Home Publications QuantumWise News ATK 12.2 released

Follow us

QuantumWise A/S on LinkedIn QuantumWise A/S on Google+ Subscribe to our YouTube channel QuantumWise A/S on YouKu

Meet us at...


QuantumWise Japan KK will attend The 38th Annual Conference on Magnetics in Japan September 2-5, in Keio University, Japan.


QuantumWise Japan KK and Anders Blom will attend the 2014 edition of SISPAD (Simulation of Semiconductor Processes and Devices) in Yokohama, Japan, September 9-11.


QuantumWise Japan KK invites for a seminar to introduce ATK 14.2. at M-school, Tokyo, Japan, September 12.


FermiTech will be representing Quantumwise at the Chinese Physical Society 2014 Fall Meeting in Harbin Institute of Technology from September 11th to 14th


QuantumWise Japan will attend the 75th JSAP Autum in Hookaido September 17-20.


FermiTech and Quantumwise will be holding an ATK Workshop 2014 at Fudan University from October 17th to 19th (ATK Workshop 2014). Click to sign up!


QuantumWise Japan will attend the 7th International Symposium on Surface Science in Matsue, Shimane, Japan, November 2-6.

ATK 12.2 released PDF Print E-mail
Monday, 23 April 2012 09:20

The long-awaited brand new ATK 12.2 is now available for download. The most important new feature: a brand new Builder, that you can use to easily build very advanced nanostructures.

Before downloading and installing the new version, please read carefully the note about new defaults, and the information about the new license system.


Download ATK 12.2 here

New features and other important changes


The primary new feature in ATK 12.2 is the new Builder. To summarize all its functionality requires more space than befits a release letter, so it has been collected in a separate article. A good resource is the new Builder Tutorials page.

Here are some highlights:

  • Select atoms in a variety like in a drawing program (use the mouse to draw a rectangle, cricle, lasso) or using mouse-commands (double-click to select all connected atoms, for instance), or using rule-based logic expressions (all carbon atoms with x-coordinate > 5 Ångström, etc).
  • Move and rotate fragments by mouse; snap-to functionality to align with or dock with other atoms
  • Easily passivatate dangling bonds with hydrogen
  • Full undo/redo
  • Merge different structures to form new systems (add a molecule to a crystal surface, make a nanotube junction, etc)
  • Build heterogeneous device configurations
  • Optimize hydrocarbon (graphene or nanotube) or Si systems with a single click (using the classical Brenner potential)
  • Extended import/output filters for external file formats
  • Advanced builder for NEB calculations - automatically generate images, then manipulate individual images or all images simultaneously, refine the path, remove/add images.
  • Surface Builder: cleave any crystal cell, regardless of symmetry, and then interactively specify the top layer, surface cell, directly turn it into a slab with vacuum, etc
  • The most advanced Interface Builder in the world? We like to think so - see it in action if you don't believe us!
  • Plugin-based interface (think apps for a smartphone): if you are missing some functionality - write a plugin for it, or download one from the addon server



The Builder in 12.2 also features a completely new high-performance 3D graphics engine, written from scratch in C++ and OpenGL.

In short, the aim is that you should be able to build pretty much any complicated nanostructure in the new Builder. You can then proceed to simulate it with ATK (using DFT or one of the many semi-empirical methods in ATK-SE), but there is also ample support for exporting the structures and use them with any other atomic-scale modeling code. And if you don't find an export filter for  your favorite code - write a plugin for it (or ask us nicely to do it for you).

We have created a getting started guide for the new Builder, which is also recommended for experienced VNL users, as many things have changed (incl. how to use mouse in the 3D views).


More new features

There are also new features in ATK itself, like new analysis functions:

  • Electron localization function (ELF) is a useful tool to analyse the electronic structure of both molecular and periodic structures (see e.g. http://en.wikipedia.org/wiki/Electron_localization_function). In ATK 12.2, this new feature can also be used for device configurations to provide further understanding of the transport mechanism.
  • Current density - visualize the current flow in a device, at a particular energy, as in the figure on the right, which is taken from our recent article on the interface betwee Ni and graphene.
  • Improved Halgren-Lipscomb method for pre-optimizing NEB paths. This can often reduce the number of ionic optimization steps by as much as 50%, compared to the conventional linear interpolation. Our approach builds on the original method by T. A. Halgren and W. N. Lipscomb (Chem. Phys. Lett., 49, 225 (1977)), but also features some additional improvements which will be presented separately.
  • The molecular dynamics (MD) and general ion dynamics functionality in ATK has been improved:
    • An MD simulation now returns an MDTrajectory object which can be queried for energies etc as a function of MD step.
    • There is also a possibility to set the initial velocities explicitly (from a previous run, for instance)
    • The performance of the force calculations has been optimized, and is now 2x faster in many cases, meaning that it constitutes less of a bottleneck in particular for small systems or when using semi-empirical methods, where the electronic structure itself is very fast to compute.
    • The FIRE optimization method in ASE has been exposed.
  • The computation of optical properties has been parallelized, and notably uses considerably less memory and produces a much smaller NetCDF file now. A few bugs in this feature were fixed as well.
  • New basis sets - more information will follow!


Special notes for users upgrading from ATK 11.8 or 11.2

The default tolerance of the self-consistent loop has been changed to 1e-4 in ATK 12.2, and we have also increased the default number of history steps to 15. Recent studies have shown that the convergence rate improves dramatically with more history steps. At the same time, physical quantities like total energy, forces, etc are normally well converged long before you reach the old default tolerance (4e-5), and thus this additional computation time provides no real benefit for the accuracy of the results.

The combined result of these two changes should therefore be that calculations now converge faster and better.

It may however be a good idea to explicitly set the tolerance to the old default value in your scripts, if you are in the middle of a series of calculations, so that the update to version 12.2 does not alter the results. This point is the ONLY place in the code that has been touched in a way that would influence the results of calculations. Thus, it is perfectly safe to upgrade to ATK 12.2, as long as this precaution is observed.

Moreover, the default for the transmission spectrum is now to use the DirectSelfEnergy method. This is significantly slower than the earlier default, the KrylovSelfEnergy method; however, the approximations made in the Krylov method can sometimes give the wrong results. We have therefore chosen to use a safer default, but if you compare the methods for a few test cases and find that Krylov works fine for your system, then you can gain a lot in performance by switching to this method.


Smaller changes and improvements

  • nlprint(bulk_configuration) now prints also fractional coordinates
  • BulkConfigurations are represented using fractional coordinates
  • EquivalentBulk now runs with default tolerance in non-self-consistent mode in DFT. This improves the accuracy of forces when optimizing device geometries by way of optimizing the central cell
  • Transmission spectrum from bulk (electrode-like) geometries is now more accurate (and hence may be a bit slower in some cases) by automatically repeating the structure in the Z direction as needed, to avoid erroneous results due to a too short system
  • Support for monoclinic structures has been added to the CIF import filter. CIF import is also less strict on formatting of symmetry operations (allow x+0.5, earlier it had to be x+1/2)
  • Minimal scripts generated by the Script Generator are now "more minimal" and thus easier to read
  • You can now query for all lattice parameters (a(), b(), c(), alpha(), beta() and gamma()) on all Bravais lattices
  • Appending data to NetCDF files now possible, resulting in much smaller I/O overhead for e.g. trajectories
  • A new explicit function to calculate the Fermi energy
  • Device relaxation reuses the electrode calculators
  • Prettier unit cells in generated script (0 instead of 1e-32 when it should be)
  • Improvements in the installer, license server tool separated out as a standalone package
  • Description of built-in Slater-Koster basis sets added to the manual
  • Improved convergence for semi-empirical calculations


Bug fixes

A few bugs have been found in ATK 11.8 and fixed  in 12.2:

  • Optical spectrum, corrected for symmetry and evaluation twice
  • Force and coordinate report in optimization runs is more consistent with the internal state of the calculation
  • Keyword bands_above_fermi_level was broken, fixed
  • CUBE files are now exported without the non-standard header which could confuse some programs that try to import the file
  • Improvements in symmetry recognition to avoid unnecessary error messages for "unusual" cells
  • The unit of LDOS was wrong in several places
  • Corrected the unit cell in the relaxation report for devices
  • Base-centered orthorhombic lattices had a wrong conversion matrix to conventional cell
  • A bug in the mapping of k-points has been fixed. This could give wrong pictures when plotting the k-points resolved transmission coefficients.
  • A handful other minor bugs


Known bugs

There are no known bugs in the backengine, i.e. the actual calculations performed by ATK.

A few minor issues in VNL and the logic part of ATK (work flow) are known and will be addressed very soon.

  • Selecting to optimize the end-point images in a NEB path, in combination with the Halgren-Lipscomb or BC method to generate the path, does not work as expected.
  • Not really a bug, but keep in mind that what we call RPBE is actually revPBE (Y. Zhang and W. Yang, Phys. Rev. Lett. 80, 890 (1998)) not RPBE by Hammer et al., Phys. Rev. B 59, 7413 (1999). The latter (RPBE) is obtained using
    exchange_correlation=ExchangeCorrelation(     exchange=HammerHansenNorskov,     correlation=PerdewBurkeErnzerhofCorrelation, )
  • Device structures cannot be passivated. You need to passivate the central region before converting it to a device.


Changes in licensing

Anyone who wants to try 12.2 - existing QuantumWise customer or not - can just download and install it without obtaining a license for it beforehand. This version will run using a built-in demo license which lasts for 14 days.


Note for customers

 Even if your license file contains a version number 12.2 or higher, you will not be fully able to run calculations with the final version of ATK 12.2 (and above) unless you get an upgraded license file. This is due to changes in the feature names. Customer with active maintenance contract will of course be able to switch license files free of charge - just contact your sales representative!


Note for trial users

For those who are interested in trying ATK, the trial procedure has been somewhat simplified. You can now download and run a fully functional* demo version of ATK for free, for 14 days, without first applying for a trial license. Just download and install!

* Really - fully functional. The only limitation of the trial version is that you are not allowed to publish any results obtained with it. For that, you must purchase a license.

After the demo period expires, you can apply for an extended trial by providing your proper and verifiable affiliation and a few other details using the trial form.

Free Trial