Chemical Engineering Journal 167 (2011) 718–726 Contents lists available at ScienceDirect Chemical Engineering Journal journal homepage: www.elsevier.com/locate/cej Scale-up concept of single-channel microreactors from process development to industrial production Norbert Kockmann ∗ , Michael Gottsponer, Dominique M. Roberge Lonza AG, Exclusive Synthesis, Process Research, 3930 Visp, Switzerland a r t i c l e i n f o Article history: Received 16 April 2010 Received in revised form 1 August 2010 Accepted 4 August 2010 Keywords: Process development Scale-up Pilot-scale production Microstructured reactors Microchannel flow Pressure loss Energy dissipation rate Mixing time Heat transfer Reaction kinetics Flow reactor Reactor safety Reactor design a b s t r a c t Microreactors can perform chemical reactions in tiny channels using continuous-flow processes. The microreactor team at Lonza has designed and tested a series of microstructured devices in continuousflow plants, and performed lab studies of pharmaceutical reactions with successful transfer to commercial production. Microreactor design and scale-up concept is guided by simple correlations, which are described here and displayed in comprehensive diagrams for hydraulic diameter over typical range of flow rate. This leads to a consistent and straightforward scale-up pathway for single-channel microreactors avoiding parallelization from lab development to pilot-scale production. © 2010 Elsevier B.V. All rights reserved. 1. Introduction Biology is a paradigm for many technical systems, since nature has generated many efficient, perfectly adapted systems, continuously improving them through evolution. Organisms use tiny channels to transport fluids to supply cells and limbs or to perform chemical reactions and separations [1]. Microreactors constitute a similar system, where complex chemical reactions are performed in tiny, highly adapted channels for improved process conditions [2]. Microreactor technology is a new field in chemical engineering and organic synthesis that embodies the principles of Green Chemistry [3,4]. In tiny channels smaller than a millimeter in diameter, chemical and biochemical transformations can be carried out that dramatically enhance mixing and heat transfer. The small internal volume also lowers the consumption of energy and raw materials, thereby increasing safety and economy. To achieve optimal performance of the microreactors, engineers and chemists have to know exactly how the interplay between flow, mixing, heat transfer and ∗ Corresponding author. Tel.: +41 027 948 65 80; fax: +41 027 947 65 80. E-mail address: norbert.kockmann@lonza.com (N. Kockmann). 1385-8947/$ – see front matter © 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.cej.2010.08.089 chemical reaction works. Typical reactions are rapid or hazardous [5] or with unstable intermediates, but can be safely operated under intensified process conditions [6]. This contribution describes a modular, multipurpose microreactor platform and the consistent design for scale-up from process development to ton-scale production of pharmaceuticals. The reactor consists of modular, microstructured plates in close contact with the heat transfer medium. The microstructured plates ensure the closed handling of hazardous reagents in a single pass with defined mixing and residence time conditions. Only one single channel is employed due to better flow and process control of meta-stable reagents, which can precipitate and plug the reactor [7]. The flow rate through the single channel determines flow velocity, Reynolds number, flow regimes, and pressure loss in the system. The pressure loss in the channel elements is a measure for the energy dissipation rate and the typical mixing time for convective mixing. Heat transfer in the reactor setup is composed of contributions from the reactor channel, cooling channel, and wall resistance and has to be correlated to the energy release from the reaction [8]. Here, only two typical cases are discussed, the reaction runaway in case of stopped flow, and the internal heat transfer coefficient due to convective flow. The derived correlations are displayed in design diagrams indicating reactor size for typical flow N. Kockmann et al. / Chemical Engineering Journal 167 (2011) 718–726 Nomenclature af AC AS b cp CA Cf Ch Cm EA d dh (−Hr ) h k0 L m n N Nu p P Pr q̇ r R Re Sc tm tR TW T Tad V̇ w temperature conductivity of the fluid (m2 s−1 ) cross-sectional area (m2 ) wetted surface area (m2 ) channel width (m) heat capacity (J kg−1 K−1 ) concentration of key component (mol m−3 ) friction coefficient heat transfer constant mixing coefficient activation energy (J mol−1 ) post or column diameter (m) hydraulic diameter (m) reaction enthalpy (J mol−1 ) channel height (m) reaction rate coefficient (s−1 ) channel length (m) fluid mass (kg) exponent number of mixing channel elements Nußelt number pressure difference (Pa) wetted perimeter (m) Prandtl number specific heat flux (W m−3 ) reaction rate (mol m−3 s−1 ) universal gas constant (J mol−1 K−1 ) Reynolds number Schmidt number mixing time scale (s) reaction time scale (s) temperature at the wall (K) temperature difference (K m−1 ) adiabatic temperature increase (K) volumetric flow rate (m3 s−1 ) mean (fluid) velocity (m s−1 ) Greek symbols ˛ heat transfer coefficient (W m−2 K−1 ) ˇ eigenvalue ε energy dissipation rate (m2 s−3 )  pressure loss coefficient  dynamic viscosity (N s m−2 )  heat conductivity (W m−1 K−1 ) f friction factor of channel flow  kinematic viscosity (m2 s−1 )  density (kg m−3 ) rates leading to a consistent scale-up procedure. Together with the modular concept of pumps, heat exchangers, Lonza reactors allow flexible and versatile set-up for laboratory development up to pilot plant production. In one such Lonza plant, a multi-ton campaign for a pharmaceutical intermediate was carried out in 2009. 2. Microreactor characteristics Complex microstructured channels with meandering curves, corrugated walls, or repeated contracting and diverging channel elements generate secondary flow structures at high flow velocities, which lead to efficient and fast mixing. Large internal specific surface enables enhanced heat transfer and good temperature control leading to good control of reaction rates and heat release. 719 Together with the small internal volume, microreactors allow for safer process conditions compared to batch processes. The above described characteristics indicate what kind of chemical reactions are suitable for continuous flow, microstructured reactors. The reagents have to flow through the tiny channels without precipitation. Hence, starting material, intermediates, side products and products have to be soluble in the working fluid. Reaction kinetics and enthalpy determine characteristic reaction time and adiabatic temperature rise of the reagents. Competitive reactions such as consecutive or parallel reactions lead to side product formation, which lowers the yield and may complicate work-up processes. Based on the characteristic reaction time, the following classification was set up to facilitate the reactor design [9]. • Type A reactions have a characteristic reaction time below 1 s and are mixing controlled. The generated reaction heat has to be removed to avoid hot spot formation and side products from parallel-competitive reactions. Rapid mixing and correct control of the stoichiometry also avoids consecutive-competitive reactions. Typical reactions are of cryogenic type such as organometallic reactions [10–14]. • Type B reactions are rapid in the range of several minutes (<10 min), but mixing in microchannels is always faster. Enhanced heat exchange over the entire reaction period leads to good yield in temperature sensitive reactions. Proper control over the stoichiometry leads to high yield and low side product formation in consecutive reactions. Examples are coupling reactions or Simmons–Smith reactions [15]. • Type C reactions are slow and show hazardous tendency. Autocatalytic reactions or decomposition potential of intermediates or product belong to this class. The excellent temperature control in microchannels as well as the low internal volume gives higher process safety [16]. • Type D reactions are all reactions not belonging to the above described classes. These reactions can be accelerated by harsh process conditions [17], such as high reaction temperature, high pressure, enhanced reaction activation, or high active reagents. From the above classification, the question arises, how to design continuous-flow systems with microreactors, which fulfils the following purposes: proper control of stoichiometry, robust to plugging or at least fast plugging detection, modular for different reaction types, rapid mixing and volume providing, modular for different phases involving gas/liquid and liquid/liquid. The answer to these requirements is partially included in problem formulation: a modular reactor plate setup with single microstructured channel for excellent mixing and proper flow control. It has to be flexible for process development in the lab, reactor development, and production on different scales with a consistent scale-up approach. 3. Equipment overview for single channel microreactor The microstructured reactor plates are made from corrosion resistive material and can fulfill various task in modular set up. Plates are designed for heat exchange to bring the reagents to reaction temperature. Mixing plates include a mixing channel as well as wider channel elements to provide reactor volume for residence time. Finally, reactor plates have only wide channels for heat exchange and residence time. In an approach to standardize Lonza’s reactor design, the sizes chosen for the production plates are based on the European paper sheet format DIN A4, A5, and A6 standard. The plate area is doubled by each size step with the result that also heat exchange area and reactor volume are doubled. The scale-up concept becomes apparent and is related to the reaction classes. Thus, for Type A reactions, the aim will be: 720 N. Kockmann et al. / Chemical Engineering Journal 167 (2011) 718–726 Fig. 1. Left hand side: details of a typical mixing channel with contacting element, here a nozzle type, and succeeding tangential mixing elements. The characteristic channel diameter is taken at the narrowest element. Right hand side: mixing channel with test reaction in the Lab-Plate reactor. Fig. 2. Details of Lab-Plate reactor for development purposes. Left hand side: DN50 flange with view glass for optical access, cooling tube connections point to bottom side; right hand side: explosive sketch with cover plate, cooling block, microchannel plate, view glass and flange (from left). – to ensure sufficient cooling between the reactor plates and mixing channel, – to provide short mixing times in tiny, complex channels with comparable high pressure drop. For Type B reactions, the aim is different and will be: – to maintain volume for sufficient residence time for the reaction and the same area-to-volume ratio for enhanced heat transfer, – to optimize mixing quality by choosing a pressure drop as low as possible. Type C reactions can be performed in conventional equipment such as static mixers. However, for Type A reactions, all the reactor plates will be microstructured as mixing and heat exchange are the dominant factors, while for Type B reactions the same plate depth will be used for the different plate format keeping constant the surface-to-volume ratio. The gradual size increase of reactor plates (multi-scale approach) and appropriate channel geometry allow operating the microreactor at very high flow rates up to 600 mL/min. A small, compact plate device called the Lab-Plate reactor has been realized to visualize the flow inside the microstructured channel. It enables reactor channel design and process development with low flow rates, when reagent availability is still limited. Conditions are similar to capillary chemistry as well as in larger reactor devices with the advantage that the reaction zone can be inspected and viewed. The fluid entering the microchannel within reactor plate passes through the entrance, contacting element, several mixing and residence channel elements, each with individual design, see Fig. 1. The entire reactor consists of cooling block with cover plate, microstructured plate, view glass, and flange housing, see Fig. 2. The fluids are sealed by conventional O-rings against the environment, no direct sealing between reagents and heat exchange medium is necessary. The modular set up allows the integration of several microstructured plates as well as reactor integration into other flow equipment. Mainly for Type B reactions, a plate stack reactor was developed by Lonza [18], see Fig. 3. The reactor was initially based on the multi-scale approach, where differently sized plates are used and adapted to the reaction needs [7]. For example a tiny channel may be used at reaction start (when heat generation is strong) followed by a gradual size increase of the plates to accommodate slower reaction rates (less heat evolution). With such a design, heat transfer is optimized, while pressure drop is minimized coupled with a large gain in volume (up to several mL). In addition, the reactor may be combined with conventional heat exchangers and tube equipment to gain volume of several liters for residence times of several minutes. Fig. 3. Typical setup of plate stack reactor with A6 and A5 size. The reagents are precooled or preheated, contacted with the correct stoichiometry, completely mixed and hold on cooling temperature for a certain time. N. Kockmann et al. / Chemical Engineering Journal 167 (2011) 718–726 The modularity and versatility of the single plate approach allow the development of plates appropriate to any kind of application. Plates for rapid mixing, gas–liquid dispersion, and multi-injection applications were designed and manufactured demonstrating the multipurpose character of this technology. In Fig. 3, the individual reactor plates made from Hastelloy are sandwiched between aluminum plates with high thermal conductivity for the thermal fluid passage leading to a very compact reactor. In this way, the thermal fluid layer is not directly fixed onto the reactor plates allowing a cost effective manufacture as well as quick and easy adaptation to different reaction conditions. The overall reactor is robust, allows high flow rates of the heat exchange fluid, and can sustain pressures above 100 bar on the reaction side. In many cases, especially with viscous systems and low temperature applications, the pressure drop may become very important at high flow rates. In addition, the mixing zone is often the plate section that consumes the larger pressure drop. Consequently an enlargement of mixer elements at higher flow rates drastically reduces the overall pressure drop. In general, no loss of performance is observed as long as the same energy dissipation rate in the mixing zone is maintained (Watt per liter). Thus, the mixing zone becomes the only scaled factor that is considered in this reactor technology and it must be properly designed and adapted. The operation of a microreactor with one single channel and complete avoidance of device parallelization is a must for successful scale-up of processes during their development from lab to pilot scale. The main goal in lab development is to achieve a robust process as chemical systems are often meta-stable and tend to form deposits that are more or less stable over time. Precipitation or fouling creates an unpredictable pressure drop behavior. The reactor technology must facilitate timeliness and flexibility in process development and keep the chemical engineering aspects to a minimum. At this project stage the reactor should not be design and fabricated, but chosen out of few standard equipment for good mixing and heat transfer in a certain range of flow rate and liquid viscosities. The necessary residence time for the chemical reaction is provided by enough process volume in the reactor and succeeding tubes. The modularity of the equipment plays the major role, plate and tube elements must be simply added in series to gain the necessary volume for the reaction. Here the use of one single channel together with robust, pulsation-free high pressure pumps always ensures correct feed balance and stoichiometry, as well as ability to clean during and after operation. In the following, channel design correlations for mixing and heat transfer are derived from volumetric flow rate considering flow velocity, pressure drop, mixing time, and heat transfer. 721 5. Convective flow In a single channel with rectangular cross-section, the volumetric flow rate determines the flow velocity in a given geometry with nominal diameter, hydraulic diameter, and cross-sectional area. Microchannels typically have rectangular shaped cross-sections with a width b and a depth h. The hydraulic diameter indicates the characteristic length for the flow through this cross-section AC = bh and is determined for a rectangular cross-section with perimeter P to dh = 4AC 2bh = P b+h (1) For a circular cross-section, the hydraulic diameter equals the geometric diameter, for an infinite wide slit (b → ∞) the hydraulic diameter is twice the depth of the channel. The cross-sectional area of the rectangular channel AC = bh can be approximated by the square of the hydraulic diameter. Ac ≈ dh2 (2) The equality holds for square shaped channels, and the error is less than 10% for aspect ratios b/h from 0.52 to 1.92. For larger aspect ratios the approximation has to be replaced by the correct correlation. To show some trends, we will work with this correlation in the following. The mean flow velocity in a cross-section is determined by the volumetric flow rate V̇ and the cross-sectional area V̇ = AC w (3) Using the approximation in Eq. (2) the hydraulic diameter dh can be correlated with the volumetric flow rate and the mean flow velocity. dh =  1/2 V̇ w (4) The Reynolds number determines the flow regimes in the mixing channel and is defined as the ratio of momentum forces to viscous forces. Re = d w dh w = h   (5) With Eq. (3) follows dh = V̇  Re (6) Eqs. (4) and (6) are displayed in Fig. 4 with typical flow velocities of water, 20 ◦ C, in the microchannel of 0.1, 1.0, 5.0, and 30 m/s 4. Design pathway from flow to kinetics The aim of this part is to describe the design parameters of the single channel microreactor in terms of hydraulic diameter and volumetric flow rate. Together with channel geometry and fluid properties, the mean flow velocity and Reynolds number Re can be determined in the channel. Both give an indication for the flow regime, pressure drop, and energy dissipation rate in the mixing channel. Together with the diffusivity of the reacting species, the energy dissipation rate is the measure for the mixing time in the mixing channel. Temperature control of the reagents by enhanced heat transfer is the second important design parameter. Kinetic data for reaction rate as well as thermodynamic data for reaction enthalpy have to be compared with the typical transport characteristics of species and heat. Simple correlations are used to show trends and general design characteristics. Important is to note, that the following generic correlations have to be adjusted to specific microchannels with data from literature or own experimental investigations. Fig. 4. Typical flow velocities (m/s) and Reynolds numbers in rectangular channels with water at 20 ◦ C as working fluid. The horizontal bar indicates the typical flow rate range for a channel with a diameter of 0.5 mm as orientation. 722 N. Kockmann et al. / Chemical Engineering Journal 167 (2011) 718–726 Table 1 Properties of organic solvents in relation to water properties at 295 K. solv water 20 ◦ C solv water 20 ◦ C Prsolv Prwater 20 ◦ C Toluene 295 K 235 K 0.24 0.89 0.22 0.25 0.38 1.23 Ethanol 295 K 235 K 1.63 6.56 0.28 0.31 2.50 7.44 THF 295 K 235 K 0.55 1.14 0.23 0.45 1.29 1.35 diameter dh and length l. The ratio of both length scales should be kept constant during scale up or down to have similar flow conditions. For mixing channel, the pressure loss coefficient  can be expressed with the following correlation =N and typical Re numbers of 200, 2000, 10,000, and 100,000, respectively. For orientation, the horizontal bar indicates the typical flow range in a channel with 0.5 mm internal diameter. This bar of typical flow range at 0.5 mm diameter is also shown in the following three figures. In Fig. 4 and in the following diagrams, water at 20 ◦ C is taken as reference fluid for the properties. However, other fluids than water are often used in chemical synthesis as well as for heating or cooling purposes. In Table 1, the property ratio of further typical organic solvents such as toluene, ethanol or tetrahydrofurane (THF) are listed to the properties of water at 20 ◦ C. The kinematic viscosity, heat conductivity and Prandtl number (Pr = /a) are main properties in the trend correlations given in this contribution. The ratios are given at two different temperature levels of 295 K and 235 K with data from [19]. The maximum ratio is given by 7.44 for Pr number ratio of ethanol to water at 235 K, the minimum ratio with a value 0.22 for heat conductivity at room temperature. All ratios are less than one order of magnitude, hence, trends are also valid for the mentioned fluids to a certain range. The ratios in Table 1 can also be read in the following way: the Re number, which is determined by the kinematic viscosity, is 4 times higher for toluene at 295 K or 6.5 times lower for ethanol at 235 K as water at 295 K, same flow velocity and hydraulic diameter assumed. The ratio of heat conductivity is always lower than unity indicating lower heat transfer in organic solvents compared to water. Due to the high velocities in complex winding and meandering structure of the channels, the flow is not laminar for Re > 100, but shows transitional behavior to turbulent characteristics. Under these flow conditions, vortices appear in bend flow, which start to fluctuate with increasing Re number and lead to chaotic flow structures, see [20]. Flow characteristics are often determined by highest flow velocity in narrowest channel, hence smallest hydraulic diameter. 6. Pressure loss and mixing time in microchannels Cf Li Ren dh,i (8) The exponent n of the Re number in the dominator depends on the flow regime in the channel. For straight laminar flow and fully turbulent flow, n is 1 or 0, respectively. In complex channel elements, transition flow between straight laminar and fully turbulent flow is often dominant, leading to non-integer number of n. For this flow regime with Re number from 100 to 1000 [21], the exponent was determined to 1/3 in a T-shaped micromixer, while Majumdar et al. [22] found a value of 1/4, which correlates with the well-known Blasius equation. With a pressure loss coefficient for N channel elements, the correlation for the pressure loss in the mixing channel is simplified to p = N Li Cf  · w2 dh,i Ren 2 (9) Using the approximation in Eq. (2) gives the following correlation dh =  NLi Cf 2p n  V̇ 2−n 1/(5−n) (10) This correlation is depicted in Fig. 5 for n = 0.25 (transitional flow) and a pressure loss of 1.0, 5.0, and 20 bar. Mixing depends mainly on the local energy dissipation rate [23] and geometry of the channel. The channel shape guides the flow and causes flow deflection. Besides shear forces, new flowperpendicular forces act on the fluid and generate secondary flow structures, vortices, and recirculation zones [8]. From the theory of chaotic advection, the center of vortex formation is called elliptic point; intersection points of separation lines between two vortices with another separation line or with the channel wall are called hyperbolic points. The position of the interface between two components or two fluids in relation to vortex separation line of characteristic points is a very important issue for the design of efficient mixing channels. A rapid change of flow vortices by alternating channel elements or by repeated deflecting flow leads to efficient flow mixing. To generate these flow structures and vortices, the fluid needs mechanical energy consumed from the pressure of the fluid. Hence, the pressure drop per unit volume is The pressure driven flow in microchannels needs mechanical energy to drive fluids through the channel. This mechanical energy is spent to the system by a pump ahead of the reactor and is dissipated within the channel. The dissipation leads to shear flow, vortex generation, and internal flow mixing, depending on the typical channel geometry. The pressure loss is typically correlated with the flow velocity and to the geometry related pressure loss coefficients. Considering long straight channel elements with length L and short elements, where the flow is turbulent, the pressure loss is represented by Bernoulli’s equation with kinetic energy and neglecting potential energy. The pressure loss consists of long straight elements (i (Li /dh,i )) and short elements with  i . p =  i i Li + i dh,i   2 w 2 i (7) Typically, mixing happens in the first channel part with N identical channel elements with mean characteristic fluid velocity w. The element can be characterized with a characteristic hydraulic Fig. 5. Typical values of pressure loss in bar and mixing time in seconds together with Re = 2000 and flow velocities of 1 and 5 m/s. N. Kockmann et al. / Chemical Engineering Journal 167 (2011) 718–726 723 the measure for mixing, expressed in the energy dissipation rate. ε= pV̇ pw = m L (11) The channel length, over which the pressure loss occurs, is equal to the length of the sum of all mixing elements L = NLi . With more energy dissipated in the mixing channel, the mixing time gets shorter due to smaller fluid structures, where diffusion of the species occurs as last measure of mixing [17,23,24]. The species diffusion is represented by the Schmidt number, the ratio of the kinematic viscosity to the diffusivity of the main species. tm = Cm Sc   1/2 (12) ε The mixing coefficient Cm was given by Bourne [23] as engulfment rate to 17.3. With the approximation in Eq. (2) and the above correlations, the hydraulic diameter can be correlated to the volumetric flow rate and the typical mixing time dh =  t 2/(7−n)  C f m Cm Sc 2Li  n−1 V̇ 3−n 1/(7−n) (13) Both Eqs. (10) and (13) are displayed in Fig. 5 for typical pressure losses of 1.0, 5.0, and 20 bar and mixing times of 0.1, 0.01, and 0.001 s. For many industrial applications, these values display an appropriate range. The horizontal bar indicates the typical flow rate range at 0.5 mm channel diameter, see also Fig. 4. Fig. 5 shows for transitional flow (n = 0.25) in the mixing channel, the pressure drop and the mixing time have nearly the same incline. Hence, slightly higher pressure drop in larger channels will lead to similar mixing times in these channels. In pure laminar flow (n = 1.0), where the pressure loss is proportional to inverse of the Re number, similar mixing time in reactors with larger diameter consumes remarkable higher pressure loss. In turbulent flow (n = 0), pressure drop and mixing times scale with the same incline with the volumetric flow rate and depends only on channel geometry and surface roughness. For most industrial applications, a mixing time of 0.1 s is sufficient. Often, heat transfer will be the limiting step. Only for analytical devices, rapid mixing below 1 ms is important [22]. 7. Heat transfer and reaction runaway in microchannels Heat transfer and runaway of exothermic reactions are closely related with the energy balance in a channel element. Two cases are discussed here, the convective heat transfer in the microchannel and the heat conduction in a case of stopped flow of the reacting fluid to avoid reaction runaway. The latter case is important for safety discussion in case of pump failure or similar event, when the reaction should be under control. Because the convective heat transfer in a channel is always higher than heat conduction in a non-moving system, this case gives an upper limit of the channel size [25]. Larger channels may lead to reaction runaway in case of interrupted flow. The internal convective heat transfer on the reaction side is described by the heat transfer coefficient ˛i , in dimensionless form with the Nußelt number Nu ˛i = Nu dh (14) Textbooks describe many ways to determine the heat transfer coefficient or Nu number in channel flow. Here, we use the Lévêque correlation [26]:  Nu = Ch total · Re2 Pr dh L 1/3 (15) Fig. 6. Internal heat transfer coefficient (water, 20 ◦ C) with 10 and 50 kW/m2 K and typical reaction conditions, adiabatic temperature rise of 50 and 150 K, typical characteristic reaction time of 1 and 10 s, describing the difference between Type A and Type B reactions. The constant is Ch = 0.404, which is valid for complex geometries such as fixed beds or tube bundles in cross flow with Re numbers above 1000 [26]. The pressure loss coefficient is introduced from the transitional flow regime between laminar and turbulent flow from Eq. (8). Combination of the correlations above yields for the characteristic hydraulic diameter depending on flow rate and internal heat transfer coefficient. dh =   2−n  Cf Pr V̇  Ch f ˛i 3 1/(5−n) (16) Other pressure loss correlations as well as heat transfer correlations can also be integrated into the above correlations to adjust the correlations to other geometries or flow conditions. The purpose here is to show trends of heat transfer scale-up characteristics with increasing flow rate and increasing hydraulic diameter. Typical values of internal heat transfer coefficients for water 20 ◦ C with 10 kW/m2 K and 50 kW/m2 K are given in Fig. 6. The horizontal bar indicates the typical flow rate range at 0.5 mm channel diameter, see also Fig. 4. In the second case we discuss the critical heat transfer to avoid reaction runaway. Starting with the thermal energy balance for an exothermic reacting flow in a channel, we have two issues to consider, the reaction heat  E  A q̇ = (−Hr )CA k0 exp − RT (17) and the cooling heat transfer by conduction within the fluid to the channel wall during stopped flow with fluid thermal conductivity  and temperature difference T between fluid and wall temperature. q̇ = −T (18) The wall often consists of metal; hence the main thermal resistance presents the fluid itself. It is assumed, that half of the hydraulic diameter is necessary to transfer the heat to the wall. Both equations above can be graphically analyzed in the Semenov curve, because an analytical solution is hard to find due to the non-linear behavior [17] of the Arrhenius term in Eq. (16). Here, we follow the approach of Frank-Kamenetzkii [27], who derived the heat balance and introduced dimensionless parameters for temperature and the spatial coordinate. The solution for the dimensionless differential equation in steady state can be given from the eigenvalues of the following characteristic parameter, also called Frank-Kamenetzkii 724 N. Kockmann et al. / Chemical Engineering Journal 167 (2011) 718–726 parameter. ı=  2 (−Hr ) EA dh  RT02 2  E  A k0 exp − (19) RT0 The critical value of ıcrit is reached, when the vessel temperature holds a steady value and no reaction run away occurs. These values are 2.00 and 3.32 for an infinite cylinder and spherical vessel, respectively. Now we can define the critical diameter of a cylinder, where internal heat conduction is not sufficient to remove heat release from reaction. dh =  2 8af RTW EA Tad tR 1/2 (20) The characteristic reaction time tr is determined from the reaction rate r, where the adiabatic temperature rise is determined by the reaction enthalpy and initial concentration CA,0 . tR = CA,0 1 ; = m−1 r(TW ) k0 CA,0 Tad = CA,0 (−Hr ) ; cp af =  cp (21) The last parameter af is the temperature conductivity of the reacting fluid. In Eq. (20) we see no influence of the flow rate anymore, because the heat transfer is only by thermal conduction. This case occurs for example for pump failure, when the reacting fluid is not moving anymore. No reaction runaway occurs in this case. For a streaming reacting fluid, the internal heat transfer is much higher than pure conduction and also no reaction runaway will occur. In Fig. 6, the internal heat transfer coefficient is between 10 and 50 kW/m2 K in the displayed diameter and related range of flow rate. For reactions with an adiabatic temperature rise below 150 K, a reaction run away will not occur, if they are slower than 10 s. For faster reactions, a more detailed analysis is necessary to guaranty safe operation. For example, one reagent can be split up in several feed streams to individual injection points, forming the socalled cross flow reactor [28]. Involving a second phase as well as adjustment of concentration or component reactivity can trim the transport characteristics and kinetics to coherent values. 8. Plate and reactor design The above derived correlations in Eqs. (4), (6), (10), and (13) guide the proper and consistent design of microchannel reactors for flow velocity, Re number, pressure drop, and mixing time. The internal channel size and channel cross-section area are also growing with plate size allowing higher flow rates through the reactor, see Fig. 7. The typical hydraulic diameter of the mixing channel is qualitatively displayed over the flow rate in double-logarithmic scale Fig. 7. Plate size and channel diameter for different flow rates with comparable mixing time. for constant flow velocity w, constant Reynolds numbers, constant pressure loss p over the mixing channel, and typical mixing time tm . The horizontal bars indicate the typical flow rates for a single channel diameter and size for necessary mixing time and allowable pressure loss. The reactor stacks were already displayed in Fig. 3. The reactor size in Fig. 7 is named according to the outer dimensions of the plates. For example, the A6 Lonza reactor corresponds to DIN A6 paper sheet format of 148 mm × 105 mm. The A5 reactor has double the area of A6 reactor and also roughly twice the heat transfer capacity. The stepwise scale-up of the plates is accompanied by stepwise scale-up of the channel cross-section, leading to a scaling factor of approx. 1.4 between each step. For example, the typical hydraulic diameters of the mixing channel in the A6 reactor plate are 0.35 and 0.5 mm, hence, the A5 reactor plate contains mixing channel structures with typical diameters of 0.7 and 1.0 mm. Larger plates are accordingly scaled up leading to comparable mixing conditions. The flow velocity reaches several m/s and the Re number indicates turbulent flow regime in the mixing channel. Heat transfer area and internal volume to gain residence time are increased in corresponding steps of connected plates in series. These measures result in a consistent, versatile, and flexible scale-up strategy completely avoiding parallelization over a wide range of flow rates [7,13]. Beside the reactors size, the operational conditions are changing accordingly. Small reactors size of Lab-Plate and A6 Lonza reactor are well suited for projects in process development. They fulfill the needs of early phase studies in a highly consistent and straightforward manner. Once a project has survived and progressed to commercial manufacture, the aim changes to detail engineering and piloting of the process. More resources can be invested in the project. Chemical engineering plays an important role and the scale-up is performed using either conventional static mixers with mini-heat exchangers, higher cross-sectional reactors (as small as needed), or as a last measure numbering-up and equipment parallelization. 9. Scale-up example All reactor types displayed in Figs. 2 and 3 are frequently used in Lonza’s laboratories and production environment. The Lab-Plate and A6 reactor are suitable for early feasibility studies for given chemical routes to check main parameters and precipitation potential. Process development is performed in A6 and A5 reactors, where main influence parameters such as temperature, stoichiometry, solvents, and reagents are investigated and optimized. Sample production in the laboratory and large scale production are performed in the A5 reactor. However, the A5 reactor can be operated for short time in laboratory under simulated process conditions similar to later production conditions. As conclusion, we will demonstrate the scale up of microreactors with a real-case example consisting of a two-steps organo-metallic reaction (Li–H exchange and coupling) with three feeds having following composition: Feed-1 with substrate at 15 wt.%, Feed-2 with first reagent at 30 wt.%, and Feed-3 with second reagent at 16 wt.%. The reaction is stoichiometric and operated at two temperature levels. Feed-1 and Feed-2 were precooled to the cryogenic reactor temperature, while the second reaction could be performed without cooling at room temperature. The flow diagram and reaction scheme are depicted in Fig. 8. The first reaction is of Type A with an adiabatic temperature rise of more than 75 ◦ C. Three reactors were tested: a static mixer, a glass microreactor, and the Lonza plate microreactor, see Table 2. The second reaction is of Type B and less demanding in terms of heat exchange (Tad < 25 ◦ C) and mixing. A microreactor and static mixer were tested with identical performances. Mixing and heat N. Kockmann et al. / Chemical Engineering Journal 167 (2011) 718–726 725 Fig. 8. Reaction scheme and process scheme with image of pilot-scale production setup. Table 2 Comparison of different reactors and flow rates for scale-up of the lithiation reaction, see Fig. 8. Reactor Static mixer 3/8 Static mixer 3/8′ Glass MR 0.5 mm Glass MR 0.5 mm Lonza MR-A6 0.5 mm Lonza MR-A6 0.5 mm Lonza MR-A5 0.7 mm Lonza MR-A5 1.0 mm Lonza MR-A5 1.0 mm ′′ ṁMR [g/min] Tout [◦ C] p [bar] Isolated yield [%] 33 148 33 148 33 140 150 150 237 9 41 −14 15 −22 −16 −20 −21 −16 0.3 1.6 0.4 3.2 0.9 8.8 3.4 2.0 4.5 88 84 86 88 89 90 90 88 88 transfer are not critical issues, only sufficient residence time has to be provided for the reagents. Investigating the lithiation reaction (Table 2), no adequate temperature control was observed within the static mixer; more or less the complete adiabatic temperature rise was detected at medium flow rate (148 g/min). Un-wanted side products were formed to large extend, visible also in a drop of yield (84% versus 89% on average). Certain concentration of the critical byproduct leads to problems during workup. The results with the glass microreactor show a loss of temperature control at higher flow rates, but not yet reflected in the product yield. For lower flow rates, the mixing was not fast enough and led to higher byproduct formation. Besides rapid mixing, a short residence time is also very important. Nevertheless, the operational robustness of the glass reactor was not adequate for this particular reaction. The results with Lonza’s reactor technology show a good equivalence of performance for the smaller (A6) and larger (A5) reactors. It was possible to operate the A5 reactor with an internal mixing channel of 1 mm diameter up to 237 g/min with a total flow rate over 700 g/min for the second reaction by keeping the pressure drop well under control. 700 kg of isolated material were produced in a pilot campaign leading to more than 10 m3 of processed solution through the reactor setup. A second campaign was performed to produce more than 2 tons of isolated material in 2009. Both campaigns showed the long-term robustness of the process and reliability of the installed reactor equipment. 10. Conclusion Microreactors have found their place in research laboratories in academia and industry for development and small scale production. Chemical and pharmaceutical companies are now starting to invest more and more effort in setting up lab devices and get- ting first chemical results. The question, how to transfer the lab results to pilot plant and large commercial production is still a challenging task. This contribution shows a consistent pathway from lab development to pilot plant production of continuous-flow processes with microstructured devices. The key factor of the reactor is that it has only one single channel, which characteristic dimension, the hydraulic diameter, is scaled up in correlation with mixing time and heat transfer characteristics. Pressure drop in the mixing channel plays the key role to determine mixing time and heat transfer rates. The reaction kinetics and enthalpy are correlated with the above results and indicate the operability of the reactor. A modular microreactor system developed at Lonza is presented, which can be used for process research in the lab as well as for production campaigns on pilot scale. The scale-up is shown for an industrial example. With this approach, valuable laboratory experience can be easily transferred with success into production environment without the loss in productivity. Simple and robust scale up with modularity for multipurpose applications will give new drive for microreactor development and applications. References [1] G.A. Truskey, F. Yuan, D.F. Katz, Transport Phenomena in Biological Systems, 2nd ed., Pearson Prentice Hall, Boston, 2010. [2] N. Kockmann (Ed.), Micro Process Engineering, Wiley-VCH, Weinheim, 2006. [3] S.J. Haswell, P. Watts, Green chemistry: synthesis in micro reactors, Green Chem. 5 (2003) 240–249. [4] J.-I. Yoshida, Microreactor synthesis in green chemistry, Chem. Ind. 57 (2006) 461–464. [5] V. Hessel, Novel process windows—gate to maximizing process intensification via flow chemistry, Chem. Eng. Technol. 32 (2009) 1655–1681. [6] J.-I. Yoshida, Flash Chemistry: Fast Organic Synthesis in Microsystems, Wiley, New York, 2008. [7] D.M. Roberge, M. Gottsponer, M. Eyholzer, N. Kockmann, Scale-up approach and production of pharmaceuticals in microreactors, Chem. Today 27 (July–August) (2009) 8–11. [8] N. Kockmann, Transport Phenomena in Micro Process Engineering, Springer, Berlin, 2008. [9] D.M. Roberge, L. Ducry, N. Bieler, P. Cretton, B. Zimmermann, Microreactor technology: a revolution for the fine chemical and pharmaceutical industries, Chem. Eng. Technol. 28 (2005) 318–323. [10] A. Nagaki, H. Kim, J.-I. Yoshida, Nitro-substituted aryl lithium compounds in microreactor synthesis: switch between kinetic and thermodynamic control, Angew. Chem. IE 48 (2009) 8063–8065. [11] C. Wiles, P. Watts, Continuous flow reactors, a tool for the modern synthetic chemist, Eur. J. Org. Chem. 10 (2008) 1655–1671. [12] K. Geyer, J.D.C. Codee, P.H. Seeberger, Microreactors as tools for synthetic chemists—the chemists’ round-bottom flask of the 21st century? Chem. Eur. J. 12 (2006) 8434–8442. [13] N. Kockmann, M. Gottsponer, B. Zimmermann, D.M. Roberge, Enabling continuous-flow chemistry in microstructured devices for pharmaceutical and fine-chemical production, Chem. Eur. J. 14 (2008) 7470–7477. [14] L. Ducry, D.M. Roberge, H. Dibal, Reduction of methyl butyrate into butyraldehyde using microreactors, Org. Process Res. Dev. 12 (2008) 163–167. 726 N. Kockmann et al. / Chemical Engineering Journal 167 (2011) 718–726 [15] F. Rainone, N. Kockmann, B. Zimmermann, Industrial fine-chemical and pharmaceutical production with microreactor technology (MRT), CPAC Satellite Workshop, Rome, 2008, March 26–28. [16] L. Ducry, D.M. Roberge, Controlled autocatalytic nitration of phenol in a microreactor, Angew. Chem. IE 44 (2009) 7972–7975. [17] N. Kockmann, D.M. Roberge, Harsh reaction conditions in continuous-flow microreactors for pharmaceutical production, Chem. Eng. Technol. 32 (2009) 1682–1694. [18] D.M. Roberge, N. Bieler, B. Zimmermann, R. Forbert, Micro-Reactor System WO2007112945, 2007. [19] H.G. Hirschberg, Handbuch Verfahrenstechnik und Anlagenbau. Chemie, Technik und Betriebswirtschaft, Springer, Berlin, 1999. [20] N. Kockmann, D.M. Roberge, Transitional flow and related transport phenomena in curved microchannels, Heat Transfer Eng (2011). [21] N. Kockmann, Pressure loss and transfer rates in microstructured devices with chemical reactions, Chem. Eng. Technol. 31 (2008) 1188–1195. [22] Z.K. Majumdar, J.D. Sutin, R.M. Clegg, Microfabricated continuous-flow, turbulent, microsecond mixer, Rev. Sci. Instrum. 76 (2005), 125103. [23] J.R. Bourne, Mixing and selectivity of chemical reactions, Org. Process Res. Dev. 7 (2003) 471–508. [24] L. Falk, J.-M. Commenge, Characterization of mixing and segregation in homogenous flow systems, vol. 1, in: V. Hessel (Ed.), Micro Process Engineering, Wiley-VCH, Weinheim, 2009, Chapter 6. [25] M. Gödde, C. Liebner, H. Hieronymus, Sicherheit in der Mikroreaktionstechnik, Chemie-Ing. Technik 81 (2009) 73–78. [26] H. Martin, The generalized Lévêque equation and its practical use for the prediction of heat and mass transfer rates from pressure drop, Chem. Eng. Sci. 57 (2002) 3217–3223. [27] D.A. Frank-Kamenetskii, Diffusion and Heat Transfer in Chemical Kinetics, Russian eds. 1947, 1967, 1987, German ed. 1959, English eds. 1969. [28] D.M. Roberge, N. Bieler, M. Mathier, M. Eyholzer, B. Zimmermann, P. Barthe, C. Guermeur, O. Lobet, M. Moreno, P. Woehl, Development of an industrial multiinjection microreactor for fast and exothermic reactions—part II, Chem. Eng. Technol. 31 (2008) 1155–1161.