hyperdynamics : error with fix hyper/local

LAMMPS LAMMPS Mailing List Mirror
1

Dear all,

I’m trying to simulate atomic diffusion using hyperdynamics method. I use the atom_style full. I have run simulations with both fix hyper\global and fix hyper\local. For both fixes, I get the following warning message :

WARNING: Hyper local for molecular systems requires care in defining hyperdynamic bonds (…/fix_hyper_local.cpp:245)

I use the same input parameters (cutbond, qfactor, Vmax, Tequil) for both fixes, and Dcut alpha Btarget as recommended at https://lammps.sandia.gov/doc/fix_hyper_local.html.

When I try to run my simulation with fix hyper/global, despite the warning message above, everything seems to work correctly. However, when I use fix hyper/local I always get the following error :

job aborted:
rank: node: exit code[: error message]
0: LAPTOP-FICK1ETR: 123
1: LAPTOP-FICK1ETR: 123
2: LAPTOP-FICK1ETR: -1073741819: process 2 exited without calling finalize.

Can anyone assist me with this issue?

Best regards

P.S. Here is my input file :

2

The warning is b/c you should not allow atom pairs with
actual bonds between them to be treated via the hyperdynamics
biasing. I can’t tell if you are excluding those or not.

I have never seen the error message before. It is not
a LAMMPS error.

Can you run your model on a single processor?

Steve

3

Dear Steve,

Thank you for considering my question. There is only one bond type in my system, and atoms pairs with bonds between them are very far from the atoms group to which the bias is applied.

When run my model on a single processor the simulation stops without error message.

Best regards

4

Can you run it with valgrind and see if there is some internal
error causing a crash?

Steve

5

Dear Steve,

I’ve checked my simulation and I noticed that the error occurs when it comes to applying a hyperdynamic bias to a bonded atom. I would like to know if it is possible to prevent atom pairs with actual bonds between them to be treated via the hyperdynamics without turning off pairwise interactions between these bonded atoms and other atoms with which they form a hyperdynamic bond?

Best regards