QuantumATK-2018,convergence problems

[hagelberg]

Hello,
I have a question related to QuantumATK-2018. Enclosed is an input file (aGNR-2Fe….) for a device calculation involving an armchair graphene nanoribbon (aGNR) with substitutional Fe atoms. This is part of a follow-up project of an earlier study on the spin-filtering efficiency of aGNRs with embedded transition metal impurities (F. Hagelberg, A.Kaiser, I.Sukuba, M. Probst, Mol. Phys. 115, 2231 (2017)). The earlier work was done with ATK-2015. Trying to reproduce some of the previous results with QuantumATK-2018, I fail completely. The atomic populations seem to oscillate between a physically reasonable state and an unphysical state that includes extremely high atomic populations at the ends of the central region. I enclose the output file corresponding (atkjob-aGNR-2Fe…I reduced the size a bit) for illustration (see, for instance, the last population block in that file). I tried to attain convergence by varying the DFT methd (i.e. GGA versus LDA, DFT+U) as well as the exchange-correlation functional, but this didn’t improve the erroneous output. Are there some critical DFT-related features in QuantumATK-2018 that were not present in ATK-2015, and that I possibly mishandle?

Thanks,
Frank Hagelberg

[Ulrik G. Vej-Hansen]

Hi Frank

To me, the most obvious candidate is probably that the default pseudopotentials and basis sets were changed between the two versions, and this could certainly result in the changed convergence behavior you describe here. I suggest that you either manually specify the pseudpopotentil/basis set combination the was used in the previous calculations - I believe the default in 2015 was FHI-DZP orhave a look at our NEGF convergence guide if you wish to keep the 2018 defaults and solve the issue: https://docs.quantumwise.com/technicalnotes/negf_convergence_guide/negf_convergence_guide.html

[hagelberg]

Hi Ulrik,

Thanks for the quick response. I ran a first test, and the outcome suggests that indeed, making FHI-DZP the pseudopotential-basis set choice in ATK-2018 takes care of the unphysical states.

Frank

[Ulrik G. Vej-Hansen]

Hi Frank

Thanks for the follow-up. My guess is that you will need longer electrode extensions in the O-2018.06 release, to accommodate the more long-ranged basis functions that are default now.