Necessity of Heteroatoms for Realizing Hypothetical Aluminophosphate Zeolites: A High-Throughput Computational Approach

Aluminophosphate zeolites (AlPOs) have important applications in adsorption, separation, and catalysis. Millions of hypothetical zeolite structures have been predicted, but experimentally realizing them as AlPOs requires a priori knowledge on whether heteroatom incorporations are necessary to stabilize their frameworks. Previous computations focus on the energy difference before and after heteroatom incorporation, which are not applicable for high-throughput computations because of the combinatorial explosion of possible incorporation sites. Here, we establish a new model to estimate the probability of a hypothetical structure being a pure or a heteroatom-stabilized AlPO, which is based on the Mahalanobis distances between a hypothetical structure and its neighboring reference structures in distortion–energy plots. Our approach avoids numerous attempts at heteroatom incorporation and is therefore applicable for high-throughput structure evaluation. Using this model, we have predicted 17 050 hypothetical structures being realizable as pure AlPOs and 12 039 structures realizable only via heteroatom incorporation. This will provide important guidance toward the synthesis of new AlPOs.

磷酸铝分子筛（Aluminophosphate zeolites, AlPOs）在吸附、分离与催化领域具有重要应用价值。目前已预测出数百万种假想分子筛结构，但要通过实验将其合成为磷酸铝分子筛，需要先验知识以明确是否需通过杂原子掺入稳定其骨架结构。以往的计算研究多聚焦于杂原子掺入前后的能量差，但由于可能的掺入位点存在组合爆炸问题，该方法无法适用于高通量计算场景。本研究提出一种新模型，用于估算假想结构能否作为纯相磷酸铝分子筛，或是需通过杂原子掺入方可稳定存在的磷酸铝分子筛的概率；该模型基于假想结构与其在畸变-能量图中邻近参考结构之间的马氏距离。本方法无需开展大量杂原子掺入尝试，因此可适用于高通量结构评估任务。利用该模型，我们预测得到17050种可合成为纯相磷酸铝分子筛的假想结构，以及12039种仅能通过杂原子掺入方可合成的假想结构。该研究结果将为新型磷酸铝分子筛的合成提供重要指导。