Skip to content

Exercise 0: Perform Singlepoint Calculations

Warning

In the following exercises, computational settings including the reciprocal space grid (tag k_grid), the basis set, and supercell's size, have been chosen to allow for a rapid computation of the exercises in the limited time and within the CPU resources available during the tutorial session. Without loss of generality, these settings allow to demonstrate trends of the lattice dynamics of materials. In the production calculation, all computational parameters should be converged.

Estimated total CPU time: 3 min

In this exercise, you will learn:

  • How to set up input files of FHI-vibes.
  • How to use the command line interface of FHI-vibes.
  • How to perform single-point calculations with FHI-aims by utilizing FHI-vibes interface.

For this part we would use silicon (Si) structure. In the directory:

phonons-with-fhi-vibes/Tutorial/phonons/0_singlepoint/input
you will find the geometry.in file with a primitive unit cell of Fm\(\bar3\)m-Si. Additionally you would find a file aims.in, which contains the settings of the interface between FHI-vibes and FHI-aims.

Let's inspect the input file:

[files]
geometry:                      geometry.in

[calculator]
name:                          aims
socketio:                      true

[calculator.parameters]
xc:                            pw-lda
compute_forces:                true

[calculator.kpoints]
density:                       3

[calculator.basissets]
Si:                            light

The input settings are defined by the so-called sections ([files], [calculator]) with argument names on the left hand side, delimited by : or =, from the according values on the right hand side. The right hand side can be any JSON datatype1, but for us the most important are just plain strings, numbers, and arrays.

Section files

The section files contains a geometry reference to the structure we want to investigate. In our case this is file: geometry.in, which means that the input geometry is taken from geometry.in

Section calculator and calculator options

Next sections contain all the commands you would usually write to a FHI-aims control.in file, only that you delimit each keyword-value-pair by :. We are going to use FHI-aims as it is written in the calculator-name line.

Section calculator.parameters

These keywords correspond one-to-one to the FHI-aims keywords that are written to control.in. Keyword-only arguments like vdw_correction_hirshfeld should be given with the value true:

[calculator.parameters]
xc:                            pw-lda
vdw_correction_hirshfeld:      true
spin:                          collinear
use_gpu:                       true

Section calculator.kpoints

Instead of giving a k_grid explicitly, FHI-vibes can compute a density such that the density of k-points does not fall below this value in Å\(^{-3}\). This is optional, setting k_grid in calculator.parameters is equally valid. If the density is used the k-point value along each reciprocal lattice vector is computed in proportion to the length of this lattice vector.

In order to understand how ASE generates k_grid based on density variable, we could manually perform the calculations for primitive unit cell of Si.

Manual transformation of density into k_grid

The Si lattice vectors have the following values:

\[ a_x = [0.0, 2.74925169, 2.74925169] \]
\[ a_y = [2.74925169, 0.0, 2.74925169] \]
\[ a_z = [2.74925169, 2.74925169, 0.0] \]

The reciprocal lattice vectors would then be:

\[ b_x = [-0.18186767, 0.18186767, 0.18186767] \]
\[ b_y = [0.18186767, -0.18186767, 0.18186767] \]
\[ b_z = [0.18186767, 0.18186767, -0.18186767] \]

The number of kpoints along each reciprocal lattice vector (\(k_x\), \(k_y\), \(k_z\)) is proportional to the density variable (\(d\)) and lenght of this particular reciprocal lattice vector (\(b_x\), \(b_y\), \(b_z\)):

\[ k_x = 2 \cdot \pi \cdot |b_x| \cdot d \]
\[ k_y = 2 \cdot \pi \cdot |b_y| \cdot d \]
\[ k_z = 2 \cdot \pi \cdot |b_z| \cdot d \]

Note, that after the calculations using formulas above, the number of kpoints is by default rounded to the nearest even value.

Consequently, for primitive unit cell of Si and density = 3 (after rounding), we would obtain k_grid = \(6\times6\times6\).

Section calculator.basissets

Here you can name the type of basis set you would like to use for your computation. In our case, by default FHI-vibes, wrote light basis set, but for production calculations, the basis set needs to be converged by also trying intermediate and tight, if not really_tight. For the purpose of the tutorial, we are going to utilize light_spd basis set. Thus, please change light to light_spd.

[calculator.basissets]
Si:                            light_spd

Section calculator.socketio

Sets up socket communication via SocketIOCalculator. This has the potential to speed up calculations since a complete restart of FHI-aims after each completed SCF cycle is avoided. This feature is optional, but recommended to use when performing calculations for related structures, e.g., for phonon calculations as would be shown in next exercises.

The settings file, we just explored, contains all the necessary settings to set up single-point calculations with FHI-vibes using FHI-aims as the calculator. This file could be created by using FHI-vibes CLI functionality for templates generation:

vibes template calculator aims
Results of this command were used to create aims.in file.

Run the calculation

We can run the calculations using command line interface, by typing:

vibes run singlepoint aims.in | tee log.aims

Once the calculations are done the traditional input files for FHI-aims (control.in and geometry.in) as well as the output file aims.out can be found in aims/calculations/. Additionally, vibes produces a trajectory file aims/trajectory.son, which contains all information for postprocessing and is particularly useful for the adavanced tasks tackled in the next tutorials. To inspect the output using the CLI, type:

vibes output trajectory aims/trajectory.son

which would generate a trajectory.nc file2.

Submit calculations on a cluster

Note that it is possible to submit calculations to the queue on a computing cluster. To submit FHI-vibes calcuations to your cluster you have to set up calculations as you have done earlier on your laptop, and submit vibes run command to the queue. As an example for slurm system, you could use:

#!/bin/bash -l

#SBATCH -J md|vibes
#SBATCH -o log/md.%j
#SBATCH -e log/md.%j
#SBATCH --mail-type=all
#SBATCH --mail-user=your@mail.com
#SBATCH --nodes=1
#SBATCH --ntasks-per-node=32
#SBATCH --ntasks-per-core=1
#SBATCH -t 24:0:00

vibes run singlepoint aims.in
Alternatively, the defaults tasks (single-point calculations, geometry optimization, phonons calculations, and even molecular dynamics simulations) can be submitted by specifing slurm parameters as a section in the input file. In this case, you should add slurm section to the input:

[slurm]
name:         aims
tag:          vibes
mail_type:    none
mail_address: <userid>@rzg.mpg.de
nodes:        1
cores:        72
queue:        general
timeout:      5

To familiarize yourself with the functionality of FHI-vibes you should visit FHI-vibes documentation page. For the time being, you could progress with the next exercises of this tutorial. The solutions to each exercise are stored in the folder solutions for each exercise.

  1. see e.g. https://www.geeksforgeeks.org/json-data-types/ ↩

  2. In this case trajectory.nc is generated just to show the functionality of FHI-vibes, we are not going to extract any tangible information from it. ↩