create_atom, mol keyword

What is the status for using fix deposit with reax/c? I’ve used fix deposit with reax without problems, but trying the same simulation with reax/c very quickly leads to nan pressure and segmentation fault.

Sincerely,

Eirik Hjertenæs

Ph.D. student

Department of Chemistry

Faculty of Natural Science and Technology (NT)

Norwegian University of Science and Technology (NTNU)



I believe someone else is already doing this (depositing ReaxFF molecules).
I think it should work.

Steve



> > no, because i can not compile new version of lammps until now
> unfortunately
>
> you can always use one of the precompiled binaries, if needed you can
> run it in a virtual machine.
>
> > and i use lammps-8Jul13 version, but if it is not possible to use it with
> > reax/c pair_style i dont want to compile it because of unknown erros and
> the
> > lack of time now.
>
> create_atoms should work, but fix deposit is not likely to work, since
> reax/c is not a "native" LAMMPS code, but an interface to a standalone
> code and which does its own memory management.
>
> axel.
>
> > Best Regard.
> >
> >
> > ________________________________
> > From: Axel Kohlmeyer <[email protected]...>
> > To: Negar Amiri <[email protected]...>
> > Cc: LAMMPS Users Mailing List <[email protected]...>
> > Sent: Tuesday, 28 January 2014, 14:58:09
> > Subject: Re: [lammps-users] create_atom, mol keyword
> >
> >
> >>
> >> Dear LAMMPS users
> >> is it possible to use create_atoms, mol keyword  or fix deposit with
> >> reax/c pair_style?
> > Have you tried?
> > Axel
> >
> >> Best Regard.
> >
> >>
> >>
> >>
> ------------------------------------------------------------------------------
> >> WatchGuard Dimension instantly turns raw network data into actionable
> >> security intelligence. It gives you real-time visual feedback on key
> >> security issues and trends.  Skip the complicated setup - simply import
> >> a virtual appliance and go from zero to informed in seconds.
> >>
> >>
> 
> >> _______________________________________________
> >> lammps-users mailing list
> >> [email protected]...
> >> [https://lists.sourceforge.net/lists/listinfo/lammps-users](https://lists.sourceforge.net/lists/listinfo/lammps-users)
> >
> >>
> >
> >
>
>
>
> --
> Dr. Axel Kohlmeyer  [email protected]...  [http://goo.gl/1wk0](http://goo.gl/1wk0)
> College of Science & Technology, Temple University, Philadelphia PA, USA
> International Centre for Theoretical Physics, Trieste. Italy.
>
>
> ------------------------------------------------------------------------------
> WatchGuard Dimension instantly turns raw network data into actionable
> security intelligence. It gives you real-time visual feedback on key
> security issues and trends.  Skip the complicated setup - simply import
> a virtual appliance and go from zero to informed in seconds.
>
> 
> _______________________________________________
> lammps-users mailing list
> [email protected]...
> [https://lists.sourceforge.net/lists/listinfo/lammps-users](https://lists.sourceforge.net/lists/listinfo/lammps-users)

Under most circumstances it should work fine. An example that could lead to crash would be depositing with small number of atoms per MPI process. You could reduce the MPIs used, enlarge your system, or use a larger safezone and mincap keywords.

Ray

Ray - why does the outcome of the deposit depend on

atoms/MPI process, i.e. on the number of processors

being used?

Steve

Not the outcome but the robustness of the reax/c code. When the change in atoms per MPI process changes dramatically due to fix deposit, reax/c code is prone to break. A larger memory buffer would help most of the time.

Ray

That sounds like a code issue, not a physics issue

with the deposit. Since fix deposit is only adding a

few atoms on a single timestep, is there a way to

have the pair style check if the atom count on that

proc has grown and realloc buffers? How is that

any different than what happens on any reneighbor

step (which is also when fix deposit adds atoms),

when a proc acquires a few new atoms from its

neighbor procs?

Steve