Organic Solvent Nanofiltration and Data-Driven Approaches

Abstract

Industrial acceptance of organic solvent nanofiltration (OSN) as a separation technique is hampered by the slow process of membrane screening based on trial and error for each solute-solvent couple. Such extensive experimental screening is still necessary due to limitations of predictive models, which are generally based on separations in water. The complexity and variety of competing interactions of all solute-solvent-membrane affinities challenge our understanding and accuracy of predictions of the underlying separation mechanism.

Recently, data-driven techniques showed their potential to enhance predictability in the field of OSN. We explore the use of data-driven techniques specifically for ceramic membranes, both native and functionalized, which have been one of VITO’s technological focus points in OSN activities for many years. More specifically, we explore the dataspace of ceramic membranes to determine the key physico-chemical properties to keep track of, as well as the least important features to discard. This aids the optimal development of data-driven models to predict OSN performance by reducing the data-dimension, and, at the same time, it provides much needed physical insight into the correlation between descriptors and the separation. Physico-chemical understanding of the separation process is also connected to the explainablity of trained prediction models. The non-swelling property of ceramic membranes reduces the complexity of the separation process and consequently that of the data-driven prediction models required. At the same time, the known sensitivities of data-driven techniques are no less present here: the amount of (experimental) data needed to represent the chemical space of interest, dimensionality of input data, quality, and completeness of data. Understanding and predictability of OSN is what can lead to well understood designs and leverage the acceptance of the technology and its contribution to sustainable chemistry. We believe data-driven models can provide the predictive power looked for in OSN, as well as help unravel its complex transport process.