Release Notes

v9.8¶

Version 9.8, released on December 23, 2022. List of changes with respect to version 9.6.

Many thanks to the contributors to the ABINIT project between October 2021 and August 2022, and some late contributions up to April 2023 ! These release notes are relative to modifications/improvements of ABINIT v9.8 with respect to v9.6.

The list of contributors includes: B. Amadon, G. Antonius, L. Baguet, S. Bandyopadhyay, L. Bastogne, J.-M. Beuken, J. Bieder, A. Blanchet, F. Bottin, J. Bouchet, E. Bousquet, F. Brieuc, V. Brousseau-Couture, N. Brouwer, F. Bruneval, M. Cote, C. Espejo, Ph. Ghosez, M. Giantomassi, O. Gingras, X. Gonze, B. Guster, P. Kesterner, R. Outerovich, Ch. Paillard, M. Royo, A. Sasani, B. Sataric, M. Schmitt, F. Soubiran, M. Torrent, M. Verstraete, He Xu, J. Zwanziger.

It is worth to read carefully all the modifications that are mentioned in the present file, and examine the links to help files or test cases. This might take some time …

Xavier

A. Important remarks and warnings. Also, hotfixes for v9.8.3 (A.4 to A.10).¶

A.1 Warning : the input variables prtefg and prtfc have been renamed nucefg and nucfc.

By J. Zwanziger (MR850)

A.2 Within PAW in ABINIT, the ZORA relativistic factor for all PAW+spin-orbit coupling has been corrected: A ^2 was missing for the 1/(1-(v/c)^2) factor. This changes slightly all the PAW+SOC results (because the factor is very small), without any influence on physical results. However, previous results cannot be anymore obtained.

By M. Torrent (MR849)

A.3 The default value for rfdir is now (1 1 1), instead of (0 0 0).
The one of rfatpol is now (1 natom).

By X. Gonze (MR852)

A.4 The default values of dossmear (@anaddb) and dosdeltae (@anaddb and @atdep) have changed, to smaller values, to improve default accuracy.

By X. Gonze

A.5 Correction and cleaning of DFT+U with magnetism (nspden=4), following the merge 881. Introduce “optdcmagpawu” input to control different choices of DC term (for tests and code comparisons, not useful for production). Versions before 9.8 is equivalent to optdcmagpawu=1 (no magnetism in the DC term). Now the default is 3 (magnetism in the DC term). Some refs are changed accordingly. Change of the tests v9/76,77 and 78 for better precision. Add the tests v9/88 and 89 to tests the combination of PAW+U with different pawxcdev.

By L. Baguet (MR898)

A.6 Some users have reported that with the new optcell=4,5 or 6 option, the calculation continues when the criterion should be already reached. For example, with optcell=5 (relaxing b only), the xx component of the stress could still be large but shouldn’t be checked. This has been fixed.

By Xu He (MR897)

A.7 More fix on the generation of symmetry-adapted terms in Multibinit. A more complete set of terms are now generated, with improved time and memory performance.

By Xu He and A. Sasani (MR899)

A.8 m4 detection netcdf(-f) //: AC_RUN_IFELSE hangs with some compilers then switch to AC_LINK_IFELSE

By JM Beuken (MR900)

A.9 Several improvements for external libs: FFTW3 thread safety, openBLAS multithreading, netCDF Fortran parallel

By M. Torrent, with patch from P. Kesteneer (MR902)

A.10 Several fixes, including documentation for icutcoul and related input variables

By. X. Gonze (MR903)

B. Most noticeable achievements¶

B.1 Lattice Wannier functions can be computed, using the SCDM-k algorithm, or the projected Wannier function algorithm.

The lattice_wannier tutorial has been written to teach how to construct such lattice Wannier functions. They are used in localized bases for atomic distortions. One typical use case is to build an effective Hamiltonian of collective, localized, atomic displacements (see the next achievement).

A script (compare_phbands.py) to compare phonon/LWF band structures is added to the scripts/post_processing directory.

Related topic topic_LWFModel. See tests lwf[1], v9[110], and v9[111].

This feature is still under heavy development. The current version should be regarded as a “technology preview”.

By He Xu (MR844)

B.2 Lattice Wannier function dynamics is available inside the second-principle engine MULTIBINIT.

Once Lattice Wannier functions are available, they can be used with different dynamical algorithm to deduce heat capacity, susceptibility, structural phase transition, critical temperature, instead of doing standard molecular dynamics with all atoms. Related input variables : lwf_constraint@multibinit, lwf_dt@multibinit, lwf_dynamics@multibinit, lwf_init_state@multibinit, lwf_init_hist_fname@multibinit, lwf_mc_avg_amp@multibinit, lwf_nctime@multibinit, lwf_ntime@multibinit, lwf_pot_fname@multibinit, lwf_taut@multibinit, lwf_temperature@multibinit, lwf_temperature_end@multibinit, lwf_temperature_nstep@multibinit, lwf_temperature_start@multibinit, lwf_var_temperature@multibinit.

See the tutorial lwf_model tutorial and related tests.

This feature is still under heavy development. The current version should be regarded as a “technology preview”.

By He Xu (MR851)

B.3 Numerous miscellaneous improvements of the second-principle engine MULTIBINIT.

Hybrid Monte-Carlo Mover with NPT ensemble (ionmov=25).

MULTIBINIT can now be used without ‘files’ file, as the main ABINIT or ANADDB. Related input variables latt_pot_fname@multibinit, latt_harm_pot_fname@multibinit, latt_anharm_pot_fname@multibinit, latt_training_set_fname@multibinit, latt_test_set_fname@multibinit, spin_pot_fname@multibinit, spin_init_hist_fname@multibinit, slc_pot_fname@multibinit, outdata_prefix@multibinit.

The MULTIBINIT tutorial has been improved. In particular, it is now starting with a global introduction (multibinit tutorial).

The following MULTIBINIT features and input variables have been introduced : define oblique supercells (ncellmat@multibinit), print Goal-Function values in CSV format (prt_GF_csv@multibinit), impose the seed for MULTIBINIT spin/LWF dynamics to obtain reproducible results (randomseed@multibinit), specify three weights for Energy, Forces and Stresses in the calculation of the Goal Function (bound_factors@multibinit), specify three weights for Energy, Forces and Stresses in the calculation of the Goal Function during the fit process (fit_factors@multibinit), specify three weights for Energy, Forces and Stresses in the calculation of the Goal Function during the optimization process (opt_factors@multibinit), specify the relative penalty for the determination of bounding coefficient values (bound_penalty@multibinit), activate the generation of pure displacement coefficients (fit_dispterms@multibinit), specify the number of anharmonic coefficients per symmetric irreducible atom to add during fit process (fit_ncoeff_per_iatom@multibinit), specify the number of coefficients imposed with fixed value as in the input xml during the fit process for the mode (fit_nimposecoeff@multibinit), specify the indices of the imposed coefficients with fixed coefficient value during the fit process for the model (fit_imposecoeff@multibinit),

See the tests in which these input variables are used.

By He Xu, M. Schmitt, A. Sasani, L. Bastogne, S. Bandyopadhyay, and P. Ghosez (MR812, 851, 868, 894)

B.4 The TDEP formalism implemented in ABINIT (aTDEP), allowing to compute temperature-dependent phonon band structures, has been made more robust, and tested extensively. One tutorial (atdep1 tutorial) is now available. See the thirty-seven tests from atdep[01] to atdep[37] and the tests mentioned in the tutorial atdep1 tutorial. Related publication, see [Bottin2020].

By F. Bottin, J. Bieder and J. Bouchet (MR 836).

B.5 The conducti utility can treat Spin-Orbit Coupling for transport properties (within PAW) : conductivity, XANES, transport coefficients, … See [Brouwer2021]. The writing of documentation and tutorial is in progress.

By M. Torrent and N. Brouwer (MR849).

C. Changes for the developers (including information about compilers)¶

C.1 Added support for NAG 7.1

From J.-M. Beuken (MR830).

C.2 Update build system to allow the use of NVTX library, providing profiling annotations (only when gpu is enabled). This makes more readable profiling and tracing information when viewed with nsys-ui. Add new parameter use_nvtx to enable/disable nvtx annotations at runtime. If abinit is built without gpu, annotations completely vanish at compile time.

From P. Kestener (MR843)

C.3 Improve detection of inlined macros. The developers should now use

ABI_SFREE(allocateble_array)

instead of

if(allocated(allocatable_array)) ABI_FREE(allocatable_array)

From M. Giantomassi (MR859)

C.4 The DDB IO routines have been refactored throughout the different main codes (respfn, gstate, anaddb, nonlineal, longwave, ddb_interpolate, gruns_new, thmeig). The ddb merging routines have been refactored (merge_ddb replaces mblktyp1 and mblktyp5). New MPI treatment of ddb reading (no more free-for-all reading). Removed several arguments of ddb_from_file: natom, natifc and atifc. Moved routine dfptnl_doutput from m_ddb to m_nonlinear. Remove dfpt_lw_doutput .

From G. Antonius (MR872)

C.5 Suppressed CHILDREN and PARENTS sections in ROBODOC header. They were not maintained automatically anymore, so many had become misleading.

From X. Gonze (MR874)

C.6
A new type of file ‘GSTORE’ has been introduced, for the electron-phonon matrix element storage, where it is taken advantage of filters (on energy or band or wavevectors) to reduce the size compared to the more usual EIG1 file type. It contains also the related metadata. Related input variables : getgstore_filepath, gstore_cplex, gstore_with_vk, gstore_kzone, gstore_qzone, gstore_kfilter, gstore_brange, gstore_erange.

By M. Giantomassi (MR870)

D. Other changes (or on-going developments, not yet finalized)¶

D.1 New tutorial on the use of the Z2pack postprocessor (z2pack tutorial). In this tutorial, the topological transition from Z2 trivial to Z2 non-trivial under pressure is reproduced. A test has been added but one needs to install z2pack to test it. See doc/tests/tutoplugs/z2.py, and z2pack tutorial.

From O. Gingras, V. Brousseau-Couture and M. Cote (MR 871)

D.2 Improved the Wannier90 tutorial (wannier90 tutorial). There is a new part, that describes projecting the band structure of La $_2$ CuO $_4$ to a single orbital ( $d_{x^2-y^2}$ ) Hamiltonian. This is the first step in order the study the Mott transition in La $_2$ CuO $_4$ using DFT+DMFT. From this step, one can use TRIQS to study the Mott transition.

From O. Gingras (MR 871)

D.3 Make abinit compile and run with libxc v6

From M. Torrent (MR885)

D.4 Modified optcell=4,5,6 to allow for relaxation of a vectors length and angle without constraining it to be orthogonal to the three others

From C. Paillard (MR858)

D.5 New input variable for Optic, nband_sum@optic, allowing to select the maximal number of bands included. This is convenient for convergence studies.

From M. Giantomassi (MR817)

D.6 Implementation of Chebyshev filtering algorithm, version 2.

From B. Sataric, J. Bieder, and M. Torrent (MR 826)

D.7 Miscellaneous improvements of the iterative Boltzmann transport equation coding. Input variable ibte_alpha_mix.

From M. Giantomassi (MR821)

D.8 Improvements related to electron-phonon interaction.

Miscellaneous improvement of the electron-phonon part of ABINIT (documentation, bug fixes, improved parallelism)

Work in progress : a new tutorial eph4isotc tutorial to demonstrate the computation of superconducting properties within the isotropic Eliashberg formalism. See tests in the eph4isotc tutorial.

By M. Giantomassi (MR870)

D.9 Improvements of the cRPA determination of the U and J parameters (default keywords, tests and tutos).

From R. Outerovich (MR835)

D.10 Work on Real-time Time-Dependent Density Functional Theory implementation within ABINIT.

From F. Brieuc (MR853)

D.11 Test of the Zero-point renormalization for hexagonal systems (Frohlich generalized model). See v8[60].

From B. Guster (MR815)

D.12 Implement chksymtnons=3 : FFT grid and spatial symmetry operations are coherent.

From X. Gonze (MR828)

D.13 Improvement of VdW-DF. Corrections to the implementation of vdW-DF non-local functional found in files 56_xc_/m_xc_vdw.F90 and 98_main/vdw_kernelgen have been done, also several debugging lines have been included. From a careful comparison of the computed quantities from a pre-calculated density and gradient obtained with Siesta, it has been observed that Abinit is getting the correct numbers up to the final 3D FFTs which still present differences that lead to incorrect values for the vdW-DF correlation energy.

From C. Espejo (MR829)

D.14 Make use of gvnlxc optional in getghc, which allows memory savings in lobpcg2 and chebfi2. Could be useful in other parts of the code.

From L. Baguet (MR831)

D.15 Miscellaneous changes in DMFT: keyword for Wannier orthonormalisation, implementation of calculation of the weight of configuration in CTMQC, double counting for chargeonly DFT+DMFT (dmft_dc=6), work in progress concerning alternate calculation of electronic entropy in DMFT.

From B. Amadon and R. Outerovich (MR833)

D.16 Move DDK reading outside of loop for non var matrix element calculations. Should be much more efficient IO.

From M. Verstraete (MR840)

D.17 Use wfdgw_t subclass in GW/BSE code. This is needed so that the wfd in EPH does not allocate bks_tab whose size scales badly with nprocs and nkibz.

From M. Giantomassi (MR841)

D.18 Add spinat to GSR.nc

From M. Giantomassi (MR842)

D.19 Orbital magnetism progress

From J. Zwanziger (MR847)

D.20 Fix prtwf and prtpot in DFPT

From M. Giantomassi (MR848).

D.21 Improved ABINIT+TRIQS python invocation interface.

From O. Gingras (MR851)

D.22 Bug fixes for PAW+Hybrid

From F. Bruneval and F. Soubiran (MR854)

D.23 Improvements of the Frohlich model implementation (e.g. dielectric average decomposition).

From B. Guster (MR860)

D.24 Bug fix of DFT+U in the non-collinear case

From E. Bousquet (MR881)

D.25 GPU coding : inverse overlap matrix, and non-local operator

From P. Kestener (MR843 and 869)

D.26 Restructuring of the tutorial index page, doci/tutorial/index.md ..

From X. Gonze

D.27 A CITATION.cff file has been created. Also, a LICENCE file (pointing toward COPYING).

From X. Gonze

D.28 Updated the tarball for the fallbacks

From JM Beuken (MR911)

D.29 Improve the procedure to examine the convergence with respect to ecut and pawecutdg in tutorial PAW1.

From X. Gonze (20230424)

D.30 Miscellaneous additional bug fixes, typos fixes, or upgrade of build system.

By F. Goudreault (MR816), M. Giantomassi (MR821 and 845), P. Kestener (MR827 and 843), A. Blanchet (MR832), C. Paillard (MR834), M. Verstraete (MR837), M. Torrent (MR838 and 873), B. Seddon and X. Gonze (MR839 and 855), L. Baguet (MR857), J.-M. Beuken (MR882).

v9.6¶

Version 9.6, released on October 4, 2021. List of changes with respect to version 9.4.

Many thanks to the contributors to the ABINIT project between February 2021 and September 2021. These release notes are relative to modifications/improvements of ABINIT v9.6 with respect to v9.4.

The list of contributors includes: L. Baguet, J.-M. Beuken, J. Bieder, A. Blanchet, J. Clerouin, C. Espejo, M. Giantomassi, O. Gingras, X. Gonze, F. Goudreault, B. Guster, Ch. Paillard, Y. Pouillon, M. Rodriguez-Mayorga, M. Royo, F. Soubiran, M. Torrent, M. Verstraete, J. Zwanziger.

It is worth to read carefully all the modifications that are mentioned in the present file, and examine the links to help files or test cases. This might take some time …

Xavier

A. Important remarks and warnings.¶

(nothing to mention for this v9.6)

B. Most noticeable achievements¶

B.1 Band-parallel implementation of DFPT: the memory footprint is now distributed over different processors. Previously, the memory was distributed only for k-point parallelism. This is automatically managed, no user action is to be taken to activate this memory saving.

See test dfpt[04].

By M. Verstraete (MR784, 803)

B.2 The Iterative Boltzmann Transport Equation (IBTE) to compute the electric conductivity has been implemented. To activate the IBTE, use ibte_prep = 1 with eph_task -4. The IBTE solver can also be invoked in standalone mode by providing a SIGEPH file with eph_task = 8. Related input variables: ibte_niter, ibte_abs_tol and ibte_alpha_mix. See test v9[65].

By M. Giantomassi (MR794)

B.3 The computation of dynamical quadrupoles and flexoelectricity is now available within the GGA. Test for GGA + longwaves v9[46].

Also, the usage of the quadrupoles has been rationalized (and made easier) in anaddb as the default value of dipquad@anaddb and quadquad@anaddb has been changed to 1. This means that dipole-quadrupole and quadrupole-quadrupole contributions are always included in the Fourier interpolation of the dynamical matrix provided the DDB provides these terms.

This “default behaviour” is similar to the one used for the dipole-dipole treatment. Indeed, the default value of dipdip@anaddb is 1 hence the dipolar term is automatically included if the DDB contains the Born effective charges and the electronic dielectric tensor. Still, the user can deactivate the inclusion of the different terms by setting the corresponding variable to zero for testing purposes. See the ANADDB input variables and test lw[6]

By M. Royo with contribution by M. Giantomassi for the change of default (MR795)

B.4 Stresses are available within cDFT (constrained DFT). See tests v9[01], v9[02] and v9[03].

By X. Gonze (MR802)

B.5 The computation of effective mass renormalization due to electron-phonon coupling, treated in the generalized Frohlich model, is now available, for cubic materials. An article has been submitted, see . Activate it using eph_task=10.

See test v9[66].

By B. Guster (MR800)

B.6 Important speed-up of the PAW calculations is allowed thanks to the storage of “cprj” coefficients. See the input variable cprj_update_lvl. However, at present this is only possible for ground-state calculations, with several restrictions, spelled in cprj_update_lvl. So, this is not activated by default. There is also an internal variable %cprj_in_memory exposed in the documentation. Other input variables have been introduced in the development process : fft_count and nonlop_ylm_count. They allow one to monitor better the number of FFTs and non-local operator applications.

See tests v9[71], v9[72], v9[73] and v9[74].

By L. Baguet (MR793).

B.7 The Extended First-Principles Molecular Dynamics has been implemented. This method allows one to drastically reduce the needed number of bands for high temperature simulations, using pure single plane waves description based on the Fermi gas model beyond explicitly computed bands. The implementation and usage will be described in an upcoming paper which is currently under review (Authors: A. Blanchet, J. Clérouin, M. Torrent, F. Soubiran).

See topic_ExtFPMD, as well as the input variables useextfpmd and extfpmd_nbcut, and test v9[92].

By A. Blanchet, J. Clérouin, M. Torrent, F. Soubiran. (MR788).

C. Changes for the developers (including information about compilers)¶

C.1 Supported compilers

gfort (GNU) compiler: v11 newly supported.
ifort (Intel) compiler: v21.4 newly supported. Two new bots introduced in the test farm : alps_intel_21.4_elpa and graphene_gnu_11.2_macports .

By JM Beuken

D. Other changes (or on-going developments, not yet finalized)¶

D.1 New input variable for “optic”: prtlincompmatrixelements@optic.

Added this flag in order to make it possible to print the different elements that are used to build the susceptibility of the linear component of the dielectric tensor. These elements are namely: the matrix elements, the renormalized electronic eigenvalues, the occupations and the kpt weights. Everything is dumped into the _OPTIC.nc file by the main process. Thus optimization could be done memory-wise and speed wise if MPI-IO is implemented for this nc file.

v9[49] was created which is the same as v9[48] except with the aforementioned flag set to 1. This test checks that everything works well even though we print the matrix elements. It does not test that matrix elements are well printed because that would require testing of the OPTIC.nc file. Although it is possible to check that it works well using a simple python script (see the figure in the merge request on Gitlab). (Note that the tests v9_13 and v9_14 have been moved to v9_47 and v9_48 in this change).

By F. Goudreault (MR776)

D.2 New radial sine transform for the vdW-DF kernel.

By C. Espejo (MR 797)

D.3 New test of orbital magnetism, v9[37]. Also, on-going work on orbital magnetism, including use with DDK wavefunctions.

By J. Zwanziger (MR767, MR775, MR779 and MR787)

D.4 Migration to mkdocs==1.1.2 and mkdocs-material==7.0.6. mksite.py now requires python >= 3.6 . Activated search capabilities, available in the new mkdocs version.

By M. Giantomassi (MR774)

D.5 Fixed bug in make_efg_onsite for nspden=2 case.

By J. Zwanziger (MR783)

D.6 Correction of tutorials Rf1 and Rf2 for version 9

By O. Gingras (MR785)

D.7 Fixed errors and bugs detected by using -ftrapuv intel option

By M. Giantomassi (MR789)

D.8 Bug fix in nspden=4 DFPT for Fe

By M. Verstraete (MR790)

D.9 Fixed a spurious test line 711 of m_occ.F90 that caused abinit to abort in the case of occopt=9, if the number of conduction bands was enough to accommodate nqFD but not enough to accommodate nelect.

By Ch. Paillard (MR791)

D.10 GW methodology with Kohn-Sham density matrix. Solving a bug producing a segmentation fault when using bdgw and gw1rdm. New test v9[37].

By M. Rodriguez-Mayorga (MR792)

D.11 Introduced new input variable use_oldchi. This input variable is temporary, for testing purposes. It is documented, but not tested.

By Wei Chen (modified line 743 in src/95_drive/screening.F90 on 23 April 2021).

D.12 The input variable rfstrs_ref has been introduced, but not yet documented and tested, as this is on-going work.

By M. Royo

D.13 Miscellaneous additional bug fixes, or upgrade of build system. in the upgrade of tutorials). By J. Bieder, M. Giantomassi, Y. Pouillon, M. Torrent, J. Zwanziger.

v9.4¶

Version 9.4, released on February 25, 2021. List of changes with respect to version 9.2.

Many thanks to the contributors to the ABINIT project between November 2020 and April 2021. These release notes are relative to modifications/improvements of ABINIT v9.4 with respect to v9.2.

The list of contributors includes: B. Amadon, L. Baguet, J.-M. Beuken, J. Bieder, E. Bousquet, V. Brousseau, F. Bruneval, W. Chen, M. Cote, M. Giantomassi, O. Gingras, X. Gonze, F. Goudreault, B. Guster, T. Karatsu, A. H. Larsen, O. Nadeau, R. Outerovich, Ch. Paillard, G. Petretto, S. Ponce, Y. Pouillon, G.-M. Rignanese, M. Rodriguez-Mayorga, M. Schmitt, M. Torrent, M. Verstraete, He Xu, J. Zwanziger.

It is worth to read carefully all the modifications that are mentioned in the present file, and examine the links to help files or test cases. This might take some time …

Xavier

A. Important remarks and warnings.¶

A.1 The charge variable is obsolete, and has been replaced by cellcharge. Indeed, charge was quite ambiguous, and the string charge present in some other input variables. For the time being, ABINIT still recognizes charge in the input file, but this might not last longer than in ABINITv9.

A.2 There is a new check, governed by the input variable chksymtnons, to examine whether the tnons of all symmetry operations is zero or a rational number with small denominator, which is required for GW calculations as implemented in ABINIT. It is always possible to choose the origin of coordinate with such properties, and ABINIT can gives suggestions. If you do not want to change your coordinate origin (e.g. you have no intention to perform a GW calculation), set chksymtnons to zero, or (but this is more dangerous) set the meta-variable expert_user to one to disable several checks of input variables at once.

By X. Gonze (MR712)

A.3 The input variable npkpt has been changed to np_spkpt . Indeed the parallelism governed by npkpt was about spin and k points, not only k points. For the time being npkpt is still admitted, but will become obsolete at the next major version change.

By X. Gonze

A.4 When nimage>1, the default value of prtgsr is now 0, like for several prt* variables.

By X. Gonze

A.5 The code does not stop anymore at the first occurence of overlap between PAW spheres being larger than pawovlp in case of ionmov/=0 or imgmov/=0, but only at the second occurrence per dataset. Indeed, such trespassing might only be transient. See the description of pawovlp.

By X. Gonze

B. Most noticeable achievements¶

B.1 The RMM-DIIS algorithm has been implemented. This SCF (ground-state) algorithm is faster, but potentially more unstable, than the CG or LOBPCG algorithms, for medium to large size systems, as it has less cubic scaling steps (e.g. orthogonalisation). Typically used for molecular dynamics or structural relaxations as the restart from the previous time step gives RMM-DIIS less opportunities to fail. See input variables rmm_diis and rmm_diis_savemem. Several tests exist (paral[32], paral[63], paral[64], v9[29], v9[30]) covering many cases, including NC/PAW and spin-orbit. Note that the PAW version of RMM-DIIS is more unstable than the NC one and extra operations are needed to make it convergence. So the speedup for PAW calculations is not as good as the one observed for NC.

By M. Giantomassi (MR757, MR719, MR718)

B.2 The treatment of quasi-Fermi energies in the valence and conduction band, with populations of electrons (in the conduction bands) and holes (in the valence bands) has been implemented (gapped materials only, of course). This has been used e.g. in [Paillard2019]. See the variables occopt=9, and nqfd. See also the related input variable : ivalence. Internal variables ne_qFD and nh_qFD are presently initialized to nqfd, which is NOT INTERNAL. See test v9[91].

By Ch. Paillard (MR755).

B.3 All the tutorials have been carefully reexamined and improved, when appropriate. In particular:

the old pseudopotentials have been replaced by new ones from e.g. pseudodojo or JTH;
the text of the tutorial uses the new convention for launching abinit (e.g. abinit input_file instead of abinit < files_file) and the pseudopotentials are mentioned in the input file;
the input file suffix has been changed from .in to .abiand the output file suffix has been changed from .outto .abo;
the input files have been cleaned when adequate, and many have been restructured using a template; with populations of electrons (in the conduction bands) and holes (in the valence bands)

Also, nlo tutorial and eph4zpr tutorial have been enlarged to new developments (see eph4zpr[8] and nlo[6]).

By B. Amadon, L. Baguet, J. Bieder, E. Bousquet, F. Bruneval, W. Chen, M. Cote, O. Gingras, M. Giantomassi, X. Gonze, F. Goudreault, B. Guster, O. Nadeau, R. Outerovich, S. Ponce, M. Schmitt, M. Torrent, M. Verstraete, He Xu, J. Zwanziger (numerous MRs)

B.4 The GW 1-body reduced density matrix (1RDM) from the linearized Dyson equation has been implemented. Its effect on the Hartree-Fock expectation values and therefore on the GW quasiparticle energies can be evaluated. The resulting total energy parts, kinetic energy (including correlation), electron-nucleus, Hartree, Exchange, can be calculated. Together with the Galitskii-Migdal correlation, it gives a new approximation the self-consistent GW total energy. See input variables gw1rdm, x1rdm, also irdchkprdm, prtchkprdm and gwgmcorr. See tests v9[33] to v9[36].

Also, some missing tests have been added: - GW calculations based on Hartree-Fock wavefunctions can use mini Brillouin Zone integration technique, see v9[31]. - A new test v9[40] has been provided for the computation of the susceptibility matrix $\chi_0$ with inclvkb.

By Mauricio Rodriguez-Mayorga and F. Bruneval (MR722).

B.5 The pSIC (polaron self-interaction corrected) methodology has been implemented. See [Sadigh2015] and [Sadigh2015a]. This is based on the images capability of ABINIT, that has been extended to different values of the input variable cellcharge for different images, and also parallelized. To activate pSIC, use imgmov=6 with the proper occupation numbers. See the test examples v9[22], psic[01], psic[02] and psic[03].

By X. Gonze (initial test from C. Tantardini) (MR770).

B.6 The computation of the electric conductivity has been implemented for metals in the relaxation-time approximation with transport lifetimes computed from the imaginary part of the Fan-Migdal self-energy. See tests v9[62] to v9[65].

By O. Nadeau (MR756, MR716)

B.7 Implementation of the i-pi client-server protocol as described in [Kapil2019]. This option requires ionmov 28 and the specification of the socket via command line options. For UNIX socket, use: –ipi {unixsocket}:UNIX . For INET socket, use –ipi {host}:{port} . Usage example:

 abinit run.abi --ipi {unixsocket}:UNIX > run.log

Note that, at present, this feature is mainly used to interface ABINIT with the ASE optimization routines. Moreover the user is responsible for creating an input file with tuned tolerances to prevent Abinit from exiting when internal convergence is reached. See examples available in the ASE documentation

By M. Giantomassi and A. H. Larsen.

C. Changes for the developers (including information about compilers)¶

C.2 Test farm: new and obsolete bots

Bots introduced in the test farm: scope_gnu_10.2_paral
Bots removed: cronos2 (replaced by scope_gnu_10.2_paral)

Bigdft tests have been activated on ALPS.

By JM Beuken

C.3 Supported compilers

gfort (GNU) compiler: v10 newly supported, v5 obsolete
ifort (INTEL) compiler: v15 obsolete (but to be reintroduced for licence reasons)

Support for AOCC has been added in the build system.

By JM Beuken

D. Other changes (or on-going developments, not yet finalized)¶

D.1 Calculation of Luttinger parameters (in the Frohlich model) and echo (see v8[57]). By V. Brousseau (MR736)

D.2 New test v9[90] of the treatment of the Coulomb interaction for low dimensional materials (0D and 2D). By B. Guster.

D.3 DFPT (including ddk perturbation) can now be done in the presence of nucdipmom. By J. Zwanziger (MR749)

D.4 Bug fix and new test v9[43] for the use of ANADDB in the presence of a large tolsym value. By G. Petretto

D.5 Several bug fixes related to the treatment of inaccurate atomic positions (and large tolsym). Several test have been created (v9[17] to v9[20]).

D.6 AiiDA+ABINIT developments

A AiiDA plugin for Abinit has been developed: https://github.com/sponce24/aiida-abinit, also indexed at https://aiidateam.github.io/aiida-registry/.
AiiDA is now supporting psp8 type pseudopotentials, https://github.com/aiidateam/aiida-pseudo. This also implied modifications of the pseudodojo, that now includes .djrepo for each type of pseudopotential.
Work on a common relaxation workflow for a dozen of codes, including ABINIT.

By S. Ponce, also with G.-M. Rignanese, G. Petretto, M. Giantomassi.

D.7 The new input variable dmft_wanorthnorm has been introduced, see v6[07] and v6[46]. However, it should still be documented. By B. Amadon.

D.8 Increase stack size limit inside xmpi_init using POSIX C-API By M. Giantomassi. MR 770.

D.9 Document i-pi interface with links to ASE docs. By M. Giantomassi. MR 770.

D.10 Correction (from a message on the forum) related to the forces in electron-positron mode. By M. Torrent. MR 780

D.11 Miscellaneous additional bug fixes, improvements of documentation including for the build system (many other were made in the upgrade of tutorials).. By B. Amadon, L. Baguet, F. Bruneval, T. Karatsu, G. Petretto, Y. Pouillon, M. Torrent, J. Zwanziger.

v9.2¶

Version 9.2, released on September 30, 2020. List of changes with respect to version 8.10. Release notes updated on November 10, 2020.

Many thanks to the contributors to the ABINIT project between October 2018 and November 2020. These release notes are relative to modifications/improvements of ABINIT v9.2 with respect to v8.10. Merge requests up to and including MR692 are taken into account, also MR 697-702, 705, 707-710, 712, 715.

The list of contributors includes: B. Amadon, G. Antonius, L. Baguet, J.-M. Beuken, J. Bieder, J. Bouchet, E. Bousquet, F. Bruneval, G. Brunin, Wei Chen, J.-B. Charraud, Ph. Ghosez, M. Giantomassi, O. Gingras, X. Gonze, F. Goudreault, B. Guster, G. Hautier, Xu He, N. Helbig, F. Jollet, H. Miranda, F. Naccarato, R. Outerovitch, G. Petretto, N. Pike, Y. Pouillon, F. Ricci, M. Royo, M. Schmitt, M. Stengel, M. Torrent, J. Van Bever, M. Verstraete, J. Zwanziger.

It is worth to read carefully all the modifications that are mentioned in the present file, and examine the links to help files or test cases. This might take some time …

Xavier

A. Important remarks and warnings.¶

A.1 At the occasion of the switch from ABINITv8 to ABINITv9, many improvements of the formats and content of files written by ABINIT have been made, so the backward compatibility of ABINITv9 may be broken.

In particular:

The build system relies on new .ac9 files (see B.6), superceeding the v8 .ac files. A bash script (upgrade-build-config-file.sh) located in the top level directory of the package can be used to convert from the old .acformat to .ac9.
The build system of ABINITv9 does not build anymore the hard dependencies (Linalg, NetCDF4, HDF5, LibXC, …), as this was not sustainable (see B.6) and nowadays most users install prerequisite libraries themselves. See also the specialized INSTALL notes for CentOS, EasyBuild, MacOS, and Ubuntu.
The main ABINIT output file now contains sections written in YAML (sometimes replacing text sections, sometimes adding information). This means that some user-developed parsing tools might not work anymore, and should be adapted to the new ABINITv9 output file (see B.9). Note that the YAML output is still under development and modifications may appear in the next versions. A python API to extract the results of the calculation will be provided when the implementation is finalized.
Several default values have been changed, see A.3.

A.2 A new account of the ABINIT effort has been published in Computer Phys. Comm. [Gonze2020]. It provides description of several new features. A version of this paper that is not formatted for Computer Phys. Comm. is also available. The licence allows the authors to put it on the Web.

A second new account of the ABINIT effort has been published in J. Chem. Phys. [Romero2020]. The scope of this second paper is different from the first one. It is more a survey of ABINIT, focusing on its specific capabilities. Still, it contains also some description of some new features. A version of this paper that is not formatted for J. Chem. Phys. is also available. The licence allows the authors to put it on the Web.

Other specific publications are mentioned in the Suggested acknowledgment page.

A.3 The default values of the following ABINIT input variables have been changed: ixcrot, chneut, ntime, prtkden, symsigma and tolsym. In particular the new default value of tolsym, 1e-5, is more in line with the tolerances of other codes, so that for users of such codes, one barrier to the use of ABINIT is removed. By the same token, some bug in the recognition of symmetries has been fixed, when tolsym is close to the default, see the new tests v9[15] and v9[16]. The new input variable chksymtnons has been introduced, to govern the possible automatic alignment of the tnons with the FFT grid (actually needed for GW calculations). By X. Gonze (MR 689 and others)

A.4 The initialization of the wavefunctions when paral_kgb=1 and nspinor=2 has been changed, since the previous one could prevent the code to converge. By M Torrent (MR 562).

A.5 The input variable xangst has been disabled. Use xcart instead, and specify the unit, namely Angstrom.

A.6 Work is on-going concerning the Coulomb singularity treatment, see D.32. The usage of the input variable icutcoul is changing. For the time being, use gw_icutcoul instead.

A.7 The name of the t-DEP main executable has been changed from tdep to atdep, in line with [Romero2020]. By J. Bieder (MR 642, 641).

B. Most noticeable achievements¶

B.1 Electron-phonon interaction (mobilities in the self-energy relaxation time approximation, temperature-dependent electronic band structures including the zero-point renormalization, etc.)

The new capabilities of ABINITv9 related to electron-phonon calculations are described fully in the Sec. 3.3.2 of [Gonze2020], as follows.

In abinit v9, it is possible to compute the EPH self-energy in the Kohn–Sham representation using the EPH matrix elements. The code employs optimized algorithms to compute either the full self-energy (needed for QP corrections and spectral functions) or just the imaginary part that is then used to evaluate mobilities within the self-energy relaxation time approximation (SERTA). The computation of the mobility is fully integrated inside abinit, and is an automatic output of the computation of the imaginary part of the self-energy, bypassing the need to post-process results. When computing the full self-energy, it is possible to reduce the number of empty states required for convergence by using the first-order wavefunctions obtained by solving the relevant Sternheimer equation.

In the case of lifetime computations, the code takes advantage of the tetrahedron method to filter contributing q-points, a double-grid integration technique to accelerate the convergence at marginal additional computational cost, and samples the relevant regions in the Brillouin zone contributing to transport properties thus leading to a significant reduction of the computational effort. Crystalline symmetries are used throughout the code in order to reduce the number of k- and q-points that must be explicitly included in the integrals. To achieve good parallel efficiently, the most CPU demanding parts are parallelized with MPI employing a distribution schemes over k/q-points, perturbations and bands (the band level is available only when computing the full self-energy).

Moreover, the interpolation of the DFPT potential, described in Sec. 3.3.1 of [Gonze2020] is fully operational, with many tests, and three tutorials provided.

List of tests: v8[44], v9[50], v9[53], v9[56], v9[60], v9[61], eph4mob[1], eph4mob[2], eph4mob[3], eph4mob[4], eph4mob[5], eph4mob[6], eph4mob[7], eph4zpr[1], eph4zpr[2], eph4zpr[3], eph4zpr[4], eph4zpr[5], eph4zpr[6], eph4zpr[7].

New input variables: brav, dvdb_add_lr, dvdb_qcache_mb, dvdb_qdamp, dvdb_rspace_cell, eph_doping, eph_phrange, eph_tols_idelta, eph_ecutosc, eph_restart, eph_stern, eph_use_ftinterp, eph_phwinfact, getdvdb, getdvdb_filepath, getkerange_filepath, getsigeph_filepath, irddvdb, prteliash, rifcsph, sigma_bsum_range, sigma_erange, sigma_ngkpt, sigma_nshiftk, sigma_shiftk, symv1scf.

Note that the new EPH processing unit of ABINIT optdriver=7 has a different implementation than the one implemented in anaddb. Three new tutorials are availables, eph_intro tutorial, eph4mob tutorial and eph4zpr tutorial, and supercede the legacy tutorials eph_legacy tutorial and eph_tdep_legacy tutorial. For further details about the implementation and usage, please consult [Brunin2020b].

By G. Brunin, H. Miranda, M. Giantomassi, G.-M. Rignanese, G. Hautier.

B.2 Flexoelectricity and dynamical quadrupoles

A new driver has been included in abinit that allows one to compute 4 spatial dispersion tensorial quantities: the clamped-ion flexoelectric tensor, the dynamical quadrupoles, the first moment of IFC matrix and the first moment of the piezoelectric force response tensor. Precalculation of ground state, first and second (d2_dkdk) order response functions is required. After execution, the driver creates a 3^rd order energy derivative database file that is used by anaddb to compute the mixed and lattice-mediated flexoelectric tensors or to include the dipole-quadrupole and quadrupole-quadrupole electrostatic interactions in the calculation of the dynamical matrix.

See the complementary description in the Sec. V. D of [Romero2020], with underlying theory and test calculations presented in [Royo2019]. At the practical level, see [Romero2020]:

In this way, both perturbations are generalized to finite q, as is already the case for atomic displacements. This enables us to carry out an analytical third order derivative of the energy with respect to two of the standard perturbations, and to the momentum q, which directly provides the sought-after spatial dispersion tensors. Remarkably, by virtue of the 2n+1 theorem, the third-order energies are computed in one shot using precalculated first-order response functions to the standard perturbations, without the necessity of self-consistently computing any response function to a perturbation gradient. After execution, the long-wave DFPT routines generate a derivative database that is subsequently used by post-processing tools implemented in ANADDB to compute and print the different contributions to the FxE tensor.

The dynamical quadrupoles are the spatial dispersion counterparts of the Born effective charges, and can be used in lattice dynamics calculations to improve the prevalent dipole-dipole treatment of the long-range interactions. The ANADDB routines that carry out the process of interpolating the dynamical matrix following Ref. 34 have been adapted to incorporate the dipole-quadrupole and quadrupole-quadrupole electrostatic interactions derived in Ref. 102. This new functionality results in a faster convergence of the phonon bands calculation with respect to the density of q points and, in some materials, represents the only route to obtain the correct sound velocities.

Currently, the implementation is restricted to the use of norm-conserving pseudopotentials without non-linear core corrections, and the LDA functional.

A tutorial is in preparation, with tests lw[1] to lw[7].

See the topic_longwave. The relevant input variable is optdriver==10. New input variables have been defined: lw_flexo, lw_qdrpl, prepalw, flexoflag@anaddb, dipquad@anaddb, quadquad@anaddb.

This capability is still under development and not completely stable. Interested users are strongly recommended to contact Miquel Royo (mroyo@icmab.es) or Massimiliano Stengel (mstengel@icmab.es) before start using it.

By M. Royo, M. Stengel

B.3 DFT+DMFT

The new capabilities of ABINITv9 related to DFT+DMFT calculations are described fully in the Sec. 3.7 of [Gonze2020], as follows.

The DFT+DMFT parallelism was improved for large systems. In particular, it is now possible to parallelize the calculation on both k-points and bands/g-vectors by using the input variable paral_kgb = 1 and related input variables.

Two new approaches to CT-QMC have been added to solve the AIM. In the first one, the density–density CT-QMC code available in abinit [Gonze2016], [Bieder2014] was generalized in order to take into account off-diagonal elements of the hybridization function. This implementation is activated with the input variable dmft_solv = 8. Spin–orbit coupling calculations are possible, but using a real valued imaginary time hybridization function. This solver was used in Refs. [Amadon2015], [Amadon2016].

In the second approach, we use the Toolbox for Research on Interacting Quantum System (TRIQS)library [Parcollet2015], which is an open-source project that provides a framework for many-body quantum physics and more specifically for strongly-correlated electronic systems. TRIQS provides an open source implementation of the continuous-time hybridization expansion quantum impurity solver (CT-HYB) [Seth2016], considered a state-of-the art solver for multi-orbital AIM. An interface between abinit and the impurity solver TRIQS/CT-HYB is now available and will make use of the independent progress made by the TRIQS library. <…>

Also, the DMFT k-resolved spectral function is available (MR 529, 490).

List of tests: paral[84], paral[86], paral[99], v8[01]. New input variables: dmft_charge_prec and dmft_kspectral_func (test to be provided for the latter). Also dmft_occnd_imag, but only for keeping backward compatibility for tests.

By T. Cavignac, B. Amadon and O. Gingras.

B.4 Spin model within Multibinit

The new capabilities of Multibinit within ABINITv9 are described fully in the Sec. 4.1 of [Gonze2020]. See also Sec. D.1. In particular, a spin model, described specifically in Sec. 4.1.2 of [Gonze2020], is available, as follows.

Multibinit implements the most commonly used model for spin systems, via the Heisenberg Hamiltonian including magnetic exchange and Dzyaloshinskii Moriya interactions. Single ion anisotropy and dipole–dipole interactions are also included, and all terms bear a very strong similarity to the quadratic part of the lattice model Hamiltonian. A number of open source spin dynamics codes already exist, such as UPPASD, VAMPIR, OOMF; the distinguishing features of multibinit are the integration with abinit, to fit parameters, and the simultaneous dynamics with other degrees of freedom (in particular using the inter-atomic force constants).

A tutorial for the multibinit spin model has been written, spin_model tutorial.

Many new input variables are present, tested and documented: slc_coupling, spin_calc_thermo_obs, spin_damping, spin_init_orientation, spin_init_qpoint, spin_init_rotate_axis, spin_init_state, spin_ntime_pre, spin_projection_qpoint, spin_sia_add, spin_sia_k1amp, spin_sia_k1dir, spin_temperature_start, spin_temperature_end, spin_temperature_nstep, spin_var_temperature, spin_write_traj.

List of tests in addition to those of the tutorial: v8[16], v8[23], v9[81], v9[82], v9[86], v9[87].

By Xu He, N. Helbig, J. Bieder, E. Bousquet, Ph. Ghosez, M. Verstraete

B.5 Constrained DFT

Constrained Density-Functional Theory (see topic_ConstrainedDFT) is available, with a new algorithm allowing to impose the constraints to arbitrary precision, whether it relates to the charge, magnetization, magnetization direction, or magnetisation size, or a combination thereof for different atoms. The constraints are smeared spherical integrals with ajustable sphere radius, centered on atoms. The algorithms has been demonstrated for norm-conserving pseudopotentials as well as PAW. Forces and derivatives with respect to the constraints are available (i.e. magnetic torque for the non-collinear spin case). Stresses are still to be coded, will be available in ABINITv9.4 or ABINITv9.6.

New tests: v8#24-29, v8#95-97 and v9#1-3. New input variables: chrgat, constraint_kind, ratsm.

By X. Gonze.

B.6 Large modifications of the build system

The build system relies on new .ac9 files, superceeding the v8 .ac files. Fully documented example files can be found in doc/build/config-examples. A bash script (upgrade-build-config-file.sh) located in the top level directory of the package can be used to convert from the old .acformat to .ac9.

The build system of ABINITv9 does not build anymore the (hard and soft) dependencies (Linalg, NetCDF4, HDF, LibXC, Wannier90, …), as this was not sustainable. Three libraries are now mandatory: linalg, NetCDF4/HDF5 and LibXC. Failing to link to them will prevent building ABINIT. The other libraries are optional, there will only be a warning if they are not available. If the user does not provide the path to these libraries, the build system will try to find them in the “usual” directories, and inform the user that it has done so. The build system also can make suggestions to the user, to complete its *.ac9 file.

Specialized INSTALL notes are available to help the user for CentOS, EasyBuild, MacOS, and Ubuntu.

By Y. Pouillon and JM Beuken

B.7 New command line interface

There is a new (recommended) command line interface to run ABINIT, without the “files” file. The new syntax is:

abinit run.abi

or

abinit run.abi > run.log 2> run.err &

where run.abi is the Abinit input file that now provides all the information related to pseudos and the prefixes that were previously passed via the “files” file. For comparison, the old syntax is

abinit < run.files > run.log 2> run.err &      ! This is the old syntax

A file extension for the input file is highly recommended (in this example we use .abi) as by default the parser will use the string before the file extension as root to build the prefixes for the input/output/temporary files.

The user can specify the name of the main output file thanks to the output_file input variable, the list of pseudopotentials thanks to the pseudos input variable and the directory where all pseudos are located with pp_dirpath. The prefix for other input, output or temporary files can be specified with indata_prefix, outdata_prefix and tmpdata_prefix, respectively. A default set of prefixes computed from the basename of the input file is used if these variables are not specified in the input.

For some examples, see tests v8[90], v7[45], and v5[54]. See also topic_Control.

A similar command line interface can also be used for the anaddb code. In this case, the relevant variables are: output_file@anaddb, ddb_filepath@anaddb, ddk_filepath@anaddb, gkk_filepath@anaddb, eph_prefix@anaddb.

The new syntax is:

anaddb run.in > run.log 2> run.err &

See tests v8[52] for a standard analysis of the DDB file and v7[94] for the (old implementation) of electron-phonon calculations in anaddb.

Important

The old “files file” interface is still operational although deprecated and will be REMOVED in Abinit v10.

By M. Giantomassi (MR 586).

B.8 Reading strings from the input file

A new mechanism to read strings enclosed between double quotation marks from the input file has been activated. So, many new input keywords are reading strings as data, and, often, can be used alternatively to similar input keywords that were expecting numerical values such as the get* and ird* variables. The goal is to encourage a new approach for performing ABINIT calculations in which multiple datasets and get* variables are replaced by independent input files that are connected together via file paths.

List of new input variables that rely on this feature:

getddb_filepath, an alternative to getddb or irdddb, see test v9[60]
getden_filepath, an alternative to getden or irdden, see test v8[36] and v8[41]
getscr_filepath, an alternative to getscr or irdscr, see test v67mbpt[51]
getwfkfine_filepath, an alternative to getwfkfine or irdwfkfine, see tests v9[55], v9[56]
getwfk_filepath, an alternative to getwfk or irdwfk, see test v9[60]
getwfq_filepath, an alternative to getwfq or irdwfq, see test v7[98]
getkerange_filepath, see test v9[60]
getpot_filepath, see test v8[44]
pseudos, indata_prefix, outdata_prefix, tmpdata_prefix, output_file, see test v9[04]
pp_dirpath: Used in all the input files of the Test Suite
output_file@anaddb, ddb_filepath@anaddb, gkk_filepath@anaddb, eph_prefix@anaddb. See tests v8[52] for a standard analysis of the DDB file and v7[94] for the (old implementation) of electron-phonon calculations in anaddb.

By M. Giantomassi

B.9 YAML sections in the output file

YAML sections are now generated in the output file, sometimes replacing text sections, sometime providing new information. At present there is a YAML section for the components of the total energy, the GS results including forces and stresses as well as a YAML section for GW calculations, and some YAML sections giving information about the iteration status.

At the occasion of the development of this capability, and its adaptation to the test farm, the perl script fldiff.pl has been replaced by a Python version. See related information in Sec. 5.5 of [Gonze2020].

By T. Cavignac, M. Giantomassi, GM Rignanese, X Gonze.

B.10 New approach to define crystalline structures in the Abinit input

The new variable structure can be used to initialize the lattice vectors and the atomic positions from an external file. Variables such as natom, ntypat, typat and znucl are automatically initialized and need not to be specified in the ABINIT input. At present, the code can read netcdf files produced by ABINIT (GSR.nc, WFK.nc, DEN.nc, HIST.nc) and POSCAR files in VASP-5 format. See the documentation for the syntax and limitations.

By M. Giantomassi.

B.11 New capabilities of abipy and abiflows

The abipy and abiflows projects have been significantly extended. See Sec. 6 of [Gonze2020], as well as the gallery of plotting scripts and the gallery of abipy workflows .

By M. Giantomassi, G. Petretto, F. Naccarato.

C. Changes for the developers (including information about compilers)¶

C.1 A python script to help ABINIT developers and development.

The new python script abisrc.py located in the top directory of the ABINIT package has been developed. It has superceded abilint.py in the makemake procedure.

Try

./abisrc.py --help

then follow the suggestions, to get info about files, directories, interfaces, to visualize dependencies of the ABINIT subroutines, etc.

Note that there are dependencies of abisrc.py, to be installed prior being able to use some of its capabilities. Use:

pip install -r requirements.txt --user

to install the dependencies in user mode.

By M. Giantomassi.

C.2 New characteristics of input variables

In the description of input variables (e.g. files abimkdocs/variables_abinit.py), a new field added_in_version has been introduced, for example,

 added_in_version="9.2.0"

or, for variables introduced prior to v9,

 added_in_version="before_v9"

C.2 Test farm: new and obsolete bots

Bots introduced in the test farm: alps_nag_7.0_openmpi, atlas_gnu_9.1_openmpi, buda2_gnu_8.2_cuda, cronos2_gnu_7.4_paral, scope_gnu_10.2_mpich, scope_gnu_7.5_dep.
Bots upgraded: abiref_gnu_5.3_* to abiref_gnu_9.2_; bob_gnu_5.3_openmp to bob_gnu_7.5_openmp; buda2_gnu_8.1_mpich3 to buda2_gnu_8.2_mpich3; graphene_gnu_6.4_macports to graphene_gnu_9.2_macports; max2_gnu_5.3_ to max2_gnu_6.5_*; ubu_gnu_5.3_openmpi to ubu_gnu_9.2_openmpi.
Bots removed: abiref_nag_6.2_openmpi (superceded by alps_nag_7.0_openmpi), atlas_gnu_7.2_fb.ac (no nightly test of the fallbacks anymore), cronos_gnu_5.3_paral(replaced by cronos2), inca_gnu_6.3_py3k (inca too old), tikal_gnu_* (tikal too old). Working also on a cronos-cronos2 cluster.

By JM Beuken

C.3 Supported compilers

gfort (GNU) compiler: v9 newly supported, v4 obsolete
ifort (INTEL) compiler: v19 newly supported.
NAG 7.0 instead of 6.2

By JM Beuken

C.4 Unitary ttransposer#1 . Test of the transposer for linear algebra to KGB parallelisation.

By J. Bieder.

C.5 Linkchecker has been reenabled, only for internal link checking.

By JM Beuken (MR 513).

C.6 Enable the generation of a HTML side-by-side diff on the test farm when fldiff fails with a line count error and it was not caused by a crash of Abinit. The diff algorithms uses a specialized heuristic to improve line synchronization and prevent weird matching.

By Th. Cavignac (MR 526)

C.7 Split of the source tree (ongoing).

In order to improve modularity, the source tree must be split in two parts, one for low-level routines, largely independent of ABINIT, and one for more specific routines to ABINIT. The low-level routines should become a separate library, with its own build system and make. At present the low-level library have been moved out of src, inside the shared/common/src directory. See related information in Sec. 5.4 of [Gonze2020].

C.8 New FFT specifications for the build system See https://gitlab.abinit.org/pouillon/abinit/-/issues/33 .

By Y. Pouillon (MR 619)

C.9 Reorganisation of the tests/Psps_for_tests directory, in preparation of the beautification. Clarification on which pseudopotentials are of recent format, and which are legacy pseudopotentials. Preparation of beautification.

By X. Gonze (MR 683)

D. Other changes (or on-going developments, not yet finalized)¶

D.1 Miscellaneous improvements of Multibinit (lattice part)

Miscellaneous improvements have been made to the lattice part of Multibinit. See the new input variables below, also see the Sec. 4.1.1 of [Gonze2020].

New tests: v8[38], v8[94], v8[98], v8[99], v8[101], v8[102], v8[103], v9[83], v9[84], v9[85]. New input variables are listed below. Not all these new input variables are present in automatic tests, though.

Finally, several input variables of the main ABINIT application are also reused for Multibinit, without modification of meaning, like iatfix, natfix (and related similar input variables), and also tolmxf.

By M. Schmitt, Xu He, F. Ricci, M. Verstraete, Ph. Ghosez

D.2 Miscellaneous improvements in the Chern number and orbital magnetization calculations, including parallelization over k points of the Chern number calculation.

By J. Zwanziger (MR 469, 500, 545, 588)

D.3 Calculation of Debye-Waller tensor. New test v8[58].

By M. Giantomassi

D.4 Speed-up of susceptibility matrix calculations and GW analytic continuation calculations. See the new input variable gwaclowrank and new test v9[32].

By F. Bruneval (MR 687).

D.5 NCPP Wavefunction mixing with Variational Energy and minor improvements to prepare PAW+Hybrid variational energy. New test v7[73], simple system for testing Hartree-Fock and the SCF algorithms.

By X. Gonze (MR 434, 444, 445).

D.6 New weight distribution of the Fourier components of the force constants. Test tolerance in the new integration weights, tests v8[52], v8[53], v8[54].

By H. Miranda and M. Giantomassi

D.7 Test calculation of velocity matrix elements (DDK) with optdriver 8 and wfk_task “wfk_ddk”, see v8[59]. By the way, the other capabilities linked to wfk_task (“wfk_fullbz”, “wfk_einterp”, “wfk_optics_fullbz”, “wfk_kpts_erange”) seem not to have been properly advertised.

By M. Giantomassi

D.8 Upgraded paral_bandpw tutorial, new version of auto paral (with threads)

By M. Torrent (MR502).

D.9 Test wannier90 interface with nsppol=2 and nspden=2, wannier90[04].

By Xu He

D.10 Mixed precision for FFT transforms. New input variable mixprec see v8[44], v9[57], v9[60], and v9[61].

From M. Giantomassi (MR491).

D.11 Multibinit has been interfaced with scale-up.

By Marcus Schmitt, Jordan Bieder, Matthieu Verstraete and Philippe Ghosez

D.12 The following units are now also allowed in input files:

S Sec Second for the ABINIT input file;
nm (for nanometer) for the ABINIT and ANADDB input files.

D.13 aTDEP utility: added aTDEP guide, aTDEP topic, and corresponding input variable documentation. References: aTDEP paper. Also, see Sec. 4.2 of [Gonze2020].

By F. Bottin, J. Bouchet, J. Bieder (MR491,422).

D.14 Improvements of NLO calculations. Optimize memory allocation. In particular for usepead = 1. Write Raman susceptibilities to netcdf in anaddb.

By G. Petretto (MR 599).

D.15 MetaGGA + PAW in progress. All internal and unit tests are OK. The implementation seems close to the end. Still need to perform physical tests.

By M. Torrent and J.-B. Charraud (MR 587, 558, 625).

D.16 Implementation of an alternate version of MPI reductions valid when the number of data exceeds 2^32.

By M. Torrent (MR 587)

D.17 Build system and src support for llvm (clang/flang) and ARM (armflang/armclang/armpl).

By M Torrent (MR 571)

D.18 Improvement of core WF reading (optics + electron-positron)

By M Torrent (MR 557)

D.19 Fix bootstrap kernel convergence issues

By Wei Chen (MR 546)

D.20 Upgrade versions of libraries used with ABINIT. Upgrade atompaw to 4.1.0.6. Upgrade Libxc to 4+. Prepare the interface to LibXC 5 and a bit of LibXC 6.

By M. Torrent, JM Beuken (MR 649, 532, 470, 465, 441)

D.21 Write yaml file for fatbands (phonopy format) with aTDEP

By J. Bieder (MR510)

D.22 Write polarization vectors in GSR.nc.

By H. Miranda (MR 462).

D.23 Updated user interface of Raman_spec.py . Added one section to tutorial NLO to describe use of Raman_spec.py.

By N. Pike (MR 460, MR 581)

D.24 XML format for core wavefunctions

By F. Jollet (MR 423)

D.25 Wavefunction prediction for molecular dynamics.

By F. Jollet (MR 412)

D.26 Added a preview for the toptic_4.abi file in the optic tutorial.

By F. Goudreault (MR 408)

D.27 Improve ELPA detection; make abinit compatible with ELPA2019

By M. Torrent (MR 626)

D.28 Upgrade of base3 tutorial and basepar tutorial.

By X. Gonze (MR628)

D.29 New input variable prtprocar, see test v5[40].

By M. Verstraete (MR630)

D.30 The ABINIT input variable supercell_latt is documented. See also v8[94].

By X. Gonze

D.31 New implementation of the cRPA (to compute U and J in DFT+U or DFT+DMFT) that can consider any subset of orbitals in the system. Compilation was a bit long on certain compilers, so new keywords have been added to the build system: enable_crpa_optim and enable_crpa_no_optim

By R. Outerov and B. Amadon (MR622).

D.32 On-going work on refactoring the Coulomb interaction part of ABINIT.

New input variables fock_icutcoul, and gw_icutcoul, that should superceed icutcoul. New test added for the mini-Brillouin Zone integration, gw_icutcoul=14, 15, 16, see v9[21].

By B. Guster, M. Giantomassi, F. Bruneval and X. Gonze (MR 627, 633, 673, 679, 686).

D.33 New TB2J python script to compute the superexchange (J) and the Dzyaloshinskii-Moriya (DMI) interactions. The script can be found in http://gitlab.abinit.org/xuhe/TB2J with doc and tutorials. The script is interfaced with wannier90 and use the w90 output files. The J calculation works in production, the DMI is much more sensitive to disentanglement noise and have to be use with care. An article is under construction to describe the method and its implementation. The script can deliver input data file for the spin model of Multibinit.

By He Xu, M. Verstraete and E. Bousquet (MR 639).

D.34 Test linear electro-optical coefficient, tutorespfn optic[5].

By N. Pike (MR 575).

D.35 Optic: print spin-decomposition when applicable. Fixed reflectivity screwed results. Test more thoroughly optics (incl. interfacing using NetCDF), see new tests v9[05] to v9[12], also v9[47] and v9[48]

By X. Gonze (MR 654, 674).

D.36 Fixed DFPT+PAW+GGA+usexcnhat=1+q<> bug, and also add new related tests v9[41] and v9[42].

By M. Torrent and X. Gonze (MR 657).

D.37 Update PAW tutorial

By F. Jollet (MR 643, 645, 652, 653).

D.38 Nonlinear xc preliminary

Preliminary work for some changes in the exchange correlation terms in nonlinear (3^rd order DFPT). One term was implemented in both pead and dfptnl routines. Now it is merged in one routine (dfptnl_exc3). Due to a subtle reordering of nonlinear core correction terms, some pead refs are changed, but the final result (“First order change in electronic dielectric susceptibility tensor”) remain the same.

By L. Baguet (MR 650).

D.39 Anaddb output data prefix

The former writing of anaddb.nc (fixed path and fixed name) has been made more flexible, by introducing a prefix

By J. Bieder and He Xu (MR702)

D.40 New “macro” input variable expert_user

When non-zero expert_user automatically switch off all checks done by chkprim, chkdilatmx, chksymbreak and chksymtnons.

By X. Gonze (MR715)

D.41 Miscellaneous additional bug fixes, improvements of documentation including for the build system. G. Antonius, L. Baguet, JM Beuken, J. Bieder, E. Bousquet, F. Bruneval, T. Cavignac, M. Giantomassi, X. Gonze, F. Jollet, R. Outerovitch, N. Pike, Y Pouillon, M. Royo, M. Torrent, J. Van Bever, M. Verstraete, Xu He.

v9.0¶

Version 9.0, released on March 29, 2020. List of changes with respect to version 8.10.

Many thanks to the contributors to the ABINIT project between October 2018 and March 2020. These release notes are relative to modifications/improvements of ABINIT v9.0 with respect to v8.10 (merge requests up to, and including, MR636 are taken into account)

The list of contributors includes: B. Amadon, L. Baguet, J.-M. Beuken, J. Bieder, J. Bouchet, E. Bousquet, F. Bruneval, G. Brunin, Wei Chen, J.-B. Charraud, Ph. Ghosez, M. Giantomassi, O. Gingras, X. Gonze, F. Goudreault, B. Guster, G. Hautier, Xu He, N. Helbig, F. Jollet, H. Miranda, F. Naccarato, R. Outerov, G. Petretto, N. Pike, Y. Pouillon, F. Ricci, M. Royo, M. Schmitt, M. Stengel, M. Torrent, J. Van Bever, M. Verstraete, J. Zwanziger.

It is worth to read carefully all the modifications that are mentioned in the present file, and examine the links to help files or test cases. This might take some time …

Xavier

A. Important remarks and warnings.¶

A.1 At the occasion of the switch from ABINITv8 to ABINITv9, many improvements of the formats and content of files written by ABINIT have been made, so the backward compatibility of ABINITv9 may be broken. The present ABINITv9.0 is NOT to be considered a production version. It is a beta release, allowing developers to get feedback from the users. Many features will work correctly, of course. Still, beginners are advised to stick to ABINITv8.10.3 except if ABINITv8.10.3 is not appropriate (or not working) for them.

In particular:

The build system relies on new .ac9 files (see B.6), superceeding the v8 .ac files. A bash script (upgrade-build-config-file.sh) located in the top level directory of the package can be used to convert from the old .acformat to .ac9.
The build system of ABINITv9 does not build anymore the hard dependencies (Linalg, NetCDF4, HDF5, LibXC, …), as this was not sustainable (see B.6) and nowadays most users install prerequisite libraries themselves.
The main ABINIT output file now contains sections written in YAML (sometimes replacing text sections, sometimes adding information). This means that some user-developed parsing tools might not work anymore, and should be adapted to the new ABINITv9 output file (see B.9). Note that the YAML output is still under development and modifications may appear in the next versions. A python API to extract the results of the calculation will be provided when the implementation is finalized.
Several default values have been changed, see A.3.

A.2 A new account of the ABINIT effort has been published in Computer Phys. Comm. [Gonze2020] It provides description of several new features. A version of this paper that is not formatted for Computer Phys. Comm. is also available. The licence allows the authors to put it on the Web.

A second new account of the ABINIT effort has been published in J. Chem. Phys. [Romero2020]. The scope of this second paper is different from the first one. It is more a survey of ABINIT, focusing on its specific capabilities. Still, it contains also some description of some new features. A version of this paper that is not formatted for J. Chem. Phys. is also available. The licence allows the authors to put it on the Web.

Other specific publications are mentioned in the Suggested acknowledgment page.

A.3 The default values of the following ABINIT input variables have been changed: ixcrot, chneut, ntime, symsigma, prtkden.

A.4 The initialization of the wavefunctions when paral_kgb=1 and nspinor=2 has been changed, since the previous one could prevent the code to converge. By M Torrent (MR 562).

A.5 The input variable xangst has been disabled. Use xcart instead, and specify the unit, namely Angstrom.

A.6 The name of the t-DEP main executable has been changed from tdep" toatdep`, in line with [Romero2020].

B. Most noticeable achievements¶

B.1 Electron-phonon interaction (mobilities in the self-energy relaxation time approximation, temperature-dependent electronic band structures including the zero-point renormalization, etc.)

The new capabilities of ABINITv9 related to electron-phonon calculations are described fully in the Sec. 3.3.2 of [Gonze2020], as follows.

In abinit v9, it is possible to compute the EPH self-energy in the Kohn–Sham representation using the EPH matrix elements. The code employs optimized algorithms to compute either the full self-energy (needed for QP corrections and spectral functions) or just the imaginary part that is then used to evaluate mobilities within the self-energy relaxation time approximation (SERTA). The computation of the mobility is fully integrated inside abinit, and is an automatic output of the computation of the imaginary part of the self-energy, bypassing the need to post-process results. When computing the full self-energy, it is possible to reduce the number of empty states required for convergence by using the first-order wavefunctions obtained by solving the relevant Sternheimer equation.

In the case of lifetime computations, the code takes advantage of the tetrahedron method to filter contributing q-points, a double-grid integration technique to accelerate the convergence at marginal additional computational cost, and samples the relevant regions in the Brillouin zone contributing to transport properties thus leading to a significant reduction of the computational effort. Crystalline symmetries are used throughout the code in order to reduce the number of k- and q-points that must be explicitly included in the integrals. To achieve good parallel efficiently, the most CPU demanding parts are parallelized with MPI employing a distribution schemes over k/q-points, perturbations and bands (the band level is available only when computing the full self-energy).

Moreover, the interpolation of the DFPT potential, described in Sec. 3.3.1 of [Gonze2020] is fully operational, with many tests provided.

List of tests: v9[50], v9[61] and v8[44].

New input variables: dvdb_qcache_mb, eph_phrange, eph_tols_idelta, eph_ecutosc, eph_restart, eph_stern, eph_use_ftinterp, getdvdb, getdvdb_filepath, getkerange_filepath, irddvdb, prteliash, sigma_bsum_range, sigma_erange, sigma_ngkpt, sigma_nshiftk, sigma_shiftk, symv1scf.

Note that the new EPH processing unit of ABINIT optdriver=7 has a different implementation than the one implemented in anaddb. A new set of tutorials are in preparation and they will be made available in the forthcoming versions. For further details about the implementation, please consult this preprint.

By G. Brunin, H. Miranda, M. Giantomassi, G.-M. Rignanese, G. Hautier.

B.2 Flexoelectricity and dynamical quadrupoles

A new driver has been included in abinit that allows one to compute 4 spatial dispersion tensorial quantities: the clamped-ion flexoelectric tensor, the dynamical quadrupoles, the first moment of IFC matrix and the first moment of the piezoelectric force response tensor. Precalculation of ground state, first and second (d2_dkdk) order response functions is required. After execution, the driver creates a 3^rd order energy derivative database file that is used by anaddb to compute the mixed and lattice-mediated flexoelectric tensors or to include the dipole-quadrupole and quadrupole-quadrupole electrostatic interactions in the calculation of the dynamical matrix.

See the complementary description in the Sec. V. D of [Romero2020], with underlying theory and test calculations presented in [Royo2019].

At the practical level, see [Romero2020]:

In this way, both perturbations are generalized to finite q, as is already the case for atomic displacements. This enables us to carry out an analytical third order derivative of the energy with respect to two of the standard perturbations, and to the momentum q, which directly provides the sought-after spatial dispersion tensors. Remarkably, by virtue of the 2n+1 theorem, the third-order energies are computed in one shot using precalculated first-order response functions to the standard perturbations, without the necessity of self-consistently computing any response function to a perturbation gradient. After execution, the long-wave DFPT routines generate a derivative database that is subsequently used by post-processing tools implemented in ANADDB to compute and print the different contributions to the FxE tensor.

The dynamical quadrupoles are the spatial dispersion counterparts of the Born effective charges, and can be used in lattice dynamics calculations to improve the prevalent dipole-dipole treatment of the long-range interactions. The ANADDB routines that carry out the process of interpolating the dynamical matrix following Ref. 34 have been adapted to incorporate the dipole-quadrupole and quadrupole-quadrupole electrostatic interactions derived in Ref. 102. This new functionality results in a faster convergence of the phonon bands calculation with respect to the density of q points and, in some materials, represents the only route to obtain the correct sound velocities.

Currently, the implementation is restricted to the use of norm-conserving pseudopotentials without non-linear core corrections, and the LDA functional.

A tutorial is in preparation, with tests lw[1] to lw[7], as well as a specific topic.

New input variables have been defined: lw_flexo, lw_qdrpl, prepalw, flexoflag@anaddb, dipquad@anaddb, quadquad@anaddb.

This capability is still under development and not completely stable. Interested users are strongly recommended to contact Miquel Royo (mroyo@icmab.es) or Massimiliano Stengel (mstengel@icmab.es) before start using it.

By M. Royo, M. Stengel, M. Giantomassi.

B.3 DFT+DMFT

The new capabilities of ABINITv9 related to DFT+DMFT calculations are described fully in the Sec. 3.7 of [Gonze2020], as follows.

The DFT+DMFT parallelism was improved for large systems. In particular, it is now possible to parallelize the calculation on both k-points and bands/g-vectors by using the input variable paral_kgb = 1 and related input variables.

Two new approaches to CT-QMC have been added to solve the AIM. In the first one, the density–density CT-QMC code available in abinit [Gonze2016], [Bieder2014] was generalized in order to take into account off-diagonal elements of the hybridization function. This implementation is activated with the input variable dmft_solv = 8. Spin–orbit coupling calculations are possible, but using a real valued imaginary time hybridization function. This solver was used in Refs. [Amadon2015], [Amadon2016].

In the second approach, we use the Toolbox for Research on Interacting Quantum System (TRIQS)library [Parcollet2015], which is an open-source project that provides a framework for many-body quantum physics and more specifically for strongly-correlated electronic systems. TRIQS provides an open source implementation of the continuous-time hybridization expansion quantum impurity solver (CT-HYB) [Seth2016], considered a state-of-the art solver for multi-orbital AIM. An interface between abinit and the impurity solver TRIQS/CT-HYB is now available and will make use of the independent progress made by the TRIQS library. <…>

Also, the DMFT k-resolved spectral function is available (MR 529, 490).

List of tests: paral[84], paral[86], paral[99], v8[01]. New input variables: dmft_charge_prec and dmft_kspectral_func (test to be provided for the latter). Also dmft_occnd_imag, but only for keeping backward compatibility for tests.

By T. Cavignac, B. Amadon and O. Gingras.

B.4 Spin model within Multibinit

The new capabilities of Multibinit within ABINITv9 are described fully in the Sec. 4.1 of [Gonze2020]. See also Sec. D.1. In particular, a spin model, described specifically in Sec. 4.1.2 of [Gonze2020], is available, as follows.

Multibinit implements the most commonly used model for spin systems, via the Heisenberg Hamiltonian including magnetic exchange and Dzyaloshinskii Moriya interactions. Single ion anisotropy and dipole–dipole interactions are also included, and all terms bear a very strong similarity to the quadratic part of the lattice model Hamiltonian. A number of open source spin dynamics codes already exist, such as UPPASD, VAMPIR, OOMF; the distinguishing features of multibinit are the integration with abinit, to fit parameters, and the simultaneous dynamics with other degrees of freedom (in particular using the inter-atomic force constants).

A tutorial for the multibinit spin model has been written, spin_model tutorial.

Many new input variables are present. Not all these new input variables are present in automatic tests, though, in this beta-release. These “non-tested” input variables are indicated below with ‘NT’. This will be completed for the production version v9.2 . List of tests in addition to those of the tutorial: v8[16], v8[23], v9[81], and v9[82].

New input variables tested and documented: spin_calc_thermo_obs, spin_damping, spin_init_orientation, spin_init_qpoint, spin_init_rotate_axis, spin_init_state, spin_ntime_pre, spin_projection_qpoint, spin_sia_add, spin_sia_k1amp, spin_sia_k1dir, spin_temperature_start, spin_temperature_end, spin_temperature_nstep, spin_var_temperature, spin_write_traj. Additionnal new input variables: slc_coupling (NT), spin_calc_correlation_obs (NT and not documented), spin_calc_traj_obs (NT and not documented), spin_projection_qpoint (NT).

By Xu He, N. Helbig, J. Bieder, E. Bousquet, Ph. Ghosez, M. Verstraete

B.5 Constrained DFT

Constrained Density-Functional Theory (see topic_ConstrainedDFT) is available, with a new algorithm allowing to impose the constraints to arbitrary precision, whether it relates to the charge, magnetization, magnetization direction, or magnetisation size, or a combination thereof for different atoms. The constraints are smeared spherical integrals with ajustable sphere radius, centered on atoms. The algorithms has been demonstrated for norm-conserving pseudopotentials as well as PAW. Forces and derivatives with respect to the constraints are available (i.e. magnetic torque for the non-collinear spin case). Stresses are still to be coded, will be available in ABINITv9.2.

New tests: v8#24-29, v8#95-97 and v9#1-3. New input variables: chrgat, constraint_kind, ratsm.

By X. Gonze.

B.6 Large modifications of the build system

The build system relies on new .ac9 files, superceeding the v8 .ac files. Fully documented example files can be found in doc/build/config-examples. A bash script (upgrade-build-config-file.sh) located in the top level directory of the package can be used to convert from the old .acformat to .ac9.

The build system of ABINITv9 does not build anymore the (hard and soft) dependencies (Linalg, NetCDF4, HDF, LibXC, Wannier90, …), as this was not sustainable. Three libraries are now mandatory: linalg, NetCDF4/HDF5 and LibXC. Failing to link to them will prevent building ABINIT. The other libraries are optional, there will only be a warning if they are not available. If the user does not provide the path to these libraries, the build system will try to find them in the “usual” directories, and inform the user that it has done so. The build system also can make suggestions to the user, to complete its *.ac9 file.

By Y. Pouillon and JM Beuken

B.7 New command line interface

There is a new (recommended) command line interface to run ABINIT, without the “files” file. The new syntax is:

abinit run.abi

or

abinit run.abi > run.log 2> run.err &

where run.abi is the Abinit input file that now provides all the information related to pseudos and the prefixes that were previously passed via the “files” file. For comparison, the old syntax is

abinit < run.files > run.log 2> run.err &      ! This is the old syntax

A file extension for the input file is highly recommended (in this example we use .abi) as by default the parser will use the string before the file extension as root to build the prefixes for the input/output/temporary files.

The user can specify the name of the main output file thanks to the output_file input variable, the list of pseudopotentials thanks to the pseudos input variable and the directory where all pseudos are located with pp_dirpath. The prefix for other input, output or temporary files can be specified with indata_prefix, outdata_prefix and tmpdata_prefix, respectively. A default set of prefixes computed from the basename of the input file is used if these variables are not specified in the input.

For some examples, see tests v8[90], v7[45], and v5[54]. See also topic_Control.

A similar command line interface can also be used for the anaddb code. In this case, the relevant variables are: output_file@anaddb, ddb_filepath@anaddb, ddk_filepath@anaddb, gkk_filepath@anaddb, eph_prefix@anaddb.

The new syntax is:

anaddb run.in > run.log 2> run.err &

See tests v8[52] for a standard analysis of the DDB file and v7[94] for the (old implementation) of electron-phonon calculations in anaddb.

Important

The old “files file” interface is still operational although deprecated and will be REMOVED in Abinit v10.

By M. Giantomassi (MR 586).

B.8 Reading strings from the input file

A new mechanism to read strings enclosed between double quotation marks from the input file has been activated. So, many new input keywords are reading strings as data, and, often, can be used alternatively to similar input keywords that were expecting numerical values such as the get* and ird* variables. The goal is to encourange a new approach for performing ABINIT calculations in which multiple datasets and get* variables are replaced by indipendent input files that are connected together via file paths.

List of new input variables that rely on this feature:

getddb_filepath, an alternative to getddb or irdddb, see test v9[60]
getden_filepath, an alternative to getden or irdden, see test v8[36] and v8[41]
getscr_filepath, an alternative to getscr or irdscr, see test v67mbpt[51]
getwfkfine_filepath, an alternative to getwfkfine or irdwfkfine, see tests v9[55], v9[56]
getwfk_filepath, an alternative to getwfk or irdwfk, see test v9[60]
getwfq_filepath, an alternative to getwfq or irdwfq, see test v7[98]
getkerange_filepath, see test v9[60]
getpot_filepath, see test v8[44]
pseudos, indata_prefix, outdata_prefix, tmpdata_prefix, output_file, see test v9[04]
pp_dirpath: cannot be tested EXPLICITLY because system dependent but used by runtests.py when generating the input file.
output_file@anaddb, ddb_filepath@anaddb, gkk_filepath@anaddb, eph_prefix@anaddb. See tests v8[52] for a standard analysis of the DDB file and v7[94] for the (old implementation) of electron-phonon calculations in anaddb.

By M. Giantomassi

B.9 YAML sections in the output file

YAML sections are now generated in the output file, sometimes replacing text sections, sometime providing new information. At present there is a YAML section for the components of the total energy, the GS results including forces and stresses as well as a YAML section for GW calculations, and some YAML sections giving information about the iteration status.

At the occasion of the development of this capability, and its adaptation to the test farm, the perl script fldiff.pl has been replaced by a Python version. See related information in Sec. 5.5 of [Gonze2020].

By T. Cavignac, M. Giantomassi, GM Rignanese, X Gonze.

B.10 New approach to define crystalline structures in the Abinit input

The new variable structure can be used to initialize the lattice vectors and the atomic positions from an external file. Variables such as natom, ntypat, typat and znucl are automatically initialized and need not to be specified in the ABINIT input. At present, the code can read netcdf files produced by ABINIT (GSR.nc, WFK.nc, DEN.nc, HIST.nc) and POSCAR files in VASP-5 format. See the documentation for the syntax and limitations.

By M. Giantomassi.

B.11 New capabilities of abipy and abiflows

The abipy and abiflows projects have been significantly extended. See Sec. 6 of [Gonze2020], as well as the gallery of plotting scripts and the gallery of abipy workflows .

By M. Giantomassi, G. Petretto, F. Naccarato.

C. Changes for the developers (including information about compilers)¶

C.1 A python script to help ABINIT developers and development.

The new python script abisrc.py located in the top directory of the ABINIT package has been developed. It has superceded abilint.py in the makemake procedure.

Try

./abisrc.py --help

then follow the suggestions, to get info about files, directories, interfaces, to visualize dependencies of the ABINIT subroutines, etc.

Note that there are dependencies of abisrc.py, to be installed prior being able to use some of its capabilities. Use:

pip install -r requirements.txt --user

to install the dependencies in user mode.

By M. Giantomassi.

C.2 New characteristics of input variables

In the description of input variables (e.g. files abimkdocs/variables_abinit.py), a new field added_in_version has been introduced, for example,

 added_in_version="9.0.0"

or, for variables introduced prior to v9,

 added_in_version="before_v9"

C.2 Test farm: new and obsolete bots

Bots introduced in the test farm: atlas_gnu_9.1_openmpi, buda2_gnu_8.2_cuda, cronos2_gnu_7.4_paral.
Bots upgraded: abiref_gnu_5.3_* to abiref_gnu_9.2_* ; abiref_nag_6.2_openmpi to abiref_nag_7.0_openmpi; bob_gnu_5.3_openmp to bob_gnu_7.5_openmp; buda2_gnu_8.1_mpich3 to buda2_gnu_8.2_mpich3; graphene_gnu_6.4_macports to graphene_gnu_9.2_macports; max2_gnu_5.3_* to max2_gnu_6.5_*; ubu_gnu_5.3_openmpi to ubu_gnu_9.2_openmpi.
Bots removed: atlas_gnu_7.2_fb.ac (no nightly test of the fallbacks anymore), cronos_gnu_5.3_paral(remplaced by cronos2), inca_gnu_6.3_py3k (inca too old), tikal_gnu_* (tikal too old). Working also on a cronos-cronos2 cluster.

By JM Beuken

C.3 Supported compilers

gfort (GNU) compiler: v9 newly supported, v4 obsolete
ifort (INTEL) compiler: v19 newly supported.
NAG 7.0 instead of 6.2

By JM Beuken

C.4 Unitary ttransposer#1 . Test of the transposer for linear algebra to KGB parallelisation.

By J. Bieder.

C.5 Linkchecker has been reenabled, only for internal link checking.

By JM Beuken (MR 513).

C.6 Enable the generation of a HTML side-by-side diff on the test farm when fldiff fails with a line count error and it was not caused by a crash of Abinit. The diff algorithms uses a specialized heuristic to improve line synchronization and prevent weird matching.

By Th. Cavignac (MR 526)

C.7 Split of the source tree (ongoing).

In order to improve modularity, the source tree must be split in two parts, one for low-level routines, largely independent of ABINIT, and one for more specific routines to ABINIT. The low-level routines should become a separate library, with its own build system and make. At present the low-level library have been moved out of src, inside the shared/common/src directory. See related information in Sec. 5.4 of [Gonze2020].

C.8 New FFT specifications for the build system See https://gitlab.abinit.org/pouillon/abinit/-/issues/33 .

By Y. Pouillon (MR 619)

D. Other changes (or on-going developments, not yet finalized)¶

D.1 Miscellaneous improvements of Multibinit (lattice part)

Miscellaneous improvements have been made to the lattice part of Multibinit. See the new input variables below, also see the Sec. 4.1.1 of [Gonze2020].

New tests: v8[38], v8[94], v8[98], v8[99], v9[83], v9[84], and v9[85]. New input variables are listed below. Not all these new input variables are present in automatic tests, though, in this beta-release. These “non-tested” input variables are indicated below with ‘NT’. Also, not all of these are documented, or only partly documented (e.g. variable type, acronym, default, but no more). This will be completed for the production version v9.2 .

analyze_anh_pot
fit_SPC_maxS@multibinit (NT)
fit_iatom@multibinit paral[81], paral[82], v8[13], v8[14], but not documented.
latt_compressibility@multibinit, NOT TESTED, NOT DOCUMENTED
latt_friction
latt_mask@multibinit, NOT TESTED, NOT DOCUMENTED
latt_taup Despite the link, NOT TESTED, only partly DOCUMENTED
latt_taut Despite the link only partly DOCUMENTED
opt_coeff
opt_effpot
opt_ncoeff
prt_names, NOT TESTED, NOT DOCUMENTED
test_effpot v8#98

By M. Schmitt, Xu He, F. Ricci, M. Verstraete, Ph. Ghosez

D.2 Miscellaneous improvements in the Chern number and orbital magnetization calculations, including parallelization over k points of the Chern number calculation.

By J. Zwanziger (MR 469, 500, 545, 588)

D.3 Calculation of Debye-Waller tensor. New test v8[58].

By M. Giantomassi

D.4 Test linear electro-optical coefficient, tutorespfn optic[5].

By N. Pike (MR 575).

D.5 NCPP Wavefunction mixing with Variational Energy and minor improvements to prepare PAW+Hybrid variational energy. New test v7[73], simple system for testing Hartree-Fock and the SCF algorithms.

By X. Gonze (MR 434, 444, 445).

D.6 New weight distribution of the Fourier components of the force constants. Test tolerance in the new integration weights, tests v8[52], v8[53], v8[54].

By H. Miranda and M. Giantomassi

D.7 Test calculation of velocity matrix elements (DDK) with optdriver 8 and wfk_task “wfk_ddk”, see v8[59].

By M. Giantomassi

D.8 Upgraded paral_bandpw tutorial, new version of auto paral (with threads)

By M. Torrent (MR502).

D.9 Test wannier90 interface with nsppol=2 and nspden=2, wannier90[04].

By Xu He

D.10 Mixed precision for FFT transforms. New input variable mixprec see v8[44], v9[57], v9[60], and v9[61].

From M. Giantomassi (MR491).

D.11 Multibinit has been interfaced with scale-up, https://www.secondprinciples.unican.es

By Marcus Schmitt, Jordan Bieder, Matthieu Verstraete and Philippe Ghosez

D.12 The following units are now also allowed in input files:

S Sec Second for the ABINIT input file;
nm (for nanometer) for the ABINIT and ANADDB input files.

D.13 aTDEP utility: added aTDEP guide, aTDEP topic, and corresponding input variable documentation. References: aTDEP paper. Also, see Sec. 4.2 of [Gonze2020].

By F. Bottin, J. Bouchet, J. Bieder (MR491,422).

D.14 Improvements of NLO calculations. Optimize memory allocation. In particular for usepead = 1. Write Raman susceptibilities to netcdf in anaddb.

By G. Petretto (MR 599).

D.15 MetaGGA + PAW in progress. All internal and unit tests are OK. The implementation seems close to the end. Still need to perform physical tests.

By M. Torrent and J.-B. Charraud (MR 587, 558, 625).

D.16 Implementation of an alternate version of MPI reductions valid when the number of data exceeds 2^32.

By M. Torrent (MR 587)

D.17 Build system and src support for llvm (clang/flang) and ARM (armflang/armclang/armpl).

By M Torrent (MR 571)

D.18 Improvement of core WF reading (optics + electron-positron)

By M Torrent (MR 557)

D.19 Fix bootstrap kernel convergence issues

By Wei Chen (MR 546)

D.20 Upgrade versions of libraries used with ABINIT. Upgrade atompaw to 4.1.0.6. Upgrade Libxc to 4+.

By M. Torrent, JM Beuken (MR 532, 470, 465, 441)

D.21 Write yaml file for fatbands (phonopy format) with aTDEP

By J. Bieder (MR510)

D.22 Write polarization vectors in GSR.nc.

By H. Miranda (MR 462).

D.23 Updated user interface of Raman_spec.py . Added one section to tutorial NLO to describe use of Raman_spec.py.

By N. Pike (MR 460, MR 581)

D.24 XML format for core wavefunctions

By F. Jollet (MR 423)

D.25 Wavefunction prediction for molecular dynamics.

By F. Jollet (MR 412)

D.26 Added a preview for the toptic_4.abi file in the optic tutorial.

By F. Goudreault (MR 408)

D.27 Improve ELPA detection; make abinit compatible with ELPA2019

By M. Torrent (MR 626)

D.28 Upgrade of base3 tutorial and basepar tutorial.

By X. Gonze (MR628)

D.29 New input variable prtprocar, see test v5[40].

By M. Verstraete (MR630)

D.30 The ABINIT input variable supercell_latt is documented. See also v8[94].

By X. Gonze

D.31 New implementation of the cRPA (to compute U and J in DFT+U or DFT+DMFT) that can consider any subset of orbitals in the system. Compilation was a bit long on certain compilers, so new keywords have been added to the build system: enable_crpa_optim and enable_crpa_no_optim

By R. Outerov and B. Amadon (MR622).

D.32 Work on refactoring the Coulomb interaction part of ABINIT.

By B. Guster, M. Giantomassi and X. Gonze (MR 627&633).

D.33 New TB2J python script to compute the superexchange (J) and the Dzyaloshinskii-Moriya (DMI) interactions. The script can be found in http://gitlab.abinit.org/xuhe/TB2J with doc and tutorials. The script is interfaced with wannier90 and use the w90 output files. The J calculation works in production, the DMI is much more sensitive to disentanglement noise and have to be use with care. An article is under construction to describe the method and its implementation. The script can deliver input data file for the spin model of Multibinit.

By He Xu, M. Verstraete and E. Bousquet

D.34 Miscellaneous additional bug fixes and improvements of documentation. L. Baguet, JM Beuken, J. Bieder, E. Bousquet, F. Bruneval, T. Cavignac, M. Giantomassi, X. Gonze, F. Jollet, N. Pike, Y Pouillon, M. Torrent, J. Van Bever, M. Verstraete, Xu He.

v8.10¶

Version 8.10, released on October 15, 2018. List of changes with respect to version 8.8.

Many thanks to the contributors to the ABINIT project between April 2018 and October 2018. These release notes are relative to modifications/improvements of ABINIT v8.10 with respect to v8.8. The merge request #408 is the first MR not reported in these release notes. Then, #410-411, #413-416 have also been included.

The list of contributors includes: B. Amadon, G. Antonius, L. Baguet, J.-M. Beuken, J. Bieder, E. Bousquet, F. Bruneval, Wei Chen, M. Cote, J. Denier, G. Geneste, Ph. Ghosez, M. Giantomassi, O. Gingras, X. Gonze, F. Goudreault, B. Guster, Xu He, Y. Jia, F. Jollet, A. Lherbier, A. Martin, H. Miranda, F. Naccarato, G. Petretto, N. Pike, S. Ponce, Y. Pouillon, S. Prokhorenko, F. Ricci, M. Torrent, M. van Setten, B. Van Troeye, M. Verstraete, J. Zwanziger.

It is worth to read carefully all the modifications that are mentioned in the present file, and examine the links to help files or test cases. This might take some time …

Xavier

A. Warnings and important remarks¶

A.1 The correct definition of the temperature has been implemented in the isokinetic algorithm ionmov=12.

A.2 The multibinit input variable “supercell” has been renamed “supercell_latt”.

A.3 Tests v8#100-107 have been moved to v8#81-88. Tests paral#100-103 have been moved to paral#80-83

B. Most noticeable achievements¶

B.1 The computation of the Raman intensity in DFPT with PAW is now possible (it was only available with norm conserving psps previously). This is based on the second-order Sternheimer equation for the derivative with respect to an electric field. See tests v8[81] to v8[89]. By L. Baguet and M. Torrent.

B.2 There has been a large effort to clean the documentation, that was needed and made possible thanks to the recent move to the mkdocs system. In particular, many new hyperlinks have been created in dozens of markdown files, references have been centralized in the bibliography. By B. Amadon, G. Antonius, L. Baguet, J. Bieder, E. Bousquet, F. Bruneval, Wei Chen, M. Cote, G. Geneste, M. Giantomassi, X. Gonze, Xu He, F. Jollet, A. Lherbier, H. Miranda, F. Naccarato, G. Petretto, N. Pike, S. Ponce, Y. Pouillon, M. Torrent, M. van Setten, B. Van Troeye, M. Verstraete, J. Zwanziger.

B.3 The multibinit application (for second-principles calculations) has considerably progressed. Documentation has thus been set up: “topics” have been written, as well as a tutorial, in addition to the already existing input variable documentation and test cases. See topic_LatticeModel, topic_BoundingProcess, topic_FitProcess and topic_DynamicsMultibinit, that are hub to the relevant tutorial, input variables and test cases (e.g. lattice_model lesson,v8[15], v8[16]…). By A. Martin, in collaboration with Fabio Ricci and Ph. Ghosez

B.4 Several new options are available for the ionmov input variable governing ionic dynamic or geometry optimization:

ionmov=15 for the FIRE algorithm, v8[17];
for ionmov=12, isokinetic ensemble, the fixing of atomic positions is now allowed, v8[21] and v8[22].

Also, the documentation for the hybrid Monte Carlo algorithm has been improved, see v8[34], and the new input variables hmcsst and hmctt.

By Xu He, S. Prokhorenko and X. Gonze.

B.5 The linear combination of images is now allowed, with the new value for input variable imgmov=6, and mixing factor given by mixesimgf. In this case, the total energy and forces are also assembled as a linear combination, and the geometry is optimized using algorithms selected with the usual ionmov input variable. See test v8[20]. The wavefunctions from the previous itimimage value (see ntimimage input variables) can be stored, using the new input variable imgwfstor. This allows saving CPU time at the expense of memory, in all the image based algorithms. By X. Gonze, testing by Y. Jia.

B.6 Tutorial nuc tutorial has now a section for the computation of the isomer shift (Mossbauer spectroscopy) based on Fermi contact interaction. By J. Zwanziger.

B.7 The Frohlich model is now implemented in the electron-phonon part of ABINIT, optdriver=7. The Frohlich average of effective masses is computed with the DFPT computation of effective masses, see v8[56]. Also, the zero-point renormalization of the band extrema is computed using a general formula valid for isotropic and anisotropic solids, as well as for non-degenerate or degenerate extrema, see eph_frohlichm and v8[57]. By X. Gonze.

C. Changes for the developers (also compilers)¶

C.1 All F90 ABINIT sources are now inside modules. Makemake now aborts if F90 procedures outside modules. By M. Giantomassi, with some help by J. Zwanziger, M. Torrent and B. Amadon.

C.2 Prepared the removal of the bindings subsystem. By Y. Pouillon and M. Torrent.

C.2 New Howto for developers (variables, mkparents, robodoc, test_suite). By M. Giantomassi.

C.3 New Howto for the test suite. By M. Giantomassi.

D. Other changes (or on-going developments, not yet finalized)¶

D.1 New input variable prtkbff. This input variable activates the output of the Kleynman-Bylander form factors in the netcdf WFK file produced at the end of the ground-state calculation. The form factors are needed to compute the matrix elements of the commutator [Vnl, r] of the non-local part of the (NC) pseudopotentials. This WFK file can therefore be used to perform optical and/or many-body calculations with external codes such as DP/EXC and Yambo. The option is ignored if PAW. By H. Miranda and M. Giantomassi.

D.2. A new routine to calculate and write the DDK matrix elements in a EVK.nc files for norm-conserving pseudo-potentials (and nspinor == 1), from a WFK file using the commutator [H, r] used to calculate the matrix elements for chi. These files are in the same format and contain the same information as the EVK.nc files (former DDK.nc) produced by the DFPT part of the code. This routine is parallelized over k-points and bands and requires a much smaller memory footprint than the DFPT code. By H. Miranda and M. Giantomassi.

D.3 New input variable slk_rankpp. This variable controls how the number of processes to be used in Scalapack diagonalization algorithm: np_slk will be calculated according to this value. By J. Bieder.

D.4 New input variables prtefmas, irdefmas and getefmas, to deal with the effective masses, e.g. to allow feeding computed effective masses from one dataset to another one. By X. Gonze.

D.5 The spin dynamics has been implemented in multibinit. By He Xu.

D.6 New value for input variable usepawu=4. The FLL double counting is used. However, and in comparison to usepaw=1, the calculation is done without polarization in the exchange correlation functional. By B. Amadon.

D.7 New extensive testing of the DFPT+PAW+GGA, see v8[51]. However, pawxcdev=0 is still needed. By M. Torrent.

D.8 New input variable prtfull1wf to print the full perturbed wavefunctions. By G. Antonius.

D.9 Allows one to suppress completion of the 2DTE for vanishing elements using symmetries only, negative values of rfmeth. By X. Gonze.

D.10 Interface with TRIQS_2.0. By O. Gingras.

D.11 Allows to define different occupations for different spins, when occopt=0 and nsppol=2. By X. Gonze.

D.12 NEB is now tested (at last), v6[29]. By X. Gonze.

D.13 Many abitutorials have been improved. By X. Gonze.

D.14 Work on orbital magnetization. By J. Zwanziger.

D.15 Elastic and piezoelectric tensors in netcdf format. By M. Giantomassi

D.16 Allow boxcutmin for DFPT case, after check by H. Miranda.

D.17 ABIWAN.nc files with Wannier output that can be analyzed wit AbiPy. See the AbiwanFile notebook for futher information. By M. Giantomassi.

D.18 The topic_Macroave has been created. By X. Gonze.

D.19. Include CTQMC Code using non diagonal hybridization function. See paral[83]. By B. Amadon, J. Denier and J. Bieder.

D.20. Finalize scalapack/elpa integration in lobpcg. Use ONLY if no openmp. Warning: scalapack is not thread-safe in general. By J. Bieder.

D.21 Automatic test v8[37] for TDep application. By J. Bieder.

D.22 Python scripts to calculate physical properties that can be derived from the elastic tensor (which is the result of an anaddb calculation). By N. Pike.

D.23 Miscellaneous additional bug fixes and improvements of documentation. By B. Amadon, G. Antonius, L Baguet, J.-M. Beuken, J. Bieder, F. Goudreault, F. Jollet, H. Miranda, F. Nacaratto, N. Pike, M. Torrent, M. Verstraete.

v8.8¶

Version 8.8, released on April 28, 2018. List of changes with respect to version 8.6.

Many thanks to the contributors to the ABINIT project between November 2017 and April 2018. These release notes are relative to modifications/improvements of ABINIT v8.8 with respect to v8.6. The merge request #285 is the first MR not reported in these release notes.

The list of contributors includes: B. Amadon, G. Antonius, L. Baguet, J.-M. Beuken, J. Bieder, F. Bottin, Y. Bouchet, E. Bousquet, W. Chen, C. Espejo, Ph. Ghosez, M. Giantomassi, X. Gonze, F. Jollet, A. Martin, H. Miranda, G. Petretto, N. Pike, Y. Pouillon, S. Prokhorenko, F. Ricci, G.-M. Rignanese, M. Torrent, M. Verstraete, J. Zwanziger

It is worth to read carefully all the modifications that are mentioned in the present file, and examine the links to help files or test cases … This might take some time …

Xavier

A. Warnings and important remarks¶

A.1 Due to the availability of new input variables, some obsolete input variables have been suppressed:

input variable cgtyphf, see now fockoptmix;
input variable gwfockmix, see now hyb_mixing and hyb_mixing_sr.

A.2 The algorithm used for computing the weights for the phonon band structure interpolation in ANADDB in the case brav@anaddb = 1 has changed. See B.5.

A.3 Tests v7#71-72 have been moved to v7#76-77.

A.4 Replaced tribes by relevance in doc/topics.

A.5 Replace EELF file extension by ELF. See prtelf. By Guido Petretto.

A.6 Definition of a maximal value for dilatmx, at 1.15, than can be bypassed by setting chkdilatmx=0. This is to prevent users slowing down ABINIT too much inadvertantly.

B. Most noticeable achievements¶

B.1 The whole ABINIT documentation has been placed under the control of mkdocs, and most files have been translated to markdown (well, there are still a few remaining files not placed in this system, but these are quite few). The capabilities developed for v8.6 (Topics -B.1-, central bibliography -B.2-, restructured and searcheable list of input variables -B.3-, frequency statistics -B.4-) have been maintained and consolidated. More documentation is available for developers than before. The whole system is better integrated and easier to maintain. The appearance is also new. The work on documentation is nearly complete, still not all bibliographical references of the doc have been entered in this central bibliographic database. Entry point: see the new header of any ABINIT documentation file e.g. the new user’s guide. By M. Giantomassi, with some help from X. Gonze.

B.2 The DFPT has been extended to non-collinear systems (nspden = 4), with or without spin-orbit coupling, for the ddk, electric field and atomic displacement perturbations, as well as for the Zeeman magnetic field perturbation (see B.3). See tests from v8[66] to v8[80]. For experts, see the new input variable ixcrot. By F. Ricci, S. Prokhorenko, M. Verstraete, M. Torrent and E. Bousquet.

B.3 DFPT can now treat the magnetic field perturbation (Zeeman interaction - magnetic field couple to the spin). See the input variable rfmagn, as well as tests v8[66] to v8[70]. The new input variable tim1rev has been introduced, to allow treating perturbations with non-zero q wavevectors. By S. Prokhorenko and E. Bousquet.

B.4 The python library AbiPy, for launching ABINIT (also in high-throughput mode) and the jupyter notebook based tutorials, nicknamed abitutorials are now sufficiently mature to be advertised. They have been used at the ICTP electron-phonon doctoral school in March 2018. Feedback on AbiPy and abitutorials is welcome. By M. Giantomassi.

B.5 A new algorithm (Wigner-Seitz cell based) for computing the weights for the phonon band structure interpolation in ANADDB has been implemented. It has replaced the old algorithm in case brav@anaddb = 1. The old algorithm is still available for back-compatibility purposes, now corresponding to brav@anaddb = -1, see v7[93], although there is no real reason for using it. The new algorithm is very general, respect better the symmetries, and should even supercede the use of other values of brav@anaddb. By G. Petretto following discussions with GM Rignanese and XGonze, and tests by Henrique Pereira Miranda.

B.6 The Chern number can be computed, in the norm-conserving case as well as in the PAW case. See the theory in [Ceresoli2006]. Associated input variable: orbmag. Nuclear magnetic dipole moment code has been improved for efficiency. In particular, this improvement is due to converted nucdipmom_k to complex type and explicit BLAS call. Tutorial nuc is nightly tested. By J. Zwanziger (nuc testing by X. Gonze).

C. Changes for the developers (also compilers)¶

C.1 Add support for NAG 6.2, use netcdf4 and hdf5 with NAG 6.2. New builders: abiref_nag_6.2_openmpi and atlas_gnu_7.2_fb

C.2 New version of ELPA module (2011 -> 2017 compatible) By M. Torrent

C.3 Replaced http://www.abinit.org by https://www.abinit.org everywhere in the doc. By JM Beuken

C.4 Added fake -n/–no-split option to makemake. This little change will let the test farm keep on working with all branches while the source tree is split into common + core. By Y. Pouillon

C.5 Upgrade Abinit to PSML API 1.1. By Y. Pouillon

C.6 Allowed for free-form link flags statements in configure options By Y. Pouillon

C.7 Intel 18.0 is now officially supported (one bot using it is present in the test farm). The situation is not perfect though, as in mrgscr and mrgdv, the ADVANCE=’NO’ specification for the write instruction does not work, but it works in simple programs. Thus all tests (and their chain of tests) that rely on mrgscr and mrgdv have been disabled for this compiler. Namely, v3#87-91, vv67mbpt#37-39, v7#86-88, v8#41-44, v8#63 Also, the reading of WFK files using MPIIO is not correct, for tests mpiio#26 and 62.

D. Other changes (or on-going developments, not yet finalized)¶

D.1 Implementation of the LDA-½ methodology (see the announcement B.10 of v8.6): v8[32] has been provided. By F. Jollet.

D.2 Numerous progresses have been made related to the hybrid functionals (although hybrid functionals are not yet in production).

Stresses are now computed correctly.
The downsampling of the Brillouin Zone to build the Fock operator has been implemented and tested. See the input variable fockdownsampling as well as tests libxc[72] and paral[09].
The B3LYP functional has been implemented, ixc = -402.
The new input variables hyb_mixing, hyb_mixing_sr, hyb_range_dft, and hyb_range_fock give sufficient flexibility in the PBE0 and HSE family of functionals.
GW calculations can now start on top of hybrid functional calculations.
At variance, there is also now more flexibility to run hybrid calculations using the GW infrastructure (gwcalctyp = 5, 15, 25) by the definition of the ixc_sigma input variable.
There has been also important work concerning the self-consistency, although this work is not finalized yet (one reason why hybrid functionals are not yet in production). The self-consistency at fixed ACE operator can take advantage of an auxiliary XC functional to decrease the number of inner iterations, see fockoptmix, auxc_ixc and auxc_scal, while for the outer loop, in which the ACE operator is upgraded, the wavefunction mixing has been implemented (fockoptmix and wfmix).

See the new tests v7#67-72 libxc#44, 45, 72, 73, 74, and also the updated tests v4#86, 87, v67mbpt#09, v7#65, libxc#41, 42, 43, paral#09. By X. Gonze and F. Jollet, with help by M. Torrent.

D.3 The tutorial on temperature-dependence of the electronic structure has been upgraded, and carefully tested. See all tests in tutorespfn/teph_tdep_legacy*. By X. Gonze and M. Giantomassi

D.4 Output of interpolated density in the MPI-IO case is now tested, mpiio[26] and mpiio[27].

D.5 Ongoing work on the multibinit project. New input variables fit_nfixcoeff, fit_fixcoeff, fix_generateTerm, see v8[13] and v8[14]. New input variable dipdip_prt, see v8[06], as well as tests paral[96] to paral[102]. New generator for the polynomial coefficients, debug strain for the fit process, add tolerance in the fit process, add the plot of the comparison between model and DFT. By A. Martin, M. Verstraete and Ph. Ghosez.

D.6 Adjustment of tutorial tutoparal ucrpa, see test tutoparal#tucrpa_4. By B. Amadon

D.7 The ddk file is now available in netCDF format (lightweight version without first-order wavefunctions), and test with the optic post-processor has been set up. See the new v7[49]. By M. Giantomassi

D.8 Continued development of the electron-phonon optdriver = 7 module of ABINIT. New input variable tmesh, defining a linear mesh of temperatures, see tests v8[44] and v8[45]. Also, debugging and improvement of doc. By M. Giantomassi

D.9 Added netcdf output of phonons for full grid, not just band structure. Only in tetrahedron prtdos 2 case. By M. Verstraete

D.10 On-going development: main executable atdep, for the TDEP algorithm, by Hellman and coworkers. See 98_main/tdep.F90, as well as directory 80_tdep. No automatic tests provided yet, no documentation as well … By F. Bottin, J. Bouchet, J. Bieder.

D.11 Capability to print perturbed vxc potential in response function calculations. By G. Antonius

D.12 On-going modularization of all source F90 files, to get rid off abilint. By M. Giantomassi

D.13 On-going improvements in the doc, to benefit from the new processing capabilities, like central bibliography, matjax, etc … By M. Giantomassi, X. Gonze

D.14 Post-processing script for Raman calculations (script/post-processing/Raman_spec.py). Reads the anaddb output file and extracts the Raman tensor and then calculates the Raman spectra as a function of frequency at a user-defined temperature. Additionally, the script will automatically extract the dielectric tensor as a function of frequency if it is available. By N. Pike

D.15 Refactoring for DFPT+NON_COLL: first version for PAW By M. Torrent

D.16 Fix of several bugs in constrained magnetization calculations magconon. By E. Bousquet

D.17 Wrong sign of the derivative of spherical harmonics for f orbitals. Can lead to problems in BSE and GW calculations if pseudos with explicit f-projectors are used. Detected and corrected by Henrique Pereira Miranda. See 87617d2 for the corrected expressions.

D.18 Add possibility to do DFT+U calculations without spin polarization in the exchange and correlation functional: the spin polarization thus only comes from the U and J terms. Can be used with usepawu = 4, but still under tests. By B. Amadon

D.19 GW is now available with nspinor = 2 with or without spin-orbit coupling, and with nspden = 1 or 4 (collinear or non-collinear spin-magnetisation). Implemented only in the norm-conserving case. Still under testing. See tests from v8[90] to v8[93]. By M. Giantomassi.

D.20 Miscellaneous additional bug fixes and improvements of documentation by: L. Baguet, W. Chen, C. Espejo, M. Giantomassi, Y. Pouillon, M. Torrent, J. Zwanziger.

v8.6¶

Many thanks to the contributors to the ABINIT project between May 2017 and October 2017. These release notes are relative to modifications/improvements of ABINITv8.6 with respect to v8.4.

The list of contributors includes: B. Amadon, G. Antonius, L. Baguet, J.-M. Beuken, J. Bieder, F. Bottin, Y. Bouchet, E. Bousquet, M. Giantomassi, O. Gingras, Ph. Ghosez, M. Giantomassi, X. Gonze, F. Jollet, J. Junquera, A. Martin, F. Naccarato, G. Petretto, N. Pike, Y. Pouillon, S. Prokhorenko, M. Torrent, M. Verstraete, J. Wiktor, J. Zwanziger

It is worth to read carefully all the modifications that are mentioned in the present file, and examine the links to help files or test cases. This might take some time …

Xavier

Version 8.6, released on November 3, 2017.

List of changes with respect to version 8.4 .

A. Warnings and important remarks¶

A.1 The interface between ABINIT and TRIQS has been changed, such that TRIQS 1.4 is used now instead of TRIQS 1.3. No backward compatibility with TRIQS 1.3 has been coded, sorry. See D.4.

A.2 Some changes of names:

input variable gwls_sternheimer_kmax has become gwls_stern_kmax;
input variable gwls_dielectric_model has become gwls_diel_model;
input variable prt_effpot has become prt_model (multibinit input variable);
input variable effmass has become effmass_free;
tutorial tlda has become tdftu;

Also, the input variable gwls_second_model_parameter, not used, has been suppressed.

A.3 The definition of Hund’s coupling J as computed within cRPA has changed: it now uses the same convention as the convention used (for the variable jpawu) in DFT+U and DFT+DMFT in ABINIT (automatic tests, and tutorial are accordingly changed). By B. Amadon

B. Most noticeable achievements¶

B.1 The whole ABINIT documentation has been significantly improved by the introduction of Topics, replacing the previous “lists of ABINIT features”. All the capabilities of ABINIT are now presented in about 70 short topic Web pages. Those topic web pages usually have: - a brief introduction; - the list of related tutorials -if any-; - the list of related input variables (ordered according to their importance for the topics -compulsory, basic, useful or expert-); - possibly example input files; - list of references. Entry point: see the new header of any ABINIT documentation file e.g. the new user’s guide By F. Jollet and X. Gonze (also tests/fixes by B. Amadon, M. Torrent).

B.2 A central bibliography database abiref.bib has been created, and linked to the above-mentioned topics (B.1) but also to other parts of the ABINIT documentation (e.g. input variable list, the tutorials, the theory documents, the acknowledgments). More than 200 bibliographical references are present. Not all bibliographical references of the doc have been entered in this central bibliographic database, though. By X. Gonze and F. Jollet.

B.3 The list of input variables has been restructured, and is now searchable. The input variables for anaddb, aim and optic have been included in the database. By J. Bieder, X. Gonze and F. Jollet.

B.4 The frequency of usage of each input variable (in the set of automatic tests) is now automatically computed, and mentioned in the documentation for this input variable. Examples input files are also now mentioned in the documentation. The input files for the automatic tests can now be directly accessed on the Web, as well as the reference files for the tutorials. By. X. Gonze and F. Jollet.

B.5 Several important developments related to electron-phonon matrix element computation have been made. The Frohlich interpolation procedure for electron-phonon matrix elements, as explained in PRL 115, 176401 (2015), has been implemented. The long-range part of the phonon coupling potential is modeled with the Born effective charges and the dielectric tensor. This long-range part is substracted from the potential before the Fourier interpolation then added after the interpolation. The resulting potential is in much better agreement with the full calculation, as can be verified from the el-ph matrix elements. Also, a functionality has been added in the eph driver (eph_task=5) to only interpolate the phonon potential onto a fine q-point grid. The interpolation is performed one perturbation at a time, and is thus more memory efficient than the previous procedures. There is additional testing of the new “driver” optdrive=7 specifically dealing with electron-phonon related computations (including zero-point renormalisation), especially the interpolation. The symmetries have been fixed. See new tests v8[61]-v8[65]. By G. Antonius and M. Giantomassi.

B.6 ABINIT can now read pseudopotentials in the PSML 1.1 format, as described in https://arxiv.org/abs/1707.08938. This XML-based format is produced by ONCVPSP 3.2 and 3.3, as well as SIESTA’s ATOM 4.2, and allows to perform calculations with the exact same pseudopotential files in both ABINIT and SIESTA. See the new directory ~abinit/tests/psml, tests psml[01] to psml[14]. Note: patches are provided at https://launchpad.net/pspgenpatch to enable PSML output in ONCVPSP. By Y. Pouillon, M. Verstraete, J. Junquera and A. Garcia.

B.7 ABINIT is now interfaced with Libxc 3.0. The interface with Libxc 4.0 is in preparation. Tests libxc[06], libxc[07], libxc[17], libxc[18], libxc[20], libxc[21] have been modified, because some functionals of libxc v2.0 have changed category in v3.0. By M. Torrent.

B.8 A new tutorial, called Electron-positron annihilation has been created. By J. Wiktor and M. Torrent.

B.9 The new input variable chkdilatmx has been introduced, to allow expert users to make ABINIT bypass the stopping criterion related to dilatmx. In practice, if the condition related to dilatmx is not met, ABINIT continues, and delivers an (approximate) optimized geometry and energy, that might be used by external drivers like e.g. USPEX to continue the search for global optimized structures. See input variable chkdilatmx, and test v3[42]. By X. Gonze.

B.10 Implementation of the LDA-½ methodology. Tests to be provided. By F. Jollet.

C. Changes for the developers (also compilers)¶

C.1 There are large changes of the procedure to document ABINIT, for most of the documentation files. The HTML files are now produced from YAML files, under the control of the script ~abinit/doc/generate_doc.py . The documentation that describes this procedure is available on the ABINIT wiki, at https://wiki.abinit.org/doku.php?id=developers:generate_doc . This is directly linked to the modifications in the doc presented in B1-B4. In particular, the Dokuwiki syntax is used for the hyperlinks. By F. Jollet and X. Gonze.

D. Other changes (or on-going developments, not yet finalized)¶

D.1 The “Adaptively Compressed Operator” approach for the fast application of the Fock operator has been implemented, and replaces the traditional way to apply the Fock operator in hybrid functionals (e.g. HSE06, PBE0, …). Hybrid functionals are not yet to be considered in production, though (see D.6), but likely for ABINITv8.8. See tests libxc 51, 52, 53, 67, 68, 69, 70, 71, and also v7#65, 66, 70. By F. Jollet and X. Gonze.

D.2 A set of 6 input files and accompanying references (from 32 to 2048 procs), for benchmarking high-performance computing is available in the new directory ~abinit/tests/hpc . Not yet tested automatically, but this future capability is prepared. By M. Torrent.

D.3 The tutorial on the temperature-dependent electronic structure has been imported from the ABINIT wiki to the usual location ~abinit/doc/tutorial and suppressed from the Wiki. However, it is not yet operational. Work is also going on on tutorial fold2bloch. By X. Gonze.

D.4 Interfacing with TRIQS 1.4 (instead of 1.3). By O. Gingras, B. Amadon and J.-M. Beuken.

D.5 Anaddb can now interpolate and print out the DDB onto an arbitrary set of q-point. The same procedure was used to produce the phonon band structure. Now, with the input variable prtddb, anaddb will produce both the _DDB file and the _DDB.nc files, the latter being separated for each q-point. By G. Antonius.

D.6 On-going work on hybrid functionals: speed-up of the SCF loop, computation of stresses, joint computation of forces and stresses, downsampling the wavevectors. By X. Gonze and F. Jollet.

D.7 On-going work on the implementation of the TDEP algorithm (temperature dependent sampling). By J. Bieder, F. Bottin and Y. Bouchet.

D.8 Replacements of http:// by https:// in many documentation files. By J.M. Beuken.

D.9 Test of non-magnetic DFT+U and DFT+U+SO. See the new test v5#16 By M. Torrent.

D.10 Make DFT+U and local EX-exchange compatible with nspden=1/nspinor=2 By M. Torrent.

D.11 Test of the Velocity Verlet algorithm ionmov=24 See the new test v8#13 By S. Prokhorenko.

D.12 Make thermally occupied supercell of a given size, with input variable thermal_supercell. See test v8#46 By M. Giantomassi

D.13 Write dielectric tensor to anaddb.nc when only perturbations w.r.t. electric field are present; Test for nlflag=2,3 Test the computation of the nonlinear coefficients and first change of dielectric tensor. See test v8#47-50 By F. Naccarato.

D.14 Ongoing work: Raman intensities, in the PAW case, using DFPT. By L. Baguet and M. Torrent.

D.15 Ongoing work on the multibinit project. New hist storage. New tests paral#101-102, to test the anharmonic part. Rationalization of supercell treatment with multibinit By A. Martin, M. Verstraete and Ph. Ghosez.

D.16 On-going work on the extension of DFPT within non-collinear magnetism. By F. Ricci, S. Prokhorenko, M. Verstraete, M. Torrent and E. Bousquet.

D.17 On-going work on DFPT with magnetic field perturbation (Zeeman field). By S. Prokhorenko and E. Bousquet.

D.18 Begin transport epc calculations within the eph part of the code. By M. Verstraete.

D.19 Improvements for the reading and initialization of density (esp. nspden=4). By M. Torrent.

D.20 New interface to the build system and new ac8 config file format. By Y. Pouillon.

D.21 Store fold2bloch results in NetCDF format. Add new option to cut3d to convert DEN/POT from Fortran to netcdf By M. Giantomassi.

D.22 Add mdtemp in the _HIST file. This can be useful (mandatory) for post processing MD/PIMD Add imgmov in the _HIST file. Convenient to know if it is PIMD or NEB/string for postprocessing By J. Bieder.

D.23 Use inversion symmetry if nspden == 4 and NC By M. Giantomassi.

D.24 Update elastic tutorial By J. Zwanziger.

D.25 Add LO-TO terms to netcdf files By M. Giantomassi.

D.26 Numerous miscellaneous additional bug fixes and improvements of documentation by: G. Antonius, J. Bieder, M. Giantomassi, F. Jollet, G. Petretto, N. Pike, Y. Pouillon, M. Verstraete, M. Torrent.

v8.4¶

Many thanks to the contributors to the ABINIT project between January 2017 and May 2017. These release notes are relative to modifications/improvements of ABINITv8.4 with respect to v8.2.

The list of contributors includes: F. Altvater, G. Antonius, L. Baguet, J.-M. Beuken, J. Bieder, E. Bousquet, W. Chen, G. Geneste, M. Giantomassi, Y. Gillet, X. Gonze, F. Jollet, A. Martin, F. Naccarato, G. Petretto, S. Prokhorenko, F. Ricci, M. Torrent, M. Verstraete, J. Zwanziger

It is worth to read carefully all the modifications that are mentioned in the present file, and examine the links to help files or test cases … This might take some time …

Xavier

Version 8.4, released on June 2, 2017.

List of changes with respect to version 8.2 .

A. Warnings and important remarks

A.1 The Github ABINIT Web page, at https://github.com/abinit, allows one to access the mirror of the repository of ABINIT, as well as some other projects related to ABINIT, like AbiPy or the pseudo-dojo.

A.2 The content of the ABINIT Web portal, specifically for the pages related to the presentation of ABINIT (http://www.abinit.org/about/what-is-abinit), has been upgraded.

B. Most noticeable achievements

B.1 Implementation of algorithms to interpolate the electronic band structure, based either on “star functions” of on “B-splines” (as alternatives to Wannier function interpolation). See the input variables einterp, nkpath, and prtebands, and tests old syntax: Tlibxc#42, Tv8#04 replaced by tests/libxc/Input/t41.abi, v8[04]. Work by M. Giantomassi

B.2 The Fock mixing factor for the HSE hybrid functional can be tuned thanks to the input variable gwfockmix. (Warning: this is for the GW-type approach to electronic structure only, not for total energies) See test Tlibxc#43 . Work by W. Chen.

B.3 Implementation of spatially-varying chemical potential, for each atomic species. See the input variables chempot and nzchempot, and tests Tv8#30 and 31. Work by X. Gonze

B.4 Linear geometrical constraints can now be imposed on PIMD runs. This allows ABINIT to use the blue moon sampling technique. See the input variable pimd_contraint, and test Tv8#05 The use of history (HIST) files is now possible with images. Work by G. Geneste and M. Torrent.

B.5 Computation of linear reponse (optics executable as well as BSE part of ABINIT) in the case of temperature-dependent electronic structure. See tests Tv67mbpt#50-53. WARNING: This capability of ABINIT has not been fully tested. However, the basic tests for non-spin-polarized simple semiconductors are OK. As usual, use at your own risk. Work by Y. Gillet and M. Giantomassi.

B.6 Computation of Gruneisen parameters by finite differences, within ANADDB. See test v8#45 Work by M. Giantomassi

B.7 New LOBPCG implementation wfoptalg = 114. This is the default when paral_kgb == 1. Performances are equivalent in standard use (MPI alone) and much better with openmp+multithreaded linalg. It also allows to have only one block for large system and to reduce memory copies. This version has been developed keeping in mind the next generation of HPC. Work by J. Bieder

B.8 New algorithms for the displacement of nuclei (ionmov): - Hybrid Monte Carlo (HMC) predictor (ionmov=25) - Velocity Verlet (VV) NVE molecular dynamics predictor (ionmov=24) See Tv8#12 for ionmov=24. Work by S. Prokhorenko

B.9 Refactoring of ANADDB for the production of DOS and other thermodynamic quantities, also the mean square displacement and mean square velocity. The DOS is obtained using usual DOS methods, instead of the histogram method, and converges much faster with ng2qpt. Activated with prtdos 1 or 2 in anaddb. Work by M. Verstraete.

C. Changes for the developers (also compilers)

C.1 Management of the test farm: the new bot ubu_intel_17_openmpi has been activated, so that Intel 17 is now supported. Also, replacement of shiva_gnu_6.3_py3k by inca_gnu_6.3_py3k, update of graphene (MacPorts) to gcc6.3 + scalapack. By J.M. Beuken.

D. Other changes (or on-going developments, not yet finalized).

D.1 The printing of potentials (e.g. prtvxc …) works now for DFPT. By M. Verstraete.

D.2 New input variable prtphbands. See tests v7#88 and v8#46. By M. Giantomassi

D.3 In case the ANADDB interpolation of phonon frequencies produces erroneously a negative slope around gamma, the new nsphere = -1 possibility allows ANADDB to select a set of IFCs that does not lead to such non-physical behaviour. See test v8#46. Work by M. Giantomassi.

D.4 Added new value for nlflag = 3 that computes only the non-linear susceptibility. By F. Naccarato.

D.5 Computation of forces is now possible in the Wavelets + PAW case. By M. Torrent.

D.6 Ongoing work concerning the new “driver” optdriver = 7 specifically dealing with electron-phonon related computations (including zero-point renormalisation). See new tests v8#41 to 44. Work by M. Giantomassi.

D.7 Ongoing work: Raman intensities, in the PAW case, using DFPT. By L. Baguet and M. Torrent

D.8 Ongoing work related to non-collinear DFPT. Definition of the new input variable rfmagn, as well as Tv8#20 . DFPT with static B-field (Zeeman term only) works for q=(0 0 0). Adjustmanet of dfpt corrections for metals to non-collinear case. Bug fix for GS calculations with finite magnetic field (case of collinear spins). By S. Prokhorenko, F. Ricci, and E. Bousquet

D.9 Ongoing work on the multibinit project. NPT simulations are possible. New tests v8#06 and paral#100, to check the correct IO of XML files. Note that tests Tv7#120-124 have been renumbered Tv8#07-11. Work by A. Martin

D.10 New test paral#62, to make sure paral_kgb = 0 works when there are some idle procs. By M. Giantomassi

D.11 Ongoing work on forces and stresses for hybrid functionals. By F. Jollet

D.12 Ongoing work on the electron-phonon postprocessor ElectronPhononCoupling. Fix bug affecting MPI runs. By G. Antonius.

D.13 Improvement of zheevd using MAGMA (from a msg on the forum) By M. Torrent

D.14 Implemented KSS.nc output with netcdf primitives By M. Giantomassi

D.15 Concerning the Fourier interpolation of the phonon band structure, inside ANADDB, work by G Petretto: - Updates in the calculation of sound velocity, - Small modifications for nlflag == 3 and added some quantities to the anaddb netcdf file, - Fix a bug in the implementation of the new weights

D.16 Concerning Path Integral Molecular Dynamics with Quantum Thermal Bath: allow restart from history file. By M. Torrent

D.17 Refactored the computation of the electric dipole moment. By M. Torrent

D.18 The energy width for the bands in the Boltztrap intrans file is now automatically set. Previously a constant 0.4 Ryd, and the user should have checked by hand if the value was sufficient. Should be considered a bug fix. By M Verstraete

D.19 Numerous miscellaneous additional bug fixes and improvements of documentation by: F. Altvater, G. Antonius, J. Bieder, M. Giantomassi, F. Jollet, G. Petretto, M. Verstraete, M. Torrent, J. Zwanziger.

v8.2¶

Many thanks to the contributors to the ABINIT project between June 2016 and January 2017. These release notes are relative to modifications/improvements of ABINITv8.2 with respect to v8.0.

Moreover, most of them are also described in the Computer Physics Communications 2016 ABINIT paper, doi:10.1016/j.cpc.2016.04.003

The list of contributors includes: B. Amadon, G. Antonius, L. Baguet, J.-M. Beuken, J. Bieder, E. Bousquet, F. Bruneval, W. Chen, M. Giantomassi, Y. Gillet, X. Gonze, G. Petretto, F. Jollet, A. Martin, V. Planes, Y. Pouillon, T. Rangel, F. Ricci, M. Torrent, M. Verstraete

It is worth to read carefully all the modifications that are mentioned in the present file, and examine the links to help files or test cases … This might take some time …

Xavier

Version 8.2, released on February 16, 2017.

List of changes with respect to version 8.0 .

A. WARNINGS AND IMPORTANT REMARKS

A.0 The 2016 article by the ABINIT group is now mentioned in the acknowledgments: “Recent developments in the ABINIT software package. Computer. Phys. Communications 205, 106 (2016)”. See http://www.abinit.org/doc/helpfiles/for-v8.2/users/acknowledgments.html, as well as the notice at the end of ABINIT runs.

A.1 inclvkb 1 has been removed. Now the possible values are either 0 or 2

A.2 The default strategy of so_psp has been changed (see the description of the input variable so_psp).

B. Most noticeable achievements

B.1 Implementation of the Limited-memory Broyden-Fletcher-Goldfarb-Shanno (LBFGS) minimization algorithm. Activate this algorithm using ionmov = 22. From the tests that have been run, this algorithm can be much better than the native implementation of BFGS in ABINIT when one approaches convergence, perhaps because of better treatment of numerical details. This algorithm might become the default in ABINIT, if the better behaviour is confirmed. Test case: v8#02 . The working routines were based on the original implementation of J. Nocedal available on netlib.org. They have been reshaped and translated into modern fortran, then interfaced to ABINIT by F. Bruneval (sources in 45_geomoptim/m_lbfgs.F90).

B.2 A new tutorial is available, on the calculation of the effective interactions U and J using constrained Random Phase Approximation (cRPA) for DFT+DMFT (or DFT+U) calculations. See doc/tutorial/_ucalc_crpa.md as well as the automatic tests tutorial/tucrpa#1-5 . This tutorial was prepared by B. Amadon.

B.3 Implementation of temperature-dependent spectral functions (electronic spectral function, with electron-phonon interactions), as well as real part of phonon self-energy (Pi) for gapped systems. Also, automatic test for spectral function, v7#89, and improved documentation. Work by G. Antonius.

B.4 The RPA one-shot bootstrap fxc kernel has been implemented for GW calculations (gwgamma=-8). See Rigamonti et al PRL 114, 146402 (2014) and Berger PRL 115, 137402 (2015). The test v67mbpt#36 has been updated. Work by W. Chen.

C. Changes for the developers (also compilers)

C.1 The version control system that is used for the development of ABINIT has been changed: the whole ABINIT project has been ported from bzr to git. Work by J.-M. Beuken, Y. Pouillon, M. Giantomassi, X. Gonze, with discussions with many developers.

C.2 New versions of Fortran compilers have been integrated in the test farm: - intel 16.0 - gnu 6.1 and 6.2 - IBM xlf compiler 14.1 - NAG 5.3 Corresponding examples are available in doc/build/config-examples. On the contrary, g95 is not tested anymore. Work by J.-M. Beuken

C.3 The v8 directory for tests has been initialized. By Matteo Giantomassi.

C.4 Python 3 >= 3.4 is now supported in build system scripts (compatibility with py2_ version >= is maintained). By Matteo Giantomassi.

D. Other changes (or on-going developments, not yet finalized).

D.1 The main executable “multibinit” has been created. Its goal is to perform “second-principles calculations”, building model Hamiltonians using the data provided by the DDB (or other info from ABINIT). Tests v7#120-124 (should be renamed) as well as paral#95-98. Work by A. Martin.

D.2 A new “driver” within ABINIT has been defined, specifically dealing with electron-phonon related computations (including zero-point renormalisation). Set optdriver=7 . New input variables: ddb_shiftq, eph_task, eph_transport, prtphdos, prtphsurf. See tests v7#88 and 89. Work by M. Giantomassi and G. Antonius.

D.3 The generation of k-point meshes with kptrlatt and shiftk is now tested. See test v8#03 . Work by M. Giantomassi

D.4 As a follow-up of the Achievement B3 in the release notes of ABINITv8.0 (DMFT + TRIQS), new input variables have been defined for DMFT: dmft_tolfreq and dmftctqmc_triqs_nleg. Automatic tests have been set-up, tests v8#01 and paral#99. Work by B. Amadon and V. Planes

D.5 More systematic tests of the IO in parallel (different files) have been set up, in the norm-conserving, PAW and PAW + spin-orbit cases. See tests mpiio#26, 27 and 28. Also, the case paral_kgb = 0 is now tested with idle processors. See test paral#62. Work by M. Giantomassi

D.6 The load balancing for the repartition of the plane waves among procs is now monitored, and a dynamical equilibration is made possible thanks to the new input variable pw_unbal_thresh. See test mpiio#26. Work by M. Torrent

D.7 The capability to output spin-resolved DOS in case of spin-orbit NC pseudopotential calculations has been checked, and a test has been set-up (test v7#17). Work by M. Giantomassi

D.8 Files generated by ABINIT, and used by BOLTZTRAP are now tested. See v6#11. Work by M. Giantomassi

D.9 Unit tests (fftprof) have been set up for the use of the MKL-DFTI routines: unitary#tfftmkl_03 and 04. Work by M. Giantomassi

D.10 Ongoing work: Raman intensities, in the PAW case, using DFPT. By L. Baguet and M. Torrent

D.11 prtvcbm working with all parallelizations. Tests mpiio 26:28 have been reactivated on 4 processors. By M. Torrent

D.12 On going work related to non-collinear DFPT. By F. Ricci, S. Prokhorenko, and E. Bousquet

D.13 MBPT: support for the commutator [Vnl r] in the case of NC pseudos with more than one projector per l-channel has been added. Tests in v67mbpt[40] (GW run with psp8 files). SOC is not yet available, though. The KSS file continues to use the old implementation to maintain backward compatibility hence outkss won’t produce the file if multiple-projectors are detected inclvkb 1 has been removed. Now the possible values are either 0 or 2 By M Giantomassi

D.14 For Hirshfeld and Bader: doc and warning. For Hirshfeld charges the output was unclear: the density integral (electrons only) was called the hirshfeld charge, as opposed to the net one. For Bader there was no check that the core charge file corresponded to the pseudopotential used. Now checks at least that it integrates to znucl-zion, which is a first step. Important as the default fc files on the web do not have semicore electrons, whereas many of the psp8 will. Contribution by M. Verstraete.

D.15 Updated acknowledgments, including the 2016 paper. By X. Gonze

D.16 Ongoing work on forces and stresses for hybrid functionals. By F. Jollet

D.17 Ongoing work concerning weights for the Fourier interpolation inside ANADDB. By G. Petretto

D.18 Numerous miscellaneous additional bug fixes (to the sources, as well as to the build system, including patches for the fallbacks), and improvements of documentation by: G. Antonius, L. Baguet, J. Bieder, F. Bruneval, M. Giantomassi, Y. Gillet, G. Petretto, Y. Pouillon, M. Verstraete, M. Torrent (in particular, for DFPT+PAW).