[Pw_forum] Using molecularpdos.x for adsorption on metal-doped graphene

[Guido Fratesi]

Dear Rolly,

> Since I am using k-point to compute both molecule and full-system, I
> obtained the Fermi energy as,
>
> The molecule NTNO2 Fermi energy reads -3.5478 eV,
> the full-system Fermi energy reads -2.1546 eV.
>
> So I suppose the HOMO is located at the first band to the left of the
> Fermi energy, and the LUMO is at the first band to the right of the
> Fermi eenrgy. From the scf output at K = 0 0 0, I collect these number
> for spin up and down as,
> 1) Molecule Up, HOMO -5.5984 eV, LUMO -3.2161 eV, so |delta| HOMO-LUMO
> gap is 2.3823 eV
> 2) Molecule Down, HOMO -5.5981 eV, LUMO -3.2159 eV, so |delta| HOMO-LUMO
> gap is 2.3822 eV

Your interpretation is fine and the data are consistent, for the 
isolated molecule (minimal spin dependence)

> 3) Full-system Up, HOMO -2.1696 eV, LUMO -2.1449 eV, so |delta|
> HOMO-LUMO gap is 0.0247 eV
> 4) Full-system Down, HOMO -2.1935 eV, LUMO -2.1177 eV, so |delta|
> HOMO-LUMO gap is 0.0758 eV

Uhm, your system is a molecule on graphene if I remember correctly, so 
why are you expecting a gap?
(beside the opening of a small gap occurring because of finite system size?)

It may be useful to think of molecules adsorbed on a metal surface, like 
Al(001).
What would then be the gap of the full system? --> 0
Would this change if you chemisorb / physisorb your molecule? --> no
What is the resulting HOMO-LUMO gap if no molecule at all is present, in 
principle? --> 0, it is a metal surface
What is the resulting HOMO-LUMO gap of the full system from the 
calculations? --> it will be "small" and depend on the K-point sampling; 
such dependence is mitigated by the smearing technique.
Would looking at the eigenvalues at K=\overline{Gamma} the correct way 
to find it? --> you would have to look at the full BZ. A band gap at 
Gamma could indeed be present also for a metal surface.

> I can see the substantial drop in the H0MO-LUMO gap on the full-system
> particularly for the Up spin configuration. Does this help to explain my
> previous questions Q3+Q4 on physisorption vs. chemisorption?
You may want not to compare the HOMO-LUMO of the molecule with the 
HOMO-LUMO of the full system, rather, the HOMO-LUMO of the free molecule 
with the energy difference between molecule-induced structures in the 
DOS of the full system.

> I am running molecularpdos.x but it seems run for more than 10 hours on
> a 12 cores/24 thread system without stopping and output as shown in the
> attached screenshot?
It should not be so long. At maximum it has taken few minutes for the 
largest cases I considered.
Notice, molecularpdos.x is a serial code so just run it on one CPU core.


> However can I select fewer bands so it may help to reduce the
> computation time? Is this i_bnd_beg_full = 1 refers to -28.1982 eV the
> first one? or 2.1629 eV the last one?
Yes, you could reduce the size of the calculation by not looking at deep 
bands. Suppose you have 400 electrons (~200 filled states per spin), and 
that the lowest 100 states are below the energy range of your interest, 
you could then put i_bnd_beg_full = 101.

But I'm surprised the code is hanging in your case and I would check 
something simpler as a test.
E.g., you could project the free molecule on the free molecule, (i.e., 
full system = part).
(BTW does the example espresso-5.4.0/PP/examples/MolDos_example runs 
properly?)

Kind regards,
Guido