[DFTB-Plus-User] R: Query on the use of DFTB+ for periodic transport calculations

[Alessandro Pecchia]

Dear Karen,

this is a sort of known problem, at least to me.

First of all, you should consider that the Fermi Energy of the bulk Au is not related to the Work function of the material because you cannot conceptually take out an electron from Au to the vacuum. For this you need a surface. In fact in DFT (say Quantum Espresso) you obtain a Fermi Energy of bulk Au even slightly above 0.  To obtain a reasonable number that compares to the work function you need to compute a slab of 5 or 7 atomic layers. The surface dipole is the main reason for the metal workfunction.  Unfortunately DFTB is not able to reproduce correctly this dipole, at least not the standard SCC-DFTB method. If I remember correctly the workfunction of a slab turns out about 2.7 eV. 
But you can try again if you use the shell-resolved SK files. In any case don't expect to go as low as 5.2 eV. The simple atomic charges are not flexible enough.

When I was doing molecular electronics calculations I used to shift the SK table of Au rigidly down by the right amount  in order to reproduce the workfunction of the surface and get a more realistic alignment to the typical MOs. But this procedure requires to temper the SK tables and it is not recommended. Unfortunately a rigid shift, behaving like an additional constant potential, is not supported in a simple way in the dftb+ code. 

If you plan to try this approach we should discuss further how to do it.

BUT this procedure might impact the Au-S (Au- organic elements) interaction, hence bond relaxation, energy, etc... In principles one might need to re-parameterize the whole set.
At the time I used a different parameterization, did some checks, and did not find large changes, but as I said everything needs to be carefully checked again with the Au sk-set released with dftb+.

An alternative position is to let the self-consistent charge loop to "self-align" the molecular energy levels to the Au chemical potential after charge rearrangement. I typically noticed that this procedure always results in the LUMO level right in resonance with the metal Fermi level, and this might not work well. But it depends on the molecular structure and energy levels.

Now that I am writing, I wander what happens if you set the potential of the two contacts to be -say- 2.5 V each, (using Potential[eV]=2.5). This might work because it rigidly shifts the contacts down and should self-align the surfaces to keep the system charge neutral.
The SCC loop might start out very badly and not converge though. But it is worth trying.

I hope this long dissertation has clarified a bit the not-so-easy issue of level alignment in dftb. In any case you'd better know from experiment or intuition what should be the correct alignment. This is a place where dftb requires some empirical adjustment. 

Regards,

Alessandro

 




-----Messaggio originale-----
Da: DFTB-Plus-User [mailto:dftb-plus-user-bounces at mailman.zfn.uni-bremen.de] Per conto di Karen Schafer
Inviato: giovedì 2 marzo 2023 15:13
A: dftb-plus-user at mailman.zfn.uni-bremen.de
Oggetto: [DFTB-Plus-User] Query on the use of DFTB+ for periodic transport calculations

Dear mailing list members,

I am relatively new to using DFTB+ and I can't fix a problem with self-consistent, periodic transport calculations. I am investigating a system with two gold (Au(111)) surfaces, bridged by an alkanedithiol ligand. The system is supposed to be periodic in x and y direction and I added a vacuum gap in z direction (transport direction).

When calculating the contacts, I receive a fermi energy of -2.6 eV which is higher than I would expect and also higher than what I get from non-periodic conditions (-4.2 eV), which results in a shift of the transmission compared to non-periodic and non SCC calculations.

I have included the gen structure file and the dftb_in file for the contact calculation. I have tried to fix the fermi energy by increasing the kpoints but that did not help. Is there something I am doing wrong?

Thank you in advance for your time and effort,

Karen