fluctuate of output

Hi everybody
I simulated a box of 200 BMIM PF6( ionic liquid imidazolium based) ,
I need density to fit with custom equation ,
I ran the box in 9 different pressures , in experimental data with
increasing pressure density increase too , And the result of fitting
was R^2=1 , but the output of simulation density fluctuated by the
increasing of pressure(the same pressures with experimental) ,The
maximum error of data of simulation with experimental data was 0.5
percent , but because of fluctuation of density the result of fitting
was R^2=0.87 , It’s too bad , Do you think it’s because of a lammps
bug or what else??? I used the 22 july 2014 version , in your opinion
if I changed the version This problem will be solved
thanks

Hi everybody
I simulated a box of 200 BMIM PF6( ionic liquid imidazolium based) ,
I need density to fit with custom equation ,
I ran the box in 9 different pressures , in experimental data with
increasing pressure density increase too , And the result of fitting
was R^2=1 , but the output of simulation density fluctuated by the
increasing of pressure(the same pressures with experimental) ,The
maximum error of data of simulation with experimental data was 0.5
percent , but because of fluctuation of density the result of fitting
was R^2=0.87 , It’s too bad , Do you think it’s because of a lammps
bug or what else??? I used the 22 july 2014 version , in your opinion
if I changed the version This problem will be solved

unlikely.

you don't say anything about how long your equilibrated and ran your
simulations and whether you verified that your densities are actually
well converged with both simulation time and system size. otherwise
you may be just have devised a very time consuming way of generating
"randomized numbers".

you also don't say anything about what kind of potential you are using
(and whether you include long-range electrostatics and how well you
converged that etc. etc.). this all can have a significant impact, not
to mention that your model may be neglecting significant contributions
that are necessary to have a perfect match with experimental data
(like most simple classical models do).

axel.

6400 atoms
6200 bonds
12000 angles
11800 dihedrals
800 impropers
17 atom types
17 bond types
30 angle types
31 dihedral types
3 improper types
-25 25 xlo xhi
-25 25 ylo yhi
-25 25 zlo zhi

units real
dimension 3
boundary p p p
atom_style full
bond_style harmonic
angle_style harmonic
dihedral_style opls
improper_style cvff
pair_style lj/cut/coul/long 12.000
pair_modify mix geometric tail yes
kspace_style pppm 1.0e-4
read_data ionicdata.data
neighbor 2.0 bin
neigh_modify delay 2 every 1 check yes
velocity all create 398.15 4928459 rot yes dist gaussian
group 1 type 1 2 3 4 5 6 10 11 12 13 14 15 16
group 2 type 7 8 9 17
variable etotal equal etotal
variable ke equal ke
variable pe equal pe
fix 1 all npt temp 398.15 398.15 10.0 iso 7.2045 7.2045 100.00
fix m all print 500 "\{etotal\} {ke} ${pe}" file energy.txt screen no
timestep 1
thermo 100
thermo_style custom step temp etotal ke pe press vol lx ly lz pxx
pyy pzz ecoul elong evdwl epair cpu
dump 1 all xyz 1000 ionicdata.xyz
restart 1000 restart1 restart2
minimize 1.0e-6 1.0e-8 1000 100000
run 1000000

hey,

just dumping a random input on people here doesn't answer any of the
questions i posed. even more so, those questions are not so much meant
to be answered by people here, but they are something that *you*
should be thinking about. those are questions that (good) reviewers
will want to be answered in any potential publication of yours, and
they are just part of doing a statistical mechanically sound
assessment of your simulation. after all, it is *your* work and *you*
have to know what you are doing. nobody here is going to do your work
for you.

axel.

and I average the volume of output of 50000 last step to compute density ,

and I average the volume of output of 50000 last step to compute density ,

again, that doesn't mean anything. you have to check convergence of
that yourself. and also whether your system is well equilibrated at
that point. for all i know, ionic liquids are not quite as
straightforward to handle in that respect. that also requires checks
for finite size effects.

like i mentioned before, this is all part of the standard protocol of
good MD simulation practices. thus you need to discuss this with your
adviser or whoever is training/advising you on MD simulations. this is
not something that is answered in a simple do-this-not-that kind of
e-mail on a mailing list.

axel.