Catalytic Abatement of Environmental Pollutants and Greenhouse Gases in Automotive, Natural Gas Vehicles, and Stationary Power Plant Applications

Transcription

1 Catalytic Abatement of Environmental Pollutants and Greenhouse Gases in Automotive, Natural Gas Vehicles, and Stationary Power Plant Applications Qinghe (Angela) Zheng Submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in the Graduate School of Arts and Sciences COLUMBIA UNIVERSITY 2016

2 2016 Qinghe (Angela) Zheng All rights reserved

3 ABSTRACT Catalytic Abatement of Environmental Pollutants and Greenhouse Gases in Automotive, Natural Gas Vehicles, and Flue Gas Applications Qinghe (Angela) Zheng The present dissertation covers three research topics on catalytic environmental emissions control, including (1) aging and regeneration mechanisms of Rh- and Pd- model three-way catalysts (TWC) for gasoline automotive emission control, (2) catalytic methane emissions abatement from natural gas vehicles, and (3) scale-up of CO 2 capture and methanation using dual functional catalytic materials. The study resulted in two peer-reviewed publications, two future papers and one patent application which is currently under review. Modern TWC use supported two separate catalyst layers on a monolith containing one Pd and the other Rh for the emissions control of CO, HC and NO x. The rhodium (Rh) metallic component (active for NO x reduction) experiences the most severe oxidative thermal deactivation (forming inactive Rh 3+ ) during fuel cutoff, an engine mode (e.g., at downhill coasting) used for enhancing fuel economy. In a subsequent switch to a slightly fuel rich condition, in situ catalyst regeneration is accomplished by the reduction of the Rh 3+ with H 2 generated through steam reforming catalyzed by residual Rh 0 sites. The present thesis reports the effects of the deactivation and regeneration processes on the activity, stability and structural properties of 0.5% Rh/Al 2 O 3 and 0.5% Rh/Ce x O y - ZrO 2 (CZO) as model catalysts. Both materials are used to varying extents in modern TWC. A very brief introduction of three-way catalysis and system considerations will be presented.

4 During simulated fuel cutoff, catalyst deactivation is accelerated with increasing aging temperature from 800 C to 1050 C. Rh on a CZO support experiences less deactivation and faster regeneration than Rh on Al 2 O 3. Catalyst characterization techniques including BET surface area, CO chemisorption, temperature programmed reduction, and x-ray photoelectron spectroscopy, transmission electron microscopy, scanning-electron microscopy, and x-ray diffraction measurements were applied to examine the role of metal-support interactions in each catalyst system. For Rh/Al 2 O 3, strong metal-support interactions leading to the formation of a stable rhodium aluminate (Rh(AlO 2 ) y ) complex dominates during fuel cutoff, resulting in more difficult catalyst regeneration (reduction). For Rh/CZO, Rh sites were partially oxidized to Rh 2 O 3 and were relatively easy to be reduced to active Rh 0 during regeneration. Moderate Pd and support sintering of Pd-Ce x O y is experienced upon aging, but with a minimal effect on the catalyst activity. Cooling in air, following aging, was not able to reverse the metallic Pd sintering by re-dispersing to PdO. Unlike the aged Rh-TWCs, reduction via in situ steam reforming (SR) of exhaust HCs was not effective in reversing the deactivation of aged Pd/Al 2 O 3, but did show a slight recovery of the Pd activity when CZO was the carrier. The Pd + /Pd 0 and Ce 3+ /Ce 4+ couples in Pd/CZO are reported to promote the catalytic SR by improving the redox efficiency during the regeneration, while no such promoting effect was observed for Pd/Al 2 O 3. A suggestion is made for improving the catalyst performance. The use of natural gas for vehicle applications is growing in popularity due to advanced fracking technology. Exhaust methane has been excluded from regulations since it does not participate in photochemical reactions. New vehicle environmental regulations are expected for controlling methane emissions given their contribution to the greenhouse gas effects. Methane is extremely resistant to oxidation when the natural gas-fueled engine operates in the stoichiometric

5 mode with a supported Rh-Pd three-way catalyst (TWC). Furthermore, vehicles will still operate with fuel cutoff (for enhanced fuel economy), resulting in thermal oxidative deactivation (1050 o C) of the Rh metal in TWC to inactive Rh 3+, resulting in a loss of both NO x and methane abatement activity. When the engine returns to the slightly rich mode, H 2 generated by methane steam reforming does not readily occur to reduce and regenerate the Rh. We report a solution to methane emissions abatement by catalytic reforming of an injected aqueous solution of ethanol into the simulated exhaust stream in TWC mode, which generates sufficient H 2 to regenerate especially Rh by reducing Rh 3+ to its metallic active state. Conventional CO 2 capture and sequestration (CCS) in aqueous alkaline solutions is a very energy-intensive process with relative unstable performance and low efficiency especially for power plant effluents, and therefore there is a need for new approaches to control green house gas emissions of CO 2. Here we report on progress with an advanced technology involving CO 2 adsorption from flue gas and synthetic natural gas production, via methanation, both performed at the same temperature with the addition of renewable H 2 and by using a dual functional material (DFM). The stored H 2 used is produced by water electrolysis during those times when solar, wind, and other alternative energies generate excess power out of phase with the direct use of the electricity. The DFM is composed of nano-dispersed CaO (or Na 2 CO 3 ) and Ru metal supported on γal 2 O 3 carrier, respectively functioning as the CO 2 adsorbent and methanation catalyst. The present paper focuses on a laboratory scale-up study by using a simulated flue gas and 5%Ru,10%CaO/Al 2 O 3 and 5% Ru,10%Na 2 CO 3 /Al 2 O 3 DFM samples. The effects of DFM preparation methods, Al 2 O 3 carrier materials (with different shapes and properties), and adsorption and methanation conditions (feed compositions, flow rates, reaction temperatures) on the DFM performance were examined. Samples were prepared using chloride precursor salts and showed

6 stable performance under pseudo scale-up conditions, with SASOL TH100 Al 2 O 3 (with the highest BET surface area and pore volume/radius among the support materials) exhibiting the best performance. Compared to Ru-CaO, Ru-Na 2 CO 3 based DFM materials showed improved CO 2 utilization and methanation production. Reaction conditions were explored to find optimized CO 2 adsorption and methanation.

7 Table of Contents List of Figures... iv List of Tables... xi Nomenclature Introduction Catalytic emissions abatement of carbon monoxide, hydrocarbons, and nitrogen oxides Automotive emission control using a Rhodium-Palladium three-way catalyst Catalyst deactivations during fuel cutoff process Catalyst regeneration by fuel rich operation Catalytic methane emissions abatement on natural gas vehicles Carbon dioxide emissions control in post-combustion flue gas Carbon dioxide emissions control by sorption/desorption based technologies Carbon dioxide emissions control using dual functional catalytic materials Objective of the thesis: Catalytic emission control in automotive, natural gas vehicles, and flue gas applications Experimental methodologies Aging and regeneration study of automotive three-way catalysts Catalyst materials Simulated fuel cutoff aging and fuel rich regeneration processes Catalyst regenerability as measured at simulated fuel rich condition Catalyst stability during simulated fuel cutoff aging-fuel rich regeneration cycle tests Data analysis Catalyst characterization Catalytic methane emissions abatement study Catalyst materials In situ catalyst pre-reduction, aging, and regeneration Catalytic methane conversion activity tests...26 i

8 Reaction thermodynamic modelling Scale-up CO 2 adsorption and methanation study Material preparation CO 2 adsorption-methanation cycle tests Process parametric study for the adsorption and methanation Results and discussion Aging and regeneration mechanisms of Rh-based gasoline three-way catalysts (TWC) Reaction thermodynamics at simulated engine fuel rich condition Catalyst deactivation and regeneration of Rh-TWCs Rh-TWCs stability during fuel cutoff aging-fuel rich regeneration cycle tests Catalyst deactivation and regeneration mechanisms for Rh-TWCs Aging and regeneration mechanisms of Pd-based three-way catalysts (TWC) Aging-induced Pd sintering: the primary catalyst deactivation mode Support sintering and Pd-support interaction: other catalyst deactivation modes Catalytic methane emissions abatement on natural gas vehicles by steam reforming Reaction thermodynamics for reforming of methane, propane, or ethanol Catalytic methane reforming on fresh catalysts Catalytic methane reforming on aged catalysts Methane emissions abatement by catalytic oxidation on fresh and aged catalysts Catalyst regeneration by reforming of ethanol Catalyst regeneration by reforming of propane Periodic air aging of Pd-TWCs Regeneration of Rh-TWCs by ethanol reforming in the presence of Pd-TWC Scale-up CO 2 capture and methanation with dual functional materials (DFM) Cyclic tests of CO 2 adsorption and conversion with 5%Ru,10%CaO (or Na 2 CO 3 ) DFM on various Al 2 O 3 carriers Effects of reaction parameters on the CO 2 capture and methanation capabilities Conclusions...90 ii

9 4.1. Aging and regeneration mechanisms of Rh-based three-way catalysts (TWC) Aging and regeneration mechanisms of Pd-based three-way catalysts (TWC) Catalytic methane emissions abatement on natural gas vehicles by steam reforming Scale-up CO 2 capture and methanation with dual functional materials Significance, novelty, and future work...94 References...96 iii

10 List of Figures Figure 1. A washcoated monolith automotive TWC catalyst....2 Figure 2. The TWC conversion profile as a function of air-to-fuel ratio....3 Figure 3. A schematic of the unit operations in the exhaust system for a TWC with feed back control of air-to-fuel ratio (λ)...4 Figure 4. Scanning Electron Microscopic (SEM) images of fresh (a) 0.5% Rh/Al 2 O 3 and (b) 0.5% Rh/CZO, (c) 3% Pd/Al 2 O 3, (d) 3% Pd/Al 2 O 3, (e) 1% Pd/CZO, and (f) 0.5% Pd/CZO model TWC catalysts at µm scale. SEM measurement condition: beam voltage of 20 kv, beam current of 10 µm, working distance of 12 mm, and 30 µm in scale Figure 5. Schematic of the packed bed flow reactor and analysis system. (MFC: Mass Flow Controller, GC: Gas Chromatography, TI: Temperature Indicator) Figure 6. Schematic process flow diagrams of (a) simulated fuel cutoff aging-fuel rich regeneration cycle and activity test; and (b) on-board gasoline engine fuel cutoff-fuel rich operation cycles Figure 7. Schematic catalytic flow reactor setup for in situ catalyst pre-reduction, aging and regeneration, CH 4 reforming/oxidation activity measurements, and cycle tests at different reaction conditions Figure 8. Schematic process flow for the catalyst preparation of 5%Ru,10%CaO/Al 2 O 3 DFM on different Al 2 O 3 support materials and 5%Ru,10%Na 2 CO 3 /Al 2 O 3 DFM on TH100 Al 2 O 3 support Figure 9. Reactor setup for the CO 2 adsorption-methanation cycle tests and process parameter study Figure 10. Schematic reaction process flow for CO 2 adsorption-methanation cycle tests Figure 11. Reaction Gibbs free energy as a function of reaction temperature (25 C to 700 C) at 1 atm. Assume ideal gas behavior for the reactant and product gas components. Compound thermodynamic data with temperature and pressure inputs is collected from I. Barin, Thermochemical Data of Pure Substances (3rd Edition) [157].33 Figure 12. (a) Main mole fractions of H 2, CO, and CH 4 ; and (b) theoretical reactant (propane and water) conversions as a function of reaction temperature (200 C to 550 C) at thermodynamic equilibrium conditions. Reactant feed: 500 vppm propane, 10 vol-% steam, 8 vol-% CO 2, N 2 in balance Figure 13. Catalyst activity of fresh and aged (a) 0.5% Rh/Al 2 O 3 and (b) 0.5% Rh/CZO Catalyst activity is plotted in terms of H 2 mole fraction as a function of reaction temperature (200 C 550 C) iv

11 Figure 14. Activity of fresh, aged, and regenerated (a) 0.5% Rh/Al 2 O 3 and (b) 0.5% Rh/CZO in converting simulated engine exhaust gas at fuel rich condition. Catalyst activity is plotted in terms of H 2 product mole fraction as a function of reaction temperature (200 C to 550 C). Aged catalysts were obtained by treating fresh ones in air at 1050 C for 5 min, followed by cooling to room temperature in air. Catalyst regenerations were performed by at rich condition at 550 C for 1 h Figure 15. H 2 generations during regeneration processes in simulated fuel cutoff aging- fuel rich regeneration cycle tests (First 5 cycles out of total 25 cycles) with (a) 0.5% Rh/Al 2 O 3 and (b) 0.5% Rh/CZO. In each cycle, the catalyst sample was first aged in air at 1050 C for 5 min, followed by in situ regeneration at propane rich condition at 550 C for 1 h Figure 16. Activity of regenerated (a) 0.5% Rh/Al 2 O 3 and (b) 0.5% Rh/CZO catalysts after 1, 5, 10, 15, 20, and 25 cycles in the aging-regeneration cycle tests. After every five cycles of aging/regeneration. The catalyst activity is plotted in terms of H 2 product mole fraction as a function of SR temperature from 250 C to 550 C Figure 17. BET surface areas (based on single-point measurement, 10% error allowed) of fresh and aged (a) 0.5% Rh/Al 2 O 3 and (b) 0.5% Rh/CZO, as a function of aging conditions. Aged samples were obtained by aging fresh catalysts (I) in air at the following conditions: (II) 800 C for 5 min; (III) 800 C for 1 h; (IV) 950 C for 5 min; or (V) 1050 C for 5 min. The aging processes were followed by cooling in air to room temperature. As a reference, BET surface areas of support materials were measured m 2 /g and 60.3 m 2 /g respectively for fresh Al 2 O 3 and CZO Figure 18. Normalized H 2 consumption in H 2 -Temperature Programmed Reduction (H 2 -TPR) measurements of fresh and aged (a) 0.5% Rh/Al 2 O 3 and (b) 0.5% Rh/CZO, as a function of reducing temperature. Aged samples for measurements were respectively achieved by aging the fresh ones in air at the following conditions: 800 C for 5 min, 800 C for 1 h, 950 C for 5 min, or 1050 C for 5 min Figure 19. Speculative schematic of proposed redox reaction mechanism and interaction between Rh δ+ /Rh 0 and Ce 4+ /Ce 3+ redox couple during H 2 -TPR of 0.5% Rh/CZO. The redox reactions followed the order described below. (I) After simulated fuel cutoff aging in air at 800 C, 950 C, or 1050 C, surface Rh sites are oxidized to Rh 2 O 3, while the Rh sites in close contacts with Ce x O y remained in reduced states (Rh δ+, 0 < δ < 1), with Rh δ+ /Rh 0 and Ce 4+ /Ce 3+ redox couple formed for enhancing electron transfer efficiency; (II) H 2 flow through the sample; (III) At low temperature regime around 100 C to 120 C, H 2 was chemisorbed and dissociated on the v

12 Rh 0 sites, followed by (IV) Reduction of Rh 3+ to Rh 0 ; (V) Reduction of surface Ce 4+ sites to Ce 3+ promoted by the Rh δ+ /Rh 0 and Ce 4+ /Ce 3+ redox couple; (VI) Reduction of bulk Ce 4+ sites to Ce 3+ when more H 2 molecules were chemisorbed and dissociated on Rh Figure 20. X-ray Photoelectron Spectroscopy (XPS) multiplex spectra in Rh 3d region (with BE of 318 ev 304 ev) of fresh, aged, and regenerated (a) 0.5% Rh/Al 2 O 3 and (b) 0.5% Rh/CZO powder catalysts, and with aging temperature varied. Aged samples were achieved by aging the fresh catalysts in air at 800 C, 950 C or 1050 C for 5 min. Regenerated samples were achieved by regenerating the aged ones (1050 C for 5 min) using the method as described in Section Figure 21. Proposed reaction mechanism and electron transfer pathways for steam reforming of propane on (a) Rh/Al 2 O 3 and (b) Rh/CZO catalysts Figure 22. Activity of fresh and aged (a) 3% Pd/Al 2 O 3 and (b) 3% Pd/CZO in converting simulated engine exhaust gas at fuel rich condition. Catalyst activity is plotted in terms of H 2 product mole fraction as a function of reaction temperature (200 C to 550 C). Aged catalysts were obtained by treating fresh ones in air at 800 C, 950 C, or 1050 C for 5 min, followed by cooling to room temperature in air Figure 23. Activity of fresh, aged, and attempted regenerated (a) 3% Pd/Al 2 O 3 and (b) 3% Pd/CZO in converting simulated engine exhaust gas at fuel rich condition. Aged catalysts were obtained by treating fresh ones in air at 1050 C for 5 min, followed by cooling to room temperature in air. Attempted catalyst regenerations were performed by SR at rich condition at 550 C for 1 h Figure 24. Activity of attempted regenerated (a) 3% Pd/Al 2 O 3 and (b) 3% Pd/CZO catalysts after 1, 5, 10, and 20 cycles in the aging-regeneration cycle tests. After every five cycles of aging/regeneration. The catalyst activity is plotted in terms of H 2 product mole fraction as a function of SR temperature from 250 C to 550 C Figure 25. H 2 productions as a function of time on stream (TOS) during attempted regeneration processes in simulated fuel cutoff aging-attempted fuel rich regeneration cycle tests (First 5 cycles out of total 25 cycles) with (a) 3% Pd/Al 2 O 3 and (b) 3% Pd/CZO. In each cycle, the catalyst sample was first aged in air at 1050 C for 5 min, followed by attempted in situ regeneration at fuel rich condition at 550 C for 1 h Figure 26. Activity of fresh Pd/CZO catalysts with different metal loadings (0%, 0.5%, 1%, and 3%) in converting simulated engine exhaust gas at fuel rich condition. Catalyst activity is plotted in terms of H 2 product mole vi

13 fraction as a function of reaction temperature (200 C to 550 C). Reaction feed: 500 ppm propane, 10% H 2 O, and N 2 balance, with total GHSV of 30,000 h Figure 27. Representative TEM images of fresh, aged, and attempted regenerated 3% Pd/Al 2 O 3 (a c) and 3% Pd/CZO (d f). Aged samples were achieved by aging the fresh catalyst in air or N 2 at 1050 C for 5 min. Attempted catalyst regenerations were performed by at rich condition at 550 C for 1 h Figure 28. TEM-derived Pd metal particle size distributions of fresh, aged, and attempted regenerated (a) 3% Pd/Al 2 O 3 and (b) 3% Pd/CZO. Aging and attempted regeneration conditions were as described in Figure Figure 29. BET surface areas of fresh and aged (a) 3% Pd/Al 2 O 3 and (b) 3% Pd/CZO, as a function of aging conditions. Aged samples were obtained by aging the (I) fresh catalysts in air at the following conditions: (II) 800 C for 5 min, (III) 800 C for 1 h, (IV) 950 C for 5 min, or (V) 1050 C for 5 min. The aging processes were followed by cooling in air to room temperature. As a reference, BET surface areas of support materials were measured m 2 /g and 60.3 m 2 /g respectively for fresh Al 2 O 3 and CZO Figure 30. X-ray diffraction (XRD) patterns for (a) Fresh 3% Pd/Al 2 O 3 ; (b) Aged 3% Pd/Al 2 O 3, (c) Fresh 3% Pd/CZO; (d) Aged 3% Pd/CZO; and the zoomed views of (c,d). Aging condition: 1050 C in air for 5 min. XRD patterns were obtained by using Cu-Kα 1 radiation (λ = Å) Figure 31. Normalized H 2 consumption in H 2 -Temperature Programmed Reduction (H 2 -TPR) measurements of fresh and aged (a) 3% Pd/Al 2 O 3 and (b) 3% Pd/CZO, as a function of reducing temperature. Aged samples for measurements were respectively achieved by aging the fresh ones in air at the following conditions: 800 C for 5 min, 800 C for 1 h, 950 C for 5 min, or 1050 C for 5 min Figure 32. Schematic of proposed model for deactivation and regeneration of 3% Pd/Al 2 O 3 and 3% Pd/CZO during simulated fuel cutoff aging (1050 C) and fuel rich regeneration (550 C) Figure 33. Thermodynamic model of equilibrium compositions for reforming reactions of different hydrocarbons: (a) Methane, (b) Propane, or (c) Ethanol. Equilibrium product gas mole fractions are plotted as a function of reaction temperature (200 o C and 700 o C). The theoretical model is constructed based on a feed composition (vol-%) of 500 ppm HC, 10% steam, 8% CO 2, and N 2 in balance (81.95%), at 1 bar Figure 34. Activity of methane co-reforming with steam and CO 2 on fresh Rh- or Pd- model TWC catalyst (nonprereduced). (a) CH 4 conversion, (b) H 2, and (c) CO product mole fractions are plotted against the reforming temperatures (300 o C to 550 o C) with different fresh including 0.5% Rh/Al 2 O 3, 0.5% Rh/CZO, 3% Pd/Al 2 O 3, vii

14 and 3% Pd/CZO. Reforming feed composition in vol-% is 500 ppm CH 4, 10% steam, 8% CO 2, and N 2 in balance (81.95%), with a total GHSV of 120, 000 h Figure 35. Effects of catalyst fuel cutoff aging on the activity of of Rh-TWCS on methane reforming. with fresh (nonprereduced) or aged (a) 0.5% Rh/Al 2 O 3, (b) 0.5% Rh/CZO, (c) 3% Pd/Al 2 O 3, or (d) 3% Pd/CZO catalysts. The methane reforming feed is the same as described in Figure 34. Catalyst aging protocol includes treating fresh sample with excess air exposure at 1050 o C for 5 min, followed by cooling in air Figure 36. Activity of CH 4 reforming and CH 4 oxidation in the CH 4 abatement with Rh- or Pd- model TWC catalyst. CH 4 conversion profiles are plotted against reaction temperature during the isothermal CH 4 reforming and CH 4 oxidation activity tests with fresh (non-prereduced), aged, or regenerated (a) 0.5% Rh/Al 2 O 3, (b) 0.5% Rh/CZO, (c) 3% Pd/Al 2 O 3, or 3% Pd/CZO catalyst. The oxidation activity tests were performed with a feed consisting of (in vol-%) 500 ppm CH 4, 0.96 % O 2, 10% steam, 8% CO 2, and N 2 in balance (80.99%). Aging protocol includes excess air exposure of the fresh sample at 950 o C for 5 min, followed by cooling in air. The regenerated catalysts for CH 4 oxidation reaction were achieved by ethanol reforming with the aged ones at 550 o C for 1 hr Figure 37. CH 4 reforming activity against reaction temperatures and cycles during the aging-regeneration (by ethanol reforming)-activity cycle tests with (a) 0.5% Rh/Al 2 O 3 and (b) 0.5% Rh/CZO catalysts. All fresh catalysts were pre-reduced. In each cycle, aging was performed in excess air at 1050 o C for 5 min, followed by cooling in air. Regeneration was performed after aging via ethanol reforming at 550 o C for 1 hr. CH 4 reforming activity tests were performed with regenerated catalysts after 1, 5, 10, 15, and 20 aging-regeneration cycles Figure 38. H 2 product mole fraction as a function of regeneration time on stream (TOS) during the aging-regeneration (by ethanol reforming)-activity cycle tests as described in Figure Figure 39. CH 4 reforming activity against reaction temperatures and cycles during the aging-regeneration (by propane reforming)-activity cycle tests with (a) 0.5% Rh/Al 2 O 3 and (b) 0.5% Rh/CZO catalysts. All fresh catalysts were pre-reduced. In each cycle, aging was performed in excess air at 1050 o C for 5 min, followed by cooling in air. Regeneration was performed after aging via propane reforming at 550 o C for 30 min. CH 4 reforming activity tests were performed with regenerated catalysts after 1, 3, 5, 10, 15, and 20 aging-regeneration cycles Figure 40. H 2 product mole fraction as a function of regeneration time on stream (TOS) during the aging-regenerationactivity cycle tests as described in Figure viii

15 Figure 41. Air aging length effect on the cyclic CH 4 reforming activity with (a) 3% Pd/Al 2 O 3 or (b) 3% Pd/CZO catalyst. Catalyst activity was measured after different TOSs (5, 25, 50, 75, 100, and 200 min) of aging at 950 o C or 1050 o C. For 3% Pd/Al 2 O 3, moderate catalyst deactivation was observed after 950 o C aging for 100 min. Further aging at 1050 o C for longer term resulted in slight loss of catalyst light-off activity. For 3% Pd/CZO, moderate catalyst deactivation was observed after direct aging at 1050 o C for 200 min Figure 42. Comparison between ethanol reforming (orange curves) and methane reforming (purple curves) for the regeneration of aged (1050 o C in air for 5 min) (a) 0.5% Rh/Al 2 O 3 or (b) 0.5% Rh/CZO in the presence of 3% Pd/CZO. For the aging and regeneration with Rh-Pd catalyst mixture, Rh catalyst was placed immediately underneath the 3% Pd/CZO layer, and was separated from Pd with a quartz wool layer sandwiched in between. In the presence of Pd catalyst, aging was performed in air at 1050 o C for 5 min. and regeneration was performed by ethanol or methane reforming at 550 o C form 1 hr Figure 43. CO 2 adsorption and methanation performance of 5%Ru10%CaO (or 10%Na 2 CO 3 ) on different Al 2 O 3 supports for cyclic testing. The amounts of (a) CO 2 captured and (b) O 2 consumed at the adsorption step, and (c) CH 4 produced during the methanation step for 5%Ru,10%CaO prepared on (I) Al 2 O 3 powder, (II) Al 2 O 3 pellets-sasol TH200, (III) Al 2 O 3 pellets-sasol TH100, and (IV) Al 2 O 3 beads supports, and (V) 5%Ru,10%Na 2 CO 3 on Al 2 O 3 pellets TH100 support. CO 2 adsorption and methanation temperatures for all studies was maintained at 320 o C Figure 44. CO 2 adsorption and methanation performance of 5%Ru,10%Na 2 CO 3 on different Al 2 O 3 supports for cyclic testing. CO 2 adsorption feed excluded steam and O 2, with composition of 7.5% CO 2, 0% steam, 0% O 2, N 2 bal., at a flow rate of 26 L/hr Figure 45. Effect of adsorption time on stream (TOS) (60 min, 20 min, or 10 min) on CO 2 capture and methanation capabilities. DFM material loading: 5%Ru,10%CaO/Al 2 O 3 pellets (SASOL TH200) of around 10 g. Adsorption condition: 7.5% CO 2, 15% steam, 4.5% O 2, N 2 bal. (vol-%); GHSV=11236 h -1 ; 320 o C; 1 bar. Methanation condition: 5 vol-% H 2 / N 2 ; GHSV= 2611 h -1 ; 320 o C; 1 bar, 60 min...84 Figure 46. Effect of reaction temperature (350 o C, 320 o C, 300 o C, or 280 o C) on the CO 2 capture and methanation capabilities. DFM material loading: 5%Ru,10%CaO/Al 2 O 3 pellets (SASOL TH200) of around 10 g. Adsorption condition: 7.5% CO 2, 15% steam, 4.5% O 2, N 2 bal. (vol-%); GHSV=11236 h -1 ; 1 bar; 20 min. Methanation condition: 5 vol-% H 2 / N 2 ; GHSV= 2611 h -1 ; 1 bar, 60 min. Legends refer to volumes of respective gases ix

16 Figure 47. Effect of adsorption feed flow rate (48.21, 40.00, 32.36, or L/hr) on the CO 2 capture and methanation capabilities. DFM loading: 5%Ru,10%CaO/Al 2 O 3 pellets (SASOL TH200) of around 10 g. Adsorption condition: 7.5% CO 2, 15% steam, 4.5% O 2, N 2 bal. (vol-%); 320 o C; 1 bar; 20 min. Methanation condition: 5 vol-% H 2 / N 2 ; GHSV= 2611 h -1 ; 320 o C, 1 bar, 60 min. Legends refer to volumes of respective gases Figure 48. Effect of methanation (H 2 /N 2 ) feed flow rate (22.4, 11.2, 5.6, or 2.8 L/hr) on the methanation capability. DFM catalyst loading: 5%Ru,10%CaO/Al 2 O 3 pellets (SASOL TH200) of around 10 g. Adsorption condition: 7.5% CO 2, 15% steam, 4.5% O 2, N 2 bal. (vol-%); GHSV=11236 h -1 ; 320 o C; 1 bar; 20 min. Methanation condition: 5 vol-% H 2 / N 2 ; 320 o C, 1 bar, 60 min. Legends refer to volumes of respective gases Figure 49. Effect of the presence of steam and/or O 2 in adsorption feed on the CO 2 adsorption and methanation capability. DFM material loading: 5%Ru10%CaO/Al 2 O 3 pellets (SASOL TH200) of around 10 g. Adsorption condition: GHSV=11236 h -1 ; 320 o C; 1 bar; 20 min. Methanation condition: 5 vol-% H 2 / N 2 ; GHSV= 2611 h -1 ; 320 o C, 1 bar, 60 min Figure 50. Proposed schematic mechanism of the surface reactions on Ru,CaO/Al 2 O 3 DFM for CO 2 adsorption and methanation x

17 List of Tables Table 1. Material properties of the alumina supports used for catalyst preparation Table 2. Main reaction pathways during catalytic conversions of simulated exhaust feed at fuel rich condition Table 3. T 50 (temperature at which 50% maximum equilibrium H 2 production is reached during activity tests, C) of fresh, aged, and regenerated catalysts Table 4. Metal dispersions (%) of fresh and aged 0.5% Rh/Al 2 O 3 and 0.5% Rh/CZO catalysts as measured by room temperature CO chemisorption a. After simulated fuel cutoff, loss of active metal sites occurred for both Rh catalysts, and was accelerated with aging temperature from 800 C to 1050 C Table 5. Mean metal particle sizes (nm) of fresh and aged TWCs as measured using TEM images Table 6. Reduction temperature (T R ) and H 2 consumption during H 2 -TPR for fresh and aged 0.5% Rh/Al 2 O 3 samples Table 7. Summary of detailed information of XPS spectra as shown in Figure 20, i.e., values of binding energy Rh 3d 3/2 for Rh 3+ and Rh 0 oxidation states, and Rh 3+ /Rh 0 (or Rh 3+ /Rh δ+ ) ratios a for fresh, aged, and regenerated 0.5% Rh/Al 2 O 3 and 0.5% Rh/CZO samples Table 8. T 50 s ( C) of fresh, aged, and attempted regenerated 3% Pd/Al 2 O 3 and 3% Pd/CZO, in comparison with 0.5% Rh/Al 2 O 3 and 0.5% Rh/CZO as reported Table 9. Support mean crystallite sizes of the fresh and aged (1050 C in air for 5 min) 3% Pd/Al 2 O 3 and 3% Pd/CZO measured by TEM and XRD Table 10. Main reaction pathways during reforming of a specific hydrocarbon fuel, i.e. (a) methane, (b) propane, or (c) ethanol, in the presence of H 2 O (steam) and CO 2 in the reaction feed Table 11. Average performance (in 10 cycles) for CO 2 adsorption and methanation for 5%Ru,10%CaO/Al 2 O 3 or 5%Ru,10%Na 2 CO 3 /Al 2 O 3 on different Al 2 O 3 carriers. The performance of # I~V materials were evaluated by integration of the adsorption and methanation curves in Figure 43. #VI DFM was evaluated by using Figure 44, when steam and O 2 were excluded from the CO 2 borne feed xi

18 Acknowledgements First and foremost, I would like to express my most sincere thanks and reverence to my advisor, Professor Robert Farrauto, for his valuable guidance, persistent patience, and constant encouragement throughout my entire masters-doctoral study and research. Professor Farrauto has not only provided me with extraordinary professional knowledge and expertise through his wonderful lectures and detailed research instructions, but also inspired me as a role model by being an outstanding scientist through his rigorousness, perspicacity, and efficiency in his academic study, and great passion for his work and life. He cares about every person around him, and always tries his best to offer his generous help. I have received tremendous help from him in every aspect of my study, life, and career, and have been influenced by him in every way of always being a reliable friend and family to others. Having Professor Farrauto as my PhD advisor has been one of the most wonderful experiences in my life. My heartfelt gratitude especially goes to my esteemed doctoral committee members: Professor Robert Farrauto, Professor Paul Duby, Dr. Michel Deeba (BASF), Professor Ngai Yin Yip, and Dr. Christoph Meinrenken. My doctoral study is enriched by being guided by these highly expertized scientists from different research disciplines and backgrounds. It has been my great honor to have the opportunity to learn from them. I would like to acknowledge their valuable advices, enlightening instructions, generous help, and kind concern for my research and study. Also, I m thankful to those professors who instructed me in courses from the Earth and Environmental Engineering, Chemical Engineering, Material Science, Applied Physics and Math, and Mechanical Engineering Departments. The wonderful lectures I have had at Columbia xii

19 University broadened my horizon, which greatly improved my decision-making skill during my doctoral research study. I would also like to express my sincere gratitude to Dr. Christiane Janke at BASF, Professor Martha Cobo at University de La Sabana, and Dr. Ioannis Valsamakis at Joule Scientific for their selfless training, guidance, help, and care in my earlier laboratory research at Columbia. They taught me with excellent experimental skills, and let me know how to be an independent researcher. Meanwhile, I d like to extend my sincere appreciation to my former and present colleagues and groupmates: Dr. Amanda Simson, Dr. Melis Duyar, Ms. Emi Leung, Ms. Shuoxun (Ashley) Wang, Ms. Martha Arellano, Ms. Christine Wang, Ms. Anh Nguyen, Ms. Yi Li, Mr. Kyle Misquitta, and all the other members of our Catalysis for a Sustainable Environment group. I m grateful for their selfless help, assistance, love, and care during the time we worked and studied together. They are not only my groupmates, but also family to me. In addition, many thanks are extended to BASF for its financial support, and to the laboratories, facilities, and administrative support from the Earth and Environmental Engineering Department, Columbia University. Finally, I d like to dedicate this dissertation to my family for their infinite love and support throughout my education at Columbia University. xiii

20 Dedication The PhD dissertation is lovingly dedicated to my mother, Professor Ju Xu, and my father, Dr. Yizhou Zheng. Their support, encouragement, and constant love have sustained me throughout my life. xiv

21 Nomenclature A/F: Engine air-to-fuel ratio BET: Brunauer Emmett Teller CCS: Carbon capture and sequestration Ce: Cerium CNG: Compressed natural gas CZO: Cerium oxide-zirconia oxide DFM: Dual functional material EtOH: Ethanol GHG: Greenhouse gas GHSV: Gas-hourly space velocity HC: Hydrocarbon LNG: Liquefied natural gas MOF: Metal-organic framework NGV: Natural gas vehicle NMHC: Non-methane hydrocarbons Pd: Palladium Rh: Rhodium Ru: Ruthenium SCR: Selective catalytic reduction SEM: Scanning electron microscopy SR: Steam reforming TEM: Transmission electron microscopy TOS: Time on stream TPR: Temperature-programmed reduction TWC: Three-way catalyst XPS: X-ray photoelectron spectroscopy XRD: X-ray diffraction Zr: Zirconium λ: Ratio of air-to-fuel in the exhaust to the airto-fuel at stoichiometric 1

22 1. Introduction 1.1. Catalytic emissions abatement of carbon monoxide, hydrocarbons, and nitrogen oxides Automotive emission control using a Rhodium-Palladium three-way catalyst When the gasoline engine is operated around the stoichiometric air-to-fuel ratio (14.6 wt%, ±2%), a three-way catalyst (TWC) allows simultaneous conversions (~98%) of carbon monoxide (CO), non methane hydrocarbons (NMHCs) and oxides of nitrogen (NO x ) to innocuous compounds [1]. Specifically, the oxidation of CO and HCs (non-methane HCs) to CO 2 and steam (H 2 O) is catalyzed by Pd, while NO x is reduced to N 2 catalyzed by Rh [2]. Modern TWC uses stabilized γal 2 O 3 to support Pd and Rh in separate layers (washcoats) deposited on a ceramic or metallic monolith wall [3 5]. A cartoon of a washcoated monolith is shown as Figure 1. An oxygen storage component composed of Ce x O y -ZrO 2 is also used in some cases as a carrier in place of Al 2 O 3. Figure 1. A washcoated monolith automotive TWC catalyst. The TWC conversion profile is shown in Figure 2. CO and NMHCs are essentially fully oxidized in excess O 2 (lean of the stoichiometric air-to-fuel ratio, i.e. on the right side of the 2

23 stoichiometric point). NO x reduction occurs when little or no O 2 is present, as in the rich operating mode (left side of the stoichiometric point). The ratio of air-to-fuel in the exhaust to the air-to-fuel at stoichiometric is defined as the lambda point (λ). At stoichiometric operation, λ equals 1. The λ value is controlled via a feedback control system through a signal received from the O 2 sensor as shown in Figure 3. Figure 2. The TWC conversion profile as a function of air-to-fuel ratio. Gamma-Al 2 O 3 (γ-al 2 O 3 ), stabilized by incorporation of small percentages of La 2 O 3 and/or BaO is most widely used as a support for the catalytic components due to its excellent hydrothermal stability, and high specific surface area and porosity, which provide adequate metal dispersion of the precious metals [6]. Cerium oxide (Ce x O y ), well known for its high oxygen storage capacity (OSC) due to the function of Ce 4+ /Ce 3+ redox pair, is also included in modern TWC composition. The air-to-fuel ratio reversibly oscillates during lean/rich perturbations as a consequence of the feedback control strategy. The main function of the Ce x O y is to provide O 2 when λ < 1 for oxidation, and storing O 2 when λ > 1 to allow reduction to occur [2,7 9]. It may also be used in place of Al 2 O 3 as a carrier for the catalytic components. A schematic of the catalytic emission abatement system, with feedback control, is shown in Figure 3. Note a small TWC close 3

24 couple catalyst converter (next to the engine ports) gets hot faster than the main catalyst and initiates conversion more quickly. The O 2 sensor, after the main TWC, is for on-board diagnostics to inform the driver of a malfunction in the converter. Figure 3. A schematic of the unit operations in the exhaust system for a TWC with feed back control of air-to-fuel ratio (λ). The redox chemistry for the CeO x is shown as Eq. (1) and (2). Further incorporation of zirconium oxide (ZrO 2 ) into Ce x O y crystallite structure (denoted as CZO) improves the thermal stability of Ce x O y, and enhances the mobility of lattice oxygen through the formation of oxygen vacancies [10 13]. Other proprietary elements are also added to further enhance performance. 2CeO & + CO Ce & O ) + CO & (at slightly fuel rich, λ < 1) (1) Ce & O ) + + & O & 2CeO & (at slightly fuel lean, λ > 1) (2) Catalyst deactivations during fuel cutoff process Fuel cutoff has been practiced for many years for enhancing fuel economy by 4% 5%. This operational mode is automatically implemented when the vehicle is coasting down hill, usually 4

25 lasting a short period (seconds to minutes). During this operational mode, fuel injection is discontinued and air flows into the TWC converter, exposing catalyst to high surface temperatures (up to 1050 C), resulting in severe catalyst deactivation [14 16]. Rhodium needs to be in its metallic state (Rh 0 ) to maintain its activity for NO x reduction [2,17]. The deactivation modes include metal and/or support sintering [18,19], metal and/or support oxidation [20], and metal-support interactions [19 23]. It is widely accepted that the interaction between Rh and γ-al 2 O 3 during thermal oxidative exposure leads to the formation of stable and inactive Rhodium Aluminate Rh(AlO 2 ) y (Eq. 3) [2,17]. Rh & O ) + yal & O ) + y 3 2 O 344,789 & 2 Rh AlO& : (3) Interactions between Rh and Ce x O y, in combination with ZrO 2, have been discussed previously. High surface energy of Ce x O y favors metal-support interactions. The dissolution of Rh into the bulk ceria was observed after catalyst calcination at 550 C [24,25]. Meanwhile, Rh + O Ce and [Rh O 2 ] 2 species are likely formed in an oxidative environment [26 29]. At high temperature, treatment of Rh/CeO 2 in air leads to the formation of Rh 2 O 3 [30], slight Rh metal sintering [31,32], and segregation of Rh cations into the CeO 2 lattice [33,34]. Rh may also incorporate into the sublattice ZrO 2, leading to decrease in metal redox behavior [35]. This effect has a small effect on performance relative to Rh on Al 2 O Catalyst regeneration by fuel rich operation Two practical approaches have been employed to solve (or partially solve) the deactivation problem: (1) stabilizing the support by using/adding refractory materials to prevent negative Rhsupport interactions; and (2) regenerating the deactivated catalyst after fuel cutoff by operating 5

26 engine in the fuel rich (λ < 1) condition [2,15,36 42]. For the first approach, refractory materials as supports in place of gamma Al 2 O 3 include zirconia, titania, denser forms of alumina, and alkaline metal oxides [43,44] have been suggested. The second approach returns the operational mode to slightly fuel rich at ~500 C, which allows the creation of a reducing engine exhaust atmosphere to reduce the Rh 3+ to Rh metal partially reversing catalyst deactivation [2,45]. At slightly fuel rich conditions, the O 2 concentration is very low while considerable amounts of HCs and CO, along with excess steam and CO 2, are present in the exhaust. The oxidized catalytic components can be reduced by H 2 generated mainly through catalytic steam reforming (SR). The SR reaction, with propane as a model compound for the exhaust HC being reformed, is shown in Eq. (4) [46,47]. Fuel rich regeneration allows Rh 3+ to be reduced to active Rh 0, and released from the interaction with the support. Eq. (5) represents the reverse of Rh-Al 2 O 3 interaction by H 2. It is also possible for some dry reforming to occur where H 2 O is replaced with CO 2. C ) H 3 + 3H & O 3CO + 7H & (4) 2Rh AlO & : + yh & >?@A 98BC 2Rh + yal& O ) + yh & O (5) One advantage of this process is that the endothermic SR reaction (which is both thermodynamcially and kinetically favorable at high temperature) can be catalyzed by the TWC catalyst itself. In other words, the deactivated components in TWC can be regenerated in situ by the H 2 produced through SR, as catalyzed by the precious metal sites (mainly Rh but to a small extent by Pd and Ce x O y ) remaining active after aging. The presence of the large amount of steam (9%) and CO 2 (8%) makes the thermodynamic reaction favroable at low temperatures. 6

27 1.2. Catalytic methane emissions abatement on natural gas vehicles Natural gas (NG), primarily composed of methane, is readily available, it has a high H/C ratio, high research octane number (~130), and is regarded as a promising fuel to replace gasoline and diesel for internal combustion engines [48-52]. Consequently, the development of natural gas vehicles (NGV) by fueling liquefied (LNG) and compressed natural gas (CNG) is expected to grow. NGVs generally have lower emissions of CO 2, particulate matter, SO x, and non-methane hydrocarbons (NMHC) than conventional gasoline- or diesel-fueled vehicles. On the other hand, methane emissions are much higher [53]. Current vehicular emission regulations in US and Europe have excluded methane from the hydrocarbon limits, instead have enforced standards for the emissions of NMHCs. This is mainly because unlike other hydrocarbons, methane does not contribute to the formation of photochemical smog. However, it is a potent greenhouse gas (GHG) with a global warming potential more than 25 times that of CO 2 on a molar basis [54-56]. A consequence of this fact is the expected emergence of more stringent methane emission regulations in the coming years [57]. In response to the upcoming new regulations on road GHG emissions, there is a growing need to develop improved engine and catalyst converter designs for NGVs [58,59]. Current catalytic emission control is mainly performed under lean-burn or stoichiometric conditions. Lean-burn NGVs, similar to diesel vehicles, for many years have been in use especially for urban service operations such as delivery trucks, buses, etc. Lean operation is considered to provide high thermal efficiency, with reduced knocking and emissions [48,60]. NO x emissions are considered moderate due to the cooler combustion resulting from the high air-to-fuel ratios (A/F), at which the lean engine operates. A monolithic oxidation catalyst typically composed of Pd and/or Pt supported on Ce x O y -γal 2 O 3 is used specifically for catalyzing the oxidation of NMHCs and CO compounds 7

28 [61]. The main abatement problem with lean-burn emission control is the difficulty of oxidizing methane. Methane, which comprises the largest portion of the hydrocarbon emissions, is the most difficult of all alkanes to be catalytically oxidized [62]. Typical exhaust gases are characterized by low temperatures (even below 400 o C), excess amounts of steam (10-15 %) and oxygen (< 8 %), with ppm of methane emitted [63]. Under such conditions, a Pd catalyst, with PdO x as the active phase, is considered the best catalyst in oxidizing methane at appreciable rates, while its oxidation activity is crystallite size and thermally sensitive [58, 63-75]. Pt is also reported to be an effective methane oxidation catalyst [76,77]. However, considerable methane conversion can only be reached at oxidation temperatures as high as 600 o C, which the lean exhaust environment usually does not attain. Another problem is NO x emissions abatement. In order to meet the increasingly stringent environmental regulation, i.e. Euro VI standard (< 2 g NO x /kwh, 2014), NO x after-treatment should be performed on lean-burn NGVs, e.g. by using urea-selective catalytic reduction (SCR) [78-82]. This requires undesired additional cost and space for an updated catalytic converter. Three-way catalysts (TWC), for treating the exhaust from natural gas engines at stoichiometric conditions, is a logical approach to comply with the new emission standards [83,84]. The TWC converter is universally used in automotive industry for the simultaneous conversion of unburned hydrocarbons, NO x and CO from the exhaust from gasoline-fueled engines. Modern systems are composed mainly of separate layers of Rh and Pd supported on stabilized γal 2 O 3, washcoated on a monolithic structure. Rh is active in catalyzing NO x reduction partially with H 2, while Pd functions in the oxidation of NMHCs and CO. The catalyst performance is further improved by the addition of ceria-zirconia (Ce x O y -ZrO 2, denoted as CZO), which provides the catalyst with the necessary oxygen storage capacity (OSC), for chemically assisting 8

29 the computer feed-back control of the air-to-fuel stoichiometry [50,85]. For natural gas stoichiometric engines, one can use exhaust gas recirculation (EGR) for improved NO x reduction [52,76,84,86]. Meanwhile, Pd is enriched for enhanced NMHC oxidation [75,77,87]. Stoichiometric NGVs have comparatively low exhaust concentrations of O 2 (< 2%), consequently significant reductions of NO x (by 99.9%) and NMHC emissions (by 90-97%) can be achieved [48,88], while CH 4 cannot be converted effectively [58,89]. The low conversion of methane occurs because the stoichiometric air-to-fuel ratio provides less O 2 and oxidation will require higher temperatures [76]. Furthermore, excess water in the exhaust at high temperature (>600 o C) is said to poison the Pd component, which is tentatively attributed to the slow conversion of the active PdO into a less active Pd(OH) 2 phase [90,91]. Besides the lean and stoichiometric emission control strategies, some researchers also reported high conversions of both regulated exhaust pollutants (NMHCs, CO and NO x ) and unregulated exhaust CH 4 by operating the natural gas engines slightly fuel rich of stoichiometry, but to date this technology has not proven to be effective [92]. Besides, fuel cut-off technology as mentioned before may be also applied to NGV applications for enhanced fuel economy, which results in deactivation of Rh and Pd in TWC. A common practice to improve vehicle fuel economy is to employ a fuel cut-off strategy upon deceleration of the vehicle. The energy saved by periodic operation of fuel cutoff is around 4%- 5% of total energy consumption. Although the events are typically short (2-10 s), they result in exposure of the TWC catalyst to dry air at temperatures up to 1000~1050 o C, causing severe deactivation mainly by oxidation of the Rh component [93-97]. The catalytic NO x reduction activity was then severely affected by Rh deactivation. Catalyst deactivation was found to increase with increasing oxygen concentration and temperature [98]. It is widely known that the thermal 9

30 durability of Pd is superior to that of Pt or Rh under the oxidizing atmosphere, in which some measurable Pd re-dispersion may take place [99,100]. Indeed, the recent trend of TWC formulations has been shifting from the conventional single layered Pt/Rh system to a doublelayered Pd/Rh system, due to the lower cost of Pd coupled with its enhanced thermal durability [101]. During fuel cut-off aging, moderate deactivation was observed with Pd-TWC mainly due to the metal sintering, while significant deactivation was found with Rh-TWC mainly due to the reversible metal oxidation and metal-support interactions [ ]. An effective catalyst regeneration method should be developed to maintain the TWC performance for NGV emission control. Supported Rh and Pd are both good reforming catalysts [104,105]. Our previous study [102,103] reported that after simulated fuel cut-off aging at 1050 o C for 5 min, partial regeneration of gasoline automotive TWCs was accomplished by in situ catalytic reforming of exhaust hydrocarbons, with propane as a model compound, using the H 2 product to reduce Rh 3+ to its NO x reduction active metallic state (Rh 0 ), thereby releasing it from the metal-support interactions. This strategy has been in practice for many years. Rapid regeneration was found for Rh-TWCs within the initial 5 min of steam reforming time on stream of each cycle. Compared to Rh/Al2O3, Rh/CZO showed more rapid response to regeneration and maintained higher stability. Reduction was not effective in reversing the deactivation of aged Pd-TWCs because of irreversible metal sintering. When operated slightly fuel rich, NGV exhaust contains certain amount of CH 4, excess steam and CO 2, and deficient O 2, allowing for H 2 production through reforming to take place. For catalyst regeneration it is unknown whether reforming of exhaust methane at fuel rich conditions can generate enough hydrogen to reduce Rh 3+ in the presence of Pd-TWC. Meanwhile, the 10

31 injection of an additional liquid fuel, such as ethanol, may have the potential to allow the reduction of Rh 3+ to Rh 0 by producing a sufficient amount of hydrogen Carbon dioxide emissions control in post-combustion flue gas Carbon dioxide emissions control by sorption/desorption based technologies Increasing greenhouse gas emissions (GHG) are causing concerns about global warming, ocean acidification, and other environmental problems [106,107]. Anthropogenic CO 2 emission from fossil fuel combustion and industrial processes contributed about 78% to the total GHG emission increase between 1970 and 2010 [108]. Scientists have been persistently seeking effective approaches to the capture of CO 2 from post-combustion effluents, such as flue gas. The current industrial scale separation technology is based on the CO 2 scrubbing (absorption) using alkanolamine, in a thermal swing reactor [109]. The liquid amine-based technology has several shortcomings including large equipment size, severe instrument corrosion, high regeneration energy, and solvent degradation [110,111]. Major advances have been achieved toward the development of various adsorbent materials for CO 2 capture through physisorption at low temperature (25 ~ 150 o C) and high pressure. Adsorbents studied include zeolites [112,113], carbon-based materials [114,115] and metal organic frameworks (MOFs) structures [116,117]. However, these materials often adsorb steam preferentially over CO 2, while possessing insufficient CO 2 adsorption capacity at atmospheric pressure and slightly elevated temperatures. Therefore, they have questionable value for a demanding adsorption condition such as in a post-combustion flue gas environment where high concentrations of steam and air are present [109]. Alternatively, CO 2 capture using dry alkali metal-based materials are suggested for post-combustion gas treatment [ ]. Typical 11

32 processes include chemical looping, a cyclic process of reversible carbonation and decarbonation of calcium oxide with CO 2, as described by Eq. (6) [124,125]. This cyclic process is operated at temperatures between 500 and 900 o C, with issues such as slow carbonation/decarbonation, and sorbent deterioration with an increasing number of cycles [ ]. CaO s + CO & g CaCO ) s ΔH 4 = 178 kj/mol (6) Carbon dioxide emissions control using dual functional catalytic materials The applicability of conventional sorption/desorption- based carbon capture and sequestration (CCS) technology is challenged by significant technical, energy-, and cost-related issues. A process design with both sorption and chemical conversion of CO 2 on the same material could potentially be a more feasible solution. CO 2 capture based on heterogeneous adsorption on solids can be an attractive approach with lower cost and energy consumption. A solid material which captures CO 2, from a simulated flue gas, using a dual functional material (DFM) containing an adsorbent and a catalyst has been developed in this lab. The CO 2 capture and conversion to CH 4 occur in the same reactor, at flue gas temperatures. Excess stored renewable H 2 is used to methanate the captured CO 2 which can be recycled to the inlet of the power plant or injected into the existing pipeline [129,130]. The DFM is composed of nano-dispersed calcium oxide (CaO) and ruthenium (Ru) metal dispersed on a high surface area carrier (Al 2 O 3 ). A two step-process involving adsorption of CO 2 in a steam/ air containing simulated flue gas followed by methanation with H 2 both at 320 o C and 1 bar is further examined in this paper. Methanation occurs, via a process whereby the adsorbed CO 2 spills over to the reduced Ru sites where renewable stored H 2 catalytically converts it to CH 4. The approach has the following advantages: (1) The reactor containing DFM can be positioned in 12

33 the flue gas utilizing its sensible heat; (2) Methanation occurs off-line at the same temperature generating synthetic natural gas (i.e. CH 4 ) at 100% selectivity; (3) The process utilizes renewable H 2 generated via electrolysis from electricity generated from solar and/or wind energy out of phase with its direct use; (4) Significant CO 2 capture and CH 4 production capacities with stable adsorption/methanation performance has been demonstrated in this paper. Both CO 2 adsorption and methanation are exothermic processes, according to the equations below. Eq. (7) is given according to a previously proposed model for CO 2 chemisorption on nanodispersed CaO in the temperature range 300 o C~ 350 o C. It s nano-dispersed state (~ 3 nm) and promotes chemisorption rather than carbonate formation [131]. When supported on γ-al 2 O 3, the dispersed CaO adsorbs CO 2 exclusively as a reversibly bound structure, which differs from the monodentate carbonates [ ] formed on bulk CaO at the same temperature condition (300 o C). These observations explain the rapid adsorption and regenerability of dispersed CaO/Al 2 O 3 as an effective CO 2 adsorbent in DFM. CaO/Al & O ) + CO & CO & CaO/Al & O ) (7) The hydrogenation (methanation) of CO 2, also known as the Sabatier reaction (Eq. 8), is an exothermic process that can be used to selectively produce methane catalytically, preferably from H 2 produced from renewable resources [136]. CO & + 4H & CH S + 2H & O ΔH 4 = 165 kj/mol (8) Supported Ni and PGM metals (e.g. Ru, Rh, Pd) on various supports (TiO 2, SiO 2, Al 2 O 3, CeO 2, ZrO 2 ) have been studied for this reaction. Ni-based catalysts are often preferred but deactivate at low temperatures due to the interaction of the metal particles with CO and the 13

34 formation of mobile nickel subcarbonyls [137,138]. Furthermore, in a flue gas atmosphere the Ni is oxidized and can not be adequately reduced at temperatures where methanation is thermodynamically favorable (<350 o C). Supported Ru is known to be an active and stable methanation catalyst [ ] and can reduced rapidly. Like CaO, it has been reported that the CO 2 reduction on supported Ru/Al 2 O 3 catalysts is cluster size-dependent, with higher Ru loadings (>1 wt%) and moderate dispersions favoring the selectivity toward CH 4 formation [144]. The methanation of CO 2 on Ru is consistent with an Eley-Rideal mechanism [130]. Ru metal also exhibits CO 2 adsorption in an activated process [130,145]. It has been suggested that CO 2 hydrogenation is initiated by a dissociative adsorption of CO 2 (Eq. 9) to CO and O, followed by dissociation of the latter species to C and O (Eq. 10) and successive conversion of C to CH 4 through the final reaction (Eq. 11) [146]. Compared to CaO, Ru has CO 2 adsorption capacity but only in its metallic state. The spill-over step of the pre-adsorbed CO 2 from CaO to the adjacent metallic Ru sites (generated upon the addition of hydrogen) is critical for the enhanced methanation yield in the DFM application. CO & ads. CO ads. + O(ads. ) (9) CO ads. C ads. + O(ads. ) (10) C ads. + 4H(ads. ) CH S (11) With proliferation of the exploration of natural gas resources and the development of hydraulic fracturing technology, there has been a substantial increase in the number of operational natural gas-fired power plants in U.S. over the past few years [147]. The clean nature of the natural gas reduces the concentration of SO 2 and eliminates the particulate matter characteristic of coal- 14

35 based effluent systems. The exhaust gas from the turbine will cool from about 650 o C, to atmospheric conditions [148]. Thus the DFM reactor can be positioned at any desired temperature in the exhaust. The effects of the main flue gas parameters including temperature, flow rate, O 2 and steam concentrations, gas composition on the DFM performance have been studied in this thesis Objective of the thesis: Catalytic emission control in automotive, natural gas vehicles, and flue gas applications The present study covers three research topics as mentioned above, including (1) catalytic automotive emission control, (2) catalytic methane abatement on natural gas vehicles, and (3) CO 2 adsorption and utilization using dual functional materials from stationary power sources. The research highlights for each topic are described as below. The present work for Topic (1) shows new data of the effects of fuel cutoff and subsequent regeneration on the catalytic performance and properties of Rh and Pd model TWCs on γal 2 O 3 or Ce x O y -ZrO 2 support. Fuel cutoff was simulated by aging fresh catalysts in flowing air at high temprature (800 C, 950 C, or 1050 C) for a short period (5 min), while catalyst regeneration was performed by exposing the aged catalysts to a reducing atmosphere (500 vppm propane, 10% steam, 8% CO 2, and N 2 balance at 550 C for 1 h), simulating a slightly rich exhaust composition which is close to normal engine operation. Catalyst regenerability was examined by comparing the activity of fresh, aged, and regenerated catalysts via H 2 generation. By combining various characterization techniques including BET, CO chemisorption, TEM, TPR, XRD, and XPS, the roles of catalyst properties were examined. The study (i) provides a mechanism study of catalyst deactivation during simulated fuel cutoff process; and (ii) explores the aging and support effects 15

36 on catalyst regenerability during simulated fuel rich operation. The thesis highlights the maintenance of catalyst performance through cyclic fuel cutoff-fuel rich operation. For the study on Topic (2), a novel solution to catalytic methane emissions abatement on NGVs with our patented process [pending reference number] has been proposed. Four model TWC catalysts including Rh and Pd supported on Al 2 O 3 or Ce x O y -ZrO 2 (CZO) were used for the study. The process highlights the following aspects: (1) Enhanced conversions of exhaust methane by catalytic reforming of a CH 4 -containing fuel rich exhaust, and (2) regeneration of deactivated TWC catalysts (after simulated fuel cut-off), by using in situ reforming of injected ethanol (or other hydrocarbons such as propane). LNG will likely contain only methane, since the presence of higher hydrocarbons results in a loss of compression ratio, and thereby are separated from fracked natural gas for higher value chemical processes. Consequently, the exhaust hydrocarbon of NGVs should be primarily CH 4. A gas composition of 500 ppm CH 4, 10% steam, 8% CO 2, and N 2 bal. is used in this study to simulate the engine exhaust. Topic (3) is focused on a laboratory scale-up study on the flue gas CO 2 adsorption and methanation with dual functional materials. The study includes the following aspects: (i) preparation of the 5%Ru,10%CaO DFM on several γ-al 2 O 3 carriers for effective industrial mass production and minimized reactor pressure drop; (ii) parametric and lab aging studies including DFM carrier, reaction temperature, flow rate, and feed composition for process optimization; (iii) a preliminary study with an alternative adsorbent material (Na 2 CO 3 ) on γ-al 2 O 3 carrier also is incorporated; (iv) a cyclic testing protocol utilizing a simulated gas mixture of 7.5% CO 2, 15% steam, 4.5% O 2, and N 2 in balance (all in vol-%) at flue gas temperature (320 o C) and pressure (1 bar). Methanation was conducted with 5% H 2 in N 2 to insure safe operation in the lab. The study provides direction for a process design. 16

37 2. Experimental methodologies 2.1. Aging and regeneration study of automotive three-way catalysts Catalyst materials Figure 4. Scanning Electron Microscopic (SEM) images of fresh (a) 0.5% Rh/Al2O3 and (b) 0.5% Rh/CZO, (c) 3% Pd/Al2O3, (d) 3% Pd/Al2O3, (e) 1% Pd/CZO, and (f) 0.5% Pd/CZO model TWC catalysts at µm scale. SEM measurement condition: beam voltage of 20 kv, beam current of 10 µm, working distance of 12 mm, and 30 µm in scale. The model catalysts studied included 0.5% Rh/Al2O3, 0.5% Rh/CexOy-ZrO2, 3% Pd/Al2O3, and 3% Pd/CexOy-ZrO2 (denoted as CZO with Ce:Zr atomic ratio of 1:2). The catalysts and reference support materials were supplied by BASF Iselin, NJ, USA. After impregnation of the precursor salts (proprietary) onto the support (γ-al2o3 or CZO), a 25% solid slurry was created, ball milled, and calcined at 550 C in N2 to generate a catalyst powder sample with average particle size less than 30 µm, as estimated by SEM, Figure 4). The samples were stored in ambient air. XRD of 0.5% Rh/γ-Al2O3 showed a well-defined γ-al2o3 support structure, with all the peaks indexed to a cubic unit cell (a = b = c = Å, space group symmetry Fd3 m (227)) [149]. For 17

38 0.5% Rh/CZO, most of the support zirconia was observed incorporated into the ceria fluorite structure, with the formation of cubic symmetric Ce x Zr (1 x) O 2 solid solution [68,150] Simulated fuel cutoff aging and fuel rich regeneration processes Figure 5. Schematic of the packed bed flow reactor and analysis system. (MFC: Mass Flow Controller, GC: Gas Chromatography, TI: Temperature Indicator). Immediately prior to fuel cutoff aging, the TWC catalyst bed temperature is around 1000 C generated under high load conditions. Upon fuel cutoff the introduction of air from the cylinder causes a short time increase in catalyst temperature by 15 C to 20 C due to the exothermic oxidation of adsorbed hydrocarbons on the catalyst surface. The catalyst bed temperature then begins to fall to about 800 C in less than s. During this high temperature-oxidizing environment the Rh reacts with the Al 2 O 3 causing deactivation of the NO x activity. Some of the US automobile companies use 1050 C aging in air as a simulation of what is experimentally observed to insure stable catalyst performance for 150,000 miles of driving with periodic fuel shut off. We have adapted this procedure in our paper consistent with current practice. The fuel shutoff process was simulated by aging the fresh catalysts in air at 800 C, 950 C, or 1050 C for a 5 min period. The aged catalysts were cooled to room temperature in ambient air. 18

39 The schematic reactor system is sketched in Figure 5. During aging, ml (around 0.05 g) of powdered catalyst, well mixed with 0.25 ml quartz sand as a diluent, was loaded into a quartz tube reactor (ID of 10.5 mm, OD of 12.7 mm) with a quartz frit fused in the middle to hold the sample in place. The reactor was housed in an infrared furnace. Air flowed into the reactor system at 3400 ml/h through a calibrated gas flow controller (MKS 647 C, MKS Instrument Inc., Andover, MA, USA) with multiple gas channels. Reaction temperature of the catalyst bed was monitored by a thermocouple (Omega K type) placed at the inlet to the catalyst. Catalyst regeneration was performed in situ by exposing the aged catalyst to reducing conditions. During regeneration the Rh catalyzes steam reforming (SR), generating H 2, which reduces Rh 3+ to its active metallic state Rh 0. The regeneration feed gas mixture of 500 vppm propane, 10 vol-% steam, 8 vol-% CO 2, and N 2 in balance, with a total GHSV of 120,000 h 1 was used to simulate the engine exhaust at slightly fuel rich conditions. Propane is commonly used as a model compound [30,31,151,152] for the HC species. Liquid water was injected at 0.68 ml/h by a syringe pump (Cole Parmer), vaporized at around 120 C, and mixed with the incoming gas feed. The regeneration temperature was maintained 550 C for 1 h with H 2 production continuously monitored. Temperatures of the water evaporator T W and catalyst bed T bed were monitored by thermocouples. A cold trap was placed downstream to condense the unreacted water, and a calibrated micro GC (Inficon 3000, INFICON Inc., New York, NY, USA, equipped with 10 m Molsieve 5A column, 8 m Plot U column, and thermal conductivity detectors) was used for online analysis of the gas products every three minutes. The regenerated sample was then cooled in air to room temperature, and preserved in ambient air. 19

40 Catalyst regenerability as measured at simulated fuel rich condition Activity tests were performed with the same reaction feed as that in regeneration, but with temperature scans from 200 C to 550 C, with 50 C increments, and a 30 min-hold at each temperature. The catalytic conversions were conducted far from equilibrium Catalyst stability during simulated fuel cutoff aging-fuel rich regeneration cycle tests Figure 6. Schematic process flow diagrams of (a) simulated fuel cutoff aging-fuel rich regeneration cycle and activity test; and (b) on-board gasoline engine fuel cutoff-fuel rich operation cycles. Fuel cutoff aging-fuel rich regeneration cycle tests (25 cycles in total) were performed to simulate the automotive engine operation cycles, as shown in Figure 6a, b respectively. Simulated fuel cutoff and regeneration conditions were maintained as described in Section 2.1.2, except the aging temperature of 1050 C was used. Activity of the regenerated catalysts after every five cycles were measured, as described in Section Data analysis 20

41 During catalyst activity tests, the mole flow rate Q i (mol/h) of each gas product component (H 2, N 2, CO, CH 4, CO 2, C 3 H 8, except for water, which was condensed before GC analysis) was determined by Eq. (12). Q 8 = Q [\ F 8 F [\ (12) where Q [\ is the mole flow rate of the carrier gas N 2 (also the internal standard, mol/h). F i is the mole percentage of compound i in the gas product mixture as analyzed by the online micro GC. Thermodynamic modeling at the same reaction condition was performed by HSC Chemistry 5. For isothermal activity profiles, error bars show standard errors of the mean value of up to ten repeated measurements at the same reaction condition Catalyst characterization (1) Scanning electron microscopy (SEM) SEM images were taken with a Hitachi S-4700 I Cold Field Emission Scanning Electron Microscope (Hitachi High Technologies America Inc., Schaumburg, IL, USA. Accelerating voltage of 20 kv, emission current of 10 µa, and working distance of 12 mm were used. Multiple pictures at different spots within the same measured sample were collected in each measurement. (2) Brunauer Emmett Teller (BET) surface area The Brunauer-Emmett-Teller (BET) surface areas of catalysts were determined using a Quantachrome ChemBET Pulsar TPR/TPD unit (Quantachrome Instrument, Boynton Beach, FL, USA), equipped with a TCD detector. About 0.05 g of catalyst sample was outgassed in pure N 2 at 200 C for 2 h, while subsequent N 2 adsorption was performed using 30% N 2 /He at liquid N 2 21

42 temperature ( C). The TCD signal was calibrated using an external standard method, and monolayer N 2 adsorption was evaluated by single point BET method. (3) CO chemisorption The metal dispersions of catalysts were measured by CO- selective chemisorption using the same Quantachrome unit. About 0.1 g of catalyst sample was heated in pure He at 200 C for 120 min, followed by pre-reduction in 10% H 2 /Ar at 400 C for 120 min. CO (99.9% purity) adsorption, with automatic injection volume of 285 µl per pulse, was performed at 40 C after pre-reduction. CO chemisorption capacity of each catalyst sample was evaluated based on the total volume of adsorbed CO at standard condition (V CO,std, L). Metal dispersion (D M ) is calculated according to Eq. (13): Dispersion (%) = 1 n V gh,ijk 22.4 M m m i y m 100% (13) where M M is the metal atomic weight (102.9 g/mol for Rh). The catalyst weight and metal content are designed by m s and y M respectively. The CO-to-metal site stoichiometry n was assumed to be 2 in accordance with current literature [20,153]. (4) Transmission electron microscopy (TEM) The Transmission Electron Microscopic (TEM) observations of the fresh, aged, and regenerated samples were taken with a JEOL 100CX-II TEM unit (JOEL Inc., Peabody, MA, USA). The TEM measurements were operated at an accelerating voltage of 100 kv. The catalyst powder sample was dispersed in pure ethanol (200 proof), followed by sonication for 3 h, and deposition on a Lacey carbon film supported Cu grid (200 mesh). For each sample, 50~60 TEM images with different magnifications at multiple spots were taken, and no less than 400 individual 22

43 palladium particles were counted with ImageJ software. The mean surface area-weighted palladium particle size is calculated using Eq. (14): d qrm = n ) s sd s & s n s d s (14) where n i is the number of particles in ferret diameter d i and s n s > 400. (5) H 2 -Temperature programmed reduction (H 2 -TPR) The redox properties of catalysts were studied by TPR. The measurements were carried out with the same Quantachrome unit as above. About 0.1 g of catalyst sample was first outgassed in pure Helium at 150 C for 2 h, and cooled to room temperature. TPR analysis was performed subsequently by heating a sample located in a U-tube reactor to 800 C at 5 C/min, with 4% H 2 /N 2 flowing through the sample. The TCD signal (corresponding to H 2 uptake) was then normalized to per gram of catalyst. (6) X-ray powder diffraction (XRD) The XRD patterns of the fresh and aged catalysts were generated with a PANalytical X Pert 3 Powder XRD unit. The powdered samples were single scanned between 15 and 90 with an incremental step of and a time per step of 300 s. The mean crystallite sizes d i for the bulk γ-al 2 O 3 and CZO support were determined with a precision of ±10% from the line broadening of the most intense reflections using Scherrer s equation Equation (15) [68,149,150, ]. d i = Kλ β cos θ (15) 23

44 where β is the FWHM (full width at half maximum, in radians) of the selected diffraction peak, θ (in radians) is the Bragg angle, K = 0.93 is the numerical constant, and λ = nm is the wavelength of the X-ray incident beam (Ni filtered Cu Kα). (7) X-ray photoelectron spectroscopy (XPS) Ex situ XPS spectra of catalysts were measured with a Perkin-Elmer PHI 5500 XPS instrument (Physical Electronics Inc., Chanhassen, MN, USA) equipped with a Mg Kα monochromatic source. The samples were prepared by fixing catalyst powder onto a double-sided carbon sticky tape. The XPS main chamber was evacuated to 10 9 Torr. C 1s peak with standard binding energy of ev was used for peak position calibration. AugerScan and Origin software were used for spectra data analysis. NIST XPS online database and other literature sources were used for peak assignments Catalytic methane emissions abatement study Catalyst materials The catalyst materials were supplied by BASF Iselin, NJ, United States, including four Rhor Pd- based model TWC catalysts supported on stabilized gamma alumina (γal 2 O 3 ) or ceriazirconia (Ce x O y -ZrO 2, CZO) carriers, with chemical compositions of 0.5% Rh/Al 2 O 3, 0.5% Rh/CZO, 3% Pd/Al 2 O 3, and 3% Pd/CZO. For details about the catalyst preparation and material properties, please refer to our previously published articles [102,103] In situ catalyst pre-reduction, aging, and regeneration 24

45 Figure 7. Schematic catalytic flow reactor setup for in situ catalyst pre-reduction, aging and regeneration, CH 4 reforming/oxidation activity measurements, and cycle tests at different reaction conditions. Pre-reduction, simulated fuel cut-off aging and cyclic aging-regeneration treatments were performed with fresh catalysts in situ in a fixed bed flow reactor before further examination of their catalytic methane conversion capacity. The schematic reactor setup is sketched in Figure 7. A quartz tube reactor (GM Associates, ID of 10.5 mm, OD of 12.7 mm, Length of 500 mm), with its two ends respectively connected to the pre-heated gas inlet and liquid condenser, was placed vertically and housed in a microthermal furnace (Mellen MTSC12.5R-.75x18-1Z). The reaction temperature was feed-back controlled by a temperature controller (Omega CN7800) connected to a thermal couple (Omega K type) placed slightly above the catalyst bed. A porous quartz frit was fused in the middle of the reactor for supporting the sample. Each loaded sample included a powdered catalyst (packing volume of ml, around 0.05 g) well mixed with calcined fine quartz sand (0.25 ml), used for preventing gas channeling during the tests. The reaction gaseous feed flow rate and composition were controlled by a calibrated mass flow controller (MKS 647 C) equipped with multiple gas channels. A syringe pump (Cole-Parmer C) was used to deliver 25

46 the liquid reactant (pure water or 1.26 vol-% ethanol-water solution) to the pre-heated (~125 o C) reactor system, where the vaporized steam would combine with the incoming pre-heated dry gaseous feed mixture. N 2 was used as the carrier gas, diluent, and internal standard. The reaction product mixture was condensed using a cold trap, allowing the dry product to be analyzed immediately by a calibrated online micro-gc (Inficon 3000) equipped with a 10 m-molsieve 5A column, an 8 m-plot U column, and thermal conductivity detectors. Catalyst aging, simulating the automotive fuel cut-off process, was carried out in air with a gaseous flow rate of 3400 ml/h at temperature of 950 o C or 1050 o C at 1 bar for different durations (5~200 min), followed by cooling to room temperature in the same working gas environment. For some experiments thermal aging was performed in only N 2 at 1050 o C. Catalyst regeneration/or attempted regeneration was performed after fuel cut-off aging by catalytic co-reforming of methane or ethanol or propane with steam and CO 2 to H 2 intended to reduce Rh 3+ to its active metallic state Rh 0. The regeneration conditions included a gas feed composition (in vol-%) of 500 ppm HC (methane, propane, or ethanol vapor), 10% steam, 8% CO 2, and N 2 balance, with a total GHSV of 120, 000 h -1, temperature of 550 o C at 1 bar, and a TOS of 30 min or 60 min. Fresh catalyst pre-reduction was conducted with some tests also at 550 o C using H 2 generated from ethanol reforming Catalytic methane conversion activity tests Catalytic reforming and catalytic oxidation of only methane was considered as a baseline. Activity tests were performed with fresh (w/ or w/o pre-reduction), fuel cut-off aged, and regenerated catalyst samples isothermally in the as-described flow reactor at GHSV of 120, 000 h - 1, with reaction temperatures ranging from 350 o C to 700 o C, every 10 o C. The reaction feed for 26

47 catalytic methane reforming consisted of 500 ppm CH 4, 10% steam, 8% CO 2, and N 2 bal. The reaction feed for catalytic methane oxidation was a mixture of 500 ppm CH 4, 0.96% O 2, 10% steam, 8% CO 2, and N 2 in balance (all in vol-%). Catalytic aging-regeneration-activity cycle tests with different combinations of reaction conditions were carried out, with specific process flow diagrams sketched for the corresponding experimental results in the following section. For isothermal activity profiles, error bars show standard errors of the mean value of up to five repeated measurements at the same reaction condition Reaction thermodynamic modelling Reaction thermodynamic model for the reforming reactions of different hydrocarbons (methane, propane, or ethanol) was generated with HSC Chemistry 5 software Scale-up CO 2 adsorption and methanation study Material preparation The DFM samples, with a target composition of 5%Ru,10%CaO/Al 2 O 3 were prepared by incipient wetness impregnation followed by co-precipitation with an alkaline solution. The schematic preparation process flow is shown in Figure 8. Ruthenium (III) Chloride hydrate (RuCl 3 xh 2 O, 99.9% PGM basis, Ru 38% min, Alfa Aesar) and calcium chloride (CaCl 2, anhydrous, powder, 99.99% trace metals basis, Sigma-Aldrich) were selected as the precursor salts, and dissolved in water with an appropriate amount of HCl added for ph control close to 0±0.2. Powder-, bead-, and cylindrical pellet-forms of γ-al 2 O 3 (Table 1) were tested as DFM carriers. Shaped carriers, with particulate sizes around 5 mm, allow a preferred reactor design with 27

48 minimized pressure drop. The physical properties of the carrier must also insure high effectiveness factors. The co-impregnation of ruthenium and calcium oxide precursor salts were first performed on the degassed γ-al 2 O 3 carrier, followed by drying in air at 120 o C for 3 hrs. An aqueous NaOH ( 97%, Sigma-Aldrich) solution was used to precipitate and fix the Ca and Ru hydroxides on the Al 2 O 3 at 80 o C for 3 hrs. The precipitates were water washed to remove the chloride precursors and filtered. Silver nitrate was added to the filtrate to insure chloride removal as noted by the absence of turbidity. The DFM was dried at 120 o C for 3 hrs in air, followed by calcination in air at 400 o C for 3 hrs. RuO x pre-reduction was carried out with a flow of 5 vol-% H 2 /N 2 at 400 o C for 150 min to form the catalytically active metallic state. Calcination temperatures were always lower than 400 o C to prevent formation of volatile Ru oxides. In some cases, DFM powders were pressed into pellets with a die (Specac Atlas 13 mm Evacuable Pellet Die) under a hydraulic pressure of 5 tons. Figure 8. Schematic process flow for the catalyst preparation of 5%Ru,10%CaO/Al 2 O 3 DFM on different Al 2 O 3 support materials and 5%Ru,10%Na 2 CO 3 /Al 2 O 3 DFM on TH100 Al 2 O 3 support. 28

49 For the preparation of DFM with a composition of 5%Ru,10%Na 2 CO 3 /Al 2 O 3, sequential impregnation on cylindrical SASOL TH100 γ-al 2 O 3 pellets was performed in the following order: (1) Ru impregnation with chloride (same RuCl 3 xh 2 O precursor as mentioned above), (2) drying at 120 o C for 3 hrs, (3) precipitation with NaOH at 80 o C for 3 hrs, (4) washing and filtration with water, (5) drying at 120 o C for 3 hrs followed by calcination at 400 o C for 3 hrs, (6) Na 2 CO 3 (Sigma-Aldrich) impregnation using an aqueous solution, and (7) final drying at 120 o C for 2 hrs followed by calcination in air at 400 o C for 1 hr. Table 1. Material properties of the alumina supports used for catalyst preparation. Support alumina BET surface area Pore volume Packing density Diameter Water uptake m 2 /g ml/g g/ml mm ml H 2 O/g BASF powder 143 N/A 0.93 N/A 0.74 SASOL TH200 pellets SASOL TH100 pellets BASF SAS200 spheres N/A N/A N/A CO 2 adsorption-methanation cycle tests CO 2 adsorption-methanation tests were performed in a fixed bed flow reactor (Figure 9). A standard quartz tube reactor (O.D of 25 mm, (I.D. of 20 mm, length of 570 mm) was connected to the reaction system through Swagelok Ultra-Torr Vacuum fittings housed in a microthermal furnace (Mellen MTSC12.5R-.75x18-1Z) with temperature feed-back control. The Omega thermocouples (K type) were placed close to the inlet of the DFM bed. Ten grams of DFM were secured in the middle of the quartz reactor while the downstream reactor space was filled with degassed glass beads (Fisher Scientific, DI of 6 mm) to reduce the reactor dead volume and improve analytical response times. Inlet gas flow rates and compositions were controlled by 29

50 multiple gas rotameters (Key Instruments) with precise linear calibrations using specific gas carriers. Water was injected using a syringe pump (Cole-Palmer) into the heated system with a temperature maintained at 125 o C to generate steam. This was mixed with the other pre-heated feed gas components. During the test, the unreacted water (and water produced from methanation) were condensed while the dry product gas mixture flowed into an Enerac Model 700 analyzer equipped with electrochemical and non-dispersive-infrared sensors for frequent (1 Hz) on-line remote analysis of the product gas composition. Among the dry gas products, the volume percentages of CH 4, CO, O 2, and CO 2 were detected, while the material balances were ensured by confirming the flow rate detected by a mass flow meter placed upstream of the gas analyzer. All the tests were performed at ambient pressure with no pressure drop observed from the inlet to outlet. Figure 9. Reactor setup for the CO 2 adsorption-methanation cycle tests and process parameter study. The schematic protocol of CO 2 adsorption-methanation cycle tests is shown in Figure 10. Two steps were involved in each cycle, i.e. CO 2 adsorption and CO 2 methanation (with added H 2 ). For the CO 2 adsorption step, a gas mixture (Vol %) composed of 7.5% CO 2, 15% steam, 4.5% O 2, 30

51 and N 2 in balance, at a total gas hourly space velocity (GHSV) of 11, 236 h -1, simulating the flue gas content of a natural gas fired power plant, was fed to the reactor. CO 2 adsorption was performed at 320 o C and 1 atm. for varying times as indicated. For the methanation step, the reaction gas feed was switched to 5% H 2 and N 2 balance, at a varying total GHSV. In between the adsorption and the methanation step the reactor was purged with pure N 2 for 3 min, to prevent mixing of O 2 and H 2. For ultimate commercial use it is expected that pure H 2 will be used and generated by water splitting driven by an alternative energy source, such as excess solar energy. The CO 2 methanation was carried out at varying temperatures but always at 1 atm. for around 60 min to complete conversion of adsorbed CO 2 to CH 4. The amount of CH 4 produced was determined from the integrated CH 4 volume percentage present in the dry product. The CO 2 captured was calculated by subtracting the system background (product CO 2 flow rate achieved by flowing the adsorption feed to an empty reactor) with the adsorption test result. Figure 10. Schematic reaction process flow for CO 2 adsorption-methanation cycle tests Process parametric study for the adsorption and methanation 31

52 The effects of process parameters on CO 2 adsorption and methanation were studied with the DFM supported on Al 2 O 3 pellets (SASOL TH200). The process parameters investigated included adsorption time on stream (TOS), reaction temperature, adsorption feed flow rate, methanation feed flow rate, and the presence and absence of steam and/or O 2 in the adsorption feed stream. The standard reaction condition was the same as described in Section 2.2. With each process parameter varied, the other reaction conditions were maintained the same. The parametric variables are shown in the accompanying figures. 3. Results and discussion 3.1. Aging and regeneration mechanisms of Rh-based gasoline three-way catalysts (TWC) Reaction thermodynamics at simulated engine fuel rich condition Table 2. Main reaction pathways during catalytic conversions of simulated exhaust feed at fuel rich condition. Reaction # Reaction Pathways Reaction Type (Forward Direction) ΔH r 0 (298 K) (kj/mol) a C ) H 3 + 3H & O 3CO + 7H & Propane steam reforming b C ) H 3 + 6H & O 3CO & + 10H & Propane steam reforming c CO + H & O CO & + H & Water gas shift d CO & + 4H & CH S + 2H & O Methanation of CO e CH S + H & O CO + 3H & Methane steam reforming The main reaction pathways occurring during fuel rich operation are listed in Table 2. Specifically, endothermic steam reforming of propane reactions (Eq. a and b) are thermodynamically and kinetically favorable at high temperature. Exothermic water gas shift (Eq. c) and methanation of CO 2 (Eq. d) reactions are thermodynamically favorable at relative low temperature, where reaction kinetics are slow. The CH 4 produced undergoes steam reforming (Eq. 32

53 e). The reaction thermodynamics and kinetics are also largely dependent on the feed. When 8 vol- % CO 2 is considered (as present in the exhaust), it can react with CH 4 in what is referred to dry reforming. Figure 11. Reaction Gibbs free energy as a function of reaction temperature (25 C to 700 C) at 1 atm. Assume ideal gas behavior for the reactant and product gas components. Compound thermodynamic data with temperature and pressure inputs is collected from I. Barin, Thermochemical Data of Pure Substances (3rd Edition) [157]. The Gibbs free energy for main reactions (Reactions a e) are plotted in Figure 11. The extent of methanation of CO 2 (Reaction d) and reverse water gas shift (reverse of Reaction c) increase in the presence of 8 vol-% CO 2 in the feed. Both of these reactions decrease H 2 content. It is clear SR is favored above about 450 C. Figure 12 shows the main equilibrium products (H 2, CO, and CH 4 ) and reactant (propane and water) equilibrium mole fractions at fuel rich condition (500 vppm propane, 10 vol-% steam, 8 vol-% CO 2, N 2 in balance). In the low temperature regime (T < 350 C), steam reforming, water gas shift, and methanation reactions dominate. In the high temperature regime (T > 350 C), WGS and methanation reactions become less favorable. 33

54 Figure 12. (a) Main mole fractions of H 2, CO, and CH 4 ; and (b) theoretical reactant (propane and water) conversions as a function of reaction temperature (200 C to 550 C) at thermodynamic equilibrium conditions. Reactant feed: 500 vppm propane, 10 vol-% steam, 8 vol-% CO 2, N 2 in balance Catalyst deactivation and regeneration of Rh-TWCs Figure 13. Catalyst activity of fresh and aged (a) 0.5% Rh/Al 2 O 3 and (b) 0.5% Rh/CZO Catalyst activity is plotted in terms of H 2 mole fraction as a function of reaction temperature (200 C 550 C). After aging in air at three temperatures (800 C, 950 C, or 1050 C), the activity of fresh and aged Rh/Al 2 O 3 and Rh/CZO are compared as the H 2 mole fraction in the propane rich feed gas (Figure 13). Fresh Rh/CZO (b) shows higher activity than Rh/Al 2 O 3 (a) under fresh and all aged conditions. 34

55 Not surprisingly, deactivation of both catalysts increases with aging temperature from 800 C to 1050 C. The most difficult regeneration was expected after aging at elevated temperatures. This is in agreement with a previous report [98]. Compared to Rh/Al 2 O 3, fresh and aged Rh/CZO showed higher catalytic activity in converting propane to H 2. Figure 14. Activity of fresh, aged, and regenerated (a) 0.5% Rh/Al 2 O 3 and (b) 0.5% Rh/CZO in converting simulated engine exhaust gas at fuel rich condition. Catalyst activity is plotted in terms of H 2 product mole fraction as a function of reaction temperature (200 C to 550 C). Aged catalysts were obtained by treating fresh ones in air at 1050 C for 5 min, followed by cooling to room temperature in air. Catalyst regenerations were performed by at rich condition at 550 C for 1 h. The activity of fresh, aged, and regenerated catalysts are compared in Figure 14. The regeneration method was very effective in recovering full activity of Rh catalysts. This is also shown in Table 3, which compares the T 50 of regenerated catalysts with those of fresh and aged. T 50 is the temperature at which 50% maximum equilibrium H 2 production is achieved. The temperatures of Rh/CZO were always 40 C lower than Rh/Al 2 O 3, indicating a more rapid response to in situ regeneration. 35

56 Table 3. T 50 (temperature at which 50% maximum equilibrium H 2 production is reached during activity tests, C) of fresh, aged, and regenerated catalysts. T 50 ( C) Catalyst After Aging in air for 5 min at Different temp. After Regeneration * Fresh 800 C 950 C 1050 C 0.5% Rh/Al 2 O % Rh/CZO * Catalyst regenerations were performed at 550 C in propane-containing feed gas Rh-TWCs stability during fuel cutoff aging-fuel rich regeneration cycle tests Figure 15 shows the rapid response (increased slope of H 2 production) achieved during regeneration for both Rh/Al 2 O 3 and Rh/CZO catalysts. Figure 15. H 2 generations during regeneration processes in simulated fuel cutoff aging- fuel rich regeneration cycle tests (First 5 cycles out of total 25 cycles) with (a) 0.5% Rh/Al 2 O 3 and (b) 0.5% Rh/CZO. In each cycle, the catalyst sample was first aged in air at 1050 C for 5 min, followed by in situ regeneration at propane rich condition at 550 C for 1 h. 36

57 The activity of both catalysts after every 5 cycles of aging-regeneration cycles are plotted in Figure 16. Significant losses of catalyst activity were observed in the first 5 cycles. The initial deactivations (greater for the Rh/Al 2 O 3 ) are believed caused by Rh metal sintering and dissolution of oxidized Rh into the sintered support materials (metal-support interactions). After the first 5 cycles, the performance stabilized. At this condition the major deactivation modes have been completed and no further permanent deactivation is noted after repeated cycles. The aged catalysts could then be regenerated but to a lesser extent than after 5 cycles. Compared to Rh/Al 2 O 3, Rh/CZO showed higher stability after periodic regeneration. Figure 16. Activity of regenerated (a) 0.5% Rh/Al 2 O 3 and (b) 0.5% Rh/CZO catalysts after 1, 5, 10, 15, 20, and 25 cycles in the aging-regeneration cycle tests. After every five cycles of aging/regeneration. The catalyst activity is plotted in terms of H 2 product mole fraction as a function of SR temperature from 250 C to 550 C Catalyst deactivation and regeneration mechanisms for Rh-TWCs 37

58 Figure 17. BET surface areas (based on single-point measurement, 10% error allowed) of fresh and aged (a) 0.5% Rh/Al 2 O 3 and (b) 0.5% Rh/CZO, as a function of aging conditions. Aged samples were obtained by aging fresh catalysts (I) in air at the following conditions: (II) 800 C for 5 min; (III) 800 C for 1 h; (IV) 950 C for 5 min; or (V) 1050 C for 5 min. The aging processes were followed by cooling in air to room temperature. As a reference, BET surface areas of support materials were measured m 2 /g and 60.3 m 2 /g respectively for fresh Al 2 O 3 and CZO. Figure 17 shows the BET surface areas of fresh and aged 0.5% Rh/Al 2 O 3 and 0.5% Rh/CZO at different aging treatments. In agreement with the catalyst activity result as shown in Figure 13, most significant losses in catalyst surface areas occurred at 950 C and 1050 C. The Al 2 O 3 support exhibited higher intrinsic surface area, but a slightly higher percentage of sintering relative to the CZO support (with 41.4% and 35.9% for Rh/Al 2 O 3 and Rh/CZO respectively after 1050 C aging). ZrO 2 in CZO support is likely the main contributor to the thermostability of Rh/CZO [158]. The metal dispersions of fresh and aged catalysts are shown in Table 4. Fresh Rh catalysts showed higher metal dispersions on CZO than on Al 2 O 3. Soria and Duarte et al. [159,160] reported that the enhancement of metal dispersion in Rh/CeO 2 system was achieved by the Rh-Ce interaction (ceria stabilized Rh + species formed on the support). The higher metal dispersions 38

59 likely enhance its activity in catalytic steam reforming of propane. After aging, loss of active metal sites occurs for both Rh catalysts, and is accelerated with increasing aging temperature from 800 C to 1050 C. Barbier et al. [161], reported that the decrease of Rh surface area in Rh/Al 2 O 3 system was mainly linked to the diffusion of Rh 3+ into the alumina matrix, while the presence of Ce x O y stabilizes Rh and prevents Rh 3+ from dissolving into the support. Table 4. Metal dispersions (%) of fresh and aged 0.5% Rh/Al 2 O 3 and 0.5% Rh/CZO catalysts as measured by room temperature CO chemisorption a. After simulated fuel cutoff, loss of active metal sites occurred for both Rh catalysts, and was accelerated with aging temperature from 800 C to 1050 C. Metal Dispersion (%) Catalyst Fresh 800 C 1050 C 0.5%Rh/Al 2 O %Rh/CZO a CO can only be chemisorbed on Rh 0 in Rh/Al 2 O 3 and Rh/CZO. CO chemisorption was negligible on non-reduced catalyst samples, and was zero on Al 2 O 3 and CZO support-only material. However, there likely exists an overestimation of the metal particle dispersion by measuring the CO chemisorption of Rh/CZO, due to the formation of carbonate species on CeO 2 surface even at low temperature (323 K) and by the likelihood of multiple CO molecules adsorbing on the Rh itself. Some preliminary TEM result as below (Table 5). After aging and regeneration, negligible metal crystallite size grow was observed with both Rh/Al 2 O 3 and Rh/CZO, which supports our result that metal sintering was not the major deactivation mode in Rh-TWC. The dramatically reduced CO chemisorption capacity of the aged Rh-TWC together with the TEM images suggests that the main deactivation mode during 1050 C aging was metal-support interaction. 39

60 Table 5. Mean metal particle sizes (nm) of fresh and aged TWCs as measured using TEM images. Active Particle Mean Size (nm) * Catalyst Fresh After Air 1050 C After Regeneration 0.5% Rh/Al 2 O % Rh/CZO * Feret diameter of active metal/metal oxide particle. Standard deviations for the results are measured within 1.32~2.15. During air aging, the oxidation state of Rh increased in both Rh/Al 2 O 3 and Rh/CZO, i.e., Rh 0 Rh 3+ (Eq. 16). Meanwhile, strong metal-support interactions with the formation of Rhodium Aluminate (Rh(AlO 2 ) y ) took place in Rh/Al 2 O 3 sample (Eq. 17) [158]. 2Rh O & ~344 Rh& O ) (16) Rh & O ) + yal & O ) + y 3 2 O 344,789 & 2 Rh AlO& : (17) H 2 -TPR was used to study the catalyst redox property after aging at different conditions (Figure 18). The lower the temperature of the H 2 consumption peak, the easier the reduction. The reductions of Rh 3+ to Rh 0 in both catalysts occurred around 100 C. H 2 reaction pathways on Rh sites include H 2 spill over and dissociation on Rh 0 sites (Eq. 18), and subsequent reduction of Rh 3+ Rh 0 (Eq. 19) [ ]. Rh 4 + x 2 H & Rh H (18) Rh & O ) + 6H 2Rh 4 + 3H & O (19) 40

61 Figure 18. Normalized H 2 consumption in H 2 -Temperature Programmed Reduction (H 2 -TPR) measurements of fresh and aged (a) 0.5% Rh/Al 2 O 3 and (b) 0.5% Rh/CZO, as a function of reducing temperature. Aged samples for measurements were respectively achieved by aging the fresh ones in air at the following conditions: 800 C for 5 min, 800 C for 1 h, 950 C for 5 min, or 1050 C for 5 min. Since Al 2 O 3 is non-reducible, H 2 consumption peaks in fresh and aged Rh/Al 2 O 3 are only assigned to Rh reductions. The results indicate that Rh in fresh Rh/Al 2 O 3 was already partially oxidized with a H 2 consumption peak at 90 C. After aging in air with ( C), the TPR profiles of Rh/Al 2 O 3 samples shifted to higher reducing temperatures, while the H 2 consumption peak area continued to increase. Rh was released from metal-support interaction by H 2 (Eq. 20), but this became increasingly difficult at higher aging temperature. 2Rh AlO & : + yh & >?@A 98BC 2Rh + yal& O ) + yh & O (20) The precise stoichiometry of Rh:O is dependent on many factors including metal loading, metal dispersion, aging temperature, and aging oxygen partial pressure. Hwang et al. [158], reported the phase diagram for the variation of rhodium oxide species on the dispersion of rhodium samples and the oxidation temperature. 41

62 Table 6. Reduction temperature (T R ) and H 2 consumption during H 2 -TPR for fresh and aged 0.5% Rh/Al 2 O 3 samples. Sample T R ( C) H 2 Consumption (µmol H 2 /gcat) N (H 2 )/N (Rh) Fresh C for 5 min C for 1 h C for 5 min C for 5 min The quantitative H 2 consumption for the fresh and aged 0.5% Rh/Al 2 O 3 samples during H 2 - TPR is shown in Table 6. The H 2 consumptions correspond well to our statement that the fresh 0.5% Rh/Al 2 O 3 sample was already partially oxidized. N(H 2 )/N(Rh), i.e., the ratio between consumed H 2 molecules and reduced Rh atoms, increases with increasing aging temperature, and approaches 1.5 after high temperature (950 C) aging, suggesting almost complete oxidation of Rh 0 to Rh 3+ in severely aged 0.5% Rh/Al 2 O 3. The H 2 consumption by rhodium oxides in 0.5% Rh/CZO is very difficult to quantify because the reduction of Ce 4+ to Ce 3+ also consumes H 2. For 0.5% Rh/CZO samples, qualitative TPR analysis and semi-quantitative XPS analysis (later text) are sufficient for the Rh oxidation state study. For fresh CZO support, one broad reduction peak is shown at C, which is assigned to bulk surface Ce 4+ to Ce 3+, with the following global reaction (Eq. 21) [ ]. 2CeO & + H & Ce & O ) + H & O (21) 42

63 Figure 19. Speculative schematic of proposed redox reaction mechanism and interaction between Rh δ+ /Rh 0 and Ce 4+ /Ce 3+ redox couple during H 2 -TPR of 0.5% Rh/CZO. The redox reactions followed the order described below. (I) After simulated fuel cutoff aging in air at 800 C, 950 C, or 1050 C, surface Rh sites are oxidized to Rh 2 O 3, while the Rh sites in close contacts with Ce x O y remained in reduced states (Rh δ+, 0 < δ < 1), with Rh δ+ /Rh 0 and Ce 4+ /Ce 3+ redox couple formed for enhancing electron transfer efficiency; (II) H 2 flow through the sample; (III) At low temperature regime around 100 C to 120 C, H 2 was chemisorbed and dissociated on the Rh 0 sites, followed by (IV) Reduction of Rh 3+ to Rh 0 ; (V) Reduction of surface Ce 4+ sites to Ce 3+ promoted by the Rh δ+ /Rh 0 and Ce 4+ /Ce 3+ redox couple; (VI) Reduction of bulk Ce 4+ sites to Ce 3+ when more H 2 molecules were chemisorbed and dissociated on Rh 0. 43

64 The reduction of Rh 3+ to Rh 0 on CZO (71 C) is easier than for fresh Rh/Al 2 O 3 (91 C). The presence of Rh in fresh Rh/CZO allows the reduction of Ce 4+ to Ce 3+ to occur at a lower temperature ( C), with the Ce 4+ reduction peak split into two side peaks (at 110 C and 308 C respectively). The lower reduction temperature of Ce 4+ (at 110 C) occurs following the surface reduction of Rh 3+, suggesting that the Ce 4+ sites being reduced were most likely the ones in close contact with the Rh sites. It has been reported that the redox properties of both Rh and Ce are enhanced when Rh is deposited on Ce x O y [168]. The Rh O Ce bond is likely formed, creating Rh δ+ /Rh 0 (0 < δ < 1) and Ce 4+ /Ce 3+ redox couple. Electrons transfer more efficiently during H 2 reduction [ ]. The introduction of Zr into the Ce x O y crystal lattice, now widely practiced for OSC, stabilizes the Rh Ce interaction via improving mobility of oxygen in Ce x O y or maintaining Ce x O y dispersion in nanometer scale [ ]. Furthermore, the Rh O Ce bond can be very easily dissociated [66], which makes the interaction between Rh and Ce x O y much weaker than that between Rh and Al 2 O 3. After complete reduction of Rh δ+ to Rh 0, electrons transfer from dissociated H 2 on Rh 0, allowing easier reduction of the bulk Ce 4+ to Ce 3+. The schematic mechanism of the redox reaction pathways and the promotional metal-support interaction within 0.5% Rh/CZO during H 2 -TPR are sketched in Figure 19. In agreement with previous literature [180], increasing the aging temperature, the reduction of both Rh and Ce in Rh/CZO shifted to higher temperature values, suggesting decreases in hydrogen dissociation capability after aging. The XPS Rh 3d electron orbitals were used to identify and semi-quantify the Rh oxidation states, by comparing the binding energy values and relative ratio of the corresponding states. Figure 20 profiles the XPS Rh 3d spectra of both catalysts, and Table 7 summarizes the peak information details. Rh 3d 3/2 and Rh 3d 5/2 peaks, resulting from spin-orbital splitting, with different 44

65 binding energies (BE) corresponding to the Rh valence states were assigned [ ]. For a fresh sample, Rh 3d 5/2 peak with BE of at ev ev is attributed to Rh 0 valence state, while Rh 3d 5/2 peak at ev ev is attributed to Rh 3+ valence state. Figure 20. X-ray Photoelectron Spectroscopy (XPS) multiplex spectra in Rh 3d region (with BE of 318 ev 304 ev) of fresh, aged, and regenerated (a) 0.5% Rh/Al 2 O 3 and (b) 0.5% Rh/CZO powder catalysts, and with aging temperature varied. Aged samples were achieved by aging the fresh catalysts in air at 800 C, 950 C or 1050 C for 5 min. Regenerated samples were achieved by regenerating the aged ones (1050 C for 5 min) using the method as described in Section 2.3. Fresh 0.5% Rh/Al 2 O 3 displayed (1) an intense Rh 3+ 3d 5/2 peak at ev; (2) a small Rh 0 3d 5/2 side peak at ev; and (3) Rh 3+ /Rh 0 ratio of Consistent with the TPR result, the XPS data suggests the Rh sites in the fresh samples were partially oxidized. With increasing aging temperature, Rh 3d peaks shift to higher binding energy, together with increases in Rh 3+ /Rh 0 ratio ( ), suggesting a transition to a higher Rh oxidation state, i.e., Rh 0 45

66 Rh 3+. It is known that the oxidation process increases Rh oxidation states while the reduction process has an opposite effect [54]. The non-reducible Rh phase was reported resulting from a diffusion of Rh 3+ ions in subsurface regions of the alumina matrix. The binding energy of the new Rh phase is greater than that in Rh 2 O 3, indicating a different state from that of Rh 3+ in Rh 2 O 3, which is ascribed to metal-support interaction [189]. Table 7. Summary of detailed information of XPS spectra as shown in Figure 20, i.e., values of binding energy Rh 3d 3/2 for Rh 3+ and Rh 0 oxidation states, and Rh 3+ /Rh 0 (or Rh 3+ /Rh δ+ ) ratios a for fresh, aged, and regenerated 0.5% Rh/Al 2 O 3 and 0.5% Rh/CZO samples. 0.5% Rh/Al 2 O 3 0.5% Rh/CZO Catalyst E (Rh 3d 3/2 ), ev E (Rh 3d 5/2 ), ev Rh 3+ /Rh 0 Rh 3+ Rh 0 Rh 3+ Rh 0 (Rh 3+ /Rh δ+ ) b Fresh C aged for 5 min C aged for 5 min C aged for 5 min Regenerated (I) c Regenerated (II) d Fresh C aged for 5 min C aged for 5 min C aged for 5 min Regenerated (I) c Regenerated (II) d a Rh 3+ /Rh 0 ratio was calculated by comparing the integrated area under the corresponding fitted curves of Rh 3+ 3d 3/2 and Rh 0 3d 3/2 in Figure 9; b In Rh/Al 2 O 3, Rh 3+ and Rh 0 coexist, and Rh 3+ /Rh 0 ratios are compared. In Rh/CZO, Rh 3+ and Rh δ+ (0 δ <1) coexist, and the Rh 3+ /Rh δ+ ratios are compared; c Regenerated samples (I) were achieved by performing in situ regeneration using propane steam reforming with aged samples after aging at 1050 C for 5 min; d Regenerated samples (II) were achieved after 25 cycles in the simulated fuel cutoff-regeneration cycle tests. Different from 0.5% Rh/Al 2 O 3, low Rh valence state (Rh δ+, 0 < δ < 1) dominates in fresh 0.5% Rh/CZO (Rh 3+ /Rh δ+ ratio of 0.48). For 0.5% Rh/CZO, Rh 0 3d 5/2 peaks display higher BE values. The small but definite electropositive shifts detected for Rh 0 peaks are ascribed to the 46

67 existence of both Rh 0 and Rh δ+ species, giving evidence to the existence of Rh δ+ /Rh 0 and Ce 4+ /Ce 3+ redox couple [69]. This assignment is in agreement with previous FT-IR result [190,191], which shows the existence of surface electron deficient Rh δ+ species present on CZO support. Like Rh/Al 2 O 3, the Rh 3+ /Rh δ+ ratio in Rh/CZO increased with aging temperature. It is also important to note that the way Rh3d peak is interpreted largely affects the result. The XPS Rh 3d spectra for Rh/CeO 2 system studied by Force et al. [192] was deconvoluted into three peaks, respectively assigned to Rh 0 (306.8 ev), Rh + (307.8 ev), Rh 3+ (309.2 ev) states. While other systems have different interpretations [193]. In our Rh/CZO system, assigning XPS peaks to Rh 0 and Rh δ+ species is easier for comparison. Furthermore, the reduced areas under Rh 3d peaks for aged samples suggests Rh sintering and/or Rh dissolution into sintered support during simulated fuel cutoff aging. The characterization result is in agreement with the findings by Kang et al. [20]. In their study, the effect of aging atmosphere on the sintering behavior of commercial Pd- or Rh-TWC as well as the TWC performance were investigated under straight oxidizing, reducing, and periodic cycling aging conditions. For Rh-TWC, the diffusion of Rh 2 O 3 into the support along with the agglomeration of the Rh metal were found the main causes of catalyst deactivation during high temperature oxidative aging. XPS Rh 3d spectra of regenerated 0.5% Rh/Al 2 O 3 and 0.5% Rh/CZO in both show that after the first in situ regeneration, the oxidation state of Rh was significantly lowered, exposing more active Rh 0 species to the reactant atmosphere. This explains the enhanced reforming activity resulting from H 2 reduction (regeneration). 47

68 In summary, different types of interactions between Rh and support materials exist in Rh/Al 2 O 3 and Rh/CZO during fuel cutoff aging. It is well known that strong interaction between Rh and Al 2 O 3 with the formation of Rhodium Aluminate occurs in oxidative aging of Rh/Al 2 O 3 [2]. Compared to aged Rh/Al 2 O 3, the metal-support interaction in aged Rh/CZO occurs to a much lesser extent. Haneda et al. [194] reported that high-temperature aging can alter the surface properties of Ce x O y -ZrO 2 to inhibit the formation of formate species poisoning the catalytic active Rh sites. The superior regenerability of 0.5% Rh/CZO was believed mainly contributed by the coexistence of Ce 4+ /Ce 3+ and Rh 0 /Rh δ+ redox couple [ ]. Wang et al. [68], investigated the interaction between Rh and Ce x O y in Rh-Ce x O y /Al 2 O 3 catalyst system, with enhanced electron transfer efficiency during catalytic CO 2 dry reforming of CH 4. Similar promotional effect likely occurred with 0.5% Rh/CZO catalyst during regeneration, as confirmed by TPR and XPS results. The electron transfer pathways during catalyst regeneration are proposed in Figure 21. For Rh/Al 2 O 3, electrons are first donated by hydrocarbons (reactant C 3 H 8, and product C 2 H 6 and CH 4 ), and then transferred through the redox circle of Rh 0 Rh +, and finally accepted by H 2 O. Electron transfer is accompanied by redox reactions and the formation of H 2, CO, and intermediate products. For Rh/CZO, the coexistence of the Ce 4+ /Ce 3+ and the Rh 0 /Rh δ+ redox couple allows availability of Rh δ+ species, to accept the electrons donated by HC more easily. The efficient electron transfer pathway results in the significant catalytic steam reforming performance of Rh/CZO. 48

69 Figure 21. Proposed reaction mechanism and electron transfer pathways for steam reforming of propane on (a) Rh/Al 2 O 3 and (b) Rh/CZO catalysts Aging and regeneration mechanisms of Pd-based three-way catalysts (TWC) Aging-induced Pd sintering: the primary catalyst deactivation mode Figure 22 compares the activity of the fresh and aged Pd catalysts as measured by steam reforming H 2 production (mol % in the product) at simulated fuel rich condition. Generally, catalyst deactivation was modest at the lower aging temperature (800 C), but was accelerated with increasing aging temperature up to 1050 C. The extent of deactivation varied with the support material. The reforming activity for Pd/Al 2 O 3 showed little change after 800 C aging. In contrast, 49

70 Pd/CZO continuously deactivated as the aging temperature increased. The SR conversion losses at 350 C for Pd/Al 2 O 3 and Pd/CZO were respectively 59.67% and 72.29% after aging at 1050 C. Figure 22. Activity of fresh and aged (a) 3% Pd/Al 2 O 3 and (b) 3% Pd/CZO in converting simulated engine exhaust gas at fuel rich condition. Catalyst activity is plotted in terms of H 2 product mole fraction as a function of reaction temperature (200 C to 550 C). Aged catalysts were obtained by treating fresh ones in air at 800 C, 950 C, or 1050 C for 5 min, followed by cooling to room temperature in air. The catalyst activity after attempted regeneration (reduction by SR at simulated fuel rich condition) is compared in Figure 23. Attempted regeneration was not very effective in restoring the activity of Pd catalysts, with no gain in activity for 3% Pd/Al 2 O 3 and a modest conversion increase for 3%/CZO shown. Table 8 summarizes the T 50 s (temperature at which 50% maximum equilibrium H 2 production is reached during activity tests) of the Pd catalysts before and after aging/attempted regeneration, in comparison with the Rh catalysts. It is clear that the deactivated Rh catalysts were essentially fully regenerated while Pd catalysts were not. Clearly supported Rh and Pd have different deactivation modes. 50

71 Figure 23. Activity of fresh, aged, and attempted regenerated (a) 3% Pd/Al 2 O 3 and (b) 3% Pd/CZO in converting simulated engine exhaust gas at fuel rich condition. Aged catalysts were obtained by treating fresh ones in air at 1050 C for 5 min, followed by cooling to room temperature in air. Attempted catalyst regenerations were performed by SR at rich condition at 550 C for 1 h. Table 8. T 50 s ( C) of fresh, aged, and attempted regenerated 3% Pd/Al 2 O 3 and 3% Pd/CZO, in comparison with 0.5% Rh/Al 2 O 3 and 0.5% Rh/CZO as reported. Catalyst Fresh After Aging in Air for 5 min at Different T After Attempted 800 C 950 C 1050 C Regeneration * 0.5% Rh/Al 2 O % Rh/CZO % Pd/Al 2 O % Pd/CZO Annotation: * Attempted catalyst regenerations were performed at 550 C in propane-containing feed gas. The aging study was extended to multiple cycles consistent with the vehicle operations. The activity of both catalysts after every five cycles of aging-attempted regeneration are plotted in Figure 24. Continuous losses in the catalyst ability to be regenerated were observed with increasing number of cycles. The regenerability of Pd/CZO was always slightly greater than Pd/Al 2 O 3 in each cycle, based on the lower temperature for H 2 conversion and its larger slope. However, continuous 51

72 catalyst deactivation was experienced by Pd/CZO even up to 20 cycles of aging. In comparison, Pd/Al 2 O 3 seemed to stabilize at five cycles and beyond. Figure 24. Activity of attempted regenerated (a) 3% Pd/Al 2 O 3 and (b) 3% Pd/CZO catalysts after 1, 5, 10, and 20 cycles in the aging-regeneration cycle tests. After every five cycles of aging/regeneration. The catalyst activity is plotted in terms of H 2 product mole fraction as a function of SR temperature from 250 C to 550 C. Figure 25 shows the H 2 production vs. time on stream (TOS) during the attempted regeneration in the first five cycles for both Pd catalysts. For 3% Pd/Al 2 O 3, no enhancement in SR performance was observed for each attempted regeneration cycle, confirming that the Pd/Al 2 O 3 is non-regenerable by this treatment. For 3% Pd/CZO, the SR performance was observed to slowly increase indicating a slight recovery of catalyst activity. 52

73 Figure 25. H 2 productions as a function of time on stream (TOS) during attempted regeneration processes in simulated fuel cutoff aging-attempted fuel rich regeneration cycle tests (First 5 cycles out of total 25 cycles) with (a) 3% Pd/Al 2 O 3 and (b) 3% Pd/CZO. In each cycle, the catalyst sample was first aged in air at 1050 C for 5 min, followed by attempted in situ regeneration at fuel rich condition at 550 C for 1 h. These results indicate similar deactivation mechanisms for the two catalysts but a small regenerating effect for Pd/CZO was observed. TEM studies will confirm that Pd sintering is a major cause of deactivation. Previous studies reported that Pd metal sintering was prominent at high temperature ( 800 C), and was a major contributor to Pd catalyst deactivation [ ]. Upon aging at 1050 C, metal sintering occurred for both of our Pd catalysts. Metal sintering in Pd/Al 2 O 3 was more or less complete after multiple aging-regeneration cycles (>10 Cycles), leading to relatively stable catalyst performance. In contrast, metal sintering also occurred in Pd/CZO, but continued 53

74 with each cycle resulting in a continuous decrease in catalyst activity. Metal sintering in aged Pd/CZO is more difficult to be stabilized than for Pd/Al 2 O 3. Figure 26. Activity of fresh Pd/CZO catalysts with different metal loadings (0%, 0.5%, 1%, and 3%) in converting simulated engine exhaust gas at fuel rich condition. Catalyst activity is plotted in terms of H 2 product mole fraction as a function of reaction temperature (200 C to 550 C). Reaction feed: 500 ppm propane, 10% H 2 O, and N 2 balance, with total GHSV of 30,000 h 1. It is useful to mention that even though CZO has some steam reforming activity especially at high reaction temperature, Pd is always the active sites in the in situ catalyst regeneration. In another measurement (Figure 26), it is shown that the catalytic steam reforming activity increased dramatically with Pd loaded, compared to that of the CZO support-only. Significant Pd metal sintering, after aging, was observed by TEM measurements, with representative images presented in Figure 27, and TEM-derived metal particle size distributions of 3% Pd/Al 2 O 3 and 3% Pd/CZO catalysts before and after aging and attempted regeneration are shown in Figure 28. The Pd and support (Al 2 O 3 or CZO) crystals are differentiated by their distinct shapes, crystallite size, and electron transmission ability. Pd/support (Al 2 O 3 or CZO support) crystallographic structure was previously identified by the dark particles distributed on light background in their TEM images [154, ]. The electron scattering intensity from thin 54

75 specimens follows the Z2 dependence of Rutherford s law, where Z is the atomic number. For example, one Pt atom scatters as strongly as about 100 oxygen atoms or 32 silicon atoms. The technique is highly successful in detecting clusters of catalytic active metals such as Pt, Pd, or Rh, on light supports such as zeolites, or mesoporous silica and alumina [206]. It should be admitted that the metal-support contrast can also be affected by the exposed crystallite plane orientation. The crystallite shapes and sizes are therefore used to facilitate differentiation of the Pd metal from the support. In our TEM result, fresh Pd/Al2O3 (Figure 27a) is characterized by well-dispersed dark spherical spots (Pd metal) with a narrow range of small diameters (with mean Feret diameter of 4.08 nm) supported on needle-like Al2O3 crystal. Fresh Pd/CZO (Figure 27d) is characterized by visible faceted spherical spots (Pd metal) with a mean Feret diameter of 5.73 nm, supported on cubic shaped CexOy crystallite structure with ZrO2 incorporated. Figure 27. Representative TEM images of fresh, aged, and attempted regenerated 3% Pd/Al2O3 (a c) and 3% Pd/CZO (d f). Aged samples were achieved by aging the fresh catalyst in air or N2 at 1050 C for 5 min. Attempted catalyst regenerations were performed by at rich condition at 550 C for 1 h. 55

76 After 1050 C aging, significant Pd sintering was observed in the TEM images showing increased crystallite sizes. The metal size distributions of Pd/Al 2 O 3 (Figure 27b) and Pd/CZO (Figure 27e) shifted to higher values, with predominant particle sizes of nm and maxim particle diameter up to 20 nm. Ostwald Ripening has been reported to be the dominant process causing the growth of Pd nanoparticles leading to losses of metal surface area and catalyst activity [207]. After attempted regeneration (Figure 27c,f), little change in the crystallite size distributions were found, proving that the regeneration method was neither effective in reversing metal sintering, or causing any further catalyst deactivation. Figure 28. TEM-derived Pd metal particle size distributions of fresh, aged, and attempted regenerated (a) 3% Pd/Al 2 O 3 and (b) 3% Pd/CZO. Aging and attempted regeneration conditions were as described in Figure

77 Consistent with a recent study by Chen et al. [99], Pd metal sintering was the principle mode of catalyst deactivation when Al 2 O 3 was the support upon thermal oxidation. Compared to metal sintering, the authors claim support phase change and Pd-Al 2 O 3 interactions were less significant and less responsible for catalyst deactivation. Farrauto, et al. [208], found in 1992, that well dispersed PdO on Al 2 O 3 decomposes to Pd at 800 C in air in two distinct steps. The first step occurs between 750 C and 800 C, and a second decomposition happens between 800 C and 850 C, causing a greater extent of Pd sintering and agglomeration. Subsequent cooling in air below 650 C causes some re-dispersion with the formation of a surface rich in PdO x (PdO x -Pd/Al 2 O 3 ). The re-dispersion model was confirmed by Datye, et al. [100], who further demonstrated that PdO decomposition at higher temperatures (>800 C) leads to re-dispersion being more difficult. The hysteresis (temperature difference) for decomposition and reformation is also strongly dependent on the nature of the support material [209, 210], in the following decreasing order: ZrO 2 > Al 2 O 3 > Ta 2 O 3 > TiO 2 > CeO 2. Compared to PdO/Al 2 O 3, there exists a large region of temperature stability of the PdO when dispersed on CeO 2. However, the aging temperature (1050 C) in our study was much higher than previously reported. Therefore, attempted redispersion upon air-cooling was not effective in reversing the severe Pd sintering. Clearly if PdO could be stabilized against decomposition to Pd above 800 C, extensive sintering would not likely occur. Some advances in this concept have been reported in a US patent in which PdO forms a binary oxide with praseodymium [211]. These materials are stable above 1000 C however do sacrifice some activity relative to PdO only. A stable PdO suggests that decreases in the amount of Rh in the TWC might be possible since Pd has good NO x activity. Its high NO x activity was 57

78 exploited in the mid-1990s when all Pd catalysts (Rh free) were commercialized in TWC when regulations were not as demanding as they are today [2]. Metal sintering and metal-support interactions (Pd O Ce) are suggested as the two causes of catalyst deactivation for Pd/CZO. Only the metal-support interactions (Pd O Ce) could be slightly reversed by regeneration. The slightly enhanced activity of Pd/CZO after attempted regeneration, was likely attributed to the enhanced redox between Pd x+ /Pd 0 and Ce 4+ /Ce 3+ couples. Similar to Rh, Pd was reported to promote the transformation of Ce 4+ to Ce 3+ when deposited on CZO [171]. Due to the higher energy potential of the Ce 4+ /Ce 3+ redox couple (1.61 ev) than that of the Pd 2+ /Pd 0 (0.915 ev), oxygen vacancies form easily and act as the enhancement of oxygen spillover and back-spillover processes at the Pd/CZ interfaces [212]. The formation of Pd x+ /Pd 0 -Ce 4+ /Ce 3+ redox couples therefore benefits the oxygen-buffering effect, and facilitates molecular bond dissociation during catalytic reactions [ ] Support sintering and Pd-support interaction: other catalyst deactivation modes Support sintering occurred for both Al 2 O 3 and CZO and was accelerated with increasing aging temperature, as indicated by the sharp decreases in BET surface area (Figure 29). Most significant losses in surface areas occurred above 950 C (IV in Figure 29). Aging at >950 C for a 5 min period introduced measureable sintering of Al 2 O 3, but was insufficient for the phase transformation of γ-al 2 O 3 to θ-al 2 O 3 [218,219]. CZO sintering was previously reported accelerated in the presence of supported Pd [220]. Even though the extent of support sintering after high temperature (1050 C) aging was significant metal encapsulation was negligible as confirmed by TEM (Figure 27). The observed crystallite sizes of the sintered Pd metal and Al 2 O 3 or CZO support were 58

79 comparable in aged samples. After 1050 C aging, metal sintering was sufficient and the enlarged Pd particles were not able to be encapsulated by support crystals with similar sizes. Figure 29. BET surface areas of fresh and aged (a) 3% Pd/Al 2 O 3 and (b) 3% Pd/CZO, as a function of aging conditions. Aged samples were obtained by aging the (I) fresh catalysts in air at the following conditions: (II) 800 C for 5 min, (III) 800 C for 1 h, (IV) 950 C for 5 min, or (V) 1050 C for 5 min. The aging processes were followed by cooling in air to room temperature. As a reference, BET surface areas of support materials were measured m 2 /g and 60.3 m 2 /g respectively for fresh Al 2 O 3 and CZO. Powder X-ray diffraction (XRD) was used to investigate the phase structures of the fresh and aged 3% Pd/Al 2 O 3 and 3% Pd/CZO catalysts. The XRD patterns of the 3% Pd/Al 2 O 3 samples are shown in Figure 30a, b, with XRD peaks indexed to three phases, i.e., γal 2 O 3, Pd, and PdO. Wellstructured tetragonal γal 2 O 3 are characterized by exclusive (220), (311), (400), and (440) reflections and invariable (111), (222), (511), (444) planes. In agreement with the BET result, sintering of the Al 2 O 3 support is confirmed by increased crystallite size as confirmed by the Scherrer Equation. The visibility of metal/metal oxide XRD patterns is improved by aging, due to extensive metal sintering experienced by Palladium. The XRD patterns of the tetragonal PdO and cubic Pd crystallite structures in aged samples are visible but are of relative low intensity due to the low metal loading (3%) and small crystallite size. Consistent with literature [208], PdO is 59

80 formed during the air cooling. The most significant reflections of aged PdO are (110), (112), and (211). The Pd peaks including (111) and (200) partially overlap with the Al 2 O 3 patterns, and are only identifiable in aged samples. Figure 30. X-ray diffraction (XRD) patterns for (a) Fresh 3% Pd/Al 2 O 3 ; (b) Aged 3% Pd/Al 2 O 3, (c) Fresh 3% Pd/CZO; (d) Aged 3% Pd/CZO; and the zoomed views of (c,d). Aging condition: 1050 C in air for 5 min. XRD patterns were obtained by using Cu-Kα 1 radiation (λ = Å). 60

81 The XRD patterns of the fresh and aged 3% Pd/CZO are shown in Figure 30c,d, with zoomed view of the aged Pd/PdO-related patterns shown at the bottom of Figure 30. Based on our observation and a previous report [149], it can be concluded that most of the Zr incorporated into the CeO 2 phase forms a Zr x Ce 1 x O 2 solid solution of cubic symmetry, where x is equal to the actual Zr loading (~67%). Similar to Pd/Al 2 O 3 samples, significant sintering of both the metal and the support are suggested by Scherrer equation after aging. Table 9. Support mean crystallite sizes of the fresh and aged (1050 C in air for 5 min) 3% Pd/Al 2 O 3 and 3% Pd/CZO measured by TEM and XRD. Catalyst Fresh 3% Pd/Al 2 O 3 Fresh 3% Pd/CZO Metal Crystallite Size (nm) Support Crystallite Size (nm) XRD TEM XRD TEM Fresh Aged Fresh Aged The mean metal and support crystallite size analyses by XRD and TEM are compared in Table 9. The full-width at half maxima of the (400) and (111) peaks of respectively the γal 2 O 3 and CZO phases are used for the support crystallite size calculations. PdO (110) peak is used for the metal crystallite size calculations for the aged samples for both Al 2 O 3 and CZO supports. The fresh metal crystallite sizes cannot be detected because of the instrumental limitation. The crystallite size results are consistent for both the XRD and TEM. Compared to CZO support, Al 2 O 3 presents more refractory nature towards thermal oxidative aging, which is in agreement with the BET result. 61

82 Figure 31. Normalized H 2 consumption in H 2 -Temperature Programmed Reduction (H 2 -TPR) measurements of fresh and aged (a) 3% Pd/Al 2 O 3 and (b) 3% Pd/CZO, as a function of reducing temperature. Aged samples for measurements were respectively achieved by aging the fresh ones in air at the following conditions: 800 C for 5 min, 800 C for 1 h, 950 C for 5 min, or 1050 C for 5 min. The H 2 -TPR profiles of Pd catalysts are profiled with normalized H 2 consumption as a function of programmed temperature in Figure 31. A negative H 2 consumption peak (release of H 2 ) around 71 C were ascribed to the release of hydrogen from palladium hydride β-pdh x [ ]. β-pdh x was formed during H 2 purge prior to TPR analysis (to achieve TPR baseline), when H 2 was absorbed by metastable Pd crystallite particles at ambient temperature. The peak correlated well with the average PdO crystallite size, since the hydride formation is a bulk phenomenon [224]. The larger amount of H 2 released with increasing aging temperature indicates the existence of larger size Pd crystallites with a greater capacity for adsorbing H 2 via hydride formation. This further supports the Pd metal sintering observation. Pd reduction peaks at higher temperatures (~300 C) were observed for both catalysts, and were either assigned to the hydrogen consumption by a spillover from Pd to the support material [225,226], or to the reduction of stable PdO x species in intimate contact with the 62

83 support [227]. In either case, high temperature-h 2 consumption is attributed to the interactions between the well dispersed PdO x species and the ceria support. The existence of Pd O Ce interaction is evident by both the TPR and the activity results. Note that during attempted regeneration, Pd O Ce interaction could be partially reversed assisted by the Pd x+ /Pd 0 and Ce 4+ /Ce 3+ redox couples, while Pd/Al 2 O 3 showed no detectable regeneration. Compared to the low temperature reducible PdO crystallites as mentioned before, the PdO x particles in close contact with the support have smaller crystallite size. The number of the interactive PdO x particles decreased by being agglomerated into the larger PdO, creating a more sintered particle with a decrease in the high temperature H 2 consumption. However, the intensity of the metal-support interaction became stronger, as indicated by the shift of the reduction temperature to even higher temperature. The interaction of Pd with CeO 2 in intimate contact enhances the reducibility of the support [228], and positively influences the redox state of the active metal [ ]. For fresh and aged Pd/CZO, the TPR peaks observed between 100 C and 120 C are attributed to the reduction of Ce 4+ to Ce 3+ assisted by Pd. In comparison, the reduction of ceria for CZO support (no Pd present) occurs at much higher temperature (~600 C). With increasing aging temperature, the Pd-assisted reduction of ceria becomes more difficult as evident by the increased reducing temperature due to the loss of well dispersed Pd particles (in intimate contacts with CeO 2 ) consumed by the sintering process. 63

84 Figure 32. Schematic of proposed model for deactivation and regeneration of 3% Pd/Al 2 O 3 and 3% Pd/CZO during simulated fuel cutoff aging (1050 C) and fuel rich regeneration (550 C). The proposed deactivation and attempted regeneration mechanisms for 3% Pd/Al 2 O 3 and 3% Pd/CZO after aging at 1050 C and the fuel rich H 2 treatment (550 C) is sketched in Figure 32. For Pd on both supports, metal sintering was the main cause of catalyst deactivation. The aging temperature was sufficiently high for PdO decomposition to Pd, leading to significant metal sintering. Subsequent cooling was unable to reverse sintering via PdO reformation or re-dispersion. Aging also led to modest Pd O Ce interactions, which occurred mostly at the Pd/Ce x O y interfaces. The Pd species participated in the aging-induced metal-support interactions (indicated as red arrows) are highlighted with red borders. H 2 treatment (attempted regeneration) was not effective in reversing the sintering of 3% Pd/Al 2 O 3, but did allow slight reactivation of the aged 3% Pd/CZO by releasing the Pd species from the Pd O Ce interaction assisted by the Pd x+ /Pd 0 and Ce 4+ /Ce 3+ redox couples, by enhancing the redox efficiency. The electron current flowing between the metal 64

85 and support redox couples is highlighted in purple. In any case, the metal-support effects were negligible compared to the metal sintering, with the latter being the primary cause of irreversible Pd-TWC deactivation during thermal oxidative aging Catalytic methane emissions abatement on natural gas vehicles by steam reforming Reaction thermodynamics for reforming of methane, propane, or ethanol Table 10. Main reaction pathways during reforming of a specific hydrocarbon fuel, i.e. (a) methane, (b) propane, or (c) ethanol, in the presence of H 2 O (steam) and CO 2 in the reaction feed. Reaction (FWD) (a) (b) (c) Steam reforming CH S + H & O 3H & + CO C ) H 3 + 6H & O 10H & + 3CO & C & H ƒ O + 3H & O 6H & + 2CO & H 4 = +206 kj/mol H 4 = +374 kj/mol H 4 = +174 kj/mol Dry reforming CH S + CO & 2H & + 2CO H 4 = +247 kj/mol C ) H 3 + 3CO & 4H & + 6CO H 4 = +622 kj/mol C & H ƒ O + CO & 3H & + 3CO H 4 = +297 kj/mol Water gas shift CO + H & O H & + CO & H 4 = 41 kj/mol CO methanation CO + 3H & CH S + H & O H 4 = 206 kj/mol CO 2 methanation CO & + 4H & CH S + 2H & O H 4 = 165 kj/mol In our previous study [102], H 2 was produced by catalytic steam/co 2 co-reforming of a model automotive exhaust hydrocarbon (i.e. propane) at a simulated fuel rich condition, after fuel cut-off aging at 800 o C~ 1050 o C in air for transient periods. A similar regeneration was applied in the current study on natural gas engine exhaust treatment and attempted catalyst regeneration. At slightly fuel rich, natural gas engine exhaust contains excess steam (~10%) and CO 2 (~8%), and a considerable amount of methane (~500 ppm), allowing multiple reactions including steam reforming (SR), CO 2 dry reforming (DR), water gas shift (WGS), reverse water gas shift (RWGS), and methanation to occur in the downstream TWC converter. The fuel rich operation allows (1) CH 4 emissions abatement by reforming, which is less temperature-demanding than catalytic oxidation, and (2) sustained catalytic NO X reduction performance by maintaining the metallic 65

86 active state of Rh component. In addition, injection of a hydrocarbon fuel, e.g. ethanol or propane for a short period, is expected to further increase the product H 2 concentration, considering their high reforming capability [233]. The main reaction pathways as described with standard reaction enthalpies are listed in Table 10 (NO x reduction reactions are excluded here, assuming high conversion of NO x at fuel rich condition). H 2 is mainly produced through endothermic reforming reactions, accompanied by exothermic WGS reaction. An optimum reaction temperature will thermodynamically favor both reactions, allowing optimum yield of H 2. Figure 33. Thermodynamic model of equilibrium compositions for reforming reactions of different hydrocarbons: (a) Methane, (b) Propane, or (c) Ethanol. Equilibrium product gas mole fractions are plotted as a function of reaction temperature (200 o C and 700 o C). The theoretical model is constructed based on a feed composition (vol-%) of 500 ppm HC, 10% steam, 8% CO 2, and N 2 in balance (81.95%), at 1 bar. With a feed composition (in vol-%) of 500 ppm HC, 10% steam, 8% CO 2, and N 2 balance, the equilibrium product mole fractions are plotted against reaction temperature (200 o C to 700 o C) in Figure 33. At low temperature (< 300 o C), reforming thermodynamics is limited while WGS is favored. At higher temperature (> 400 o C), both reforming and RWGS reactions are thermodynamically favored, resulting in a slightly decreased H 2 product concentration and a significantly increase in CO production. At around 350 o C, full conversion of CH 4, maximized H 2 production, and minimized CO production are achieved at equilibrium for the reforming of all 66

87 three HCs. Therefore, moderate temperatures give desired reaction thermodynamics. Note that among the HCs being reformed, the H 2 equilibrium production follows the order: propane > ethanol > methane, suggesting that more H 2 regeneration and reduction of Rh 3+ will more likely occur by reforming of propane or ethanol compared to methane Catalytic methane reforming on fresh catalysts Figure 34. Activity of methane co-reforming with steam and CO 2 on fresh Rh- or Pd- model TWC catalyst (non-prereduced). (a) CH 4 conversion, (b) H 2, and (c) CO product mole fractions are plotted against the reforming temperatures (300 o C to 550 o C) with different fresh including 0.5% Rh/Al 2 O 3, 0.5% Rh/CZO, 3% Pd/Al 2 O 3, and 3% Pd/CZO. Reforming feed composition in vol-% is 500 ppm CH 4, 10% steam, 8% CO 2, and N 2 in balance (81.95%), with a total GHSV of 120, 000 h -1. The catalytic methane reforming activity, with steam and CO 2 present in the reaction feed, was first examined with fresh (non-prereduced) Rh- and Pd- model TWC catalysts on Al 2 O 3 and CZO carrier (Figure 34). Reforming allowed CH 4 to be catalytically converted at fairly low temperatures, with T 50 (i.e. reaction temperature at which 50% of the maximum conversion occurs) of around 380 o C and 450 o C for Rh and Pd catalysts, respectively. At temperatures above 480 o C, 67

88 reaction kinetics were favored on all fresh catalysts, allowing equilibrium CH 4 conversion and H 2 production to be reached. Pd-TWCs showed higher reforming activity than Rh-TWCs at the same given space velocity, which was mainly ascribed to the higher metal loading of Pd (3% vs 0.5%). Note that the Pd activity gradually increased when the temperature was raised from 300 o C to 450 o C, whereas Rh activity dramatically increased around 450 o C. In comparison, Pd was more resistant to mild oxidizing conditions, and exhibited superior catalytic methane reforming activity. Meanwhile, close attention should be paid to the CO produced, which increased at 430 o C. In onboard applications, the reforming temperature should be optimized according to engine calibration and exhaust specific condition for maximized CH 4 conversion and minimized CO emission Catalytic methane reforming on aged catalysts Catalyst aging was examined in Figure 35. A 5 min-air aging at 1050 o C caused significant deactivation of Rh catalysts, resulting in the CH 4 conversion curves shifted to 550 o C~ 650 o C. In comparison, only moderate Rh deactivation was observed (data not shown) when air was replaced by N 2 since no Rh oxidation occurs. Pd catalysts experienced moderate deactivation after aging. 68

89 Figure 35. Effects of catalyst fuel cutoff aging on the activity of of Rh-TWCS on methane reforming. with fresh (non- prereduced) or aged (a) 0.5% Rh/Al 2 O 3, (b) 0.5% Rh/CZO, (c) 3% Pd/Al 2 O 3, or (d) 3% Pd/CZO catalysts. The methane reforming feed is the same as described in Figure 34. Catalyst aging protocol includes treating fresh sample with excess air exposure at 1050 o C for 5 min, followed by cooling in air Methane emissions abatement by catalytic oxidation on fresh and aged catalysts Catalytic methane reforming is compared with the catalytic oxidation in Figure 36. Methane conversion/abatement was superior for catalytic reforming (blue curves) than for oxidation (purple curves), especially with Rh-TWC. The T 50 of CH 4 conversion reforming was approximately 100 o C~ 150 o C lower than the oxidation pathway. Meanwhile, reforming kinetics allowed full methane 69

90 conversion to be reached at ~480 o C, while oxidation kinetics were so limited that a maximum conversion of only around 60%~80% was achieved at 650 o C. Note that the only difference between the reforming and oxidation condition was 0.96% of O 2 in the later feed. Air aging at 950 o C for 5 min deactivated the Rh catalyst, and further lowered its methane conversion capacity in the subsequent oxidation activity tests (red curves). Therefore, an oxidizing exhaust environment, either created during high temperature fuel cutoff or stoichiometric/slight fuel lean operation, would cause a significant decrease in methane conversion with Rh. Exhaust oxygen concentration should therefore be lowered in order to maintain the Rh activity, which is especially important for NO x reduction. Pd catalysts showed tolerance to both mild oxidation and a more severe aging environment. Similar methane conversions through both reaction pathways were seen on fresh and aged Pd catalysts, especially on Pd/CZO. This is important because the enhanced Pd performance will potentially allow considerable methane conversion even though Rh is temporarily deactivated due to oxidation. If sufficient H 2 is produced through reforming on Pd, Rh in the same catalyst may be partially reduced enabling NO x reduction. The regeneration of Rh with individual Rh- TWC or combined Rh-Pd catalyst systems by reforming with different fuels will be discussed in the following section. 70

91 Figure 36. Activity of CH 4 reforming and CH 4 oxidation in the CH 4 abatement with Rh- or Pd- model TWC catalyst. CH 4 conversion profiles are plotted against reaction temperature during the isothermal CH 4 reforming and CH 4 oxidation activity tests with fresh (non-prereduced), aged, or regenerated (a) 0.5% Rh/Al 2 O 3, (b) 0.5% Rh/CZO, (c) 3% Pd/Al 2 O 3, or 3% Pd/CZO catalyst. The oxidation activity tests were performed with a feed consisting of (in vol-%) 500 ppm CH 4, 0.96 % O 2, 10% steam, 8% CO 2, and N 2 in balance (80.99%). Aging protocol includes excess air exposure of the fresh sample at 950 o C for 5 min, followed by cooling in air. The regenerated catalysts for CH 4 oxidation reaction were achieved by ethanol reforming with the aged ones at 550 o C for 1 hr Catalyst regeneration by reforming of ethanol 71

92 Attempted regeneration of aged Rh (air aging at 1050 o C for 5 min) was performed by using in situ reforming of methane at 550 o C for 1 hr, with the same reaction feed as that in the methane reforming activity test. There was no regeneration of activity below 550 o C, and therefore no reaction profile can be shown. Figure 37. CH 4 reforming activity against reaction temperatures and cycles during the aging-regeneration (by ethanol reforming)-activity cycle tests with (a) 0.5% Rh/Al 2 O 3 and (b) 0.5% Rh/CZO catalysts. All fresh catalysts were pre-reduced. In each cycle, aging was performed in excess air at 1050 o C for 5 min, followed by cooling in air. Regeneration was performed after aging via ethanol reforming at 550 o C for 1 hr. CH 4 reforming activity tests were performed with regenerated catalysts after 1, 5, 10, 15, and 20 agingregeneration cycles. Rh regeneration using catalytic reforming of injected ethanol is shown in in Figure 37 and Figure 38. Using a liquid such as ethanol brings convenience to the on-board operation. The source of ethanol can also be biomass, a renewable energy source which leads to zero carbon footprint. Ethanol reforming also gave stable Rh activity after periodic aging processes. It is realized that the 72

93 fresh Rh was already partially oxidized before activity tests, resulting in insufficient reaction kinetics at lower temperatures. Pre-reduction by ethanol reforming at 550 o C for 1 hr allowed the preservation of low-temperature reaction kinetics of Rh catalysts, as represented by the more gradual increases in CH 4 conversion with temperature (compare to the fresh Rh activity data shown in Figure 34). It should be emphasized that the flow rate of liquid ethanol injected into the reactor system generated 500 vppm in the feed was notoriously small (0.011 ml/hr), and can be optimized in the future. Nevertheless, significant regenerability using ethanol was still shown especially for Rh/CZO, while Rh/Al 2 O 3 shows some deactivation within 20 cycles. Increasing the ethanol concentration may minimize this effect. When carried on-board, ethanol allows easier catalytic reforming and sufficient hydrogen production for sustained Rh-TWC performance. The successful regeneration of Rh-TWC was very important not only for the catalytic methane removal, but also for the catalytic NO x reduction. 73

94 Figure 38. H 2 product mole fraction as a function of regeneration time on stream (TOS) during the agingregeneration (by ethanol reforming)-activity cycle tests as described in Figure 37. The rapid Rh regeneration through ethanol reforming is demonstrated in Figure 38, where H 2 product mole fraction over regeneration TOS is plotted. The H 2 generation in the ethanol reforming case rapidly reached equilibrium within the initial 5 min of each regeneration period. Ethanol shows the best regeneration performance among the hydrocarbons studied, and is therefore primarily proposed for the onboard application for natural gas engine exhaust treatment Catalyst regeneration by reforming of propane Figure 39. CH 4 reforming activity against reaction temperatures and cycles during the aging-regeneration (by propane reforming)-activity cycle tests with (a) 0.5% Rh/Al 2 O 3 and (b) 0.5% Rh/CZO catalysts. All fresh catalysts were pre-reduced. In each cycle, aging was performed in excess air at 1050 o C for 5 min, followed by cooling in air. Regeneration was performed after aging via propane reforming at 550 o C for 30 74

95 min. CH 4 reforming activity tests were performed with regenerated catalysts after 1, 3, 5, 10, 15, and 20 aging-regeneration cycles. Injected propane was also studied for the Rh catalyst regeneration (Figure 39). The regeneration by propane reforming effectively reversed the Rh deactivation due to periodic thermal oxidative aging at 1050 o C. Catalytic methane conversion by Rh was partially preserved after multiple cycles (as much as 20 cycles). Rh/CZO showed better regenerability than Rh/Al 2 O 3. A certain amount of metal/carrier sintering would account for the partial activity loss of Rh/Al 2 O 3. Note that for both catalysts, cyclic regeneration not only recovered the high temperature methane reforming kinetics, but also preserved the low temperature ones (< 460 o C). The regeneration through propane reforming is mainly ascribed to the effective Rh reduction by reforming product H 2 with sufficient intensity (concentration). Figure 40 shows H 2 production as a function of regeneration time on stream (TOS) during the total 20 cycles. The recovery of the catalyst activity after cyclic aging was clearly represented by the gradually increased H 2 production over regeneration time in each cycle. During regeneration, oxidized Rh sites experienced rapid (within the first 5~10 min) recovery of their metallic active states by H 2 reduction produced by propane reforming. This rapid increase in H 2 indicates that regeneration time can be significantly reduced to less than the 30 min used in the tests. Note that the activity recovery rate slightly deteriorated with incremental cycles, due to the involvements of other deactivation modes (metal and carrier sintering) as previously mentioned. In practice, the loss in CH 4 conversion due to the irreversible Rh deactivation could be partially compensated by the relatively stable Pd performance (Pd was very refractory towards thermal oxidizing deactivation as shown earlier). However, the NO x reduction activity will be largely affected if regeneration is not periodically applied. Furthermore, Pd is subject to other deactivation modes 75

96 such as poisoning by exhaust components and thus injected hydrocarbon provides a backup for regeneration. Figure 40. H 2 product mole fraction as a function of regeneration time on stream (TOS) during the agingregeneration-activity cycle tests as described in Figure Periodic air aging of Pd-TWCs The refractory nature of Pd-TWC is further demonstrated in Figure 41, which shows the CH 4 reforming activity after periodic aging. Air aging at 950 o C or 1050 o C with various durations was applied while no regeneration was applied in between the aging processes. The CH 4 reforming activity was sustained with at least 200 min-tos of aging at 950 o C for 3 Pd/Al 2 O 3, while further accelerated aging at 1050 o C caused only slight loss of the low temperature performance. For 3% Pd/CZO, the aging-induced loss in reaction kinetics only occurred at < 400 o C. 76

97 Figure 41. Air aging length effect on the cyclic CH 4 reforming activity with (a) 3% Pd/Al 2 O 3 or (b) 3% Pd/CZO catalyst. Catalyst activity was measured after different TOSs (5, 25, 50, 75, 100, and 200 min) of aging at 950 o C or 1050 o C. For 3% Pd/Al 2 O 3, moderate catalyst deactivation was observed after 950 o C aging for 100 min. Further aging at 1050 o C for longer term resulted in slight loss of catalyst light-off activity. For 3% Pd/CZO, moderate catalyst deactivation was observed after direct aging at 1050 o C for 200 min Regeneration of Rh-TWCs by ethanol reforming in the presence of Pd-TWC The commercial TWC catalyst formula includes both Rh and Pd components. A set of experiments using a Rh-Pd combined catalyst system would give significant practical meaning (Figure 42). During catalyst aging and regeneration, the Rh catalyst was placed immediately downstream (but in contact) from Pd/CZO. The question to be answered was could aged Pd catalyze sufficient H 2 production to regenerate the Rh 3+ to Rh metal. Pd/CZO was then removed from the reactor, retaining the Rh catalyst to be tested for a CH 4 reforming activity test. The regeneration effects by methane reforming (orange curves) and ethanol reforming (purple curves) with the Rh-Pd double layered catalyst were compared. Ethanol (orange) was superior for 77

98 regenerating aged Rh. The amount of H 2 produced through methane reforming with Pd (purple) was insufficient for the rapid recovery of Rh activity. This demonstrates that periodic regeneration through ethanol reforming is effective and necessary in sustaining Rh performance for in natural gas engine emission control. Figure 42. Comparison between ethanol reforming (orange curves) and methane reforming (purple curves) for the regeneration of aged (1050 o C in air for 5 min) (a) 0.5% Rh/Al 2 O 3 or (b) 0.5% Rh/CZO in the presence of 3% Pd/CZO. For the aging and regeneration with Rh-Pd catalyst mixture, Rh catalyst was placed immediately underneath the 3% Pd/CZO layer, and was separated from Pd with a quartz wool layer sandwiched in between. In the presence of Pd catalyst, aging was performed in air at 1050 o C for 5 min. and regeneration was performed by ethanol or methane reforming at 550 o C form 1 hr Scale-up CO 2 capture and methanation with dual functional materials (DFM) Cyclic tests of CO 2 adsorption and conversion with 5%Ru,10%CaO (or Na 2 CO 3 ) DFM on various Al 2 O 3 carriers 78

99 79

100 Figure 43. CO 2 adsorption and methanation performance of 5%Ru10%CaO (or 10%Na 2 CO 3 ) on different Al 2 O 3 supports for cyclic testing. The amounts of (a) CO 2 captured and (b) O 2 consumed at the adsorption step, and (c) CH 4 produced during the methanation step for 5%Ru,10%CaO prepared on (I) Al 2 O 3 powder, (II) Al 2 O 3 pellets-sasol TH200, (III) Al 2 O 3 pellets-sasol TH100, and (IV) Al 2 O 3 beads supports, and (V) 5%Ru,10%Na 2 CO 3 on Al 2 O 3 pellets TH100 support. CO 2 adsorption and methanation temperatures for all studies was maintained at 320 o C. Table 11. Average performance (in 10 cycles) for CO 2 adsorption and methanation for 5%Ru,10%CaO/Al 2 O 3 or 5%Ru,10%Na 2 CO 3 /Al 2 O 3 on different Al 2 O 3 carriers. The performance of # I~V materials were evaluated by integration of the adsorption and methanation curves in Figure 43. #VI DFM was evaluated by using Figure 44, when steam and O 2 were excluded from the CO 2 borne feed. # Catalyst CO 2 capture capacity Methanation CO 2 conv. a capacity b efficiency c I 5%Ru,10%CaO/Al 2 O 3 powder (BASF) % II 5%Ru,10%CaO/Al 2 O 3 pellets (TH200) % III 5%Ru,10%CaO/Al 2 O 3 pellets (TH100) % IV 5%Ru,10%CaO/Al 2 O 3 beads (SAS200) % V 5%Ru,10%Na 2 CO 3 /Al 2 O 3 pellets (TH100) % d VI 5%Ru,10%Na 2 CO 3 /Al 2 O 3 pellets (TH100) % e Annotations: a CO 2 capture capacity at adsorption step (g CO 2 / kg DFM) g of CO 2 captured / kg of DFM material; b Methanation capacity at methanation step (g CH 4 / kg DFM) g of CH 4 produced/ kg of DFM material; c CO 2 conversion efficiency (mol CH 4 / mol CO 2, %) Mole amount of CH 4 produced/ Mole amount of CO 2 captured. d and e Reason for > 100% CO 2 conversion efficiency is due to the continuous catalytic conversion of the carbonate species of NaHCO 3 or Na 2 CO 3. The CO 2 adsorption-methanation cycle tests were performed with DFM samples prepared on different Al 2 O 3 supports (powder, pellets with different pore volumes, and beads). For each material, indicated in Figure 43 and Table 11, the amount of CO 2 adsorbed and O 2 consumed during the adsorption step and CH 4 production were measured as a function of reaction time on stream (TOS) for a total of 10 cycles. During the adsorption step, the nano-dispersed CaO sites 80

101 were able to adsorb and bind CO 2 molecules, resulting in the formation of a labile (reversible) structure of chemisorbed CO 2 on CaO compared to bulk carbonate. In the presence of both O 2 and steam, ruthenium metal undergoes partial oxidation, rendering it inactive for CO 2 adsorption. This is reversed by in situ reduction with H 2 added during the methanation step. Both CO 2 adsorption and ruthenium oxidation are rapid processes, occurring within the first 15 minutes. In contrast, the conversion of adsorbed CO 2 to methane is a slower process (40-60 minutes) due to the reduction of the Ru oxide to catalytically active metal. After the recovery to its active metallic state, ruthenium was able to catalyze the conversion of CO 2 that had migrated (spilled-over) from the CaO to CH 4 and 2H 2 O. The CO 2 adsorption, methanation rates and capacities varied with the various carrier materials, reflecting pore diffusion of CO 2 to the dispersed CaO sites, H 2 reduction of the RuO x and desorption and pore diffusion of CH 4 into the bulk gas. The rate limiting step at this time appears to be reduction of the RuO x to the metallic state. Generally, stable performances were observed for all four catalysts during and after 10 cycles. The slightly higher CO 2 adsorption and the lower CH 4 production in early cycles suggests the possibility of unreactive carbonate formation on the CaO causing some temporary deactivation. This can be reversed by longer methanation times. The average performance over 10 cycles is given in Table 11. Among the Ru-CaO based materials studied, DFM supported on BASF powder showed the best CO 2 adsorption, due to its small particle size free from pore diffusion, while DFM supported on SASOL TH100 alumina pellets showed the best methanation performance. The high specific surface area and larger average pore size of TH100 alumina pellets may contribute to the improved DFM performance by enhancing the metal dispersion and the gas diffusion within the catalyst. 81

102 Different CO 2 adsorption behaviors were observed on CaO and on Na 2 CO 3. It should be noted that during the H 2 pre-treatment before the cycle test, significant amounts of CH 4 production was detected (not quantified) from the carbonate sample. It is likely the CH 4 is formed by the reaction between the H 2 and the Na 2 CO 3. During the cycle test, Na 2 CO 3 captured larger amounts of CO 2 (67.38 ml CO 2 by Na 2 CO 3 versus less than ml by CaO) within a total 20 min of adsorption TOS. On the other hand, the O 2 consumption by Na 2 CO 3 impregnated DFM sample was not as significant as CaO. This might be due to their different preparation procedures. The coimpregnation and co-precipitation method for 5%Ru,10%CaO/Al 2 O 3 allowed Ru and CaO to be more averagely dispersed on the Al 2 O 3 carrier. The sequential impregnation of Ru followed by Na 2 CO 3 impregnation allowed better exposure of Na 2 CO 3 to the reactant gas than Ru, partially resulting in enhanced CO 2 adsorption and reduced Ru oxidation. Meanwhile, Na 2 CO 3 itself demonstrated better CO 2 adsorption capacity than CaO. Furthermore, 5%Ru,10%Na 2 CO 3 /Al 2 O 3 showed significantly enhanced methanation capacity than 5%Ru,10%CaO/Al 2 O 3, while longer methanation TOS was needed for the completion of the methanation step with the former. The enhancement was represented by (1) accelerated breakthrough and (2) increased methane production. The latter is attributed to the maintenance of the reduced active state of Ru. The methane produced on Na 2 CO 3 DFM sample was even higher than the CO 2 adsorbed, due to the continuous catalytic conversion of the carbonate species in NaHCO 3 (formed during CO 2 adsorption) or Na 2 CO 3 (original structure) (Eq ). This material is under further investigation. 2NaHCO ) Na & CO ) + H & O + CO & (22) Na & CO ) Na & O + CO & (23) 82

103 With steam and O 2 excluded in the CO 2 adsorption feed, the performance of 5%Ru,10%Na 2 CO 3 /DFM was evaluated and plotted in Figure 44. Note that the space velocities for CO 2 adsorption in Figure 43 and 44 were different due to an instrumental limit, therefore the two results could not be directly compared. DFM material with Na 2 CO 3 and Ru demonstrated extraordinary CO 2 adsorption and methanation capacity. More rapid methanation response and stable adsorption/methanation performance were observed over cycles when steam and O 2 were not present in the adsorption feed. Compared to CaO, Na 2 CO 3 as an adsorbent demonstrated enhanced CO 2 storage capacity Both TH200 and TH100 were the most efficient in converting adsorbed CO 2 to CH 4. Compared to Ru-CaO system, Ru-Na 2 CO 3 system showed significantly enhanced adsorption and methanation capacities. Figure 44. CO 2 adsorption and methanation performance of 5%Ru,10%Na 2 CO 3 on different Al 2 O 3 supports for cyclic testing. CO 2 adsorption feed excluded steam and O 2, with composition of 7.5% CO 2, 0% steam, 0% O 2, N 2 bal., at a flow rate of 26 L/hr Effects of reaction parameters on the CO 2 capture and methanation capabilities (1) Adsorption time on stream (TOS) 83

104 Figure 45. Effect of adsorption time on stream (TOS) (60 min, 20 min, or 10 min) on CO 2 capture and methanation capabilities. DFM material loading: 5%Ru,10%CaO/Al 2 O 3 pellets (SASOL TH200) of around 10 g. Adsorption condition: 7.5% CO 2, 15% steam, 4.5% O 2, N 2 bal. (vol-%); GHSV=11236 h -1 ; 320 o C; 1 bar. Methanation condition: 5 vol-% H 2 / N 2 ; GHSV= 2611 h -1 ; 320 o C; 1 bar, 60 min From Figure 45 it is clear that the adsorption of CO 2 and the oxidation of Ru are rapid and essentially complete within the first minutes of the reaction. Further increases in TOS did not result in significant increases in reaction. Increases beyond 20 minutes during adsorption should be avoided since this leads to more extensive oxidation of the Ru that delays onset of methanation as shown in 45c between 10 (blue) and 20 (red) minutes of air exposure. However, more methane is produced with longer hydrogenation times (compare blue, red and green curves). These differences will likely be eliminated in a commercial plant that operates with pure H 2 rather than the 5% used in these experiments. (2) Adsorption and methanation temperatures 84

105 Figure 46. Effect of reaction temperature (350 o C, 320 o C, 300 o C, or 280 o C) on the CO 2 capture and methanation capabilities. DFM material loading: 5%Ru,10%CaO/Al 2 O 3 pellets (SASOL TH200) of around 10 g. Adsorption condition: 7.5% CO 2, 15% steam, 4.5% O 2, N 2 bal. (vol-%); GHSV=11236 h -1 ; 1 bar; 20 min. Methanation condition: 5 vol-% H 2 / N 2 ; GHSV= 2611 h -1 ; 1 bar, 60 min. Legends refer to volumes of respective gases. Both adsorption and methanation are exothermic processes and thus are thermodynamically favored at low temperatures, while reaction kinetics follow the opposite trend. Figure 46 shows the effect of adsorption temperature on the DFM performance. With increasing temperature for both adsorption and methanation between 280 o C to 350 o C, CO 2 adsorption slightly decreased, while ruthenium oxidation accelerates. Little methanation occurs at 280 o C while the optimum rate occurs around 320 o C. Extensive ruthenium oxidation occurs at the highest temperature (350 o C), and therefore higher temperatures should be avoided. (3) Variation of adsorption feed flow rate The effects of adsorption feed flow rates on the performance are plotted in Figure 47. Both (a) CO 2 adsorption and (b) O 2 consumption increase with increased feed flow rate indicating adsorption and oxidation rates are relatively fast. At lower flow rates (lower linear velocities) some bulk mass transfer dominates the process. It is clear from Figure 47c that methanation has a lower 85

106 rate than CO 2 adsorption given the larger TOS required for conversion. This is undoubtedly due to the slow reduction of the Ru oxides to the metallic and active methanation state. The extent of methane produced parallels the amount of CO 2 adsorbed. An optimum flow rate, for this configuration is L/hr, but further improvements in the reactor design are possible. Figure 47. Effect of adsorption feed flow rate (48.21, 40.00, 32.36, or L/hr) on the CO 2 capture and methanation capabilities. DFM loading: 5%Ru,10%CaO/Al 2 O 3 pellets (SASOL TH200) of around 10 g. Adsorption condition: 7.5% CO 2, 15% steam, 4.5% O 2, N 2 bal. (vol-%); 320 o C; 1 bar; 20 min. Methanation condition: 5 vol-% H 2 / N 2 ; GHSV= 2611 h -1 ; 320 o C, 1 bar, 60 min. Legends refer to volumes of respective gases. (4) Methanation feed flow rate Figure 48c demonstrates the delicate balance between low and high H 2 flow rates required for methanation. Moderate flow rates provide sufficient residence time with low mass transfer resistance resulting in higher methane production. The highest flow rate produces the highest methanation rate with sharper and narrower methane peaks at low TOS. This suggests that higher average H 2 concentrations will increase the rate of the reaction by enhancing the reduction rate for RuO x. Also our published kinetic rate model for methanation (Eley-Rideal mechanism) shows a strong dependence on H 2 partial pressure [130]. However, methane production is low due to 86

107 insufficient residence time. Therefore, the reactor should be designed with a high H 2 linear velocity but with sufficient residence time to allow for more complete methanation. Figure 48. Effect of methanation (H 2 /N 2 ) feed flow rate (22.4, 11.2, 5.6, or 2.8 L/hr) on the methanation capability. DFM catalyst loading: 5%Ru,10%CaO/Al 2 O 3 pellets (SASOL TH200) of around 10 g. Adsorption condition: 7.5% CO 2, 15% steam, 4.5% O 2, N 2 bal. (vol-%); GHSV=11236 h -1 ; 320 o C; 1 bar; 20 min. Methanation condition: 5 vol-% H 2 / N 2 ; 320 o C, 1 bar, 60 min. Legends refer to volumes of respective gases. (5) Influences of steam and/or O 2 in the CO 2 feed Figure 49. Effect of the presence of steam and/or O 2 in adsorption feed on the CO 2 adsorption and methanation capability. DFM material loading: 5%Ru10%CaO/Al 2 O 3 pellets (SASOL TH200) of around 87

108 10 g. Adsorption condition: GHSV=11236 h -1 ; 320 o C; 1 bar; 20 min. Methanation condition: 5 vol-% H 2 / N 2 ; GHSV= 2611 h -1 ; 320 o C, 1 bar, 60 min. The most important consideration in the scale-up study was the presence of O 2 in the CO 2 adsorption feed. Figure 49 shows the adsorption/conversion of CO 2 as the adsorption feed composition is varied. The presence of steam (without O 2 ) slightly decreases the CO 2 adsorption (blue ml vs purple in in Figure 49a). The combination of O 2 and steam (red vs. purple) has the most dramatic effect of decreasing both the CO 2 adsorption and subsequently, decreasing methane produced. With O 2 present, the Ru sites are oxidized making them inactive towards CO 2 adsorption. Therefore, with lower CO 2 adsorbed less methane is produced (Figure 49c). The DFM is most effective for feeds without steam or O 2. This is shown by the immediate and enhanced CH 4 production with the two non-o 2 feed conditions (purple and blue lines in Figure 49c). In the real flue gas applications, the exposure of DFM to O 2 cannot be avoided. However, a more O 2 -tolerant catalyst such as Rh is an expensive alternative to Ru. Furthermore, there are applications for CO 2 capture where the feed is O 2 -free such as in brewery exhausts, rich burn engines, and water gas shift reaction processes. In summary, the optimum process parameters for the CO 2 adsorption and methanation were achieved as listed below. (1) Adsorption condition: 7.5% CO 2, 15% steam, 4.5% O 2, N 2 bal. (vol- %); GHSV=11236 h -1 ; 320 o C; 1 bar; 20 min-tos; (2) Methanation condition: 5 vol-% H 2 / N 2 ; GHSV= 2611 h -1 ; 320 o C, 1 bar, 60 min-tos. It must be mentioned that in a real application H 2 will be 100% and thus the limitations using 5% H 2 in N 2 will likely disappear. 88

109 Figure 50. Proposed schematic mechanism of the surface reactions on Ru,CaO/Al 2 O 3 DFM for CO 2 adsorption and methanation. The Ru,CaO/Al 2 O 3 DFM surface reaction mechanisms for CO 2 capture and methanation are postulated in Figure 50. The co-impregnation and co-precipitation preparation methods allow uniform dispersions of the CaO and Ru nanoparticles on the Al 2 O 3 support. During the CO 2 adsorption step, the Ru catalyst is exposed to a CO 2 -containing feed in the presence of steam and O 2. CO 2 adsorption onto the CaO sites is rapid and equally rapid is ruthenium oxidation preventing it from adsorbing CO 2. The presence of steam slightly suppresses the CO 2 adsorption by competing with CO 2 for the occupation of the CaO sites. Subsequently, the reaction feed is switched to H 2 for the methanation process, involving the following elementary steps: (1) regeneration of catalytically active ruthenium metal by reduction of its oxide; (2) spill-over of adsorbed CO 2 from the CaO to the adjacent reduced Ru metal sites; (3) dissociative chemisorption of H 2 on Ru; (4) dissociation of CO 2 and H 2 to complexes, and (5) catalytic conversion to CH 4. 89