The Elk FP-LAPW Code
An all-electron full-potential linearised augmented-plane wave (FP-LAPW) code with many advanced features. Written originally at Karl-Franzens-Universität Graz as a milestone of the EXCITING EU Research and Training Network, the code is designed to be as simple as possible so that new developments in the field of density functional theory (DFT) can be added quickly and reliably. The code is freely available under the GNU General Public License.

Latest version: 4.3.6

News | Features | Download | Documentation | Forums | Mailing list | Contributions | CECAM tutorial | Links
Elk Workshop 2018
We are pleased to announce the hands-on workshop Electronic Structure at the Cutting Edge with Elk taking place at the Max Planck Institute of Microstructure Physics, Halle (Saale), Germany from 3 to 7 September, 2018.

The tutorial will be grouped into the following main topics:

1. Density functional theory (DFT) and the linearised augmented plane wave (LAPW) method
2. Functionals, old and new
3. Magnetism and the DFT+U method
4. Many-body theory, the GW approximation and the Bethe-Salpeter equation
5. Phonons, electron-phonon coupling and superconductivity
6. Time-dependent density functional theory (TDDFT)
7. Real-time evolution and spin dynamics

Each topic will presented by experts in that field and will be followed by a comprehensive hands-on session with Elk.

If you would like to attend then please send an email request to Sangeeta Sharma.
• High precision all-electron DFT code
• FP-LAPW basis with local-orbitals
• APW radial derivative matching to arbitrary orders at muffin-tin surface (super-LAPW, etc.)
• Arbitrary number of local-orbitals allowed (all core states can be made valence for example)
• Every element in the periodic table available
• Total energies resolved into components
• LSDA and GGA functionals available
• Potential-only meta-GGA available with Libxc
• Core states treated with the radial Dirac equation
• Simple to use: just one input file required with all input parameters optional
• Multiple tasks can be run consecutively

Structure and symmetry
• Determination of lattice and crystal symmetry groups from input lattice and atomic coordinates
• Determination of atomic coordinates from space group data (with the Spacegroup utility)
• XCrysDen and V_Sim file output
• Automatic reduction from conventional to primitive unit cell
• Automatic determination of muffin-tin radii
• Full symmetrisation of density and magnetisation and their conjugate fields
• Automatic determination and reduction of the k-point set

• Spin polarised calculations performed in the most general way: only (n(r); m(r)) and (vs(r); Bs(r)) are referred to in the code
• Spin symmetry broken by infinitesimal external fields
• Spin-orbit coupling (SOC) included in second-variational scheme
• Non-collinear magnetism (NCM) with arbitrary on-site magnetic fields
• Fixed spin-moment calculations (with SOC and NCM)
• Fixed tensor moment calculations (experimental)
• Spin-spirals for any q-vector
• Spin polarised cores
• Automatic determination of the magnetic anisotropy energy (MAE) (experimental)

• Band structure plotting with angular momentum character
• Total and partial density of states with irreducible representation projection
• Charge density plotting (1/2/3D)
• Plotting of exchange-correlation and Coulomb potentials (1/2/3D)
• Electron localisation function (ELF) plotting (1/2/3D)
• Fermi surface plotting (3D)
• Magnetisation plots (2/3D)
• Plotting of exchange-correlation magnetic field, Bxc (2/3D)
• Plotting of ∇⋅Bxc (1/2/3D)
• Wavefunction plotting (1/2/3D)
• Electric field (E=-∇V) plotting (1/2/3D)
• Simple scanning tunnelling microscopy (STM) imaging based on the local density of states (LDOS) (experimental)

Forces and phonons
• Forces - including incomplete basis set (IBS) and core corrections
• Forces work with spin-orbit coupling, non-collinear magnetism and LDA+U
• Structural optimisation of both atomic positions and lattice vectors
• Iso-volumetric optimisation of unit cell
• Phonons for arbitrary q-vectors computed with density functional perturbation theory (DFPT)
• Phonons computed with the supercell method
• Phonon dispersion and density of states
• Thermodynamic quantities calculated from the phonon DOS: free energy, entropy, heat capacity
• Phonon calculations can be distributed across networked computers
• Electron-phonon coupling matrices
• Phonon linewidths
• Eliashberg function, α2F(ω)
• Electron-phonon coupling constant, λ
• McMillan-Allen-Dynes critical temperature, Tc
• Eliashberg equations solved self-consistently (experimental)

• Exact exchange (EXX) optimised effective potential (OEP) method (with SOC and NCM) (experimental)
• EXX energies (with SOC and NCM) (experimental)
• Hartree-Fock for solids (with SOC and NCM) (experimental)
• LDA+U: fully localised limit (FLL), around mean field (AFM) and interpolation between the two; works with SOC, NCM and spin-spirals
• Reduced density matrix functional theory (RDMFT) for solids (experimental)
• Bethe-Salpeter equation (BSE), including beyond the Tamm-Dankoff approximation; works with SOC and NCM
• Time-dependent density functional theory (TDDFT) for linear optical response calculations
• GW approximation spectral functions; works with SOC and NCM (experimental)

• Mössbauer hyperfine parameters: isomer shift, EFG and hyperfine contact fields (experimental)
• First-order optical response
• Kerr angle and Magneto-Optic Kerr Effect (MOKE) output (experimental)
• Generalised DFT correction of L. Fritsche and Y. M. Gu, Phys. Rev. B 48, 4250 (1993) (experimental)
• Energy loss near edge structure (ELNES)
• Non-linear optical (NLO) second harmonic generation
L, S, and J expectation values
• Effective mass tensor for any state
• Equation of state fitting (with the EOS utility)
• Iterative diagonalisation with fine-grained parallelisation
• Interface to the ETSF Libxc exchange-correlation functional library (experimental)

• Clean, simple code structure - ideal for development
• OpenMP parallelisation
• Message passing interface (MPI) parallelisation
• Efficient OpenMP+MPI hybrid parallelism
• Strict Fortran 90 compliance
• Only one language used
• Free-form style input file
• Full LaTeX documentation included with every subroutine
Elk is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. Elk is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with Elk. If not, see

Elk is updated regularly with new features and bug fixes. Features not listed as experimental may be used for production but, as with any code, please check the consistency of your results carefully. The latest releases can be downloaded from

Release notes
-GW approximation now available thanks to Arkadiy Davydov; this is a finite temperature Matsubara implementation; currently it only calculates the k-point resolved spectral function on the real axis; see the 'examples/GW' directory (experimental)
-the GW approximation works with spin-orbit coupling and non-collinear magnetism; it can also be run after the optimized effective potential (OEP) method
-significant optimisations for Hartree-Fock, OEP, RDMFT and BSE
-fixed an minor bug in BSE
-greatly improved the stability of lattice optimisation; also reduced the variable 'deltast' to 0.001
-improved the convergence of the electron-phonon coupling parameter with respect to k-points; thanks to Xianxin Wu for the careful testing
-removed the variable 'epseph' which determined the sampling window for electron-phonon coupling
-added new automatic test for TDDFT time-evolution
-upgraded to LAPACK 3.7.0
-Elk now uses the LAPACK routine zheevd for diagonalisation which is faster than zheev
-changed the variable 'rstsf' to 'ssxc'
-the smearing width 'swidth' can now be set with temperature in kelvin using the variable 'tempk'

-changed muffin-tin functions to a 'packed' convention; this improves the memory efficiency and speed for the whole code; completely backward and forward compatible
-Michael Fechner found a serious bug in DFT+U which has now been fixed; may affect systems with atoms to which DFT+U is applied and which have symmetric equivalents; MF added to list of contributors
-MF also suggested that we create a 'sticky' thread on the forum for announcements of publications which use Elk; this has now been done
-fixed a problem with the laser pulse power density plot (AFPDT.OUT) thanks to Peter Elliot and Tristan Mueller
-added the ability to randomize the orbitals at the start of a time-evolution run; use the variable 'rndevt0' to set the random amplitude
-improved the accuracy of the Hellmann-Feynman force on the atoms; this also improves the supercell phonon calculations
-added the reduced Stoner theory (RST) modification to the exchange-correlation functionals, see: L. Ortenzi, I. I. Mazin, P. Blaha, and L. Boeri, Phys. Rev. B 86, 064437 (2012); use the parameter `rstsf´ to adjust the spin scaling
-removed the Pulay mixer (mixtype=2) as it was consistently inferior to Broyden
-changed the variables 'lmaxinr' and 'lmaxvr' to 'lmaxi' and 'lmaxo', respectively; the old variables still work in

-fixed long-standing convergence problem of non-collinear GGA: should now work reliably for all cases
-removed logical variable 'ncgga' and added new real variable 'dncgga' (see manual for details)
-removed 'robust adaptive mixer' (mixtype=4); it was no great improvement over the default mixer; we may reintroduce it at a later stage after more testing
-no other changes

-lots of optimisations of both speed and memory; many tasks should be noticeably faster
-added new 'robust adaptive mixer' (mixtype=4) which should be able to converge almost anything; there are no mixing parameters for this mixer; useful for materials searches when you have to be guaranteed convergence; slower to converge than the default mixing and Broyden mixing
-added missing term to DFPT phonons corresponding to the first-order change in occupancies and the Fermi energy for q=0; the phonon dispersion for MgB2 is now in excellent agreement with experiment and previous calculations
-the electron-phonon coupling code was greatly improved; produces very good coupling constants for Al, Nb, Pb and MgB2; all of which are in the 'examples/phonons-superconductivity' directory
-new variable 'epseph' controls the Fermi surface sampling of the electron-phonon coupling matrix elements
-Markus Meinert created a script for calculating optical constants from the diagonal components of the dielectric tensor; it calculates the complex refractive index, reflectivity, absorption coefficient and EELS spectrum; available in the utilities directory
-real-time TDDFT now outputs a time-dependent DOS and effective temperature at each time step (experimental)
-setting 'uhighq=.true.' results in an ultra-high-quality calculation; when you absolutely have to have the best

-Important: fixed an inaccuracy in the scalar relativistic part of the code which has existed for around 12 years; this may slightly affect calculations of heavy elements; special thanks to Stefaan Cottenier, José A. Flores Livas and Marcin Dulak for the extensive calculations
-real-time evolution now works with DFT+U (experimental)
-added 'ramp' vector potential to time evolution code; this can simulate constant and constantly increasing electric fields; see manual for details; see the example Si-ramp
-lots of optimisations everywhere: code should be noticeably faster
-José made some changes to the 'highq' and 'vhighq' options
-fixed bug with non-collinear optimised effective potential (OEP)
-new task (150) writes out the Kohn-Sham-Dirac eigenvalues of the atomic species
-SS added two new f_xc kernels: single iteration bootstrap (fxctype=211) and revised bootstrap (RBO) (fxctype=212), see S. Rigamonti, et al., Phys. Rev. Lett. 114, 146402; added two examples: Ne-RBO and LiF-RBO
-improved the accuracy of the magnetic anisotropy energy (MAE) calculation by rotating the spherical coverings of the spherical harmonic transform; thanks to James Glasbrenner for testing
-David Ernsting found and fixed a bug in the Compton interpolation routine
-improved the stability of structural optimisation
-removed the smoothing option from the exchange-correlation magnetic field source projection routine
-the number of first-variational states for magnetic systems was twice as large as it needed to be: now fixed
-improved the vim syntax highlighting generator; run 'make vim' again
-organised the examples into categories
-minor bug fixes and improvements
-note for developers: changed some of the variable names related to atomic species (purely cosmetic)

-fixed electron-phonon coupling again; thanks to Karel Carva and Elie Moujaes for the careful testing and discussion
-DFPT phonons now work with MPI
-SS added the 'single-iteration bootstrap kernel' for linear optics calculations; this is selected with fxctype=211; see the example 'Ar-bootstrap-single'; this works well for calculating excitonic binding energies
-SS fixed a problem with RDMFT which had appeared in recent versions
-SS made minor changes to the non-linear optics code
-Important: fixed problem with the option 'cmagz' for forcing spin-orbit calculations to be collinear; this may affect MAE calculations; thanks to Martin Gmitra for finding this and the careful testing
-Important: fixed a problem with the magnetisation directions selected for MAE calculations
-reworked the 'ncgga' option for non-collinear GGA calculations; should be much more reliable now
-several important optimisations; the Hamiltonian matrix setup and density symmetrisation are now faster
-magnetic anisotropy energy (MAE) calculation now outputs the direction vectors; thanks to Galya Madzharova and Yu-ichiro Matsushita for the suggestion
-added option 'vhighq' for very high quality calculations;see example 'Os-Delta'
-'highq' and 'vhighq' options now write out the parameters which are changed
-increased default 'lmaxmat' from 6 to 7
-added new option 'nxlo'; this adds extra local-orbitals to the species in order to make the APW functions smoother; this can make calculations, particularly of excited state properties, more accurate
-syntax highlighting now available for the vim editor; just run 'make vim' and highlighting will be enabled

-real-time propagation for solids now available thanks to the considerable efforts of Kevin Krieger over the past three years; see the example 'Ni-laser-pulse'; this feature is highly experimental
-1D, 2D and 3D integrated electron momentum density (EMD) plots now available thanks to David Ernsting and Stephen Dugdale; these plots can be compared directly to Compton scattering profiles; see the examples 'Li-Compton' and 'Ni-Compton'; still experimental
-electron-phonon coupling has now been fixed and is also available for linear-response phonons
-large number of optimizations made throughout code
-code is more memory efficient in many areas
-Arkardy Davydov found a bug in OpenMP nested parallelism for BSE which has now been fixed
-fixed bug in magnetic anisotropy energy (MAE) calculation found by José A. Flores Livas
-added option for automatically converting species from APW+l.o. to LAPW; see option 'nxapwlo' in manual
-greatly improved unit cell optimisation algorithm; now should be very stable; see example 'Ge-lattice-opt'
-Markus Meinert updated his '' python code which should now work correctly with EPSILON_xx.OUT files
-modified the 'highq' defaults thanks to tests performed by Don Hamann described here
-full charge and spin response function can be written to file using task=331
-supercell phonons calculation now work with MPI
-removed use of OpenMP ATOMIC directive because of bug in ifort version 11

-Lars Nordström, Francesco Cricchio and JKD added fixed tensor moment (FTM) calculations (experimental)
-LN made changes to the mixer routine which makes restarts smoother
-the Hamiltonian is now made real for systems with inversion symmetry; this was removed in the previous release because of a bug; the FV step is now up to four times faster
-fixed problem with symmetry breaking of tetragonal systems during lattice optimisation pointed out by various users
-further improved and stabilised lattice optimisation
-fixed a long-standing bug related to electron-phonon coupling pointed out by Matthieu Verstraete and Ryotaro Arita
-fixed a problem with electron-phonon coupling introduced in a previous version, discovered by user 'wuhuagumu' and Heung-Sik Kim
-TDDFT reponse code for both charge and magnetism has been made faster and more memory efficient
-creating a STOP file in the running directory now cleanly stops a structural optimisation run; as requested by James Mudd
-problem with nested OpenMP has now been fixed
-more OpenMP parallel loops have been added
-extended the use of BLAS throughout the code
-fixed bug which affected the running of task=120
-removed the 'frozencr' option for the moment because the core energy was incorrectly calculated; this will be restored in a later release
-various optimisations and minor bug fixes

-DFT+U now works with multiple entries per atom; every mention of LDA+U has now changed to DFT+U to reflect the generality of the method, and to give it an air of respectibility; for example, the block 'lda+u' is now 'dft+u'; old input files will still work fine
-greatly improved the magnetic anisotropy energy (MAE) calculations, thanks to suggestions by various users; the crystal is now rotated instead of the magnetisation direction
-fractional atomic species can now be generated on the fly using the 'fspecies' block; see the example 'fractional-species'; note that the atomic mass is only an estimate in these species files; this is useful for the virtual crystal approximation (VCA)
-spin-orbit coupling can now be used with strictly collinear calculations by using 'cmagz=.true.'; this will speed up the calculation considerably; use only when the system is genuinely collinear
-unstable non-collinear GGA calculations can now be stabilised by setting 'ncgga=.true.'
-improved the speed and stability of the lattice vector optimisation; thanks to forum discussions regarding BaTiO3
-added the lattice optimisation of BaTiO3 to the examples directory; thanks to David Tompsett
-geometry optimisation should now work fine with MPI
-added highly converged osmium example 'Os-convergence' thanks to a lengthy discussion on the forum started by Marcin Dulak; also increased the value of 'lmaxmat' for the 'highq' option thanks to Marcin's careful testing
-Important: greatly improved the code setup and compilation thanks to suggestions by Marcin Dulak; all compiler options are now in the file 'elk/'; the Makefile in 'elk/src' no longer requires modification
-new variable 'rotavec' available in; this is a rotation specified in axis-angle convention which allows the rotation of the crystal axes
-thanks go to Rich Martin and collaborator Wen for useful testing and suggestions
-improved the DFPT phonon calculations and added the example 'Nb-phonon-DFPT'
-decreased the chance of accidentally triggering the convergence indicator in the self-consistent loop thanks to a suggestion from Markus Meinert
-fixed problem with the routine 'nonlinopt', pointed out by user Sabrina and Koichi Kitahara
-improved the OpenMP efficiency in parts by using '$OMP ATOMIC' instead of '$OMP CRITICAL'
-changed the names of various internal subroutines; for example 'seceqnfv' is now 'eveqnfv'
-thanks go to Igor Mazin, Konrad Bussmann and James Glasbrenner for some very useful improvments
-fixed problem with species generation and MPI pointed out by Marcin Dulak
-fixed problem with use of Fortran 'APPEND' flag thanks to user Alaska
-updated the fundamental constants and conversion factors to CODATA 2008 recommended values
-Important: updated code for compatibility with Libxc version 2.2.0; see the manual and note the change in compilation instructions
-fixed problem with the routine 'energyfdu' which affected DFT+U calculations with 'inpdftu=4' (input of Yukawa screening length)
-Important: temporarily switched of the conversion of the Hamiltonian to a real symmetric problem because of possible bug; the only effect of this is that the code is slower for systems with inversion symmetry; this will be fixed in the next release
-minor bug fixes and optimisations

-fixed problem reported by Sonu Sharma and Michael Fechner with calculation of expectation values L, S and J
-slightly improved the magnetic anisotropy energy (MAE) calculation
-added new variable 'socscf' which scales the spin-orbit interaction term in the Hamiltonian; this allows calculation by interpolation of the MAE for cases where the MAE is small
-added new variables to VARIABLES.OUT

-fixed problem with fixed spin moment (FSM) calculations pointed out by Jagdish Kumar
-added new output file VARIABLES.OUT; this is for developers wanting to interface their codes to Elk; it contains most of the useful variables required for reading the Elk binary files; if you would like a variable added then just send me a mail; enabled when 'wrtvars' is set to .true.
-minor cosmetic changes

-Important: fixed serious bug discovered by Jan Balluff and Markus Meinert which produced erroneous magnetic moments with GGA; this only affects version 2.2.5; sorry about that
-even more substantial speed-ups for most parts of the code
-fixed MPI problem with the magnetic anisotropy energy (MAE) routine reported by Konrad Bussmann and José A. Flores-Livas
-minor bug fixes and improvements

-dramatic speed-up for most parts of the code thanks to some (fairly obvious) optimisations
-greatly improved lattice vector optimisation
-automatic determination of the magnetic anisotropy energy (MAE) with 'task=28' (experimental); see the example 'FeCo-MAE'
-the variable 'nempty' is now a real number; see manual entry
-magnitude of total magnetisation vector written to MOMENTM.OUT
-improved the species files
-made the linearisation energy search more reliable
-reduced the annoying 'linearisation energy not found' warning

-added full structural optimisation; setting 'latvopt=1' performs unconstrained optimisation of the lattice vectors; 'latvopt=2' performs iso-volumetric optimisation; see the examples 'Ge-lattice-opt' and 'MgB2-lattice-opt'
-minor bug fixes and optimisations
-all tests should now report 'Passed'

-fixed several problems with the previous (beta) release; thanks to Markus Meinert for the exhaustive checking
-magnetic hyperfine fields are now correctly calculated thanks to MM
-added missing items to the manual
-added the example 'Si-phonon-DFPT'
-thanks to Kari Ruotsalainen, Aleksay Golovchan and Vladimir Nazarov for pointing out bugs in 2.1.22 - now fixed
-added estimation of direct gap to INFO.OUT; thanks to user Sean for suggesting this
-this is the version that will be used for the Elk Tutorial
-see you in Lausanne!

elk-2.1.22 (beta)
-Important: this is a beta version and should not be used for production work; a production release will be made in a few weeks
-Important: this version is not backwards compatible; calculations will have to be converged from scratch
-density functional perturbation theory (DFPT) phonons now available; only for spin-unpolarised calculations at the moment; spin-polarisation will be added for the next release; use with task=205; still experimental and generates incorrect results for certain symmetries - we're still working on this; see the example 'Al-phonon-DFPT'
-full frequency-dependent magnetic response function now available; works with non-collinear magnetism, etc.; use task=330; see example 'Ni-magnetic-response'
-iterative diagonalisation greatly improved; now uses the method of P. Blaha, et al., J. Comp. Phys. 229, p453 (2010)
-upgraded code to handle libxc version 2
-hybrid functionals should now work with libxc (experimental)
-upgraded LAPACK to version 3.4.2
-block name dos changed to wplot; changed variable 'nwdos' to 'nwplot'
-variable 'gmaxrpa' changed to 'gmaxrf'; this is the response function G-vector cut-off
-BSE and TDDFT now faster, more accurate and memory efficient; thanks to M. Meinert for discussions
-task 188 changed to 320
-reintroduced the variables 'lmaxinr' and 'fracinr'; this improves stability
-lots of optimisations
-added more MPI parallelism
-various minor bug fixes
-conduction state local-orbitals can now be switched on; use 'lorbcnd=.true.'; this can improve both ground-state and response function calculations; this is switched on automatically when 'highq=.true.'
-most species files have been changed; mostly larger muffin-tin radii
-Important: Markus Meinert suggested a change to 'nempty'; this variable now represents the number of states per atom and spin; this scales automatically as the number of atoms is increased
-default 'nempty' is now 4
-M. Meinert changed the default Broyden parameters after extensive testing
-LN fixed bug in XCrysDen Fermi surface plotting
-LN also reduced the number of states which contribution to the Fermi surface plot to the minimum
-supercell phonon calculations can now be restarted with 'task=202'
-Frank Wagner discovered a bug which affects certain types of symmetries in rare instances; this has now been fixed
-Important: default smearing function is now Fermi-Dirac (stype=3)
-entropy contribution now added to the total energy
-default smearing width reduced to 0.001 Ha, corresponding to room temperature
-M. Meinert added a blocks-to-columns script; this is in the 'utilities' directory
-Oscar Grånäs added to list of main authors

-Important: fixed bug in BSE calculations owing to incorrect rotation of dielectric function matrix elements; thanks to Arkardy Davydov for pointing this out
-BSE is now three separate tasks: 185 writes the BSE Hamiltonian to file, 186 diagonalises the Hamiltonian and 187 produces the BSE dielectric function; thanks to Markus Meinert for suggesting this
-adiabatic local density approximation (ALDA) for linear-response TDDFT now available; see example 'diamond-TDDFT-ALDA'
-added Markus' suggestion that the Tran-Blaha constant can now be read in with the variable 'c_tb09'
-added new variable 'highq'; set this to .true. and the most important parameters are set for a high-quality, well-converged calculation; see the settings in the file 'readinput.f90'; thanks to Igor Mazin for all the testing
-removed the variables 'lmaxinr' and 'fracinr' as they were unnecessary
-simplified TDDFT routines
-Important: the variable 'radkpt' has been redefined and should be made larger by 2π ; (sorry about this, the original scale for this variable was correct!)
-fixed problem with MPI that caused some LDA+U runs to crash; thanks to Alaska Subedi for pointing this out
-added documentation
-several minor bug fixes and optimisations

-Important:fixed severe bug affecting spin-unpolarised native GGA functionals, thanks to Jiji Pulikkotil for pointing out a problem with Mg3Sb2 which led to its discovery; this now makes structural optimisation truly variational for PBE
-Important:fixed severe bug affecting spin-polarised 'libxc' GGA functionals
-thanks to Miguel Marques and Silvana Botti for the week-long collaboration at Université Lyon where, amongst other things, we rewrote the meta-GGA interface to 'libxc'
-thanks to Markus Meinert for careful testing of meta-GGA with various systems
-MM also added questions and answers to the FAQ
-meta-GGA now works for collinear magnetism only: it is inconsistent with non-collinearity
-Henning Glawe pointed out a bug affecting the generation of the k-point set when 'autokpt=.true.'
-Important: the variable 'radkpt' has been redefined and should be made smaller by a factor of 2π
-SS fixed a bug in the plotting of spin-polarised Fermi surfaces
-electron momentum density now available with 'task=170'; thanks to S. Dugdale and D. Ernsting for discussions
-lattice vectors, atomic positions and muffin-tin magnetic fields can now be randomised by setting the random amplitudes 'rndavec', 'rndatposc' and 'rndbfcmt' to positive values; this is useful for checking stability of a calculation
-the k-points corresponding to the indirect band-gap are now reported to 'INFO.OUT'
-lots of optimisations and simplifications

-potential-only meta-GGA now available in conjunction with Libxc as an experimental feature; see the example 'Si-meta-GGA'; thanks to Miguel Marques for discussions
-finite q-vector linear response TDDFT now available thanks to SS
-Important: variable 'autormt' has been removed and automatic scaling of the muffin-tin radii is now always on; a new simpler algorithm for computing the radii was implemented; non-overlapping muffin-tins will not be rescaled; calculations may need to be reconverged
-changed parameter for estimation of the nuclear radius from Z to A; thanks to Giorgio Concas for pointing this out
-Kohn-Sham band gap written to 'GAP.OUT' after each iteration
-3D nesting function plot now available; use 'task=105'
-fixed problem with a constant in the BSE exchange term
-fixed problem of TDDFT freezing with MPI
-lots of optimisations and simplifications

-fixed problem with OpenMP 'reduction' clause in calculation of dielectric function
-no other changes

-removed factor of 2 from the BSE exchange term, this has little effect on spectra in the optical range; spotted by Markus Meinert after careful investigation of core-state BSE spectra
-M. Meinert also improved the 'CaO-BSE-core' and 'Fe-XMCD' examples
-fixed problem with symmetry discovered by Alexey Baranov
-calculation of the partial DOS is now parallel, thanks to suggestions from Anton Filanovich
-various optimisations; including some ideas from Anton Kozhevnikov
-iterative diagonalisation is now much faster
-improved the adaptive mixing scheme
-improved the initial stability of the self-consistent loop, thanks to discussions with Lars Nordstrom
Notes for developers
-inverse indices from 'idxis', 'idxia', 'idxil' and 'idxim' for species, atoms and angular momenta are now available, thanks to A. Kozhevnikov
-the arguments to hmlaa, hmlalo, hmllolo, olpaa, olpalo, olplolo have changed

-Antonio Sanna fixed a problem with the Pade approximant used in the Eliashberg code
-SS fixed a problem in Fermi surface plotting with OpenDX
-no other changes

-fixed a problem which occurs with version 10 of the Intel compiler
-simplified the 'setup' script and added OpenMP compiler options by default
-no other changes

-added directionality to the TDDFT code; thanks also to Vladimir Nazarov for discussions
-a pre-print is now available for the new TDDFT 'bootstrap' kernel: arXiv:1107.0199v1 [cond-mat.mtrl-sci]
-added the scissor correction to the RPA inverse dielectric function in EPSINV_RPA.OUT
-fixed a problem with running geometry optimisation with MPI; thanks to Arkady Davydov and Antonio Sanna for discussions
-geometry optmimisation has changed: now the geometry, interatomic distances, final total energies and forces are stored for each optimisation step in GEOMETRY_OPT.OUT, IADIST_OPT.OUT, TOTENERGY_OPT.OUT and FORCES_OPT.OUT
-removed geometry optmisation from the ground-state calculation and put it in a separate subroutine
-reduced pseudocharge density constant (lnpsd) because the spherical bessel function order could be too large; may change total energies slightly
-q-points and weights written to QPOINTS.OUT
-minor changes and optimisations

-fixed serious bug introduced in the last version affecting some tasks (for example DOS plots) for crystals with non-symorphic symmetries; thanks to Vladimir Nazarov and Daniel Rohr for pointing this out
-VM also fixed the graphene example, in which the atomic coordinates were (embarrassingly) incorrect; also included a high-precision DOS plot to demonstrate the Dirac-like band structure
-SS, JKD, Antonio Sanna and Hardy Gross added a new TDDFT functional, called the 'bootstrap' kernel, which produces remarkably good linear optical response functions (article currently in preparation); see the example 'LiF-TDDFT'
-Antonio Sanna added the Eliashberg equations for finding the superconducting gap as a function of temperature; see the example 'Al-Eliashberg'
-improved the electron-phonon coupling calculation (task=240); this is faster, more accurate and now works for spin-spirals; now also parallelised with MPI
-removed packed matrix format everwhere in the code, along with the variable 'tpmat'; the first-variational matrices are now stored in upper triangular format; this takes more memory, but is faster
-Anton Kozhevnikov fixed a small problem with calculating the nuclear-nuclear energy
-various optimisations, simplifications and minor bug fixes
-this version of the code with be used in the CECAM Tutorial
Notes for developers
-the order of indices for the q- and w-dependent inverse dielectric function, epsinv, have changed from (w,G,G') to (G,G',w)

-linear response time-dependent density functional theory (TDDFT) now available for calculating the q->0 dielectric response; see the example 'LiF-TDDFT' (experimental)
-fixed a problem with the BSE calculation which made the response function too large
-added the possibility of using arbitrary states in the BSE kernel thanks to discussions with Markus Meinert; this enables the calulation of core state BSE spectra; see the example 'CaO-BSE-core' (experimental)
-MM also added an x-ray magnetic circular dichroism (XMCD) example: 'Fe-XMCD'
-BSE calculations are now faster, thanks to discussions with MM
-Alexey Baranov made several changes to the structure factor code, including adding an energy window, 'wsfac', for the calculations; see the example 'MnO-str-factors'
-Tyrel McQueen suggested a way to speed up the Hartree-Fock calculation; this has been implemented and also considerably speeds up OEP and RDMFT
-TMcQ also made some ongoing changes to the hybrid functional code
-made the radial Dirac and Schrodinger integrators yet more stable; thanks to AB and Frank Wagner for discussions
-fixed a problem with the ordinary RPA dielectric function; only affects calculations which have a scissor shift which made epsilon slightly too small
-real symmetric diagonalisation now used for the first-variational eigenvalue problem for crystals with inversion symmetry; this can speed up the calculation by a factor of three; this is thanks to discussions with Lars Nordstrom; Important: the atomic basis may be shifted to a different position and old output files may need to be reconverged
-fixed a stability problem which occurs when using GGA functionals by removing G-vector truncation of the effective potential introduced in version 1.0.16; thanks to Greg Walker for discovering this
-LDA+U calculations are now faster thanks to fast evaluation of the atomic density matrix
-Broyden mixing scheme now available: this seems to be both fast and stable (use 'mixtype=3')
-removed Anderson mixing
-improved starting guess for the density in ground-state calculations
-upgraded to LAPACK 3.3.1
-various optimisations and simplifications
Notes for developers
-the arrays haa, halo, hlolo, oalo, ololo are now smaller and the indexing has been rearranged

-the Bethe-Salpeter equation (BSE) for linear optics now works beyond the Tamm-Dankoff approximation (use 'bsefull=.true.')
-RPA and BSE calculations now use full wavefunctions instead of plane waves
-removed the Coulomb regulator from the calculation of RPA dielectric function, instead used the analytic results for the head and the wings of the matrix (thanks to Vladimir Nazarov for discussions)
-improved the stability of radial Dirac and Schrodinger integrators (thanks to Alexei Baranov, Frank Wagner and Ondrej Certik for discussions)
-added more block descriptions to the manual (thanks to various people for pointing out omissions)
-various optimisations

-SS and JKD added the Bethe-Salpeter equation (BSE) for linear optics calculations. This feature works with LDA+U, magnetism and spin-orbit coupling. It is also parallelised with MPI and can be run across a cluster. See the 'LiF-BSE' and 'Si-BSE' examples. Currently an experimental feature
-Alexey Baranov added density and magnetic structure factors: see the 'MnO-str-factors' example
-AB also fixed a problem with output of the spacegroup code
-full frequency and G vector dependent RPA inverse dielectric function now available with 'task=180'. Works with metals, insulators, magnetism and SOC, and can be calculated for arbitrary complex frequencies. Also parallelised with OpenMP and MPI. Thanks to Anton Kozhevnikov for discussions
-added lots more MPI parallelism and made the code more MPI-friendly
-many optimisations across the whole code
-Simone Chiesa, Anton Kozhevnikov and Adolfo Eguiluz found a problem in the plotting of the partial DOS which has now been fixed
-the first-variational eigenvalue matrices can now be stored and diagonalised in non-packed storage mode: use 'tpmat=.false.'. This can speed up calculations at the expense of memory
-Tyrel McQueen found a bug in the 'findprim' routine, now fixed
-TMcQ also suggested a change to how the code connects the points of the 'plot1d' vertices
-Martin Stankovski suggested that 'ecvcut' be made into an input variable, allowing the core-valence cut-off to be adjusted
-added the phonon calculation of Ni to the examples
-Jerzy Goraus contributed a script for calculating VB-XPS spectra from PDOS* files; and also one for converting the Wien2K struct file to
-modified some species files thanks to tests done by J. Goraus
-Henning Glawe suggested including the version number in the release directory
-upgraded to LAPACK 3.3.0
-made tolerance for finding the linearisation energies (epsband) much smaller: this improves the overall stability of the self-consistent loop
Notes for developers
-arguments to 'genppts' have changed; also the non-reduced k-points are now stored in the remaining part of the reduced k-point arrays

-message passing interface (MPI) parallel runs now available; scalable to hundreds of cores across a cluster; can also be used as hybrid OpenMP+MPI parallelism for maximum efficiency; with thanks to Bhagawan Sahu for help with testing; see manual for compilation and running instructions
-S. Sharma added susceptibility tensor for non-linear optical second-harmonic generation (SHG); see example 'GaAs-NLO' for details
-added spin-orbit correction to momentum matrix elements; affects linear and non-linear optics
-optical calculations now work in combination with spin-spirals
-updated interface to version 1.0 of the ETSF exchange-correlation library, libxc; thanks to M. Marques and T. McQueen for assistance
-fixed bug spotted by T. McQueen involving using multiple tasks in conjunction with 'primcell=.true.'
-A. Kozhevnikov fixed bug in 'xc_pwca.f90'
-A. Kozhevnikov also fixed serious bug in 'getevecfv.f90'
-F. Cricchio changed the order of lattice vector angles in 'spacegroup' utility to 'bc, ac, ab' which corresponds to the convention alpha, beta, gamma
-removed scissor correction from eigenvalue solver; now it is used, as before, only in optics calculations
-T. McQueen suggested an improvement to the routine which generates the path in reciprocal space for bandstructure plots, 'connect.f90', which has been implemented
-fixed problem with XCrysDen Fermi surface plots, spotted by FC
-various optimisations and simplifications
Notes for developers
-arguments to 'zpotcoul' have changed in anticipation of linear-response phonons; should now be called in conjunction with 'genzvclmt'

-fixed problem with Fermi surface generation introduced in the last version
-E. K. U. Gross added to list of main developers
-no other changes

-fixed problem in linear optics calculations for metals - thanks to Antonio Sanna, Fabio Bernardini and Sandro Massida for pointing this out
-FC and LN added option for automatic determination of APW linearisation energies (use 'autolinengy=.true.')
-Marty Blaber made a script which allows for semi-automatic labeling of the vertex locations with gnuplot (found in the 'elk/utilities' directory)
-FC and LN fixed problem with some compilers when writing FERMISURF.OUT
-FC and LN fixed problem with writing XCrysDen Fermi surface plots
-Anton Kozhevnikov found and fixed a problem with the generation of the irreducible representation file 'ELMIREP.OUT'
-Torbjörn Björkman added a new routine which automatically determines the smearing width from the k-point density. See 'autoswidth' in the manual and the example 'Nb-autoswidth' for details
-added a constant electric field E in the form of a saw-tooth potential: set the vector 'efieldc' (experimental)
-added a constant vector potential A to the Hamiltonian to generate constant currents: set the variable 'afieldc' (experimental)
-FC and LN fixed problem with a spin-polarised GGA routine
-FC and LN pointed out that the energy term associated with the excess charge compensating background should be zero. This term has now been removed
-code now more memory efficient thanks to smarter array allocation
-the variable 'rgkmax' can now be made arbitrarily large without risk of instability
-removed variable 'cfdamp'
-Anton Kozhevnikov pointed out that the radial functions were being redundantly calculated for equivalent atoms -- now fixed
-Anton Filanovich added several questions to the FAQ
-scissors correction (given by the variable 'scissor') now applied consistently immediately after the generation of the eigenvalues
-FC and LN removed the spherical harmonic transform matrices with rank lmmaxapw
-added dielectric function calculation for arbitrary q-vector, see 'LiF-Yambo' example (experimental)
-by setting 'gmaxvr=0', the G-vector cut-off for the potential and density is automatically determined from 'gmaxvr=2*gkmax+epslat'
-various bug fixes, simplifications and optimisations

-internal releases only

-Tyrel McQueen and JKD incorporated the ETSF exchange-correlation functional library Libxc into Elk - see the manual for installation instructions. This is still an experimental feature
-Fixed problem with non-collinear GGA: this feature should still be considered as experimental
-GGA now faster and more accurate
-FC and LN included an estimation of the band gap at each iteration
-FC and LN parallelised the DOS routine
-SS updated the RDMFT code and added spectral output
-SS made the RDMFT code more memory efficient
-fixed problem with the Mössbauer hyperfine fields and in doing so also allowed the core states to be spin-polarised (use spincore=.true.)
-added a Mössbauer hyperfine field example (Fe-Mossbauer)
-removed evalmin and replaced it with a check if the lowest eigenvalue is less than the lowest linearisation energy
-Marty Blaber parallelised part of the dielectric routine and JKD also parallelised the generation of momentum matrix elements
-FC and LN fixed a symmetry problem when using reducebf
-FC and LN included the possibility of summing the partial DOS over m: just set dossum=.true. and the PDOS files will only contain the l-resolved DOS
-fixed linearisation energies can now be determined automatically thanks to FC: just set autolinengy=.true.
-Important: gkmax is now determined from rgkmax using the average muffin-tin radius, rather than the smallest
-completely removed the species program and incorporated the code directly into Elk. See the species directory and the ununhexium example
-improved and regenerated all the species files (thanks to discussions with S. Massidda) and set all atomic masses to best known values
-improved the spherical covering used for the spherical harmonic transforms (SHT): replaced the "Saff and Kuilaars" spiral with the "golden section" spiral. This results in faster convergence with respect to lmax
-removed the input variable vacuum: the box containing the molecule now has to be set up by the user when molecule=.true.. See the NaCl-monomer example
-bfcmt read in after the atomic positions in the atoms block is now optional
-alphabetised the list of subroutines in the Elk manual and added hyperlinks
-various optimisations and simplifications
-enabled the Elk Wiki pages: Anton Kozhevnikov is the Wiki administrator
-upgraded license to GPL version 3
-Elk now officially out of beta
The Elk Code Manual is available, as is a manual for the Spacegroup utility. You can ask and answer questions in the Elk forums on A mailing list is available for notification of new releases, features and bugs in the code. There is also a collection of useful links. Online lectures and presentation notes from the CECAM Elk Tutorial by experts in DFT and many-body theory can be found here.

Vitaliy Romaka has produced some very nice Elk tutorials complete with video demonstrations.

Citation of Elk
Citation of the code is not mandatory but would be appreciated by the contributors - a reference to this website will suffice. We also recommend that the input files used be made available to other researchers.
Main authors and contributors
Kay Dewhurst, Sangeeta Sharma, Lars Nordström, Francesco Cricchio, Oscar Grånäs, Hardy Gross, Claudia Ambrosch-Draxl, Clas Persson, Fredrik Bultmark, Christian Brouder, Rickard Armiento, Andrew Chizmeshya, Per Anderson, Igor Nekrasov, Frank Wagner, Fateh Kalarasse, Jürgen Spitaler, Stefano Pittalis, Nektarios Lathiotakis, Tobias Burnus, Stephan Sagmeister, Christian Meisenbichler, Sébastien Lebègue, Yigang Zhang, Fritz Körmann, Alexey Baranov, Anton Kozhevnikov, Shigeru Suehara, Frank Essenberger, Antonio Sanna, Tyrel McQueen, Tim Baldsiefen, Marty Blaber, Anton Filanovich, Torbjörn Björkman, Martin Stankovski, Jerzy Goraus, Markus Meinert, Daniel Rohr, Vladimir Nazarov, Kevin Krieger, Pink Floyd, Arkardy Davydov, Florian Eich, Aldo Romero Castro, Koichi Kitahara, James Glasbrenner, Konrad Bussmann, Igor Mazin, Matthieu Verstraete, David Ernsting, Stephen Dugdale, Peter Elliott, Marcin Dulak, José A. Flores Livas, Stefaan Cottenier, Yasushi Shinohara, Michael Fechner

Contributing to Elk
If you have made a modification to part of the code which makes it faster, simpler or produces additional results, then feel free to send it to one of the main authors, who may then incorporate it into the main branch. Groups and individuals are actively encouraged to develop and release their own specialised versions of the Elk code under the GPL. However, in order to avoid any confusion, please use a different name for your version of the code. We suggest you append the name of your location to Elk, for example: Elk-Uppsala.

Special versions of Elk/EXCITING
The exciting Code

Codes which interact with Elk
DGrid - a program for the generation and analysis of properties on equidistant grids
ASE - The Atomic Simulation Environment
Libxc - an extensive library of LDA and GGA exchange-correlation functionals
NOMAD - established to host, organize, and share materials data
Phonopy - an open source package for phonon calculations at harmonic and quasi-harmonic levels
Get Elk at Fast, secure and Free Open Source software downloads

Last modified March 2017