Siesta 4.1-b3

Release Notes for Siesta-4.1-b3 (July 2017)

This version is the third beta release of the 4.1 series.
If you are testing the 4.1-beta releases you are highly
encouraged to update to this version.
It fixes and extends a number of issues from the 4.1-b2 release:

* Fixed <>.nc file for non-collinear and spin-orbit coupling

* Added user-control of phases to the dH matrix elements in tbtrans

* Added installation scripts for hdf5+NetCDF and flook

* Added tests for the Lua interface

* Fixed orbital current calculation (lp:1672084)

* Now transiesta/tbtrans stops if the transfer matrix is zero,
  i.e. if the electrodes are molecules.

* Lots of updates to the Lua interface, more variables and more control
  - Added all energies
  - Enabled changing mixers and mixing weights etc.

* Fixed tbtrans polarized calculations (if TBT.Spin not set) (lp:1658896)

* Fixed transiesta Hartree plane for 1-electrode calculations

* Fixed bugs related to certain combinations of mixing routines

* Fixed tbtrans pivoting for electrodes (lp:1649774)

* Enabled contraction/extrapolation of spin configurations (lp:1645388)

* Fix for printout in XML format (lp:1647073)

* Better standard compliance in code structure

* Fix bug related to SlabDipoleCorrection which couldn't be turned off (lp:1630827)

* Fix for 'nodes' basis generation option (lp:1625725)

* Fixes VCA mixing of pseudos (lp:1633039)

* Fixes integer energy specifications in ProjectedDensityOfStates block (lp:1657584)

* Added print-outs when GridCellSampling is used

* Fix for Bader charge analysis (lp:1656273)

* Fix memory problem when memory usage close to limit (lp:1665294)

* Forced Diag.ParallelOverK to false if non-collinear spin configuration (lp:1666428)

* Updated Eig2DOS to be more like gnubands (options are the same)

* Added Geometry.Must.Converge flag to ensure the relaxation has converged

* Enabled internal walltime check to forcefully stop SIESTA after a given time

* Updates and fixes for STM simulation program in Util/STM/ol-stm
 - Fix wrong fftw call
 - Enabled reading WFSX files directly

* Added new interpolation option for Util/Macroave

Here are the remaining release notes for the 4.1 series:

*** Backward-compatibility issues:

    * The mixing routines have completely changed, hence the same
    convergence path cannot be expected. This, unfortunately, makes
    comparison difficult with prior versions. However, the final
    converged system should be transferable.

    * SIESTA now defaults to mixing the Hamiltonian instead of the
    density matrix. To revert to density-matrix mixing, use
    "SCF.Mix DM". The option to mix after the initial scf step is now
    'on' by default.

    * SIESTA now defaults to monitoring convergence for both the
    density matrix AND the Hamiltonian. To revert to only density
    matrix convergence, use: "SCF.H.Converge false"

    * A major number of fdf-flags concerning mixing
    parameters have changed to a more consistent naming scheme.
    However, all previous flags are still in effect but the newer
    flags have precedence. The previous flags are the default values
    for the newer flag-names.

    * Two additional files are created:
       H_DMGEN and
       H_MIXED
      these contain the Hamiltonian at various stages through the SCF.
      Currently they are intended for developers and may be removed in
      the final 4.1 release.
      You may delete these without problems.

*** New features

-- LDA+U (Javier Junquera)

    * Full incorporation of the LDA+U implementation in SIESTA
    * Two different LDA+U projectors available
    * Estimate the best U according to: Cococcioni and Gironcoli in PRB, 2005

-- Spin-Orbit coupling (Ramon Cuadrado)

    * On-site approximation for spin-orbit-coupling matrix elements.

-- MRRR method for diagonalization (Alberto Garcia)

    * This will typically be faster than divide-and-conquer algorithm
    and may be the future default. For Gamma-point calculations.

-- ELPA method for diagonalization (Alberto Garcia)

    * This provides better scalability compared to ScaLAPACK for large
    # of processors. For Gamma-point calculations.

-- Added interface to the PEXSI library for calculating the density
   matrix, DOS, and LDOS (Alberto Garcia)

    * This library provides massive parallelism and better
    scalability, but should only be used for very large systems.

-- SIESTA is now hybrid-parallelised (Nick R. Papior)

    * One may compile Siesta/Transiesta in serial, OpenMP, MPI, or
    MPI+OpenMP mode.

-- Re-write of non-equilibrium Green function code (Nick R. Papior)

    * N>=1 terminal support in transiesta
    * Improved convergence
    * Different ways of handling charge-reductions in SCF
    * All electrodes have settings _per electrode_ for full customization
    * Greatly reduced memory usage
    * Skewed axis are enabled for further reduction of complex systems
    * Implemented MUMPS method for inversion
    * Implemented LAPACK for inversion
    * Implemented BTD method for inversion (extremely fast)
    * Fully OpenMP threaded
    * Start directly from transiesta enabled
    * Temperature gradients as non-equilibrium a possibility

-- Complete rewrite of tbtrans utility (Nick R. Papior)

    * Made tbtrans a stand-alone utility for user-defined tight-binding method
    * EXTREME SCALE version (BTD-only)
      - Memory consumption _only_ dependent on "device" region
    * N>=1 electrode support
    * Region projections for transmission through "eigenstates"
    * Custom change of the self-energies
    * k -> k' transmissions
    * Interpolation of bias Hamiltonians
    * Bond-currents
    * Fully OpenMP threaded and/or MPI parallelized
    * DOS and bulk transmission for electrodes
    * Gf-DOS/spectral-DOS for device region

-- Mixing routines rewritten (Nick R. Papior)

    * New mixing schemes Pulay (Guarenteed Reduction)
    * Custom mixing restart options (full user customizability)

-- Added more constraints (Nick R. Papior)

    * Constraints are verbose and many new ways of using constraints exists

-- NetCDF4 file format for siesta -> parallel IO (Nick R. Papior)

    * Provides a standard intrinsically grouped file for retaining
    nearly all siesta related information in one file.

-- Enabled convergence control of density-, energy density matrices,
   Hamiltonian and energy.

-- LUA scripting in siesta (Nick R. Papior)

    * This is an experimental feature

-- Gate calculations (Nick R. Papior)

    * Charge and Hartree gate

-- Utilities (Nick R. Papior)

    * All make-files are now prepared to enable parallel builds
      - this makes compilation *MUCH* faster. For example:
         make -j4
        will compile using 4 cores.
    * Grimme utility
      - easy creation of FDF block for Grimme parameters
    * SpPivot, pivot sparsity pattern using several different routines
    * TS/** different utilities related to transiesta