Problem in LAMMPS-ReaxFF

Dear all,

I am running Reaxff simulations in LAMMPS with an aqueous NaCl system. I could run the simulation till 3ns. However, when I try to restart the calculation from a previous time point (not the last one), the calculation starts but gets stuck without printing any error message. This seemed to be strange to me as the calculation gets stuck at a step which was successfully executed previously. I am using LAMMPS-7Jan2022 version on 2 nodes with 48 tasks per node.

The relevant portion the input script is (full input script attached)
in.input_reax (1.6 KB)
pair_style reaxff control_lmp minhbonds 100000
pair_coeff * * field_electrolyte.ff O H Cl Na
fix charge all qeq/reaxff 1 0.0 10.0 1.0e-6 reaxff
fix spe all reaxff/species 2000 1 2000 species.out element O H Cl Na
fix b all reaxff/bonds 4000 bond.out
neighbor 2.0 bin
neigh_modify every 1 delay 0 check yes

The relevant portion of the control file:
tabulate_long_range 10000 ! denotes the granularity of long range tabulation, 0 means no tabulation
energy_update_freq 1

nbrhood_cutoff 5.0 ! near neighbors cutoff for bond calculations in A
hbond_cutoff 7.0 ! cutoff distance for hydrogen bond interactions
bond_graph_cutoff 0.3 ! bond strength cutoff for bond graphs
thb_cutoff 0.001 ! cutoff value for three body interactions

I thank you in advance for your help/suggestions.

Is there any output at all from LAMMPS?

Normally, in a situation like this, I comment out all lines but read_restart and run 0 (and possibly some other lines LAMMPS needs), to see how severe the hang is. If the simulation works, I would sequentially enable the lines one by one, to see which one causes the problem. minhbonds 100000 looks especially suspicious as it is much more than normally is needed.

Additionally, try to convert the restart file to a data file using restart2data and run the simulation from that data file using read_data. Sometimes fixes can write data to a restart file, and presumably sometimes when one fix reads that data back while other fixes start from scratch, the simulation could get stuck in some unrecoverable state.