Top Page > Browsing
Elastic Constant of Cu - 6.0Hs3p3d3
Date: 2019/06/17 16:58
Name: C. Pashartis

Hello,

I have been using openmx to calculate elastic constants of various elements. So far I have had success in this regard but occasionally I'll stumble upon a situation where I have to fine tune some input parameters in comparison to those given on your VPS/PAO convergence tests.

One such case I can't seem to resolve is the curious case of Cu-6.0H-s3p3d3. It seems I have stumbled on a region of numerical noise that fluctuates with 30 GPa. I have tried to change smearing temperature, scf tolerances, cutoff energy, k-points, and force tolerances. My assumption at this point is that perhaps I am using over-converged basis sets. It seems people in the past had trouble with Cu basis sets, not sure if they are resolved. Is there anything else you would think of that may help. I am using openmx-3.8

P.S. I appreciate your help!

Below is a sample input file:

FOR RELAXATION:

#
# SCF
#

scf.Kgrid 36 36 36
scf.XcType GGA-PBE
scf.SpinPolarization off
scf.energycutoff 700
scf.ElectronicTemperature 300
scf.maxIter 100
scf.EigenvalueSolver band
scf.Mixing.Type Rmm-Diisk
scf.criterion 1e-06
scf.restart off
scf.partialCoreCorrection on
scf.Mixing.History 20
scf.Mixing.StartPulay 6
scf.Init.Mixing.Weight 0.3
scf.Min.Mixing.Weight 0.001
scf.Max.Mixing.Weight 0.4
scf.lapack.dste dstevx
scf.Mixing.EveryPulay 1
scf.fixed.grid 0.0 0.0 0.0
scf.stress.tensor on

#
# MD
#

MD.Type RFC5
MD.maxIter 100
MD.Opt.DIIS.History 3
MD.Opt.StartDIIS 5
MD.TimeStep 0.5
MD.Opt.criterion 0.0001

FOR INTERNAL RELAXATION OF DEFORMATION (USUALLY DOESN'T CHANGE ANYTHING)

#
# MD
#

MD.Type DIIS
MD.maxIter 100
MD.Opt.DIIS.History 3
MD.Opt.StartDIIS 5
MD.TimeStep 0.5
MD.Opt.criterion 0.0001

メンテ
Page: [1]

Re: Elastic Constant of Cu - 6.0Hs3p3d3 ( No.1 )
Date: 2019/06/20 23:51
Name: T. Ozaki

Hi,

In a dense structure such as FCC, HCP, and BCC with a small lattice constant,
eigenvalues of overlap matrix approach to zero, and sometime they become even negative
values. The situation is called overcompleteness of basis functions, and a numerical
instability may appear. Then, changing parameters such as temperature, k-points, and scf
criteria does not resolve the overcompleteness.

Reducing the basis functions can solve the problem.
I wonder that s3p3d2 works well for your purpose.

Best regards,

TO
メンテ
Re: Elastic Constant of Cu - 6.0Hs3p3d3 ( No.2 )
Date: 2019/06/21 20:14
Name: C. Pashartis

Hello Prof. Ozaki,

Thank you for your response. For the most part, I have been following the basis set schemes that your project has shown on the website in comparison to WIEN2K for the equation of state (varying lattice and plotting energy). I believe this to be a good starting spot for my calculations given that I deform my structure in a non-isotropic fashion.

I believe the over-completeness to still be occurring for s2p2d2, am I correct in assuming that the s3p3d2 should then not work either? What are my other options given that it seems the s2p2d2 scheme won't work either?

I did also try s3p3d3 on the soft version, however, the results aren't as accurate in elastic properties compared to literature. Which can be inferred from the difference in the equation of state to WIEN2K.
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