skip to main content

This content will become publicly available on December 8, 2024

Title: Deep Learning Approach for High-accuracy Electron Counting of Monolithic Active Pixel Sensor-type Direct Electron Detectors at Increased Electron Dose

Electron counting can be performed algorithmically for monolithic active pixel sensor direct electron detectors to eliminate readout noise and Landau noise arising from the variability in the amount of deposited energy for each electron. Errors in existing counting algorithms include mistakenly counting a multielectron strike as a single electron event, and inaccurately locating the incident position of the electron due to lateral spread of deposited energy and dark noise. Here, we report a supervised deep learning (DL) approach based on Faster region-based convolutional neural network (R-CNN) to recognize single electron events at varying electron doses and voltages. The DL approach shows high accuracy according to the near-ideal modulation transfer function (MTF) and detector quantum efficiency for sparse images. It predicts, on average, 0.47 pixel deviation from the incident positions for 200 kV electrons versus 0.59 pixel using the conventional counting method. The DL approach also shows better robustness against coincidence loss as the electron dose increases, maintaining the MTF at half Nyquist frequency above 0.83 as the electron density increases to 0.06 e−/pixel. Thus, the DL model extends the advantages of counting analysis to higher dose rates than conventional methods.

more » « less
Award ID(s):
Author(s) / Creator(s):
; ; ; ; ; ; ;
Publisher / Repository:
Oxford University Press
Date Published:
Journal Name:
Microscopy and Microanalysis
Page Range / eLocation ID:
2026 to 2036
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. Optical metasurfaces consist of densely arranged unit cells that manipulate light through various light confinement and scattering processes. Due to its unique advantages, such as high performance, small form factor and easy integration with semiconductor devices, metasurfaces have been gathering increasing attention in fields such as displays, imaging, sensing and optical computation. Despite advances in fabrication and characterization, a viable design prediction for suitable optical response remains challenging for complex optical metamaterial systems. The computation cost required to obtain the optimal design exponentially grows as the design complexity increases. Furthermore, the design prediction is challenging since the inverse problem is often ill-posed. In recent years, deep learning (DL) methods have shown great promise in the area of inverse design. Inspired by this and the capability of DL to produce fast inference, we introduce a physics-informed DL framework to expedite the computation for the inverse design of metasurfaces. Addition of the physics-based constraints improve generalizability of the DL model while reducing data burden. Our approach introduces a tandem DL architecture with physics-based learning to alleviate the nonuniqueness issue by selecting designs that are scientifically consistent, with low error in design prediction and accurate reconstruction of optical responses. To prove the concept, we focus on the inverse design of a representative plasmonic device that consists of metal gratings deposited on a dielectric film on top of a metal substrate. The optical response of the device is determined by the geometrical dimensions as well as the material properties. The training and testing data are obtained through Rigorous Coupled-Wave Analysis (RCWA), while the physics-based constraint is derived from solving the electromagnetic (EM) wave equations for a simplified homogenized model. We consider the prediction of design for the optical response of a single wavelength incident or a spectrum of wavelength in the visible light range. Our model converges with an accuracy up to 97% for inverse design prediction with the optical response for the visible light spectrum as input. The model is also able to predict design with accuracy up to 96% and optical response reconstruction accuracy of 99% for optical response of a single wavelength of light as input. 
    more » « less
  2. Precision and accuracy of quantitative scanning transmission electron microscopy (STEM) methods such as ptychography, and the mapping of electric, magnetic, and strain fields depend on the dose. Reasonable acquisition time requires high beam current and the ability to quantitatively detect both large and minute changes in signal. A new hybrid pixel array detector (PAD), the second-generation Electron Microscope Pixel Array Detector (EMPAD-G2), addresses this challenge by advancing the technology of a previous generation PAD, the EMPAD. The EMPAD-G2 images continuously at a frame-rates up to 10 kHz with a dynamic range that spans from low-noise detection of single electrons to electron beam currents exceeding 180 pA per pixel, even at electron energies of 300 keV. The EMPAD-G2 enables rapid collection of high-quality STEM data that simultaneously contain full diffraction information from unsaturated bright-field disks to usable Kikuchi bands and higher-order Laue zones. Test results from 80 to 300 keV are presented, as are first experimental results demonstrating ptychographic reconstructions, strain and polarization maps. We introduce a new information metric, the maximum usable imaging speed (MUIS), to identify when a detector becomes electron-starved, saturated or its pixel count is mismatched with the beam current. 
    more » « less
  3. Abstract

    Two-dimensional (2D) ternary materials recently generated interest in optoelectronics and energy-related applications, alongside their binary counterparts. To date, only a few naturally occurring layered 2D ternary materials have been explored. The plethora of benefits owed to reduced dimensionality prompted exploration of expanding non-layered ternary chalcogenides into the 2D realm. This work presents a templating method that uses 2D transition metal dichalcogenides as initiators to be converted into the corresponding ternary chalcogenide upon addition of copper, via a solution-phase synthesis, conducted in high boiling point solvents. The process starts with preparation of VSe2nanosheets, which are next converted into Cu3VSe4sulvanite nanosheets (NSs) which retain the 2D geometry while presenting an X-ray diffraction pattern identical with the one for the bulk Cu3VSe4. Both the scanning electron microscopy and transmission microscopy electron microscopy show the presence of quasi-2D morphology. Recent studies of the sulfur-containing sulvanite Cu3VS4highlight the presence of an intermediate bandgap, associated with enhanced photovoltaic (PV) performance. The Cu3VSe4nanosheets reported herein exhibit multiple UV–Vis absorption peaks, related to the intermediate bandgaps similar to Cu3VS4and Cu3VSe4nanocrystals. To test the potential of Cu3VSe4NSs as an absorber for solar photovoltaic devices, Cu3VSe4NSs thin-films deposited on FTO were subjected to photoelectrochemical testing, showing p-type behavior and stable photocurrents of up to ~ 0.036 mA/cm2. The photocurrent shows a ninefold increase in comparison to reported performance of Cu3VSe4nanocrystals. This proves that quasi-2D sulvanite nanosheets are amenable to thin-film deposition and could show superior PV performance in comparison to nanocrystal thin-films. The obtained electrical impedance spectroscopy signal of the Cu3VSeNSs-FTO based electrochemical cell fits an equivalent circuit with the circuit elements of solution resistance (Rs), charge-transfer resistance (Rct), double-layer capacitance (Cdl), and Warburg impedance (W). The estimated charge transfer resistance value of 300 Ω cm2obtained from the Nyquist plot provides an insight into the rate of charge transfer on the electrode/electrolyte interface.

    more » « less
  4. In time-correlated single-photon counting (TCSPC), photons that arrive during the detector and timing electronics dead times are missed, causing distortion of the detection time distribution. Conventional wisdom holds that TCSPC should be performed with detections in fewer than 5% of illumination cycles to avoid substantial distortion. This requires attenuation and leads to longer acquisition times if the incident flux is too high. Through the example of ranging with a single-photon lidar system, this work demonstrates that accurately modeling the sequence of detection times as a Markov chain allows for measurements at much higher incident flux without attenuation. Our probabilistic model is validated by the close match between the limiting distribution of the Markov chain and both simulated and experimental data, so long as issues of calibration and afterpulsing are minimal. We propose an algorithm that corrects for the distortion in detection histograms caused by dead times without assumptions on the form of the transient light intensity. Our histogram correction yields substantially improved depth imaging performance, and modest additional improvement is achieved with a parametric model assuming a single depth per pixel. We show results for depth and flux estimation with up to 5 photoelectrons per illumination cycle on average, facilitating an increase in time efficiency of more than two orders of magnitude. The use of identical TCSPC equipment in other fields suggests that our modeling and histogram correction could likewise enable high-flux acquisitions in fluorescence lifetime microscopy or quantum optics applications.

    more » « less
  5. Context . We implement an electron avalanche photodiode (e-APD) in the MIRC-X instrument, which is an upgrade of the six-telescope near-infrared imager MIRC, at the CHARA array. This technology should improve the sensitivity of near-infrared interferometry. Aims . We aim to characterize a near-infrared C-RED ONE camera from First Light Imaging (FLI) using an e-APD from Leonardo (previously SELEX). Methods . We first used the classical mean-variance analysis to measure the system gain and the amplification gain. We then developed a physical model of the statistical distribution of the camera output signal. This model is based on multiple convolutions of the Poisson statistic, the intrinsic avalanche gain distribution, and the observed distribution of the background signal. At low flux level, this model independently constrains the incident illumination level, the total gain, and the excess noise factor of the amplification. Results . We measure a total transmission of 48 ± 3% including the cold filter and the Quantum Efficiency. We measure a system gain of 0.49 ADU/e, a readout noise of 10 ADU, and amplification gains as high as 200. These results are consistent between the two methods and therefore validate our modeling approach. The measured excess noise factor based on the modeling is 1.47 ± 0.03, with no obvious dependency with flux level or amplification gain. Conclusions . The presented model allows the characteristics of the e-APD array to be measured at low flux level independently of a preexisting calibration. With < 0.3 electron equivalent readout noise at kilohertz frame rates, we confirm the revolutionary performances of the camera with respect to the PICNIC or HAWAII technologies. However, the measured excess noise factor is significantly higher than what is claimed in the literature (< 1.25), and explains why counting multiple photons remains challenging with this camera. 
    more » « less