lp:~nickpapior/siesta/4.1-ncdf-md

Branch merges

Propose for merging

No branches dependent on this one.

Merged into lp:siesta/4.1 at revision 939

Alberto Garcia: Approve on 2018-07-06

Diff: 1187 lines (+425/-261) (has conflicts)

20 files modified

Docs/siesta.tex (+13/-5)
Src/Makefile (+8/-8)
Src/final_H_f_stress.F (+18/-5)
Src/m_ncdf_siesta.F90 (+205/-114)
Src/read_options.F90 (+16/-2)
Src/save_density_matrix.F (+13/-9)
Src/siesta_forces.F (+7/-27)
Src/state_init.F (+53/-3)
Util/COOP/Makefile (+8/-8)
Util/Denchar/Src/Makefile (+8/-8)
Util/Gen-basis/Makefile (+8/-8)
Util/Grimme/Makefile (+8/-8)
Util/Helpers/Makefile (+8/-8)
Util/STM/ol-stm/Src/Makefile (+8/-8)
Util/SpPivot/Makefile (+8/-8)
Util/TS/TBtrans/Makefile (+8/-8)
Util/TS/ts2ts/Makefile (+8/-8)
Util/TS/tshs2tshs/Makefile (+8/-8)
Util/VCA/Makefile (+8/-8)
version.info (+4/-0)

api

Related source package recipes

No recipes using this branch. (?)

Create packaging recipe

Related snap packages

Related bugs

Related blueprints

940. By Nick Papior on 2018-07-09

Added pjnl_j to the ncdf basis

939. By Nick Papior on 2018-07-09

Fixed FinalTimeStep option reads and indentation

Thanks Alberto

938. By Nick Papior on 2018-07-04

Fixed NCDF write-out for MD and FC calculations

Previously the NCDF file-write-out would crash when performing
MD/FC calculations because the sparse arrays does not really make sense
to store.

Now the file content is based on the type of run, single/MD/FC.
A relaxation is equivalent to an MD (currently).

One non-CDF related change is the change of MD.NumCGSteps to
MD.Steps keyword. The NumCGSteps is deprecated since many more
options have entered and CG was not necessarily the correct instance.

Secondly, the MD.Steps keyword is now also the default for
MD.FinalTimeStep which means that a single keyword determines the number
of steps runned.

The old flags are still used as the default values.

937. By Nick Papior on 2018-07-03

Changed the weighting for the charges to be based on the normalized weights

936. By Nick Papior on 2018-06-28

Fixed a few remaining issues with electrode options and the xa_EPS for elec-DM-copy

Now all options specified for tbtrans *must* be located at the end.

All electrode replicate flags have been removed (use Bloch instead).

Fixed a coordinate check variable to ensure we don't copy too many DM elements
from the electrode.

935. By Nick Papior on 2018-06-26

Clean-up of DM.Init.Bulk (still undocumented feature)

Now one can fine-tune the details using Tile/Repeat
and also the atomic coordinate check may now be decided
by the user.

934. By Alberto Garcia on 2018-06-26

Enable auxcell in diag*g and pdos*g. Clarify 'gamma' meaning. NaiveAuxCell

* Enable real-arithmetic operation with an auxiliary cell

The diag*g and pdos*g routines are now able to work transparently even
if an auxiliary supercell is used (either with explicit use of a
passed indxuo array or by means of an equivalent MODP helper function
in 'intrinsic_missing' (a mod with a [1,p] target domain, instead of
[0,p-1])).

* Clarify the logic flow of the 'gamma'-related variables

The 'm_gamma' module has been removed. 'Gamma' itself is reserved to
signal an actual gamma-point-only operation in relevant routines. It
is passed as an argument, with appropriate values imported from the
k-point sampling modules.

When needed, the use of an auxiliary supercell is detected on the
basis of the equality or not of 'no_s' and 'no_u'.

The HSX files now output information in 'auxiliary supercell' mode
(former 'gamma' .false.) even if the supercell is trivially the unit
cell. They always contain the 'indxuo' array.

Code in Util/COOP has been changed to avoid a spurious check on
equivalence of the two 'gamma's (one indicating whether wavefunction
info is complex or real, and the other the presence of a supercell).

It is not possible (or meaningful...) to run a PEXSI calculation with
a forced auxcell.

Some obsolete checks on indxuo have been removed.

* Re-instate 'naive-auxiliary-cell' option

The auxiliary supercell needed can be determined using the atomic
positions, the orbital extent and the lattice parameters. For a fixed
geometry configuration the tightest auxiliary supercell is uniquely
defined. However, when performing MD simulations (or relaxations) the
displacement of certain atoms may result in changes to the required
auxiliary supercell, thus inhibiting re-use of DM information across
geometry steps. When this is option is true the auxiliary supercell
will be computed with a simpler algorithm using only the orbital
extent and the lattice vectors. The resulting supercell might be
larger than strictly needed, but it will remain mostly constant during
the run (even when the MD.VariableCell option is used, if the lattice
changes are moderate).

By default this will be false when performing single-point
calculations (MD.TypeOfRun:CG with MD.NumCGsteps:1 or
below). Otherwise this will default to true.

(Thanks to Nick Papior for several ideas)

933. By Nick Papior on 2018-06-19

Corrected EDM setting when using electrode bulk DM

Also added a flag TS.Hartree.Offset which specifies a
global offset that will be added to the Hartree potential.

Also enabled TS.Elecs.DM.Init force-bulk to also do it
in V /= 0 bias calculations.

932. By Nick Papior on 2018-06-18

Fixed electrode names and TBT.*nc output

Now electrodes may not be named Device/Buffer or
having an "+" in the name.

Secondly, the atomic indices of the electrode regions are now stored
in the nc file-outputs.

This enables sisl to attach group-names.

931. By Nick Papior on 2018-06-15

Limited integer field size in ofc