December 1 - 6, 2024
Boston, Massachusetts

Event Supporters

2024 MRS Fall Meeting & Exhibit
MT04.01.06

Text-Mined Dataset of Solid-State Syntheses with Impurity Phases Using LLM

When and Where

Dec 2, 2024
11:45am - 12:00pm
Hynes, Level 2, Room 210

Presenter(s)

Co-Author(s)

Sanghoon Lee1,2,Kevin Cruse1,2,Viktoriia Baibakova1,2,Gerbrand Ceder1,2,Anubhav Jain2

University of California, Berkeley1,Lawrence Berkeley National Laboratory2

Abstract

Sanghoon Lee1,2,Kevin Cruse1,2,Viktoriia Baibakova1,2,Gerbrand Ceder1,2,Anubhav Jain2

University of California, Berkeley1,Lawrence Berkeley National Laboratory2
Solid-state synthesis is widely used to obtain various inorganic materials, such as battery materials and thermoelectrics. Despite its prevalence, the process remains a "black box" due to the lack of a general theory and well-understood underlying reaction mechanisms<sup>1,2</sup>, thereby posing inherent challenges to controlling the outcome.<br/><br/>Recent advances in machine learning have motivated data-driven approaches in materials synthesis and control, which require comprehensive synthesis datasets. While prior works have successfully extracted structured datasets from literature, they often neglect or overlook product phase purity or yield<sup>3</sup>. Traditionally, the formation of impurity phases has been viewed as undesirable or indicative of failed synthesis<sup>4</sup>. This can lead to two issues: a lack of negative data points if phase-impure syntheses are filtered out, and inaccurate results if phase-impure syntheses are incorrectly classified as phase-pure.<br/><br/>In this work, we construct a structured solid-state synthesis dataset that includes impurity phases via information extraction with LLM. We investigate the formation of these impurity phases and their interpretability using Materials Project phase diagrams. Our findings provide valuable insights into the conditions leading to impurity phase formation, contributing to a more comprehensive understanding of solid-state synthesis processes and paving the way for improved synthesis control.<br/><br/>References<br/><sup>1</sup>Kohlmann, H. Looking into the black box of solid-state synthesis. Eur. J. Inorg. Chem. 2019, 4174–4180, https://doi.org/10.1002/ejic.201900733 (2019).<br/><sup>2</sup>Aykol, M., Montoya, J. H. & Hummelshøj, J. Rational solid-state synthesis routes for inorganic materials. J. Am. Chem. Soc. 143, 9244–9259, 10.1021/jacs.1c04888 (2021).<br/><sup>3</sup>Kononova, O. et al. Text-mined dataset of inorganic materials synthesis recipes. Sci. Data 6 10.1038/s41597-019-0224-1 (2019)<br/><sup>4</sup>Raccuglia, P. et al. Machine-learning-assisted materials discovery using failed experiments. Nature 533, 73–76, 10.1038/nature17439 (2016).

Keywords

chemical synthesis | second phases

Symposium Organizers

Kjell Jorner, ETH Zurich
Jian Lin, University of Missouri-Columbia
Daniel Tabor, Texas A&M University
Dmitry Zubarev, IBM

Session Chairs

Kjell Jorner
Jian Lin

In this Session