XSPECTRA
---------
2009: First version of XSpectra
by Christos Gougoussis, Matteo Calandra, Ari P. Seitsonen and Francesco Mauri
2014: Restyling of I/O,
by Delphine Cabaret and Nadejda Mas
2015: L23 edge XAS calculation by O. Bunau and M. Calandra
-----------------------------------------------------------------------
The theoretical approach on which XSpectra is based was
described in:
L23 edges,
O. Bunau and M. Calandra
Projector augmented wave calculation of x-ray absorption spectra at the L2,3 edges
Phys. Rev. B 87, 205105 (2013)
K/L1-edge,
C. Gougoussis, M. Calandra, A. P. Seitsonen, F. Mauri,
"First principles calculations of X-ray absorption in an ultrasoft
pseudopotentials scheme: from $\alpha$-quartz to high-T$_c$ compounds",
Phys. Rev. B 80, 075102 (2009)
M. Taillefumier, D. Cabaret, A. M. Flank, and F. Mauri
"X-ray absorption near-edge structure calculations with the pseudopotentials:
Application to the K edge in diamond and αalpha-quartz"
Phys. Rev. B 66, 195107 (2002)
You should cite these three works in all publications using this software.
The implementation of the DFT+U approximation and its application to
K-edge XAS in NiO was performed in:
C. Gougoussis, M. Calandra, A. Seitsonen, Ch. Brouder, A. Shukla, F. Mauri
" Intrinsic charge transfer gap in NiO from Ni K -edge x-ray absorption spectroscopy",
Phys. Rev. B 79, 045118 (2009)
If you use DFT+U, you should cite this work too.
Finally you should cite properly the Quantum ESPRESSO distribution.
-----------------------------------------------------------------------
XSpectra is a post-processing tools that relies on the output
(the charge density) of the PWscf code (pw.x).
Thus a scf calculation needs to be done before running
xspectra.x.
To simulate core-hole effects, a pseudopotential with a hole in the s
state (1s for K-edges, 2s for L1-edges, 2p1/2 for L2-edges, 2p3/2 for
L3-edges) needs to be generated for the absorbing atom.
Some of these pseudopotentials are available
in the XSpectra examples directory, some others are available on
the pseudopotential web-page at www.quantum-espresso.org/ with the
label "*star1s*_gipaw*" for K-edges, "*star2s*_gipaw*" for L1-edges and so on.
The self-consistent calculation is then performed on a supercell including
the absorbing atom. The size of the supercell needs to be verified from
system to system, since fairly large supercells are necessary for convergence.
If core-hole effects need not to be taken into account then a calculation on
a single cell with a standard pseudopotential (i.e. without the core-hole)
is enough.
Since xspectra.x uses GIPAW reconstruction of the all electron wavefunction
the pseudopotential needs to contain information about GIPAW reconstruction.
There is no limit to the number of GIPAW projector that can be included.
Note however that at least two projectors are needed to obtain XAS spectra
converged up to 30-40 eV from the Fermi level.
The use of a single projector is discouraged, particularly when semicore
states are present. If more than two projectors are used, linear independence
of the projectors should be explicitly verified (verbosity='high').
Once the scf charge density has been obtained, the xspectra.x code can be
used as a post-processing tool. Note that the X-ray absorption spectra
can be calculated on a larger mesh, different from that used in the
PWscf scf run. Convergence need to be tested also for this second mesh.
XSpectra calculates then the XAS electric dipole (for K and L edges)
or electric quadrupole contributions (for K and L1 edges only),
using the Lanczos method and the continued fraction.
This approach does not require the explicit calculation of empty states
and it is consequently very fast (only the charge density is needed).
The code needs the radial core wavefunction of the initial core state
in input. This wavefunction is included in the pseudo
and can be extracted using the script upf2plotcore.sh
in the directory ~/espresso/XSpectra/tools/ .
Note that this script works only for UPF version 1.
This is necessary to calculate the XAS matrix element.
The output spectrum can be separated in its spin-up and
spin-down polarizations.
DFT+U calculations and collinear magnetism are possible.
Ultrasoft pseudopotentials are allowed.
Hybrid functionals not yet allowed.
--------------------------------------------------------------------------
=======================================================================
NAMELIST / input_xspectra /
calculation character (len=8) DEFAULT=''
'xanes_dipole', Perform dipolar calculation
'xanes_quadrupole', Perform quadrupolar
calculation
'hpsi', Perform the test H*psi=E*Psi (debug option)
edge character (len=16) DEFAULT='K',
specifiy the edge to be calculated.
'K' specify the standard K-edge calculation
'L2' calculates the L2 edge,
'L3' calculates the L3 edge,
'L23' calculates both.
However, it should be noted that in the single particle
approximation the L3/L2 branching ration is exactly equal two 2.
Thus a calculation of one of the edges is enough.
lplus logical DEFAULT=.false.
if lpus=.true. only transition 2p ---> d are allowed
in the dipolar cross section for L23 edges.
lminus logical DEFAULT=.false.
if lminus=.true. only transition 2p ---> s are allowed
in the dipolar cross section for L23 edges.
prefix character (len=256)
prefix of the pwscf output files
outdir character (len=256) DEFAULT='./'
directory tmp_dir or where the pwscf output files are stored
verbosity character (len=4) DEFAULT='low'
'high',it checks linear dependence of PAW projectors
and write details about the projectors.
Note that GIPAW already perform a check on the linear
dependence of the projectors even without this option.
xiabs integer DEFAULT=1
type of the absorbing atom:
position within the ATOMIC_SPECIES in pwscf input
xkvec(1:3) real(DP) DEFAULT=(1.0,0.0,0.0)
coordinates of the X-ray wave-vector k
xepsilon(1:3) real(DP) DEFAULT=(0.0,0.0,1.0)
coordinates of the incident X-ray polarization vector
xcoordcrys logical DEFAULT=.true.
.true. to use crystal coordinates for xkvec and xepsilon
xonly_plot logical DEFAULT=.false.
.false. the continued fraction is calculated
for each k-point and at the end written
on the save file
.true. uses a previously calculated continued
fraction (x_save_file) to re-plot the
spectrum with different parameters
(gamma broadening parameter,with occupied state,etc.)
x_save_file character (len=256) DEFAULT=xanes.sav
save file where results of the Lanczos
calculation (a,b coefficients, etc.) are written
If xonly_plot=.true., the x_save_file is read (read only)
to get the Lanczos parameters calculated in a previous run
Current version number is 2
ef_r (obsolete use xe0)
The Fermi energy is determined from the SCF save directory.
For an insulator, it is set to the energy of the highest occupied level.
If the calculation is spin polarized, the largest
of the Fermi energies corresponding to spin up and down is kept.
If the zero of the spectrum needs to be changed, use xe0 (see below).
xe0 real(DP) DEFAULT=1.d4
energy-zero for the spectrum in eV
- must be set to the Fermi level if xonly_plot is .true.
and the version of x_save_file is 1
(written with a previous version of the code).
If x_save_file is 2 and xe0 is not specified then the zero
energy of the spectrum is set at the Fermi level.
- can also be used to set the zero energy for the calculation
of the spectrum at an other value than the Fermi energy
(for example, for an insulator, in the middle of the gap)
xniter integer DEFAULT=2000
maximum number of iterations for Lanczos.
The maximum number of iterations allowed must be lower
than the number of vectors in the Hilbert space
(i.e. the number of plane waves).
xcheck_conv integer DEFAULT=5
number of iterations between 2 convergence tests:
Xspectra checks convergency of the spectrum every
xcheck_conv iterations of the Lanczos-basis construction.
xerror real(DP) DEFAULT=0.01
convergence threshold for Lanczos calculation (eV)
If the difference of two successive spectra
(for a given k-point) is smaller than xerror,
the convergence is achieved.
show_status logical DEFAULT=.false.
flag to show the status of the code
U_projection_type character(len=16) DEFAULT='atomic'
type of projection for DFT+U calculations
(see the PWscf input file for more info)
wf_collect logical DEFAULT=.false.
must be true if wf_collect is enabled in the scf calculation
time_limit integer DEFAULT=1.d8
time in seconds before stopping the calculation.
If XSpectra stops because of the time limit,
a and b coefficients of the incomplete continued fraction
are stored in the .sav file.
restart_mode character (len=32) DEFAULT='from_scratch'
'restart' if you want to restart from a .sav file
where a and b coefficients of an incomplete
continued fraction are stored.
===============================================================================
NAMELIST / plot /
xnepoint integer DEFAULT=100
number of energy points in the plot of the XAS spectrum
xemax real(DP) DEFAULT=10.0
maximum energy in eV for the plot of the XAS spectrum
xemin real(DP) DEFAULT=0.0
minimum energy in eV for the plot of the
XAS spectrum
cut_occ_states logical DEFAULT=.false.
.false. the occupied states are visualized
.true. the occupied states are smoothly cut
from the plot
terminator logical DEFAULT=.false.
.true. to use the terminator function for the
continued fraction
.false. no terminator is used.
gamma_mode character (len=256) DEFAULT='constant'
'constant': a constant broadening parameter (xgamma)
is used for the XAS spectrum.
'variable': an energy-dependent broadening parameter
is used: constant and equal to gamma_value(1)
from xemin to gamma_energy(1),
constant and equal to gamma_value(2)
from gamma_energy(2) to xemax and linear
from gamma_energy(1) to gamma_energy(2).
'file': the continued fraction uses an energy-dependent
broadening parameter stored in file gamma_file.
The broadening parameter (gamma of the continued fraction)
is equivalent to the half width at half maximum
of a Lorentzian used for a convolution
xgamma real(DP) DEFAULT=0.1
constant broadening parameter to be used in the spectrum (eV)
used for convergence and if gamma_mode='constant'
gamma_energy(1:2) real(DP)
energy values in eV of the 2 points of reference for variable gamma
used if gamma_mode='variable'
gamma_value(1:2) real(DP)
gamma values in eV of the 2 points of reference
used if gamma_mode='variable'
gamma_file character (LEN=256) DEFAULT='gamma.dat'
The file has to be formatted in two columns :
energy1 gamma1
energy2 gamma2
where at energy1 the broadening parameter is gamma1.
used if gamma_mode='file'
==============================================================================
NAMELIST / pseudos /
filecore, character (len=256) DEFAULT='Core.wfc'
core wavefunction file
r_paw(1:...) real(DP) DEFAULT=1.5*rc
paw radii to be used in paw reconstruction.
rpaw(1) corresponds to l quantum number=1 (electric dipole)
rpaw(2) corresponds to l quantum number=2 (electric quadrupole)
==============================================================================
In order to cut the occupied states, the program performs an integration
over the variable t in ] 0, infinity [.
For more details see ref.
Ch. Brouder, M. Alouani, K. H. Bennemann, Phys. Rev. B 54 (1996) p.7334-49.
The integration is done with t going in two opposite directions,
from the start value cut_startt. So, the integration
is done over ]cut_tinf,cut_startt] at least with step cut_stepl, and
over [cut_startt,cut_tsup[ at least with step cut_stepu.
There are two arrays of size cut_nmeml and cut_nmemu
in order to save Green functions values. There is an area near
the Fermi level of size cut_desmooth (in eV) where the cross section
is interpolated in order to avoid a divergence.
NAMELIST / cut_occ /
cut_ierror real(DP) DEFAULT=1.d-7
convergence tolerance for one step in the integral
cut_stepu real(DP) DEFAULT=1.d-2
integration initial step, upper side
cut_stepl real(DP) DEFAULT=1.d-3
integration initial step, lower side
cut_startt real(DP) DEFAULT=1.d0
integration start value of the t variable
cut_tinf real(DP) DEFAULT=1.d-6
maximum value of the lower integration boundary
cut_tsup real(DP) DEFAULT=100.d0
minimum value of the upper integration boundary
cut_desmooth real(DP) DEFAULT=1.d-2
size of the interval near the fermi energy
in which cross section is smoothed
cut_nmemu integer DEFAULT=100000
size of the memory of the values of the
green function, upper side
cut_nmeml integer DEFAULT=100000
size of the memory of the values of the
green function, lower side
=================================================================================