Highlights

• DFT & semi-empirical (extended Hückel) models for electronic structure and transport calculations
• New electrostatic model, allowing for inclusion of gates (and more!)
• Improved algorithms on many fronts, e.g. the basic two-probe model, geometry optimization, etc.
• Performance improvements
• New features, like bulk DOS and LDA+U (see the new LDA+U tutorial!)
• A new, modern graphical user interface
• See more details below

The currently released package is a beta-version of ATK 10.8, which will replace both ATK 2008.10 and the preview versions 2009.xx and 2010.02.. The final release is expected in the beginning of July, barring unforeseen complications. The information below applies to the final release, but some of the features might still be a bit rough around the edges in this beta-version. There are known bugs in this package; these will be fixed in the final release, by which time we will also provide more detailed information on the updated features.

Please report any issues you encounter, preferably via the Forum.

To download the packages, go to the download page.

License

All users will need a new license since we have changed license system to LM-X from X-Formation. Please contact support, if you have not already received one by email. Or, if you are a new user, please register to obtain a trial license.

Updated and new features

Backengine (ATK)

In the ATK back-engine we note, among other things, the following major enhancements compared to 2008.10:

• Semi-empirical extended Hückel model and DFT combined in the same package, for electronic structure and transport calculations
  
• New electrostatic model
  • Include metallic gate electrodes (arbitrary position and shape) and dielectric regions, for realistic simulations of transistor-like structures
  • Ambient dielectric constant, to simulate a solvent
  • Charged systems, enabling calculations of single-electron transistors
• Improved algorithms
  • Most consistent two-probe model ever!
  • Double-contour integration, stabilize finite bias simulations (allows for a higher bias to be applied)
  • More accurate and flexible current integration model (compute current from transmission spectrum)
  • Possibility to do study thermionics by introducing a different temperature in the two leads for the current calculation
  • Use of electronic free energy instead of total energy, as appropriate in an open system
• Geometry optimization
  • More accurate forces
  • Considerably better relaxation algorithms (ATK 10.8 uses the ASE algorithms)
  • True relaxation under bias
• Performance
  • Faster self-energy calculations via new methods
  • Lower memory usage and faster convergence for many systems
  • Better utilization of multi-core computers (more parts threaded)
• Other new features
  • DOS
    • bulk DOS
    • optimized using unit cell symmetries
    • projections (on atoms, orbitals coming soon), bulk + two-probe
  • Transmission spectrum
    • kpoint-resolved data automatically stored for later analysis
    • optimized using unit cell symmetries
  • LDA+U (tutorial)
  • Non-selfconsistent calculations for quick-and-dirty parameter checks (like k-point convergence)
  • Proper band structure calculations via symmetry points
• Miscellaneous
  
  • New license system
  • Simplified installation procedure, graphical installer also for Linux
  • The external code GPAW has been wrapped inside the ATK package, plus there is access to ASE
  • Unified, open file format (no more cryptic VNL file, just NC files)
  • Richer Python interface (more queries, less hidden variables and data)
  • Explicit boundary conditions
  • Change of naming of executable to "atkpython"
  • and more...

Graphical user-interface (VNL)

One of the immediately most visible updates is the graphical user interface, which has been completely redesigned. The workflow has been adjusted, and there is generally more functionality in the interface, to allow users to be more productive. The list of detailed new features is too long to publish - you will have to discover them as you work with the program. If there is something you can't find, ask us (preferably on the Forum), and we will consider if for future releases.

Some of the major new things in VNL 10.8 are:

• Builder
  • Unified interface for molecules, bulk and devices
  • Mirror, translate, copy/paste operations
• Plug-in modules
  
  • "Custom builder" for users to create their own builders
    • Several predefined ones are provided (this is e.g. where you now find the Nanotube Builder!)
  • "Custom analyzer" for specific analysis tasks, like I-V curves
• 3D viewer
  • improved graphical performance
  • change individual color and radius of individual atoms
  • orthographic projection
• Expanded databases for molecules and crystals
  • Added database for fullerenes
• Built-in file browser for easy access to calculations and scripts
• Miscellaneous
  • Proper band structure plots
  • Improved and extended import/export (CAR, CIF, XYZ, Cube)
    • VASP geometries can be imported and exported via a supplemental script
    • Export of raw data for band structure, electron density, and other 3D grids

Backwards compatibility

There are a couple of things to be aware of regarding this release.

• There is no compatibility with old NC or VNL files (even NC files from 2010.02).
• The Python syntax has changed, so those who are used to writing their own scripts are advised to read the manual carefully.

It is possible to import geometries (in Python files) from older versions either directly, or via a few simple intermediate steps.

• Molecules and bulk systems: as long as they are saved directly from VNL, Python scripts defining molecules and bulk systems (also PeriodicAtomConfigurations) can be directly dropped on the instuments in VNL 10.8.
• Two-probe systems: In many cases you can drop them directly, but if they contain repetitions in the electrodes or if they are constructed completely in the Atomic Manipulator (AM), you should re-open them in the AM and export the equivalent bulk system. Then open this system in the VNL 10.8 Builder and convert it to a device geometry. In fact, this work-flow is generally recommended, since it represents the new way to think of device geometries: make the central part, then specify which part is the electrode.

Platform support

To the best of our knowledge this release runs on pretty much all recent Windows versions and Linux distributions (sometimes specific libraries might be needed).

If you have a specific issue, please post it on the Forum (search first, it might already be solved!).

Known issues

The following are known issues in this beta-release:

• The Script Generator has some minor quirks; this tool is not 100% complete.
• LDA+U
  
  • is only correct for spin-polarized systems, and has
  • some issues when using "double-zeta" basis sets.
• GGA will be part of the final release, but is missing from the beta-version
• The Script Generator forgets to save the optimized geometry in the NC file, you should add this manually, by inserting a line (adjust to match exactly the name of the NC file and the variable used for the configuration, e.g. bulk_configuration if it's a bulk system):

             nlsave("analysis.nc", configuration)

• In addition the manual for OptimizeGeometry forgets to mention that the optimized geometry is returned, while the original configuration, passed to it, remains unchanged
  • So, what you need, in the simplest case, is

             optimized_geometry = OptimizeGeometry(initial_geometry)
             nlsave("file.nc", optimized_geometry)

• The interactive command-line environment doesn't always work properly (on Windows).
• Rhombohedral crystals in hexagonal setting are not imported correctly from CIF files.
• Missing documentation in the manual; also, the tutorials are not fully updated to match the new interface.
• Minor issues that might appear in certain situations, but which usually are not noticeable in the standard workflow.