Design and Control Techniques for the Numbering-up of Capillary

Jul 1, 2010 - The slug flow capillary reactor can thus be used to overcome the mass transport ..... To achieve uniform flow rates for the two-phase sy...
0 downloads 0 Views 2MB Size
Download PDF
10908

Ind. Eng. Chem. Res. 2010, 49, 10908–10916

Design and Control Techniques for the Numbering-up of Capillary Microreactors with Uniform Multiphase Flow Distribution Matthias Mendorf,* Henrik Nachtrodt, Axel Mescher, Aras Ghaini, and David W. Agar Technische Chemie B, Technische UniVersita¨t Dortmund, 44221 Dortmund, Germany

The results of experimental studies on the parallelization of both single-phase and biphasic liquid-liquid slug flow in capillary microreactors are presented. Various flow distributor designs were investigated with respect to the uniformity of liquid-liquid flow in eight parallel capillaries and the two most promising concepts were identified. A novel multichannel phase splitting unit was installed at the downstream end of the capillaries to ensure well-defined phase separation. Further investigations on more complex reaction systems, in which the hydrodynamics depend on conversion and/or selectivity (e.g., polymerization), suggest that an active flow control for individual capillaries will be necessary in such cases. The uniformity of flow rate and flow structure could be achieved using a noninvasive capacitative measurement to characterize the flow in conjunction with specially developed microvalves and actuators for the regulation of flow rates and the manipulation of slug flow structure in each capillary. 1. Introduction The looming scarcity of conventional feedstocks has stimulated extensive research into more efficient processes for the chemical industry. Over the past decade or so microstructuring has emerged as one of the most promising new process intensification technologies to meet such challenges. A plethora of literature has been devoted to the potential benefits operation of using microstructured equipment (e.g.,1-7). Apart from precisely defined reaction conditions, microstructured devices offer more uniform product quality, quicker and more complete mixing of homogeneous systems, and very high mass transport rates for both gas-liquid and liquid-liquid biphasic systems. In the liquid-liquid slug flow regime, Taylor flow, i.e., the internal circulation vortex induced, enhances the mass transport between and within the two phases. The radial flow in the vicinity of the phase interface diminishes the concentration gradients within the slugs through convective mass transport. In addition to the convective enhancement of mass transport, the capillary reactor offers very high specific surface areas. The resulting overall kLa values for capillary reactors, depending on flow rate, capillary diameter, and so forth, are generally several orders of magnitude greater than in conventional equipment and may even exceed the performance available in specialized contacting equipment (see Table 1). Furthermore, due to the suppression of coalescence, this excellent mass transfer performance is achieved with minimal energy consumption. The slug flow capillary reactor can thus be used to overcome the mass transport limitations, which often have a deleterious effect on the performance of extraction processes or of biphasic reactions, such as nitrations.13 In spite of these benefits, instances of the industrial application of this technology remain few and far between. Closer inspection of the literature in this field reveals that the task of numberingup is far more challenging than the pioneers of microtechnology originally anticipated. Even with precise fabrication techniques and advanced simulation tools, it remains extremely difficult to achieve a reliable uniform flow distribution over multiple parallel microchannels. A numbering-up approach for homo* To whom correspondence should be addressed. E-mail: [email protected].

geneous one-phase systems was described by Schenk et al. in 2003,14 in which the need for a buffer volume upstream of the distribution and a high pressure drop in the downstream equipment were identified as crucial factors in ensuring an even flow distribution, within the limits imposed by the fabrication tolerances. Furthermore, Tonomura et al. in 200815 showed that the pressure, rather than the flow rate, must be controlled, if one is to prevent a distribution malfunction, such as plugging, in one capillary from adversely affecting the remainder. The presence of a second phase in the capillary dramatically increases the complexity of the flow distribution task, because not only the flow rate but also the biphasic flow structure must be uniform. It is known that the slug flow pressure drop is higher for greater interfacial areas. This is due to both the formation of the phase interface, which consumes energy to counteract the effect of surface tension, and the greater dissipation of energy as a consequence of the internal slug circulation.16 For uniform flow in parallel channels, each individual channel downstream of the distribution unit should thus have precisely the same pressure drop characteristics. Even if the pressure drop for all capillaries is initially uniform with homogeneous flow, the presence of a second phase and smaller slugs in one capillary would increase the pressure drop and thus diminish the flow rate in this capillary. Depending on the distribution unit, a smaller flow rate could aggravate this effect, resulting in the formation of still smaller slugs, until they become so small that the internal circulation desired no longer arises, i.e., when the slug length is in the same range as the inner capillary diameter. Another scenario observed was that individual capillaries only contain a single phase, i.e., complete maldistribution. This can occur because the generation of interfacial surface area to create a slug can require more energy than is needed to maintain the separate flow of each phase through different capillaries. A Table 1. kLa Values for Different Contacting Equipment, Compared to Liquid-liquid Slug Flow in Capillaries equipment

kLa [1/s]

reference

CSTR impinging stream contactor slug flow, capillary reactor slug flow, capillary reactor slug flow, capillary reactor slug flow, capillary reactor

-5

8 9 8 10 11 12

10.1021/ie100473d  2010 American Chemical Society Published on Web 07/01/2010

1.6 × 10 - 0.017 0.07-0.5 0.02-0.5 0.067-17.35 0.88-1.67 0.1-0.6

Ind. Eng. Chem. Res., Vol. 49, No. 21, 2010

10909

Figure 1. Schematic of the experimental setup.

further effect that can lead to total maldistribution is the change of surface properties over time, i.e., changes in the surface polarity at the capillary inlet, where the phase interface is generated. Such phenomena are often self-propagating, because the extended wetting times with a given phase modify the surface properties still further, and can take place within time spans of less than an hour. In contrast to the numbering up of liquid-liquid applications, several multichannel geometries have already been investigated for gas-liquid systems. In such cases the most important challenge is to avoid channeling, i.e., only one phase filling individual channels. For internal numbering-up of gas-liquid applications this problem is well-known.17-19 Channeling seems to result from a low pressure drop accompanying the distribution of the gas flow and can be avoided by a drastic narrowing of the distribution structure or an increase in the gas flow rate. Similar difficulties will be reported in this paper for experimentation with liquid-liquid systems and compared to these earlier results. Recently a numbering up for a liquid-liquid system with 6 parallel capillaries has been reported by Kashid.12 No problems with channeling were observed in this instance and it can be concluded that the external distribution of the liquids used generated a sufficiently high pressure drop to equalize the flow of both liquids in each capillary. Rapid phase separation immediately after reaction allows the precise regulation of residence times. It can be achieved by density differences (e.g., in hydrocyclones20,21) or by membrane separation.8,22 Furthermore, it is well-known that the phase separation at the downstream end of the capillaries can be simply carried out by inserting a steel needle into the hydrophobic capillary channel to separate the aqueous phase by exploiting the preferential wettability behavior.8 Because phase separation for individual capillaries complicates the operation of highly parallelized systems, a new multichannel phase splitting concept was developed. In the final results section, a detailed description of a new control concept is provided, which is a prerequisite for applications with complex reaction systems in which the hydrodynamic behavior is coupled with and sensitive to the conversion and selectivity of the reaction. The feasibility of this concept is demonstrated in single capillary experiments. The goal of this strategy is to fine-tune both the overall flow rate and the slug flow structure simultaneously by monitoring the flow using noninvasive capacitative measurements and employing appropriate novel micro valves and actuators to manipulate the slug flow in an individual capillary. 2. Experimental Methods The experimental setup is depicted in Figure 1. The fluids A, B, and C are fed into the distribution unit using microannular gear pumps (low pressure series micro annular gear pump mzr-

Figure 2. Distribution unit no. 1.

2921, HNP Mikrosysteme GmbH) to provide a nonpulsating flow, which is essential for uniform slug generation. For purposes of generalization, it was decided that the flow parallelization concept should not be restricted to just two immiscible phases. It is therefore envisaged that one should also be able to first mix two miscible liquids, say B and C, before adding a second immiscible liquid A to create slug flow. This arrangement covers a wider range of applications, such as the extraction of reaction (by-) products (Borg + C org f Daqu), the narrowing of the residence time distributions or the enhanced mixing of A and B using an inert phase, in addition to the obvious biphasic reaction/extraction processes. The pressure within the distribution unit is monitored and a PID controller provides the set-point for the organic phase pump with the set points for the aqueous phase pumps being supplied by a proportional rate controller (National Instruments Labview 8.5). The set point for the pressure was calibrated in earlier experiments to maintain the desired overall flow rate. A pressure of 0.1-0.3 bar was required to achieve total flow rates in the range of 100-1200 mL/h (for water/kerosene (ηkerosene ) 0.56 mPas)). With response times for the PID controller in the range of one second no detectable flow fluctuations were observed. It has been observed previously that measuring the pressure in the distribution unit helps to avoid undesirable coupling between the behavior of individual capillaries. Normally, even a small change in the volumetric flow rate in one capillary of less than 10% can distort the flow rate in all the remaining capillaries. This interdependent behavior is very important for reactive systems, in particular for those exhibiting precipitation, product plugging, or sensitive side-reactions, making it essential to maintain a uniform flow rate in all capillaries following malfunction in a single capillary. Between the actual capillary reactor (L2 ) 2 m; d2 ) 0.8 mm) and the distribution unit a second much shorter capillary (L1 ≈ 0.3 m; d1 ) 0.5 mm) (subsequently referred to as “pressure drop capillary”) is interposed and its length is adjusted to ensure equal pressure drop characteristics and to compensate for variations in manufacturing tolerances and systematic differences in the connection fittings for individual capillaries. Distribution unit no. 1 (Figure 2) comprises three elongated buffer reservoirs for each liquid. The inlet to the buffer volume is located in the center at the bottom of the volume. We observed no influence to the flow rate in the pressure drop capillary, which is connected to the outlets above these inlets. The two miscible liquids are first mixed in a Y-junction and the slugs are then formed by introducing the second phase through channel A into channel D. All channels are cylindrical with an internal diameter of 1.6 mm for bores A, B, and C and 0.8 mm for bore D. The

10910

Ind. Eng. Chem. Res., Vol. 49, No. 21, 2010

Figure 3. Distribution unit no. 2.

distribution unit was fabricated out of PTFE. In the initial experiments this unit gave rise to complete maldistribution. Since no interface was being generated, the resulting flow rates were much higher than with all other distribution units tested. As explained above, the flow was following the path of least resistance, which in this case meant single phase flow through individual capillaries without interface formation. It proved possible to resolve this shortcoming by simply introducing a short narrow constriction as a resistance element in the line A to even out the flows in each reaction capillary. The constriction is a simple PTFE-capillary (dinner ) 0.3 mm, douter ) 1.6 mm, L ) 30 mm), which can be inserted into each channel (A, B, C). If, for example, the flow of the liquid A decreases in one of the 8 capillaries due to the additional energy expenditure needed to generate phase interface, as it has occurred before the insertion, the additional dynamic pressure drop in channel A diminishes and thus the pressure level in channel A will increase until the phase interface is generated. After inserting the narrow constricting elements a consistent slug formation was achieved in all experiments. Distribution unit no. 2 (Figure 3) comprises a rotor (PTFE reinforced with glass fiber) and a stator (steel). Each liquid is fed through the stator into a groove on the rotor, which is enclosed between two O-ring seals, and fills the fluid chamber at the top of the rotor at the desired pressure. At the outlet face of the rotor, the fluid chamber is again enclosed by a sealing ring and thus continuously feeds the outlets to the pressure drop capillaries. Due to the continuous rotation of this chamber it supplies each pressure drop capillary in turn. By manipulating the rotation speed and thus the switching times of the outlets, the resulting slug length can be regulated. In Figure 3, only the fluid chamber for liquid C is illustrated, the channels and chambers for liquids A and B have been omitted for reasons of clarity. To ensure the desired pressure prevails in all interstices and chambers, it is measured separately for each liquid within the unit (e.g., PI A for liquid A). The use of the proportional controller results in sudden pressure pulses due to the switching between the outlets. In addition to the performance assessment of the two different flow distribution units, a novel multichannel separator (see Figure 4) for biphasic liquid-liquid systems located at the downstream end of the capillaries was developed. The phase splitting is achieved by exploiting two distinct physical phenomena. On the one hand the different densities of the two

Figure 4. Phase separator.

phases give rise to phase separation due to gravitational forces in the buffer volume at the center of the unit, as in conventional decanters. If the phase separation must be carried out rapidly, because a prolonged residence time might falsify the reaction behavior for example, it is possible to decrease the internal volume by inserting perforated discs. On the other hand, the different wetting affinity of the organic and aqueous phase for the materials used for the upper and lower sections of the disengagement chamber also expedites phase separation. The driving force in this case is the surface tension and the capillary pressure helps prevent cross-contamination of the organic outlet with the aqueous phase, for instance. To manipulate the capillary pressure required, which is strongly dependent on the inner diameter, it is possible to insert different capillaries connecting the buffer chamber with each outlet. 3. Results 3.1. Pressure Drop Adjustment for Uniform Flow Distribution. Initially, the lengths of all pressure drop capillaries were exactly 0.3 m (see L1 in Figure 1). With these unmodified capillaries, the standard deviation of single aqueous phase (deionized water) flow rates was in the range of 6% for around 50 mL/h per reaction capillary. In all experiments the flow rate

Ind. Eng. Chem. Res., Vol. 49, No. 21, 2010

10911

Figure 5. Results of distribution unit no. 1 (left: water; right: water-kerosene).

and the organic phase ratio in each capillary were measured by determining the liquid volume (using a syringe) after filling a sample volume over a fixed time span. The parallel pressure drop capillaries are modified to achieve a uniform flow distribution for homogeneous flow. The precise adjustment requires eight individual pressure drop capillaries for each distribution unit. Using the Hagen Pouseuille formula: ∆p ) (128 η · V˙L)/(π · d4) an appropriate length L*2 is calculated for each capillary system (index i), excluding the pressure drop contribution of the pressure drop capillary (index 1) (128 η)/(π) · V˙i• (L1,i)/(d41,i). The length L*2 differs from the length of the reaction capillary (index 2), because it is calculated using the flow rates measured and thus it includes the manufacturing tolerances and their impact to the pressure drop. Since the overall pressure drop ∆ptot is maintained at a constant value it is possible to calculate how much L1,i has to be shortened to increase the flow rate to V˙max (see eq 2). In the next step, all pressure drop capillaries were trimmed accordingly, with the exception of the one exhibiting the highest flow rate (V˙max).

∆ptot ) const. ⇒ k · V˙i

(

L1,i 4 d1,i

+

L1,i(new) )

L* 2,i 4 d2,i

)

( ( V˙i

(

) k · V˙max

L1,i(new) 4 d1,i

+

) )

L* 2,i 4 d2,i

L* L* L1,i 2,i 2,i 4 + 4 - 4 d1,i 4 V˙max d1,i d2,i d2,i

)

S

(2)

It should be noted that using exactly ∆L1,i (the length which has to be shortened in capillary system i) will not result in a perfectly uniform flow distribution, due to the impact of the final fitting, which has to be tightened after shortening the pressure drop capillary. The best procedure was to shorten each capillary by about 70% of the calculated ∆L1 before rechecking the flow and, if necessary, trimming it again. In this manner it proved possible to decrease the standard deviation of the flow distribution for homogeneous single phase flow to less than 1% within two steps. The repeatability of the experiment carried out three times without any additional changes, e.g., shortening, new set-point, or structural alteration was in the range of 0.3% which can therefore be considered to be the minimum value

that can be attained with the pumps, pressure measurements, etc. employed in this setup. To achieve uniform flow rates for the two-phase system water-kerosene, the same procedure may be adopted. In actual fact, the pressure drop in the slug flow regime is not linearly dependent on the flow rate, as assumed above. The use of a more suitable model for pressure drop calculation16 is conceivable, but probably superfluous, since the differences caused by nonuniform slug length and the retightening the fittings after shortening give rise to a much more significant error in the calculation of the actual truncation length. 3.2. Distributor Performance. For single-phase experimentation the liquids A, B, and C were all water and for the two phase experiments kerosene was used as liquid A (see Figures 1-3). The results for distribution unit no. 1 (see Figure 5) show an almost perfect distribution for a homogeneous single phase flow. Compared to literature data (σstd23 ) 2%, ∆max23 ) 5%, ∆max24 ) 3%) the standard deviation and the maximal deviation (σstd ) ((1/n∑ (V˙i - V˙mean )2)/(V˙mean ) and ∆max ) (Max{(V˙i - V˙mean )2})/(V˙mean )) could be decreased significantly. The uniform flow distribution of single-phase flow was achieved for all distributors studied and no influence on the overall flow rate was observed. It can thus be concluded that the flow deviations resulting from fabrication tolerances can be virtually eliminated. For distribution unit no. 1 it also proved possible to achieve a nearly uniform, two-phase flow distribution as well. For distributor 1, the slug length is a function of the capillary used and the characteristic dimension of the geometrical structure at which the phase interface is generated. To decrease the slug size, the bore of channel D (see Figure 2) was widened and a PTFE capillary (dinner ) 0.8 mm) was inserted prior to the inlet of channel A, with a larger bore (dinner ) 1 mm) thereafter. In this manner, slug lengths in the range of 2-4 mm were generated. As expected, the capillary with the highest flow rate contained the longest slugs and gave rise to the lowest pressure drop resistance. It is conjectured that one reason for different slug lengths might be the manufacturing tolerances of the inserted capillary bores. Only minor influences to the overall flow rate were observed, but in general it can be stated that a higher level of pressure drops within the distributor structure resulted in more uniform flow distribution. In contrast to the results for distribution unit no. 1, a strong influence of the overall flow rate on the performance was found for distribution unit no. 2 (see Figure 6). For higher flow rates, it proved possible to achieve similarly uniform flow distributions. The wetting time, which can be regulated by the rotation

10912

Ind. Eng. Chem. Res., Vol. 49, No. 21, 2010

Figure 6. Results of distribution unit no. 2 (water-kerosene). Left: flow rate and organic phase ratio at 0.3 bar pressure level inside the distributor 2; right: resulting flow rates for various pressure levels.

Figure 7. Cross-contamination of the organic (org) and the aqueous (aqu) phase outlets for the system water/hexane and water/toluene vs overall organic phase ratio in the feed. Phase separator, without discs inlays and dinner ) 0.3 mm for the inserted capillaries, at a constant overall flow rate of 200 mL/h.

speed of the rotor, regulates the slug length. This gives an additional degree of freedom to the operator when setting process parameters. However, the slug length is also a function of the flow rate in each capillary. The slug flow obtained was consistent, but differed between the capillaries due to the different flow rates. As mentioned above, at a higher flow rate, and thus a higher pressure drop, the flow variances were lower and the slug structures were therefore more uniform. The greater deviations at lower flow rates (V˙mean < 20 mL/h) resulted in a much larger variation in slug length and caused an unwanted positive feedback between the decrease in flow rate and slug length in some capillaries. 3.3. Phase Separator Performance. After trimming the pressure drop capillaries to equalize the flow distribution for the two-phase flow, the phase separator was installed at the downstream end of the reaction capillaries. During the experiments no influence of different flow rates in individual capillaries on the phase separator performance was observed. In the following section the flow rates describe the sum of those from all eight capillaries. The cross-contamination was measured using a syringe to separate the phases leaving both outlets. The cross contamination in each outlet is defined as the ratio of the unwanted phase (e.g., the aqueous phase in the PTFE organic phase outlet) to the total sample volume. During the experimental measurements only a minor influence of the overall flow rates on the degree of phase separation was observed. The relative outlet pressure drop in the organic and aqueous phase outlets was found to be more decisive for the overall performance. The optimal phase separation point (located at 70% volumetric inlet organic phase content in Figure 7) can be easily manipulated by shortening the outlet capillaries, or

Figure 8. Phase separation of water/toluene, with discs inlays and dinner ) 0.3 mm for the inserted capillaries, at a constant overall flow rate of 200 mL/h, before and after rinsing the separator with H2SO4.

by using appropriate valves for fine-tuning the flow rate. Only minor differences were observed when using hexane instead of toluene, but the higher surface tension of the hexane-water system (γwater/hexane ) 51 mN/m, γwater/toluene ) 36 mN/m) resulted in less cross-contamination. It is obvious that with single-phase inlet conditions, the cross contamination in one outlet must be 100% (e.g., 100% contamination in the organic outlet for 0% Vorg in the inlet). Thus the closer the ideal separation point is to single-phase inlet conditions, the stronger the increase in crosscontamination (for example the aqu outlet for water/hexane at Vorg > 80% in Figure 7). It is always possible to obtain at least one pure phase and in the vicinity of the ideal phase separation point both outlets were free of cross-contamination. After filling the chamber of the separator with perforated discs (2× steel/2× PTFE, height ) 2 mm, douter ) 10 mm each with 5 bore holes, dinner ) 1.6 mm) no range of inlet compositions without cross-contamination was observed (see Figure 8, dotted lines for Vorg ) 75% compared to Figure 7 dotted lines for Vorg ) 60 to 70%). This can be explained by the small pressure and flow variations as a consequence of slug formation, coalescence, or simply due to fluctuations from the pumps. These small variations seem to have been dampened out in the absence of the displacement inserts, as depicted in Figure 7. Surface effects were observed after rinsing the system with 2 M sulfuric acid, which strongly affected the phase separation. Sulfuric acid was chosen for flushing because of its potential use in later applications (e.g., Villermaux/Dushman reaction for accessing micro mixing). First, the optimal phase ratio for complete disengagement was shifted to lower organic phase fractions. The pressure drop characteristic before and after the sulfuric acid rinsing also indicates this shift. It was observed that after rinsing with highly concentrated sulfuric acid, the pressure drop in the metallic aqueous phase outlet line, measured by plugging

Ind. Eng. Chem. Res., Vol. 49, No. 21, 2010

10913

Figure 9. Control concept for a single capillary.

the organic phase outlet with a pressure sensor and using pure water prior to and following rinsing with H2SO4, decreased for the same flow rate. As a consequence of this, the optimal phase ratio was shifted to higher aqueous phase contents. Additionally it seems that the PTFE material was also affected by the acid and subsequently, even at higher organic phase fractions, a slight cross-contamination in the aqueous phase was observed. The greater the internal diameters of the inserted capillaries (PTFE for the organic outlet and steel for the aqueous outlet), the more strongly the phase separation results were influenced by the acid rinsing procedure. In some experiments unstable conditions and a variation of wetting properties with time ranges of minutes to hours were observed before steady-state phase separation conditions could be established. 3.4. Measurement and Control of Slug Flow. Measuring slug velocities has previously been reported in the literature for gas-liquid applications (e.g., microscopic particle image velocimetry (µ-PIV)25 or pulsed-laser fluorescence imaging26). The dielectric conductivity of the organic and aqueous media typically differ by a factor of 30 or more and are thus eminently suitable for slug flow measurements. This method is well described in the literature and has been used in the past for microchip applications and capillary electrophoresis.27-29 The new control concept, which is intended to be used for further experimentation with more complex and sensitive reaction systems, is shown in Figure 9. It is important that the regulation of overall flow rate, the phase ratio, and the slug flow structure be based on a non-invasive flow characterization to avoid unwanted distortions. Measuring the change in dielectric permittivity between two ring electrodes can be used for the detection of the change between organic and aqueous inside the capillary and thus monitor both slug size and velocity. I)

IA

(

σl · b E+

ε·

f ∂E(t) ∂t

displacement current

)

b dA

(3)

The influence of dielectric properties on the current are wellknown (see eq 3 in Maxwell 186530). Current (I) is a function not only of the electric field strength (E) and the electric conductivity (σl) over the cross-sectional area A, but also of changing electric field strength with time E(t) , which induced the so-called displacement current. Imposing an oscillating voltage on the two ring electrodes yields the displacement current, the magnitude of which depends on the dielectric permittivity (ε) of the medium lying between them. As mentioned above, the large differences in the dielectric constants of the organic and the aqueous phase lead to a sharp change in the displacement current induced. The displacement current data acquisition was carried out by a log-in amplifier (Stanford Research Systems SR830 DSP lock-

Figure 10. Results of the capacitative measurement with different slug velocities.

Figure 11. Slug size control valve.

in amplifier), which additionally generates the optimal frequency for the impedance determination. The log-in amplifier is specially designed for the processing of signals with low signalto-noise ratio. The output signal intensity is the time derivative of the displacement current. The amplitude of signal reflects the time interval over which the relative permittivity of the medium changes (see Figure 10), which in turn is proportional to the slug velocity. The relation between the slug velocity and the signal frequency enables one to calculate the slug lengths. Thus the slug flow structure can be characterized without difficulty in real time. As the slugs are formed by a droplet break-off process, the slug length is primarily influenced by the scale of the local geometry. For a smaller channel diameter, the droplet volume and thus the slug length corresponding to the critical size for break-off is decreased. Comprehension of this process enables one to manipulate slug size by changing the cross-section of the location at which the droplet disengagement occurs. For this purpose, an adjustable set-screw was introduced into the y-mixer structure, to hasten or retard slug formation (see Figure 11). Figure 12 shows qualitatively how the slug length can be tuned to increased values by increasing the effective channel width ∆s from 0.5 to 2 mm inside the slug size control valve by withdrawing the set-screw. It proved possible by this means to obtain uniform slug lengths in range of 1-20 mm with standard deviations of slug length in the range of 5-10%. The 3 snapshots of the same capillary, taken after a 1 min delay for 3 different ∆s settings illustrate clearly that the repeatability and uniformity achieved are good (see Figure 12). Manual adjustment of the slug size will suffice, because for moderate changes in the overall flow rate the variation in slug size is minimal. Conventional valves for fine flow regulation employ specially profiled tapered needles. By progressively screwing the needle

10914

Ind. Eng. Chem. Res., Vol. 49, No. 21, 2010

Figure 12. Results of the slug size manipulation (80 mL/h, phase ratio 1:1, kerosene-water, dinner ) 1 mm). Figure 14. Laminar pressure drop characteristics of the new micro valve compared to a conventional HPLC needle valve.

Figure 13. New micro valve concept.

into a narrow orifice, the pressure resistance to flow is increased due to the diminishing free cross-sectional area available. To regulate flow rates precisely in the range of mL/h with commercially available needle valves entails costs in the range of 1000 USD per valve. Most needle valves permit one to fully shut off the flow. For the fine-tuning of parallel capillaries, however, it is only necessary to regulate the 10% or so deviations between individual capillaries. A major part of the operating range of a conventional needle valve is thus superfluous for this purpose. A new micro valve was developed which is more suitable for the task in hand and is both simple and cheap to fabricate (see Figure 13). The new micro valve no longer requires a high precision tapered needlesa simple piece of standard steel wire gauge can be used as the throttling element. To increase the pressure drop flow resistance, the constant diameter needle is inserted into a slightly larger capillary channel. Rather than constricting the free cross-sectional area, the length of the constricted channel section is increased. The micro valve can be installed in the connecting capillary upstream of the slug size control valve or at the end of the capillary reactor (if the phase ratio is not to be manipulated, a single micro valve located at the downstream end of the capillary and before the phase separator would be sufficient). To compare the new micro valve with a standard HPLC needle valve (Upchurch Micro Metering Valve P-445) experiments were carried out to identify {∆p ) V˙ · µ · Klaminar, Klaminar ) ((128 · L)/(π · d4)} (Hagen Poiseuille law) at different set-screw positions. To manipulate the flow rates in the range of interest (10-100 mL/h with the range of ∆pvalve ) 50-200 mbar), Klaminar values of 1 × 1015 m-1 to 1 × 1016 m-1 were required. To achieve such values with the HPLC valve, it is necessary to operate almost at closure and the flow could be only regulated using the last 20% of the adjusting screw mechanism (dotted line), corresponding to an absolute displacement of only 2 mm. The new microvalve on the other hand has an adjustment range of more than 50 mm. If different Klaminar values are required, the range can be shifted simply by using different steel wire gauges or by modifying the capillary sheath to increase or decrease the free cross-sectional area. Figure 14 illustrates the almost linear dependency of Klaminar on the depth of the needle insertion.

The first prototype of the new micro valve was driven by a simple actuator motor (Elliptec Resonant Actuator Module R40N), which, in contrast to normal step motors, needs no additional external encoder. By this means it is possible to keep the additional costs in the range of 100 USD per regulated capillary. 4. Conclusion For the uniform distribution and structuring of biphasic flows in parallelized capillaries the key question to be resolved is what are the most important factors influencing the distribution and flow structure (e.g., surface effects, fabrication tolerances, etc.) for a given distributor?. In all the devices studied, a strong interaction between the flow rate and the slug length was observed. Smaller slug lengths give rise to increased pressure drop and thus diminished flow rates. It can be concluded that an external numbering-up of biphasic liquid-liquid systems is feasible down to tolerances of within 4% (standard deviation), as opposed to less than 1% for single-phase flow. At the downstream end of the parallel capillaries a novel phase separator was installed. A separation exploiting gravitational forces and due to the specific preferential wettability of the organic and aqueous fluids on different materials was combined in a single unit. It is always possible to obtain at least one pure phase in one of the two outlets. Depending on precise pressure drop characteristics of the outlets and their adjustment by needle valves, the phase separation can also be operated without any cross-contamination. For specific applications in the future one must evaluate how much the expected deviations in flow rate will affect the selectivity of the desired reaction. For example, in simple mass transport limited extraction applications, deviations of the magnitude indicated would probably only have a marginal influence on the overall performance of the parallelized system. However, in other cases, such as polymerization reactions, the conversion as a function of the residence time and thus flow rate in each capillary will strongly affect the pressure drop. In the presence of such interactions, dynamic fluctuations, selfreinforcing divergence, or even complete maldistribution can occur. To overcome such problems, a noninvasive capacitative measurement technique was developed, as a basis for fine-tuning the flows in each individual capillary. By measuring the displacement current between two adjacent ring electrodes, it is possible to measure the rate of change of the liquid capacitance in the intervening section of the capillary. By processing the measurement signal, using a so-called log-in

Ind. Eng. Chem. Res., Vol. 49, No. 21, 2010

amplifier, it is then possible to characterize both the biphasic flow rates and slug structure. The second prerequisite for manipulating the flow in individual capillaries was to develop appropriate new micro valves, because commercially available valves are both expensive and exhibit unsuitable pressure drop characteristics. As the objective is only to “tweak” a flow distribution, which is already uniform to within a deviation of less than 10%, many of the demands made on conventional valves are irrelevant. Besides controlling the overall flow rates and flow ratios, the new measurement also provides the opportunity to regulate slug length. For this purpose, the geometry of the location at which the slug is formed must be adjustable. Since the slugs are formed by a reciprocating process, with each phase displacing the other until it encounters the opposite wall of the mixing element, this distance to the opposite wall and thus the resulting slug volume can be manipulated by installing a set-screw. The resulting slug sizes could be set in the range of 1-20 mm with a standard deviation in the range of 5-10% between individual slugs in a capillary. This is less critical than a similar constant deviation between capillaries, because the number of slugs along the capillary tend to compensate for discrepancies in their individual lengths. Although issues concerning the integration of the measurement and control techniques described into a single robust unit and the precise control algorithms remain to be resolved, the underlying feasibility of the concept has been demonstrated. Furthermore, the regulation of slug flow in this manner should make it possible to employ simpler liquid delivery systems in place of the costly high-precision, pumps previously used in such work. Acknowledgment We thank the Max Buchner Forschungsstiftung for the financial support of this work. Appendix NOTATION kLa ) Mass transfer coefficient [1/s] L ) Length [m] PI ) Pressure indicator [bar] ∆p ) Pressure drop [bar] η ) Dynamic viscosity [Pa s] d ) Diameter [mm] V˙ ) Flow rate [mL/h] k ) Constant [Pa s] σstd ) Standard deviation [%] ∆max ) Maximal deviation [%] γ ) Surface tension [mN/m] I ) Current flux [A] A ) Area [m2] σ ) Electric conductivity [A/(V m)] E ) Electric field [V/m] ε ) Dielectric conductivity [A s/(V m)] t ) Time [s] ∆s ) Distance [mm] Klaminar ) Constant [1/m3]

Literature Cited (1) Lerou, J. J.; Harold, M. P.; Ryley, J.; Ashmead, J.; O’Brien, T. C. Microfabricated Minichemical Systems: Technical Feasibility. DECHEMA Monogr. 1996, 132, 51–69.

10915

(2) Jensen, K. F. Microreaction engineering -- is small better? Chem. Eng. Sci. 2001, 56 (2), 293–303. (3) Gavriilidis, A.; Angeli, P.; Cao, E.; Yeong, K. K.; Wan, Y. S. S. Technology and Applications of Microengineered Reactors. Trans. IChemE 2002, 80, 3–30. (4) Hessel, V.; Lo¨we, H. Microchemical Engineering: Components, Plant Concepts, User Acceptance - Part III. Chem. Eng. Technol. 2003, 26 (5), 531–544. (5) Hessel, V. From Microreactor Design to Microreactor Process Design. Chem. Eng. Technol. 2005, 28 (3), 243. (6) Hessel, V.; Lo¨we, H.; Scho¨nfeld, F. Micromixers--a review on passive and active mixing principles. Chemical Engineering Science 5th International Symposium on Mixing in Industrial Processes (ISMIP5); 2005, 60 (8-9), pp 2479-2501. (7) Lo¨bbecke, S.; W. Ferstl, S.P.c.; Tu¨rcke, T. Concepts for Modularization and Automation of Microreaction Technology. Chem. Eng. Technol. 2005, 28 (4), 484–493. (8) Kashid, M. N.; Harshe, Y. M.; Agar, D. W. Liquid-liquid slug flow in a capillary: An alternative to suspended drop or film contactors. Ind. Eng. Chem. Res. 2007, 46 (25), 8420–8430. (9) Dehkordi, A. M. Novel Type of Impinging Streams Contactor for Liquid-Liquid Extraction. Ind. Eng. Chem. Res. 2000, 40 (2), 681–688. (10) Zhao, Y. C.; Chen, G. W.; Yuan, Q. Liquid-liquid two-phase mass transfer in the T-junction microchannels. AIChE J. 2007, 53 (12), 3042– 3053. (11) Ghaini, A.; Kashid, M. N.; Agar, D. W. Effective interfacial area for mass transfer in the liquid-liquid slug flow capillary microreactors. Chem. Eng. Process.: Process Intensification 2010, 49 (4), 358–366. (12) Kashid, M. N.; Gupta, A.; Renken, A.; Kiwi-Minsker, L. Numbering-up and mass transfer studies of liquid-liquid two phase microstructured Reactors. Chem. Eng. J. 2010, 158 2 (1), 233–240. (13) Burns, J. R.; Ramshaw, C. A Microreactor for the Nitration of Benzene and Toluene. Chem. Eng. Commun. 2002, 189 (12), 1611–1628. (14) Schenk, R.; Hessel, V.; Hofmann, C.; Lowe, H.; Schonfeld, F. Novel liquid-flow splitting unit numbering-up of liquid/liquid specifically made for chemical microprocessing. Chem. Eng. Technol. 2003, 26 (12), 1271– 1280. (15) Tonomura, O.; Tominari, T.; Kano, M.; Hasebe, S. Operation policy for micro chemical plants with external numbering-up structure. Chem. Eng. J. 2008, 135, 131–S137. (16) Kashid, M. N.; Agar, D. W. Hydrodynamics of liquid-liquid slug flow capillary microreactor: Flow regimes, slug size and pressure drop. Chem. Eng. J. 2007, 131 (1-3), 1–13. (17) de Mas, N.; Gu¨nther, A.; Schmidt, M. A.; Jensen, K. F. Microfabricated multiphase reactors for the selective direct fluorination of aromatics. Ind. Eng. Chem. Res. 2003, 42 (4), 698–710. (18) Wada, Y.; Schmidt, M. A.; Jensen, K. F. Flow Distribution and Ozonolysis in Gas-Liquid Multichannel Microreactors. Ind. Eng. Chem. Res. 2006, 45 (24), 8036–8042. (19) Yue, J.; Boichot, R.; Luo, L.; Gonthier, Y.; Chen, G.; Yuan, Q. Flow distribution and mass transfer in a parallel microchannel contactor integrated with constructal distributors. AIChE J. 2010, 56 (2), 298–317. (20) Colman, D. A.; Thew, M. T. Correlation of Separation Results from light Disperion Hydrocyclones. Chem. Eng. Res. Des. 1983, 61 (a), 233– 240. (21) Schu¨tz, S.; Gorbach, G.; Piesche, M. Modeling fluid behavior and droplet interactions during liquid-liquid separation in hydrocyclones. Chem. Eng. Sci. 2009, 64 (18), 3935–3952. (22) Kralj, J. G.; Sahoo, H. R.; Jensen, K. F. Integrated continuous microfluidic liquid-liquid extraction. Lab Chip 2007, 7 (2), 256–263. (23) Schenk, R.; Hessel, V.; Hofmann, C.; Kiss, J.; Lo¨we, H.; Ziogas, A. Numbering-up of micro devices: a first liquid-flow splitting unit. Chemical Engineering Journal 7th International Conference on Microreaction Technology, 2004, 101 (1-3), pp 421-429. (24) Togashi, S.; Tetsuro Miyamoto, Y.A.a.Y.E. Yield Improvement of Chemical Reactions by Using a Microreactor and Development of a Pilot Plant Using the Numbering-Up of Microreactors. J. Chem. Eng. Jpn. 2009, 42, 512–519. (25) van Steijn, V.; Kreutzer, M. T.; Kleijn, C. R. [mu]-PIV study of the formation of segmented flow in microfluidic T-junctions. Chemical Engineering Science 8th International Conference on Gas-Liquid and GasLiquid-Solid Reactor Engineering, 2007, 62 (24), pp 7505-7514. (26) de Mas, N.; Gu¨nther, A.; Kraus, T.; Schmidt, M. A.; Jensen, K. F. Scaled-Out Multilayer Gas-Liquid Microreactor with Integrated Velocimetry Sensors. Ind. Eng. Chem. Res. 2005, 44 (24), 8997–9013.

10916

Ind. Eng. Chem. Res., Vol. 49, No. 21, 2010

(27) Zemann, A. J.; Schnell, E.; Volgger, D.; Bonn, G. K. Contactless Conductivity Detection for Capillary Electrophoresis. Anal. Chem. 1998, 70 (3), 563–567. (28) Pumera, M.; Wang, J.; Opekar, F.; Jelinek, I.; Feldman, J.; Lowe, H.; Hardt, S. Contactless Conductivity Detector for Microchip Capillary Electrophoresis. Anal. Chem. 2002, 74 (9), 1968–1971. (29) Liu, S.; Gu, Y.; Roux, R. B. L.; Matthews, S. M.; Bratton, D.; Yunus, K.; Fisher, A. C.; Huck, W. T. S. The electrochemical detection of droplets in microfluidic devices. Lab Chip 2008, 8 (11), 1937–1942.

(30) Maxwell, J. C. A Dynamical Theory of the Electromagnetic Field. R. Soc. Trans. 1865, 155, 459–512.

ReceiVed for reView March 3, 2010 ReVised manuscript receiVed June 4, 2010 Accepted June 21, 2010 IE100473D