Design, simulation and application of a new micromixing device for time resolved infrared spectroscopy of chemical reactions in solution

Abstract

We present a novel micromachined fast diffusion based mixing unit for the study of rapid chemical reactions in solution with stopped-flow time resolved Fourier transform infrared spectroscopy (TR-FTIR). The presented approach is based on a chip for achieving lamination of two liquid sheets of 10 μm thickness and ∼1 mm width on top of each other and operation in the stopped-flow mode. The microstructure is made on infrared transmitting calcium fluoride discs and built up with two epoxy negative photoresist layers and one silver layer in between. Due to the highly laminar flow conditions and the short residence time in the mixer there is hardly any mixing when the two liquid streamlines pass through the mixing unit, which allows one to record a mid-IR transmission spectrum of the analytes prior to reaction. When the flow is stopped, the reactant streams are arrested in the flow-cell and rapidly mixed by diffusion due to the reduced interstream distances and the reaction can be directly followed with hardly any dead time. On the basis of two model reactions—neutralisation of acetic acid with sodium hydroxide as well as saponification of methyl monochloroacetate—the performance of the mixing device was tested revealing proper functioning of the device with a time for complete mixing of less than 100 ms. The experimental results were supported by numerical simulations using computational fluid dynamics (CFD), which allowed a reliable, quantitative analysis of concentration, pressure and flow profiles in the course of the mixing process.

---

Introduction

There is a clear trend toward miniaturisation of devices and tools needed to perform analytical and synthetic chemistry as well as integration of these units into compact systems. This development allows one to perform more unit operations of synthesis and analysis in less time by consuming less reagents, samples by reduced waste production and energy consumption.^1,2 The design, fabrication and application of miniaturised chemical systems have gained tremendous progress in recent years meeting the requirements for robust systems with reduced reagent consumption, ease of automation and handling of toxic or harmful reagents. Integrating different microfluidic devices on a single chip enabled the development of fully implemented chemical reaction systems for preparative,^3,4 analytical and bioanalytical chemistry^5–7 and in particular capillary electrophoresis based separation techniques.^8–10 Mixing is one of the most important procedures in both analytical and synthetic chemical reaction systems. Especially when downscaling the fluid components to the micrometer scale, rapid mixing becomes a challenging task, as due to strictly laminar flow conditions this is achieved by diffusion only, which is a rather slow process, even over short distances. However, several miniaturised mixers have been developed so far. One of the most common methods is based upon increasing the contact surface of fluids by dividing and rearranging the fluid streams, which results in reduced diffusion distances.^11–17 Other approaches use simple T-connectors to achieve fast and reproducible diffusion based mixing.^18–21 Using electrokinetically driven microfluidic components with charged analytes and buffered solution mixing can also be achieved by applying a voltage by means of nobel metal films integrated into the microfluidic pattern.²² The vast majority of such devices use silicon or glass as substrate materials, as the fabrication procedures—photolithography and wet etching—are well-established processes. For other support materials such as polymers and metals, radiation induced etching or moulding procedures are the methods of choice.²³

Time resolved Fourier transform infrared spectroscopy (TR-FTIR) has evolved to an efficient technique to study fast chemical processes providing structural information on the dynamics of the system under investigation in the time frame from milli- to picoseconds. One key feature of these state-of-the-art-techniques is that the reaction or event under study must be highly precisely repeatable up to several hundred times, which is mostly achieved using laser^24–26 or voltage pulses or by stretching the sample,^27,28 which certainly limits the application to a few chemical systems. However, little work has been done so far to develop a tool that enables one to investigate any chemical reaction by rapidly mixing the reagent solutions. Most of these devices make use of a mixing chamber prior to an appropriate flow cell, which implies a certain dead time from the point of mixing to the point of measurement.^29–32 In order to develop a mixing device applicable to study fast reactions in solution at the point of mixing with TR-FTIR one has to consider, that mixing has to be achieved within the shortest time possible with minimal or even no dead time. In addition, the pathlength of the device at the point of measurement has to be less than 50 μm due to the high absorption of water and organic solvents in the mid IR spectral region. Little reagent consumption is of crucial importance too, as for TR-FTIR the mixing event has to be repeated several times to achieve a good S/N ratio. Finally a mechanically robust system with high chemical resistance is necessary to be able to use high flow rates of aqueous solutions, but also organic solvents. The method of choice is a miniaturised mixing unit on the basis of an IR transmitting material. Recently we proposed, for the purpose of fast mixing of two liquids, a micromachined device based on multilamination which out of two feeding channels produced an alternating sequence of 50 ‘spaghetti like’ streamlines adjacent to each other with a width of 20 μm each.³³ Whereas this device worked properly with a mixing time of about 200 ms there are several drawbacks related with that initial design, which have now been improved in the new design with respect to mixing time, susceptibility to blockage and micro machining process. The new microstructure is, as the previous one, constructed using an epoxy negative photoresist resin material on IR transmitting calcium fluoride discs and is also operated in the stopped-flow mode. The new design is based on superimposing two reactant streams into two ‘lasagne like’ flow sheets (∼10 μm each) that pass attached to each other through the infrared probed main channel. Due to the highly laminar flow conditions and the short residence time in the mixing unit hardly any mixing occurs prior to stopping the flow. In that way it is possible to obtain a spectrum of the reactants prior to reaction. When the flow is stopped, the liquids are rapidly mixed by diffusion due to the small fluid-layer thickness and the reaction can be directly followed. To make the new design possible it was furthermore necessary to develop a technique to build up three-dimensional closed microstructures on IR transparent support materials.

Recently, the application of computational fluid dynamics (CFD) methods in chemical engineering^34–36 has become increasingly important for the solution of engineering flow problems and the optimisation of complex geometries. However, the need to look at various species, reaction kinetics, phase change, mass transfer between phases, particle-fluid flows limits the number of commercially available software products in chemical engineering. Among the frequently used products are CFX™, FIDAP™ and FLUENT™.

Due to laminar flow conditions in microfluidic devices³⁷ CFD calculations are less complicated and often more reliable than turbulent flow, multiphase flow or particulate flow calculations in other areas of chemical engineering. However, many microfluidic geometries exhibit large aspect ratios (length ratio of largest and smallest scales within a geometry), thus leading to very fine meshes (>10⁶ cells) to achieve the required accuracy for the discretisation.

Experimental

Preparation of the mixing device

For IR transmission measurements conventional materials that are usually used for microstructuring are disadvantageous due to their low transmittance in the mid-IR region (see Fig. 1). Therefore, a special process has been developed to build fluid handling structures consisting of two polymer layers and one metal layer in between. The negative working photoresist SU-8 is used for the polymer layers. This epoxy-based material allows forming structures with an aspect ratio of 10∶1–15∶1. The structures show high mechanical strength and very good chemical resistance. For the metal layer silver is used because films of this metal can be easily deposited by evaporation with a thickness of 2 μm. The device is built up as follows: A CaF₂ disc (1 mm thickness, 20 mm diameter) is spin-coated with diluted SU-8-50 to obtain a layer thickness of ∼10 μm and soft-baked. The disc is then UV-exposed to pattern the bottom layer. After the post-exposure-bake the metal layer is deposited by thermal evaporation. This metal layer is spin-coated with positive photoresist (AZ 1529), which is dried, exposed and developed as usual. During this second exposure the metal layer prevents the SU-8 layer being exposed. Therefore the metal layer can be structured freely, independent of the structure of the first SU-8 layer. The metal layer is patterned by wet chemical etching. After removal of the AZ photoresist the second (top) layer of SU-8 is deposited, soft-baked and exposed to form the structure of the top layer. Top and bottom layer are now developed, i.e. the non-exposed areas are dissolved in an appropriate solvent. After a final hardbake a second CaF₂ disc which contains the holes for two inlets and one outlet for the liquids, is placed on top of the structure and sealed with light-curing epoxy resin. This technology allows one to combine polymer and metal layers to build ‘closed’ structures such as channels, chambers etc. as well as ‘free’ metal structures such as membranes, sieves etc. and therefore it is well suited for miniaturised fluid handling devices. Fig. 2a shows a schematic view of the whole device with two inlet and one outlet hole and the IR probed main channel, where the fluid sheets are superimposed. A detailed insight into the sandwich construction of polymer layers and silver layer is given in Fig. 2b (the poor quality of the edges is due to the low-cost masks used here). The fluid entering through inlet 1 is forced to flow under the separating silver layer, whereas the fluid entering from inlet 2 passes above the boundary layer. The fluid streams enter the main channel, where they flow attached to each other without hardly any premixing as long as the flow is maintained.

Fig. 1 Comparison of the transmittance in the mid-infrared region (wavelength 14–2.5 μm, wavenumber 700–4000 cm⁻¹) of calcium fluoride and different materials commonly used for microfluidic devices.

Fig. 2 (a) Schematic view of the whole micromixing pattern with a close up view of the microchannels. (b) Scanning electron micrographs to illustrate the sandwich construction of polymer and silver layer.

Computational fluid dynamics (CFD)

For a better understanding of the fluid flow in the mixer device and the mixing behaviour, fluid dynamic simulations were performed. The whole mixer geometry was implemented into a CFD model (Fig. 2a) using the finite volume solver FLUENT™ V5.5 to discretise the equations of fluid flow on the mesh achieved from GAMBIT™ 1.3 preprocessor. Fluid properties were set to the physical and thermodynamic properties of water at 293 K. For the simulation of the mixing behaviour, a diffusing species “A” was introduced and added to one of the feed streams at a concentration of w = 0.01. Physical and thermodynamic properties of “A” were also set to water at 293 K, and the binary diffusion coefficient of “A” in water was set to D = 1 × 10⁻⁹ m² s⁻¹.

Experimental setup

A Bruker Equinox 55 Fourier transform infrared spectrometer (Bruker GmbH, Germany) was used throughout all experiments, operated in the rapid scan acquisition mode. The scanner velocity was set to 280 kHz modulation frequency of the HeNe laser and the spectral resolution to 8 cm⁻¹ resulting in an average time difference between subsequent spectra of 65 ms. These settings were sufficient for a rough characterisation of the performance of the mixing device and are also realistic conditions for “real applications” planned in the future. However, if necessary the time resolution could be further increased either by decreasing the spectral resolution, which would lead to a reduced mirror displacement and thus to a reduced average time for one scan, yet losing spectral information, or by using the step-scan instead of the rapid-scan technique for data acquisition.The liquid handling systems consisted of a Cavro XL 3000 double-channel pump, a high-speed pneumatic switching valve and PTFE tubing with an internal diameter of 254 μm. The double-channel syringe pump for pumping the reagent solutions through the mixing disc was equipped with two 500 μl syringes that are operated with a single upstroke to ensure uniform pumping through both feeding channels of the mixer. To avoid blockage of microchannels, 0.5 μm PTFE microfilters (Upchurch Scientific, Oak Harbor, WA) were placed immediately before the mixing chip. The operation mode of the switching valve as well as the optical setup to focus the IR beam onto the main channel of the mixer can be found elsewhere.³³ Typical operating conditions for the syringe pumps were 100 μl min⁻¹.

For repeated stopped flow shots, which are necessary for time resolved FTIR measurements, the control software of the spectrometer (OPUS™) and the software to operate the pump and the valve (AnalySIA, Center for Biotechnology, Turku, Finland) were coupled via an electronic interface. A measurement sequence was performed in the following way (see Fig. 3): The pump started to aspirate the reactant solutions, switched to the outlet connection and pumped the liquid through the valve to the mixing unit. Now the FTIR spectrometer started to record 5 interferograms during the flow on, which corresponds to 5 spectra prior to reaction after Fourier transformation. Scanning was continued when a signal from the spectrometer software switches the valve to the stopped flow position which initiated diffusion based mixing of the reactants and thus the reaction under study could be directly followed by continued scanning. Thereupon the valve was switched back to the flow on position and this sequence was repeated as often as necessary, depending on the S/N ratio required. The interferograms corresponding to the same time slices during the repeated experiments were averaged and Fourier transformed to obtain high quality FTIR spectra with high time resolution.

Fig. 3 Scheme of a stopped-flow rapid scan measurement sequence

Results and discussion

Prior to the manufacture of the micromixer, fluid dynamic simulations should verify the proper geometric design of the device. Besides a short mixing time, even flow distribution and a tolerable pressure drop at typical feed flow rates are of great importance. The velocity and pressure profiles in the centre plane of the 20 μm mixing device (Fig. 4a and 4b) show an even velocity distribution across the width of the mixing channel at an overall pressure drop of 0.8 bar compared to a pressure drop of 2.3 bar achieved with the ‘spaghetti-like’ micromixer at the same feed flow rate. The local concentrations along the mixing channel during flow operation indicate a diffusion-coefficient dependent premixing which may be significant at high diffusion coefficients (Fig. 5, D = 1 × 10⁻⁹ m² s⁻¹) and larger distances from the inlet due to the residence time of the mixed fluids. The mixing time—the time after which the concentration at any point across the thickness of the mixer is within a range of ±5% of the average concentration after mixing—is a major indicator for the mixer performance. Due to premixing, the mixing time is a function of the distance from the inlet. For the analysis of the results obtained from the unsteady CFD simulation in stopped flow mode, a representative distance of d = 1.0 mm from the channel inlet (centre of FTIR beam) was chosen. Assuming a diffusion coefficient of component “A” of D = 1 × 10⁻⁹ m² s⁻¹, a mixing time of less than 100 ms can be achieved at this position (Fig. 6a and 6b), which is less than 50% of the mixing time achieved with a spaghetti-like mixer.³³

Fig. 4 (a) Velocity profile in the centre plane of the mixing device [m s⁻¹]. (b) Pressure profile in the centre plane of the mixing device [Pa].

Fig. 5 Contour plot of the mass fraction of a component A with a diffusion coefficient of D = 1 × 10⁻⁹ m² s⁻¹ across the z-axis of the mixer at different distances from the inlet (0, 0.5, 1, 1.5, 2, 2.5 and 3 mm).

Fig. 6 (a) Concentration profile of a component A with a diffusion coefficient of D = 1 × 10⁻⁹ m² s⁻¹ at different distances from the inlet (0, 0.5, 1, 1.5, 2, 2.5 and 3 mm) after stopping the flow (0, 50 and 100 ms). (b) Contour plot of the mass fraction of a component A with a diffusion coefficient of D = 1 × 10⁻⁹ m² s⁻¹ across the z-axis of the mixer at a distance from the inlet of 1 mm after stopping the flow (0, 50, 100, 150, 200 ms).

Time resolved FTIR measurements of model reactions

In order to demonstrate the feasibility of the micromachined mixing device for time resolved FTIR measurements two different model reactions were carried out. The first one is a simple neutralisation reaction of acetic acid with sodium hydroxide. This simple reaction represents an extremely fast reaction with strong spectral features in the mid-IR region. Due to its high reaction rate this reaction is proper to determine the mixing time, as any delay in the reaction originates just from diffusion (Fig. 7). From these data the following conclusions can be derived: First, the time for complete mixing of the two reactant solutions is about 100 ms, which is in good agreement with the results obtained from the numerical simulations. Secondly it is demonstrated that for very fast reactions—such as proton exchange reactions—premixing effects cannot be neglected. The strong absorption bands at 1552 and 1415 cm⁻¹, which are caused by the asymmetric and symmetric stretching vibration of the acetate ion are already developed during the flow, indicating a certain extent of prereaction. Nevertheless, the absorption bands at 1710 and 1280 cm⁻¹ corresponding to the stretch vibrations of the carbonyl and the C–O bond of the protonated acid are still present and disappear immediately after stopping the flow. As there is no mechanical separation between the two fluid streams, the formation of reaction products during the flow on is inevitable, especially with high reaction rates and small molecules with high diffusion coefficients.

Fig. 7 Stack plot of FTIR spectra obtained from a time resolved run of the reaction of acetic acid and sodium hydroxide. Time delay between subsequent spectra is 65 ms.

The second reaction presented here is the saponification of methyl monochloroacetate with sodium hydroxide (Fig. 8). In that case no formation of the reaction product is observed during the flow on. Upon stopping the flow, the reaction products—chloroacetate and methanol—occur. The reaction is completed after a few hundred milliseconds demonstrating the proper performance of the device.

Fig. 8 Stack plot of FTIR spectra obtained from a time resolved run of the reaction of methyl monochloroacetate and sodium hydroxide. Time delay between subsequent spectra is 65 ms.

In both examples the accessible spectral region extends from 950–1600 and from 1700–3000 cm⁻¹. Between 1600 and 1700 cm⁻¹ reliable measurements are made difficult to impossible because of the strong IR-absorption of the bending vibration of water which has its maximum at 1640 cm⁻¹. Whereas for many chemical systems this does not present a serious problem, this fact limits the applicability of the current device for biochemical systems because the information rich amide I band of proteins (1680–1620 cm⁻¹), from which e.g. information on the secondary structure of proteins can be derived, is masked by the water band. Using D₂O as solvent instead of H₂O would circumvent this problem by shifting the bending vibration to lower wavenumbers. However, due to proton exchange reactions and the high cost of D₂O this option is not a real solution to this problem.

Conclusion and outlook

We present a new micromachined mixing chip suitable for time resolved FTIR measurements of rapid reactions in solution with a mixing time of about 100 ms. To carry out mixing directly at the place of IR measurements a new micromachining process was developed based on a sandwich construction of epoxy polymer layers and a silver layer on infrared transmitting CaF₂ supports, which can be an interesting micromachining procedure for other microfluidic applications. The performance of the device was demonstrated with time resolved FTIR measurements of simple model reactions and confirmed by numerical fluid simulations. Computational fluid dynamic simulation proved to be a highly valuable and powerful tool to give an insight into details of microfluidic devices, especially for the prediction of the velocity distribution, the pressure drop and the mixing time and will be of high interest for future attempts as reaction kinetics can be easily implemented into the model. The developed micromixer will provide information on rapid processes in all areas of chemistry that were not accessible so far. The course of reactions as well as the occurrence of short living intermediates can be investigated. Reducing the thickness of the polymer layers below 5 μm would reduce the pathlength of the main channel to ∼10 μm, which on the one hand will reduce the mixing time by a factor of 4 and on the other hand will extend the applicability of the developed system to many biochemical systems because changes in the amide I band of proteins can then be observed also in water. Future research will therefore be directed to reduce the pathlength of the main channel to 10 μm. Applications of this methods are anticipated in the field of bio-ligand interaction studies, proteonomics as well as catalysis research.

Acknowledgement

The authors acknowledge financial support received from the Austrian Science Fund within the project P13350 ÖCH.

References

1. A. Manz, N. Grabner and H.M. Widmer, Sens. Actuators B, 1990, 1, 244.
2. J. M. Ramsey, Nat. Biotech., 1999, 17, 1061.
3. G. M. Greenway, S. J. Haswell, D. O. Morgan, V. Skelton and P. Styring, Sens. Actuators B, 2000, 63, 153.
4. W. Ehrfeld, V. Hessel and H. Lehr, Top. Curr. Chem., 1998, 194, 233 .[*] Significant reference.
5. G. H. W. Sanders and A. Manz, Trends Anal. Chem., 2000, 19, 364 .[*] Significant reference.
6. B. Lendl, R. Schindler, J. Frank, R. Kellner, J. Drott and T. Laurell, Anal. Chem., 1997, 69, 2877.
7. A. van den Berg and T. S. J. Lammerink, Top. Curr. Chem., 1998, 194, 21 .[*] Significant reference.
8. D. J. Harrison, K. Fluri, K. Seiler, Z. Fan, C. S. Effenhauser and A. Manz, Science, 1993, 261, 895.
9. S. C. Jacobson, R. Hergendröder, A. W. Moore Jr. and J. M. Ramsey, Anal. Chem., 1994, 66, 4127.
10. C. S. Effenhauser, G. J. M. Bruin and A. Paulus, Electrophoresis, 1997, 18, 2203.
11. F. G. Bessoth, A. J. de Mello and A. Manz, Anal. Commun., 1999, 36, 213.
12. M. Koch, D. Chatelain, A. G. R. Evans and A. Brunnschweiler, J. Micromech. Microeng., 1998, 8, 123.
13. M. Koch, C. G. J. Schabmueller, A. G. R. Evans and A. Brunnschweiler, Sens. Actuators A, 1999, 74, 207.
14. V. Hessel, W. Ehrfeld, H. Möbius and T. Richter, Proceedings of the International Conference on Solid-State Sensors and Actuators, Stockholm, Sweden, 1995.
15. W. Ehrfeld, K. Golbig, V. Hessel, H. Löwe and T. Richter, Ind. Eng. Chem. Res., 1999, 38, 1075.
16. B. He, B. J. Burke, X. Zhang, R. Zhang and F. E. Regnier, Anal. Chem., 2001, 73, 1942.
17. C. Ehrbacher, F. G. Bessoth, M. Busch, E. Verpoorte and A. Manz, Mikrochim. Acta, 1999, 131, 19.
18. J. B. Knight, A. Vishwanath, J. P. Brody and R. H. Austin, Phys. Rev. Lett., 1998, 80, 3863.
19. D. Bökenkamp, A. Desai, X. Yang, Y. Tai, E. M. Marzluff and S. L. Mayo, Anal. Chem., 1998, 70, 232.
20. A. E. Kamholz, B. H. Weigl, B. A. Finlayson and P. Yager, Anal. Chem., 1999, 71, 5340.
21. D. Gobby, P. Angeli and A. Gavriilidis, J. Micromech. Microeng., 2000, 11, 126.
22. S. C. Jacobson, T. E. McKnight and J. M. Ramsey, Anal. Chem., 1999, 71, 4455.
23. T. McCreedy, Trends Anal. Chem., 2000, 19, 369.
24. R. Rammelsberg, B. Heßling, H. Chrongiewski and K. Gerwert, Appl. Spectrosc., 1997, 51, 558.
25. G. D. Smith, P. Chen, J. L. Chao, K. M. Omberg, D. A. Kavaliunas, J. A. Treadway, T. J. Meyer and R. A. Palmer, Laser Chem., 1999, 19, 291.
26. C. Rodig, I. Chizhov, O. Weidlich and F. Siebert, Biophys. J., 1999, 76, 2687.
27. H. Wang, R. A. Palmer and C. J. Manning, Appl. Spectrosc., 1997, 51, 1245.
28. B. Hinterstoisser and L. Salmen, Vib. Spectrosc., 2000, 22, 111.
29. B. C. Dunn and E. M. Eyring, Appl. Spectrosc., 1999, 53, 292.
30. A. J. White, K. Drabble and C. W. Wharton, Biochem. J., 1995, 306, 843 .[*] Significant reference.
31. A. Troullier, D. Reinstädler, Y. Dupont, D. Naumann and V. Forge, Nat. Struct. Biol., 2000, 7, 78.
32. N. S. Marinkovic, A. R. Adzic, M. Sullivan, K. Kovacs, L. M. Miller, D. L. Rousseau, S.-R. Yeh and M. R. Chance, Rev. Sci. Instrum., 2000, 71, 4057.
33. P. Hinsmann, M. Haberkorn, J. Frank, P. Svasek, M. Harasek and B. Lendl, Appl. Spectrosc., 2001, 55, 241 .[*] Significant reference.
34. J. A. M. Kuipers and W. P. M. Van Swaaij, Rev. Chem. Eng., 1997, 13, 1.
35. V. V. Ranade, Rev. Chem. Eng., 1995, 11, 229.
36. S. A. Orszag and I. Staroselsky, Comput. Phys. Commun., 2000, 127, 165.
37. I.-M. Hsing, R. Srinivasan, M. P. Harold, K. F. Jensen and M. A. Schmidt, International Conference on Solid-State Sensors and Actuators, Proceedings, 1997, vol. 2, pp. 1015–1018.

---