Filtern
Dokumenttyp
- Zeitschriftenartikel (29)
- Vortrag (21)
- Posterpräsentation (2)
- Buchkapitel (1)
- Sonstiges (1)
Schlagworte
- Synchrotron (30)
- BAMline (26)
- XRF (18)
- Gold (7)
- Machine learning (7)
- XANES (7)
- Bayesian Statistics (5)
- Gaussian process (4)
- Machine Learning (4)
- Neural network (4)
Organisationseinheit der BAM
- 6 Materialchemie (54) (entfernen)
Paper des Monats
- ja (1)
Eingeladener Vortrag
- nein (21)
Enhancing efficiency at bamline: employing data science and machine learning for x-ray research
(2023)
This talk discusses how data science and machine learning techniques are being applied at the BAM Federal Institute for Materials Research and Testing to enhance efficiency and automation at the BAMLine synchrotron facility. The methods presented include Gaussian processes and Bayesian optimization for beamline adjustment and optimization of X-ray measurements. These statistical techniques allow automated alignment of beamline components and active learning scanning to reduce measurement time.
Additional machine learning methods covered are neural networks for quantification of X-ray fluorescence (XRF) data and decoding coded apertures.
This contribution provides an overview of the BAMline synchrotron radiation beamline, which specializes in hard X-ray spectroscopy techniques for materials research. The BAMline offers X-ray absorption spectroscopy (XAS), x-ray fluorescence spectroscopy (XRF), and tomography to study materials' electronic structure, chemical composition, and structure. Key capabilities include standard and dispersive XAS for electronic structure, micro-XRF for elemental mapping, coded aperture imaging, and depth-resolved grazing exit XAS. The BAMline enables in situ characterization during materials synthesis and functions for energy, catalysis, corrosion, biology, and cultural heritage applications.
Ongoing developments like the implementation of machine learning techniques for experiment optimization and data analysis will be discussed. For instance, Bayesian optimization is being used to improve beamline alignment and scanning. An outlook to the future, where the BAMline will continue pioneering dynamic and multi-scale characterization, aided by advanced data science methods, to provide unique insights into materials research, will be given.
To gain insight into the possible origin of the gold used in the production of tesserae containing gold leaf less than 0.5 μm thick placed between two layers of glass, we propose a non-destructive synchrotron radiation (SR) XRF protocol based on sequential analysis under optimised analytical conditions. Using this protocol, trace element analysis is achieved with detection limits of 1–6 mg/kg. As Pt and Au have adjacent fluorescence energies, we tested the most challenging situation, when Pt is present in very low concentrations in gold. Data obtained by double-dispersive XRF (D2XRF) and μXRF for fourth–ninth-century mosaics decorating nine Eastern and Western religious buildings show that the Eastern and Western tesserae are made from different alloys.
However, these alloys are identical to those used to make gold leaf for gilding, because plastic deformation requires the use of gold alloys with high ductility and malleability. Although trace element composition of gold used in the concerned period is only available for coins, by comparing the amounts of Pt contained in the tesserae and in the coins we show that Roman tesserae are made from Roman gold, as described in the documentary sources. We observe for the Byzantine period the use of a Byzantine gold and of gold supposedly from different stages of recycling, and we suggest the use of Umayyad and Abbasid gold for the production of Islamic tesserae.