DFT+U in two probe system

[rw yan]

Hi ATK-VNL team:
        When I calculate the transmission spectrum of BDT at 0V bias in two probe system (the electrodes are gold), the value of it at E_f is larger than experiment about two order of magnitudes due to the underestimation of HOMO-LUMO gap in DFT calculation. So I wonder if DFT+U can imporve the calculation, if yes, then I have two questions:
        (1)How to set Hubbard U parameter in ATK? For example, there are 3s3p4d for S atom , shoud we set U parameter only for 3p orbit or for all of these three orbits? And how to choose the value of U?
        (2)When I set 2eV for 3p orbit of S atom, the calculation is hard to converge. Is there any way to make it converge?
   Waiting for your kind reply!
              Thanks!

[zh]

The use of DFT+U may be meaningless to treat BDT molecule, because it is a very weak correlated system.

[rw yan]

The use of DFT+U may be meaningless to treat BDT molecule, because it is a very weak correlated system.

    Thank you for your reply!
    By the way, is there any method to improve the HOMO-LUMO gap in ATK? Since the position of energy level is very important in studying electron transfer.

[zh]

At the present stage, no better solution, because the implementation of hybrid functional is till under development.

[Anders Blom]

I don't know if Meta-GGA has been tried for molecules, but that's another possibility to at least experiment with. It is available in ATK already, and doesn't add to the computation time substantially.

[rw yan]

I don't know if Meta-GGA has been tried for molecules, but that's another possibility to at least experiment with. It is available in ATK already, and doesn't add to the computation time substantially.

    Thank you for your kind reply! But I am still confused about what you have said. As I know, Meta-GGA isn't suitable for metals (the electrodes are gold), can we use Meta-GGA for BDT molecule and GGA for gold  separately in ATK?

[Anders Blom]

As I know, Meta-GGA isn't suitable for metals

Why not? That is not true, there are many calculations using MGGA on metals. Ok, most are using TPSS and not TB09, but as far as I have seen from the metals I tried, MGGA works just fine.

[rw yan]

The use of DFT+U may be meaningless to treat BDT molecule, because it is a very weak correlated system.

Recently，I have read a paper about Hubbard U.
http://www.sciencedirect.com/science/article/pii/S0022369713003296#
In this paper, the authors set U parameter for C/H/S elements. So does that mean Hubbard U is also feasible in BDT molecule?
Thank you for your kind help!

[rw yan]

Why not? That is not true, there are many calculations using MGGA on metals. Ok, most are using TPSS and not TB09, but as far as I have seen from the metals I tried, MGGA works just fine.

Thank you for correcting my view.
I have tried to use MGGA in my two probe system, but it is hard to converge. ???The py file is attached

[zh]

You should first do the test calculation of MGGA functional for an isolated molecule to check whether the HOMO-LUMO gap of this molecule is improved or not. If not, it's no better reason to choose it to do the further calculation of MGGA functional for  a two-probe system.

[zh]

The use of DFT+U may be meaningless to treat BDT molecule, because it is a very weak correlated system.

Recently，I have read a paper about Hubbard U.
http://www.sciencedirect.com/science/article/pii/S0022369713003296#
In this paper, the authors set U parameter for C/H/S elements. So does that mean Hubbard U is also feasible in BDT molecule?
Thank you for your kind help!

I think that it's not a good way to use U on such molecule.