Publications by authors named "H Graafsma"
HO transforms to two forms of superionic (SI) ice at high pressures and temperatures, which contain highly mobile protons within a solid oxygen sublattice. Yet the stability field of both phases remains debated. Here, we present the results of an ultrafast X-ray heating study utilizing MHz pulse trains produced by the European X-ray Free Electron Laser to create high temperature states of HO, which were probed using X-ray diffraction during dynamic cooling.
View Article and Find Full Text PDFTEMPUS, a Timepix4-based system for the event-based detection of X-rays.
J Synchrotron Radiat
September 2024
Article Synopsis
- TEMPUS is a new detector system designed for photon science, utilizing the Timepix4 chip.
- It operates in two modes: a photon-counting mode for high-speed readout at 40 kfps and an event-driven mode for precise time-stamping in the nanosecond range.
- The paper discusses the prototype's development, its readout system, and presents test results from evaluations conducted at PETRA III and ESRF.
Pulsed laser heating of an ensemble of Pd nanoparticles, supported by a MgO substrate, is studied by x-ray diffraction. By time-resolved Bragg peak shift measurements due to thermal lattice expansion, the transient temperature of the Pd nanoparticles is determined, which quickly rises by at least 100 K upon laser excitation and then decays within 90 ns. The diffraction experiments were carried out using a Cu x-ray tube, giving continuous radiation, and the hybrid pixel detector Timepix3 operating with single photon counting in a time-of-arrival mode.
View Article and Find Full Text PDFArticle Synopsis
- Serial crystallography at synchrotron and XFEL sources generates large data sets, but only a small fraction is useful for analysis, necessitating an efficient data classification system to identify 'hit' (useful) and 'miss' (non-useful) images.
- The proposed solution includes a real-time feature extraction algorithm called modified and parallelized FAST (MP-FAST), paired with an image descriptor and a machine learning classifier to sort images effectively.
- Performance testing shows that the MP-FAST-based classification outperforms traditional feature extractors and classifiers by leveraging various processing units for improved speed and accuracy.
Article Synopsis
- Serial crystallography at X-ray free-electron laser facilities generates large amounts of data, but only some is useful for analysis, requiring differentiation between 'hit' (useful) and 'miss' (not useful) data.
- Artificial intelligence techniques, especially convolutional neural networks (CNNs), have shown quantitative success in classifying this data, but the internal workings of these networks remain poorly understood and have not been visualized.
- This study aims to provide qualitative insights by visualizing which parts of an image influence CNN predictions, helping to demystify these 'black box' models in the context of serial crystallography.