crosslinked polymer with low-density problem calculated by dreiding

Dear Lammps users,

without going into detail, there are three major issues with your message:
- you don't specify which LAMMPS version you are using, which is
crucial information
- your coulomb handling is not likely to produce a converged stress
tensor due to your lax kspace convergence of 1.0e-3. also, for optimal
performance, you should use pppm instead of ewald. i suggest a
conservative choice of the convergence parameter at 1.0e-6
- you use a timestep of 1.0fs which is quite large for a system
containing light hydrogen atoms and no means to constrain bonds
containing them.

also, why do you need to rescale velocities all the time?

Dear Lammps users,

I recently simulate crosslinked epoxy resin and need to analyse its bulk
property based on Dreiding forcefiled. However I encounter a difficulty at
the first step that the density calculated is about 10% lower than the
target value. I believe the relaxation process should be fine. Somebody in

target value in comparison to what? other simulations? experiment?

the previous discussion state there may be problems of setting parameters
for the improper terms. Since the contribution of improper energy in my
simulation is not significant compared to other energy component, and the
improper energy I calculated using lammps is consistent with MS results
(actually all the energy components I calculated are consistent with the MS
restuls),I dont think it is the main reason of leading to low density.

i don't think so either.

Another issue is about the Van Der Wals parameters. Mayo's paper gives the
van der Waals bond length ,R0, and I think it should corresponds to the
equilibrium position in the plot of van der Waals energy. In Lammps, the van
der Waals distance does not corresponds to the equilibrium position, but
corresponds to the cross-section point (between the energy cure and x axis)
where the van der Waals energy is zero, so the van der Waals length in
Lammps should be shorter than RO in Mayo's paper, which should be equal to
R0/(2^(1/6)). The value of 2^(1/6) should be around 1.1225. Therefore I
divide all R0 by 1.1225 and then input all the parameters in the pair_coeff
command in the lammps input file.

if you want to confirm that you have been converting parameters
correctly, you need to reproduce data that has been provided as
reference with the parameter set.

One thing is, in another data file written by my friend, he used van der
Waals length even shorter than I used. The length he use is about
R0/(2^(1/3)). I believe that if using his parameters the density surely will
be higher, but I can not see any reason why he choose R0/(2^(1/3)).

neither do i. but since you have not done anything to validate your
translation of parameters to LAMMPS, nobody can tell whether you have
done it correctly or not. rather than worrying about your system of
interest, you should worry about reproducing data that is more easily
reproduced and readily available to check. furthermore, your input is
running at a temperature of 600K. have you checked whether dreiding is
even meant to be transferable to that temperature? or what kind of
error with respect to density you need to expect. since this is not a
new force field, there should be a large host of publications out
there using it, so you can review how the authors of those
publications have validated their choice of parameters and what kind
of errors, especially for elevated temperatures they are seeing and to
the kind of material you are looking at.

Below is my input file (please notice the atom types and help to check the
pair_coeff), anyone could see any problem and teach me how to fix the
low-density issue ?

actually, this is more an issue of learning how to validate a
classical MD simulation and parameter set in general. that is
something that you need to discuss with your adviser and other local
experience MD users. i've given you some hints and pointed you in a
few directions that you can investigate, but this is not a forum meant
to teach people how to do MD. this is mean to discuss installing,
extending, and using LAMMPS.

axel.

Thanks Axel, for relaxation I use a Windows version. I think it is a version of 2017, will recheck the details and inform you soon.

I will try pppm instead of ewalds as you suggestted. Actually I also tried to calculate without any charges since in Mayo's work it recommend the charges can be either calculated using a Gageister method or just be ignored. However I still got a low density. So I'm not sure if the charges influence much.

For the timestep, do you think 0.5fs is a proper choice? Or could you suggest any lammps command that can constrain bonds including hydrogen ?

I used a script downloaded online and it rescales velocities all the time. Seems it will cause stress discontinuities, right? I will make a change of it.

Hao

Thanks Axel, for relaxation I use a Windows version. I think it is a version of 2017, will recheck the details and inform you soon.

I will try pppm instead of ewalds as you suggestted. Actually I also tried to calculate without any charges since in Mayo's work it recommend the charges can be either calculated using a Gageister method or just be ignored. However I still got a low density. So I'm not sure if the charges influence much.

For the timestep, do you think 0.5fs is a proper choice? Or could you suggest any lammps command that can constrain bonds including hydrogen ?

I used a script downloaded online and it rescales velocities all the time. Seems it will cause stress discontinuities, right? I will make a change of it.

this all sounds a lot like you do need significant advising and
tutoring. you are skipping over important questions that i am posing
and you are ignoring my *very important* recommendation to validate
your parameters and instead are changing things without knowing why
that you put there in the first place without knowing/understanding
why they are there. that is bad. in a computer simulation, *all*
choices need to be justified. if you don't know how, you need learn
it. do not just do something because somebody mentions it. do things
for a reason.

put this all together, there is a *lot* of homework that you have to
do. do not treat this as a case of "i just need the right magical
change and everything will work fine". things in simulations are never
that easy. most problems are much more subtle and don't manifest
themselves as easily.

axel.