Lost Atoms with fix wall/reflect

Dear Lammps users,

When using a p p f boundary conditions, I used fix wall/reflect in z direction. However, I got the following errors:
fix wall all wall/reflect zlo EDGE zhi EDGE

ERROR: Lost atoms: original 12 current 10 (…/thermo.cpp:438)

Furthermore, with fix wall all wall/lj126 zlo EDGE 1.0 1.0 2.5, the following errors came out:

ERROR on proc 0: Particle on or inside fix wall surface (…/fix_wall_lj126.cpp:91)

To mitigate the errors, I tried in a huge simulation box and small number atoms and visually no atoms are outside the simulation box or at the boundaries.I can not find the reason of this problem.

Could you please help me with this?

Regards,
Michael

In the same system, with SPCE water and Na+ and Cl- ions, the error information would disappear if the charge of Na and Cl are set to 0. The coordinates are the same and I am sure no atoms locate on the wall or inside the wall. It is so confusing.

Best regards,
Michael

when you turn off charges, you significantly change the forces (usually you lower them).

the lost atoms errors in both cases mean, that atoms move too far in a single time step, which means that either the time step is to large, or forces/velocities are huge, e.g. when having bad/inconsistent force field parameters.

as noted before, you are confused because you are lacking understanding of some of the fundamentals of doing MD simulations. to get those, you need in person oversight and tutoring. continue without at your own risk, but be aware that you are unlikely to receive much (more) help from the mailing list beyond what you already did, as your problems are not really LAMMPS specific problems but of general nature. please also keep in mind that there are many subtle ways (i.e. where the simulation does not crash) where MD simulations are be very wrong, and without proper tutoring, you are likely to (re-)make many of the mistakes that people have done and learned from in the last 20-30. MD is as much a craft as it is science and thus you need - like in other crafts - transfer of knowledge from a practitioner to learn the many little things (the so-called “residual knowledge”) that are passed from one practitioner to the next.

axel.

Thank you, Axel. I really need knowledge before the operation. I’ll go to read some text books and also find some help from people around.
For the above case, when the pair style changes from lj/cut/coul/wolf to lj/cut/coul/dsf, the simulation can be run without errors. My previous setting is problematic in pair forces calculation.

Best Regards,
Michael

lj/cut/coul/wolf and lj/cut/coul/dsf are very similar, if one of them triggers problems and the other does not, it means that both are marginal and neither simulation is to be trusted.

as I pointed out before: if a simulation finishes without errors, it by no means guarantees that it was a meaningful simulation.

axel.

Thank you, Axel. I need to get work on their theories behind. I am really confused with these pair_styles.
In the simulation case, it is very simple with only SPCE water and NaCl electrolyte. The reason may be due to the high NaCl concentration (3 mol/L).

Best Regards,
Michael