WATER POISEUILLE IN LAMMPS

Hi dear
Excuse , I am simulating SPCE poiseuille water flow with 1058 water molecules in an nanochannel with copper(cu) walls with lammps ,with real units , the channel is a cubic with P P P boundary conditions,the height of channel is 31.66 angstrome that is equal to 10 times bigger than water molecules diameter . but excuse i have 2 problem in my work , i thank you very much if you help me . the first is this that i fix the temperature of water molecules at 300 K in this way :
fix 1 all nve # all group means cu group(walls) and the water group (oxygen and hydrogen group)(flow)
fix 2 cu nvt temp 300.0 300.0 100.0
fix 3 water temp/rescale 1 300.0 300.0 0.001 1.0
program is run in 200000 time steps,the temperature raises from 1K in first timestep and in timestes almost 20 in remained on 219.8891K and doesnot chane after that and stayes fixed on 219.8891K till program be finished(timestep 200000)! But if i fix the temperature on almost 410K : (fix 3 water temp/rescale 1 410.0 410.0 0.001 1.0) then so the temperature will be stayed on 300K till last time step,with no change,meaning the favorite temperature,but I donot know the reason , i want to fix on 300K and get the 300K . I donot know why this happens,please if you know its reason or you can give me a propose please tell me . thank you so much
The another prob is this that the reference article applied unitless force(lj units)=2.5 to the mass center of each water molecules and , 2.5 unitless for force is equal to 0.123 kcal/mol for real units in SPCE water model . so I apply 0.123/3 = 0.041 kcal/mol to each atom in water molecules,that cause to apply 0.123kcal/mol to each mass center of water molecules . but the maximum velocity that I reach (in center of cannel)is 0.0049 m/s = 490m/s while the reference tells maximum velocity is 8 (unitless) that is equal to 1490 m/s . in your idea this my way for applying the force differs with that in your article? if yes so why? And whether if it differs,can it effect this much on velocity?(490 to 1490)please if you can answer me or give me an idea . thanks a lot .

fix 1 all nve # all group means...
fix 2 cu nvt temp 300.0 300.0 100.0
fix 3 water temp/rescale 1 300.0 300.0 0.001 1.0

I am not familliar with this type of simulation, although there were
some discussions of it here:

However I do see one problem with your input script: You are
integrating the equations of motion on the cu group twice. (since the
"all" group includes the "cu" group). Both "fix nve" and "fix nvt"
integrate the equations of motion. This will cause problems.

1) One way to fix this is to exclude the "cu" group from "fix nve" by
using the "water" group:

fix 1 water nve

2) Just out of curiousity, what happens if you run the simulation
without any thermostating? (using only "fix nve". No "nvt" or
"temp/rescale")

Before you run the simulation, you can initialize the velocities
randomly using the "velocity" command (consistent with thermal motion
at 300K):

You can also use the "velocity" command (again) to add a non-random
vector to your random velocities to create an initial flow in any
direction you choose (using "velocity ... sum ...")

When you do this, does the temperature of the water increase too much
over time? If not, then try that.

3) Because "fix langevin" does not ou can use "fix langevin" on the
"cu" group (assuming you want to continue to use "temp/rescale" to
maintain the temperature of the water group.
fix 2 cu langevin 300.0 300.0 100.0 48279

Alternately, you can use "fix langevin" on the water as well ("all"
group). But I'm not sure this is a good idea. In that case, you
should probably use a much weaker damping-time parameter (ie larger
than "100.0"), to prevent it from slowing down the particle flow
significantly.
fix 2 all langevin 300.0 300.0 50000.0 48279
(You will have to adjust this number.)

Hopefully someone else more knowledgeable will reply.
Cheers
Andrew