Dear LAMMPS users,
Hello to all,
I have trouble performing reactive molecular dynamics using ReaxFF.
To see the lithiation at the interface between anode and lithium,
I constructed the interface system after enough relaxation of anode and lithium slabs.
All of my test simulations were performed at 1 atm and 298 K using LAMMPS version 13Mar2017
The problem was that all simulations were terminated with the following error message before hundreds to thousands of steps have passed.
¡°ERROR: Non-numeric pressure - simulation unstable (…/fix_nh.cpp:1029)¡±
(please find the attached log files)
I know that ReaxFF is a very complex potential and that kind of error can occur if the initial structure is unstable.
However, when I printed out thermodynamic quantity and configuration every step, I could not find anything particularly strange (e.g., nan).
Also, when I continued the job using the atomic position and velocity, printed several hundred steps before termination,
the calculation proceeded normally hundreds or thousands of steps with the error message at the end.
Any help would be helpful.
Thank you very much
Best regards,
Se Hun Joo
log.lammps (1.16 MB)
Dear LAMMPS users,
Hello to all,
I have trouble performing reactive molecular dynamics using ReaxFF.
To see the lithiation at the interface between anode and lithium,
I constructed the interface system after enough relaxation of anode and
lithium slabs.
All of my test simulations were performed at 1 atm and 298 K using LAMMPS
version 13Mar2017
The problem was that all simulations were terminated with the following
error message before hundreds to thousands of steps have passed.
“ERROR: Non-numeric pressure - simulation unstable (../fix_nh.cpp:1029)”
(please find the attached log files)
I know that ReaxFF is a very complex potential and that kind of error can
occur if the initial structure is unstable.
However, when I printed out thermodynamic quantity and configuration every
step, I could not find anything particularly strange (e.g., nan).
the error message you quote is exactly caused by your forces and thus your
pressure becoming NaN, i.e. non-numeric.
due to using the npt integrator, this is detected and the simulation
stopped immediately, as with the box update, all coordinates would become
NaN and thus the simulation meaningless.
the most common causes for NaN are bad geometry, bad/unsuitable force field
parameters, and too large a time step.
Also, when I continued the job using the atomic position and velocity,
printed several hundred steps before termination,
the calculation proceeded normally hundreds or thousands of steps with the
error message at the end.
does your simulation conserve energy sufficiently well, when you perform
such a restart after a crash with fix nve instead of fix npt?
axel.