» Back to overview
Proposing Institution

Professorship for Computational Photonics, Department for Electrical and Computer Engineering, TUM
Project Manager

Michael Riesch
Arcisstrasse 21
80333 München
Quantum Cascade Lasers (QCLs) are a novel type of semiconductor laser. Their frequency range typically lies in the terahertz or mid-infrared regime, offering a variety of applications in several fields. In order to get a better understanding of the dynamic processes in a QCL (carrier transport, light-matter interaction, ...), numerical simulations are carried out. Our simulation model is a semiclassical approach that combines Maxwell's equations for the optical field with the Liouville-von Neumann equation for the quantum mechanical treatment of the carrier transport.The effort to solve the Liouville-von Neumann equation is proportial to N^2, where N is the number of energy levels under consideration. For small level count, the computational workload can be handled by shared memory systems or graphics processing units. However, we want to include up to 1000 energy levels in order to incorporate a more accurate model. This yields a demand for more compute resources.In our existing simulation code different numerical methods to solve the Maxwell-Liouville-von Neumann equations are implemented. At the moment Maxwell's equations are discretized using the finite difference time domain (FDTD) method. The drawback of this method is the numerical dispersion that arises unless very fine discretization is used. The pseudo-spectral time domain (PSTD) method is a promising alternative which heavily uses the computationally intensive fast fourier transform (FFT) but is practically free of numerical dispersion even for coarser discretization.In the scope of this project we address two objectives. The first objective is to compare the different numerical methods with respect to performance. On the basis of this comparison the most suitable method is selected. Secondly, we aim to develop a new version of our code that uses the message passing interface (MPI) to distribute the computational workload between the compute resources. Once it is confirmed that the new version yields realistic results, the performance scalability of the simulation code is evaluated.

Impressum, Conny Wendler