BoltzTraP Tutorial for Quantum Espresso

Hello from -15°C Boston!

This is a tutorial that I feel needs to be made, especially as I have found not clear tutorial on this available for Quantum Espresso. I apologize for the brevity; in this tutorial I wanted to create a succinct guide on how to use BoltzTraP. Please note, I am not the developer of BoltzTraP and have no formal affiliation with the software creators.

What is BoltzTraP?

BoltzTraP is a code that can calculate the Boltzmann transport properties of a material from such properties as the band-structure of a material. This method, as more accurate methods such as BoltzWann, are being increasingly used in research to calculate such quantities as conductivity, etc, of the material.

You can find all the great science behind this method right here: http://www.sciencedirect.com/science/article/pii/S0010465506001305. The DOI is 10.1016/j.cpc.2006.03.007 if that does not work.

I have started using it because we are looking at conductivity values of some of the materials we are studying at for solar properties. However, there is no clear documentation on how to run this with Quantum Espresso.

Installation

This is not a guide on how to install BoltzTraP, as that documentation is actually pretty straight forward. You can find the software here: BoltzTrap Website. The main thing I will mention is that it will generate two executables: x_trans and BoltzTraP. You can leave x_trans anywhere in your path, but it appears to need BoltzTraP to be placed in /usr/local/bin. I could not tell you why this is, but it was the only way I could get it installed.

 

Running with QE

Starting off, you will need to generate the following in order to successfully run BoltzTraP:

• Relaxed structure
• Dense K-Point nscf calculation

I will do this example with Silicon so that I can run it locally on my laptop. Here are the input files that I used:

relax.in:

&control
 calculation='relax'
 restart_mode='from_scratch'
 prefix='BoltzTraP_Tutorial'
 pseudo_dir = '.'
/
&system
 ibrav= 2
 celldm(1) =10.187
 nat= 2
 ntyp= 1
 ecutwfc =16
/
&electrons
 conv_thr = 1.0d-10
/
 &ions
 ion_dynamics = 'bfgs'
 /
ATOMIC_SPECIES
Si 0.0 Si.pbe-n-rrkjus_psl.0.1.UPF
ATOMIC_POSITIONS {crystal}
Si 0.00 0.00 0.00
Si 0.25 0.25 0.25
K_POINTS AUTOMATIC
4 4 4 1 1 1

 

 

nscf.in:

&control
 calculation='nscf.out'
 restart_mode='from_scratch'
 prefix='BoltzTraP_Tutorial'
 pseudo_dir = '.'

verbosity='high'
/
&system
 ibrav= 2
 celldm(1) =10.187
 nat= 2
 ntyp= 1
 ecutwfc =16
/
&electrons
 conv_thr = 1.0d-10
/
 &ions
 ion_dynamics = 'bfgs'
 /
ATOMIC_SPECIES
Si 0.0 Si.pbe-n-rrkjus_psl.0.1.UPF
ATOMIC_POSITIONS {crystal}
Si 0.00 0.00 0.00
Si 0.25 0.25 0.25
K_POINTS AUTOMATIC
10 10 10 1 1 1

The three main takeaway here are in the nscf file. The first thing, and this is important: you need to set the verbosity to ‘high’. I will explain why in a second. The second is the weird name of my prefix. The prefix name must be the same as the directory that you are running in (in this case ‘BoltzTraP_Tutorial’). This is not 100% true 100% of the time, but if you are new to BoltzTraP, leave it this way to prevent confusion. The third is that your output from the nscf calculation must be PreFix.nscf.out. In this example it would be BoltzTraP_Tutorial.nscf.out.

Next, you need to use the code qe2boltz.py that is located in the util folder of the BoltzTrap compilation. The version of the .py file they give works 90% of the time, but there is a slight error if your system has no symmetry (like most of my systems do). I have updated the script, and you can find it here: http://levilentz.com/BoltzTraP/qe2boltz.py.

The basics of this program are as follows:

Converts the output of Quantum Espresso 5.0 or BerkeleyGW 1.0
 to the input of BoltzTraP 1.2.1. Written by Georgy Samsonidze,
 An Li, Daehyun Wee, Bosch Research (October 2011).

Usage: qe2boltz.py prefix format efermi nbnd_exclude [fn_pw [fn_energy]]

* prefix = name of the system, same as in espresso input file
 * format = pw | bands | inteqp (read energies from the output of
 espresso/PW/pw.x, espresso/PP/bands.x or BerkeleyGW/BSE/inteqp.flavor.x)
 * efermi = Fermi energy in eV (if abs(efermi) > 1e6 and format = pw | bands
 efermi is read from fn_inp)
 * nbnd_exclude = number of the lowest energy bands to exclude in the output
 * fn_pw = name of a file containing the output of pw.x (the default is
 prefix.nscf.out, must be run with verbosity = high and ibrav = 0)
 * fn_energy = name of a file containing the output of bands.x or
 inteqp.flavor.x (the default is bands.out or bandstructure.dat,
 not used if format = pw)

Creates files BoltzTraP.def, prefix.intrans, prefix.energy, prefix.struct.
 File names and parameters written to BoltzTraP.def and prefix.intrans
 are hard-coded into the script.

 

 

In this case I will run it as follows:

qe2boltz.py BoltzTraP_Tutorial pw 6.7320 0

This then gives you a bunch of files that BoltzTrap needs. The main one you will need to edit, if you choose, is the *.intrans file. The main recommendation I could give you is if you need to expand the energy level around the Fermi Energy (the efcut shown below). 

GENE # Format of DOS
0 0 0 0.0 # iskip (not presently used) idebug setgap shiftgap
0.494792895338 0.0005 0.4 8.0 # Fermilevel (Ry), energygrid, energy span around Fermilevel, number of electrons
CALC # CALC (calculate expansion coeff), NOCALC read from file
5 # lpfac, number of latt-points per k-point
BOLTZ # run mode (only BOLTZ is supported)
0.35 # (efcut) energy range (Ry) of chemical potential     <------This is what you need to edit
800.0 50.0 # Tmax, temperature grid
-1.0 # energyrange of bands given individual DOS output sig_xxx and dos_xxx (xxx is band number)
TETRA

You then run it in the simpliest way you could imagine:

$: x_trans BoltzTraP
 ================ BoltzTraP vs 1.2.5 =============
94.6u 0.1s 1:35.14 99.6% 0+0k 1112+30040io 4pf+0w

The output is for you to browse through for your own purposes. The *.trace file has the trace of the tenser that BoltzTraP calculates. This can be plotted as I have plotted the Conductivity at various temperatures at a chemical potential equal to the fermi level:

 

I have zipped up all the inputs and outputs of this tutorial (save the .save directory), and you can download it here: http://levilentz.com/BoltzTraP/BoltzTrap_Tutorial.zip 

Please feel free to contact me with any questions via email at levi lentz at gmail com