A 10-minute read
4D STEM is currently present in a significant share of TEM projects, and it is certainly on the “must-watch” list of every electron microscopist. Its wide adoption in Materials Science can be associated with the latest developments in TEM technology, most notably the hybrid-pixel direct electron detector.
Regardless of naming conventions, or the method’s roots in diffraction mapping, 4D STEM is now the trailblazing technique for improving (virtual) STEM imaging; revealing hidden electromagnetic fields; crystallography and localized defects analyses. It also brought us new records in the resolution with ptychography reconstructions.
STEM imaging is based on rastering a high-energy electron beam over a thin sample and detecting “selected” electrons that travel through it. The careful selection and detection of transmitted electrons result in high-resolution images with meaningful information to the sample characterization. Chemical (Z-) contrast is achieved by collecting electrons that are scattered to high angles (>80 mrad), and enhanced visibility of crystalline domains can be obtained by collecting electrons that are scattered at intermediary angles – within diffraction angles as defined by the sample’s crystalline structure.
Beyond the features that are directly observable with conventional STEM imaging, a much more accurate picture of TEM samples can be drawn using complete electron scattering distribution (or diffraction pattern) analysis. For instance, the precise measurement of local electromagnetic fields and details of a sample’s crystal structure and its defects (akin to EBSD maps on SEMs) are among the local characteristics that can be retrieved from diffraction patterns in TEM.
For years, TEM users have been decoding samples’ features from recorded diffraction patterns, and some of the more enthusiastic experts attempted to scan a small electron beam over a region of interest to record localized diffraction patterns as a map. Despite their limited success (spoiler alert: their detectors’ technology was their main limitation!), their efforts correlate 1-to-1 with today’s widely recognized 4D STEM  practice – recording a 2D array of electron probe positions, each containing a (2D) localized diffraction pattern that is captured with a pixelated detector. Whether brand-new or rebranded, 4D STEM has recently become a hot topic in materials characterization on the nanoscale.
STEM imaging using multiple annular detectors is an effective way to capture and sum up the scattering signals from delimited regions of a diffraction pattern. The detectors effectively work as a “one-pixel camera” by returning the total electron count (as a gray level) from their region of collection, and this allows for an extremely fast time per pixel (also known as “dwell time”) – usually in the µs range. For instance, the recording of a typical STEM image with a 1,024 x 1,024 scan array and a dwell time of 5 µs takes only about 5 seconds. However, the integration of the signals by STEM detectors results in reduced information, in comparison to the ideal recording of full diffraction patterns as 2D images.
The evolution of electron recording devices (skipping the photographic plates for logistical reasons) brought a remarkable improvement in the acquisition speed for images and diffraction patterns alike. The change in detector architecture, from CCDs to CMOS technologies, allowed TEM cameras to speed up their acquisition time from around 1 s to a few milliseconds for a snapshot of 512 x 512 pixels. However, even with this significant speed-up, 4D STEM experiments remained out of reach: the same 1,024 x 1,024 scan array, now with a dwell time of 3.3 ms, would take around 55 minutes to complete! This is clearly an impractical experiment, and it would possibly result in a distorted dataset due to sample drift, or an “unhappy” sample with the excessive electron dose.
A possible compromise on 4D STEM data acquisition with conventional CMOS cameras is to reduce the scan array (to 128 x 128 points, for instance) in order to allow for quicker acquisitions (around one minute). This, though, would occur at the expense of spatial resolution and/or field of view for the measurement.
The recent introduction of hybrid-pixel technology for electron detection is quickly changing perspectives on 4D STEM experiments. Besides the remarkable enhancement in the acquisition speed of TEM images and diffraction patterns, hybrid-pixel direct electron detectors feature enhanced sensitivity. This improved capability can discriminate single-electron events and offers an extended dynamic range that allows for flexible diffraction experiments: farewell to beam stoppers!
DECTRIS hybrid-pixel detectors can operate at 4,500 Hz (with an 8-bit dynamic range, as well as a readout of 1,024 x 512 pixels for DECTRIS ELA detectors), which is equivalent to recording a full diffraction pattern every 220 µs. The same 4D STEM experiment that was mentioned above (with a 1,024 x 1,024 scan array frame) now lasts around 4 minutes and is suitable even for atomic resolved maps – the typical drift rates are well below 0.5 nm/minute for modern TEM microscopes. In case faster acquisitions are necessary, speeds up to 18,000 Hz can be reached easily by selecting a smaller detector region for readout.
Beyond acquisition speed (and the scope of this blog entry), hybrid-pixel detectors’ efficiency (DQE) impacts the quality of the recorded diffraction patterns. It is safe to state that the increasing acquisition speeds, in combination with moderate beam currents – keeping in mind possible damage to the sample – will result in fewer electrons per electron diffraction snapshot, and possibly a maximum of few counts for most pixels. In this context, the effective recording of fast 4D STEM datasets requires noise-free acquisition with single-electron sensitivity, as is allowed by DECTRIS hybrid-pixel technology.
A new direction that is starting to take shape is time-resolved 4D STEM, or “5D STEM” if we loosely add time as a dimension. The acquisition of a 128 x 128 scan array can be concluded in less than a second at the current maximum speed, and the new-generation detectors are not getting any slower. So, buckle up and fine-tune your data analysis algorithms, as complete diffraction patterns will come pouring in!
References and Acknowledgments
 Sample provided courtesy of Cristina Bran, Institute of Materials Science Madrid (ICMM-CSIC).
 Liam Spillane and Ana Pakzad’s (AMETEK, Gatan) support during experiments and discussions is greatly appreciated.
 B. Plotkin-Swing et al., Ultramicroscopy (2020), DOI: 10.1016/j.ultramic.2020.113067