[lammps-users] Reax/c Per-atom energy and stress

Dear users:

I am wondering if anyone has developed a way to compute per atom energy and stress for reax/c. The documentation states these are currently not implemented in LAMMPS. If someone is working on this I would be willing to help out anyway that I can.

Thank you for your help,

Shawn P Coleman

University of Arkansas
Mechanical Engineering

Metin is the person to talk to. Or Aidan.

Steve

Thanks, I forgot to add why I want to use reax/c versus reax. I have tried running small jobs on reax and have been getting an error that I have exceeded MBONDDEF. MBONDDEF is a parameter listed in the original reax fortran code.

I have successfully compiled the original reax code and linked it to the 2Feb11 version when I compiled LAMMPS. However, I have been unable to start from scratch, edit reax_defs.h in the reax library, compile the reax code, and link reax to LAMMPS when compiling.

Shawn P Coleman

University of Arkansas
Mechanical Engineering

What is the most convenient method for calculating a standard deviation on the quantity which is spatially averaged (ke/atom) in each bin? I am interested in the standard deviation which has accumulated by computing the spatial average over the prescribed number of timesteps. In my case:

Nevery = 1
Nrepeat = 19000
Nfreq = 20000
[ave = 1]

So every 20000 time steps I’ll get a new average and a new standard deviation for each bin.

I hope I didn’t skip over this in the literature or archived message threads,

Thanks,
Mike

This is something that we would like to have, but there is no current effort
under way. Metin? One way to proceeed is to generate restart files, and then
compute per-atom energy on those using pair reax, as a post-processing step.

From: Steve Plimpton <[email protected]>
Date: Thu, 24 Feb 2011 08:58:25 -0700
To: Shawn Coleman <[email protected]>, Aidan Thompson <[email protected]>,
Hasan Metin Aktulga <[email protected]>
Cc: "[email protected]" <[email protected]>
Subject: Re: [lammps-users] Reax/c Per-atom energy and stress

Metin is the person to talk to. Or Aidan.

Steve

Dear users:
I am wondering if anyone has developed a way to compute per atom energy and
stress for reax/c. The documentation states these are currently not
implemented in LAMMPS. If someone is working on this I would be willing to
help out anyway that I can.
Thank you for your help,

Shawn P Coleman
University of Arkansas
Mechanical Engineering

----------------------------------------------------------------------------->>

Aidan,

This sounds like a reasonable get-around for other simulations but Shawn has indicated that he was not able to use reax with his system due to the MBONDDEF restriction. So it would not be applicable in his case, I think.

Thanks,
Metin

Yes, it can be difficult to size all the arrays used by pair_style reax to
accomodate a large system. But if you can do it, it is better than nothing.

There is no way to do this currently with a LAMMPS option - you would
have to post-process the file output by fix ave/spatial.

Steve