UC Berkeley

A novel, nearly isothermal, nonselective-membrane separation process, membrane vapor extraction (MVE), efficiently recovers butanol from a dilute aqueous solution, for example, from a fermentation broth (Liu et al., 2015). In MVE, feed and solvent liquids are not in contact; they are separated by vapor. Therefore, compared to conventional extraction, MVE avoids formation of difficult-to-separate emulsions. In MVE, a semi-volatile aqueous solute (e.g., butanol) vaporizes at the upstream side of a membrane, diffuses as a vapor through the membrane pores, and subsequently condenses and dissolves into a high-boiling nonpolar solvent, favorable to the solute but not to water. Design analysis of a 1.5-m long, 30-m2membrane-area countercurrent MVE unit for processing 2-wt% aqueous butanol by dodecane solvent at 40 °C indicates over 90% recovery of the feed butanol with essentially no water loss and with very low energy requirement (Liu et al., 2015). The separation factor is over 1500. However, the published design study gives no experimental evidence for the calculated MVE separation. Here, we present experimental data to validate the MVE process. We use an omniphobic (i.e., hydrophobic and oleophobic), 0.2-µm pore-diameter Versapor®200 R membrane (Pall Corporation, Exton, PA) housed in a 6-cm wide by 10-cm long plate-and-frame channeled flow cell with 0.8-cm gap thickness. Membrane transfer area is 28 cm2. The membrane flow cell is designed for minimal axial concentration change and is operated in the transient mode between two recirculating flow loops. 2-wt% aqueous butanol is extracted into dodecane or mesitylene at 25 or 40 °C. Also, 1.5-wt% furfural is extracted into dodecane at 40 °C. Since vapor transport across the membrane contributes minimal resistance, MVE performance is governed by mass transfer through feed and solvent boundary layers. Mass-transfer coefficients are determined from the Graetz-Lévêque analysis of laminar thin-slit flow (Bird et al., 2002). Predicted extraction performance agrees well with experiment using no adjustable parameters. Consistent with the initial multistage-design analysis (Liu et al., 2015), our new bench-scale experimental results confirm that MVE is a viable separation process to recover dilute semi-volatile biosolutes from water with minimal energy requirement. Preliminary analysis of downstream solute recovery from the extract via distillation is more efficient than that for pervaporation because of insignificant water carry over through the MVE membrane.