Data-Driven Discovery of Metal-Organic Frameworks for Sieving Separation of Quinary C5 Olefin Mixtures

C5 olefins are widely used as chemical feedstocks for producing high-performance polymers and pharmaceuticals, and their separation requires complex and energy-intensive processes. Herein, we report the adsorptive separation of a quinary C5 mixture through the data-driven discovery of sieving sorbents. By systematically analyzing 1067 literature sources using large language models, a data set of 345 metal-organic frameworks (MOFs) and their corresponding sieving performance was established. This data set enabled the determination of effective pore sizes for MOFs, which subsequently guided the discovery of ZIF-8 as an exceptional sorbent capable of producing 99.99% pure isoprene from the C5 mixture. In a pilot-scale breakthrough experiment, 350 mL of high-purity isoprene was obtained by using 1.8 kg of shaped MOF. Extending this approach, three additional MOFs (Co-FA, Co(pz-NH₂)Ni, and Co-gallate) were discovered for the separation of remaining four C5 components. Finally, integrating these four MOFs in a sequential separation process achieved complete isolation of all five C5 compounds (isoprene, 1-pentene, cis-2-pentene, trans-2-pentene, and n-pentane) with high purity and demonstrated remarkable cycling stability over consecutive operations.