‘Meet the Researcher’ is a series in which we spotlight different researchers in academia who use NVIDIA technologies to accelerate their work.
This month we spotlight Dr. Emanuel Gull, Associate Professor of Physics at University of Michigan, whose research focuses on the development of theoretical and computational methods for strongly correlated quantum systems.
Gull is the recipient of a Sloan Research Fellow, Ralph E. Powe Junior Faculty Enhancement Award, DOE Early Career Research Award, SCES early career Nevill F. Mott Prize, and APS Outstanding Referee Program.
What are your research areas?
The physics of materials in which many quantum particles strongly interact with each other. These are the systems out of which we build our newest generation of magnets, superconductors, solar cells, and systems for standard approximative methods.
When did you know that you wanted to be a researcher and pursue this field?
I was always open to having a career in the software/computing side of industry/finance — but, when I had to decide whether to go for a postdoc, the financial crisis hit. Instead, I did a postdoc in the U.S. and managed to get hired into an academic position afterwards.
What motivated you to pursue your recent research area of focus?
‘Quantum’ theory is the reason why many of our recent technological breakthroughs work. After all, NVIDIA chips are just an application of quantum theory. However, taking just theory and predicting and improving material properties without further input is incredibly difficult, even though we believe we understand the theory very well. I have always been fascinated by the challenge of combining computers and theoretical methods to bring calculations closer to reality. This started with an internship I did at a high performance computer center back when I was a high school student.
What problems or challenges does your research address?
While we know the equations that govern the physics of systems with many interacting quantum particles well, they are impossibly difficult to solve. This is why we need to find approximations that are both numerically tractable and accurate. My research spans the entire gamut from theoretical derivations, to implementation of new algorithms, to HPC, to comparisons with experiments. All of my research aims to make quantum theories more predictive and more accurate.
What challenges did you face during the research process, and how did you overcome them?
Time management is probably the most crucial. It’s easy to have many ideas, but testing them, improving upon them, and revising them takes time. In research, you’re constantly juggling finding resources, training people, having and revising ideas, publishing, going to conferences, etc. Finding quiet intervals to work deeply on a problem is essential, but difficult. I don’t believe I’ve overcome that limitation.
What is the impact of your work on the field/community/world?
Stronger magnets, higher temperature superconductors, and better materials for sensors and chips.
How have you used NVIDIA technology either in your current or previous research?
Yes! In fact, our home-written ab-initio simulation toolkit uses NVIDIA codes to simulate the physics of real materials and their excitations. Most of our calculations would be either impossible or borderline without the NVIDIA fast and double-precision arithmetics on the V100 and A100. Our codes run at just about 50% of theoretical peak flop, and are parallelized with streams within each GPU and with MPI between different GPUs and nodes.
What research breakthroughs or interesting results can you share?
We did, and we’re just now writing a paper on a new high-temperature superconductor.
What’s next for your research?
We’re currently doing a big push for driving systems out of equilibrium. We’re exciting them with a laser, ‘quenching’ them with a short current pulse, or probing them in other nonequilibrium conditions. The nonequilibrium physics of quantum materials is very different from the equilibrium conditions, and many exciting new phenomena appear. Besides, most sensors work out of equilibrium. How to generalize our computational toolkit to these situations is currently an open question that we’re working on.
Any advice for new researchers, especially to those who are inspired and motivated by your work?
Ask the big questions. Why is this interesting? Why will it work or how will it not work? What have we learned if it does work? But, don’t lose sight of the small details. Pounce at the details that don’t quite make sense, that’s where there’s something that needs to be understood. When something turns into a dead end, learn to let it go (even if you’ve invested a lot of resources into it).
Also, know the established ways of thinking about a problem, but question them always. Know the limitations of your tools and theories, and invest in your toolkit. New tools (computer codes, theoretical methods, experimental setups) lead to new discoveries, so make sure you have the best ones available for your application.