[Wien] query on optimization

Sat Feb 10 11:36:33 CET 2018

Dear Prof. Gerhard and other,

Sorry for the delay in response. One of our members already discussed this
on the mailing list so I was doing some test to make sure about the
authenticity of the calculation.

The origin of the issue is from this thread:
https://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/msg17059.html

As the matter was to see a significant effect of the procedure of
optimization on band gap. So, before answering all of your questions, in
below link please have a look here for the system and other details
(experimental data are taken from JACS paper):

https://docs.google.com/viewer?a=v&pid=sites&srcid=ZGVmYXVsdGRvbWFpbnxiaGFtdWNtc2x8Z3g6NGYzMGM0M2IzZjc1ODYyYQ

In above link, I have provided some test from two approaches, port
(min_lapw -j "run_lapw ....")  and MSR1a (run_lapw ....-min) on two
structure one for 10K and other at 300K.

Your Que 1.
what does it mean
" the program do not produce the exact relaxed structure on the optimized
lattice parameter" ?????

Ans.

Lets say I am doing optimization for -10, -5, 0, 5, 10: (in %). The
optimize.job script takes all deformed structure (at different volumes) one
by one and after optimization at the particular deformed structure it saves
it with some name (say VolXX_default.*, where "XX" stands for %vol and "*"
stands for all files at respective volume).

The last  VolXX_default.* among five will be from the one we put at the
last in optimize.job.

So the script gives us five optimized files at five specified volumes but
the optimized lattice parameters may or may not be from the one that the
script produced.

Most of the time we get the optimised lattice parameters in between any two
deformed structures (say between 0% and 2.5% or -2.5% and 0%).

So, the optimize.job file will not produce a structure file with printed
optimized lattice parameters instead it gives at the end after some
processing!!

Now, I come to relaxed positions:  When the optimization finishes, it
leaves the relaxed case.structure from last deformed structure (here it is
from 10% vol change).

So if we say that case.struct file is the final relaxed structure after
optimization finishes, it seems wrong to me as this is one which relaxed at
10% vol changes. #I may be wrong!!

Q. 2
what is wrong ?
how did you decide that it is wrong ?
[This question comes from bottom of this thread:
https://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/msg17072.html ]

Ans. It is believed that the simultaneous optimization of positions and
lattice parameters a good approach but as my results vary at the way I
proceed so I am having doubt for the present case. Although for my previous
results were not varied much (ABO2 type system).

Q.3:

where do you have your initial parameters from ?

Ans. data are in above link and taken from JACS paper.

#-----------

>From the data provided in the link, for low temperature (at exp parameter
at 10K), I am getting slightly overestimated lattice parameters (~0.03 Ang)
while for high temperature (300K) I am getting slightly underestimated
lattice parameters (~0.1 Ang). All calculations were performed with
PBESol.

If I do optimization without relaxation then the optimized lattice
parameters are underestimating the Exp. parameters by ~0.01 Ang. only and
Bulk modules was ~20 GPs wherein the present case it is ~10GPa.

1. So from my answer to Qes 1, I am not sure that the optimize.job gives us
final relaxed structure instead it only give optimized lattice parameter
and we need to do relax our initial structure at these optimized lattice
parameters during scf.

2. What is the best way to optimize the unit cell parameters (in this
present case only) and to proceed to scf as here we see that bulk module
also vary significantly in addition to the bandgap when we optimize with
relaxation and without relaxation!!

I apologize if I make this draft a little bit longer but it is to make a
clear understanding.

Please let me know what additional information I can provide you!

Thanks and regards

Bhamu

On Fri, Feb 9, 2018 at 12:10 PM, Fecher, Gerhard <fecher at uni-mainz.de>
wrote:

>
>
> read the last question first !
>
>
> Ciao
> Gerhard
>
> DEEP THOUGHT in D. Adams; Hitchhikers Guide to the Galaxy:
> "I think the problem, to be quite honest with you,
> is that you have never actually known what the question is."
>
> ====================================
> Dr. Gerhard H. Fecher
> Institut of Inorganic and Analytical Chemistry
> Johannes Gutenberg - University
> 55099 Mainz
> and
> Max Planck Institute for Chemical Physics of Solids
> 01187 Dresden
> ________________________________________
> Von: Wien [wien-bounces at zeus.theochem.tuwien.ac.at] im Auftrag von Dr. K.
> C. Bhamu [kcbhamu85 at gmail.com]
> Gesendet: Freitag, 9. Februar 2018 06:24
> An: A Mailing list for WIEN2k users
> Betreff: [Wien] query on optimization
>
> Hello,
>
> This is in continuation to my previous email. As I got unusual trend in
> band gap so this is just to verify the procedure: I do not want to make any
> mistake!!
>
> For an optimization of  system having SG = 225 with at five volumes, say
> -10, -5, 0, 5 10% with: min_lapw -j "run_lapw -cc 0.0001 -ec 0.00001 -fc
> 0.1 -p" I got nice parabola and optimized lattice parameter.
>
> From here, we can not take relaxed structure and we only can take
> optimized lattice parameters as the program do not produce the exact
> relaxed structure on the optimized lattice parameter.
>
>
> If it is not giving the final relaxed structure at optimized lattice
> parameter then  what is the mean of relaxing the structure during
> optimization?
>
>
> Using the optimized lattice parameter, do we need to relax the structure
> during scf (say with any suitable -vxc)?  #I would say, yes!!
>
>
>
> regards
> Bhamu
>
>
>
>
>
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://zeus.theochem.tuwien.ac.at/pipermail/wien/attachments/20180210/6101473e/attachment.html>