Slater-Koster method for charge transfer calculations

[hadi9827]

Dear QuantumATK staff,

I am using the semi-empirical method, particularly the Slater-Koster method (dftb.org parameters) to calculate the interactions between a ZnO surface with a molecule (look at the attached pic). I am calculating the charge transfer, HOMO-LUMO, binding energy and so on. I have two important and almost independent questions and I will be so grateful if you can help me:

FIRST QUESTION- Is the Slater-Koster method reliable for these calculations? I know that each method has its approximations but I want to know that generally, the Slater-Koster method (dftb.org parameters) performs correctly in calculating the parameters I am calculating or I must use for example Huckel method or even DFT?

The SECOND QUESTION- I would like to investigate the impacts of implicit solvent on my calculations. I mean I want my molecule to be in a solvent above the slab surface and not in a vacuum. I read the tutorial and I know that in the DFTB method I can change the dielectric constant in the poisson solver and in this way the solvent will be considered. I choose Multigrid solver and leave the boundary conditions for A and B periodic and choose Neumann and Dirichlet for C left and right respectively and 80 for dielectric constant. However, I never can see any differences between with-solvent and without-solvent calculations in any way and it seems to me that no solvent is considered at all. What could be the reason?

I am looking forward to your help. Thanks.

[Petr Khomyakov]

I am not sure SK would be work for this kind of problems. Did you see any example in literature? You may always verify that by benchmarking SK against DFT, e.g., calculating absorption energy or any physical quantity of interest, at least for some configuration, to see how that performs. 

For the second one, you may introduce a solvent dielectric constant in the Poisson solver for DFT as well. Also, did you do self-consistent calculation with SK (you would need to unselect "No SCF iteration" in the Calculator)?  Seeing your script might be helpful.

[hadi9827]

Thank you very much for your responses.

If I want to benchmark it against DFT, what DFT parameters and predefined functions are suitable for doing it? LDA, GGA, MGGA, and PBE , ...?

Yes I assigned the dielectric constant and unselected "No SCF". I have attached my script. please find the attachment.

Thank you very much.

[Petr Khomyakov]

For DFT calculations, I would start with GGA-PBE. You may also need this correction, https://docs.quantumatk.com/tutorials/dispersion_corrections_and_bsse/dispersion_corrections_and_bsse.html, this can now be set in the GUI. Also, you might consider including van der Waals interaction. But that needs to be checked.

[hadi9827]

Thank you again for your response.
I am using ATK version P-2019.03. In the link you sent me I can see "The DFT-D dispersion corrections" and "BSSE and the counterpoise correction". The first one is accessible in GUI as DFT-D2 and DFT-D3 but how about the second one? I can not see anything about it and the tutorial says it should be added manually! would you please guide me a bit more? Thanks for your patience.

[hadi9827]

Sorry, would you mind telling me where in GUI I can consider BSSE correction?! I am using version P-2019.03.
And would you please guide me on the script I sent you?
Thank you very much for being so supportive.

[Petr Khomyakov]

If you do not see Counterpoise in the Calculator, see a png-image enclosed, it means you have to do it manually as described in the tutorial.

[hadi9827]

Thank you very much. I can not see it, so I will do it manually.
Have you checked my script as well? I have no idea why the dielectric constant has no impact on the optimization trajectory. I change it from 1 to 50 but when I check the optimization trajectories the identical things happen. I hoped to see some changes when I change the dielectric constant since I am implementing the molecule in implicit solvent!

[Petr Khomyakov]

I do not see anything peculiar in your script. You may try calculating ElectrostaticDifferencePotential for epsilon=1 and epsilon>>1, and compare that for the 2 cases to see if there is any effect of changing epsilon on the electrostatic potential in your calculations. 

[hadi9827]

Thank you very much for your response. I will do so. Sorry, I asked a new question titled "Asking for help; I can not relax a ZnO slab". Would you mind checking it to help me, please? It has been a week I am trying to solve that problem. Thank you for being supportive