Variation in x y z coordinates in dump files due to restart

LAMMPS LAMMPS Mailing List Mirror
1

Dear all,

I am running a polymer simulation with bond, angle, dihedral and nonbonded potentials. The sample was deformed under nvt ensemble.

the specific commands are as folows:

#read_data 1000_PE_LAMMPS_INPUT
read_restart pmr.restart.5000000

pair_modify tail yes

neighbor 1.0 bin
timestep 1.0

run_style verlet

fix 1 all deform 1 x vel 0.0003 units box
fix 2 all nvt temp 100 100 500

When the simulation was run directly from the data file and then by using a intermediate restart file, the output at a given step say 5183000 were compared and were found to be varying.

First few lines from the dump file of both runs are shown below,
For run from original datafile,
ITEM: ATOMS id x y z

1 -416.997 9.7084 11.587
2 -418.323 8.93339 11.6771
3 -419.256 9.61928 12.6947
4 -420.546 8.83569 12.88
5 -420.327 7.45046 13.5581
6 -421.494 6.50961 13.3391
7 -421.181 5.04872 13.6829
8 -421.296 4.74522 15.1938
9 -421.424 3.26868 15.6285
10 -421.471 3.11456 17.1354

For run from restart file at 5000000

ITEM: ATOMS id x y z

1 -417.077 9.70336 18.4055
2 -417.566 9.08762 17.1175
3 -418.873 9.73894 16.5819
4 -419.58 9.15077 15.3765
5 -419.859 7.66144 15.5719
6 -420.442 6.9203 14.349
7 -420.553 5.42464 14.6078
8 -421.461 5.14037 15.7876
9 -421.411 3.69951 16.2402
10 -422.2 3.43722 17.526

The variation in the y and z coordinates seem significant. The correlation between them along each axis is given below,
X_original and X_ restart --> 0.997742
Y_original and Y_restart --> 0.940218
Z_original and Z_restart --> 0.935896

Why do these variations occur, and are variations in the acceptable range?

Thanks
Vivek Ananth
IMSc,Chennai

2

Dear all,

I am running a polymer simulation with bond, angle, dihedral and nonbonded
potentials. The sample was deformed under nvt ensemble.

the specific commands are as folows:

#read_data 1000_PE_LAMMPS_INPUT
read_restart pmr.restart.5000000

pair_modify tail yes
neighbor 1.0 bin
timestep 1.0
run_style verlet

fix 1 all deform 1 x vel 0.0003 units box
fix 2 all nvt temp 100 100 500

When the simulation was run directly from the data file and then by using
a intermediate restart file, the output at a given step say 5183000 were
compared and were found to be varying.

First few lines from the dump file of both runs are shown below,
For run from original datafile,
ITEM: ATOMS id x y z
1 -416.997 9.7084 11.587
2 -418.323 8.93339 11.6771
3 -419.256 9.61928 12.6947
4 -420.546 8.83569 12.88
5 -420.327 7.45046 13.5581
6 -421.494 6.50961 13.3391
7 -421.181 5.04872 13.6829
8 -421.296 4.74522 15.1938
9 -421.424 3.26868 15.6285
10 -421.471 3.11456 17.1354

For run from restart file at 5000000
ITEM: ATOMS id x y z
1 -417.077 9.70336 18.4055
2 -417.566 9.08762 17.1175
3 -418.873 9.73894 16.5819
4 -419.58 9.15077 15.3765
5 -419.859 7.66144 15.5719
6 -420.442 6.9203 14.349
7 -420.553 5.42464 14.6078
8 -421.461 5.14037 15.7876
9 -421.411 3.69951 16.2402
10 -422.2 3.43722 17.526

The variation in the y and z coordinates seem significant. The correlation
between them along each axis is given below,
X_original and X_ restart --> 0.997742
Y_original and Y_restart --> 0.940218
Z_original and Z_restart --> 0.935896

Why do these variations occur, and are variations in the acceptable range?

​there is not enough information here to confirm that you ran your
simulations and restart correctly, but you should be aware, that LAMMPS
uses floating point math and thus any operation that changes the order of
operation can lead to small differences, e.g. when summing up forces due to
floating point math not being associative. for more details on that, i
recommend reading the following two blog posts:
http://blog.reverberate.org/2014/09/what-every-computer-programmer-should.html
http://blog.reverberate.org/2016/02/06/floating-point-demystified-part2.html

​restarting enforces building a new neighbor list and ​thus may change the
order of how positions are stored. same goes for using a different number
of processors.
after restarting positions and energies should be the same as right before
the restart and then slowly diverge exponentially.

http://lammps.sandia.gov/doc/Section_errors.html#err-1

axel.

3

Dear Axel,

Thanks for the reply.
The dump file shown is 200,000 steps from the restart.
Given the floating point math, does the variation in z-coordinate from 11.587 to 18.405 acceptable? Or does this indicate an error in simulation.

Thanks again,
Vivek Ananth