Liquid residence time distribution in micro-reactors with complex geometries By Alexandra Hopley Thesis submitted to the Faculty of Graduate and Postdoctoral Studies as required for the degree of Master of Applied Science In Department of Chemical and Biological Engineering Faculty of Engineering University of Ottawa © Alexandra Hopley, Ottawa, Canada, 2018 Abstract Micro-reactors, enabling continuous processes at small scales, have been of growing interest due to their advantage over batch. These advantages include better scaling, as well as improved mass and heat transfer, though many new challenges arise due to the small scales involved such as non-negligible entrance effects and significant pressure drops. The flow in coils, rectangular channel serpentine plates, mix-and-reside plates, and complex liquid-liquid mixing plates was investigated and characterized using residence time distribution (RTD) tests. A pulse test was used to determine the RTD curve shape of these reactors at flowrates ranging from 20 to 100 g/min. A semi-empirical, multi-parameter model was used to describe the asymmetrical curves, while the axial dispersion model was used to describe the symmetrical ones. The Peclet number is given in function of the Reynolds number for the liquid-liquid plates that were found to be near-plug flow (Pe > 100). In a continuous mixing plate, the Pe ranged from 190 to 475 with Pe increasing as Re increased. The effect of straight channel sections in micro-reactors is also evaluated. Longer straight segments between micromixers resulted in the development of unidirectional flow and the occurrence of tailing in the RTD. Finally, the suitability of a liquid-liquid plate for a reactive liquid-solid system is evaluated. The plugging is determined visually and by measuring pressure increase; pressure started to increase after 5 minutes and the experiment had to be halted after 10 minutes due to plugging. Parallels between the particle size distribution and the residence time distribution curves are drawn. The particle size distribution of silver chloride at low flow rates is much wider ii than at high flowrates. The average particle size at high flowrates was also much lower (≈69nm) than at low flowrates (≈112nm). iii Table of Contents 1 Introduction ............................................................................................................................... 1 1.1 Background ...................................................................................................................... 1 1.2 Motivations and Research Topic ...................................................................................... 4 1.3 Nomenclature ................................................................................................................... 7 2 Experimental ............................................................................................................................. 8 2.1 Experimental setup ........................................................................................................... 8 2.2 Micro-reactors Studied ................................................................................................... 12 2.2.1 Teflon Reactor Coils ............................................................................................... 12 2.2.2 Micro-reactor Plates ................................................................................................ 14 2.3 Data Analysis Method .................................................................................................... 18 2.4 Modelling ....................................................................................................................... 21 2.4.1 Single parameter turbulent models ......................................................................... 21 2.4.2 Laminar Models ...................................................................................................... 24 2.4.3 Multi-parameter Models ......................................................................................... 25 3 Results and Discussion ........................................................................................................... 26 3.1 Coil Reactors .................................................................................................................. 26 3.2 Serpentine Channel Reactor Plates ................................................................................ 29 3.2.1 Without mixing zone............................................................................................... 29 3.2.2 With mixing zone .................................................................................................... 31 3.3 Liquid-Liquid Reactor Plates ......................................................................................... 33 3.3.1 Effect of Micro-Mixer Spacing ............................................................................... 34 3.4 Comparison of the geometries. ...................................................................................... 41 4 Implications: Nanoparticle Crystallization in Micro-reactors ................................................ 44 4.1 Background .................................................................................................................... 44 4.2 Experimental .................................................................................................................. 45 4.3 Results and Discussion ................................................................................................... 47 5 Conclusions and Future Work ................................................................................................ 49 5.1 Conclusions .................................................................................................................... 49 5.2 Future Work ................................................................................................................... 51 6 References ............................................................................................................................... 52 iv List of Figures Figure 1 – Experimental setup ........................................................................................................ 8 Figure 2 – Pulse absorbance with pre-set volume setting on the syringe pump at a system flow rate of 50 g/min (left) and 100 g/min (right) without line being flushed (first 2 pulses) and with line being flushed (subsequent pulses) with tracer dye. ................................................................. 9 Figure 3 – A) Fluorecine dye injected into the water stream via a Tee, B) A5 micro-reactor plate, and C) light cell attached to tubes (north-south) and optical fiberglass (east-west). .................... 10 Figure 4 – Inlet and outlet signal response at 15 g/min in an LL plate with grouped mixers (pulse volume = 0.04 mL; dye concentration = 0.25 g/L; injection duration = 160 ms)......................... 11 Figure 5 – Dye concentration vs measured absorbance at 50 g/min. ............................................ 12 Figure 6 – Teflon coil reactors: single direction coil (left) and figure-8/inverting flow (right). .. 14 Figure 7 – Micro-reactor plates. .................................................................................................... 16 Figure 8 - Pressure loss in LL and serpentine plates .................................................................... 17 Figure 9 – LL (left), TG (centre) and SZ (right) mixers. .............................................................. 18 Figure 10 – Effect of analysis cut-off on RTD moments at 42 g/min in an LL plate with grouped mixers at a) first negative (most consistent method), b) first block of 3 negatives, c) about half the values are negative. ................................................................................................................. 19 Figure 11 – Time definitions......................................................................................................... 20 Figure 12 – (a) Dispersion model at different Pe and (b) boundary conditions [34]. ................... 22 Figure 13 – Ideal laminar model in a tubular reactor [20]. ........................................................... 24 Figure 14 – Single coil (left) and figure-8 coil (right) (1/4”) RTD curves. .................................. 27 Figure 15 – 1/4” Single coil (empty diamonds) and figure-8 coil (filled circles) sensor response curves at 60 g/min. ........................................................................................................................ 29 Figure 16 – RTD of serpentine channel plate with depth of 0.5 mm (left) and 1.0 mm (right). .. 30 Figure 17 – RTD of LL continuous mixing plate (left) and model RTD curves (right). .............. 34 Figure 18 – RTD of LL plate (equally spaced) and model RTD curves....................................... 36 Figure 19 – Pe vs Re for LL plates (continuous and equally spaced mixers) ............................... 36 Figure 20 – D vs flowrate for LL plates (continuous and equally spaced mixers) .................... 38 ax Figure 21 – RTD curves at various flow rates in LL plates with grouped mixers (left) and minimal mixers (right). ................................................................................................................. 39 Figure 22 – Semi-empirical model fit to the RTD of the LL plate with minimal mixers at 55 g/min. ............................................................................................................................................ 40 Figure 23 – Comparison of RTDs of geometries at 60 g/min...................................................... 43 Figure 24 – experimental setup for crystallization reaction with LL plate. .................................. 46 Figure 25 – Inlet pressure during crystallization experiments at 15 and 60 mL/min. .................. 47 Figure 26 – Plugging at reactor entrance at 60 mL/min at 1 minute (left) and 8 minutes (right). 48 Figure 27 – Particle size distribution in size 300 A7 LL plate after 5 minutes of runtime. .......... 49 v List of Tables Table 1 – Micro-reactor plate dimensions. ................................................................................... 15 Table 2 – Average time t , variance σ2, and skewness s3 of Teflon coils. ................................... 26 m Table 3 – Average time t , variance σ2, and skewness s3 of serpentine channel plates with and m without mixing zones. ................................................................................................................... 32 Table 4 – Average time t , variance σ2, and skewness s3 of LL reactor plates. ........................... 35 m Table 6 – Semi-empirical model parameters for LL plates with grouped and minimal mixers ... 41 Table 7 – Semi-empirical model parameters for LL plates with continuous and equally spaced mixers ............................................................................................................................................ 41 vi 1 Introduction 1.1 Background Micro-reactors are used for production of milligrams [1] to kilograms [2] of product per day as well as the steps in between. This makes them suitable for preclinical and clinical development in the pharmaceutical industry. During this time, new products are continuously produced and tested. As a result, modular setups are used, allowing for flexibility, speed and scalability. [3] While small-scale batches are highly flexible, the transition from batch to continuous production has been of growing interest in the pharmaceutical industry [4]. At production scale, discontinuous processes are expensive to run, partly due to vessels and equipment needing to be emptied and decontaminated between batches. Continuous processes require less human intervention and cleaning steps, resulting in less product losses, time losses, and use of water/cleaning supplies than batch processes. [5] Reduced handling also reduces potential safety hazards of the process. Safety is further increased by requiring smaller equipment and generally having better control over process parameters such as temperature and pressure. [6] Overall, transitioning to continuous processes increases efficiency and decreases cost, particularly as the process is scaled-up [5]. It is reported that if continuous production is desired at large production scale, it is better to start with continuous production at bench-scale [7]. Process scaling via both numbering-up (adding identical reactors in parallel) and sizing-up (increasing channel size and reactor volume) micro-reactors has been studied. Su et al. explored the numbering-up method for a photochemical reaction where sizing-up would have impeded the “homogenous irradiation of the entire reaction medium” [8]. However, sizing-up is often more 1 practical than numbering-up reactors [7]. It is therefore the ability to size-up micro-reactors while maintaining efficient heat and mass transfer that makes this technology so much more attractive than batch processes. Micro-reactors have efficient heat transfer due to their high surface to volume ratios [9] which comes from having a small hydraulic diameter [10]. Increased heat transfer is especially important with highly exothermic reactions and/or heat sensitive products [2]. To avoid temperature spikes, the maximum diameter can be calculated based on an overall heat transfer coefficient, heat of reaction and rate of reaction. The heat transfer coefficient is a combination of the convective resistance of the reactive medium, the wall resistance, contact resistance, and convective resistance of the thermal fluid. [10] Micro-reactors have short diffusion paths therefore it is widely assumed that they have good radial mixing when calculating conversion profiles, although this is not quite accurate [9]. Mixing strategy is one of the key issues with micro-reactors. At this scale, creating turbulence cannot be achieved purely by increasing flow rate as it results in pressure drops that are too high for most pumps. This is because the pressure drop is proportional to ~1/d5 in turbulent flow. To t induce radial mixing at more achievable flow rates, one of two mixing strategies are employed: active or static mixing. Mixing using external sources, such as vibrations, impellers, or pulsing, is defined as active mixing. Static mixing takes advantage of the flow energy, inducing mixing using specialized channel design. [11] The reactors studied in this paper use static mixing, specifically, curvature, expansion/contraction, and obstacles for changes to the velocity field magnitude and direction. The purpose of these geometries is to generate chaotic secondary flow patterns, thereby causing 2 reactants to more rapidly come into contact [12]. This is important because local concentrations have effects on selectivity and reaction rate [13]. However, mixing happens at different scales and is conventionally divided into micromixing, mesomixing, and macromixing [14]. i. Micromixing or molecular mixing occurs in large part at the Batchelor scale, which is defined as: 1 (1) 𝜂 = 𝜂 𝑆𝑐−3 𝐵 𝐾 Diffusion only becomes significant at this scale and leads to the homogenization of the mixture. [15] ii. Mesomixing occurs at the Kolmogoroff scale, at which energy is dissipated by the viscous forces, and is defined as [15]: 𝜈3 0.25 (2) 𝜂 = ( ) 𝐾 𝛹 iii. Macromixing occurs over the entire reactor volume and is associated to the flow regime and patterns within the reactor [14]. Just as with macro-scale channels, flow regime is determined by the Reynolds number 𝑢𝑑 𝜌 (3) 𝑡 𝑅𝑒 = 𝜇 using the hydraulic diameter d and with transition between the laminar and turbulent flow t regimes occurring at Re between 2000 and 4000 for a straight channel [16]. It should also be noted that the flow pattern is dependent on the vessel geometry while flow regime is not. 3 Therefore when referring to a laminar flow regime, it is understood that the Re is below 2000; referring to a laminar flow pattern means that the fluid has a parabolic profile [14]. 1.2 Motivations and Research Topic While micro-reactor technology in general has advantages over batch technology, there is an array of commercially available micro-reactors, not all of which will be appropriate for a particular application. Plouffe et al. developed a toolbox approach for choosing micro-reactors. This approach involves first determining the rate of reaction which has been divided into three categories. Type A reactions have reaction rates of milliseconds to seconds and are therefore, in general, mass transfer limited. Reactors with rapid mixing and heat transfer are most appropriate. Type B reactions have a reaction rate of several seconds to minutes and are therefore largely kinetically rate limited when in single phase. While good heat transfer and mixing may still be required at the entrance of the reactor, the appropriate reactors must have enough volume to allow the reactions to come to completion. Type C reactions have a reaction rate of several minutes to hours. Reactions of this type especially benefit from increased temperature and pressure control as a means to safely increase the reaction rate. [17] The second consideration of the Plouffe et al. approach is the phases involved. For a single phase, a mix-and-reside approach is suggested. For an immiscible liquid-liquid system, phases can separate if left to reside, therefore continuous mixing is suggested. When dealing with gas- liquid systems, higher pressure is favourable and mixing methods which break up gas bubbles are used at the entrance of the reactor. In solid-liquid systems, reactors which are not prone to plugging are used. [17] 4