[gmx-users] vsite as interaction site for COM of benzene

Thomas Schlesier schlesi at uni-mainz.de
Sat Jun 9 01:18:16 CEST 2012

Dear Richard,

thanks for the idea, i will try this.

Relating to your side note:
I perform pulling simulations of a calixarene-catenane dimer (both together 600 atoms)
and i mesitylene as a solvent. To access longer time-scales I wanted to coarse-grain the
solvent only. My supervisor said that the mesitylene rotates on such a short time-scale,
that it would be ok, so coarse-grain it as one particle. But i also thought 3 particles
would be an option.
But since it is easier to coarse-grain the mesitylene as one particle, i started with that,
at this stage, my first interest to get a somewhat reasonable simulation, which runs, since
it is my first project involving coarse-graining and vsites.
If the results are not good enough, i still can make the system more complex :) but it is easier
to start with the simple case.
I'm now at the to construct the AA-CG interaction. For this is started with an atomistic mesitylene
in coarse-grained mesitylene (strictly speaking it's still an CG-CG interaction, but the single mesitylene
has an atomistic representation, as later the calixarene).

For the question here i used benzene instead of mesitylene, since it is smaller but the concept is the same.

Dear Thomas,

Whilst in principle a constraint should do what you are asking, you might be
better off using multiple vsites to solve the problem so that there is an
algebraic as apposed to numerical mapping of the forces. My immediate
thought is to use two type 3 (or if necessary type 3fd) virtual sites to
define two points near the centre of the benzene. You could then place your
tabulated potential on a type 2 virtual site between these two. I believe
that should algebraically link all the atomic sites together making the
process more stable and it should be quite easy to transfer it to other

On a separate note why if you are coarse graining the benzenes to a single
point particle are you then maintaining all the atoms? Could you go a stage
further and map the atoms onto 3 non interacting point masses? That would
maintain angular momentum etc. whilst reducing the computational overhead.
Although it becomes very complex to implement if your benzene's are in the
backbone of a polymer.

Hope that helps,


On 08/06/2012 18:18, "Thomas Schlesier"<schlesi at uni-mainz.de  <http://lists.gromacs.org/mailman/listinfo/gmx-users>>  wrote:

>/  Hi all,
/>/  i have a more conceptional question, for using vsites as
/>/  interaction-centers for coarse-grained particles:
/>/  First the simple case:
/>/  I want to simulate one benzene molecule (atomistic - aa) in
/>/  coarse-grained (cg-) benzene (each benzene molecule as a single
/>/  particle). For the cg-cg interaction of the benzene i have calculated a
/>/  table potential.
/>/  But for the aa-cg interaction I'm a little bit clueless:
/>/  I want to use the COM of the aa-benzene as an interaction-site. But
/>/  there is no vsite, which i could construct from 6 atoms.
/>/  Side-note: The vsite would not interact with the with the atoms, only
/>/  with cg-benzenes
/>/  So there are two ideas:
/>/  1) Using one '3out'-vsite, which is construted from 3 non-neighbour
/>/  atoms (if we had mesitylene instead of benzene, it would be the three
/>/  C-atoms to which only hydrogens are bound). The the force from the aa-cg
/>/  interaction would be distributed to these three atoms and i would hope
/>/  that the bond-constraints would do 'the rest' (i.e. they would mimick
/>/  that the whole aa-benzene molecule interacts with the cg-particles)
/>/  2) Using two '3out'-vsites. First viste the same as in (1), the second
/>/  comes from the set of the other three C-atoms. These two vsites, should
/>/  be at the same position (more or less, for a perfect energy-minimized
/>/  benzene molecule it would be so; but i would assume the error is only
/>/  minimal). Since the potential is pair-additive, i would use for this
/>/  table potential only half of the potential, so that the potential of
/>/  both particles would add up to the 'true' potential. The forces would be
/>/  calculated from the 'half-potential'.
/>/  Big questions is now:
/>/  Are (1) and (2) equivalent?
/>/  (2) Seems like the 'right way' to do.
/>/  But (1) would be easier to implement :) But the big question is, if the
/>/  bond-constraints would correct the approximation, that only 3 out of 6
/>/  atoms have a direct interaction with the cg-particles.
/>/  Side note:
/>/  The real system is a little more complex. There are fracments of
/>/  molecules where i could not really construct two vsites which would have
/>/  more or less the same position. In this case, method (1) would really help.
/>/  Any ideas?
/>/  Greetings
/>/  Thomas

More information about the gromacs.org_gmx-users mailing list