Copyright (C) 2001 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 following contributors to the ABINIT project
between May 2001 and October 2001:
Christophe Audouze, Jean-Michel Beuken, Razvan Caracas,
Rex Godby, Francois Jollet, Masayoshi Mikami,
Valerio Olevano, Giovanni Onida, Jean-Yves Raty,
Lucia Reining, Gian-Marco Rignanese, Aline Roy,
Luc Sindic, Jay Sullivan, Yasuko Suzukawa
Stefan Tausendpfund, Marc Torrent, Matthieu Verstraete
Gilles Zerah.

I think 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 3.2, released on October 18, 2001.

Changes with respect to version 3.1 :

A. Changes for the installation.
B. Most noticeable changes in the package
C. WARNINGS.
D. Other changes.

__________________________________________________________________

A.  Changes for the installation.

A1. Note that the directories Src_defs and Src_main have been created.

    The Src_main directory contains the sources of all the main
    programs (abinit.f, cut3d.f, ...), that are no more
    in the ~abinit top directory.
    The files basis_defs.f and psp_defs.f have been placed
    in the Src_defs directory, and their name changed to defs_basis.f
    and defs_psps.f . Other defs file have been placed there.
    See Infos/dirs_and_files

A2. The files makefile_macros have been changed a lot.
    All the makefile_macros files provided in the package have been
    updated (and most must still be tried on the proper platform !)
    See the help.makefile_macros file for the up-to-date description of
    these files (in order to adapt your makefile_macros file,
    if it is not one of those provided in the package).

    (1) Default values have been provided for most macros in the
    makefile_macros files.
    (2) The meaning of the BLAS and LAPACK macros has changed,
    while BLAS_A and LAPACK_A have disappeared.
    (3) The BLAS library has been placed on the same
    footing as the other libraries (no need of "make blas"
    before "make allseq")

    Also, it is now possible to give tuned compiler options for the
    specific directories Src_2ffts, Src2_nonlocal and Src_4wfs.


A3. New makefile_macros files have been created, noticeably
    for clusters of Intel/Linux and DecAlpha/Linux machines :
    ~abinit/Machine_dept_files/Intel_P6/makefile_macros.PGI_MPICH
    and
    ~abinit/Machine_dept_files/Intel_P6/makefile_macros.DecLinux.tux ,
    as well as for the use of the new Intel compiler (ifc) for Linux
    (see http://www.intel.com/software/products/compilers/f50/linux),
    ~abinit/Machine_dept_files/Intel_P6/makefile_macros.Intel .

    The file
    ~abinit/Machine_dept_files/Intel_P6/makefile_macros.PGI_paral
    has been moved to
    ~abinit/Machine_dept_files/Intel_P6/makefile_macros.PGI_Lam .

    ABINIT has been modified for compilation under the 3.2-3
    release of the PGI compiler, and some makefile_macros
    are based on that release, noticeably
    ~abinit/Machine_dept_files/Intel_P6/makefile_macros.PGI_prof

    A makefile_macros for MacOS X has been created, but more tests
    must still be done to release this Mac version of ABINIT.

    (JMB + XG)

__________________________________________________________________

B.  Most noticeable changes.

B1. Calculations of the band structure in the GW approximation are now
    possible. This is based on a program originally started
    by R.W. Godby, developped over the years by V. Olevano, G. Onida,
    L. Reining, M. Torrent, and incorporated in ABINIT
    by G.-M. Rignanese. The adequate output of ABINIT
    has been coded by M. Torrent, and modified by V. Olevano.

    For the time being, the documentation is
    still to be written (and this will take time !).
    An example of GW calculation is given in Test_v3, tests 30 to 32.

B2. The recognition of the space group number (1 to 230)
    from the list of symmetry operations works now : it is
    tested extensively for all the 230 space groups
    in Test_v3, cases 15 to 19.
    (RC + XG)

B3. The response-function calculations work with nsppol=2
    (spin-polarized calculations). See Test_v3, cases 22 to 24.

B4. For antiferromagnetic materials, the CPU time can be divided
    by a factor of two, thanks to the use of Shubnikov symmetries.
    For ground-state calculation, use nsppol=1 with nspden=2
    (see the help file, as well as Test_v3/t20.in).
    For response-function calculations, Shubnikov symmetries
    are used to decrease the number of perturbations,
    with nsppol=1 and nspden=2 , see Test_v3/t24.in .
    The recognition of Shubnikov symmetries is implemented, as
    well as their generation, but the latter needs still some testing,
    and the corresponding input variables are still to be documented.
    (XG + RC)

B5. The format of the ifc input file has been completely changed.
    Keywords are used now. These are documented in ifc_help.htm .
    The old input files do NOT work anymore.
    Examples of the new input files can be found in Test_v2 :
    cases 13,15,16,17,18,19,20,21,22,24,25,28,29,32,39 .

B6. The implementation of non-collinear spin-magnetization
    calculations.
    The exchange-correlation part is computed by changing the
    direction of the magnetization, to make it directed
    along z (and always positive). This involves no approximation
    in the LDA case, while for the GGA, other implementations
    are possible. In particular, a small deviation from
    anti-ferromagnetism will not exactly reduce to the antiferromagnetic
    case when the deviation tends to zero. However, this effect
    is expected to be small.
    WARNING : Forces, stresses and response-functions
    are not yet coded, and symmetries cannot be used yet !
    See nsppol==4 and Test_v3/t27.in .
    (Contributed by Gilles Zerah).

B8. In the framework of the self-documentation project, the Infos/Theory
    directory has been created. At present, it contains three documents
    that present the basic theory underlying the code, and that
    concerns (I) the definition of the cell and atoms geometry
    (and symmetries), (II) the definitions, storage, and
    conversion of one wavefunction,
    (III) the definitions, storage, and conversion of a
    set of wavefunctions. In the future, these documents should
    be linked to the sources. Investigation continue on how
    to make this in the framework of ROBODOC and TtH.
    (Contributed by MMikami, Yasuko Suzukawa , and XG)

__________________________________________________________________

C.  WARNINGS

C1. For developpers : there has been big changes due to the
    introduction of structured (derived) datatypes, and
    the associated modules defs_common.f, ...
    For example, the number of arguments of gstate.f has changed
    from 90 to 25 !
    It is strongly adviced to the developpers to carefully read
    the defs_*.f files contained in the subdirectory Src_defs .

C2. For developpers : the name of the module
    basis_defs.f has been changed to defs_basis.f . This module
    is used in all ABINIT routines.

C3. For developpers : the unit number associated with the main abinit
    output, previously "iout" is now defined
    in the module defs_basis.f , and no more
    in the abinit.f file.  It is called ab_out .
    Since the defs_basis module is used in all abinit routines,
    one can suppress the "iout" argument in many routines
    (but this is still to be done for most routines).
    The unit number associated with the temporary files, previously
    defined in abinit.f as "unit00" is now also defined in the module
    defs_basis.f, with the name tmp_unit . Many routines have left
    their "unit00" argument.

C4. For developpers : the routine rhohxc.f has been moved to
    rhohxc_coll.f , due to the implementation of the non-collinear
    magnetism. A new rhohxc.f is the interface towards rhohxc_coll.f
    in non-spin-polarized and collinear magnetization cases.
    See B6.

__________________________________________________________________

D.  Other changes (or current developments, not yet finalized).


D0. In the case nsppol=2, the magnetic moment of the unit cell
    can be constrained to a predefined value. See the input
    variable fixmom. (Contributed by Jay Sullivan).

D1. In the input file, the sign "!" is now considered as the
    beginning of a comment, similarly to "#".
    See abinis_help.html .

D2. The analysis of interatomic force constants has been speeded
    by a large factor (the sorting algorithm scaling was N2, it is
    now N Log N). For example the test 39, that was 2 minutes
    on a PIII at 450 MHz is now a dozen of seconds.

D3. A new scheme allowing the computation of electronic structure
    at finite temperature, with special k point smearing,
    has been devised by M. Verstraete. See the input variables
    tphysel and tsmear.

D4. Oscillator strengths are now available in TDDFT. It is
    possible also to compute the dynamic polarisability (for
    which a small utility program exist,
    see ~abinit/Utilities/dynamic_pol.f . This should work,
    but is considered to be in development (more tests are needed).
    See iscf=-1 .
    Contributed by JYRaty .

D5. It is possible to determine the inner eigenvalues of the
    energy spectrum,
    thanks to the use of the conjugate gradient algorithm
    on the operator (H-eshift)^2. See the input variables wfoptalg
    and eshift. Not very efficient yet, so to be considered
    under development.
    Contributed by CAudouze, GZerah and XG.

D6. A bug in the computation of the maximum residual has been
    removed : nbdbuf was not properly taken into account.

D7. Introduction of kssform input variable, for further developments
    related to GW, by Valerio Olevano

D8. Change int to nint in cut3d, by MMikami and MVerstraete.

D9. The Makefile and makemake scripts have been extensively rewritten,
    and simplified.

D10. The arguments of many routines have been changed, in preparation
     of the PAW implementation. (contribution by FJollet and MTorrent)

D11. A script for the preparation of input files for testing the
     pseudopotentials of the whole periodic table is in preparation,
     see Test_physics/Solid.pl (contribution by RCaracas).

D12. A package (in development) towards the use of autoconf
     has been contributed by STausendpfund. It is in the directory Configure.