Microdevices

2.4.1 H eat Transfer Devices

Microengineered reactors offer significant opportunities for high heat transfer rates when coupled with microengineered heat exchangers or heating elements. One o f the first realisations o f an enhanced microscale heat transfer system was an integrated microchannel combustor/evaporator which could be used as a portable or residential heating device as well as to provide heat for endothermie reactions (Drost, et al, 1997). It produced 30 W o f thermal energy per square centimetre o f external combustor area, with more than 70 % combustion efficiency operating on methane. The microchannels for the water and the combustion gases were 300 pm wide and 500 pm deep and resulted in heat transfer coefficients an order o f magnitude larger for single phase flow and two orders o f magnitude for two-phase flow than predicted for macroscale conditions. Using smaller microchannels (hydraulic diameters 70-160 pm) in a cross-flow arrangement, Schubert et al (2001) measured heat transfer coefficients for heat exchange between cold and hot water o f up to 25 kW/m^K. Heat transfer coefficients were doubled by continuously splitting and recombining the fluid flow, through channel segmentation. The only drawback was the higher pressure drop. Similarly Hardt et al, (2000) using a checkerboard array o f 100x100 pm micro fins in a

500 jim deep channel, observed a theoretical overall heat transfer enhancement o f more than an order o f magnitude as compared to unstructured channels.

Another approach to provide heat in microengineered reactors is via electrical heating elements. Through a combination o f microchannel structures and cartridge resistor heaters, Brandner, et al (1999) achieved heat transfer coefficients o f 20 kW/m^K, while avoiding overheating. These devices were used to control processing temperatures up to several hundred degrees Kelvin. Srinivasan, et al (1997) deposited thin film Pt serpentine heaters on the external wall o f microreactors.

Heat transfer efficiency in microchannels can be affected by the wall conductivity, due to the presence o f axial conduction. Low wall heat conductivity can result in better efficiency as axial heat conduction along the walls o f the microchannels decreases (Bier et al, 1993). However, as the heat conductivity o f the wall material reduces, radial heat transfer is also hindered. As a result, an optimal value exists for maximum heat exchanger efficiency based on the heat conductivity o f the wall material (Stief, et al, 2000).

2.4.2 Mixing Devices

Before reaction between two components can take place, the corresponding molecules must be brought into intimate contact by mixing. Because laminar flow is usually encountered within microdevices, mixing between fluids is accomplished mainly through diffusion and not through the much faster convective processes present in turbulent flow. The diffusive mixing efficiency is usually described by the Fourier number, Fo, defined as follows:

Fo = ^

where D = diffusion coefficient

t = contact time

1 = characteristic length over which diffusion takes place

Good mixing occurs when Fo is between 0.1 and 1. From the above equation it is obvious that mixing time increases with the characteristic dimension. Fast mixing times can therefore be achieved by minimising the diffusion distance. Lowe, et al (2000) presented a comprehensive review on the different types o f micromixers and identified various mixing approaches, which will be discussed as follows:

a) contacting o f two substreams

Here, two streams merge and sufficient length is allowed afterwards for the mixing to be completed. Gobby, et al (2001) studied both T- and Y- geometries using CFD simulations and found that for gas-phase mixing, different Y-junction angles had negligible effect o f on the length required for complete mixing (mixing length) and the pressure drop. Mixing lengths less than 3 mm were obtained for fluid velocities 0.3 m/s in 500 pm wide channels.

b) decrease o f the diffusion path perpendicular to flow direction

After two fluids are brought together, decreasing the thickness o f the resulting stream decreases mixing time according to the Fourier number definition mentioned above. Veenstra, et al (1999) achieved this by combining two fluids in a 300 pm wide channel which was then reduced to 100 pm. Complete mixing was achieved in the outlet at total flowrate 10 pL/min while for flowrates above 50 pL/min there was no mixing. A similar idea was also investigated by Gobby, et al (2001) where a “throttle” mixer was found to reduce the length required for

complete mixing compared to a 45° Y-mixer, at the expense o f higher pressure drops. Knight, et al (1998) hydrodynamically ‘pinched’ the main liquid stream using two side liquid streams.

c) injection o f small substreams o f one component in a main stream o f another component

In this approach, a liquid stream is guided through the mixing chamber while a second liquid stream is split into many small jets and injected perpendicularly through small holes in the mixing chamber. Koch, et al (1998) achieved complete mixing within 1 s for flowrates o f about 1 pL/min using 42 100x100 pm holes, while Elwenspoek, et al (1994) split the flow into a larger number o f smaller holes (400 15x15 pm holes) to achieve similar mixing times for larger flowrates (240 pL/min).

d) injection o f many sub streams o f the two components

This approach has widely been used in microsystems. In these mixers, the two streams are first split into sub streams with reduced width and then brought into contact. In the mixer used by Bessoth, et al (1999) the combined sub streams were brought sequentially into contact (from 16 to 8 to 4 to 2, and finally to 1, see Figure 2.2), until all partial flows were united in one broad outlet channel. The presence o f walls and bends affected the velocity profile in each substream and thus the mixing time. In other multilamination-based micromixers all the combined sub streams simply join in a large channel as shown in Figure 2.3 (Ehrfeld, et al, 1999, Zech, et al., 2000, Schubert, et al., 2001). Depending on

m ixer configuration, flow rates and fluids m ixed, m ixing tim es and lengths o f 30 f.is and 150 pm , respectively, have been reported.

Figure 2.2 E xam ple o f a m icrom ixer utilising the m ultilam ination concept (B essoth , et al, 1999) Tiütpur

1 H o u sin g

H o u sin g

--- 1000 unn

Figure 2.3: E xam ple o f a interdigital m icrom ixer utilising the m ultilam ination con cep t (IM M , 1999)

e) m anifold splitting and recom bination o f a stream consisting o f tw o fluid lam ellae By careful splitting and recom bining stream s using a 3-dim ensional flow pattern, the surface area available for m ass transfer can be increased, w hile diffusion distances decrease. Ehrfeld, et al (1999b) developed a “caterpillar” m ix er w ith a ram p-like channel architecture in w hich the fluid lam ellae w idth is reduced to 2 pm after 6-8 passages. O ther ways to accom plish splittting and recom bination

include the use o f fork-like elements (Schwesinger, et al, 1996) or separation plates (Branebjerg, et al, 1996).

In the above mixers, the flowrates o f the phases to be combined affect mixing efficiency. In addition, producing thin fluid layers usually results in high pressure drops. Comparing the different types o f mixers is not easy since different fluids, channel geometries and flowrates have been used. A number o f techniques have been implemented for characterising mixing efficiency in micromixers such as colour changing reactions (Koch, et al, 1998, 1999), fluorescence quenching (Bessoth, et al 1999) and a system o f fast and slow competing reactions (Erhfeld, et al 1999).

It is worth noting that it is possible to enhance diffusive mixing by various methods. Bokenkamp, et al (1998) obtained submillisecond mixing times in T-mixers by increasing flowrates so that turbulent flow was achieved. Woias, et al (1999) proposed an “active” micromixer, where the pulsation o f a membrane enhances mass transfer. Compared to the static mixers, which use fluid multilamination, mixing time would not depend on the flowrates.

The multilamination approach has been used for mixing o f two different phases (gas- liquid or liquid-liquid). Hessel, et al (1998) studied the bubble sizes formed in micromixers during gas-aqueous solution mixing (bubble coalescence was prevented by using a surfactant). The mixer produced mean bubble sizes in the range 120-800 pm. Size was affected by liquid flowrate, viscosity and mixing channel width. The same mixer was also used for the mixing o f two immiscible liquids (Schiewe, et al, 2000). Compared to a standard laboratory shaker, the micromixer gave a smaller

average drop size ( 1 pm for an oil-in-water dispersion) and narrower size distribution. It was also found that such a mixer required less energy compared to a stirred tank to produce the same average drop size and distribution (Bayer, et al, 2000).

2.4.3 Chemical Analysis Devices

Analysis o f product streams from microengineered reactors can be carried out by conventional analytical equipment. However, integrating analysis with reaction can provide advantages such as portability, continuous on-line measurement and faster process control. It is worth noting that the first reported attempt at creating a micro fluidic system was a micro-GC (Terry et al, 1979). Recently, interest in microreactors has re-kindled the desire to produce micro-GCs. Besides having a shorter column length, the micro-GC developed at the Institut fur Mikrotechnik Mainz (EMM) features improved heating and cooling systems over conventional systems. A planar module measuring 80 x 80 x 75 mm^ comprising a capillary column support sandwiched between 4 resistive heaters (50 W each) and an axial-flow cooling fan was constructed. The heating rate was 2 °C/s, while the cooling rate was 0.5 °C/s. The total analysis time for ethylene oxidation products was 8 min, as compared to 25- 35 min in conventional analysis (Schiewe, et al, 1999). To increase cooling rate, a cylindrical system was designed where the capillary column was coiled helically around axial cooling microchannels and a heating cartridge. A heating rate o f 4.7 °C/s and a cooling rate o f 5 °C/s (using air flowrate o f 5.8 1/s) were obtained (Richter, et al, 2000). Another improvement to micro-GC technology was reported by Lehman, et al (2 0 0 0) who developed a new method for coating the analysis column with the stationary phase. The column was etched into a silicon wafer (2 m length on a 20 x 25 mm wafer) and covered with a 500 nm thick silicon organic stationary phase

deposited using PECVD. The corresponding areas o f a glass lid were also covered in this way before being bonded by anodic bonding. An injector and a miniaturised TCD detector were also incorporated. Heating was supplied by a Peltier element. Excellent separation o f methane and ethane was achieved.

In-situ analysis can be often performed using ER-spectroscopy. Guber, et al (1999) cut a reaction channel straight through a metal substrate and covered it with IR- transparent discs (AgCl). The assemblage was then placed in the object carrier o f an IR-microscope and coupled to an FTIR spectrometer. Micrometer-precise movement o f the object carrier allowed nearly any number o f measuring points. It is also possible to do away with transparent discs as silicon is ER transparent. Microreactors made entirely from different types o f silicon were tested for IR-transmitance. Undoped silicon was found to be the best (48.4 %), followed by boron-doped wafers (41.7 %). Phosphorous doped samples had almost no transmittance. Use o f anti- reflective coating was found to increase transmittance by 30 % (Floyd, et al, 2000).

An interesting development was the application o f photoacoustic detection to a microreactor. W hen a gas is illuminated, if the optical wavelength couples to an energy transition in the gas, the light will be absorbed and the gas will exhibit periodic expansion. Firebaugh, et al (2000) placed an optical fibre beneath the reaction channel etched through silicon and capped by thin silicon nitride layers on the top and bottom. A displacement sensor was placed on top o f the reactor to monitor movement o f the membrane due to gas expansion. This microsensor produced a larger signal than a larger analogue.