Example_SrO-Phonopy

Example: SrO Phonopy Calculation

In this example, we will perform a harmonic Phonopy calculation on SrO using a 2×2×2 supercell, with a set of successively larger finite-displacement steps, and use the OUTCARToGULP_ModeMap.py script to extract data from the OUTCAR files generated from the single-point force calculations.

Harmonic Phonopy Calculation

We assume users following this example have set up Phonopy and are familiar with the basics of using it. If not, you are encouraged to first look through the documentation and examples on the code website.

First, prepare a supercell finite-displacement calculation on the optimised SrO structure using Phonopy:

phonopy -c POSCAR.vasp -d --dim="2 2 2"

Perform single-point force calculations on the two displaced structures (POSCAR-001, POSCAR-002). Example INCAR and KPOINTS files are provided, and should be paired with the Sr_sv (PAW_PBE Sr_sv 07Sep2000) and O (PAW_PBE O 08Apr2002) PAW pseudopotentials from the VASP database. The output should be similar to the reference OUTCAR files in the Step-0.01 directory.

Next, collect the forces from the vasprun.xml files and create a FORCE_SETS file:

phonopy -f 001/vasprun.xml 002/vasprun.xml

Finally, perform some post processing with Phonopy to check the phonon density of states (DoS)...:

phonopy -c POSCAR.vasp -p -s Phonopy_PhononDoS.conf

... and also the phonon dispersion:

phonopy -c POSCAR.vasp -p -s Phonopy_PhononDispersion.conf

The result should look similar to the following:

Different displacement steps

To obtain more data sets for fitting, we will prepare a set of Phonopy calculations with displacement steps from 0.01 Å (the default) to 1 Å. The Phonopy_Prepare.sh script creates a subdirectory for each of a set of step sizes and calls Phonopy to prepare the displaced structures:

#!/bin/bash

for step in 0.01 0.05 0.10 0.50 1.00
do
    dir="Step-${step}"
    
    mkdir "${dir}"
    
    cd "${dir}"
    
    phonopy -d --dim="2 2 2" -c ../POSCAR.vasp --amplitude=${step}
    
    cd ..
done

After running, the script(s) will generate five directories named Step-0.01, Step-0.05, etc., each containing the basic setup for a harmonic Phonopy calculation.

We now need to run the five sets of single-point force calculations, which might be done with a suitably-modified version of the Phonopy_Run.sh script:

#!/bin/bash

for step in 0.01 0.05 0.10 0.50 1.00
do
    phonopyDir="Step-${step}"
    
    if [ -d "${phonopyDir}" ] ; then
        cd "${phonopyDir}"
        
        for poscarFile in `ls POSCAR-*`
        do
            runDir="${poscarFile/POSCAR-/}"
            
            if [ ! -d "${runDir}" ] ; then
                mkdir "${runDir}"
                
                cd "${runDir}"
                
                cp "../${poscarFile}" "POSCAR"
                
                for inputFile in "INCAR" "KPOINTS" "POTCAR"
                do
                    cp "../../${inputFile}" .
                done
                
                mpirun -np 16 vasp_std | tee "../${runDir}.out"
                
                cd ..
            fi
        done
        
        cd ..
    fi
done

Finally, for each displacement step, we need to collect the forces and perform the post processing as in the previous step (Phonopy_PostProcess.sh):

#!/bin/bash

for step in 0.01 0.05 0.10 0.50 1.00
do
    dir="Step-${step}"

    cd "${dir}"

    phonopy -f 001/vasprun.xml 002/vasprun.xml

    phonopy -c ../POSCAR.vasp --dim="2 2 2" --mesh="2 2 2" --gamma_center --fc_symmetry=1
    mv mesh.yaml mesh-2x2x2.yaml

    phonopy -c ../POSCAR.vasp -p -s ../Phonopy_PhononDoS.conf
    phonopy -c ../POSCAR.vasp -p -s ../Phonopy_PhononDispersion.conf

    cd ..
done

Comparing the DoS and dispersion curves calculated with the five different steps shows that whereas the smaller displacement steps produce similar results, the larger displacements lead to the atoms being pushed into the anharmonic part of the potential, which generally leads to higher frequencies:

This is also evident in comparing the phonon frequencies at the three symmetry-inequivalent phonon wavevectors (q-points) commensurate with the 2×2×2 supercell expansion (taken from the mesh-2x2x2.yaml files generated during the post processing):

	0.01 Å	0.05 Å	0.10 Å	0.50 Å	1.00 Å
Γν=4-6	6.172	6.187	6.236	7.456	8.895
Xν=1,2	3.735	3.734	3.731	3.463	1.432
Xν=3	5.653	5.660	5.682	6.028	4.950
Xν=4,5	7.027	7.039	7.080	8.206	10.033
Xν=6	9.249	9.254	9.269	9.861	11.622
Lν=1,2	3.166	3.171	3.187	3.599	4.236
Lν=3	6.852	6.854	6.859	7.000	7.338
Lν=4,5	8.223	8.231	8.258	8.958	9.782
Lν=6	12.702	12.707	12.723	13.176	13.814

In most cases, the larger displacement steps yield larger phonon frequencies, although the X-point acoustic modes (X_ν=1,2 and X_ν=3) appear to be interesting exceptions, as can also be seen in the dispersion curves.

Preparing GULP input files

We can use OUTCARToGULP_ModeMap.py to extract the total energies, gradients and strain derivatives from the pairs of OUTCAR files generated during each of the Phonopy calculations (Phonopy_OutputGULP.sh):

#!/bin/bash

for step in 0.01 0.05 0.10 0.50 1.00
do
    dir="Step-${step}"

    cd "${dir}"

    python OUTCARToGULP_ModeMap.py */OUTCAR -o "Step-${step}.gulp" -n "Phonopy (${step} A)" --add_commands

    cd ..
done

We use the -n command-line argument to have the displacement step added to the header comments above the data blocks, and we also set the --add_arguments flag to have some useful GULP commands included in the output files.

Finally, we might want to join the GULP input generated from each of the five Phonopy calculations into a single file, which can be done using the second half of Phonopy_OutputGULP.sh:

head -n 3 Step-0.01/Step-0.01.gulp > Phonopy.gulp

for step in 0.01 0.05 0.10 0.50 1.00
do
    tail -n +4 Step-${step}/Step-${step}.gulp | tail -r  | tail -n +4 | tail -r >> Phonopy.gulp

    # The version of `head` macOS 10.12.3 doesn't accept negative -n arguments; on Unix machines using versions that do, the following works and avoids the two calls to `tail -r`.
    # tail -n +4 Step-${step}/Step-${step}.gulp | head -n -3 >> Phonopy.gulp
done

tail -n 3 Step-0.01/Step-0.01.gulp >> Phonopy.gulp