Self-Driving Laboratory

Self-Driving Laboratory Could Accelerate Chemical Discovery Processes A team of chemical engineering researchers has developed an autonomous laboratory that can identify and optimize new complex multi-step reaction pathways for the synthesis of advanced functional materials and molecules. This system, currently in proof-of-concept demonstration, has found a more efficient method for producing high-quality semiconductor nanocrystals used in optical and photonic devices. Milad Abolhasani, a professor of chemistry and biomolecular engineering at North Carolina State University and author of the study, stated: "Progress in materials and molecular discovery is slow because traditional techniques for discovering new chemistry rely on changing one parameter at a time using siloed processes in chemistry and materials science laboratories." Abolhasani added: "If complex chemistry involves dozens of parameters, developing a new target material or devising a more efficient way to produce a desired chemical could take decades. Our system, called AlphaFlow, when combined with automated microfluidic devices, uses an artificial intelligence technique called reinforcement learning, which accelerates the materials discovery process. We demonstrated that AlphaFlow could conduct more than 100 experiments in the same timeframe as 100 human chemists, using less than 0.01% of the relevant chemicals. This effectively miniaturizes experiments and performs the same laboratory operations that would require an entire wet chemistry laboratory on an end-to-end experimental platform the size of a bag. This is extremely efficient." AlphaFlow's artificial intelligence model determines which experiment to conduct based on two factors: data developed from experiments it has already run and predictions about what the results of the next few experiments will be. Amanda Volk, first author of the paper and a doctoral student at NC State, stated: "We use this moving window of previous action steps and the predicted outcomes of future action sequences to inform AlphaFlow's decision-making process. From this, AlphaFlow can account for actions with delayed effects and simultaneously direct the decision-making process in real time according to the latest experimental results," she added, noting "Essentially, the system has the ability to learn instantly from unexpected results and adapt to them." This is true whether the system focuses on discovering a new chemical or optimizing a production process for a known chemical. The difference is that for discovery, the system attempts to determine which precursors need to be added to find the best-performing chemistry and the optimal order in which to add them. For optimization, the artificial intelligence model already knows which precursors must be added and in what order. As a result, AlphaFlow's optimization focus means determining how much of each precursor to use and the time required for each reaction to reach optimum performance most efficiently. Abolhasani stated: "This integration of artificial intelligence and chemistry reduces the time required to develop new chemistries by at least an order of magnitude," adding "Think in terms of hours rather than months or years." Volk noted: "AlphaFlow also provides new insights into fundamental chemistry. For example, in a proof-of-concept demonstration, AlphaFlow developed a new method for producing a semiconductor nanocrystal with a cadmium selenide core and cadmium sulfide shell. These nanocrystals are used in photonic and optical technologies. The new chemistry discovered by AlphaFlow involves fewer steps than previously human-discovered chemistry, making the process more efficient. Additionally, one of the steps eliminated by AlphaFlow was previously believed to be a critical step in such multi-step chemistry, which was surprising. Being able to produce the same high-quality nanocrystal without this step expands our understanding of the relevant chemistry." Abolhasani added: "In fact, AlphaFlow showed that a step researchers thought was critical was unnecessary, and developed this more efficient chemistry—which changes what we thought we knew about the multi-step chemistry of core/shell semiconductor nanocrystals—in just 30 days of continuous operation compared to 15 years of academic literature." Currently AlphaFlow is configured to conduct experiments related to colloidal atomic layer deposition. This type of multi-step chemistry is particularly experimentally challenging because it involves many different parameters, potentially accounting for more than 40 variables in this context. Professor Milad Abolhasani stated: "However, AlphaFlow can be modified to conduct any series of experiments involving the performance of chemical reactions in solution." Volk noted: "AlphaFlow is the first example we know of that integrates reinforcement learning with an autonomous laboratory, and it emphasizes how much physical sciences can benefit from artificial intelligence." Researchers are now seeking partners in both the research community and the private sector to begin using AlphaFlow to address challenges in the field of chemistry. Abolhasani said: "Ideally, we want to reach a point where multiple AlphaFlow platforms are used to address different large-scale challenges related to energy transition and sustainability, but share data that would enable everyone to discover and develop new materials and molecules faster. AlphaFlow is open source. We believe it is important to share high-quality, reproducible, standardized experimental data from both failures and successes. We think this is important because we want to accelerate the discovery of new materials and chemical processes," he stated. Source: The paper appears in Nature Communications. More information: "AlphaFlow: Autonomous Discovery and Optimization of Multi-Step Chemistry using a Self-Driven Fluidic Lab Guided by Reinforcement Learning" Nature Communications, DOI: 10.1038/s41467-023-37139-y. www.nature.com/articles/s41467-023-37139-y Journal information: Nature Communications Provided by North Carolina State University / https://phys.org/news/2023-03-self-driven-laboratorychemical-discovery.html