(235a) Horizontal Taylor Flow Hydrodynamics, Pressure Drop and Overall Mass Transfer in a Capillary Reactor

Introduction. The interest for applying micro structured reactors and monoliths for performing gas-liquid-solid (catalytic) reactions has increased considerably over the past years [1,2,3,4]. The hydrodynamics of the gas-liquid flow in the small channels (diameter < 1 mm) and the relation to pressure drop and gas-liquid and liquid-solid mass transfer has to be well understood for a proper design of these reactors. Taylor flow is the most studied regime in these systems in the literature. Pressure drop studies have been mostly performed in optically transparent channels without chemical reaction and without a catalytic coating on the channel walls. High speed imaging techniques were then used to obtain information about the hydrodynamic parameters. Three phase mass transfer studies have been performed mostly in devices which are not transparent. The hydrodynamic parameters were then obtained from visual observation of the flow patterns outside the actual reaction channel, assuming that the conditions in the reaction channel were similar to those in the viewing section. Mass transfer and pressure drop have therefore seldom been studied simultaneously in the same device and/or in the same experiment. We believe however that this combined measurement will contribute to a proper interpretation and understanding of mass transfer processes in micro channels. Objective of this work. Therefore, in this work, both overall mass transfer and pressure drop measurements have been simultaneously performed during Taylor flow at different operating conditions. High speed imaging techniques were used to obtain information on the hydrodynamic parameters (Figure 1). The catalytic hydrogenation of phenyl-acetylene to styrene and ethylbenzene was performed in a catalytically active fused silica capillary with an inner channel diameter of 250 µm and a length of 4.5 m. The capillary wall was coated with a 120 nm thick porous TiO2 layer loaded with 1 wt.% Pd. Catalyst activity was demonstrated by hydrogenating a 10 vol.% phenylacetylene in methanol solution in the capillary. At a temperature of 30 0C the catalyst activity was in the order of 1 molH2molPd-1s-1 [5]. In subsequent experiments isopropanol was used because of its higher boiling point compared to methanol. The catalyst activity could then be increased by increasing the reaction temperature without reaching the boiling point of the solvent. The hydrogen partial pressure was controlled by diluting the hydrogen flow with nitrogen. Pressure drop. First, pressure drop experiments were performed with in a similar capillary with a length of 1 m. A water /nitrogen system was chosen instead of an isopropanol/ phenylacetylene/hydrogen/nitrogen system. More literature data are available for the first system allowing for a better comparison of the data and model described in this work to the work of other authors. Nitrogen was fed and the superficial gas velocity (Ug0) was varied between 0.03 and 0.52 m/s (the velocities are given at a pressure of 1 bar and a temperature of 298 K). The superficial velocity (Ul) of water was varied between 0.11 and 0.36 m/s. The experiments were performed at a temperature of 292 K. The liquid flow was split into two equal flows which were contacted with the gas flow in a cross-mixer (Figure 2). A differential pressure sensor (Honeywell 26PC 05 KD, response time of 1 ms) measuring the pressure difference with ambient pressure was connected close to the capillary inlet and 13 cm upstream from the cross-mixer to ensure a fully developed Taylor flow passing the sensor. All connectors and tubing had an inner diameter of 250 µm. The flow exited the capillary under ambient conditions. The capillary is optically transparent, even with the catalytic coating present. Images of the Taylor flow were recorded using a high-speed camera connected to a microscope at a frame rate of 2500 frames per second for 2 seconds. Images were recorded for 58 combinations of gas and liquid velocities and analyzed in Matlab using scripts developed by the authors. The gas bubble velocity, the number of gas bubbles formed per unit time and the gas and liquid slug size distributions were estimated directly from the recorded images. A mass balance based model for Taylor flow was then applied to the data to obtain the gas hold-up and the film thickness of the liquid film surrounding the gas bubble [6]. The recordings of the pressure signals and the video images were synchronized. The pressure drop over the capillary is caused by the liquid flow in the liquid slug, assuming that the liquid film surrounding the gas bubble is stagnant and that the pressure drop in the gas bubble is negligible. The gas bubbles influence the pressure drop because their presence disturbs the otherwise parabolic velocity profile in the liquid slug. The average velocity in the liquid slug is equal to the average superficial velocity in the channel (Ug+Ul) (Figure 3). Note that the superficial gas velocity (Ug) in equations 1-4 is the velocity at the conditions in the channel. Calculations show that the pressure drop caused by the acceleration of the flow due to expansion of the gas phase is in the order of 0.1% of the total pressure drop. Acceleration effects are therefore not taken into account. The pressure drop ((dP/dz)exp) over the capillary can then be described in terms of an apparent friction factor fobs times the Reynolds (Re) number, see Equation 1. Other parameters in equation 1 are: the liquid slug length (Ls), the gas bubble length (Lb), the hydrodynamic diameter (Dh) and the liquid viscosity (µl).Figure 4 shows that the apparent friction factor multiplied with the Reynolds number decreases with increasing liquid slug length. This value approaches the single phase limit value of 16 for infinitely long slugs. This indicates that the number of bubbles per unit channel length determines the excess pressure drop over single phase liquid flow. This was already reported earlier for channels with an approximately 10 times larger diameter and for various liquids [1]. Mass transfer. A model was derived describing the pressure (P) and the hydrogen molar flow rate (FH2) as a function of axial position z in the capillary based on the following assumptions: 1. The gas and liquid phases exhibit plug flow behavior. Since both the gas bubbles and the liquid slugs move through the capillary as isolated packages with internal recirculation cells, both of them are considered to be well mixed without axial dispersion. 2. No nitrogen dissolves into the liquid phase. Since hydrogen is consumed from the liquid phase and nitrogen is not, the nitrogen flux from the gas into the liquid is considerably smaller than the hydrogen flux. The partial pressure of hydrogen will therefore hardly be influenced by nitrogen dissolving in the liquid phase. 3. The reaction is fully mass transfer limited. The reaction rate is then zero order in phenylacetate and first order in hydrogen. 4. Only the liquid flow in the liquid slugs contributes to the pressure drop. The same assumptions were made that lead to equation 1. These assumptions lead to the following set of equations (2-4) from which the overall mass transfer coefficient kov and the observed friction factor fobs were estimated. The model reduces to two independent equations with these two unknowns if equation 2 is substituted into equations 3 and 4. The product concentrations at the exit of the capillary were determined by GC measurements. The total amount of reacted hydrogen was then calculated from the overall mass balance. Thus, both the inlet and outlet hydrogen molar flows (FH2) and the pressure drop were known from the measurements. The bubble velocity (ub), the gas bubble length (lb) and the liquid slug length (ls) were obtained from image analysis. The molar flow of nitrogen (FN2), the cross-sectional channel area (Ar), the channel diameter (Dh), the liquid viscosity (µl), and the hydrogen Henry coefficient (HH2) were constant. The mass transfer coefficient kov and the observed friction factor (fobs) were obtained by simultaneously solving equations 2, 3 and 4 with the appropriate boundary conditions using a boundary problem solver routine in Matlab. The paper will report the dependences of these measured overall mass transfer coefficients and observed friction factors on the Taylor gas bubble velocity, the gas bubble and liquid slug size distributions, and the film thickness of the liquid film surrounding the gas bubble, at different hydrodynamic operating conditions. [1] M.T. Kreutzer et al., Chem.Eng.Sc., 60 (22) (2005) 5895-5916. [2] M.W. Losey et al., Ind.Eng.Chem.Res., 40 (12) (2001) 2555-2562. [3] V. Haverkamp et al., Ind.Eng.Chem.Res. (2007). [4] V. Hessel et al., Ind.Eng.Chem.Res., 44 (25) (2005) 9750-9769. [5] E.V. Rebrov et al., Conference proceedings IMRET 10, 2008, AIChE spring national meeting. [6] M.J.F. Warnier et al., Chem.Eng,J., 135S(-), S153-S158