- Basic tags
- Displacement creation tags
- Band structure tags
- Mesh sampling tags
- Phonon density of states (DOS) tags
- Thermal properties related tags
- Thermal displacements
- Specific q-points
- Non-analytical term correction
- Group velocity
- Symmetry
- Force constants
- Create animation file
- Create modulated structure
- Characters of irreducible representations
- Input/Output file control

Most of the setting tags have corresponding command-line options (Command options).

For specifying real and reciprocal points, fractional values
(e.g. `1/3`

) are accepted. However fractional values must not
have space among characters (e.g. `1 / 3`

) are not allowed.

`DIM`

¶The supercell is created from the input unit cell. When three integers are specified, a supercell elongated along axes of unit cell is created.

```
DIM = 2 2 3
```

In this case, a 2x2x3 supercell is created.

When nine integers are specified, the supercell is created by multiplying the supercell matrix \(M_\mathrm{s}\) with the unit cell. For example,

```
DIM = 0 1 1 1 0 1 1 1 0
```

the supercell matrix is

\[\begin{split}M_\mathrm{s} = \begin{pmatrix}
0 & 1 & 1 \\
1 & 0 & 1 \\
1 & 1 & 0
\end{pmatrix}\end{split}\]

where the rows correspond to the first three, second three, and third three sets of numbers, respectively. When lattice parameters of unit cell are the column vectors of \(\mathbf{a}_\mathrm{u}\), \(\mathbf{b}_\mathrm{u}\), and \(\mathbf{c}_\mathrm{u}\), those of supercell, \(\mathbf{a}_\mathrm{s}\), \(\mathbf{b}_\mathrm{s}\), \(\mathbf{c}_\mathrm{s}\), are determined by,

\[( \mathbf{a}_\mathrm{s} \; \mathbf{b}_\mathrm{s} \; \mathbf{c}_\mathrm{s} )
= ( \mathbf{a}_\mathrm{u} \; \mathbf{b}_\mathrm{u} \;
\mathbf{c}_\mathrm{u} ) M_\mathrm{s}\]

Be careful that the axes in `POSCAR`

is defined by three row
vectors, i.e., \(( \mathbf{a}_\mathrm{u} \; \mathbf{b}_\mathrm{u}
\; \mathbf{c}_\mathrm{u} )^T\).

`PRIMITIVE_AXIS`

or `PRIMITIVE_AXES`

¶```
PRIMITIVE_AXIS = 0.0 0.5 0.5 0.5 0.0 0.5 0.5 0.5 0.0
```

Likewise,

```
PRIMITIVE_AXIS = 0 1/2 1/2 1/2 0 1/2 1/2 1/2 0
```

The primitive cell for building the dynamical matrix is created by multiplying primitive-axis matrix \(M_\mathrm{p}\). Let the matrix as,

\[\begin{split}M_\mathrm{p} = \begin{pmatrix}
0.0 & 0.5 & 0.5 \\
0.5 & 0.0 & 0.5 \\
0.5 & 0.5 & 0.0
\end{pmatrix}\end{split}\]

where the rows correspond to the first three, second three, and third three sets of numbers, respectively.

When lattice parameters of unit cell (set by `POSCAR`

) are the
column vectors of \(\mathbf{a}_\mathrm{u}\),
\(\mathbf{b}_\mathrm{u}\), and \(\mathbf{c}_\mathrm{u}\),
those of supercell, \(\mathbf{a}_\mathrm{p}\),
\(\mathbf{b}_\mathrm{p}\), \(\mathbf{c}_\mathrm{p}\), are
determined by,

\[( \mathbf{a}_\mathrm{p} \; \mathbf{b}_\mathrm{p} \; \mathbf{c}_\mathrm{p} )
= ( \mathbf{a}_\mathrm{u} \; \mathbf{b}_\mathrm{u} \;
\mathbf{c}_\mathrm{u} ) M_\mathrm{p}\]

Be careful that the axes in `POSCAR`

is defined by three row
vectors, i.e., \(( \mathbf{a}_\mathrm{u} \; \mathbf{b}_\mathrm{u}
\; \mathbf{c}_\mathrm{u} )^T\).

`ATOM_NAME`

¶Chemical symbols

```
ATOM_NAME = Si O
```

The number of chemical symbols have to be same as that of the numbers
in the sixth line of `POSCAR`

.

Chemical symbols read by phonopy are overwritten by those written in
`POSCAR`

. See `POSCAR`

examples. In WIEN2k mode,
you don’t need to set this tag, i.e., chemical symbols are read from
the structure file.

`EIGENVECTORS`

¶When this tag is ‘.TRUE.’, eigenvectors are calculated. With `-p`

option, partial density of states are calculated.

`MASS`

¶This tag is not necessary to use usually, because atomic masses are automatically set from the chemical symbols.

Atomic masses of a **primitive cell** are overwritten by the values
specified. The order of atoms in the primitive cell that is defined by
`PRIMITIVE_AXIS`

tag can be shown using `-v`

option. It must be
noted that this tag does not affect to the symmetry search.

For example, when there are six atoms in a primitive cell, `MASS`

is
set as follows

```
MASS = 28.085 28.085 16.000 16.000 16.000 16.000
```

`MAGMOM`

¶Symmetry of spin such as collinear magnetic moments is considered
using this tag. The number of values has to be equal to the number of
atoms in the unit cell, not the primitive cell or supercell. If this
tag is used with `-d`

option (`CREATE_DISPLACEMENTS`

tag),
`MAGMOM`

file is created. This file contains the `MAGMOM`

information of the supercell used for VASP. Unlike `MAGMOM`

in VASP,
`*`

can not be used, i.e., all the values (the same number of times
to the number of atoms in unit cell) have to be explicitly written.

```
MAGMOM = 1.0 1.0 -1.0 -1.0
```

`FREQUENCY_CONVERSION_FACTOR`

¶Unit conversion factor of frequency from input values to your favorite unit can be specified, but the use should be limited and it is recommended to use this tag to convert the frequency unit to THz in some exceptional case, for example, a special force calculator whose physical unit system is different from the default setting of phonopy is used. If the frequency unit is different from THz, though it works just for seeing results of frequencies, e.g., band structure or DOS, it doesn’t work for derived values like thermal properties and mean square displacements.

The default values for calculators are those to convert frequency
units to THz. The default conversion factors for `wien2k`

,
`abinit`

, `pwscf`

, `elk`

, and CRYSTAL are 3.44595, 21.49068, 108.9708,
154.1079, and 15.633302 respectively. These are determined following the physical
unit systems of the calculators. How to calcualte these conversion
factors is explained at Physical unit conversion.

`CREATE_DISPLACEMENTS`

¶Supercells with displacements are created. This tag is used as the post process of phonon calculation.

```
CREATE_DISPLACEMENTS = .TRUE.
DIM = 2 2 2
```

`DISPLACEMENT_DISTANCE`

¶Finite atomic displacement distance is set as specified value when
creating supercells with displacements. The default displacement
amplitude is 0.01 \(\textrm{\AA}\), but when the `wien2k`

or
`abinit`

option is specified, the default value is 0.02 Bohr.

`DIAG`

¶When this tag is set `.FALSE.`

, displacements in diagonal directions
are not searched, i.e. all the displacements are along the lattice
vectors. `DIAG = .FALSE.`

is recommended if one of the lattice
parameter of your supercell is much longer or much shorter than the
other lattice parameters.

`PM`

¶This tag specified how displacements are found. When `PM = .FALSE.`

,
least displacements that can calculate force constants are found. This
may cause less accurate result. When `PM = .TRUE.`

, all the
displacements that are opposite directions to the least displacements
are also found, which is called plus-minus displacements here. The
default setting is `PM = AUTO`

. Plus-minus displacements are
considered with this tag. If the plus and minus displacements are
symmetrically equivalent, only the plus displacement is found. This
may be in between `.FALSE.`

and `.TRUE.`

. You can check how it
works to see the file `DISP`

where displacement directions on atoms
are written.

`BAND`

and `BAND_POINTS`

¶`BAND`

gives sampling band paths. The reciprocal points are
specified in reduced coordinates. The given points are connected for
defining band paths. When comma `,`

is inserted between the points,
the paths are disconnected.

`BAND_POINTS`

gives the number of sampling points including the path
ends. The default value is `BAND_POINTS = 51`

.

An example of three paths, (0,0,0) to (1/2,0,1/2), (1/2,1/2,1) to (0,0,0), and (0,0,0) to (1/2,1/2,1/2), with 101 sampling points of each path are as follows:

```
BAND = 0 0 0 1/2 0 1/2, 1/2 1/2 1 0 0 0 1/2 1/2 1/2
BAND_POINTS = 101
```

`BAND_LABELS`

¶Labels specified are depicted in band structure plot at the points of
band segments. The number of labels has to correspond to the number of
band paths specified by `BAND`

plus one. When LaTeX math style
expression such as \(\Gamma\) (`\Gamma`

) is expected, it is
probably necessary to place it between two $ characters.

```
BAND = 1/2 0 1/2 0 0 0 1/2 1/2 1/2
BAND_LABELS = X $\Gamma$ L
```

The colors of curves are automatically determined by matplotlib. The same color in a band segment shows the same kind of band. Between different band segments, the correspondence of colors doesn’t mean anything.

`BAND_CONNECTION`

¶With this option, band connections are estimated from eigenvectors and
band structure is drawn considering band crossings. In sensitive
cases, to obtain better band connections, it requires to increase
number of points calculated in band segments by the `BAND_POINTS`

tag.

```
BAND = 1/2 0 1/2 0 0 0 1/2 1/2 1/2
BAND_POINTS = 101
BAND_CONNECTION = .TRUE.
```

Mesh sampling tags are used commonly for calculations of thermal properties and density of states.

`MESH`

, `MP`

, or `MESH_NUMBERS`

¶`MESH`

numbers give uniform meshes in each axis. As the default
behavior, the center of mesh is determined by the Monkhorst-Pack
scheme, i.e., for odd number, a point comes to the center, and for
even number, the center is shifted half in the distance between
neighboring mesh points.

Examples of an even mesh with \(\Gamma\) center in two ways,

```
MESH = 8 8 8
GAMMA_CENTER = .TRUE.
```

```
MESH = 8 8 8
MP_SHIFT = 1/2 1/2 1/2
```

`MP_SHIFT`

¶`MP_SHIFT`

gives the shifts in direction along the corresponding
reciprocal axes (\(a^*\), \(b^*\), \(c^*\)). 0 or 1/2
(0.5) can be used as these values. 1/2 means the half mesh shift with
respect to neighboring grid points in each direction.

`GAMMA_CENTER`

¶Instead of employing the Monkhorst-Pack scheme for the mesh sampling,
\(\Gamma\) center mesh is used. The default value is `.FALSE.`

.

```
GAMMA_CENTER = .TRUE.
```

`WRITE_MESH`

¶With a dense mesh, with eigenvectors, without mesh symmetry, sometimes
its output file `mesh.yaml`

or `mesh.hdf5`

can be huge. However
when those files are not needed, e.g., in (P)DOS calculation,
`WRITE_MESH = .FALSE.`

can disable to write out those files. With
(P)DOS calculation, DOS output files are obtained even with
`WRITE_MESH = .FALSE.`

. The default setting is `.TRUE.`

.

```
WRITE_MESH = .FALSE.
```

Phonon density of states (DOS) is calcualted either with smearing method (default) or tetrahedron method. Phonons are calculated on a sampling mesh, therefore these tags must be used with Mesh sampling tags. The physical unit of horizontal axis is that of frequency that the user employs, e.g., THz, and that of vertical axis is {no. of states}/({unit cell} x {unit of the horizontal axis}). If the DOS is integrated over the frequency range, it will be \(3N_\mathrm{a}\) states, where \(N_\mathrm{a}\) is the number of atoms in the unit cell.

Phonon-DOS is formally defined as

\[g(\omega) = \frac{1}{N} \sum_\lambda \delta(\omega - \omega_\lambda)\]

where \(N\) is the number of unit cells and \(\lambda = (\nu, \mathbf{q})\) with \(\nu\) as the band index and \(\mathbf{q}\) as the q-point. This is computed on a set of descritized sampling frequency points for which \(\omega\) is specified arbitrary using DOS_RANGE. The phonon frequencies \(\omega_\lambda\) are obtained on a sampling mesh whose the number of grid points being \(N\). In the smearing method, the delta function is replaced by normal distribution (Gaussian function) with the standard deviation specified by SIGMA. In the tetrahedron method, the Brillouin integration is made analytically within tetrahedra in reciprocal space.

`DOS`

¶This tag enables to calculate DOS. This tag is automatically set when
`PDOS`

tag or `-p`

option.

```
DOS = .TRUE.
```

`DOS_RANGE`

¶```
DOS_RANGE = 0 40 0.1
```

Total and partial density of states are drawn with some parameters. The example makes DOS be calculated from frequency=0 to 40 with 0.1 pitch.

FMIN, FMAX, and FPITCH can be alternatively used to specify the minimum and maximum frequencies (the first and second values).

`FMIN`

, `FMAX`

, and `FPITCH`

¶The uniform frequency sampling points for phonon-DOS calculation are
specified. `FMIN`

and `FMAX`

give the minimum, maximum frequencies
of the range, respectively, and `FPITCH`

gives the frequency pitch
to be sampled. These three values are the same as those that can be
specified by `DOS_RANGE`

.

`PDOS`

¶Projected DOS is calculated using this tag. The formal definition is written as

\[g^j(\omega, \hat{\mathbf{n}}) = \frac{1}{N} \sum_\lambda
\delta(\omega - \omega_\lambda) |\hat{\mathbf{n}} \cdot
\mathbf{e}^j_\lambda|^2,\]

where \(j\) is the atom indices and \(\hat{\mathbf{n}}\) is the unit projection direction vector. Without specifying PROJECTION_DIRECTION or XYZ_PROJECTION, PDOS is computed as sum of \(g^j(\omega, \hat{\mathbf{n}})\) projected onto Cartesian axes \(x,y,z\), i.e.,

\[g^j(\omega) = \sum_{\hat{\mathbf{n}} = \{x, y, z\}} g^j(\omega,
\hat{\mathbf{n}}).\]

The atom indices \(j\) are specified by

```
PDOS = 1 2, 3 4 5 6
```

These numbers are those in the primitive cell. `,`

separates the
atom sets. In this example, atom 1 and 2 are summarized as one curve
and atom 3, 4, 5, and, 6 are summarized as another curve.

`EIGENVECTORS = .TRUE.`

and `MESH_SYMMETRY = .FALSE.`

are
automatically set, therefore the calculation takes much more time than
usual DOS calculation. With a very dense sampling mesh, writing data
into `mesh.yaml`

or `mesh.hdf5`

can be unexpectedly huge. If only
PDOS is necessary but these output files are unnecessary, then it is
good to consider using `WRITE_MESH = .FALSE.`

(WRITE_MESH).

`PROJECTION_DIRECTION`

¶Eigenvectors are projected along the direction specified by this tag.
Projection direction is specified in reduced coordinates, i.e., with
respect to *a*, *b*, *c* axes.

```
PDOS = 1, 2
PROJECTION_DIRECTION = -1 1 1
```

`XYZ_PROJECTION`

¶PDOS is calculated using eigenvectors projected along x, y, and z
Cartesian coordinates. The format of output file `partial_dos.dat`

becomes different when using this tag, where phonon-mode-frequency and
x, y, and z components of PDOS are written out in the order:

```
frequency atom1_x atom1_y atom1_z atom2_x atom2_y atom2_z ...
```

With `-p`

option, three curves are drawn. These correspond to
sums of all projections to x, sums of all projections to y, and sums
of all projections to z composents of eigenvectors, respectively.

```
XYZ_PROJECTION = .TRUE.
```

`SIGMA`

¶This tag specifies the smearing width. The unit is same as that used for phonon frequency. The default value is the value given by the difference of maximum and minimum frequencies divided by 100.

```
SIGMA = 0.1
```

`TETRAHEDRON`

¶Tetrahedron method is used instead of smearing method.

`DEBYE_MODEL`

¶By setting `.TRUE.`

, DOS at lower phonon frequencies are fit to a
Debye model. By default, the DOS from 0 to 1/4 of the maximum phonon
frequencies are used for the fitting. The function used to the fitting
is \(D(\omega)=a\omega^2\) where \(a\) is the parameter and
the Debye frequency is \((9N/a)^{1/3}\) where \(N\) is the
number of atoms in unit cell. Users have to unserstand that this is
**not** a unique way to determine Debye frequency. Debye frequency is
dependent on how to parameterize it.

```
DEBYE_MODEL = .TRUE.
```

`MOMEMT`

and `MOMENT_ORDER`

¶Phonon moments for DOS and PDOS defined below are calculated using
these tags up to arbitrary order. The order is specified with
`MOMENT_ORDER`

(\(n\) in the formula). Unless `MOMENT_ORDER`

specified, the first and second moments are calculated.

The moments for DOS are given as

\[M_n(\omega_\text{min}, \omega_\text{max})
&=\frac{\int_{\omega_\text{min}}^{\omega_\text{max}} \omega^n
g(\omega) d\omega} {\int_{\omega_\text{min}}^{\omega_\text{max}}
g(\omega) d\omega}.\]

The moments for PDOS are given as

\[M_n^j(\omega_\text{min}, \omega_\text{max})
&=\frac{\int_{\omega_\text{min}}^{\omega_\text{max}} \omega^n
g^j(\omega) d\omega} {\int_{\omega_\text{min}}^{\omega_\text{max}}
g^j(\omega) d\omega}.\]

\(\omega_\text{min}\) and \(\omega_\text{max}\) are specified :using ref:dos_fmin_fmax_tags tags. When these are not specified, the moments are computed with the range of \(\epsilon < \omega < \infty\), where \(\epsilon\) is a small positive value. Imaginary frequencies are treated as negative real values in this computation, therefore it is not a good idea to set negative \(\omega_\text{min}\).

```
MOMENT = .TRUE.
MOMENT_ORDER = 3
```

`TDISP`

, `TMAX`

, `TMIN`

, and `TSTEP`

¶Mean square displacements projected to Cartesian axes as a function of
temperature are calculated from the number of phonon excitations. The
usages of `TMAX`

, `TMIN`

, `TSTEP`

tags are same as those in
thermal properties tags. Phonon
frequencies in THz, which is the default setting of phonopy, are used to
obtain the mean square displacements, therefore physical units have to
be set properly for it (see Interfaces to calculators.) The result
is given in \(\textrm{\AA}^2\) and writen into
`thermal_displacements.yaml`

. See the detail of the method,
Thermal displacement. These tags must be used with
Mesh sampling tags

`CUTOFF_FREQUENCY`

tag with a small value is recommened to be set
when sampling \(\Gamma\) point or using very dense sampling mesh
to avoid divergence.

The projection is applied along arbitrary direction using
`PROJECTION_DIRECTION`

tag (PROJECTION_DIRECTION).

`mesh.yaml`

or `mesh.hdf5`

is not written out from phonopy-1.11.14.

```
TDISP = .TRUE.
PROJECTION_DIRECTION = 1 1 0
```

`TDISPMAT`

, `TMAX`

, `TMIN`

, and `TSTEP`

¶Mean square displacement matricies are calculated. The definition is
shown at Thermal displacement. Phonon frequencies in THz, which
is the default setting of phonopy, are
used to obtain the mean square displacement matricies, therefore
physical units have to be set properly for it (see
Interfaces to calculators.) The result is given in
\(\textrm{\AA}^2\) and writen into
`thermal_displacement_matrices.yaml`

where six matrix elements are
given in the order of xx, yy, zz, yz, xz, xy. In this yaml file,
`displacement_matrices`

and `displacement_matrices_cif`

correspond
to \(\mathrm{U}_\text{cart}\) and \(\mathrm{U}_\text{cif}\)
defined at Mean square displacement matrix, respectively.

`CUTOFF_FREQUENCY`

tag with a small value is recommened to be set
when sampling \(\Gamma\) point or using very dense sampling mesh
to avoid divergence.

The 3x3 matrix restricts distribution of each atom around the equilibrium position to be ellipsoid. But the distribution is not necessarily to be so.

`mesh.yaml`

or `mesh.hdf5`

is not written out from phonopy-1.11.14.

```
TDISPMAT = .TRUE.
```

`TDISPMAT_CIF`

¶This tag specifis a temperature (K) at which thermal displacement is
calculated and the mean square displacement matrix is written to the
cif file `tdispmat.cif`

with the dictionary item `aniso_U`

. Phonon
frequencies in THz, which is the default setting of phonopy, are used
to obtain the mean square displacement matricies, therefore physical
units have to be set properly for it (see
Interfaces to calculators.) The result is given in
\(\textrm{\AA}^2\).

`mesh.yaml`

or `mesh.hdf5`

is not written out from phonopy-1.11.14.

```
TDISPMAT_CIF = 1273.0
```

`CUTOFF_FREQUENCY`

¶Frequencies lower than this cutoff frequency are not used to calculate thermal displacements.

`QPOINTS`

¶When q-points are supplied, those phonons are calculated. Q-points are specified successive values separated by spaces and collected by every three values as vectors in reciprocal reduced coordinates.

```
QPOINTS = 0 0 0 1/2 1/2 1/2 1/2 0 1/2
```

With `QPOINTS = .TRUE.`

, q-points are read from `QPOITNS`

file
(see the file format at QPOINTS) in curret directory
phonons at the q-points are calculated.

```
QPOINTS = .TRUE.
```

`WRITEDM`

¶```
WRITEDM = .TRUE.
```

Dynamical matrices \(D\) are written into `qpoints.yaml`

in the following \(6N\times3N\) format, where *N* is the number of atoms in
the primitive cell.

The physical unit of dynamical matrix is ```
[unit of force] / ([unit of
displacement] * [unit of mass])
```

, i.e., square of the unit of phonon
frequency before multiplying the unit conversion factor
(see FREQUENCY_CONVERSION_FACTOR).

\[\begin{split}D =
\begin{pmatrix}
D_{11} & D_{12} & D_{13} & \\
D_{21} & D_{22} & D_{23} & \cdots \\
D_{31} & D_{32} & D_{33} & \\
& \vdots & & \\
\end{pmatrix},\end{split}\]

and \(D_{jj'}\) is

\[\begin{split}D_{jj'} =
\begin{pmatrix}
Re(D_{jj'}^{xx}) & Im(D_{jj'}^{xx}) & Re(D_{jj'}^{xy}) &
Im(D_{jj'}^{xy}) & Re(D_{jj'}^{xz}) & Im(D_{jj'}^{xz}) \\
Re(D_{jj'}^{yx}) & Im(D_{jj'}^{yx}) & Re(D_{jj'}^{yy}) &
Im(D_{jj'}^{yy}) & Re(D_{jj'}^{yz}) & Im(D_{jj'}^{yz}) \\
Re(D_{jj'}^{zx}) & Im(D_{jj'}^{zx}) & Re(D_{jj'}^{zy}) &
Im(D_{jj'}^{zy}) & Re(D_{jj'}^{zz}) & Im(D_{jj'}^{zz}) \\
\end{pmatrix},\end{split}\]

where *j* and *j’* are the atomic indices in the primitive cell. The
phonon frequencies may be recovered from `qpoints.yaml`

by writing a
simple python script. For example, `qpoints.yaml`

is obtained for
NaCl at \(q=(0, 0.5, 0.5)\) by

```
phonopy --dim="2 2 2" --pa="0 1/2 1/2 1/2 0 1/2 1/2 1/2 0" --qpoints="0 1/2 1/2" --writedm
```

and the dynamical matrix may be used as

```
#!/usr/bin/env python
import yaml
import numpy as np
data = yaml.load(open("qpoints.yaml"))
dynmat = []
dynmat_data = data['phonon'][0]['dynamical_matrix']
for row in dynmat_data:
vals = np.reshape(row, (-1, 2))
dynmat.append(vals[:, 0] + vals[:, 1] * 1j)
dynmat = np.array(dynmat)
eigvals, eigvecs, = np.linalg.eigh(dynmat)
frequencies = np.sqrt(np.abs(eigvals.real)) * np.sign(eigvals.real)
conversion_factor_to_THz = 15.633302
print frequencies * conversion_factor_to_THz
```

`NAC`

¶Non-analytical term correction is applied to dynamical
matrix. `BORN`

file has to be prepared in the current directory. See
BORN (optional) and Non-analytical term correction.

```
NAC = .TRUE.
```

`Q_DIRECTION`

¶This tag is used to activate NAC at
\(\mathbf{q}\rightarrow\mathbf{0}\), i.e. practically
\(\Gamma\)-point. Away from \(\Gamma\)-point, this setting is
ignored and the specified **q**-point is used as the **q**-direction.

```
MESH = 1 1 1
NAC = .TRUE.
Q_DIRECTION = 1 0 0
```

`GROUP_VELOCITY`

¶Group velocities at q-points are calculated by using this tag. The group velocities are written into a yaml file corresponding to the run mode in Cartesian coordinates. The physical unit depends on physical units of input files and frequency conversion factor, but if VASP and the default settings (e.g., THz for phonon frequency) are simply used, then the physical unit will be Angstrom THz.

```
GROUP_VELOCITY = .TRUE.
```

Technical details are shown at Group velocity.

`GV_DELTA_Q`

¶The reciprocal distance used for finite difference method is specified. The default value is 1e-4.

```
GV_DELTA_Q = 0.01
```

`MESH_SYMMETRY`

¶Symmetry search on the reciprocal sampling mesh is disabled by setting
`MESH_SYMMETRY = .FALSE.`

.

`FC_SYMMETRY`

¶This tag is used to symmetrize force constants partly. The number of iteration of the following set of symmetrization applied to force constants is specified. The default value is 0. In the case of VASP, this tag is usually unnecessary to be specified.

```
FC_SYMMETRY = 1
```

From the translation invariance condition,

\[\sum_i \Phi_{ij}^{\alpha\beta} = 0, \;\;\text{for all $j$, $\alpha$, $\beta$},\]

where *i* and *j* are the atom indices, and \(\alpha\) and
\(\beta\) are the Catesian indices for atoms *i* and *j*,
respectively. Force constants are symmetric in each pair as

\[\Phi_{ij}^{\alpha\beta}
= \frac{\partial^2 U}{\partial u_i^\alpha \partial u_j^\beta}
= \frac{\partial^2 U}{\partial u_j^\beta \partial u_i^\alpha}
= \Phi_{ji}^{\beta\alpha}\]

These symmetrizations break the symmetry conditions each other. Be
careful that the other symmetries of force constants, i.e., the
symmetry from crystal symmetry or rotational symmetry, are broken to
force applying `FC_SYMMETRY`

.

`FORCE_CONSTANTS`

¶```
FORCE_CONSTANTS = READ
```

There are three values to be set, which are `READ`

and `WRITE`

,
and `.FALSE.`

. The default is `.FALSE.`

. When ```
FORCE_CONSTANTS =
READ
```

, force constants are read from `FORCE_CONSTANTS`

file. With
`FORCE_CONSTANTS = WRITE`

, force constants calculated from
`FORCE_SETS`

are written to `FORCE_CONSTANTS`

file.

The file format of `FORCE_CONSTANTS`

is shown
here.

`READ_FORCE_CONSTANTS`

¶`READ_FORCE_CONSTANTS = .TRUE.`

is equivalent to ```
FORCE_CONSTANTS =
READ
```

.

`WRITE_FORCE_CONSTANTS`

¶`WRITE_FORCE_CONSTANTS = .TRUE.`

is equivalent to ```
FORCE_CONSTANTS =
WRITE
```

.

`ANIME_TYPE`

¶```
ANIME_TYPE = JMOL
```

There are `V_SIM`

, `ARC`

, `XYZ`

, `JMOL`

, and `POSCAR`

settings. Those may be viewed by `v_sim`

, `gdis`

, `jmol`

(animation), `jmol`

(vibration), respectively. For `POSCAR`

, a set
of `POSCAR`

format structure files corresponding to respective
animation images are created such as `APOSCAR-000`

,
`APOSCAR-001`

,….

There are several parameters to be set in the `ANIME`

tag.

`ANIME`

¶**The format of ``ANIME`` tag was modified after ver. 0.9.3.3.**

```
ANIME = 0.5 0.5 0
```

The values are the *q*-point to be calculated. An animation file of
`anime.ascii`

is generated.

Phonon is only calculated at \(\Gamma\) point. So *q*-point is not
necessary to be set.

`anime.arc`

, `anime.xyz`

, `anime.xyz_jmol`

, or `APOSCAR-*`

are generated according to the `ANIME_TYPE`

setting.

```
ANIME = 4 5 20 0.5 0.5 0
```

The values are as follows from left:

- Band index given by ascending order in phonon frequency.
- Magnitude to be multiplied. In the harmonic phonon calculation, there is no amplitude information obtained directly. The relative amplitude among atoms in primitive cell can be obtained from eigenvectors with the constraint of the norm or the eigenvectors equals one, i.e., number of atoms in the primitive is large, the displacements become small. Therefore this has to be adjusted to make the animation good looking.
- Number of images in one phonon period.
- (4-6) Shift of atomic points in reduced coordinate in real space. These
values can be omitted and the default values are
`0 0 0`

.

For `anime.xyz_jmol`

, the first and third values are not used,
however dummy values, e.g. 0, are required.

`MODULATION`

¶The `MODULATION`

tag is used to create a crystal structure with
displacements along normal modes at q-point in the specified supercell
dimension.

Atomic displacement of the *j*-th atom is created from the real part
of the eigenvectors with amplitudes and phase factors as

\[\frac{A} { \sqrt{N_\mathrm{a}m_j} } \operatorname{Re} \left[ \exp(i\phi)
\mathbf{e}_j \exp( \mathbf{q} \cdot \mathbf{r}_{jl} ) \right],\]

where \(A\) is the amplitude, \(\phi\) is the phase,
\(N_\mathrm{a}\) is the number of atoms in the supercell specified
in this tag and \(m_j\) is the mass of the *j*-th atom,
\(\mathbf{q}\) is the q-point specified, \(\mathbf{r}_{jl}\)
is the position of the *j*-th atom in the *l*-th unit cell, and
\(\mathbf{e}_j\) is the *j*-th atom part of eigenvector. Convention of
eigenvector or dynamical matrix employed in phonopy is shown in
Dynamical matrix.

If several modes are specified as shown in the example above, they are
overlapped on the structure. The output filenames are
`MPOSCAR...`

. Each modulated structure of a normal mode is written
in `MPOSCAR-<number>`

where the numbers correspond to the order of
specified sets of modulations. `MPOSCAR`

is the structure where all
the modulations are summed. `MPOSCAR-orig`

is the structure without
containing modulation, but the dimension is the one that is specified.
Some information is written into `modulation.yaml`

.

The first three (nine) values correspond to supercell dimension
(supercell matrix) like the CELL_FILENAME tag. The following
values are used to describe how the atoms are modulated. Multiple sets
of modulations can be specified by separating by comma `,`

. In each
set, the first three values give a Q-point in the reduced coordinates
in reciprocal space. Then the next three values are the band index
from the bottom with ascending order, amplitude, and phase factor in
degrees. The phase factor is optional. If it is not specified, 0 is
used.

Before multiplying user specified phase factor, the phase of
the modulation vector is adjusted as the largest absolute value,
\(\left|\mathbf{e}_j\right|/\sqrt{m_j}\), of element of
3N dimensional modulation vector to be real. The complex modulation
vector is shown in `modulation.yaml`

.

```
MODULATION = 3 3 1, 1/3 1/3 0 1 2, 1/3 1/3 0 2 3.5
```

```
MODULATION = 3 3 1, 1/3 1/3 0 1 2, 1/3 0 0 2 2
```

```
MODULATION = 3 3 1, 1/3 1/3 0 1 1 0, 1/3 1/3 0 1 1 90
```

```
MODULATION = -1 1 1 1 -1 1 1 1 -1, 1/2 1/2 0 1 2
```

`IRREPS`

¶Characters of irreducible representations (IRs) of phonon modes are
shown. For this calculation, a primitive cell has to be used. If the
input unit cell is a non-primitive cell, it has to be transformed to a
primitive cell using `PRIMITIVE_AXIS`

tag.

The first three values gives a *q*-point in reduced coordinates
to be calculated. The degenerated modes are searched only by the closeness of
frequencies. The frequency difference to be tolerated is specified by
the fourth value in the frequency unit that the user specified.

```
IRREPS = 0 0 0 1e-3
```

Only the databases of IRs for a few point group types at the \(\Gamma\) point are implemented. If the database is available, the symbols of the IRs and the rotation operations are shown.

`SHOW_IRREPS`

¶Irreducible representations are shown along with character table.

```
IRREPS = 1/3 1/3 0
SHOW_IRREPS = .TRUE.
```

`LITTLE_COGROUP`

¶Show irreps of little co-group (point-group of wavevector) instead of little group.

```
IRREPS = 0 0 1/8
LITTLE_COGROUP = .TRUE.
```

`FC_FORMAT`

, `READFC_FORMAT`

, `WRITEFC_FORMAT`

¶There are two file-formats to store force constants. Currently
text style (`TEXT`

) and hdf5 (`HDF5`

)
formats are supported. The default file format is the text
style. Reading and writing force constants are
invoked by FORCE_CONSTANTS tag. Using
these tags, the input/output formats are switched.

`FC_FORMAT`

affects to both input and output, e.g.:

```
FORCE_CONSTANTS = WRITE
FC_FORMAT = HDF5
```

`READFC_FORMAT`

and `WRITEFC_FORMAT`

can be used to control
input and output formats separately, i.e., the following setting to
convert force constants format is possible:

```
READ_FORCE_CONSTANTS = .TRUE.
WRITE_FORCE_CONSTANTS = .TRUE.
WRITEFC_FORMAT = HDF5
```

`BAND_FORMAT`

, `MESH_FORMAT`

, `QPOINTS_FORMAT`

¶There are two file-formats to write the results of band structure,
mesh, and q-points calculations. Currently YAML (`YAML`

) and hdf5
(`HDF5`

) formats are supported. The default file format is the YAML
format. The file format is changed as follows:

```
BAND_FORMAT = HDF5
```

```
MESH_FORMAT = HDF5
```

```
QPOINTS_FORMAT = HDF5
```

`HDF5`

¶**This tag is deprecated.**

The following output files are written in hdf5 format instead of their
original formats (in parenthesis) by `HDF5 = .TRUE.`

. In addition,
`force_constants.hdf5`

is read with this tag.

`force_constants.hdf5`

(`FORCE_CONSTANTS`

)`mesh.hdf5`

(`mesh.yaml`

)`band.hdf5`

(`band.yaml`

)`qpoints.hdf5`

(`qpoints.yaml`

)

```
HDF5 = .TRUE.
```

`force_constants.hdf5`

¶With `--hdf5`

option and `FORCE_CONSTANTS = WRITE`

(`--writefc`

), `force_constants.hdf5`

is written.
With `--hdf5`

option and `FORCE_CONSTANTS = READ`

(`--readfc`

),
`force_constants.hdf5`

is read.

`mesh.hdf5`

¶In the mesh sampling calculations (see Mesh sampling tags),
calculation results are written into `mesh.hdf5`

but not into
`mesh.yaml`

. Using this option may reduce the data output size and
thus writing time when `mesh.yaml`

is huge, e.g., eigenvectors are
written on a dense sampling mesh.

`qpoints.hdf5`

¶In the specific q-points calculations (QPOINTS),
calculation results are written into `qpoints.hdf5`

but not into
`qpoints.yaml`

. With WRITEDM, dynamical matrices are also
stored in `qpoints.hdf5`

. Using this option may be useful with large
set of q-points with including eigenvector or dynamical matrix output.

`band.hdf5`

¶In the band structure calculations (Band structure tags),
calculation results are written into `band.hdf5`

but not into
`band.yaml`

.