Controlling morphology of nanocrystals is one of the central tasks of nanoscience. In this work, we studied nanoscale metal-organic frameworks (nMOFs) from Hf-oxo clusters and linear dicarboxylate ligands with the aid of machine-learning methods for data analysis. Ligand solubility and modulator concentration were found to quantitatively predict the growth of nMOFs with a specific morphology, such as ultrathin two-dimensional film, hexagonal nanoplate, octahedron, cuboctahedron, concave octahedron, or hollow octahedron morphology. With these insights, we use epitaxy growth sequences to design nMOFs of desirable nanostructures with enhanced substrate transport and, hence, increased activities for catalytic olefin hydrogenation. This work highlights new opportunities in using machine learning to guide morphology engineering of nMOFs and other nanomaterials.

控制纳米晶体的形态是纳米科学的中心任务之一。在这项工作中，我们借助机器学习方法进行数据分析，研究了来自Hf-氧代簇和线性二羧酸酯配体的纳米级金属有机框架（nMOF）。发现配体溶解度和调节剂浓度可定量预测具有特定形态的nMOF的生长，例如超薄二维薄膜，六边形纳米板，八面体，立方八面体，凹面八面体或空心八面体形态。有了这些见解，我们使用外延生长序列来设计具有增强的底物传输并因此增加了催化烯烃氢化活性的所需纳米结构的nMOF。