‘Self-Driving’ Lab Speeds Up Research, Synthesis of Energy Materials

Conceptual image of the self-driving lab

This is a modified version of an article written by Matt Shipman, Research Lead in University Communications.

Metal halide perovskites (MHPs) are an unconventional family of crystalline materials with continuously expanding compositional and structural spaces. Their nanocrystals are an emerging class of semiconductor materials that, because of their solution-processability and unique size- and composition-tunable properties, are thought to have potential for use in printed photonic devices and energy technologies including solar cells, luminescence solar concentrators, light-emitting diodes (LEDs), and photodetectors.

Professor Milad Abolhasani
Prof. Milad Abolhasani

Professor Milad Abolhasani and his colleagues have developed and demonstrated a ‘self-driving lab’ that uses artificial intelligence (AI) and fluidic systems to advance our understanding of MHP nanocrystals. This self-driving lab can also be used to investigate a broad array of other semiconductor and metallic nanomaterials.

For their proof-of-concept demonstrations, the researchers focused on cesium lead halide (CsPbX3, X=Cl, Br), an all-inorganic MHP nanocrystal.

“We’ve created a self-driving laboratory that can be used to advance both fundamental nanoscience and applied engineering,” says Prof. Abolhasani, corresponding author of a paper on the work.

Solution-processed materials are materials that are made using liquid chemical precursors, including high-value materials such as quantum dots, metal/metal oxide nanoparticles and metal organic frameworks. And because they can be made using solution processing, they have the potential to be made in a cost-effective way.

However, MHP nanocrystals are not in industrial use yet.

“In part, that’s because we’re still developing a better understanding of how to synthesize these nanocrystals in order to engineer all of the properties associated with MHPs,” Abolhasani says. “And, in part, because synthesizing them requires a degree of precision that has prevented large-scale manufacturing from being cost-effective. Our work here addresses both of those issues.”

The new technology expands on the concept of the Artificial Chemist 2.0, which Abolhasani’s lab unveiled in 2020. The Artificial Chemist 2.0 is completely autonomous, and uses AI and automated robotic systems to perform multi-step chemical synthesis and analysis. In practice, that system focused on tuning the properties of MHP quantum dots, allowing users to go from requesting a custom quantum dot to completing the relevant R&D and beginning manufacturing in less than an hour.

“Our new self-driving lab technology can autonomously dope MHP nanocrystals, adding manganese atoms into the crystalline lattice of the nanocrystals on demand,” Abolhasani says.

Doping the material (introduction of impurities into an intrinsic semiconductor for the purpose of modifying some of its properties) with varying levels of manganese changes the optical and electronic properties of the nanocrystals and introduces magnetic properties to the material. For example, doping the MHP nanocrystals with manganese can change the wavelength of light emitted from the material.

Quantum dots with different colors
Quantum Dots with Different Colors

“This capability gives us even greater control over the properties of the MHP nanocrystals,” Abolhasani says. “In essence, the universe of potential colors that can be produced by MHP nanocrystals is now larger. And it’s not just color. It offers a much greater range of electronic and magnetic properties.”

The new self-driving lab technology also offers a much faster and more efficient means of understanding how to engineer MHP nanocrystals in order to obtain the desired combination of properties. Video of how the new technology works can be found at https://www.youtube.com/watch?v=2BflpW6R4HI.

“Let’s say you want to get an in-depth understanding of how manganese-doping and bandgap tuning will affect a specific class of MHP nanocrystals, such as CsPbX3,” Abolhasani says. “There are approximately 160 billion possible experiments that you could run, if you wanted to control for every possible variable in each experiment. Using conventional techniques, it would still generally take hundreds or thousands of experiments to learn how those two processes – manganese-doping and bandgap tuning – would affect the properties of the cesium lead halide nanocrystals.”

But the new system does all of this autonomously. Specifically, its AI algorithm selects and runs its own experiments. The results from each completed experiment inform which experiment it will run next – and it keeps going until it understands which mechanisms control the MHP’s various properties.

“We found, in a practical demonstration, that the system was able to get a thorough understanding of how these processes alter the properties of cesium lead halide nanocrystals in only 60 experiments,” Abolhasani says. “In other words, we can get the information we need to engineer a material in hours instead of months.”

While the work demonstrated in the paper focuses on MHP nanocrystals, the autonomous system could also be used to characterize other nanomaterials that are made using solution processes, including a wide variety of metallic and semiconductor nanomaterials.

“We’re excited about how this technology will broaden our understanding of how to control the properties of these materials, but it’s worth noting that this system can also be used for continuous manufacturing,” Abolhasani says. “So you can use the system to identify the best possible process for creating your desired nanocrystals, and then set the system to start producing material nonstop – and with incredible specificity.

“We’ve created a powerful technology. And we’re now looking for partners to help us apply this technology to specific challenges in the industrial sector.”

The paper, “Autonomous Nanocrystal Doping by Self-Driving Fluidic Micro-Processors,” is published open access in the journal Advanced Intelligent Systems. The paper was co-authored by Fazel Bateni, a Ph.D. student in Prof. Abolhasani’s research group; Robert Epps and Jeffery Bennett, postdoctoral researchers in the group; Kameel Antami, a Ph.D. graduate from the group; Rokas Dargis, a CBE undergraduate member of the group; and Kristofer Reyes, an assistant professor at the University at Buffalo.

The work was done with support from the National Science Foundation, under grant number 1940959, and from the UNC Research Opportunities Initiative.