Hello everyone.
I was running the example script that using EMD method to compute thermal conductivity for Ar. Here’s part of the script that I’m curious about.
variable dt equal 4.0 # timestep
variable p equal 200 # correlation length
variable s equal 10 # sample interval
variable d equal $p*$s # dump interval
#fix NVE all nve
reset_timestep 0
compute myKE all ke/atom
compute myPE all pe/atom
compute myStress all stress/atom virial
compute flux all heat/flux myKE myPE myStress
variable Jx equal c_flux[1]/vol
variable Jy equal c_flux[2]/vol
variable Jz equal c_flux[3]/vol
fix JJ all ave/correlate $s $p d &
c_flux[1] c_flux[2] c_flux[3] type auto file J0Jt.dat ave running
variable scale equal {convert}/${kB}/$T/$T/$V*s*{dt}
variable k11 equal trap(f_JJ[3]){scale}
variable k22 equal trap(f_JJ[4])*{scale}
variable k33 equal trap(f_JJ[5])${scale}
I found that the change value of Nevery, Nrepeat, Nfreq and even timestep can have significant infulence on the result. Why is that? The variable scale has Nevery and timestep right? Actually I need to use this method to compute thermal conductivity of other material so I need to figure it out. Can someone help me? Thank you very much.
value of Nevery, Nrepeat, Nfreq and even timestep can have significant infulence on the result. Why >is that?
If you make the time step too big a good simulation can become
a bad one (e.g. loss of energy conservation). Making the timestep
smaller should be OK, but you’d have to run longer to get good
statistics. Changing the params of the correlation (ave/correlate)
can’t be done arbitrarily. You have to worry about sampling
correctly and for long enough to gather good statistics
Perhaps Reese or Jeremy can give some rules-of-thumb
for correlation params for thermal conductivities. The
script on the web page is simply meant to be an illustration,
not necessarily a production-ready script.
You can get under resolution of the correlation if your sample interval is too late
You can get artificial correlation of the samples of the sample interval is too small
4-10 stable time steps is a rule of thumb for the sample interval
BUT
The largest errors are due to
Insufficient averaging of the correlation function
Using a max correlation time that is too small and cuts off meaningful correlations
See jones & mandadapu j chem phys 2012
You can tell a lot my just graphing at the output of ave/correlate