Problem with neb

LAMMPS
1

Hello guys I am suck to run the neb on my laptop with following error

“mpirun -oversubscribe -np 6 lmp -partition 6x1 -in in.neb.w.vac.lmp
LAMMPS (11 Feb 2026 - Development - patch_11Feb2026-99-gb75dfcc930)
Running on 6 partitions of processors

MPI_ABORT was invoked on rank 1 in communicator MPI_COMM_WORLD
with errorcode 1

since system is not allowing to upload the files I am giving inputs to generate initial and final and neb. NOTE: I have verified MPI works fine for simple min.

units metal
dimension 3
boundary p p p
atom_style atomic

variable a equal 3.1652

lattice bcc ${a}
region box block 0 8 0 8 0 8 units lattice
create_box 1 box
create_atoms 1 box

mass 1 183.84

pair_style eam/alloy
pair_coeff * * /home/irslanullahashraf/lammps/potentials/W_zhou.eam.alloy W
neighbor 2.0 bin
neigh_modify delay 10
min_style cg
minimize 1e-12 1e-12 10000 10000
write_dump all atom dump.refrence

region vac sphere 12.6608 12.6608 12.6608 0.2 units box
region vac2 sphere 9.43344 12.6608 12.6608 0.2 units box
group gv region vac
group gv2 region vac2
delete_atoms group gv compress no
#delete_atoms overlap 3.6 gv gv2 compress no

thermo 50
thermo_style custom step pe etotal press vol

min_style cg
minimize 1e-12 1e-12 10000 10000
write_dump all atom dump.init
write_data init_vac.data
clear

---------- build_final.lmp ----------

units metal
dimension 3
boundary p p p
atom_style atomic

variable a equal 3.1652

lattice bcc ${a}
region box block 0 8 0 8 0 8 units lattice
create_box 1 box
create_atoms 1 box

mass 1 183.84

pair_style eam/alloy
pair_coeff * * /home/irslanullahashraf/lammps/potentials/W_zhou.eam.alloy W
neighbor 2.0 bin
neigh_modify delay 10

region vac2 sphere 9.43344 12.6608 12.6608 0.2 units box
group gv region vac2

delete_atoms group gv compress no

thermo 50
thermo_style custom step pe etotal press vol

min_style cg
minimize 1e-12 1e-12 10000 10000

write_data final_vac.data
write_dump all atom dump.final
clear

---------- neb_w_vac.lmp ----------

units metal
atom_style atomic
atom_modify map array
atom_modify sort 0 0.0
dimension 3
boundary p p p

read_data init_vac.data

mass 1 183.84

pair_style eam/alloy
pair_coeff * * /home/irslanullahashraf/lammps/potentials/W_zhou.eam.alloy W
neighbor 2.0 bin
neigh_modify delay 10 every 1 check yes

thermo 10
thermo_style custom step pe fnorm
reset_timestep 0
fix 1 all neb 5.0
timestep 0.01
min_style quickmin

neb 1.0e-10 1.0e-8 1000 1000 50 final final_vac.data

2

Please add the -nb command line flag to LAMMPS to (temporarily) disable output buffering and check for any error messages. LAMMPS does not stop without printing an error, but it may be in a buffer that is not flushed when MPI_Abort() is called.

Your copied input is useless because it is not properly quoted.

1 Like
3

Dear Akohlmey thank you very much for suggestion

I have applied the -nb command line flag but still cant see the printed error message that is potentially stopping the simulation.

here is the error

mpirun -oversubscribe -np 6 lmp -partition 6x1 -nb -in in.neb.w.vac.lmp
LAMMPS (11 Feb 2026 - Development - patch_11Feb2026-99-gb75dfcc930)
Running on 6 partitions of processors

MPI_ABORT was invoked on rank 3 in communicator MPI_COMM_WORLD
with errorcode 1.

NOTE: invoking MPI_ABORT causes Open MPI to kill all MPI processes.

You may or may not see output from other processes, depending on
exactly when Open MPI kills them.

here are quoted inputs

---------- build_initial.lmp ----------

units metal
dimension 3
boundary p p p
atom_style atomic

variable a equal 3.1652

lattice bcc ${a}
region box block 0 8 0 8 0 8 units lattice
create_box 1 box
create_atoms 1 box

mass 1 183.84

pair_style eam/alloy
pair_coeff * * /home/user/lammps/potentials/W_zhou.eam.alloy W
neighbor 2.0 bin
neigh_modify delay 10
min_style cg
minimize 1e-12 1e-12 10000 10000
write_dump all atom dump.refrence

region vac sphere 12.6608 12.6608 12.6608 0.2 units box
region vac2 sphere 9.43344 12.6608 12.6608 0.2 units box
group gv region vac
group gv2 region vac2
delete_atoms group gv compress no
#delete_atoms overlap 3.6 gv gv2 compress no

thermo 50
thermo_style custom step pe etotal press vol

min_style cg
minimize 1e-12 1e-12 10000 10000
write_dump all atom dump.init
write_data init_vac.data
clear

---------- build_final.lmp ----------

units metal
dimension 3
boundary p p p
atom_style atomic

variable a equal 3.1652

lattice bcc ${a}
region box block 0 8 0 8 0 8 units lattice
create_box 1 box
create_atoms 1 box

mass 1 183.84

pair_style eam/alloy
pair_coeff * * /home/user/lammps/potentials/W_zhou.eam.alloy W
neighbor 2.0 bin
neigh_modify delay 10

region vac2 sphere 9.43344 12.6608 12.6608 0.2 units box
group gv region vac2

delete_atoms group gv compress no

thermo 50
thermo_style custom step pe etotal press vol

min_style cg
minimize 1e-12 1e-12 10000 10000

write_data final_vac.data
write_dump all atom dump.final
clear

---------- neb_w_vac.lmp ----------

units metal
atom_style atomic
atom_modify map array
atom_modify sort 0 0.0
dimension 3
boundary p p p

read_data init_vac.data

mass 1 183.84

pair_style eam/alloy
pair_coeff * * /home/user/lammps/potentials/W_zhou.eam.alloy W
neighbor 2.0 bin
neigh_modify delay 10 every 1 check yes

thermo 10
thermo_style custom step pe fnorm
reset_timestep 0
fix 1 all neb 5.0
timestep 0.01
min_style quickmin

neb 1.0e-10 1.0e-8 1000 1000 50 final final_vac.data

4

Have a look at the created files. As I already said, there must be an error message.

Have a look at the guidelines post how to do this properly.

1 Like
5

Dear @akohlmey

thank you very much for your guidance I am following accordingly.

kind regards

Laiba Ejaz

6

Hello Laiba Ejaz,

A quick comment. Usually, when I perform a NEB calculation, my ‘final’ file is not a data file (instead, a list of the atoms with their atomic number and modified coordinates, as detailed in the doc). Could this be an issue in your script?

Hope this helps,

Julien.

7

Thank you very much this is exactly how it should be.

8

Dear @akohlmey , I have read alot documentation and eventually able to run NEB fine, here is a problem you mentioned whenever Lammps quit it mentions errors on logfiles, unfortunatley i couldnt see error on any of logfiles when it exit with following error on screen. what could be the reason behind here is snip of log “liba@liba:~/md/tangsten/potentials_study/mason2023eam/nebiH20260407$ mpirun -oversubscribe -np 6 lmp -partition 6x1 -in in.neb_w_vac.lmp
LAMMPS (11 Feb 2026 - Development - patch_11Feb2026-99-gb75dfcc930)
Running on 6 partitions of processors
Reading NEB coordinate file(s) …
Setting up regular NEB …
Step MaxReplicaForce MaxAtomForce GradV0 GradV1 GradVc EBF EBR RDT RD1 PE1 RD2 PE2 RD3 PE3 RD4 PE4 RD5 PE5 RD6 PE6
0 2.0097488 1.4304226 8.067647e-07 0.67721206 1.5549398 0.62260406 0.61172434 2.5074103 0 -9160.3263 0.2 -9159.8936 0.4 -9159.7037 0.6 -9159.7235 0.8 -9159.9384 1 -9160.3154
[liba:21644] *** An error occurred in MPI_Wait

liba:21644\] \*\*\* reported by process \[1113587713,1

[liba:21644] *** on communicator MPI_COMM_WORLD
[liba:21644] *** MPI_ERR_TRUNCATE: message truncated
[liba:21644] *** MPI_ERRORS_ARE_FATAL (processes in this communicator will now abort,
[liba:21644] *** and potentially your MPI job)”

9

Impossible to say from remote. I notice that you didn’t use the -nb flag, though, so your log files may be truncated.

1 Like
10

Hi thank you very much, even with the flag the log files are truncated without a specific error it simply truncated. since I am studying defects I wonder it worked fine for vacancy case but show unusual truncation for interstitial. Do you have a suggestion how to approach this problem ?

11

In this case you have to do old fashioned debugging at the source code level.
I have no idea from the available information where the issue happens and what happens.
My first hypothesis would be that one of the replica is aborting or getting into an invalid state and then some inter-replica communication is out-of-sync.
This is not something that can be done from remote.