Dear Lammps Developer and users,
I’m using reaxff and there is no other difference in the geometry file other than the atom type. I have just substituted few of the Fe atoms with Cr atoms. The simulation works perfectly for Fe+Ni atoms but gives me Non-numeric atom coords error for Fe+Ni+Cr system. How can I solve this, since I believe that geometry is not the problem here. I have also checked for overlapping atoms. My options here were to play with the bin size, frequency of build-up, timestep size in the input script. I have tried all but nothing helped.
Has the ReaxFF parameter file been parameterized for all the elements you are using (that is, does the publication describing it confirm this, it is irrelevant whether all elements are listed in the file, they may just been left in the file for convenience).
ffield_CrNiAlFe (32.9 KB)
I’m not sure about it. I obtained this parameters from Dr. van Duin. I tried to work on system which was 9 times smaller than this one, which is 400 metal atoms reacting with 100 water molecules. It does work for smaller systems.
Then you better get some certainty. The comment in the file says:
Reactive MD-force field: Shin et al. ACS Catalysis 2015 Cr/Fe/C/O/H/S
That casts a doubt on the parameters for Ni. But it also mentions a publication that you can look up and double check.
That is irrelevant. At any rate, you have “debug” your input (rather than just modifying things systematically). One possible way could be to just do a “run 0” and then a “write_dump” command with custom format where you output the types, atom IDs, charges, positions and forces and try to find atoms with unusually high forces or crazy charges or forces that are already NaN or Inf. Those atoms are obviously giving you problems, so you can have a closer look at atom types or geometry for close contacts.
Which is the exact problem is impossible to say from remote. Please consult local people have have some experience in “debugging” MD simulations. It is not likely that somebody from remote can (or want to) do this for you.
