Dec 2, 2024
2:30pm - 3:00pm
Sheraton, Third Floor, Fairfax B
Bryan Reed1,Ruth Bloom1,Gonzalo Eyzaguirre1,Jonathan Victorino1,Abdolreza Moghadam1,Curt Henrichs1,Daniel Masiel1,Hiroki Hashiguchi2,Kazuki Yagi2,Yu Jimbo2,Jonathan Peters3,Matthew Mosse3,Lewys Jones3
IDES, Inc.1,JEOL Ltd.2,Trinity College Dublin, The University of Dublin3
Bryan Reed1,Ruth Bloom1,Gonzalo Eyzaguirre1,Jonathan Victorino1,Abdolreza Moghadam1,Curt Henrichs1,Daniel Masiel1,Hiroki Hashiguchi2,Kazuki Yagi2,Yu Jimbo2,Jonathan Peters3,Matthew Mosse3,Lewys Jones3
IDES, Inc.1,JEOL Ltd.2,Trinity College Dublin, The University of Dublin3
This presentation will focus on how electrostatic beam blanking enables a surprising range of capabilities for precise, time-resolved, dose-controlled, intelligent transmission electron microscopy (TEM).<br/>TEM has progressed enormously in recent decades. Measurements that used to be heroic are now routine, often limited not so much by the microscope as by the sample. Sources are brighter, columns are more stable, aberration correction is widespread, and detectors have advanced to where 4D-STEM is replacing traditional imaging methods. If the sample can survive the intense scrutiny of the electron beam, modern instruments and techniques can draw out truly enormous amounts of information. But if the sample is more fragile, we need to be smarter about how we draw the information out.<br/>This brings us to the linked frontiers of automation, dose control, data analysis, and intelligent microscopy. The TEM is not just a big, expensive camera for taking pictures. It’s a tool for answering questions about the properties and behavior of materials. The way we probe the sample must be attuned to the questions we want answered. If the sample is fragile, we must allocate the dose where it matters and make the most of every bit of information we can catch. If we have <i>a priori</i> knowledge, we should use it not just to analyze the data but to direct the measurement itself, preferably in real time using strategies rooted in information theory. This is especially true if the material state we’re interested in only lasts a short time.<br/>Of all the microscope functions that have improved over the years, it’s easy to overlook one of the most basic: the way we turn the electron beam on and off. Old-fashioned magnetostatic beam blankers were fine in the days of cameras that could only capture about one frame per second, but by today’s standards they’re terribly slow and imprecise. Simply replacing this function with an electrostatic beam blanker, able to operate on nanosecond scales with zero hysteresis, yields surprising benefits. The beam blanker should be designed for integration into complex workflows, including both direct high-speed timing control and external software automation interfaces.<br/>The speed and lack of hysteresis of an electrostatic blanker means one can turn the beam on and off at essentially any time without affecting focus or alignment (apart from brief transients, negligible on the typical time scale of TEM measurements). This means you can use pulse width modulation (PWM) to turn down the beam current without sacrificing resolution, and you can freely change the PWM settings as often as you like—even for every single pixel in a STEM scan, a mode we call “dose painting.” You can blank the beam whenever it would produce poor or useless data, such as during flyback or even the inter-pixel settling time in STEM. You can even respond to signal levels in real time, blanking the beam when either a fast detector indicates you may be striking a high-energy-absorbing part of the sample, or upon reaching a criterion of accumulated signal level sufficient for your purpose. You can allocate dose in time and space so as to take advantage of the nonlinear, time-dependent aspect of beam-sample damage. And you can allocate dose specifically to regions of spacetime that are relevant for the questions you’re asking of the sample, no more and no less. These decisions can be made by human operators, high-speed circuitry, machine-learning algorithms operating in either open or closed loops, or any combination thereof.