A problem with the error Non-numeric box dimensions - simulation unstable

Dear all

    I'm trying to simulate the evaporation of water from nanoporous graphene. The whole system is simulated in NVT ensemble about 100,000 timesteps for equlibrium and in NPT ensemble followed about 20,000,000 timesteps for evaporation. However, the error ( Non-numeric box dimensions - simulation unstable ) occurred at about 4,000,000 timesteps. There is no abnormal temperature or pressure along the whole simulation. I supposed that the pressure of the box became higher with evaporation and finally the lammps couldn't change the box size any more. I don't know if I am right and my input is as followed. Hope and thanks for your advice!

# Initialzation

units real
atom_style full
boundary p p p

# Force fields

pair_style lj/cut/coul/long 10 12
bond_style harmonic
angle_style harmonic
read_data watergraphene-$rnm.data
kspace_style pppm 1.0e-6

set type 2 charge 0.41
set type 3 charge -0.82

pair_coeff 1 1 0.1479999981 3.6170487995 # C
pair_coeff 2 * 0.0000000000 0.0000000000 # H
pair_coeff 3 3 0.1554164124 3.1655200879 # O
pair_coeff 1 3 0.1048294600 3.190 # C & O

group C type 1
group H type 2
group O type 3
group water union O H

bond_coeff 1 480.0000 1.4200
bond_coeff 2 540.6336 0.9600

angle_coeff 1 90.0000 120.0000
angle_coeff 2 50.0000 104.5000
angle_coeff 3 0 0

delete_bonds all angle 3 remove

# Settings

neighbor 2.0 bin
neigh_modify delay 5

velocity water create ${temp1} 123456
velocity C set 0 0 0
timestep 1
write_dump all atom dump.setup_Watergraphene_rnm\_{temp2}K.lammpstrj

# Run for equlibrium

# Fix

group none subtract water water
change_box none z final -300 400

fix mem C setforce 0 0 0
fix 1 water nvt temp \{temp1\} {temp1} 100 drag 0.2

# Output

thermo 1000
restart 5000000 restartWatergraphene_rnm\_{temp2}K.*

# Dump

dump 1 all custom 10000 dump_Watergraphene_rnm\_equil\_{temp2}K.lammpstrj type element mass x y z # 10000

# Run

run 100000
unfix 1
undump 1

# Heating & Evaporation

dump 1 all custom 100000 dump_Watergraphene_rnm\_{temp2}K.lammpstrj type element mass x y z #1000000
fix 1 water npt temp \{temp2\} {temp2} 100 x 1.0 1.0 0.2 y 1.0 1.0 0.2 z 1.0 1.0 0.2 drag 0.2
run 20000000

Dear all

    I'm trying to simulate the evaporation of water from nanoporous graphene. The whole system is simulated in NVT ensemble about 100,000 timesteps for equlibrium and in NPT ensemble followed about 20,000,000 timesteps for evaporation. However, the error ( Non-numeric box dimensions - simulation unstable ) occurred at about 4,000,000 timesteps. There is no abnormal temperature or pressure along the whole simulation. I supposed that the pressure of the box became higher with evaporation and finally the lammps couldn't change the box size any more. I don't know if I am right and my input is as followed. Hope and thanks for your advice!

i suspect that this is due to a very bad model. i don't quite
understand the purpose of using an NPT nose-hoover thermostat and
barostat, which is designed for equilibrium simulations for something
that is so very much a non-equilibrium system. moreover, have you
watched your bond energy? with allowing the system to expand in every
direction, your water is likely to try ripping your graphene apart,
but in a very unphysical way through the stretching of the box and the
corresponding remapping of the graphene coordinates.

in short:

bad model -> bad simulation.
bad simulation -> crash

axel.

Dear Axel,

Thanks for your reply!

I chose the NPT ensenmble, because I want to perform a simulation at a normal temperature and pressure the same as my experiment, and I' ve read some other articles about water evaporation using NPT ensemble likewise. If I only change the z-box length, would it behave as normal and corresponding with my experiment? Or I have to substitute another ensemble?

Thanks!

--Chen Bo

Dear Axel,

Thanks for your reply!

I chose the NPT ensenmble, because I want to perform a simulation at a normal temperature and pressure the same as my experiment, and I' ve read some other articles about water evaporation using NPT ensemble likewise. If I only change the z-box length, would it behave as normal and corresponding with my experiment? Or I have to substitute another ensemble?

that is a question, that you would have to answer for yourself. after
all, if you publish your results, you will have to justify your
choices. you will need something that has a physically sound
foundation. if you want to model and experiment, the core problem is
usually the different length scales. it is more than likely, that the
process you want to study is not happening on the accessible length
scale for atomistic MD simulations. that in turn usually means, that
you cannot extract all useful information from just a single
simulation, but will have to set up multiple runs, each of which
representing one particular aspect of your problem well.

this is all that can be said on such an abstract and general level.

axel.