Letters to Nature

Vol. 394, pp. 350-353, 1998

High throughput screening of combinatorial catalyst libraries

Selim M. Senkan

Department of Chemical Engineering

University of California

Los Angeles, CA 90095-1592

Tel: (310) 206-4106; Fax: (310) 206-4107; e.mail: senkan@seas.ucla.edu

Abstract: One of the barriers that impede the use of combinatorial methods for the discovery of catalysts is the availability of high-throughput, sensitive and selective library screening techniques. Here we report the first in situ method for the simultaneous determination of the activities and selectivities of all the sites on a large catalyst library. The technique involves the selective photoionization of reaction products over spatially addressable catalyst cluster rows using tunable UV lasers under resonance-enhanced multiphoton ionization (REMPI) conditions followed by the detection of photoelectrons or photoions by an array of microelectrodes. The feasibility of the method has been demonstrated to screen a 72 site (8x9) solid-state library for the catalytic dehydrogenation of cyclohexane to benzene. Each site was about 5 mm by 5 mm, corresponding to a library density of 10 sites per square inch. Further miniaturization and scale-up of the technique is possible using monolithic or honeycomb structures. The ability to rapidly screen combinatorial libraries for catalytic activity, coupled with empirical knowledge and theory of catalysis, could significantly increase the rate at which new and improved catalysts can be developed.

Catalysts are used in the manufacture of a vast array of chemicals and fuels, and as such significantly contribute to our economy and high living standards (1). However, in spite of their significance and broad utility, the development of new and improved catalysts continues to be an arduous and rather unpredictable trial-and-error process (2,3). Combinatorial or diversity synthesis methods, in which large libraries of catalysts are rapidly prepared, processed and screened for activity and selectivity, can significantly accelerate the process of catalyst discovery and/or optimization. To date, thin-film and solution-based combinatorial synthesis methods have been used for the discovery of superconducting (4), magnetoresistive (5) and luminescent (6-8) materials. These and related techniques provide a good control of chemistry and are well-suited to generate catalytic materials.

We are not aware of any prior work in which combinatorial techniques have been applied for the discovery of new and/or improved catalysts. One of the impediments for this has been the availability of sensitive, selective and high-throughput screening techniques. Unlike superconductivity and magneto-resitivity, both of which can be tested by contact probes, or luminescence that can be tested by light emission, catalyst screening requires the unambiguous detection of a specific molecule (i.e., a product) above a small catalyst site on a large library. To date two methods have been proposed to screen catalyst libraries: 1. In situ infrared (IR) thermography (9) and 2. Microprobe sampling mass spectrometry (MS, 10,11). The IR method, although it provides data on activity, generally provides no information on products. Microprobe MS requires the physical removal and transport of very small quantities of gas samples from each site. In addition, the sampling probe must physically be moved from site to site. As a consequence of these issues, IR and MS are of marginal utility to screen large catalyst libraries.

Here we report a new method to rapidly screen entire catalyst libraries for activity and selectivity without the need to remove samples. The method developed exploits tunable lasers for the selective photoionization of product molecules in the vicinity of catalytic sites, followed by the detection of the resulting photoions or photoelectrons by an array of microelectrodes. When the laser frequency is tuned to a real intermediate electronic state of a molecule, the cross section for ionization is significantly enhanced. This is called resonance-enhanced multiphoton ionization (REMPI, 12). If the laser is not tuned to a real electronic state, the probability for photoionization is smaller. Consequently, a desired reaction product can be selectively ionized with high efficiency while avoiding the simultaneous photoionization of the reactants, other products and background gases.

The most common REMPI approach is resonant 2-photon ionization (R2PI) in which one photon (h1) energizes the molecule to an excited electronic state and the second photon (h2) ionizes the molecule:

P + h1 = P* (A)

P* + h2 = P+ + e- (B)

where P* is a real electronic excited state, P+ is the photoion and e- is the photoelectron. The absorption of two or more photons in each step may also be utilized. Ionization occurs if (h1+h2)>IP, where IP is the ionization potential. By varying the photon energies, the ionization spectrum of P (as well as other molecules of interest) can be mapped out to determine suitable laser frequencies for its exclusive ionization. The two photons used can have the same or different frequencies. The REMPI process can also be repeated to sequentially detect different products (using different laser frequencies) for the determination of product selectivities. REMPI is a sensitive technique with real-time detection at low parts per billion (13) and high parts per trillion (14) already demonstrated. However, e ach application of REMPI requires separate analysis building upon the spectral properties of molecules that are either known or determined in separate experiments.

In Figure 1 a sketch of the overall system developed to screen catalyst libraries is presented. In Figure 2, the details of the reactant and catalyst contact, and the relative positions of the laser beam, the microelectrodes and the catalyst sites are shown. REMPI signals are detected by the use of dedicated microelectrodes placed above each site.

Figure 1. The screening system consists of a reactor chamber, a temperature controlled furnace and the tunable UV laser source. The library consisted of a 7.5 cm square ceramic substrate containing 72 sites (8 by 9 rows). Each site was 5 mm by 5 mm and separated 2 mm from one another, corresponding to a library density of about 10 sites per square inch. The reactor was designed to gently introduce the reactant gases in and to remove the products out of the vessel. Photoionization was achieved by passing a pulsed, tunable UV laser beam over all the sites in a row as indicated. Suitable laser light frequencies were obtained from a dye laser (Laser Analytical Systems, LDL 2051) that was pumped by an excimer laser (Lambda Physik, Compex 102). BBO I crystal was used for second harmonic generation. Optical access to the library was provided through MgF2 windows placed at the side walls of the reactor.

Figure 2. The reactant gases are forced through individual sites on the library to generate product plumes, if the sites are catalytic. The laser beam is then passed over the airspaces of all the sites, intercepting the plumes, thereby generating the product photoions and photoelectrons. Dedicated microelectrodes placed above the sites in the close vicinity of the laser beam simultaneously collect either the photoions or photoelectrons depending on the DC bias voltage applied. The use of other library geometries, such as those based on monolithic or honeycomb structures with well defined channels are also feasible. Microelectrodes were 0.5 mm diameter and 3 cm long and possessed both the anode and the cathode (15). They were mounted through a ceramic plate that was attached to the top of the reactor chamber. A multichannel switch was used to individually apply DC power (Stanford Research Systems, PS 350) to each microelectrode and to screen the sites. Signal detection was accomplished by a digital oscilloscope (Hewlett-Packard 54540C).

The technical feasibility of the screening method was demonstrated by considering the catalytic dehydrogenation of cyclohexane into benzene:

C6H12 ----> C6H6 + 3H2 (C)

This is an established reaction catalyzed by transition and precious metals in the temperature range 250-350oC (16,17). Pt and Rh catalysts were acquired from a commercial vendor (Precious Metals Corp., TN). Catalysts were crushed into 0.2-0.5 mm pieces, and about 40 mg batches were manually placed to the fifth row of the library as shown in Figures 1 and 2. The catalyst library was then placed into the reactor and the entire system was heated to 300oC in the presence of argon gas flow. The feed stream composition was then changed to 13% cyclohexane in argon carrier gas, prepared by bubbling argon through cyclohexane at about 25oC.

The screening protocol demands the unambiguous detection of benzene in cyclohexane, hydrogen and argon matrix. In order to accomplish this goal one must identify a suitable UV laser wavelength that selectively produces benzene REMPI ions. This wavelength was identified using a time-of-flight mass spectrometer (TOF-MS). In Figure 3a the TOF-MS photo-ionization spectra of benzene and cyclohexane are presented.



Figure 3a. Cyclohexane and benzene, each at a concentration of about 500 ppm in argon, were introduced into the vacuum chamber of the TOF-MS using a pulsed valve, and the resulting jet/molecular beam was crossed by a pulsed UV laser light in the 258-262 nm range to generate their photoionization spectra. The UV light had about 100 µJ/pulse energy over the entire wavelength range and was obtained by frequency doubling the beam from the dye laser (see Figure 1). These measurements led to the conclusion that the REMPI ions produced by 258-262 nm light were exclusively due to the photoionization of benzene (mass 78), with no photoions detected at masses 84 (cyclohexane), 40 (argon), or 2 (hydrogen). Furthermore, no peaks other than the benzene parent at mass 78 were detected.

As evident from Figure 3a, benzene exhibits major REMPI peaks starting at 259.7 nm, and there are no contributions from cyclohexane. Following the TOF-MS experiments, REMPI spectra were also determined at ambient conditions using the microelectrode detection system. As seen in Figure 3b, the REMPI spectra obtained by the microelectrode technique were similar to the TOF-MS spectra, with the expected broadening in room temperature experiments. Figure 3b also illustrates that a significant broadening of the REMPI spectra can be tolerated in catalyst screening when the REMPI features of the reactants and products are well separated.

Figure 3b. The REMPI spectra obtained under ambient conditions using microelectrode detection. Experiments were done by flowing cyclohexane or benzene bearing argon gas over the microelectrode and by photoionizing the gases within 1-2 mm of the probe tip using the same laser source as in the TOF-MS experiments. A DC bias of +500 V was applied to the anode to collect the photoelectrons.

In Figure 4, the results of the catalyst library screening process using 259.7 nm light are presented for the 5th row. These measurements correspond to data acquired by one laser shot and do not involve signal averaging. The signals exhibited decay times on the order of microseconds, suggesting the feasibility to screen large libraries. As seen in Figure 4, microelectrodes located at sites 2,4,7,8 picked up appreciable benzene REMPI signals, consistent with the presence of Pt and Pd catalysts at these sites.

It is also important to note that while some benzene REMPI signals were detected at sites 1,3,5,6, they were significantly lower, consistent with the absence of catalysts at these sites. Evidently some benzene remained in the reactor because of low gas flow rates used and recirculation patterns present, both of which prevent the rapid removal of products from the reactor. A smaller chamber design and higher gas flow rates, or the use of monolithic or honeycomb channel structures should minimize this problem. Nevertheless, the technique clearly distinguished between the active and inactive sites on the catalyst library. The reactor

Figure 4. Screening results of the 5th row in the catalyst library. The addresses of the various catalysts used in the experiments were: Site 1: Inert; Site 2: 0.5% Pt; Site 3: Inert; Site 4:1.0% Pd; Site 5: Inert; Site 6: Inert; Site 7: 1.0%Pt; Site 8:0.5% Pd.

exhaust gases were also analyzed by the TOF-MS using the 259.7 nm laser light to explore if species other than benzene could have contributed to the measured signals. No photoions, other than those with mass 78, were detected. Photoionization MS analysis of the reactor exhaust gases by the same laser used in the screening process insures that microelectrode measurements are exclusively due to desired product and not due to the formation of another by-product catalyzed by the library.

Based on the magnitudes of the REMPI signals measured (Figure 4), the relative activities of the catalytic sites appear to be 7>2>4>8. Although fortuitous, as the microelectrodes were not calibrated, these results are consistent with the relative loadings of the Pd and Pt in catalysts used. In addition, they also suggest that Pt is a more active dehydrogenation catalyst than Pd, in good agreement with the results of earlier studies using conventional catalytic reactor systems (16,17).

As evident from the foregoing discussion, quantitative data on the activities of each site in the catalyst library can be obtained, provided each microelectrode system is calibrated. This calibration can be done electronically to produce the same signal intensity when using a calibration gas mixture (or mixtures) in the absence of the catalyst, i.e., no reaction. During the actual library screening process, the signals measured will be proportional to the activities of the catalyst sites, after correcting for background contributions. The latter can easily be done from signal measurements at known non-catalytic sites. Clearly, differences in catalyst loading must also be taken into account in assessing the relative activities of the catalytic sites.

It is significant to note that the technique developed can readily be extended to monitor multiple reaction products, and thus can be used to acquire information on catalyst selectivity as well. This can be accomplished by using different laser frequencies to sequentially generate the REMPI signals of different products. The REMPI signals can then be converted into absolute concentrations, using calibration standards, for the determination of selectivities. In addition, the technique developed should be useful to study issues related to operational lifetimes, resistance to poisoning, regeneration and loss of catalysts.

References:

1. National Research Council, "Catalysis Looks to the Future", National Academy Press, Washington, DC 1992.

2. Poncelet, G., Martens,J., Delmon,B., Jacobs,P.A., Grange,P., "Preparation of Catalysts VI", Elsevier, Amsterdam, The Netherlands 1995.

3. Heinemann, H., "A Brief History of Industrial Catalysis", in "Catalysis: Science and Technology", J.R. Anderson and M. Boudart, Eds., Chapter 1, Springer-Verlag, Berlin 1981; Satterfield, C.N., "Heterogeneous Catalysis in Practice", 2nd Ed., McGraw Hill, New York 1991.

4. Xiang,X-D., Sun,X., Briceno,G., Lou,Y., Wang,K-A., Chang,H., Wallace-Freedman,W.G., Chen,S-W., Schultz,P.G., "A Combinatorial Approach to Materials Discovery", Science, 268, 1738-1740 (1995).

5. Briceno,G., Chang,H., Sun,X., Schultz, P.G., Xiang,X-D.,"A Class of Cobalt Oxide Magnetoresistance Materials Discovered with Combinatorial Synthesis", Science, 270, 273-275 (1995).

6. Sun,X-D, Wang, K-A., Yoo,Y., Wallace-Freedman,W.G., Gao,C., Xiang,X-D., Schultz,P.G., "Solution-Phase Synthesis of Luminescent Materials Libraries", Adv. Mater., 9, 1046-1049 (1997).

7. Sun,X-D., Gao,C., Wang,J., Xiang,X-D., "Identification and Optimization of Advanced Phosphors Using Combinatorial Libraries", App. Phys. Lett., 70, 3353-3355 (1997).

8. Danielson,E., Golden,J.H., McFarland,E.W., Reaves,C.M., Weinberg, W.H., Wu,X-D., "A Combinatorial Approach to the Discovery and Optimization of Luminescent Materials", Nature, 389, 944-948 (1997).

9. Moates,F.C., Somani,M., Annamalai,J., Richardson,J.T., Luss,D., Willson,R.C., "Infrared Thermographic Screening of Combinatorial Libraries of Heterogeneous Catalysts", Ind. Eng. Chem. Res., 35, 4801-4803 (1996).

10.Kassem, M., Senkan, S.M.,"Chemical Structure of fuel-Rich 1,2-C2H4Cl2/CH4/O2/Ar Flames: Effects of Microprobe Cooling on the Sampling of Flames of Chlorinated Hydrocarbons", Combust. Sci. Tech., 67, 147-157 (1989).

11. Service, R.F., "High-Speed Materials Design", Science, 277, 474-475 (1997).

12. Lubman,D., Li, L.,"Resonant Two-Photon Ionization Spectroscopy of Biological Molecules in Supersonic Jets Volatilized by Pulsed Laser Desportion", in Lasers and Mass Spectrometry, D.M. Lubman, Ed., Oxford Univ. Press, New York 1990.

13. Gittins,C.M., Castaldi,M.J., Senkan,S.M., Rohlfing,E.A,"Real-Time Quantitative Analysis of Combustion Generated Polycyclic Aromatic Hydrocarbons by Resonance Enhanced Multiphoton Ionization Time of Flight Mass Spectrometry", Anal. Chem., 69, 286-293 (1997).

14. Senkan, S.M., Castaldi,M.J., "Real-time, Ultrasensitive Monitoring of Air Toxics by Laser Photoionization Time of Flight Mass Spectrometry", J. Air and Waste Mgmgt. Assoc., 48, 77-81 (1998).

15. Cool,T.A., "Quantitative measurement of NO density by resonance three-photon ionization", App. Optics, 23, 1559-1572 (1984).

16. Rehbon,D.M., and Haensel,V.,"A Kinetic and Mechanistic Study of Cyclohexane Disproportionation: An Example of Irreversible Hydrogen Transfer", J. Catalysis, 111, 397-408 (1988).

17. Ahmed,K., Chowdhury, H.M.,"Dehydrogenation of Cyclohexane and Cyclohexene over Supported Nickel and Platinum Catalysts", Chem. Eng. J., 50, 165-168 (1992).

Acknowledgements: The author would like to thank LCS Inc. for funding and use of the facilities.