Bugfix update VNL-ATK 2017.1

Sep 08 2017

A bugfix update to VNL-ATK 2017 has been released (version 2017.1). 

Download the update from http://quantumwise.com/products/download as usual. Your license for 2017.0 will work for this version too.

For more information about VNL-ATK 2017, see the original release letter.

2017.1 updates

This update contains a substantial amount of bug fixes, in particular several that fix various crashes. We are not aware of any bugs (solved or not) that would cause decidedly wrong results or errors in calculations (except very minor ones), but there were quite a few scenarios in which ATK 2017.0 would crash. The new 2D plot framework also received a touch-up in this release, and we spent some time solving minor annoying issues in VNL to improve the user experience.

(The number in parenthesis on each line is our internal tracking number.)

Important changes

  • Two-center integrals are now calculated on a much finer grid. This gives improved accuracy for force forces and stress, which improves relaxations, phonon frequencies etc. It does, however, also mean that you might get slightly different results for these quantities in 2017.1 compared to 2017.0 and earlier (22476)
  • New default PPS parameters for Ge, giving better band gaps (22745)
  • Syntax and other errors now appear in stdout also when running in parallel, as was the case always before (22690)
  • Warn, don't crash, if we fail to write a checkpoint file; also check if a checkpoint file can be written before the calculation starts (22107)

Added functionality

We normally don't add any new features in updates, but some things were really missing in 2017.0, or we had strong requests from customers.

  • Phonon bandstructure plot tool (existed in 2016, went missing in 2017.0) (22181)
  • Added option Supercell tool deletes overlapping atoms (22186)
  • Added the compound chemical formula to the last page of the Cleaver (20275)
  • Enabled tag selection for MDTrajectories in the Viewer (22593)
  • It is now possible to specify an "account" string in the PBS machine settings (22726)

Fixed crashes

  • VNL crashed (or at least does not display 3D graphics properly) showing GL_SCISSOR_TEST error messages (22177)
  • Could not save DOS in HDF5 files when using 'save_partial=False' (22250)
  • Spin-polarized LDA+1/2 was broken for devices/surfaces (22270)
  • Viewer isosurface: +/- isosurface option was broken (22333)
  • Using RandomSpin crashed when saving into HF5 file (22723)
  • PW segmentation fault for Sn (SG15) (21762)
  • Restarting PW calc with initial_state crashed when using MPI (22145)
  • Don't crash if PYTHONSTARTUP is defined (and fails); appears to be common on OpenSUSE (22360)

Wrong results corrected

Although these issues could give wrong results, they are in reality small or would rarely appear in a real calculation.

  • Noncollinear GGA and MGGA gave slightly incorrect density in parallel (22718)
  • Energies, forces and stress were wrong immediately after a DynamcialMatrix calculation using classical potentials (22061)
  • Using integer occupation numbers in Slater-Koster gave wrong occupations in the density matrix (using floats worked fine) (22133)
  • OptimizeGeometry with bonded force fields could return wrong connectivity (22163)

Corrected and improved functionality

  • Use caching to make the fat bandstructure plugin faster (22766)
  • Default integral_lower_point was not updated in the Script Generator when changing pseudopotential (in 2017, this parameter is automatically adjusted to match the lowest eigenvalues of the specific pseudopotential used) (22265)
  • Print forces smarter in the log file during relaxation, for easier grepping (grep for "OPT") (20027)
  • Tag selection in Viewer is now active (22593)
  • mpiexec messed up the license configuration when running in parallel on the local machine (22132)
  • Calculating the mobility with just a single q-point failed (22152)
  • IVCurve scripter: "Self-energy calculator" drop-down menu was not functional (22252)
  • Translate plugin stopped working for devices (22298)
  • "Phonon energy intervals" method did not work with LOE (22376)
  • Delete atoms in devices or surface could lead to index errors (22403, 22421)
  • NEB did not set initial_spin correctly after reading (21951)
  • Changing the fuzz factor would reset other graphical properties (21773)
  • Error in Kerker preconditioner scripts (22272)
  • Bad formatting of tags for 1000+ tagged atoms fixed (21878)
  • NEB with only a start and a final (no interior) image failed (21921)
  • PDOS crashed for molecules (22058)
  • Improved error message when an error occurs while opening licenses.licconf (22096)
  • Local machine in Job Manager: disabled threading options for serial jobs (22167)
  • Choosing an electrode with a negative score (but same length as the actual "best" electrode) would make an error message appear in the Electrode Size plugin (22266)
  • Isovalue indicator did not update for FermiSurface (22488)
  • Plot2D issues
    • Dropping a graph drag icon on the toolbar was allowed and failed (21864)
    • Incompatible dimensions error could appear when arrow is present in a 2D plot (22659)
    • Made it possible to also annotate a 2D plot outside plot area (22059)
    • Plot interaction failed in logarithmic scaling (22325)
    • Ensure plt files can be dropped on the Editor (22050)
    • Fixed an issue with editing item properties on Windows (22628)
    • Made plot titles more consistent (22681)
    • Inconsistent legend for 2D plot after editing fixed (22060)
    • Arrow moves outside plot area in some case (22062)
    • Arrow handle could not be removed (22063)
    • It was possible to remove lines in the band structure analyzer (22680)
    • Overlay windows disappeared when changing the style in PDOS many times (22683)
    • Could not resize the internal parts of the PDOS analyzer on Windows (21991)
    • DOS and Bandstructure plot now use the same label fonts (22003)
    • An undesirable behavior in Color Selection could not be fixed, it's a Qt bug but a work-around is available (22679)
    • Could not measure in combined BS/DOS plot (22066)
    • Error when cancelling the creating of a combined BS/DOS plot (22067)
    • Fix an error with "scroll locked axis" and zooming (22514)

Updates to the FHI-aims interface

  • Electron density object was incorrectly processed in VNL (22083)
  • "Internal" relaxation was wrongly scripted by Scripter (22262)

Updates to the VASP interface

  • Fixed an issue with ordering of atoms when exporting NEB for VASP, by only sorting the atoms by element when they are not already in contiguous order (21778)

Platform-specific issues

  • 2017 installer complains about icon generation at the end on some Linux distribution which are missing xdg-utils; the problem was harmless, but the installation appeared to fail (only the desktop icon would be missing) (22106)
  • On Windows, the installer appears to fail when MSVC 2017 present - notably, VNL itself was correctly installed, just the Visual C++ runtime was not (and was not needed, since it exists already) (22477)
  • The discontinued "chrome" cluster removed from the ATK On-Demand plugin (22388)

Documentation

  • PhotoCurrent module added (22259)
  • Description of how to use "metatext" in HDF5 files added (22537)
  • The working name OneShotDisplacement is properly replaced with SpecialThermalDisplacement (22258)
  • InGaN potential references was incomplete (22328)

 

Known remaining bugs

  • Slightly wrong total energy in a LCAO spin-orbit calculation with processes_per_kpoint > 1. However, this only affects non-converged results, specifiaclly if you just take 1 SCF step (22862)
  • NEB calculation crashes with preoptimization=True if you set processes_per_neb_image=None (workaround: optimize the end-points before creating the NEB path - generally a good idea anyway) (22868)
  • System fonts with incorrect permission crash matplotlib on startup. Not really something we are responsible for fixing... (workaround: make sure your fonts are all readable by all users) (22487)
  • Unicode support is still patchy, there are some areas that still don't accept unicode characters (22760 and a few others)
  • Back Engine Exception if NEB and HTST run in the same script (workaround: run them in separate scripts) (22698)
  • FHI-aims hangs in ELPA solver with MPI (we have a workaround, contact support if you need it) (22907)
  • Measure tool can be a bit jumpy in devices (22000)
  • TransmissionEigenvalues cannot be nlsaved in parallel (22902)
  • SLURM plugin does not correctly enable OpenMP threading (22899)
  • ... and minor stuff