Reax/c : error opening the force filed file! terminating... (sic)

Dear all,

I have an issue that is giving me a strong headache…

I use reac/c ff in my simulation and I encounter the error given below (Output 2).

My pb is that, when seeking in the cpp files, I only see that error when Lammps cannot open the file (you could tell me it is quite obvious) but Lammps does read it since when I modify the Reaxff file, the output is modified (Output 1, related to the X element in the ffield file).

That means that the error is elsewhere, doesn’t it ?

Any advice would be appreciated, I’m becoming crazy

Thank you,

Output 1

Hi Alexandre,

It appears you are using the special_bonds command, which is not allowed with ReaxFF. Please see the http://lammps.sandia.gov/doc/special_bonds.html doc page.

Remove irrelevant/incorrect commands and try again. Also, it is not advised to alter the force field parameter files.

Cheers,
Ray

Dear all,

I have an issue that is giving me a strong headache...

I use reac/c ff in my simulation and I encounter the error given below
(Output 2).

My pb is that, when seeking in the cpp files, I only see that error when
Lammps cannot open the file (you could tell me it is quite obvious) but
Lammps *does* read it since when I modify the Reaxff file, the output is
modified (Output 1, related to the X element in the ffield file).

That means that the error is elsewhere, doesn't it ?

impossible to tell without seeing a (complete, minimal) input deck
that reproduces the issue.

axel.

Dear Ray and Axel,

I’m not using special_bonds but read_data, so maybe the special_bonds came from that. Anyway it is not the problem here since I verified I still have the pb with no special_bonds.

You will find attached files that reproduces the pb (ffield.reax is ffield.reax.Fe_O_C_H from Lammps distribution).

Do you reproduce the error I get (test.out)?

That is a dummy test (data are not consistent, such as charge, element, etc…) because I work on Si and I wanted to use a legit ffield file.

Thank you for your help

Alexandre

test.out (4.33 KB)

ffield.reax (12.7 KB)

test.in (974 Bytes)

That seems is not coming from my files, I tested with example/reax/Fe_O_C_H for 2 Lammps distributions (6Apr13 and 3Feb13) and I got the same error.

Can it be the way I compile Lammps ?

That is what I did :

  1. in lib/reax

make -f Makefile.gfortran

edit Makefile.lammps : replace all the -xxx by -lgfortran

delete -L/opt/intel/fce/10.0.023/lib in the syspath

  1. in src

export PATH=$PATH:/softs/OPENMPI/openmpi-1.6.0-gcc/bin

export MPI_HOME=/softs/OPENMPI/openmpi-1.6.0-gcc/

export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/softs/OPENMPI/openmpi-1.6.0-gcc/lib

make openmpi yes-user-reaxc yes-reax yes-user-misc yes-manybody yes-kspace yes-granular FFT_INC= FFT_PATH= FFT_LIB=

Thank you

Alexandre

Dear Ray and Axel,

I'm not using special_bonds but read_data, so maybe the special_bonds came
from that. Anyway it is not the problem here since I verified I still have
the pb with no special_bonds.

You will find attached files that reproduces the pb (ffield.reax is
ffield.reax.Fe_O_C_H from Lammps distribution).

Do you reproduce the error I get (test.out)?

no. that input works for me without a hitch. perhaps you have a
permission mismatch or some other strange thing going on at the
machine you are running on. have you tried on a desktop?

That is a dummy test (data are not consistent, such as charge, element,
etc...) because I work on Si and I wanted to use a legit ffield file.

that is fine.

axel.

Dear Ray and Axel,

I'm not using special_bonds but read_data, so maybe the special_bonds came

the special_bonds command defaults to lj/coul 0.0 0.0 1.0 however,
since reaxff computes bonding on the fly, you either must not have any
bonds defined in your data file or reset it to lj/coul 1.0 1.0 1.0.
otherwise, you are missing entries in your neighbor list and your
calculation will be completely bogus.

from that. Anyway it is not the problem here since I verified I still have
the pb with no special_bonds.

yes, this is a different problem. the special_bonds (or rather having
bonds in your data file) problems will hit you later.

axel.

That seems is not coming from my files, I tested with example/reax/Fe_O_C_H
for 2 Lammps distributions (6Apr13 and 3Feb13) and I got the same error.

Can it be the way I compile Lammps ?

That is what I did :

1. in lib/reax

make -f Makefile.gfortran

edit Makefile.lammps : replace all the -xxx by -lgfortran

delete -L/opt/intel/fce/10.0.023/lib in the syspath

2. in src

export PATH=$PATH:/softs/OPENMPI/openmpi-1.6.0-gcc/bin

export MPI_HOME=/softs/OPENMPI/openmpi-1.6.0-gcc/

export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/softs/OPENMPI/openmpi-1.6.0-gcc/lib

make openmpi yes-user-reaxc yes-reax yes-user-misc yes-manybody yes-kspace
yes-granular FFT_INC= FFT_PATH= FFT_LIB=

you are using reax/c, so there is no need to compile the reax lib and
install the reax package.

axel.

I thought about some permission stuff some days ago and set chmod777 on the file, but with no effect...
And I often used material files (for tersoff or sw) without a pb

I'll check with my machine admin

Alexandre

-----Message d'origine-----

The issue is solved, Axel you were right (as always I guess) : it was a "strange thing going on at the machine".
Apparently reax/c (on the contrary of what Lammps usually does) wants the ffield file on all the cluster nodes, not only on the master, and we set up our MPI by copying input files only to the master...

Thanks again
Have a nice day,

Alexandre

-----Message d'origine-----