Compute msd

I want to calculate the diffusion coefficient of azote in In0.1Ga0.9N . I used compute msd in lammps but but the curve is not linear and different with article. this my input file:units metal
dimension 3
boundary p p p
atom_style atomic
read_restart file.restart.100000000
#group delete id 7516
#delete_atoms group delete

#region delete block 0.0 57.1409188377 0.0 76.977424703 0.0 31.1035765747

#delete_atoms porosity delete 0.05 45348

group Ga type 2
group N type 1
group In type 3

#pair_style polymorphic
#pair_coeff * * InGaN_sw.poly N Ga In
#pair_style sw
#pair_coeff * * GaNAl.sw Ga N Al
pair_style sw
pair_coeff * * InGaN.sw N Ga In

neighbor 2.0 bin
neigh_modify delay 10 check yes

timestep 0.001
compute new all temp

#delete_atoms overlap 0.1 all all

#velocity all create 3100 12345 temp new
#fix F1 all box/relax iso 0 vmax 0.01 fixedpoint 0 0 0
fix 1 all npt temp 3100 3100 0.1 iso 0.0 0.0 1
compute mymsd N msd

thermo 100000
thermo_style custom step ke etotal temp vol lx ly lz c_mymsd[4]

#fix msdout N ave/time 100000 1 100000 c_mymsd[4] file msd1.out
#dump msdout all custom 100 msd.out c_mymsd[4]

dump MyDump all custom 10000000 dump.atoms id type x y z vx vy vz fx fy fz
dump position all custom 100000 id type x y z

#min_style cg
#minimize 1.0e-10 1.0e-10 5000 10000
restart 100000000 file.restart
run 10000000
#fix F1 all box/relax iso 0 vmax 0.01 fixedpoint 0 0 0
variable cohesive_energy equal pe/atoms
#variable a equal "lx/
#variable c equal "lz/
print “Calculated cohesive energy: ${cohesive_energy} eV/atom”
any suggestion!

Sorry, but “I don’t get the result that I want” is not something that you can expect much help here.

Debugging your input and figuring out why you get the results you get, is your job, as is the case for everybody else. The forum can help you if you have more specific questions about LAMMPS. This seems to be more about that particular research topic.

You have to make certain that you are replicating all conditions and settings that are mentioned in the reference publication. You have to make certain, that you have a meaningful simulation with meaningful results. For the most part, those are topics to discuss with your adviser/tutor/mentor/colleagues or anybody that is familiar with your research and cares about you getting meaningful results. From remote there is not much that can be done, anyway.