Copyright (C) 2006 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 .
Many thanks to the contributors to the ABINIT project
between September 2005 and June 2006. These release notes
are relative to the progresses that have been made in ABINITv5.1 compared to v4.6.
This combines "normal" contributions, on top of v4.6, that were included
in an internally available v4.7,
and a completely redesigned build mechanism, tested in v5.0 .

Concerning the latter, most of the work has been done by Yann Pouillon who deserves
very special thanks... There has been contributions to the build mechanism
due to a few testers also
(Pierre-Matthieu Anglade, Thierry Deutsch, Masayoshi Mikami,
 Gian-Marco Rignanese, ...).

The list of "normal" contributors includes :
P.-M. Anglade, J. Bhattacharjee, J.-M. Beuken, F. Bottin,
P. Boulanger, F. Bruneval, M. Cote, T. Deutsch, M. Giantomassi, T. Hoefler, J. Jansen
F. Jollet, S. Le Roux, Z. Levine, Mikami-san,
P. Plaenitz, Y. Pouillon, G.-M. Rignanese,
A. Roy, R. Shaltaf, L. Sindic, B. Tardif, M. Torrent, M. Verstraete,
U. Waghmare, E. Walter, G. Zerah

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 5.1, released on February 22, 2006.

Changes with respect to version 4.6 :

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. See the changes described in C1-C4.

A2. The electron-phonon calculation still contained some bugs, that have been
    corrected in the v5.1.2, and then v5.1.3 by M. Verstraete (see D9). The tutorial
    has been updated.

A3. Some small numerical changes have been done in the PAW, by MTorrent (on the order
    of 0.2 mHa on the total energy, see tests/tutorial paw1_2, paw1_3, paw1_3).

A4. The default value of mqgrid and mqgriddg has been raised from 1201 to 3001.

__________________________________________________________________

B.  Most noticeable achievements

B1. The build system has been completely redesigned, as well as the structure
    of the ABINIT package. These changes are described in C1-C4.

B2. GW part of ABINIT  (improvements by F. Bruneval):
    (1) The GW part of ABINIT is now able to treat metals
    (2) The GW calculations can be done without plasmon-pole model
    (3) Hartree-Fock, screened exchange and COHSEX are implemented
    (4) The Faleev method, see PRL 93, 126406 (2004), is implemented.
    Moreover, it is possible to do some self-consistency steps, by exploiting
    different datasets (of course this is still rather primitive ...).
    See the input variable gwcalctyp , as well as automatic tests Test_v4#84 to 88.
    See also Fabien Bruneval's thesis, for results obtained for Cuprous Oxyde (Cu2O) thanks
    to these techniques. A paper is in preparation.

B3. GW part of ABINIT  (improvements by R. Shaltaf, supervised by GM Rignanese):
    (1) Introduction of two new Plasmon Pole models
            for GW calculations. See the input variable "ppmodel".
    (2) Parallelisation of GW calculations on k points. Test case M has been created
            in tests/paral to show how this is working.
    (3) Added automatic tests for mrgscr utility,

B4. New utilities have been created : AbinitInputMaker, StructureViewer, BandStructureMaker
    They can be found in util/users.
    They have been provided by Benjamin Tardif (under supervision of M. Cote, with also some help by Paul Boulanger)
    BandStructureMaker can be used in production (see the help file doc/users/AbinitBandStructureMaker_manual.pdf),
    the other utilities are still under development.
    BandStructureMaker provides an easy way to rapidly plot electronic band structures
    from Abinit output files.

__________________________________________________________________

C. Changes for the developers

C1. The package has been completely reordered by Yann Pouillon,
    with the appearance of
    five main subdirectories of the top directory : src, lib, tests,
    doc, config .

    "src" now contains all the old "Src_xx" directories, whose names
    have been changed by the suppression of the prefix "Src". Inside
    these subdirectories, the structure is the same as previously

    "lib" now contains all the old "Lib_xx" directories, whose names
    have been changed by the suppression of the prefix "Lib". Inside
    these subdirectories, the structure is the same as previously

    "tests" now contains the old directories Tutorial, and Test_XX .
    Inside these subdirectories there has been some reordering,
    including the creation of "Input" subdirectories, containing
    all the input files.

    "doc" replaces the old "Infos". New subdirectories have been
    created inside "doc", to group the information for the
    users, the one for the maintainers, etc .

    "config" is a brand new directory, that contains most of the
    scripts and information needed to build ABINIT using the new build
    system.

    There are also less important subdirectories, like "util" or
    "extras" in which some previously existing files have been grouped,
    with some new ones.

C2. The new build mechanism, set up by Yann Pouillon, is completely
    different with respect to the one of v4.6. The basic mechanism
    is described in part in the ~abinit/doc/install_notes/install_v5.1.html
    file. The reading of this file is mandatory to understand
    the modifications of the present v5.1 with respect to v4.6 .
    Automake and autoconf are used. For a more elaborate
    understanding of the build mechanism, the interesting person
    should dive inside ~abinit/config ...

C3. There are now Makefile files in each directory. These Makefile
    files have been produced automatically by automake and autoconf. Most of them
    include only the basic targets defined by automake. The Makefile
    of several directories (~abinit, ~abinit/tests, ~abinit/doc, ~abinit/src)
    also contain additional, more elaborate targets.
    They can be obtained by typing "make help" or also
    "make help_dev" in the corresponding directory.

C4. Two new src subdirectories have been created : 06geomoptim and 04occeig.
    Also, many routines have been moved from one of the src subdirectories
    to another, that corresponds better to their level of use, or in view of
    minimizing the time needed to compile partly the sources.

C5. It is possible now to produce an executable that only deals with the
    GW part of abinit. Compilation is much faster, since it avoids compiling
    many routines e.g. those needed for the response functions.
    Simply type 'make abinitgw' just after 'configure', in the top
    ABINIT directory. An executable named 'abinitgw' will be produced.

C6. A new directory ~abinit/src/02nlstrain has been created, and contains
    all the cont*str*.F90 routines previously stored in 02nonlocal.
    This is to allow changing the options of compilations of these routines,
    in order to allow for faster compilation, if they are not used.

C7. Not everything is working yet ... For a list of problems,
    see the file ~abinit/KNOWN_PROBLEMS. This is the new name of the
    old ~ABINIT/Infos/known_problems file of v4.6 .

__________________________________________________________________

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

D1. Localized Wannier functions can be computed, thanks to new capabilities
    of cut3d, provided by Joydeep Bhattacharjee, supervised by Umesh Waghmare.
    See Test_v4#47-49, and the cut3d help file on the Web,
    or ~ABINIT/Infos/cut3d_help.html
    Note : not yet maximally localized Wannier functions ; still somehow in development :
    should be in main abinit instead of cut3d.

D2. New input variable inclvkb, to allow some speed up of the GW-screening calculation,
    at the expense of some loss of accuracy.

D3. Parallelisation on the perturbations, by Philipp Plaenitz
    See new input variables paral_rf and ngroup_rf
    See Test_paral case K  (warning : to be tested more carefully)

D4. Parallelisation on the FFT based on MPI, by Torsten Hoefler.
    See new input variable gpara.
    Warning : still in development.

D5. There is a new utility, named abilint, currently developed,
    by TDeutsch. This has the aim of building automatically
    interfaces between routines, to ease the checkings by compilers.
    (in development, not integrated in v5.1.1, although present in the GNUArch repository)

D6. Improvements in the PAW part of ABINIT, by M. Torrent, F. Jollet and F. Bottin
    Tests of differeznt mixing schemes for PAW (see tests/v4/Input/t07.in).
    Actually, the PAW of ABINIT can now be used in production. Announcement
    by M. Torrent will follow...

D7. New version of lobpcg : wfoptalg=5 (lobpcgII)  (S. Le Roux, ARoy and GZerah)
    See GZ mail on Nov 7

D8. Work on the self-consistency, by PM Anglade.
    (in development, not integrated in v5.1.1, although present in the GNUArch repository)

D9. Bugs have been correctred in the electron-phonon part : in some cases,
    the computed phonon lifetimes of some branches were exchanged.

D10. Parallelisation of the FFT in the case wfoptalg=4 (ARoy and GZerah)
     Still undergoing changes, also connected to modifications D3 and D4.

D11. Bug fixes, from v5.1.2 to v5.1.4, by YPouillon, FBruneval, JJanssen, MBoulet,
     SPesant, MVerstra, MTorrent, JJansen, Lu Jianmin, ZLevine, EWalter, PPlaenitz, MMikami,
     LSindic, MGiantomassi.

D12. New input variables for electron-phonon case have been added, from MVerstraete.
     These new input variables are : gkk2exist, gkk2write, gkk_rptexist, gkk_rptwrite,
     gkqexist, gkqwrite, phfrqexist, phfrqwrite.

D13. It is now possible to read the starting electronic density from a _DEN file
     for self-consistent GS calculations (it was only possible for non SC runs before).
     In this case, both are read, and one obtains a perfect restart.
     Test have been changed : tests/v2/Input/t11.in and tests/v3/Input/t18.in
     (From MTorrent)