Copyright (C) 2000 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 March 2000 and July 2000:
Jean-Michel Beuken, Francois Detraux, Razvan Caracas, Alessio Filipetti,
Martin Fuchs, Francois Jollet, Bernard Magne, Masayoshi Mikami,
Gian-Marco Rignanese, Jean-Yves Raty,
Aline Roy, Luc Sindic, Massimiliano Stengel, 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 2.3, released on July 31, 2000.

Changes with respect to version 2.2 :

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

__________________________________________________________________

A.  Changes for the installation.

A1. (nothing for this version)

__________________________________________________________________

B.  Most noticeable changes.

B1. The abinis_help has been split into two parts (a list of
    input variables, and the explanation part), and translated
    to HTML, with the creation of adequate hyperlinks.
    It is available from the Web site, as well as from the
    usual ~abinit/doc/users directory (see ~abinit/doc/users/abinis_help.html
    and ~abinit/doc/input_variables/keyhr.html).
    Also, the tutorial is now available under HTML, with links
    to the above-mentioned files
    (see ~abinit/doc/tutorial/welcome.html).
    This work has been done by Razvan Caracas.

B2. The spin-orbit coupling has been coded.
    Presently, the total energy, the density and the electronic
    structure should be OK. Forces and stresses are not yet working.
    There are new input variables : "nspinor", "nspden", "pspso".
    Also, "spinat" has been modified, as the magnetization
    becomes a vector.
    See tests 74 and 75 of Test_v2 .
    This work has been done by G. Zerah, F. Jollet and X. Gonze.

B3. ABINIT has been placed under CVS, locally, as well
    as on the SourceForge Web site (http://www.sourceforge.net).
    Developpers might have a look
    at  ~abinit/doc/developers/collab .

B4. The determination of symmetries has been improved :
    the Bravais lattice and the point group are automatically
    recognized and written ; the non-primitive cells
    can be automatically reduced to primitive cells
    (see option -1 to the input variable "brvltt").
    Also, the k point generator is better :
    nkpt can be automatically computed ;
    other grids than Monkhorst-Pack one are available
    (see the input variable "kptrlatt");
    the best k point grid can be automatically selected
    (see the input variable "prtkpt").
    In the process of doing this, the meaning of the
    input variable "kptpot" had to be changed, see C5.

B5. Introduction of Nose-Hoover thermostat (ionmov=8),
    by J.-Y. Raty. Related input parameters :
    "noseinert", "noseit", "noseft".

B6. An additional parallelisation for SMP machines,
    using OpenMP, has been coded by A. Roy.
    Note, however, that it should NOT yet be used, since
    it has not been optimized, and many things
    are still to be be understood in order to provide speed-up.
    FFTs are parallelized on one of the dimensions ;
    non-local pseudopotential part of the Hamiltonian is parallelized
    over plane waves, as are orthogonalisation operations.

__________________________________________________________________

C.  WARNINGS

C1. The new input variable "nbdbuf" (see D17) has been introduced.
    With its default value, it modifies the stopping criterion
    of non-self-consistent calculations and metallic response-function
    calculations, compared to version 2.2.
    In order to recover the old behaviour in these cases, use
    nbdbuf 0

C2. The input variable "spinat" is now a vector (see B2), instead
    of a scalar in version 2.2.
    In order to recover the old behaviour in this case, use
    vectors having the old slcalar value for the z-component.

C3. The header of wavefunction, density and potential files
    has additional variables : pspso, nspden, nspinor, fermie .
    Also, the body of the wavefunction files now contains the
    occupation numbers.
    ABINITv2.2 and before are not able to read the new files.
    ABINITv2.3 is able to read all files.
    Cut3d from the web site does NOT work with the new header.
    A new version of cut3d is in preparation.

C4. The mean of the potential has been redefined, to give agreement
    with the Janak's theorem. Eigenenergies are changed with respect
    to the previous version. You can recover the old values
    by hand, by subtracting the mean of the XC potential.

C5. There has been large modifications of the routines
    that preprocess the input file.
    Previously, inprep.f called indims.f, and inmain.f
    called invars.f .
    Now, there is first invars0.f, then invars1m.f that
    calls invars1.f, then invars2m.f that calls invars2.f
    (the 'm' stands for 'multi').

C6. The meaning of the input variable "kptopt" has changed :
    a band structure calculation needs a negative value,
    and initialization of the input variables "nkdiv" and "kptbounds".
    The input variable "ngkpt" has dimension
    ngkpt(1:3) instead of ngkpt(1:4).
    The input variable "nshiftk" must be used instead of ngkpt(4)
    Additional WARNING : ABINIT is unable to see whether
    you have modified your old files to the new setting ! It will
    default nshiftk to 1 , and so use only one shift, even
    if you are listing 2 or 4 shifts.


__________________________________________________________________

D.  Other changes.

D1. The parallel version on a Cray T3E has been considerably
    speeded up by A. Roy

D2. The susceptibility and dielectric matrix can now been computed
    for imaginary frequencies, by M. Fuchs
    New input variables : "ikhxc", "nfreqsus", "freqsusin", "freqsuslo"

D3. A tentative makefile_macros file for parallel version of SP2
    has been communicated by B. Magne .

D4. Bug corrected : the use of XF history is now correct
    in the parallel version, by M. Stengel.

D5. GW development is going on, by G.-M. Rignanese.
    See the new input variable "optdriver".

D6. ABINIT has been installed on NEC, by G.-M. Rignanese.
    Not yet optimized. See ~abinit/Machine_dept_files/NEC .

D7. The MPI parallel version is available for HP S-Class and N-Class
    machines, by L. Sindic.

D8. Additional new input variables, related to timing and automatic
    exit : "chkexit", "timopt".

D9. Ifc code : for FCC lattices, it is now possible to sample the
    q Brillouin Zone with the Gamma point
    only (using a 1x1x1 BCC sampling), see Test_v2#39.

D10. Additional miscellaneous informations, about
     visualisation of densities (VASP Data Viewer, XCrysDen),
     and phonon modes (PLOTMTV), by M. Mikami, see ~abinit/Infos/Miscellaneous

D11. For response-function calculations, the f-sum rule is now
     correct for partial occupation numbers

D12. Test_paral : tests #5 and #6 created, test #2 modified.

D13. Test_cpu : increased the run time, to improve statistics. Also,
     suppresses ngfft 16 and 18.

D14. The time analysis has been improved, and is now OK for RF

D15. A database of atomic masses has been created.
     See the input variable "amu".

D16. The self-documentation project is now documented (!),
     in ~abinit/doc/developers/rules_coding , section 10.e .

D17. The input variable "nbdbuf" has been introduced.
     It defines a buffer of wavefunctions, to stabilize calculations
     of metallic response-function, and to get rid off
     spurious effects in the non-self-consistent optimization
     of wavefunctions.