Dear All:
You can already find in the Download area of your
SIESTA account the Release Candidate number 2
of SIESTA-3.0.
Some important fixes have been put in place so we
suggest for you to update if you are using siesta-3.0-beta
or siesta-3.0-rc1.
Find below the release notes, that can be also found
in Src/Docs and at the same download area.
Yours,
Jose A. on behalf of the SIESTA Team
--
Jose A. Torres,
SIESTA Manager
PS: release notes:
===========================================
Release Notes for SIESTA-3.0-rc2, July 5, 2010
GENERAL NOTES
Since the release of Siesta-3.0-rc1 in April 2010, we have fixed
a few bugs and made other improvements (see below). This version,
tagged siesta-3.0-rc2, (which stands for "3.0 release-candidate number 2"),
is one step closer to the "stable" siesta-3.0 release.
Users of versions 11.X of the Intel fortran compiler should note that
we have been unable to find robust workarounds for the bugs in that
compiler. Until the bugs are fixed we recommend users to stick to
version 10.X of the compiler.
CHANGES SINCE SIESTA-3.0-rc1
(A complete list of changes can be found in the Docs/CHANGES file.)
SUMMARY
* Compute and print band-structure energies
* Make the HSX format the default for H,S files
* New hook to output the initial charge density
* Fixes and new functionality in Pulay mixing
* Clarify the meaning of the supercell options in the Vibra manual
* Fixes and improvements in the optical spectra code
* Updates and fixes to the TranSiesta/tbtrans
* Fix bug in re-positioning of MD file
* Use larger output fields for atm diagnostic files
BUG FIXES AND IMPROVEMENTS
* Fixes and improvements in the optical spectra code
* Fixed a bug in optical.F: the call to subroutine kgridinit nwas
causing problems and providing results that were not compatible any
longer with those of subroutine kgrid.
* Modified call to subroutine transition_rate: wmin and wmax passed to
subroutine so that only the relevant transition probabilities are
explicitly calculated.
* Updates and fixes to the TranSiesta/tbtrans code
This patch includes bug fixes (in TranSiesta and tbtrans), a cleanup
involving removal of dead code, proper declarations of variables, and
a new prototype test framework for TranSiesta and tbtrans.
* The new_dm module now has a cleaner Transiesta related structure.
The possibility of doing a "continuation run" (where a .TSDE file is
read and the run starts with a TS calculated density matrix), is
handled by calling the routine read_dynamic_ts_dm. Since Transiesta
also stores and reads Escf, the variable EDM_tmp has been created and
is used in the same way as D_tmp. The routine change_sparsity is also
used for Escf. These changes fix a bug introduced in 3.0-rc1, but they
have not been heavily tested yet in all circumstances.
* The original version of TBTrans (by M. Brandbyge) was not
parallelized. Some of the original properties computed by tbtrans
(the transmission function and the pdos) have been "ported" to work
also in the parallelized version of the code, but some have not. Those
who have not, have been, for now, deleted (atompdos, coop and NetCDF
related things).
* The transmission function is now multiplied by two in the
case of non spin-polarized calculations. As a consequence, the
computed current is multiplied too.
* The printed value of the pdos (averaged over all the k-points),
when running in parallel with more than one node, was not correct:
only the values coming from the node=0 were being printed. This has
been fixed.
* Some smaller changes were done on the way the results are written
in the output files (e.g. including blank lines in between values
for each k-point, so that the file can be directly plotted with
xmgrace). Also changed the name of the routine "Tsiokp"
(m_tbt_kpoints.F90) to "get_kp_on_node" since it represents better
what it really does.
* Fixes and new functionality in Pulay mixing
* The "kick" functionality intended to "shake" periodically the
Pulay mixing procedure has been re-implemented, with the added
feature that the Pulay history is completely reset after a kick.
* Additionally, support has been added to specify whether the first
density-matrix residual of the SCF cycle and the first residual after
a kick are included in the Pulay history. It can be argued that in
these cases the "output" DM might be significantly different from the
"input" DM. To preserve backwards compatibility, these residuals are
kept in the Pulay history unless the fdf variable
DM.Pulay.Avoid.First.After.Kick
is set to "true" in the input file.
* The above changes have forced a temporary disabling of the "Pulay on
file" feature, which will be re-implemented in a form that works
well in parallel.
* Fix bug in repositioning of MD file
The MD file repositioning code in iomd.f gave problems in some
systems. It has been removed, and substituted by an explicit
'position="append"' clause in the open statement.
* Use larger output fields for atm diagnostic files
The atm program generates, among others, the files FOURIER_QMAX and
ECONF_DIFFS, which contain information useful for optimization
runs. The output fields were quite tight and sometimes
overflowed. They have beeen made larger.
NEW FEATURES
* Make the HSX format the default for H,S files
The old HS format was highly inefficient, as each piece of data was
written to its own record. In addition, it only contained information
about the inter-orbital vectors Xij if k-points were used in the
calculation.
The new HSX file format was introduced at the time of the first
COOP/COHP implementation. It is more tightly packed, and the real
arrays (H, S, and Xij) are written in single-precision.
The HSX format is now the default in Siesta (TranSiesta has its own
format).
The programs hsx2hs and hs2hsx in Util/HSX are provided to facilitate
the transition. Note however that an HS file written by Siesta during
a Gamma-only calculation will not contain information about the
inter-orbital vectors Xij, and hs2hsx will refuse to create an HSX
file from it.
* Hook to output the initial charge density
If the fdf variable SaveInitialChargeDensity is "true", the program
generates a RHO_INIT (and a RhoInit.grid.nc file if netCDF is compiled
in) containing the charge density used to start the first
self-consistency step, and it stops. Note that if an initial density
matrix (DM file) is used, it is not normalized. This is useful to
generate the charge density associated to "partial" DMs.
* Created a test directory (ts_fast) to test TranSIESTA and TBTrans.
* Compute and print band-structure energy. Updates to Util/COOP/mprop
* The band-structure energy Ebs = Tr(DM*H) is now computed at every
iteration and printed at the beginning and end of the SCF cycle, both
in the output file and in the CML file. The calculation is done in
compute_dm, so DM is the "output DM" from the diagonalization of H.
* The mprop program has been updated to output the integrated
band-structure energy as a function of energy as a third column in
the integrated-dos file. The " -n " option has been documented.
* Creation of Util/COOP/Docs, and addition of some notes on the units
and meaning of the COOP and COHP curves.
COSMETIC CHANGES
* Clarify the meaning of the supercell options in the Vibra manual
New wording to make it clear that the supercell options refer
to the number of augmentation shells around the unit cell.
