Towards Cinematic STEM and Beyond: Fast Frame Rates Using Overdriven Scan Shaping

Faster frame rate imaging in scanning transmission electron microscopy (STEM) has long been desired to mitigate the effects of specimen drift [1], capture dynamic events [2], or for controlling electron dose [3]. Until recently, conventional STEM imaging synced each line with the mains electrical frequency to minimize the effects of external fields. However, this limits frame rates to ∼0.1 frames per second (fps) for typical image sizes (512x512 pixels). In more recent years, multi-frame imaging has been employed with frame rates on the order of 2 fps, where distortions in the stack of frames can be diagnosed and removed [1]. One of the major roadblocks to faster scanning is the requirement for a flyback time between scan lines, allowing time for the beam to travel from the right to the left of the image. As dwell-times are reduced more and more, the fixed flyback time represents a larger and larger percentage of the overall total acquisition time and dose. This flyback hysteresis can be reduced using computational post-processing [4], but further frame rate improvements require new hardware. The new generation of scan controllers are capable of arbitrary scans, allowing the possibility of new scanning strategies [5-7] for reducing flyback or reducing the number of exposed pixels. Pixel times on the order of 10 ns are also now achievable, though the deflection systems in most STEMs are not capable of such speeds. This is from a combination of the inductive nature of the scan coils, and the limitations of the electronics controlling them. New coil designs can reduce scan coil inductance [2], though with a limited field of view and not widely available on all microscopes. To achieve the highest frame rates on existing STEMs and at low magnifications, a method to mitigate the effects of the deflection system whilst scanning near the maximum speed of the scan controller is needed. For this goal, we propose the serpentine scanning pattern where alternating lines are scanned in opposite directions. This removes the large discontinuities in beam position of conventional scanning and has a minimal number of discontinuities in the beam velocity compared to other scanning patterns (e.g. Hilbert scans). Despite this, a fast serpentine scanning pattern will still suffer due to the imperfect deflection system as the beam cannot change direction instantaneously. To account for this, we propose an overdrive system to shape the scan to the desired output, in this case a modified serpentine with deceleration at the line ends. By overdriving the beam (i.e. programming the beam to go to a further position) we can force the beam to be in the desired position. We measure this using a CCD camera to directly measure the beam position in a confocal configuration (Fig. 1), and apply an iterative correction to account for the non-linear response of the beam (Fig. 2). We demonstrate this on a Thermo Fisher Titan G2 300 kV equipped with a point electronic REVOLON scan controller and Gatan UltraScan 1000 CCD. Using our method we can move towards imaging at ∼50 ns per pixel, and frame rates of 50 fps with fully sampled images. We also discuss the current problems with fast scanning and paths towards even greater frame rates in the future [8]. Example measured beam positions following a serpentine with 1024 STEM points (pixels) per period (equivalent to a serpentine scan with 512 left-to-right and 512 right-to-left pixels). Scanning speeds are 1 μs per pixel (blue), 200 ns per pixel (green), 100 ns per pixel (yellow), and 50 ns per pixel (red). As the scan becomes faster, it deviates more and more from the ideal triangular wave. Pre, (a), and post, (b), corrected deviation of measured beam position (dashed black) from desired beam position (solid orange).

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.