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

[Guido Fratesi]

Dear Rolly,

let me try to expand a bit my previous concise answers to your questions 
(Q1-Q2):

As it is implemented, molecularpdos.x needs the two systems (full & 
part) to be described by the very same parameters as of:
(1) unit cell dimensions
(2) k-point sampling
(3) number spin populations considered
(4) dimension of the localized basis set (atomic functions)

In principle, such close correspondence could be avoided in some cases 
like where your reference (i.e., part) system is an isolated molecule. 
Its states could be computed at the gamma point for a cell of your 
choice, then the Bloch states (as would be obtained by computing them 
for the same cell as for the full system) could be reconstructed 
analytically. This is not implemented so far, so an explicit calculation 
with the very same k-grid is needed (I generally copy the k-points as 
given in output by pw.x for the full system).

Similarly, the spin-up states of the full system will be projected on 
the spin-up states of the part (same for spin down); this requires you 
to make a calculation with nspin=2 (but no magnetization) also when the 
molecule by itself is not spin-polarized if it is embedded in a 
spin-polarized system.

Smearing / not smearing is not relevant to molecularpdos.x : just use 
the one adequate to each of the two calculations (full & part).

As for the dimension of the localized basis set, molecularpdos.x assumes 
that the atomic orbitals of the full & part calculation correspond 1-1 
to each other, keeping the same order. The range of the orbitals to be 
considered is specified by input variables 
i_atmwfc_{beg,end}_{full,part} so the atoms to be considered have to be 
consecutive and listed in the same order in the full & part systems. 
BEWARE (common mistake) this range specifies the atomic wavefunctions 
(see the output of projwfc.x for a list), not the atoms. You may want to 
add atoms to the "part" calculation, e.g., add H to saturate a dangling 
bond: in that case, just do not include the corresponding atomic states 
in the range for molecularpdos.x.

Hope this helps,
Guido

On 09/08/2016 18:46, Rolly Ng wrote:
> Dear QE experts,
>
> I am trying to refine computation of 3-NT adsorption on metal doped 
> graphene using QE. This is according to our previous work
> http://dx.doi.org/10.1016/j.commatsci.2011.07.045
>
> With the help of Dr. Guido Fratesi, I am exploring molecularpdos.x to 
> find the change of HOMO and LUMO of the nitrated tyrosine molecule on 
> adsorption to metal-doped graphene.
>
> There are my questions and answers, your comments are welcome.
>
> Q1) Spin polarization. The full system contains a metal doped graphene
> with single metallic atom of Au and Ni at the center and a 3-NT molecule
> adsorbed onto the graphene sheet, so I included nspin=2 for the full
> system. But the adsorbate (3-NT) molecule is likely to be nonmagnetic,
> so I did nspin=1 for its gas phase. Can molecularpdos.x cope with
> different nspin for the full system and adsorbate?
> A1) Yes, since nspin=2 can handle a nonmagnetic case
> R1) So, I am adding nspin=2 to the 3-NT molecules
>
> Q2) K-points. I used smearing for the full system since it is
> semi-metallic. Should I use smearing for the molecule in gas phase? I
> believe this is not a good idea but can molecularpdos.x works with
> k-point and non k-point?
> A2) Smearing: no problem, K-points: should be the same
> R2) So, I am using the same K-points for both full system and single 
> molecule.
>
> Q3) I would like to evaluate the change of the HOMO-LUMO gap of the
> adsorbate (3-NT) on adsorption to the metal-doped graphene. Can
> molecularpdos.x do that?
> A3) Dr. Fratesi and his team used that code also to study 
> molecules/graphene...
> http://dx.doi.org/10.1038/srep24603
> R3) Very useful and thank you!
>
> Q4) I would like to determine if physisorption or chemisorption occurs as
> with DMol3, can molecularpdos.x do the same job? For the physisorption 
> vs. chemisorption, the adsorption energy may be a good indicator. 
> However, I believe the electronic structure of the bond may also 
> indicate which type of adsorption it suppose to be?
>
> With regards,
> Rolly
>

-- 
Guido Fratesi

Dipartimento di Fisica
Universita` degli Studi di Milano Italy