In this interview, we had the pleasure of speaking with Dr. Stephen Exarhos, a TEM expert at Seagate. Dr. Exarhos shared valuable insights into TEM research for electronic devices at the nanoscale, offering a deep dive into sample preparation, in situ experiments in industry, and his perspective on the future of the field.
You have done a lot of research in the past, on a variety of topics including plasmonic nanoparticles. Could you describe your background and how you got into your current field of research?
I gradually found my way into the world of microscopy during my studies in physics at Lawrence University. I had the opportunity to work together with one of the faculty members, Matthew Stoneking, working on plasma confinement. The main focus was on methods of diagnostics and characterization of plasma columns, how they respond to excitement and how they decay.
From there, I moved to UC Riverside, where I continued working with plasma—this time for material synthesis. My research under Prof. Lorenzo Mangolini focused on two key systems: CZTS for photovoltaics, where I used in situ Raman spectroscopy, and zirconium nitride–silicon oxynitride nanoparticles for plasmonic applications, which led me to electron microscopy. This was my first deep dive into TEM, where I not only synthesized materials but also characterized and analyzed them.
I then joined Prof. Peter Bruggeman’s group at the University of Minnesota as a PostDoc, furthering my interest in plasma-driven nanomaterial synthesis. Our team even developed a custom microscopy holder to introduce both liquid and plasma, allowing us to better understand the nanomaterial synthesis processes.
Now, I work as a microscopist at Seagate in Minnesota, specializing in failure analysis of recording heads for hard drives used in data centers and cloud storage. I work in failure analysis with Seagate’s only corrected STEM tool and am also involved in in-house R&D, which keeps me engaged in pushing the boundaries of our capabilities.
You have used TEM research both in academia and in industry; do you feel your approach to experimental design for TEM is different in the two instances?
There’s definitely a big difference in how academia and industry approach experimental design for TEM. In academia, you’re responsible for the entire process—experimental design, sample synthesis, and analysis—so you develop a deep, hands-on understanding of each step. In industry, however, we work as part of a highly specialized team. For example, FIB engineers handle sample preparation, while dedicated analysts focus on microscope data interpretation. This division of expertise allows for more efficient workflows, but I’ve found that certain aspects, like the later stages of FIB preparation, still require the ‘magic touch’ of an experienced operator. Some steps simply can’t be fully automated yet.
From your perspective, what value does in situ TEM offer in your field that might be challenging to achieve through other methods?
As surprising as it may sound, in situ electron microscopy actually saves us a lot of time! We primarily study degradation processes in combined materials and thermal corrosion. Traditionally, we would test dozens of samples on a larger scale, identify failures, and then use TEM to analyze what went wrong. By integrating heating experiments directly into the microscopy process, we can bypass much of that lengthy workflow. Of course, we still validate our findings with bulk-scale studies, but in situ microscopy significantly streamlines the process.
For someone first starting heating and or biasing in situ experiments; are there any considerations when designing experiments to ensure the nanoscale results are relevant?
In my PostDoc, I primarily conducted biasing experiments, integrating liquids and plasma, whereas my current focus is on heating studies. Let me share insights on both.
For biasing studies, two major challenges stood out. First, introducing both gas and liquid into the small sample holder was tricky. We tackled this in two ways: either by bubbling gas through the liquid-filled cell to create a controlled thin liquid layer or by using biasing to generate gas via electrochemical reactions. The latter, however, offered less control over bubble size and gas composition. Overcoming this challenge alone was a significant milestone for us.
The second challenge was plasma ignition, which required 1–5 kV of power. This brought up concerns about current leakage, contact resistance, and circuit calibration. For anyone working with high-current or high-voltage experiments, I’d strongly recommend fully understanding your circuit—where leakage happens and whether your setup is properly calibrated to ensure accurate conditions at the sample.
For our heating experiments, we primarily work with FIB-prepared samples to study defect formation at the nanoscale. One challenge with this approach is that extracting a FIB lamella from the bulk can alter grain interactions, making them behave differently than in the original material. To address this, we encapsulate the lamella in a dielectric material, which helps limit grain movement while still allowing for high-resolution analysis. This method enables us to capture meaningful insights into material dynamics despite the challenges.
Ultimately, I believe that successful experiments come down to thoughtful design. If you clearly define the questions you want to answer, every step—from sample preparation to analysis—should be aligned with that goal.
What role do you see automation, machine learning, or AI playing to make TEM and in situ TEM more accessible to all TEM expertise levels?
In the microscopy lab at Seagate, we see automation in two key areas: sample preparation and data analysis. On the sample preparation side, we use FIB tools that can automate certain basic preparation steps. For both FIB and STEM, we also leverage image analysis software to process large datasets, helping us identify defects more efficiently. This allows for automatic screening of samples before they are passed on to the TEM team for more detailed analysis.
For in situ work, I believe automation can be pushed even further with machine learning and AI. This could enable longer microscope sessions without requiring constant operator supervision. Safety is a key area where AI could make a difference—for example, by detecting anomalies and automatically shutting TEM column valves or cooling the sample to prevent damage. At the very least, it could alert the operator to intervene when necessary. Combining human expertise with machine learning has the potential to significantly enhance both efficiency and safety in microscopy.
How do you stay up to date with the latest developments in the field?
In industry, keeping up with literature tends to be less of a priority compared to academia. Research is usually done on a need-to-know basis—when specific questions arise. Instead, vendors and webinars play a key role in staying informed about the latest developments, often sparking excitement for new capabilities.
Conferences are attended more selectively, with a focus on microscopy advancements and emerging tools that could enhance industry capabilities. The more functionalities we can integrate into our workflow, the greater the impact on efficiency and innovation.
How do you foster collaboration across different research disciplines, and what opportunities are there for interdisciplinary work?
At Seagate, most research and development efforts are self-contained, and interdisciplinary collaboration typically stays in-house unless a specific challenge requires external expertise. Since our materials deposition techniques and sample preparation processes follow similar methodologies, much of the work remains within the company.
That said, collaboration does happen in other ways. We have a strong internship program and actively mentor students, with PhD students often engaging in more fundamental research. Additionally, Seagate has a dedicated research group that collaborates with academic groups on various projects, allowing for deeper exploration of scientific challenges beyond immediate industry applications.
Looking ahead, what advancements or capabilities would you most like to see in (in situ)TEM technology, especially in its application for electronic devices?
Looking ahead, I’d love to see continued advancements in in situ TEM technology, particularly for electronic device applications. One major area for improvement is automation—drift and focus correction still require a lot of manual intervention, which can be tedious. Enhanced correction algorithms would make long-duration experiments much more efficient.
Another exciting direction is expanding in situ capabilities to allow multiple simultaneous interactions with a sample. Right now, advanced holders like Fusion enable heating and biasing, but incorporating additional functionalities—such as tensile testing or magnetic field control—would open up new possibilities. For example, during a webinar with Protochips, Darius Pohl discussed Hall measurements. If we could better control and manipulate magnetic fields within the TEM while also simulating device conditions like heat and excitation, it would be a game-changer for studying hard disc drive storage technologies.
Finally, integrating more benchtop-compatible tools into the TEM workflow would improve efficiency and accessibility, allowing for quicker testing and optimization before committing to high-resolution in situ studies.
Thank you so much Dr. Exarhos for sharing your insights and giving us a really interesting perspective on the future!
- Would you like to learn more about our heating and biasing capabilities for in situ TEM? Find our Fusion AX system here!
- Want to learn more about our automation software? Find our AXON Synchronicity software here!
- Interested in speaking to a sales person? Contact us here!
