Hi,
I have not run your inputs, but a quick look at them reveals some problems.
I may test them when I have time, but you can first have a look at the
following comments:
1) The first few lines in your xyz.in file reads:
22075 3 2.1 0 0 1 1
1 1 1 185.726000 107.229000 40.000000
0 0.00000000 0.00000000 5.00000000 12.01 1 1
0 1.30333990 -0.56436271 5.00000000 12.01 1 1
There are several problems here:
A) The maximal number of neighbors for one atom is only 3, which is too
small for a multi-layer system. You can change it to 200.
B) You initial cutoff distance is also too small. From your LJ potential
file, your largest cutoff distance is around 8.538 A. I suggest you set the
initial cutoff distance to (8.538+1) A, where the extra 1 A can be
considered as a skin distance. You can choose to update the neighbor list
or not, depending on your model.
C) You want to use a vacuum layer? That is not good in GPUMD. I suggest you
simply use free boundaries for the z direction. That is, change "1 1 1" in
the second line to "1 1 0". The reason is that GPUMD will automatically
control the pressure in all directions with periodic boundary conditions if
you use the npt_ber ensemble, which might shrink your box in the direction
with vacuum layer. If you do not use the NPT ensemble, it is OK to have a
vacuum layer.
2) In the run.in file:
There are three instances of the "compute" command within a run:
compute 0 100 200 temperature
compute 0 100 200 jp
compute 0 100 200 jk
This is not the intended grammar. You can change it to
compute 0 100 200 temperature jp jk
=======
Zheyong
On Wed, Jun 19, 2019 at 6:47 PM 1830969 <1830969@xxxxxxxxxxxxx> wrote:
I've been looking for a long time, but I can't figure out what went
wrong.This is a sandwich structure with three potential files,one xyz.in,one
run.in.An error is as follows:
[1]+ Segmentation fault (core dumped) nohup
/public/Wrok/GPUMD/GPUMD-master/src/gpumd < input.txt &
