lp:~albertog/siesta/psml-chlocal

Branch merges

Propose for merging

No branches dependent on this one.

Merged into lp:siesta/psml-support at revision 549

Javier Junquera: Approve on 2017-02-21

Diff: 7877 lines (+6555/-916)

31 files modified

Pseudo/vnl-operator/psop.f90 (+4/-1)
Src/Makefile (+4/-3)
Src/atom.F (+315/-75)
Src/hamann.f90 (+245/-0)
Src/ncps/src/m_ncps_reader.f (+2/-0)
Src/psoplib/src/flib_spline.f90 (+0/-257)
Src/psoplib/src/m_interpol.F90 (+0/-189)
Src/psoplib/src/m_localgen.f (+7/-304)
Src/psoplib/src/m_psop.f90 (+1/-1)
Src/psoplib/src/makefile (+1/-2)
Tests/Pseudos/O.tm2-siesta-vnl.psml (+3212/-0)
Tests/Pseudos/O.tm2.psml (+2677/-0)
Tests/Pseudos/Si.soft-siesta-vnl.psml (+4/-3)
Tests/Pseudos/Si.tm2.nrl.vnl.psml (+4/-3)
Tests/fe-psml/fe-psml.pseudos (+1/-1)
Tests/fe-psml/makefile (+1/-1)
Tests/h2o-psml/h2o-psml.fdf (+24/-0)
Tests/h2o-psml/h2o-psml.pseudos (+1/-0)
Tests/h2o-psml/makefile (+6/-0)
Tests/si64-psml-full/makefile (+1/-1)
Tests/si64-psml-full/si64-psml-full.pseudos (+1/-1)
Tests/si64-psml-sl/makefile (+1/-1)
Tests/si64-psml-sl/si64-psml-sl.pseudos (+1/-1)
Tests/si64-soft-psml-full/makefile (+1/-1)
Tests/si64-soft-psml-full/si64-soft-psml-full.pseudos (+1/-2)
Tests/si64-soft/makefile (+1/-1)
Tests/si64-soft/si64-soft.pseudos (+1/-1)
Tests/test.mk (+5/-3)
Tests/test.psml.mk (+0/-46)
Tests/test_tddft.mk (+29/-17)
version.info (+4/-1)

api

Related source package recipes

No recipes using this branch. (?)

Create packaging recipe

Related snap packages

Related bugs

Related blueprints

558. By Alberto Garcia on 2017-02-14

Use consistent epsilon for red_vlocal cutoff

Use the same 1.0e-5 value as in the "vlocal_as_fit" routines.

557. By Alberto Garcia on 2017-02-14

Use 'vlocal tail' cutoff by default for foreign codes

For consistency with the 'siesta-fit' cases, when
constructing chlocal from foreign pseudos use the
'vlocal tail' criterion for the chlocal cutoff.

( Search from the end towards r=0, to make it more
robust)

556. By Alberto Garcia on 2017-02-13

Add vlocal_from_chlocal file output

555. By Alberto Garcia on 2017-02-06

Streamlined pseudo file handling in test.mk

When an item in the name.pseudos file contains the
extension .psf or .psml, it is taken directly as a file name
in Tests/Pseudos. Otherwise, for backward compatibility, the
.psf extension is added.

Updated test files for PSML runs and added 'h2o-psml'

554. By Alberto Garcia on 2017-02-06

Create chlocal_from_vlocal to simplify logic. Update psml files

Streamlined the logic in get_psml_vlocal_chlocal.
Increase psop version to 1.0.1
Print uuid of the PSML file at reading time in Siesta
Update psml files from psop.
Add O psml files (pending update of test framework to support
mixed psf and psml species)

553. By Alberto Garcia on 2017-02-05

Use annotation for chlocal cutoff. Clean psoplib

Psop now generates an annotation for the local-potential
element which stores the chlocal cutoff.

Unpacking the annotation is the preferred method to extract
chlocal from a psop-generated PSML file.

The "chlocal-from-vlocal" functionality has been removed
from psoplib.

552. By Alberto Garcia on 2017-02-02

Refine heuristics for parameters

551. By Alberto Garcia on 2017-02-02

New heuristics. Mostly works, with tighter parameters

For Si, a good combination of parameters is:

      psml.chlocal.build.exp_al 1.012
      psml.chlocal.build.rr1 0.001
      psml.chlocal.build.r0 0.004
      psml.chlocal.build.delta 0.005
      psml.chlocal.build.poly_order_rlin 4

although there could be some improvement.

modified:
  Src/Makefile
  Src/atom.F
  Src/hamann.f90

550. By Alberto Garcia on 2017-02-02

New (faulty) version with finite-diffs in atom's grid.

For some reason, there is an upward (but smooth!) tilt in chlocal at
small r.
One more commit comes...

modified:
  Src/atom.F
  Src/hamann.f90

549. By Alberto Garcia on 2017-02-01

New method to determine chlocal

It is based in the calculation of the laplacian of Vlocal by finite
differences, using a routine by Don Hamann, and subsequent direct
interpolation into the linear tables used by Siesta with another
Hamann routine.

The weak point is still the determination of the chlocal cutoff. The
use of a Hamann-style grid forces a change in the order in which the
selection of the final radial point and the normalization is carried
out. For full compatibility with Siesta a chlocal computed directly
on the Froyen grid will be needed (next commit).

added:
  Src/hamann.f90
modified:
  Src/Makefile
  Src/atom.F