Representation of Inorganic Synthesis Reactions and Prediction: Graphical Framework and Datasets
Samuel Andrello, Daniel Alabi, Simon J. L. Billinge
TL;DR
This work introduces ActionGraph, a graph-based representation of inorganic synthesis pathways that jointly encodes precursors, intermediate operations, and final products. By transforming 13,017 text-mined solid-state reactions from Materials Project into ActionGraph graphs and applying a PCA-reduced adjacency in a k-nearest neighbors framework, the authors demonstrate improvements in predicting synthesis operations and substantially higher operation-length fidelity compared to a baseline composition-only approach. The results reveal a trade-off: precursor prediction benefits primarily from compositional information, while operation sequencing and procedural fidelity gain from richer structural encoding. The study highlights the potential of graph-based synthesis planning for accelerating discovery in inorganic materials, and outlines directions for richer data, improved featurization, and closed-loop experimentation.
Abstract
While machine learning has enabled the rapid prediction of inorganic materials with novel properties, the challenge of determining how to synthesize these materials remains largely unsolved. Previous work has largely focused on predicting precursors or reaction conditions, but only rarely on full synthesis pathways. We introduce the ActionGraph, a directed acyclic graph framework that encodes both the chemical and procedural structure, in terms of synthesis operations, of inorganic synthesis reactions. Using 13,017 text-mined solid-state synthesis reactions from the Materials Project, we show that incorporating PCA-reduced ActionGraph adjacency matrices into a $k$-nearest neighbors retrieval model significantly improves synthesis pathway prediction. While the ActionGraph framework only results in a 1.34% and 2.76% increase in precursor and operation F1 scores (average over varying numbers of PCA components) respectively, the operation length matching accuracy rises 3.4 times (from 15.8% to 53.3%). We observe an interesting trade-off where precursor prediction performance peaks at 10-11 PCA components while operation prediction continues improving up to 30 components. This suggests composition information dominates precursor selection while structural information is critical for operation sequencing. Overall, the ActionGraph framework demonstrates strong potential, and with further adoption, its full range of benefits can be effectively realized.