Variable atom-style cannot express chunkID of each atom defined by compute chunk/atom

LAMMPS LAMMPS General Discussion
1

Hi all,

Command line ‘fix 1 all rigid/small custom v_’ requires an atom-style variable, and I hope the variable can represent the chunkID of each atom.

I don’t know why ‘variable chunkID atom c_ chunk’ cannot assign the chunkID of each atom but the total number of chunks to the variable.

However, in ‘fix ave/chunk’, c_chunk can represent the chunkID of each atom.

2

What does

compute chunk all chunk/atom [... settings ...]
variable chunkID atom c_chunk

do in your script? I would expect it to return a per-atom chunk ID variable as expected.

3

fix 1 groupID rigid custom v_chunkID

4

To set each chunk as a rigid body:

fix 1 groupID rigid custom v_chunkID

but failed.

5

Failed how?

Also, when reporting unexpected behavior, please always report which LAMMPS version you are using and what platform you are running on.

6

Here is a brief test in.file about variable chunkID atom c_Nchunk and the test results of the lammps-2022Jun23 compiled by Intel on a single processor on MobaXterm.

Obviously, the variable only assigns the total number of chunks to each atom instead of chunkID.
c_Nchunk is considered as a global scalar(total number of chunks) instead of per-atom vector(chunkID).

in.chunktest (869 Bytes)
CH.airebo (905.0 KB)
Si.sw (795 Bytes)
9-540Ncell_Sicore_9.5.data (356.0 KB)

chunktest.log (5.9 KB)
preNchunk.txt (552 Bytes)
System.data (356.0 KB)

7

Your analysis is incorrect. The problem is that you are connecting the atom style variable to the compute, but the compute is “non-current” in between runs and thus cannot be accessed and will produce inconsistent or random data.
You need a “cache” that makes the data properly permanent. Fix store/state can be used for that.

compute cluster Ncell cluster/atom 2.5
compute Nchunk Ncell chunk/atom c_cluster compress yes nchunk once ids once
fix storechunk all store/state 0 c_Nchunk
variable chunkID atom f_storechunk
run 0 post no

fix R Ncell rigid/nve custom v_chunkID
run 2 post no
write_dump all custom dump.chunk id v_chunkID
8

the same limitation doesn’t seem to apply to compute cluster/atom though:

compute cluster Ncell cluster/atom 2.5
compute chunk Ncell chunk/atom c_cluster #nchunk once ids once
fix storechunk all store/state 0 c_chunk
variable chunkID1 atom c_chunk
variable chunkID2 atom c_cluster
variable chunkID3 atom f_storechunk
compute cmin Ncell reduce min v_chunkID1 v_chunkID2 v_chunkID3
compute cmax Ncell reduce max v_chunkID1 v_chunkID2 v_chunkID3
thermo_style custom c_cmin[1] c_cmax[1] c_cmin[2] c_cmax[2] c_cmin[3] c_cmax[3]
# outputs 5657, 5657, 1, 5657, 1, 5657 for c_chunk, c_cluster, f_storechunk in order
run 0
9

It does. You are creating thermo output which is during a run, if you would add “c_cluster” to the write_dump command in my example, you will get:

ERROR: Compute used in dump between runs is not current (src/dump_custom.cpp:567)

And if you use the compute in a variable and add the variable to the write_dump line you get:

ERROR: Variable clusterID: Compute used in variable between runs is not current (src/variable.cpp:1571)

Using fix store/state can give you access to both with the example input changed as follows:

compute cluster Ncell cluster/atom 2.5
compute Nchunk Ncell chunk/atom c_cluster compress yes nchunk once ids once
fix storechunk all store/state 0 c_Nchunk c_cluster
variable chunkID atom f_storechunk[1]
variable clusterID atom f_storechunk[2]
run 0 post no

fix R Ncell rigid/nve custom v_chunkID
run 2 post no
write_dump all custom dump.chunk id v_chunkID f_storechunk[1] f_storechunk[2]
10

You mean that compute Nchunk only performs calculation at the 0 timestep and cannot be invoked at other timesteps, right? But even in timestep 0, the value of v_chunkID is also incorrect in preNchunk.txt.