Data-driven high-throughput screening of anion-pillared metal-organic frameworks for hydrogen storage

doi:10.11949/0438-1157.20250178

Abstract

Abstract:

Hydrogen storage is a core issue that hinders the development of the hydrogen energy industry, and improving hydrogen storage density is a key technical difficulty. Metal-organic frameworks (MOFs) exhibit excellent hydrogen storage density, and have become one of the most promising energy storage materials. However, conventional computational approaches, including traditional molecular simulations and high-throughput screening methods, encounter significant limitations when applied to MOF hydrogen storage research. These methods are particularly constrained by their excessive computational time requirements and substantial resource consumption, which hinder efficient material discovery and optimization. To address these challenges, this study developed a data-driven high-throughput screening strategy for the rapid prediction of hydrogen storage performance in anion-templated metal-organic frameworks (AP-MOFs). The proposed method achieved exceptional predictive accuracy, with R² values exceeding 0.99 for both the training and test sets, and required only 63 s of computation time. Through this approach, 20 AP-MOFs with a hydrogen storage density exceeding 5.5%(mass) at 77 K and 5 MPa were identified, surpassing the hydrogen storage target set by the United States Department of Energy. Among these, the ALFFIVE_2_Fe structure, which is potentially synthesizable, exhibited a remarkable hydrogen storage density of 9.75%(mass) under the same conditions, along with a deliverable hydrogen storage density of 3.05%(mass). The study also revealed that the volumetric porosity has the greatest impact on hydrogen storage performance, followed by gravimetric surface area, density and porosity. These findings provide theoretical insights for the future application of AP-MOFs in hydrogen storage technologies.

Key words: metal-organic frameworks, high-throughput screening, machine learning, hydrogen storage, molecular simulation

摘要：

氢存储是掣肘氢能产业发展的核心问题，提高储氢密度是关键技术难点。金属有机框架表现出优异的储氢密度，成为最具潜力的储氢材料之一，然而传统的分子模拟技术和常规的高通量筛选方法研究金属有机框架储氢，面临计算耗时长且计算资源消耗大的问题。本文构建了一种数据驱动辅助高通量筛选策略，用于快速预测阴离子柱撑金属有机框架储氢性能，其在训练集和测试集上的R²值均超过了0.99，且耗时仅为63 s。筛选出20个金属有机框架，在77 K、5 MPa下的质量储氢密度均超过了美国能源部设定的储氢目标5.5%（质量）。潜在可能合成的ALFFIVE_2_Fe结构在77 K、5 MPa下，质量储氢密度达到9.75%（质量），可交付质量储氢密度达到了3.05%（质量）。通过特征值贡献度发现孔体积对储氢容量影响最大，依次是质量表面积、密度以及孔隙率。研究结果可为阴离子柱撑金属有机框架在储氢应用方面提供一定的理论参考。

关键词: 金属有机框架, 高通量筛选, 机器学习, 储氢, 分子模拟

CLC Number:

O 604

Zheng GAO, Hui WANG, Zhiguo QU. Data-driven high-throughput screening of anion-pillared metal-organic frameworks for hydrogen storage[J]. CIESC Journal, 2025, 76(8): 4259-4272.

高正, 汪辉, 屈治国. 数据驱动辅助高通量筛选阴离子柱撑金属有机框架储氢[J]. 化工学报, 2025, 76(8): 4259-4272.

Figures/Tables 16

Fig.1 Schematic diagram of the data-driven-assisted high-throughput screening protocol

Fig.2 Periodic model and cluster structure of ALFFIVE-3-Cu

Table 1 Lennard-Jones parameters of MOFs

Atom	(ε/k_B)/K	σ/Å	Atom	(ε/k_B)/K	σ/Å	Atom	(ε/k_B)/K	σ/Å	Charge/e
H	7.65	2.85	Ti	8.55	2.83	Ge	190.69	3.81	—
B	47.81	3.58	V	8.05	2.8	Zr	34.72	2.78	—
C	47.86	3.47	Fe	6.54	2.59	Nb	29.69	2.82	—
N	38.95	3.26	Co	7.05	2.56	Cd	114.73	2.54	—
O	48.16	3.03	Ni	7.55	2.52	In	301.43	3.98	—
F	36.48	3.09	Cu	2.52	3.11	Sn	285.28	3.91	—
Al	155.99	3.91	Zn	62.4	2.46	H_H₂	—	—	0.468
Si	156	3.8	Ga	208.84	3.90	H_com	36.7	2.958	-0.936

Fig.3 Pearson correlation matrix between features

Table 2 Model accuracy comparison

Model	Abbr.	MAE	RMSE	R²
Auto-Gluon-stacking	AG	0.2793	0.4004	0.9997
H₂O-stacking	H₂O	0.4745	0.6367	0.9994
Extra Trees Regressor	ET	0.9966	1.5235	0.9959
Gradient Boosting Regressor	GB	1.3043	1.8446	0.9941
Random Forest Regressor	RF	1.3000	1.9372	0.9933
Light Gradient Boosting Machine	LightGBM	1.5946	2.5524	0.9882
AdaBoost Regressor	Ada	2.1591	2.7458	0.9869
Linear Regression	LR	2.1809	2.7753	0.9867
Bayesian Ridge	BR	2.2090	2.7953	0.9865
Decision Tree Regressor	DT	1.6932	2.8758	0.9853
K Neighbors Regressor	KNN	2.2578	3.4469	0.9788
Ridge Regression	Ridge	2.9051	3.7187	0.9759
Lasso Regression	Lasso	3.5638	4.535	0.9644
Lasso Least Angle Regression	LLAR	3.5638	4.535	0.9644
Elastic Net	EN	3.5786	4.5482	0.9642
Huber Regressor	Huber	3.7659	4.9356	0.9575
Orthogonal Matching Pursuit	OMP	5.8721	6.6563	0.9237
Passive Aggressive Regressor	PAR	11.3854	14.1384	0.4173
Dummy Regressor	Dummy	20.603	24.2602	-0.0084

Fig.4 Comparison between training set, test set and GCMC of AG,H2O,ET,GB and RF

Fig.5 Time statistics based on molecular simulation

Fig.6 (a) Feature importance; (b) Relationship between feature and adsorption capacity

Fig.7 Molecular structure of high-performance hydrogen storage MOFs

Table 3 MOFs for high-performance hydrogen storage

No.	MOFs	ρ/(g/cm³)	GSA/(m²/g)	VP/(cm³/g)	VF	GCMC/(mg/g)	ML/(mg/g)
1	BFFIVE_2_Ni	0.52	3943.96	1.42	0.74	116.9	113.94
2	BFFIVE_2_Zn	0.52	3765.55	1.40	0.73	115.95	114.19
3	BFFIVE_7_Zn	0.54	3551.53	1.34	0.72	115.7	114.19
4	BFFIVE_2_Cu	0.52	3882.58	1.40	0.73	115.12	114.12
5	BFFIVE_7_Cd	0.54	3533.66	1.37	0.74	114.85	114.12
6	BFFIVE_2_Cd	0.53	3695.77	1.40	0.74	114.41	114.12
7	BFFIVE_7_Cu	0.55	3568.01	1.31	0.72	113.38	113.82
8	BFFIVE_2_Fe	0.52	3737.87	1.40	0.73	112.88	113.82
9	BFFIVE_2_Co	0.54	3741.83	1.35	0.73	112.42	113.82
10	BFFIVE_7_Co	0.56	3423.04	1.27	0.71	111.52	111.00
11	BFFIVE_7_Fe	0.54	3427.77	1.33	0.72	109.86	111.25
12	BFFIVE_7_Ni	0.56	3300.72	1.26	0.71	107.05	106.65
13	BFFIVE_16_Cu_i	0.55	3715.38	1.22	0.67	101.39	101.22
14	BFFIVE_16_Zn_i	0.56	3561.92	1.18	0.66	97.64	97.50
15	BFFIVE_16_Co_i	0.56	3711.88	1.18	0.67	97.57	97.21
16	ALFFIVE_2_Fe	0.59	3203.36	1.18	0.69	97.51	97.42
17	BFFIVE_16_Ni_i	0.57	3599.53	1.16	0.66	97.45	97.36
18	BFFIVE_16_Fe_i	0.56	3632.94	1.18	0.66	97.31	97.50
19	ALFFIVE_2_Zn	0.6	3162.78	1.17	0.70	96.86	96.18
20	ALFFIVE_2_Cd	0.59	3051.31	1.22	0.72	96.65	96.12

Fig.8 Organic ligands for high-performance hydrogen storage MOFs

Fig.9 (a) ALFFIVE_2_Fe adsorption isotherm; (b) Adsorption heat versus pressure of ALFFIVE_2; (c) Deliverable mass capacity versus adsorption heat of a typical MOF[36]

Fig.10 Density distribution of ALFFIVE_2_Fe at 77 K

Fig.11 Van der Waals potential distribution on the surface of ALFFIVE_2_Fe

Fig.12 Typical adsorption configurations of H2 on MOF

Table 4 Adsorption energies at different adsorption sites

位置	构型	位点	$E H 2$ /a.u	E_MOF/a.u	$E M O F - H 2$ /a.u	ΔE/(kJ/mol)
吡啶	水平（H）	Fe-N	-1.166	-3463.457	-3464.626	-8.590
	垂直（V）	Fe-N	-1.166	-3463.457	-3464.625	-7.319
	水平（H）	Hollow	-1.166	-3463.457	-3464.628	-8.595
	垂直（V）	Hollow	-1.166	-3463.457	-3464.625	-6.975
乙炔	水平（H）	—	-1.166	-3463.448	-3464.624	-4.900
乙炔	垂直（V）	—	-1.166	-3463.457	-3464.624	-3.067

Table 4 Adsorption energies at different adsorption sites

位置	构型	位点	$E H 2$ /a.u	E_MOF/a.u	$E M O F - H 2$ /a.u	ΔE/(kJ/mol)
吡啶	水平（H）	Fe-N	-1.166	-3463.457	-3464.626	-8.590
	垂直（V）	Fe-N	-1.166	-3463.457	-3464.625	-7.319
	水平（H）	Hollow	-1.166	-3463.457	-3464.628	-8.595
	垂直（V）	Hollow	-1.166	-3463.457	-3464.625	-6.975
乙炔	水平（H）	—	-1.166	-3463.448	-3464.624	-4.900
乙炔	垂直（V）	—	-1.166	-3463.457	-3464.624	-3.067

References 40

[1]	Gupta A, Baron G V, Perreault P, et al. Hydrogen clathrates: next generation hydrogen storage materials[J]. Energy Storage Materials, 2021, 41: 69-107.
[2]	Li H L, Eddaoudi M, O'Keeffe M, et al. Design and synthesis of an exceptionally stable and highly porous metal-organic framework[J]. Nature, 1999, 402: 276-279.
[3]	Moghadam P Z, Chung Y G, Snurr R Q. Progress toward the computational discovery of new metal-organic framework adsorbents for energy applications[J]. Nature Energy, 2024, 9: 121-133.
[4]	Li A, Bueno-Perez R, Madden D, et al. From computational high-throughput screenings to the lab: taking metal-organic frameworks out of the computer[J]. Chemical Science, 2022, 13(27): 7990-8002.
[5]	Bobbitt N S, Chen J Y, Snurr R Q. High-throughput screening of metal-organic frameworks for hydrogen storage at cryogenic temperature[J]. The Journal of Physical Chemistry C, 2016, 120(48): 27328-27341.
[6]	Liu K H, Chen Z J, Islamoglu T, et al. Exploring the chemical space of metal-organic frameworks with rht topology for high capacity hydrogen storage[J]. The Journal of Physical Chemistry C, 2024, 128(18): 7435-7446.
[7]	Thornton A W, Simon C M, Kim J, et al. Materials genome in action: identifying the performance limits of physical hydrogen storage[J]. Chemistry of Materials, 2017, 29(7): 2844-2854.
[8]	Xie Y Y, Li X D, Zhang H D, et al. Evaluation and screening of multivariate metal-organic frameworks for hydrogen storage[J]. International Journal of Hydrogen Energy, 2024, 87: 1356-1366.
[9]	Ahmed A, Seth S, Purewal J, et al. Exceptional hydrogen storage achieved by screening nearly half a million metal-organic frameworks[J]. Nature Communications, 2019, 10(1): 1568.
[10]	Lu X Y, Xie Z Z, Wu X J, et al. Hydrogen storage metal-organic framework classification models based on crystal graph convolutional neural networks[J]. Chemical Engineering Science, 2022, 259: 117813.
[11]	Wang L M, Feng S H, Zhang C J, et al. Artificial intelligence and high-throughput computational workflows empowering the fast screening of metal-organic frameworks for hydrogen storage[J]. ACS Applied Materials & Interfaces, 2024, 16(28): 36444-36452.
[12]	Ahmed A, Siegel D J. Predicting hydrogen storage in MOFs via machine learning[J]. Patterns, 2021, 2(7): 100291.
[13]	Park J, Lim Y, Lee S, et al. Computational design of metal-organic frameworks with unprecedented high hydrogen working capacity and high synthesizability[J]. Chemistry of Materials, 2023, 35(1): 9-16.
[14]	Guo W J, Liu J, Dong F, et al. Machine learning for predicting gas adsorption capacities of metal organic framework[M]//Machine Learning and Deep Learning in Computational Toxicology. Cham: Springer International Publishing, 2023: 629-654.
[15]	Lin J, Liu Z M, Guo Y J, et al. Machine learning accelerates the investigation of targeted MOFs: performance prediction, rational design and intelligent synthesis[J]. Nano Today, 2023, 49: 101802.
[16]	Altintas C, Keskin S. On the shoulders of high-throughput computational screening and machine learning: design and discovery of MOFs for H₂ storage and purification[J]. Materials Today Energy, 2023, 38: 101426.
[17]	Neikha K, Puzari A. Metal-organic frameworks through the lens of artificial intelligence: a comprehensive review[J]. Langmuir, 2024, 40(42): 21957-21975.
[18]	Li Z, Bucior B J, Chen H Y, et al. Machine learning using host/guest energy histograms to predict adsorption in metal-organic frameworks: application to short alkanes and Xe/Kr mixtures[J]. The Journal of Chemical Physics, 2021, 155(1): 014701.
[19]	Shi K H, Li Z, Anstine D M, et al. Two-dimensional energy histograms as features for machine learning to predict adsorption in diverse nanoporous materials[J]. Journal of Chemical Theory and Computation, 2023, 19(14): 4568-4583.
[20]	Liu X, Wang H, Liu C, et al. Recent advances of multidentate ligand-based anion-pillared MOFs for enhanced separation and purification processes[J]. Chem & Bio Engineering, 2024, 1(6): 469-487.
[21]	Cui X L, Chen K J, Xing H B, et al. Pore chemistry and size control in hybrid porous materials for acetylene capture from ethylene[J]. Science, 2016, 353(6295): 141-144.
[22]	Gu C K, Yu Z Z, Liu J, et al. Construction of an anion-pillared MOF database and the screening of MOFs suitable for Xe/Kr separation[J]. ACS Applied Materials & Interfaces, 2021, 13(9): 11039-11049.
[23]	Willems T F, Rycroft C H, Kazi M, et al. Algorithms and tools for high-throughput geometry-based analysis of crystalline porous materials[J]. Microporous and Mesoporous Materials, 2012, 149(1): 134-141.
[24]	Martin R L, Haranczyk M. Construction and characterization of structure models of crystalline porous polymers[J]. Crystal Growth and Design, 2014, 14(5): 2431-2440.
[25]	Wu Y, Duan H P, Xi H X. Machine learning-driven insights into defects of zirconium metal-organic frameworks for enhanced ethane-ethylene separation[J]. Chemistry of Materials, 2020, 32(7): 2986-2997.
[26]	Orhan I B, Daglar H, Keskin S, et al. Prediction of O₂/N₂ selectivity in metal-organic frameworks via high-throughput computational screening and machine learning[J]. ACS Applied Materials & Interfaces, 2022, 14(1): 736-749.
[27]	Xue X Y, Cheng M, Wang S H, et al. High-throughput screening of metal-organic frameworks assisted by machine learning: propane/propylene separation[J]. Industrial & Engineering Chemistry Research, 2023, 62(2): 1073-1084.
[28]	Dubbeldam D, Calero S, Ellis D E, et al. RASPA: molecular simulation software for adsorption and diffusion in flexible nanoporous materials[J]. Molecular Simulation, 2016, 42(2): 81-101.
[29]	Colón Y J, Fairen-Jimenez D, Wilmer C E, et al. High-throughput screening of porous crystalline materials for hydrogen storage capacity near room temperature[J]. The Journal of Physical Chemistry C, 2014, 118(10): 5383-5389.
[30]	Hutter J, Iannuzzi M, Schiffmann F, et al. cp2k: atomistic simulations of condensed matter systems[J]. WIREs Computational Molecular Science, 2014, 4(1): 15-25.
[31]	LeDell E, Poirier S. H₂O AutoML: scalable automatic machine learning[C]//Proceedings of the AutoML Workshop at ICML. Association for Computing Machinery, 2020: 1-16.
[32]	Paldino G M, De Stefani J, De Caro F, et al. Does AutoML outperform naive forecasting?[J]. Engineering Proceedings, 2021, 5(1): 36.
[33]	Ahmed A, Liu Y Y, Purewal J, et al. Balancing gravimetric and volumetric hydrogen density in MOFs[J]. Energy & Environmental Science, 2017, 10(11): 2459-2471.
[34]	Allendorf M D, Stavila V, Snider J L, et al. Challenges to developing materials for the transport and storage of hydrogen[J]. Nature Chemistry, 2022, 14(11): 1214-1223.
[35]	Cadiau A, Belmabkhout Y, Adil K, et al. Hydrolytically stable fluorinated metal-organic frameworks for energy-efficient dehydration[J]. Science, 2017, 356(6339): 731-735.
[36]	García-Holley P, Schweitzer B, Islamoglu T, et al. Benchmark study of hydrogen storage in metal–organic frameworks under temperature and pressure swing conditions[J]. ACS Energy Letters, 2018, 3(3): 748-754.
[37]	Srivastava S, Shet S P, Shanmuga Priya S, et al. Molecular simulation of copper based metal-organic framework (Cu-MOF) for hydrogen adsorption[J]. International Journal of Hydrogen Energy, 2022, 47(35): 15820-15831.
[38]	Lu T. A comprehensive electron wavefunction analysis toolbox for chemists, Multiwfn[J]. The Journal of Chemical Physics, 2024, 161(8): 082503.
[39]	Lu T, Chen Q X. Visualization analysis of weak interactions in chemical systems[M]//Comprehensive Computational Chemistry. Amsterdam: Elsevier, 2024: 240-264.
[40]	Humphrey W, Dalke A, Schulten K. VMD: visual molecular dynamics[J]. Journal of Molecular Graphics, 1996, 14(1): 33-38.