Colvars Angles

LAMMPS LAMMPS Mailing List Mirror
1

Hi All,
I am new to using the colvars package/fix.

I would like to apply two constraints to a pair of atoms.

First, I would like to apply a harmonic potential between the two atoms, based on their distance from one another. This seems to work fine.

I would also like to apply a constraint to keep the two particles parallel to a single axis.
The relevant block for defining this angle in the .colvars file is below:

colvar {
name angle

orientationangle {
atoms {
atomnumbers 1 9022
}
refpositions (0, 0, 0) (10, 0, 0)

}
}

This block produces output that makes sense most of the time, but intermittently has values that are wrong. The output from the .traj file is below:

step one fa_one angle E_h_pot x0_one W_h_pot

0 2.63443126732797e+01 4.96706198016085e+00 7.10991472947676e+00 4.11195078579324e+00 2.80000000000000e+01 -2.15239352473637e+00
1 2.63445235947703e+01 4.96642921568898e+00 7.10980348107483e+00 4.11090319240818e+00 2.80000000000000e+01 -4.30451285153493e+00
2 2.63447172415882e+01 4.96584827523553e+00 7.10970573171377e+00 4.10994151544328e+00 2.80000000000000e+01 -6.45638043747032e+00
3 2.63448937414113e+01 4.96531877576597e+00 7.10962199287174e+00 4.10906509082901e+00 2.80000000000000e+01 -8.60801857363558e+00
4 2.63450534150514e+01 4.96483975484581e+00 7.10955270220001e+00 4.10827229854957e+00 2.80000000000000e+01 -1.07594491340688e+01
5 2.63451967319910e+01 4.96440980402706e+00 7.10949824538948e+00 4.10756078371999e+00 2.80000000000000e+01 -1.29106933824805e+01
6 2.63453243392690e+01 4.96402698219292e+00 1.79979388626225e+02 4.10692731332322e+00 2.80000000000000e+01 -1.50617717414308e+01
7 2.63454369997417e+01 4.96368900077495e+00 7.10943484838600e+00 4.10636808273571e+00 2.80000000000000e+01 -1.72127036417666e+01
8 2.63455356106389e+01 4.96339316808321e+00 7.10942624765969e+00 4.10587862349585e+00 2.80000000000000e+01 -1.93635073479360e+01
9 2.63456211470618e+01 4.96313655881467e+00 7.10943305980519e+00 4.10545408357378e+00 2.80000000000000e+01 -2.15141998567556e+01
10 2.63456946650961e+01 4.96291600471174e+00 7.10984839588891e+00 4.10508921163732e+00 2.80000000000000e+01 -2.36647967921307e+01
11 2.63457572996944e+01 4.96272810091666e+00 1.79304379628549e+02 4.10477836727131e+00 2.80000000000000e+01 -2.58153123025279e+01
12 2.63458102315154e+01 4.96256930545368e+00 7.10954396489181e+00 4.10451568523850e+00 2.80000000000000e+01 -2.79657590015579e+01

The value of 7.1 for the angle is correct, based on the positions of the particles, but at random times the value jumps to ~ 180 as it does at steps # 6 and 11. This produces warning messages:
colvars: Warning: one molecular orientation has changed by more than 0.01: discontinuous rotation ?

It seems this could be because I’ve done something incorrectly, but then the correct values confuse me. Is this a bug or have I done something wrong?
The problem seems consistent between LAMMPS versions 9 Dec 2014 and 14 Jun 2014.

Thanks,
Mike

2

mike,

do you mind updating your sources using the colvars code from colvars.github.io?
there has been a bugfix commit in relative to the very latest released
version just yesterday that seems related.

if you have a compatible OS, you may save the effort to compile by
downloading a precompiled (serial) binary rpm from rpm.lammps.org
(just updated them) and then extracting the executable with rpm2cpio
lammps-latest.x86_64.rpm | cpio -ivd

if the issue persists, i suggest you better report the issue on the
github issue tracker and attach example inputs to reproduce it.

axel.

3

Hey Axel, luckily the fix you mentioned it is not related.

The problem is that with only two atoms it is difficult to calculate a superposition matrix, which is the basis of orientationAngle. Use at least 4-5 atoms whose structure doesn’t change significantly, and it should work.

Giacomo

4

Thank you both for your rapid responses. This is consistent with what I see, since the colvars update doesn’t fix the problem. Is there a simple way to do what I want with two atoms that I am overlooking?

Thanks,
Mike

5

Thank you both for your rapid responses. This is consistent with what I
see, since the colvars update doesn't fix the problem. Is there a simple
way to do what I want with two atoms that I am overlooking?

if you don't need any biasing force information, you could try using
fix aveforce on the in-plane force components for the two atoms (in
addition to the harmonic restraint orthogonal to it) to achieve the
effect you desire.

axel.