Input File Description

a string describing the task to be performed:
   'scf',
   'nscf',
   'bands',
   'relax',
   'md',
   'vc-relax',
   'vc-md'

   (vc = variable-cell).
         

reprinted on output.
         

Currently two verbosity levels are implemented:
  'high' and 'low'. 'debug' and 'medium' have the same
  effect as 'high'; 'default' and 'minimal', as 'low'
         

'from_scratch'  : from scratch. This is the normal way
                  to perform a PWscf calculation
'restart'       : from previous interrupted run. Use this
                  switch only if you want to continue an
                  interrupted calculation, not to start a
                  new one. See also startingpot, startingwfc
         

This flag controls the way wavefunctions are stored to disk :

.TRUE.  collect wavefunctions from all processors, store them
        into the output data directory outdir/prefix.save,
        one wavefunction per k-point in subdirs K000001/,
        K000001/, etc.

.FALSE. do not collect wavefunctions, leave them in temporary
        local files (one per processor). The resulting format
        will be readable only by jobs running on the same
        number of processors and pools. Useful if you do not
        need the wavefunction or if you want to reduce the I/O
        or the disk occupancy.
         

number of ionic + electronic steps
         

band energies are written every iprint iterations
         

calculate stress. It is set to .TRUE. automatically if
calculation='vc-md' or 'vc-relax'
         

print forces. Set to .TRUE. if calculation='relax','md','vc-md'
         

time step for molecular dynamics, in Rydberg atomic units
(1 a.u.=4.8378 * 10^-17 s : beware, the CP code use
 Hartree atomic units, half that much!!!)
         

input, temporary, output files are found in this directory,
see also 'wfcdir'
         

this directory specifies where to store files generated by
each processor (*.wfc{N}, *.igk{N}, etc.). The idea here is
to be able to separately store the largest files, while
the files necessary for restarting still go into 'outdir'
(for now only works for stand alone PW )
         

prepended to input/output filenames:
prefix.wfc, prefix.rho, etc.
         

If .false. a subdirectory for each k_point is not opened
in the prefix.save directory; Kohn-Sham eigenvalues are
stored instead in a single file for all k-points. Currently
doesnh't work together with wf_collect
         

jobs stops after max_seconds CPU time
         

convergence threshold on total energy (a.u) for ionic
minimization: the convergence criterion is satisfied
when the total energy changes less than etot_conv_thr
between two consecutive scf steps. Note that etot_conv_thr
is extensive, like the total energy.
See also forc_conv_thr - both criteria must be satisfied
         

convergence threshold on forces (a.u) for ionic minimization:
the convergence criterion is satisfied when all components of
all forces are smaller than forc_conv_thr.
See also etot_conv_thr (note that the latter is extensive,
forc_conv_thr is not) - both criteria must be satisfied
         

Specifies the amount of disk I/O activity
'high':    save all data at each SCF step

'default': save wavefunctions at each SCF step unless
           there is a single k-point per process

'low' :    store wfc in memory, save only at the end

'none':    do not save wfc, not even at the end
           (guaranteed to work only for 'scf', 'nscf',
            'bands' calculations)

If restarting from an interrupted calculation, the code
will try to figure out what is available on disk. The
more you write, the more complete the restart will be.
         

directory containing pseudopotential files
         

If .TRUE. a sawlike potential simulating an electric field
is added to the bare ionic potential. See variables
edir, eamp, emaxpos, eopreg for the form and size of
the added potential.
         

If .TRUE. and tefield=.TRUE. a dipole correction is also
added to the bare ionic potential - implements the recipe
of L. Bengtsson, PRB 59, 12301 (1999). See variables edir,
emaxpos, eopreg for the form of the correction, that must
be used only in a slab geometry, for surface calculations,
with the discontinuity in the empty space.
         

If .TRUE. a homogeneous finite electric field described
through the modern theory of the polarization is applied.
This is different from "tefield=.true." !
         

In the case of a finite electric field  ( lelfield == .TRUE. )
it defines the number of iterations for converging the
wavefunctions in the electric field Hamiltonian, for each
external iteration on the charge density
         

If .TRUE. perform a Berry phase calculation
See the header of PW/bp_c_phase.f90 for documentation
         

For Berry phase calculation: direction of the k-point
strings in reciprocal space. Allowed values: 1, 2, 3
1=first, 2=second, 3=third reciprocal lattice vector
For calculations with finite electric fields
(lelfield==.true.), gdir is the direction of the field
         

For Berry phase calculation: number of k-points to be
calculated along each symmetry-reduced string
The same for calculation with finite electric fields
(lelfield==.true.)
         

  Bravais-lattice index. In all cases except ibrav=0,
  either [celldm(1)-celldm(6)] or [a,b,c,cosab,cosac,cosbc]
  must be specified: see their description. For ibrav=0
  you may specify the lattice parameter celldm(1) or a.

ibrav      structure                   celldm(2)-celldm(6)
                                     or: b,c,cosab,cosac,cosbc
  0          free
      crystal axis provided in input: see card CELL_PARAMETERS

  1          cubic P (sc)
      v1 = a(1,0,0),  v2 = a(0,1,0),  v3 = a(0,0,1)

  2          cubic F (fcc)
      v1 = (a/2)(-1,0,1),  v2 = (a/2)(0,1,1), v3 = (a/2)(-1,1,0)

  3          cubic I (bcc)
      v1 = (a/2)(1,1,1),  v2 = (a/2)(-1,1,1),  v3 = (a/2)(-1,-1,1)

  4          Hexagonal and Trigonal P        celldm(3)=c/a
      v1 = a(1,0,0),  v2 = a(-1/2,sqrt(3)/2,0),  v3 = a(0,0,c/a)

  5          Trigonal R, 3fold axis c        celldm(4)=cos(alpha)
      The crystallographic vectors form a three-fold star around
      the z-axis, the primitive cell is a simple rhombohedron:
      v1 = a(tx,-ty,tz),   v2 = a(0,2ty,tz),   v3 = a(-tx,-ty,tz)
      where c=cos(alpha) is the cosine of the angle alpha between
      any pair of crystallographic vectors, tx, ty, tz are:
        tx=sqrt((1-c)/2), ty=sqrt((1-c)/6), tz=sqrt((1+2c)/3)
 -5          Trigonal R, 3fold axis <111>    celldm(4)=cos(alpha)
      The crystallographic vectors form a three-fold star around
      <111>. Defining a' = a/sqrt(3) :
      v1 = a' (u,v,v),   v2 = a' (v,u,v),   v3 = a' (v,v,u)
      where u and v are defined as
        u = tz - 2*sqrt(2)*ty,  v = tz + sqrt(2)*ty
      and tx, ty, tz as for case ibrav=5

  6          Tetragonal P (st)               celldm(3)=c/a
      v1 = a(1,0,0),  v2 = a(0,1,0),  v3 = a(0,0,c/a)

  7          Tetragonal I (bct)              celldm(3)=c/a
      v1=(a/2)(1,-1,c/a),  v2=(a/2)(1,1,c/a),  v3=(a/2)(-1,-1,c/a)

  8          Orthorhombic P                  celldm(2)=b/a
                                             celldm(3)=c/a
      v1 = (a,0,0),  v2 = (0,b,0), v3 = (0,0,c)

  9          Orthorhombic base-centered(bco) celldm(2)=b/a
                                             celldm(3)=c/a
      v1 = (a/2, b/2,0),  v2 = (-a/2,b/2,0),  v3 = (0,0,c)
 -9          as 9, alternate description
      v1 = (a/2,-b/2,0),  v2 = (a/2,-b/2,0),  v3 = (0,0,c)

 10          Orthorhombic face-centered      celldm(2)=b/a
                                             celldm(3)=c/a
      v1 = (a/2,0,c/2),  v2 = (a/2,b/2,0),  v3 = (0,b/2,c/2)

 11          Orthorhombic body-centered      celldm(2)=b/a
                                             celldm(3)=c/a
      v1=(a/2,b/2,c/2),  v2=(-a/2,b/2,c/2),  v3=(-a/2,-b/2,c/2)

 12          Monoclinic P, unique axis c     celldm(2)=b/a
                                             celldm(3)=c/a,
                                             celldm(4)=cos(ab)
      v1=(a,0,0), v2=(b*cos(gamma),b*sin(gamma),0),  v3 = (0,0,c)
      where gamma is the angle between axis a and b.
-12          Monoclinic P, unique axis b     celldm(2)=b/a
                                             celldm(3)=c/a,
                                             celldm(5)=cos(ac)
      v1 = (a,0,0), v2 = (0,b,0), v3 = (c*sin(beta),0,c*cos(beta))
      where beta is the angle between axis a and c

 13          Monoclinic base-centered        celldm(2)=b/a
                                             celldm(3)=c/a,
                                             celldm(4)=cos(ab)
      v1 = (  a/2,         0,                -c/2),
      v2 = (b*cos(gamma), b*sin(gamma), 0),
      v3 = (  a/2,         0,                  c/2),
      where gamma is the angle between axis a and b

 14          Triclinic                       celldm(2)= b/a,
                                             celldm(3)= c/a,
                                             celldm(4)= cos(bc),
                                             celldm(5)= cos(ac),
                                             celldm(6)= cos(ab)
      v1 = (a, 0, 0),
      v2 = (b*cos(gamma), b*sin(gamma), 0)
      v3 = (c*cos(beta),  c*(cos(alpha)-cos(beta)cos(gamma))/sin(gamma),
           c*sqrt( 1 + 2*cos(alpha)cos(beta)cos(gamma)
                     - cos(alpha)^2-cos(beta)^2-cos(gamma)^2 )/sin(gamma) )
  where alpha is the angle between axis b and c
         beta is the angle between axis a and c
        gamma is the angle between axis a and b
         

number of atoms in the unit cell
         

number of types of atoms in the unit cell
         

number of electronic states (bands) to be calculated.
Note that in spin-polarized calculations the number of
k-point, not the number of bands per k-point, is doubled
         

total charge of the system. Useful for simulations with charged cells.
By default the unit cell is assumed to be neutral (tot_charge=0).
tot_charge=+1 means one electron missing from the system,
tot_charge=-1 means one additional electron, and so on.

In a periodic calculation a compensating jellium background is
inserted to remove divergences if the cell is not neutral.
         

total majority spin charge - minority spin charge.
Used to impose a specific total electronic magnetization.
If unspecified then tot_magnetization variable is ignored and
the amount of electronic magnetization is determined during
the self-consistent cycle.
         

starting spin polarization on atomic type 'i' in a spin
polarized calculation. Values range between -1 (all spins
down for the valence electrons of atom type 'i') to 1
(all spins up). Breaks the symmetry and provides a starting
point for self-consistency. The default value is zero, BUT a
value MUST be specified for AT LEAST one atomic type in spin
polarized calculations, unless you constrain the magnetization
(see "tot_magnetization" and "constrained_magnetization").
Note that if you start from zero initial magnetization, you
will invariably end up in a nonmagnetic (zero magnetization)
state. If you want to start from an antiferromagnetic state,
you may need to define two different atomic species
corresponding to sublattices of the same atomic type.
starting_magnetization is ignored if you are performing a
non-scf calculation, if you are restarting from a previous
run, or restarting from an interrupted run.
If you fix the magnetization with "tot_magnetization",
you should not specify starting_magnetization.
         

kinetic energy cutoff (Ry) for wavefunctions
         

kinetic energy cutoff (Ry) for charge density and potential
For norm-conserving pseudopotential you should stick to the
default value, you can reduce it by a little but it will
introduce noise especially on forces and stress.
If there are ultrasoft PP, a larger value than the default is
often desirable (ecutrho = 8 to 12 times ecutwfc, typically).
PAW datasets can often be used at 4*ecutwfc, but it depends
on the shape of augmentation charge: testing is mandatory.
The use of gradient-corrected functional, especially in cells
with vacuum, or for pseudopotential without non-linear core
correction, usually requires an higher values of ecutrho
to be accurately converged.
         

kinetic energy cutoff (Ry) for the exact exchange operator in
EXX type calculations. By default this is the same as ecutrho
but in some EXX calculations significant speed-up can be found
by reducing ecutfock, at the expense of some loss in accuracy.
Currently only implemented for the optimized gamma point only
calculations.
         

if (.TRUE.) symmetry is not used. Note that
- if the k-point grid is provided in input, it is used "as is"
  and symmetry-inequivalent k-points are not generated;
- if the k-point grid is automatically generated, it will
  contain only points in the irreducible BZ for the bravais
  lattice, irrespective of the actual crystal symmetry.
A careful usage of this option can be advantageous
- in low-symmetry large cells, if you cannot afford a k-point
  grid with the correct symmetry
- in MD simulations
- in calculations for isolated atoms
         

if(.TRUE.) symmetry is not used but the k-points are
forced to have the symmetry of the Bravais lattice;
an automatically generated k-point grid will contain
all the k-points of the grid and the points rotated by
the symmetries of the Bravais lattice which are not in the
original grid. If available, time reversal is
used to reduce the k-points (and the q => -q symmetry
is used in the phonon code). To disable also this symmetry set
noinv=.TRUE..
         

if (.TRUE.) disable the usage of k => -k symmetry
(time reversal) in k-point generation
         

if (.TRUE.) disable the usage of magnetic symmetry operations
that consist in a rotation + time reversal.
         

if (.TRUE.) force the symmetry group to be symmorphic by disabling
symmetry operations having an associated fractionary translation
         

if (.TRUE.) do not discard symmetry operations with an
associated fractionary translation that does not send the
real-space FFT grid into itself. These operations are
incompatible with real-space symmetrization but not with the
new G-space symmetrization. BEWARE: do not use for phonons!
The phonon code still uses real-space symmetrization.
         

'smearing':     gaussian smearing for metals
                requires a value for degauss

'tetrahedra' :  especially suited for calculation of DOS
                (see P.E. Bloechl, PRB49, 16223 (1994))
                Requires uniform grid of k-points,
                automatically generated (see below)
                Not suitable (because not variational) for
                force/optimization/dynamics calculations

'fixed' :       for insulators with a gap

'from_input' :  The occupation are read from input file.
                Requires "nbnd" to be set in input
         

This flag is used for isolated atoms (nat=1) together with
occupations='from_input'. If it is .TRUE., the wavefunctions
are ordered as the atomic starting wavefunctions, independently
from their eigenvalue. The occupations indicate which atomic
states are filled.
The order of the states is written inside the UPF
pseudopotential file.
In the scalar relativistic case:
S -> l=0, m=0
P -> l=1, z, x, y
D -> l=2, r^2-3z^2, xz, yz, xy, x^2-y^2
In the noncollinear magnetic case (with or without spin-orbit),
each group of states is doubled. For instance:
P -> l=1, z, x, y for spin up, l=1, z, x, y for spin down.
Up and down is relative to the direction of the starting
magnetization.
In the case with spin-orbit and time-reversal
(starting_magnetization=0.0) the atomic wavefunctions are
radial functions multiplied by spin-angle functions.
For instance:
P -> l=1, j=1/2, m_j=-1/2,1/2. l=1, j=3/2,
     m_j=-3/2, -1/2, 1/2, 3/2.
In the magnetic case with spin-orbit the atomic wavefunctions
can be forced to be spin-angle functions by setting
starting_spin_angle to .TRUE..
         

In the spin-orbit case when domag=.TRUE., by default,
the starting wavefunctions are initialized as in scalar
relativistic noncollinear case without spin-orbit.
By setting starting_spin_angle=.TRUE. this behaviour can
be changed and the initial wavefunctions are radial
functions multiplied by spin-angle functions.
When domag=.FALSE. the initial wavefunctions are always
radial functions multiplied by spin-angle functions
independently from this flag.
When lspinorb is .FALSE. this flag is not used.
         

value of the gaussian spreading (Ry) for brillouin-zone
integration in metals.
         

'gaussian', 'gauss':
    ordinary Gaussian spreading (Default)

'methfessel-paxton', 'm-p', 'mp':
    Methfessel-Paxton first-order spreading
    (see PRB 40, 3616 (1989)).

'marzari-vanderbilt', 'cold', 'm-v', 'mv':
    Marzari-Vanderbilt cold smearing
    (see PRL 82, 3296 (1999))

'fermi-dirac', 'f-d', 'fd':
    smearing with Fermi-Dirac function
         

nspin = 1 :  non-polarized calculation (default)

nspin = 2 :  spin-polarized calculation, LSDA
             (magnetization along z axis)

nspin = 4 :  spin-polarized calculation, noncollinear
             (magnetization in generic direction)
             DO NOT specify nspin in this case;
             specify "noncolin=.TRUE." instead
         

if .true. the program will perform a noncollinear calculation.
         

ecfixed, qcutz, q2sigma:  parameters for modified functional to be
used in variable-cell molecular dynamics (or in stress calculation).
"ecfixed" is the value (in Rydberg) of the constant-cutoff;
"qcutz" and "q2sigma" are the height and the width (in Rydberg)
of the energy step for reciprocal vectors whose square modulus
is greater than "ecfixed". In the kinetic energy, G^2 is
replaced by G^2 + qcutz * (1 + erf ( (G^2 - ecfixed)/q2sigma) )
See: M. Bernasconi et al, J. Phys. Chem. Solids 56, 501 (1995)
         

Exchange-correlation functional: eg 'PBE', 'BLYP' etc
See Modules/functionals.f90 for allowed values.
Overrides the value read from pseudopotential files.
Use with care and if you know what you are doing!
         

Fraction of EXX for hybrid functional calculations. In the case of
input_dft='PBE0', the default value is 0.25, while for input_dft='B3LYP'
the exx_fraction default value is 0.20.
         

screening_parameter for HSE like hybrid functionals.
See J. Chem. Phys. 118, 8207 (2003)
and J. Chem. Phys. 124, 219906 (2006) for more informations.
         

Specific for EXX. It selects the kind of approach to be used
for treating the Coulomb potential divergencies at small q vectors.

gygi-baldereschi : appropriated for cubic and quasi-cubic supercells
vcut_spherical : appropriated for cubic and quasi-cubic supercells
vcut_ws : appropriated for strongly anysotropic supercells, see also
          ecutvcut.
none : sets Coulomb potential at G,q=0 to 0.0
         

Reciprocal space cutoff for correcting
Coulomb potential divergencies at small q vectors.
         

Specify lda_plus_u = .TRUE. to enable DFT+U calculations
See: Anisimov, Zaanen, and Andersen, PRB 44, 943 (1991);
     Anisimov et al., PRB 48, 16929 (1993);
     Liechtenstein, Anisimov, and Zaanen, PRB 52, R5467 (1994).
You must specify, for each species with a U term, the value of
U and (optionally) alpha, J of the Hubbard model (all in eV):
see lda_plus_u_kind, Hubbard_U, Hubbard_alpha, Hubbard_J
         

Specifies the type of DFT+U calculation:
                  0   simplified version of Cococcioni and de Gironcoli,
                      PRB 71, 035105 (2005), using Hubbard_U
                  1   rotationally invariant scheme of Liechtenstein et al.,
                      using Hubbard_U and Hubbard_J
         

Hubbard_U(i): U parameter (eV) for species i, DFT+U calculation
         

Hubbard_alpha(i) is the perturbation (on atom i, in eV)
used to compute U with the linear-response method of
Cococcioni and de Gironcoli, PRB 71, 35105 (2005)
(only for lda_plus_u_kind=0)
         

Hubbard_J(i,ityp): J parameters (eV) for species ityp,
used in DFT+U calculations (only for lda_plus_u_kind=1)
For p orbitals:  J = Hubbard_J(1,ityp);
For d orbitals:  J = Hubbard_J(1,ityp), B = Hubbard_J(2,ityp);
For f orbitals:  J = Hubbard_J(1,ityp), E2 = Hubbard_J(2,ityp),
                 E3= Hubbard_J(3,ityp).
If B or E2 or E3 are not specified or set to 0 they will be
calculated from J using atomic ratios.
         

In the first iteration of an DFT+U run it overwrites
the m-th eigenvalue of the ns occupation matrix for the
ispin component of atomic species I. Leave unchanged
eigenvalues that are not set. This is useful to suggest
the desired orbital occupations when the default choice
takes another path.
         

Only active when lda_plus_U is .true., specifies the type
of projector on localized orbital to be used in the DFT+U
scheme.

Currently available choices:
'atomic': use atomic wfc's (as they are) to build the projector

'ortho-atomic': use Lowdin orthogonalized atomic wfc's

'norm-atomic':  Lowdin normalization of atomic wfc. Keep in mind:
                atomic wfc are not orthogonalized in this case.
                This is a "quick and dirty" trick to be used when
                atomic wfc from the pseudopotential are not
                normalized (and thus produce occupation whose
                value exceeds unity). If orthogonalized wfc are
                not needed always try 'atomic' first.

'file':         use the information from file "prefix".atwfc that must
                have been generated previously, for instance by pmw.x
                (see PP/poormanwannier.f90 for details)

NB: forces and stress currently implemented only for the
'atomic' choice.
         

The direction of the electric field or dipole correction is
parallel to the bg(:,edir) reciprocal lattice vector, so the
potential is constant in planes defined by FFT grid points;
edir = 1, 2 or 3. Used only if tefield is .TRUE.
         

Position of the maximum of the sawlike potential along crystal
axis "edir", within the  unit cell (see below), 0 < emaxpos < 1
Used only if tefield is .TRUE.
         

Zone in the unit cell where the sawlike potential decreases.
( see below, 0 < eopreg < 1 ). Used only if tefield is .TRUE.
         

Amplitude of the electric field, in ***Hartree*** a.u.;
1 a.u. = 51.4220632*10^10 V/m). Used only if tefield=.TRUE.
The sawlike potential increases with slope "eamp" in the
region from (emaxpos+eopreg-1) to (emaxpos), then decreases
to 0 until (emaxpos+eopreg), in units of the crystal
vector "edir". Important: the change of slope of this
potential must be located in the empty region, or else
unphysical forces will result.
         

The angle expressed in degrees between the initial
magnetization and the z-axis. For noncollinear calculations
only; index i runs over the atom types.
         

The angle expressed in degrees between the projection
of the initial magnetization on x-y plane and the x-axis.
For noncollinear calculations only.
         

Used to perform constrained calculations in magnetic systems.
Currently available choices:

'none':
         no constraint

'total':
         total magnetization is constrained by
         adding a penalty functional to the total energy:

         LAMBDA * SUM_{i} ( magnetization(i) - fixed_magnetization(i) )**2

         where the sum over i runs over the three components of
         the magnetization. Lambda is a real number (see below).
         Noncolinear case only. Use "tot_magnetization" for LSDA

'atomic':
         atomic magnetization are constrained to the defined
         starting magnetization adding a penalty:

         LAMBDA * SUM_{i,itype} ( magnetic_moment(i,itype) - mcons(i,itype) )**2

         where i runs over the cartesian components (or just z
         in the collinear case) and itype over the types (1-ntype).
         mcons(:,:) array is defined from starting_magnetization,
         (and angle1, angle2 in the non-collinear case). lambda is
         a real number

'total direction':
          the angle theta of the total magnetization
          with the z axis (theta = fixed_magnetization(3))
          is constrained:

          LAMBDA * ( arccos(magnetization(3)/mag_tot) - theta )**2

          where mag_tot is the modulus of the total magnetization.

'atomic direction':
          not all the components of the atomic
          magnetic moment are constrained but only the cosine
          of angle1, and the penalty functional is:

          LAMBDA * SUM_{itype} ( mag_mom(3,itype)/mag_mom_tot - cos(angle1(ityp)) )**2

N.B.: symmetrization may prevent to reach the desired orientation
      of the magnetization. Try not to start with very highly symmetric
      configurations or use the nosym flag (only as a last remedy)
         

value of the total magnetization to be maintained fixed when
constrained_magnetization='total'
         

parameter used for constrained_magnetization calculations
N.B.: if the scf calculation does not converge, try to reduce lambda
      to obtain convergence, then restart the run with a larger lambda
         

It is the number of iterations after which the program
write all the atomic magnetic moments.
         

if .TRUE. the noncollinear code can use a pseudopotential with
spin-orbit.
         

Used to perform calculation assuming the system to be
isolated (a molecule of a clustr in a 3D supercell).

Currently available choices:

'none' (default): regular periodic calculation w/o any correction.

'makov-payne', 'm-p', 'mp' : the Makov-Payne correction to the
         total energy is computed. An estimate of the vacuum
         level is also calculated so that eigenvalues can be
         properly aligned.
         Theory:
         G.Makov, and M.C.Payne,
         "Periodic boundary conditions in ab initio
         calculations" , Phys.Rev.B 51, 4014 (1995)

'dcc' :  density counter charge correction CURRENTLY DISABLED
         The electrostatic problem is solved in open boundary
         conditions (OBC). This approach provides the correct
         scf potential and energies (not just a correction to
         energies as 'mp'). BEWARE: the molecule should be
         centered around the middle of the cell, not around
         the origin (0,0,0).
         Theory described in:
         I.Dabo, B.Kozinsky, N.E.Singh-Miller and N.Marzari,
         "Electrostatic periodic boundary conditions and
         real-space corrections", Phys.Rev.B 77, 115139 (2008)

'martyna-tuckerman', 'm-t', 'mt' : Martyna-Tuckerman correction.
         As for the dcc correction the scf potential is also
         corrected. Implementation adapted from:
         G.J. Martyna, and M.E. Tuckerman,
         "A reciprocal space based method for treating long
         range interactions in ab-initio and force-field-based
         calculation in clusters", J.Chem.Phys. 110, 2810 (1999)

'esm' :  Effective Screening Medium Method.
         For polarized or charged slab calculation, embeds
         the simulation cell within an effective semi-
         infinite medium in the perpendicular direction
         (along z). Embedding regions can be vacuum or
         semi-infinite metal electrodes (use 'esm_bc' to
         choose boundary conditions). If between two
         electrodes, an optional electric field
         ('esm_efield') may be applied. Method described in
         M. Otani and O. Sugino, "First-principles
         calculations of charged surfaces and interfaces:
         A plane-wave nonrepeated slab approach," PRB 73,
         115407 (2006).
         NB: Requires cell with a_3 lattice vector along z,
         normal to the xy plane, with the slab centered
         around z=0. Also requires symmetry checking to be
         disabled along z, either by setting 'nosym' = .TRUE.
         or by very slight displacement (i.e., 5e-4 a.u.)
         of the slab along z.
         See 'esm_bc', 'esm_efield', 'esm_w', 'esm_nfit'.
         

If assume_isolated = 'esm', determines the boundary
conditions used for either side of the slab.

Currently available choices:

'pbc' (default): regular periodic calculation (no ESM).

'bc1' : Vacuum-slab-vacuum (open boundary conditions)

'bc2' : Metal-slab-metal (dual electrode configuration).
        See also 'esm_efield'.

'bc3' : Vacuum-slab-metal
         

If assume_isolated = 'esm', determines the position offset
[in a.u.] of the start of the effective screening region,
measured relative to the cell edge. (ESM region begins at
z = +/- [L_z/2 + esm_w] ).
         

If assume_isolated = 'esm' and esm_bc = 'bc2', gives the
magnitude of the electric field [Ryd/a.u.] to be applied
between semi-infinite ESM electrodes.
         

If assume_isolated = 'esm', gives the number of z-grid points
for the polynomial fit along the cell edge.
         

if .TRUE. compute semi-empirical dispersion term (DFT-D).
See S. Grimme, J. Comp. Chem. 27, 1787 (2006), and
V. Barone et al., J. Comp. Chem. 30, 934 (2009).
         

global scaling parameter for DFT-D. Default is good for PBE.
         

cutoff radius (a.u.) for dispersion interactions
         

maximum number of iterations in a scf step
         

Convergence threshold for selfconsistency:
estimated energy error < conv_thr
(note that conv_thr is extensive, like the total energy)
         

If .TRUE. this turns on the use of an adaptive conv_thr for
the inner scf loops when using EXX.
         

When adaptive_thr = .TRUE. this is the convergence threshold
used for the first scf cycle.
         

When adaptive_thr = .TRUE. the convergence threshold for
each scf cycle is given by:
min( conv_thr, conv_thr_multi * dexx )
         

'plain' :    charge density Broyden mixing

'TF' :       as above, with simple Thomas-Fermi screening
            (for highly homogeneous systems)

'local-TF':  as above, with local-density-dependent TF screening
             (for highly inhomogeneous systems)
         

mixing factor for self-consistency
         

number of iterations used in mixing scheme
         

For DFT+U : number of iterations with fixed ns ( ns is the
  atomic density appearing in the Hubbard term ).
         

'david':  Davidson iterative diagonalization with overlap matrix
          (default). Fast, may in some rare cases fail.

'cg' :    conjugate-gradient-like band-by-band diagonalization
          Typically slower than 'david' but it uses less memory
          and is more robust (it seldom fails)

'cg-serial', 'david-serial': obsolete, use "-ndiag 1 instead"
          The subspace diagonalization in Davidson is performed
          by a fully distributed-memory parallel algorithm on
          4 or more processors, by default. The allocated memory
          scales down with the number of procs. Procs involved
          in diagonalization can be changed with command-line
          option "-ndiag N". On multicore CPUs it is often
          convenient to let just one core per CPU to work
          on linear algebra.
         

Convergence threshold for the first iterative diagonalization
(the check is on eigenvalue convergence).
For scf calculations, the default is 1.D-2 if starting from a
superposition of atomic orbitals; 1.D-5 if starting from a
charge density. During self consistency the threshold (ethr)
is automatically reduced when approaching convergence.
For non-scf calculations, this is the threshold used in the
iterative diagonalization. The default is conv_thr /N elec.
         

For conjugate gradient diagonalization:
max number of iterations
         

For Davidson diagonalization: dimension of workspace
(number of wavefunction packets, at least 2 needed).
A larger value may yield a somewhat faster algorithm
but uses more memory. The opposite holds for smaller values.
Try diago_david_ndim=2 if you are tight on memory or if
your job is large: the speed penalty is often negligible
         

If .TRUE. all the empty states are diagonalized at the same level
of accuracy of the occupied ones. Otherwise the empty states are
diagonalized using a larger threshold (this should not affect
total energy, forces, and other ground-state properties).
         

Amplitude of the finite electric field (in Ry a.u.;
1 a.u. = 36.3609*10^10 V/m). Used only if lelfield=.TRUE.
and if k-points (K_POINTS card) are not automatic.
         

Finite electric field (in Ry a.u.=36.3609*10^10 V/m) in
cartesian axis. Used only if lelfield=.TRUE. and if
k-points (K_POINTS card) are automatic.
         

'atomic': starting potential from atomic charge superposition
          ( default for scf, *relax, *md )

'file'  : start from existing "charge-density.xml" file
          For nscf and bands alculation this is the default
          and the only sensible possibility.
         

'atomic': start from superposition of atomic orbitals
          If not enough atomic orbitals are available,
          fill with random numbers the remaining wfcs
          The scf typically starts better with this option,
          but in some high-symmetry cases one can "loose"
          valence states, ending up in the wrong ground state.

'atomic+random': as above, plus a superimposed "randomization"
          of atomic orbitals. Prevents the "loss" of states
          mentioned above.

'random': start from random wfcs. Slower start of scf but safe.
          It may also reduce memory usage in conjunction with
          diagonalization='cg'

'file':   start from a wavefunction file
         

If .true., use the real-space algorithm for augmentation
charges in ultrasoft pseudopotentials.
Must faster execution of ultrasoft-related calculations,
but numerically less accurate than the default algorithm.
Use with care and after testing!
         

Specify the type of ionic dynamics.

For different type of calculation different possibilities are
allowed and different default values apply:

CASE ( calculation = 'relax' )
    'bfgs' :   (default)   use BFGS quasi-newton algorithm,
                           based on the trust radius procedure,
                           for structural relaxation
    'damp' :               use damped (quick-min Verlet)
                           dynamics for structural relaxation
                           Can be used for constrained
                           optimisation: see CONSTRAINTS card

CASE ( calculation = 'md' )
    'verlet' : (default)   use Verlet algorithm to integrate
                           Newton's equation. For constrained
                           dynamics, see CONSTRAINTS card
    'langevin'             ion dynamics is over-damped Langevin

CASE ( calculation = 'vc-relax' )
    'bfgs' :   (default)   use BFGS quasi-newton algorithm;
                           cell_dynamics must be 'bfgs' too
    'damp' :               use damped (Beeman) dynamics for
                           structural relaxation
CASE ( calculation = 'vc-md' )
    'beeman' : (default)   use Beeman algorithm to integrate
                           Newton's equation
         

'default '  : if restarting, use atomic positions read from the
              restart file; in all other cases, use atomic
              positions from standard input.

'from_input' : restart the simulation with atomic positions read
              from standard input, even if restarting.
         

'full' :           the full phase-space is used for the ionic
                   dynamics.

'coarse-grained' : a coarse-grained phase-space, defined by a set
                   of constraints, is used for the ionic dynamics
                   (used for calculation of free-energy barriers)
         

   Used to extrapolate the potential from preceding ionic steps.

   'none'        :  no extrapolation

   'atomic'      :  extrapolate the potential as if it was a sum of
                    atomic-like orbitals

   'first_order' :  extrapolate the potential with first-order
                    formula

   'second_order':  as above, with second order formula

Note: 'first_order' and 'second-order' extrapolation make sense
only for molecular dynamics calculations
         

    Used to extrapolate the wavefunctions from preceding ionic steps.

   'none'        :  no extrapolation

   'first_order' :  extrapolate the wave-functions with first-order
                    formula.

   'second_order':  as above, with second order formula.

Note: 'first_order' and 'second-order' extrapolation make sense
only for molecular dynamics calculations
         

This keyword is useful when simulating the dynamics and/or the
thermodynamics of an isolated system. If set to true the total
torque of the internal forces is set to zero by adding new forces
that compensate the spurious interaction with the periodic
images. This allows for the use of smaller supercells.

BEWARE: since the potential energy is no longer consistent with
the forces (it still contains the spurious interaction with the
repeated images), the total energy is not conserved anymore.
However the dynamical and thermodynamical properties should be
in closer agreement with those of an isolated system.
Also the final energy of a structural relaxation will be higher,
but the relaxation itself should be faster.
         

Specify the type of dynamics for the cell.
For different type of calculation different possibilities
are allowed and different default values apply:

CASE ( calculation = 'vc-relax' )
  'none':    no dynamics
  'sd':      steepest descent ( not implemented )
  'damp-pr': damped (Beeman) dynamics of the Parrinello-Rahman
             extended lagrangian
  'damp-w':  damped (Beeman) dynamics of the new Wentzcovitch
             extended lagrangian
  'bfgs':    BFGS quasi-newton algorithm (default)
             ion_dynamics must be 'bfgs' too
CASE ( calculation = 'vc-md' )
  'none':    no dynamics
  'pr':      (Beeman) molecular dynamics of the Parrinello-Rahman
             extended lagrangian
  'w':       (Beeman) molecular dynamics of the new Wentzcovitch
             extended lagrangian
         

Target pressure [KBar] in a variable-cell md or relaxation run.
         

Fictitious cell mass [amu] for variable-cell simulations
(both 'vc-md' and 'vc-relax')
         

Used in the construction of the pseudopotential tables.
It should exceed the maximum linear contraction of the
cell during a simulation.
         

Convergence threshold on the pressure for variable cell
relaxation ('vc-relax' : note that the other convergence
thresholds for ionic relaxation apply as well).
         

Select which of the cell parameters should be moved:

all     = all axis and angles are moved
x       = only the x component of axis 1 (v1_x) is moved
y       = only the y component of axis 2 (v2_y) is moved
z       = only the z component of axis 3 (v3_z) is moved
xy      = only v1_x and v_2y are moved
xz      = only v1_x and v_3z are moved
yz      = only v2_x and v_3z are moved
xyz     = only v1_x, v2_x, v_3z are moved
shape   = all axis and angles, keeping the volume fixed

BEWARE: if axis are not orthogonal, some of these options do not
 work (symmetry is broken). If you are not happy with them,
 edit subroutine init_dofree in file Module/cell_base.f90