[lammps-users] minimize with buck/coul potential

LAMMPS LAMMPS Mailing List Mirror
1

dear all,

I want to simulate lithium metalisicate with buck/coul potential , my problem is that minimization is not efficient , the input for minimizaed all the configuration is:

input:

label minimizeloop
clear
variable a loop 200
variable r equal $a*1000
read_restart config.$r

reset_timestep 0
thermo_style custom step atoms pe temp enthalpy press

neighbor 0 bin
neigh_modify every 1 delay 0 check yes
pair_style buck/coul/long 8
kspace_style pppm 0.0001
pair_coeff 1 1 40533.77691 0.351320 4954.646367
pair_coeff 1 2 55098.73098 0.248870 765.3239006
pair_coeff 1 3 247362.358 0.208510 1632.163337
pair_coeff 2 2 154857.0464 0.146420 0
pair_coeff 2 3 5515573.834 0.106060 0
pair_coeff 3 3 20126639610 0.065700 0

thermo 10
dump fotos all custom 1000 confmin$r type id xu yu zu

minimize 1.0e-10 1e-5 500 10000

next a
jump loop-minimize minimizeloop

always the stopping criterion = linesearch alpha is zero
Is it compatible the k_space and pair_style with the minimization?
any suggestions are appreciated !
Cristian Balbuena
Universidad Nacional del Sur, Argentina

2

this could help.

increase the force tolerance.
Use the line quadratic option in min_modify command.

3

In addition to Vikas's suggestions, you can

a) try pair_modify shift yes
b) increase the PPPM tolerance
c) make sure you are using the most current patched version

The reason for a) and b) are explained on the minimize doc page.

Steve

4

thanks for the suggestion, I will consider it.
But I still confused with the use of pair_style command.
The problem is as follows:
My potential is buck/coul, if I want to use a cutoff for the Buck (dispersion forces) and a sum of Ewald or PPPM for Coulombic interactions , the syntax would be, according to the manual:

- pair_style buck/coul cut long 7.5
kspace_style ewald 0.0001

but this make the followins error: invalid pair_style

if i put:
- pair_style buck/coul/cut long 7.5
kspace_style ewald 0.0001
generate this error: KSpace style is incompatible with Pair style

testing alternatives there are two options that work:
- pair_style buck/coul/long 7.5
kspace_style ewald 0.0001

- pair_style buck/coul/long long 7.5
kspace_style ewald 0.0001

question: Do any of the situation is that I want? generate, both runs, different thermodynamic values

lammps version I use is may2009 with the corresponding manual.
thanks for any suggestions,have a nice day !

Cristian

my input is:

units real
boundary p p p
atom_style charge

lattice bcc 2.78
region superbox block 0 6 0 6 0 6
create_box 3 superbox
create_atoms 1 box

set group all type/fraction 2 0.477 3242
group oxigeno type 1
group resto type 2
set group resto type/fraction 3 0.19 3242
group silicio type 3
group litio subtract all oxigeno silicio

mass 1 16
mass 2 6.9
mass 3 28

set group oxigeno charge -1.38
set group litio charge +0.87
set group silicio charge +2.40

neighbor 0 bin
neigh_modify every 1 delay 0 check yes

thermo_style custom step pe ke temp press

pair_style buck/coul/long 7.5
kspace_style ewald 0.0001

pair_coeff 1 1 40533.77691 0.351320 4954.646367
pair_coeff 1 2 55098.73098 0.248870 765.3239006
pair_coeff 1 3 247362.358 0.208510 1632.163337
pair_coeff 2 2 154857.0464 0.146420 0 #118.2164436
pair_coeff 2 3 5515573.834 0.106060 0 #252.1135755
pair_coeff 3 3 20126639610 0.065700 0 #537.6684751

velocity all create 4000 87287
fix 01 all nve/limit 0.1
thermo 100
timestep 1
run 5000
unfix 01

fix 1 all nvt 4000 4000 20
thermo 500
run 2000
unfix 1

fix 2 all nvt 4000 3000 20
thermo 500
run 200
unfix 2

fix 3 all nvt 3000 3000 20
thermo 500
run 2000
unfix 3

fix 4 all nvt 3000 2000 20
thermo 500
run 200
unfix 4
…follows !

5

The pair style buck/coul and the pair styles buck/coul/cut and buck/coul/long
are different and take different args. The former does a long-range calculation
for both the Coulomb and exp() terms. The latter two do not. The doc
pages explain all of this. The former is also a user add-on package, which
you have to include before you build LAMMPS if you want to use it.
That is probably why you got the error below.

Steve