Errors in using EDIP/C portable model with LAMMPS

OpenKIM
1

Dear all,

I want to perform simulations with the EDIP/C potential in LAMMPS at remote super-computer center. It is unable to connect network for the super-computer center so I have to install software using source code. I first installed the open-kim API and links to the LAMMPS at the super-computer. Then, I installed the EDIP model driver and portable model in my own computer using: kim-api-collections-management install user EDIP_LAMMPS_MD_783584031339_000 (model driver installation), and kim-api-collections-management install user EDIP_LAMMPS_Marks_2000_C_MO_374144505645_000 (portable model installation). Then, I copy the mode driver and portable model to the super-computer (directory: lib64/kim-api/model-drivers and lib64/kim-api/portable-models)

Then, I test two simulation script using LAMMPS. The first one is using the in.kim-ex.melt from example/kim directory of LAMMPS (using LenneardJones_Ar portabal model). The log file indicates a successful simulation, which means the installation of LAMMPS and open-kim API is ok.

Then, I try to run my own lammps script with EDIP portable model. Here is my script for potential part:
kim init EDIP_LAMMPS_Marks_2000_C__MO_374144505645_000 metal
boundary p p p
atom_style full
atom_modify map hash
read_data ${file_name}
kim interactions C

However, I get error message: Unable to load KIM Simulator Model. It is strange because EDIP_LAMMPS_Marks_2000_C__MO_374144505645_000 is a portable model rather than a simulator model according to the file description from openKIM.

Thank you very much for your help!

2

Hi Ke,

Then, I installed the EDIP model driver and portable model in my own computer using: kim-api-collections-management install user EDIP_LAMMPS_MD_783584031339_000 (model driver installation), and kim-api-collections-management install user EDIP_LAMMPS_Marks_2000_C_MO_374144505645_000 (portable model installation). Then, I copy the mode driver and portable model to the super-computer (directory: lib64/kim-api/model-drivers and lib64/kim-api/portable-models)

You should compile and install the model driver and the portable model on the cluster. To do that, first please remove the copied files, then you can download both (EDIP model driver & EDIP/C portable model) on your laptop and copy them to the cluster (e.g., using scp). You should have KIM-API installed on your cluster, and there please follow the below steps (the commands given below are for the bash shell),

- tar Jxvf EDIP_LAMMPS__MD_783584031339_000.txz
- tar Jxvf EDIP_LAMMPS_Marks_2000_C__MO_374144505645_000.txz
- kim-api-collections-management install user EDIP_LAMMPS__MD_783584031339_000
- kim-api-collections-management install user EDIP_LAMMPS_Marks_2000_C__MO_374144505645_000

I think this should resolve the issue.

3

Hi yafshar,

I followed the commands, but when I run the third and fourth command, the following error occurred! It seems that this comand try to download source file from openkim.org rather than using the source file.
*** ERROR *** Unable to successfully contact openkim.org to check EDIP_LAMMPS_MD_783584031339_000 comp atibility.
‘wget’ returned error code 4.
Aborting!

4

Hi,
Try this instead (note the added ./)

1 Like
5

Hi, relliott,

Still the same errors!

6

Hello,
I don’t understand why you’re getting the error. I tried this and both the commands yafshar and I gave work for me.

Make sure you are running the 3rd & 4th commands with the working directory set to the location where you extracted the tarballs. Please try again and see if it works. If not, please post some output from your shell history.

7

Dear relliott,

Thank you for your reply. I have figure out the install problem by using kim-api-collections-management install system EDIP… The reason is that I have no authority to access the usr/bin in the cluster.

During my simulation, I find other problems. My lammps script, lammps log file, and cluster output information are as follows. According to the lammps log file, the kim potential involves 19 mutable parameters without abl potential parametes described in EDIP/C mode (zbl cutoff skin shift). How to use the EDIP/C mode? Another problem is the simulation terminates without run even one step. I tried in.kim-ex.melt from example/kim of lammps, and it runs ok.

Lammps script used:

variable file_name string start.data
variable tstep equal 0.001 # 0.5 fs timestep
variable start_temp equal 300
variable anneal_temp equal 1500
variable time_stage1 equal 10
variable time_stage2 equal 200
variable time_stage3 equal 1000 # anneal at this stage
variable time_stage4 equal 200 # cool the model down to start temp at this stage

variable dump_gap equal round(20/v_tstep) # dump one lammpstrj every dump_gap
variable fix_data_gap equal round(0.05/v_tstep) # output data every fix_data_gap

variable ptime equal stepv_tstep
variable Tdamp equal 100v_tstep
variable Pdamp equal 1000*v_tstep
variable runs_stage1 equal round(v_time_stage1/v_tstep)
variable runs_stage2 equal round(v_time_stage2/v_tstep)
variable runs_stage3 equal round(v_time_stage3/v_tstep)
variable runs_stage4 equal round(v_time_stage4/v_tstep)
#######################common settings############################
kim init EDIP_LAMMPS_Marks_2000_C__MO_374144505645_000 metal
boundary p p p
atom_style full
atom_modify map hash

read_data ${file_name}

mass * 12.0

kim interactions C

group diamond_shpere type 1

neighbor 2.0 bin
neigh_modify every 10 one 10000

###################define initial vel##############################
velocity all create ${start_temp} 4928459 mom yes rot yes dist gaussian

###################settings for thermo_style and dump#############
variable etotal equal etotal
compute 1 all displace/atom
compute 2 all stress/atom NULL

thermo_style custom step temp etotal press density lx ly lz
thermo 1000

shell mkdir results_info
shell cd results_info

fix info_print all print {fix_data_gap} "{ptime} ${etotal}" file energy.txt screen no

dump 1 all custom ${dump_gap} *.lammpstrj id type mol x y z c_1[4] c_2[4]

################################Constrain mmomentum of graphene###################
fix control2 diamond_shpere momentum 100 linear 1 1 1 angular

###################################Dynamic run#####################################
fix 1 all nvt temp {start_temp} {start_temp} {Tdamp} # x 0 0 {Pdamp} y 0 0 {Pdamp} z 0 0 {Pdamp}

#timestep {tstep} run {runs_stage1}

The lammps log file

#######################common settings############################
kim init EDIP_LAMMPS_Marks_2000_C__MO_374144505645_000 metal
#=== BEGIN kim init ==========================================
units metal
neighbor 2.0 bin # Angstroms
timestep 1.0e-3 # picoseconds

This model has 19 mutable parameters.
No. | Parameter name | data type | extent

1 | eps | “Double” | 1
2 | BB | “Double” | 1
3 | beta | “Double” | 1
4 | sigma | “Double” | 1
5 | a | “Double” | 1
6 | aprime | “Double” | 1
7 | Z0 | “Double” | 1
8 | lambda0 | “Double” | 1
9 | lambdaprime | “Double” | 1
10 | gamma | “Double” | 1
11 | q | “Double” | 1
12 | flow | “Double” | 1
13 | fhigh | “Double” | 1
14 | alpha | “Double” | 1
15 | plow | “Double” | 1
16 | phigh | “Double” | 1
17 | Zdih | “Double” | 1
18 | Zrep | “Double” | 1
19 | c0 | “Double” | 1
#=== END kim init ============================================

boundary p p p
atom_style full
atom_modify map hash

read_data ${file_name}
read_data start.data
Reading data file …
orthogonal box = (-39.940409 -39.035132 -39.142455) to (39.092289 39.178505 39.128494)
6 by 4 by 4 MPI processor grid
reading atoms …
11121 atoms
Finding 1-2 1-3 1-4 neighbors …
special bond factors lj: 0 0 0
special bond factors coul: 0 0 0
0 = max # of 1-2 neighbors
0 = max # of 1-3 neighbors
0 = max # of 1-4 neighbors
1 = max # of special neighbors
special bonds CPU = 0.007 seconds
read_data CPU = 0.031 seconds

mass * 12.0

kim interactions C
#=== BEGIN kim interactions ==================================
pair_style kim EDIP_LAMMPS_Marks_2000_C__MO_374144505645_000
pair_coeff * * C
#=== END kim interactions ====================================

group diamond_shpere type 1
11121 atoms in group diamond_shpere

####################define the potential###########################
#pair_style edip
#pair_coeff * * C.edip C

#pair_style airebo 3.0 1 1 #
#pair_coeff * * CH.airebo C

neighbor 2.0 bin
neigh_modify every 10 one 10000
#neigh_modify exclude group graphene graphene check no

######################Energy minimization##########################
#min_style cg
#minimize 1e-20 1e-20 1000 100000
#write_data minimized.data

###################define initial vel##############################
velocity all create ${start_temp} 4928459 mom yes rot yes dist gaussian
velocity all create 300 4928459 mom yes rot yes dist gaussian

###################settings for thermo_style and dump#############
variable etotal equal etotal
compute 1 all displace/atom
compute 2 all stress/atom NULL

thermo_style custom step temp etotal press density lx ly lz
thermo 1000

shell mkdir results_info
WARNING: Shell command ‘mkdir’ failed with error ‘File exists’ (…/input.cpp:1214)
shell cd results_info

fix info_print all print {fix_data_gap} "{ptime} {etotal}" file energy.txt screen no fix info_print all print 50 "{ptime} ${etotal}" file energy.txt screen no

dump 1 all custom ${dump_gap} *.lammpstrj id type mol x y z c_1[4] c_2[4]
dump 1 all custom 20000 *.lammpstrj id type mol x y z c_1[4] c_2[4]

################################Constrain mmomentum of graphene###################
fix control2 diamond_shpere momentum 100 linear 1 1 1 angular

###################################Dynamic run#####################################
fix 1 all nvt temp {start_temp} {start_temp} {Tdamp} # x 0 0 {Pdamp} y 0 0 {Pdamp} z 0 0 {Pdamp}
fix 1 all nvt temp 300 {start_temp} {Tdamp}
fix 1 all nvt temp 300 300 ${Tdamp}
fix 1 all nvt temp 300 300 0.1

#timestep {tstep} run {runs_stage1}
run 10000

CITE-CITE-CITE-CITE-CITE-CITE-CITE-CITE-CITE-CITE-CITE-CITE-CITE

Your simulation uses code contributions which should be cited:

@Article{tadmor:elliott:2011,
author = {E. B. Tadmor and R. S. Elliott and J. P. Sethna and R. E. Miller and C. A. Becker},
title = {The potential of atomistic simulations and the {K}nowledgebase of {I}nteratomic {M}odels},
journal = {{JOM}},
year = 2011,
volume = 63,
number = 17,
pages = {17},
doi = {10.1007/s11837-011-0102-6}

Cluster output file:
LAMMPS (30 Jul 2021)
OMP_NUM_THREADS environment is not set. Defaulting to 1 thread. (…/comm.cpp:98)
using 1 OpenMP thread(s) per MPI task
Reading data file …
orthogonal box = (-39.940409 -39.035132 -39.142455) to (39.092289 39.178505 39.128494)
6 by 4 by 4 MPI processor grid
reading atoms …
11121 atoms
Finding 1-2 1-3 1-4 neighbors …
special bond factors lj: 0 0 0
special bond factors coul: 0 0 0
0 = max # of 1-2 neighbors
0 = max # of 1-3 neighbors
0 = max # of 1-4 neighbors
1 = max # of special neighbors
special bonds CPU = 0.007 seconds
read_data CPU = 0.031 seconds
11121 atoms in group diamond_shpere
WARNING: Shell command ‘mkdir’ failed with error ‘File exists’ (…/input.cpp:1214)

CITE-CITE-CITE-CITE-CITE-CITE-CITE-CITE-CITE-CITE-CITE-CITE-CITE

Your simulation uses code contributions which should be cited:

CITE-CITE-CITE-CITE-CITE-CITE-CITE-CITE-CITE-CITE-CITE-CITE-CITE

Neighbor list info …
update every 10 steps, delay 10 steps, check yes
max neighbors/atom: 10000, page size: 100000
master list distance cutoff = 8.8
ghost atom cutoff = 8.8
binsize = 4.4, bins = 18 18 18
1 neighbor lists, perpetual/occasional/extra = 1 0 0
(1) pair kim, perpetual
attributes: full, newton off, ghost, cut 5.4
pair build: full/bin/ghost
stencil: full/ghost/bin/3d
bin: standard
Setting up Verlet run …
Unit style : metal
Current step : 0
Time step : 0.001
srun: error: cc0906: tasks 0,3,12,15,80,83,92,95: Segmentation fault
srun: First task exited 60s ago
srun: StepId=187275.0 tasks 1-2,4-11,13-14,16-79,81-82,84-91,93-94: running
srun: StepId=187275.0 tasks 0,3,12,15,80,83,92,95: exited abnormally
srun: launch/slurm: _step_signal: Terminating StepId=187275.0
srun: Job step aborted: Waiting up to 62 seconds for job step to finish.
slurmstepd: error: *** STEP 187275.0 ON cc0906 CANCELLED AT 2022-03-22T15:23:13 ***
srun: error: cc0906: tasks 1-2,4-11,13-14,16-79,81-82,84-91,93-94: Killed

8

image
image1488×760 69.6 KB

Hi relliott,

My simulation runs normally when I reduces the parallel cpu num from 96 to 24. But I find the calculation is relatively slow (7 minutes for 1000 runs of a diamond sphere of 11121 atoms). Also, is there a way to use the EDIP/C mode?

Thanks,

Best regards,

Ke Duan

9

Hi Ke,

My simulation runs normally when I reduces the parallel cpu num from 96 to 24.

The maximum number of CPUs you should use is usually dependent on your simulation size, number of atoms, etc. (here, you are taking advantage of domain decomposition in LAMMPS). So you can not use any number of processors to speed up a small-size simulation since there is not enough work for each processor. In addition, the cost of communication between processors exceeds your computational cost.

To find the optimum number for your simulation, you should do some experiments on your cluster, and it might happen that, e.g., using eight processors gives you the better performance for the current simulation.

But I find the calculation is relatively slow (7 minutes for 1000 runs of a diamond sphere of 11121 atoms).

It seems that you are able to run your simulation using the EDIP/C portable model. However, this model is costly (computationally) compared to the edip or edip/multi model.

Also, is there a way to use the EDIP/C mode?

When you are using EDIP_LAMMPS_Marks_2000_C__MO_374144505645_000 portable model in your script, you only use the EDIP/C model. This portable model is an interatomic potential for Carbon (C).

I hope this helps.

10

Hi yafshar,

Thanks for your reply! T tested, for the same simulation model, the speed of using pair_style edip in lammps is at least one order of magnitude higher than that of using EDIP_LAMMPS_Marks_2000_C_MO_374144505645_000. But, for the pair_style edip in lammps (
pair_style edip
pair_coeff * * C.edip C),

the parameters of C.edip is as follows.

image
image910×448 14.6 KB

the evolution of diamond nanoparticle is quit strange, which is quite different from the final onion-like morphology reported by your group obtained by EDIP potential.

image
image831×858 141 KB

Therefore, I am confused about whether I can use the pair_style edip (lammps) to simulate the carbon-carbon evolution.

11

Thanks for your reply! T tested, for the same simulation model, the speed of using pair_style edip in lammps is at least one order of magnitude higher than that of using EDIP_LAMMPS_Marks_2000_C_MO_374144505645_000. But, for the pair_style edip in lammps (
pair_style edip
pair_coeff * * C.edip C),

The EDIP/C is a different model than edip, it is built upon the edip and extends its generalized form. Please read Marks’ paper to know what are the differences (please note that the original edip model is developed for Silicon and not Carbon)
The computational cost of this model is higher than the edip as you observed.
I can not suggest which model to use, please read the original papers and then decide which potential to use.

12

Dear yafshar and relliott,

Thank you for your kind help in stalling EDIP/C openkim potential. I have addressed all related issues now.

Best regards,

Ke Duan

13

Hi ke,
I also want to use this edip/c potential. Do you think this edip/c potential can be used to simulate the friction behavior of carbon surface. And can I use this potential with other kinds of potential like Lennard-Jones potential or EAM.
Thanks!

Shusen Guo

14

Hi Shusen,

The EDIP model driver (software implementation) in OpenKIM (which is an extension of edip/c in LAMMPS) does not include van der Waals interactions and therefore on its own cannot be used to model friction between carbon surfaces. Also, KIM portable models (parameterizations based on the model driver) do not support hybrid mode and so cannot be combined with other potentials.

Best,

Ellad Tadmor