badges
- Science method