Autonomous laboratory uses Bayesian optimization to closed-loop search homogeneous catalysis conditions

1/ AI utilization method: Bayesian optimization-based autonomous experimental agent
2/ Applied task: closed-loop optimization of ligands and reaction conditions for homogeneous catalysis
3/ Reported result: more than 360 experiments improved search efficiency and reported condition-dependent selectivity inversion and scale-up validation
4/ The 2026-09-11 Nature Communications study built an autonomous laboratory for data-driven homogeneous catalysis and used Bayesian optimization to explore ligands and reaction conditions in a closed loop.
5/ The system performed optimization using more than 360 experimental data points, and reported improved search efficiency together with selectivity inversion depending on conditions and scale-up validation.
6/ This is a case in which Bayesian optimization and autonomous experimentation were combined to iteratively explore a catalytic reaction space in practice, showing how experimental automation and AI can be integrated in materials and chemistry research.
7/ However, these results cannot be assumed to reproduce directly in other catalytic systems or with other instrument configurations, and performance may vary depending on the reaction mechanism, measurement noise, and experimental failure rate.
8/ An approach materials researchers can try is to first limit the variables under exploration, then connect an autonomous experiment with a surrogate model to receive recommended next conditions.
9/ If experimental noise and measurement error are not accounted for, a condition that appears optimal may in fact be a coincidence, so independent repetition and external validation are important. [1]