Set k-point for low-dimensional systems with ngridk

 

ZEYU JIANG - 2020-05-30

Dear developers,

I'm using ELK to perform the standard band calculation for a 1D-chain. The lattice constants of supercell are 13.7x13.7x6.5 (angstrom) and I used ngridk=1 1 10. However, I found that the code never used this mesh, but instead it's determined by the lines below as printed in the "INFO.OUT" file:

Radius of sphere used to determine k-point grid density : 50.00000000
k-point grid : 3 3 5

So it seems that the k-point is still determined automatically, but even if that's the case, the default value of radkpt should be 40.0, but not 50.0......

Attached is my input file, I will really appreciate if anyone could tell me where it goes wrong .

Thanks very much,

Zeyu Jiang, Postdoctoral Fellow
Department of Physics, Applied Physics, and Astronomy
Rensselaer Polytechnic Institute

 

• elk.in

  • 
    

• 

  Youzhao Lan - 2020-05-30

  Dear Zeyu,
  Becasue you set "highq" to true.
  You can check the souce code "readinput.f90", then search "highq" and you can find the "radkpt" is set to "50" and "autokpt" to "true" by DEFAULT.

  Best regards.
  Youzhao Lan

   

• • 
    

J. K. Dewhurst - 2020-05-31

Dear Zeyu,

Youzhao is correct: the 'highq' option sets 'autokpt' to .true.

However, any block set after 'highq' in elk.in overides high-quality option.

For instance, using

highq
 .true.

ngridk
  10  10  1

will switch 'autokpt' to .false. and use the 10x10x1 k-point set instead.

Regards,
Kay.

 

https://sourceforge.net/p/elk/discussion/897820/thread/2233aeb1fd/