Die DPG-Frühjahrstagung in Dresden musste abgesagt werden! Lesen Sie mehr ...

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

O: Fachverband Oberflächenphysik

O 62: Focus Session: Big Data in Aquisition in ARPES (joint session O/CPP)

O 62.3: Vortrag

Mittwoch, 18. März 2020, 11:15–11:30, REC C 213

Handling Big Multidimensional Experimental Data on Small Desktop Computers — •Michael Hartelt, Benjamin Frisch, Tobias Eul, Eva Prinz, Marten Wiehn, Benjamin Stadtmüller, and Martin Aeschlimann — Department of Physics and Research Center OPTIMAS, TU Kaiserslautern, Germany

In the pursuit of discovering new phenomena, photoemission experiments have evolved to capture ever more information about the electronic properties of materials. Major progress was made in the parallel detection of more degrees of freedom, which can include real- or k-space, energies and spin states of the emitted electrons. Additional dimensions of the parameter space are opened up by state-of-the-art experimental techniques that vary the sample temperature, the photon energy of the light source, or the time-delay between ultrashort laser pulses. As a result, experimental datasets of a single experiment can nowadays be 4-dimensional or even more. This makes the analysis of experimental data a non-trivial task, both conceptually and computationally.

We present our approach for the easy handling of these multi-dimensional, bigger-than-memory datasets on a conventional office computer. Using the Python programming language gives us access to powerful open-source packages like h5py, opencv, numpy, pint, and pycuda. Taking advantage of these, we integrated them into a toolbox package, which manages storage of large datasets for optimized I/O performance. The user is provided with an interface based on physical context, to perform data evaluation procedures with high efficiency.