As a computational scientist, my primary research interest is to accelerate numerical simulations in science and engineering as efficiently as possible. Given my background and my current affiliation, my primary area of application is micro– and nanoelectronics. However, my interests – thus my concrete research topics – are broad. Here, I discuss my current and most important research topics.
I am particularly interested in utilizing high performance computing resources, i.e., supercomputers (like the VSC-3 in Vienna) as well as multi-core workstations and many-core accelerators and co-processors. To that end I investigate and develop algorithms, data structures, and simulators based on shared-memory parallelization techniques (e.g. OpenMP) as well as distributed-memory approaches (e.g. MPI). In particular, I am a long-term contributor to the ViennaCL project, a free open source linear algebra library for multi- and many core architectures.
Simulating an expanding front is a fundamental step in many computational science and engineering applications, such as image segmentation, brain connectivity mapping, medical tomography, seismic wave propagation, geological folds, semiconductor process simulation, or computational geometry. In my case, I focus my research on applications in the area of semiconductor process simulation, where a redistancing step is a vital and time-critical component of the level set-based surface evolution simulations used, for instance, for plasma etching.
In general, an expanding front originating from a start position is described by its first time of arrival to the points of a domain. This problem can be described by solving the Eikonal equation. I investigate novel methods for solving this equation targeting shared-memory multi- and many-core architectures via, for instance, OpenMP. I particularly focus on the fast iterative method, the semi-ordered fast iterative method, and the fast marching method.
An attractive new approach to model future nanoelectronic devices is based on the Wigner formalism, which provides an attractive alternative to the non-equilibrium Green’s function formalism. Both stochastic and deterministic methods have been applied to solve the one-dimensional Wigner equation. However, only the Wigner Monte Carlo method, using the signed-particle technique, has made multi-dimensional Wigner simulations viable thus far; a multi-dimensional approach is essential for the simulation of realistic semiconductor devices.
However, the computational efforts of solving the signed-particle-based Wigner Monte Carlo system is significant, introducing the dire need for parallelization. One option is to partition the spatial domain and distribute the thus partitioned workload among the compute units. To allow for a scalable approach, distributed-memory architectures (e.g. supercomputers) are targeted using MPI. The concepts of classical domain decomposition have been adapted and applied to the signed-particle Wigner Monte Carlo simulator yielding excellent speed-up allowing to obliterate the otherwise multi-day simulation period to the minutes to hours regime. Aside of the parallelization techniques I oversee the free open source ViennaWD simulation package, which contains the simulator.
To be able to solve mathematical models for complicated structures (e.g. a FlexFET), the geometrical description – describing the structure – needs to be discretized, i.e., the simulation domain needs to be partitioned into a finite set of elements. The mathematical models, e.g. PDEs, are then evaluated on those elements respecting a certain set of boundary conditions and possible initial guesses. The step of discretizing the simulation domain is called mesh generation which introduces its own set of intricate challenges, like accuracy and speed. In the earlier days I started investigating software aspects to interface with several mesh generation tools and use them to build advanced meshing workflows, aiding the end-users in creating adequate meshes for their simulations. This work birthed the free open source ViennaMesh software package. Nowadays, I am focusing on supervising younger colleagues in this area.