[lammps-users] correlation function per atom

LAMMPS LAMMPS Mailing List Mirror
1

Hello I have problem in lammps with fix ave/correlate because is
correlation function is for global variables, i need correlation function
per atom
atom as incoherent/self-intermediate scattering function and velocity
auto-correlation function in order of 0.1 microseconds because dynamics of
supercooled liquids is very slow, i use FFT by calculation correlation
function but dump file by post-processing is very large 1T and read the
dump file is most time of calculation the correlation function, algorithm
use to calculation the correlation function on fly and have some ideas to
modify the code.

I hope your help
thank you very much
Bye
Samuel

2

The problem with doing a per-atom correlation is that
you need to keep a huge amount of data around to do
the correlation. That is feasible for global variables, b/c
the data is smaller. But for per-atom it would mean many,many
copies of atom coordinates or other per-atom quantities.
Keeping LAMMPS from blowing out memory would be problematic.

So I think post-processing is your best bet. But maybe others
have alternate ideas.

Steve