TiO2 simulation take long time

Dear Sir,

I am trying to do simulate TiO2 radiation damage using Matsui_akoagi potential. I am running this simulation in my desktop. After lmp_daily < input.txt command, I just found, it took almost 3 hours for the first step. Do you think, I need a cluster computer (HPC) for running this simulation.

-------------- Initialize Simulation ------------

units real
dimension 3
#newton on
boundary p p p
atom_style charge
atom_modify map array

----------------------- ATOM DEFINITION ----------------------------

read_data data_TiO2.lmp
write_dump all atom all_atom.lammpstrj
set atom 1 type 1
set atom 2 type 2
set type 1 charge 2.196
set type 2 charge -1.098

------------------Define Interatomic Potential -----------------

pair_style buck/coul/long 2.5
pair_coeff 1 1 717654 0.154 120.997
pair_coeff 1 2 391053 0.194 290.392
pair_coeff 2 2 271719 0.234 696.941
kspace_style ewald 1.0e-6

Groups

region rinterior sphere 57.5 57.5 57.5 53 # define inner box region used for simulation domain
region rexterior sphere 57.5 57.5 57.5 47.5 side out # define thermostat region outside of simulation domain
region outer intersect 2 rinterior rexterior
region rPKA sphere 57.5 57.5 60.4581 1.4 # Create spherical region ‘rpka’ at x, y, x with radius (73361 particle index)
group interior region rinterior # define group called ‘interior’ comprised of region ‘rinterior’
group exterior region rexterior
group heatsink region outer
group PKA region rPKA

Recoil atom

set PKA (40Kev) velocities to correspond to ~ 3715.13 Ang/ps

velocity PKA set 1000 NULL NULL units box
compute kin all ke/atom
compute kin_all interior ke
variable kine equal c_kin_all
compute pot all pe/atom
compute eatoms1 all reduce sum c_pot
variable pot equal c_eatoms1
compute tempk all temp
variable temp_all equal c_tempk
variable simu_time equal “time”
fix val1 all print 100 “{simu_time} {pot} {kine} {temp_all}” file Values_100_kev.txt screen no
dump 1 all custom 100 test_all.dump type x y z c_kin c_pot
dump 2 PKA custom 100 PKA_traj.dump type x y z c_kin c_pot
thermo 100
thermo_style custom step dt temp pe etotal press vol
thermo_modify lost warn line one flush yes format 1 “ec %8lu” format float “%20.10g”

— Run —

reset_timestep 0
timestep 0.00001
fix MYNVE1 all nve
fix MYNVT1 heatsink temp/berendsen 300 300 0.3
run 30000
unfix MYNVE1
unfix MYNVT1
timestep 0.0001
fix MYNVE1 all nve
fix MYNVT1 heatsink temp/berendsen 300 300 0.3
run 10000
unfix MYNVE1
unfix MYNVT1
timestep 0.001
fix MYNVE1 all nve
fix MYNVT1 heatsink temp/berendsen 300 1 0.3
run 10000
print “"
print “Atom radiated successully”
print "

Dear Sir,

I am trying to do simulate TiO2 radiation damage using Matsui_akoagi potential. I am running this simulation in my desktop. After lmp_daily < input.txt command, I just found, it took almost 3 hours for the first step. Do you think, I need a cluster computer (HPC) for running this simulation.

your pair style cutoff of 2.5 angstrom seems wrong? what did it say in the paper where you took the parameters from?
more typical would be a value of 10-14 angstrom. and that would change the balance between real space and kspace parts of the long-range coulomb calculation to be more efficient.
also, you can try using pppm instead of ewald, which has more favorable O(N log(N)) scaling vs. O(N**3/2).
also, you probably have multiple CPUs on your local machine and should try to see how much speedup you can get from that.
running on a cluster might help, but you first need to verify, that you are running your input efficiently and correctly.

axel.