MAGICS Workshop

Thermal Conductivity Hands-On Session

This work was supported as part of the Computational Materials Sciences

Program funded by the U.S. Department of Energy, Office of Science, Basic
Energy Sciences, under Award Number DE-SC00014607
The goal of this session is to calculate DOS,
Cv and thermal conductivity of
2-D materials with quantum correction
 Outline

Ø Thermal conductivity calculation in MoS2

Ø Length scaling
Ø Temperature scaling

Ø PDOS, DOS and Cv calculation

Ø Quantum correction for thermal conductivity

 Software Download
LAMMPS Plugins for Thermal Conductivity Calculation
https://magics.usc.edu/thermal-conductivity-plugin/
LAMMPS Plugins for Thermal Conductivity
Contents of your Home Directory

~/Home

DOS_Cv Thermal_Conductivity

Length_Scaling Temperature_Scaling Isotope_TC

600A 800A 1000A 100K 200K 300K

LAMMPS Plugins for Thermal Conductivity
Each sub directory of Length_Scaling and Temperature_Scaling following files

1. in.relax: LAMMPS script to create a relaxed system at a given Temperature

2. in.heatflux: LAMMPS script to for thermal conductivity calculation
3. in.variables: Variable definitions that are used in the LAMMPS Script
4. MoS2.data: Input unit cell coordinate of MOS2 monolayer
5. MoS2.sw: SW Interaction potential for MOS2
6. MoS2.restart: Restart file to be used by LAMMPS.
7. job.pbs: PBS file used to submit jobs to the HPC
8. calthermal_conductivity.py: Python code to post-process LAMMPS data to
compute thermal conductivity value
9. input.txt: Input parameters for calthermal_conductivity.py
 First Set of Job Submissions

STEP 1: Go to Thermal_Conductivity folder in your home directory

cd ~/Thermal_Conductivity

STEP 2: Navigate to Length_Scaling folder

cd ./Length_Scaling

STEP 3: Navigate to 1000A folder

cd ./1000A

STEP 4: Submit your first Job

qsub job.pbs
First Set of Job Submissions (Contd.)

STEP 1: Get back to Length_Scaling folder

cd ~/Thermal_Conductivity/Length_Scaling

STEP 2: Navigate to 800A folder

cd ./800A

STEP 3: Submit your second Job

qsub job.pbs
Steps for Thermal Conductivity Calculation

Step 1: Create a relaxed system at Temperature = TK

use LAMMPS script: in.relax

Step2: Add/remove heat to/from the system for 10-12ns

use LAMMPS script: in.heatflux

Step3: Compute thermal conductivity

post process LAMMPS output file using calthermal_conductivity.py
LAMMPS Walkthrough: Simulation Units

Units used in LAMMPS here for Molecular dynamics Simulation – Metal.

Example : Timestep is 1 Femtosecond = 0.001 (in LAMMPS Metal unit)

Parameter Units

Distance Angstroms (Å)

Time Picoseconds (ps)

Energy Electron Volts (eV)

Temperature Kelvin (K)

 Step 1: Create a relaxed system
Input script: in.relax
units metal
atom_style atomic
1. Create a system
boundary p p p 2. Do energy minimization
processors ${px} ${py} ${pz}
read_data MoS2.data Input Crystal structure
replicate ${xnum} ${ynum} ${znum}
group watom type 1
group seatom type 2
neighbor 2.0 bin
neigh_modify delay 0 every 1 check yes
pair_style sw Interaction potential
pair_coeff * * MoS2.sw Mo S
thermo_style custom step temp pe ke etotal press vol
thermo 50
dump 1 all atom 100 dump.min
min_style cg Energy minimization
minimize 1.0e-8 1.0e-8 5000 10000
 Step 1: Create a relaxed system
Input script: in.relax
3. Heat to temperature T
4. Relax to temperature T

velocity all create ${SysTemp100} 156467 mom yes rot yes dist gaussian
fix 2a all nve
run ${Steps40ps} Heat to T
fix 5 all nve
dump 4 all atom 50000 dump300.nve Relax at T
run ${Steps600ps}
unfix 5
write_restart MoS2.restart Input file for in.heatflux
for thermal conductivity
 LAMMPS Walkthrough: Variables
File: in.variables
Ø Contains variable definitions
Ø Here are some of the important ones

variable px equal 20 # Number of processors in the x direction

variable py equal 4 # Number of processors in the y direction
variable pz equal 1 # Number of processors in the z direction
variable SimTimestep equal 0.001 # Timestep used. Here 1 femtosecond
variable Steps300K equal 300000 # Number of steps to run. Here 300000
variable HeatAdded equal 0.00431*100 # Amount of heat added. 0.431eV
variable HeatRemoved equal -0.00431*100 # Amount of heat removed
0.431eV
 Schematic of Thermal Conductivity
Simulation Setup
#
E0=Heat given, ! = $

E0= Heat Taken Out

%#
!= $

Length =W
=L h
idt
W
Ø Define Hot and Cold region in the system at X=L/4 and X=3L/4
respectively
Ø E0 heat is given to the hot region and E0 heat is taken away from
cold region.
 Step 2: Add/Remove heat to/from system
Input script: in.heatflux
Read relaxed structure coordinate
1. Read relaxed structure coordinate
boundary p p p
processors ${px} ${py} ${pz} 2. Define two strips L/2Å apart
read_restart MoS2.restart
neighbor 2.0 bin
3. Add/remove heat from these strip
4. Run system for 4ns
5. Take temperature average over 4ns
6. Continue this process at least thrice

Define strips and add/remove heat from these regions

region hot block 240 260 INF INF INF INF units box
region cold block 740 760 INF INF INF INF units box
group hot region hot
group cold region cold
fix 1 hot heat 100 ${HeatAdded} region hot
fix 2 cold heat 100 ${HeatRemoved} region cold
 Step 2: Add/Remove heat to/from system
Input script: in.heatflux
1. Read relaxed structure coordinate
2. Define two strips L/2Å apart
3. Add/remove heat from these strip
4. Run system for 4ns
5. Take temperature average over 4ns
6. Continue this process for at least 3 time
Take Temperature average for 4ns
compute myKE all ke/atom
compute Thot hot temp/region hot
compute Tcold cold temp/region cold
variable atemp atom c_myKE/(1.5*8.621738*0.00001)
timestep ${SimTimestep}
fix 6b all ave/spatial 1 ${T_avg4ns} ${T_avg4ns} x lower
20.0 v_atemp file Temperature.txt units box
run ${T_avg4ns}
 Hands on Calculations
Length Scaling:
1. Pre-relaxed system at 300K for 12ns is inside L_scaling folder:
a) 600Å×100Å : MOS2_600L.restart
b) 800Å×100Å: MOS2_800L.restart
Already Submitted Earlier
c) 1000Å×100Å: MOS2_1000L.restart
2. Run in.heatflux for 300000 steps
4. Compute thermal conductivity using calthermal_conductivity.py
 Job Submission: Job 3 (Length Scaling)

STEP 1: Navigate to Thermal_Conductivity folder in your home

directory
cd ~/Thermal_Conductivity

STEP 2: Navigate to Length_Scaling folder

cd ./Length_Scaling

STEP 3: Navigate to 600A folder

cd ./600A

STEP 4: Submit your Third Job

qsub job.pbs
 Hands on Calculations

Temperature Scaling:
1. Pre-relaxed system at 100K, 200K, 300K for 12ns is inside T-scaling folder:
a) 800Å×100Å , 200K : MOS2_200T.restart
b) 800Å×100Å , 100K : MOS2_100T.restart
Already Submitted Earlier
c) 800Å×100Å , 300K : MOS2_200T.restart
2. Run in.heatflux for 300,000
3. Compute thermal conductivity using calthermal_conductivity.py
Job Submission: JOB1 (Temperature Scaling)

STEP 1: Get to Temperature_Scaling folder

cd ~/Thermal_Conductivity/Temperature_Scaling

STEP 2: Navigate to 100K folder

cd ./100K

STEP 3: Submit your fourth Job

qsub job.pbs
 Step 3: Compute Thermal Conductivity

Compute thermal conductivity using calthermal_conductivity.py

Ø Input parameters for calthermal_conductivity.py is defined

in input.txt

Note: you can find these parameters in in.heatflux and lammps

dump file
 Sourcing Python

Before we start plotting anything we must source python. Follow these steps

STEP 1: Go to your home directory

cd ~/

STEP 2: Source Python

source /usr/usc/python/default/setup.sh
 Step 3: Compute Thermal Conductivity
Compute thermal conductivity using calthermal_conductivity.py
$ python3.5 calthermal_conductivity.py Temperature.txt

Note: Temperature.txt is the output file created by in.heatflux

Output:
a)K and 1/K value
Step 3: Compute Thermal Conductivity (4ns data)

Compute thermal conductivity using calthermal_conductivity.py

python3.5 calthermal_conductivity.py Temperature.txt

File: Temperature.txt Also Outputs : 300K_steps.png

1. Contains Temperature
profile averaged over 300K
Steps (0.3 ns)
2. Noisy
Output : Length: 39.999983500000006
1/Length 0.02500001031250425
Mean Thermal Conductivity value and
standard deviation 249.719363936 0.0
1/(Thermal_Conductivty) value and
standard deviation 0.00400449522312
0.0
Step 3: Compute Thermal Conductivity (4ns data)

Compute thermal conductivity using calthermal_conductivity.py

python3.5 calthermal_conductivity_4ns.py 20Temperature4.txt

File: 20Temperature4.txt Also Outputs :4ns.png

Much Better
1. Contains Temperature Fit compared
profile averaged over 4ns to 300Ksteps
2. Much Less Noise

Output : Length: 39.999983500000006

1/Length 0.02500001031250425
Mean Thermal Conductivity value and
standard deviation 18.6176254622 0.0
1/(Thermal_Conductivty) value and
standard deviation 0.0537125425599
0.0
Job Submission: JOB 2 Temperature Scaling
STEP 1: Navigate to Thermal_Conductivity folder in your home
directory
cd ~/Thermal_Conductivity

STEP 2: Navigate to Temperature_Scaling folder

cd ./Temperature_Scaling

STEP 3: Navigate to 200K folder

cd ./200K

STEP 4: Submit your Fifth Job

qsub job.pbs
Job Submissions : Isotope (JOB 1)

STEP 1: Get to Isotope_TC directory

cd ~/Thermal_Conductivity/Isotope_TC

STEP 2: Submit job

qsub job.pbs
 Step 4: Plot Length scaling
For 600Å×100Å, 800Å×100Å, 1000Å×100Å
compute K value and plot K-1 vs L-1

1. Poor Length Scaling because of insufficient time averaging

2. Extremely noisy temperature gradient
Step 4: Plot Length scaling (4ns System)
For 600Å×100Å, 800Å×100Å, 1000Å×100Å compute K
value and plot K-1 vs L-1 using 20Temperature4.txt
Note: Temperature profiles have been averaged over 4 ns
instead of 0.3ns (300Ksteps)
 Step 5: Plot Temperature scaling
For 800Å×100Å at temperature 100K, 200K and 300K
compute K value and plot K vs T
Note: For 800Å×100Å at 300K use the Thermal conductivity
value from the Length Scaling System
PDOS, DOS and Cv Calculation
 Software Download
LAMMPS Plugins for Thermal Conductivity Calculation
https://magics.usc.edu/thermal-conductivity-plugin/
LAMMPS Plugins for Thermal Conductivity

DOS_Cv: Contains following files

1. in.dos: LAMMPS script to create a relaxed system at Temperature=T K
2. MOS2.data: Input unit cell coordinate of MOS2 monolayer
3. MoS2.sw: SW Interaction potential for MOS2
4. caldos.py: Python script to run dos.c and save images of of velocity
autocorrelation, PDOS, FDOS, Cv in the folder image.
5. dos.c: C program to compute velocity autocorrelation, PDOS, TDOS,
Cv from LAMMPS dump file
6. dos.h: header file for dos.c
7. input.txt: contains input parameters for DOS calculation , used by
dos.c
Major steps involved in calculating DOS

Step 1: Create a relaxed system at a given Temperature

use LAMMPS script in.dos

Step2: Compute DOS from LAMMPS Dump file

use caldos.py
Step 1: Create a relaxed system
Input script: in.dos
1. Create a system
2. Do energy minimization
3. Heat and relax to temperature T
4. Save coordinates for analysis

Used to calculate Velocity autocorrelation,

PDOS, DOS and Cv by caldos.py
 Hands on Calculations
Compute Velocity autocorrelation, Density of states using LAMMPS dump file

Velocity autocorrelation Density of State

Ø Compute Velocity autocorrection (Za) for Mo & using 50 initial condition

Ø Partial Density of State:
%& /
!" # = " ∫0 )" * +,- #* .*
'
Ø Total Density of State :
! # = 1 !" #
"
 Hands on Calculations
Velocity autocorrelation, Density of states, Cv calculation:
1. Dump file for a relaxed system is inside DOS_cv_plugins folder:
Ø dumpdos.nve
2. Compute velocity autocorrelation, density of states and Cv
python3.5 caldos.py dumpdos.nve
Note: caldos.dy run a C program called dos.c which does this calculations.
dos.h is header file for dos.c(no need to edit this file).Various parameters
required for the calculation is defined inside input.txt file(you will need to
change parameters here).

Output : images and values of DOS, Velocity autocorrelations of each

element and Specific Heat.
 Input parameters for dos.c
Input parameter file: input.txt

from in.dos file:

TFREQ in input.txt
NFRAME

Note: !"#"$"%& × !)%* + Corlength ≥ !-./01

If this condition is not satisfied you will get error message
 Hands on Calculations

1. Calculate Velocity autocorrelations, PDOS,DOS and Cv

with one initial condition (Ninitial =1 ) and with 10 initial
condition (Ninitial =10 )
2. Compare the results
Quantum Corrected Thermal Conductivity

1 )* +)
!" ∫2 . / 0/ ℏ/
+) − - *
= 1 , 4 =
#$%& ∫2 . / 0/ %6 7

!"
8 9:;;<9=<> = ×8 @A
#$%&

Hands-on Calculation:
1. Take the value of Cv at 100K,200K and
300K from Specific_heat.txt file
2. Take the value of K computed at 100K,
200K and 300K for 800Å×100Å system
3. Multiply these two number to get
quantum corrected K value
 Hands on Calculations : Isotope Effect

Ø Compute thermal conductivity value of 800Å×100Å at 100K

with isotopes
python2.7 calthermal_conductivity.py Temperature.txt

Ø Compare the thermal conductivity value of 800Å×100Å at

100K with and without isotopes
 Stuff to keep in mind -1

Average temperature over longer time

Averaging for longer gives a very Extremely Noisy Temperature

clean gradient and almost a gradient. Gives erroneous slope
linear gradient values
 Stuffs to keep in mind -2

Heating rate is important : Very high heating rate can give

anomalous result

Note the weird bump at heat

source. This can be avoided
using a reasonable heating
rate.
Also Note the rather flat
heat sink.