University of Tennessee Positioned as Leader in Automated Manufacturing and Characterization

Automated characterization is necessary to close the loop between synthesis and discovery of new materials. Through its award-winning faculty, state-of-the-art facilities, and location, the University of Tennessee, Knoxville is at the forefront of this transformative process.

UT's framework serves as a model for how machine learning-driven automated materials discovery and scalable experimentation can shorten the path from concept to deployment, advancing the nation's competitiveness in nuclear, aerospace, and advanced manufacturing technologies.

UT is one of only three programs in the country that works on automated characterization, alongside Oak Ridge National Laboratory (ORNL) and Pacific Northwest National Laboratory (PNNL). The practice accelerates the materials discovery and development process, which can be time consuming and expensive, by using automated hardware that can execute experiments, collect and analyze data, and interpret results.

"These capabilities position UT as a leader in translating fundamental materials innovation into scalable, real-world applications," said Mohsen Asle Zaeem, the department head of the Department of Materials Science and Engineering(MSE). "The integration of cloud-enabled microscopy, automated synthesis, and high-performance computing directly supports the development of next-generation materials for nuclear energy systems, aerospace structures and systems, and national defense."

As one of the first institutions in the US to work on high-throughput automated synthesis and characterization, UT is helping advance the capabilities of AI materials discovery while training the next generation of the workforce, scientists, and engineers.

"It doesn't matter how much stuff you make if you cannot characterize it," said MSE Weston Fulton Professor Sergei Kalinin. "Characterization is the missing link from the accelerated material discovery in the future. The unique piece of UT is being the place where the accelerated characterization is being done."

Bringing AI to the Labs

UT has a long tradition of excellence in the field of materials characterization, spearheaded by faculty members like MSE Professor Gerd Duscher and Dayakar Penumadu, the Fred N. Peebles Professor and IAMM Chair of Excellence.

Currently, UT is the only academic institution that has built fully cloudified electron and scanning probe microscopy and nanoindentation. These instruments can stream imaging data to the cloud, analyze it, and return control commands back to the instrument. This enables the bridge between the ML-agents and real-world instrumentation, opening the door to the ML algorithms to act on the nanometer and atomic scales.

UT is also home to one of the first automated high throughput synthesis labs in the country, directed by Associate Professor Mahshid Ahmadi, and combinatorial research labs, directed by Philip Rack, MSE professor, Leonard G. Penland Chair, and Director of he Center for Materials Processing (CMP).

"The uniqueness of our approach at UT is that we always start bottom up and not the top down," Kalinin said. "We don't come up with the idealized suggestion, and try to squeeze it into real world, which is the reason why things seldom work for machine learning in real world applications. What we try to do is start with the real tools in the lab and make them faster, make them talk to each other faster, make them solve the problems faster."

Ahmadi has been working at the intersection of AI and autonomous material synthesis and characterization since 2019 and is one of the pioneers in adapting self-driving labs (SDLs) trying to harness the power of SDL technologies to revolutionize the approach to the world's energy challenges.

Her lab in the Institute of Advanced Materials and Manufacturing (IAMM) houses three liquid handling robots capable of printing optoelectronic materials, as well as a fully automated spin bot system capable of spin coating 10 films and perform in-situ characterization of optoelectronic materials, all integrated with large language model (LLM) agents. These uniquely developed tools are not only essential for understanding how materials grow and crystallize but also enable unprecedented insights into their fundamental properties.

"It's one of the most unique labs in the nation that is fully automated," said Ahmadi, who focuses on materials for energy and solar application. "We're now capable of integrating LLM agents and the use of ML for decision making to help us close the loop of hypothesis formation, experimental planning, data analysis and provide the feedback in the loop and then do a much more intelligent decision making in order to optimize or discover materials."

Uniquely Positioned for Innovation

UT's microscopy facilities at IAMM are all combined to have a uniform platform so that machine learning tools can be deployed to all of them. Among the tools are automated microscopes, scanning probe microscopes, electron microscopes, ion microscopes, and a transmission electron microscope that has a resolution of 42 picometers–approximately one trillionth of a meter.

The transmission electron microscope is controlled by a supercomputer. The facility is also planning to control the scanning electron microscope and focused ion beam microscope with the same platform and machine learning tools.

"The supercomputer takes images and looks at various specific features in that image. Then, you can jump around depending on what you find of interest," said Duscher, whose research areas includes atomic and electronic structure on the atomic level. "That means that you are concentrating more on scientific questions instead of the operation of the instrument. The performance is much quicker and efficient, because the machine is often much quicker to select specific points of and characterize under very small areas on a large sample."

Through the university's support and faculty expertise, Duscher believes UT is ideally suited to take advantage of the resources.

"The key thing is we not only have these instruments, but we also have the environment for these microscopes to operate all this up to their specifications," Duscher said. "The facilities we have at UT are extremely good and, combined with the knowledge we have here in the area, we are very uniquely positioned to do microscopy and machine learning together on a world-class level."

Accelerating Future Discovery

The infrastructure in place at UT enhances its position as a national hub for advanced materials research and innovation.

"The collaboration between UT, ORNL, and local and statewide industry partners is driving economic and workforce development across East Tennessee and the broader state," Asle Zaeem said. "These efforts directly support Tennessee's priorities in technology-based growth, manufacturing competitiveness, and energy security."

The research being performed at UT can lead to the discovery of new materials and possibly change the properties of known materials for new applications. The impact spans from lightweight alloys for aircraft engines to heat shields of spacecrafts to quantum sensors.

Before the current push to integrate AI into materials discovery, UT had the foresight that transformation was necessary to spark innovation.

"Materials science doesn't usually change over time. But now it's a different time. I can even say that it's the inflection point," Ahmadi said. "We started, we learned from the history, and we had this vision early on. Being early is very important because we have gained knowledge through experience and are close to tackling all the challenges."

Provided by University of Tennessee at Knoxville