npt and wall/lj126

Hi all,

I have a trouble to start a simulation with fix npt and fix wall/lj126. System contains dendrimer, Na & Cl atoms,1200 H2O molecules (TIP3P) and input looks like this :

echo log
log den_eq.4.log
dimension 3
units metal
boundary p p p
timestep 0.001500
atom_style full
read_data data.den_eq.4

group aaa type 1 2 3 4 5 6
group bbb type 7 8 9 10 11 12 13 14
group ccc type 15 16 17 18 19
group Na type 20
group Cl type 21
group H2O type 22 23
group dendrymer union aaa bbb ccc

fix bal all balance 10000 xyz 20 1.1

pair_style lj/cut/coul/long 12.000000
pair_coeff 1 1 0.004743 3.399670 #c3_a-c3_a

kspace_style ewald 1.0e-4

bond_style harmonic
bond_coeff 1 29.127733 1.093000

angle_style harmonic
angle_coeff 1 4.020141 110.070000

dihedral_style multi/harmonic
dihedral_coeff 1 0.054631 0.039022 -0.041624 -0.052030 0.000000

variable zllo equal zlo
variable zhhi equal zhi
variable Etot equal etotal

fix wallo all wall/lj126 zlo v_zllo 0.3 1. 5. units box pbc yes
fix walhi all wall/lj126 zhi v_zhhi 0.3 1. 5. units box pbc yes

thermo_style custom temp press pe lx ly lz
thermo 500

fix shake H2O shake 0.0001 20 0 b 11 a 20
fix fxnpt all npt temp 310.0 310.0 0.1 iso 1. 1. 1.

dump xyz all xyz 20 den_eq.4.xyz
fix tave all ave/time 2 1 2 c_thermo_temp c_thermo_pe c_thermo_press v_Etot file den_eq.4.avs
run 2000
write_restart den_eq.4.restart

and I get

Temp Press PotEng Lx Ly Lz
0 -425.41473 113.93615 100 100 200
470.54449 -282.16068 -6039.3123 91.885824 91.885824 183.77165
324.48267 -241.03243 -7798.1488 85.942783 85.942783 171.88557
-nan -nan -nan -nan -nan -nan
-nan -nan -nan -nan -nan -nan
Loop time of 403.871 on 32 procs for 2000 steps with 42164 atoms

When fix wall is off everything is ok.
any suggestions why fix wall is not working here?

I suggest you print out thermo every step and include the timmestep.
Something is blowing up. Probably b/c your atoms are overlapping
the wall too closely. It also makes little sense to have a wall
in the z dimension when you are also doing NPT in the z dimension.
Your box is shrinking by 2x as well in 500 steps, which is bad dynamics.

Steve

I changed NPT, now doing in x & y dimension and I also changed fix wall/lj to wall/reflect-simulations run.

What I’m trying to do is first equilibrated my system in NPT. Next switch to NVT and add a gold surface in z dimension.
it is a good way, or it can be done better?

there are many ways to do that. My earlier point
is that it is odd to do NPT on a dimension where
you also have a wall. The two commands are competing
with each other.

Steve