The Ohio State University, Department of Physics, 191 West Woodruff Ave., Columbus, OH 43210 USA
ResCon Technologies, LLC, 1275 Kinnear Rd., Columbus, OH 43212 USA

Growing up, I had a passion for science and spent most early Saturday mornings with my father watching NASA films on the space program. I knew that I wanted to go on to be a rocket engineer! At the same time, I was an avid reader of Scientific American’s Amateur Scientist, which suggested projects you could do on your own with some basic instructions, and a list of parts and vendors. Several projects involved building lasers or using them to study different physical effects, which developed into another passion beyond my interest in space.

In high school, I applied to a few universities to follow my dream of becoming an aerospace engineer but was not accepted or put on the wait list. I was crushed. I then received a Reserve Officer Training Corp scholarship from the U.S. Navy and needed to identify a university that would accept the scholarship as soon as possible. I discovered that the University of Rochester had a program in Optics where I could study lasers – how fun could that be?!

The Institute of Optics gave me many opportunities to learn about lasers and required more “applied” mathematics courses in comparison to other majors on campus. This is where I learned how the language of mathematics could give a greater understanding of science and engineering topics. By my sophomore year, I knew I wanted to go to graduate school and resigned from my scholarship program. Later, I decided to stay at the Institute of Optics for graduate studies.

There, I studied nonlinear optics – a branch of optical physics - with Prof. Robert Boyd. In optical physics, the tradition was that experimentalists were expected to undertake the (usually fairly simple) theoretical research needed to explain their observations. In the mid 1980’s, the topic of chaos was of great interest and many experimental groups were searching for chaos in their favorite system. Prof. Boyd thought it would be a lot of fun to find chaos in one of our nonlinear optical systems to give us an excuse to learn more about the mathematics behind chaos. He would come to the lab every few weeks to see if we had found chaos but I really wasn’t focused on the search.

Finally, I stumbled on a nonlinear optical system that displayed chaos [1] and this led to a search to find the “simplest” nonlinear optical system to display chaos. We found that laser beams counterpropagating through a nonlinear optical medium is enough, displaying pattern formation, multi-stability, and chaos. The figure shows the transverse intensity profile of one of the beams emerging from the medium when the system was above the instability threshold [2].

After completing a post-doctoral research associateship at the University of Oregon where I studied quantum optical systems, I moved to the Physics Department at Duke University. One attraction was the Center for Nonlinear and Complex Systems, a large interdisciplinary group focused on dynamical systems. While I was mainly focused on developing a quantum optics laboratory, interactions with other faculty in the Center convinced me to also setup experiments on controlling chaos in optical and electronic systems. This expanded to work on controlling cardiac dynamics, where I formed a collaboration with biomedical engineers and applied mathematicians. This experience taught me that it is possible to greatly expand your scientific horizons and that you can’t be an expert in everything: collaboration is the key to rapidly move into a new area.

My research expanded to synchronizing chaotic systems – two at first and then to small networks. I was very interested in dynamics on complex networks but was searching for a practical application. In the mid 2000’s, research in machine learning was really taking off and many experimental groups were starting to explore whether physical systems can be used for information processing in general and machine learning specifically. This seemed to be a natural direction that allowed me to continue to study the fundamental behaviors of complex networks as well as satisfying my interest in applications.

In the early 2010’s, one of my collaborators, Prof. Ingo Fischer at the Institute for Cross-Disciplinary Physics and Complex Systems at the Universitat de les Illes Balears, Spain, was very excited about building optical systems for a type of machine learning known as reservoir computing and he infected me with his excitement. A reservoir computer is an artificial recurrent neural network that is especially well suited for “machine” learning of dynamical systems and can be trained rapidly with little data. As earlier in my career, moving into a new direction would give me an excuse to learn about machine learning and the underlying mathematics.

I was using field-programmable gate arrays for studying dynamics of Boolean-like complex networks [3] and this seemed to be a good physical substrate for reservoir computing [4]. Several of my students were also interested and the government agencies funding my research encouraged me to move into this new area. Around this time, I moved to The Ohio State University for family reasons and continued to develop my program on reservoir computing.

As this program was developing, I started to consider commercialization of the ideas being developed and was active in protecting our intellectual property through the on-campus Technology Commercialization Office. This culminated in co-founding ResCon Technologies, LLC [5], which is focused on applying reservoir computing to create “digital twins” of dynamical systems for application in control, integration with larger systems, system health monitoring, and maintenance management. I currently spend half my time with ResCon and the other half with the university.

To commercialize reservoir computing, I realized that I had to make it even simpler. While I had some thoughts along these lines, it was hearing applied mathematician Prof. Erik Bollt’s (Clarkson University) virtual presentation (virtual because of the Covid-19 pandemic) on the success of reservoir computing that helped steer me in the right direction. In collaboration with my (past) student Dr. Aaron Griffith, post-doctoral research associate Dr. Wendson Barbosa, and Prof. Bollt, we developed the “next generation” approach to reservoir computing [6]. (Perhaps you can see a thread in my research over the years of searching for simplicity.) I believe that this result will have both fundamental and practical implications and shows that entrepreneurship activities can flow back to academic activities.

[1] D.J. Gauthier, P. Narum and R.W. Boyd, ‘Observation of deterministic chaos in a phase conjugate mirror,’ Phys. Rev. Lett. 58, 1640 (1987).
[2] D.J. Gauthier, M.S. Malcuit and R.W. Boyd, ‘Polarization instabilities of counterpropagating laser beams in sodium vapor,’ Phys. Rev. Lett. 61, 1827 (1988); D.J. Gauthier, M.S. Malcuit, A.L. Gaeta and R.W. Boyd, ‘Polarization bistability of counterpropagating laser beams,’ Phys. Rev. Lett. 64, 1721 (1990).
[3] D. P. Rosin, D. Rontani, D. J. Gauthier, and E. Schöll, ‘Experiments on autonomous Boolean networks,’ Chaos 23, 025102 (2013).
[4] N. D. Haynes, M. C. Soriano, D. P. Rosin, I. Fischer, D. J. Gauthier, ‘Reservoir computing with a single time-delay autonomous Boolean node,’ Phys. Rev. E 91, 020801 (2015).
[5] https://flyrescon.com/
[6] D. J. Gauthier, E. Bollt, A. Griffith, W. A. S. Barbosa, ‘Next generation reservoir computing,’ Nat. Commun. 12, 5564 (2021).