Microdevice-Based System for Catalyst Development

Transcription

1 Microdevice-Based System for Catalyst Development R.S. Besser Chemical Engineering and Institute for Micromanufacturing Louisiana Tech University Ruston, LA (318) (240) (fax) Abstract Catalytic chemical reaction processes have played a major role in the growth of today s international economy. Because of this impact, technologies that reduce the development time of new catalyst materials are highly sought. We describe a new experimental system for the characterization of candidate catalysts that exploits the many advantages of microfabricated chemical devices. The core component of the system is a silicon microreactor in which the reaction of study is executed under highly controlled conditions. The system was used to acquire data on a model hydrocarbon hydrogenation reaction over platinum catalyst. The measured values of turnover frequency and reaction probability compare well with the literature and illustrate the value of the system as a tool for rapid and efficient catalyst development. Keywords catalysis, catalyst, heterogeneous, microreactor, microsystem, microfluidic 1

2 Introduction For the past century and a half, catalytic materials have played a critical role in the tremendous growth in the worldwide economy by enabling the low-cost large-scale production of a multitude of valuable chemical products. For example, the proliferation of the internal combustion engine in all modes of transportation would not have occurred without the simultaneous expansion of the petroleum refining industry that depends heavily on catalytic chemical processes. Catalysts have had similar large impact in textiles, plastics, agricultural chemicals, pharmaceuticals, food, and elsewhere. As a result of this impact, new catalysts and improved versions of existing catalysts are continually being sought. These discoveries will lead to the development of chemical processes with even greater efficiency in producing critical products as well as making possible the widespread production of important chemicals currently considered prohibitively expensive. The economic impact of catalyst technology is evident from the leverage that catalysts exert on the chemical process industry. The total annual investment in commercial catalysts by chemical producers results in revenues from catalytic processes far exceeding this investment. This factor was recently estimated in the U.S. to be sixty or so [1]. Because of this leverage, even incremental improvements in catalyst efficiency will result in substantial increases in profitability as a given catalyst investment will produce greater revenues through increased yield, less waste, and lower energy requirements. Catalyst science has moved from a purely empirical to a more theory-based discipline, especially over the past three decades or so [2]. Now, models based on molecular-dynamics and other computational methods are aiding prediction and preselection of candidate catalysts [3]. Combinatorial methods have also been adapted in an attempt to accelerate catalyst screening, however, efforts to date have mostly focused on schemes that assess catalytic activity without directly measuring conversion [4, 5]. The most heavily relied upon method of catalyst assessment today still consists of measuring reactant conversion under conditions closely matching those expected in the ultimate industrial process system. 2

3 Traditionally, catalyst screening has been performed in small laboratory reactors. For gas-solid catalysis, these reactors typically consist of a tube packed with a catalyst sample and operated as a plug flow reactor. The reactor is interfaced to suitable equipment for control of temperature, pressure, and temperature. Conversion and selectivity are monitored with analytical instruments. These experimental setups typically require moderate laboratory space and require support personnel for their setup and monitoring during an experiment. In addition to time for setup, adequate time is needed for stabilization whenever reaction conditions are modified. It is not uncommon for this stabilization time period to last several hours, and often only a single set of experimental conditions can be explored per day. As a result, the experimental evaluation of a single catalyst may require a week or more. The desired characteristics of laboratory reactors were defined in a frequently cited paper by Weekman over twenty-five years ago [6]. These characteristics relate to the quality and applicability of the data obtained by using the reactor, as well as the relative convenience and cost of operation. It was a goal of this work to construct a laboratory reaction system that would possess these attributes as applied to the screening of candidate catalysts and for taking kinetic data useful for the large-scale design of reactors utilizing these catalysts. We determined to incorporate a chemical microdevice for the reactor in this system in order to exploit the many advantages of microscale systems. The advantages of chemical microreactors have been well documented in the literature [7]. Included in these advantages and of particular interest to this project are the following: 1. High surflace-to-volume ratio that suppresses heterogeneous, free-radical chain reactions in favor of the heterogeneous reactions under study. The large surface area relative to volume also facilitates thermal management. 2. Precise control over geometry of the fluidic structures on the device resulting in a well-defined and narrow distribution of residence time. 3. Fast equilibration due to minuscule heat and mass transfer resistances that derive from microscale geometry. 3

4 4. Materials of construction that are robust mechanically and chemically under harsh conditions of elevated temperature and pressure. These materials can be manipulated and built into practical devices with a well-known and versatile set of fabrication processes. We chose to adopt conventional silicon micromachining procedures for these reasons. 5. Compatibility with a variety of catalyst formation processes including physical and chemical vapor deposition, sol-gel deposition, solution impregnation, and nanoparticle layered assembly. 6. Low consumption of reactants. 7. Inherent safety of operation due to minimal likelihood of explosion and reduced hazard of toxic leaks. Experimental Several system design parameters were defined at the outset in order to be successful in fabricating a microreactor for gas-solid catalytic process evaluation. Maximum temperature of operation was set at 250 C as the goal for an initial design, with 500 C as the ultimate goal. Likewise, the initial pressure specification was taken as 5 atm, with an ultimate goal of 30 atm. Other factors affecting the design were the need for catalysts to be flexibly and rapidly deposited to facilitate quick experiment turnaround. A further consideration was for the microreactor to be based on a standard chip footprint to allow future designs to take advantage of the peripheral equipment setup constructed for the initial system. The outcome of the design process was a simple microreactor module consisting of a bulk micromachined silicon chip with a glass cover (Figure 1). The chip is 1 x 3 cm 2 in size and has a reaction zone consisting of a number of parallel microchannels coated with catalyst. Devices with 100-µm and 5-µm channel widths have been fabricated. Gasphase reactant chemicals are introduced from the bottom side of the chip, passing upward through vias that connect to access channels leading to the reaction zone. Similar vias take the reaction products out the bottom side of the exit end of the chip where they can 4

5 be analyzed. The layout can accommodate up to three separate feeds and two separate product streams. The heat transfer characteristics of the microreactor were assessed with a simple thermal model. A two-dimensional finite difference approach was taken to obtain a solution to the steady-state heat equation with suitable boundary conditions. The effect of enthalpy of reaction was investigated by assuming that the heat generated thereby was dissipated uniformly around the perimeter of the microchannel cross-section and ignoring advection. We found that the reactor can manage very large heats of reaction without appreciable temperature increase. The graph in Figure 2 shows that even exothermic heats as high as 5000 kcal/mole result in a manageable temperature of only tens of degrees or less above the temperature of the heat sink below the device. The microreactor was sized to accommodate a range of reaction probability (RP). The RP is defined as the probability that an impinging reactant molecule will undergo transformation on any given collision with the catalyst surface and varies between 10-5 and or so for many classes of hydrocarbon reactions. RP values for reactions and catalysts of importance are documented in many literature sources [2]. A simple model for a continuous flow reactor taking into account geometry, temperature, pressure, and flow rate was used to determine conversion as a function of reaction probability. The result of this analysis for typical flow conditions and reactor geometry is shown in Figure 3. We took 10% as an arbitrary level of conversion at which catalysts can be reasonably compared. From the Figure, we see that the 1.9 cm reaction zone length supports 10% conversion for reactions with RP as small as and larger. The requirement that the microreactor be fabricated with few steps and high yield led to the selection of well-known bulk micromachining processes [8,9]. The process steps shown in Figure 4 are based on simple photolithography and etching of a silicon dioxide layer that acts as a mask during subsequent KOH etching of silicon [10]. Fourinch silicon wafers of (110) orientation were used in order to produce vertical microchannel sidewalls as illustrated in Figure 5. The device photomask set incorporates several options to allow for flexibility in fabrication. As few as two masking steps can be used to complete the silicon process, however, greater control over the depth of 5

6 individual fluidic structures on the device can be gained by using up to four separate masking steps. Catalyst films are deposited in the microreactors after the fluidic passages are formed in the silicon. This step may be completed on all devices on the wafer simultaneously (eighteen microreactors per wafer), or on individual chips after sawing the wafer. For the model reaction study described later in this paper, a Pt film 200 Å thick was deposited by DC magnetron sputtering. The final step of fabrication is sealing the chip with a glass cover. Anodic bonding was used to bond the Pyrex glass cover to the silicon. The resulting seal was found to be hermetic under all conditions tested (temperature and pressure as high as 300º C and 100 psig). The microreactor was interfaced to an experimental setup for controlling temperature, introducing and removing reactants and products, and analyzing the composition of the gas streams. A diagram of the setup is shown in Figure 6. The block supporting the microreactor was custom fabricated. It is fitted with cartridge heaters and thermocouples for temperature control. The gas passages in the block seal directly to the bottom side of the silicon chip using 2-mm silicone o-rings. The flow controllers and pressure gauges are commercially available components. Chemical analysis was performed by an Extrel C-50 research grade mass spectrometer housed in a custom-built vacuum chamber. A Varian Saturn gas chromatograph has been connected but was not used in the experiments described in this paper. We conducted conversion experiments in order to assess the usefulness of the system for determining catalytic activity. The hydrogenation of cyclohexene (C 6 H 10 ) over platinum catalyst was selected as a model reaction for these experiments. This reaction is representative of the important class of hydrocarbon hydrogenation reactions ubiquitous in the chemical and petroleum processing industries. In the experiment described here, C 6 H 10 vapor was introduced into the reactor by bubbling argon through the liquid hydrocarbon. The reaction conditions were the following: 0.1 sccm of hydrogen, 0.1 sccm of argon saturated with C 6 H 10, 1 atm pressure, and temperature settings applied in the order 473 K, 523 K, and 423 K. Temperature was held constant for a period of six hours before adjusting to a new setting. 6

7 A typical mass spectrum taken during the experiment is shown in Figure 7. The spectrum indicates the presence of the cyclohexene reactant, and cyclohexane (C 6 H 12 ), the hydrogenation product. Also visible in the spectrum is evidence for the dehydrogenation product, benzene (C 6 H 6 ). Table I shows the measured conversion during the temperature cycles. We observe a monotonic decrease in conversion with increasing time on stream, irrespective of temperature. It was suspected that this dominant effect is due to catalyst deactivation. To examine whether surface chemical changes led to deactivation of the catalyst, we removed the pyrex cover and performed x-ray photoelectron spectroscopy (XPS) analysis on the reaction zone area. Figure 8 shows the results of XPS analysis combined with sputter depth profiling of the sample surface. The plot reveals that the platinum signal is absent upon initial survey of the surface. The high carbon signal in the nearsurface depth range indicates the presence of a carbonaceous overlayer. The formation of this layer is consistent with a decrease in catalytic activity due to the elimination of active catalytic surface available to participate in the reaction. This phenomenon (known as coking ) is well known in catalytic hydrocarbon reactions [11]. The measured conversion data permit the determination of catalytic parameters that can be compared to literature values. Figure 9 is a plot of conversion vs. RP for the geometry and reaction conditions of the experiment. We consider the conversion for the 473 K case since conversion was as yet relatively unaffected by catalyst deactivation. The RP responsible for this conversion level was determined to be This result was then used in conjunction with the density of active catalytic sites to calculate turnover frequency (TOF). The site density was estimated for the sputtered platinum layer by assuming a perfectly planar morphology and taking every surface atom as an active site. The TOF calculated thereby was 1.7 molecules/site/s. These values of RP and TOF compare well with literature values for hydrogenation reactions [12]. Discussion The above results demonstrate the potential of the experimental system as a tool for rapid catalyst development. An important aspect of the demonstration is that data 7

8 were obtained under conditions approximating an industrial reaction. Data quality was ensured by adequate levels of single-pass conversion and throughput. The catalytic reaction parameters obtained (RP, TOF) allow catalyst candidates to be confidently compared with literature values, or with one another in screening experiments. A second important attribute is the ability to conduct experiments rapidly. We used six-hour cycles in this experiments to make certain we had achieved steady-state operation. The data taken in the experiment made it clear that the cycle length could be reduced without loss of quality. Because of the fast heat and mass transfer kinetics, the limiting factor in turnaround time is the ability to fill the system volume with chemical streams at low flow rates. We are presently working to eliminate excess volume in the system to verify that turnaround times can be reduced to only minutes per data point. Ultra-low consumption of reactants was verified in the above experiment. In the 18 hour time period of the experiment, a total of less then 32 mg of cyclohexene was consumed. This low usage means that the cost of reactants is not a limitation in these kinds of experiments. Studies on expensive fine or pharmaceutical reactants will be significantly more economical with other methods. Moreover, costly isotopically marked chemicals can be used routinely. The low volume of chemical usage also implies the inherent safety of the system from the standpoint of explosion or toxicity hazard. Potential negative impact on the environment is similarly minimal. Another useful attribute of the system is the ability to rapidly perform surface analysis on the catalyst for correlation with reaction results. The XPS data above support the conclusion that deactivation was due to the formation of a carbonaceaous overlayer that had deposited on the catalyst. The rapidity of obtaining the spectroscopic result was due to the ability to easily remove the microreactor from the experimental system and to quickly remove the cover. The intial surface spectroscopic information was available within minutes of completing the reaction experiment. This aspect of rapid post-reaction characterization applies to other analyses as well, for example BET surface area measurement. 8

9 Conclusions In summary, we have demonstrated a flexible, microsystem-based test platform for heterogeneous catalysis development. The design, modeling, and fabrication approaches used for the silicon micromachined reactor chip led to a useful module that was integrated with analysis and flow control equipment to form the complete system. The model hydrogenation reaction that we characterized yielded catalytic reaction parameters that compare well with literature values. The system was also useful for gaining understanding of the catalyst surface after the reaction as the entire reactor was able to be quickly loaded into a surface analysis system for spectroscopy. The experimental reaction system was shown to have faster turnaround, less chemical usage, and greater safety than conventional laboratory systems. Acknowledgements The author gratefully acknowledges the support of Jonathon Fort, Sean Ouyang, Michelle Prevot, Harshal Surangalikar, Wanjun Yang, and Shihuai Zhao. We also thank Scott Williams and Shawn Moncrief for their help with metal fabrication and Dr. Mike Vasile for his assistance with the mass spectrometer. We also acknowledge many useful discussions with Dr. Frank Jones of the University of Tennessee. This work was performed under the support of the Louisiana Board of Regents, LEQSF-RD-A-20. References 1. J.D. Hewes, L.P. Herring, M.A. Schen, B. Cuthill, and R. Sienkiewicz, Combinatorial Chemistry-The Discovery of Catalysts and New Materials, ATP Position Paper, Advanced Technology Program, National Institute of Standards and Technology, Gaithersburg, MD, G.A. Somorjai, Introduction to Surface Chemistry and Catalysis, (Wiley, New York, 1994), 1 st ed., Chap. 7, pp

10 3. L. Deng, T. Ziegler, and L. Fan, Computer Design of Living Olefin Polymerization Catalysts: A Combined Density Functional Theory and Molecular Mechanics Study, Organometallics, 17 (15), 3240 (1998). 4. F.C. Moates, M. Somani, J. Annamalai, J.T. Richardson, D. Luss, and R.C. Willson, Infrared Thermographic Screening of Combinatorial Libraries of Heterogeneous Catalysts, Ind. Eng. Chem. Res., 35 (12), (1996). 5. S.M. Senkan and S. Ozturk, Discovery and Optimization of Heterogeneous Catalysts Using Combinatorial Chemistry, Angew. Chem. Int. Ed., 38, 791 (1999). 6. V.W. Weekman, Laboratory Reactors and Their Limitations, AIChE Journal, 20, 833 (1074). 7. For an exhaustive coverage of the subject, see W. Ehrfeld, V. Hessel, and H. Lowe, Microreactors: New Technology for Modern Chemistry, Wiley, New York, 2000, and references contained therein. 8. M. Prevot and R.S. Besser, Fabrication Process for a Microreaction Device, Abstracts of the Materials Research Society Spring Meeting, Materials Research Society, p. 491, R.S. Besser and M. Prevot, Linear Scale-Up of Micro-Reaction Systems, Topical Conference Proceedings of the Fourth International Conference on Microreaction Technology, American Institute of Chemical Engineers, pp , For a review on the subject, see M. Madou, Wet Bulk Micromachining, Fundamentals of Microfabrication, (CRC Press, New York, 1997), 1 st ed., Chap. 4, and references contained therein. 10

11 11. Reference [2], p G. A. Somorjai and J. Carrazz, Structure Sensitivity of Catalytic Reactions, Industrial Engineering Chemistry Fundamentals, 25, 63 (1986). 11

12 Figure and Table Captions Figure 1: Photos of microreactor chip sealed with Pyrex cover and microreactors on 100-mm silicon wafer. Figure 2: (A) Results of solving two-dimensional steady-state heat equation for reaction zone cross-section, assuming a supporting baseplate temperature of 500 K. (B) Subsection of reaction zone area analyzed in model. (C) Temperature profile obtained for the 220 kcal/mole case. Figure 3: Results of conversion model of microreactor. The 1.9-cm reaction zone will support 10% conversion at reaction probabilities as small as Figure 4: Fabrication process for the silicon microreactor chip. The process consists of sequential lithographic steps to form patterns in SiO 2 layers that act as masks for KOH etching of silicon. In the final step, a Pyrex cover is bonded to the silicon chip. Figure 5: Scanning electron micrographs of microchannel reaction zone in the microreactor chip. Figure 6: Schematic of entire experimental catalyst characterization system. Figure 7: Typical mass spectrum from cyclohexene hydrogenation experiment. Signature peaks of various species are labeled. Figure 8: X-ray photoelectron spectroscopy sputter depth profile of deactivated platinum catalyst film. 12

13 Figure 9: Conversion measurement vs. concomitant reaction probability at 473 K and the other conditions of the experiment. Table I: Measured reactant conversion during the progress of the cyclohexene hydrogenation experiment. 13

14 Figure 1

15 Temperature Rise (K) Glass Silicon Exothermic Heat of Reaction (kcal/mole) Figure 2 A

16 700 µm 500 µm 200 µm Figure 2 B

17 Figure 2 C

18 3 2.5 Reactor Length for 10% Conversion (cm) Log(Reaction Probability) Figure 3

19 Figure 4

20 Figure 5

21 Figure 6

22 Counts m/z Figure 7

23 70 60 Atomic Concentration (%) O-1s Si-2p C-1s 10 Pt-4f Sputter Time (min.) Figure 8

24 Conversion (%) Log 10 (Reaction Probability) Figure 9

25 Cycle Temperature (K) Time (h) Conversion (%) Table 1