wannier90: Tutorial

Version 3.1

Contents

Preliminaries 3

Parallel execution 3

About this tutorial 4

Contact us 4

1: Gallium Arsenide – MLWFs for the valence bands 5

2: Lead – Wannier-interpolated Fermi surface 6

3: Silicon – Disentangled MLWFs 6

4: Copper – Fermi surface, orbital character of energy bands 8

Examples Using the pwscf Interface 10

5: Diamond – MLWFs for the valence bands 10

6: Copper – Fermi surface 11

7: Silane (SiH4 ) – Molecular MLWFs using Γ-point sampling 12

8: Iron – Spin-polarized WFs, DOS, projected WFs versus MLWFs 13

9: Cubic BaTiO3 15

10: Graphite 17

11: Silicon – Valence and low-lying conduction states 17

1
2 wannier90: Tutorial

12: Benzene – Valence and low-lying conduction states 19

13: (5,5) Carbon Nanotube – Transport properties 20

14: Linear Sodium Chain – Transport properties 21

15: (5,0) Carbon Nanotube – Transport properties 24

16: Silicon – Boltzmann transport 26

17: Iron – Spin-orbit-coupled bands and Fermi-surface contours 28

18: Iron – Berry curvature, anomalous Hall conductivity and optical conductivity 30

19: Iron – Orbital magnetization 34

20: Disentanglement restricted inside spherical regions of k space 36

21: Gallium Arsenide – Symmetry-adapted Wannier functions 37

22: Copper – Symmetry-adapted Wannier functions 38

23: Silicon – G0 W0 bands structure interpolation 39

24: Tellurium – gyrotropic effects 41

25: Gallium Arsenide – Nonlinear shift current 42

26: Gallium Arsenide – Selective localization and constrained centres 44

27: Silicon – Selected columns of density matrix algorithm for automated MLWFs 45

28: Diamond – plotting of MLWFs using Gaussian cube format and VESTA 48

29: Platinum – Spin Hall conductivity 49

30: Gallium Arsenide – Frequency-dependent spin Hall conductivity 53

31: Platinum – Selected columns of density matrix algorithm for spinor wavefunctions 55

32: Tungsten — SCDM parameters from projectability 56

wannier90: Tutorial 3

Preliminaries

Welcome to wannier90! The examples contained in this tutorial are designed to help you become
familiar with the procedure of generating, analysing and using maximally-localised Wannier functions
(MLWFs). As a first step, install wannier90 following the instructions in the README file of the
wannier90 distribution. For an introduction to the theory underlying MLWFs, you are encouraged
to refer to the brief overview given in the wannier90 User Guide [1], to the two seminal papers of
Refs. [2, 3], a recent review article [4] and to a paper [5] describing wannier90.
The following additional programs may be installed in order to visualise the output of wannier90 (they
are optional, not all of them are necessary)

• gnuplot is used to plot bandstructures. It is available for many operating systems and is often
installed by default on Unix/Linux distributions
http://www.gnuplot.info

• xmgrace may also be used to plot bandstructures.

http://plasma-gate.weizmann.ac.il/Grace

• XCrySDen is used to visualise crystal structures, MLWFs, and Fermi surfaces. It is available
for Unix/Linux, Windows (using cygwin), and OSX. To correctly display files from wannier90,
version 1.4 or later must be used.
http://www.xcrysden.org

• vmd can also be used to visualise crystal structures and MLWFs.

http://www.ks.uiuc.edu/Research/vmd

• python with the numpy and matplotlib modules is used in examples 17–19
http://www.python.org
http://www.numpy.org
http://matplotlib.org

Parallel execution

postw90.x and wannier90.x can be run in parallel to speed up the calculations, using the MPI libraries.
To enable the parallel version to be built, you must specify some flags in the make.inc file of wannier90
and postw90; for further information, please refer to the README.install file in the top directory of
the wannier90 distribution.
Then, to run e.g. with 8 processors, you typically need to run a command similar to postw90 as follows:

mpirun -np 8 postw90.x seedname

(the mpirun command and its flags may differ depending on the MPI libraries installed on your system:
refer to your MPI manual and/or to your system administrator for further information).
4 wannier90: Tutorial

About this tutorial

The first part of this tutorial comprises four examples taken from Refs. [2, 3]: gallium arsenide, lead,
silicon and copper. All of the wannier90 input files have been provided.
The second part of the tutorial covers the generation of wannier90 input files starting from a full
electronic structure calculation. We have provided input files for the pwscf interface (http://www.
quantum-espresso.org) to wannier90. Therefore, you will need to install and compile elements of
the quantum-espresso package, namely pw.x and pw2wannier90.x, in order to run these examples.
Please visit http://www.quantum-espresso.org to download the package, and for installation instruc-
tions. The tutorial examples work with pwscf v5.1.x and v6.0.x. The exception are the examples
on symmetry adapted Wannier functions which require v6.0.x together with the very latest version of
pw2wannier90.f90. This can be found in the directory pwscf/v6.0 in the wannier distribution. It
should be moved to PP/src in the pwscf distribution and compiled using make pp. Later versions
v6.x.x should have the most up-to-date version of pw2wannier90.f90 already included in the Quantum
ESPRESSO distribution.
There are interfaces to a number of other electronic structure codes including abinit (http://www.
abinit.org), fleur (http://www.flapw.de), OpenMX (http://www.openmx-square.org/), GPAW
(https://wiki.fysik.dtu.dk/gpaw/), VASP (http://www.vasp.at), and Wien2k (http://www.
wien2k.at)

Contact us

If you have any suggestions regarding ways in which this tutorial may be improved, then send us an
email.
For other questions, email the wannier90 forum at wannier@quantum-espresso.org. Note that first
you will need to register in order to post emails. Emails from non-registered users are deleted auto-
matically. You can register by following the links at
http://www.wannier.org/forum.html.
wannier90: Tutorial 5

1: Gallium Arsenide – MLWFs for the valence bands

• Outline: Obtain and plot MLWFs for the four valence bands of GaAs.

• Generation details: From pwscf, using norm-conserving pseudopotentials and a 2×2×2 k-point
grid. Starting guess: four bond-centred Gaussians.

• Directory: examples/example1/

• Input Files

– gaas.win The master input file

– gaas.mmn The overlap matrices M(k,b)
– gaas.amn Projection A(k) of the Bloch states onto a set of trial localised orbitals
– UNK00001.1 The Bloch states in the real space unit cell. For plotting only.

1. Run wannier90 to minimise the MLWFs spread

wannier90.x gaas

Inspect the output file gaas.wout. The total spread converges to its minimum value after just a
few iterations. Note that the geometric centre of each MLWF lies along a Ga-As bond, slightly
closer to As than Ga. Note also that the memory requirement for the minimisation of the spread
is very low as the MLWFs are defined at each k-point by just the 4×4 unitary matrices U(k) .

2. Plot the MLWFs by adding the following keywords to the input file gaas.win

wannier_plot = true

and re-running wannier90. To visualise the MLWFs we must represent them explicitly on a real
space grid (see Ref. [1]). As a consequence, plotting the MLWFs is slower and uses more memory
than the minimisation of the spread. The four files that are created (gaas_00001.xsf, etc.) can
be viewed using XCrySDen,1 e.g.,

xcrysden --xsf gaas_00001.xsf

For large systems, plotting the MLWFs may be time consuming and require a lot of memory.
Use the keyword wannier_plot_list to plot a subset of the MLWFs. E.g., to plot the 1st and
3rd MLWFs use

wannier_plot_list = 1 3

The MLWFs are plotted in a supercell of the unit cell. The size of this supercell is set through
the keyword wannier_plot_supercell. The default value is 2 (corresponding to a supercell with
eight times the unit cell volume). We recommend not using values great than 3 as the memory
and computational cost scales cubically with supercell size.
Plot the 3rd MLWFs in a supercell of size 3. Choose a low value for the isosurface (say 0.5). Can
you explain what you see?
Hint: For a finite k-point mesh, the MLWFs are in fact periodic and the period is related to the
spacing of the k-point mesh. For mesh with n divisions in the ith direction in the Brillouin zone,
the MLWFs “live” in a supercell n times the unit cell.
1
Once XCrySDen starts, click on Tools → Data Grid in order to specify an isosurface value to plot.
6 wannier90: Tutorial

2: Lead – Wannier-interpolated Fermi surface

• Outline: Obtain MLWFs for the four lowest states in lead. Use Wannier interpolation to plot the
Fermi surface.

• Generation Details: From pwscf, using norm-conserving pseudopotentials and a 4×4×4 k-point
grid. Starting guess: atom-centred sp3 hybrid orbitals

• Directory: examples/example2/

• Input Files

– lead.win The master input file

– lead.mmn The overlap matrices M(k,b)
– lead.amn Projection A(k) of the Bloch states onto a set of trial localised orbitals
– lead.eig The Bloch eigenvalues at each k-point. For interpolation only

The four lowest valence bands in lead are separated in energy from the higher conduction states (see
Fig. 1). The MLWFs of these states have partial occupancy. MLWFs describing only the occupied
states would be poorly localised.

1. Run wannier90 to minimise the MLWFs spread

wannier90.x lead

Inspect the output file lead.wout.

2. Use Wannier interpolation to generate the Fermi surface of lead. Rather than re-running the
whole calculation we can use the unitary transformations obtained in the first calculation and
restart from the plotting routine. Add the following keywords to the lead.win file:

restart = plot
fermi_energy = 5.2676
fermi_surface_plot = true

and re-run wannier90. The value of the Fermi energy (5.2676 eV) was obtained from the initial
first principles calculation. wannier90 calculates the band energies, through
interpolation, on a dense mesh of k-points in the Brillouin zone. The density of this grid is
controlled by the keyword fermi_surface_num_points. The default value is 50 (i.e., 503 points).
The Fermi surface file lead.bxsf can be viewed using XCrySDen, e.g.,

xcrysden --bxsf lead.bxsf

3: Silicon – Disentangled MLWFs

• Outline: Obtain disentangled MLWFs for the valence and low-lying conduction states of Si. Plot
the interpolated bandstructure
wannier90: Tutorial 7

15

10
Energy (eV)

−5

L G X K G

Figure 1: Bandstructure of lead showing the position of the Fermi level. Only the lowest four bands
are included in the calculation.

• Generation Details: From pwscf, using norm-conserving pseudopotentials and a 4×4×4 k-point
grid. Starting guess: atom-centred sp3 hybrid orbitals

• Directory: examples/example3/

• Input Files

– silicon.win The master input file

– silicon.mmn The overlap matrices M(k,b)
– silicon.amn Projection A(k) of the Bloch states onto a set of trial localised orbitals
– silicon.eig The Bloch eigenvalues at each k-point

The valence and lower conduction states can be represented by MLWFs with sp3 -like symmetry. The
lower conduction states are not separated from the higher states by an energy gap. In order to form
localised WF, we use the disentanglement procedure introduced in Ref. [3]. The position of the inner
and outer energy windows are shown in Fig. 2.

1. Run wannier90.

wannier90.x silicon

Inspect the output file silicon.wout. The minimisation of the spread occurs in a two-step
procedure [3]. First, we minimise ΩI – this is the extraction of the optimal subspace in the
disentanglement procedure. Then, we minimise ΩD + ΩOD .

2. Plot the energy bands by adding the following commands to the input file silicon.win

restart = plot
bands_plot = true
8 wannier90: Tutorial

and re-running wannier90. The files silicon_band.dat and silicon_band.gnu are created. To
plot the bandstructure using gnuplot

myshell> gnuplot
gnuplot> load ‘silicon_band.gnu’

The k-point path for the bandstructure interpolation is set in the kpoint_path block. Try
plotting along different paths.

15

10

Outer Window
5
Inner Window

−5

L G X K G

Figure 2: Bandstructure of silicon showing the position of the outer and inner energy windows.

4: Copper – Fermi surface, orbital character of energy bands

• Outline: Obtain MLWFs to describe the states around the Fermi-level in copper

• Generation Details: From pwscf, using ultrasoft pseudopotentials [6] and a 4×4×4 k-point grid.
Starting guess: five atom-centred d orbitals, and two s orbitals centred on one of each of the two
tetrahedral interstices.

• Directory: examples/example4/

• Input Files

– copper.win The master input file

– copper.mmn The overlap matrices M(k,b)
– copper.amn Projection A(k) of the Bloch states onto a set of trial localised orbitals
– copper.eig The Bloch eigenvalues at each k-point

1. Run wannier90 to minimise the MLWFs spread

wannier90.x copper
wannier90: Tutorial 9

Inspect the output file copper.wout.

2. Plot the Fermi surface, it should look familiar! The Fermi energy is at 12.2103 eV.

3. Plot the interpolated bandstructure. A suitable path in k-space is

begin kpoint_path
G 0.00 0.00 0.00 X 0.50 0.50 0.00
X 0.50 0.50 0.00 W 0.50 0.75 0.25
W 0.50 0.75 0.25 L 0.00 0.50 0.00
L 0.00 0.50 0.00 G 0.00 0.00 0.00
G 0.00 0.00 0.00 K 0.00 0.50 -0.50
end kpoint_path

4. Plot the contribution of the interstitial WF to the bandstructure. Add the following keyword to
copper.win

bands_plot_project = 6,7

The resulting file copper_band_proj.gnu can be opened with gnuplot. Red lines correspond to
a large contribution from the interstitial WF (blue line are a small contribution; ie a large d
contribution).

Investigate the effect of the outer and inner energy window on the interpolated bands.
40

35

30

25
Outer Window

20

15
Inner Window

10

G X W L G K

Figure 3: Bandstructure of copper showing the position of the outer and inner energy windows.
10 wannier90: Tutorial

Examples Using the pwscf Interface

The pwscf plane-wave, density-functional theory code, which is available as part of the quantum-
espresso distribution (http://www.quantum-espresso.org), is fully interfaced to wannier90 via the
pw2wannier90 post-processing code that is also available as part of quantum-espresso. The latest
version of pw2wannier90 is included as part of the wannier90 distribution. Please see the pwscf
directory for instructions on how to incorporate it into pwscf.
Note that both the PWSCF executable pw.x and pw2wannier90.x can be run in parallel, which for
large calculations can reduce the computation time very significantly. This requires compiling the
code in its parallel version, using the MPI libraries. Refer to the quantum-espresso package for the
documentation on how to do so. Note that, unless you specify wf_collect=.true. in your pw.x input
file, you must run pw2wannier90 with the same number of processors as pw.x.
Moreover we remind here that both the wannier90 executable and postw90.x can be run in parallel.
In this case any number of processors can be used, independently of the number used for pw.x and
pw2wannier90.x.

5: Diamond – MLWFs for the valence bands

• Outline: Obtain MLWFs for the valence bands of diamond

• Directory: examples/example5/

• Input Files

– diamond.scf The pwscf input file for ground state calculation

– diamond.nscf The pwscf input file to obtain Bloch states on a uniform grid
– diamond.pw2wan The input file for pw2wannier90
– diamond.win The wannier90 input file

1. Run pwscf to obtain the ground state of diamond

pw.x < diamond.scf > scf.out

2. Run pwscf to obtain the Bloch states on a uniform k-point grid

pw.x < diamond.nscf > nscf.out

3. Run wannier90 to generate a list of the required overlaps (written into the diamond.nnkp file).
wannier90.x -pp diamond

4. Run pw2wannier90 to compute the overlap between Bloch states and the projections for the
starting guess (written in the diamond.mmn and diamond.amn files).
pw2wannier90.x < diamond.pw2wan > pw2wan.out

5. Run wannier90 to compute the MLWFs.

wannier90.x diamond
wannier90: Tutorial 11

6: Copper – Fermi surface

• Outline: Obtain MLWFs to describe the states around the Fermi-level in copper

• Directory: examples/example6/

• Input Files

– copper.scf The pwscf input file for ground state calculation

– copper.nscf The pwscf input file to obtain Bloch states on a uniform grid
– copper.pw2wan Input file for pw2wannier90
– copper.win The wannier90 input file

1. Run pwscf to obtain the ground state of copper

pw.x < copper.scf > scf.out

2. Run pwscf to obtain the Bloch states on a uniform k-point grid

pw.x < copper.nscf > nscf.out

3. Run wannier90 to generate a list of the required overlaps (written into the copper.nnkp file).
wannier90.x -pp copper

4. Run pw2wannier90 to compute the overlap between Bloch states and the projections for the
starting guess (written in the copper.mmn and copper.amn files).
pw2wannier90.x < copper.pw2wan > pw2wan.out

5. Run wannier90 to compute the MLWFs.

wannier90.x copper

Inspect the output file copper.wout.

1. Use Wannier interpolation to obtain the Fermi surface of copper. Rather than re-running the
whole calculation we can use the unitary transformations obtained in the first calculation and
restart from the plotting routine. Add the following keywords to the copper.win file:

restart = plot
fermi_energy = [insert your value here]
fermi_surface_plot = true

and re-run wannier90. The value of the Fermi energy can be obtained from the initial first
principles calculation. wannier90 calculates the band energies, through Wannier interpolation,
on a dense mesh of k-points in the Brillouin zone. The density of this grid is controlled by
the keyword fermi_surface_num_points. The default value is 50 (i.e., 503 points). The Fermi
surface file copper.bxsf can be viewed using XCrySDen, e.g.,

xcrysden --bxsf copper.bxsf

2. Plot the interpolated bandstructure. A suitable path in k-space is

12 wannier90: Tutorial

begin kpoint_path
G 0.00 0.00 0.00 X 0.50 0.50 0.00
X 0.50 0.50 0.00 W 0.50 0.75 0.25
W 0.50 0.75 0.25 L 0.00 0.50 0.00
L 0.00 0.50 0.00 G 0.00 0.00 0.00
G 0.00 0.00 0.00 K 0.00 0.50 -0.50
end kpoint_path

Further ideas

• Compare the Wannier interpolated bandstructure with the full pwscf bandstructure. Ob-
tain MLWFs using a denser k-point grid. To plot the bandstructure you can use the pwscf
tool bands.x or the small FORTRAN program available at http://www.tcm.phy.cam.ac.uk/
~jry20/bands.html.

• Investigate the effects of the outer and inner energy windows on the interpolated bands.

• Instead of extracting a subspace of seven states, we could extract a nine dimensional space (i.e.,
with s, p and d character). Examine this case and compare the interpolated bandstructures.

7: Silane (SiH4 ) – Molecular MLWFs using Γ-point sampling

• Outline: Obtain MLWFs for the occupied states of molecular silane. Γ-point sampling

• Directory: examples/example7/

• Input Files

– silane.scf The pwscf input file for ground state calculation

– silane.nscf The pwscf input file to obtain Bloch states on a uniform grid
– silane.pw2wan Input file for pw2wannier90
– silane.win The wannier90 input file

1. Run pwscf to obtain the ground state of silane

pw.x < silane.scf > scf.out

2. Run pwscf to obtain the Bloch states on a uniform k-point grid

pw.x < silane.nscf > nscf.out

3. Run wannier90 to generate a list of the required overlaps (written into the silane.nnkp file).
wannier90.x -pp silane

4. Run pw2wannier90 to compute the overlap between Bloch states and the projections for the
starting guess (written in the silane.mmn and silane.amn files).
pw2wannier90.x < silane.pw2wan > pw2wan.out

5. Run wannier90 to compute the MLWFs.

wannier90.x silane
wannier90: Tutorial 13

8: Iron – Spin-polarized WFs, DOS, projected WFs versus MLWFs

• Outline: Generate both maximally-localized and projected Wannier functions for ferromagnetic
bcc Fe. Calculate the total and orbital-projected density of states by Wannier interpolation.

• Directory: examples/example8/

• Input Files

– iron.scf The pwscf input file for the spin-polarized ground state calculation
– iron.nscf The pwscf input file to obtain Bloch states on a uniform grid
– iron_{up,down}.pw2wan Input files for pw2wannier90
– iron_{up,down}.win Input files for wannier90 and postw90

• Note that in a spin-polarized calculation the spin-up and spin-down MLWFs are computed sep-
arately. (The more general case of spinor WFs will be treated in Example 17.)

1. Run pwscf to obtain the ferromagnetic ground state of bcc Fe

pw.x < iron.scf > scf.out

2. Run pwscf to obtain the Bloch states on a uniform k-point grid

pw.x < iron.nscf > nscf.out

3. Run wannier90 to generate a list of the required overlaps (written into the .nnkp files).
wannier90.x -pp iron_up
wannier90.x -pp iron_dn

4. Run pw2wannier90 to compute the overlap between Bloch states and the projections for the
starting guess (written in the .mmn and .amn files).
pw2wannier90.x < iron_up.pw2wan > pw2wan_up.out
pw2wannier90.x < iron_dn.pw2wan > pw2wan_dn.out

5. Run wannier90 to compute the MLWFs.

wannier90.x iron_up
wannier90.x iron_dn

Density of states

To compute the DOS using a 25 × 25 × 25 k-point grid add to the two .win files

dos = true
dos_kmesh = 25

run postw90,

postw90.x iron_up
postw90.x iron_dn

and plot the DOS with gnuplot,

14 wannier90: Tutorial

myshell> gnuplot
gnuplot> plot ‘iron_up_dos.dat’ u (-$2):($1-12.6256) w l,‘iron_dn_dos.dat’ u
2:($1-12.6256) w l

Energies are referred to the Fermi level (12.6256 eV, from scf.out). Note the exchange splitting
between the up-spin and down-spin DOS. Check the convergence by repeating the DOS calculations
with more k-points.

Projected versus maximally-localized Wannier functions

In the calculations above we chose s, p, and d-type trial orbitals in the .win files,

Fe:s;p;d

Let us analyze the evolution of the WFs during the gauge-selection step. Open one of the .wout files
and search for “Initial state”; those are the projected WFs. As expected they are atom-centred,
with spreads organized in three groups, 1+3+5: one s, three p, and five d. Now look at the final state
towards the end of the file. The Wannier spreads have re-organized in two groups, 6+3; moreover,
the six more diffuse WFs are off-centred: the initial atomic-like orbitals hybridized with one another,
becoming more localized in the process. It is instructive to visualize the final-state MLWFs using
XCrySDen, following Example 1. For more details, see Sec. IV.B of Ref. [7].
Let us plot the evolution of the spread functional Ω,

myshell> grep SPRD iron_up.wout > sprd_up

myshell> gnuplot
gnuplot> plot ‘sprd_up’ u 6 w l

Figure 4: Evolution of the Wannier spread Ω of the minority (spin-up) bands of bcc Fe during the
iterative minimization of Ω,
e starting from s, p, and d-type trial orbitals.

The first plateau corresponds to atom-centred WFs of separate s, p, and d character, and the sharp
drop signals the onset of the hybridization. With hindsight, we can redo steps 4 and 5 more efficiently
using trial orbitals with the same character as the final MLWFs,
wannier90: Tutorial 15

Fe:sp3d2;dxy;dxz,dyz

With this choice the minimization converges much more rapidly.

Any reasonable set of localized WFs spanning the states of interest can be used to compute physical
quantities (they are “gauge-invariant”). Let us recompute the DOS using, instead of MLWFs, the WFs
obtained by projecting onto s, p, and d-type trial orbitals, without further iterative minimization of
the spread functional. This can be done by setting

num_iter = 0

But note that we still need to do disentanglement! Recalculate the DOS to confirm that it is almost
identical to the one obtained earlier using the hybridized set of MLWFs. Visualize the projected WFs
using XCrySDen, to see that they retain the pure orbital character of the individual trial orbitals.

Orbital-projected DOS and exchange splitting

With projected WFs the total DOS can be separated into s, p and d contributions, in a similar way to
the orbital decomposition of the energy bands in Example 4.
In order to obtain the partial DOS projected onto the p-type WFs, add to the .win files

dos_project = 2,3,4

and re-run postw90. Plot the projected DOS for both up- and down-spin bands. Repeat for the s and
d projections.
Projected WFs can also be used to quantify more precisely the exchange splitting between majority
and minority states. Re-run wannier90 after setting dos=false and adding to the .win files

write_hr_diag = true

This instructs wannier90 to print in the output file the on-site energies h0n|H|0ni. The difference
between corresponding values in iron_up.wout and in iron_dn.wout gives the exchange splittings
for the individual orbitals. Compare their magnitudes with the splittings displayed by the orbital-
projected DOS plots. In agreement with the Stoner criterion, the largest exchange splittings occur for
the localized d-states, which contribute most of the density of states at the Fermi level.

9: Cubic BaTiO3

• Outline: Obtain MLWFs for a perovskite

• Directory: examples/example9/

• Input Files

– batio3.scf The pwscf input file for ground state calculation

– batio3.nscf The pwscf input file to obtain Bloch states on a uniform grid
16 wannier90: Tutorial

– batio3.pw2wan Input file for pw2wannier90

– batio3.win The wannier90 input file

To start with, we are going to obtain MLWFs for the oxygen 2p states. From the bandstructure [8],
these form an isolated group of bands. We use the wannier90 keyword exclude_bands to remove all
but the 2p bands from the calculation of the overlap and projection matrices (we don’t have to do this,
but it saves time).

1. Run pwscf to obtain the ground state of BaTiO3

pw.x < BaTiO3.scf > scf.out
2. Run pwscf to obtain the Bloch states on a uniform k-point grid
pw.x < BaTiO3.nscf > nscf.out
3. Run wannier90 to generate a list of the required overlaps (written into the BaTiO3.nnkp file).
wannier90.x -pp BaTiO3
4. Run pw2wannier90 to compute the overlap between Bloch states and the projections for the
starting guess (written in the BaTiO3.mmn and BaTiO3.amn files).
pw2wannier90.x < BaTiO3.pw2wan > pw2wan.out
5. Run wannier90 to compute the MLWFs.
wannier90.x BaTiO3

Inspect the output file BaTiO3.wout.

Plot the second MLWF, as described in Section 1, by adding the following keywords to the input file
BaTiO3.win

wannier_plot = true
restart = plot
wannier_plot_list = 2
wannier_plot_supercell = 3

and re-running wannier90. Visualise it using XCrySDen,

xcrysden --xsf BaTiO3_00002.xsf

We can now simulate the ferroelectric phase by displacing the Ti atom. Change its position to

Ti 0.505 0.5 0.5

and regenerate the MLWFs (i.e., compute the ground-state charge density and Bloch states using
pwscf, etc.) and look at the change in the second MLWF.

Further ideas

• Look at MLWFs for other groups of bands. What happens if you form MLWFs for the whole
valence manifold?
• Following Ref. [8], compute the Born effective charges from the change in Wannier centres under
an atomic displacement.
wannier90: Tutorial 17

10: Graphite

• Outline: Obtain MLWFs for graphite (AB, Bernal)

• Directory: examples/example10/

• Input Files

– graphite.scf The pwscf input file for ground state calculation

– graphite.nscf The pwscf input file to obtain Bloch states on a uniform grid
– graphite.pw2wan Input file for pw2wannier90
– graphite.win The wannier90 input file

1. Run pwscf to obtain the ground state of graphite

pw.x < graphite.scf > scf.out

2. Run pwscf to obtain the Bloch states on a uniform k-point grid

pw.x < graphite.nscf > nscf.out

3. Run wannier90 to generate a list of the required overlaps (written into the graphite.nnkp file).
wannier90.x -pp graphite

4. Run pw2wannier90 to compute the overlap between Bloch states and the projections for the
starting guess (written in the graphite.mmn and graphite.amn files).
pw2wannier90.x < graphite.pw2wan > pw2wan.out

5. Run wannier90 to compute the MLWFs.

wannier90.x graphite

11: Silicon – Valence and low-lying conduction states

Valence States

• Outline: Obtain MLWFs for the valence bands of silicon.

• Directory: examples/example11/

• Input Files

– silicon.scf The pwscf input file for ground state calculation

– silicon.nscf The pwscf input file to obtain Bloch states on a uniform grid
– silicon.pw2wan Input file for pw2wannier90
– silicon.win The wannier90 input file

1. Run pwscf to obtain the ground state of silicon

pw.x < silicon.scf > scf.out

2. Run pwscf to obtain the Bloch states on a uniform k-point grid. Note that we request the lower
4 (valence) bands
pw.x < silicon.nscf > nscf.out
18 wannier90: Tutorial

3. Run wannier90 to generate a list of the required overlaps (written into the silicon.nnkp file).
wannier90.x -pp silicon

4. Run pw2wannier90 to compute the overlap between Bloch states and the projections for the
starting guess (written in the silicon.mmn and silicon.amn files).
pw2wannier90.x < silicon.pw2wan > pw2wan.out

5. Run wannier90 to compute the MLWFs.

wannier90.x silicon

Inspect the output file silicon.wout. The total spread converges to its minimum value after just a
few iterations. Note that the geometric centre of each MLWF lies at the centre of the Si-Si bond. Note
also that the memory requirement for the minimisation of the spread is very low as the MLWFs are
defined by just the 4×4 unitary matrices U(k) .
Plot the MLWFs by adding the following keywords to the input file silicon.win

wannier_plot = true

and re-running wannier90. Visualise them using XCrySDen, e.g.,

xcrysden --xsf silicon_00001.xsf

Valence + Conduction States

• Outline: Obtain MLWFs for the valence and low-lying conduction-band states of Si. Plot the
interpolated bandstructure. Apply a scissors correction to the conduction bands.

• Input Files

– silicon.scf The pwscf input file for ground state calculation

– silicon.nscf The pwscf input file to obtain Bloch states on a uniform grid
– silicon.pw2wan Input file for pw2wannier90
– silicon.win The wannier90 input file

The valence and lower conduction states can be represented by MLWFs with sp3 -like symmetry. The
lower conduction states are not separated by an energy gap from the higher states. In order to form
localised WF we use the disentanglement procedure introduced in Ref. [3]. The position of the inner
and outer energy windows are shown in Fig. 2.

1. Modify the input file and run pwscf and wannier90.

Inspect the output file silicon.wout. The minimisation of the spread occurs in a two-step proce-
dure. First, we minimise ΩI – this is the extraction of the optimal subspace in the disentanglement
procedure. Then, we minimise ΩO + ΩOD .

2. Plot the bandstructure by adding the following commands to the input file silicon.win

restart = plot
bands_plot = true
wannier90: Tutorial 19

and re-running wannier90. The files silicon_band.dat and silicon_band.gnu are created. To
plot the bandstructure using gnuplot

myshell> gnuplot
gnuplot> load ‘silicon_band.gnu’

The k-point path for the bandstructure interpolation is set in the kpoint_path block. Try
plotting along different paths.

Further ideas

• Compare the Wannier-interpolated bandstructure with the full pwscf bandstructure. Recom-
pute the MLWFs using a finer k-point grid (e.g., 6×6×6 or 8×8×8) and note how the accuracy
of the interpolation increases [9].

• Compute four MLWFs spanning the low-lying conduction states (see Ref. [3]).

12: Benzene – Valence and low-lying conduction states

Valence States

• Outline: Obtain MLWFs for the valence states of benzene

• Directory: examples/example12/

• Input Files

– benzene.scf The pwscf input file for ground state calculation

– benzene.pw2wan Input file for pw2wannier90
– benzene.win The wannier90 input file

1. Run pwscf to obtain the ground state of benzene

pw.x < benzene.scf > scf.out

2. Run wannier90 to generate a list of the required overlaps (written into the benzene.nnkp file).
wannier90.x -pp benzene

3. Run pw2wannier90 to compute the overlap between Bloch states and the projections for the
starting guess (written in the benzene.mmn and benzene.amn files).
pw2wannier90.x < benzene.pw2wan > pw2wan.out

4. Run wannier90 to compute the MLWFs.

wannier90.x benzene

Inspect the output file benzene.wout. The total spread converges to its minimum value after just a
few iterations.
Plot the MLWFs by adding the following keywords to the input file benzene.win
20 wannier90: Tutorial

restart = plot
wannier_plot = true
wannier_plot_format = cube
wannier_plot_list = 2-4

and re-running wannier90. Visualise them using, e.g., XCrySDen.

Valence + Conduction States

• Outline: Obtain MLWFs for the valence and low-lying conduction states of benzene.

• Input Files

– benzene.scf The pwscf input file for ground state calculation

– benzene.nscf The pwscf input file to obtain Bloch states for the conduction states
– benzene.pw2wan Input file for pw2wannier90
– benzene.win The wannier90 input file

In order to form localised WF we use the disentanglement procedure. The position of the inner energy
window is set to lie in the energy gap; the outer energy window is set to 4.0 eV. Modify the input file
appropriately.

1. Run pwscf and wannier90.

Inspect the output file benzene.wout. The minimisation of the spread occurs in a two-step
procedure. First, we minimise ΩI . Then, we minimise ΩO + ΩOD .

2. Plot the MLWFs by adding the following commands to the input file benzene.win

restart = plot
wannier_plot = true
wannier_plot_format = cube
wannier_plot_list = 1,7,13

and re-running wannier90. Visualise them using, e.g., XCrySDen.

13: (5,5) Carbon Nanotube – Transport properties

• Outline: Obtain the bandstructure, quantum conductance and density of states of a metallic (5,5)
carbon nanotube

• Directory: examples/example13/

• Input Files

– cnt55.scf The pwscf input file for ground state calculation

– cnt55.nscf The pwscf input file to obtain Bloch states for the conduction states
– cnt55.pw2wan Input file for pw2wannier90
– cnt55.win The wannier90 input file
wannier90: Tutorial 21

In order to form localised WF that describe both the occupied and unoccupied π and π ∗ manifolds,
we use the disentanglement procedure to extract a smooth manifold of states that has dimension equal
to 2.5 times the number of carbon atoms per unit cell [10]. The positions of the energy windows are
shown in Fig. 5.
The part of the wannier90 input file that controls the transport part of the calculation looks like:

transport = true
transport_mode = bulk
one_dim_axis = z
dist_cutoff = 5.5
fermi_energy = -1.06
tran_win_min = -6.5
tran_win_max = 6.5
tran_energy_step = 0.01
dist_cutoff_mode = one_dim
translation_centre_frac = 0.0 0.0 0.0

Descriptions of these and other keywords related to the calculation of transport properties can be found
in the User Guide.

1. Run pwscf and wannier90.

Inspect the output file cnt55.wout. The minimisation of the spread occurs in a two-step proce-
dure. First, we minimise ΩI . Then, we minimise ΩO + ΩOD .

2. Note that the initial pz projections on the carbon atoms are oriented in the radial direction with
respect to the nanotube axis.

3. The interpolated bandstructure is written to cnt55_band.agr (since bands_plot_format = xmgr

in the input file).

4. The quantum conductance and density of states are written to the files cnt55_qc.dat and
cnt55_dos.dat, respectively. Note that this part of the calculation may take some time. You
can follow its progress by monitoring the output to these files. Use a package such as gnuplot
or xmgrace in order to visualise the data. You should get something that looks like Fig. 6.

14: Linear Sodium Chain – Transport properties

• Outline: Compare the quantum conductance of a periodic linear chain of Sodium atoms with that
of a defected chain

• Directories: examples/example14/periodic
examples/example14/defected

• Input Files

– Na_chain.scf The pwscf input file for ground state calculation

– Na_chain.nscf The pwscf input file to obtain Bloch states for the conduction states
22 wannier90: Tutorial

Eouter
6

Band Energy (eV) 4

2 Einner

-2

-4

-6

-8
G A
Figure 5: Bandstructure of (5,5) carbon nanotube showing the position of the outer and inner energy
windows.

– Na_chain.pw2wan Input file for pw2wannier90

– Na_chain.win The wannier90 input file

The periodic system contains two unit cells evenly distributed along the supercell. Transport cal-
culations are performed using transport_mode = bulk and so the resulting quantum conductance
represents that of an infinite periodic chain.
The part of the wannier90 input file that controls the transport part of the calculation looks like:

transport = true
transport_mode = bulk
tran_read_ht = false
one_dim_axis = x
fermi_energy = -2.7401
tran_win_min = -5.0
wannier90: Tutorial 23

0 4 8 12 16
6
5
4
3
2

Energy (eV)
1
0
-1
-2
-3
-4
-5
-6
2
Γ A G / (2e /h) DoS

Figure 6: Wannier interpolated bandstructure, quantum conductance and density of states of (5,5)
carbon nanotube. Note that the Fermi level has been shifted by 1.06eV with respect to Fig. 5.

tran_win_max = 5.0
tran_energy_step = 0.01
translation_centre_frac = 0.5 0.5 0.5
tran_num_bb = 2

The defected system uses a 13 atom supercell with the central atom position altered to break symmetry.
Setting transport_mode = lcr with tell wannier90 to treat the system as an infinite system with the
defect at its centre. The supercell is chosen so that is conforms to the 2c2 geometry (see User Guide
for details). Each principal layer is 2 atoms long so that the conductor region contains the defected
atom plus a single atom on either side.
The transport section of the input file contains these key differences:

transport_mode = lcr
tran_num_ll = 2
tran_num_cell_ll = 2

Descriptions of these and other keywords related to the calculation of transport properties can be found
in the User Guide.

1. Run pwscf and wannier90 for the periodic system.

2. Run pwscf and wannier90 for the defected system.

3. The quantum conductance is written to the files periodic/Na_chain_qc.dat and

defected/Na_chain_dos.dat, respectively. Compare the quantum conductance of the periodic
(bulk) calculation with the defected (LCR) calculation. Your plot should look like Fig. 7.
24 wannier90: Tutorial

G (2e /h)
2
0.5

0
-4 -2 0 2 4
Energy (eV)

Figure 7: Quantum conductance of periodic Sodium chain (black) compared to that of the defected
Sodium chain (red).

15: (5,0) Carbon Nanotube – Transport properties

Note that these systems require reasonably large-scale electronic structure calculations.

Bulk Transport properties

• Outline: Obtain the quantum conductance of a pristine single-walled carbon nanotube

• Directory: examples/example14/periodic

• Input Files

– cnt.scf The pwscf input file for ground state calculation

– cnt.nscf The pwscf input file to obtain Bloch states for the conduction states
– cnt.pw2wan Input file for pw2wannier90
– cnt.win The wannier90 input file

First we consider a single unit cell, with 10 k-points. With transport_mode = bulk we compute the
transport properties of a pristine, infinite, periodic (5,0) carbon nanotube. Later, we will compare the
quantum conductance of this system with a defected nanotube.

1. Run pwscf and wannier90.

2. The quantum conductance and density of states are written to the files cnt_qc.dat and cnt_dos.dat,
respectively.

LCR transport properties – Defected nanotube

• Outline: Use the automated LCR routine to investigate the effect of a single silicon atom in a
infinite (5,0) carbon nanotube.

• Directory: examples/example15/defected
wannier90: Tutorial 25

• Input Files

– cnt+si.scf The pwscf input file for ground state calculation

– cnt+si.nscf The pwscf input file to obtain Bloch states for the conduction states
– cnt+si.pw2wan Input file for pw2wannier90
– cnt+si.win The wannier90 input file

In this calculation an 11-atom supercell is used with a single silicon substitutional defect in the central
unit cell. The supercell is chosen so that is conforms to the 2c2 geometry (see User Guide for details)
with principal layers set to be two unit cells long.

1. Run pwscf and wannier90. Again these are large calculations, progress can be monitored by
viewing respective output files.

2. The quantum conductance is written to cnt+si_qc.dat. Compare the quantum conductance

with the periodic (bulk) calculation. Your plot should look like Fig. 8.

8
Quantum Conductance (2e /h)
2

0
-2 0 2 4
Energy (eV)

Figure 8: Quantum conductance of infinite pristine nanotube (black) compared to that of the infinite
nanotube with the substitutional silicon defect (red).

Further ideas

• Set write_hr = true in the bulk case. Consider the magnitude of Hamiltonian elements between
Wannier functions in increasingly distant unit cells. Are two unit cell principal layers really large
enough, or are significant errors introduced?

• Does one unit cell either side of the defected unit cell shield the disorder so that the leads are
ideal? Does the quantum conductance change if these ‘buffer’ regions are increased?
26 wannier90: Tutorial

16: Silicon – Boltzmann transport

• Outline: Obtain MLWFs for the valence and low-lying conduction states of Si. Calculate the elec-
trical conductivity, the Seebeck coefficient and the thermal conductivity in the constant relaxation
time approximation using the BoltzWann module.

If you want to use Quantum ESPRESSO

• Directory: examples/example16-withqe/

• Input Files

– Si.scf The pwscf input file for ground state calculation

– Si.nscf The pwscf input file to obtain Bloch states on a uniform grid
– Si.pw2wan Input file for pw2wannier90
– Si.win The wannier90 and postw90 input file

If you do not want to use Quantum ESPRESSO

• Directory: examples/example16-noqe/

• Input Files

– Si.win The wannier90 and postw90 input file

– Si.mmn The overlap matrices M(k,b)
– Si.amn Projection A(k) of the Bloch states onto a set of trial localised orbitals
– Si.eig The Bloch eigenvalues at each k-point. For interpolation only

Note the first five steps in the following are the same of Example 11, and are needed only if you want
to use the PWscf code of Quantum ESPRESSO. Otherwise, if you have already run Example 11 with
Quantum ESPRESSSO (in particular, the section “Valence + Conduction States”) you can start from
those files and continue from point 6, after having added the BoltzWann flags to the input file.
If instead you do not have Quantum ESPRESSO installed, or you do not want to use it, you can start
from step 5 using the files in the examples/example16-noqe/ folder.

1. Run pwscf to obtain the ground state of silicon

pw.x < Si.scf > scf.out

2. Run pwscf to obtain the Bloch states on a uniform k-point grid. Details on the disentanglement
procedure are discussed in Example 11.
pw.x < Si.nscf > nscf.out

3. Run wannier90 to generate a list of the required overlaps (written into the Si.nnkp file).
wannier90.x -pp Si

4. Run pw2wannier90 to compute the overlap between Bloch states and the projections for the
starting guess (written in the Si.mmn and Si.amn files).
pw2wannier90.x < Si.pw2wan > pw2wan.out
wannier90: Tutorial 27

5. Run wannier90 to compute the MLWFs.

wannier90.x Si
Inspect the output file Si.wout and check if the convergence was reached both in the disentan-
glement and in the wannierisation steps (as discussed in further detail in Example 11). You may
also want to plot the Wannier functions and the interpolated band structure.

6. Run postw90 to calculate the transport coefficients.

postw90.x Si (serial execution)
mpirun -np 8 postw90.x Si (example of parallel execution with 8 MPI processes)

Inspect the output file Si.wpout. It summarizes the main details of the calculation (more details can
be obtained by setting a larger value of the iprint flag). Check if no warnings are issued. Note that
if no special flags are passed to BoltzWann, it assumes that the ab-initio calculation did not include
magnetization effects, and thus it sets to 2 the number of electrons per state.
Note also that the value of the relaxation time τ = 10 fs in the example is set only as a representative
value; note also that only the electrical and thermal conductivity depend on τ , while the Seebeck
coefficient is independent of τ .
Using your favourite plotting program, plot the Si_boltzdos.dat file to inspect the DOS.
Using your favourite plotting program, plot columns 1 and 3 of the Si_seebeck.dat file to inspect the
Sxx component of the Seebeck coefficient as a function of the chemical potential µ, at T = 300 K.

Further ideas

• Change the interpolation to a 60 × 60 × 60 mesh and run again postw90 to check if the results
for the transport properties are converged.

• Change the Si.win input file so that it calculates the transport coefficients for temperatures from
300 to 700 K, with steps of 200 K. Rerun postw90 and verify that the increase in execution time
is neglibile (in fact, most of the time is spent to interpolate the band structure on the k mesh).
Plot the Seebeck coefficient for the three temperatures T = 300 K, T = 500 K and T = 700 K.
To do this, you have to filter the Si_seebeck.dat to select only those lines where the second
column is equal to the required temperature. A possible script to select the Sxx component of the
Seebeck coefficient for T = 500 K using the awk/gawk command line program is the following:

awk ‘{if ($2 == 500) {print $1, $3;}}’ < Si_seebeck.dat \

> Si_seebeck_xx_500K.dat

Then, you can plot columns 1 and 2 of the output file Si_seebeck_xx_500K.dat.

• Try to calculate the Seebeck coefficient as a function of the temperature, for a n−doped sample
with, e.g., n = 1018 cm−3 . Note that to this aim, you need to calculate consistently the value
µ(T ) of the chemical potential as a function of the temperature, so as to reproduce the given value
of n. Then, you have to write a small program/script to interpolate the output of BoltzWann,
that you should have run on a suitable grid of (µ, T ) points.
28 wannier90: Tutorial

17: Iron – Spin-orbit-coupled bands and Fermi-surface contours

Note: It is recommended that you go through Example 8 first (bcc Fe without spin-orbit).
Note: This example requires a recent version of the pw2wannier90 interface.

• Outline: Plot the spin-orbit-coupled bands of ferromagnetic bcc Fe. Plot the Fermi-surface con-
tours on a plane in the Brillouin zone.

• Directory: examples/example17/

• Input files

– Fe.scf The pwscf input file for ground state calculation

– Fe.nscf The pwscf input file to obtain Bloch states on a uniform grid
– Fe.pw2wan The input file for pw2wannier90
– Fe.win The wannier90 and postw90 input file

Note that num_wann =18 in Fe.win, but only nine trial orbitals are provided. The line

spinors = true

tells wannier90 to use in step 3 below the specified trial orbitals on both the up- and down-spin
channels, effectively doubling their number.

1. Run pwscf to obtain the ferromagnetic ground state of iron2

pw.x < Fe.scf > scf.out

2. Run pwscf to obtain the Bloch states on a uniform k-point grid

pw.x < Fe.nscf > nscf.out

3. Run wannier90 to generate a list of the required overlaps (written into the Fe.nnkp file)
wannier90.x -pp Fe

4. Run pw2wannier90 to compute:

– The overlaps hunk |umk+b i between spinor Bloch states (written in the Fe.mmn file)
– The projections for the starting guess (written in the Fe.amn file)
– The spin matrix elements hψnk |σi |ψmk i, i = x, y, z (written in the Fe.spn file)

pw2wannier90.x < Fe.pw2wan > pw2wan.out

5. Run wannier90 to compute the MLWFs.

wannier90.x Fe

6. Run postw90 to compute the energy eigenvalues and spin expectation values.
postw90.x Fe (serial execution)
mpirun -np 8 postw90.x Fe (example of parallel execution with 8 MPI processes)
2
Please note the following counterintuitive feature in pwscf: in order to obtain a ground state with magnetization
along the positive z-axis, one should use a negative value for the variable starting_magnetization.
wannier90: Tutorial 29

In this example we use the module kpath to plot the energy bands coloured by the expectation value
of the spin along [001]:

kpath = true
kpath_task = bands
kpath_bands_colour = spin
kpath_num_points=500

To plot the bands using gnuplot (version 4.2 or higher) issue

myshell> gnuplot
gnuplot> load ‘Fe-bands.gnu’

or, using python,

myshell> python Fe-bands.py

Next we plot the Fermi-surface contours on the (010) plane ky = 0, using the kslice module. Set
kpath = false and uncomment the following instructions in Fe.win,

kslice = true
kslice_task = fermi_lines
fermi_energy = [insert your value here]
kslice_corner = 0.0 0.0 0.0
kslice_b1 = 0.5 -0.5 -0.5
kslice_b2 = 0.5 0.5 0.5
kslice_2dkmesh = 200 200

taking the Fermi level value from scf.out. The energy eigenvalues are computed on a 200 × 200
k-point grid covering the BZ slice. The lines of intersection between the Fermi surface and the (010)
plane can be visualized with the gnuplot or python scripts generated at runtime,

myshell> gnuplot
gnuplot> load ‘Fe-kslice-fermi_lines.gnu’

or

myshell> python Fe-kslice-fermi_lines.py

The Fermi lines can be colour-coded by the spin expectation value hSz i of the states on the Fermi
surface. Add to Fe.win the line

kslice_fermi_lines_colour = spin

and re-run postw90. The names of the gnuplot and python scripts generated at runtime are unchanged.
(However, the plotting algorithm is different in this case, and the lines are not as smooth as before.
You may want to increase kslice_2dkmesh.)
30 wannier90: Tutorial

Further ideas

• Redraw the Fermi surface contours on the (010) plane starting from a calculation without spin-
orbit coupling, by adding to the input files iron_{up,down}.win in Example 8 the lines

kslice = true
kslice_task = fermi_lines
fermi_energy = [insert your value here]
kslice_corner = 0.0 0.0 0.0
kslice_b1 = 0.5 -0.5 -0.5
kslice_b2 = 0.5 0.5 0.5
kslice_2dkmesh = 200 200

before running postw90,

postw90.x iron_up
postw90.x iron_dn

The python scripts generated at runtime draw the up- and down-spin Fermi lines on separate
figures. To draw them together, use the script iron_updn-kslice-fermi_lines.py provided
with Example 17 (or merge the two generated scripts). Compare the Fermi lines with and
without spin-orbit, and note the spin-orbit-induced avoided crossings.

• In Example 8 we obtained MLWFs separately for the up- and down-spin channels of bcc Fe
without spin-orbit. The Wannier-interpolated DOS was therefore automatically separated into
minority and majority contributions. For a spinor calculation we can still spin-decompose the
DOS, using

dos = true
spin_decomp = true
dos_kmesh = 25 25 25

The data file Fe-dos.dat created by postw90 contains the up-spin and down-spin contributions
in the third and fourth columns,

myshell> gnuplot
gnuplot> plot ’Fe-dos.dat’ u (-$3):($1-12.6285) w l,’Fe-dos.dat’ u ($4):($1-12.6285)
w l

(You should replace 12.6285 with your value of the Fermi energy). An alternative approach is
to project the DOS onto the up-spin and down-spin WFs separately. To find the DOS projected
onto the up-spin (odd-numbered) WFs replace spin_decomp = true with

dos_project = 1,3,5,7,9,11,13,15,17

and re-run postw90. This approach has the advantage that it does not require the Fe.spn file.

18: Iron – Berry curvature, anomalous Hall conductivity and optical

conductivity

Note: This example requires a recent version of the pw2wannier90 interface.

wannier90: Tutorial 31

• Outline: Calculate the Berry curvature, anomalous Hall conductivity, and (magneto)optical con-
ductivity of ferromagnetic bcc Fe with spin-orbit coupling. In preparation for this example it may
be useful to read Ref. [11] and Ch. 11 of the User Guide.

• Directory: examples/example18/

• Input files

– Fe.scf The pwscf input file for ground state calculation

– Fe.nscf The pwscf input file to obtain Bloch states on a uniform grid
– Fe.pw2wan The input file for pw2wannier90
– Fe.win The wannier90 and postw90 input file

The sequence of steps below is the same of Example 17. If you have already run that example, you can
reuse the output files from steps 1–5, and only step 6 must be carried out again using the new input
file Fe.win.

1. Run pwscf to obtain the ground state of iron

pw.x < Fe.scf > scf.out

2. Run pwscf to obtain the Bloch states on a uniform k-point grid

pw.x < Fe.nscf > nscf.out

3. Run wannier90 to generate a list of the required overlaps (written into the Fe.nnkp file)
wannier90.x -pp Fe

4. Run pw2wannier90 to compute the overlaps between Bloch states and the projections for the
starting guess (written in the Si.mmn and Si.amn files)
pw2wannier90.x < Fe.pw2wan > pw2wan.out

5. Run wannier90 to compute the MLWFs

wannier90.x Fe

6. Run postw90
postw90.x Fe (serial execution)
mpirun -np 8 postw90.x Fe (example of parallel execution with 8 MPI processes)

Berry curvature plots

The Berry curvature Ωαβ (k) of the occupied states is defined in Eq. (11.18) of the User Guide. The
following lines in Fe.win are used to calculate the energy bands and the Berry curvature (in bohr2 )
along high-symmetry lines in k-space.

fermi_energy = [insert your value here]

berry_curv_unit = bohr2
kpath = true
kpath_task = bands+curv
kpath_bands_colour = spin
kpath_num_points = 1000
32 wannier90: Tutorial

After executing postw90, plot the Berry curvature component Ωz (k) = Ωxy (k) along the magnetization
direction using the script generated at runtime,

myshell> python Fe-bands+curv_z.py

and compare with Fig. 2 of Ref. [11].

In Example 17 we plotted the Fermi lines on the (010) plane ky = 0. To combine them with a heatmap
plot of (minus) the Berry curvature set kpath = false, uncomment the following lines in Fe.win,

kslice = true
kslice_task = curv+fermi_lines
kslice_corner = 0.0 0.0 0.0
kslice_b1 = 0.5 -0.5 -0.5
kslice_b2 = 0.5 0.5 0.5
kslice_2dkmesh = 200 200

re-run postw90, and issue

myshell> python Fe-kslice-curv_z+fermi_lines.py

Compare with Fig. 3 in Ref. [11]. Note how the Berry curvature “hot-spots” tend to occur near
spin-orbit-induced avoided crossings (the Fermi lines with and without spin-orbit were generated in
Example 17).

Anomalous Hall conductivity

The intrinsic anomalous Hall conductivity (AHC) is proportional to the BZ integral of the Berry
curvature. In bcc Fe with the magnetization along ẑ, the only nonzero components are σxy = −σyx .
To evaluate the AHC using a 25×25×25 k-point mesh, set kslice = false, uncomment the following
lines in Fe.win,

berry = true
berry_task = ahc
berry_kmesh = 25 25 25

and re-run postw90. The AHC is written in the output file Fe.wpout in vector form. For bcc Fe with
the magnetization along [001], only the z-component σxy is nonzero.
As a result of the strong and rapid variations of the Berry curvature across the BZ, the AHC converges
rather slowly with k-point sampling, and a 25 × 25 × 25 does not yield a well-converged value.

– Increase the BZ mesh density by changing berry_kmesh.

– To accelerate the convergence, adaptively refine the mesh around spikes in the Berry curvature,
by adding to Fe.win the lines
wannier90: Tutorial 33

berry_curv_adpt_kmesh = 5
berry_curv_adpt_kmesh_thresh = 100.0

This adds a 5 × 5 × 5 fine mesh around those points where |Ω(k)| exceeds 100 bohr2 . The percentage
of points triggering adaptive refinement is reported in Fe.wpout.
Compare the converged AHC value with those obtained in Refs. [7] and [11].
The Wannier-interpolation formula for the Berry curvature comprises three terms, denoted D-D, D-A,
and Ω in Ref. [7], and J2, J1, and J0 in Ref. [12]. To report in Fe.wpout the decomposition of the
total AHC into these three terms, set iprint (verbosity level) to a value larger than one in Fe.win.

Optical conductivity

The optical conductivity tensor of bcc Fe with magnetization along ẑ has the form
   
σxx 0 0 0 σxy 0
σ = σ S + σ A =  0 σxx 0  +  −σxy 0 0 
0 0 σzz 0 0 0
where “S” and “A” stand for the symmetric and antisymmetric parts and σxx = σyy 6= σzz . The dc
AHC calculated earlier corresponds to σxy in the limit ω → 0. At finite frequency σxy = −σyx acquires
an imaginary part which describes magnetic circular dichoism (MCD).
To compute the complex optical conductivity for h̄ω up to 7 eV, replace

berry_task = ahc

with

berry_task = kubo

add the line

kubo_freq_max = 7.0

and re-run postw90. Reasonably converged spectra can be obtained with a 125 × 125 × 125 k-point
mesh. Let us first plot the ac AHC in S/cm, as in the lower panel of Fig. 5 in Ref. [11],

myshell> gnuplot
gnuplot> plot ‘Fe-kubo_A_xy.dat’ u 1:2 w l

Comapare the ω → 0 limit with the result obtained earlier by integrating the Berry curvature.3
Next we plot the MCD spectrum. Following Ref. [11], we plot Im[ωσxy (h̄ω)], in units of 1029 sec−2 .
The needed conversion factor is 9 × 10−18 × e/h̄ ' 0.0137 (e and h̄ in SI units),

gnuplot> set yrange[-5:15]

gnuplot> plot ‘Fe-kubo_A_xy.dat’ u 1:($1)*($3)*0.0137 w l
3
The calculation of the AHC using berry_task = kubo involves a truncation of the sum over empty states in the
Kubo-Greenwood formula: see description of the keyword kubo_eigval_max in the User Guide. As discussed around
Eq. (11.17) of the User Guide, no truncation is done with berry_task = ahc.
34 wannier90: Tutorial

Further ideas

• Recompute the AHC and optical spectra of bcc Fe using projected s, p, and d-type Wannier
functions instead of the hybridrized MLWFs (see Example 8), and compare the results.

• A crude way to model the influence of heterovalent alloying on the AHC is to assume that its
only effect is to donate or deplete electrons, i.e., to shift the Fermi level of the pure crystal [13].
Recalculate the AHC of bcc Fe for a range of Fermi energies within ±0.5 eV of the true Fermi
level. This calculation can be streamlined by replacing in Fe.win

fermi_energy = [insert your value here]

with

fermi_energy_min = [insert here your value minus 0.5]

fermi_energy_max = [insert here your value plus 0.5]

Use a sufficiently dense BZ mesh with adaptive refinement. To plot σxy versus εF , issue

myshell> gnuplot
gnuplot> plot ‘Fe-ahc-fermiscan.dat’ u 1:4 w lp

19: Iron – Orbital magnetization

Note: This example requires a recent version of the pw2wannier90 interface.

• Outline: Calculate the orbital magnetization of ferromagnetic bcc Fe by Wannier interpolation.

• Directory: examples/example19/

• Input files

– Fe.scf The pwscf input file for ground state calculation

– Fe.nscf The pwscf input file to obtain Bloch states on a uniform grid
– Fe.pw2wan The input file for pw2wannier90
– Fe.win The wannier90 and postw90 input file

The sequence of steps below is the same of Examples 17 and 18. If you have already run one of those
examples, you can reuse the output files from steps 1–3 and 5. Steps 4 and 6 should be carried out
again using the new input files Fe.pw2wan and Fe.win.

1. Run pwscf to obtain the ground state of iron

pw.x < Fe.scf > scf.out

2. Run pwscf to obtain the Bloch states on a uniform k-point grid

pw.x < Fe.nscf > nscf.out

3. Run wannier90 to generate a list of the required overlaps (written into the Fe.nnkp file).
wannier90.x -pp Fe
wannier90: Tutorial 35

4. Run pw2wannier90 to compute:

– The overlaps hunk |umk+b i (written in the Fe.mmn file)
– The projections for the starting guess (written in the Fe.amn file)
– The matrix elements hunk+b1 |Hk |umk+b2 i (written in the Fe.uHu file)
pw2wannier90.x < Fe.pw2wan > pw2wan.out
5. Run wannier90 to compute the MLWFs.
wannier90.x Fe
6. Run postw90 to compute the orbital magnetization.
postw90.x Fe (serial execution)
mpirun -np 8 postw90.x Fe (example of parallel execution with 8 MPI processes)

The orbital magnetization is computed as the BZ integral of the quantity Morb (k) defined in Eq. (11.20)
of the User Guide. The relevant lines in Fe.win are

berry = true
berry_task = morb
berry_kmesh = 25 25 25
fermi_energy = [insert your value here]

After running postw90, compare the value of the orbital magnetization reported in Fe.wpout with the
spin magnetization in scf.out. Set iprint = 2 to report the decomposition of Morb into the terms
J0, J1, and J2 defined in Ref. [12].
To plot Mzorb (k) along high-symmetry lines set berry = false and uncomment in Fe.win the block
of instructions containing

kpath = true
kpath_task = bands+morb

After running postw90, issue

myshell> python Fe-bands+morb_z.py

Compare with Fig. 2 of Ref. [12], bearing in mind the factor of −1/2 difference in the definition of
Morb (k) (see Ch. 11 in the User Guide).
To plot Mzorb (k) together with the Fermi contours on the (010) BZ plane set kpath = false, uncom-
ment in Fe.win the block of instructions containing

kslice = true
kslice_task = morb+fermi_lines

re-run postw90, and issue

myshell> python Fe-kslice-morb_z+fermi_lines.py

Mzorb (k) is much more evenly distributed in k-space than the Berry curvature (see Example 18). As a
result, the integrated orbital magnetization converges more rapidly with the BZ sampling.
36 wannier90: Tutorial

20: Disentanglement restricted inside spherical regions of k space

LaVO3

• Outline: Obtain disentangled MLWFs for strained LaVO3 .

• Directory: examples/example20/

• Input Files

– LaVO3.scf The pwscf input file for ground state calculation

– LaVO3.nscf The pwscf input file to obtain Bloch states on a uniform grid
– LaV03.pw2wan Input file for pw2wannier90
– LaVO3.win The wannier90 input file

1. Run pwscf to obtain the ground state of LaVO3 .

pw.x < LaVO3.scf > scf.out

2. Run pwscf to obtain the Bloch states on a uniform k-point grid.

pw.x < LaVO3.nscf > nscf.out

3. Run wannier90 to generate a list of the required overlaps (written into the LaVO3.nnkp file).
wannier90.x -pp LaVO3

4. Run pw2wannier90 to compute the overlap between Bloch states and the projections for the
starting guess (written in the LaVO3.mmn and LaVO3.amn files).
pw2wannier90.x < LaVO3.pw2wan > pw2wan.out

5. Run wannier90 to compute the MLWFs.

wannier90.x LaVO3

Inspect the output file LaVO3.wout. In the initial summary, you will see that the disentanglement was
performed only within one sphere of radius 0.2 arount the point A = (0.5, 0.5, 0.5) in reciprocal space:

| Number of spheres in k-space : 1 |

| center n. 1 : 0.500 0.500 0.500, radius = 0.200 |

Compare the band structure that Wannier90 produced with the one obtained using Quantum ESPRESSO.
You should get something similar to Fig. 9. Notice how the t2g -bands are entangled with other bands
at A and the Wannier-interpolated band structure deviates from the Bloch bands only in a small region
around that k-point. It is important to keep in mind that all symmetry equivalent k-points within the
first Brillouin zone must be written explicitly in the list of sphere centers. For instance, the A point
in the simple tetragonal lattice of this example is non-degenerate, while the X point has degeneracy
two, hence one must specify both (1/2, 0, 0) and (0, 1/2, 0) (see the SrMnO3 example here below).

Further ideas

• Try to obtain the Wannier functions using the standard disentanglement procedure (without
spheres, dis_spheres_num = 0). You will notice that the Wannier-interpolated band structure
wannier90: Tutorial 37

18

Energy (eV) 16

14

12

Γ M X Γ Z A R X

Figure 9: Band structure of epitaxially-strained (tetragonal) LaVO3 . Black: Bloch bands; red circles:
Wannier-interpolated band structure. The disentanglement was performed only for k-points within a
sphere of radius 0.2 Å−1 centered in A.

now shows deviations also in regions of k-space far away from A, where disentanglement is
actually not necessary. If you disable the disentanglement completely, instead, the Wannierisation
procedure does not converge.

• In order to illustrate all possible cases, it is instructive to apply this method to SrMnO3 , where
the t2g bands are entangled with the above-lying eg bands, and also with the deeper O-2p states.
In the SrMnO3 subfolder, you can find input files for building three different sets of Wannier
functions: only t2g states, only eg states, or all V-3d-derived states (t2g + eg ). In each case
one needs to specify different disentanglement spheres, according to which region(s) in k-space
show entanglement of the targeted bands. Also the index dis_sphere_first_wan needs to be
adapted to the new disentanglement window, which here contains also states below the lowest-
lying Wannier function (at variance with the LaVO3 case).

21: Gallium Arsenide – Symmetry-adapted Wannier functions

Note: This example requires a recent version of the pw2wannier90 interface.

• Outline: Obtain symmetry-adapted Wannier functions out of four valence bands of GaAs. For
the theoretical background of the symmetry-adapted Wannier functions, see R. Sakuma, Phys.
Rev. B 87, 235109 (2013).

• Directory: examples/example21/atom_centered_As_sp/
examples/example21/atom_centered_Ga_p/
examples/example21/atom_centered_Ga_s/
examples/example21/atom_centered_Ga_sp/
examples/example21/bond_centered/

• Input Files

– GaAs.scf The pwscf input file for ground state calculation

38 wannier90: Tutorial

– GaAs.nscf The pwscf input file to obtain Bloch states on a uniform grid
– GaAs.pw2wan The input file for pw2wannier90
– GaAs.win The wannier90 input file

1. Run pwscf to obtain the ground state of GaAs

pw.x < GaAs.scf > scf.out

2. Run pwscf to obtain the Bloch states on a uniform k-point grid

pw.x < GaAs.nscf > nscf.out

3. Run wannier90 to generate a list of the required overlaps (written into the GaAs.nnkp file).
wannier90.x -pp GaAs

4. Run pw2wannier90 to compute the overlap between Bloch states, the projections for the start-
ing guess, and the symmetry information needed for symmetry-adapted mode (written in the
GaAs.mmn, GaAs.amn, and GaAs.dmn files, respectively).
pw2wannier90.x < GaAs.pw2wan > pw2wan.out

5. Run wannier90 to compute the MLWFs.

wannier90.x GaAs

Each directory creates different kind of symmetry-adapted Wannier function. See more detail in
examples/example21/README.

22: Copper – Symmetry-adapted Wannier functions

Note: This example requires a recent version of the pw2wannier90 interface.

• Outline: Obtain symmetry-adapted Wannier functions for Cu. By symmetry-adapted mode, for
example, we can make atomic centered s-like Wannier function, which is not possible in the usual
procedure to create maximally localized Wannier functions. For the theoretical background of the
symmetry-adapted Wannier functions, see R. Sakuma, Phys. Rev. B 87, 235109 (2013).

• Directory: examples/example22/s_at_0.00/
examples/example22/s_at_0.25/
examples/example22/s_at_0.50/

• Input Files

– Cu.scf The pwscf input file for ground state calculation

– Cu.nscf The pwscf input file to obtain Bloch states on a uniform grid
– Cu.pw2wan The input file for pw2wannier90
– Cu.sym Used only in examples/example22/s_at_0.25/. pw2wannier90 reads this file when
“read_sym = .true.” in Cu.pw2wan. By default, “read_sym = .false.” and Cu.sym
is the output of pw2wannier90, in which the symmetry operations employed in the cal-
culation are written for reference.
– Cu.win The wannier90 input file
wannier90: Tutorial 39

1. Run pwscf to obtain the ground state of Cu

pw.x < Cu.scf > scf.out

2. Run pwscf to obtain the Bloch states on a uniform k-point grid

pw.x < Cu.nscf > nscf.out

3. Run wannier90 to generate a list of the required overlaps (written into the Cu.nnkp file).
wannier90.x -pp Cu

4. Run pw2wannier90 to compute the overlap between Bloch states, the projections for the starting
guess, and the symmetry information needed for symmetry-adapted mode (written in the Cu.mmn,
Cu.amn, and Cu.dmn files, respectively).
pw2wannier90.x < Cu.pw2wan > pw2wan.out

5. Run wannier90 to compute the MLWFs.

wannier90.x Cu

Each directory creates s-like symmetry-adapted Wannier function centered at different position on top
of atomic centered d-like Wannier functions. See more detail in examples/example22/README.

23: Silicon – G0 W0 bands structure interpolation

Note: This example requires a recent version of the ypp post-processing code of yambo.

• Outline: Interpolate the bands structure of silicon obtained from many-body perturbation theory
at the G0 W0 level. Using the yambo code, the quasi-particle corrections (QP) are summed to
Kohn-Sham eigenvalues, while the wavefunctions remain the same.

• Directory: examples/example23/

• Input Files

– silicon.scf The pwscf input file for the ground state calculation
– silicon.nscf The pwscf input file to obtain Bloch states on a uniform grid
– silicon.gw.nscf The pwscf input file to obtain Bloch states on a reduced grid with many
empty bands
– silicon.pw2wan The input file for pw2wannier90
– silicon.win The wannier90 input file
– silicon.gw.win The wannier90 input file (for the G0 W0 step)
– yambo.in The yambo input file
– ypp.in The ypp input file

1. Copy the input files from the INPUT directory into a working directory (e.g. WORK)

2. Run pwscf to obtain the ground state charge of silicon

pw.x < silicon.scf > scf.out
40 wannier90: Tutorial

3. Run pwscf to obtain the Bloch states reduced grid. We use a 8x8x8 with many bands (many
empty bands are needed to perform a G0 W0 with yambo)
pw.x < silicon.gw.nscf > nscf.gw.out

4. Use the k_mapper.py utility to find the indexes of a 4x4x4 uniform grid into the 8x8x8 reduced
grid
./k_mapper.py 4 4 4 "../examples/example23/WORK/nscf.gw.out"
Use the output to complete the yambo.in input file (you also need to specify the on how many
bands you want to compute the QP corrections, here you can use all the bands from 1 to 14).
Then, you should have obtained something like:
1| 1| 1|14|
3| 3| 1|14|
5| 5| 1|14|
13| 13| 1|14|
...

5. Enter the si.save directory and run p2y. A SAVE folder is created, you can move it up in the
/WORK/ directory.

6. Run a G0 W0 calculation from the /WORK/ directory (remember, we are using a 8x8x8 grid but
computing QP corrections only on a 4x4x4 grid)
yambo

7. Run pwscf to obtain the Bloch states on a uniform k-point grid

pw.x < silicon.nscf > nscf.out

8. Run wannier90 to generate a list of the required overlaps (written into the silicon.nnkp file).
wannier90.x -pp silicon

9. Run pw2wannier90 to compute the overlap between Bloch states, the projections for the starting
guess (written in the silicon.mmn and silicon.amn respectively).
pw2wannier90.x < silicon.pw2wan > pw2wan.out

10. Run wannier90 to compute the MLWFs.

wannier90.x silicon
At this point, you should have obtained the interpolated valence bands for silicon at the DFT
level.

11. Run a ypp calculation (just type ypp)

You should obtain a file silicon.gw.unsorted.eig which contains the QP corrections on a
uniform 4x4x4 grid.

12. Run the gw2wannier90.py script to reorder, align and correct all matrices and files using the QP
corrections
../../../utility/gw2wannier90.py silicon mmn amn

13. Run wannier90 to compute the MLWFs.

wannier90.x silicon.gw
At this point, you should have obtained the interpolated valence bands for silicon at the G0 W0
level.
wannier90: Tutorial 41

After you completed the tutorial for the valence bands only, you can repeat the final steps to interpolate
also some conduction bands using disentanglement (the code is already present as comments in the
input files).

24: Tellurium – gyrotropic effects

• Outline: Calculate the gyrotropic effects in trigonal right-handed Te Similar to the calculations
of [14]

• Directory: examples/example24/

• Input files

– Te.scf The pwscf input file for ground state calculation

– Te.nscf The pwscf input file to obtain Bloch states on a uniform grid
– Te.pw2wan The input file for pw2wannier90
– Te.win The wannier90 input file

To make things easy, the example treats Te without spin-orbit

1. Run pwscf to obtain the ground state of tellurium

pw.x < Te.scf > scf.out

2. Run pwscf to obtain the Bloch states on a uniform 3x3x4 k-point grid
pw.x < Te.nscf > nscf.out

3. Run wannier90 to generate a list of the required overlaps (written into the Te.nnkp file).
wannier90.x -pp Te

4. Run pw2wannier90 to compute:

– The overlaps hunk |umk+b i (written in the Te.mmn file)

– The projections for the starting guess (written in the Te.amn file)
– The matrix elements hunk+b1 |Hk |umk+b2 i (written in the Te.uHu file)
– The spin matrix elements hψnk |σi |ψmk i (would be written in the Te.spn file, but only if
spin-orbit is included, which is not the case for the present example)

pw2wannier90.x < Te.pw2wan > pw2wan.out

5. Run wannier90 to compute the MLWFs.

wannier90.x Te

6. Add the following lines to the wannier90.win file:

gyrotropic=true
gyrotropic_task=-C-dos-D0-Dw-K
fermi_energy_step=0.0025
fermi_energy_min=5.8
fermi_energy_max=6.2
gyrotropic_freq_step=0.0025
42 wannier90: Tutorial

gyrotropic_freq_min=0.0
gyrotropic_freq_max=0.1
gyrotropic_smr_fixed_en_width=0.01
gyrotropic_smr_max_arg=5
gyrotropic_degen_thresh=0.001
gyrotropic_box_b1=0.2 0.0 0.0
gyrotropic_box_b2=0.0 0.2 0.0
gyrotropic_box_b3=0.0 0.0 0.2
gyrotropic_box_center=0.33333 0.33333 0.5
gyrotropic_kmesh=50 50 50

7. Run postw90
to compute the gyrotropic properties: tensors D, D,
e K, C (See the User Guide):.
postw90.x Te (serial execution)
mpirun -np 8 postw90.x Te (example of parallel execution with 8 MPI processes)

The integration in the k-space is limited to a small area around the H point. Thus it is valid only
for Fermi levels near the band gap. And one needs to multiply the results by 2, to account for the
H’ point. To integrate over the entire Brillouin zone, one needs to remove the gyrotropic_box_. . .
parameters

8. Now change the above lines to

gyrotropic=true
gyrotropic_task=-NOA
fermi_energy=5.95
gyrotropic_freq_step=0.0025
gyrotropic_freq_min=0.0
gyrotropic_freq_max=0.3
gyrotropic_smr_fixed_en_width=0.01
gyrotropic_smr_max_arg=5
gyrotropic_band_list=4-9
gyrotropic_kmesh=50 50 50

and compute the interband natural optical activity

postw90.x Te (serial execution)

mpirun -np 8 postw90.x Te (example of parallel execution with 8 MPI processes)

25: Gallium Arsenide – Nonlinear shift current

• Outline: Calculate the nonlinear shift current of inversion asymmetric fcc Gallium Arsenide. In
preparation for this example it may be useful to read Ref. [15]

• Directory: examples/example25/

• Input files:

– GaAs.scf The PWSCF input file for ground state calculation

wannier90: Tutorial 43

– GaAs.nscf The PWSCF input file to obtain Bloch states on a uniform grid
– GaAs.pw2wan The input file for pw2wannier90
– GaAs.win The wannier90 and postw90 input file

1. Run PWSCF to obtain the ground state of Gallium Arsenide

pw.x < GaAs.scf > scf.out
2. Run PWSCF to obtain the ground state of Gallium Arsenide
pw.x < GaAs.nscf > nscf.out
3. Run Wannier90 to generate a list of the required overlaps (written into the GaAs.nnkp file)
wannier90.x -pp GaAs
4. Run pw2wannier90 to compute:
– The overlaps hunk |unk+b i between spinor Bloch states (written in the GaAs.mmn file)
– The projections for the starting guess (written in the GaAs.amn file)
pw2wannier90.x < GaAs.pw2wan > pw2wan.out
5. Run wannier90 to compute MLWFs
wannier90.x GaAs
6. Run postw90 to compute nonlinear shift current
postw90.x GaAs (serial execution)
mpirun -np 8 postw90.x GaAs (example of parallel execution with 8 MPI processes)

Shift current σ abc

The shift current tensor of GaAs has only one independent component that is finite, namely σ xyz . For
its computation, set

berry = true
berry_task = sc

Like the linear optical conductivity, the shift current is a frequency-dependent quantity. The frequency
window and step is controlled by kubo_freq_min, kubo_freq_max and kubo_freq_step, as explained
in the users guide.
The shift current requires an integral over the Brillouin zone. The interpolated k-mesh is controlled
by berry_kmesh, which has been set to

berry_kmesh = 100 100 100

We also need to input the value of the Fermi level in eV:

fermi_energy = [insert your value here]

Due to the sum over intermediate states involved in the calculation of the shift current, one needs to
consider a small broadening parameter to avoid numerical problems due to possible degeneracies (see
parameter η in Eq. (36) of Ref. [15] and related discussion). This parameter is controlled by sc_eta.
44 wannier90: Tutorial

It is normally found that values between 0.01 eV and 0.1 eV yield an stable spectrum. The default
value is set to 0.04 eV.
Finally, sc_phase_conv controls the phase convention used for the Bloch sums. sc_phase_conv=1
uses the so-called tight-binding convention, whereby the Wannier centres are included into the phase,
while sc_phase_conv=2 leaves the Wannier centres out of the phase. These two possible conventions
are explained in Ref. [16]. Note that the overall shift-current spectrum does not depend on the chosen
convention, but the individual terms that compose it do.
On output, the program generates a set of 18 files named SEED-sc_***.dat, which correspond to the
different tensor components of the shift current (note that the 9 remaining components until totaling
3 × 3 × 3 = 27 can be obtained from the 18 outputed by taking into account that σ abc is symmetric
under b ↔ c index exchange). For plotting the only finite shift-current component of GaAs σ xyz (units
of A/V2 ) as in the upper panel of Fig. 3 in Ref. [15],

myshell> gnuplot
gnuplot> plot ’GaAs-sc_xyz.dat’ u 1:2 w l

26: Gallium Arsenide – Selective localization and constrained centres

• Outline: Application of the selectively localised Wannier function (SLWF) method to gallium
arsenide (GaAs), following the example in Ref. [17], which is essential reading for this tutorial
example.

• Directory: examples/example26/

• Input files:

– GaAs.scf The PWSCF input file for ground state calculation

– GaAs.nscf The PWSCF input file to obtain Bloch states on a uniform grid
– GaAs.pw2wan The input file for pw2wannier90
– GaAs.win The wannier90 and postw90 input file

1. Run PWSCF to obtain the ground state of Gallium Arsenide

pw.x < GaAs.scf > scf.out
2. Run PWSCF to obtain the ground state of Gallium Arsenide
pw.x < GaAs.nscf > nscf.out
3. Run Wannier90 to generate a list of the required overlaps (written into the GaAs.nnkp file)
wannier90.x -pp GaAs
4. Run pw2wannier90 to compute:
– The overlaps hunk |unk+b i between Bloch states (written in the GaAs.mmn file)
– The projections for the starting guess (written in the GaAs.amn file)
pw2wannier90.x < GaAs.pw2wan > pw2wan.out
5. Inspect the .win file.
– Make sure you understand the new keywords corresponding to the selective localisation
algorithm.
wannier90: Tutorial 45

– Run wannier90 to compute the SLWFs, in this case using one objective Wannier func-
tion.
wannier90.x GaAs

To constrain the centre of the SLWF you need to add slwf_constrain = true and
slwf_lambda = 1 to the input file and uncomment the slwf_centres block. This will add a
penalty functional to the total spread, which will try to constrain the centre of the SLWF to be
on the As atom (as explained in Ref. [17], particularly from Eq. 24 to Eq. 35).
Look at the value of the penalty functional, is this what you would expect at convergence? Does
the chosen value of the Lagrange multiplier slwf_lambda give a SLWF function centred on the
As atom?
Alternatively, you can modify the slwf_centres block to constrain the centre of the SLWF to
be on the Ga atom. Do you need a different value of slwf_lambda in this case to converge? Take
a look at the result in Vesta and explain what you see. Do these functions transform like the
identity under the action of the Td group?

27: Silicon – Selected columns of density matrix algorithm for auto-

mated MLWFs

Note: This example requires a recent version of the pw2wannier90.x post-processing code of Quantum
ESPRESSO (v6.4 or above).

• Outline: For bulk crystalline Silicon, generate the Amn matrices via the selected columns of
density matrix (SCDM) algorithm and the corresponding MLWFs for 1) Valence bands 2) Valence
bands and 4 low-lying conduction bands 3) Conduction bands only. To better understand the input
files and the results of these calculations, it is crucial that the Reader has familiarized with the
concepts and methods explained in Ref. [18]. More info on the keywords related to the SCDM
method may be found in the user_guide.

• Directory: examples/example27/

• Input Files: input_files, and in the three subfolders isolated, erfc and gaussian. The
input_files folder contains:

– si.scf The pwscf input file for the ground state calculation
– si_4bands.nscf The pwscf input file to obtain Bloch states on a uniform grid for 4 bands.
– si_12bands.nscf The pwscf input file to obtain Bloch states on a uniform grid for 12
bands.

• Whereas the three subfolders isolated, erfc and gaussian contain the si.win wannier90 in-
put files and si.pw2wan pw2wannier90 input files each corresponding to one of the scenarios
listed in the outline.

1 Valence bands: In this case we will compute 4 localized WFs corresponding to the 4 valence
bands of Silicon. These 4 bands constitute a manifold that is separated in energy from other
bands. In this case the columns of the density matrix are already localized in real space and no
extra parameter is required.
46 wannier90: Tutorial

1. Copy the input files si.scf and si_4bands.nscf from the input_files directory into the
isolated folder
2. Run pwscf to obtain the ground state charge of bulk Silicon.
pw.x < si.scf > scf.out
3. Run pwscf to obtain the Bloch states on a uniform k-point grid of 4x4x4 for 4 bands.
pw.x < si_4bands.nscf > nscf.out
4. Inspect the si.win input file and make sure that the auto_projections flag is set to .true..
Also, make sure that no projections block is present.
5. Run wannier90 to generate a list of the required overlaps and also info on the SCDM method
(written into the si.nnkp file).
wannier90.x -pp si
6. Inspect the si.nnkp file and make sure you find the auto_projections block and that no
projections have been written in the projections block.
7. Inspect the .pw2wan input file. You will find two new keywords, i.e. scdm_proj and
scdm_entanglement. The former, will instruct pw2wannier90.x to use the SCDM method
when generating the Amn matrix. The latter, defines which formula to adopt for the function
f (εnk ) (see [18] and point below).
8. Run pw2wannier90 to compute the overlap between Bloch states and the projections via
the SCDM method (written in the si.mmn and si.amn respectively).
pw2wannier90.x < si.pw2wan > pw2wan.out
9. Run wannier90 to compute the MLWFs.
wannier90.x si
At this point, you should have obtained 4 Wannier functions and the interpolated valence
bands for Silicon. Inspect the output file si.wout. In particular, look at the geometric
centres of each WF, do they lie at the centre of the Si-Si bond as for the MLWFs computed
from user-defined initial s-like projections (see Example11)? Plot these WFs using Vesta.
Do they show the σ character one would expect from chemical arguments?

2 Valence bands + conduction bands: In this case we will compute 8 localized WFs corresponding to
the 4 valence bands and 4 low-lying conduction bands. Here, we don’t have a separate manifold,
since the conduction bands are entangled with other high-energy bands and the columns of the
density matrix are not exponentially localized by construction. A modified density matrix is
required in this case[18], and it is defined as:
X
P (r, r0 ) = ∗
ψnk (r)f (εn,k )ψnk (r0 ),
n,k

where ψnk and εn,k are the energy eigestates and eigenvalues from the first-principle calculation
respectively. The function f (εn,k ) contains two free parameters µ and σ and is defined as a
complementary error function:
 
1 εn,k − µ
f (εn,k ) = erfc .
2 σ

1. Copy the input files si.scf and si_12bands.nscf from the input_files folder into the
erfc folder
2. Run pwscf to obtain the ground state charge of bulk Silicon.
pw.x < si.scf > scf.out
wannier90: Tutorial 47

3. Run pwscf to obtain the Bloch states on a uniform k-point grid of 4x4x4 for 12 bands this
time.
pw.x < si_12bands.nscf > nscf.out
4. Inspect the si.win input file and make sure that the auto_projections flag is set to .true..
Also, make sure that no projection block is present.
5. Run wannier90 to generate a list of the required overlaps and also info on the SCDM method
(written into the si.nnkp file).
wannier90.x -pp si
6. Inspect the si.nnkp file and make sure you find the auto_projections block and that no
projections have been written in the projections block.
7. Inspect the .pw2wan input file. You will find other two new keywords, i.e. scdm_mu and
scdm_sigma. These are the values in eV of µ and σ in f (εn,k ), respectively.
8. Run pw2wannier90 to compute the overlap between Bloch states and the projections via
the SCDM method (written in the si.mmn and si.amn respectively).
pw2wannier90.x < si.pw2wan > pw2wan.out
9. Run wannier90 to compute the MLWFs.
wannier90.x si
At this point, you should have obtained 8 localized Wannier functions and the interpolated
valence and conduction bands for Silicon. Again, compare the results for the geometric
centres and the individual spreads with the ones from Example11. Is the final value of total
spread bigger or smaller than the one from Example11? Look at the WFs with Vesta. Can
you explain what you see? Where do the major lobes of the sp3-like WFs point in this case?

3 Conduction bands only: In this case we will compute 4 localized WFs corresponding to the 4
low-lying conduction bands only. As for the previous point, we need to define a modified density
matrix[18]. Since we are only interested in a subset of the conduction states, within a bounded
energy region, a good choice for f (εn,k ) is:

(εn,k − µ)2
 
f (εn,k ) = exp − .
σ2
1. Copy the input files si.scf and si_12bands.nscf from the input_files directory into the
gaussian folder
2. Run pwscf to obtain the ground state charge of bulk Silicon.
pw.x < si.scf > scf.out
3. Run pwscf to obtain the Bloch states on a uniform k-point grid of 4x4x4 for 12 bands this
time.
pw.x < si_12bands.nscf > nscf.out
4. Inspect the si.win input file and make sure that the auto_projections flag is set to .true..
Also, make sure that no projections block is present.
5. Run wannier90 to generate a list of the required overlaps and also info on the SCDM method
(written into the si.nnkp file).
wannier90.x -pp si
6. Inspect the si.nnkp file and make sure you find the auto_projections block and that no
projections have been written in the projections block.
7. Run pw2wannier90 to compute the overlap between Bloch states, the projections for the
starting guess via the SCDM method (written in the si.mmn and si.amn respectively).
pw2wannier90.x < si.pw2wan > pw2wan.out
48 wannier90: Tutorial

8. Run wannier90 to compute the MLWFs.

wannier90.x si
At this point, you should have obtained 4 localized Wannier functions and the interpolated
conduction bands for Silicon. From chemical intuition, we would expect these functions to
be similar to anti-bonding orbitals of molecules with tetrahedral symmetry. Plot the WFs
and check if this is confirmed.

28: Diamond – plotting of MLWFs using Gaussian cube format and

VESTA

• Outline: Obtain MLWFs for the valence bands of diamond and output them in Gaussian cube
format

• Directory: examples/example28/ The input files for this examples are the same as the ones in
example05

• Input Files

– diamond.scf The pwscf input file for ground state calculation

– diamond.nscf The pwscf input file to obtain Bloch states on a uniform grid
– diamond.pw2wan The input file for pw2wannier90
– diamond.win The wannier90 input file

1. Run pwscf to obtain the ground state of diamond

pw.x < diamond.scf > scf.out

2. Run pwscf to obtain the Bloch states on a uniform k-point grid

pw.x < diamond.nscf > nscf.out

3. Run wannier90 to generate a list of the required overlaps (written into the diamond.nnkp file).
wannier90.x -pp diamond

4. Run pw2wannier90 to compute the overlap between Bloch states and the projections for the
starting guess (written in the diamond.mmn and diamond.amn files).
pw2wannier90.x < diamond.pw2wan > pw2wan.out

5. When the lattice vectors are non-orthogonal, not all the visualisation programs are capable to
plot volumetric data in the Gaussian cube format. One program that can read volumetric data
for these systems is VESTA. To instruct wannier90 to output the MLWFs data in Gaussian cube
format you need to add the following lines to the .win file

wannier_plot = .true.
wannier_plot_supercell = 3
wannier_plot_format = cube
wannier_plot_mode = crystal
wannier_plot_radius = 2.5
wannier_plot_scale = 1.0

Run wannier90 to compute the MLWFs and output them in the Gaussian cube file.
wannier90.x diamond
wannier90: Tutorial 49

6. Plot the first MLWF with VESTA vesta diamond_00001.cube

Extra: Instead of using wannier_plot_mode = crystal try to use the molecule mode as wannier_plot_mode
= molecule (see the user guide for the definition of this keyword). Add the following line to the .win
file:

restart = plot

and re-run wannier90. Use VESTA to plot the resulting MLWFs, do you see any difference from
the crystal mode case? Can you explain why? Try to change the size of the supercell from 3 to
5, do you expect the results to be different? (Hint: When using the Gaussian cube format the code
outputs the WF on a grid that is smaller than the super unit-cell. The size of the grid is specified by
wannier_plot_scale and wannier_plot_radius.)

29: Platinum – Spin Hall conductivity

• Outline: Calculate spin Hall conductivity (SHC) and plot Berry curvature-like term of fcc Pt
considering spin-orbit coupling. To gain a better understanding of this example, it is suggested
to read Ref. [19] for a detailed description of the theory and Ch. 12.5 of the User Guide.

• Directory: examples/example29/

• Input files

– Pt.scf The pwscf input file for ground state calculation

– Pt.nscf The pwscf input file to obtain Bloch states on a uniform grid
– Pt.pw2wan The input file for pw2wannier90
– Pt.win The wannier90 and postw90 input file

1. Run pwscf to obtain the ground state of platinum

pw.x < Pt.scf > scf.out

2. Run pwscf to obtain the Bloch states on a uniform k-point grid

pw.x < Pt.nscf > nscf.out

3. Run wannier90 to generate a list of the required overlaps (written into the Pt.nnkp file)
wannier90.x -pp Pt

4. Run pw2wannier90 to compute the overlaps between Bloch states and the projections for the
starting guess (written in the Pt.mmn and Pt.amn files)
pw2wannier90.x < Pt.pw2wan > pw2wan.out

5. Run wannier90 to compute the MLWFs

wannier90.x Pt

6. Run postw90
postw90.x Pt (serial execution)
mpirun -np 8 postw90.x Pt (example of parallel execution with 8 MPI processes)
50 wannier90: Tutorial

Spin Hall conductivity

spinγ
The intrinsic spin Hall conductivity σαβ is proportional to the BZ integral of the Berry curvature-like
term. To evaluate the SHC using a 25 × 25 × 25 k-point mesh, set the following lines in Pt.win,

berry = true
berry_task = shc
berry_kmesh = 25 25 25

When calculating SHC, adaptive smearing can be used by commenting the following two lines,

#kubo_adpt_smr = false
#kubo_smr_fixed_en_width = 1

Then set the Fermi energy εF to a specific value

fermi_energy = [insert your value here]

or invoke Fermi energy scan by setting

fermi_energy_min = [insert here your lower range]

fermi_energy_max = [insert here your upper range]
fermi_energy_step = [insert here your step]

and re-run postw90. The SHC is written in the output file Pt-shc-fermiscan.dat. If only fermi_energy
is set, the output file will contain SHC at this specific energy; if a list of Fermi energies are set, the
output file will contain SHC calculated at each energy point in the list: we call this the “Fermi energy
scan” of SHC.
spinz
To plot the Fermi energy scan of SHC σxy versus εF , issue

myshell> gnuplot
gnuplot> plot ‘Pt-shc-fermiscan.dat’ u 2:3 w lp

As a result of the strong and rapid variations of the Berry curvature-like term across the BZ, the SHC
converges rather slowly with k-point sampling, and a 25×25×25 kmesh does not yield a well-converged
value.

– Increase the kmesh density by changing berry_kmesh.

– To accelerate the convergence, adaptively refine the kmesh around spikes in the Berry curvature-
like term, by adding to Pt.win the lines

berry_curv_adpt_kmesh = 5
berry_curv_adpt_kmesh_thresh = 100.0

This adds a 5×5×5 fine mesh around those points where |Ωspinγ
αβ (k)| exceeds 100 berry_curv_unit.
The percentage of points triggering adaptive refinement is reported in Pt.wpout.
wannier90: Tutorial 51

Compare the converged SHC value with those obtained in Refs. [19] and [20].
Note some rough estimations of computation progress and time are reported in Pt.wpout (see the SHC
part of the Solution Booklet). These may be helpful if the computation time is very long.

Notes

• Since the Kubo formula of SHC involves unoccupied bands, we need to include some unoccupied
bands and construct more MLWF. Thus the following parameters should be increased accordingly:

dis_froz_max
dis_win_max
projections
spinz
• Normally we calculate the SHC σxy , i.e. α = x, β = y, γ = z. To calculate other components,
the following parameters can be set as 1, 2, 3

shc_alpha = [insert here the α direction]

shc_beta = [insert here the β direction]
shc_gamma = [insert here the γ direction]

with 1, 2, 3 standing for x, y, z respectively.

Berry curvature-like term plots

The band-projected Berry curvature-like term Ωspinγn,αβ (k) is defined in Eq. (12.22) of the User Guide. The
following lines in Pt.win are used to calculate the energy bands colored by the band-projected Berry
spinγ spinγ
curvature-like term Ωn,αβ (k) (in Å2 ), as well as the k-resolved Berry curvature-like term Ωαβ (k)
along high-symmetry lines in k-space, i.e. the kpath plot. First comment the line berry = true and
then set

kpath = true
kpath_task = bands+shc
kpath_bands_colour = shc
kpath_num_points = 400
kubo_adpt_smr = false
kubo_smr_fixed_en_width = 1
fermi_energy = [insert your value here]
berry_curv_unit = ang2

After executing postw90, four files are generated: Pt-bands.dat, Pt-path.kpt, Pt-shc.dat and
Pt-bands+shc.py. Then plot the band-projected Berry curvature-like term Ωspinγ
n,αβ (k) using the script
generated at runtime,

myshell> python Pt-bands+shc.py

52 wannier90: Tutorial

and compare with Fig. 2 of Ref. [19]. Note a large fixed smearing of 1 eV is used to recover the result
in Ref. [19]. You can adjust the kubo_smr_fixed_en_width as you like to draw a visually appealing
figure. A kpath plot of 0.05 eV smearing is shown in the Solution Booklet.
Besides, you can set kpath_task = shc to only draw k-resolved term Ωspinγαβ (k) (the lower panel of
the figure), or set kpath_task = bands and kpath_bands_colour = shc to only draw energy bands
colored by the band-projected term Ωspinγ
n,αβ (k) (the upper panel of the figure).

Similar to that of AHC, we can get a heatmap plot of the k-resolved Berry curvature-like term Ωspinγ
αβ (k),
i.e. the kslice plot. To move forward, set kpath = false and uncomment the following lines in
Pt.win,

kslice = true
kslice_task = shc+fermi_lines
kslice_corner = 0.0 0.0 0.0
kslice_b1 = 1.0 0.0 0.0
kslice_b2 = 0.3535533905932738 1.0606601717798214 0.00
kslice_2dkmesh = 200 200
√ √
Note the kslice_b2 is actually ( 42 , 3 4 2 , 0.0) which leads to a square slice in the BZ, making it easier
to plot in the generated python script. Re-run postw90, and issue

myshell> python Pt-kslice-shc+fermi_lines.py

Compare the generated figure with Fig. 3 in Ref. [19], or the Solution Booklet.

Notes

• Adaptive smearing depends on a uniform kmesh, so when running kpath and kslice plots
adaptive smearing should not be used. A fixed smearing is needed to avoid near zero number in
the denominator of the Kubo formula, Eq. (12.22) in the User Guide. To add a fixed smearing
of 0.05 eV, add the following keywords in the Pt.win,

kubo_adpt_smr = .false.
kubo_smr_fixed_en_width = 0.05

Input parameters for SHC

Finally, we provide a complete list of input parameters that can be used to control the SHC calculation,
including the calculation of alternating current (ac) SHC which will be introduced in the next tutorial.

• general controls for SHC

shc_freq_scan, shc_alpha, shc_beta, shc_gamma,

kubo_eigval_max, exclude_bands, berry_curv_unit
wannier90: Tutorial 53

• kmesh

berry_task, berry_kmesh,
berry_curv_adpt_kmesh, berry_curv_adpt_kmesh_thresh

• ac SHC

kubo_freq_min, kubo_freq_max, kubo_freq_step,

shc_bandshift, shc_bandshift_firstband, shc_bandshift_energyshift,
scissors_shift, num_valence_bands

• smearing

[kubo_]adpt_smr, [kubo_]adpt_smr_fac, [kubo_]adpt_smr_max,

[kubo_]smr_fixed_en_width

• Fermi energy

fermi_energy, fermi_energy_min, fermi_energy_max, fermi_energy_step

• kpath

kpath, kpath_task, kpath_num_points, kpath_bands_colour

• kslice

kslice, kslice_task, kslice_corner, kslice_b1, kslice_b2, kslice_2dkmesh,

kslice_fermi_level, kslice_fermi_lines_colour

Their meanings and usages can be found in Ch. 11.5 of the User Guide.

30: Gallium Arsenide – Frequency-dependent spin Hall conductivity

• Outline: Calculate the alternating current (ac) spin Hall conductivity of gallium arsenide consid-
ering spin-orbit coupling. To gain a better understanding of this example, it is suggested to read
Ref. [19] for a detailed description of the theory and Ch. 12.5 of the User Guide.

• Directory: examples/example30/

• Input files

– GaAs.scf The pwscf input file for ground state calculation

– GaAs.nscf The pwscf input file to obtain Bloch states on a uniform grid
– GaAs.pw2wan The input file for pw2wannier90
– GaAs.win The wannier90 and postw90 input file

1. Run pwscf to obtain the ground state of gallium arsenide

pw.x < GaAs.scf > scf.out
54 wannier90: Tutorial

2. Run pwscf to obtain the Bloch states on a uniform k-point grid

pw.x < GaAs.nscf > nscf.out

3. Run wannier90 to generate a list of the required overlaps (written into the GaAs.nnkp file)
wannier90.x -pp GaAs

4. Run pw2wannier90 to compute the overlaps between Bloch states and the projections for the
starting guess (written in the GaAs.mmn and GaAs.amn files)
pw2wannier90.x < GaAs.pw2wan > pw2wan.out

5. Run wannier90 to compute the MLWFs

wannier90.x GaAs

6. Run postw90
postw90.x GaAs (serial execution)
mpirun -np 8 postw90.x GaAs (example of parallel execution with 8 MPI processes)

ac spin Hall conductivity

The spin Hall conductivity is also dependent on the frequency ω in the Eq. (12.22) of the User Guide.
spinγ
The direct current (dc) SHC calculated in the previous example corresponds to σαβ in the limit
spinγ
ω → 0 and it is a real number. At finite frequency σαβ acquires an imaginary part.
To compute the ac spin Hall conductivity for h̄ω up to 8 eV, add the lines

shc_freq_scan = true
kubo_freq_min = 0.0
kubo_freq_max = 8.0
kubo_freq_step = 0.01

and re-run postw90. The file GaAs-shc-freqscan.dat contains the calculated ac SHC. Reasonably
converged spectra can be obtained with a 250 × 250 × 250 k-point mesh. To plot the ac SHC, issue
the following commands

myshell> gnuplot
gnuplot> plot ‘GaAs-shc-freqscan.dat’ u 2:3 w l title ‘Re’, ‘GaAs-shc-freqscan.dat’
u 2:4 w l title ‘Im’

and then compare the result with Fig. 4 in Ref. [19] or the Solution Booklet.

Notes

• When calculating ac SHC, adaptive smearing can be used by add the following keywords in the
GaAs.win,

kubo_adpt_smr = true
kubo_adpt_smr_fac = [insert here your smearing factor]
kubo_adpt_smr_max = [insert here your maximum smearing]
wannier90: Tutorial 55

• Adaptive kmesh refinement is not implemented for ac SHC calculation.

• The first 10 semi-core states are excluded from the calculation by using the following keywords

exclude_bands = 1-10

and in the case of GaAs disentanglement is not adopted so

num_bands = 16
num_wann = 16

• Since the band gap is often under estimated by LDA/GGA calculations, a scissors shift is applied
to recover the experimental band gap by using the following keywords

shc_bandshift = true
shc_bandshift_firstband = 9
shc_bandshift_energyshift = 1.117

or by

num_valence_bands = 8
scissors_shift = 1.117

31: Platinum – Selected columns of density matrix algorithm for spinor

wavefunctions

Note: This example requires a recent version of the pw2wannier90.x post-processing code of Quantum
ESPRESSO (v6.3 or above).

• Outline: For bulk crystalline platinum with spin-orbit coupling, generate the Amn matrices via
the selected columns of density matrix (SCDM) algorithm and the corresponding spinor-MLWFs.
To better understand the input files and the results of these calculations, it is crucial that the
Reader has familiarized with the concepts and methods explained in Ref. [18]. More info on the
keywords related to the SCDM method may be found in the user_guide.
This example focuses on the use of the SCDM method for spin-noncollinear systems. For the
overview of the use of SCDM method to spinless systems, please refer to example27.

• Directory: examples/example31/
The input files for this examples are similar to the ones in example 29, except that a coarser
k-point grid is used and that the keywords related to postw90.x are removed.

• Input Files:

– Pt.scf The pwscf input file for the ground state calculation
– Pt.nscf The pwscf input file to obtain Bloch states on a uniform grid
– Pt.pw2wan The input file for pw2wannier90 with keywords related to the SCDM method
– Pt.win The wannier90 input file
56 wannier90: Tutorial

We will compute 18 localized WFs. Since the band structure of platinum is metallic, the low-lying
bands are entangled with other high-energy bands, and the columns of the density matrix are not
exponentially localized by construction. Thus, we use a modified density matrix [18], with the function
f (εn,k ) defined as a complementary error function. Refer to example 27 for the definition of the
modified density matrix and the functional form of f (εn,k ).

1. Run pwscf to obtain the ground state of platinum

pw.x < Pt.scf > scf.out
2. Run pwscf to obtain the Bloch states on a uniform 7 × 7 × 7 k-point grid
pw.x < Pt.nscf > nscf.out
3. Inspect the Pt.win input file and make sure that the auto_projections flag is set to .true..
Also, make sure that no projection block is present.
4. Run wannier90 to generate a list of the required overlaps (written into the Pt.nnkp file)
wannier90.x -pp Pt
5. Inspect the Pt.nnkp file and make sure you find the auto_projections block and that no pro-
jections have been written in the projections block.
6. Inspect the Pt.pw2wan input file. You will find four SCDM-related keywords: scdm_proj,
scdm_entanglement, scdm_mu and scdm_sigma. In particular, the keyword scdm_proj will in-
struct pw2wannier90.x to use the SCDM method when generating the Amn matrix. The re-
maining three keywords defines the formula and parameters to define the function f (εnk ) (see
Ref. [18] and example 27).
7. Run pw2wannier90 to compute the overlap between Bloch states and the projections via the
SCDM method (written in the Pt.mmn and Pt.amn respectively).
pw2wannier90.x < Pt.pw2wan > pw2wan.out
8. Inspect the pw2wan.out output file. Compared to the spinless case, you will find the following
two additional lines.

Number of pivot points with spin up : 9

Number of pivot points with spin down: 9

These lines give information on the pivots obtained by the QR decomposition with column
pivoting (QRCP) in the SCDM algorithm. Each pivot determines a point in the real-space grid
and a spin state. The basis of the spin state is determined by the basis used in the electronic
structure code. In pwscf, the basis states are spin up and down states along the Cartesian
z-axis.
9. Run wannier90 to compute the MLWFs
wannier90.x Pt

32: Tungsten — SCDM parameters from projectability

• Outline: Compute the Wannier interpolated band structure of tungsten (W) using the SCDM
method to calculate the initial guess (see Example 27 for more details). The free parameters in
wannier90: Tutorial 57

the SCDM method, i.e., µ and σ, are obtained by fitting a complementary error function to the
projectabilities. The number of MLWFs is given by the number of pseudo-atomic orbitals (PAOs)
in the pseudopotential, 21 in this case. All the steps shown in this example have been automated
in the AiiDA[21] workflow that can be downloaded from the MaterialsCloud website[22].

• Directory: examples/example31/

• Input files

– W.scf The pwscf input file for ground state calculation

– W.nscf The pwscf input file to obtain Bloch states on a uniform grid
– W.pw2wan The input file for pw2wannier90
– W.proj The input file for projwfc
– generate_weights.sh The bash script to extract the projectabilities from the output of
projwfc
– W.win The wannier90 input file

1. Run pwscf to obtain the ground state of tungsten

pw.x -in W.scf > scf.out

2. Run pwscf to obtain the Bloch states on a 10 × 10 × 10 uniform k-point grid

pw.x -in W.nscf > nscf.out

3. Run wannier90 to generate a list of the required overlaps (written into the W.nnkp file)
wannier90.x -pp W

4. Run projwfc to compute the projectabilities of the Bloch states onto the Bloch sums obtained
from the PAOs in the pseudopotential
projwfc.x -in W.proj > proj.out

5. Run generate_weights to extract the projectabilitites from proj.out in a format suitable to be

read by Xmgrace or gnuplot
./generate_weights.sh

6. Plot the projectabilities and fit the data with the complementary error function
1 µ−
f (; µ, σ) = erfc(− ).
2 σ
We are going Xmgrace to plot the projectabilities and perform the fitting. Open Xmgrace
xmgrace
To Import the p_vs_e.dat file, click on Data from the top bar and then Import -> ASCII....
At this point a new window Grace: Read sets should pop up. Select p_vs_e.dat in the Files
section, click Ok at the bottom and close the window. You should now be able to see a quite
noisy function that is bounded between 1 and 0. You can modify the appearence of the plot by
clicking on Plot in the top bar and then Set appearance.... In the Main section of the pop-up
window change the symbol type from None to Circle. Change the line type from straight to
none, since the lines added by default by Xmgrace are not meaningful. For the fitting, go to
Data -> Transformations -> Non-linear curve fitting. In this window, select the source
from the Set box and in the Formula box insert the following
58 wannier90: Tutorial

y = 0.5 * erfc( ( x - A0 ) / A1 )

Select 2 as number of parameters, give 40 as initial condition for A0 and 7 for A1. Click Apply.
A new window should pop up with the stats of the fitting. In particular you should find a
Correlation coefficient of 0.96 and a value of 39.9756 for A0 and 6.6529 for A1. These
are the value of µf it and σf it we are going to use for the SCDM method. In particular,
µSCDM = µf it − 3σf it = 20.0169 eV and σSCDM = σf it = 6.6529 eV. The motivation for
this specific choice of µf it and σf it may be found in Ref. [23], where the authors also show vali-
dation of this approach on a dataset of 200 materials. You should now see the fitting function,
as well as the projectabilities, in the graph (see Fig. 10-(a)).

7. Open W.pw2wan and append the following lines

scdm_entanglement = 'erfc'
scdm_mu = 20.0169
scdm_proj = .true.
scdm_sigma = 6.6529
/

8. Run pw2wannier90 to compute the overlaps between Bloch states and the projections for the
starting guess (written in the W.mmn and W.amn files)
pw2wannier90.x -in W.pw2wan > pw2wan.out

9. Run wannier90 to obtain the interpolated bandstructure (see Fig. 10-(b)).

wannier90.x W
Please cite Ref. [23] in any publication employing the procedure outlined in this example to
obtain µ and σ.

References
[1] A. A. Mostofi, G. Pizzi, I. Souza, and J. R. Yates, User Guide to wannier90, available at
http://www.wannier.org/user_guide.html.

[2] N. Marzari and D. Vanderbilt, Phys. Rev. B 56, 12847 (1997).

[3] I. Souza, N. Marzari, and D. Vanderbilt, Phys. Rev. B 65, 035109 (2001).

[4] N. Marzari, A. A. Mostofi, J. R. Yates, I. Souza, and D. Vanderbilt, Rev. Mod. Phys. 84, 1419
(2012).

[5] A. A. Mostofi, J. R. Yates, Y.-S. Lee, I. Souza, D. Vanderbilt, and N. Marzari, Comput. Phys.
Commun. 178, 685 (2008).

[6] D. Vanderbilt, Phys. Rev. B 41, 7892 (1990).

[7] X. Wang, J. R. Yates, I. Souza, and D. Vanderbilt, Phys. Rev. B 74, 195118 (2006).

[8] N. Marzari and D. Vanderbilt, arXiv:9802210 (1998).

[9] J. R. Yates, X. Wang, D. Vanderbilt, and I. Souza, Phys. Rev. B 75, 195121 (2007).
wannier90: Tutorial 59

60

40

20

-20

-40
DFT
W90
Γ H N Γ

(a) (b)

Figure 10: a) Each blue dot represents the projectability as defined in Eq. (22) of Ref. [23] of the
state |nki as a function of the corresponding energy nk for tungsten. The yellow line shows the fitted
complementary error function. The vertical red line represents the value of σf it while the vertical green
line represents the optimal value of µSCDM , i.e. µSCDM = µf it − 3σf it . b) Band structure of tungsten
on the Γ-H-N-Γ path from DFT calculations (solid black) and Wannier interpolation using the SCDM
method to construct the initial guess (red dots).

[10] Y.-S. Lee, M. B. Nardelli, and N. Marzari, Phys. Rev. Lett. 95, 076804 (2005).

[11] Y. Yao, L. Kleinman, A. H. MacDonald, J. Sinova, T. Jungwirth, D.-S. Wang, E. Wang, and
Q. Niu, Phys. Rev. Lett. 92, 037204 (2004).

[12] M. G. Lopez, D. Vanderbilt, T. Thonhauser, and I. Souza, Phys. Rev. B 85, 014435 (2012).

[13] Y. Yao, Y. Liang, D. Xiao, Q. Niu, S.-Q. Shen, X. Dai, and Z. Fang, Phys. Rev. B 75, 020401
(2007).

[14] S. S. Tsirkin, P. Aguado Puente, and I. Souza, ArXiv e-prints (2017), arXiv:1710.03204 [cond-
mat.mtrl-sci] .

[15] J. Ibañez-Azpiroz, S. S. Tsirkin, and I. Souza, ArXiv e-prints (2018), arXiv:1804.04030 .

[16] T. Yusufaly, D. Vanderbilt, and S. Coh, “Tight-Binding Formalism in the Context of the PythTB
Package,” http://physics.rutgers.edu/pythtb/formalism.html.

[17] R. Wang, E. A. Lazar, H. Park, A. J. Millis, and C. A. Marianetti, Physical Review B 90 (2014),
10.1103/PhysRevB.90.165125.

[18] A. Damle and L. Lin, ArXiv e-prints (2017), 1703.06958 .

[19] J. Qiao, J. Zhou, Z. Yuan, and W. Zhao, Phys. Rev. B 98, 214402 (2018).

[20] G. Y. Guo, S. Murakami, T.-W. Chen, and N. Nagaosa, Phys. Rev. Lett. 100, 096401 (2008).

[21] G. Pizzi, A. Cepellotti, R. Sabatini, N. Marzari, and B. Kozinsky, Computational Materials

Science 111, 218 (2016).

[22] V. Vitale, G. Pizzi, A. Marrazzo, J. R. Yates, N. Marzari, and A. A. Mostofi, Materials Cloud
Archive (2019), doi:10.24435/materialscloud:2019.0044/v2.
60 wannier90: Tutorial

[23] V. Vitale, G. Pizzi, A. Marrazzo, J. Yates, N. Marzari, and A. Mostofi, “Automated high-
throughput wannierisation,” (2019), arXiv:1909.00433 .