Iron Self Diffusion Mean Squared displacement

msdplot

I’ve been attempting to replicate some published results on Iron self-diffusion by running MD simulation in Lammps on an iron lattice with vacancies and plotting the mean square displacement. I am puzzled as to why when I plot the data I see a large jump at the first timestep followed by more expected behavior (becoming more linear after a long time).

My input script is as follows:

Define variables

variable T equal 350
units metal
dimension 3
boundary p p p

atom_style atomic
neigh_modify delay 0 every 1

Define Structure

Structure is a bcc iron lattice with vacancies

created in atomsk with name data.diff_Fe.lmp

read_data data.diff_Fe.lmp

pair_style meam
pair_coeff * * library.meam Fe Fe.meam Fe

mass 1 55.845

reset_timestep 0

minimize 1.0e-6 1.0e-6 100 1000
fix 1 all box/relax iso 1.0 vmax 0.1
run 10000

reset_timestep 0

unfix 1

Equilibration Run

fix 2 all npt temp $T $T 100.0 iso 1.0 1.0 1000.0
thermo 100
run 100000

unfix 2

####### Data Gathering ##############################################

velocity all create $T 91897

fix 3 all nvt temp $T T 100.0 group fe type 1 compute msd all msd com yes average yes fix 9 all vector 10 c_msd[4] variable D equal slope(f_9)/6/(dt) variable msdsc equal c_msd[4] fix 7 all print 100 "{msds}" file Femsd.txt

compute velocity autocorrelation function

compute 2 all vacf
fix 15 all ave/time 500 1 5000 c_2[4] file vacfavg.txt
fix 5 all vector 1 c_2[4]
variable vacf4 equal c_2[4]

fix print to output data

fix 14 all print 100 “${vacf4}” file vacf4.txt

thermo_style custom step temp c_msd[4]

fix 8 all print 1000 “${D}” file D.txt

thermo 100
run 2000000

Could be multiple reasons; iron atoms in the generated data file might be highly dense or opposite, and initial relaxation or aggregation might cause this until the equilibration. Also, you assign new velocities to the system, this is an another factor. Either neglect the initial timesteps or run some equilibration runs before MSD analysis, depending on your problem.

1 Like