I have been working on sesqui-chalcogenides, i.e., M2X3 structure where M=transition metal and X=chalcogen with VASP. After simulating the bandstructure for both the non spin-polarised and spin-polarized case, I found two different bandgaps that are far different. I tried to reproduce the data on Nb2Se3 and found the bandstructure and Density of States to be the similar. hence I hope my results are accurate. Only problem is I got a negetive value (-2.9 eV) for ISPIN=1 and 0.007 for ISPIN=2 and their TDOS is also different. I have attached the plots here.
Bandgap and TDOS are solely dependent on the crystal structure and material itself, aren't they? Then why do I get different results? Which one is accurate,polarised or non-polarized case? The situation really gets worse, as a negative bandgap might be an indication of topological insulation (band inversion). How do I know if the band is really inverted and it is topological insulator or it is metal and the negative value is due to the underestimation of bandgap by LDA and GGA?
Hello Farhat Tasnim
I never used VASP, but I've been using Quantum ESPRESSO for a while. Generally, DFT is known to no give correct bandgaps, so you must be careful about it, and trust experimental results.
About your calculations, I could try to replicate it in Quantum ESPRESSO to check what it gives, but I have some questions for you.
Did you optimize your structure before the bands calculations? Ignore this step gives completely inaccurate result, even for the expected inaccuracy of DFT bandgaps. You are using a lanthanide, so did you use Spin-orbit coupling method? For a lanthanide, this step cannot be ignored. And there are also the other regular parameters of DFT, like energy cutoff, k-points, convergence threshold, etc.
Could you provide the structure of Nb2Se3 you used?
Best regards,
Ricardo Tadeu
Ricardo Tadeu Maia
Yes I optimized my structure. And Nb is not a lanthanide and SOC was not included in paper I used to reproduce the results. Here I added the POSCAR and the paper.
Hello,
Sorry, I misread Niobium with Neodymium. So, I read the paper, and your band structure is very similar, you just need to adjust the scale of the graph. And probably you got confused because in the paper result, there are these "vertical lines", that is just something the author wanted to include, but if you ignore them, you can see your result is very similar.
About the DOS, I couldn't get the parameters the paper used, but I tried to replicate the non-polarized calculation in Quantum ESPRESSO. My result was this one attached. It doesn't have this much ups and downs, but I did get a result consistent with a conductor material. It looks like you have a "semiconductor" result, once your energy goes to 0 at some point close to fermi energy, and this is not what the paper shows. In Quantum ESPRESSO, this is adjusted by the occupations of the electrons. In QE, it's like it calculates the energies of the orbitals, and if you set the electrons occupations to be "fixed", it will really try to show a result where the HOMO and LUMO orbitals will not overlap. If you set the occupations as "smearing", kind of the HOMO and LUMO orbitals will overlap, and then you get a "metallic" behavior.
This is what I checked in some self made calculations with known materials like Silicon, in which if I didn't set the occupations as "fixed", QE returned a "metallic" behavior. Well, unfortunately, I don't know how to do it in VASP, but I think this is the point you should investigate.
Bests,
Ricardo
Actually, I focused on something else. Yes, the material is metallic which is evident from the density of states.I have also provided the bandgap relatted information in the first picture. The bandgap for both polarized and non polarized case are different although their bandstructure is same. Why this occurs ? please note that bandgap is shown to be negetive in non polarizd case.
For a "bandgap" to exist, there must be a gap between two bands, the valence and conduction bands. In metals, this doesn’t apply because there is no gap. I have never heard about this negative bandgap. I apologize if I am missing some concept, but to me, it's not correct to attribute a bandgap to a conducting material.
I don't know that much of polarization, but it would be better if you know previously if the material is magnetic, so you can turn the polarization on. But I really want to know how to determine if a material is magnetic with DFT.
- Badges
- Science topic
- Similar topics
- Physical Sciences
- Condensed Matter Physics
- Soft Matter
- Liquids