GGA+U bandstructure calculation of monolayer mos2

[zmisha]

Dear Sir/Madam,
 
          I used GGA-PBE (no spin) to calculate the band structure of monolayer MoS2. The band gap is 1.689eV.
          We know that GGA and LDA underestimate the band gap. And people usually use GGA+U to get the correct band gap.
          Then I used GGA+U with U=3eV on d orbital of Mo atom. The band gap decreases to 1.566eV.
         Based on reference “Tuning the electronic properties of bondings in monolayer MoS2 through Au O co-doping”, the band gap should increase we use a positive U values. In this reference paper, they uses CASTEP and U(7.5 eV) for Mo 4d orbital (no spin). The band gap increases to 1.97eV.
         Although the software is different, but the general trend should be the same, I think.
         When I use U=7.5 eV in ATK, the SCF does not converge within 300 steps. 

         Could you please give some comments on this?
         Thanks.

[zh]

Please refer to the details of LDA+U (or GGA+U) methodology. The implementation of counting (calculating) occupation of local orbitals would also the affect the calculation results even if the same U is applied on the same orbital in the two different codes.

[Anders Blom]

Try a negative value of U instead. The definition of +U is not exactly the same in planewave and LCAO.

[zmisha]

Try a negative value of U instead. The definition of +U is not exactly the same in planewave and LCAO.

Hi,
     I want to mention that the ATK version I am using is 13.8.1, since our group only bought the license of this version.  Does this give a difference?
     I tried to use a negative value of U.
     Without U, the GGA-PBE (no spin) gives Eg 1.689eV.
     With U=-2eV on d orbital of Mo atom only, the GGA-PBE (no spin) gives Eg 1.698eV.
     With U=-3eV on d orbital of Mo atom only, the GGA-PBE (no spin) gives Eg 1.691eV.
     I tried U=-3eV or U=3eV on both d orbital of Mo atom and p orbital of S atom. The band gap still did not change much.

      Anyway, I did not see a large Eg increase of monolayer MoS2.

      However, I did  LDA+U (no spin) for orthorhombic phase of HfO2.
      Without U, the band gap is 4.03 eV.
      With U=5eV on d orbital of both Hf and O atoms, the band gap increases to 5.4eV.

     Could you give some suggestions on how to get a correct band gap(around 2eV) of monolayer MoS2 using LDA+U or GGA+U?   
     Thanks.

[zh]

The difference may be not related to the ATK version.

Please don't misuse DFT+U. Even some U value (in most case they may be very large) would give the band gap close to the experimental data, the physics meaning of this U may lose. To get a better prediction on the band gap, the DFT+U is not the best solution.

My suggest is to first try the use of linear response method in Quantum espresso or ABINT codes to estimate the U value for MoS2. The obtained U value would be more physics meaningful even such U values couldn't give the band gap to match the experimental results.

[zmisha]

Thanks.