From a Prototype to a Fully Functioning Detector in a Few Months

A 15-minute read

Dr. Benedikt Haas obtained his Ph.D. in method development for Scanning Transmission Electron Microscopy (STEM) at the French Alternative Energies and Atomic Energy Commission (CEA) in Grenoble and completed a postdoc at the Humboldt University of Berlin in the group of Prof. Christoph Koch. Then, in 2020, he joined the latter institution to take charge of its new microscope. 

This Nion HERMES 200 kV is equipped with the very first DECTRIS ELA hybrid-pixel electron detector, which makes Benedikt and his colleagues the first users of this technology in the field of Electron Microscopy (EM). In this interview with the DECTRIS team, Benedikt shares his experience as an early adopter, as well as his future plans for how to get the maximum out of each electron with the ELA detector.

DECTRIS: Can you tell us a bit about the research you are doing in the EM center of Humboldt University? What instruments are you using?

Dr. Benedikt Haas, Staff Scientist at the Humboldt University of Berlin: We are a very versatile group. We have a Jeol 2200FS microscope – our workhorse – which we use for all kinds of routine work, but also for off-axis and in-line holography, among other experiments. We also operate a modified Scanning Electron Microscope (SEM), which can do transmission diffraction or 4D STEM at very low electron energies. At the Joint Laboratory for Structural Research (JLSR), in cooperation with the Helmholtz Centre Berlin and the Institute of Crystal Growth, we have access to a cryo Transmission Electron Microscope (TEM), which is equipped with a TVIPS camera and an objective-lens-corrected FEI Titan. 

And – last, but not least – we have the Nion HERMES 200 kV with an IRIS spectrometer and a DECTRIS ELA direct detector, on which we perform high-resolution Electron Energy Loss Spectroscopy (EELS). It is one of the few machines in the world with an energy resolution that is good enough for us to capture very low-energy phonons. 

In addition to EELS, we use this setup for 4D STEM experiments, in particular for ptychography reconstructions. But we also develop and apply other methods: for example, a technique called LARBEB, Large-Angle Rocking Beam Electron Diffraction. This is a crystallographic technique that is used to retrieve the structure factors of a material. With our instruments, we can work on a very versatile range of materials: from classic semiconductor structures to 2D materials, and even proteins.

DECTRIS: What was your role in building such a versatile facility?

Benedikt: From 2017, when I joined the microscopy group, I was heavily involved – together with my supervisor, Prof. Christoph Koch – in the planning and setup of the Nion HERMES lab. This work ranged from room planning, such as active mechanical vibration and passive electromagnetic field damping, to the installation and upgrading of critical components like the spectrometer and the detector.

DECTRIS: How did you end up with the DECTRIS ELA detector?

Benedikt: At the time that we installed our setup, ELA was not yet a product, but a prototype. We got it as a loan from Nion, shortly after the installation of the microscope. I was very excited about the opportunity to try out this new technology. The more experiments my colleagues and I did with ELA, the more convinced we became that we should buy it. In the end, Christoph, the head of the facility, managed to raise funds for the purchase. Now, we are proud owners of the DECTRIS ELA detector with serial number 01. 
 

Low Doses and Noise-Free Readout 

DECTRIS: When you decided to get a hybrid-pixel electron detector, what kinds of experiments did you have in mind?

Benedikt: The hybrid-pixel detector is excellent for any experiments that are limited by readout noise or low doses. It is perfect for EELS, due to its high saturation current and relatively high number of pixels, as compared to other hybrid-pixel detectors. That’s what we mostly intended it for at first. But we had also realized that, when the detector was operated at a very high frame rate, its signal-to-noise ratio was very good. Therefore, we adapted it for 4D STEM applications like ptychography. Together with its speed of up to 18,000 frames per second in windowed mode, ELA gave us a huge advantage in this field. 

Today, we also use ELA for our LARBED experiments, in which we need to acquire quantitative, noise-free diffraction patterns under different tilted illuminations to reconstruct the examined material’s scattering-factor matrix. And, most recently, I have been working on an energy- and momentum-resolved experiment, the so-called “5D STEM” technique (Author’s Note: Dr. Haas will present the results of his research during a poster session at the M&M 2022 in Portland). This is also something that greatly benefits from this detector’s high saturation current, relatively high number of pixels, and zero readout noise.

Finally, we are continuously developing new methods and techniques, and the parameters of our instruments are a key factor here. We are either empowered or limited by them, and this is why we strive to acquire the most powerful technology available.

DECTRIS: How did you deal with the amount of data that the detector acquired?

Benedikt: Last year, together with Nion, we developed and tested live data-processing approaches for the ELA detector [1]; this was the first-ever demonstration of 4D STEM live processing. The detector produces enormous amounts of data, and we think that it is crucial to learn how to extract and store the information that one needs without having to keep all the raw data. This approach of live virtual detectors keeps all the benefits of a pixelated detector, and also helps you not to drown in the data. 

This year, we went one step further and will demonstrate at the M&M in Portland how the data can be live-compressed efficiently – for example, based on Zernike polynomials [2]. Thus, data can still be used in analyses like ptychographic reconstructions, but are compressed by a factor of many thousands.

DECTRIS: What are the key requirements for the detector in your facility?

Benedikt: Since we perform very diverse research, there are a lot of requirements that need to be met. After the “obvious” expectations of direct detectors – a very high Detective Quantum Efficiency, DQE (ideally, making each electron impact count versus inherent noise) and a low point-spread function (the spatial “smear” of information across pixels from electron impacts) – the detector should meet several other criteria. 

First and foremost, the detector needs to be robust. We sometimes work with high currents, and we cannot have them damage the detector. However, it is one thing when the detector withstands the current; it is something else when the detector can deliver a meaningful readout from it. Hence, a high saturation per pixel is also highly appreciated. Another parameter that is important for EELS, for example, is the number of pixels, especially in the dispersive direction. 

Finally, speed – despite often being overlooked – is also a very relevant parameter. It allows us to obtain several maps consecutively from the same material, and quickly enough to overcome the drift and scanning artifacts, while spreading the dose over many frames, instead of just one. The registration of multiple fast maps, instead of one slow map, is very beneficial, as we have recently demonstrated [3]. Currently, many direct detectors are relatively slow, but ELA can go up to 18 kHz; this is a huge step forward.

“Almost an Unfair Advantage”

DECTRIS: What was your experience like, working with the ELA detector and DECTRIS?

Benedikt: When we got the detector, it was still a prototype, so we did not know what to expect from it. Its development and integration were still a work in progress for DECTRIS and Nion, and we were happy to be part of the process. We felt supported by both companies; we always received clear instructions and maintained close contact with the teams. Within a few months, we – all working together – turned our prototype into a fully functioning detector. 

Once the development was over, we had less interaction with DECTRIS, which was a good thing: the detector was performing as it was supposed to! Later on, however, I reached out with a research idea and some questions about the detector’s hardware. The team was very helpful and put me in contact with another DECTRIS user, with whom we now have an ongoing collaboration.

As for working with the detector itself, sometimes I feel that it gives us almost an unfair advantage over other research groups. We have had several instances when our fellow researchers would encounter a problem with their indirect detection devices, but we could solve it with the ELA detector. 

DECTRIS: What would you want to see next in the detector’s development?

Benedikt: On the one hand, I would wish for more speed. We have a great starting point with 18 kHz, but faster detectors would open up new possibilities. 

On the other hand – energy resolution. Since hybrid-pixel detectors work by counting, they inherently have a certain energy resolution. However, at the moment, it is not good enough to perform energy-resolved electron microscopy. I believe that energy resolution of the detector itself would improve our capabilities – and, even more importantly, EM methods’ ease of use.

About Benedikt Haas, Ph.D.

After obtaining a Ph.D. in method development for STEM (centered on strain and mesoscopic E-field measurement techniques) at the CEA in Grenoble (France) from 2013-2017, Benedikt Haas was a postdoc for three years in the group of Prof. Christoph T. Koch at the Department of Physics & IRIS Adlershof at the Humboldt-Universität zu Berlin. Since 2020, he has been a staff scientist there. He is responsible for the Nion HERMES ultra-high-resolution scanning transmission electron microscope (0.7 Å spatially and 6 meV in energy), which he utilizes to develop novel EM techniques that he applies to relevant Materials Science questions. His focus is on STEM, in the form of “4D STEM” and (momentum-resolved) high-resolution EELS. Materials he is, or has been, working on comprise 2D materials, topological insulators, nanowires, nanoparticles, different semiconductor structures, and many other systems. In addition, he is working on the registration and live processing of 4D STEM data (GBs/s) and “FAIR” (Findable, Accessible, Interoperable, and Reusable) data structures.

References

[1] Haas et al., High-Fidelity 4D-STEM Enabled by Live Processing at 15’000 Detector Frames Per Second, Microscopy and Microanalysis, 2021 (doi:10.1017/S1431927621003779)
[2] Gladyshev et al., Comparison of Compression Methods for Ptychographic Reconstructions through Decomposition of the Diffraction Patterns in Orthonormal Bases, Microscopy and Microanalysis, 2022 (doi:10.1017/S1431927622002306)
[3] O'Leary et al., Increasing Spatial Fidelity and SNR of 4D-STEM Using Multi-Frame Data Fusion, Microscopy and Microanalysis, 2021 (doi:10.1017/S1431927621012587)