Copyright (C) 2010-2014 ABINIT group (XG)
This file is distributed under the terms of the GNU General Public License, see
For the initials of contributors, see ~abinit/doc/developers/contributors.txt .
Many thanks to the contributors to the ABINIT project between
January 2014 and June 2014. These release notes
are relative to improvements of ABINITv7.8 with respect to v7.6.
The list of contributors includes :
F. Abreu, B. Amadon, T. Applencourt, J.-M. Beuken, J. Bieder, E. Bousquet, F. Bottin, J. Bouchet,
F. Bruneval, D. Caliste, M. Delaveau, B. Dorado, C. Espejo, G. Geneste, M. Giantomassi, Y. Gillet, X. Gonze,
D. Hamann, F. Jollet, A. Levitt, I. Lukacevic, A. Martin, C. Martins, S. Ponce, Y. Pouillon, T. Rangel, G.-M. Rignanese,
A. Romero, A. Shukri, M. Torrent, M. Van Setten, M. Verstraete, Bin Xu, V. Zecca, 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 ...
Version 7.8, released on June 30, 2014.
Changes with respect to version 7.6 :
B. Most noticeable achievements (for users)
C. Changes in the package, for developers
D. Other changes (or on-going developments, not finalized)
A. WARNINGS AND IMPORTANT REMARKS
A.1 As advertised in the recent announcements of ABINITv7.6, some developers are presently "cleaning" ABINIT,
i.e. removing little used code, or rarely used capabilities, that were slowing down the
developments by requiring too much maintenance.
As a consequence, in ABINITv7.8, some backward compatibilities have been broken.
In particular, the input variable mkmem (and associated mkqmem, mk1mem, mffmem) have been removed.
Calculations using out-of-core storing of wavefunctions are not possible anymore.
Instead, the user should rely on the parallelisation, with distributed memory.
The computation of lattice wannier function has also been disabled, as well as the self-consistent
phonon calculations (see C.7).
A.2 Although the availability of NetCDF, libxc, and MPI-IO (in case of parallelism) is not yet mandatory,
v7.8 is likely the last version of ABINIT for which NetCDF is not mandatory (it will be mandatory
for ABINITv8). On the other hand the requirement of libxc and MPI-IO will likely not be enforced in ABINITv8 .
B. Most noticeable achievements
B.1 Computation of the U parameter (for LDA+U or DMFT), using the cRPA screened interaction formalism.
This is available based on two models : the d-dp(b) model or the energy range model.
For the latter, see the new input variable ucrpa_window.
Tests v7#68 and 69.
Published in Phys. Rev.B 89, 125110 (2014).
By B. Amadon and T. Applencourt
This can be used in LDA+U as well as in DMFT
(Note : the CTQMC solver has been included in the trunk (dmft_solv=5), see the release notes for ABINITv7.6, point B.2)
B.2 The Path-Integral Molecular Dynamics in NVT ensemble is now available
with two thermostats (Langevin and Nose-Hoover chains) and possible coordinate transformations (primitive, staging, normal modes).
See the test v7#08 .
Adiabatic Path-Integral Molecular Dynamics is under development. New input variables : adpimd and adpimd_gamma.
Work by G. Geneste and M. Torrent.
B.3 Computation of the random stopping power, in the RPA and in the TD-LDA.
Input variables npvel and pvelmax.
Test v7#67 .
Work by A. Shukri and F. Bruneval.
B.4 Response function from Molecular Dynamics
Phonon spectrum obtained from trajectories (with a minimal set of coefficients)
Included in APPA.
Work by Francois Bottin, Johann Bouchet, and Alexandre Martin
B.5 ABINIT has the capability to read the pseudopotentials created by the ONCVPSP code by D. Hammann.
Also including the spin-orbit interaction. Documentation available in doc/psp_infos/psp8_info_txt .
See test v7#07
Work by D. Hamann
B.6 Computation of the electron-phonon transport in the inelastic variational approximation are in beta testing.
See test v6#95.
Work by Bin Xu and M. Verstraete
B.7 FFT : completion of the parallel implementation for fftalg=312 (FFTW3) and fftalg=401 and 402 (SGoedecker) ,
also with unitary testing.
Also, paral_kgb=1 can take advantage of composite operations (fftalg=402, 312)
New MPI-FFT version based on zero-padding and powered by FFTW3/MKL (312)
Automatic selection of the best FFT algorithm depending on the value of npfft
See tests tfftfftw3 #05,06 and tfftsg #05,06
NOTE : at present the default value of fftalg is still 401 when FFTW3 is not available and paral_kgb==1.
Users are encourages to use fftalg 402 or even better 312 for their calculations. Later, the default
will be changed to 402 - automatic test references have to be updated.
Work by M. Giantomassi
B.8 Constrained magnetization calculations can now be performed.
Input variables magconon and magcon_lambda
From M. Verstraete, I. Lukacevic and E. Bousquet
C. Changes for the developers
C.1 The test farm has undergone many modifications.
New bots :
- tikal_gcc49 : gfort v4.9 is thus officially supported. This bot is used for full debugging.
- petrus_nag (NAG 5.3.1) : NAG compiler is officially supported. The debugging is quite aggressive.
- buda_intel14 (14.0.2) : quite aggressive optim : -O3 -xhost
- bigmac_gcc48_sl : testing MacPorts, and Scalapack is also activated.
Rendering ABINIT NAG enabled was a significant debugging work for ABINIT !
Updated bots :
- ktulu_gcc48 (ubuntu packages) : OS ubuntu 14.04 instead of ubuntu 10.04
- inca_gcc44_sdebug (for buildsys and abirules) : use of make (version 4) able to synchronize output with -j option
- cronos_gcc46_paral : executes the entire test with np=2
Suppressed different bots : amd-g95_milou_ftrace_nofb, amd-g95_milou_ftrace_serial_nofb, amd-gfortran4.8_milou_fpetrap_nofb,
amd-gfortran4.8_milou_fpetrap_serial_nofbi, intel-gfortran4.4_ktulu, intel-gfortran4.7_bigmac_sernofb
The following compilers are either obsolete, or considered too unstable/buggy for ABINIT :
g95, pathscale, open64 . This does not mean that ABINIT will not work, but there is no guarantee for ABINIT to run correctly.
Work by JM Beuken, and MGiantomassi, with contributions from Vittorio Zecca for the NAG enabling.
C.2 The modularization of the build system is progressing : bindings is autonomous ;
creation of the "watch" subdirectory ; "Special" tests have been separated from usual non-regression tests.
Work by Y. Pouillon
C.3 The modularization of the sources is progressing : many subroutines have been gathered in modules.
About two dozen of new modules have been created : m_alloc.F90, m_anaddb_dataset.F90,
m_array.F90, m_atmdata.F90, m_ddb.F90,
m_dfpt_db.F90, m_doublegrid.F90, m_dynmat.F90, m_eig2d.F90, m_ewald.F90, m_fftcore.F90, m_fstrings.F90, m_ifc.F90, m_kxc.F90,
m_psps.F90, m_sg2002.F90, m_special_funcs.F90, m_xieee.F90,
Work by Matteo Giantomassi (and others)
C.4 The script mkrobodoc_dirs.py, to provide headers of directories for ROBODOC, has been written.
Work by Matteo Giantomassi
C.5 There is a new naming scheme for the procedures exported by m_xmpi (e.g. old: xsum_mpi, new: xmpi_sum)
Work by Matteo Giantomassi.
C.6 The creation of sections of output written in YAML (see e.g. http://en.wikipedia.org/wiki/YAML ) is progressing.
See the use of yaml_out in abinit.F90, anaddb.F90 and optic.F90
(Note : in ABINITv7.6, The build system supports YAML, and The YAML library has been integrated in the build system)
Work by M. Giantomassi and Y. Pouillon
C.7 Cleaning : suppression of lattice wannier function (LWF) calculations, and self-consistent phonon calculations (SCPhon).
The automatic tests for LWF, v3#96 and 97, and for SCPhon, v5#100 have been suppressed.
Input variables for SCPhon have been suppressed : prepscphon, scphon_temp, scphon_supercell.
By M. Giantomassi and M. Verstraete
D. Other changes
(or on-going developments, not yet finalized).
D.1 The input variable gw_qprange has been introduced, to allow selecting the
set of bands and k-points for which the GW correction has to be computed (optdriver=4 only).
Work by M. Giantomassi
D.2 The computation of oscillator matrix elements within the PAW formalism can not be done
following two different methodologies : Arnaud and Alouani or Shishkin and Kresse.
Input variable pawoptosc .
Tests that explicitely specify these : v67mbpt#29 (for Arnaud Alouani expression), and
v7#68,69 for Shishkin Kresse .
Work by M. Giantomassi.
D.3 Temperature-dependence of the gap
Addition of current qpoint inside the EIGR2D.nc and adaptation of the temp dep python script to read nc files
Output EIGR2D file (second order derivative wrt phononic displacement) in netcdf format following the ETSF-io specifications
Work by S. Ponce
D.4 Tutorial files have been updated/corrected/checked, for the ABINIT tutorial in CECAM Lyon, May 2014.
Work by E. Bousquet, M. Giantomassi, GM Rignanese, M. Torrent, M. Verstraete, J. Zwanziger
D.5 Point symmetrize all efg tensor contributions, update tests accordingly.
Work by J. Zwanziger
D.6 A new flavour of VdW_WF (vdw_xc=14) was added, which implements the recent vdW-QHO-WF method, a combination
of the MBD method of Tkatchenko with the localized Wannier functions approach. This method includes many body
dispersion contributions to the van der Waals correction to the DFT energy.
See test wannier90#14 .
Work by C. Espejo and A. Romero
D.7 Generalize prt1dm for spinors
Work by M. Verstraete
D.8 Improvements and debugging of the Dynamical Mean Field Theory (DMFT) with Continuous Time Quantum Monte Carlo solver
is going on.
Work by J. Bieder and B. Amadon
D.9 Work on ABINIT GUI by F. Abreu and Y. Gillet
D.10 On-going efforts on the development of Hartree-Fock and hybrid functionals.
Hartree-Fock calculations are now coded within Norm-Conserving PSP formalism,
with the use of symetries and parallelism.
PAW + Hartree-Fock is under development.
Still in the testing phase.
New input variables : cgtyphf, nbandhf, nkpthf, nnsclohf, nphf.
Tests v7#65 and 66 for PBE0, and Tests paral#93 (HF in parallel) and 94 (PBE0 in parallel).
From C. Martins (helped by M. Torrent and B. Amadon)
D.11 Development of scripts for high-throughput GW calculations
Work by M. Van Setten and G.-M. Rignanese
D.12 Work on Berry phase calculations with PAW, by J. Zwanziger (creation of the routine dsdr_k_paw.F90)
Forces with electric field and PAW now working, more tests to come
D.13 Contribution from M Delaveau: improvement of parallelization over atoms for DFPT
D.14 Debugging of PAW+DFPT+metals, and also for Born Effective charges.
Phonon band structures are now OK within PAW+DFPT, but not yet sure of the BEC.
By M. Torrent and A. Martin
D.15 Work on elastic and piezoelectric tensor in PAW.
By A. Martin and M. Torrent
D.16 Implementation of several flavours of k-space interpolation technique for Bethe-Salpeter.
Work by Y. Gillet.
D.17 PAW + Bigdft project
The ABINIT PAW library (libpaw) is in progress, see src/42_libpaw.
Work by T. Rangel + M Torrent.
D.18 Work is continuing on the JTH PAW dataset table.
Soon the JTH 2.0 table will be released.
Work by F. Jollet.
D.19 On-going work on PAW+DFPT : improving the stability, the parallelism has been fully tested and improved.
See test mpiio#62
Work by M. Torrent and B. Dorado.
D.20 Continued efforts on the development of the abipy library
From M. Giantomassi
D.21 Implementation of the Chebishev algorithm, allowing to scale to 10000 procs (and beyond) with the band-FFT parallelism.
Still, not yet completely ready for mass production (missing istwfk > 1, ecutsm, nspinor)
By A. Levitt and M. Torrent
D.22 The Bethe-Salpeter Equation with spin-polarization is now available in the case of direct diagonalization.
See test v67mbpt#29.
Later, it will be available with the Haydock method as well.
The model dielectric function can be computed on-the-fly, allowing a large saving of memory
See test v67mbpt#11.
Work by M. Giantomassi
D.22 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 :
B. Amadon, J. Bieder, F. Bruneval, D. Caliste, M. Giantomassi, Y. Gillet, F. Jollet,
S. Ponce, Y. Pouillon, T. Rangel, M. Torrent, M. Verstraete, V. Zecca, J. Zwanziger