Hi all,
I just encountered a question like this:
For two atoms, one is at timestep 0 with atom id of 1, the other is at timestep 1000 with atom id of 1.
Are these two atoms the same one?
Besides, during an energy minimization or a normal run or a neb calculation, dose the id of atom change?
If it changes, how to keep the id of atoms consistent during a simulation?
2011/1/18 Qingjie <[email protected]...>:
Hi all,
I just encountered a question like this:
For two atoms, one is at timestep 0 with atom id of 1, the other is at
timestep 1000 with atom id of 1.
Are these two atoms the same one?
yes
Besides, during an energy minimization or a normal run or a neb calculation,
dose the id of atom change?
no
If it changes, how to keep the id of atoms consistent during a simulation?
why should it change? do you see any indication of that?
axel.
Thanks, Axel.
To be honest, I haven’t seen any indications of the change.
It’s only my guess when I read the command “atom_modify” which involves “atom sorting or reordering”. (it’s my misunderstanding)
在 2011年1月19日 上午10:27，Axel Kohlmeyer <[email protected]>写道：
2011/1/18 Qingjie <[email protected]...>:
Thanks, Axel.
To be honest, I haven't seen any indications of the change.
It's only my guess when I read the command "atom_modify" which involves
"atom sorting or reordering". (it's my misunderstanding)
the sorting is with respect to the internal atom indexing
in the individual MPI tasks i.e. the domains of the domain
decomposition. the atom ids (internally called tags) are
reordered in sync.
the purpose is to have atoms that are close in space
also "close in memory" resulting in better use of the
CPU cache. atom ids/tags are a global atom property
just like as mass or type and thus have stay the same.
axel.