Hujsak Karl A, Roth Eric W, Kellogg William, Li Yue, Dravid Vinayak P
Department of Materials Science & Engineering, Northwestern University, Evanston, Illinois, 60208-3108, United States.
BioCryo Facility, NUANCE Center, Northwestern University, Evanston, Illinois, 60208-3108, United States.
Micron. 2018 May;108:31-40. doi: 10.1016/j.micron.2018.03.001. Epub 2018 Mar 10.
Technological advances in electron microscopy, particularly improved detectors and aberration correctors, have led to higher throughput and less invasive imaging of materials and biological structures by enhancing signal-to-noise ratios at lower electron exposures. Analytical methods, such as electron energy loss spectroscopy (EELS) and energy dispersive x-ray spectrometry (EDS), have also benefitted and offer a rich set of local elemental and bonding information with nano-or atomic resolution. However, spatially resolved spectrum imaging with EELS and EDS continue to be difficult to scale due to limited detector collection angles or high signal background, requiring hours or even days for full maps. We present the principle and application of a Multi-Objective Autonomous Dynamic Sampling (MOADS) method which can accelerate spectrum mapping in EELS or EDS by over an order of magnitude. Initial guesses about the true spectrum images are constructed as measurements are collected, which allows the prediction of points which contribute information/contrast. In this fashion, an intelligently selected and reduced set of points which best approximate the true spectrum image are autonomously collected on-the-fly to save considerable time and/or radiative area dose. We implemented MOADS as a software add-on to arbitrary commercial Scanning Transmission Electron Microscopes (STEMs) equipped with a Gatan Digital Micrograph (DM, Gatan ©) interface. We demonstrate the efficacy of our proposed method on several prototypical analytical specimens, as well as dose sensitive materials. It is expected that MOADS and similar supervised dynamic sampling approaches may open the exploration of large area analytical maps or the imaging of beam reactive materials not previously thought feasible.
电子显微镜技术的进步，特别是改进的探测器和像差校正器，通过在较低电子曝光下提高信噪比，实现了对材料和生物结构的更高通量和侵入性更小的成像。诸如电子能量损失谱（EELS）和能量色散X射线光谱（EDS）等分析方法也从中受益，并提供了具有纳米或原子分辨率的丰富的局部元素和键合信息。然而，由于探测器收集角度有限或信号背景较高，EELS和EDS的空间分辨光谱成像仍然难以扩展，完整的图谱需要数小时甚至数天时间。我们介绍了一种多目标自主动态采样（MOADS）方法的原理和应用，该方法可以将EELS或EDS中的光谱映射速度提高一个数量级以上。在收集测量数据时构建对真实光谱图像的初始猜测，这使得能够预测有助于提供信息/对比度的点。通过这种方式，可以动态自主地收集一组经过智能选择和缩减的、最接近真实光谱图像的点，从而节省大量时间和/或辐射面积剂量。我们将MOADS实现为一个软件插件，用于配备Gatan Digital Micrograph（DM，Gatan ©）接口的任意商用扫描透射电子显微镜（STEM）。我们在几个典型分析样品以及对剂量敏感的材料上展示了我们提出的方法的有效性。预计MOADS和类似的监督动态采样方法可能会开启对大面积分析图谱的探索，或者对以前认为不可行的束反应材料进行成像。
