12 October 2022 in Communications Materials
A grand challenge of materials science is predicting synthesis pathways for novel compounds. Data-driven approaches have made significant progress in predicting a compound’s synthesizability; however, some recent attempts ignore phase stability information. Here, we combine thermodynamic stability calculated using density functional theory with composition-based features to train a machine learning model that predicts a material’s synthesizability. Our model predicts the synthesizability of ternary 1:1:1 compositions in the half-Heusler structure, achieving a cross-validated precision of 0.82 and recall of 0.82. Our model shows improvement in predicting non-half-Heuslers compared to a previous study’s model, and identifies 121 synthesizable candidates out of 4141 unreported ternary compositions. More notably, 39 stable compositions are predicted unsynthesizable while 62 unstable compositions are predicted synthesizable; these findings otherwise cannot be made using density functional theory stability alone. This study presents a new approach for accurately predicting synthesizability, and identifies new half-Heuslers for experimental synthesis.
22 March 2024 in Integrating Materials and Manufacturing Innovation
15 February 2024 in Journal of Chemical Information and Modeling
30 January 2024 in ChemRxiv
Predicting the synthesizability of a new molecule remains an unsolved challenge that chemists have long tackled with heuristic approaches. Here, we report a new method for predicting synthesizability using a simple, yet accurate thermochemical descriptor. We introduce E_min, the energy difference between a molecule and its lowest energy constitutional isomer, as a synthesizability predictor that is accurate, physically meaningful, and first-principles based. We apply E_min to 134,000 molecules in the QM9 dataset and find that E_min is accurate when used alone and reduces incorrect predictions of "synthesizable" by up to 52% when used to augment commonly-used prediction methods. Our work illustrates how first-principles thermochemistry and heuristic approximations for molecular stability are complementary, opening a new direction for synthesizability prediction methods
24 November 2025 in ECS Meeting Abstracts
Fuel cell electrode catalyst layers are typically manufactured using colloidal ink casting and coating techniques such as slot-die coating or spray coating. While ink deposition and drying require careful optimization of their own, a crucial prerequisite is tuning the catalyst ink to have appropriate rheological and microstructural properties like viscosity, particle/cluster dispersion, and aggregate size and size distribution. Despite studies that have elucidated mechanisms for fine-tuning ink properties for subsequent coating and drying process control, formulating brand new inks with specific properties still requires slow trial-and-error studies. The difficulty of discovering new ink formulations lies within three main challenges: 1) fuel cell catalyst inks have many components (ionomer, catalyst, catalyst support, solvent), leading to a large formulation design space; 2) ink properties are determined by complex multi-scale physics that simultaneously involve almost all components; and 3) ink property optimization is multi-objective in nature, requiring one to make calculated property trade-offs. This work leverages uncertainty-aware machine learning (ML) models to correlate carbon-based fuel cell ink formulation parameters with experimentally measured ink properties. Statistical sampling is first used to define a design of experiments for an initial batch of experiments. This initial dataset is used to begin a multi-objective Bayesian Optimization campaign that designs new experiments to improve ML accuracy and obtain target ink properties. After iterations of experiment-ML development, we report our refined ML models with the elucidated ink material-property correlations.