Machine Learning-Assisted Elucidation of the Adsorption Structure-Mechanism Relationship of Uranyl on Covalent Organic Frameworks★

doi:10.6023/A26060214

Abstract

Covalent organic frameworks (COFs) have shown significant potential in the adsorption and separation of uranyl ions ($\mathrm{UO}_{2}^{2+}$) due to their structural designability, ordered pore architectures, and tunable functional groups. However, existing research has largely focused on improving the performance of individual materials, while a systematic understanding of the relationship between COFs structural features and uranyl adsorption mechanisms remains limited. In this study, we constructed a COFs-uranyl adsorption dataset comprising 105 valid samples based on literature data. We standardized the adsorption mechanisms reported in the literature into five types: coordination, redox, coordination-redox coupling, coordination-electrostatic coupling, and weak interactions. By integrating statistical analysis, machine learning classification, and SHapley Additive exPlanations (SHAP) analysis, we systematically investigated the associations between COFs structural parameters and uranyl adsorption mechanisms. The results showed that coordination-related interactions were dominant in uranyl adsorption by COFs, accounting for over 80% of the samples, indicating the importance of nitrogen- and oxygen-containing coordination sites and pore accessibility in uranyl enrichment. SHAP analysis further showed that bandgap, pore size, and specific surface area made substantial contributions to model-based mechanism classification. Samples with lower band gaps were more likely to be classified as involving redox-related mechanisms, while moderate pore sizes and higher specific surface areas were closely associated with pore accessibility and the exposure of coordination sites. Framework ordering may also contribute to electrostatically assisted enrichment by affecting pore connectivity and interfacial charge distribution. This work reveals data-driven associations between COFs structural features and uranyl adsorption mechanisms, providing valuable insights into mechanism identification and the rational design of COFs-based uranyl adsorbents.

Key words: machine learning, covalent organic frameworks, uranyl ion, adsorption material

Cite this article

Shibo Sun, Sheng Wang, Hanyang Wang, Chenkuo Lan, Jing Ma, Qizhao Wang, Zhenli Sun, Xiangke Wang. Machine Learning-Assisted Elucidation of the Adsorption Structure-Mechanism Relationship of Uranyl on Covalent Organic Frameworks^★[J]. Acta Chimica Sinica, 2026, 84(9): 1418-1427.

Export EndNote|Reference Manager|ProCite|BibTeX|RefWorks

share this article

Fig. & Tab. 7

References 26

[1]	Sun Z.; Chen Z.; Wang S.; Tai X.; Wang X. Innovation 2026, 7, 101093.
[2]	Ma J.; Zhang T. Y.; Yu F. Coord. Chem. Rev. 2026, 556, 217641. doi: 10.1016/j.ccr.2026.217641
[3]	Wang N.; Pang H.; Yu S.; Gu P.; Song S.; Wang H.; Wang X. Acta Chim. Sinica 2019, 77, 143 (in Chinese). doi: 10.6023/A18090404
	(王宁, 庞宏伟, 于淑君, 顾鹏程, 宋爽, 王宏青, 王祥科, 化学学报, 2019, 77, 143.) doi: 10.6023/A18090404
[4]	Yang L.; Xiao H.; Qian Y.; Zhao X.; Kong X.-Y.; Liu P.; Xin W.; Fu L.; Jiang L.; Wen L. Nature Sustain. 2022, 5, 71. doi: 10.1038/s41893-021-00792-6
[5]	Geng K.; He T.; Liu R.; Dalapati S.; Tan K. T.; Li Z.; Tao S.; Gong Y.; Jiang Q.; Jiang D. Chem. Rev. 2020, 120, 8814. doi: 10.1021/acs.chemrev.9b00550
[6]	Sun Z. L.; Chen Z. S.; Tai X. S.; Wang X. K. Sci. China Chem. 2025, 68, 3923. doi: 10.1007/s11426-025-2784-1
[7]	Sun Z.; Liao Y.; Zhang Y.; Sun S.; Kan Q.; Wu Z.; Yu L.; Dong Z.; Wang Z.; He R.; Wang L.; Meng Q.; Wang H.; Wang Q.; Mao L.; Pan D.; Wang S.; Zhang Z.; Zhu W.; Liu S.; Wakeel M.; Hu B.; Duan T.; Tai X.; Wang X. Sustain. Carbon Mater. 2025, 1, e007.
[8]	Liu S.; Tao B.; Zuo B.; Zheng K.; Abdelfattah W.; Bao J.; Yan G.; Wang T.; Wang R.; Li Z.; Melhi S.; Alshahrani W. A.; Xu X. Chem. Eng. J. 2024, 500, 156783. doi: 10.1016/j.cej.2024.156783
[9]	Guo L.; Liu Y.; Jia Y.; Huang Z. C.; Gong L.; Xie X.; Luo F. J. Am. Chem. Soc. 2025, 147, 31340. doi: 10.1021/jacs.5c10932
[10]	Niu C.; Zhang C.; Liu X.; Liang R.; Qiu J. Nat. Commun. 2023, 14, 4420. doi: 10.1038/s41467-023-40169-1
[11]	Yang H.; Hao M.; Xie Y.; Liu X.; Liu Y.; Chen Z.; Wang X.; Waterhouse G. I.; Ma S. Angew. Chem. Int. Ed. 2023, 62, e202303129.
[12]	Cui W.; Zhang C.; Xu R.; Chen X.; Jiang W.; Li Y.; Liang R.; Zhang L.; Qiu J. Appl. Catal. B-Environ. 2021, 294, 120250. doi: 10.1016/j.apcatb.2021.120250
[13]	Zhang H. L.; Liu W.; Li A.; Zhang D.; Li X. Y.; Zhai F. W.; Chen L. H.; Chen L.; Wang Y. L.; Wang S. Angew. Chem. Int. Ed. 2019, 58, 16110. doi: 10.1002/anie.v58.45
[14]	Lopez A. M.; Wells A.; Fendorf S. Environ. Sci. Technol. 2021, 55, 352. doi: 10.1021/acs.est.0c05591
[15]	Allen A. E. A.; Tkatchenko A. Sci. Adv. 2022, 8, eabm7185.
[16]	Liu X.; Lu D.; Zhang A.; Liu Q.; Jiang G. Environ. Sci. Technol. 2022, 56, 2124. doi: 10.1021/acs.est.1c06157
[17]	Li J.; Song Q.; Liu Z.; Wang D. Acta Chim. Sinica 2024, 82, 387 (in Chinese). doi: 10.6023/A23100473
	(李珺卿, 宋千禧, 刘子义, 王东琪, 化学学报, 2024, 82, 387.) doi: 10.6023/A23100473
[18]	Sun Z.; Sun S.; Wang S.; Wang H.; Zhang Y.; Huang J.; Fan F.; Yang Y.; Wang X. Sci. China Chem. 2026, 69, 3477. doi: 10.1007/s11426-025-3233-y
[19]	Xie Y.; Wu Y.; Liu X.; Hao M.; Chen Z.; Waterhouse G. I.; Wang X.; Yang H.; Ma S. Cell Rep. Phys. Sci. 2023, 4, 101220.
[20]	Cui W.; Zhang C.; Xu R.; Chen X.; Yan R.; Jiang W.; Liang R.; Qiu J. Small 2021, 17, 2006882. doi: 10.1002/smll.v17.6
[21]	Xu R.; Cui W.; Zhang C.; Chen X.; Jiang W.; Liang R.; Qiu J. Chem. Eng. J. 2021, 419, 129550. doi: 10.1016/j.cej.2021.129550
[22]	Hao M.; Chen Z.; Liu X.; Liu X.; Zhang J.; Yang H.; Waterhouse G. I.; Wang X.; Ma S. CCS Chem. 2022, 4, 2294. doi: 10.31635/ccschem.022.202201897
[23]	Hao M.; Xie Y.; Liu X.; Chen Z.; Yang H.; Waterhouse G. I.; Ma S.; Wang X. JACS Au 2023, 3, 239. doi: 10.1021/jacsau.2c00614
[24]	Zhang J.; Zhou L.; Jia Z.; Li X.; Qi Y.; Yang C.; Guo X.; Chen S.; Long H.; Ma L. Nanoscale 2020, 12, 24044. doi: 10.1039/d0nr06854c pmid: 33295920
[25]	van der Maaten, L. J. Mach. Learn. Res. 2014, 15, 3221.
[26]	Jolliffe I. T.; Cadima J. Philos. Trans. R. Soc. A 2016, 374, 20150202. doi: 10.1098/rsta.2015.0202

机制类别	原始机制标签	判定依据	典型证据
配位作用	配位、螯合、表面络合、配位键作用	$\mathrm{UO}_{2}^{2+}$与COFs骨架中的N/O/S/P等供电子位点形成稳定结合	傅里叶变换红外光谱法(FTIR)/X射线光电子能谱(XPS)中N, O, S和P等官能团峰位或结合能变化; 扩展X射线吸收精细结构(EXAFS)证实U—O/U—N等配位结构; 密度泛函理论(DFT)显示较强结合能或稳定配位构型; 吸附前后官能团特征变化
氧化还原作用	氧化还原、化学还原、光催化还原、电化学还原	U(VI)发生电子转移, 形成 U(IV)或低价铀物种	XPS出现U(IV)信号; 光催化或电化学还原证据; 光电流、EPR或能带分析等电子转移证据
配位-氧化还原作用	配位-氧化还原、螯合-还原、配位- 化学还原、配位- 光催化还原	$\mathrm{UO}_{2}^{2+}$先通过配位作用实现界面富集, 随后发生电子转移或价态转化	配位证据与U(IV)或低价铀证据同时存在; 或光/电催化还原过程伴随明显官能团络合证据
配位-静电作用	配位-静电、配位- 离子交换、静电- 离子交换-配位	配位结合同时伴随表面电荷吸引、离子交换或pH依赖富集过程	Zeta电位变化; pH依赖吸附行为; 离子竞争实验; 表面电荷分析; FTIR/XPS配位证据
弱相互作用	物理吸附、物理嵌合、氢键、范德华力、π-π作用、非特异性吸附	缺乏明确强配位或价态变化证据, 吸附主要依赖孔结构匹配或弱界面作用	孔结构匹配分析; 氢键、范德华力或π-π作用讨论; 较弱光谱变化; 非特异性吸附现象

机制类别	原始机制标签	判定依据	典型证据
配位作用	配位、螯合、表面络合、配位键作用	$\mathrm{UO}_{2}^{2+}$与COFs骨架中的N/O/S/P等供电子位点形成稳定结合	傅里叶变换红外光谱法(FTIR)/X射线光电子能谱(XPS)中N, O, S和P等官能团峰位或结合能变化; 扩展X射线吸收精细结构(EXAFS)证实U—O/U—N等配位结构; 密度泛函理论(DFT)显示较强结合能或稳定配位构型; 吸附前后官能团特征变化
氧化还原作用	氧化还原、化学还原、光催化还原、电化学还原	U(VI)发生电子转移, 形成 U(IV)或低价铀物种	XPS出现U(IV)信号; 光催化或电化学还原证据; 光电流、EPR或能带分析等电子转移证据
配位-氧化还原作用	配位-氧化还原、螯合-还原、配位- 化学还原、配位- 光催化还原	$\mathrm{UO}_{2}^{2+}$先通过配位作用实现界面富集, 随后发生电子转移或价态转化	配位证据与U(IV)或低价铀证据同时存在; 或光/电催化还原过程伴随明显官能团络合证据
配位-静电作用	配位-静电、配位- 离子交换、静电- 离子交换-配位	配位结合同时伴随表面电荷吸引、离子交换或pH依赖富集过程	Zeta电位变化; pH依赖吸附行为; 离子竞争实验; 表面电荷分析; FTIR/XPS配位证据
弱相互作用	物理吸附、物理嵌合、氢键、范德华力、π-π作用、非特异性吸附	缺乏明确强配位或价态变化证据, 吸附主要依赖孔结构匹配或弱界面作用	孔结构匹配分析; 氢键、范德华力或π-π作用讨论; 较弱光谱变化; 非特异性吸附现象

类别	参数	单位	缩写	数据范围	样本数
吸附剂结构	比表面积	m²/g	SA	0.80~1713.00	98
	孔径	nm	pore_size	0.40~7.49	68
	孔容	cm³/g	pore_vol	0.02~0.90	33
	带隙	eV	bandgap	1.25~3.58	35
	结晶有序性	—	crystallinity	高/中/低	102
实验条件	初始浓度	mg/L	C₀	0.0033~500.00	105
	pH	—	pH	1.50~9.00	105
	温度	K	T	293.00~318.00	105
	平衡时间	h	eq_time	0.03~288.00	98

类别	参数	单位	缩写	数据范围	样本数
吸附剂结构	比表面积	m²/g	SA	0.80~1713.00	98
	孔径	nm	pore_size	0.40~7.49	68
	孔容	cm³/g	pore_vol	0.02~0.90	33
	带隙	eV	bandgap	1.25~3.58	35
	结晶有序性	—	crystallinity	高/中/低	102
实验条件	初始浓度	mg/L	C₀	0.0033~500.00	105
	pH	—	pH	1.50~9.00	105
	温度	K	T	293.00~318.00	105
	平衡时间	h	eq_time	0.03~288.00	98

机器学习辅助解析共价有机框架材料吸附铀酰的结构-机制关系^★