Yeong, Kay Kin; (2003) Design and development of microreactors. Doctoral thesis (Ph.D), UCL (University College London).

Text Design_and_development_of_micr.pdf Download (22MB)

Abstract

The study of microreactors involves the miniaturisation of entire reactor systems on to structures ranging from several millimetres to several centimetres in length and height. Their large surface area to volume ratio results in efficient heat and mass transfer. As such, they allow better control of processes and quick quenching, which can lead to improved selectivity and conversion. Safety is improved as their flame trap dimensions prevent flame propagation. Also, previously unfeasible reaction routes may be opened up. Because of the way they are put together, there will be shorter lag time between R and D and industrial application. This is because scale-up is unnecessary, Instead, numbering-up of reactors can be used to increase production capacities. The first step in fabricating a microreactor is to appreciate the varied methods available. The most important and applicable methods are: photolithography, wet bulk machining, dry etching, chemical vapour deposition (CVD), LIGA, laser machining, ultraprecision machining, soft lithography and glass microfabrication. Combinations of these techniques will allow the fabrication of most reactors. In addition, since microfluidic systems typically take the form of thin fabricated sheets carrying different unit operations which can be stacked to form a complete reaction system, bonding techniques are also very important. Examples include anodic bonding, fusion bonding, eutectic bonding and adhesive bonding. A T-microreactor was designed for the gas phase partial oxidation of methanol to formaldehyde over a silver catalyst. It consists of a T-channel etched in silicon with the top of the channel sealed by a piece of glass. While simple, it served to demonstrate the main procedures involved in microfabrication: CVD, photolithography, wet and dry etching, and anodic bonding. Since methanol is a liquid at room temperature, the reactor had to incorporate a pre-heating section to vaporise methanol prior to it entering the reaction section. Three reactor designs (channel width 600 μm and depth 300 μm) were made with differing reaction channel lengths (15, 104 and 207 mm). This was done to widen the range of residence times. Catalyst deposition methods used included evaporation, manual placement of silver wires/foils, electrodeposition and an in-situ chemical reaction, all of which proved inadequate due to combinations of poor adhesion, activity and stability. The sealing of the reactor and the development of suitable interfacing with external equipment proved to be major challenges too, mostly due to the high temperature operation of the reactor. The final product was a silicon microreactor fabricated using Deep Reactive Ion Etching or KOH etching, with silver catalyst deposited through sputtering and sealed by a Pyrex plate through anodic bonding. High conversion was achieved using a mixture of 8.0 - 8.6% methanol and balance oxygen, without any helium to quench the reaction. A micro-falling film reactor (μ-FFR) was used to study the gas-liquid hydrogenation of nitrobenzene over a solid palladium catalyst. A robust catalyst had to be developed that was compatible with the reactor system, in terms of incorporation, suitability for continuous use, longevity and reproducibility. Several different catalyst incorporation methods were tested (sputtering, UV-decomposition of palladium acetate, wet impregnation on γ-alumina, incipient wetness on γ-alumina) with the final method satisfying all of these requirements. It was also shown that there were negligible external or internal mass transfer limitations. A batch reactor (3 m3, 0.4 mol nitrobenzene/1 ethanol, 20 bar H2 pressure, 125 °C, 7.89*10^-5 g cat/ml, 85% conversion Farrauto, et al (1997)) was estimated to produce 204 kmol aniline/m3 reactor daily (assuming no downtime). In comparison, the μ-FFR using the incipient wetness catalyst (0.1 mol nitrobenzene/1 ethanol, 0.5 ml/min., 1 bar H2 pressure, 60 °C, 5.4 g Pd/ml, 82% conversion) would have a daily production rate of 394 kmol aniline/m3 reactor. While a larger amount of catalyst was used, the process conditions were significantly milder and the catalyst was unoptimised, proving that the μ-FFR could enhance gas-liquid reactions that were mass transfer limited in practice. The liquid film thickness obtained in the μ-FFR at a range of flowrates (0.5-2.0 ml/min) was studied using confocal microscopy. It was found that the Nusselt equation greatly underpredicted film thicknesses. The temperature profile μ-FFR was measured using an IR-camera under non-reacting conditions by passing hot water through its heat exchanger (30-60 °C). The temperature distribution was reasonably uniform. However, the uniformity decreased slightly as the temperature increased (3% at 30 °C, 5% at 60 °C), with the central area of the plate cooler than the sides, and the top and bottom being the warmest. A simple model of nitrobenzene hydrogenation in the μ-FFR was made using gPROMS. Three Langmuir-Hinselwood-type kinetic expressions were used to fit the experimental data. No close matches were found and the model appeared to underpredict nitrobenzene conversions at high flowrates. This was likely due to inaccuracies in extrapolating liquid film thicknesses (i.e. the thicknesses were too large). In addition, mass transfer enhancement factors (e.g. the Marangoni effect, hydrodynamic instability, wave formation), which were not taken into account in the model could have affected the experimental results. The μ-FFR was also used to perform the asymmetric transfer hydrogenation of acetophenone to (S)-phenylethanol catalysed by a homogeneous rhodium complex that was activated by sodium isopropoxide, with isopropanol as the hydrogen source. The aim was to use the μ-FFR as an evaporator-reactor to remove acetone in order to drive the equilibrium forward. It was shown that the μ-FFR was much more efficient at evaporating acetone than a batch reactor (31% at a theoretical residence time of 9.3 s vs. 50% in 60 min for a batch reactor with nitrogen bubbling through). However, when the reaction was performed, the μ-FFR was found to have no impact on the conversion. This was most likely because the reactants had been pre-mixed in a flask (with nitrogen bubbling through), had reacted within it and the acetone produced had been evaporated within the flask itself. This is an example of a reaction system where a batch reactor was perfectly capable of driving the reaction to its limits, thus making a microreactor unnecessary. Based on the knowledge gained in this project, a general design algorithm was proposed in order to bring some structure to the discipline of microreactor engineering. It represents a summary and guide to the microreactor design route. A key aspect of this route is the use of modelling to aid in designing microreactors.

Download activity - last month

Export as XMLCSVJSON

Download activity - last 12 months

Export as JSONXMLCSV

Downloads by country - last 12 months

Export as XMLCSVJSON

Archive Staff Only

View Item