[Wien] Questions about difference electron density map

[Tran, Fabien]

The way lstart works to generate case.sigma may be a bit confusing. The Ecut asked by lstart does not matter for case.sigma. Ecut is used to separate the core electrons from the others (semicore and valence) during the SCF calculation. In order to generate case.sigma with selected (semicore/valence) orbitals, you need to replace N by P in case.inst (not in case.inst_sigma) next to the selected orbitals and execute "x lstart" without -sigma. "x lstart -sigma" is used to generate case.sigma with all semicore/valence orbitals without having to replace all N by P in case.inst.



________________________________
From: Wien <wien-bounces at zeus.theochem.tuwien.ac.at> on behalf of Ding Peng <ding.peng at monash.edu>
Sent: Monday, May 4, 2020 2:46 PM
To: A Mailing list for WIEN2k users
Subject: Re: [Wien] Questions about difference electron density map

Dear Dr Fabien Tran,

Thank you very much for your feedback. This is very helpful and solve my puzzle that confused me for a long time. I ask this question because I am not sure if my calculation on the difference charge density on CeB6 is correct.

I'm still a bit confused about the superposed valence electron density. In the calculation of difference charge density, firstly I did "x lapw2" with Emin to separate the semi-core and valence state. Then I did x lstart -sigma. I'm not sure if the value of Ecut for this step should be the same as the one in the DFT initialisation step (i.e. set the Ecut to separate the core-state and the valence to avoid the core charge leaks out of MT-sphere). However, the valence electrons defined in the DFT initialisation step is not the same as the ones intrinsically defined by the chemists. For the case of CeB6, the valence electrons in chemistry are 4f, 5d and 6s electrons. This is different from the ones in the DFT initialisation step since we have to truncate 5s, 5p and sometimes 4d as well (i.e. treat them as the valence electrons) to avoid the charge leaks out of MT-sphere.

So my question is: does this mean I should not use the "default valence states" module and have to turn to the "non-default valence states" module in the function of "electron density plots", and manually set 5s, 5p and 4d as core states to calculate the superposed atomic charge densities?

If I follow this process, then I found an another problem, in the file lstart.def, the file case.inst replaced case.inst_sigma. But as Prof Peter Blaha had suggested (if my understanding of his words is correct), both case.inst and case.inst_sigma should be presented in lstart.def. Does that mean I should manually change included file names in lstart.def (add both "case.inst_sigma" and "case.inst" in it)?

Many thanks and sorry for asking so many questions, which may seem to be silly.

Ding

On 4 May 2020, at 7:26 pm, Tran, Fabien <fabien.tran at tuwien.ac.at<mailto:fabien.tran at tuwien.ac.at>> wrote:

Hi,
Ideally, the integrated density difference should be zero, but due to numerical imprecision it won't be exactly the case. Since I don't know exactly what you want to do and the procedure you want to follow, it's difficult to give you more advice. Just be aware of two things:
1) Integrating accurately a quantity generated from lapw5 (uniform 3D grid) is difficult (or maybe impossible), because  the grid close to the nuclei (where the density is very spiky) needs to be extra dense. But, this should not be a problem for a density difference because the spikes of the two densities should cancel such that the difference can be integrated easily.
2) There is no particular problem at the edge of the cell. But, by integrating one should not count twice the same points: those which are on the opposite planes of the cell.
​
________________________________
From: Wien <wien-bounces at zeus.theochem.tuwien.ac.at<mailto:wien-bounces at zeus.theochem.tuwien.ac.at>> on behalf of Ding Peng <ding.peng at monash.edu<mailto:ding.peng at monash.edu>>
Sent: Monday, May 4, 2020 12:45 PM
To: A Mailing list for WIEN2k users
Subject: Re: [Wien] Questions about difference electron density map

Dear Dr Fabien Tran,

I had tried your suggestions and it indeed has worked. The difference charge density map looks reasonable. Many thanks for that!

Can I ask you a question that if the net difference charge density in the unit cell (i.e. summation of the difference charge densities within the 3D unit cell) has to be zero? Because I know the errors introduced by the density samplings may be presented in LAPW5, especially at the edge of the unit cell. Also it can be affected by how many points are used to sample the grid-segment of the unit cell when using lapw5. I don't know if this sampling issue will affect the correctness of the charge density as well as the net difference charge density in the unit cell.

Many thanks,
Ding



On 3 May 2020, at 2:50 am, Tran, Fabien <fabien.tran at tuwien.ac.at<mailto:fabien.tran at tuwien.ac.at>> wrote:

If I'm not wrong (I have never tried this combination of keywords), you can do that by specifying
DIFF ADD
at the 6th line of case.in5(c). See Sec. 8.14.3 of the user's guide for explanation.​

________________________________
From: Wien <wien-bounces at zeus.theochem.tuwien.ac.at<mailto:wien-bounces at zeus.theochem.tuwien.ac.at>> on behalf of Ding Peng <ding.peng at monash.edu<mailto:ding.peng at monash.edu>>
Sent: Saturday, May 2, 2020 6:41 PM
To: A Mailing list for WIEN2k users
Subject: Re: [Wien] Questions about difference electron density map

Dear Dr Fabien Tran,

Thanks for your reply. Very clear explanation as it is always.

I got another question about the difference charge density calculation.
For magnetic materials (including ferromagnetic, anti ferromagnetic or quadrupolar-antiferromagntic materials etc.), what I can think of to calculate the difference charge density is

1) calculate the valence charge density (option “rho”) for spin-up electrons (x lapw5 -up)

2) calculate the valence charge density (Option “rho” for spin-down electrons (x lapw5 -dn)

3) sum 1 and 2 and subtract them by the superposed atomic valence densities (option “Over”.


Can I ask if this is correct way to do the calculation of the difference charge density for magnetic materials? and also if there is an automatic option to directly calculate it in Wien2k “, just like option “Diff” for non-magnetic materials ?

Many thanks,
Ding

On Sun, 3 May 2020 at 12:16 am, Tran, Fabien <fabien.tran at tuwien.ac.at<mailto:fabien.tran at tuwien.ac.at>> wrote:
As I already said in my previous email, the last column is the occupancy.
An occupancy of zero means that the corresponding band is not included
in the valence electron density (stored case.clmval, that is used for plotting).
"Energy to separate semicore and valence states​" seems to be a prediction
of what it means.


From: Wien <wien-bounces at zeus.theochem.tuwien.ac.at<mailto:wien-bounces at zeus.theochem.tuwien.ac.at>> on behalf of Ding Peng <ding.peng at monash.edu<mailto:ding.peng at monash.edu>>
Sent: Saturday, May 2, 2020 5:23 PM
To: A Mailing list for WIEN2k users
Subject: Re: [Wien] Questions about difference electron density map

Dear Dr Fabien Tran,


Thanks for your reply. Actually, I set the Emin as -0.2 Ry, which means I discarded the states that are below -0.2 Ry. Could I ask you if the last number of the "band lines" is the occupancy?  If it is 0.0000000D+00 (double precision of 0), does  it mean that the corresponding band is included as the valence state? Also I noticed there is a line "Energy to separate semicore and valence states: -3.971792594507138E-002". Could  I ask if the value "-3.971792594507138E-002" in this line the predicted Emin by Wien2k?


Sorry for asking so many questions in this email.


Best regards,
Ding



On 2 May 2020, at 10:12 pm, Tran, Fabien <fabien.tran at tuwien.ac.at<mailto:fabien.tran at tuwien.ac.at>> wrote:



From your output2, we can see that you discarded the states that are below -0.3 Ry



(the occupancy in the last column is zero for them) to calculate the electron  density.
You have to check that the number of remaining electrons used to calculate the
valence electron density is the same as the number of electrons obtained from lstart to
create the valence electron density from superimposed ​atomic  densities.



From: Wien <wien-bounces at zeus.theochem.tuwien.ac.at<mailto:wien-bounces at zeus.theochem.tuwien.ac.at>>  on behalf of Ding Peng <ding.peng at monash.edu<mailto:ding.peng at monash.edu>>
Sent: Saturday, May 2, 2020 1:11 PM
To: A Mailing list for WIEN2k users
Subject: Re: [Wien] Questions about difference electron density map





Dear Prof Peter Blaha and other WIEN2k experts,



Many thanks for your advice. Your explanation is very clear and I can understand most of them. But I still get some questions to ask:





By following your advices, I've checked the band ranges in output2 and they are :

 ​

 Compared to the information in case.scf1, I was thinking the Emin should be set slightly above  the energy band 10, which is from -0.3524 to -0.2916 (please correct me if I'm wrong). Does that mean the Emin of the whole valence electrons should be set slightly above Emax of energy band 10, which is -0.2916?


Many thanks,
Ding​            _______________________________________________


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://zeus.theochem.tuwien.ac.at/pipermail/wien/attachments/20200504/90fe1d33/attachment.html>