read_restart, read_data result in differenttemperature at the start of new run

I wish there is something like that but that is all I've got. I've just performed conversion from binary to ASCII and this is what I got after restart - two completely different results 1st - cluster remains consistent, 2nd - cluster explodes.

I wish there is something like that but that is all I've got. I've just performed conversion from binary to ASCII and this is what I got after restart - two completely different results 1st - cluster remains consistent, 2nd - cluster explodes.

what is your original input?
i.e. the one that created the restart.

lammps will recreate more than just the coordinates
from the restart file. restart2data will drop all of that.

there *has* to be something missing.

otherwise please provide the necessary means
to easily reproduce this from scratch so that it
can be properly debugged.

thanks,
    axel.

Example below is part of larger sim so it would be difficult to provide full original input
however I've managed to reproduce this problem with relatively easy lj-case:

first input - that writes restart file at the end of sim:

units real
atom_style atomic

lattice fcc 4.73
region box block -50 50 -50 50 -50 50 units lattice

create_box 1 box

lattice fcc 4.73

region LJDrop sphere 0 0 0 5 units lattice
create_atoms 1 region LJDrop

mass 1 40
velocity all create 10 1

pair_style lj/cut 6.0
pair_coeff * * 0.249792 3.345 6.0

neighbor 2 bin

compute pe all pe/atom
compute ke all ke/atom
dump mydump all custom 100 dump.lammpstrj id type x y z fx fy fz vx vy vz c_pe c_ke
log log.thermo.lammpsfirst

timestep 1

thermo 10

fix 1 all nve
fix 2 all temp/rescale 100 10 10 0.0 1.0
run 10000
unfix 2
run 10000

write_restart restart.lj

Second one that restart sim form restart.lj (and goes very well):

read_restart restart.lj

pair_style lj/cut 6.0
pair_coeff * * 0.249792 3.345 6.0

neighbor 2 bin

compute pe all pe/atom
compute ke all ke/atom
dump mydump all custom 100 dump_r1.lammpstrj id type x y z fx fy fz vx vy vz c_pe c_ke
log log.thermo.lammpssecond_r1

timestep 1

thermo 10

fix 1 all nve
fix 2 all temp/rescale 100 10 10 0.0 1.0
run 10000
unfix 2
run 10000

Third that restart from data file created form restart.lj with restart2data (it requires pair_style definition before read_data):

pair_style lj/cut 6.0

read_data data.lj

neighbor 2 bin

compute pe all pe/atom
compute ke all ke/atom
dump mydump all custom 100 dump_r2.lammpstrj id type x y z fx fy fz vx vy vz c_pe c_ke
log log.thermo.lammpssecond_r2

timestep 1

thermo 10

fix 1 all nve
fix 2 all temp/rescale 100 10 10 0.0 1.0
run 10000
unfix 2
run 10000

Example below is part of larger sim so it would be difficult to provide full original input
however I've managed to reproduce this problem with relatively easy lj-case:

Third that restart from data file created form restart.lj with restart2data (it requires pair_style definition before read_data):

pair_style lj/cut 6.0

here is the problem!
you have to add "units real" to the input
when reading the data file.
LAMMPS defaults to "units lj"

this is one of the settings that gets stored
in the restart and applied from it, but is
not in the data file. if i add the line,
the log files produce identical outputs.

cheers,
    axel.