How to fix bond length without bond coefficients?

Dear lammps-users:
I want to use TraPPE force field in my simulation. TraPPE force field uses fixed bond length but doesn’t fix angle.
The fix shake command in LAMMPS can be used to fix bond length but it needs specific bond style and bond coefficient including Kb. However, TraPPE does not give Kb.
So how to fix bond length without Kb but keep angle not rigid at the same time in LAMMPS? Can I set the bond style and bond coefficient at random?
Any suggestion could help, thanks.

Yinan Chen

Department of Energy and Resources Engineering;

College of Engineering;

Peking University;Beijing;P.R.China 100871

Cell:156-5271-0581

Dear lammps-users:
        I want to use TraPPE force field in my simulation. TraPPE force
field uses fixed bond length but doesn't fix angle.
        The fix shake command in LAMMPS can be used to fix bond length but
it needs specific bond style and bond coefficient including Kb. However,
TraPPE does not give Kb.

​if you constrain a bond with fix shake, only the bond length parameter is
used, so you can give any number as Kb. however, there is another problem
with TraPPE in LAMMPS. TraPPE was derived for MonteCarlo simulations, where
keeping bond lengths constant is easy to do and parallelization is no a
concern. in LAMMPS, however, you can use fix shake only on isolated "shake
clusters", not on "chains". so you will get an error and LAMMPS will stop
for anything but a system, where multiple constrained bonds are connected
to the same atom. since TraPPE is a united atom force, field, tert. butane
or isoropane are the largest molecules you can do, regular propane is not
possible.

        So how to fix bond length without Kb but keep angle not rigid at
the same time in LAMMPS? Can I set the bond style and bond coefficient at
random?

​you may want to look up and use some other force field, e.g. OPLS or
OPLS-AA, which is designed with classical MD​ in mind.

axel.

But it is not expected for an apparently robust code as LAMMPS.
Why do developers not work on new algorithms? Any user desires to utilize a specific force field and the software should not dictate the kind of FF.

Am I right?

You must be joking? LAMMPS has far more features than any other widely used MD package, and it is free.

Amber costs 500 USD (or 20,000 USD for commercial licenses!) and also does not support constraining bonds over large molecules, and it supports far fewer force fields than LAMMPS does (without intending any disrespect towards Amber)

But it is not expected for an apparently robust code as LAMMPS.
Why do developers not work on new algorithms? Any user desires to utilize
a specific force field and the software should not dictate the kind of FF.

Am I right?

​you are wrong!

LAMMPS is largely a community based effort, so if *you* want to see
something happen, it is *your* job to come up with a good idea to implement
it and do the work to implement and contribute it. many of the files in the
LAMMPS distribution were not written by the LAMMPS developers themselves.
have a look.
the LAMMPS developers work on what *they* themselves consider relevant. we
are open for suggestions, but since we do this mostly as a hobby or as
contributions to specific project, where there are specific goals that need
to be implemented, those take precedence.

besides, there is a *very* good reason why fix shake (and rattle) have this
restriction: it collides with with efficient parallelism​. applying
shake/rattle to connected chains of atoms is a "hard" problem to do
efficiently in parallel. MD codes that implement this, have thus limited
parallel efficiency or have to make some approximations or simply cheat. in
monte carlo, this is not a concern, since monte carlo moves can be
prescribed any which way the programmer wants. at the same time, monte
carlo is inherently incompatible/inefficient with domain decomposition and
long-range electrostatics.

also, one can turn your statement around and say: why should somebody
expect that a force field designed with monte carlo simulations in mind
should work (well) on a molecular dynamics code? this is like expecting to
drive a car on a bicycle track or a bicycle on a highway.

finally, you *still* get an infinite value for what *you* have invested in
LAMMPS so far, since you didn't pay or contribute anything. if you don't
like it, don't use it. nobody forces you. and if you want it changed, you
can pay back the LAMMPS community by implementing what you think it should
be able to do yourself.

i do not speak for other LAMMPS developers, only for myself here: my take
is, that i don't take anybody serious that says "LAMMPS should do this! or
LAMMPS must have this feature! or LAMMPS is bad because it cannot do that!"
unless that person has actually actively and significantly contributed to
the LAMMPS code or community. put up or shut up! in short, before you
publicly criticize LAMMPS or the LAMMPS developers, you have to earn the
right to do so.

axel.

I think Gromacs does have an algorithm for doing SHAKE across

large molecules - maybe it thus implements the Trappe FF.

No one needs to take offense here. It’s fine to ask

if/why LAMMPS does not have a feature. And when

it doesn’t, we always welcome users to work on

it and contribute code for everyone to use. That’s

a key way LAMMPS gains new features. It’s not

all from the core developers (not by a long shot).

See the Authors web page …

Steve