Dear Mostafa,
I (and others) also noticed that the spectral heat current (SHC) method as
implemented in GPUMD might give weird results for some systems. The exact
reason for the failure of this method for some systems is not entirely
clear to me, but it is not likely due to a hidden bug is the code, but due
to the applicability of the method as implemented in GPUMD. This method was
proposed by Saaskilahti et al, and can be traced back to some earlier works
in Volz's group. It might be the case that the heat current *should be*
calculated from the force constants, up to a given order, say the second
order as in most previous works. In contrast, I have implemented the full
heat current in GPUMD.
Can I ask a few questions on your model and potential?
-- Are there Ga-Ga and N-N bonds in your structure?
-- Are there any bond lengths r within the corresponding two cutoff
distances R1 and R2: R1 < r < R2?
-- The same question for your AlN system.
I mentioned to you that Alex has implemented a new set of spectral
decomposition methods in GPUMD, mostly based on the Green-Kubo modal
analysis (GKMA) method developed in Henry's group. We will document this
method soon.
Zheyong
On Tue, Nov 5, 2019 at 7:29 AM Mostafa Valadkhani <mostafa.v68@xxxxxxxxx>
wrote:
Dear all,
I'm calculation SHC of AlN and GaN with implementing tersoff potential to
GPUMD,
The results for AlN was reasonable but for GaN I have some sin noises as
attached on the result don't know why! The run.in file is as:
potential potentials/tersoff/NGa.tersoff.txt 0 1
velocity 300
ensemble nvt_ber 300 300 0.01
fix 0
time_step 1
dump_thermo 1000
run 1000000
ensemble heat_lan 300 200 10 1 10
fix 0
dump_position 10000
compute 0 10 1000 temperature
compute_shc 2 500 10000 5 6
run 20000000
I doubled the run time and averaging in compute_shc and the results became
even worse.
Any idea to fix the results?
thanks,
M. V.
________________
Mostafa Valadkhani
Department of Mechanical Engineering,
University of Tehran,
Tehran, Iran
mostafa.v68@xxxxxxxxx | valadkhani.m@xxxxxxxx
