Time:
10:00
Location:
Sala Sancrotti, Building Q2, Strada Statale 14 – km 163,5 in AREA Science Park – 34149 Basovizza, Trieste ITALY
Speaker:
Kirsty Paton and Xiangyu Xie (Photon Science Detector Group, Paul Scherrer Institut, Switzerland)
Abstract:
Direct electron detectors (DEDs) have caused a step chance across all branches of electron microscopy, enabling new insights across a range of fields, from the life sciences to condensed matter physics. In biology, monolithic active pixel sensors have led to the “resolution revolution” in cryogenic electron microscopy (cryo-EM), due to their small (≤ 15 µm) pixel pitch and excellent imaging performance when using high-energy (≥ 200 keV) electrons. However, their dynamic range, radiation-hardness and frame-rates are limited. Furthermore, their imaging performance deteriorates at lower (≤ 120 keV) electron energies, which are of increasing interest for better image contrast at a given electron dose and lower operational costs.
Hybrid pixel detectors (HPDs) are an alternative type of DED, which have been widely adopted for applications requiring large dynamic range, kHz frame-rates and MHz count-rates, namely diffraction-based modalities and high-speed filming of dynamical studies. Their imaging performance for low-energy electrons is good but is in general limited by their large (typically ≥ 55 µm) pixels. With increasing electron-energy, the trajectories of incident electrons in the thick (≥ 300 µm) Si sensors of HPDs get longer, so that multiple pixels register each incident electron, resulting in a deterioration in imaging performance. Given the other advantageous properties of HPDs, the question naturally arises as to how their performance for imaging, especially with high-energy electrons can be improved. We will discuss two strategies to achieve this, with the aim of developing highly versatile detectors that perform well across wide range of experimental modalities in transmission electron microscopy (TEM).
The first approach utilizes deep learning methods to reconstruct the impact points of incident electrons with sub-pixel resolution. This has been realized with MÖNCH, a general-purpose, charge integrating HPD which is notable for its small, 25 µm pitch pixels that make it possible to extract the maximum amount of information from incident electrons’ trajectories to localize their entry point. For this strand of research, details of the deep learning framework, experimental set-up data and acquisition will be presented, along with results in the degree improvement in imaging performance achieved, as quantified by measurements of the detector’s modulation transfer function and detective quantum efficiency.
The second strategy takes advantage of the fact that the sensors of HPDs do not need to be made of Si. We will outline ongoing efforts to investigate the potential of high-Z sensors, which should improve the imaging performance of HPDs due to their greater stopping power compared with Si. Our current research efforts currently focus on using the JUNGFRAU HPD to investigate a new variety of GaAs:Cr, and we will present results comparing the characteristics of this new form of GaAs:Cr with earlier varieties and some initial TEM measurements.