Jean Anne Incorvia works as an Assistant Professor and Fellow of Advanced Micro Devices (AMD) in Computer Engineering in the Department of Electrical and Computer Engineering at The University of Texas at Austin. She and her research group specialize in designing and fabricating electronics that use magnetism and spin, called spintronics, that will improve in-memory computing and computing in extreme conditions. We spoke with Jean Anne about her experience at the University and her latest scientific findings – particularly the development of practical nanodevices for the future of computing using emerging physics and materials that will assist with memory.

‍

What is your research on, and what is your innovation?

My group works on realizing new physics for computing, which is quite frankly a really broad topic that covers an even wider ranger of technologies and sciences. As such, we take a very vertical approach to our research. This means that we do fundamental work on quantum transport and new materials, as well as creative design of new nanodevice prototypes that use these new materials, and then we actually must be able to make, test, and understand these new prototypes’ functions while simultaneously ensuring that we are not simply stopping at that device level and push our technologies into the circuit or systems level. At these higher levels, we need to be able to develop and design these circuits or system models informed by our experiments, enabling us to better understand how we can best use these new materials to improve computing capacity.

How can society benefit from your research?

In terms of specific applications, we have a big focus on neuromorphic or brain-inspired computing. Traditional computers are really good at doing certain things, such as adding large numbers or handling basic math. Unfortunately, basic computers are not good at doing more unstructured tasks, such as recognizing images or learning from their environments. Luckily, however, these are the exact types of jobs that our brains are excellent at handling. We hope that we can translate those behavioral tasks into our new materials and systems. 

Furthermore, we work in a similar but different area called in-memory computing. Currently, our traditional computers have one significant bottleneck, which is between computing and memory. To remove that bottleneck, there has already been a lot of work done to bring these two aspects closer together. Our team, however, is seeking to go even further by having them on one single device to handle both computing and memory tasks. This could be particularly applicable for edge computing and even more so for energy-constrained or extreme environments, where you really cannot afford to have multiple devices doing an assortment of tasks while remaining energy-efficient. For example, we have done some projects to replace current technology in either zero-gravity or high-radiation environments, which could prove particularly useful for our technologies because materials like silicon do not act reasonably in these extreme environments. 

In what way can magnetic and 2D materials improve on existing technology?

In terms of the materials front, we have primarily been focused on spintronic and magnetic base materials. These materials are extremely promising because there are just so many rich dynamics at the nano-scale. At this scale, we can consider magnetic field interactions, oscillatory dynamics and can observe magnetization movement in time. It is even more interesting that in this arena, we can create coupled dynamics, whereby one magnet will appear oscillating while another placed next to it will begin to oscillate in synchronized tandem to the other. Frankly, there is a lot of super-rich physics that we can play around with and interact with at this scale, making it possible to study these emerging behaviors and how we can make use of them. A specific example of something we’ve done was observing these effects in 2D materials, such as graphene, to generate a spontaneous spin current that can be used in our applications.

Does this mean that your work can be coupled with AI and Deep Learning?

Yes! Some of our projects are very much in the AI space. We currently have a paper under review where we show in simulation that we can use these thin topological textures, called skyrmions, for context-aware data analysis of images. Specifically, we took this technology and applied it to breast cancer analysis. We created a model based on different risk factors of a patient’s history that changed our designed systems’ configuration and allowed us to analyze existing tumor data more efficiently with our new hardware made from skyrmions. We have done more large-scale simulations and AI-based analyses with these new materials but are eager to expand their application further.

‍

What do you think are the biggest hurdles to energy-efficient computing?

There are so many things, but the biggest hurdle is that we continue to take a system-level approach to energy-efficient computing. It is no longer acceptable to simply wonder whether a widget will switch with low current or low voltage in this day and age. Instead, we need to consider how these material technologies will be engineered in a circuit and a system and make sure we are aware of what new problems might arise that we need to address to make sure the system is energy efficient. Furthermore, we need to make sure we choose the proper application that will showcase these materials and that the application’s unique properties fit the specific design. This requires interdisciplinary research. For example, we cannot simply have physicists only studying the materials while the architects or computer scientists only handle design architecture. There needs to be a shared comprehension between these groups, and it has been surprising how different the understanding barriers can be between those varying industries.

Therefore, the obstacle I’ve been working to overcome the most are these language barriers that prevent full cooperation and comprehension. This required me to work with people at the circuits and systems levels and work closely with people in neuroscience since these are the individuals who understand the brain. This self-cross-pollination has meant that these scientists can talk to me and translate their findings into something tangible, which is only possible because I took the time to learn their language. It’s when these groups can come together that great things are possible, which is why I have sought to facilitate this interdisciplinary understanding because it overcomes one of the biggest obstacles to sustained growth.

One other big hurdle is making sure that we continue to get the funding for this interdisciplinary research. It can be challenging to train new researchers or students because it does require a lot of knowledge and can be difficult for many to grasp right off the bat. Maintaining that excitement and funding is critical. Although this field can be technically challenging, it can be creative, fun and super rewarding!

What do you think the timescale is for seeing these magnetic and 2D materials in our commercial electronics?‍

This is a continual process, and even if we’re looking at past technologies, we can see that from when an idea was first imagined to when it becomes mainstream had a lot of growth points over time. Magnetic materials are already used in hard disk drives, which is a massive industry for data centers. A newer technology that is already in production is called spin-transfer torque magnetic random access memory, shortened to STT-MRAM, which has immense market potential because it is a nonvolatile memory that can be switched electrically, giving it the best of both worlds, where it maintains the stability and endurance of hard disk drives while also not requiring any moving parts since everything inside can be switched electronically. 

These technologies are continuously coming out, and our team is constantly thinking about what’s next for the future. We’ve even started thinking in terms of bio-inspired brain applications with 2D materials. We still haven’t landed on a timescale for this application since the material demands are unclear. Other emerging technologies, however, are much further along in this area and are coming into production. This includes phase-change memory and resistive random access memory. More research is also being done on how 2D materials can assist in computing, but again, these will come out slowly over time, and more research is required. That being said, some of the nanodevices we’re currently working on, such as the Domain Wall Magnetic Tunnel Injunction, are already showing immense potential in improving computing capacity and could represent the first steps to the technologies we may see ten years from now.