Institute Quantum Phenomena in Novel Materials

Data Science

In 2016, the Helmholtz Incubator Information & Data Science, initiated by Helmholtz President Prof. Otmar Wiesler, sent an important signal for forward-looking research in data science. Experts from all 18 Helmholtz Centers provided new pulses for the use of big data. As part of this process, I was able to incorporate the ideas and strategies discussed there into our department and develop them further into my own research aspects. An important part for the digitization of our research processes is the optimization of digital and scientific workflows. On the other hand, we are handling a large amounts of data, in particular energy-resolved neutron scattering data can gain data sizes that make classical analysis difficult.

We prefer to use the NeXus/HDF5 data format, which was introduced and promoted with the flat-cone diffractionmeter E2 operating by our department until the final shutdown of the BER II research reactor (see also decommissioning). We provide the TVneXus software package for the analysis and evaluation of E2 data. The NeXus format is also used to store data during the data analysis.

Our department has been involved in the development of software for theory-based modeling of measurement data such as diffuse neutron scattering data based on Main-Field theory (La_1-xSr_xMnO₃), Crystal Disorder calculation (Zirconia) or random walks Monte-Carlo simulation of Spin-Ice (Dy₂Ti₂O₇) and water Ice (D₂O) approaches. Spin waves can be modeled with the MatLab package SpinW. 

For large or complex data sets, new prospects for machine learning (deep learning) must be sought. Here, a procedure was developed that starts with the simulation of measurement data to train the neural networks, through systematic testing of different types of ANNs, to the comparison with the measurement data itself and the determination of the wanted parameters. This was put together in Python library, called the ANNtoolbox project. This is illustrated by the BaNi₂V₂O₈ compound.

In order to be able to implement these new ideas, appropriate computer hardware was purchased and linked together. Databases such as NOMAD Oasis are in planning. We had created the blueprint for an IT-secure scientific and laboratory network.