Copyright (C) 2004 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 2004 and September 2004:
F. Bottin, J. Bouchet, R. Caracas, C. Darrigan, M. Delaveau, T. Deutsch, F. Detraux,
P. Ghosh, D. Hamann, J. Iniguez, J. Janssen, F. Jollet, R. Meyer, Mikami-san, Y.-M. Niquet,
V. Olevano, Y. Pouillon, A. Roy, S. Sharma, M. Torrent, D. Vanderbilt, M. Veithen,
M. Verstraete.
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 4.4, released on August 28, 2004.
Changes with respect to version 4.3 :
A. WARNINGS.
B. Most noticeable achievements (for users)
C. Most noticeable changes in the package (for developers)
D. Other changes (or on-going developments, not finalized)
__________________________________________________________________
A. WARNINGS
A1. Switch to the 2002 adjustment of fundamental constants from the NIST,
replacing the 1998 values. See Src_defs/defs_basis.f .
Changes are at the 1.0d-6 relative level.
A2. In the GW calculation, the Vxc that is used in v4.4 is no more the
relativistic version of the Ceperley-Alder functional (that was actually
present only in this part of ABINIT).
See D6 for more information. Contributed by YM Niquet.
The choice of XC functional
in this part of ABINIT is now governed by the input variable ixc.
The default value is ixc=1 , that gives a different value of Vxc than
in v4.3 or earlier versions. Thus, the results of a GW calculation
will be slightly different in v4.4 than in v4.3 (on the order of
0.01 ... 0.02 eV for absolute value of eigenenergies, and much
less -about 1 meV- for the gap values).
A3. The smoothing function for the kinetic energy of plane waves
has been changed. See D9 for more information. Contributed by DHamann.
Now, its second derivative is continuous at the
cut-off energy. This change affect all calculations in which
the input variable ecutsm is non-zero. In this case, note that
the present version gives slightly different results than
previously. Moreover, one should not use
ground-state wavefunction files produced with version v4.3 or older
to start response function calculations with v4.4 with non-zero ecutsm.
This has been forbidden in the hdr_io.f routine.
A4. The PAW input variable pawmqgrdg has been suppressed.
A5. The name of tar files has been changed.
The main tar is no more src_tests.x.y.z.tar.gz but has become
abinit-x.y.z.tar.gz . Similar changes have been performed
with other tar files (binaries, etc ... ).
In particular, all "underline"
before version number have been replaced by a "dash".
__________________________________________________________________
B. Most noticeable achievements
B1. The computation of linear and non-linear optics has been implemented,
by S. Sharma. See "optic" utility.
Still to be documented, although Test_v4#57 provides an example.
The interface of "optic" with ABINIT has been implemented by M. Verstraete.
References :
1. S. Sharma, J. K. Dewhurst and C. Ambrosch-Draxl, Phys. Rev. B {\bf 67} 165332 2003
2. J. L. P. Hughes and J. E. Sipe, Phys. Rev. B {\bf 53} 10 751 1996
3. S. Sharma and C. Ambrosch-Draxl, Physica Scripta T 109 2004
4. J. E. Sipe and Ed. Ghahramani, Phys. Rev. B {\bf 48} 11 705 1993
B2. The electron-phonon coupling is now treated : there is a new utility "mrggkk",
and new capabilities of anaddb (input variable "elphflag" - however, still
to be documented), developed by Matthieu Verstraete.
This new capability of ABINIT is available for beta-testing.
In case you are interested, please contact mverstra@pcpm.ucl.ac.be .
Only EXPERT users, strongly motivated, should contact him.
No automatic test is yet available.
B3. PAW : GGA is implemented (M. Torrent, F. Jollet and F. Bottin (CEA-Bruyeres-le-Chatel))
Other modifications related to PAW : note the new variables
pawmixtyp, pawsphmix, pawxcdev, pawnzlm, pawvlbox, ngfftdg, bxctmindg .
The documentation of the PAW input variables has been improved.
B4. The development of the features related to strain perturbation continues :
spin-polarized response function calculations for the strain perturbation
are available (by D. Hamann)
See Test_v4 #79.
Remember, there are still limitations
in the use of this strain perturbation : LDA only, no spin-orbit coupling.
B5. MPI parallelization over k-points of the non-linear responses computations
(by M. Veithen).
B6. Finite electric field calculations : MPI parallelism over k points,
and use of symmetry operations, by M. Veithen and J. Iniguez.
See Test_v4#55 (among others).
__________________________________________________________________
C. Most noticeable changes for the developers
C1. The GNU-Arch repository is now effective, and has been used to merge
the present version 4.4 (set up by Y. Pouillon).
C2. Due to C1, change of names for .mod -> _mod
.so -> _so
Creation of many new files, to conform with GNU style
(THANKS, INSTALL, INSTALLME, NEWS, ...). However, the old
mechanisms are still used.
The temporary directories in which tests are executed have changed
name : the first character is now "," instead of "_" (e.g. ,test_20040724
instead of _test_20040724). This is because all files starting with
"," are treated as temporary by GNU-Arch by default).
C3. The use of 'intent' declarations in all routines is now the rule
rather than the exception ! This is the outcome of a global "beautify"
operation, lead between August 17 and August 20, with the following developers :
J. Bouchet, R. Caracas, T. Deutsch, F. Detraux, D. Hamann,
Mikami-san, M. Torrent, F. Jollet, Y. Pouillon,
M. Veithen, M. Verstraete, and my self. V. Olevano also contributed, later.
Also, during this "beautify" operation, the description of each routine
has been checked, as well as the description of all arguments.
This was quite heroic ! Lot of similarity with the initial porting
under ROBODOC, in December 2000, by a dozen of developers also ...
C4. Changed the name of all source files : switched from .f suffix to .F90 suffix .
(by Y. Pouillon)
__________________________________________________________________
D. Other changes
(or on-going developments, not yet finalized).
D1. New version of anaddb notes, by D. Vanderbilt
D2. Possibility to use a new pseudopotential format, from DHamann.
See ~abinit/doc/psp_infos/psp8.info
The format 8 for ABINIT pseudopotentials is designed to allow users
who wish to experiment with pseudopotentials, possibly with non-standard
features, to have great flexibility in doing so. It does not correspond
to any publicly available tabulation, or the output format of any
publicly available program, although codes such as FHI98PP (see
psp6.info) could easily be modified to produce this format. An extended
discussion follows below the annotated example.
D3. Band-by-band decomposition of the electronic dielectric tensor
(by M. Veithen). See Test_v3#77.
D4. FFT parallelism implementation continues : work by ARoy.
D5. Use of NetCDF for molecular dynamics trajectories. From MDelaveau.
See input variable "nctime".
Also, NetCDF for wavefunction, density, potential files. From MVerstraete.
Still to be documented.
D6. (See also the Warning A2)
This modification has been implemented by Y.-M. Niquet.
In the GW calculation, the Vxc that is used in v4.4 is no more the
relativistic version of the Ceperley-Alder functional (that was actually
present only in this part of ABINIT). The choice of XC functional
in this part of ABINIT is now governed by the input variable ixc.
The default value is ixc=1 , that gives a different value of Vxc than
in v4.3 or earlier versions.
In order to recover the older results, the value ixc=0 must be used
specifically in the GW calculation part. However, this practice
is discouraged, since none of the pseudopotentials has been
generated using this XC functional.
More information : the eigenenergy changes (with respect to DFT), in the usual
GW approximation (diagonal approximation) are computed using
Delta eig = < wavefunction | Sigma - Vxc | wavefunction > / Z .
The Vxc in this formula is specifically the one that is affected
by the new implementation.
D7. Computation of the kinetic energy contribution to the band gap energy
in the GW (ACFD-RPA) approximation. Implemented by YM Niquet.
See routine csigme.f and fkin.f .
D8. The opportunity to symmetrize the dynamical matrix, inside anaddb,
is now offered. See the new input variable for anaddb "symdynmat".
Thanks to this option, artificial symmetry breaking in case of
body-centered tetragonal lattices can be avoided.
D9. Change of smoothing function for the kinetic energy, governed by ecutsm.
Modification by Don Hamann.
Now, its second derivative is continuous at the
cut-off energy. This change affect all calculations in which
the input variable ecutsm is non-zero. See warning A3.
D10. New input variable, to activate use of Thomas-Fermi kinetic energy functional :
tfkinfunc.
D11. Performance improvements (sdirot.f, orthon.f and vtowfk.f routines) by Ralf Meyer.
D12. Possibility to print eigenenergies in a separate file, with suffix _EIG
See the input variable "prteig". Contributed by M. Verstraete
D13. The document "netiquette.html", by Mikami-san, describe the ABINIT netiquette.
D14. The conducti utility works now for the full conductivity tensor, not only its trace.
Contributed by P. Ghosh.
D15. The new ROBODOC version 4.0.24 has been used, thanks to Mikami-san
(who produced the robodoc.rc file).
D16. Bug fixes, and additional small modifications : C. Darrigan, D. Hamann, M. Verstraete, J. Janssen ...
D17. When iscf=-2, the eigenvalues for all the k points are printed anyway (like with prtvol>=2) ...
D18. In the input file, the sign '=' is now treated as a blank (a separator),
to avoid some beginner's errors,
especially that OCTOPUS uses = signs (so, '=' signs are now admitted as separators !).