I am trying to calculate the Fermi level of a DNA-graphene system. I was able to get the calculations for graphene and DNA strand to converge separately. However, when I try to calculate the same for DNA-graphene system, the SCC cycles do not converge even after multiple restarts. Any information on why is this happening would be much appreciated.

The input file can be found here.

Further context: The coordinates are obtained from clustering of a Molecular Dynamics trajectory, and the graphene segment and DNA strand is cut out from the original full sheet and 20-nucleotide long single-strand DNA from the MD trajectory. Therefore, i am not allowed to perform an extra geometry optimization to relax the structure.

  • 1
    $\begingroup$ The hole that you cut out in the graphene is not hydrogen-terminated, which means that you are studying a system with a huge number of aryl radicals, which will immediately destroy your DNA if this system is prepared experimentally. Also, you didn't add counteranions to the DNA, which means that you are calculating a DNA with phosphate radicals, instead of the usual phosphate anions. Do you really intend to calculate such a system with so many radicals located in unconventional positions? If not, please revise the structure. The convergence difficulty will likely resolve itself $\endgroup$
    – wzkchem5
    Jun 1, 2023 at 14:22
  • $\begingroup$ I agree that neither the edges are not hydrogen-terminated nor the phosphate groups are charge balanced. What we are interested is only in the contribution of the ssDNA segment towards the Fermi of the graphene patch. $\endgroup$ Jun 1, 2023 at 15:31
  • $\begingroup$ Do you mean the Fermi level? Then I think at least the charge state of the DNA should be correct, because negative charges will shift the Fermi level up by a lot. Even then, it may be problematic to use non-hydrogen-terminated graphene, because the non-terminated carbon atoms may introduce edge states that screw up the band structure. $\endgroup$
    – wzkchem5
    Jun 1, 2023 at 15:50

1 Answer 1


I had posted the same question on the dftb+ forum and Prof. Aradi responded to the question with the following answer.

A common strategy is for large metallic systems to start with a very high electronic temperature (e.g. 100000 K), converge it at that temperature, then start a new calculation with lower electronic temperature but using the converged charges of the previous calc as initial guess (via ReadInitialCharges = Yes).

Additional comments:

  1. You should not use the 3rd order correction with the mio-set. If you wish to have 3rd order correction, use the 3ob set instead.

  2. If you use 3rd order, you should use ThirdOrderFull (and not ThirdOrder), as suggested in the manual.

  3. Your cell is quite large, so I don't think you need any k-points apart of Gamma. (Which would make the calculation significantly faster).


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .