[lammps-users] Thermal conductivity of Ar-Cu nanofluids using fix thermal/conductivity command


Hi, everyone

I’m doing simulations of thermal conductivity of Ar-Cu nanofluids using fix thermal/conductivity, but the results of nanofluids is even smaller than that of Ar base fluids, could anyone give me some suggestions, thanks

The method that I place the Cu atoms is that I carve a sphere in the fcc Ar crystals(888), and then place Cu atoms according to the density of Cu.

And I divide the simulation box into 32 layers to compute temperature gradient, and I also name the region of high and low temperature as up and down.

The timestep is 1fs.

In the following is my in file:
*****************************

# MD simulation of Ar-1%Cu thermal conductivity
# Initialization
units lj
dimension 3
newton on
boundary p p p
atom_style atomic
neighbor 0.3 bin
neigh_modify check yes
lattice fcc 0.844
region box block -4 4 -4 4 -4 4 units lattice
create_box 2 box
create_atoms 1 box
region sphere sphere 0.0 0.0 0.0 1.06920369405758 units lattice
delete_atoms region sphere
lattice fcc 3.35456791535199
create_atoms 2 region sphere
group ar type 1
group cu type 2
lattice fcc 0.844
region up1 block INF INF INF INF -0.5 -0.25 units lattice
region up2 block INF INF INF INF 0.5 0.75 units lattice
region up union 2 up1 up2
region down1 block INF INF INF INF -3.5 -3.25 units lattice
region down2 block INF INF INF INF 3.5 3.75 units lattice
region down union 2 down1 down2
mass 1 1.0
mass 2 1.59053218754711
mass0 6.633e-26
epsilon0 1.67e-21
sigma0 3.405e-10
velocity all create 0.71 458127641 mom yes rot yes dist gaussian units box
# LJ potential *****************************************************
pair_style lj/cut 2.8
pair_modify mix arithmetic
pair_coeff 1 1 1.0 1.0 # LJ parameters for Ar-Ar
pair_coeff 2 2 39.2964071856287 0.686549192364170 # LJ parameters for Cu-Cu
fix temp all temp/berendsen 0.71 0.71 0.0466
fix nve all nve
compute ke all ke/atom
variable temp atom c_ke/(1.5
1.0)
fix temp_profile all ave/spatial 1 100000 100000 z lower 0.25 v_temp file temp.profile units lattice
compute up_temp all temp/region up
compute down_temp all temp/region down
variable delta_temp equal c_up_temp-c_down_temp
fix delta_out all ave/time 1 100000 100000 v_delta_temp file delta_temp.dat
thermo_style custom step temp etotal vol
thermo_modify lost warn
thermo 100
# Run

timestep 0.000466
run 100001

fix heat_swap all thermal/conductivity 100 z 32
fix e_exchange all ave/time 100 1000 100000 f_heat_swap file e_exchange.dat
variable thermal_conductivity equal f_e_exchange/(0.000466
100.0
4.0
f_delta_out)1.3806504e-23/3.405e-10/3.405e-10sqrt(1.67e-21/6.633e-26)*6.0/8.0
fix thermal_conductivity_out all ave/time 100000 1 100000 v_thermal_conductivity file thermal_conductivity.dat

# Run
run 10000000
|

In fix thermal/conductivity, the velocities need to be swapped between atoms with same mass for energy and momentum conservation. Are you doing that?

I’ve modified the fix_thermal_conductivity.cpp file to make the Muller Plathe method can be used to execute heat flux among atoms with different masses, and I can get the same results as that using “fix thermal/conductivity” command only for Ar atoms.

If I can get good results these days, I’ve enjoy the codes with everyone.

Xiaoliang

10年4月14日,周三, Vikas Varshney <[email protected]…24…> 写道:

Dear Xiaoliang,

Not to be rude, modifying the code doesn’t mean that is will be 100% correct unless you verified it. Have you verified the code with some simulations which give similar results as other form of non-equilibrium (direct method) MD simulations?

Best Regards,
Vikas

Yes, I’ve used the codes to investigate the interfacial thermal conductance between CNT and Si, and the results is in the same dimension with other results.
And I also find that the total energy of the simulation system conserves, because I introduce energy and momentum conservation into Muller Plathe method when the atoms involved has different masses.

10年4月14日,周三, Vikas Varshney <[email protected]…24…> 写道:

Great to hear that!. May be Steve could answer you question better.

Dear Xiaoliang:

Is your system in liquid phase? You seem to have a temperature of 85K which is higher than the melting point, but you should still double check the phase.

If your system is in solid, then Cu add more scattering to the phonons and you are supposed to get smaller thermal conductivity.

Even if your system is in liquid phase, the validity of the potential may also prevent you from getting the enhancement by add Cu atoms. Also, how did you place the Cu atoms? Randomly or in a symmetric way?

Regards,

Tengfei

Dear Tengfei,
I didn’t consider the phase, because I find someone used the Muller Plathe method to solid systems, e.g. “Estimating thermal conductivity of amorphous silica nanoparticles
and nanowires using molecular dynamics simulations” written by Sanket S. Mahajan Ganesh Subbarayan and Bahgat G. Sammakia

And also I find that in the article “Molecular dynamics simulation of effective thermal conductivity and study of enhanced thermal transport mechanism in nanofluids”, Suranjan Sarkara and R. Panneer Selvam obtained higher value of thermal conductivity of Ar-Cu nanofluids using Green-Kubo method and using LJ potential.

Do you mean GK method is appropriate for computing thermal conductivity of Ar-Cu nanofluids but Muller Plathe method is not?

Thanks

Xiaoliang Zhang

10年4月14日,周三, Tengfei Luo <[email protected]…64…> 写道:

And also I’ve used the Muller Plathe method to compute the thermal conductivity of Ar base fluids, the result is reliable and close to other’s, but not for Ar-Cu nanofluids

10年4月14日,周三, Vikas Varshney <[email protected]…24…> 写道:

Dear Xiaoliang:

Both method should work on both liquid and solid materials. Physically, if you add defects to a solid system, they scatter phonons and thus effectively reduce the mean free path which leads to a lower thermal conductivity.

I suggest you run your simulation at a higher temperature. In your NEMD setup, if you have mean T=85, it’s possible that the low temperature end of your system is below the melting point. You can also easily try GK method to see if you can reproduce the results in the second paper you cited.

Tengfei