pair_style lj/long/dipole/long

Dear lammps users,
I tried to make NH3 system using MolMod database,

Here is my lammps input:

units metal
atom_style full
boundary p p p
region mybox block 0 10 0 10 0 10
create_box 2 mybox
molecule NH3 NH3_I.molecule
create_atoms 0 random 8 100 mybox mol NH3 1000

fix 1 all rigid/small molecule
  pair_style lj/long/dipole/long off off 17.5
pair_modify mix arithmetic

#int
pair_coeff 1 1 0.01576109721015 3.376
pair_coeff 2 * 0.01 0.01

#charge
set type 1 charge -0.9993
set type 2 charge +0.3331

#mass
mass 1 14.007
mass 2 1.008

neighbor 2.5 bin
thermo 100
thermo_style custom step lx ly lz vol press temp
timestep 5e-04

run 1000

write_data NH3_box_full

And here is the output:

LAMMPS (3 Mar 2020)
Created orthogonal box = (0 0 0) to (10 10 10)
   2 by 2 by 4 MPI processor grid
Read molecule NH3:
   4 atoms with max type 2
   0 bonds with max type 0
   0 angles with max type 0
   0 dihedrals with max type 0
   0 impropers with max type 0
Created 32 atoms
   create_atoms CPU = 0.0211558 secs
   create bodies CPU = 0.00193787 secs
8 rigid bodies with 32 atoms
   1.01361 = max distance from body owner to body atom
Setting atom values ...
   8 settings made for charge
Setting atom values ...
   24 settings made for charge
Neighbor list info ...
   update every 1 steps, delay 10 steps, check yes
   max neighbors/atom: 2000, page size: 100000
   master list distance cutoff = 20
   ghost atom cutoff = 20
   binsize = 10, bins = 1 1 1
   1 neighbor lists, perpetual/occasional/extra = 1 0 0
   (1) pair lj/long/dipole/long, perpetual
       attributes: half, newton on
       pair build: half/bin/newton
       stencil: half/bin/3d/newton
       bin: standard
Setting up Verlet run ...
   Unit style : metal
   Current step : 0
   Time step : 0.0005

why use this pair style at all? and why the "off off" setting?
you don't have any dipoles. you don't use a long-range solver or
coulomb. so you could just use pair style lj/cut and have the same
forces and energies.
the remainder of the input looks as if the intended pair style would
be lj/cut/coul/cut.

the error is because the implementation is (incorrectly) only checking
for the presence of the mu and torque properties if the settings are
not "off off", but the computation is accessing the respective data
structures regardless. so to avoid the crash you are seeing, you must
use:

atom_style hybrid full dipole sphere

that has other issues: when using atom style sphere, you must also include:

compute temp all temp/sphere
thermo_modify temp temp

or else the monitored and computed temperature is wrong.

axel.