(582a) Crystal Engineering of a Zeolite Using Machine Learning

It is shown that machine learning (ML) algorithms can capture the effect of crystallization inputs on key microstructural characteristics (outputs) of faujasite (FAU), a widely used zeolite catalyst and adsorbent. This work focuses primarily on using the technique known as geometric harmonics to learn input-output relationships of interest, but we also provide a brief comparison with neural networks and Gaussian process regression, as alternative approaches. Through ML, synthesis conditions were identified to enhance the Si/Al ratio of (FAU) zeolite prepared via direct template-free synthesis to the hitherto highest level (i.e., Si/Al = 3.5). Our analysis of the ML algorithms’ results offers the insight that reduced Na₂O content is a key parameter to achieve the enhanced Si/Al ratio. An acid catalyst prepared by partial ion exchange of the new high-Si/Al-ratio FAU (Si/Al=3.5) exhibits improved proton reactivity in propane cracking and dehydrogenation compared to the catalyst prepared from the previously reported highest Si/Al ratio (Si/Al=2.8).