Active Learning-Guided Polymorph Control in Co-Precipitation Synthesis

Phase control of polymorphic materials is important yet challenging due to the vast synthesis parameter space and the sensitivity of certain phases to specific conditions. Recent advances in integrating artificial intelligence with laboratory automation offer promising solutions to address experimentation challenges like this. In this work, we developed an active learning-guided robotic synthesis workflow to achieve phase control during co-precipitation synthesis. This workflow was demonstrated using FeC₂O₄·2H₂O, a polymorphic compound with diverse applications. The optimal synthesis conditions for obtaining pure α-FeC₂O₄·2H₂O were identified using Bayesian optimization. Building on this, an active learning-guided workflow that can predict phase outcomes based on given synthesis parameters was showcased, enabling more efficient exploration of selective synthesis. The influence of synthesis parameters on the morphology of FeC₂O₄·2H₂O was also preliminarily examined. This study highlights how artificial intelligence with robotic synthesis can accelerate the uncovering of synthesis-phase relationships and advance controllable material synthesis.