question regarding reaxff

Dear lammps dev.

The force field file in reax (e.g. rdx), is it transferrable? If I have a molecule or crystal which is consists of C, N, O, H system can it be used?

regards
min

It is supposed to be trasferrable, but whether it
works well is really a ReaxFF question. I would
search the literature for your system.

Steve

Dear lammps dev.

The force field file in reax (e.g. rdx), is it transferrable? If I

this is not a question that can be answered with a simple "yes" or "no".

in general, it is always up to you to validate it applicability
before you use any kind of simulation technique. the same applies,
for example, to the quantum chemistry methods and basis set sizes,
one which a large part of the reaxff parametrization is based.

have a molecule or crystal which is consists of C, N, O, H system can
it be used?

as far as i know, there are several generations of reaxff
parameters and they are validated for different types of
applications, compounds, and simulations. for example,
there are some that are only useful for high-temperature
gas-phase systems with small molecules.

i think it is unrealistic to expect an empirical model
to be applicable and transferable to any kind of system
in the most general way, however, anybody that develops
models and parameters, has a desire to make them as
transferable as possible, without losing accuracy for
specific use cases. so it is a bit of a balancing act,
and it is very difficult to predict whether that balance
is good enough for your specific use case.

axel.

Dear lammps dev,

I attach the log file of my simulation on my crystal after performing MD using reaxff. I equilibrate the system by using temp/rescale. The total energy of the system seems very stable and is very minimal, does that means the force field applicable in my system?

P/S: Sorry I am still beginner in MD

regards
Min

log.lammps (5.83 KB)

Dear lammps dev,
I attach the log file of my simulation on my crystal after performing MD
using reaxff. I equilibrate the system by using temp/rescale. The total

i strongly discourage using temp/rescale unless
you know exactly what you are doing. it is about
the worst option to thermalize a system, since
it doesn't dissipate excess energy. you are much
better off using fix langevin with a rather short
time constant, that can be made longer in 1-2 steps
and then switched to fix nvt for a correct isothermal
ensemble with nose-hoover chains.

energy of the system seems very stable and is very minimal, does that means
the force field applicable in my system?

not at all. there are multiple options that you need
to test. running a stable and energy conserving MD
is simply a requirement of the method overall and
doesn't tell you much about the applicability of the
force field parameters.

what you need to do is to validate that your molecules
are the correct average structure as individual molecules
and as crystal or liquid or solution, whatever your system
is. you can compute properties like atom-atom radial
distribution functions g(r), self-diffusion, average density
(when running in npt ensemble). for inhomogeneous
systems you can also compute property profiles across
the direction of the inhomogeniety. you want to compare
this to available experimental data and published (or
self-generated) simulation results from a comparable
non-reactive force field.

P/S: Sorry I am still beginner in MD

not a problem, for as long as you don't mistake
the mailing list as a replacement for a text book
(which you haven't so far). for as long as you
research and study the basics of the methods and details,
that you are interested in, with a couple of MD +
stat mech text books and the occasional original
literature article on a method, you will find people
very responsive to questions discussing issues
beyond what is elementary knowledge.

in many cases, these kind of discussions can
turn out to be educational even for more
experienced people. learning how to ask the
right questions is also an important skill.

cheers,
    axel.