MIT Libraries logoDSpace@MIT

MIT
View Item 
  • DSpace@MIT Home
  • MIT Libraries
  • MIT Theses
  • Doctoral Theses
  • View Item
  • DSpace@MIT Home
  • MIT Libraries
  • MIT Theses
  • Doctoral Theses
  • View Item
JavaScript is disabled for your browser. Some features of this site may not work without it.

Accelerating Novel Energy Catalyst Discovery Using Automation, Active Learning, and AI

Author(s)
Ren, Zhichu
Thumbnail
DownloadThesis PDF (53.84Mb)
Advisor
Li, Ju
Terms of use
Attribution-NonCommercial 4.0 International (CC BY-NC 4.0) Copyright retained by author(s) https://creativecommons.org/licenses/by-nc/4.0/
Metadata
Show full item record
Abstract
The discovery of novel energy catalysts is a critical challenge in the field of materials science. Traditional methods for materials discovery are labor-intensive and time-consuming, hindering the rapid development of new catalysts. To address this issue, we introduce a comprehensive approach that integrates automation, active learning, and artificial intelligence (AI) to accelerate the discovery process. Our approach introduces the Copilot for Real-world Experimental Scientist (CRESt) system, which combines a large multimodal model (LMM) with an active learning-guided robotic system. CRESt streamlines the workflow of composition selection, high-throughput materials synthesis, electrochemical screening and characterization for the optimization of high-entropy alloy catalysts. The system allows researchers, regardless of their programming skills, to interact with the robotic platform using voice commands, making it highly accessible and user-friendly. We demonstrate the effectiveness of our approach by experimentally exploring over 700 chemistries and 1300 samples. The optimized 8-dimensional alloy (Pd-Pt-Cu-Au-Ir-Ce-Nb-Cr) achieved approximately 10 times the cost-specific performance of commercial catalysts for the direct formate fuel cell. This breakthrough highlights the potential of our approach to accelerate the discovery of novel energy catalysts across various domains. Furthermore, we discuss the challenges and considerations associated with implementing active learning in real-world experiments. We provide guidance on addressing model-centric and data-centric issues, such as model customization and data irreproducibility, to ensure the successful application of active learning in materials research projects. Looking ahead, we explore the role of human experimentalists in the era of AI-driven discovery. While AI and automation are poised to transform many aspects of experimental research, we argue that human experimentalists remain irreplaceable for now. Our ability to exercise critical thinking and engage in complex real-world interactions sets us apart from abiotic intelligence. However, as AI becomes more deeply integrated into research practices, the experimental landscape is bound to undergo significant changes.
Date issued
2024-05
URI
https://hdl.handle.net/1721.1/162738
Department
Massachusetts Institute of Technology. Department of Materials Science and Engineering
Publisher
Massachusetts Institute of Technology

Collections
  • Doctoral Theses

Browse

All of DSpaceCommunities & CollectionsBy Issue DateAuthorsTitlesSubjectsThis CollectionBy Issue DateAuthorsTitlesSubjects

My Account

Login

Statistics

OA StatisticsStatistics by CountryStatistics by Department
MIT Libraries
PrivacyPermissionsAccessibilityContact us
MIT
Content created by the MIT Libraries, CC BY-NC unless otherwise noted. Notify us about copyright concerns.