Multiple Time Steps

LAMMPS LAMMPS Mailing List Mirror
1

Dear LAMMPS user

Last time I asked a question about Multiple time-step* but I did not received answer because people just tried to know why I'm using some specific potential.

I am using two different potential (one short range - one long range) to investigate interface phenomena in BCC lattice in presence of a flat wall. So without Multiple Time Steps it takes long time!

I'm not sure if it's possible in LAMMPS (this is the answer I have never received!) or I should implement that for myself.
And If it is the second case, can any one give me some clues?!

Cheers
Mojtaba

* Understanding Molecular Simulation, D. Frenkel, chapter 15.3 P 425

2

I don’t recall details of the earlier post. If by potentials

you mean you are using 2 pair styles and using pair hybrid,

then pair hybrid uses run_style respa the same as other

pair styles. Which means if the pair styles define

inner/middle/outer options, then rRESPA can invoke

them at different time scales. The individual pair

styles document whether they support this, only a few

do, e.g. pair lj/cut.

Note that if your “long-range” potential still has

a short-range part (most do), then there may be

little advantage to using rRESPA with it.

Steve

3

Dear Steve

Thanks for the answer.
The thing is I’m using a pair potential (Yukawa) and a fix wall (assume a extremely short-range). So I guess the rRESPA is not the case.

cheers
Moji

4

Dear LAMMPS user

Last time I asked a question about Multiple time-step* but I did not
received answer because people just tried to know why I'm using some
specific potential.

people were asking this, because you refused to provide a convincing
argument why what you insisted on doing was a good idea and you flat
out dismissed the suggested alternative. please note that this mailing
list is populated by volunteers and that you are in no way entitled to
get an answer if nobody cares about it and you don't make an effort to
care about the answers you get. that is just a fact of life.

I am using two different potential (one short range - one long range) to
investigate interface phenomena in BCC lattice in presence of a flat
wall. So without Multiple Time Steps it takes long time!

I'm not sure if it's possible in LAMMPS (this is the answer I have never
received!) or I should implement that for myself.

you already received an answer on that as the *very* first answer: you
need to do some programming and write a r-RESPA integrator for your
specific needs.

And If it is the second case, can any one give me some clues?!

read the source. there is no alternative.

axel.

5

Dear Axel

Thanks for your answer.
I know people do it voluntarily and I'm grateful for this.
As I said before this is part of my PhD project and I'm trying to investigate the solid-wall interfacial free energy.
the idea is using a short range wall like 1/(z^256).

during checking the source as you suggest, I have faced with this part of fix wall codes

void FixWall::setup(int vflag)
{
   if (strstr(update->integrate_style,"verlet")) {
     if (!fldflag) post_force(vflag);
   } else {
     ((Respa *) update->integrate)->copy_flevel_f(nlevels_respa-1);
     post_force_respa(vflag,nlevels_respa-1,0);
     ((Respa *) update->integrate)->copy_f_flevel(nlevels_respa-1);
   }
}

But I thought is not possible to use fix wall in rRespa.
I did not get it or I'm totally wrong with this part of code?!

cheers
Mojtaba

6

Dear Axel

Thanks for your answer.
I know people do it voluntarily and I'm grateful for this.
As I said before this is part of my PhD project and I'm trying to
investigate the solid-wall interfacial free energy.
the idea is using a short range wall like 1/(z^256).

during checking the source as you suggest, I have faced with this part of
fix wall codes

void FixWall::setup(int vflag)
{
  if (strstr(update->integrate_style,"verlet")) {
    if (!fldflag) post_force(vflag);
  } else {
    ((Respa *) update->integrate)->copy_flevel_f(nlevels_respa-1);
    post_force_respa(vflag,nlevels_respa-1,0);
    ((Respa *) update->integrate)->copy_f_flevel(nlevels_respa-1);
  }
}

But I thought is not possible to use fix wall in rRespa.

nobody said that. you have to keep reading, as this is only part of the code.

I did not get it or I'm totally wrong with this part of code?!

you didn't say what you understood this code is doing. how should i be
able to tell whether you misread it or not.
steve had already explained in the previous exchange how fix wall fits
into the r-RESPA scheme and i have nothing to add to this and to what
aidan and i had stated previously.

if you want some more constructive help, you have to act more
constructive as well.

axel.