VSR-Seminar

Vortrag 1: Topological Spintronics: guided by ab-initio research

Referent: Prof. Dr. Yuriy Mokrousov, IAS-1

Nanoelectronics is facing major challenges in the future, and the latest experimental advances mark the advent of the new concepts, which are believed to bring us to a new level of technology. These so-called topological concepts deal with the dynamics of quantum systems via exploiting the framework of geometric phases. The benefits of the topological framework became apparent when some years ago it was discovered that accounting for the topological nature of Bloch electrons in solids leads to a clear understanding and formulation of the electron and spin dynamics under applied external electric and magnetic fields. In my talk I will demonstrate how state-of-the-art ab initio research of topological properties of solids can assist in shaping the future of topological spintronics by providing crucial insights into the physics and properties of such important phenomena as Hall effects, magnetic interactions, skyrmions, spin torques, and terahertz spin currents. A crucial aspect I will emphasize is the importance of material-specific aspect of theoretical research in this field, based on first principles techniques, for guiding the experiment and technology towards the realization of highly efficient and ultrafast magnetic applications.

Vortrag 2: Inverse Modelling for Atmospheric Earth observation

Referent: Dr. Hendrik Elbern, IEK-8

With the advent on an ever increasing fleet of environmental monitoring satellites, the task of surveying the climate and earth system states by data assimilation, i.e. combining models with data in a mathematically rigorous way, is emerging. Objectives like early climate change signal detection and attribution are only one of the driving factors, as are better analyses of costly field campaigns. Another objective is the urgent requirement of energy meteorology to provide significantly improved short term predictions of wind and photovoltaic power, both resting on atmospheric state analyses by assimilation. The presentation aims to introduce advanced inverse modelling algorithms for trace gases and aerosols in their full atmospheric context, which are developped and applied. The main focus is placed on the numerical implementation features, which require the use of the HPC platforms.

The underlying EURopean Air pollution Disperion - Inverse Model (EURAD-IM) solves an advection-diffusion-reaction equation with a manifold variety of solvers, specific to different chemical and aerosol constituents. It simulates 70 reactive trace gases and additionally aerosol parameters of about the same number per grid point as optimisation parameters, along with 23 emitted species per surface grid point. For chemical state and emission strengths inversion, this implies to an optimisation problem with ≈ O(10⁷) degrees of freedom.

This problem is solved by the four-dimensional variational data assimilation (4D-var) procedure as the underlying technique which involves adjoint modelling for least square optimisation by quasi-Newton minimisation. A special challenge is the preconditioning task which is solved by a diffusion paradigm approach, replacing an intractably large covariance matrix of size ≈ O(10¹⁴).

Further, it will be demonstrated how the model development of both the chemistry transport model and the meteorological driver model WRF proceed to the transition to stochastic modelling, and to solve the data assimilation problem by particle filtering and particle smoothing. Related techniques with focus on parallelisation will be given, with emphasis placed on ensemble approaches on JUQUEEN, also involving ultra large ensemble sizes and data mining for capturing exceptional events, with low likelyhood and high impact for energy meteorology forecast needs. These include gross errors in wind power prediction, but also volcanic ash dispersion.

Alle Interessierten sind herzlich eingeladen.