Hello to whom it may concern,
I am using the pair_style lj/cut/coul/long to treat the separation of some mixture solution of alcohol and water. I used to use the pppm, but I found it did not work well in my inhomogeneous system. It blows up when the molecule hits the rigid membrane framework. I guess maybe it makes wrong simulations for the interface because atoms on the framework have charges and when I turn off the coul/long, it could run. Generally, the pppm treats those long range forces as they were within the cutoff. So it simply thinks that all around it is the strong interface force, which blows up the velocity and energy. So my question is how to choose the right kspace that can calculate inhomogeneous system well. Or another deal is to simple use the coul/cut with a larger cutoff. Can someone with such experience gives me some help?
Best regards,
Andy
Hello to whom it may concern,
I am using the pair_style lj/cut/coul/long to treat the separation of some
mixture solution of alcohol and water. I used to use the pppm, but I found
it did not work well in my inhomogeneous system. It blows up when the
molecule hits the rigid membrane framework.
what blows up how?
I guess maybe it makes wrong simulations for the interface because atoms
on the framework have charges and when I turn off the coul/long, it could
run.
guessing and speculation is pointless. since you do simulations on a
computer, you can study the generated data, you can monitor all properties
of all parts of the system, you can draw meaningful conclusions. it is
impossible to comment on such a vague system description. there are
countless reasons, why a simulation will diverge or is simply bogus.
turning off forces due to charges is comparing apples an oranges and the
resulting simulation is likely completely bogus.
Generally, the pppm treats those long range forces as they were within the
cutoff.
huh?
So it simply thinks that all around it is the strong interface force,
which blows up the velocity and energy.
algorithms don't think. your system will behave according to the physics
of the model you created.
So my question is how to choose the right kspace that can calculate
inhomogeneous system well. Or another deal is to simple use the coul/cut
with a larger cutoff. Can someone with such experience gives me some help?
not with so little information about your actual system. couldn't it just
be that your force field parameters are bad? or your timestep too large? or
your geometry unphysical? or your input script buggy? or ???
axel.
Believe me. I have tried everything I can do. Creating new initial data file, shrinking the tilmestep to 0.5 ftsec and change fix by nve or nvt. None of them works. It used to work well, because the previous model have a relatively lower charges on the framework. I monitored the energy, velocity and temp in the thermout. They blow up but not my guess. And I used the vmd to monitor to visualize the very first hundreds step before blowing up. And the steps just before collapsing show that the water molecules get teared up and move in a relatively high speed right after they interact with the framework. Now, I used the ewald/disp to treat the long range force. It is working well right now. That could only be the pppm.
Believe me. I have tried everything I can do. Creating new initial data
file, shrinking the tilmestep to 0.5 ftsec and change fix by nve or nvt.
None of them works. It used to work well, because the previous model have a
relatively lower charges on the framework. I monitored the energy, velocity
and temp in the thermout. They blow up but not my guess. And I used the vmd
to monitor to visualize the very first hundreds step before blowing up. And
the steps just before collapsing show that the water molecules get teared
up and move in a relatively high speed right after they interact with the
framework. Now, I used the ewald/disp to treat the long range force. It is
working well right now. That could only be the pppm.
this is as unscientific and thus mostly useless a description as the
previous one.
so what is it that you are claiming:
- the PPPM *algorithm* is at fault?
- the PPPM *implementation* in LAMMPS is broken?
- something else?
what you describe about your visualization suggests that you have bad force
field or input parameters. possibly both.
at any rate, i doubt that anybody can give you any serious advice based on
vague descriptions and unverifiable speculation.
axel.