Gregory Joseph Panuccio. Lanny D. Schmidt

Transcription

1 UNIVERSITY OF MINNESOTA This is to certify that I have examined this copy of a doctoral dissertation by Gregory Joseph Panuccio and have found that it is complete and satisfactory in all respects, and that any and all revisions required by the final examining committee have been made. Lanny D. Schmidt Name of Faculty Adviser Signature of Faculty Adviser Date GRADUATE SCHOOL

2 Heterogeneous and Homogeneous Chemistry in the Catalytic Partial Oxidation of Liquid Hydrocarbon Feedstocks A DISSERTATION SUBMITTED TO THE FACULTY OF THE GRADUATE SCHOOL OF THE UNIVERSITY OF MINNESOTA BY Gregory Joseph Panuccio IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY Lanny D Schmidt, Adviser August 26

3 Gregory Joseph Panuccio 26

4 Acknowledgements I would like to sincerely thank Professor Lanny Schmidt for his guidance and direction over the past four years. His boundless enthusiasm for the work is contagious and his encouragement for me to pursue my own ideas has made this project both very challenging and very rewarding. My colleagues in the Schmidt group have also been an invaluable asset. I would like to thank Dr. Jakob Krummenacher for my experimental training and indoctrination into the lab. Thank you to Bradon Dreyer for being patient, flexible, and understanding with the co-habitation of the reaction hood for the past two years. Thanks to Corey Leclerc, Rama Subramanian, Nick Degenstein, and Anders Bitsch-Larsen for drinking coffee every morning and getting me excited about my day. Thanks to Karthik Venkataraman, Gregg Deluga, James Salge, Ken Williams, Abhishek Jhalani, Paul Dauenhauer, and Raimund Horn for making the Schmidt Group the epicenter of lively discussion and for being great people to work with. Thanks to my friends in the Twin Cities for making my time inside Amundson Hall that much more manageable by helping me to make the best of the time I spent outside of it. Thanks for all the Saturday night Mike Tyson s Punch Out (Contra, RBI baseball, and Super Tecmo Bowl) marathons, the Metrodome tailgating, the weekly roundtables at Legends, Twins games, cookouts, sled riding, and animal petting at the State Fair. Your contributions to my maintained sanity are incalculable. Thanks to my family for supporting me throughout my education. Thanks to my Mom for making me take the hard writing courses in High School (turns out that I would have to write a lot when I grew up) and my Dad for instilling in me a sense of discipline, of confidence, and of humility. My final and most emphatic thanks go to my wife, Michelle. Thanks for your encouragement in difficult times and your companionship at all times. The culmination of this degree is just as much due to your efforts as it is to mine. i

5 Abstract Catalytic partial oxidation (CPO) is emerging as an alternative to industrial steam reforming and steam cracking for the production of high value chemicals such as H 2 or olefins (ethylene, propylene, etc.). These products can be generated in very high yields at millisecond contact times from the reaction of liquid alkane fuels on Rh or Pt catalysts. However, before catalytic partial oxidation can be commercialized, the effects of fuel structure on product distribution and fuel conversion must be studied along with the contributions of heterogeneous and homogeneous chemistry in the reaction of heavy alkanes. In Chapter 2, the CPO of C 1 through C 16 normal alkanes is examined on Pt and Rh-coated α-al 2 O 3 ceramic foam supports. New data for heavy liquid fuels ( C 6 ) is combined with previously reported data for lighter hydrocarbons to explore the effects of reactant fuel molecular weight, catalyst metal, and support structure on the CPO of normal paraffins. These results show that, independent of catalyst metal and support geometry, fuel conversion and the total selectivity to olefin products increase with increasing chain length of the reacting fuel. Conversely, the selectivities of H 2 and CO decrease with increasing molecular weight of the reacting fuel. Pt catalysts generate higher selectivities to ethylene and other olefins than Rh catalysts, but Rh is the better catalyst for synthesis gas (H 2 + CO) production. Catalyst supports with lower internal surface area to volume ratio produce lower selecitivities of H 2 and CO and higher selectivities of H 2 O, ethylene, and other olefins than high surface area to volume catalysts. Catalytic partial oxidation experiments with n-octane, 2,2,4-trimethylpentane (ioctane), and an n-octane:i-octane (1:1) mixture are performed in Chapter 3 in order to explore the effects of chemical structure for single components and binary mixtures of linear and branched alkanes. When reacted as single components, the conversion of i- octane is greater than n-octane at C/O > 1.1 (both fuel conversions are 1% for C/O < 1.1). However, when reacted in an equimolar mixture, the conversion of n-octane is greater than i-octane. All three fuels give high selectivity to syngas (H 2 and CO) on 8 ii

6 ppi supports for C/O < 1. For C/O > 1, n-octane produces high selectivity to ethylene while i-octane makes mostly i-butylene and almost no ethylene. The fuel mixture produces these species proportional to the mole fractions of n-octane and i-octane within the reacting mixture. Increasing the support pore diameter decreases the selectivity to syngas and increases H 2 O and olefin selectivity. The reforming of all three octane fuels from Chapter 3 is modeled in Chapter 4 in an effort to determine the effects of homogeneous and heterogeneous chemistry in the CPO process. Detailed homogeneous reaction mechanisms with several thousand elementary reactions steps and several hundred species are used to simulate experimentally observed olefin selectivities for all three fuels on different catalyst supports at several reactant flow rates. These results show that a majority of the observed olefins are made through gas-phase chemistry if H 2, H 2 O, CO, and CO 2 are made on the surface. In Chapter 5, a method is developed by which temperature and species flow rate profiles are measured for the (CPO) of n-octane in a fixed bed of Pt and Rh-coated alumina spheres in order to determine the mechanism of the heterogeneous reactions. Results indicate that two regions of catalytic activity are present in the bed: a short oxidation zone in the front of the bed and a longer reforming zone downstream. The lengths of the reforming and oxidation zones are dependent on the stoichiometry of the reactor feed. O 2 is consumed rapidly on both Rh and Pt catalysts; however full O 2 conversion is reached faster on Rh than on Pt. Both steam reforming and dry (CO 2 ) reforming are observed in the bed downstream of the oxidation zone, but steam reforming is more extensive because higher conversions of H 2 O are observed than CO 2 on both Rh and Pt catalysts. The conversion of H 2 O and CO 2 in the reforming zone is always higher on Rh than on Pt, therefore Rh is a better reforming catalyst than Pt and produces higher selectivity to H 2 and CO and lower selectivity to H 2 O and CO 2. Low molecular weight hydrocarbons like methane and ethylene are more readily reformed than octane. Results indicate that sphere beds are better supports for syngas production than foam monoliths. In Chapter 6, the CPO of n-octane over Rh and Pt-coated Al 2 O 3 catalysts is examined with H 2 and CH 4 addition to the reactor feed. The aim is to preferentially iii

7 oxidize H 2 or CH 4 and allow more octane to react in the gas phase to form high value chemicals such as ethylene and other olefins. The addition of H 2 increases olefin selectivities on both Rh and Pt foams, and the highest increase in olefin selectivity occurs at H 2 /O 2 = 3/1 on Pt-coated 45 ppi catalysts. In this case, the selectivity to ethylene increases from a maximum of 38% to 51%, and total olefin selectivity increases from 75% to 83% without a decrease in octane conversion. The effect of adding CH 4 is heavily dependent on the catalyst metal and average pore size of the support structure. On Pt catalysts and 45 ppi Rh-coated catalysts, the addition of methane can increase olefin selectivities by several percent. However, on 8 ppi Rh-coated foams, adding methane actually suppresses the formation of olefins and increases selectivity to H 2 and CO. The results of this thesis are summarized in Chapter 7. The goal of this thesis has been to understand the role of fuel structure along with the contributions of homogeneous and heterogeneous chemistry in the CPO of liquid hydrocarbon fuels. One of the most important future directions for this work is to develop heterogeneous reaction mechanisms for heavier hydrocarbons that can be combined with the homogeneous mechanisms in order to successfully simulate the CPO reaction for fuels greater than methane. iv

8 Table of Contents Acknowledgements...i Abstract...ii Table of Contents...v List of Tables...ix List of Figures...x Chapter 1 Introduction What is catalytic partial oxidation? CPO vs. current industrial techniques CPO overview CPO of a wide variety of fuels Experimental variables and their contributions Goals of this work Effect of chain length and branching on CPO Heterogeneous and homogeneous chemistry Application: Increasing olefin yields from the CPO of heavy hydrocarbons...5 References...6 Chapter 2 Comparison of the millisecond contact time catalytic partial oxidation of C 1 through C 16 normal paraffins Introduction Experimental Results Temperature Fuel Conversion...17 v

9 2.3.3 Syngas selectivity H 2 O and CO 2 selectivity Olefins Discussion Effect of catalyst support Differences between Rh and Pt Effects of fuel chain length Conclusions...24 References...24 Chapter 3 Comparison of the catalytic partial oxidation of octane isomers and mixtures over Rh-coated foams Introduction Experimental Results Effect of fuel (single component vs. mixture) Effect of pore size Discussion Fuel reactivity Heterogeneous chemistry Olefins through gas phase chemistry Conclusions...46 References...46 Chapter 4 Modeling the contributions of heterogeneous and homogeneous chemistry in the catalytic partial oxidation of octane isomers and mixtures on Rh-coated foams Introduction...55 vi

10 4.2 Simulation Method Heterogeneous approximation Detailed homogeneous model Simulation results Discussion: Relaxing the 2-zone model Conclusions...61 References...62 Chapter 5 Species and temperature profiles in a differential sphere bed reactor for the catalytic partial oxidation of n-octane Introduction Experimental Results Temperature Reactant profiles H 2 and H 2 O CO and CO Lower molecular weight hydrocarbon profiles Discussion Quantitative interpretation of profile data Rh vs. Pt catalysts Effect of feed stoichiometry Reformation of smaller hydrocarbons Comparison of sphere bed to monolith experiments Conclusions...83 References...83 vii

11 Chapter 6 Increasing olefins by H 2 and CH 4 addition to the catalytic partial oxidation of n-octane Introduction Experimental Results Rh on 8 ppi (w/ washcoat) catalyst Rh on 45 ppi (no washcoat) catalyst Pt on 8 ppi (w/ washcoat) catalyst Pt on 45 ppi (no washcoat) catalyst Discussion Two-zone reaction model H 2 addition CH 4 addition Flames and explosions Conclusions References Chapter 7 Thesis summary and future directions Thesis summary Future directions Heterogeneous mechanism Mixture experiments Profile measurements Catalyst morphology References viii

12 List of Tables Table 1-1 Advantages of CPO vs. conventional industrial reactors...8 Table 1-2 Summary of tunable experimental parameters to maximize syngas or olefin production in the CPO process...8 Table 2-1 Summary of the experimental conditions for the data presented in Chapter Table 6-1 Maximum ethylene and total olefins selectivities for no co-feed, with H 2, and with CH 4 co-feed ix

13 List of Figures Figure 1-1 Illustration of the catalytic partial oxidation process...9 Figure 1-2 Picture of the CPO reactor in operation...1 Figure 1-3 The effects of feed stoichiometry on product distribution...11 Figure 1-4 Photograph and SEM micrograph of an 8 ppi Rh-coated monolith...12 Figure 2-1 Schematic of the CPO reactor for liquid fuels...28 Figure 2-2 Layout of the CPO reaction system...29 Figure 2-3 Measured catalyst back-face temperature for CPO of C 1 through C 16 hydrocarbons...3 Figure 2-4 C 1 through C 16 fuel conversion...31 Figure 2-5 H 2 selectivity for the CPO of C 1 through C 16 hydrocarbons...32 Figure 2-6 CO selectivity for the CPO of C 1 through C 16 hydrocarbons...33 Figure 2-7 H 2 O selectivity for the CPO of C 1 through C 16 hydrocarbons...34 Figure 2-8 C 2 H 4 selectivity for the CPO of C 1 through C 16 hydrocarbons...35 Figure 2-9 Total olefins selectivity for the CPO of C 1 through C 16 hydrocarbons...36 Figure 3-1 Fuel conversion and catalyst back-face temperature for the CPO of n-octane, i-octane, and n-octane:i-octane (1:1)...49 Figure 3-2 Partial oxidation and combustion product selectivities for CPO of n-octane, i-octane, and n-octane:i-octane (1:1)...5 Figure 3-3 Olefin selectivities for CPO of n-octane, i-octane, and n-octane:i-octane (1:1)...51 Figure 3-4 Partial oxidation and combustion product selectivities for the CPO of n-octane on different supports...52 Figure 3-5 Methane and olefin selectivities for the CPO of n-octane on different supports...53 Figure 3-6 Structure of n-octane and i-octane...54 x

14 Figure 4-1 Schematic of the catalytic foam model...64 Figure 4-2 Results comparing the predicted and experimental selectivities of H 2, CO, H 2 O, and CO 2 for different heterogeneous approximations...65 Figure 4-3 Comparison of the predicted and experimental fuel conversion and olefin selectivity for the CPO of n-octane at 2, 4, and 6 SLPM...66 Figure 4-4 Comparison of the predicted and experimental fuel conversion and olefin selectivity from the CPO of i-octane...67 Figure 4-5 Comparison of the predicted and experimental olefin selectivity for the CPO of n-octane:i-octane (1:1)...68 Figure 5-1 Schematic of the sphere bed catalyst assembly...86 Figure 5-2 Catalyst bed temperature profile as a function of catalyst metal and feed stoichiometry...87 Figure 5-3 Fuel flow rate profile for Rh and Pt at different C/O ratios...88 Figure 5-4 O 2 flow rate profile for Rh and Pt at different C/O ratios...89 Figure 5-5 H 2 flow rate profile for Rh and Pt at different C/O ratios...9 Figure 5-6 H 2 O flow rate profile for Rh and Pt at different C/O ratios...91 Figure 5-7 CO flow rate profile for Rh and Pt at different C/O ratios...92 Figure 5-8 CO 2 flow rate profile for Rh and Pt at different C/O ratios...93 Figure 5-9 C 2 H 4 flow rate profile for Rh and Pt at different C/O ratios...94 Figure 5-1 Ratio of the rates of formation of H 2 and consumption of H 2 O as a function of mass of catalyst for Rh and Pt at feed C/O = 1., 1.5, and Figure 5-11 Schematic of the reaction steps in the CPO of n-octane...96 Figure 5-12 Comparison of product selectivities as a function of feed C/O ratio for the CPO of n-octane at 4 SLPM after 1 mm of Rh catalyst for spheres, 8 ppi monoliths, and 45 ppi monoliths...97 Figure 6-1 Schematic of the coupling between heterogeneous and homogeneous chemistry in CPO Figure 6-2 Results of H 2 addition to n-octane CPO on 8 ppi Rh catalyst xi

15 Figure 6-3 Results of CH 4 addition to n-octane CPO on 8 ppi Rh catalyst Figure 6-4 Measured catalyst back-face temperature for the H 2 and CH 4 addition to n-octane CPO on 8 ppi Rh catalyst Figure 6-5 Results of H 2 addition to n-octane CPO on 45 ppi Rh catalyst...12 Figure 6-6 Results of CH 4 addition to n-octane CPO on 45 ppi Rh catalyst Figure 6-7 Results of H 2 addition to n-octane CPO on 8 ppi Pt catalyst Figure 6-8 Results of CH 4 addition to n-octane CPO on 8 ppi Pt catalyst Figure 6-9 Results of H 2 addition to n-octane CPO on 45 ppi Pt catalyst Figure 7-1 SEM micrographs of used 45 ppi 5% Pt catalyst at the same location using SEI and BSE detectors xii

16 Chapter 1 Introduction 1.1 What is catalytic partial oxidation? Catalytic partial oxidation (CPO) is a process for autothermally reforming low value petrochemical feedstocks into high value chemicals such as hydrogen or olefins. This is accomplished by reacting vaporized hydrocarbon fuel and oxygen over a noble metal-coated ceramic support at high space velocities as illustrated in Figure 1-1. A picture of a partial oxidation reactor in operation is shown in Figure 1-2. The process is autothermal because the overall reaction for CPO is exothermic. C8H O 2 8CO + 9H2 rxn ΔH = -676 kj/mol (1-1) Therefore, reactor temperatures on the order of 1 C are maintained with no external heating as long as the reactor is sufficiently insulated. A certain amount of combustion also occurs in the catalyst which adds to the exothermicity of the overall reaction. C H + 25 O 8CO + 9H O rxn ΔH = kj/mol (1-2) Catalytic partial oxidation is industrially relevant because it is possible to tune the reactor to produce high yields of either synthesis gas (syngas H 2 + CO) or olefins such as ethylene and propylene. Olefins are the biggest commodities produced in the chemicals industry with 1 million tons of ethylene produced annually [1, 2], and syngas could be further reformed into a suitable gas mixture for powering fuel cells. A gas stream that has 1 ppm of CO or less can be generated from the syngas that exits the CPO reactor for use in a PEM fuel cell by converting poisonous CO into inert CO 2 via the water-gas shift and preferential oxidation reactions 1

17 Water-gas shift CO + H O CO + H2 ΔH = -41 kj/mol rxn Pt/Ceria 2 2 (1-3) Preferential oxidation 1 Rh or Ru CO + O 2 CO 2 2 ΔH = -281 kj/mol (1-4) rxn with a total residence time of approximately 1 ms [3, 4]. 1.2 CPO vs. current industrial techniques The most widely used industrial methods for producing syngas (steam reforming) and olefins (steam cracking) require large furnaces to drive the endothermic reactions [1, 2, 5]. Steam reforming CH 4 + H2O Ni CO + 3H2 rxn ΔH = +2 kj/mol (1-5) Homogeneous Cracking CH CH + H ΔH = +137 kj/mol (1-6) rxn In the cracking process, steam is added to suppress coking on the catalyst as in H2 O + C (s) CO + H2 ΔH = +131 kj/mol. rxn (1-7) Since these processes are limited by the rate of heat flux from the furnace into a tubular reactor, long residence times on the order of.1 1 second are required. Furthermore, flames inside the industrial furnace can reach temperatures of 15 2 C, which are high enough to produce significant quantities of harmful pollutants like NO x. In comparison, the temperatures inside the CPO reactor are not high enough to produce 2

18 NO x, and the elimination of heat transfer barriers reduces residence times required for high conversion from 1 s to approximately 5 ms. A reduction in the residence time of this magnitude means the size of the CPO reactor can be 1 times smaller than industrial reformers and hydrocrackers. It also allows for better scalability and faster response times if transient operation is needed. This is important if a reactor is designed to feed hydrogen to a fuel cell to power a laptop, cellular phone, or automobile where the electricity demand is transient. The catalysts for CPO are also more stable than their industrial counterparts because they do not need to be periodically shut down in order to de-coke and regenerate the active surface. The summary for the advantages of CPO over conventional industrial reactors is shown in Table 1-1. One disadvantage of the CPO process is that the noble metal catalysts that are used (i.e. Rh and Pt) are much more expensive than the Ni catalyst that is used in industrial steam reforming. 1.3 CPO overview CPO of a wide variety of fuels The formation of syngas from the catalytic partial oxidation of methane over Rh and Pt-coated ceramic foam catalysts was first published by Hickman and Schmidt in 1992 [6, 7]. Since then, a variety of fuels have been reacted in the CPO system including diesel fuel [8], JP-8 jet fuel [9], biodiesel [1], ethanol [11], glycerol [12], and vegetable oil [13]. These experiments have shown that the CPO process is robust in that it can successfully reform a wide array of fuels into valuable chemicals like H 2 or olefins at millisecond contact times without catalyst degradation Experimental variables and their contributions There are some tunable parameters in the CPO process that control the reactor s performance. The simplest variable to manipulate is the feed stoichiometry. In this thesis, the stoichiometry is reported as the carbon to oxygen ratio (C/O) in the feed stream which is defined as the ratio of the moles of atomic carbon from the fuel to the moles of atomic oxygen from O 2. By this definition, the stoichiometric ratio for the partial oxidation reaction is C/O = 1. (see Equation 1-1). The illustration in Figure 1-3 3

19 shows how the feed stoichiometry typically affects the product distribution for liquid alkane fuels [8]. The production of syngas is maximized for C/O ~ 1.. The production of combustion products (H 2 O and CO 2 ) increases as more O 2 is added (C/O decreases) and the reactant mixture approaches combustion stoichiometry. When the reactant stream is more fuel rich (C/O > 1), the production of syngas decreases and the amount of olefins in the product stream increases. The shaded region in Figure 1-3 shows the typical range of experimental C/O ratios. Experiments are not performed for C/O <.7 because the reactor temperature is greater than 125 C and the amount of the desired syngas product decreases. For C/O > 2, the reactor performance is poor because fuel and O 2 conversion drop significantly. Another variable in the CPO process is the catalyst metal. CPO experiments have been performed with a variety of metals including Rh, Pt, Ir, Ni, Pd, Fe, Co, Re, and Ru [14]. Previous results that compare catalyst metals show that Rh is the best catalyst to use for syngas production while Pt is best for making olefins [14-16]. The other metals do not perform as well for a variety of reasons that include coking, sintering, or inactivity. The ceramic foam support structure also affects the product distribution. Bodke and Schmidt have shown that supports with small pore diameters produce higher selectivity to syngas and lower selectivity to olefins than supports with large pore diameters [17]. The application of a γ-alumina washcoat also increases syngas and decreases olefins. These trends are summarized in Table Goals of this work Effect of chain length and branching on CPO For use in an automotive fuel cell, H 2 is likely to be reformed from a distributed fuel like gasoline or diesel fuel. Approximately 5% of the ethylene generated from industrial steam cracking is produced from naphtha [2]. These feedstocks are complicated mixtures of linear, branched, and cyclic alkanes, alkenes, and aromatic compounds. In order to properly design a CPO reactor that can efficiently reform complicated mixtures, it is important to understand how each of these different classes of 4

20 components affects the system performance. It is also important to understand how these different types of reactants interact with one another. The work in Chapters 2 and 3 of this thesis examines the effects of different fuels in the CPO process in detail. The results in Chapter 2 show how the molecular weight of the reacting fuel affects the product distribution and the fuel conversion for the CPO of linear alkanes. In Chapter 3, the reaction of a linear alkane (n-octane) is directly compared to the reaction of a highly branched alkane (i-octane). Also in Chapter 3, the interactions of linear and branched alkanes are explored as an equimolar mixture of n-octane and i-octane is reacted Heterogeneous and homogeneous chemistry A picture and an SEM micrograph of a Rh-coated ceramic foam monolith are shown in Figure 1-4. These show that there is both a continuous void space and solid phase inside the catalyst. Therefore, it is possible that reactions can occur on the surface (heterogeneously) and in the gas-phase (homogeneously). Simulations that combine elementary-step heterogeneous surface mechanisms with homogeneous gas-phase mechanisms show that homogeneous chemistry is not important for the CPO of methane at atmospheric pressure [18]. However, recent experimental and theorectical work has shown that both heterogeneous and homogeneous chemistry are important for the CPO of ethane on Pt catalysts [19-21]. The work described in Chapters 4 and 5 of this thesis describes the mechanism and relationship between gas-phase and surface chemistry in the CPO of liquid hydrocarbons like octane Application: Increasing olefin yields from CPO of heavy hydrocarbons As previously noted, the CPO process could be used as an alternative method for industrial production of ethylene and other commodity chemicals because it can produce high selectivities of olefins in very short contact times. Chapters 4 and 5 explore the effects of homogeneous and heterogeneous chemistry in the CPO system. The goal of the work described in Chapter 6 is to use the knowledge gained from the work in Chapters 4 and 5 to improve the design of the reactor and increase the yield of ethylene 5

21 and other olefins obtained from the catalytic partial oxidation of a naptha-like hydrocarbon feedstock. References [1] Sundaram KM, Shreehan MM, Olszewski EF. Ethylene. In: Kirk-Othmer Encyclopedia of Chemical Technology. John Wiley & Sons, Inc, 21. [2] Zimmermann H, Walzl R. Ethylene. In: Ullmann's Encyclopedia of Industrial Chemistry. Wiley-VCH Verlag GmbH & Co. KGaA, 22. [3] Wheeler C, Jhalani A, Klein EJ, Tummala S, Schmidt LD. The water-gas shift reaction at short contact times. Journal of Catalysis. 24; 223: [4] Jhalani A, Schmidt LD. Preferential CO oxidation in the presence of H 2, H 2 O, and CO 2 at short contact-times. Catalysis Letters. 25; 14: [5] Baade WF, Parekh UN, Raman VS. Hydrogen. In: Kirk-Othmer Encyclopedia of Chemical Technology. John Wiley & Sons, Inc., 21. [6] Hickman DA, Schmidt LD. Production of syngas by direct catalytic oxidation of methane. Science. 1993; 259: [7] Hickman DA, Schmidt LD. Syngas production by direct oxidation of methane over Pt monoliths. Journal of Catalysis. 1992; 138: [8] Krummenacher JJ, West KN, Schmidt LD. Catalytic partial oxidation of higher hydrocarbons at millisecond contact times: decane, hexadecane, and diesel. Journal of Catalysis. 23; 215: [9] Dreyer BD, Lee IC, Krummenacher JJ, Schmidt LD. Autothermal steam reforming of higher hydrocarbons: n-decane, n-hexadecane, and JP-8. Applied Catalysis A. 26; 37: [1] Subramanian R, Schmidt LD. Renewable olefins from biodiesel by autothermal reforming. Angewandte Chemie. 24; 44: [11] Deluga GA, Salge JR, Schmidt LD, Verykios XE. Renewable hydrogen from ethanol by autothermal reforming. Science. 24; 33: [12] Dauenhauer PJ, Salge JR, Schmidt LD. Renewable hydrogen by autothermal steam reforming of volatile carbohydrates. Journal of Catalysis. Submitted; 6

22 [13] Salge JR, Dreyer BD, Dauenhauer PJ, Schmidt LD. Renewable hydrogen from nonvolatile fuels by reactive flash volatilization. Science. Submitted; [14] Torniainen PM, Chu X, Schmidt LD. Comparison of monolith-supported metals for the direct oxidation of methane to syngas. Journal of Catalysis. 1994; 146: 1-1. [15] Huff M, Torniainen PM, Hickman DA, Schmidt LD. Partial oxidation of CH 4, C 2 H 6, and C 3 H 8 on monoliths at short contact times. Natural Gas Conversion II. 1994; [16] Huff M, Torniainen PM, Schmidt LD. Partial oxidation of alkanes over noble metal coated monoliths. Catalysis Today. 1994; 21: [17] Bodke A, Bharadwaj S, Schmidt LD. Effect of ceramic supports on partial oxidation of hydrocarbons over noble metal coated monoliths. Journal of Catalysis. 1998; 179: [18] Goralski CT, O'Connor RP, Schmidt LD. Modeling homogeneous and heterogeneous chemistry in the production of syngas from methane. Chemical Engineering Science. 2; 55: [19] Beretta A, Ranzi E, Forzatti P. Experimental and theoretical investigation on the roles of heterogeneous and homogeneous phases in the oxidative dehydrogenation of light paraffins in novel short contact time reactors. Studies in Surface Science and Catalysis. 2; 13B: [2] Huff M, Androulakis IP, Sinfelt JH, Reyes SC. The contribution of gas-phase reactions in the Pt-catalyzed conversion of ethane-oxygen mixtures. Journal of Catalysis. 2; 191: [21] Donsi F, Williams KA, Schmidt LD. A multistep surface mechanism for ethane oxidative dehydrogenation on Pt- and Pt/Sn-Coated Monoliths. Industrial and Engineering Chemistry Research. 25; 44:

23 Table 1-1. Advantages of CPO vs. conventional industrial reactors Property CPO Conventional τ (s) Reactor Adiabatic Wall heated tubes Carbon No carbon Cokes Operation Robust Narrow range of operation NO x Produced No Yes Table 1-2. Summary of tunable experimental parameters to maximize syngas or olefin production in the CPO process as described in Section To maximize Property Syngas Olefins Feed C/O.8-1. > 1.2 Catalyst Metal Rh Pt Pore diameter Small Large w/ washcoat Yes No 8

24 N 2 + O 2 Quartz Tube τ ~ 1ms 1 cm 1.8 cm Catalytic Foam Monolith H 2 + CO or C 2 H 4 + C 3 H Figure 1-1. Illustration of the catalytic partial oxidation process. Vaporized hydrocarbon fuel (n-octane and i-octane in this figure) and air are passed over a noble metal-coated ceramic foam monolith to produce valuable chemicals at millisecond contact times. The reaction can be tuned by changing the feed stoichiometry, the catalyst metal, or the catalyst support structure to produce high selectivity of either syngas or olefins. 9

25 Quartz tube Heat shield (See Fig. 2-1) Insulation Autothermal catalyst T ~ 1 C Figure 1-2. Photo of the CPO reactor in operation. The process is autothermal because the reaction is exothermic and produces enough heat to sustain itself when properly insulated. 1

26 Stoichiometric partial oxidation Combustion Products Syngas Olefins Feed C/O 3 Figure 1-3. The effect of the feed stoichiometry on the product distribution. For C/O ~ 1., syngas is produced in the highest selectivity. As C/O decreases and more O 2 is fed into the reactor, the mixture approaches the combustion reaction stoichiometry and H 2 O and CO 2 are more readily formed. When the O 2 in the feed is decreased, the selectivity of syngas drops and the olefin selectivity increases. The shaded area is the typical range of experimental C/O operation. 11

27 (A) (B) Void space Rh-coated Al 2 O 3 Figure 1-4(A-B). Photograph (A) and SEM micrograph (B) of an 8 ppi Rh-coated ceramic foam monolith. The catalysts have highly interconnected pore geometry with continuous solid phase and void space. Reactions can occur on the surface (heterogeneously) or in the gas-phase (homogeneously). The relationship between these two types of reactions for the CPO of liquid hydrocarbons is explored in this thesis. 12

28 Chapter 2 Comparison of the millisecond contact time catalytic partial oxidation of C 1 to C 16 normal paraffins * 2.1 Introduction There are several variables in the CPO process that can determine the characteristics of the reaction. Since industrial raw materials like natural gas and petroleum-derived feedstocks are complex mixtures of linear and branched alkanes, alkenes, and aromatics, one of the most important factors to understand is the behavior of the CPO system in relation to the reacting fuel. The first step in successfully designing a process for a fuel that is a complicated mixture is to understand how the components of that mixture behave separately. To this end, a wide variety of fuels have been reacted in CPO reactors including normal alkanes ranging from methane to n-hexadecane [1-1], branched alkanes like i-butane and i-octane [8, 11, 12], and cyclic fuels like cyclohexane [7, 12]. Results show that each of these different classes of compounds behaves differently in the CPO system. This chapter focuses on the reaction of linear alkanes and describes the effect of chain length on experimental product selectivities, fuel conversions, and catalyst temperatures. A variety of metals have been used to catalyze the partial oxidation reactions including (but not limited to) rhodium, platinum, nickel, iridium, and palladium. Previous results on lighter alkanes have shown that Rh catalysts give the best performance for syngas production, while reactions on Pt giver higher selectivities to olefins, water, and carbon dioxide [2, 4, 6, 13]. The other metals have shown poorer performance than Rh and Pt because of coking, sintering, or inactivity [2, 6, 11]. For this reason, this chapter focuses on examining the effects of Rh and Pt catalysts on the CPO process for normal alkanes. * Portions of this chapter appear in Panuccio G.J., Dreyer B.J., and Schmidt L.D., Comparison of the millisecond contact time catalytic partial oxidation of C 1 C 16 normal paraffins, Submitted to AIChE Journal (26). American Institute of Chemical Engineers. 13

29 The third parameter that is explored in this chapter is the effect of the catalyst support geometry. Honeycomb monoliths, foam monoliths, sphere beds, and wire gauze have all been previously studied in the millisecond contact time catalytic partial oxidation process. Each of these supports has its own advantages. For example, honeycomb monoliths have a very regular geometry and are good supports to use for modeling experimental results and developing reaction mechanisms [14]. Wire gauzes have very rapid quenching times and so they can generate significant quantities of oxygenates from alkanes [15, 16]. Foam monliths and sphere beds have highly interconnected pore structures that improve radial mixing and heat transfer and therefore give the best overall performance in terms of syngas (or olefin) production and fuel conversion. This chapter focuses on α-al 2 O 3 foam monoliths because they have the largest amount of data in the literature for lighter alkanes that can be used for comparison purposes. Specifically, experiments performed on 8 pores per linear inch (ppi) monoliths that are prepared with a γ-al 2 O 3 washcoat are compared to 45 ppi foams that have no washcoat. The 8 ppi foams have a smaller average pore diameter than the 45 ppi supports and therefore have a higher surface area to volume ratio [17]. The addition of a washcoat to the 8 ppi support roughens the surface and further increases the internal surface area to gas volume ratio within the catalyst [3]. Comparison of the results of the experiments performed on these two different supports gives insight into the relationship between heterogeneous and homogeneous chemistry in the CPO process. 2.2 Experimental The catalyst was a 17 mm diameter, 1 mm length α-al 2 O 3 ceramic foam monolith with either 8 or 45 pores per linear inch (ppi) that was coated with 5 wt% Rh or Pt metal. The catalysts were prepared by the dropwise addition of the aqueous metal salt solution (Rh(NO 3 ) 3 or H 2 PtCl 6 ) onto the alumina support, drying in air, and calcining in a closed oven at 6 C for 5 hours. The 8 ppi supports were also coated with a 5 wt% γ-al 2 O 3 washcoat in order to roughen the surface and increase the effective surface area to volume ratio within the catalytic foam [3]. The washcoat was applied by preparing a slurry of γ-alumina in water and adding it dropwise to the 8 ppi foams. The 14

30 foams were then dried and calcined in a closed oven for 4 hours at 6 C. Four different catalysts were studied in this Chapter: 8 ppi 5% Rh (or Pt) with 5% washcoat and 45 ppi 5% Rh (or Pt) with no washcoat. A schematic of the reactor is shown in Figure 2-1. The catalytic foam was placed between two blank α-al 2 O 3 monoliths to prevent axial radiative heat losses, wrapped in alumina-silicate paper to prevent gas bypass, and placed inside a 19 mm ID quartz tube which was wrapped with insulation. A K-type thermocouple was placed between the back-face of the catalytic foam and the downstream heat shield to measure the temperature. For the liquid hydrocarbon experiments, the fuel was delivered to the reactor via a low-flow automotive fuel injector from a 5 psig pressurized fuel tank. The flowrate of fuel was determined from the desired C/O ratio for the experiment which was defined as the ratio of the moles of carbon atoms from the fuel divided by the moles of oxygen atoms in the feed. By this definition, the stoichiometric feed composition for the partial oxidation reaction (shown in Chapter 1) was C/O = 1. for all fuels. N 2, O 2, and gaseous hydrocarbons were delivered to the reactor via calibrated Brooks mass flow controllers. The upstream portion of the tube was wrapped with a Variac-controlled resistive heating tape that served to preheat the gases and vaporize the liquid hydrocarbon when applicable. A schematic of the layout of the CPO system is shown in Figure 2-2. The fuel injector and mass flow controllers were controlled by LabView software on the computer. The software also recorded thermocouple readings at the catalyst back-face and in the preheat zone of the reactor tube. Products were sampled downstream with a gas tight syringe and injected into an HP 589 GC for separation and analysis. All gases that were not sampled were incinerated in a Bunsen burner before they were vented into the hood. Species flow rates (F i ) were calculated from the peak areas (A i ) in the resulting chromatogram using N 2 as an internal standard according to Equation 2-1. F = i R R i A A i N2 N2 F N2 (2-1) 15

31 The response factors for each species (R i ) were determined from calibration gases purchased from Matheson Tri Gas. The response factor for N 2 was defined to be 1. Experimental carbon and hydrogen atom balances typically closed within ± 5%. Results are reported in terms of product carbon (or hydrogen) atom selectivity and fuel conversion. Selectivity is defined as the amount of carbon (or hydrogen) in product i divided by the total amount of carbon (or hydrogen) in all the products and is calculated according to Equation 2-2. S = i nf i i nf products j j (2-2) Here n i is the number of carbon (or hydrogen) atoms in species i. Fuel and oxygen conversion (X) are calculated according to Equation 2-3. X = Fuel F - F Fuel, In F Fuel, In Fuel, Out (2-3) 2.3 Results Table 2-1 shows the nominal total reactant flow rate and oxidant for all the combinations of fuel and catalyst that are investigated in this study. All results that are indicated as being performed for the current work are performed at 4 standard liters per minute (SLPM at 25 C and 1 atm) total flow rate with air (N 2 /O 2 = 3.76) as the oxidant. At a typical reactor temperature of 9 C, this flow rate corresponds to an average catalyst contact time of approximately 5 ms. Data that is acquired from the literature is as close to these conditions as is available. No data is plotted for experiments on 8 ppi Pt with washcoat catalysts for fuels lighter than hexane because the only data available in the literature is for catalysts that were not prepared with a washcoat. All the data that is presented as new for the current work is an average of at least three data points obtained on a minimum of two catalysts that were similarly prepared. 16

32 2.3.1 Temperature The measured catalyst back-face temperature is plotted as a function of reactant C/O ratio, catalyst metal, support geometry, and reacting fuel in Figure 2-3. The measured catalyst back-face temperature is not often reported in the literature for the lighter alkanes, therefore most of the data presented in Figure 2-3 was obtained as current results for liquid normal alkane fuels. There are no prevailing trends that relate reactant chain length to catalyst temperature. In general, the back-face temperature cools from a range of 15 to 115 C at C/O < 1. to approximately 75 to 85 C at C/O = 2.. There are a few notable exceptions. On 45 ppi Rh no washcoat catalyst, the temperature measured for ethane CPO is several hundred degrees hotter than the temperatures for other fuels. This is also the case for the reaction of methane on 45 ppi Pt no washcoat catalysts. The methane experiment was carried out in O 2 with a 5% N 2 dilution and the ethane experiment had a 2% N 2 dilution while the other temperatures are obtained from reactions in air (N 2 /O 2 = 3.76/1). The temperatures observed in the ethane and methane experiments are higher because of the absence of N 2 which acts as a diluent and a heat sink Fuel Conversion Fuel conversion is plotted as a function of reactant C/O ratio, catalyst metal, and support geometry for normal alkanes ranging from methane to hexadecane in Figure 2-4. In general, the conversion of each fuel is 1% for C/O < 1. and decreases as the reactant mixture becomes more fuel rich. These results also show that fuel conversion increases with increasing molecular weight independent of catalyst metal or support pore size. For example, at C/O = 2. on 8 ppi Rh with washcoat catalysts, fuel conversion increases from 45% for hexane to 91% for hexadecane. The exception to the trend is the conversion of ethane on 45 ppi Rh no washcoat catalyst where the ethane conversion is greater than octane but less than decane. This is most likely due to the difference in the experimental conditions because the ethane CPO was conducted in only 2% N 2 (instead of air stoichiometry) which resulted in higher catalyst temperatures than the other fuels 17

33 (see Figure 2-3). The increase in temperature increases reaction rates which leads to higher fuel conversion than expected if the reaction were carried out in air. The average pore diameter in the support structure does not have a large effect on fuel conversion for Rh catalysts. The conversion of decane, octane, hexane, and methane is approximately the same on 8 ppi and 45 ppi Rh catalysts. For Pt coated monoliths, the 8 ppi catalysts have a higher fuel conversion than the 45 ppi catalysts. For example, at C/O = 2., the conversion of octane on 8 ppi Pt (85%) is greater than the conversion on 45 ppi Pt catalyst (64%). For each of the C 6 through C 16 liquid fuels, the conversion on 8 ppi Pt catalysts is higher than the corresponding conversion on 8 ppi Rh catalysts. For example, the conversion of hexane at C/O = 2. increases from 45% for Rh to 61% for Pt. However, on 45 ppi foams, the catalyst metal does not have an effect on the fuel conversion. The octane conversion at C/O = 2. on 45 ppi Pt catalyst (64%) is approximately equal to the conversion on 45 ppi Rh catalyst (65%) Syngas selectivity Hydrogen and CO selectivities are plotted in Figures 2-5 and 2-6 as a function of catalyst metal, support geometry, reacting fuel, and feed stoichiometry. For a single fuel, syngas selectivity decreases with increasing C/O inlet stoichiometry on every catalyst studied. For example, experiments performed on 8 ppi Pt with washcoat catalysts show that the selectivity of H 2 produced from hexane decreases from 8% at C/O =.8 to 31% at C/O = 2.. H 2 and CO selectivities decrease with increasing molecular weight on 8 ppi catalysts. Carbon monoxide selectivity at C/O = 1.5 decreases from 79% for hexane to 35% for hexadecane for experiments performed on 8 ppi Rh with washcoat catalysts. For the 45 ppi no washcoat catalysts, syngas selectivity is approximately equal for fuels heavier than ethane. On a catalyst with equivalent specific surface area, experiments performed on Rh produce higher H 2 and CO selectivities than experiments performed with Pt. The CPO of octane on 8 ppi catalysts at C/O = 2. show that reactions on Rh produce 39% H 2 selectivity while Pt only produces 6%. Similarly, octane CPO experiments performed at 18

34 C/O =.8 on 45 ppi foams coated with Rh produce 39% H 2 selectivity while Pt gives only 23%. For the same metal catalyst, reactions carried out on 8 ppi washcoated foams generate much higher selectivities to syngas than 45 ppi foams that are prepared without a washcoat. The CPO of decane on 8 ppi foams at C/O = 1. produces 75% CO selectivity while 45 ppi foams only produce 47% H 2 O and CO 2 selectivity The selectivity of H 2 O is plotted as a function of inlet stoichiometry, catalyst metal, catalyst support structure, and reacting fuel in Figure 2-7. In general, the trends in water selectivity are the opposite of the trends for H 2 and CO selectivity. For the same metal and reacting fuel, the selectivity of H 2 O is greater on 45 ppi catalysts that are prepared without a washcoat than on 8 ppi foams that are prepared with a washcoat. For example, the CPO of octane at C/O = 1. on Rh-coated foams generates only 6% selectivity on 8 ppi supports and 29% selectivity on 45 ppi catalysts. Also, reactions performed on the same support with the same fuel yield higher selectivities on Pt catalysts than on Rh catalysts. Further, the selectivity of H 2 O is fairly constant with respect to the feed C/O ratio. The selectivity of CO 2 (not shown) from the CPO of normal paraffins is not strongly dependent on fuel molecular weight, support pore size, or catalyst metal. In general, the CO 2 selectivity varies between 5 and 15% and increases with decreasing inlet C/O as the stoichiometry approaches that of combustion. For C/O < 1., the CO 2 selectivity increases by less than 5% as the reacting fuel chain length increases from methane to hexadecane Olefins Ethylene selectivity is plotted as a function of catalyst metal, support geometry, reacting fuel, and inlet stoichiometry in Figure 2-8. On 8 ppi Rh and Pt catalysts, the ethylene selectivity increases with increasing molecular weight for alkanes lighter than octane. For example, on 8 ppi Rh catalysts at C/O = 1.2, ethylene selectivity increases 19

35 from < 1% for butane, to 3% for hexane, to 15% for hexadecane. The CPO of octane, decane, and hexadecane produce very similar ethylene selectivities. On 8 ppi Rh catalysts, the maximum ethylene selectivity obtained from reaction with octane, decane, and hexadecane are 2%, 21%, and 19%, respectively. Similarly, on 8 ppi Pt catalysts, the maximum ethylene selectivity for octane (37%), decane (36%), and hexadecane (38%) are very close even though they occur at different fuel C/O ratios. The effects of the catalyst metal and support geometry on ethylene selectivity are opposite of their effects on syngas selectivity. On the same support, Pt produces higher ethylene selectivity than Rh. Also, 45 ppi supports prepared without a washcoat produce higher selectivity to ethylene than 8 ppi catalysts prepared with a washcoat. Figure 2-9 shows the total olefin selectivity as a function of inlet stoichiometry, catalyst metal, support geometry, and reacting fuel. The total olefin selectivity increases with increasing molecular weight on all catalysts, though the effect is more pronounced on the 8 ppi supports that are prepared with a washcoat. For example, on 8 ppi Rh with washcoat catalyst, the maximum total olefin selectivity increases from 2% for butane, to 14% for hexane, to 43% for octane, to 61% for decane, to 66% for hexadecane. This trend continues on 45 ppi no washcoat catalysts, but difference in total olefin selectivities is not as pronounced. 2.4 Discussion Effect of catalyst support Previous results on lighter hydrocarbons have shown that catalyst supports with high specific surface areas produce larger amounts of syngas and fewer olefins and water than supports with low specific surface areas [3]. In the current work, the comparison of the selectivities obtained from 8 ppi with washcoat catalyst to 45 ppi without washcoat catalyst with the same metal for a single fuel in Figures 2-5 through 2-9 shows that this trend also holds for the CPO of heavy alkanes. These results can be explained through a 2-zone reaction scheme. A recent study on the simulation of the CPO of octane isomers and mixtures on Rh-coated ceramic foam monoliths has shown that the CPO system can be accurately 2

36 described by splitting the reactor into a heterogeneous oxidation zone at the front of the catalyst that is followed by a reforming zone downstream [8]. In the first zone, fuel and oxygen react heterogeneously to form H 2, H 2 O, CO, CO 2 and heat. The heat generated in the oxidation zone is utilized in the reforming zone to homogeneously pyrolyze the remaining fuel into smaller hydrocarbons like ethylene and other olefins. The ratio of the active surface area to the volume of the gas phase inside the catalytic foam is higher for the 8 ppi catalysts than the 45 ppi foams because the average pore diameter is smaller (.25 vs..47 mm, respectively) and the surface is roughened by the application of a γ- alumina washcoat. Therefore, the ratio of heterogeneous to homogenous reactions taking place in the 8 ppi foams should be higher as well. Since H 2 and CO are formed heterogeneously and ethylene and other olefins are formed largely in the gas phase, experiments performed on 8 ppi catalysts should generate higher selectivities of heterogeneously-derived syngas while reactions on 45 ppi foams produce greater amounts of homogeneously-generated olefins. The observed experimental selectivity of H 2 O also supports the hypothesis that olefins are made through the gas phase via the pyrolysis of heavier hydrocarbons. The selectivities of ethylene and other olefins increase with increasing inlet C/O for all fuels and catalysts (Figures 2-8 and 2-9). However, the selectivity of H 2 O remains fairly constant with respect to feed C/O over this same range (Figure 2-7). If olefins are formed through an oxidative dehydrogenation mechanism, 1 CH x 2x+2 + O 2 CH x 2x + HO 2 2 then the selectivity of water should increase as the olefin selectivities increase. However, if olefins are generated from the pyrolysis of heavier hydrocarbons, then water would not be formed as a coproduct, and the selectivity of H 2 O would not increase as the selectivity of olefins increases. 21

37 2.4.2 Differences between Rh and Pt For the reaction of the same fuel on the same catalyst support, Rh catalysts always generate higher selectivities to syngas and lower water and olefin selectivities than Pt catalysts (Figures 2-5 through 2-9). This suggests that there are consistent mechanistic differences between the heterogeneous reactions taking place on Rh and Pt catalysts for n-paraffins ranging from methane to hexadecane. Hickman and Schmidt compared the elementary step surface reaction mechanisms of the direct formation of H 2 from the partial oxidation of methane on Rh and Pt [18]. They found that the primary difference that leads to the large discrepancy in the experimentally measured H 2 selectivities for Pt and Rh lies in the difference in the activation energy for formation of OH on the surface as in: H (s) + O (s) OH (s) + * They found that the activation energy for this reaction on Rh is significantly higher than on Pt. This energy barrier dictates that less adsorbed hydrogen atoms are going to combine with adsorbed O atoms and are therefore more likely to combine with other H atoms to form H 2 instead of forming water through OH. It is also possible that the H 2 is formed indirectly. This means that the catalyst promotes the deep oxidation of methane to H 2 O and CO 2 which are then catalytically reformed to H 2 and CO through steam reforming and CO 2 reforming overall reactions. CH (Combustion) 4 + 2O 2 2H2O + CO2 CH + H O H + CO (Steam reforming) CH + CO H + 2CO (CO2 reforming) 22

38 If the indirect formation of H 2 and CO is considered, the discrepancy in the performance of Rh and Pt catalysts with respect to syngas production could be a function of the difference in the activities of the metals. It is possible that both Rh and Pt are efficient oxidation catalysts, but Rh is a much better reforming catalyst than Pt. Leclerc and coworkers have recently shown that a dual catalyst bed of Pt followed by Ni can generate similar selectivities to pure Rh catalyst for the CPO of methane [19]. However, a dual bed catalyst of Ni followed by Pt does not perform as well as the Rh catalyst. In the Pt/Ni arrangement, the Pt catalyst combusts the methane and generates H 2 O, CO 2, and heat. The Ni catalyst (which is a very good steam reforming catalyst and is typically used in industrial steam reforming reactors [2]) then reforms the H 2 O and CO 2 into H 2 and CO. This supports the hypothesis that both Rh and Pt can function as a combustion catalyst, but that Rh is a better reforming catalyst than Pt Effect of fuel chain length There are several trends that are observed as a function of the chain length (or molecular weight) of the reacting fuel. The results displayed in Figure 2-4 show that fuel conversion increases with increasing molecular weight for C/O > 1.. This is probably due to a combination of two factors. First, the gas phase reactivity increases for heavier hydrocarbons. This means that, at the same temperature, a heavier fuel is more likely to pyrolyze in the second zone of the catalyst and therefore will have a higher conversion. Also, a previous study on the ignition behavior of n-alkanes on Rh during catalytic partial oxidation shows that methane should have a sticking coefficient that is 2 to 4 times smaller than n-octane [21]. This means that a smaller alkane is much less likely to absorb when it comes in contact with the surface which also decreases the probability that it will react on the surface compared to a larger hydrocarbon. The results displayed in Figures 2-5 and 2-6 show that syngas selectivity decreases with increasing chain length. Conversely, the total olefin selectivity increases with increasing molecular weight as shown in Figure 2-9. Both of these phenomena can be explained by the relative reactivity of the reactant fuel in the gas phase. Simulations that combine heterogeneous and homogeneous chemistry to study the CPO of methane at 23

39 millisecond contact times show that almost no methane is consumed through gas-phase reaction pathways at atmospheric pressure and experimentally observed temperatures [22]. However, other simulations performed for heavier fuels show that a large amount of octane is consumed through gas-phase reaction pathways under these experimental conditions [8]. Since it has been shown that syngas is formed primarily on the surface and olefins are formed primarily through the gas-phase, it follows that the selectivity of the species that are formed in the gas phase should increase relative to the selectivity of the species formed on the surface as the homogeneous reactivity of the fuel increases. 2.5 Conclusions The catalytic partial oxidation of n-alkanes ranging from methane to hexadecane has been performed on Pt and Rh-coated ceramic foam monoliths with different specific surface areas. Catalyst supports with high internal surface area to volume ratios produce high selectivities to syngas while those with low specific surfaces areas produce high selectivities of olefins. Over the entire range of fuels, Rh is consistently the best catalyst for syngas production while Pt catalysts generate high selectivities to olefins. Significant trends are observed with respect to the molecular weight of the reacting fuel. As molecular weight increases, fuel conversion and olefin selectivities increase while syngas selectivity decreases. This is due to the relative reactivity of the fuel in the gas phase. Heavier fuels are more homogeneously reactive and therefore have a higher conversion and produce more olefins than lighter hydrocarbons. Overall, the process is very robust and also very selective. Depending on the application, a system could be designed to generate high yields of either syngas or olefins from a wide range of starting materials. References [1] Hickman DA, Schmidt LD. Syngas formation by direct catalytic oxidation of methane. Science. 1993; 259: [2] Torniainen PM, Chu X, Schmidt LD. Comparison of monolith-supported metals for the direct oxidation of methane to syngas. Journal of Catalysis. 1994; 146:

40 [3] Bodke A, Bharadwaj S, Schmidt LD. Effect of ceramic supports on partial oxidation of hydrocarbons over noble metal coated monoliths. Journal of Catalysis. 1998; 179: [4] Huff M, Torniainen PM, Hickman DA, Schmidt LD. Partial oxidation of CH 4, C 2 H 6, and C 3 H 8 on monoliths at short contact times. Natural Gas Conversion II. 1994; [5] Huff M, Schmidt LD. Ethylene formation by oxidative dehydrogenation of ethane over monoliths at very short contact times. J. Phys. Chem. 1993; 97: [6] Huff M, Torniainen PM, Schmidt LD. Partial oxidation of alkanes over noble metal coated monoliths. Catalysis Today. 1994; 21: [7] Dietz III AG, Carlsson AF, Schmidt LD. Partial oxidation of C 5 and C 6 alkanes over monolith catalysts at short contact times. Journal of Catalysis. 1996; 176: [8] Panuccio GJ, Williams KA, Schmidt LD. Contributions of heterogeneous and homogeneous chemistry in the catalytic partial oxidation of octane isomers and mixtures on rhodium coated foams. Chemical Engineering Science. 26; 61: [9] Krummenacher JJ, West KN, Schmidt LD. Catalytic partial oxidation of higher hydrocarbons at millisecond contact times: decane, hexadecane, and diesel. Journal of Catalysis. 23; 215: [1] Degenstein NJ, Subramanian R, Schmidt LD. Partial oxidation of n-hexadecane at short contact times: Catalyst and washcoat loading and catalyst morphology. Applied Catalysis A. 26; 35: [11] Liebman LS, Schmidt LD. Oxidative dehydrogenation of isobutane at short contact times. Applied Catalysis A. 1999; 179: [12] O'Connor RP, Klein EJ, Schmidt LD. High yields of synthesis gas by millisecond partial oxidation of higher hydrocarbons. Catalysis Letters. 21; 7: [13] Hickman DA, Haupfear EA, Schmidt LD. Synthesis gas formation by direct oxidation of methane over Rh monoliths. Catalysis Letters. 1993; 17: [14] Raja LL, Kee RJ, Deutschmann O, Warnatz J, Schmidt LD. A critical evaluation of navier stokes, boundary layer, and plug flow models of the flow and chemistry in a catalytic-combustion monolith. Catalysis Today. 2; 59:

41 [15] Iordanoglou D, Schmidt LD. Oxygenates from alkanes in single gauze reactors at short contact times. Journal of Catalysis. 1999; 87: [16] O'Connor RP, Schmidt LD. C 6 oxygenates from n-hexane in a single gauze reactor. Chemical Engineering Science. 2; 55: [17] Twigg MV, Richardson JT. Theory and applications of ceramic foam catalysts. Trans IChemE. 22; 8: [18] Hickman DA, Schmidt LD. Steps in CH 4 oxidation on Pt and Rh surfaces: Hightemperature reactor simulations. AIChE Journal. 1993; 39: [19] Tong GCM, Flynn J, Leclerc CA. A dual catalyst bed for the autothermal partial oxidation of methane to synthesis gas. Catalysis Letters. 25; 12: [2] Baade WF, Parekh UN, Raman VS. Hydrogen. In: Kirk-Othmer Encyclopedia of Chemical Technology. John Wiley & Sons, Inc., 21. [21] Williams KA, Schmidt LD. Catalytic autoignition of higher alkane partial oxidation on Rh-coated foams. Applied Catalysis A: General. 26; 299: [22] Goralski CT, O'Connor RP, Schmidt LD. Modeling homogeneous and heterogeneous chemistry in the production of syngas from methane. Chemical Engineering Science. 2; 55:

42 Table 2-1. Summary of experimental conditions for the data presented. Catalyst Fuel Flow Oxidant Reference Methane 5 SLPM 3% N 2 [3] Rh 8 ppi w/ washcoat Pt 8 ppi w/ washcoat Rh 45 ppi no washcoat Pt 45 ppi no washcoat Butane 5 SLPM 3% N 2 [3] Hexane 4 SLPM Air Current Work Octane 4 SLPM Air Current Work Decane 4 SLPM Air Current Work Hexadecane 4 SLPM Air [1] Hexane 4 SLPM Air Current Work Octane 4 SLPM Air Current Work Decane 4 SLPM Air Current Work Hexadecane 4 SLPM Air Current Work Methane 5 SLPM 3% N 2 [3] Ethane 5 SLPM 2% N 2 [5] Butane 5 SLPM 3% N 2 [3] Hexane 4 SLPM Air Current Work Octane 4 SLPM Air Current Work Decane 4 SLPM Air Current Work Methane 4 SLPM 5% N 2 [2] Ethane 4.5 SLPM Air [4] Propane 5 SLPM Air [6] Butane 5 SLPM Air [6] Hexane 4 SLPM Air Current Work Octane 4 SLPM Air Current Work Decane 4 SLPM Air Current Work 27

43 Fuel Injector N 2 and O 2 Heating Tape Mixer Preheat Thermocouple Insulation Heat Shields Catalyst Septum Sampling Port Products Back-face thermocouple Figure 2-1. Schematic of the CPO reactor for liquid fuels. Fuel is introduced into the reactor through a low-flow automotive fuel injector, vaporized, and mixed with air. The heat shields prevent radiative heat losses in the axial direction and the tube is wrapped in insulation to prevent radial heat losses. The reactor used for fuels that are gases at room temperature is the same except that there is no fuel injector and the fuel is introduced to the reactor from the side port along with the N 2 and O 2. 28

44 System Control Fuel Injector Control Fuel Tank Mass Flow Control GC Signal Thermocouples Sampling Port Incinerator N 2 O 2 HP 589 GC Figure 2-2. Layout of the CPO reaction system. N 2 and O 2 mass flow controllers and the automotive fuel injector are controlled by LabView software on the computer. The software also records back-face and upstream thermocouple readings. Product gases are sampled downstream from the center of the tube and injected into an HP 589 GC for separation and analysis. All gases that are not sampled are incinerated. 29

45 12 8 ppi w/ WC 45 ppi no WC 12 Rh Back-face Temp. ( C) C 16 C 8 C 1 C 6 Back-face Temp. ( C) C 8 C 1 C 2 C C/O C/O Pt Back-face Temp. ( C) C 6 C 8 C 16 C 1 Back-face Temp. ( C) CH 4 C 1 C 4 C 2 C 6 C C/O C/O Figure 2-3. Measured catalyst back-face temperature for CPO of C 1 C 16 hydrocarbons on 8 ppi with washcoat and 45 ppi no washcoat Rh and Pt catalysts. 3

46 8 ppi w/ WC 45 ppi no WC Rh Fuel Conv. (%) CH 4 C 4 C 6 C 16 C 1 C 8 Fuel Conv. (%) CH 4 C 4 C 2 C 6 C 1 C C/O C/O 1 C 16 1 C 1 Pt Fuel Conv. (%) C 6 C 8 C 1 Fuel Conv. (%) C 2 C 4 CH 4 C 3 C6 C C/O C/O Figure 2-4. C 1 C 16 fuel conversion on 8 ppi with washcoat and 45 ppi no washcoat Rh and Pt catalysts. Under fuel rich conditions (C/O > 1.), fuel conversion increases with increasing molecular weight on all catalysts. 31

47 Rh Pt H 2 Select. (%) H 2 Select. (%) C/O C 16 C 1 8 ppi w/ WC 45 ppi no WC C 8 C C 4 CH 4 C/O C 6 C 8 C 1 C 16 H 2 Select. (%) H 2 Select. (%) C 6 CH 4 C 8 C 1 C 2 C C 3 C 8 C 1 CH 4 C 6 C 4 C 2 C/O C/O Figure 2-5. H 2 selectivity for CPO of C 1 C 16 hydrocarbons on 8 ppi with washcoat and 45 ppi no washcoat Rh and Pt catalysts. H 2 selectivity decreases with increasing molecular weight for experiments on 8 ppi catalysts. For 45 ppi catalysts, H 2 selectivity is approximately equal for fuels heavier than ethane. 32

48 Rh Pt CO Select. (%) CO Select. (%) C/O C 16 8 ppi w/ WC 45 ppi no WC C 1 C 8 C C 4 CH 4 C/O C 6 C 8 C 1 C 16 CO Select. (%) CO Select. (%) CH 4 C 8 C 6 C 1 C C 3 C 8 C 6 C 4 CH 4 C 1 C/O C 4 C/O C 2 Figure 2-6. CO selectivity for CPO of C 1 C 16 hydrocarbons on 8 ppi with washcoat and 45 ppi no washcoat Rh and Pt catalysts. CO selectivity decreases with increasing molecular weight for experiments on 8 ppi catalysts and is approximately the same for fuels heavier than ethane on 45 ppi foams. 33

49 Rh Pt H 2 O Select. (%) H 2 O Select. (%) CH 4 8 ppi w/ WC 45 ppi no WC C/O C 16 C 1 C16 C 1 C 8 C C 6 C/O C 8 H 2 O Select. (%) H 2 O Select. (%) C8 C 6 CH 4 C CH 4 C1 C/O C 4 C/O C 2 C 2 C 3 C 8 C 6 Figure 2-7. H 2 O selectivity for CPO of C 1 C 16 on 8 ppi with washcoat and 45 ppi no washcoat Rh and Pt catalysts. For the same fuel and catalyst metal, the selectivity of H 2 O is greater on 45 ppi with no washcoat supports than on 8 ppi with washcoat supports. Pt catalysts produce more H 2 O than Rh catalysts when the same fuel is reacted on the identical supports. 34

50 8 ppi w/ WC 45 ppi no WC 5 5 Rh C 2 H 4 Select. (%) 4 3 C 8 2 C1 1 C C 16 6 C C/O C 2 H 4 Select. (%) 4 C 2 3 C 6 2 C 8 1 C C/O 5 5 C 2 Pt C 2 H 4 Select. (%) C 4 C 16 C 8 2 C 8 C 1 C 1 1 C 6 C 2 H 4 Select. (%) 4 C 3 C C/O C/O Figure 2-8. C 2 H 4 selectivity for CPO of C 1 C 16 hydrocarbons on 8 ppi with washcoat and 45 ppi no washcoat Rh and Pt catalysts. On 8 ppi catalysts, C 2 H 4 selectivity increases with increasing molecular weight for fuels lighter than octane. Ethylene selectivity is nearly independent of molecular weight for experiments performed on 45 ppi catalysts. 35

51 Rh Tot. Olefin Select. (%) C/O C C C C 6 4 C 2 5 C 16 5 Pt C Tot. Olefin Select. (%) C 1 8 ppi w/ WC 45 ppi no WC C 8 C C/O C 16 C 1 C 8 C 4 C 6 Tot. Olefin Select. (%) Tot. Olefin Select. (%) C 4 C/O C/O C 2 C 1 C 6 C 8 Figure 2-9. Total olefin selectivity for CPO of C 1 C 16 hydrocarbons on 8 ppi with washcoat and 45 ppi no washcoat Rh and Pt catalysts. Total olefin selectivity increases with increasing molecular weight for all catalysts. In general, Pt catalysts produce more olefins than Rh, and 45 ppi no washcoat catalysts have higher olefin selectivities than 8 ppi with washcoat catalysts. 36

52 Chapter 3 Comparison of the catalytic partial oxidation of octane isomers and mixtures on rhodium coated foams * 3.1 Introduction As stated in Chapter 1, CPO is an emerging technology that can be used to generate high selectivities to either H 2 or olefins in millisecond contact times. For use in a transportation fuel cell application, H 2 will likely be reformed from distributed liquid fuels like gasoline or diesel. Olefins will be made from refinery sources like naptha [1, 2]. These chemical feedstocks are complicated mixtures that typically contain linear and branched hydrocarbons as well as aromatic compounds. The relationship between CPO reactor performance and reactant fuel molecular weight for straight chain hydrocarbons was explored in Chapter 2. The comparison of the reaction of linear and branched alkanes as individual components and in binary mixtures in a CPO reactor will be explored in this chapter and can offer insight as to how actual distributed feedstocks will perform and what contributions linear and branched alkanes make to the overall process. There are two motivations for the work in this chapter. First, a direct comparison of the CPO of n-octane, 2,2,4-trimethylpentane (i-octane), and an equimolar mixture of n- octane and i-octane is performed in order to explore the effects that branching has on fuel reactivity and product distribution when isomers are reacted as single components and as a binary mixture. These two fuels were chosen because they are model components of gasoline, and they have the same chemical formula but very different chemical structures. Previous work on binary mixtures of linear alkanes where the components have different molecular weights has argued that the specific C/O ratio for each component (local C/O ratio) and diffusive flux are important parameters for fuel conversion [3]. These * Portions of this chapter appear in Panuccio G.J., Williams K.A., and Schmidt L.D., Contributions of heterogeneous and homogeneous chemistry in the catalytic partial oxidation of octane isomers and mixtures on rhodium coated foams, Chemical Engineering Science, 61, (26). 26 Elsevier Science Ltd. 37

53 variables are eliminated in this work because n-octane and i-octane have the same chemical formula (C 8 H 18 ). For this analysis, the most important parameter to consider when comparing the competitive and non-competitive conversion of these isomers is the structural differences between a straight-chain and a highly branched alkane. Previous work on the CPO of heavy alkanes shows that straight-chain hydrocarbons like decane and hexadecane produce the highest olefin selectivity to ethylene [4], while branched and cyclic hydrocarbons like i-octane and cyclohexane produce high selectivities to branched and cyclic olefins like i-butylene and cyclohexene [5]. Comparing the olefin selectivities resulting from the reaction of a binary mixture of octane isomers to the single components will aid in the analysis of the reaction pathways required for olefin generation. Secondly, the effect of the monolith support is explored. Previous experiments on butane, i-octane and n-decane have shown that increasing the pore diameter increases selectivity to olefins and combustion products and decreases syngas selectivity [5-7]. Increasing the pore size increases the extent of homogeneous chemistry because the ratio of surface area to volume is decreased. A detailed examination of the dependence of product distribution on pore size gives further insight as to which products are made heterogeneously and which are formed homogeneously in the CPO system. 3.2 Experimental The fuels used in these experiments are HPLC grade (> 99% purity) n-octane, i- octane, and n-octane:i-octane (1:1). The reactor is a quartz tube with an inner diameter of 19 mm. A detailed sketch of the reactor setup has been provided elsewhere [4, 8, Figure 2-1]. The fuel is delivered to the reactor through a low flow automotive fuel injector from a 5 psig pressurized liquid fuel tank. The fuel injector sprays a conical dispersion of droplets onto the wall of the tube creating a thin film of fuel that is vaporized by heating the walls of the tube with a Variac-controlled resistive heating tape. High purity O 2 and N 2 are delivered separately to the reactor through calibrated mass flow controllers. The fuel, O 2, and N 2 mix in this arrangement such that homogeneous auto-ignition of the fuel is avoided upstream of the catalyst [4]. The total flow rate of fuel and air to the reactor is 38

54 varied between 2 and 6 standard liters per minute (SLPM at 25 C and 1 atm) corresponding to residence times on the order of 5 to 15 ms at average catalyst back-face temperatures between 8 and 1 C. The feed stoichiometry is given as the carbon to oxygen ratio (C/O), and is defined as the moles of carbon atoms divided by the moles of oxygen atoms in the feed. By this definition, the stoichiometric feed composition for partial oxidation is at C/O = 1.. The experimental system is controlled by a personal computer with a program written in LabView. The software regulates the N 2 and O 2 mass flow controllers and the automotive fuel injector. The software also reads and stores the temperatures from chromel-alumel K-type thermocouples located at the back-face of the catalyst and in the upstream mixing zone. The products are analyzed with an HP589 Gas Chromatography (GC) instrument fitted with a GS-GasPro 6 m capillary column and a thermal conductivity detector. Liquid nitrogen is used to cool the GC oven to an initial temperature of -8 C so that all products can be analyzed using a single separation column. The resulting carbon and hydrogen atom balances typically close to within ± 5%. The catalyst support structures are cylindrical α-alumina macroporous ceramic foam monoliths of 18 mm diameter and 1 mm length with a void fraction of approximately 8% [9]. They have both a continuous pore phase and alumina phase. The supports are characterized by the average pores per linear inch (ppi). Both 8 and 45 ppi monoliths are used for these experiments. All catalysts have approximately 5 wt% Rh loading and are prepared by coating the surface with Rh(NO 3 ) 2 solution and calcining in air at 6 C for 6 hours. The 8 ppi monoliths are also coated with 5 wt% γ-alumina washcoat before the application of Rh in order to roughen the surface [6], while the 45 ppi foams are not prepared with a washcoat. The addition of washcoat further increases the disparity in the extent of heterogeneous reaction between the 8 and 45 ppi foams. Previous results on the CPO of smaller alkanes have shown that washcoat promotes the formation of syngas over H 2 O and CO 2 [6]. The temperature at which n-octane and i-octane spontaneously begin reacting catalytically (light off) on rhodium is approximately 25 C [8], which is marked by a steep increase of the back-face temperature of the catalyst. In order to raise the catalyst 39

55 to the light-off temperature, nitrogen is passed through the quartz tube and resistively heated by heating tape wrapped around the upstream portion of the tube. The nitrogen in turn heats the monolith. Once the catalyst reaches sufficient temperature, the oxygen and fuel are turned on simultaneously and the reaction catalytically ignites. The system reaches steady state in less than one minute at which point a product sample can be taken and analyzed. The reactor is shut down by turning off the oxygen flow first and then the fuel injector. 3.3 Results Experiments were repeated on three similarly prepared catalysts for each fuel and flow rate studied. Each data point shown is an average of the results from those experiments. Experimental variability was normally within ± 5% of the reported value. In general, the O 2 conversion was > 95% for the 8 ppi monoliths and > 9% for 45 ppi. Also, at a given feed stoichiometry, the back-face temperature of the catalyst and the fuel conversion increased with increasing flow rate. In this section, the differences in effects of fuels and foam supports are examined in detail Effect of fuel (single component vs. mixture) Figures 3-1 through 3-3 compare the fuel conversion, back-face temperature, and product selectivity resulting from the reaction of n-octane, i-octane, and n-octane:i-octane (1:1) with air. This experiment was performed on 45 ppi 5 wt% Rh catalyst at 6 SLPM. The trends observed for this case are also observed at 2 and 4 SLPM and on 8 ppi monoliths (not shown). The fuel conversions and catalytic back-face temperatures of n- octane and i-octane are displayed in Figure 3-1. Figure 3-1(A) shows conversion vs. C/O for single components and Figure 3-1(B) is the conversion of each fuel within the mixture. The fuel conversion is 1% from C/O =.7 to C/O ~ 1.1 after which it decreases with increasing C/O. For the pure fuels, i-octane has a higher conversion than n-octane at C/O > 1.1 with values of 71% and 6%, respectively at C/O = 2.. This trend reverses when the fuels are reacted in a 1:1 mixture: n-octane has a higher conversion than i-octane (75% vs 6%) at C/O = 2.. 4

56 Figure 3-1(C) displays the catalytic back-face temperature as a function of C/O for n-octane, i-octane, and the 1:1 mixture. The temperature increases as C/O decreases and approaches combustion stoichiometry (C/O =.32), and decreases as C/O increases and the process operates under fuel rich conditions. The observed temperature falls from 115 C at C/O =.8 to 8 C at C/O = 2.. All three fuels studied give approximately the same catalytic back-face temperature. The selectivity of syngas (H 2 and CO) and combustion products (H 2 O and CO 2 ) are displayed vs. C/O for all three fuels in Figure 3-2. At low C/O ratios, the H 2 and CO selectivity is slightly larger for n-octane than i-octane. The corresponding selectivity for the mixture experiment falls between the pure fuels. At C/O > 1., the syngas selectivity differences between fuels are not statistically significant. This suggests that there is a small effect of fuel structure on catalytic activity at low C/O when comparing species of identical molecular weight. The CO 2 selectivity also has a small dependence on the fuel used in the reaction because i-octane generates 4% more CO 2 than n-octane at low C/O. All fuels produce approximately 3% selectivity to water with no statistically significant difference between fuels. The selectivity of ethylene, propylene, n-butylene and i-butylene are displayed in Figure 3-3 for different fuels. The selectivity of ethylene in the product stream is greatly dependent on both C/O and the reacting fuel. For n-octane, the maximum ethylene selectivity is 37% at C/O = 1.1. The maximum ethylene selectivity obtained from reacting i-octane is only 12% at C/O =.9. The equimolar feed mixture produces a maximum ethylene selectivity of 24% at C/O = 1.; which is almost an exact average of the results of the constituent fuels. i-butylene is the principal olefin produced from reacting pure i-octane generating a maximum selectivity of 39% at C/O = 2.. The maximum selectivity of i-butylene generated from the reaction of n-octane is only 3%. The reaction of an equimolar mixture generates an average of the i-butylene selectivities from the constituent components with a maximum selectivity of 19% observed at C/O = 2.. There is approximately 5-1% more propylene made by reacting i-octane than n- octane at all C/O ratios studied. As with the other olefins, the mixture generates an average amount of the two pure components. Other minor products produced in these 41

57 reactions include methane, acetylene, ethane, and C 5 and C 6 olefins. The selectivity of these species is never greater than a few percent and is not shown Effect of pore size Figures 3-4 and 3-5 compare resulting product selectivity for both 8 ppi (5% Rh, 5% γ-alumina washcoat) and 45 ppi (5% Rh) catalytic foam monoliths. The experiment plotted was completed by reacting n-octane at a nominal flow rate of 4 SLPM; but the same trends are observed for experiments performed at 2 and 6 SLPM and with i-octane and n-octane:i-octane (1:1) (not shown). The catalytic back-face temperature is the same (within experimental error) for both catalysts and decreases from 112 C at C/O =.7 to 82 C at C/O = 2.. The fuel conversion is also equivalent for both catalysts and decreases from 1% at C/O =.7 to 5% at C/O = 2. (data not shown). The product distribution is largely dependent on the average pore size of the support used in the catalytic partial oxidation system. Figure 3-4 displays the selectivity of H 2, CO, H 2 O, and CO 2 as a function of C/O for 8 and 45 ppi catalytic foam monoliths. The maximum H 2 selectivity generated from 8 ppi foams is 88% compared to a maximum of 48% selectivity on 45 ppi monoliths. The largest difference in selectivity to H 2 is 64% (88% vs. 24%) at C/O =.9. A similar trend is observed for CO. More CO is made with the 8 ppi foam with the maximum difference of 41% (86% vs. 45%) at C/O of.9. The exact opposite effect is observed when comparing combustion product selectivities. Much more water is produced from reactions on 45 ppi foams. The maximum difference is 25% at C/O =.9. The difference is not as pronounced in the resulting selectivity of CO 2 where the largest difference is 3%. Figure 3-5 displays the selectivity of CH 4, C 2 H 4, C 3 H 6, and n-c 4 H 8 as a function of C/O for 8 and 45 ppi catalytic foam monoliths. The selectivity is higher for each of these species for experiments carried out on 45 ppi supports over most of the domain. The maximum observed methane selectivity on 45 ppi is 8.5% compared to 4% on 8 ppi. Up to 35% C 2 H 4 selectivity is generated on 45 ppi compared to only 17% on 8 ppi. The general trend of a two factor difference continues with maximum selectivities to 42

58 propylene of 13% and 8% on 45 and 8 ppi, respectively. The maximum selectivity to n- butylene is 8% on 45 ppi and 4% on 8 ppi. 3.4 Discussion Fuel reactivity Experimental data shows that n-octane and i-octane give the same conversion (nearly 1%) up to C/O ~ 1.1 both when reacted individually and in an equimolar mixture. For C/O > 1.1, the amount of O 2 in the feed is limiting, and decreasing the O 2 in the feed decreases the amount of exothermic surface reactions that occur and leaves more fuel remaining to react homogeneously. Homogeneous reaction rates are very sensitive to temperature, and the decrease in exothermic surface reactions decreases the gas temperature in the ceramic foam and causes fuel conversion to decrease with increasing C/O since a large amount of the fuel must be converted in the gas phase. i-octane should have a higher conversion than n-octane at high experimental C/O because i-octane has one quaternary, one tertiary, one secondary, and five primary carbons while n-octane has two primary and 6 secondary carbons as shown in Figure 3-6. The quaternary-secondary and tertiary-secondary C-C bonds (and also the secondary and tertiary C-H bonds) in i-octane are weaker than the corresponding secondary-secondary C-C bonds (and secondary C-H bonds) in n-octane because the radicals formed from the cleavage of those bonds have resonance stabilization [1]. According to the experimental data, this analysis holds true when the fuels are reacted as single components. The conversion of i-octane is 1-15% higher than n-octane at C/O = 2. when the components are reacted individually (Figure 3-1(A)). However, when they are reacted in an equimolar mixture, the conversion of n-octane is approximately 1-15% higher than i- octane at C/O = 2. (Figure 3-1(B)). Examination of the olefin selectivity shows that n- octane and i-octane proportionally produce the same selectivity to olefins when reacted in an equimolar mixture as they do when reacted as pure fuels. These results suggest that the individual components have the same gas phase reactivity in the mixture as in the pure fuel, so the only difference in conversion for the mixture must occur at the surface. Previous results on binary mixtures of fuels with a large disparity in molecular weight 43

59 indicate that the C/O ratio specific to both fuels and relative rates of molecular flux to the surface are of primary importance when considering individual component surface reactivity [3]. Since n-octane and i-octane have the same chemical formula, another factor must be critical. Studies of the adsorption and desorption of straight chain alkanes on single crystal catalysts have shown that activation energy for desorption increases with increasing chain length [11, 12]. A higher activation energy for desorption indicates that a molecule is less likely to desorb and more likely to stay on the surface and react. The effective chain length for i-octane is five carbons long compared to eight carbons for n-octane. This suggests that i-octane has a lower desorption activation energy and would be more likely to desorb from the surface back into the gas phase. It has also been shown that the enthalpy for adsorption of n-octane is 3 kcal/mol less than i-octane on smectite clay at 15 C [13] which indicates that n-octane should adsorb better under competitive conditions. Qualitatively, differences in steric effects may hinder the adsorption of i- octane more than n-octane because of its highly branched geometry. Any combination of these effects could explain why n-octane would have a higher heterogeneous conversion than i-octane when they are competitively reacting with O 2 on the surface Heterogeneous chemistry Increasing the average pore size of the supporting foam decreases the ratio of surface area to volume within the catalyst. Consequently, the relative ratio of heterogeneous to homogeneous chemistry decreases as well. The results shown in Figures 3-4 and 3-5 indicate that as the pore size is increased, the selectivity of H 2 and CO decreases substantially while the selectivity of H 2 O markedly increases. This is also observed in the CPO of fuels lighter [6, 14, 15] and heavier [7] than octane. This suggests that a portion of the syngas formed in the reactor is not made by direct partial oxidation but by steam reforming. Beretta and Forzatti have shown that both direct and indirect reactions pathways are important in the CPO of ethane and propane on Rh/α- Al 2 O 3 catalysts [16]. The selectivity of CO 2 is independent of the pore size of the support 44

60 suggesting that CO 2 reforming reactions are not as important as steam reforming reactions in the CPO process Olefins through gas-phase chemistry Above C/O = 1., the O 2 in the feed is limiting enough that a substantial amount of fuel is left over to react homogeneously. At this point, a significant amount of ethylene is produced from n-octane because the high reactor temperature generates reaction rates fast enough for complete thermolysis of the fuel not consumed by surface reactions to form ethylene. However, as C/O is further increased, the temperature in the reactor decreases. The complete cracking of the fuel is not possible because the reaction rates are too slow at the temperature of the reactor. Therefore, the maximum C-atom selectivity for ethylene is typically found at C/O ~ 1.3. The same trend holds for propylene and butylene as the temperature continues to decrease with increasing C/O. Since these species are formed first and then further react to produce smaller olefins, a maximum selectivity to each should be found at successively higher C/O as the temperature in the reactor continues to decrease and the homogeneous reaction rates decrease. The maximum selectivity to propylene and butylene is usually observed at C/O ~ 1.5 and ~ 2., respectively. These maxima have also been observed for n-decane and n-hexadecane experiments on Rh-coated ceramic foams [4]. The selectivity to combustion products and syngas does not vary much as a function of fuel structure. However, the type of olefin produced is very much a function of the structure of the reacting fuel. Previous experiments on linear alkanes such as propane, butane, pentane, hexane, decane, and hexadecane have shown that ethylene is generated in the highest selectivity [4, 6, 15, 17]. Continuing on this trend, fuel rich n- octane CPO also produces high selectivity to ethylene. Ethylene is generated from n- octane through cascading C-C bond cleavage followed by βh elimination. Propylene and larger olefins are the result of incomplete reaction along this pathway. Conversely, branched and cyclic alkanes produce branched and cyclic olefins, respectively. Fuel rich CPO of i-butane generates i-butylene [14, 18]; cyclohexane produces an array of unsaturated cyclic C 6 species like cyclohexene, cyclohexadiene, and benzene; and i- 45

61 octane gives the highest selectivity to i-butylene [5, 15]. The mechanism for i-butylene formation from the pyrolysis of i-octane has been described previously and involves tertiary-secondary C-C bond scission followed by βh elimination [5, 19, 2]. The olefin distribution from the fuel rich CPO of an n-octane:i-octane (1:1) mixture is an average of the distribution from the single components. The selectivity of each species is directly proportional to the amount of the fuel it is generated from in the reacting mixture. This suggests that there are no strong interactions between the gasphase reactions of the different components within the fuel. 3.5 Conclusions Comparison of the partial oxidation of n-octane and i-octane shows that the selectivities of H 2, H 2 O, CO, and CO 2 are independent of the structure of the reacting fuel. However, the olefin distribution is strongly dependent on the chemical structure of the fuel in that n-octane produces mostly ethylene and propylene and i-octane produces mostly i-butylene and propylene. Different trends in fuel conversion are observed when comparing the reaction of an equimolar mixture of n-octane and i-octane to the reaction of pure components. Under fuel rich stoichiometries, n-octane has a higher conversion than i-octane in the mixture, but i-octane has a higher conversion when reacted as a pure fuel. The surface must selectively react with n-octane in the presence of i-octane, perhaps from an imbalance in the competitive absorption/desorption of both species due to steric hindrances for i- octane. Experiments carried out on 8 and 45 ppi monoliths show that increasing pore size strongly decreases H 2 and CO selectivity while increasing H 2 O and olefins. This suggests that both direct and indirect heterogeneous reactions are important in the formation of syngas. References [1] Sundaram KM, Shreehan MM, Olszewski EF. Ethylene. In: Kirk-Othmer Encyclopedia of Chemical Technology. John Wiley & Sons, Inc,

62 [2] Zimmermann H, Walzl R. Ethylene. In: Ullmann's Encyclopedia of Industrial Chemistry. Wiley-VCH Verlag GmbH & Co. KGaA, 22. [3] Subramanian R, Panuccio GJ, Krummenacher JJ, Lee IC, Schmidt LD. Catalytic partial oxidation of higher hydrocarbons: reactivities and selectivities of mixtures. Chemical Engineering Science. 24; 59: [4] Krummenacher JJ, West KN, Schmidt LD. Catalytic partial oxidation of higher hydrocarbons at millisecond contact times: decane, hexadecane, and diesel. Journal of Catalysis. 23; 215: [5] O'Connor RP, Klein EJ, Schmidt LD. High yields of synthesis gas by millisecond partial oxidation of higher hydrocarbons. Catalysis Letters. 21; 7: [6] Bodke A, Bharadwaj S, Schmidt LD. Effect of ceramic supports on partial oxidation of hydrocarbons over noble metal coated monoliths. Journal of Catalysis. 1998; 179: [7] Krummenacher JJ, Schmidt LD. High yields of olefins and hydrogen from decane in short contact time reactors: rhodium versus platinum. Journal of Catalysis. 24; 222: [8] Williams KA, Schmidt LD. Catalytic autoignition of higher alkane partial oxidation on Rh-coated foams. Applied Catalysis A: General. 26; 299: [9] Twigg MV, Richardson JT. Theory and applications of ceramic foam catalysts. Trans IChemE. 22; 8: [1] Whitesell JK, Fox MA. Organic Chemistry Jones & Bartlett Publishing, [11] Wetterer SM, Lavrich DJ, Cummings T, Bernasek SL, Scoles G. Energetics and kinetics of the physisorption of hydrocarbons on Au(111). J. Phys. Chem. B. 1998; 12: [12] Tait SL, Dohnalek Z, Campbell CT, Kay BD. n-alkanes on MgO(1). II. Chain length dependence of kinetic desorption parameters for small n-alkanes. The Journal of Chemical Physics. 25; 122: [13] Keldsen GL, Nicholas JB, Carrado KA, Winans RE. Molecular modeling of the enthalpies of adsorption of hydrocarbons on smectite clay. J. Phys. Chem. 1994; 98:

63 [14] Liebman LS, Schmidt LD. Oxidative dehydrogenation of isobutane at short contact times. Applied Catalysis A. 1999; 179: [15] Dietz III AG, Carlsson AF, Schmidt LD. Partial oxidation of c5 and c6 alkanes at millisecond contact times. Journal of Catalysis. 1998; 176: [16] Beretta A, Forzatti P. Partial oxidation of light paraffins to synthesis gas in short contact-time reactors. Chemical Engineering Journal. 24; 99: [17] Huff M, Schmidt LD. Olefin formation by direct oxidation of propane and butane at short contact times. Journal of Catalysis. 1994; 149: [18] Huff M, Schmidt LD. oxidative dehydrogenation of isobutane on monoliths at short contact times. Journal of Catalysis. 1995; 155: [19] Ranzi E, Faravelli T, Gaffuri P, Garavaglia E, Goldaniga A. Primary pyrolysis and oxidatoin reactions of linear and branched alkanes. Industrial and Engineering Chemistry Research. 1997; 36: [2] Walker JQ, Maynard JB. Analysis of vapor-phase pyrolysis products of the four trimethylpentane isomers. Analytical Chemistry. 1971; 43:

64 Conversion (%) Conversion (%) Back-face Temp ( o C) 1 (A) 9 8 i-c 8 H 18 7 n-c 8 H 18 6 Pure Fuels C/O 1 9 n-c 8 H i-c 8 H 18 6 (1:1) Mixture (B) C/O n-c 8 H (C) (1:1) Mixture i-c 8 H C/O Figure 3-1. Conversion of n-octane and i-octane as a function of C/O ratio when reacted as (A) single components and in a (B) 1:1 mixture. (C) Measured catalytic back-face temperature vs C/O for n-octane ( ), i-octane ( ), and an equimolar mixture ( ). Experiments performed on 45 ppi 5% Rh catalysts (no washcoat) and nominal flow rate of 6 SLPM. 49

65 Selectivity (%) n-c 8 H 18 i-c 8 H 18 (1:1) Mix H i-c 8 H 18 n-c 8 H 18 (1:1) Mix CO Selectivity (%) C/O 4 2 C/O (1:1) Mix H 2 O CO (1:1) Mix 2 n-c 8 H i-c 8 H 18 1 n-c8 H 18 i-c 8 H C/O C/O Figure 3-2. Partial oxidation and combustion product selectivities for n-octane ( ), i- octane ( ), and an equimolar mixture ( ) as a function of C/O ratio. Experiments performed on 45 ppi 5% Rh catalysts (no washcoat) and nominal flow rate of 6 SLPM. The selectivity of syngas and combustion products resulting from the CPO of octane isomers is independent of the structure of the reacting fuel. 5

66 Selectivity (%) C 2 H 4 n-c 8 H 18 C 3 H 6 15 (1:1) Mix (1:1) Mix 1 5 i-c 8 H 18 n-c 8 H 18 i-c 8 H 18 Selectivity (%) C/O 1 n-c 4 H 8 8 n-c 8 H 18 6 (1:1) Mix 4 i-c 8 H C/O C/O 5 i-c 4 H 8 4 i-c 8 H (1:1) Mix 1 n-c 8 H C/O Figure 3-3. Olefin selectivities from the CPO of octane isomers as a function of C/O ratio. Experiments performed on 45 ppi 5% Rh catalysts (no washcoat) and nominal flow rate of 6 SLPM. The CPO of n-octane results in the highest olefin selectivity to ethylene while i-octane CPO generates highest olefin selectivity to i-butylene. Olefin selectivities obtained from the reaction of an equimolar mixture are an average of the results from the constituent components. 51

67 1 8 H CO Selectivity (%) ppi 45 ppi ppi 45 ppi Selectivity (%) C/O C/O 4 15 H 2 O CO ppi 8 ppi ppi 45 ppi C/O C/O Figure 3-4. Partial oxidation and combustion product selectivity for different catalyst supports as a function of C/O ratio. 8 ppi catalysts are prepared with 5 wt% γ-alumina washcoat while 45 ppi catalysts are prepared with no washcoat. Experiments performed with n-octane fuel and nominal flow rate of 4 SLPM. Increasing pore diameter significantly decreases the selectivity of H 2 and CO while increasing the selectivity of H 2 O. 52

68 Selectivity (%) CH 4 C 2 H ppi 2 8 ppi 1 45 ppi 8 ppi Selectivity (%) C/O 1 C/O C 3 H 6 45 ppi 12 8 n-c 4 H 8 45 ppi 9 8 ppi ppi C/O C/O Figure 3-5. Methane and olefin selectivities for different catalyst supports as a function of C/O ratio. 8 ppi catalysts are prepared with 5 wt% γ-alumina washcoat while 45 ppi catalysts are prepared with no washcoat. Experiments performed with n-octane fuel and nominal flow rate of 4 SLPM. Increasing the pore diameter significantly increases the selectivity of methane and olefins. 53

69 n-octane structure 2 o 2 o i-octane structure 4 o 3 o 2 o Figure 3-6. Structure of n-octane and i-octane. The 4-2 and 3-2 C C bonds and the 3 C H and 2 C H bonds in i-octane are weaker than the 2-2 C C bonds and 2 C H bonds in n-octane with respect to homogeneous bond cleavage. 54

70 Chapter 4 Modeling the contributions of heterogeneous and homogeneous chemistry in the catalytic partial oxidation of octane isomers and mixtures on rhodium coated foams * 4.1 Introduction In this chapter the CPO of octane isomers and mixtures is modeled using the data from Chapter 3 in order to demonstrate that this process can be accurately described by combining heterogeneous and homogeneous chemistry. Previous research has demonstrated that ethylene can be produced with 7% selectivity from the reaction of ethane with O 2 on noble metal coated ceramic foam monoliths [1]. It was first thought that ethylene was formed heterogeneously from ethane, and a surface reaction mechanism was developed to describe the reaction of ethane on Pt foams [2]. Recent experimental and theoretical work on the oxidative dehydrogenation of ethane on Pt/γ-Al 2 O 3 catalysts has shown that ethylene is primarily produced through gas-phase reactions and suggests that the catalyst is critical in initiating gas-phase reactions by heterogeneously oxidizing the fuel to CO, CO 2, and H 2 O and generating the heat that drives the homogeneous reactions that produce ethylene [3-5]. Simulations of these experiments that integrate heterogeneous and homogeneous chemistry agree with this relationship between surface and gas-phase chemistry [6-8]. These ideas on the CPO of ethane are extended to the development of a model for the reaction of octane isomers. A two-zone model is developed by which H 2, CO, H 2 O, and CO 2 are assumed to form through heterogeneous reactions between absorbed fuel and O 2 in the first zone at the entrance of the catalytic bed. Olefins and other hydrocarbons are then formed downstream in a second zone through homogeneous (O 2 -free) pyrolysis of unreacted fuel. These two reaction pathways are combined to accurately predict fuel conversion and * Portions of this chapter appear in Panuccio G.J., Williams K.A., and Schmidt L.D., Contributions of heterogeneous and homogeneous chemistry in the catalytic partial oxidation of octane isomers and mixtures on rhodium coated foams, Chemical Engineering Science, 61, (26). 26 Elsevier Science Ltd. 55

71 product distribution for all three fuels studied at total flow rates from 2 to 6 standard liters per minute (SLPM) on 45 and 8 pores per linear inch (ppi) supports. 4.2 Simulation Method One motivation for performing these simulations is to demonstrate that the catalytic partial oxidation of large alkanes can be accurately described through the combination of heterogeneous and homogeneous chemistry. This is accomplished by decoupling the surface and gas-phase chemistry and modeling them independently in a two-zone model. The hypothesis is that heterogeneous chemistry accounts for the production of H 2, CO, H 2 O, and CO 2 through surface reactions between absorbed fuel and O 2. Any fuel that does not react on the surface utilizes the heat generated by the exothermic heterogeneous chemistry to crack into smaller alkanes and alkenes through oxygen-free gas-phase pyrolysis. The overall process is modeled by separating the reactor into a zone of catalytic activity on the front face followed by a region of homogeneous reaction downstream (Figure 4-1). The catalytic foam monolith is modeled as a single pore plug flow reactor for 8 ppi (average pore diameter is.25 cm) and 45 ppi (d p =.47 cm) monoliths. The highly interconnected pore geometry of the foam monolith increases radial mixing and makes the plug flow model acceptable Heterogeneous approximation While elementary-step heterogeneous reaction mechanisms have been developed for the reaction of methane [9-11] and ethane [2, 6, 7, 12] on noble metal catalysts, no mechanism has been proposed that can correctly predict species selectivity and ignition behavior for the CPO of heavier fuels like n-octane and i-octane. In this work, the heterogeneous portion of this two-zone model is approximated. Several different methods were explored to approximate the front (heterogeneous) zone of the model. First, the surface chemistry was approximated by calculating the amount of each species predicted by thermodynamic equilibrium when H 2, H 2 O, CO, CO 2, and n-c 8 H 18 were allowable products. This calculation predicted too much H 2 and CO and not enough H 2 O and CO 2. Another approach was to approximate the surface chemistry by using a set of 56

72 global psuedo-homogeneous reactions. This technique predicted species mole fractions that were closer to the experimental values, but the predicted mole fractions did not vary as strongly with inlet C/O as the experimentally observed values did. Since these methods did not sufficiently reflect experimental trends and did not lend any further insight into the kinetics of the surface phenomena, it was decided to simply approximate the heterogeneous portion of the model directly with the experimental data. It was observed in our simulations that the combustion and partial oxidation products produced in the heterogeneous section (the first zone) did not react homogeneously in an O 2 -free environment. Therefore, the molar flows of these species after the heterogeneous section should be the same as the flows at the exit of the catalyst (see Figure 4-1). The flow rates (F i ) of H 2, CO, H 2 O, and CO 2 entering the homogeneous section of the model can be calculated from the observed experimental selectivity (S i ) of that species, fuel conversion (X) at the reactor outlet, the flow of fuel (F Fuel, ) at the reactor inlet, and the ratio of C (or H) atoms in the fuel and in species i (n F and n i, respectively) according to Equation 4-1: n F = S F X F i i Fuel, ni (4-1) The fuel flow rate entering the homogeneous section of the model is then calculated according to a carbon atom balance as shown in Equation 4-2: F = Fuel,1 nf - (F + F ) F Fuel, CO CO 2 n F (4-2) Here, FCO and F CO are the flow rates of carbon monoxide and carbon dioxide calculated 2 according to Equation 4-1. The fuel flow rate is calculated by the carbon balance and not the hydrogen balance because the experimental carbon balance error is generally smaller. 57

73 4.2.2 Detailed homogeneous model Several mechanisms are available to model the homogeneous reactions of octane isomers. In this study, mechanisms published by researchers at E.N.S.I.C. Department de Chimie Physique des Reactions at Universite de Nancy and by the combustion group at Lawrence Livermore National Laboratories (LLNL) are used. The Nancy group has published mechanisms for n-octane, i-octane, and n-heptane:i-octane (1:1) [13, 14]. Similarly, the LLNL group has published mechanisms for i-octane and n-heptane:ioctane (1:1) [15] (but no mechanism for n-octane). The n-octane:i-octane (1:1) experiments are modeled using the n-heptane:i-octane mechanisms because an n- octane:i-octane mechanism is not available, and n-octane and n-heptane should behave similarly in the gas phase. In this case, the calculated molar flow of n-octane (Equation 4-2) used the homogeneous section of the model is treated as n-heptane. The homogeneous zone of the reactor is modeled using the Plug subroutine in Chemkin 3.7 software [16]. Solution time depended on the number of reactions in the gas-phase mechanism and typically varied from a few seconds to a few minutes. An isothermal temperature profile is applied to the homogeneous section of the reactor at the experimental back-face value. This is done because the temperature profile and the heat loss characteristics of the catalyst are not well known. The Plug subroutine in Chemkin uses Differential-Algebraic System Solver (DASSL) to integrate the coupled differentialalgebraic equations and outputs species mole fractions as a function of axial distance [17, 18]. When the simulation is completed, the predicted fuel conversion and olefin selectivities at the reactor outlet are calculated and compared to their corresponding experimental values. 4.3 Simulation Results In these simulations, the heterogeneous and homogeneous chemistry are decoupled and modeled independently in series as described previously. It is assumed that all H 2, H 2 O, CO, and CO 2 are generated through surface reactions and that the unreacted fuel utilizes the heat generated by the surface reactions to pyrolize into smaller olefins and alkanes in the gas-phase. The predicted and experimental selectivities of H 2, 58

74 CO, H 2 O, and CO 2 for the CPO of n-octane at 4 SLPM on 8 ppi washcoated catalyst are plotted in Figure 4-2 for the different methods of heterogeneous approximations. This shows that using the experimental data for the heterogeneous approximation best predicts the syngas and combustion product selectivities and the correct amount of fuel that does not react on the surface that is left over to react in the gas phase. This approximation works because when there is little or no oxygen in the homogeneous zone, the syngas and combustion products do not react homogeneously (according to the gas-phase mechanisms used), so the simulated selectivities of those species match exactly with the experimental values used as inputs. Figure 4-3 shows the fuel conversion and C 2 through C 4 olefin selectivity predicted from the simulation of n-octane CPO at nominal flow rates of 2, 4, and 6 SLPM. The n-octane mechanism developed by the Nancy group is used in the homogeneous plug flow portion of the model. This mechanism incorporates 1699 elementary reactions involving 146 species and was validated by comparing to experimental results obtained in the adiabatic n-octane oxidation in a PFR at 8 C [14]. The predicted fuel conversion matches well with the experimental values at low C/O for all flow rates. However, as the C/O increases, the predicted conversion is approximately 1% higher than the experimentally observed value at 4SLPM. Ethylene, propylene, and n-butylene selectivities are also plotted in Figure 4-3. The agreement between experiment and simulation is good irrespective of C/O and flow rate. The results obtained when modeling the 45 ppi experiments also match very well to experiment (data not shown). The homogeneous i-octane mechanisms developed by the Nancy group and LLNL are both used in the gas-phase plug flow portion of the simulation of i-octane CPO. The Nancy mechanism incorporates 2411 elementary reactions involving 473 species and was validated at C/O = 2. and 7 C [13]. The LLNL mechanism contains approximately 87 species and five thousand elementary reactions and is validated from 1-45 atm and 3 to 14 C [15]. Figure 4-4 displays the fuel conversion and C 2 through C 4 olefin selectivity predicted from the simulation of the CPO of i-octane at 6 SLPM. Both mechanisms accurately predict the final conversion of i-octane, but the 59

75 conversion predicted by the LLNL mechanism is 15% higher than Nancy (7% vs. 55%) at C/O = 2.. The LLNL mechanism matches the selectivity of ethylene, propylene, and i-butylene almost exactly to experiment while the Nancy mechanism prediction is too high for ethylene and too low for propylene and i-butylene. No mechanisms are available for n-octane:i-octane (1:1) mixtures, however both the Nancy group and LLNL have developed a mechanism for n-heptane:i-octane (1:1) because they are a primary reference fuel for gasoline. These mechanisms can be used to model the mixture of n-octane/i-octane because n-heptane and n-octane should behave similarly under the operating conditions in this study. The simulated conversions of n- octane and i-octane match very well with experiment for both mechanisms (not shown). Figure 4-5 compares experimental and predicted olefin selectivity vs. C/O for the CPO of n-octane:i-octane (1:1) on 45 and 8 ppi Rh-coated foam monoliths at 4 SLPM. The agreement between experiment and model on both monoliths using either the Nancy or LLNL mechanism is very good. The modeling results mirror the experimental trends with ethylene and propylene displaying a maximum in selectivity between C/O = 1.1 and 1.5 while i-butylene levels off at C/O = Discussion: Relaxing the 2-zone model The results from the homogeneous modeling indicate that the gas-phase mechanisms compiled by both the Nancy and LLNL groups accurately describe olefin formation in the gas phase for n-octane, i-octane, and the 1:1 mixture in short contact time reactors. Specifically, the Nancy mechanism clearly describes the gas-phase n- octane reactions while the LLNL mechanisms best approximate the homogeneous activity for i-octane and the mixture. They predict the correct selectivity and the prevailing trends by generating maxima for ethylene and propylene at appropriate C/O ratios. The mechanisms are also fairly robust in that they accurately predict olefin selectivities from experiments performed with all three fuels at varying flow rates and on different catalyst supports. The agreement between experiment and model also suggests that the hypothesis that fuel reacts catalytically with O 2 to generate syngas, combustion products and heat 6

76 followed by homogeneous pyrolysis of unreacted fuel has some merit. Simulations that only use gas-phase chemistry suggest that the catalyst has a pivotal role in the CPO process and that gas-phase reactions alone cannot predict all species selectivities accurately. A one-zone homogeneous simulation (no initial heterogeneous section) was performed with the Nancy n-octane mechanism. These computations predicted H 2 O and CO 2 concentrations that are higher than experimental values and H 2 and CO selectivities that are much lower than those observed from experiment. Consequently, more fuel remains to form olefins, and predicted olefin yields are much higher than those obtained from experiment. This suggests that CPO can not be explained solely through gas-phase reaction pathways and that the catalyst is an important factor in the resulting product distribution. A more thorough model of the CPO of higher alkanes will relax the 2-zone scheme and allow for coupled heterogeneous and homogeneous chemistry. A heterogeneous reaction mechanism for octane on noble metals is under development and will be a modification of the existing mechanisms for methane and ethane. It can be concluded from the experimental results that the mechanisms for n-octane and i-octane should not be very different as they generate similar selectivity to partial oxidation and combustion products. The results of the simulations presented here show that if a surface mechanism can be developed to predict experimentally observed selectivities of H 2, H 2 O, CO, and CO 2 through heterogeneous chemistry, then that surface mechanism could be combined with gas phase mechanisms developed by groups in Nancy and LLNL to accurately describe the CPO of heavy alkanes like octane over a wide range of experimental conditions. 4.5 Conclusions The catalytic partial oxidation process is simulated by decoupling the heterogeneous and homogeneous chemistry into two separate zones and modeling the reactor as a PFR. Simulations agree with experiment and show that the formation of olefins can be explained through the homogeneous pyrolysis of unreacted fuel if H 2, CO, H 2 O, and CO 2 are produced heterogeneously. Mechanisms compiled by groups in Nancy, 61

77 France and LLNL accurately predict olefin selectivity resulting from the partial oxidation of n-octane, i-octane, and an equimolar mixture of n-octane and i-octane at flow rates between 2 and 6 SLPM on 8 and 45 ppi Rh-coated ceramic foam monoliths. References [1] Huff M, Schmidt LD. Ethylene formation by oxidative dehydrogenation of ethane over monoliths at very short contact times. J. Phys. Chem. 1993; 97: [2] Huff M, Schmidt LD. Elementary step model for the partial oxidation of ethane on Pt coated monoliths. AIChE Journal. 1996; 42: [3] Beretta A, Ranzi E, Forzatti P. Experimental and theoretical investigation on the roles of heterogeneous and homogeneous phases in the oxidative dehydrogenation of light paraffins in novel short contact time reactors. Studies in Surface Science and Catalysis. 2; 13B: [4] Beretta A, Ranzi E, Forzatti P. Production of olefins via oxidative dehydrogenation of light paraffins at short contact times. Catalysis Today. 21; 64: [5] Beretta A, Ranzi E, Forzatti P. Oxidative dehyrdrogenation of light paraffins in novel short contact time reactors. Experimental and theoretical investigation. Chemical Engineering Science. 21; 56: [6] Huff M, Androulakis IP, Sinfelt JH, Reyes SC. The contribution of gas-phase reactions in the Pt-catalyzed conversion of ethane-oxygen mixtures. Journal of Catalysis. 2; 191: [7] Donsi F, Williams KA, Schmidt LD. A multistep surface mechanism for ethane oxidative dehydrogenation on Pt- and Pt/Sn-Coated Monoliths. Industrial and Engineering Chemistry Research. 25; 44: [8] Donsi F, Caputo T, Russo G, Di Benedetto A, Pirone R. Modeling ethane oxydehydrogenation over monolithic combustion catalysts. AIChE Journal. 24; 5: [9] Hickman DA, Schmidt LD. Steps in CH 4 oxidation on Pt and Rh surfaces: Hightemperature reactor simulations. AIChE Journal. 1993; 39:

78 [1] Mhadeshwar AB, Vlachos DG. Hierarchical multiscale mechanism development for methane partial oxidation and reforming and for thermal decomposition of oxygenates on Rh. J. Phys. Chem. B. 25; 19: [11] Schwiedernoch R, Tischer S, Correa C, Deutschmann O. Experimental and numerical study on the transient behavior of partial oxidation of methane in a catalytic monolith. Chemical Engineering Science. 23; 58: [12] Zerkle DK, Allendorf MD, Wolf M, Deutschmann O. Understanding homogeneous and heterogeneous contributions to the platinum catalyzed partial oxidation of ethane in a short-contact-time reactor. Journal of Catalysis. 2; 196: [13] Come GM, Warth V, Glaude PA, Fournet R, Battin-Leclerc F, Scacchi G, 26th (Int.) Symposium on Combustion, The Combustion Institute, 1997, pp [14] Glaude PA, Warth V, Fournet R, Battin-Leclerc F, Scacchi G, Come GM. Modelling of the oxidation of n-octane and n-decane using an automatic generation of mechanisms. Int. J. Chem. Kin. 1998; 3: [15] Curran HJ, Gaffuri P, Pitz WJ, Westbrook CK. A comprehensive modeling study of iso-octane oxidation. Combustion and Flame. 22; 129: [16] Kee RJ, Rupley FM, Miller JA, Coltrin ME, Grcar JF, Meeks E, Moffat HK, Lutz AE, Dixon-Lewis G, Smooke MD, Warnatz J, Evans GH, Larson RS, Mitchell RE, Petzold LR, Reynolds WC, Caracotsios M, Stewart WE, Glaborg P, Wang C, Adijun O, Houf WG, Chou CP, Miller SF. Chemkin Collection, Release 3.7.1, Reaction Design, Inc., San Diego, CA (23) [17] A Fortran Program for the Analysis of Plug Flow Reactors with Gas-Phase and Surface Chemistry. Sandia National Laboratories Report. [18] A description of DASSL. Sandia National Laboratories Report. 63

79 Quartz Tube Catalytic Monolith Reactants Products Pore Heterogeneous Approx. Homogeneous Chemkin PFR Simulation 1 mm 9 mm F F 1 F 2 N 2 Fuel O 2 N 2 Fuel H 2 H 2 O CO CO 2 d p u z Species flows are equal N 2 Fuel H 2 H 2 O CO CO 2 CH 4 C 2 H 4...etc. Figure 4-1. Schematic of the catalytic foam model. Fuel reacts with O 2 to form H 2, H 2 O, CO, and CO 2 in the heterogeneous section. These species are inert in the homogeneous (O 2 -free) portion of the model so the molar flows at point 2 are the same as point 1. Unreacted fuel from the heterogeneous section undergoes pyrolysis in the homogeneous section to form olefins and other hydrocarbons. 64

80 1 (A) Selectivity (%) H 2 CO 2 H 2 O CO C/O (B) Selectivity (%) H 2 O H 2 CO 2 CO Selectivity (%) H 2 C/O H 2 O CO 2 (C) CO C/O Figure 4-2(A-C). Results comparing the predicted and expeirmental selectivities of H 2, CO, H 2 O, and CO 2, for the heterogeneous approximation using (A) equilibrium, (B) a set of global psuedo-homogeneous reactions, and (C) experimental data for the CPO of n- octane on 8 ppi w/ washcoat catalysts at 4 SLPM. Lines are the predicted species selectivities at the end of the PFR simulation that result from the different heterogeneous approximations. 65

81 Conversion (%) Selectivity (%) SLPM 6 SLPM 4 SLPM 4 2 n-octane Conversion C/O 25 4 SLPM 2 15 C 2 H 4 1 C 3 H 6 5 n-c 4 H C/O Selectivity (%) Selectivity (%) 25 2 SLPM 2 15 C 2 H 4 1 C 3 H 6 5 n-c 4 H C/O 25 6 SLPM 2 C 2 H 4 15 C 3 H n-c 4 H C/O Figure 4-3. Comparison of experimental and predicted fuel conversion and olefin selectivity vs. C/O for n-octane CPO on 8 ppi 5% Rh, 5% γ-alumina catalysts at 2, 4, and 6 SLPM. The Nancy n-octane homogeneous mechanism is used for this simulation[14]. There is good agreement between predicted and experimental olefin selectivities and fuel conversion at all three flow rates. 66

82 Conversion (%) 1 LLNL 8 6 Nancy 4 2 i-octane Conversion C/O Selectivity (%) 1 C 2 H 4 8 Nancy 6 LLNL C/O Selectivity (%) C 3 H 6 LLNL Nancy Selectivity (%) i-c 4 H 8 LLNL Nancy C/O C/O Figure 4-4. Comparison of experimental and predicted i-octane conversion and olefin selectivity vs C/O for i-octane CPO on 8 ppi 5% Rh, 5% γ-alumina monoliths at 6 SLPM. Both the LLNL [15] and Nancy [13] i-octane homogeneous mechanisms are used for this model. The Nancy mechanism works better for conversion while the LLNL mechanism more accurately predicts experimentally observed selectivities. 67

83 25 45 ppi 8 ppi 2 Selectivity (%) Selectivity (%) 2 15 LLNL 1 Nancy C/O LLNL Nancy Selectivity (%) Selectivity (%) C 2 H 4 C 2 H C/O LLNL Nancy LLNL Nancy C/O 25 2 C 3 H 6 C 3 H C/O Selectivity (%) LLNL Nancy Selectivity (%) 15 1 LLNL Nancy C/O i-c 4 H 8 i-c 4 H Figure 4-5. Comparison of experimental and simulated olefin selectivity for n-octane:ioctane (1:1) CPO on 45 (left column) and 8 ppi (right column) Rh-coated catalysts at 4 SLPM. Both the LLNL [15] and Nancy [13] homogeneous mechanisms for n-heptane:ioctane (1:1) are used in the simulations. The LLNL mechanism more accurately predicts experimentally observed selectivities. 68 C/O

84 Chapter 5 Species and temperature profiles in a differential sphere bed reactor for the catalytic partial oxidation of n-octane 5.1 Introduction Understanding the chemistry of the CPO system is paramount to the design of an efficient reactor that can produce the highest yields of the desired products. In Chapter 4, it was shown that the CPO of octane isomers and mixtures can be successfully modeled by combining heterogeneous and homogeneous chemistry [1]. Specifically, the formation of olefins can be explained through the homogeneous pyrolysis of the hydrocarbon fuel when H 2, H 2 O, CO, and CO 2 are formed through heterogeneous reactions. However, the nature of the surface reactions could not be fully addressed in the simulation methods. For the work in this Chapter, an experiment has been designed by which the mechanism of the surface reactions can be explored by inspecting the temperature and species flow rate profiles within the catalyst bed. There is a long standing debate on the nature of the heterogeneous chemistry in the catalytic partial oxidation process. Some argue that the formation of H 2 and CO is through a direct mechanism whereby hydrocarbon fuel adsorbs on the catalyst surface, pyrolyzes into adsorbed C and H atoms that then recombine with adsorbed O atoms to form a combination of H 2, H 2 O, CO, and CO 2 depending on the overall stoichiometry of feed stream. Others argue that the formation of syngas is through an indirect route where the hydrocarbon is first completely oxidized to H 2 O and CO 2 at the front of the catalyst and then undergoes steam and dry reforming to produce H 2 and CO downstream. Reviews on both the experimental and simulation literature for methane CPO have been recently done that show that arguments can be made for both mechanisms [2, 3]. Most recently, a technique has been developed which can measure temperature and species flow rate profiles inside a Rh-coated catalytic foam monolith for the CPO of methane using a moveable direct-sampling capillary that is connected to a mass spectrometer for 69

85 species analysis [2]. Spatial temperature and flow rate measurements can offer further insight into the chemistry of the reaction because it can be possible to see different reaction zones develop within the catalyst. These results show that syngas is formed through both direct and indirect means and that there is a significant amount of steam reforming inside the catalyst bed, but an insignificant amount of CO 2 reforming on Rh catalysts when methane is used as the reacting fuel. Similarly in this work, temperature and species flow rate profiles are measured within a catalytic sphere bed for the CPO of n-octane to examine the heterogeneous chemistry for heavier hydrocarbon fuels. A method is developed whereby the products can be analyzed with a gas chromatography instrument (GC) and profiles can be constructed by changing the length of the catalyst bed (see Section 5.2). The resulting temperature and flow rate profiles are analyzed and the mechanism of the heterogeneous reactions is discussed. Several different metals have been used to catalyze the partial oxidation reaction including Rh, Pt, Ni, Pd, and Ir. Previous results have shown that Rh and Pt give the best performance because Rh generates high selectivity to syngas while Pt is the best catalyst to use for olefin production [4-7]. The other metals do not perform as well as Rh or Pt for various reasons including coking, sintering, and inactivity [5, 6, 8]. In this work, the contribution of the catalyst metal to the chemistry of the CPO process is examined by measuring temperature and species flow profiles for both Rh and Pt catalysts. The final variable considered in this work is the effect of feed stoichiometry. Previous results on the CPO of n-octane give several general trends in relation to the C/O ratio in the reactant stream [1, 9]. The maximum selectivity of syngas is usually observed for C/O ratios slightly less than 1.. In this region, the fuel conversion is 1% and measured catalyst temperatures are approximately 1 C. As the amount of O 2 in the feed decreases (C/O increases), reactor temperatures fall to 8 or 9 C, fuel conversion decreases, syngas selectivity decreases, and the selectivities of ethylene and other olefins increase. Measuring the temperature and species flow rate profiles as a function of the inlet stoichiometry will offer some insight as to the changing chemistry in the reactor as the feed mixture becomes more fuel rich. 7

86 5.2 Experimental Experiments were carried out in a 19 mm ID (inside diameter) quartz tube at atmospheric pressure. The upstream portion of the reactor for this experiment was constructed in a similar fashion as has been described and sketched previously [1, Figure 2-1]. n-octane (> 99% purity) was introduced through a low flow automotive fuel injector at the top of the quartz tube and mixed with N 2 and O 2 (in air stoichiometry) that was introduced through a side port via electronically controlled mass flow controllers. The portion of the reactor tube in this section was wrapped in a Variac-controlled resistive heating tape that vaporized the thin film of fuel produced on the wall from the spray of the fuel injector. A thermocouple was inserted into the reactor after the vaporization and mixing zone and before the catalyst assembly to ensure that a constant vapor preheat temperature of ~ 16 C was maintained. The catalysts used in these experiments were low surface area 1.3 mm diameter α-alumina spheres that were coated with either Rh or Pt metal. The catalysts were prepared via dropwise addition of aqueous metal salt solution (Rh(NO 3 ) 3 or H 2 PtCl 6 ) directly onto the spheres. The total amount of solution was applied to the spheres over the course of several doses where the water was allowed to evaporate and the spheres were thoroughly mixed between applications. Enough solution was added to ensure that the catalysts would contain 5 wt% metal loading based on the initial mass of the spheres. The coated spheres were calcined in a closed furnace at 6 C for 5 hours. A sketch of the catalyst assembly is shown in Figure 5-1. A 1 mm ID 2 mm OD (outside diameter) quartz sampling tube was pushed through a septum, wrapped in Fiberfrax paper, and inserted into a side port in the quartz reactor so that the end of the sampling tube was in the middle of the reactor. The sampling tube was inserted through a septum in order to ensure that the reactor was sealed and was wrapped in Fiberfrax paper (not shown in the diagram) to ensure that none of the spheres from the catalytic bed fell into the side port in the reactor tube. Next, a K-type thermocouple and a 17 mm diameter, 1 mm length, 8 pores per linear inch (ppi), cylindrical α-alumina ceramic foam monolith were inserted through the bottom of the reactor. The thermocouple was positioned so that the tip was close to the end of the quartz sampling tube and the 71

87 monolith was wrapped in Fiberfrax paper to ensure a tight fit inside the reactor. To ensure that the sampling tube was at the very end of the catalytic bed, uncoated (blank) 1.3 mm alumina spheres were added from the top of the reactor until the sampling tube was half covered. Next, a specified mass of either Rh or Pt-coated 1.3 mm alumina spheres was added from the top of the reactor. Another blank 8 ppi monolith that was wrapped in Fiberfrax was pushed down through the top of the reactor tube onto the catalytic sphere bed in order to fix the bed in position and act as a heat shield to prevent axial radiative heat losses. The outside of the quartz tube was wrapped with insulation to prevent radial heat losses. Gases were sampled from the center of the very end of the active catalyst bed by inserting a gas-tight syringe into the 1 mm ID quartz sampling tube. Since the gas product samples were taken at the very end of the catalytic bed, a differential reactor profile could be generated as a function of the mass of the bed by simply changing the mass of the Rh or Pt coated spheres inside the quartz reactor tube. The bed length was varied from a single monolayer of active spheres (~.5 g) to 1 mm of spheres (~ 4.5 g). The insert in the top left panel of Figure 5-2 shows the approximate length of the bed as a function of the mass. When the mass of the catalyst bed was changed, the used catalyst was discarded. The reactor was rebuilt with fresh catalyst that was aged by reacting fuel and air on the fixed bed for approximately an hour before new data was obtained. The product gases were analyzed by injecting them into an HP 589 GC that was fit with a single GS Gas Pro 6 m capillary column and a TCD detector. The GC oven was cooled with liquid nitrogen so that all species from H 2 to n-octane could be separated on a single column and analyzed with a single detector. The nitrogen from air was used as an internal standard to calculate the flows of all other species present. Experiments were performed at 4 standard liters per minute (SLPM) total flow rate of fuel and air. At typical reactor temperatures of 9 C, this corresponded to an average residence time of 5 ms at the full 1 mm of catalyst bed length assuming 4% bed porosity [11, 12]. The flow rate of n-octane into the tube was set by the desired experimental C/O ratio which was defined as the ratio of the moles of carbon in the feed from n-octane to the moles of oxygen in the feed from O 2. By this definition, the 72

88 stoichiometric C/O ratio for the partial oxidation reaction was 1.. Experiments were performed at C/O ratios of.7, 1., 1.5, and 2. so that profiles could be obtained over the entire range of stoichiometries that are typical for CPO experiments. 5.3 Results For each catalyst weight (W cat ), the data points presented in this work are the average of three experiments that were performed on a single catalyst bed that had a total time on stream between one and seven hours. In the following section, temperature and species flow rate profiles are presented as a function of mass of the catalyst bed, catalyst metal, and feed stoichiometry. The flow rates that are plotted at g catalyst are the values at the inlet of the quartz tube reactor Temperature The measured temperature within the catalyst bed is plotted as a function of the mass of the bed, the catalyst metal, and the feed stoichiometry in Figure 5-2. For both Rh and Pt catalysts, the reactor temperature decreases with increasing feed C/O at all points in the catalytic bed. At 4.5 g of Rh catalyst, the measured temperature decreases from 11 C to 77 C as the feed C/O increases from.7 to 2.. Reactor temperature decreases because the amount of O 2 and heat generated from exothermic oxidation reactions decrease with increasing C/O. The temperature profiles within the Rh and Pt sphere beds are different. The difference between the maximum observed temperature inside the bed and the temperature at the end of the bed is on average 1 C greater for Rh catalysts than for Pt. For example, at C/O = 1.5, the measured temperature of the Rh catalyst decreases 18 C from 99 C at.9 g catalyst to 81 C at 4.5 g catalyst. At the same C/O ratio in the Pt bed, the measured temperature only decreases 7 C from 93 C at 1.3 g catalyst to 86 C at 4.1 g catalyst. The maximum temperature in the bed is always greater for Rh than for Pt, but the temperature at the end of the bed is always higher for Pt than for Rh. Also, the maximum temperature for the Pt catalyst is always further downstream in the catalytic bed for Pt than for Rh. At C/O = 1., the maximum temperature in the Rh 73

89 catalyst bed of 116 C is observed at.9 g catalyst, but the corresponding maximum temperature of 17 C is observed at 1.3 g of Pt catalyst Reactant profiles The flow rate of the reactant n-octane is plotted as a function of the catalyst depth, the catalyst metal, and the feed stoichiometry in Figure 5-3. The flow rate of fuel at the end of the bed increases (and the fuel conversion therefore decreases) with increasing C/O ratio for both Rh and Pt coated spheres. This agrees with previous results observed for the CPO of n-octane on noble metal coated ceramic foam monoliths [1] and occurs because the extent of exothermic oxidation reactions and the reactor temperature decrease as the reactant mixture becomes more fuel rich. It should be noted that the flow rate of the reactant fuel does not decrease with increasing catalyst mass past ~ 1 g of catalyst. As we will show later, the bulk of the reforming reactions occur after ~ 1 g of catalyst and therefore implies that the heavier molecular weight fuel is not reformed in this zone of the catalytic bed. The flow rate of O 2 is shown as a function of catalyst mass, catalyst metal, and feed stoichiometry in Figure 5-4. After the full length of catalyst bed, the conversion of O 2 is 1% for both Rh and Pt catalysts at all C/O ratios. For reactions on Rh-coated spheres, all the O 2 is consumed before the end of the first monolayer of catalyst (~.5 g Rh catalyst) for C/O 1.5. Some O 2 breakthrough is observed after.5 g Rh catalyst at C/O = 2., but all the O 2 is fully converted by ~ 1 g of catalyst. For reactions on Ptcoated spheres, O 2 breakthrough is observed at all C/O ratios at small bed mass. The amount of O 2 remaining after 1 monolayer of spheres increases with increasing C/O, but is fully converted after ~ 1 g catalyst for C/O 1.5. At C/O = 2., O 2 breakthrough is observed until the mass of the catalyst bed reaches 1.3 g H 2 and H 2 O The flow rates of H 2, H 2 O, CO, and CO 2 are shown in Figures as a function of catalyst mass, catalyst metal, and feed stoichiometry. The calculated equilibrium flow rates of these species are plotted in these figures as well (heavy dashed 74

90 lines). These values were calculated at every point in the catalyst using the average of the experimentally observed Rh and Pt-coated sphere bed temperature at that point (Figure 5-1). A constant temperature (experimentally observed) and pressure (p = 1 atm) equilibrium calculation was performed at each point in the bed using the EQUIL subroutine in the CHEMKIN software package [13] with N 2, O 2, and n-c 8 H 18 as the initial species (in the appropriate stoichiometry according to the experimental C/O ratio), and allowing for N 2, O 2, H 2, H 2 O, CO, CO 2, and n-c 8 H 18 as equilibrium products. As shown in Figure 5-5, the flow rate of H 2 reaches the equilibrium value at C/O =.7 on Rh catalysts within the first monolayer of spheres and at 1 g of catalyst on Pt spheres. For C/O 1., the flow rate of H 2 in the catalytic bed is higher over Rh catalysts than Pt at all points in the bed. The H 2 flow rate reaches the equilibrium value at all C/O ratios for Rh experiments, but does not reach equilibrium for experiments performed with Pt catalysts for C/O > 1.. These plots show that the rate of formation of H 2 is significantly faster on Rh catalysts than on Pt catalysts. The experimentally observed flow rate of H 2 O over Rh and Pt catalysts is plotted as a function of catalyst mass and feed stoichiometry in Figure 5-6 along with the calculated equilibrium flow rate. The observation of a maximum in the water flow rate at a point inside the catalytic bed indicates that there is a zone of net production of H 2 O that is followed by a zone of net consumption of H 2 O. Comparison of Figures 5-4 and 5-6 shows that the transition from net production to net consumption of H 2 O corresponds very closely to the point in the bed where the O 2 flow rate first approaches zero. At C/O =.7 and 1., the flow rate of H 2 O reaches the equilibrium value on both Rh and Pt but gets there faster on the Rh catalysts. For C/O > 1., the flow rate of H 2 O still reaches equilibrium inside the Rh catalyst bed but does not reach equilibrium in the Pt bed. The average conversion of H 2 O from the peak flow rate to the end of the catalyst bed for C/O 1. is 99% for Rh catalysts but only 78% for reactions carried out over Ptcoated spheres. 75

91 5.3.4 CO and CO 2 The experimentally observed flow rate of CO over Rh and Pt-coated spheres is plotted as a function of catalyst mass and feed stoichiometry and compared to the calculated equilibrium flow rates in Figure 5-7. The flow rate of CO reaches the equilibrium value at C/O =.7 within the first monolayer of Rh-coated spheres and after approximately 1 g of Pt coated spheres. For C/O 1., the flow rate of CO is greater for Rh than for Pt at all points inside the catalytic bed. However, unlike H 2 the flow of CO never quite reaches equilibrium for C/O 1. on either Rh or Pt. Furthermore, the difference between the observed flow rate and the calculated equilibrium value increases with increasing C/O for both Rh and Pt catalysts. The experimental flow rates and equilibrium flow rates of CO 2 are plotted as a function of catalyst mass, catalyst metal, and feed stoichiometry in Figure 5-8. Similar to H 2 O, the profile of CO 2 flow rate has a maximum for both Rh and Pt catalysts that closely corresponds to the catalyst depth at which point the O 2 flow rate first approaches zero (Figure 5-4). This also indicates that there is a net production of CO 2 in the region of the catalyst bed where O 2 is present and a net consumption of CO 2 after all the O 2 is converted. Similar to H 2 O, the experimental flow rate of CO 2 reaches the calculated equilibrium value after ~ 1 g of catalyst for both Rh and Pt at C/O =.7. However, unlike H 2 O, the flow rate of CO 2 never quite reaches the calculated equilibrium value for Rh or Pt at C/O 1. and the difference between the observed flow and the equilibrium flow at the end of the catalyst bed increases with increasing C/O. The average conversion of CO 2 from the peak flow rate to the end of the catalytic bed is 62% for Rh and 5% for Pt for C/O 1.. These values are significantly lower than the corresponding values for H 2 O conversion on the same catalysts Low molecular weight hydrocarbon profiles Figure 5-3 shows that there is not much conversion of fuel after ~ 1 g of catalyst. The results in Figures show that there is a significant amount of reaction occurring in the catalyst bed downstream of 1 g that is producing H 2 and CO and 76

92 consuming H 2 O and CO 2. Since n-octane is not being consumed with the water and CO 2, it must be smaller hydrocarbons that are being reacted. The flow rate of C 2 H 4 is plotted as a function of catalyst mass, catalyst metal, and feed stoichiometry in Figure 5-9. The peak in the C 2 H 4 profile indicates that ethylene is first produced in the upstream zone of the catalytic bed and then reformed downstream along with the water and the CO 2. The location of the maximum in the flow rate of ethylene closely corresponds with the location of the maxima in the flow rates of CO 2 and H 2 O. The peak in the ethylene flow rate is higher and also further downstream in the catalytic bed for Pt than for Rh. The profiles for other smaller hydrocarbons are very similar and are not shown. The flow rate of C 2 H 4 is the largest in magnitude. The magnitude of the flow rates of the lower molecular weight hydrocarbons progresses as C 2 H 4 > CH 4 > C 3 H 6 > C 4 H 8 with the flow rate of C 4 H 8 approximately one order of magnitude less than C 2 H 4 flow rate. 5.4 Discussion Quantitative interpretation of profile data The results of these experiments qualitatively indicate that there are two distinct regions of activity within the catalyst bed. There is a short oxidation zone at the front of the catalyst that is followed by a reforming zone downstream. In order to numerically confirm these zones, the net rates of production of various species can be analyzed as a function of position in the reactor. The rate of formation of species A at W i grams of catalyst ( r A W i ) is defined as the derivative of the flow rate of species A (F A ) with respect to catalyst mass at W i grams of catalyst. This can be approximated by using the central difference formula for the interior points in the bed as in Equation 5-1. Here the units for r A are mol min -1 gcatalyst -1. r A Wi df F - F dw W - W A A W A i+1 Wi-1 Wi i+1 i-1 (5-1) 77

93 Equation 5-1 is evaluated at the left hand side of the domain (W i = g) by substituting the values of W i for W i-1. Similarly, W i is substituted for W i+1 in Equation 5-1 for evaluation at the right hand side of the domain (W i = 4.5 g). Since H 2 O, CO 2, CH 4, C 2 H 4, and other olefins are being consumed in the downstream portion of the reactor bed, it is most likely that steam or CO 2 reforming reactions are occurring as shown in Equations 5-2 and 5-3. Steam reforming: CH 4 + H2O CO + 3H2 rco -r H2O = 1 CXH 2X + xh2o xco + 2xH2 H2 H2O 2 r -r 3 (5-2) CO 2 (Dry) reforming: CH 4 + CO2 2CO + 2H2 rco -r CO = 2 2 CXH 2X + xco2 2xCO + xh2 1 r -r 2 H2 CO2 (5-3) The ratio of the rate of formation of H 2 to the rate of consumption of H 2 O ( r -r ) is H2 H2O plotted as a function of catalyst mass in Figure 5-1 for C/O 1.. This is not calculated for C/O =.7 because the reactions reach equilibrium faster than the minimum resolution of the experiment. According to Equation 5-2, the ratio of these rates should be between 2 and 3 if the steam reforming reaction is dominating. The arrows and vertical lines in Figure 5-1 indicate the region in the catalyst bed that this ratio is observed and therefore the region where steam reforming is governing the overall reaction. The value of r -r is negative at the front of the catalyst bed because both H 2 and H 2 O are being H2 H2O formed in the oxidation zone. There is some noise in the H 2 and H 2 O flow profiles for the Pt catalyst (see Figures 5-5 and 5-6) that gets compounded in the calculation of the derivative and leads to a few outlying points in the top two panels in Figure 5-1. The ratio of CO production to H 2 O consumption, is approximately 1 in the same region that r -r, is also calculated (not shown) and the steam reforming reaction stoichiometry and further strengthens the argument for the 78 CO H2 H2O H2O r -r is between 2 and 3. This agrees with

94 domination of steam reforming in the overall chemistry in this region of the catalyst. The comparison of Figure 5-1 with Figures 5-4 through 5-9 shows that the part of the catalyst bed where steam reforming is dominating overlaps with the region where O 2 is no longer present and H 2 and CO are being formed while H 2 O, CO 2, and smaller hydrocarbons are being consumed. The calculated values for the ratio of the rates of formation of CO and H 2 to the consumption of CO 2 ( rco -rco and r 2 H -r 2 CO ) are much higher in the reforming region 2 than the stoichiometric values for Equation 5-3. This means that not enough CO 2 is being consumed to account for all the production of H 2 and CO and suggests that while some H 2 and CO are being formed from reactions with CO 2, that the bulk of the formation is from steam reforming. This is confirmed from the observation that the conversion of H 2 O from the peak value inside the bed to the end of the catalyst is significantly higher than the conversion of CO 2. For example, on Rh catalyst the average H 2 O conversion for C/O 1. is 99% while the average conversion of CO 2 is only 62% Rh vs. Pt catalysts Combination of the quantitative results from section and Figure 5-1 with the qualitative results observed in Figures 5-2 through 5-9 shows that there is a difference between Rh and Pt catalysts in the CPO system. Figure 5-4 shows that the oxidation reactions are faster on Rh than Pt because oxygen breakthrough is observed after a single monolayer of Pt catalyst at all C/O ratios but is not observed after a monolayer of Rh catalyst for C/O < 2.. Since the oxidation reactions are happening faster on Rh, the maximum temperature observed inside the bed is higher for Rh than for Pt. Rh is also a better reforming catalyst than Pt with respect to both the rate and the extent of reforming reactions. With the exception of C/O = 1., results show that the length of the reforming zone is shorter and therefore the rate of the reforming reactions is faster on Rh than on Pt. At C/O =.7, H 2, H 2 O, CO, and CO 2 flow rates reach equilibrium values within the first monolayer of Rh catalysts but not until 2 mm (~ 1 g) of Pt catalysts. Figure 5-1 shows that the length of the reforming zone for Rh is shorter than Pt at C/O = 1.5 and C/O = 2. also. The results displayed in Figure 5-9 also suggest 79

95 that the reforming reactions are slower on Pt because the flow rate of ethylene is higher and the peak in the ethylene flow rate is further downstream on Pt than on Rh. This trend is observed for the other reformed smaller hydrocarbons like methane, propylene, butylene, etc. The extent of the reforming is higher on Rh than on Pt as well. For C/O 1., the average conversion of H 2 O from the internal maximum to the end of the bed is 99% for Rh but only 78% for Pt catalyst. Similarly, the conversion of CO 2 is 62% for Rh and only 5% for Pt. Since there is a larger extent of endothermic reforming reactions on Rh, the average temperature drop from the maximum inside the bed to the value at the end of the bed is approximately 1 C higher for Rh than for Pt. This also causes the measured temperature at the exit of the bed to be higher for Pt than for Rh Effect of feed stoichiometry Temperature and species profiles are measured for inlet C/O ratios ranging from.7 to 2. in order to examine the effect of stoichiometry on the chemistry in the CPO system. Both the length of the oxidation zone and the reforming zone increase as the C/O ratio increases and the feed becomes more fuel rich. Figure 5-4 shows that the flow rate of O 2 after 1 monolayer of Pt catalyst increases with increasing C/O and O 2 breakthrough is not observed for C/O < 2. on Rh catalyst after a single monolayer. The length of the reforming zone increases with increasing C/O as well. The exception is at C/O = 1.. At C/O =.7, the flow rates of H 2, H 2 O, CO, and CO 2 reach equilibrium within a single monolayer of Rh catalyst and at ~ 2 mm (~ 1 g) of Pt catalyst (Figures ). The analysis in section and the results displayed in Figure 5-1 show that the length of the reforming zone on Rh catalyst increases with increasing C/O from.5 g at C/O =.7 to 2. g at C/O = 1.5 to 3.5 g at C/O = 2.. A similar trend is observed on Pt catalysts. This is an expected observation because decreasing the O 2 in the feed decreases the extent of exothermic oxidation chemistry and the temperature in the catalyst. The decrease in temperature decreases the rate of the endothermic reforming reactions which increases the length over which they take place and decreases the extent to which they occur. Therefore, the flow rates of H 2 and CO should decrease and the 8

96 flows of H 2 O, CO 2, and smaller hydrocarbons should increase at all points in the catalyst bed for increasing feed C/O. An effect other than the catalyst bed temperature is governing the reforming reactions at C/O = 1. and causes the reforming zone length to be longer than expected Reformation of smaller hydrocarbons A schematic of the suggested overall reaction is given in Figure In the presence of O 2, octane reacts to form H 2 O, CO 2 and some CO and H 2 or it breaks apart into smaller hydrocarbons like methane, ethylene, propylene, etc. The smaller hydrocarbons then react with H 2 O and CO 2 in reforming reactions to make more H 2 and CO. The profile data suggests that the smaller hydrocarbons and not octane are reformed with H 2 O and CO 2 because their flow rates decrease in the downstream section of the reactor while the flow rate of octane is constant after the O 2 is consumed. While there is not much literature available to confirm that olefins can be efficiently steam reformed on Pt or Rh catalysts, Resini and coworkers have shown that the direct steam reforming of propylene generates high selectivity to H 2 and CO with high conversions of C 3 H 6 on Pd- Cu/γ-Al 2 O 3 catalysts [14]. Perhaps smaller hydrocarbons like methane and ethane are more likely to be reformed than large hydrocarbons like octane because of the stoichiometric hindrances of high molecular weight reactants. The higher the molecular weight of the reforming species, the more water or CO 2 it is going to take to reform it. Therefore, octane must have to break down into more stoichiometrically manageable sizes before it can be efficiently reformed into H 2 and CO Comparison of sphere bed to monolith experiments Figure 5-12 compares the product selectivities that result from of the CPO of n- octane in air at 4 SLPM after 1 mm of Rh catalyst on different supports. The current results for a fixed bed of 1.3 mm α-alumina spheres are compared to previous results obtained from reactions on 8 ppi (γ-alumina washcoated) and 45 ppi (unwashcoated) α- alumina foam monoliths [1, Chapter 3]. The selectivities of H 2, H 2 O, CO 2, and C 2 H 4 are plotted vs. feed C/O ratio for all three supports. The important parameter to consider 81

97 when comparing these supports is the relative amount of heterogeneous chemistry occurring on the support. The 8 ppi monoliths have an average pore diameter of.21 mm while the 45 ppi foams have average pore diameters of.45 mm [15]. The surface area to volume ratio (and therefore the relative amount of heterogeneous chemistry) is higher on the 8 ppi foams because they have a smaller pore diameter. Hohn and Schmidt have shown that sphere beds produce higher selectivity of syngas at higher conversion of fuel than foams in the CPO of methane [16] at atmospheric pressure. Other studies have shown that homogeneous chemistry is insignificant for the CPO of methane at these conditions [17], therefore the sphere bed must have higher heterogeneous activity than the foams. The supports should be ordered in terms of the relative extent of heterogeneous chemistry as 1.3 mm spheres > 8 ppi foams > 45 ppi foams. Accordingly, the selectivity of H 2 (and CO not shown) is highest for the sphere bed support and decreases with increasing pore diameter for the foam catalysts. The selectivities of H 2 O, CO 2, C 2 H 4, and other smaller hydrocarbons increase with decreasing surface activity because there is not enough active catalyst available to reform them into syngas. In terms of the schematic shown in Figure 5-11, a decrease in the active catalyst surface in the bed decreases the extent of the reaction of smaller hydrocarbons with H 2 O and CO 2 to form H 2 and CO because these reactions are slower and need longer catalyst contact times. Therefore, the concentrations of the intermediate species H 2 O, CO 2, and smaller hydrocarbons increase and the selectivities of H 2 and CO decrease for supports with decreasing catalyst activity. Also, the conversion of n-octane (not shown) is the same for all three supports after 1 mm of catalyst. This confirms that the low molecular weight hydrocarbons are being reformed downstream to make syngas since the selectivities of ethylene and other small hydrocarbons are lower for reactions on sphere beds than on foams but the fuel conversion is the same. This comparison shows that sphere beds are the optimum support if H 2 is the desired product of the CPO reaction. 5.5 Conclusions 82

98 For the first time, temperature and species flow profiles are obtained for the CPO of a high molecular weight hydrocarbon fuel on both Rh and Pt catalysts. The results show that there is first a very short oxidation zone in the front of the catalyst where fuel and O 2 and consumed to generate H 2 O, CO 2, CO, H 2, smaller hydrocarbons, and heat and a longer reforming zone downstream where heat, H 2 O, CO 2, and smaller hydrocarbons are consumed to form more H 2 and CO. Steam and CO 2 reforming of low molecular weight hydrocarbons is observed on both Rh and Pt catalysts with steam reforming occurring more extensively than CO 2 reforming. The flow rate of octane is constant in the reforming zone of the catalyst and it is not consumed in the reforming reactions. Rh produces higher flow rates of H 2 and CO than Pt because Rh is a better reforming catalyst. The stoichiometry of the reactor feed has an effect on the extent of reforming reactions and length of the reforming zone. Increasing the inlet C/O ratio increases the length until 1% O 2 conversion, decreases the temperature in the catalyst, decreases the extent of the reforming reactions, and increases the length of the reforming zone in the catalyst. These results show that metal-coated sphere beds produce higher selectivities of H 2 and CO and lower selectivities of H 2 O, CO 2, and olefins than metal-coated foam monoliths because the extent of heterogeneous chemistry is higher. References [1] Panuccio GJ, Williams KA, Schmidt LD. Contributions of heterogeneous and homogeneous chemistry in the catalytic partial oxidation of octane isomers and mixtures on rhodium coated foams. Chemical Engineering Science. 26; 61: [2] Horn R, Williams KA, Degenstein NJ, Schmidt LD. Syngas by catlytic partial oxidation of methane on rhodium: Mechanistic conclusions from spatially resolved measurements and numerical simulations. Journal of Catalysis. 26; 242: [3] Yorck APE, Xiao T, Green MLH. Brief overview of the partial oxidation of methane to synthesis gas. Topics in Catalysis. 23; 22: [4] Hickman DA, Haupfear EA, Schmidt LD. Synthesis gas formation by direct oxidation of methane over Rh monoliths. Catalysis Letters. 1993; 17:

99 [5] Torniainen PM, Chu X, Schmidt LD. Comparison of monolith-supported metals for the direct oxidation of methane to syngas. Journal of Catalysis. 1994; 146: 1-1. [6] Huff M, Torniainen PM, Schmidt LD. Partial oxidation of alkanes over noble metal coated monoliths. Catalysis Today. 1994; 21: [7] Huff M, Torniainen PM, Hickman DA, Schmidt LD. Partial oxidation of CH 4, C 2 H 6, and C 3 H 8 on monoliths at short contact times. Natural Gas Conversion II. 1994; [8] Liebman LS, Schmidt LD. Oxidative dehydrogenation of isobutane at short contact times. Applied Catalysis A. 1999; 179: [9] O'Connor RP, Klein EJ, Schmidt LD. High yields of synthesis gas by millisecond partial oxidation of higher hydrocarbons. Catalysis Letters. 21; 7: [1] Krummenacher JJ, West KN, Schmidt LD. Catalytic partial oxidation of higher hydrocarbons at millisecond contact times: decane, hexadecane, and diesel. Journal of Catalysis. 23; 215: [11] Jaeger HM, Nagel SR. Physics of the Granular State. Science. 1992; 255: [12] Torquato S, Truskett TM, Debendetti PG. Is random close packing of spheres well defined. Physical Review Letters. 2; 84: [13] Kee RJ, Rupley FM, Miller JA, Coltrin ME, Grcar JF, Meeks E, Moffat HK, Lutz AE, Dixon-Lewis G, Smooke MD, Warnatz J, Evans GH, Larson RS, Mitchell RE, Petzold LR, Reynolds WC, Caracotsios M, Stewart WE, Glaborg P, Wang C, Adijun O, Houf WG, Chou CP, Miller SF. Chemkin Collection, Release 3.7.1, Reaction Design, Inc., San Diego, CA (23) [14] Resini C, Arrighi L, Delgado MCH, Vargas MAL, Alemany LJ, Riani P, Berardinelli S, Marazza R, Busca G. Production of hydrogen by steam reforming of C3 organics over Pd-Cu/γ-Al 2 O 3 catalyst. International Journal of Hydrogen Energy. 26; 31: [15] Twigg MV, Richardson JT. Theory and applications of ceramic foam catalysts. Trans IChemE. 22; 8:

100 [16] Hohn KL, Schmidt LD. Partial oxidation of methane to syngas at high space velocities over Rh-coated spheres. Applied Catalysis A. 21; 211: [17] Goralski CT, O'Connor RP, Schmidt LD. Modeling homogeneous and heterogeneous chemistry in the production of syngas from methane. Chemical Engineering Science. 2; 55:

101 Reactant Flow Quartz Reactor 19 mm ID Blank 8 ppi Monoliths Metal-coated 1.3 mm α-al 2 O 3 spheres Sampling tube 1 mm ID 2 mm OD Blank 1.3 mm α-al 2 O 3 spheres Side port 4 mm ID Septum K-type Thermocouple Figure 5-1. Schematic of the catalyst assembly. The temperature is measured with a K- type thermocouple and the product gases are sampled with a gas-tight syringe through a quartz sampling tube at the center of the end of the Rh or Pt-coated sphere bed. Since temperatures and species flows are measured at the very end of the catalyst bed, differential temperature and species flow profiles are measured by varying the mass of the bed. 86

102 Rh Pt C/O =.7 C/O = Rh Pt 9 8 Approx. length of catalyst bed (mm) W cat (g) W cat (g) C/O = 1.5 C/O = Rh Pt Rh Pt W cat (g) W cat (g) Figure 5-2. Measured temperature as a function of the depth of the catalyst bed, the catalyst metal, and the feed stoichiometry. The insert in the upper left panel shows the approximate length of the catalyst bed as a function of the bed mass. The maximum temperature that is observed in the bed is higher for Rh than Pt and the change in temperature from the maximum to the value at the end of the bed is on average 1 C larger for Rh than Pt catalysts. The temperature in the bed decreases with increasing feed C/O ratio. 87

103 C/O =.7.16 C/O = Rh Pt.4 Pt.2.2 Rh W cat (g) W cat (g) C/O = C/O = Rh Pt Rh Pt W cat (g) W cat (g) Figure 5-3. Fuel (n-octane) flow rate profile as a function of the depth of the catalyst bed, the catalyst metal, and the feed stoichiometry. Fuel conversion at the end of the bed decreases with increasing feed C/O and is slightly higher for Pt than for Rh at C/O 1.5. The conversion of fuel stops after ~ 1 g of catalyst for both Rh and Pt. 88

104 C/O =.7.3 C/O = Rh Pt Pt.5 Rh W (g) W cat (g) cat C/O = C/O = Pt.1 Pt.5 Rh.5 Rh W cat (g) W cat (g) Figure 5-4. Oxygen flow rate profile as a function of the depth of the catalyst bed, the catalyst metal, and the feed stoichiometry. The oxygen conversion is 1% for all catalysts and feed stoichiometries after approximately.5 1. grams of catalyst. The length of the oxidation zone increases with increasing C/O ratio and is longer for Pt catalysts than for Rh catalysts. 89

105 Equil.6.5 Equil Rh.6.5 Rh Pt Pt C/O =.7.1 C/O = W cat (g) W cat (g).8.8 Equil.7.7 Equil Rh.3 Rh.3 Pt.2.2 Pt.1.1 C/O = 1.5 C/O = W cat (g) W cat (g) Figure 5-5. Hydrogen flow rate profile as a function of the depth of the catalyst bed, the catalyst metal, and the feed stoichiometry. The flow rate of H 2 on Rh is greater than the flow on Pt for C/O >.7. The flow of H 2 reaches equilibrium on Rh catalysts for all feed stoichiometries but does not reach equilibrium for reactions on Pt at C/O > 1.. 9

106 Pt.35 C/O =.7.3 C/O = 1. Equil.25.2 Pt cat Pt Rh.1 Rh.5 Equil W (g) W cat (g) C/O = C/O = 2. Pt Rh.15.1 Rh.5.5 Equil Equil W cat (g) W cat (g) Figure 5-6. Water flow rate profile as a function of the depth of the catalyst bed, the catalyst metal, and the feed stoichiometry. The presence of a maximum in the water flow profile indicates that steam reforming is taking place after all O 2 is consumed. The average conversion of water (over all C/O ratios) is higher on Rh catalysts (99%) than on Pt catalysts (78%). 91

107 Equil Rh Equil Rh Pt C/O =.7.1 C/O = W cat (g) W cat (g).7.7 Equil Equil.6.6 Pt.5 Rh Rh.3.2 Pt.1 C/O = C/O = W cat (g) W cat (g) Figure 5-7. CO flow rate profile as a function of the depth of catalyst bed, the catalyst metal, and the feed stoichiometry. Unlike H 2, the flow rate of CO does not quite reach equilibrium for C/O > 1. for either Rh or Pt. Similarly to H 2, the flow rate of CO is greater over Rh catalyst than Pt catalyst..3.2 Pt 92

108 .1.8 Pt.1 C/O =.7 C/O = Equil Rh.6.4 Rh Pt Equil W cat (g) W cat (g).1.1 C/O = 1.5 C/O = Rh Pt.6.4 Rh Pt.2.2 Equil Equil W cat (g) W cat (g) Figure 5-8. Carbon dioxide flow rate as a function of the depth of the catalyst bed, the catalyst metal, and the feed stoichiometry. The decrease in CO 2 flow rate with catalyst depth indicates that the CO 2 is being converted in reforming reactions, though not as extensively as H 2 O. 93

109 C/O =.7 C/O = Rh.6.6 Pt.3 Rh Pt.3 cat W (g) W cat (g) C/O = 1.5 C/O = Pt.12.9 Pt Rh.6.3 Rh W cat (g) W cat (g) Figure 5-9. Ethylene flow rate profile as a function of the depth of the catalyst bed, the catalyst metal, and the feed stoichiometry. Ethylene flow rate first increases and then decreases with catalyst bed length which indicates that it is being consumed in the reforming reactions. The peak in ethylene flow rate is further downstream on Pt catalysts than on Rh catalysts. 94

110 1 8 Pt 6 4 Rh W cat (g) C/O = Pt 6 4 Rh W cat (g) C/O = Rh Pt W cat (g) C/O = 2. Figure 5-1. Ratio of the rates of formation of H 2 and consumption of H 2 O as a function of mass of catalyst for Rh and Pt at feed C/O = 1., 1.5, and 2.. When steam reforming is dominating the overall reaction in the bed, the ratio of H 2 production to H 2 O consumption should be between 2 and 3. Values can be slightly higher because of the additional H 2 formed from CO 2 reforming. The vertical lines indicate the regions in which steam reforming is dominating in either the Rh or Pt sphere bed. 95

111 n-c 8 H 18 CH 4, C 2 H 4, etc. O 2 Reforming H 2 CO H 2 O CO 2 H 2 CO Figure Schematic of the reaction steps in the CPO of n-octane. In the presence of O 2, octane forms H 2 O, CO 2, some CO and H 2 or breaks apart into smaller hydrocarbons. The smaller hydrocarbons are then catalytically reformed with H 2 O or CO 2 to make more H 2 and CO. 96

112 1 8 Spheres 6 8 ppi ppi C/O ppi ppi 5 Spheres C/O ppi 5 8 ppi Spheres C/O ppi 8 ppi Spheres Figure Comparison of product selectivities as a function of feed C/O ratio for the CPO of n-octane at 4 SLPM after 1 mm of Rh catalyst for different supports: 1.3 mm spheres, 8 ppi foams with washcoat, and 45 ppi foams without washcoat [1]. The H 2 (and CO - not shown) selectivity increase while H 2 O, CO 2, and ethylene selectivity decrease with increasing heterogeneous activity. The same trends are observed on Pt catalysts. C/O 97

113 Chapter 6 Increasing olefins by H 2 and CH 4 addition to the catalytic partial oxidation of n-octane * 6.1 Introduction As mentioned in Chapter 1, olefins are the largest commodities in the chemicals industry with approximately 1 million tons of ethylene being produced annually [1, 2]. The process of catalytic partial oxidation (CPO) is an alternative to the industrial method of steam cracking (Equations 1-4 and 1-5) for the production of ethylene and other olefins. Approximately 5% of the ethylene made from steam cracking utilizes naptha as a feedstock [2]. Napthas are mixtures of n-alkanes, isoalkanes, olefins, naphthenes, and aromatics that have a boiling point range between 3 and 2 C. The quality and composition of naphtha can vary over a wide range, but a typical feedstock might contain approximately 4 wt% n-paraffins and generate approximately 3 wt% yield to ethylene. We have recently shown that CPO can generate high yields to olefins for naphtha boiling point range n-alkanes [3, 4, Chapters 2 and 3] branched alkanes [3, 5, Chapter 3], and cyclic alkanes [5]. In this Chapter, we will use n-octane (bp 126 C) as a model naphtha feedstock for olefin production. It has been shown in Chapters 4 and 5 that the CPO of liquid alkanes like n-octane can be described through a coupling of heterogeneous surface reactions and homogeneous gas-phase reactions as sketched in Figure 6-1. The basic premise is that octane reacts on the surface with adsorbed oxygen in the first few mm of the catalyst to form H 2 O, CO 2, heat and some H 2 and CO. The heat created by these exothermic surface reactions is used in the gas-phase to drive the O 2 -free homogeneous octane pyrolysis reactions that result in the formation of olefins and other smaller hydrocarbons. The * Portions of this chapter appear in Panuccio G.J. and Schmidt L.D., Increasing olefins by H 2 and CH 4 addition to the catalytic partial oxidation of n-octane, Applied Catalysis A: General (In Press). Elsevier Science Ltd. 98

114 fundamental goal in this study is to increase the selectivity to olefin species in the product stream by co-feeding another reacting species with octane that would preferentially consume the O 2 on the surface and leave more of the octane to react in the gas phase. Hydrogen and methane are chosen as possible sacrificial fuels for this work. Hydrogen is a by-product of the CPO reaction (in cases where olefins are the desired product) and therefore could be recycled to the inlet of the reactor and be oxidized. Methane is often flared during the refining process and so should be readily available for use as an additive to the CPO of heavy fuels. The combustion of both species is highly exothermic and should therefore be able to provide plenty of heat to drive the homogeneous cracking of the heavy alkane to form olefins (see Equations 6-1 and 6-2). H O 2 HO 2 2 ΔH R = -242 kj/mol (6-1) CH + 2O CO + 2H O ΔH R = -82 kj/mol (6-2) In the presence of noble metal catalyst, methane can also react with water to form hydrogen and carbon monoxide in the endothermic steam reforming reaction as shown in Equation 6-3. CH 4 + H2O CO + 3H 2 ΔH R = +25 kj/mol (6-3) The balance between the extents of reaction in Equations 6-2 and 6-3 is critical in determining the efficiency of increasing olefin yields by adding methane to the CPO of n- octane for a particular catalyst. Previous research has focused on the effects of adding sacrificial fuels to small alkane feedstocks. The effect of H 2 addition to the millisecond contact time catalytic oxidative dehydrogenation (ODH) of ethane to ethylene has been investigated on Pt/α- Al 2 O 3 catalysts [6, 7]. These results show that adding H 2 to the feed of ethane and O 2 can marginally increase ethylene selectivity, but the overall yield decreases because ethane 99

115 conversion decreases. Similar experiments performed with propane over Pt/α-Al 2 O 3 catalysts show that ethylene and propylene selectivities can be increased, but that propane conversion decreases with increasing H 2 [8]. Hein and Jess have shown that H 2 and CO addition to propane ODH on Pt/Ni catalysts can improve the propylene and total olefin selectivity [9], while other researchers have compared the effects of adding H 2 and CO to ethane ODH over Pt catalysts [1, 11]. There is however, very little work relating to improving olefin yields in heavier hydrocarbons. O Connor and co-workers showed that benzene and cyclohexene selectivities can be improved on Pt-Sn/α-Al 2 O 3 at low cyclohexane conversion with H 2 addition [12], and Liu and co-workers successfully increased the selectivity of olefins from H 2 addition to hexane on Pt/SiO 2 catalysts at much lower throughputs than considered here [13]. All of these studies examined the effects of addition of a sacrificial fuel at only a single hydrocarbon stoichiometry and catalyst configuration. In this chapter, the effects of H 2 and CH 4 addition to the CPO of a naphtha-like n-octane feed are explored as a function of octane stoichiometry, H 2 (or CH 4 ) stoichiometry, catalyst metal, and support pore size. The optimum operating conditions to maximize olefins will be defined and mechanisms of the reaction will be discussed. 6.2 Experimental The catalysts used in these experiments were α-al 2 O 3 ceramic foam monoliths that were impregnated with Rh or Pt metal. The monoliths were 17 mm diameter and 1 mm length cylinders with 8 or 45 pores per linear inch (ppi). The 8 ppi supports were coated with 5 wt% γ-al 2 O 3 washcoat to further roughen the surface while the 45 ppi monoliths were prepared without a washcoat. By roughening the surface with a washcoat, the difference in the extent of heterogeneous chemistry on 8 ppi foams compared to 45 ppi foams was magnified [3, 14]. The washcoat was applied by preparing a slurry of γ-al 2 O 3 in water and adding it dropwise to the 8 ppi α-al 2 O 3 supports. They were then dried in air and calcined in an oven at 6 C for 4 hours. Rhodium catalysts were prepared by dropwise addition of an aqueous solution of Rh(NO 3 ) 3 to a foam monolith, drying in air, and calcining in an oven under air for 6 1

116 hours at 6 C. Platinum catalysts were prepared by dropwise addition of H 2 PtCl 6 aqueous solution to a foam monolith, drying in air, and calcining in an oven under N 2-1% H 2 atmosphere for 6 hours at 6 C. This prevented a significant quantity of Pt from being lost in the form of PtCl 2 during the calcining process [15]. The final products were 8 ppi 5 wt% Rh (or Pt) 5 wt% γ-al 2 O 3 and 45 ppi 5 wt% Rh (or Pt) catalysts. A detailed description and sketch of the experimental reactor setup has been provided elsewhere [4, 16]. The catalytic foams were placed between two blank (no metal or washcoat) monoliths that acted as radiation heat shields. This assembly was wrapped in Fiberfrax cloth and placed in a quartz tube. A K-type thermocouple was placed between the back-face of the catalyst and the downstream heat shield so that the catalyst temperature may be recorded. A low-flow automotive fuel injector sprayed a conical dispersion of high purity (> 99%) liquid n-octane on the walls at the top of the quartz tube which created a thin film of octane that was vaporized by heating the walls of the tube with a Variac-controlled resistive heating tape. The vaporizing fuel mixed with the other gases in the reactant stream (N 2, O 2, and H 2 or CH 4 ) which were introduced through a side port at the top of the tube via calibrated electronic mass flow controllers. Experiments were performed at atmospheric pressure and N 2 and O 2 were always fed to the reactor in air stoichiometry (N 2 /O 2 = 3.76/1). The amount of octane in the reactant stream was determined from the desired C/O ratio for the experiment. The C/O ratio is defined as the moles of carbon atoms (from octane) divided by the moles of oxygen atoms in the feed. By this definition, the stoichiometric feed composition for the partial oxidation reaction is at C/O = 1.. The experimental octane C/O ratio was varied between.8 and 2.. The stoichiometric H 2 /O 2 and CH 4 /O 2 ratios for combustion are 2. and.5, respectively (see Equations 6-1 and 6-2). For every C/O, the H 2 /O 2 ratio was adjusted from. to 3. (hydrogen addition experiments) and CH 4 /O 2 was varied between. and 2. (methane addition experiments). The sum of the flow rates of N 2, O 2, and n-octane was 4 standard liters per minute (SLPM). Hydrogen (or methane) was added to the reacting mixture according to the desired H 2 /O 2 (or CH 4 /O 2 ) ratio while keeping all other flow rates the same. Therefore, the total flow rate in the reactor increased with H 2 (or CH 4 ) addition which resulted in a maximum decrease in residence 11

117 time from 8 ms with no H 2 addition to 5 ms at H 2 /O 2 = 3/1 at an average catalyst temperature of 9 C. Product gases were sampled with a Hamilton gas-tight syringe downstream of the catalyst and analyzed with an HP589 Gas Chromatography (GC) instrument fit with a 6 m GS-Gas Pro capillary column and a thermal conductivity detector (TCD). The GC oven was also fit with liquid N 2 cooling so that an initial oven temperature was set to -8 C. This allowed for efficient separation of all species from H 2 to n-octane using one GC separation column in less than 2 minutes. Nitrogen was used as the internal standard to calculate the flow rates of all other species. When they are co-fed as reactants, both octane and methane can react to form CO, CO 2, H 2, and H 2 O. Therefore, the calculation of the selectivity of those species should include both octane and methane as fuels. The C-atom selectivity (S i ) of CO and CO 2 was calculated from the number of carbon atoms in species i (n i ), and the flow rate of species i (F i ) as in Equation 6-4. S = i nf i i n F + n F - (n F + n F ) C8H18 C8H 18,In CH4 CH 4,In C8H18 C8H 18,Out CH4 CH 4,Out (6-4) A similar equation was used to calculate the H-atom selectivities to H 2 and H 2 O (with n i as the number of hydrogen atoms in species i). The selectivities of all other species were calculated based only on the carbon in octane (Equation 6-5) because the amount of those species formed from methane was insignificant compared to the amount formed from octane. S = i nf i i n F - n F CH 8 18 CH 8 18,In CH 8 18 CH 8 18,Out (6-5) All C-atom species selectivities (including CO and CO 2 ) were calculated according to Equation 6-5 for the hydrogen addition experiments. 12

118 6.3 Results Each data point in the following figures is an average of three experiments performed on at least two similarly prepared catalysts. Data is plotted as a function of H 2 /O 2 (or CH 4 /O 2 ) gas ratio in the feed (x-axis) and octane C/O ratio (in series). In the following section, the differences between H 2 and CH 4 addition to the CPO of n-octane on Rh and Pt coated foam monoliths of different pore sizes is examined Rh on 8 ppi (with washcoat) catalyst Figure 6-2 shows the n-octane conversion, net H 2 production, ethyelene selectivity, and total olefin selectivity that result from the addition of H 2 to the CPO of n- octane on 8 ppi 5% Rh 5% γ-al 2 O 3 washcoat catalysts. Octane conversion remains nearly constant with increasing H 2 in the feed at all C/O ratios. Conversion decreases from 1% at C/O =.8 to 7% at C/O = 2., which is in agreement with previously observed results of n-octane CPO [3]. Ethylene and total olefin selectivities increase with increasing feed H 2 /O 2 at all octane C/O ratios. For example, at C/O = 2., the ethylene selectivity increases from 19% at H 2 /O 2 =. to 24% at H 2 /O 2 = 3. and the total olefin selectivity increases from 42% when no hydrogen is fed to the reactor to 49% at H 2 /O 2 = 3.. The selectivity of olefins produced in this reaction follows as ethylene > propylene > n-butylene > i-butylene in the same relative distribution as previously reported [3]. The trends in selectivity with respect to H 2 (and CH 4 ) addition for higher molecular weight olefins are the same as those observed for ethylene. Methane and other low molecular weight paraffins are produced in similar quantities as previously reported [3] and are not shown for brevity. The top right panel in Figure 6-2 compares the moles of hydrogen produced with the moles of hydrogen fed to the reactor on a per mole of oxygen basis. This particular catalyst configuration is a net producer of H 2 since all experimental points lie above the H 2,Out = H 2,In line. This shows that no outside hydrogen would need to be purchased to run this reaction in an industrial setting. 13

119 The results from the addition of methane to the CPO of n-octane on 8 ppi 5% Rh 5% γ-al 2 O 3 catalysts are summarized in Figure 6-3. Adding methane to this reaction yields very different results than the addition of hydrogen. Now, the conversion of octane decreases with increasing methane in the feed for all octane C/O ratios. Also, the selectivity of H 2 (and CO not shown) increases with CH 4 addition, while the formation of ethylene and other olefins is suppressed by the addition of CH 4 to the reacting mixture. Examination of the differences in the measured catalytic back-face temperature for H 2 addition and CH 4 addition experiments (Figure 6-4) shows that temperatures increase slightly when H 2 is added and decrease significantly with the addition of CH 4. For example, at n-c 8 C/O = 1., the measured catalyst back-face temperature increases from 924 to 954 C when H 2 is added and decreases from 924 to 83 C with the addition of CH 4. These results indicate that methane and hydrogen behave very differently on 8 ppi Rh catalysts and will be discussed later Rh on 45 ppi (no washcoat) catalyst Figure 6-5 shows the results of adding H 2 to the CPO of n-octane on 45 ppi 5% Rh (no washcoat) foams. Octane conversion increases from 86 to 94% at C/O = 1.3 and from 64 to 7% at C/O = 2. when H 2 is added to the reacting mixture. The maximum ethylene selectivity increases with H 2 /O 2 from 34% at C/O = 1. and H 2 /O 2 =. to 46% at C/O = 1. and H 2 /O 2 = 3.. Ethylene selectivities also increase for other octane C/O ratios when H 2 is added. The maximum total olefin selectivity also increases from 6% at C/O = 2. and H 2 /O 2 =. to 75% at C/O = 2. and H 2 /O 2 = 3.. The selectivity to olefins is significantly higher for the 45 ppi foam than the 8 ppi support both with and without H 2 addition. The top-right panel in Figure 6-5 shows that the overall reaction is a net producer of H 2 for octane C/O =.8 and a net consumer of H 2 for C/O 1.. The results of adding CH 4 to the CPO of n-octane on 45 ppi 5% Rh (no washcoat) catalysts are summarized in Figure 6-6. Octane conversion decreases with increasing methane in the feed, which follows the same trends observed for methane addition on 8 ppi catalysts. However, the selectivity of H 2 decreases for CH 4 /O 2.25 at C/O =.8 and remains nearly constant for all CH 4 /O 2 ratios at octane C/O 1., which is not the 14

120 same trends as observed on 8 ppi Rh coated catalysts. Furthermore, ethylene and total olefin selectivities increase with CH 4 addition (more so at C/O 1. than C/O 1.3). This result suggests that the extent of heterogeneous chemistry is very important when considering the effects of adding CH 4 to the CPO of liquid fuels on Rh-coated foams Pt on 8 ppi (with washcoat) catalyst Figure 6-7 summarizes the results of adding H 2 to n-octane CPO on 8 ppi 5% Pt 5% γ-al 2 O 3 catalysts. As with the reactions on 8 ppi Rh catalysts, the conversion of n- octane is nearly constant with increasing H 2 in the feed stream. It should be noted that the conversion of octane on 8 ppi Pt catalyst is higher than the conversion on 8 ppi Rh catalysts. For example, the conversion of octane is ~7% on 8 ppi Rh and ~ 85% on 8 ppi Pt catalysts at octane C/O = 2.. Ethylene selectivity increases with increasing H 2 /O 2 for all octane C/O. The global maximum for ethylene selectivity without hydrogen addition is 38% at C/O = 1.3. This maximum shifts to 45% at C/O = 2. when hydrogen is added in an H 2 /O 2 = 3. ratio. Total olefin selectivity also increases with increasing H 2. The effect is more pronounced for C/O < 1.3 where the total olefin selectivities increase from 5% to 24% at C/O =.8 and 23% to 4% at C/O = 1.. Overall, the reaction results in a net production of hydrogen for C/O 1.3 and a net consumption of hydrogen for C/O = 2.. The results from the addition of CH 4 to n-octane CPO on 8 ppi 5% Pt 5% γ- Al 2 O 3 washcoat are shown in Figure 6-8. As is the case with Rh catalysts, octane conversion decreases with increasing CH 4 /O 2 for all C/O ratios. While H 2 selectivity increases with increasing methane in the feed on 8 ppi Rh catalysts (Figure 6-3), the selectivity to H 2 decreases with increasing CH 4 /O 2 on 8 ppi Pt catalysts. Ethylene selectivity increases with increasing methane for C/O 1. and remains constant for C/O 1.3. The total olefin selectivity increases for C/O 1.3 and remains constant for C/O = 2.. These trends in olefin selectivity are the opposite of those observed for 8 ppi Rh catalysts where olefin selectivities decrease with increasing methane for all octane C/O ratios. 15

121 6.3.4 Pt on 45 ppi (no washcoat) catalyst Finally, the results of H 2 and CH 4 addition to n-octane CPO are investigated on 45 ppi 5% Pt (no washcoat) catalysts. The results of adding H 2 to the reaction on this catalyst configuration are shown in Figure 6-9. Octane conversion is independent of H 2 /O 2 ratio for C/O 1.3 and increases from 64% at C/O = 2. with no H 2 to 75% at H 2 /O 2 = 1.. Ethylene and total olefin selectivity increase when hydrogen is added to the reacting mixture. The maximum selectivity of ethylene when H 2 /O 2 =. is 38% at C/O = 1.3. With H 2 addition, the maximum ethylene selectivity increases to 51% at H 2 /O 2 = 3. and C/O = 1.. There is a net consumption of hydrogen at all H 2 /O 2 and octane C/O ratios on 45 ppi Pt catalysts. Adding methane to n-octane CPO on 45 ppi Pt catalysts gives very different results than the addition of hydrogen. Methane addition to the reactant mixture on this particular catalyst configuration severely limits the range over which the reaction will operate under stable conditions (data not shown). For example, when any methane is added to the CPO of n-octane at C/O > 1.3, the reaction is no longer autothermal and it extinguishes itself. This is marked by a continuous decline in the measured catalyst back-face temperature with time until the temperature on the catalyst is the same as the temperature in the upstream mixing zone of the quartz tube (which indicates that there is no reaction taking place). The addition of methane at lower octane C/O ratios severely hinders the reactor performance as octane conversion and H 2 selectivity fall sharply with increasing CH 4 /O 2. The maximum ethylene selectivity increases from 38% at C/O = 1.3 and CH 4 /O 2 =. to 42% at C/O = 1. and CH 4 /O 2 =.25. But, as conversion (and consequently reactor temperature) drops, ethylene selectivity also decreases. The total olefin selectivity increases for C/O 1. with the addition of methane, but remains constant for C/O = 1.3. The results described in this section show that the performance of the CPO reaction with the addition of a sacrificial fuel is very dependent on the catalyst metal, the catalyst support structure, and whether the additional feed component is hydrogen or methane. Table 6-1 shows the maximum ethylene and total olefin selectivities obtained with no co-feed, with H 2, or with CH 4 co-feed for all four catalysts studied in these 16

122 experiments. In general, co-feeding H 2 with n-octane results in a larger increase in olefin selectivities than CH 4 addition. The global maximum for ethylene (51%) and total olefin (83%) selectivity occurs on 45 ppi Pt catalysts with a co-feed of hydrogen at H 2 /O 2 = 3/1. The implication of these results is discussed in the next section. 6.4 Discussion Two-zone reaction model A recent computational study on the oxidative dehydrogenation of ethane with hydrogen addition over Pt-coated foam monoliths combines elementary step heterogeneous (2 reversible reactions) and homogeneous (44 reversible reactions) mechanisms [17]. This study shows that ethane ODH with H 2 addition can be described accurately by two sequential zones in the catalyst. In the first zone, hydrogen and ethane are consumed on the surface with O 2 which creates oxidation products and heat. This is followed by a zone where the gas-phase dehydrogenation of ethane dominates and ethylene is formed. The simulations and experiments shown in Chapters 4 and 5 show that a 2-zone reaction scheme can accurately describe the CPO of n-octane on noble metal-coated foams. As with ethane ODH, the oxidation reactions occur in the first few millimeters of the catalyst where adsorbed oxygen and fuel react on the surface to produce H 2, H 2 O, CO, CO 2, and heat. The heat is used in endothermic gas-phase pyrolysis reactions of octane that is not consumed on the surface to produce olefins and other smaller hydrocarbons. The experiments with H 2 and CH 4 addition to n-octane CPO can also be qualitatively explained with this mechanism. Now, octane and sacrificial fuel can both react catalytically with oxygen in the first few millimeters of the catalyst to generate oxidation products and heat. The extents to which either octane or sacrificial fuel react with oxygen in a competitive manner is dependent on the catalyst metal, the surface area to volume ratio in the catalytic bed, and the relative reactivity of the preferential oxidant in relation to the heavy hydrocarbon fuel. 17

123 6.4.2 H 2 addition The selectivity of ethylene and other olefins resulting from the CPO of n-octane can be improved by adding H 2 to the reactor feed stream. While ethylene selectivity increases by 5% (65 to 7%) from reactions with ethane [6, 7] and 6% with propane (32 to 38%) [8], the maximum ethylene selectivity increases by 12% (34 to 46%) on 45 ppi Rh catalysts and 13% (38 to 51%) on Pt 45 ppi catalysts for H 2 addition to octane CPO (see Table 6-1). The total olefin selectivity increases by 14% (61 to 75%) on 45 ppi Rh and 8% (75 to 83%) on 45 ppi Pt catalysts when hydrogen is added. Furthermore, experiments performed with H 2 addition to ethane and propane ODH show that the conversion of hydrocarbon decreases as H 2 /O 2 and ethylene selectivity increase [6-8]. The results of this current work show that conversion of octane does not change with increasing H 2 /O 2 on all the catalysts studied. This shows that the effect of adding H 2 to the catalytic partial oxidation process improves reactor performance (in terms of olefin yields) more efficiently for heavier fuels than for lighter hydrocarbons because olefin selectivity is increased more substantially and conversion is not decreased. There is no drop in conversion with H 2 addition to n-octane because of the gas phase reactivity of the hydrocarbon fuel. Since some of the oxygen is reacting with hydrogen in the front zone of the catalyst, there is more hydrocarbon remaining to react in the gas-phase. At the same temperature, a heavy molecular weight hydrocarbon like octane is more likely to pyrolyze than a light paraffin like ethane or propane. Therefore, conversion of a light paraffin will decrease because the gas phase reactivity cannot compensate for the decrease in surface conversion. The heavier paraffin is more likely to react homogeneously and therefore has the same overall conversion. It should be noted that the maximum for ethylene selectivity and the maximum for total olefin selectivity do not occur at the same C/O ratio. For example, for reactions on 45 ppi Pt-coated foams, the maximum ethylene selectivity is found at octane C/O = 1. and H 2 /O 2 = 3/1. However, the maximum olefin selectivity occurs at C/O = 2. and H 2 /O 2 = 3/1. So, the conditions that are chosen for reactor operation will be a function of the end goal of maximizing ethylene or total olefin yields. 18

124 For all catalysts (excluding 8 ppi Rh w/ WC), there is a net consumption of hydrogen at the conditions under which the selectivities of ethylene and other olefins are maximized (H 2 /O 2 = 3/1). This means that hydrogen must be purchased and cannot simply be recycled from the product stream. The increased operating cost from the hydrogen may not justify the improved yields observed in these conditions. However, adding hydrogen at H 2 /O 2 = 1/1 may be a viable option because olefin selectivities are increased by a few percent, but under nearly neutral hydrogen production conditions CH 4 addition The effect of methane addition is heavily dependent on catalyst metal and pore geometry. The addition of methane to the CPO of n-octane on 8 ppi Rh-coated foams causes the temperature in the reactor to decrease, the selectivity of H 2 and CO to increase, and the selectivity of ethylene and other olefins to decrease (Figure 6-3). These results can be explained by the consumption of methane via the steam reforming reaction as in Equation 6-3. This reaction explains the increase in H 2 and CO selectivity as well as the decrease in measured catalyst back-face temperature since the overall reaction is endothermic. It is shown in Chapter 5 that this reforming reaction is likely to take place in the catalyst after the O 2 has been consumed (i.e. in the 2 nd zone of the catalyst in Figure 6-1) because the catalytic reforming reactions are kinetically slower than the oxidation reactions [18]. A decrease in reactor temperature precipitates a decrease in the kinetic rates of the pyrolysis reactions and the suppression of the formation of olefins. This decrease in olefins is not observed on 45 ppi Rh-coated foams. In fact, the selectivities of olefins increase and H 2 and CO decrease when methane is added on this catalyst (more so at octane C/O < 1.3 see Figure 6-6). This is probably because the ratio of the catalyst surface area to the volume of the gas is not large enough for the extent of catalytic steam reforming reactions occurring in the downstream section to be as large as the extent of homogeneous reactions in the gas phase. Methane is still oxidized somewhat in the front of the catalyst but is not reformed in the downstream section. The increase in olefin selectivities is therefore due to the fact that some of the oxygen is consumed in reactions with methane instead of octane, and the kinetics of the pyrolysis 19

125 reactions are not hindered because the gas temperature is not lowered by the endothermic catalytic reforming reactions in the downstream zone. The addition of methane to octane CPO on 8 ppi Pt catalysts causes the selectivity of ethylene and other olefins to increase while the selectivity of H 2 and CO decreases (see Figure 6-8). The results of Chapters 2 and 5 along with previous studies that compare the partial oxidation of alkanes over Pt and Rh-coated monoliths have shown that Rh catalysts produce much higher selectivities to syngas than Pt catalysts and Pt catalysts produce higher selectivity to H 2 O, CO 2, and olefins than Rh [19-24]. This suggests that Pt is a poorer reforming catalyst than Rh. Therefore, some methane is oxidized instead of octane in the first few mm of the Pt bed but is not reformed in the downstream section of the catalyst. In the absence of the reforming reactions, the reactor temperature does not drop substantially and the pyrolysis reaction rates remain high enough to form significant quantities of olefins. Since some of the oxygen reacts with methane instead of octane, more hydrocarbon is left to react through gas-phase chemistry and olefin selectivity increases. The addition of methane to the catalytic partial oxidation of octane on 45 ppi Ptcoated catalysts severely limits the range over which the reactor will operate autothermally. The reaction quenches itself when the rate of exothermic surface reactions is not fast enough to keep the overall process autothermal. When methane is added, the partial pressure of oxygen decreases and the surface reaction rate decreases. This is not observed on 8 ppi Pt catalysts because the higher surface area to volume ratio increases the overall reaction rate. Experiments performed on the ignition behavior of different fuels in the CPO process have shown that the sticking coefficient (the statistical probability that a species will adsorb when it comes in contact with a surface) for methane should be 2 to 4 times smaller than the sticking coefficient for n-octane [16]. Therefore, methane causes the reaction to quench on the 45 ppi Pt foams because it acts as a heat load, decreases the oxygen partial pressure, and also does not stick to the surface and react. Combine these effects with a lower surface area to volume ratio and the overall heterogeneous reaction rate is not fast enough to keep the process autothermal. 11

126 6.4.4 Flames and explosions Experiments where hydrogen and oxygen are both fed as reactants are potentially dangerous because of the possibility of ignition or detonation. Conditions in the H 2 addition to n-octane CPO experiments are typically within the explosion limits of both hydrogen in air (18% - 59%) [25, 26] and n-octane in air (1% - 7%) [27]. However, no flames or explosions are observed over the range of H 2 /O 2 and C/O ratios used in these experiments. Hydrogen addition experiments with lighter alkane fuels also do not result in homogeneous ignition upstream of the catalyst [6-8]. As with ethane and propane, the presence of octane in the feed must stabilize the mixture because the heavy hydrocarbon can successfully quench any radicals that may lead to propagation reactions. If this process was adopted for industrial production of olefins, pure oxygen will probably be used instead of air in order to eliminate the cost of separating N 2 from the product stream. The use of pure O 2 increases the flammability and explosive limits and makes reactor operation more dangerous. Extra care should be taken in designing a reactor that does not operate with N 2 dilution. 6.5 Conclusions The addition of a sacrificial fuel such as H 2 or CH 4 to the catalytic partial oxidation of naphtha-like fuels like n-octane can increase the yields of ethylene and other olefins by catalytically oxidizing some of the sacrificial fuel and allowing more n-octane to undergo homogeneous pyrolysis. Hydrogen addition gives the best performance by increasing olefin selectivities while keeping the octane conversion constant. The choice of catalyst metal and support pore size is flexible because 45 ppi Rh, 45 ppi Pt, and 8 ppi Pt catalysts all generate high selectivity to olefins when H 2 is co-fed with octane. However, this is accomplished at a net consumption of hydrogen, and the increased operational cost from purchasing hydrogen may not justify the increase in olefin yields. The performance of the CPO of n-octane with methane addition is very dependent on the catalyst metal and the pore size of the support structure. When 8 ppi Rh catalysts are used, the selectivity to olefins actually decreases while the selectivity to syngas increases. This is most likely due to methane steam reforming in the second zone of the 111

127 catalyst bed. Methane addition to reactions on 45 ppi Rh and 8 ppi Pt catalysts slightly increases the selectivity to ethylene and other olefins. It may be worthwhile to add methane to a catalytic partial oxidation feed in order to increase olefin selectivities a few percent in chemical plants that have access to a methane waste stream that is ordinarily flared or is otherwise not utilized. References [1] Sundaram KM, Shreehan MM, Olszewski EF. Ethylene. In: Kirk-Othmer Encyclopedia of Chemical Technology. John Wiley & Sons, Inc, 21. [2] Zimmermann H, Walzl R. Ethylene. In: Ullmann's Encyclopedia of Industrial Chemistry. Wiley-VCH Verlag GmbH & Co. KGaA, 22. [3] Panuccio GJ, Williams KA, Schmidt LD. Contributions of heterogeneous and homogeneous chemistry in the catalytic partial oxidation of octane isomers and mixtures on rhodium coated foams. Chemical Engineering Science. 26; 61: [4] Krummenacher JJ, West KN, Schmidt LD. Catalytic partial oxidation of higher hydrocarbons at millisecond contact times: decane, hexadecane, and diesel. Journal of Catalysis. 23; 215: [5] O'Connor RP, Klein EJ, Schmidt LD. High yields of synthesis gas by millisecond partial oxidation of higher hydrocarbons. Catalysis Letters. 21; 7: [6] Bodke AS, Henning D, Schmidt LD, Bharadwaj SS, Maj JJ, Siddall J. Oxidative dehydrogenation of ethane at millisecond contact times: Effect of H2 addition. Journal of Catalysis. 2; 191: [7] Bodke A, Olschke D, Schmidt LD, Ranzi E. High selectivities to ethylene by partial oxidation of ethane. Science. 1999; 285: [8] Bodke A, Henning D, Schmidt LD. A comparison of H 2 addition to 3 ms partial oxidation reactions. Catalysis Today. 2; 61: [9] Hein O, Jess A. Heterogeneous and homogeneous processes in oxidative dehydrogenation of propane in a catalytic fixed bed reactor. Erdol Erdgas Kohle. 22; 5:

128 [1] Donsi F, Cimino S, Pirone R, Russo G. Autothermal oxidative dehydrogenation of ethane on LaMnO 3- and Pt-based monoliths: H 2 and CO addition. Industrial and Engineering Chemistry Research. 25; 44: [11] Chen S, McDonald SR, Chen Z, ConocoPhillips Company, United States, 25. [12] O'Connor RP, Klein EJ, Henning D, Schmidt LD. Tuning millisecond chemical reactors for the catalytic partial oxidation of cyclohexane. Applied Catalysis A: General. 23; 238: [13] Liu X, Li W, Zhu H, Ge Q, Chen Y, Xu H. Light alkenes preparation by the gas phase oxidative cracking or catalytic oxidative cracking of high hydrocarbons. Catalysis Letters. 24; 94: [14] Bodke A, Bharadwaj S, Schmidt LD. Effect of ceramic supports on partial oxidation of hydrocarbons over noble metal coated monoliths. Journal of Catalysis. 1998; 179: [15] Schweizer AE, Kerr GT. Thermal decomposition of hexachloroplatinic acid. Inorganic Chemistry. 1978; 17: [16] Williams KA, Schmidt LD. Catalytic autoignition of higher alkane partial oxidation on Rh-coated foams. Applied Catalysis A: General. 26; 299: [17] Donsi F, Williams KA, Schmidt LD. A multistep surface mechanism for ethane oxidative dehydrogenation on Pt- and Pt/Sn-Coated Monoliths. Industrial and Engineering Chemistry Research. 25; 44: [18] Panuccio GJ, Schmidt LD. Species and temperature profiles in a differential sphere bed reactor for the catalytic partial oxidation of n-octane. In preparation; [19] Hickman DA, Haupfear EA, Schmidt LD. Synthesis gas formation by direct oxidation of methane over Rh monoliths. Catalysis Letters. 1993; 17: [2] Torniainen PM, Chu X, Schmidt LD. Comparison of monolith-supported metals for the direct oxidation of methane to syngas. Journal of Catalysis. 1994; 146: 1-1. [21] Huff M, Torniainen PM, Hickman DA, Schmidt LD. Partial oxidation of CH 4, C 2 H 6, and C 3 H 8 on monoliths at short contact times. Natural Gas Conversion II. 1994;

129 [22] Huff M, Torniainen PM, Schmidt LD. Partial oxidation of alkanes over nobel metal coated monoliths. Catalysis Today. 1994; 21: [23] Krummenacher JJ, Schmidt LD. High yields of olefins and hydrogen from decane in short contact time reactors: rhodium versus platinum. Journal of Catalysis. 24; 222: [24] Panuccio GJ, Dreyer BJ, Schmidt LD. Comparison of the millisecond contact time catalytic partial oxidation of C 1 to C 16 normal paraffins. AIChE Journal. Submitted; [25] Hord J. Is hydrogen a safe fuel? International Journal of Hydrogen Energy. 1978; 3: [26] Lewis B, Von Elbe G. Combustion, Flames, and Explosions of Gases (2nd 2nd). New York: Academic Press Inc, [27] n-octane Material Safety Data Sheet. Sigma Aldrich Corporation. 114

130 Table 6-1. Maximum ethylene and total olefin selectivities with no co-feed, with H 2, and with CH 4 co-feed. The values in parenthesis are the octane C/O and H 2 /O 2 (or CH 4 /O 2 ) ratios at which the maximum occurs. C 2 H 4 Sel. (%) Total Olefin Sel. (%) Catalyst No Add w/ H 2 w/ CH 4 No Add w/ H 2 w/ CH 4 Pt 45 ppi Pt 8 ppi Rh 45 ppi Rh 8 ppi 38 (1.3, -) 51 (1., 3) 42 (1.,.25) 75 (2., -) 83 (2., 3) 71 (1.3,.5) 37 (1.3, -) 44 (2., 3) 38 (1., 2.) 7 (2., -) 73 (2., 3) 74 (2., 2.) 34 (1., -) 46 (1., 3) 37 (1.,.25) 61 (2., -) 75 (2., 3) 65 (2.,.25) 2 (2., -) 24 (2., 3) 2 (2., ) 42 (2., -) 49 (2., 3) 42 (2., ) 115

131 H 2 or CH 4 Heat, CO, H 2, H 2 O, CO 2 Surface O 2 Reactants Quartz tube Oxidation zone C 8 H 18 1 mm Catalytic Foam Monolith Gas Phase Reforming zone C 2 H 4, C 3 H 6, C 4 H 8 Products Figure 6-1. Schematic of the coupling between heterogeneous and homogeneous chemistry in CPO. Hydrogen and methane are added to the reactant stream so that they might preferentially react with O 2 on the surface and thereby increase the amount of octane that is converted through the gas-phase pathway to form olefins. 116

132 n-c 8 Conv. (%) C 2 H 4 Select. (%) n-c 8 C/O = 2. n-c 8 C/O = H 2 /O H 2 /O 2 H 2,Out /O 2,In Tot. Olf. Select. (%) H 2 /8 ppi/rh,wc n-c 8 C/O = H 2,In /O 2,In n-c 8 C/O =.8.8 H 2,Out = H 2,In H 2 /O 2 Figure 6-2. n-octane conversion, net H 2 production, C 2 H 4 selectivity, and total olefin selectivity for H 2 addition to n-octane CPO on 8 ppi Rh w/ washcoat catalysts. Conversion is nearly constant at all octane C/O ratios with H 2 addition. Ethylene and total olefin selectivities increase with increasing H 2 in the feed and the overall reaction is a net producer of H 2 at all C/O and H 2 /O 2 ratios studied on 8 ppi Rh coated catalysts. 117

133 n-c 8 Conv. (%) CH 4 /8 ppi/rh,wc n-c 8 C/O = CH 4 /O 2 H 2 Select. (%) CH 4 /O 2 n-c 8 C/O = C 2 H 4 Select. (%) n-c 8 C/O = CH 4 /O 2 Tot. Olf. Select. (%) n-c 8 C/O = CH 4 /O 2 Figure 6-3. n-octane conversion, H 2 selectivity, C 2 H 4 selectivity, and total olefin selectivity for the addition of CH 4 to n-octane CPO on 8 ppi Rh w/ washcoat catalysts. The addition of methane gives very different results from those of H 2 addition on 8 ppi Rh catalysts. Increasing the amount of methane in the feed decreases octane conversion, increases H 2 (and CO) selectivity, and suppresses the formation of olefins at all octane C/O ratios studied. 118

134 Temperature ( C) n-c 8 C/O = H 2 Addition Temperature ( C) H 2 /O 2 CH 4 Addition n-c 8 C/O = CH 4 /O 2 Figure 6-4. Measured catalyst back-face temperature for H 2 (top) and CH 4 (bottom) addition to n-octane CPO on 8 ppi Rh w/ washcoat catalysts. While catalyst back-face temperature increases slightly when H 2 is added to the feed, the temperature drops precipitously when CH 4 is added. This suggests that the methane is converted to H 2 and CO through endothermic steam reforming reactions. 119

135 n-c 8 Conv. (%) C 2 H 4 Select. (%) H 2 /O 2 n-c 8 C/O = H 2 /O n-c 8 C/O = 2. H 2,Out /O 2,In Tot. Olf. Select. (%) H 2 /45 ppi/rh n-c 8 C/O = n-c 8 C/O =.8 H 2,In /O 2,In H 2,Out = H 2,In H 2 /O 2 Figure 6-5. n-octane conversion, net H 2 production, C 2 H 4 selectivity, and total olefin selectivity for H 2 addition to n-octane CPO on 45 ppi Rh no washcoat catalysts. Octane conversion increases slightly at C/O 1.3 with increasing H 2 in the feed. Ethylene and total olefin selectivities also increase with increasing H 2 for all octane C/O ratios studied. The overall reaction is a net consumer of H 2 for octane C/O

136 n-c 8 Conv. (%) C 2 H 4 Select. (%) CH 4 /O n-c 8 C/O = CH 4 /O 2 n-c 8 C/O = H 2 Select. (%) Tot. Olf. Select. (%) CH 4 /O 2 n-c 8 C/O =.8 n-c 8 C/O = CH 4 /O 2 CH 4 /45 ppi/rh Figure 6-6. n-octane conversion, H 2 selectivity, C 2 H 4 selectivity, and total olefin selectivity for CH 4 addition to n-octane CPO on 45 ppi Rh no washcoat catalysts. Octane conversion decreases with increasing methane in the feed at all octane C/O ratios. H 2 selectivity remains constant with increasing feed methane for C/O 1.. Ethylene and total olefin selectivities increase with methane addition for octane C/O

137 n-c 8 Conv. (%) C 2 H 4 Select. (%) H 2 /O n-c 8 C/O = H 2 /O n-c 8 C/O = 2. H 2,Out /O 2,In Tot. Olf. Select. (%) H 2 /8 ppi/pt,wc n-c 8 C/O = n-c 8 C/O =.8 H 2,In /O 2,In H 2,Out = H 2,In H 2 /O 2 Figure 6-7. n-octane conversion, net H 2 production, C 2 H 4 selectivity, and total olefin selectivity for H 2 addition to n-octane CPO on 8 ppi Pt w/ washcoat catalysts. Octane conversion is nearly constant with H 2 addition at all C/O ratios studied. Ethylene and total olefin selectivities increase with increasing H 2 in the feed and the overall reaction is a net producer of H 2 for octane C/O 1.3 and a net consumer of H 2 for C/O >

138 n-c 8 Conv. (%) C 2 H 4 Select. (%) CH 4 /O n-c 8 C/O = CH 4 /O 2 n-c 8 C/O = H 2 Select. (%) Tot. Olf. Select. (%) n-c 8 C/O = CH 4 /O 2 2. CH 4 /8 ppi/pt,wc CH 4 /O 2 n-c 8 C/O =.8 Figure 6-8. n-octane conversion, H 2 selectivity, C 2 H 4 selectivity, and total olefin selectivity for CH 4 addition to n-octane CPO on 8 ppi Pt w/ washcoat catalysts. Octane conversion and H 2 selectivity decrease with increasing methane in the feed. Ethylene and total olefin selectivities increase with increasing methane for octane C/O 1.3 and remain constant for octane C/O =

139 n-c 8 Conv. (%) C 2 H 4 Select. (%) H 2 /O n-c 8 C/O =.8 n-c 8 C/O = H 2 /O H 2,Out /O 2,In Tot. Olf. Select. (%) H 2 /45 ppi/pt H 2,Out = H 2,In n-c 8 C/O = n-c 8 C/O =.8 H 2,In /O 2,In H 2 /O 2 Figure 6-9. n-octane conversion, net H 2 production, C 2 H 4 selectivity, and total olefin selectivity for H 2 addition to n-octane CPO on 45 ppi Pt no washcoat catalysts. Octane conversion is constant with increasing feed H 2 for C/O 1.3 and increases and then decreases for increasing H 2 at C/O = 2.. Ethylene and total olefin selectivities increase with increasing H 2 in the feed while the overall reaction is a net consumer of H 2 for all octane C/O ratios studied. 124

140 Chapter 7 Thesis summary and future directions 7.1 Thesis summary The results integrated in this thesis have aided in the understanding of the mechanism for the catalytic partial oxidation of liquid alkane feedstocks. Both homogeneous and heterogeneous chemistry are significant in the CPO of heavy hydrocarbons. Reactions on the surface produce H 2, H 2 O, CO, and CO 2 via the direct oxidation of reactant fuel and from the dry and steam reforming of smaller hydrocarbons. Rh is the best metal to use if syngas is the desired product because the reforming reactions occur at faster rates and higher extents than Pt. Conversely, Pt is the best catalyst for olefin production because the reforming reactions are slower and the smaller hydrocarbons are not consumed. Homogeneous reactions produce olefins and low molecular weight paraffins from the gas-phase pyrolysis of heavier hydrocarbons. Olefin selectivity can be maximized by using catalyst supports with large pore diameters and by co-feeding a sacrificial fuel like H 2 that reacts with O 2 in the front of the catalyst and leaves more hydrocarbon fuel to react in the gas phase. The molecular weight and structure of the reactant fuel are important parameters for the product distribution and the fuel conversion. Comparing the reaction of different chain length normal paraffins shows that fuel conversion and olefin selectivity increase while syngas selectivity decreases with increasing molecular weight. Comparing linear and branched alkanes shows that the heterogeneous chemistry is not dependent on the fuel structure, but the degree of branching is very important on the product distribution resulting from the homogeneous pyrolysis of the fuel. Furthermore, the reaction of mixtures of linear and branched hydrocarbons shows that there is not much interaction in the gas-phase chemistry for the different isomers because the mixture produces an average of the selectivities obtained from the single component reaction. However, the structure of the fuel is very important when considering the competitive oxidation in the first zone of the catalyst. Steric hindrances cause the conversion of i-octane to decrease 125

141 in the presence of n-octane because the normal paraffin can more easily adsorb and react on the surface. 7.2 Future directions Heterogeneous mechanism The results of Chapter 4 show that the homogeneous chemistry in the CPO of heavy alkanes can be satisfactorily explained by gas-phase pyrolysis mechanisms developed by other researchers. However, a corresponding heterogeneous reaction mechanism must be developed if the CPO of heavy alkanes is to be thoroughly modeled. Attempts were made to fit the ignition and surface species generation of the CPO of octane by modifying an elementary step surface mechanism for the CPO of methane on Rh and Pt catalysts that was first developed by Hickman and Schmidt [1, 2]. The methane mechanism was modified innumerable times by adding octane adsorption/desorption and pyrolysis reactions similar to Equations 7-1 and 7-2. CH 8 18(g) + Rh (s) CH 8 18(s) (7-1) C8H 18(S) + 25Rh (s) 8C (s) + 18H(s) (7-2) These mechanisms never successfully encapsulated the physics of the reaction of octane on Rh. The results of Chapter 5 show that it is probably important to add adsorption/desorption, steam reforming, and CO 2 reforming reactions for small hydrocarbons like ethylene, methane, propylene, etc. This could probably be accomplished by a researcher familiar with global minimization techniques by developing algorithms that write every possible surface reaction and then find the parameter space that best fits the data Mixture experiments As explained previously, distributed feedstocks like gasoline and naphtha are very complicated mixtures of linear, branched, and cyclic alkanes, alkenes, and aromatics. 126

142 The results of Chapter 3 show that there are some interesting catalytic effects when mixtures of branched and linear alkanes are partially oxidized. Other results have shown that mixtures of normal paraffins with different molecular weights also give interesting results [3]. There are still several combinations of hydrocarbon mixtures that have not been investigated which include aromatic/alkane, alkene/alkane, cyclic alkane/normal alkane, and cyclic alkane/aromatic mixtures (among others). The competitive adsorption and reaction of these types of compounds in a mixture should be interesting and offer further insight as to the performance that can be expected from the reaction of distributed fuels Profile measurements The profile measurements shown and discussed in Chapter 5 give insight as to the mechanism of the surface chemistry for the CPO of n-octane on Rh and Pt catalysts. This experiment could be repeated for any other number of situations to give interesting results. This would be especially relevant for the CPO of fuel mixtures. For example, aromatic/alkane mixture temperature and species flow rate profile could be measured to examine the relative reactivity inside the catalyst of those two classes of compounds. Furthermore, the profile measurements should be done for oxygenated fuels like ethanol, and ethylene glycol. CPO experiments performed on those fuels show that the mechanism for oxygenated compounds should be very different than the mechanism for fuels that do not have oxygen in the chemical formula [4, 5] Catalyst morphology In the process of collecting the data presented in this thesis, I have noticed that catalyst performance can change slightly over time. It would be interesting to do a study of the morphology of the catalyst surface as a function of the time on stream of the CPO reaction. Two SEM images of the same location of a used 45 ppi 5% Pt catalyst are shown in Figure 7-1. The SEI detector image shows the structure of the catalyst and support while the BSE detector shows different solid phases. The lighter phase in the BSE detector is the metal Pt while the dark phase is alumina. This shows that the Pt 127

143 metal is dispersed on the alumina catalyst in particle sizes ranging from approximately 1 nm to 1 or 2 μm. Studying how the dispersion and the loading of the catalyst changes over time and also comparing the physical characteristics of Rh and Pt metal on alumina are interesting topics that should be explored. References [1] Hickman DA, Schmidt LD. Production of syngas by direct catalytic oxidation of methane. Science. 1993; 259: [2] Hickman DA, Schmidt LD. Steps in CH 4 oxidation on Pt and Rh surfaces: Hightemperature reactor simulations. AIChE Journal. 1993; 39: [3] Subramanian R, Panuccio GJ, Krummenacher JJ, Lee IC, Schmidt LD. Catalytic partial oxidation of higher hydrocarbons: reactivities and selectivities of mixtures. Chemical Engineering Science. 24; 59: [4] Deluga GA, Salge JR, Schmidt LD, Verykios XE. Renewable hydrogen from ethanol by autothermal reforming. Science. 24; 33: [5] Dauenhauer PJ, Salge JR, Schmidt LD. Renewable hydrogen by autothermal steam reforming of volatile carbohydrates. Journal of Catalysis. Submitted; 128

144 (A) (B) Figure 7-1. SEM micrographs of a used 45 ppi 5% Pt catalyst at the same location using (A) SEI and (B) BSE detectors. The SEI detector shows physical structure. The white areas in the BSE micrograph is Pt and the dark area is alumina. 129