ewald/disp vs ewald/n

Hi everyone,

I am trying to reproduce a mean square displacement calculation carried out by a previous student in our group. This calculation was performed with an older version of lammps using the ewald/n kspace_style and buck/coul/long pair style.

When I try to perform the same calculation with “ewald/disp” I get much higher mean square displacement values. I tried both the buck/coul/long and the buck/long/coul/long pair_styles.

It would be great if someone could give me some ideas as to what kspace_style – pair_style combination I should use to obtain similar results. The input file is below.

Prateek

LiLaZrO garnet

Initiation

units metal

atom_style charge

boundary p p p

read_data LLZ.data

Atom definition

group La type 1

group Zr type 2

group O type 3

group Li type 4

set group La charge 3.00

mass 1 138.91

set group Zr charge 4.00

mass 2 91.224

set group O charge -2.00

mass 3 15.999

set group Li charge 1.00

mass 4 6.941

Interatomic potential

kspace_style ewald/n 1.0e-5

pair_style buck/coul/long 8.0

pair_coeff * * 0.00 0.100000 0.000000

pair_coeff 1 3 5436.827 0.293900 0.000000

pair_coeff 2 3 7290.347 0.261000 0.000000

pair_coeff 3 3 25.410 0.693700 32.320000

pair_coeff 3 4 426.480 0.300000 0.000000

Simulation parameters (set Tdamp~30-50 for T < 500 K)

fix 1 all nvt temp 1700 1700 100

compute msd all msd

compute msdLi 4 msd

compute msdLa 1 msd

compute msdZr 2 msd

compute msdO 3 msd

thermo 200

thermo_style custom step time temp press px py pz pe ke etotal c_msd c_msdLi c_msdLa c_msdZr c_msdO

dump 1 all custom 200 xyz.dat id type x y z

dump 2 all custom 200 velocity.dat id type vx vy vz

dump 3 all custom 200 forces.dat id type fx fy fz

Run simulation (dt = 1 fs, t = 10 ns)

timestep 0.001

run 10000

Hi everyone,

I am trying to reproduce a mean square displacement calculation carried out
by a previous student in our group. This calculation was performed with an
older version of lammps using the ewald/n kspace_style and buck/coul/long
pair style.

When I try to perform the same calculation with “ewald/disp” I get much
higher mean square displacement values. I tried both the buck/coul/long and
the buck/long/coul/long pair_styles.

you get a reliable MSD form a 10ps trajectory???
i see this is at high temperature, but 10ps is awfully short.

axel

Sorry about that, I am actually using 10 ns.

This is a sample file the previous group member sent me, its missing a few zeros. She also used 10 ns.

PS. I resent this email as I only replied to Axel before.

Hi Prateek,

I really have no idea whether this might help, but some time ago there was an
issue that pair/buck/long did not provide the same results as in older
versions. This could be solved when using the

pair_modify table 0

command.

In case you still have the log files of the old simulations, I would also
recommend you to check whether the Ewald parameter and the
number of k vectors for the Ewald summation in the old simulation
were equal to those in the new simulation.

Best,
Rolf