Developing Machine Learning Potential to Unravel Quantum Effect on Ionic Hydration and Transport in Nanoscale Confinement

碳纳米管等纳米多孔材料已成为创建可持续支持人类社会的先进分离技术的基础;例如，通过生产清洁水或回收锂等关键材料。纳米多孔材料的一个独特的结构相关性质来自其纳米级孔，其对孔内的离子和分子施加量子力学效应，导致性质与日常大量观察到的性质大不相同。全面理解纳米尺度的限制效应对于合理设计纳米多孔膜分离技术是至关重要的。然而，人们对纳米尺度限制诱导的量子效应对离子和分子性质的影响了解有限，阻碍了分离技术的发展。本计画旨在发展计算模型，以阐明量子效应对碳奈米管中离子与分子的结构与动力学性质的影响。拟议的研究和教育活动通过本科生研究机会紧密结合在这个项目中，重点是促进STEM领域的多样性和课程内容的开发，以培养未来的STEM劳动力先进的计算分析技术。该项目由ENG/CBET的界面工程项目和刺激竞争力研究的既定项目（EPSCoR）共同资助。由于纳米约束流体在开发先进的分离和反应技术中的重要性，其性质一直是科学和工程研究的主题。这项研究表明，理解纳米约束流体行为的关键在于流体的非体性质。由于纳米尺度的物理约束和各向异性环境，纳米尺度的限制可以激发量子效应。量子效应在确定纳米约束流体的非体特征方面起着至关重要的作用。然而，由于缺乏有效的计算工具，在理解离子和分子的非体特征的量子效应方面仍然存在知识差距。该研究小组假设，训练有素的机器学习潜力可以像量子力学计算一样准确地预测大量和碳纳米管内离子的水合作用。在这一假设的驱动下，该项目将开发一个基于机器学习的力场，并研究在一定直径和手性范围内碳纳米管（CNT）中离子水合和传输的量子效应。该项目包括三项研究任务：（1）开发机器学习力场，以调查选定阳离子的水合结构，动力学和热力学（Li+、Na+、K+、Mg 2+和H+）和阴离子（2）开发一个机器学习力场来研究碳纳米管内部水分子的非本体性质;以及（3）扩展机器学习力场并研究碳纳米管中的离子水合和输运效应。拟议的研究将使用分子动力学（MD），从头算MD，增强采样和机器学习方法进行。最重要的成果将包括（a）更深入地了解纳米级限制中离子水合和传输的量子效应，以及（B）体相和CNT内部离子和水分子的可扩展机器学习力场。该奖项反映了NSF的法定使命，并已被认为值得通过使用基金会的智力价值和更广泛的影响审查标准进行评估的支持。

Nanoporous materials such as carbon nanotubes have served as the foundation for creating advanced separation technologies that sustainably support human society; for example, by producing clean water or recovering critical materials such as lithium. A unique structure-related property of nanoporous materials arises from their nanoscale pores, which exert quantum mechanical effects on the ions and molecules within the pores resulting in properties vastly differing from those observed in everyday bulk quantities. A complete understanding of the nanoscale confinement-induced effects is crucial for rational design of separation technologies using nanoporous membranes. However, fundamental knowledge of the nanoscale confinement-induced quantum effects on the properties of ions and molecules is limited, hampering the development of separation technologies. This project aims to develop computational models to elucidate the quantum effects on the structural and dynamic properties of ions and molecules confined in carbon nanotubes. The proposed research and education activities are closely integrated in this project through undergraduate research opportunities that focus on promoting diversity within the STEM field and the development of course content to train the future STEM workforce in advanced computational analysis techniques. This project is jointly funded by the Interfacial Engineering program of ENG/CBET and the Established Program to Stimulate Competitive Research (EPSCoR).The properties of nanoconfined fluids have been topics for science and engineering investigation because of their importance in developing advanced separation and reaction technologies. This research has demonstrated that the key to understanding nanoconfined fluid behavior lies in the fluid’s non-bulk properties. Nanoscale confinement can incite quantum effects due to the physically constraining and anisotropic environment. Quantum effects play an essential role in determining the non-bulk features of nanoconfined fluids. However, a knowledge gap remains in understanding the quantum effect on the non-bulk features of ions and molecules due to the lack of efficient computational tools. This research group hypothesizes that a well-trained machine learning potential can predict the hydration of ions in bulk and within carbon nanotubes as accurately as quantum mechanical calculations. Driven by this hypothesis, this project will develop a machine learning-based force field and investigate the quantum effect on ionic hydration and transport in carbon nanotubes (CNT) over a range of diameters and chiralities. The project includes three research tasks: (1) develop a machine learning force field to investigate the hydration structure, dynamics, and thermodynamics of selected cations (Li+, Na+, K+, Mg2+ and H+) and anions (F-, and Cl-) in the bulk phase; (2) develop a machine learning force field to investigate the non-bulk properties of water molecules inside carbon nanotubes; and (3) extend the machine learning force field and investigate ionic hydration and transport effects in carbon nanotubes. The proposed research will be conducted using molecular dynamics (MD), ab initio MD, enhanced sampling, and machine learning methods. The most significant outcomes will include (a) a deeper understanding of the quantum effect on ionic hydration and transport in the nanoscale confinement and (b) an extensible machine learning force field for ions and water molecules in the bulk phase and inside CNTs. The outcomes will provide the knowledge and cyber-infrastructure for model-based design of nanoporous separation membranes.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.