[lammps-users] low temperature in NPT

LAMMPS LAMMPS Mailing List Mirror
1

Hello All:
The in file is:
units metal
atom_style charge
read_data gdc510.data
neighbor 2.0 bin
neigh_modify delay 0 every 20 check yes
kspace_style ewald 1e-8
pair_style buck/coul/long 10.0
pair_coeff 1 1 9547.96 0.2192 32.0
pair_coeff 1 2 1885.75 0.3399 20.34
pair_coeff 1 3 1809.68 0.3547 20.40
pair_coeff 2 2 0.00000 1.0000 0.000
pair_coeff 2 3 0.00000 1.0000 0.000
pair_coeff 3 3 0.00000 1.0000 0.000
velocity all create 5 87287 dist gaussian
fix 1 all npt 5 5 0.1 aniso 0.0 0.0 0.0 0.0 0.0 0.0 0.1 drag 1.0
thermo_style custom step temp lx ly lz pxx pyy etotal pzz
thermo 100
timestep 0.001
run 20000
When the temperature in NPT ensemble is 5 or lower , the job will terminate, and the massage in log is:
ERROR on proc 0: Failed to allocate -1077060736 bytes for array ewald:eg
ERROR on proc 1: Failed to allocate -1077060736 bytes for array ewald:eg
[cli_0]: aborting job:
application called MPI_Abort(MPI_COMM_WORLD, 1) - process 0
[cli_1]: aborting job:
application called MPI_Abort(MPI_COMM_WORLD, 1) - process 1
rank 0 in job 1 compute-1-3.local_47103 caused collective abort of all ranks
exit status of rank 0: killed by signal 9
But if the desired temperature is higher, f.e. 20 or 30. the job will run correctly. What’s wrong with the model when T is low in the simulation?

2

2010/3/22 yunjunchen1986 <[email protected]...>:

Hello All:
The in file is:

this is better, but to _reproduce_ your calculation one also would
need a data file, right?

units    metal
atom\_style      charge
read\_data       gdc510\.data
neighbor   2\.0 bin
neigh\_modify   delay 0 every 20 check yes
kspace\_style     ewald 1e\-8
pair\_style buck/coul/long  10\.0
pair\_coeff      1 1 9547\.96    0\.2192    32\.0
pair\_coeff      1 2 1885\.75    0\.3399    20\.34
pair\_coeff      1 3 1809\.68    0\.3547    20\.40

the following parameters look odd:

pair\_coeff      2 2 0\.00000    1\.0000    0\.000
pair\_coeff      2 3 0\.00000    1\.0000    0\.000
pair\_coeff      3 3 0\.00000    1\.0000    0\.000

what is the purpose of them?

velocity  all  create 5  87287 dist gaussian
 fix          1  all npt 5 5 0\.1 aniso 0\.0 0\.0 0\.0 0\.0 0\.0 0\.0 0\.1 drag

1.0
thermo_style custom step temp lx ly lz pxx pyy etotal pzz
thermo 100
timestep 0.001
run 20000
When the temperature in NPT ensemble is 5 or lower , the job will terminate,
and the massage in log is:
ERROR on proc 0: Failed to allocate -1077060736 bytes for array
ewald:eg
ERROR on proc 1: Failed to allocate -1077060736 bytes for array
ewald:eg

indeed you are running out of address space. have you checked
if your system is collapsing? did you visualize the trajectory?

   \[cli\_0\]: aborting job:
   application called MPI\_Abort\(MPI\_COMM\_WORLD, 1\) \- process 0
   \[cli\_1\]: aborting job:
  application called MPI\_Abort\(MPI\_COMM\_WORLD, 1\) \- process 1
   rank 0 in job 1  compute\-1\-3\.local\_47103   caused collective abort of

all ranks
exit status of rank 0: killed by signal 9
But if the desired temperature is higher, f.e. 20 or 30. the job will run
correctly. What's wrong with the model when T is low in the simulation?

what is your simulation supposed to model?

cheers,
   axel.

3

When the temperature in NPT ensemble is 5 or lower , the job will terminate, and the massage in log is:

> ERROR on proc 0: Failed to allocate -1077060736 bytes for array ewald:eg

This is probably b/c your system is undergoing some very large volume
fluctuation and the Ewald solver is attemping to reallocate huge arrays that
are overflowing. Have you run, printing out thermodyamics very frequently, and
verified that your system is running in a stable manner, up to the point
that the error occurs?

Steve