Autonomous and dynamic precursor selection for solid-state materials synthesis

Abstract

Solid-state synthesis plays an important role in the development of new materials and technologies. While in situ characterization and ab-initio computations have advanced our understanding of materials synthesis, experiments targeting new compounds often still require many different precursors and conditions to be tested. Here we introduce an algorithm (ARROWS3) designed to automate the selection of optimal precursors for solid-state materials synthesis. This algorithm actively learns from experimental outcomes to determine which precursors lead to unfavorable reactions that form highly stable intermediates, preventing the target material’s formation. Based on this information, ARROWS3 proposes new experiments using precursors it predicts to avoid such intermediates, thereby retaining a larger thermodynamic driving force to form the target. We validate this approach on three experimental datasets, containing results from over 200 synthesis procedures. In comparison to black-box optimization, ARROWS3 identifies effective precursor sets for each target while requiring substantially fewer experimental iterations. These findings highlight the importance of domain knowledge in optimization algorithms for materials synthesis, which are critical for the development of fully autonomous research platforms.

Accelerating Materials Discovery: Automated Identification of Prospects from X‐Ray Diffraction Data in Fast Screening Experiments

“…In Table S3, Supporting Information, we compared our developed network to a similar network presented by Szymanski et al for the identification of XRD patterns. [ 13,19 ] Our model performed better on both presented datasets. While the performance of their model nearly matches ours on the spinels dataset, the network by Szymanski and colleagues fails to successfully extract the almost overlapping peaks that appear in the doped copper oxides dataset, resulting in considerably worse performance metrics.…”

“…[16][17][18] Additionally, Szymanski et al deployed a neural network to identify target and intermediate phases in material synthesis experiments, enabling their optimization algorithm to determine the most suitable precursors and experimental parameters for the effective synthesis of the target phase. [19] To effectively train neural networks for phase identification in XRD datasets, previous research has predominantly utilized simulated training data. [9][10][11][12][13][14][15][16][17][18][19] This approach involves generating synthetic diffraction patterns from crystallographic database entries, incorporating variations and experimental artifacts characteristic of actual experimental patterns, to ensure that models trained on simulated data effectively transfer their performance to actual experimental scans.…”

“…[19] To effectively train neural networks for phase identification in XRD datasets, previous research has predominantly utilized simulated training data. [9][10][11][12][13][14][15][16][17][18][19] This approach involves generating synthetic diffraction patterns from crystallographic database entries, incorporating variations and experimental artifacts characteristic of actual experimental patterns, to ensure that models trained on simulated data effectively transfer their performance to actual experimental scans. The primary aim of this methodology is to address the difficulty of obtaining a sufficiently large dataset containing high-quality XRD scans with their specific phase identification results, crucial for training the neural network.…”

“…[ 16–18 ] Additionally, Szymanski et al deployed a neural network to identify target and intermediate phases in material synthesis experiments, enabling their optimization algorithm to determine the most suitable precursors and experimental parameters for the effective synthesis of the target phase. [ 19 ]…”

Text Mining the Literature to Inform Experiments and Rationalize Impurity Phase Formation for BiFeO3

“…A significant bottleneck in this effort is the acquisition of sufficient data. The rapid gathering of relevant synthesis data can be accomplished directly through autonomous, high-throughput synthesis, − where a synthesis machine learns optimal synthesis conditions for a specific target material or property by taking patterns from historical syntheses and their results into account. Existing studies show the promise of autonomous synthesis in accelerating the drive toward efficient materials discovery, though there are still pitfalls such as a need for condition initialization and informing experiments based on historical data .…”