We detected a possible problem using the AIREBO potential as implemented in the LAMMPS version of December 7.
We simulated the collision of two small clusters of diamond (see the input below). Three atoms at the collision interface (with ids 639, 727 and 1275) suddenly acquire a very high potential energy, starting at step 7.500, resulting in a rapid and increasing departure from the desired NVE simulation. The atoms’ coordination increases to 5, 6 and 7, before step 10.000. The simulation was made with a variable time step with values which should be fine for the chosen initial cluster velocities, with a large margin to spare.
We will be thankful for any comments and suggestions to help us with this problem.
Sincerely,
Dalía Bertoldi
There have been several reported "problems" with AIREBO
in LAMMPS. To date, none of them have resulted in finding
a bug that we can fix. This includes efforts by Steve Stuart's
group to track down small energy/force discrepancies vs their
original AIREBO code. The problem you are describing sounds
dramatic. Hence my guess is it is more due to the
model you are trying to create, than to AIREBO itself. When
you slam things into each other, bad results are always possible.
Maybe AIREBO is not parameterized for such high energy
collisions.
You are welcome to correspond with one of the people
in S Stuart's group and see what he thinks.
(Marcell is CCd on this message.)
Maybe you
have a test case that will illuminate some issue for them.
Or maybe they will tell you that you can't use AIREBO
in this manner.
Steve
2011/1/1 Dalía Bertoldi <[email protected]...>: