force cutoff in LAMMPS

Hi,

Could anyone tell me how the force cutoff is set for a system without any pair interaction (i.e. no pair_style cmds) but only harmonic bonds (bond_style harmonic). Under this circumstance, there isn’t any cutoff explicitly set neither in pair or bond.
I know such system is rare but for some theoretical model it is possible.

Also this is critical for me as I got such error when I run simulation under NVE
ERROR: Cannot use neighbor bins - box size << cutoff (…/neighbor.cpp:1619)

which obviously due to some very big force cutoff implicitly set or not initialized?

Best

Ming

Hi,

Could anyone tell me how the force cutoff is set for a system without any
pair interaction (i.e. no pair_style cmds) but only harmonic bonds
(bond_style harmonic). Under this circumstance, there isn't any cutoff
explicitly set neither in pair or bond.
I know such system is rare but for some theoretical model it is possible.

​usually to make this kind of system work, you need to set the
communication cutoff manually.​
​please note that the commands required to do this have changed in the 13
May 2014 patch.​

Also this is critical for me as I got such error when I run simulation
under NVE
ERROR: Cannot use neighbor bins - box size << cutoff (../neighbor.cpp:1619)
which obviously due to some very big force cutoff implicitly set or not
initialized?

​if the above doesn't work, you can always set the bin size manually or add
an otherwise unused atom type to your system and ​set a cutoff for it

​axel.​

Hi Axel,

Thanks for your advice. Just after I sent the 1st email, I made it work now but through some ugly way. In terms of parallel efficiency, it is very low and for this kind of system I’m sure it could be highly tuned.

Another 3 questions about this is

  1. In terms of setting the bin size manually, do you mean using atom_modify?

  2. In terms of add an otherwise unused atom type to your system and ​set a cutoff for it, do you mean use atom_style hybrid and then set the binsize in atom_modify?

  3. How should I choose the binsize (say using atom_modify), via try-and-error or is there anything like an equation?

Best
Ming

Hi Axel,

Thanks for your advice. Just after I sent the 1st email, I made it work
now but through some ugly way. In terms of parallel efficiency, it is very
low and for this kind of system I'm sure it could be highly tuned.

Another 3 questions about this is
1. In terms of *setting the bin size manually*, do you mean using
atom_modify?

​no.

2. In terms of *add an otherwise unused atom type to your system and ​set
a cutoff for it*, do you mean use atom_style hybrid and then set the
binsize in atom_modify?

​no.

3. How should I choose the binsize (say using atom_modify), via
try-and-error or is there anything like an equation?

​default is (max. cutoff)/2 ​

Ok, now I’m confused. I have the following few questions, and I keep in mind that in my system (no pairwise and only harmonic bond), there is such pairwise force cutoff (i.e. it is 0.0) and I don’t want to set it as it will increase computing cost greatly which is not necessary.
1, Which bin size do you mean? Is it not the binsize in atom_modify? Also I need to set it at the beginning mannualy or LAMMPS will report
ERROR on proc 0: Bond atom missing in image check (…/domain.cpp:625). If it not the binsize in atom_modify, than which one you refer to?

  1. I know that there are different cutoffs used in LAMMPS, e.g. ghost cutoff, neighbor cutoff (= pairwise force cutoff + neighbor skin), pairwise force cutoff. And apparently the pairwise force cutoff is more fundamental as the other two can be generated based on this using default settings. How should I set these cutoffs? Keep pair force cutoff =0.0 and set the other two manually?

Best

Ming

......
Ok, now I'm confused.

​you confuse yourself, because you don't pay attention to *all* of what is
said, but just jump on the first item you recognize and also don't properly
research the documentation before coming to conclusion. ​

I have the following few questions, and I keep in mind that in my system
(no pairwise and only harmonic bond), there is such pairwise force cutoff
(i.e. it is 0.0) and I don't want to set it as it will increase computing
cost greatly which is not necessary.

​not, if you do it right. but i don't even think it is necessary.

1, Which bin size do you mean? Is it not the binsize in atom_modify? Also
I need to set it at the beginning mannualy or LAMMPS will report
ERROR on proc 0: Bond atom missing in image check (../domain.cpp:625). If
it not the binsize in atom_modify, than which one you refer to?

​this error​ has *nothing* to do with binsize, but the communication
cutoff. this error has been reported many times to the mailing list and the
proper advice has been given (almost) as often.

2. I know that there are different cutoffs used in LAMMPS, e.g. ghost

cutoff, neighbor cutoff (= pairwise force cutoff + neighbor skin), pairwise
force cutoff. And apparently the pairwise force cutoff is more fundamental
as the other two can be generated based on this using default settings. How
should I set these cutoffs? Keep pair force cutoff =0.0 and set the other
two manually?

​you are not making sense. your conclusions are not correct. again, this is
something that has been explained many times, and i am in no mood to
explain it again to somebody that doesn't pay attention.​

Hi Axel,

Thanks for your reply. I notice that I just copied the wrong error message, the right one with its condition is

For my system, I need to set the binsize using atom_modify at the beginning mannualy or LAMMPS will report
ERROR: Atom sorting has bin size = 0.0 (…/atom.cpp:1622).

, does it mean that I should set it manualy? also I just notice that there is another binsize which I can set using neigh_modify, I’ll try it soon.

Best

Ming

Hi Axel,

Thanks for your reply. I notice that I just copied the wrong error
message, the right one with its condition is

For my system, I need to set the binsize using atom_modify at the
beginning mannualy or LAMMPS will report
ERROR: Atom sorting has bin size = 0.0 (../atom.cpp:1622).
, does it mean that I should set it manualy?

​without non-bonded interactions, there is not much of a point in sorting.
just turn it off. ​

also I just notice that there is another binsize which I can set using
neigh_modify, I'll try it soon.

​...again *again* you are ignoring my advice. why do you ask, if you do
not follow the answer?

Hi Axel,

I’m really sorry that if you feel I’m ignoring you advice. I always respect your suggestions/advice and find them really useful during the past few years.

For this particular question, I’ll read all your advice carefully again and try them.

B.t.w., I’m quite excited about the forthcoming tiled style as it will probably increase the computing efficiency for inhomogeneous systems significantly, like one of the systems I’m studying.

Thanks again for your suggestions :slight_smile:

Best
Ming

Hi Axel,

Just to tell follow your suggestion, I solved the problem using neigh_modify binsize.

Thanks for your help again :slight_smile:

Best
Ming

See the comm_modify cutoff command, when you have

no pair style.

Steve