lp:~albertog/siesta/efield-2.5-jjunquera
- Get this branch:
- bzr branch lp:~albertog/siesta/efield-2.5-jjunquera
Branch merges
Related source package recipes
Branch information
Recent revisions
- 244. By Alberto Garcia
-
Implementation of forces with bulk electric fields
This commit updates the branch with a snapshot of the
state of Javier Junquera's code as of Feb 25, 2010.A few cosmetic changes were needed for a simple compilation.
Note that the Examples directory has not been updated yet.The detailed documentation for the implementation is not
yet committed.added:
Docs/efield. CHANGES
Src/derdelk.f
Src/grdelk.f
modified:
Docs/CHANGES
Src/Makefile
Src/chempot.F
Src/constr.f
Src/delk.f
Src/dfscf.f
Src/dhscf.F
Src/diagef.F
Src/diagpol.f
Src/efield.F
Src/finitefield.F
Src/indexkgrid.F
Src/initdm.F
Src/iodm.F
Src/linpack.F
Src/m_dimefield. f90
Src/m_forces.F90
Src/m_overkkisig. f
Src/m_pkisigma.f
Src/m_psitilde.f
Src/meshmatrix.F
Src/new_dm.F
Src/overfsm.f
Src/setup_hamiltonian. F
Src/siesta.F
Src/siesta_efield. F
Src/siesta_forces. F
Src/siesta_init.F
Src/write_subs.F - 243. By Alberto Garcia
-
Implementation of finite electric field for periodic systems (J Junquera)
-------------- ------- ------- ----
In kgrid.F and find_kgrid.F
-------------- ------- ------- ---- In order to have the k-points ordered,
a vector of the reciprocal lattice has been subtracted to some of them.
The algorithm is esentially the same, with the exception of the following
lines:C Find total range of grid indexes
C Find total range of grid indexes
! JJ: If the origin of the Monkhorst-Pack mesh has been displaced,
! substract a vector of the reciprocal lattice to some of the points,
! in order to have an ordered mesh of k-points.
! Nothing is done if the origin of the mesh of k-point is undisplaced.
do j = 1,3
ng(j) = kdsc(j,j)
if (abs(dkg(j)) .lt. tiny) then
igmin(j) = -( (ng(j)-1) / 2)
igmax(j) = ng(j) / 2
else
igmin(j) = -( (ng(j)-1) / 2) - 1
igmax(j) = ng(j) / 2 - 1
endif
enddoIf the mesh of k-points is not displaced, nothing is done.
If the mesh of k-points is displaced, and we have even numbers
in the diagonal, the k-points are the same as before and only their order
in the list is altered.
If the mesh of k-points is displaced, and we have odd numbers
in the diagonal, some k-points are different from the ones obtained
with the original subroutine. Some of the components might have changed
their sign.-------
------- ------- ------- ----
In kpoint_grid.F90
-------------- ------- ------- ----
The variables kscell and kdispl have been declared public
(they are going to be passed to finiteef).-------
------- ------- ------- ----
In alloc.F90:
-------------- ------- ------- ---- New subroutines:
- realloc_z3
- realloc_z4
- dealloc_z3
- dealloc_z4Complex versions of the previous subroutines to reallocate or deallocate
complex arrays of 3 and 4 dimensions.The complex pointers are supposed to be in every case
of double precision type.realloc_z3 and realloc_z4 are copies of the previous
realloc_d3 and realloc_d4 where the definition of the array has been changed
from double precision to complex.dealloc_z3 and dealloc_z4 are copies of the previous
dealloc_z1 where the dimension of the array to be deallocated
has been increased to 3 and 4 respectively.
from double precision to complex.-------
------- ------- ------- ----
In dhscf.F:
-------------- ------- ------- ---- New argument in the call to dhscf:
C integer iexpikr : 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 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 siesta_options.F90
-------------- ------- ------- ---- The default value for MD.TypeOfRun has been changed from Verlet to CG.
Two new variables:
! logical :: bulkfinitefieldflag ! Determine whether there is a finite electric
! ! field in a periodic system or not.
! logical :: infinitefield ! Determines if the control of the program is
! ! taken by the module to compute the
! ! finite electric field or not.-------
------- ------- ------- ----
In state_analysis.F
-------------- ------- ------- ---- The KS energy is written only if the calculation is at zero
external electric field in bulk.Under an external electric field, the electric enthalpy is
written instead.-------
------- ------- ------- ----
In local_DOS.F:
-------------- ------- ------- ---- 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
to compute the local DOS,
iexpikr is set up to zero in local_DOS-------
------- ------- ------- ----
In setup_hamiltonian.F
-------------- ------- ------- ---- New variable:
integer:: iexpikr ! 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,
iexpikr is set up to zero in setup_hamiltonian.F-------
------- ------- ------- ----
In siesta_analysis.F
-------------- ------- ------- ---- The call 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
for the analysis,
iexpikr is set up to zero in siesta_analysis.F-------
------- ------- ------- ----
In grdsam.F:
-------------- ------- ------- ---- 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,
iexpikr is always set up to zero in grdsam.F-------
------- ------- ------- ----
In m_dimefield.f90:
-------------- ------- ------- ---- Module where all the relevant variables,
including:- Derived types for the list of k-point neighbours
are defined.-------
------- ------- ------- ----
In indexkgrid.F
-------------- ------- ------- ---- New subroutine that initializes some lists that allow us
to relate the index of a given k-point with the indices
of neighbouring k-points along the three directions of the
reciprocal lattice.Look for instance at Eq. (53) of the paper by
I. Souza et al., Phys. Rev. B 69, 085106 (2004),
where the overlap matrices between a given k-point and
the six first neighbours are required.The algorithm is inspired in the Appendix
of the technical paper of Siesta.
Basically:- We reintroduce again the points that were cutted due to the
time reversal symmetry (k = -k).
- Construct a supercell in reciprocal space.
- Look for neighboring points in the supercell.
- Once the neighbouring point is localized, go back to the normal mesh
of k-points.
- Localize the equivalent point due to time reversal symmetry.Paralleli
zation issues:
The subroutine is called only once, and it takes no more than one second
to compute the list of k-point neighbours.
Therefore, no parallelization is required.The only place where parallelization might play a role
is in the broadcasting of the derived types
dictionaryneighkafter and dictionaryneigh kbefore to all the nodes. -------
------- ------- ------- ----
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)].Paralleli
zation issues:
If vmat is parallelized, delk should work as well, since the
basic loops are maintained without any change.-------
------- ------- ------- ----
In efield.F
-------------- ------- ------- ---- New variables:
! logical bulkfinitefield
flag: switch that determines whether we are going to
! perform a calculation of the bulk system
! under the presence of an external electric field.Behaviour:
Read the electric field and generate the corresponding potential
in the grid if and only if we are not dealing with periodic systems.If we are dealing with an open system
(molecule, cluster, wire, tube, surface)
and we have a field normal to the periodic direction, then
the potential corresponding to the field is introduced as a ramp with
the discontinuity in the vacuum, in the usual way.-------
------- ------- ------- ----
In m_energies.F90
-------------- ------- ------- ---- New variable: DUextbulk (interaction energy with external electric field
in a bulk material) -------
------- ------- ------- ----
In post_scf_work.F
-------------- ------- ------- ---- New term added to the sum of the total energy:
Etot = E0 + DEna + DUscf + DUext + Exc + Ecorrec + Emad + Emm +
. Emeta + DUextbulk-------
------- ------- ------- ----
In scfconvergence_test.F
-------------- ------- ------- ---- New term added to the sum of the total energy:
Etot = E0 + DEna + DUscf + DUext + Exc + Ecorrec + Emad + Emm +
. Emeta + DUextbulk-------
------- ------- ------- ----
In siesta_init.F
-------------- ------- ------- ---- The variable DUextbulk is initialized
DUextbulk = 0.0_dp
-------
------- ------- ------- ----
In write_subs.F
-------------- ------- ------- ---- The variable DUextbulk is written in the same way as DUext was
written before.Besides, the writing of the decomposition of the energy,
forces, and stresses has been changed. Now, the "prefix" of everyline
(as it was now "siesta :") depends on whether the calculation
is at zero electric field (in this case "siesta :") is maintained,
or at a finite electric field (in this case, the prefix is "finief:").The Harris energy is only written under zero electric field in bulk.
-------
------- ------- ------- ----
m_noccbands.f
-------------- ------- ------- ---- New module (extracted from Daniel's subroutine ksv.f)
that computes the number of occupied bands.-------
------- ------- ------- ----
diagef.F
-------------- ------- ------- ---- New subroutine of diagonalization.
All the matrices are defined as complex arrays.
The matrix that is diagonalized is Tk and not the KS hamiltonian.
It builds the density and energy density matrices,
but now the loop in k-points is outside the subroutine
inside finitefield).-------
------- ------- ------- ----
m_overkisig.f
-------------- ------- ------- ---- File that contains the subroutine that computes the overlap
between the periodic part of the Bloch-like resonances in neighboring
k-points, following the definition given right before
Eq. (53) of the paper by I. Souza et al. Phys. Rev. B 69, 085106 (2004).To adapt this to Siesta, we have followed Eq. (5) of the paper
by D. Sanchez-Portal et al., Fundamental Physics of Ferroelectrics,
535, 111, 2000.-------
------- ------- ------- ----
m_psitilde.f
-------------- ------- ------- ---- Compute the dual basis of the Bloch-like resonances,
vtilde [see Eq. (53) of the paper by
Ivo Souza et al. Phys. Rev. B 69, 085106, (2004) ]-------
------- ------- ------- ----
m_pkisigma.f
-------------- ------- ------- ---- Compute the projectors defined in Eq. (61) of the paper by
Ivo Souza et al. Phys. Rev. B 69, 085106, (2004)-------
------- ------- ------- ----
siesta_efield.F
-------------- ------- ------- ---- Interface between siesta.F and the module that computes
the response of a dielectric to a finite electric field. - 241. By Alberto Garcia
-
Proper fix of cell management for zmatrix case. Enthalpy cosmetics
A recent previous patch broke the unit cell handling in the zmatrix case.
It has been fixed.The assorted "enthalpy" output has been rationalized,
reserving the use of the word "enthalpy" for the original case
of "target enthalpy".(Fixed syntax error in std-test.mk)
- 240. By Alberto Garcia
-
Make diagonalization the default solution method regardless of size
Make diagonalization the default solution method regardless of system size. - 239. By Alberto Garcia
-
Fix test.mk scripts
The std-test.mk script did not copy the output file to the main test directory.
The bsc-test.mk did not remove all the scratch files.Fixed a typo in the input to the si2x1h test.
- 238. By Alberto Garcia
-
Fix for structural supercell construction
Due to the accidental mis-use of a module variable, the routine coor
did not generate the supercell correctly when a %supercell block was used.Added and additional warning if such a block is used when the structure is
input in Zmatrix form. - 237. By Alberto Garcia
-
Fix memory leak in dhscf
The array dvxcdn was allocated but never deallocated in dhscf. Actually,
it was never used in cellxc, so it has now been dimensioned trivially
to (1,1,1). The bug was introduced in the bsc-master-2.1 branch, at
patch 27. - 236. By Alberto Garcia
-
New 'fdf' target in Makefile for tests.
(E. Anglada)New target for the test option: fdf calls.
Requested by BSC so they can test the new fdf.New files (new tests and missing xml output files for reference)
A/ Tests/var_cell_stress
A Tests/Reference-xml/fire_ benzene. xml
A Tests/Reference-xml/si_ coop.xml
A Tests/var_cell_stress/ .arch-ids/ =id
A Tests/var_cell_stress/ .arch-ids/ makefile. id
A Tests/var_cell_stress/ .arch-ids/ var_cell_ stress. fdf.id
A Tests/var_cell_stress/ .arch-ids/ var_cell_ stress. pseudos. id
A Tests/var_cell_stress/ makefile
A Tests/var_cell_stress/ var_cell_ stress. fdf
A Tests/var_cell_stress/ var_cell_ stress. pseudos Modified files
M Src/siesta_move.F : remove debug info
M Src/remove_intramol_ pressure. f90 : remove debug info
M Tests/std-Makefile Xmlcheck enabled by default. New targets: fdf, all
M Tests/std-test.mk
M Tests/si2x1h/si2x1h. fdf New option
M Src/atom.f Remove debug info.(replayed patch 6 from outlier branch ref-2.4)
- 235. By Alberto Garcia
-
xml tester update (Eduardo Anglada)
New files:
A Src/xmlparser/corresponding_ node.f90
A Src/xmlparser/string_ utilities. f90 Changes in the xml output:
M Src/state_analysis. F
M Src/siesta_move.F
M Src/setup_hamiltonian. F
M Src/siesta_forces.F
M Src/scfconvergence_test. F
M Src/post_scf_work.FChanges in the xml tester:
M Src/wxml/m_wxml_ core.f90
M Src/wxml/flib_wstml. f90
M Src/xmlparser/makefile
M Src/xmlparser/m_dom_element. f90
M Src/xmlparser/m_strings. f90
M Src/xmlparser/compare_ m.f90
M Src/xmlparser/test.f90 Changes in the tests makefiles
M Tests/std-Makefile
M Tests/std-test.mkUpgraded xml reference outputs:
...(replayed patch-5 from outlier branch ref-2.4)
Branch metadata
- Branch format:
- Branch format 7
- Repository format:
- Bazaar repository format 2a (needs bzr 1.16 or later)
- Stacked on:
- lp:siesta
