Integrated open software suite for nanoscale modeling
At the moment DFTB+XT - the core package is ready for public use, CP2K+XT is available for testing.
Other parts of the suite are under development.
Theoretical and computational modeling of real systems requires multiscale and multiphysics approach linking atomistic ab initio description, quantum transport methods and 3D continuous methods for classical electric fields (also stress, thermal flow etc.). For atomistic modeling the semi-empirical, ab initio, Densitiy Functional Theory (DFT) and hybrid methods are used. The Density Functional Tight-Binding (DFTB) approach is especially effective for large (devices, bio) systems, being orders of magnitude faster than full DFT approach with similar accuracy. Atomic and molecular scale systems are naturally described by discrete-level models, based, for example, on atomic orbitals or molecular orbitals. Starting from discrete-level representations we arrive at matrix Green functions, being the main theoretical tool at the nanoscale, convenient for numerical implementation. We develop our own open source DFTB+XT/TraNaS package as a core of the TraNaS OpenSuite. We also develope the extended version of the CP2K package, called CP2K+XT.
The other significant peculiarity of nanoscale systems is the enhanced role of interactions. Both electron- electron and electron-vibron interactions may be strong and the Landauer approach for coherent transport can not be used anymore. Fortunately, Nonequilibrium Green Function (NGF) methods are able to treat the many-body problems. We are working on effective parallel methods to solve the complicated equations of the many-body theory within the DFTB+XT/TraNaS package.
Finally, for real device geometries, the influence of external systems (electrodes, gates, STM tip) should be taken into account. In many cases there is possible to use classical Finite Element Method (FEM) to model the electric potentials produced by the charges in the classical part, as well as electric and mechanic state of the classical part. We work on integration with FEM methods using the ElmerSolver code.