LAMMPS version comparison

Hello Ray,

referring to an earlier thread regarding strange output from the 29 Jan
2014 LAMMPS version (latest on our cluster):
http://lammps.sandia.gov/threads/msg44161.html
I would like to provide an input set which has produced similar issues
with a pure water system (files attached).
Thank you,

Osvalds

control.lmp (1.44 KB)

ffield (25.2 KB)

h2o.dat (3.74 KB)

in.combined (1.31 KB)

species.reaxc (44.6 KB)

Hello Ray,

referring to an earlier thread regarding strange output from the 29 Jan
2014 LAMMPS version (latest on our cluster):
http://lammps.sandia.gov/threads/msg44161.html
I would like to provide an input set which has produced similar issues
with a pure water system (files attached).

​i don't see any problems with this input on my desktop with the current
version of LAMMPS (11 May​ 2014).

a few comments:

- please note that the input as such requires creating a directory. in
order to make it easy for people to run this, please either don't write
files to subdirectories or include "shell mkdir" commands to make certain
that directory is created, if needed.

- why the excessive minimization, if you're going to run an equilibration
afterwards and heating it up again?

- since you have a tiny system, it is not a good idea to start with NPT
right away. that can result in excessive volume fluctuations. better to run
with constant volume for a bit and using either fix langevin or temp/csvr
to thermalize the system and then switch to fix npt later when things have
settled a bit.

axel.

It ran fine for me with a 2May14 version as well.

On top of what Axel suggested, to make it easy for you to get help
from the mailing list, it would be best if you provide an input deck
that reproduces the problem "right away", instead of requiring people
to run tens of thousands of steps.

Also, reiterate what is the problem and what exactly is the error
message that you see is helpful, too.

Ray

I have not tried the simulation with versions after Jan 29, but it ran
fine with the 11 Jul 2013-ICMS version. The problem could be detected by
occurrence of species other than H2O (!) in the species output file (not
observed with 11 Jul 2013-ICMS) starting a few hundreds of steps after
minimization (file attached to previous e-mail). In pure water
simulations I did not see any error messages. Thank you,

-Osvalds

I have not tried the simulation with versions after Jan 29, but it ran

I strongly suggest you try with the most current version (12May14) since no
one is likely to spend time in debugging an outdated version.

fine with the 11 Jul 2013-ICMS version. The problem could be detected by
occurrence of species other than H2O (!) in the species output file (not

I don't understand what this means. What do you mean by "detected"? What
exactly was the error message?

observed with 11 Jul 2013-ICMS) starting a few hundreds of steps after
minimization (file attached to previous e-mail). In pure water

Isn't the data file you attached pure water?

simulations I did not see any error messages. Thank you,

Sorry for the confusion. The dissociation of pure water (with files that
were sent) at 300K/4000 atm was considered as an indication of error
indeed. Nevertheless I will try a newer version.

-Osvalds

Sorry for the confusion. The dissociation of pure water (with files that
were sent) at 300K/4000 atm was considered as an indication of error
indeed. Nevertheless I will try a newer version.

​i would say, this is *much* more likely due to the small system size,
particularly in y-direction. here the water molecules have no choice but to
mostly interact with their own periodic images. this will result in very
large pressure fluctuations which in turn will produce large relative
volume changes, but also all kinds of finite size artifacts. while ~64
waters is generally considered the lower bound for reasonable simulations
of bulk water, that would apply to a cubic simulation box.

axel.

Dear Sirs,

I did a comparison between the 13 May 2014 and 11 Jul 2013-ICMS verions
(input and species output attached) with an expanded system (132 H2O) to
reduce small cell size effect. Since energies of both were in agreement
(using the same minimized structure as input) I could suggest that the
problem might be related to fix species (checking with my own bonding
table based species script confirms that). I should add that upon
restart the species output seems to be correct again.

Osvalds

h2o.zip (138 KB)

Dear Osvalds,

17.05.2014, 00:01, "Osvalds Verners" <[email protected]>:

Dear Sirs,

I did a comparison between the 13 May 2014 and 11 Jul 2013-ICMS verions
(input and species output attached) with an expanded system (132 H2O) to
reduce small cell size effect. Since energies of both were in agreement
(using the same minimized structure as input) I could suggest that the
problem might be related to fix species (checking with my own bonding

what is this "problem"? Do you have different results with different LAMMPS versions? Since you didn't provide the output for identical input, there is no direct comparison between two species files... Anyway, AFAIR, there were no changes in fix reax/c/species that would result in different molecules. If you call the "problem" the fact that you observe water dissociation - as Axel said, it is most likely related to the dynamics of your system, not the molecule finding algorithm. You can try to tune it by changing bond order cutoff in fix reax/c/species line in input script, but keep in mind that it will just give different numbers for the same physics.

table based species script confirms that). I should add that upon
restart the species output seems to be correct again.

Since you don't provide your script nor its parameters or results, there is nothing to say on that. You know yourself what is the difference between it and fix reax/c/species.

Best regards,
Oleg