4. Running ALAMODE

4.1. Program alm

Program alm estimates harmonic and anharmonic interatomic force constants (IFCs) based on the supercell approach.

  1. Perform usual SCF calculations for a primitive cell
Before performing phonon calculations, one needs to perform usual self-consistent field calculations and check the convergence with respect to the cutoff energy and the \(k\) point density. After that, please optimize the internal coordinate so that the atomic forces are negligibly small. Optimization of cell parameters may also be necessary, but please note that phonon properties are relatively sensitive to the cell parameters in polar materials such as perovskites.
  1. Decide the size of supercell
Next, please decide the size of a supercell. Here, one may use a conventional cell. When the primitive cell is fairly large (\(a \sim 10\) Å), one may proceed using the primitive cell.
  1. Prepare an input file for alm

Please make an input file for alm, say alm.in, which should contain &general, &interaction, &cutoff, and &position entries. For details of available input variables, please refer to here. Once the input file is properly prepared with MODE = suggest, necessary displacement patterns can be generated by executing alm as follows:

$ alm alm.in > alm.log

This produces the following files containing the pattern of atomic displacements.

  • PREFIX.HARMONIC_pattern
  • PREFIX.ANHARM?_pattern (If NORDER > 1)

In pattern files, all necessary displacement patterns are given in Cartesian coordinates.

Note

Pattern files just indicate the direction of displacements. The magnitude of displacements should be specified by each user. (\(\Delta u \sim 0.01\) Å is usual for calculating harmonic force constants.)

  1. Perform SCF calculations to generate displacement-force data set

Then, please prepare necessary input files for a DFT engine (or a classical force field engine) and calculate atomic forces for each displaced configuration. Once the atomic forces are calculated for all configurations, please collect the atomic displacements and atomic forces to separate files, say disp_all.dat and force_all.dat, in Rydberg atomic units. The detail of the file format is described on this page.

Note

We provide some auxiliary Python scripts to expedite the above procedure for VASP, Quantum ESPRESSO, and xTAPP users. The script files can be found in the tools/ directory. We are willing to support other software if necessary.

  1. Estimate IFCs by linear regression

In order to perform a fitting, please change the variable MODE of the input file alm.in to MODE = optimize. In addition please add the &optimize entry with appropriate DFSET. Then, IFCs can be estimated by executing

$ alm alm.in > alm.log

which makes the following two files in the working directory.

  • PREFIX.fcs : The list of force constants
  • PREFIX.xml : XML file containing necessary information for subsequent phonon calculations

4.2. Program anphon

  1. Prepare an input file for anphon

To perform phonon calculations and thermal conductivity calculations, one needs to prepare another input file, say anphon.in, for the program anphon.

If one wants to perform (harmonic) phonon calculations, one should write MODE = phonons in the &general entry of anphon.in. Please make sure that FCSXML variable is set to the XML file generated by alm.

If one wants to conduct thermal conductivity calculations instead of usual phonon calculations, please switch to MODE = RTA with appropriate FCSXML containing cubic IFCs.

For details of input variables of anphon, please refer to the list of input variables for anphon.

  1. Execute anphon

Phonon properties and lattice thermal conductivity can be calculated via executing

$ anphon anphon.in > anphon.log

or

$ mpirun -np NPROCS anphon anphon.in > anphon.log

Here, NPROCS is the number of MPI threads. If the code is compiled with the OpenMP option, the OpenMP parallelization can also be used by setting the OMP_NUM_THREADS variable as

$ export OMP_NUM_THREADS=16

The number 16 should be modified appropriately for your environment.

Note

MPI+OpenMP hybrid parallelization is available when calculating thermal conductivity with MODE = RTA, in which anharmonic self-energies of all \(N_{\boldsymbol{q},irred}\times N_{j}\) phonon modes need to be calculated. Here \(N_{\boldsymbol{q},irred}\) and \(N_{j}\) are the number of irreducible \(q\) points and the number of phonon branches, respectively. These phonon modes are distributed across NPROCS MPI threads, and phonon self-energies are calculated in parallel. OpenMP is used for the double loop over the \(N_{j}\) branches inside the calculation of each phonon self-energy. Therefore, a good performance is expected when OMP_NUM_THREADS is a divisor of \(N_{j}^{2}\).

When the calculation finishes normally, various files are generated in the working directory.

  • PREFIX.bands : Phonon dispersion along designated Brillouin zone paths
  • PREFIX.dos : (Atom projected) phonon DOS
  • PREFIX.thermo : Thermodynamic functions
  • PREFIX.msd : Mean-square displacement of atoms
  • …

The complete list of output files can be found here.

  1. Analyze the result

One can plot the phonon dispersion relation or phonon DOS using gnuplot. Alternatively, one can use a small script in the tools/ directory for visualizing these results. For example,

$ plotband.py target.bands

shows the phonon dispersion relation. Available command line options can be displayed by

$ plotband.py -h

We also provide a similar script for phonon DOS. Another script analyze_phonons.py may be useful to analyze the result of thermal conductivity calculations. For example, phonon lifetimes and mean-free-path at 300 K can be extracted by

$ analyze_phonons.py --calc tau --temp 300 target.result

It can also estimate a cumulative thermal conductivity by

$ analyze_phonons.py --calc cumulative --temp 300 --direction 1 target.result

For details, see the tutorial.