Deposited atoms don't fall reaxff

Hello, I am trying to use reaxff in the deposition model(lammps-23Jun2022), but those deposited atoms didn’t fall. The previous similar model used lj force field can work normally with falling atoms. But after I changed to reaxff and added
fix 1 all qeq/reaxff 1 0.0 10.0 1.0e-6 reaxff ,

those deposited atoms can’t drop.
And also the there are warnings like

WARNING: Van der Waals parameters for element MO indicate shielding without inner wall, but earlier elements indicate a different van der Waals method. This may cause division-by-zero errors. Keeping van der Waals setting for earlier atoms. (src/REAXFF/reaxff_ffield.cpp:279)
WARNING: Van der Waals parameters for element NI indicate shielding without inner wall, but earlier elements indicate a different van der Waals method. This may cause division-by-zero errors. Keeping van der Waals setting for earlier atoms. (src/REAXFF/reaxff_ffield.cpp:279)
WARNING: Van der Waals parameters for element B indicate shielding without inner wall, but earlier elements indicate a different van der Waals method. This may cause division-by-zero errors. Keeping van der Waals setting for earlier atoms. (src/REAXFF/reaxff_ffield.cpp:279)
WARNING: Van der Waals parameters for element F indicate shielding without inner wall, but earlier elements indicate a different van der Waals method. This may cause division-by-zero errors. Keeping van der Waals setting for earlier atoms. (src/REAXFF/reaxff_ffield.cpp:279)
WARNING: Van der Waals parameters for element P indicate shielding without inner wall, but earlier elements indicate a different van der Waals method. This may cause division-by-zero errors. Keeping van der Waals setting for earlier atoms. (src/REAXFF/reaxff_ffield.cpp:279)
WARNING: Changed valency_val to valency_boc for F (src/REAXFF/reaxff_ffield.cpp:296)

Will these warnings from force field file relate with the suspending deposited atoms?
Or could you help me find anything wrong with the below code? Thanks a lot.
The following file is the reaxff file and log file.
jz0c03720_si_002.txt (31.4 KB)
log.lammps (4.5 KB)

units real
atom_style charge

boundary p p f

lattice bcc 3.51

read_data Li111-25x24x80-rg.lmp
#mass 1 6.94 # Li
pair_style reaxff NULL checkqeq no
pair_coeff * * ffield.reax Li

neighbor 2 bin
neigh_modify every 1 delay 0 check yes

min_style cg
minimize 1e-8 1e-8 10000 10000

region slab block 0 25 0 24 70 75 units box

group addatoms region slab

fix 1 all qeq/reaxff 1 0.0 10.0 1.0e-6 reaxff

fix 2 all nvt temp 313.0 313.0 $(100.0*dt)

fix 3 addatoms deposit 100 1 100 12345 region slab near 1.0 vz -8.0 -8.0

fix 4 addatoms wall/reflect zhi EDGE

thermo_style custom step atoms temp epair etotal press
thermo 100

dump 1 all atom 50 dump.deposit.Li

run 10000

What do you mean by “falling atoms”? Fall where and how? Within classical MD atoms are just points and points cannot fall.

This means that whoever created the potential parameter file was doing a careless job and included parameters from a previous parameterization that are not compatible with the current one and are using a different variant of the ReaxFF potential function. So to get correct results you have to be careful which elements you use.

No. They only impact the correctness of the potential parameters.

This is missing the data file and also the input using the LJ potential that was working as expected. Only with full disclosure of your simulation inputs somebody can reproduce them and try to understand what kind of differences you are talking about and whether you adapted the input that behaves as expected into an input with unexpected behavior.

1 Like

Li111-25x24x80-rg.lmp (51.5 KB)

Sorry that I forgot to upload this frame data file.

And here I don’t understand “the input using the LJ potential”, do you mean I set LJ potential in my input file? But I set reaxff pair_style here.
pair_style reaxff NULL checkqeq no
pair_coeff * * ffield.reax Li

And “falling atoms” I mean by using “fix deposit”, atoms can fall from a layer to the substrate as the below picture. I expect this as the normal case, which can be seen in a simpler model with LJ force field. I attach its input file here. Thank you.
iamtrying1_Li_deposition.lmp (1.2 KB)
Li001-21x21x80.lmp (31.7 KB)
dump.deposit.Li (2.9 MB)

There is no “falling” because there is no gravity, and thus no up and down. What you mean is “motion toward the surface”.

There is a very significant difference between your input using the LJ potential and the input using the ReaxFF potential: the first is using “metal” units and the second uses “real” units.
Those two settings differ in the time unit (picoseconds versus femtoseconds), thus your velocity units also differ by a factor of 1000.0 and thus you would need to run for 1000 times as many timesteps to see your deposited atoms move the same distance, or use a 1000 times larger numbers for the velocity values.

Apropos velocity. Using a thermostat on group all for this kind of system makes no sense at all. Only the lower part of your surface slab should be time integrated with a thermostat. The rest (and especially the deposited atoms) should be time integrated with fix nve since they are not within a bulk that is coupled to a large number of other atoms (which is what a thermostat algorithm emulates).

Yet your fix deposit command uses values of the same magnitude, thus in the ReaxFF input, deposited atoms are given about a 1000 times slower velocity than in the LJ input.

1 Like

Yes, thanks a lot :rose:, after changing the deposit velocity to 1000 smaller, those deposited atoms can move again.
And I wanna ask if I continue to use the reaxff force field with warnings, but choose the atoms that don’t generate warnings, in this case like Li doesn’t generate a warning, but Mo, Ni, B, F, P, F have warnigs.
When I use Li, can my system run normally?

That is difficult to say from remote. You have to check what kind of systems and compounds the publication describing the parameter set mentions it was generated for. Then you should make some tests trying to reproduce the reference data from that publication.

That should be common procedure for any calculation using downloaded parameter sets.

1 Like