model=sqw_phonons(configuration, calculator, ..., options)
iFunc/sqw_phonons: computes phonon dispersions using the ASE.
A model which computes phonon dispersions from the Hellmann-Feynman forces acting
between atoms. The input argument is any configuration file describing the
material, e.g. CIF, PDB, POSCAR, ... supported by ASE.
This models can compute the coherent inelastic phonons dispersions for
any crystalline (powder or single crystal) material, in the harmonic and
adiabatic approxiimation.
The phonon spectra is computed using one of the calculator supported by the
Atomic Simulation Environment (ASE) <https://wiki.fysik.dtu.dk/ase>.
Supported calculators are:
ABINIT Plane-wave pseudopotential code
Dacapo Plane-wave ultra-soft pseudopotential code
ELK Full Potential LAPW code
EMT Effective Medium Theory calculator (Al,Cu,Ag,Au,Ni,Pd,Pt,H,C,N,O)
GPAW Real-space/plane-wave/LCAO PAW code
NWChem Gaussian based electronic structure code
QuantumEspresso_ASE Plane-wave pseudopotential code (with ASE/QE-util)
QuantumEspresso Plane-wave pseudopotential code (with PHON)
VASP Plane-wave PAW code (when installed and licensed)
We recommend QuantumEspresso, GPAW, ABINIT and Elk.
Refer to https://molmod.ugent.be/deltacodesdft for codes comparison.
The simplest usage is to call: sqw_phonons('gui') which displays an entry dialog box
and proceeds with a fully automatic computation, and plots final results.
The calculators can be specified by just giving their name as a parameter,
or using e.g. options.calculator='GPAW'. Except for EMT, other calculators must
be installed separately, and optionally specific pseudo-potentials.
See https://wiki.fysik.dtu.dk/ase/ase/calculators/calculators.html
Benchmarks indicate that, for phonon dispersions:
* QuantumEspresso/PHON is the fastest, with excellent parallelization and accuracy.
* QuantumEspresso/ASE is excellent.
* ABINIT, VASP are also fast codes.
* The all-electrons Elk code is about 10 times slower than QuantumEspresso.
* Using k-points=3 is both fast and ensures a reasonable accuracy in phonons
* Phonon dispersions are not too sensitive on the energy cut-off. 340 eV is good.
* Elk and ABINIT can not handle large supercells without recompiling.
* NWChem and Elk are not sensitive to the energy cut-off.
The model must first be created (which triggers e.g. a DFT computation), and
once created, it can be evaluated in the whole HKLE reciprocal space.
MODEL CREATION ===============================================================
sqw_phonons(configuration, calculator, options, ...)
is the usual syntax, such as sqw_phonons('POSCAR','qe','metal')
sqw_phonons('gui')
shows a dialogue to select the phonon calculator configuration.
sqw_phonons('defaults')
uses a simple fcc Aluminium as example.
sqw_phonons(dir, calculator, options, ...)
re-use an existing directory containing any of POSCAR, ASE atom object, PhonoPy
files, ...
The arguments for the model creation should be:
input (model creation):
configuration: file name or directory to an existing material configuration
Any A.S.E supported format can be used (POSCAR, CIF, SHELX, PDB, ...).
See <https://wiki.fysik.dtu.dk/ase/ase/io.html#module-ase.io>
Alternatively, the 'bulk','molecule', 'crystal', and 'nanotube' ASE constructors can be
used, using the Python syntax, e.g.
'bulk("Si", "diamond", a=5.4)'
'bulk("Cu", "fcc", a=3.6, cubic=True)'
'molecule("H2O")'
'nanotube(6, 0, length=4)'
'crystal(["Na", "Cl"], [(0, 0, 0), (0.5, 0.5, 0.5)], spacegroup=225,
cellpar=[5.64, 5.64, 5.64, 90, 90, 90])'
See <https://wiki.fysik.dtu.dk/ase/ase/structure.html>
<https://wiki.fysik.dtu.dk/ase/tutorials/spacegroup/spacegroup.html>
'metal','insulator','semiconductor': indicates the type of occupation for
electronic states, which sets smearing.
'html' or 'report': generate a full report of the simulation and export data.
'plot' or 'autoplot': plot the results after building the model.
options: an optional structure with optional settings, as follows:
General options
options.target =path Where to store all files and FORCES
a temporary directory is created when not specified.
options.supercell=scalar or [nx ny nz] supercell size. Default is 0 (auto mode).
options.calculator=string EMT,GPAW,Elk,NWChem,Dacapo,ABINIT,Quantum
We recommend ABINIT,QE and Elk. Default set from installed software.
options.potentials=string Basis datasets or pseudopotentials.
NWChem: see http://www.nwchem-sw.org/index.php/Release64:AvailableBasisSets
e.g. 'ANO-RCC'
Dacapo: path to potentials, e.g. /usr/share/dacapo-psp (DACAPOPATH)
Elk: path to potentials, e.g. /usr/share/elk-lapw/species (ELK_SPECIES_PATH)
ABINIT: path to potentials, e.g. /usr/share/abinit/psp/ (ABINIT_PP_PATH)
or any of 'NC' 'fhi' 'hgh' 'hgh.sc' 'hgh.k' 'tm' 'paw'
QuantumEspresso: path to potentials, e.g. /usr/share/espresso/pseudo
options.command='exe' Path to calculator executable.
options.mpi=scalar use multi-cpu, requires e.g. OpenMPI to be installed.
when available and not specified, MPI is used with all CPU's.
options.machinefile=filename file containing the list of MPI machines to use
options.autoplot=0|1 when set, automatically evaluates the Model
and plot results.
options.htmlreport=0|1 when set, automatically generates a full
report on the computation results.
options.optimizer when set, performs a geometry optimization
before computing the forces (move atoms in the cell). Can be set to
'BFGS' or 'LBFGS' (low memory BFGS)
'MDMin' or 'FIRE'
options.accuracy 'fast', 'very fast' (default) or 'accurate'.
The 'fast' choice uses the symmetry operators to lower the number of atom
displacements. The force gradient uses central difference.
The 'very fast' option halves the number of displacements. The equilibrium
forces are used to improve the accuracy of the force gradient.
The 'accurate' choice is longer to execute (e.g. 3-6 times slower), but computes all forces.
options.use_phonopy=0|1 requests to use PhonoPy when installed.
This choice also sets accuracy='very fast'
options.disp=value the atom displacement in Angs. Default is 0.01.
DFT specific options
options.kpoints=scalar or [nx ny nz] Monkhorst-Pack grid. Default is 0 (auto mode).
options.xc=string Type of Exchange-Correlation functional to use
'LDA','PBE','revPBE','RPBE','PBE0','B3LYP' for GPAW
'PZ', 'PBE','revPBE','RPBE','PW91','VWN' for Dacapo/Jacapo
'LDA','PBE','revPBE','RPBE','GGA' for ABINIT
'LDA','PBE','REVPBE','PBESOL','WC06','AM05' for ELK
'LDA','PBE','RHF','MP2','B3LYP' for NWChem
'LDA','PBE','PW91' for VASP
Default is 'PBE'.
options.raw='name=value, ...' Additional arguments passed to the ASE
using the Python syntax.
options.nbands=int Number of valence bands for which
wavefunctions are being computed. (= # of electrons /2); for a metal, 20% more
(minimum 4 more).
options.ecut=scalar Kinetic energy cutoff for wavefunctions.
Large value improves convergence. Estimate is 200*n_typ_atoms in [eV].
Usually one runs at calculations at various ecut to investigate convergence
ABINIT paw/pawxml potentials require larger ecut (1000 eV) and nsteps (100).
options.diagonalization='dav' or 'cg' or 'rmm-diis'
The Davidson method is faster than the conjugate-gradient but uses more
memory. The RMM-DIIS method allows parallelization.
options.occupations='metal' for metals ('smearing') help converge
'insulator' for insulators
'semiconductor' sets 0 eV FermiDirac smearing
'auto' for Elk (automatic smearing)
or 0 for semi-conductors
or a value in eV for the distribution width e.g. 0.3
options.nsteps=scalar Max number of iterations for SCF.
Typical: 30. Large value improves convergence.
options.toldfe=scalar Convergence threshold on the energy for
selfconsistency, e.g. 1e-8 [eV].
Options specific per calculator
options.mode='pw','fd', or 'lcao' GPAW computation mode as Plane-Wave,
Finite Difference, or LCAO (linear combination of atomic orbitals). Default is 'pw'.
options.iscf='NC','PAW' Type of SCF cycles (ABINIT)
options.pps = 'fhi' 'hgh' 'hgh.sc' 'hgh.k' 'tm' 'paw' 'pawxml' Type of database (ABINIT)
a structure such as struct('Ca','_pv') ... (VASP)
options.mixing_beta=scalar mixing factor for self-consistency
default=0.7. use 0.3 to improve convergence (QuantumEspresso)
options.mixing_ndim=scalar number of iterations used in mixing
default=8. If you are tight with memory, you may reduce it to 4. A larger
value will improve the SCF convergence, and use more memory (QuantumEspresso)
The options can also be entered as strings with 'field=value; ...'.
output (model creation):
model: iFunc 4D model S(qh,qk,ql,w) with parameters p.
The syntax sqw_phonons(model,'html') allows to re-create the HTML report about
the 4D phonon model.
You may look at the following resources to get material structure files:
<http://phonondb.mtl.kyoto-u.ac.jp/>
<https://www.materialsproject.org/>
<http://nomad-repository.eu/cms/>
WARNING:
This model is suitable to compute phonon dispersions for e.g solid-
state materials.
The Atomic Simulation Environment must be installed.
The temporary directories (UserData.dir) are not removed.
Momentum is expressed in reduced lattice unit, and energy is in meV.
1 meV = 241.8 GHz = 11.604 K = 0.0965 kJ/mol = 8.0657 cm-1
References: https://en.wikipedia.org/wiki/Phonon
Atomic Simulation Environment
S. R. Bahn and K. W. Jacobsen, Comput. Sci. Eng., Vol. 4, 56-66, 2002
<https://wiki.fysik.dtu.dk/ase>. Exists as Debian package 'python-ase'.
Repository: http://download.opensuse.org/repositories/home:/dtufys/
GPAW J. J. Mortensen et al, Phys Rev B, Vol. 71, 035109 (2005).
<http://wiki.fysik.dtu.dk/gpaw>. Exists as Debian package 'gpaw' and 'gpaw-setups' (gpaw-data).
NWChem M. Valiev et al, Comput. Phys. Commun. 181, 1477 (2010).
<http://www.nwchem-sw.org/>. Exists as Debian package 'nwchem' and 'nwchem-data'.
Elk <http://elk.sourceforge.net>. Exists as Debian package 'elk-lapw'.
The Elk executable should be compiled with e.g.
elk/src/modmain.f90:289 maxsymcrys=1024 or larger
DACAPO B. Hammer et al, Phys.Rev. B 59, 7413 (1999)
<https://wiki.fysik.dtu.dk/dacapo/>. Exists as Debian package 'dacapo' and 'dacapo-psp'.
ABINIT X. Gonze et al, Computer Physics Communications 180, 2582-2615 (2009).
<http://www.abinit.org/>. Exists as Debian package 'abinit' and 'abinit-doc' (abinit-data).
Install potentials from http://wiki.fysik.dtu.dk/abinit-files/abinit-pseudopotentials-2.tar.gz
into e.g. /usr/share/abinit
PHON D. Alfe, Computer Physics Communications 180,2622-2633 (2009).
<http://chianti.geol.ucl.ac.uk/~dario>.
Quantum Espresso P. Giannozzi, et al, J.Phys.:Condens.Matter, 21, 395502 (2009).
<http://www.quantum-espresso.org/>.
VASP G. Kresse and J. Hafner. Phys. Rev. B, 47:558, 1993.
<https://www.vasp.at/> Requires a license.
PhonoPy: A. Togo and I. Tanaka, Scr. Mater., 108, 1-5 (2015)
<https://atztogo.github.io/phonopy/>
MODEL EVALUATION (once created) ==============================================
Once the model has been created, its use requires that axes are given on
regular qx,qy,qz grids (in rlu along reciprocal axes). The model evaluations
does not require to recompute the forces, and is very fast.
When all axes are vectors of same orientation, the HKL locations is assumed to be a q-path.
When axes are not all vectors, not same length, nor orientation, a 3D HKL cube is used.
To generate a powder 2D S(q,w) you may use: sqw_powder(model)
To generate the dispersion curves along principal crystallographic directions,
use: sqw_kpath(model)
The phonon density of states (DOS) is automatically computed when evaluating the
model onto a grid HKL=[-.5 : 0.5]. It is stored in model.UserData.DOS as an
iData object. Partial density of states (per mode) are stored in UserData.DOS_partials
The neutron scattering cross sections should be automatically determined from
the chemical formula in order to properly evaluate the neutron scattering
intensity.
In case this would fail, or other cross sections are needed, it is required
to set the coherent neutron scattering length (b_coh in [fm]) as a vector in the
model.UserData.properties.b_coh = [vector / number of atoms in the cell]
or alternatively the scattering cross section in [barn]
model.UserData.properties.sigma_coh = [vector / number of atoms in the cell]
where negative values set b_coh < 0 (e.g. Hydrogen).
Setting b_coh=0 unactivates the intensity estimate.
This assignement should be done after creating the model, before performing
further HKL evaluations. Default is to use guessed b_coh from the formula.
Once the model has been created:
input: p: sqw_phonons model parameters (double)
p(1)=Amplitude
p(2)=Gamma dispersion DHO half-width in energy [meV]
p(3)=Background (constant)
p(4)=Temperature of the material [K]. When 0, the intensity is not computed.
p(5)=Energy_scaling. All frequencies are multiplied by this value.
or p='guess'
qh: axis along QH in rlu (row,double)
qk: axis along QK in rlu (column,double)
ql: axis along QL in rlu (page,double)
w: axis along energy in meV (double)
signal: when values are given, a guess of the parameters is performed (double)
output: signal: model value
Example (model creation and evaluation):
s=sqw_phonons('bulk("Cu", "fcc", a=3.6, cubic=True)','EMT','metal');
qh=linspace(-.5,.5,50);qk=qh; ql=qh; w=linspace(0.01,50,51);
f=iData(s,[],qh,qk,ql',w); scatter3(log(f(1,:, :,:)),'filled');
figure; plot(s.UserData.DOS); % plot the DOS
sqw_phonons(s, 'report') generates a full report for the given model.
sqw_kpath(s) plots the dispersion curves and density of states.
s=sqw_phonons('bulk("Si", "diamond", a=5.4, cubic=True)','semiconductor');
s=sqw_phonons([ ifitpath 'Data/POSCAR_Al'],'metal','EMT');
Version: Aug. 22, 2017
See also iData, iFunc/fits, iFunc/plot, gauss, sqw_cubic_monoatomic, sqw_sine3d, sqw_vaks
sqw_powder, <a href="matlab:doc(iFunc,'Models')">iFunc:Models</a>
(c) E.Farhi, ILL. License: EUPL.