Fourier Transform Combinatorial Chemistry applied to the Discovery of Novel Catalysts for the Water-Gas-Shift Reactions

High-throughput synthesis and screening methods have been developed for the discovery of families of high activity water-gas-shift (WGS) catalysts. The discovery libraries, for primary screening, consisted of 16x16 arrays of 256 catalysts on 4” quartz wafers. Catalysts were prepared by robotic liquid dispensing techniques and screened for catalytic activity in Symyx’ Scanning Mass Spectrometer in the temperature range of 200°C to 450°C. The ScanMS is a fast serial screening tool that uses flat wafer catalyst surfaces, local laser heating, a scanning/sniffing nozzle and a quadrupolar mass spectrometer to compare relative catalytic activities. The feed consisted of CO, CO2, H2, H2O with Kr as internal standard in Ar carrier gas. More than 250,000 experiments were conducted to comprehensively examine catalyst performance for various binary, ternary and higher-ordered compositions. The discovered lead compounds encompass supported noble metal systems as well as non-noble metal compositions. Linear grid search strategies as well as modulation techniques using periodic metal concentration profiles across the wafer with different frequency for each metal ingredient have been applied. By using the Pt-Ce-Fe ternary as an example, it is demonstrated that the Fourier transform of sinusoidal metal concentrations represents an effective tool of averaging over experimental fluctuations and non-uniformities across the wafer as well as directly finding correlations from the Fourier coefficients (e.g. synergistic metal combinations). INTRODUCTION Fuel processing of fossil fuels is an economically viable route to produce hydrogen for power generation in fuel cells. A typical fuel processor consists of 3 stages, a reformer for syngas production followed by water-gas-shift stages (CO + H2O = CO2 + H2) and preferential oxidation for clean up of CO that would poison the fuel cell’s electrocatalyst. Highly active, stable and non-pyrophoric WGS catalysts are needed to meet the requirements for smaller fuel processor volumes in mobile fuel cell applications. Ceria supported Pt has been proposed as alternative to traditional Cu-Zn based systems for the medium temperature shift reaction at around 300°C. We have conducted a primary screening program in ScanMS searching for novel catalyst compositions for the low, medium and high temperature shift reactions. Several search strategies including linear grid search, statistical correlation finders (χ-test), predictive modeling and modulation techniques have been applied and evaluated in order to identify correlations and synergistic building blocks among the metal combinations. In this paper, we will discuss the application of Fourier Transform combinatorial chemistry to the discovery of multi metal WGS catalysts. EXPERIMENTAL METHODS Our integrated workflow for high-throughput combinatorial synthesis and screening is described below [1-10]. This workflow allows for quick identification of active and selective catalyst lead compounds within large compositional spaces for the water gas shift reactions over supported noble and base metals, cf., Fig. 1 and 2. General Procedure for Wafer Synthesis Wafer formatted catalyst libraries were designed using Symyx Library Studio software, and synthesized in situ using Symyx’ Impressionist software [11]. Commercial 4” round quartz wafers were cleaned, silanized, and bead blasted through masks to produce 16x16 arrays of wells for catalyst synthesis [12]. These wells were then precoated with a porous carrier layer (ceria, zirconia, alumina) using a slurry dispensing technique. Linear gradients or sinusoidal profiles of metal precursor solutions were chosen in order to obtain an array of solution mixtures within a single microtitre (MT) plate. Catalyst library substrates (wafers) were defined as arrays of wells to which a portion of the premixed solutions from the microtitre plate are transferred. Once the library design was complete, Library Studio generated a text file containing instructions that Impressionist software uses to control liquid dispensing robots. These robots physically prepared the wafer libraries. Standards (Pt/ZrO2) were slurried into several first row and last column wells. The remaining spots in first row/last column were left blank to correct for the background signal during data processing, cf., Fig. 1. Scanning Mass Spectrometer The scanning mass spectrometer, a rapid serial primary screen, was used to test wafer formatted catalyst libraries for the WGS reactions, cf., Fig. 2. Details of the reactor have been published previously [1-10]. The reactor setup permitted the introduction of up to four different gases and one liquid feed. The wafers were mounted on an x-y stage that registers the catalyst to be tested below the reactor head. The catalyst being tested was heated with a CO2 laser to the desired reaction temperature. Temperatures in the range from 200-500°C were obtained using standard quartz wafers, with a thickness of ~1300μm. The chamber pressure was maintained at constant pressure of 30 psig, using argon dilution gas and a pressure controller. The reactants were mixed in the probe head and delivered through two small ports on either side of the catalyst element being analyzed. At the center of the probe head, a small fraction of the reaction gas was removed through a heated ‘sniffer’ capillary. This capillary delivers a constant sample stream to the Bruker Quadrupole Mass Spectrometer. Custom Symyx ScanMS software was written to operate both the reactor and QMS, enabling independent reactor operation (with one wafer) for periods of over a week. Up to 20 quadrupole masses could be monitored for a given experiment. With both heat and reactant streams delivered to each catalyst independently, initial activity and selectivity were measured. For WGS, the feed gas consisted of 7.4% CO, 7.4% CO2, 51.6% H2, 26.2% H2O, 7.4% Kr, with monitored product M/Z of 16 (CH4), 18 (H2O), 28 (CO), 44 (CO2) and 84 (Kr). The Kr signal at M/Z = 84 was used as the internal standard. Each wafer contained at least 6 blank elements and 6 standard catalysts, in order to assess the quality of the data as well as the relative conversion and selectivity for the catalyst array within each wafer. The measurement required from one to three minutes per sample, depending on the catalyst stabilization time and number of masses being analyzed. A single 256 element wafer could be screened at three temperatures over the course of one day. Fourier Transform approach Fourier transform combinatorial chemistry has been described for peptide synthesis on cotton threads wrapped on cylinders of different sizes [13]. We have developed and implemented a Fourier transform combinatorial strategy for the evaluation of mixed metal catalyst libraries. Different and incommensurable fundamental frequencies are assigned to individual metals. Because interactions between metals show up as sums and differences of two or more fundamental frequencies, a numerological selection of frequencies is required that gives little or no ambiguity for at least second-order terms. Suitable ternary frequency combinations are (2, 3, 10) or (4, 5, 7). For the (2, 3, 10) frequency combination, possible second order frequencies are given by 4 (2+2) 1 8 5 6 (3+3) 7 12 13 20 (10+10) showing differences above the diagonal and sums on and below the diagonal. A Fourier transformation on the resulting responses will give distinct frequencies for all firstand second-order effects. Linear terms show up as positive coefficients in the sine transform at the fundamental metal frequency. Quadratic terms show up as sum and difference frequencies in the cosine transform, ideally with opposite signs. A synergy between metals is associated with a positive coefficient on the difference of the metal frequencies and a negative coefficient on their sum. Higher order effects could cause non-zero values at frequencies 9, 11, 14 and above (except for 20). Higher order effects could also show up as inequalities between the sum and difference values (i.e. 1 versus 5, 7 versus 13, 8 versus 12). The magnitude of the Fourier coefficient at a given frequency indicates the significance to WGS activity, averaged over the library, of the particular metal or metal combination. For the Pt-Ce-Fe ternary library to be discussed in more detail below, we have chosen the (4, 5, 7) frequency combination and synthesized the following composition modulated library in “time” domain: Pt = 1 + sin(4*θ), Ce = 1 + sin(5*θ), Fe = 1 + sin(7*θ) where θ varies from 0 to 2π as well position varies from 1 to 225 (with blanks and Pt/ZrO2 standards in first row and last column of 16x16 array). Fe(NO3)3, Ce(NO3)3, H2PtCl6 precursor solutions and water were premixed in a microtitre plate and then dispensed to the ZrO2 carrier precoated wafer. The library was dried at 110°C for 12h, calcined at 500°C for 1h in air, reduced at 300°C for 4h in 5% H2/N2 and screened in ScanMS for WGS activity. The Fourier sine transform of the ScanMS “time” domain screening data into “frequency” domain is calculated by: sinv (frequency, gas, reaction temperature) = Σ screening signal (gas, reaction temperature) * sin (frequency*θ), for the reactant and product gases CO2, CO, H2O, CH4, and correspondingly for the cosine transform cosv. RESULTS Figure 3 shows the ScanMS response signals of the 3-component Pt-Ce-Fe/ZrO2 library with (4-5-7) frequency distribution for CO2 and CO at reaction temperatures of 250°C, 300°C and 350°C. The Fourier coefficients are displayed in figure 4 and also listed in table 1. The sine transform values (linear terms) reveal that generally Pt at frequency 4 is positive/promoting the WGS reaction (i.e. high CO2, low CO), Ce at frequency 5 is neutral or slightly activating, and Fe at frequency 7 is negative (deactivating). At first glance the moderator function of Fe may be surprising given the fact that FeCrAl systems are efficient high temperature shift catalysts at about 400°C, however, the WGS activity of Fe is rapidly declining with decreasing temperature whereas the partly reduced (electron donating) Fe oxide attenuates the much more active metal Pt. On the other hand, Pt-Fe systems are very WGS selective due to the moderating Fe whereas the more active Pt-Ce tends to be unselective (promoting the methanation side reaction) at higher temperatures. The cosine transform values provide information on quadratic terms. A synergy would be associated with a positive coefficient on the difference and a negative coefficient on the sum of the particular metal frequencies. In the CO2 plots, for example, there is almost always a higher-than-noise signal at frequency 8, which corresponds to Pt*Pt (4+4). That means there is a “negative” synergy of Pt with itself. In practical terms, that means that there is a negative curvature in the response, with the CO2 production leveling off as the dispensed Pt amount gets higher. Higher-order terms show up as progressively more complicated sin and cos terms in the combination of frequencies. There is some evidence that higher-order terms might be significant because the sinv plot often shows values at the “wrong” frequencies (especially 1). CONCLUSIONSFourier Transform combinatorial approaches using libraries with modulated metalconcentration profiles have been employed to classify metals and identify correlations formulti metal catalysts for the water-gas-shift reactions. Wafer formatted combinatoriallibraries (16x16) were synthesized by automated liquid or slurry dispensing and screenedin Symyx’ fast serial scanning mass spectrometer. For the Pt-Ce-Fe ternary, Ce functionsas a promoter and Fe as a moderator to the active metal Pt.FTCC represents an effective tool to average over experimental fluctuations across thewafer as well as directly deriving correlations from the Fourier coefficients. Responsesignals from all across the wafer are combined into (few) Fourier coefficients. Whendifferent fundamental frequencies (for the oscillating metal contents in compositionalspace) are chosen for each metal, linear terms show up as positive coefficients in thesine transform whereas quadratic terms manifest themselves as sum and differencefrequencies in the cosine transform. References[1] US 6514764, US 6410331, US 5959297; EP 1019947 to Symyx Technologies, Inc.[2] P. Cong, R. D. Doolen, Q. Fan, D. M. Giaquinta, S. Guan, E. W. McFarland, D. M.Poojary, K. Self, H. W. Turner, W. H. Weinberg, Angew. Chem. Int. Ed. 38 (1999)484.[3] Y. Liu, P. Cong, R.D. Doolen, H.W. Turner, W.H. Weinberg, Catal. Today 61 (2000)87-92.[4] K. Yaccato, A. Hagemeyer, A. Lesik, A. F. Volpe, H. W. Turner, W. H. Weinberg,“High Throughput Screening of SCR and Direct Decomposition DeNOx Catalysts inScanning Mass Spectrometer”, 9 NICE Workshop on "High Temperature Catalysis",Durdent Court, UK, February 11-12, 2003.[5] K. Yaccato, R. Carhart, A. Hagemeyer, A. Lesik, H. W. Turner, A. F. Volpe, W. H.Weinberg, R. K. Grasselli, “CO and CO2 Methanation over Supported Noble MetalCatalysts in High Throughput Scanning Mass Spectrometer”, paper 166d, AIChESpring Meeting, New Orleans, March 30 – April 3, 2003. [6] K. Yaccato, A. Hagemeyer, A. Lesik, A. F. Volpe, H. W. Turner, W. H. Weinberg,“High Throughput Screening of Low Temperature SCR and Direct DecompositionDeNOx Catalysts in Scanning Mass Spectrometer”, Preprints 6 InternationalCongress on Catalysis and Automotive Pollution Control (CAPoC 6), Brussels,Belgium, October 22-24, 2003; Topics in Catalysis 30/31 (2004) 127-132.[7] K. Yaccato, A. Hagemeyer, Y. Liu, A. F. Volpe, W. H. Weinberg, “High ThroughputScreening of Low Temperature CO Oxidation Catalysts for Cold-Start AutomotiveEmissions Control in Scanning Mass Spectrometer”, paper 516a, AIChE AnnualMeeting, San Francisco, November 16-21,2003.[8] A. Guram, A. Hagemeyer, C. G. Lugmair, H. W. Turner, A. F. Volpe Jr, W. H.Weinberg, K. Yaccato, Adv. Synth. Catal. 346 (2+3) (2004) 215-230.[9] K. Yaccato, A. Hagemeyer, A. F. Volpe, W. H. Weinberg, “High ThroughputScreening for the Discovery of More Efficient Catalysts for Emissions Control”,ACS National Meeting, Anaheim, CA, March 28 – April 1, 2004.[10] K. Yaccato, A. Hagemeyer, L. Lefort, H. W. Turner, A. F. Volpe, W. H. Weinberg,“High Throughput Screening of VOC Removal Catalysts in Scanning MassSpectrometer”, 13 International Catalysis Conference (ICC), Paris, France, July11-16, 2004.[11] US 6507945, EP 1175645, EP 108435. Additional patents pending.[12] US 6004617, US 5985356, US 6326090.[13] A. W. Schwabacher, Y. Shen, C. W. Johnson, J. Am. Chem. Soc. 1999, 121,8669-8670.