Atomistic simulations
for the non-expert ctd:
- These calculations take quite a long time and require
a reasonable amount of memory. Typically,
on a Pentium with at least a giga memory one gets a few real
picoseconds in 24 hours for 2000 atom samples with classical potentials.
- We prefer to write our own
c or fortran 95 code for LINUX nowadays, but canned codes like LAMMPS also good.
- Supercomputers are great, commodity clusters also work. Any CPU
thats around helps if it has enough memory. TAMNUN is PERFECT for these types of calculations.
- The codes can be parallelised with different parts of the sample on different machines if there are good communications. Trivial parallelization, running different realizations on different processors is also good.
- Sometimes we switch off the dynamics and use the same
potentials for static simulations, e.g. with Metropolis Monte Carlo
to find a minimum energy state or a groundstate. Also can use simulated
tempering etc.