Microfabricated Multiphase Packed-Bed Reactors: Characterization of

May 11, 2001 - Development of Pilot-Scale Continuous Production of an LY2886721 Starting Material by Packed-Bed Hydrogenolysis .... Using a Multijunct...
0 downloads 9 Views 414KB Size
Ind. Eng. Chem. Res. 2001, 40, 2555-2562

2555

Microfabricated Multiphase Packed-Bed Reactors: Characterization of Mass Transfer and Reactions Matthew W. Losey,† Martin A. Schmidt,‡ and Klavs F. Jensen*,† Department of Chemical Engineering and Microsystems Technology Laboratories, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139

A microchemical device has been built in silicon and glass by using microfabrication methods including deep-reactive-ion etch technology, photolithography, and multiple wafer bonding. The microchemical system consists of a microfluidic distribution manifold, a microchannel array, and a 25-µm microfilter for immobilizing solid particulate material within the reactor chip and carrying out different heterogeneous chemistries. Multiple reagent streams (specifically, gas and liquid streams) are mixed on-chip, and the fluid streams are brought into contact by a series of interleaved, high-aspect-ratio inlet channels. These inlet channels deliver the reactants continuously and cocurrently to 10 reactor chambers containing standard catalytic particles. The performance of the microfabricated “packed-bed” reactor is compared to that of traditional multiphase packed-bed reactors in terms of fluid flow regimes, pressure drop, and mass transfer. The hydrogenation of cyclohexene is used as a model reaction to measure the mass transfer resistances. Overall mass transfer coefficients (KLa) are measured to range from 5 to 15 s-1snearly 2 orders of magnitude larger than values reported in the literature for standard laboratory-scale reactors. Introduction Multiphase reactions present a unique opportunity for microfabricated reactors.1 In addition to efficient thermal control, these systems have the added complexity of forcing a reactant of one phase to mix, diffuse, and react with that of another. For fast gas-liquid-solid reactions, the chemical kinetics are often limited by the mass transfer rate of the gaseous species through the liquid to the surface of the catalyst. As a class of reactions, hydrogenations represent a typical and ubiquitous gas-liquid-solid reaction. Hydrogen is reacted with an organic substrate over a supported noble metal catalyst in either a slurry reactor or a packed-bed arrangement. The limited solubility of hydrogen in organic substrates and solvents makes mass transfer a primary concern.2 Consequently, reactions are often operated at high pressures, sometimes as high as 100 atm3, to increase the reaction rate and offset the low hydrogen solubility. However, in that case, the explosion potential requires extreme safety measures. Continuous industrial multiphase reactors are usually classified according to their flow dynamics.2 In tricklebed reactors, for example, the volume of the reactor is predominantly the gas phase, and the liquid forms a thin film around the catalyst pellet. Reaction engineers are concerned with the uniform distribution of the fluids within the reactor. Uneven flows can lead to incomplete utilization or local zones of varying reaction rate and heat transfer. Poor distribution of the fluids can thus lead to local “hot spots”, which can decrease selectivity, reduce catalyst life, or lead to side reactions that can cause reactor runaway.4,5 Poor thermal management can also decrease the life of a catalyst and further reduce * Corresponding author. Fax: (617) 258-8224. E-mail: [email protected]. † Department of Chemical Engineering. ‡ Microsystems Technology Laboratories.

the efficiency of a reactor. Hydrogenation is exothermic, and consequently thermal uniformity and adequate temperature control is of primary concern. Efforts in process miniaturization and microreaction technology have emerged as a means for improving process capability and control in chemical synthesis and as a means for conducting safer and more efficient chemical kinetics investigations.6-9 Miniaturized chemical systems provide the opportunity for distributed, ondemand manufacturing that could eliminate the hazards of transportation and storage of toxic or hazardous chemical intermediates. As a testing platform, microchemical systems offer inherently safer conditions because of their reduced active volume. Microfabrication technology (involving photolithography, deep-reactive-ion etching, thin-film growth and deposition, and multiple wafer bonding) offers the opportunity to manufacture novel chemical reactor designs. Moreover, integration of sensing elements could allow for more efficient control and rapid probing of chemical kinetics. Reactor dimensions can be scaled with high-aspect-ratio features smaller than 100 µm of any given planar complexity. The resulting high surfaceto-volume ratios offer improvements in thermal management and mass transfer but also provide challenges resulting from increases in the pressure drop across channels having dimensions on the order of tens of microns. The ability to incorporate particles or spherical beads into chip devices for analytical applications has been demonstrated;10 here, catalyst particles are immobilized within the silicon microchannels with the aid of a 25-µm filter. Several liquid-phase micromixer designs have emerged in recent years, most of which rely on lamination and rapid diffusion at short length scales.11,12 By increasing the interfacial contact area between two mixing fluids and by reducing the characteristic length scale, the rate of transfer can be improved. For rapid multiphase reactions that are limited

10.1021/ie000523f CCC: $20.00 © 2001 American Chemical Society Published on Web 05/11/2001

2556

Ind. Eng. Chem. Res., Vol. 40, No. 12, 2001

by mass transfer effects, a reduction in scale commensurate with increased surface area should improve process efficiency. Particularly for multiphase reactions, the opportunity to manipulate, at the micron scale, the means and conditions by which two or more phases contact, mix, react, and separate affords the reaction engineer a high level of control over the chemical reaction. Batch fabrication allows for designs to be manufactured efficiently with a high level of dimensional tolerance. Operating several microreactor units in parallel could achieve desired production capacities and offer distinct advantages over traditional modes of processing. In this work, we describe the design and characterization of a microfabricated reactor for heterogeneous catalytic processes. Two designs of the reactor have been constructed: a single-channel reactor and a multiplechannel reactor that has 10 reaction channels connected in parallel through a microfluidic distribution network. Both reactors are constructed using silicon micromachining methods and have features as small as 25 µm. We characterized the mass transfer rates in these reactors using the classical heterogeneous hydrogenation of cyclohexene to cyclohexane over Pt/Al2O3 catalysts. The measured reaction rates in the microreactor compare favorably with the reported results of conventional trickle-bed reactors. In fact, the overall mass transfer coefficient in the microreactor exceeds typical values for conventional reactors by 2 orders of magnitude. Device Design and Fabrication Design Motivation. The design of the microreactor device was motivated by the method for incorporating the catalytic solid phase, the means for ensuring dispersion of the gas and liquid phases, and the minimization of the pressure. From a fabrication standpoint, integrating the catalyst could be accomplished easily through thin-film deposition methods, but planar metal films have relatively low surface areas compared to catalysts on porous supports. Standard catalysts, such as noble metals supported on inert porous materials, are readily available, and there is an established knowledge base related to chemical kinetics. Thus, with the use of standard porous catalysts, catalyst preparation becomes a matter of separating discrete particle size ranges and loading the reaction channels. In addition, the packedbed approach ensures better mixing of the gas and liquid phases, as opposed to multichannel planar surface architectures. However, with the packed-bed approach, the reactor and catalyst dimensions must be scaled to maintain an acceptable pressure drop. For microreaction systems in general, the engineering design challenge is to balance the gains made in heat and mass transfer against the losses in pressure drop. The pressure drop for a packed bed of spherical particles can be estimated from the Ergun equation

[

]

∆P G (1 - ) 150(1 - )µ + 1.75G ) L FDp 3 Dp

(1)

where G is the superficial mass velocity and  is the void fraction.13 For the relatively small velocities in this work, the second term is negligible. Substituting G with the volumetric flow rate Q, the cross-sectional area of the reactor As, and the density gives the following

functional dependence: 2 ∆P µQ (1 - ) ∝ 2 L dp As 3

(2)

As an example, for a void fraction of 0.4, a single-phase flow rate of 1.0 mL/min of ethanol, a reactor of dimensions 625 µm × 300 µm × 2 cm, and particles 64 µm in diameter, the calculated pressure drop is 0.48 MPa. The experimentally measured pressure drop for these conditions using glass spheres in with diameters the range of 53-74 µm was 0.56 MPa. Judging from eq 2, for a given volume of packing with constant residence time, a short, large-diameter reaction channel would minimize the pressure drop. If reactions are operated such that there are no mass transfer limitations, then the capacity for conversion depends on the amount of catalyst and not on the specific geometry. The problems with operating such a design are obtaining an even distribution of reactants over the inlet diameter and controlling the temperature profile. Because the temperature of most tubular reactors is controlled through the exterior radial walls, a largediameter reactor will become increasingly hindered by radial gradients in temperature. Therefore, splitting the flow into multiple channels so that the effective crosssectional area is large will ultimately reduce the pressure drop while maintaining the same reactor throughput and high surface-to-volume ratio. Microfabrication methods have the potential for efficiently realizing such a reactor design: one that reduces mass transfer limitations, ensures thermal uniformity, and employs a reactor geometry that reduces the pressure drop. Microfluidic Design. The microfluidic design contains two main parts: the inlet manifold for each reaction channel and the filter unit at the exit of the device. Figure 1 shows a photomicrograph of the singlechannel reactor chip filled with active carbon catalyst. Figure 2 illustrates the inlet portion for each reaction chamber. Gas and liquid inlet flows are split among several channels (25 µm wide) that meet at the main reaction chamber (625 µm wide). Perpendicular to these inlet channels is a 400 µm wide channel that is used to deliver the catalyst slurry to the reaction chamber. Figure 2b is a scanning electron micrograph of a cross section of the inlet channels where they meet at the main reaction chamber. At the outlet of the 20-mm-long reaction chamber, a series of posts etched in the silicon retain the packing material (see Figure 3). Figure 4 shows the multichannel device in which 10 reaction chambers and gas-liquid inlets have been connected in parallel. The fluid manifolds are constructed on-chip: the gas is distributed on the same silicon layer as the reaction chambers, and the liquid is distributed in a second silicon layer beneath the reactor layer. Fabrication Process. The fabrication processes involved multiple photolithographic and etch steps, a silicon fusion bond, and an anodic bond. The fluid channels were formed in the silicon substrate (100 mm in diameter, 500 µm thick) by using a time-multiplexed inductively coupled plasma etch process.14 The depth was controlled to approximately 300 µm by timing the etch. As is apparent in Figure 2, a variation in the etch rate from small to large features caused the 25-µm-wide channels to be approximately 20% less deep than the main reaction channel. The wafer was then patterned from the backside to create ports for access to the

Ind. Eng. Chem. Res., Vol. 40, No. 12, 2001 2557

Figure 1. Photomicrograph of the single-channel reactor loaded with activated carbon. The hash-marks indicate where two images have been spliced to view the entire device. (Photographs by Felice Frankel.)

Figure 3. (A) Reaction channel outlet design. A grating of posts 40 µm wide separated by 25 µm forms the filter. (B) SEM of filter unit. Figure 2. Fixed-bed microreactor design. (A) Nine separate 25µm inlets distribute the gas and liquid phases to the reaction channel. (B) Scanning electron micrograph of inlet channels forming the gas-liquid distributor.

various channels. A second silicon wafer containing the fluidic manifolds was processed in a similar manner.

Optionally, protective films (silicon dioxide, silicon nitride) could be grown on the silicon wafers as protection against aggressive reagents that would react with silicon. The two silicon wafers were aligned and fusion bonded.15 The final step in the process was to cap the channels on the first layer with a Pyrex 7740 glass wafer

2558

Ind. Eng. Chem. Res., Vol. 40, No. 12, 2001

Figure 4. Multichannel reactor design. (A) Top silicon layer illustrating fluid manifold to 10 parallel fixed-bed reaction channels (B) Photomicrograph of multichanel reactor chip.

Figure 5. System configuration and reactor packaging.

using an anodic bond.15 From a 100-mm silicon wafer, 8 multichannel reactors (15 mm × 40 mm × 0.5 mm) or 12 single-channel reactors (10 mm × 40 mm × 0.5 mm) could be obtained. Experimental Section Setup and Packaging. Figure 5 shows the experimental setup and packaging for the packed-bed microreactor. The silicon-glass reactor is compressed against a thin elastomeric sheet (0.8-mm-thick Viton) to form a fluidic connection to the stainless steel base, which was machined for standard high-pressure fittings. The cover plate, which compresses the chip, can be either aluminum with cartridge heaters (3.2 mm, 25 W, Omega,

Stamford, CT) for temperature control or plexiglass for visualization. Thermocouples (0.25 mm, type K, Omega) are inserted into 400-µm channels of the microreactor (see Figure 5). In-line pressure transducers (Omegadyne, Stamford, CT) connected with a low-volume stainless steel “T” type connector (Upchurch Scientific) measure the pressure. A mass flow controller (Unit Systems) delivers the hydrogen while a syringe pump (Harvard Apparatus PHD2000) delivers the liquid reactant. Fractions of the effluent are collected and analyzed off-line using a Hewlett-Packard 6890 GC with a mass-selective detector. Flow visualization is obtained with the use of a Leica MZ12 stereomicroscope connected to a Cohu CCD camera positioned directly above the plexiglass compression plate. We loaded different materials into the microreactor, including glass microspheres (MO-Science Inc., MN), polystyrene beads, and catalyst powders. All chemicals and catalysts, unless otherwise noted, were used as obtained from Sigma-Aldrich (Milwaukee, WI). Using standard sieving equipment, a 50-75-µm fraction was filtered from the catalyst powder, and a slurry was formed using ethanol. The slurry was delivered to the microreactor using the 400-µm-wide side channels (Figure 1). Once the reactor was loaded with material, the catalyst inlets were closed by either substituting a different gasket or capping the external fittings. Cyclohexene Hydrogenation. The hydrogenation of cyclohexene was used as a model reaction to measure the mass transfer rate. Cyclohexene was purified according to procedures reported elsewhere to prevent deactivation of the catalyst.4 The catalyst employed was standard platinum supported on alumina powder, where the platinum content was either 1 or 5 wt %. The metal surface area for the 1 wt % Pt/Al2O3 catalyst was measured as 0.57 m2/g using CO chemisorption in a Micromeritics ASAP 2010 instrument. The catalyst powder was sieved, and fractions of sizes 53-75 µm, 36-38 µm, or