Top Page > Browsing
Forces of same structure gotten from openmx are a lot different from that from vasp
Date: 2019/12/17 21:07
Name: Zuzhang Lin   <linzz16@mails.tsinghua.edu.cn>

Dear all
Thank you for checking my message. Recently I tried to calculate phonon spectra by openmx with the method called finite displacement and I got an issue.
I have tested two structures.
Structure I: this structure is optimized by vasp. Then I use it as an input structure in openmx calculations. When I calculate the force of the structure which was derived from structure I with a finite displacement of one atom, I found the force I got is very different from that calculated form vasp(about 0.1 eV/Angstron).
Structure II: this structure is optimized by openmx. Structure II is just a little different from structure I. The force of the structure which was derived from structure II with a finite displacement of one atom are similar to that of Structure I.
I wonder why there is a large difference?

With best regards
Zuzhang Lin
メンテ
Page: [1]

Re: Forces of same structure gotten from openmx are a lot different from that from vasp ( No.1 )
Date: 2019/12/18 09:47
Name: T. Ozaki

Hi,

The two codes, VASP and OpenMX, use different basis functions and pseudopotentials.
Thus, it is difficult to expect that the two codes give an exactly equivalent result.
After getting the optimized structure by one of the codes, the optimized structure
cannot be optimum for the other code, and vice versa. This is what you saw.
Even with the single code, VASP or OpenMX, the optimum structure depends on computational
parameters such as quality of basis functions, choice of pseudopotential, and how the
real and k-spaces are discretized. So, once you change the numerical parameters,
you should observe residual forces even for a structure optimized by the same code but with
a set of different parameters.

To properly calculate the phonon dispersion using the finite difference method, one needs
to keep all the computational parameters in a single code for all the calculations to avoid
any numerical noise. Also, rather severe computational parameters have to be chosen to avoid
numerical artifacts such as egg-box effect on the energy surface.

We have implemented a code to calculate phonon dispersion using the finite difference
method, and obtained results well compared to published ones.
http://www.openmx-square.org/workshop/meeting15/Phonon_OpenMX.pdf
http://t-ozaki.issp.u-tokyo.ac.jp/meeting16/OMX-YTLee-2016Nov.pdf

Also, please notice that ALAMODE has been interfaced with OpenMX.
https://alamode.readthedocs.io/en/latest/
which may provide a way to calculate phonon dispersions via OpenMX.

Regards,

TO
メンテ
Re: Forces of same structure gotten from openmx are a lot different from that from vasp ( No.2 )
Date: 2019/12/25 01:27
Name: Zuzhang Lin  <linzz16@mails.tsinghua.edu.cn>

Dear T. Ozaki
Thank you for your suggestions. I have used ALAMODE and solved the problems. I found the dfference I got before may be due to unsuitable orbital bases I chose.

Best wishes
Zuzhang Lin
メンテ

Page: [1]

Thread Title (must) Move the thread to the top
Your Name (must)
E-Mail (must)
URL
Password (used in modification of the submitted text)
Comment (must)

   Save Cookie