Copyright (C) 2008 ABINIT group (XG)
This file is distributed under the terms of the GNU General Public License, see
~abinit/COPYING or
http://www.gnu.org/copyleft/gpl.txt .
For the initials of contributors, see ~abinit/doc/developers/contributors.txt .
Many thanks to the contributors to the ABINIT project
between August 2006 and February 2007. These release notes
are relative to improvements of ABINITv5.3 with respect to v5.2.
The list of contributors includes :
B. Amadon, P.-M. Anglade, M. Arai, M. Beland, F. Bottin, P. Boulanger, M. Boulet,
F. Bruneval, D. Caliste, J.-P. Crocombette, T. Deutsch, L. Diaz, L. Genovese,
A. Ghasemi, M. Giantomassi, S. Goedecker, D. Hamann, T. Hoefler, F. Jollet,
S. Mazevet, M. Mohr, A. Neelov, V. Olevano, Y. Pouillon, V. Recoules,
G.-M. Rignanese, D. Sangalli, S. Sarholz, R. Shaltaf, M. Torrent,
M. Verstraete, G. Zerah, O. Zilberberg
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 ...
Please note the WARNINGS !
Xavier
__________________________________________________________________
Version 5.3, released on January 21, 2007.
Changes with respect to version 5.2 :
A. WARNINGS.
B. Most noticeable achievements (for users)
C. Changes in the package, for developers
D. Other changes (or on-going developments, not finalized)
__________________________________________________________________
A. WARNINGS
A1. Several default values have been changed.
The old behaviour of the code can be obtained by specifying by hand, in the
input file, the old default values.
The list of changed variables for abinis/abinip is as follows ...
iscf 7 in the norm-conserving case, 17 in the PAW case (previously 5 and 14)
prtden 1 (previously 0)
prteig 1 (previously 0)
chkexit 0 (previously 1 for the sequential version, and 2 for the parallel version)
nstep 30 (previously 1)
For anaddb, some input variables were also changed :
asr 1 (previously 0)
symdynmat 1 (previously 0)
The input files of the automatic tests have been appended, as needed, by the old default values
of variables (in most cases, only iscf !). Gradually, this should be suppressed, to test preferably the default
values, but it will take some time to update all the reference files.
A2. The automatic test directory for the tutorial "~abinit/tests/tutorial"
has been split in two directories : one with the old name, and a new
"~abinit/tests/tutorespfn", with all the tutorials that use
the response function (phonon/homogeneous electric field/strain) part of ABINIT.
A3. In GW calculations, the plasmon-pole model 2 was coded for materials with an inversion symmetry,
but applied irrespective of this constraint. Its generalisation to materials
without an inversion symmetry has been coded, see D12. As a consequence
numerical results for materials without an inversion symmetry might change. For SiC,
changes are on the order of 0.05 - 0.1 eV.
A4. The default value for the configure option smart-config has been set to "no"
A5. There has been improvement of the MPI support, by Yann Pouillon.
Note that "--enable-mpi=yes" is now the default and triggers auto-detection.
Note that this auto-detection nevertheless expects "--with-mpi-prefix" to be filled
(and no other "--with-mpi-*" will be read). By contrast with "--enable-mpi=force",
auto-detection is by-passed, and "--with-mpi-*" are to be set manually.
So, the behaviour of the build system has changed in this respect.
The user/developer might have to update his config file (hostname.ac file)
accordingly. More information to be found in the directory ~abinit/doc/config ,
in the file addendum-mpi.html .
__________________________________________________________________
B. Most noticeable achievements
B.1 DFT+U (LDA+U / GGA+U) has been implemented within the PAW framework,
and is now available in production.
A short documentation is available in dev/users/HowtodoLDA+U.txt
See also test case v5/t08 and input variable usepawu.
Work done by B. Amadon and coworker F. Jollet.
B.2 The Band/FFT parallelisation is available, for testing purposes,
for a specific combination of run parameters, see the input variables
npfft, npband (+ others, like fft_opt_lob). This parallelisation
has been developed by F. Bottin and G. Zerah, using the S. Goedecker
FFT routines, with also some initial help by M. Boulet (and T. Hoefler).
Preprocessing flags -DSCALAPACK and -DMPI_FFT are needed.
Do not use the -DMPI preprocessing flag.
A typical input file can be found in ~abinit/tests/paral/t_bandfft.in .
With this input file, a 300 speed-up has been demonstrated, on 468 processors.
B.3 The thermal conductivity (phonon transport, limited by electron-phonon
collisions), and the electrical conductivity (electron transport,
limited by electron-phonon collisions) can now be computed inside anaddb.
This has been developed by JP Crocombette, with help by M. Verstraete and F. Jollet.
These functionalities are activated by the anaddb input variable ifltransport .
Automatic test should be provided.
B.4 The electrical conductivity, the thermal conductivity, and the thermopower
(electron transport, high temperature, Kubo-Greenwood formalism), can be computed
using PAW. The input variable prtnabla must be activated.
Notes on the use of these features and their implementation can be found in
~abinit/doc/users/conductivity_paw_manu.tex (see also the
similar notes for the norm-conserving pseudopotential case,
~abinit/doc/users/conducti_manual.tex).
This implementation has been realized by S. Mazevet, with help by
V. Recoules, M. Torrent, and F. Jollet.
Automatic test should be provided.
B.5 Spin-polarized systems can now be treated in GW (nsppol=2).
Also, Symmorphic symmetries are now used in the GW part of ABINIT in order to reduce
the time needed for the computation of chi and sigma at q points
that are invariant under some symmetry operations.
See the input variables symsigma and symchi.
An automatic test is provided : v5#62 (BCC hydrogen)
Note 1 : the default is such that these savings are not activated,
this should be changed ...
Note 2 : parallelism over bands in GW is not compatible with nsppol=2 as of v5.3.0)
Contribution by M. Giantomassi.
B.6 ( WARNING : the parallelism over bands appear not to give exactly the same
result as the sequential version. Hence, it has been temporarily made unavailable,
except for expert users ... )
Parallelism over bands is available for GW. See the input variable gwpara.
This parallelism allows for memory savings, and is quite efficient.
At present, either the band parallelism or the k point parallelism is available
in GW, they cannot be used simultaneously.
This has been implemented by R. Shaltaf.
(Note, however, that parallelism over bands in GW is not compatible with nsppol=2 as of v5.3.0)
Automatic test for band parallelism : paral#N
B.7 ABINIT is able to produce the different kind of files that follow the Nanoquanta/ETSF
file format specification (http://www.etsf.eu/fileformats) :
density/potential files and wavefunction files.
This NetCDF-based fileformat is intended to be the standard file format for exchanging
data between the softwares of the nanoquanta EU Network of Excellence.
At present, they can be read by the V_Sim visualization program
(http://www-drfmc.cea.fr/sp2m/L_Sim/V_Sim/index.en.html).
The implementation has been done by D. Caliste, with some help by Y. Pouillon,
and is based on a library developed by D. Caliste for Nanoquanta.
See the tests contained in the new directory tests/etsf-io .
The use of Nanoquanta/ETSF file specification is activated by the
use of the input variable accesswff set to 3.
ABINIT needs to be compiled with the NetCDF and etsf-io libraries in order
to produce these files.
B.8 This version of ABINIT is able to deal with wavelets
for representing the wavefunctions. At present, the full SCF loop is working:
density and total energies are computed. Forces are also computed.
Only HGH and GTH pseudopotentials are supported.
This version is still only for developpers.
Effort from the BigDFT community : S. Goedecker, L. Genovese, T. Deutsch,
A. Neelov, D. Caliste, A. Ghasemi, O. Zilberberg.
See C1
Automatic tests are available in tests/bigdft
B.9 The Casida approach to excitation energies in case of finite systems
has been generalized to the (collinear) spin-polarized case, by D. Sangalli,
in collaboration with GM Rignanese and XGonze.
See the automatic tests v5#61 and #62.
B.10 The atomic temperature factors (root mean square expectation values of the
atomic displacements due to thermal fluctuations) are available.
This had been initially implemented by C. Lee in RESPFN, the precursor of
optdriver=2 part of ABINIT (before 1997),
but has been recently debugged and reinstalled by P. Boulanger.
This is automatically activated in ANADDB with the input variable thmflag .
An automatic test (v5#22) has been introduced by P. Boulanger
B.11. A new tutorial lesson on GW is available.
It concerns the metallic case, and the computation of spectral functions.
It has been developed by F. Bruneval.
__________________________________________________________________
C. Changes for the developers
C1. All the names of the directories in src have been changed.
Most of the directories underwent a trivial change : change
of the first digit of the name, usually a "0" to "1", e.g. :
01util is now 32_util . The directory 01managempi
has been split in three parts : 51_manage_mpi, 12_hide_mpi
and 59_io_mpi . Some other directories had their level adjusted,
e.g. 02psp is now 13psp (instead of simply 12psp),
and 04rsprc is now 65_rsprc. Some other dirs has been created,
like 21paral_md .
These changes were needed in order to treat correctly the
dependencies between directories, for which several misplacements
had been seen thanks to the use of abilint.
C2. New directories have been created for the wavelets :
lib/bigdft to host the BigDFT library
src/62_wvl_wfs to host new ABINIT routines needed to drive wavelet computations
See D1
C3. The dataset content is now copied outsied of driver.F90, by the routine dtsetcopy.F90.
Warning : the allocated dtset arrays must be deallocated in dtsetfree.F90
C4. It is now possible to launch sets of tests focused either on PAW,
GW or TDDFT (Casida approach). In the test directory, issue :
make tests_paw (or the other possibilitites tests_gw or tests_tddft)
Documented : in tests dir, issue make help
__________________________________________________________________
D. Other changes
(or on-going developments, not yet finalized).
D1. Many bug fixes for the build system, from Y. Pouillon.
D2. Hessian and DIIS for geometry optimization, by D. Caliste .
When ionmov=20 . Under development, not yet working.
No documentation, no automatic tests.
D3. Improvements of abilint.py , to make it faster. By T. Deutsch.
D4. Introduced new values for input variable iprcel, for testing
and development purposes, by PM Anglade
D5. Modify ABINIT to be able to perform quadruple precision computations,
by PM Anglade
D6. New input variable tolrff (SCF convergence tolerance criterion,
similar aim as toldff, but easier to use), by MVerstraete.
D7. Cleaning and beautification of the elphon part of ABINIT, by Matteo Giantomassi.
D8. Improvements of MPIO parallelism, by M Boulet and G. Zerah
D9. It is possible to compute an "average frequency" for phonons
(as a function of temperature), that might be used to obtain
an estimate of the Gruneisen parameter.
See the associated input variable iavfrq. Developed by M. Torrent.
Should provide an automatic test ...
D10. Two new input variables have been introduced in anaddb, for the electron-phonon
treatment : ep_b_min and ep_b_max . Still to be described.
D11. A new utility, anascr, is available to analyse the screening
files produced by GW computations.
Work done by M. Giantomassi.
Should provide an automatic test and some documentation ...
D12. The plasmon pole model 2 needed a modification to work for
systems without inversion symmetry. This has been implemented
by M. Giantomassi.
See tests/v3, case 31.
D13. A new approach to the computation of the commutator of the non-local
part of the pseudopotential with the position operator has been
implemented by M. Giantomassi, with inclvkb=2 .
Not yet available with the parallelisation over bands.
Automatic test needed.
D14. Simple mixing damping of the density mixing for GW self-consistency,
from F. Bruneval. See the input variable rhoqpmix.
See tests/v4, case 88 .
D15. Introduce thmflag=2 in anaddb, by P. Boulanger.
Used to print phonon frequencies associated with grid 2.
There should be an automatic test.
D16. Test of core-level shifts, for NH2-CH3 vs NH3, from GMRignanese
D17. The PAW approach can take benefit of the antiferromagnetic symmetry.
This has been coded by M Torrent.
Still need an automatic test.
D18. The build system has undergone many minor changes, mostly bug fixes, by Y. Pouillon.
D19. Cleaning of the old NetCDF implementation, by Y. Pouillon
D20. Moments of the Eliashberg function are now computed, from M. Verstraete
D21. Introduction of a new input variable pawstgylm, to allow CPU/storage trading, from M. Torrent
There should be an automatic test.
D22. Other bug fixes, by D. Hamann, M. Arai, S. Sarholz, GM Rignanese, V Olevano, D. Caliste,
T. Deutsch, M. Mohr, L. Diaz, B. Amadon, M. Beland, R. Meyer
D23. make multi has been improved
D24. There is a new option to the "configure" script : "enable-cclock", which triggers the use of the C clock for timings.