[lammps-users] ERROR: Dihedral coeff for hybrid has invalid style

Hello,

A previous version of lammps:

   LAMMPS (8 Oct 2010-ICMS)

runs a particular script + input data correctly. A recent version:

   LAMMPS (18 Jan 2011-ICMS)

does not (I kept the old binary version for comparison). This is taking
place on the same machine using the same makefile, input data file, and
script. Nothing was change except for a git pull and a rebuild of the
executable with the packages updated.

The error appears to be in this block of the input data file:

Dihedral Coeffs
# TorsionType hybrid: opls, class2, harmonic
  1 opls 1.30000 -0.05000 0.20000 0.00000 # opls
  2 opls 0.00000 0.00000 0.30000 0.00000 # opls
  3 opls 0.00000 0.00000 0.30000 0.00000 # opls
  4 class2 0.0 0.0 1.2 180.0 0.25 180.0 # class2
  5 class2 0.0 0.0 1.2 180.0 1.15 180.0 # class2
  6 harmonic 1.40000 1 3 # harmonic
  7 harmonic 1.40000 1 3 # harmonic
  8 harmonic 1.40000 1 3 # harmonic
  9 harmonic 1.40000 1 3 # harmonic
10 harmonic 1.40000 1 3 # harmonic
11 harmonic 1.40000 1 3 # harmonic
12 harmonic 1.40000 1 3 # harmonic

The error listed in the log file is:

read_data data.input
  4 = max bonds/atom
  6 = max angles/atom
  15 = max dihedrals/atom
  orthogonal box = (0 0 0) to (58.6281 58.6281 58.6281)
  4 by 2 by 4 processor grid
ERROR: Dihedral coeff for hybrid has invalid style

I will be glad to send additional data if anyone is interested.

Thanks for inputs.

valmor,

you are running the ICMS version of the code. can you please
first check if your input does run with the current official LAMMPS
version? i would first like make sure whether this is something
due to changes in the base code or in my modifications.

if the issue only happens in the ICMS branch, please send me
files that can reproduce the problem, so i can track it down.

thanks,
     axel.

Axel,

It does happen with the regular version too

LAMMPS (18 Jan 2011-ICMS)

read_data data.ndd_tbp_h2o
  4 = max bonds/atom
  6 = max angles/atom
  15 = max dihedrals/atom
  orthogonal box = (0 0 0) to (58.6281 58.6281 58.6281)
  1 by 1 by 2 processor grid
ERROR: Dihedral coeff for hybrid has invalid style

in a couple of machines I have tried. I know this script+data have
worked back in October with both ICMS and regular LAMMPS.

Looking back old e-mails I've seen one mentioning a patch that may have
had an effect on dihedrals for hybrid FF. It seems the problem is not
related to the ICMS version.

Thanks,

Attached Message (3.1 KB)

I meant to paste

LAMMPS (18 Jan 2011)

If I put these lines into bench/data.rhodo in place
of the D Coeffs already there (and change dihedral types to 12)

Dihedral Coeffs
# TorsionType hybrid: opls, class2, harmonic
1 opls 1.30000 -0.05000 0.20000 0.00000 # opls
2 opls 0.00000 0.00000 0.30000 0.00000 # opls
3 opls 0.00000 0.00000 0.30000 0.00000 # opls
4 class2 0.0 0.0 1.2 180.0 0.25 180.0 # class2
5 class2 0.0 0.0 1.2 180.0 1.15 180.0 # class2
6 harmonic 1.40000 1 3 # harmonic
7 harmonic 1.40000 1 3 # harmonic
8 harmonic 1.40000 1 3 # harmonic
9 harmonic 1.40000 1 3 # harmonic
10 harmonic 1.40000 1 3 # harmonic
11 harmonic 1.40000 1 3 # harmonic
12 harmonic 1.40000 1 3 # harmonic

and also change the in.rhodo line to
dihedral_style hybrid opls class2 harmonic

I then get no errors reading the data file until later in the
file when the dihedrals themselves use types > 12.

So what am I missing? I don't see the error you flag.

Steve