Fix/wall and fix NVT with different time steps

Dear all

I would like to simulate a box which includes N particles. The interaction between them is based on “Yukawa potential” and also I have the fix wall with “1/®” potential. for this simulation I would like to use different time steps: dt1 = 0.5 for “Yukawa” dt2 = 0.0001 for the “1/®” But I could not find out how can I do that. SO would be grateful if one can help me with that. Best regards Moji

Dear all

I would like to simulate a box which includes N particles. The interaction between them is based on “Yukawa potential” and also I have the fix wall with “1/(r^256)” potential.
for this simulation I would like to use different time steps:

dt1 = 0.5 for “Yukawa”
dt2 = 0.0001 for the “1/(r^256)”

But I could not find out how can I do that.

This is not possible. You need to derive and develop a new multi-timestep integrator for this.

“Respa” is not helpful???

Cheers
Moji

"Respa" is not helpful???

​why not just use a reflecting wall?​

Ok,

I guess I should make my case more clear.
I have a relaxed “BCC” with “Npart=2N_xN_y*N_z” particles which has been relaxed in a elongated box with “NVT" and with “PBC” in all direction.
The interaction with this particles is Yukawa potential with time step of “DT=0.5”.
Now I have to case of study:
1- Take first and last layer of the relaxed “BCC” in “Z-direction” and change the type of them and give them interaction with inner particle with “U=1/(r^256)” and switch off the “PBC” in “Z-direction”.
2- Make fix wall behind the box (in “Z-direction") with “U=1/(r^256)” and also Switch off the “PBC”.
With this two I need different “DT”.

This is the case.
I thought maybe “repsa” run style is helpful but I did not understand it.

cheers
Moji

Ok,

I guess I should make my case more clear.
I have a relaxed “BCC” with “Npart=2*N_x*N_y*N_z” particles which has been
relaxed in a elongated box with “NVT" and with “PBC” in all direction.
The interaction with this particles is Yukawa potential with time step of
“DT=0.5”.
Now I have to case of study:
1- Take first and last layer of the relaxed “BCC” in “Z-direction” and
change the type of them and give them interaction with inner particle with
“U=1/(r^256)” and switch off the “PBC” in “Z-direction”.
2- Make fix wall behind the box (in “Z-direction") with “U=1/(r^256)” and
also Switch off the “PBC”.
With this two I need different “DT”.

This is the case.

​so where does this come from? a publication? has something like this been
run before? with LAMMPS or a different code? where does the code for the
r^256 potential come from? ​

I thought maybe “repsa” run style is helpful but I did not understand it.

​nobody can help you unless you explain, *what* it is exactly about
run_style respa, that you don't understand.

in any case, *before* even trying to use it, you should set up your system
to just run with the smaller time step for *everything* and then add the
changes you list one by one. it is pointless to discuss optimizations
(after all that is what respa is doing​) you need to get the basic physics
of your model right.

​so, do you have the relaxed bcc structure already?​
​have you been able to run it with two particle types and PBC?

axel.​

This is what we did in “Fortran” but we have tried to do it in LAMMPS.
There is some paper about this hard wall.

This is what we did in "Fortran" but we have tried to do it in LAMMPS.

​"fortran"?? that is a programming language not a simulation software.
tried what in LAMMPS? and how?

your e-mails remain irritatingly vague.

There is some paper about this hard wall.

​does it say why it is better than a reflecting wall?​ i can only see
disadvantages.

And yes I relaxed the BCC with dt=0.5.
I checked it with Yukawa wall and that was ok so next step is multiple
time step.

​no. next step is doing the entire system with short time step. then
convert and the "hard" particles.
this has to work *first* and properly conserve energy and so on.​
multi-time step *always* is the last thing to do.
don't you have an adviser? this entire exchange, except for the very first
inquiry, is one that you should have with your adviser.

as far as i am concerned, you have now exhausted my quota of "goodwill"
responses. don't expect an answer from me until you actually have a real
LAMMPS problem, i.e. can demonstrate that LAMMPS doesn't work as its
documentation says, or when it crashes (with a sufficient input deck to
reproduce that) or something similar.

axel.