Source-linked AI summary
Inorganic Materials Synthesis Planning with Literature-Trained Neural Networks
Edward Kim, Zach Jensen, Alexander van Grootel, Kevin Huang, Matthew Staib, Sheshera Mysore, Haw-Shiuan Chang, Emma Strubell, Andrew McCallum, Stefanie Jegelka, Elsa Olivetti
TL;DR
Because inorganic synthesis knowledge is largely locked in journal text, the paper builds an automated literature-to-synthesis workflow using language processing and a CVAE. It predicts precursors for two later-reported perovskites from pre-2005 routes and screens proposed ABO3 compounds by literature-based synthesizability.
Problem
Inorganic synthesis knowledge is largely locked in journal articles, and no general framework automatically draws synthesis insights from the literature at large.
Source: p. 1Method
The workflow identifies synthesis sections and entities in literature text, then trains an unsupervised CVAE to learn and generate synthesis routes.
Source: p. 1Results
19 of 83 proposed ABO3 compounds were selected by CVAE screening; the model also suggested precursors for InWO3 and PbMoO3 using routes published by 2005.
Source: pp. 2, 4Takeaways & Limitations
Literature-derived synthesis screening adds a practical synthesizability axis that complements existing domain knowledge when filtering theoretically proposed materials.
Source: p. 4The broad route representation omits temperatures, solvents, and morphologies and assumes a one-to-one relation between elements in precursors and targets.
Source: p. 4
Abstract
from arXiv · showhide
Leveraging new data sources is a key step in accelerating the pace of materials design and discovery. To complement the strides in synthesis planning driven by historical, experimental, and computed data, we present an automated method for connecting scientific literature to synthesis insights. Starting from natural language text, we apply word embeddings from language models, which are fed into a named entity recognition model, upon which a conditional variational autoencoder is trained to generate syntheses for arbitrary materials. We show the potential of this technique by predicting precursors for two perovskite materials, using only training data published over a decade prior to their first reported syntheses. We demonstrate that the model learns representations of materials corresponding to synthesis-related properties, and that the model's behavior complements existing thermodynamic knowledge. Finally, we apply the model to perform synthesizability screening for proposed novel perovskite compounds.