[Pw_forum] Core-Level Shifts and Charged Supercells Related Questions

[Giuseppe Mattioli]

Dear Evrem
Your problem has a very simple solution... When you use a PP having a core hole, the code found always an extra electron. Use tot_charge=+1.0 (and no 
smearing, unless your system requires it also without core holes) and your energy differences (between the "standard" and the "core hole" calculations) 
will be significant. I've performed such kind of calculation using PPs generated by ld1.x, and the results can be quite accurate (you may want to flick 
through this paper: Bolognesi et al., J. Phys. Chem. A  2009, 113, 13593-13600). Only two further suggestions

a) The above energy differences do not generally reproduce well "absolute" ionization energies. The best way to provide IE values is to reference the 
energy differences to that calculated for a known molecule. For C(1s) lines, e.g., I use the CO2 reference, embedded in the same supercell. As you are 
dealing with surfaces, you could still use a molecule as reference. but if your insulator is, e.g., NaCl, you might be forced to provide "relative" IE 
values only. I hope this is clear...

b) GGA functionals can behave rather poorly, or VERY poorly. When the valence charge displaces to screen the core hole, the system behaves as a 
strongly correlated one. In the case of molecules, I've obtained very accurate estimates by using the B3LYP functional, which is tailored around 
molecules. HSE, for instance, is not very good for the benzene molecule. But if you are going to perform EXX calculations, you cannot use PAW PPs 
(yet), but only NC ones.

HTH

Giuseppe


On Monday 25 November 2013 14:21:00 Evren Ataman wrote:
> Dear Quantum Espresso Users, Developers and Everybody in Between
> 
> (Sorry for the loong e-mail, which may take time equivalent of drinking a
> cup of espresso, to read)
> (I have tried to follow the "posting guidelines" in the Quantum Espresso
> page before posting. I hope this e-mail is an appropriate one)
> (To make the e-mail easy to read for time-lacking people I have put $$$ in
> front of every question)
> (If I solve my problems and understand everything, I promise to take the
> responsibility to reply similar questions in the forum in 2014)
> 
> Thank you very much in advance.
> 
> INTRO:
> I have a problem which results from my partial or complete lack of
> understanding related to couple of points.
> My ultimate purpose is to calculate surface core-level shifts for couple of
> elements in an insulator (I prefer to keep the material confidential) to
> compare/guide to XPS measurements.
> I eventually want to use PAW pseudopotentials (PP) and generate my own
> (with and without the core-hole on the appropriate energy level) by using
> AtomPaw code. I have not used ld1.x because of two reasons. Firstly, with
> my all respect to ld1.x, I personally found it a bit more complicated for a
> beginner to begin with. Secondly, for AtomPaw code the input files for
> well-tested PPs are very easily available so that one can just remove one
> core-electron and good to go (I hoped).
> 
> THE CASE STUDY 1 (Benzene Molecule):
> To understand the concept and warm myself up I first made some trivial
> calculations with a benzene molecule using PPs generated by experts
> (details are coming).
> When I calculate the benzene molecule with ordinary PPs, in the output file
> I get number of electrons as 30. This number makes sense: 6 carbon atoms
> each with 4 valence electrons and 6 hydrogen atoms each with 1; add up to
> 30.
> When I replace one of the carbon PP with a corresponding PP with 1s core
> hole, in the output file I get 31 electrons.
> The PP I use with the core-hole has the following info among others:
> 
> --------------------
> <PP_INFO>
> Generated using "atomic" code by A. Dal Corso  (espresso distribution)
> Author: Paolo Giannozzi   Generation date:  7Feb2008
> Info:   C PBE with a 1s core-hole: 1s1 2s2 2p2 RRKJ3 US
> nl pn  l   occ               Rcut            Rcut US             E pseu
> 2S  1  0  2.00      1.40000000000      1.60000000000 -2.19814028365
> 2S  1  0  0.00      1.40000000000      1.60000000000 0.05000000000
> 2P  2  1  2.00      1.30000000000      1.60000000000 -1.56800948749
> 2P  2  1  0.00      1.30000000000      1.60000000000 0.05000000000
> 3D  3  2 -2.00      1.30000000000      1.30000000000 0.15000000000
> <PP_HEADER>
> 5.00000000000      Z valence
> Wavefunctions         nl  l   occ
>                        2S  0  2.00
>                        2P  1  2.00
> --------------------
> 
> When I run this (an SCF calculation) I get the error message saying:
> "charge is wrong: smearing is needed"
> and a bit below a warning:
> "WARNING: integrated charge=    30.30000000, expected= 31.00000000"
> When I add smearing (occupations='smearing', degauss = 0.1,
> smearing='gauss') fairly calculation runs without any problem but unfairly
> WARNING disappears.
> 
> QUESTION SET NUMBER ONE:
> $$$ Is this supercell charged with and/or without smearing?
>     $$$ If yes how the charged-supercell is taken care of (is "a
> compensating jellium background is inserted" automatically without
> prompting the user)?
> $$$ What is integrated charge and what is expected charge?
> $$$ Why z_valance is 5?
>     $$$ If one puts one of the 1s core electrons into the one of the
> valence levels should not this make 2p level occupied by 3 electrons
> instead of 2 and write this explicitly in the PP file as 1s2 2s2 2p3?
> $$$ Does ld1.x calculate Z_valence as (atomic number) - (number of core
> electrons) or it just counts the electrons in the valence level given by
> the user?
> 
> ONE COMMENT AND TWO QUESTIONS:
> I assume that a PP with a core electron moved to valence level can be quite
> suitable for X-ray absorption process and with certain approximations for
> XPS, too. But let say that I want to try the case with 4 valence electrons
> and that somehow I create a PP for C atom with a core-hole and Z_valence
> value of 4 and used in my trivial benzene calculation.
> $$$ Would this supercell be charged?
>     $$$ If yes how the charged-supercell is taken care of (is "a
> compensating jellium background is inserted" automatically)?
> $$$ In general, is it possible to generate a charged-supercell with Quantum
> Espresso (or any other planeweve DFT code) just with ionized PPs or the
> only way is to use appropriate keyword, total_charge?
> 
> CASE STUDY 2 (Generating PPs with AtomPaw) AND ONE QUESTION:
> As far as I understood when one creates a PP with ld1.x one need to input
> atomic symbol Xor atomic number. So inputting both "C" and "6" is not
> allowed.
> However in AtomPaw input, one has to write both C and 6 but in this case 6
> represents the total number of electrons (this is what I understood from
> the user manual).
> The AtomPaw generates a lot of output files two of which are the actual PP
> files: one with an .xml extension and one with an .upf extension for
> Quantum Espresso.
> 
> When I generate a PP with a core-hole I assumed that total number of
> electrons for C should be 5 with 1s1 2s2 2p2 configuration.
> In this case in the .xml file I read: "<atom symbol= "C" Z="5" core="1"
> valence="4"/>" and in the corresponding .upf file in the <PP_HEADER> I read
> z_valence = 4.
> 
> When I generate a PP with a core-hole but this time with C and 6 electrons
> (keeping everything else the same).
> In the .xml file I read: "<atom symbol= "C" Z="6" core="1" valence="5"/>"
> and in corresponding .upf file in the <PP_HEADER> I read z_valence = 4.
> $$$ Why z_valence value is not correctly inserted into the .upf file, i.e.,
> always becomes 4?
> 
> 
> With my Kind regards


********************************************************
- Article premier - Les hommes naissent et demeurent
libres et ègaux en droits. Les distinctions sociales
ne peuvent être fondèes que sur l'utilitè commune
- Article 2 - Le but de toute association politique
est la conservation des droits naturels et 
imprescriptibles de l'homme. Ces droits sont la libertè,
la propriètè, la sùretè et la rèsistance à l'oppression.
********************************************************

   Giuseppe Mattioli                            
   CNR - ISTITUTO DI STRUTTURA DELLA MATERIA   
   v. Salaria Km 29,300 - C.P. 10                
   I 00015 - Monterotondo Stazione (RM)          
   Tel + 39 06 90672836 - Fax +39 06 90672316    
   E-mail: <giuseppe.mattioli at ism.cnr.it>
   ResearcherID: F-6308-2012