Machine learning-assisted design of metal–organic frameworks for hydrogen storage - A high-throughput screening and experimental approach - read and study

Aug 6, 2026

Upload by Aurora

Big data and high-throughput screening methods have accelerated materials discovery, yet translating ML predictions into experimentally realizable compounds remains difficult. This study presents a machine-learning-assisted design strategy for porous metal–organic frameworks (MOFs) aimed at hydrogen storage and targeted synthesizability. Using ML models trained on MOF databases, the approach predicts whether MOF structures are likely feasible to synthesize. Experimental validation includes a newly synthesized vanadium-based MOF showing strong cryogenic H2 storage performance.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages33

Reading Time1 h 23 min

AccessFree to download

Abstract

Introduction

Hydrogen energy storage challenges

MOFs for hydrogen storage

MOF databases and high-throughput screening

UC Berkeley UC Berkeley Previously Published Works Title Machine learning-assisted design of metal–organic frameworks for hydrogen storage: A high-throughput screening and experimental approach Permalink [https://escholarship.org/uc/item/1h0372vv](https://escholarship.org/uc/item/1h0372vv) Journal Chemical Engineering Journal, 507 ISSN 1385-8947 Authors Kim, Wan-Tae Lee, Weon-Gyu An, Hong-Eunet al. Publication Date 2025-03-01 DOI 10.1016/j.cej.2025.160766 Copyright Information This work is made available under the terms of a Creative Commons AttributionNonCommercial License, available at [https://creativecommons.org/licenses/by-nc/4.0/](https://creativecommons.org/licenses/by-nc/4.0/) Peer reviewed [eScholarship.org](eScholarship.org) Powered by the California Digital Library University of California 1 Machine Learning-Assisted Design of Metal–Organic Frameworks for Hydrogen 2 Storage: A High-Throughput Screening and Experimental Approach 3 4 Wan-Tae Kim,a,‡ Weon-Gyu Lee,b,‡ Hong-Eun An,a,c Hiroyasu Furukawa,d,e WooSeok Jeong,f,† 5 Sung-Chul Kim,g Jeffrey R. Long,d,e,h Sohee Jeong,a,* and Jung-Hoon Leeb,g* 6 7 a Materials Architecturing Research Center, Korea Institute of Science and Technology (KIST), 8 Seoul, 02792, Republic of Korea 9 bComputational Science Research Center, Korea Institute of Science and Technology (KIST), 10 Seoul, 02792, Republic of Korea 11 c Department of Materials Science and Engineering, Korea University, Seoul, 02841, Republic of Korea 12 d Department of Chemistry and Institute for Decarbonization Materials, University of California, Berkeley, 13 California, 94720, United States 14 e Materials Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, California, 94720, USA 15 fCenter for AI and Natural Sciences, Korea Institute for Advanced Study (KIAS), 16 Seoul, 02455, Republic of Korea 17 gAdvanced Analysis Center, Korea Institute of Science and Technology (KIST), 18 Seoul, 02792, Republic of Korea 19 h Department of Chemical and Biomolecular Engineering and Department of Materials Science and 20 Engineering, University of California, Berkeley, California, 94720, USA 21 gKU-KIST Graduate School of Converging Science and Technology, 22 Korea University, Seoul, 02841, Republic of Korea 23 24 *Corresponding author: [soheejeong@kist.re.kr](soheejeong@kist.re.kr) (Sohee Jeong), [jhlee84@kist.re.kr](jhlee84@kist.re.kr) (Jung-Hoon Lee) 25 26 †Present address: Computational Science & Engineering Lab. , Korea Institute of Energy Research 27 (KIER), Daejeon, 34129, Republic of Korea 28 29 ‡Wan-Tae Kim and Weon-Gyu Lee contributed equally to this work. 30 31 Abstract 32 Various theoretical approaches, including big data and high-throughput screening 33 techniques, have been explored in developing new materials due to their significant 34 potential time-saving advantages. However, it remains a significant challenge to 35 experimentally realize new materials that are predicted. In this study, we propose a novel 36 materials design strategy that utilizes machine-learning (ML) techniques to predict new 37 porous materials that show promise for hydrogen storage and are likely to be feasible to 38 synthesize. By leveraging ML techniques and metal−organic framework (MOF) databases, 39 we are able to predict the synthesizability of MOF structures. This is evidenced by the 40 successful synthesis of a new vanadium-based MOF that exhibits excellent performance 41 for cryogenic H2 storage. Notably, the total gravimetric and volumetric H2 uptakes are as 42 high as 9.0 wt % and 50.0 g/L at 77 K and 150 bar. This ML-assisted materials design 43 offers an efficient and promising approach for developing hydrogen storage materials. 44 45 Keywords: metal-organic frameworks, hydrogen storage, synthesizability, high- 46 throughput screening 47 48 1. Introduction 49 Hydrogen energy holds immense poten

FAQs

What problem does the study address in hydrogen-storage materials development?

It targets the challenge of experimentally realizing new materials that are predicted by theory or screening, despite the speed advantages of big-data and high-throughput approaches.

How does the proposed method use machine learning?

The method leverages ML techniques with MOF databases to predict the synthesizability of candidate MOF structures, guiding which porous materials are likely feasible to make.

What experimental result supports the proposed strategy?

A new vanadium-based MOF was successfully synthesized and demonstrated excellent performance for cryogenic H2 storage, with reported gravimetric and volumetric uptakes at 77 K.

Aug 6, 2026

Upload by Aurora