~nickpapior/siesta/trunk-buds-format0.92

____________________________________________________________

** CHANGES ALONG Javier Junquera's Wannier branch

------------------------------------------------------------

------------------------------------------------------------

2013-12-04   Javier Junquera trunk-438--radfunc-5--wannier-4

The parallelization of the interface with Wannier90 has finished.

This version runs the test ranging from number 1 to number 12 distributed

with the Wannier90 suite, and the results are perfectly compatible with

those obtained using QuantumEspresso

added:

  Src/reordpsi.F

modified:

  Src/Makefile

  Src/alloc.F90

  Src/amn.F90

  Src/compute_pw_matrix.F90

  Src/m_digest_nnkp.F90

  Src/m_noccbands.f

  Src/m_overkkneig.F90

  Src/m_trialorbitalclass.f90

  Src/mmn.F90

  Src/mneighb.f

  Src/siesta2wannier90.F90

  Src/siesta_analysis.F

  Src/siesta_options.F90

  Src/write_inp_wannier.F90

------------------------------------------------------------

2013-08-02   Alberto Garcia  trunk-438--radfunc-5--wannier-3

Move wannier analysis to 'siesta_analysis'. Reinstate call to siesta_end

A previous patch had removed the call to siesta_end in the main program.

It has been re-instated. 

The call to siesta2wannier90 has been moved to 'siesta_analysis'. In

case the user requests a local_DOS calculation (which destroys the scf

DM), the new call is placed just before that section.

modified:

  Src/Makefile

  Src/siesta.F

  Src/siesta_analysis.F

------------------------------------------------------------

2013-08-01   Javier Junquera  trunk-438--radfunc-5--wannier-2

Complete the SIESTA-WANNIER90 interface

* Use the new function registry.

* Added a new test

  removed:

    Src/initswan.F90

    Src/parallelsubswan.F90

    Src/trialOrbitalClass.f90

  added:

    Src/m_overkkneig.F90

    Src/write_inp_wannier.F90

    Tests/wannier/

    Tests/wannier/kremik.fdf

    Tests/wannier/kremik.nnkp

  modified:

    Src/Makefile

    Src/alloc.F90

    Src/amn.F90

    Src/compute_pw_matrix.F90

    Src/m_trialorbitalclass.f90

    Src/matel_registry.F90

    Src/mmn.F90

    Src/register_rfs.F90

    Src/siesta2wannier90.F90

------------------------------------------------------------

2013-07-09 14:30 GMT Alberto Garcia trunk-438--radfunc-5--wannier-1

Merge with trunk-register-5 to pull in registry for new matel

(A. Garcia)

This merge pulled in also the trunk changes up to 438 (see CHANGES).

See Docs/radfunc.CHANGES for details about the registry.

NOTE: the functionality of the module "trialorbitalClass" is duplicated, but

the two versions are not really compatible. One of them should be chosen once

and for all.

------------------------------------------------------------

2013-07-09 14:10 GMT Javier Junquera    trunk-434--wannier-1

Implementation of the interface between Siesta and Wannier90

(Javier Junquera)

Commited by A. Garcia (See description below)

added:

  Src/amn.F90

  Src/broadcast_projections.F

  Src/compute_pw_matrix.F90

  Src/delk.F

  Src/initswan.F90

  Src/m_digest_nnkp.F90

  Src/m_noccbands.f

  Src/m_planewavematrix.F90

  Src/m_planewavematrixvar.F90

  Src/m_trialorbitalclass.f90

  Src/m_writedelk.F

  Src/mmn.F90

  Src/parallelsubswan.F90

  Src/siesta2wannier90.F90

modified:

  Src/Makefile

  Src/atom.F

  Src/dhscf.F

  Src/diagpol.f

  Src/grdsam.F

  Src/ksv.f

  Src/local_DOS.F

  Src/m_mpi_utils.F

  Src/mesh.F

  Src/meshdscf.F

  Src/meshsubs.F

  Src/savepsi.F

  Src/setup_hamiltonian.F

  Src/siesta.F

  Src/siesta_analysis.F

  Src/siesta_options.F90

-----------------------------------------------------------------------

Description of the changes

-----------------------------------------------------------------------

--------------------------------

    In m_planewavematrix.F90

--------------------------------

New module that contains all the ingredients to compute the matrix elements

of a plane wave in the grid.

--------------------------------

    In dhscf.F:

--------------------------------

      New argument in the call to dhscf:

C integer isigneikr     : Switch which fixes whether the matrix elements

C                         of exp(i k r) are calculated or not.

C                          0 = not computed.

C                          1 = compute the matrix elements of exp(i k r).

C                         -1 = compute the matrix elements of exp(-i k r).

      New call to the subroutine delk, where the matrix elements

 of a plane wave are computed in real space.

--------------------------------

    In grdsam.F:

--------------------------------

     The calls to dhscf have been modified adding the

 new variable isigneikr in the list of arguments.

     Since the matrix elements of the plane wave are not required

 during the normal self-consistent-cicle, 

 isigneikr is always set up to zero in grdsam.F

--------------------------------

    In local_DOS.F:

--------------------------------

     The calls to dhscf have been modified adding the

 new variable isigneikr in the list of arguments.

     Since the matrix elements of the plane wave are not required

 to compute the local DOS, 

 isigneikr is set up to zero in local_DOS.F

--------------------------------

    In setup_hamiltonian.F

--------------------------------

     New variable:

     integer:: isigneikr ! Calculate matrix element of a plane wave exp(ikr)

                       !  0 => no,

                       !  1 => calculate exp(ikr)

                       ! -1 => calculate exp(-ikr)

     The calls to dhscf have been modified adding the

 new variable iexpikr in the list of arguments.

     Since the matrix elements of the plane wave are not required

 during the normal self-consistent-cicle, 

 isigneikr is set up to zero in setup_hamiltonian.F

--------------------------------

    In siesta_analysis.F

--------------------------------

     The call to dhscf have been modified adding the

 new variable isigneikr in the list of arguments.

     Since the matrix elements of the plane wave are not required

 for the analysis, 

 isigneikr is set up to zero in siesta_analysis.F

--------------------------------

    In mesh.F:

--------------------------------

    A new variable, iatfold, is defined in the module mesh.

 This is an integer variable of dimension 3 * number of atoms in the supercell.

 It is defined and computed in InitAtomMesh (subroutine meshsubs.F, see below) 

 independently of whether the finite field

 capability is used or not. The computational cost in CPU time and memory

 is negligeable, though.

    In the subroutine setupExtMesh, the variable nem is deleted

 from the list of internal variables and introduced as a module variable.

 It is defined in module mesh.

--------------------------------

    In meshsubs.F:

--------------------------------

    A new variable, iatfold, is defined in the module mesh.

 This is an integer variable of dimension 3 * number of atoms in the supercell.

 It is defined and computed independently of whether the finite field

 capability is used or not. The computational cost in CPU time and memory

 is negligeable, though.

    The variable is allocated and computed in the subroutine InitAtomMesh

 This variable stores the supercell vector required to fold the position

 of an atom back into the unit cell. 

--------------------------------

    In meshdscf.F

--------------------------------

    New subroutine matrixMtoOC.

 This is a copy of matrixMtoO, but modified to handle complex matrices.

--------------------------------

    In m_mpi_utils.F

--------------------------------

    New subruoutine broadcast_vvv_int added.

 This is required to broadcast the array nnfolding readed from the .nnkp file

--------------------------------

    In delk.f

--------------------------------

      New subroutine where the matrix elements of a plane wave are

 computed in real space.  

      The matrix elements of a plane wave in a basis of numerical

 atomic orbitals are computed in the grid, instead of in

 the former approach implemented by Daniel Sanchez-Portal,

 where a Taylor expansion of the plane wave was made assuming

 small values of the exponent:

      exp(i k r) \sim 1 + i k r

      The reason for this change is the fact that, within the new

 algorithm, there is no reason to assume that k (or more precisely

 the vector between consecutive k-points \Delta k)

 would be small enough to justify the cut of the Taylor expansion.

 Remember that, in the algorithm described by I. Souza et al.,

 the larger the number of k-points (that is, the smaller the

 interval between neighboring points) the smaller the value of the

 threshold energy for Zener tunneling

 [see the end of the first paragraph in the

 left column of the third page of the paper by

 I. Souza et al., Phys. Rev. Lett. 89, 117602 (2002)].

      Parallelization issues:

 If vmat is parallelized, delk should work as well, since the 

 basic loops are maintained without any change.