Segmentation fault using hybrid reax/c NULL reax/c NULL lj/charmm/coul/long

LAMMPS LAMMPS Mailing List Mirror
1

Dear LAMMPS Users,

I am trying to calculate morphology of polymer in solvent (ex water) using two different reax/c force fields and interaction between polymer and water atoms using lj/charmm/coul/long.
But I am getting segmentation error…

my input file is:

units real
dimension 3
boundary p p p
atom_style charge

---------- Create Atoms ---------------------

read_data data_right_middle_left.FC

group PssPedot id <= 14112
group water id >= 14113

---------- Define Interatomic Potential ---------------------

pair_style hybrid reax/c NULL checkqeq no safezone 16 mincap 1000 reax/c NULL checkqeq no safezone 16 mincap 1000 lj/charmm/coul/long 8.0 10.0 10.0
pair_coeff * * reax/c 1 ffield_Castro_CombFlame C C O H S C O O H H S NULL NULL
pair_coeff * * reax/c 2 ffield.protein NULL NULL NULL NULL NULL NULL NULL NULL NULL NULL NULL H O

pair_coeff 4 12 lj/charmm/coul/long 0.02 2.
pair_coeff 910 12 lj/charmm/coul/long 0.02 2.
pair_coeff 12 12 lj/charmm/coul/long 0.055 3.
pair_coeff 6 12 lj/charmm/coul/long 0.055 3.
pair_coeff 3 12 lj/charmm/coul/long 0.063 2.6
pair_coeff 78 12 lj/charmm/coul/long 0.063 2.6
pair_coeff 5 12 lj/charmm/coul/long 0.063 3.
pair_coeff 11 12 lj/charmm/coul/long 0.063 3.
pair_coeff 4 13 lj/charmm/coul/long 0.063 2.6
pair_coeff 910 13 lj/charmm/coul/long 0.063 2.6
pair_coeff 12 13 lj/charmm/coul/long 0.173 3.6
pair_coeff 6 13 lj/charmm/coul/long 0.173 3.6
pair_coeff 3 13 lj/charmm/coul/long 0.2 3.2
pair_coeff 78 13 lj/charmm/coul/long 0.2 3.2
pair_coeff 5 13 lj/charmm/coul/long 0.2 3.6
pair_coeff 11 13 lj/charmm/coul/long 0.2 3.6
kspace_style pppm 1.0e-4
dielectric 80.
neighbor 2.10 bin # build the neighbor list with bin style (~N/P)
neigh_modify delay 100 check yes

dump 10 all xyz 1000 dump.xyz
dump mydump all custom 10000 dump.pedot id type q x y z
dump_modify mydump sort id

---------- Run Minimization ---------------------

velocity water create 100 1234 dist gaussian mom yes rot yes

fix 5 PssPedot setforce 0.0 0.0 0.0

fix 2 all npt temp 100 100 1. iso 0.1 0.1 1000.
thermo 1000
thermo_style custom step temp pe lx ly lz press
timestep 0.2
run 50000

any help will be greatly appreciated.

2

Ray can look at this and figure out if reax/c supports it correctly.

He can also comment on whether mixing 2 different ReaxFF parameterizations

for different parts of a single model is a good idea. I’m guessing not.

What does the QEq part for each do?

Steve

3

Yes, reax/c supports being called more than once in hybrid or hybrid/overlay.

I think mixing 2 sets of reaxff parameters is the same as hybridizing reaxff with other pair styles - not a big problem if you define interactions between cross terms properly.

Sandeep turned off fix qeq/reax, so he avoided this problem. If qeq/reax were to be used, then two fixes must be defined and be separately applied to different groups.

Ray

4

Sandeep,

I can run a test problem with 2 reax/c styles over hybird/overy. Can you try updating your LAMMPS version to a more current one, then report back when you see the seg fault? You can also post the data file so I can run your script and take a look.

Ray

5

Dear Ray,

Thanks for your kind mail. I updated the Lammps to 10 August 2015 stable version but still facing segmentation problem. I am attaching my data and input file. Please have a look and can you kindly post your test run so I can also have a look at it.

data_right_middle_left.FC (735 KB)

ffield.protein (30.4 KB)

ffield_Castro_CombFlame (23.6 KB)

opt.in (2.99 KB)

submit.sh (251 Bytes)

6

When you said seg fault, I expected it to occur right away after the simulation starts. However, I can run the first 300 steps okay. If it seg faults later, it is highly likely a result of bad dynamics. After all, mixing 2 reaxff potentials in this way is highly risky and not recommended.

If you still want me to look at it, please generate a restart file that fails right before (within a couple hundred steps is okay) the seg fault.

Ray

7

“After all, mixing 2 reaxff potentials in this way is highly risky and not recommended.”

Ray is being polite.

Any calculations you make mixing two ReaxFF will not be considered valid unless you go through a complete validation process and even then you are violating the fundamental basis for the method. Read Adri van Duin’s papers and create a new ReaxFF for your system.

Jim Kress

8

Dear Lammps Users,

Thanks for your kind advises. Now I hybrid reax/c with SPC/E (water) potential which is working fine.