Inconsistent results from different lammps versions

Dear sir,

I am running an NVT simulation of 100 water molecules using three different binaries: lammps-1Feb2014, lammps-7Dec2015, and lammps-30Jul2016, with the exact same input and initial structure (water_nvt.in and water_nvt.data).

The first two binaries produce identical results (plz see the attached water_nvt_Feb_2014.log, and water_nvt_Dec_2015.log files);

however there is significant difference with the results obtained from the lammps-30Jul2016 binary (plz see water_nvt_Jul_2016.log).

In particular, you can see at Step 0 all energies being identical with a significant difference in the pressure term.

The 30Jul2016 binary computes a significantly higher pressure compared to the Dec15 and Feb14 binaries. Of course at the very end of the simulation the computed energies are also significantly different.

thank you very much for your time and attention, and looking forward for your guidance on this problem

kind regards,

Mihalis Kavousanakis

water_nvt.data (33.5 KB)

water_nvt.in (937 Bytes)

water_nvt_Dec_2015.log (3.57 KB)

water_nvt_Feb_2014.log (2.81 KB)

water_nvt_Jul_2016.log (3.57 KB)

Dear sir,

I am running an NVT simulation of 100 water molecules using three different
binaries: lammps-1Feb2014, lammps-7Dec2015, and lammps-30Jul2016, with the
exact same input and initial structure (water_nvt.in and water_nvt.data).

The first two binaries produce identical results (plz see the attached
water_nvt_Feb_2014.log, and water_nvt_Dec_2015.log files);

however there is significant difference with the results obtained from the
lammps-30Jul2016 binary (plz see water_nvt_Jul_2016.log).

In particular, you can see at Step 0 all energies being identical with a
significant difference in the pressure term.

The 30Jul2016 binary computes a significantly higher pressure compared to
the Dec15 and Feb14 binaries. Of course at the very end of the simulation
the computed energies are also significantly different.

thank you very much for your time and attention, and looking forward for
your guidance on this problem

when noticing differences between LAMMPS versions, please be certain
to first have a good look at: http://lammps.sandia.gov/bug.html
(and the corresponding pages for preceding years, if needed).

in your case, please have a particular look at the comments for the 14
March 2016 patch, which fixes a long-standing bug in pair style
lj/charmm/coul/charmm.

axel.

You can also use the compute pressure command

with its different options to isolate exactly which

term(s) in the pressure are different in the newer

version. If it is the pair term for the pair style

Axel mentioned, then that is undoubtably the change.

Steve