Chemical Engineering Science 58 (2003) 3961 – 3971 www.elsevier.com/locate/ces Reactive mass transfer at gas–liquid interfaces: impact of micro-scale uid dynamics on yield and selectivity of liquid-phase cyclohexane oxidation Johannes G. Khinast a ; * , Athanas A. Koynov a , Tiberiu M. Leib b a Department of Chemical and Biochemical Engineering, Rutgers University, 98 Brett Road, Piscataway, NJ 08854-8058, USA b DuPont Engineering, Wilmington, DE 19880-0304, USA Received 16 September 2002; received in revised form 11 April 2003; accepted 15 May 2003 Abstract The impact of single-bubble wake dynamics on the reaction-enhanced mass transfer and on the yield and selectivity of the cyclohexane oxidation reaction was studied using a two-dimensional CFD-reaction model that was developed by our group. Temperature and the concentrations of the (desired) intermediate and (undesired) nal products of this autocatalytic reaction were the parameters of this study. Two bubble types were studied: (a) a circular bubble with closed wake, and (b) an elliptical bubble with an unsteady, vortex-shedding wake. The main results of our work are: (1) Film theory over-predicts reaction-enhanced mass transfer since the assumption of an average lm thickness is not justied. In order to study fast reaction systems on a reactor scale using coarse-grid CFD codes, a full bubble model, or correlations based on it, should be incorporated as a sub-grid micro model. (2) The bubble wake does not contribute to mass transfer in systems where reaction rates are low. For fast reactions, the local mass transfer rate in the wake can increase by several thousand percent. (3) Vortex shedding causes qualitatively dierent mixing since patches rich in the dissolved gas are quickly convected away from the bubble. Bubbles that cause vortex shedding will lead to a signicantly higher conversion per volume than spherical bubbles. (4) Parallel–consecutive reactions with a high liquid-phase reactant concentration and with reaction rates that depend in an identical way on the dissolved gas concentration, are not micro-mixing sensitive in terms of selectivity. Since bubble shapes and sizes can be controlled by changing operating and design parameters, the yield of this reaction can be controlled. ? 2003 Elsevier Ltd. All rights reserved. Keywords: Reactive ows; Bubble dynamics; Yield and selectivity; Vortex shedding; Cyclohexane oxidation; Bubble column reactor 1. Introduction Bubble (slurry) columns are reactors that are widely used to carry out catalytic and non-catalytic reactions, such as hydrogenations, liquid-phase oxidations, hydroformylations and carbonylations. Reactor design, scale up, hydrody- namics and mass-transfer correlations have been reviewed by many authors, including Shah (1979), Ramachandran and Chaudhari (1983), Gianetto and Silveston (1986), Fan (1989), Beenackers and van Swaaij (1986), Krishna and Ellenberger (1995), Nigam and Schumpe (1996), Mills and Chaudhari (1999), and Dudukovic, Larachi, and Mills (1999). While these earlier studies were mainly based on ∗ Corresponding author. Tel.: +1-732-445-2970; fax: +1-732-445-2481. E-mail address: khinast@sol.rutgers.edu (J. G. Khinast). experimental work in small-scale laboratory columns, in re- cent years is has become possible to numerically study the reactor hydrodynamics using 2D and 3D Euler–Euler or Euler–Lagrange models. The main advantage of numerical simulation models is their applicability to large scale, indus- trially sized reactors. Euler–Lagrange models describe the liquid phase as a continuum and track each individual bubble as it rises due to buoyancy (Lapin & L  ubbert, 1994; Delnoij, Lammers, Kuipers, & Van Swaaij, 1997; Delnoij, Kuipers, & Van Swaaij, 1999; Lain, Br oder, & Sommerfeld, 1999). In contrast, Euler–Euler models represent the two phases as interpenetrating continua (e.g., Sokolichin & Eigenberger, 1994; Peger, Gomes, Gilbert, & Wagner, 1999; Peger & Becker, 2001; Pan, Dudukovic, & Chang, 1999; Pan & Dudukovic, 2000). The results of either one of these hydro- dynamic simulations can be used, rst, to study mixing in the reactor and, second, to predict the conversion, yield and selectivity towards the dierent reaction products. 0009-2509/$ - see front matter ? 2003 Elsevier Ltd. All rights reserved. doi:10.1016/S0009-2509(03)00311-7