lammps reax/c problems

Dear All Lammps Experts,

Thanks so much for your attention to this letter. I am stuck in 2 problems which call for your advices!

I use lammps-17Nov16 with Reax/c package to run an oxidation reaction between H2O and Ni-Cr-Fe grain boundary cluster. My executable is lmp_ubuntu on an ubuntu system.

Problem-1

This whole work mainly includes 2 sub-simulations, which input scripts are both attached.

Sub-simulation-1

step-(1) initialization (513 H2O + 5 O2 + 5980 Ni-Cr-Fe atoms=7834 atoms as shown in the attached picture)

step-(2) minimization

step-(3) non-reactive NVT equilibration dynamics of all the system atoms at 650 K.

In order to do the non-reactive NVT equilibration dynamics, I switched off the H-Ni, H-Cr, H-Fe, O-Ni, O-Cr, O-Fe bonds in the force field file by setting the corresponding parameters to 0 0 0, to prevent the system from any chemical reactions between (H2O + O2) group and Ni-Cr-Fe group.

In this part, I plan to get a restart file about the equilibrated system for the following Sub-simulation-2.

Sub-simulation-2

step-(4) reactive NVE production dynamics of all system atoms at 650 K.

In this step, I read the restart file from the sub-simulation-1.

In order to switch from the non-reactive equilibration dynamics to the reactive production dynamics, I switched on the H-Ni, H-Cr, H-Fe, O-Ni, O-Cr, O-Fe bonds in the same force field file.

However, the calculation always stopped updating the thermodynamic data at step 129 in the attached log.lammps file, without reporting any errors!

I once thought this problem may be caused by the switch-off effect of the H-Ni, H-Cr, H-Fe, O-Ni, O-Cr, O-Fe bonds, but it stilled stopped updating the data after I switched on these bonds.

Then I recompile the executable and got a new lmp_ubuntu executable, the same problem occurred again at the step 129.

Then I changed the computer from ubuntu PC to a supercomputer and run the simulation on 128 processors. I got the same problem.

Actually, this is not the first time that I got such problem. I checked the lammps mailing list and found others also got the similar problem. Someone said there is not such strange phenomenon in the lammps reax, and recommend lammps reax. However, the lammps reax is not recommended any more.

I don’t know why?

Is it because my initial configuration is too bad to run ?

Problem-2

I set the 5980 Ni-Cr-Fe cluster to include 60 Ni (atom type 1) + 30 Cr (atom type 2) + 10% Fe (atom type 3) by the following set commend:

set region box type/fraction 1 0.6 123 type/fraction 2 0.3 236 type/fraction 3 0.1 256 # Ni60Cr30Fe10

And I got the following data in the log.lammps file

3608 settings made for type/fraction
1808 settings made for type/fraction
585 settings made for type/fraction

However, when I checked the output species file, it shows a totally different information: Ni2239Cr1621Fe2125.

It looks so strange! Why I got a different number of Ni, Cr, Fe atoms?

Hope all you valued tutors can give me some advices, which will be highly appreciated.

I have been interested in the MD simulation, but truly frustrated. I don’t want to give up this work.

You all have my many thanks!

Best regards,

Xiaolong LIU from Korea Aerospace University.

in.stage1 (2.93 KB)

in.stage2 (1.63 KB)

log.lammps (43.9 KB)

SCCspecies.out (230 Bytes)

initial model.png

  1. The alloy should not be in a vacuum – in other words, it should be a slab with 2 directions being periodic.

  2. How did you “switch off” bonds in the force field file?

  3. The error is likely related to fix reax/c/species. My recommendation is to try the Kokkos version as both ReaxFF and Fix reax/c/species have been improved with Kokkos.

  4. The alloy concentration error is likely a bug in fix reax/c/species. Try the Kokkos version as well.

Ray