dihedral distribution no symmetrical

LAMMPS LAMMPS Mailing List Mirror
#1

Hello Lammps users:

I used the following commands to dump information related to dihedral torsions in a molecular system:

compute diheinfo all property/local dtype datom1 datom2 datom3 datom4
compute dihevalu all dihedral/local phi
dump 1 all local 20000 dump.*.dihe index c_diheinfo[1] c_diheinfo[2] c_diheinfo[3] c_diheinfo[4] c_diheinfo[5] c_dihevalu

Then I extracted a specific type of torsions and tried to explore the torsions angle distribution. However it is not symmetric with respect to 0 degree. Please see the attachment for the plot.

Do you have any idea why this is the case? Thanks.

Shaorui

bpator_plot.PNG

#2

Not all LAMMPS dihedral styles are symmetric. These are not:

charmm
class2
table
helix
cosine/shift/exp

It also depends on the dihedral parameters (coeffs).

Please post the "dihedral_style" and "dihedral_coeff" commands (from
your lammps input script), and the "Dihedral Coeffs" section from your
data file.
Cheers

Andrew

P.S. I think "dihedral_style class2" still has a bug. Hopefully Steve
added the patch by now. If not, I'll email it. However this bug
would not cause the problem you are having.

#3

Hello Andrew:

Thanks for replying. I am using class2 style for dihedral. The formulation for dihedral, as you may know, is

E = SUM(n=1,3) { V(n) * [ 1 - cos(n*Phi - Phi0(n)) ] }

Here is the parameters for the specific type of dihedral I am looking at:

[K(1), phi0(1), K(2), phi0(2), K(3), phi0(3)] = [-0.2802 0.0000 -0.0678 0.0000 -0.0122 0.0000]

I see a bit in-consistency between the formulation used by LAMMPS and that found in the pcff force field file. In pcff.frc, it is
E = SUM(n=1,3) { V(n) * [ 1 + cos(n*Phi - Phi0(n)) ] }

This means if pcff.frc gives Phi0(n) as 0 degree, then the parameters used in my datafile should be 180 degree. However, in the datafile I obtained by converting Accelrys .mdf and .car files using msi2lmp tool, Phi0(n) are all 0 degrees. Is this in-correct?

Thanks.

#4

follow up to the previous mail.

I guess it is indeed minus. I checked with Sun’s 1995 paper. Then the expression in tools/msi2lmp/biosym_frc_files/pcff.frc should be wrong.

Shaorui

#5

Hello Andrew:

Thanks for replying. I am using class2 style for dihedral. The formulation
for dihedral, as you may know, is

E = SUM(n=1,3) { V(n) * [ 1 - cos(n*Phi - Phi0(n)) ] }

Here is the parameters for the specific type of dihedral I am looking at:

[K(1), phi0(1), K(2), phi0(2), K(3), phi0(3)] = [-0.2802 0.0000
-0.0678 0.0000 -0.0122 0.0000]

You are correct. If all of the phi0 parameters are zero, then the
angular distribution should be symmetric. First verify that "msi2lmp"
is generating the correct LAMMPS parameters. Can you post the
dihedral_coeff commands that msi2lmp created? (...and/or the "Dihedral
Coeffs" section from your DATA file?)

Next: it might be easier to find your problem if you have a small
system. For your convenience I am posting a small 4-atom test system
I used to test "dihedral_style class2"
Edit the "in.test" file and replace the following line:
dihedral_coeff 1 10.0 -60 0.0 -60 0.0 -60
with:
dihedral_coeff 1 -0.2802 0.0 -0.0678 0.0 -0.0122 0.0
(These are the parameters you said you are using.)

After these modifications, does the molecule behave the way you
expect? I'm reasonably confident that this small molecule should have
a symmetric distribution of dihedral angles. Please try it and
report back.

I see a bit in-consistency between the formulation used by LAMMPS and that
found in the pcff force field file. In pcff.frc, it is
E = SUM(n=1,3) { V(n) * [ 1 + cos(n*Phi - Phi0(n)) ] }

This means if pcff.frc gives Phi0(n) as 0 degree, then the parameters used
in my datafile should be 180 degree. However, in the datafile I obtained by
converting Accelrys .mdf and .car files using msi2lmp tool, Phi0(n) are all
0 degrees. Is this in-correct?

Alas, I am not familiar with the FRC file format.
In LAMMPS 0 degrees corresponds to the "cis" conformation. I do not
know whether 0 degrees in an FRC file corresponds to the "cis" or
"trans" conformation.
(terminology:
http://en.wikipedia.org/wiki/Cis-trans_isomerism
http://lammps.sandia.gov/doc/dihedral_style.html.)

Andrew
(P.S. If you know of a web site containing a description of the FRC
file format, please post it.)

test_system_class2.tar.gz (1.47 KB)