Use of Oxygen Isotopic Exchange to Explore Catalytic Activity and the Mechanism of Oxygen Reduction on Oxides

Material Information

Use of Oxygen Isotopic Exchange to Explore Catalytic Activity and the Mechanism of Oxygen Reduction on Oxides
Kan, Cynthia
Place of Publication:
[Gainesville, Fla.]
University of Florida
Publication Date:
Physical Description:
1 online resource (115 p.)

Thesis/Dissertation Information

Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Materials Science and Engineering
Committee Chair:
Wachsman, Eric D.
Committee Members:
Perry, Scott S.
Nino, Juan C.
Norton, David P.
Weaver, Jason F.
Graduation Date:


Subjects / Keywords:
Adsorption ( jstor )
Cathodes ( jstor )
Cobalt ( jstor )
Conceptual lattices ( jstor )
Flow velocity ( jstor )
Oxides ( jstor )
Oxygen ( jstor )
Signals ( jstor )
Surface temperature ( jstor )
Vapor phases ( jstor )
Materials Science and Engineering -- Dissertations, Academic -- UF
catalysis, cathode, isotopic, lscf, lsm, oxygen, solid
bibliography ( marcgt )
theses ( marcgt )
government publication (state, provincial, terriorial, dependent) ( marcgt )
born-digital ( sobekcm )
Electronic Thesis or Dissertation
Materials Science and Engineering thesis, Ph.D.


The electrochemical performance of intermediate temperature solid oxide fuel cells is limited by high polarization losses incurred from the oxygen reduction reaction at the cathode. The mechanism of oxygen reduction as well as the key characteristics responsible for high oxygen reactivity are currently not well understood and represent a hurdle in the targeted development of electro-catalytically active cathode materials. These properties were studied using heterogeneous catalysis techniques coupled with labeled oxygen. The primary materials studied were Lanthanum Strontium Manganite (LSM) and Lanthanum Strontium Cobalt Iron Oxide (LSCF). Temperature programmed isotopic exchange indicated that LSM was less active for surface exchange than LSCF, in agreement with previous results from the literature. This confirms the validity of isotopic exchange as a means to gauge the activity for oxygen reduction. Samples of LSM and LSCF were infiltrated with metal oxides to modify the surface properties and tested with this technique to identify trends in catalytic activity. Of the samples examined, LSCF was the most active. Addition of cobalt to LSM improved the activity while iron infiltration was detrimental. According to XPS, the iron on the surface of LSCF and LSM were of different oxidation states, which could explain why it caused a negative effect in LSM. These results indicate that optimization of catalytic activity is possible through surface modification; however, it is not only the surface composition, but also the electronic properties of the surface which determine activity. A two-step mechanism consisting of dissociative adsorption of molecular oxygen followed by incorporation into the lattice was proposed and used to model the behavior obtained from isothermal isotopic switching. Effective rate constants for each step were obtained by fitting the model to the gas phase isotopologue distribution. These parameter values indicate which step is rate-limiting. LSCF was shown to be limited by dissociative adsorption. The reaction for LSM was slower than predicted by the model, suggesting that the diffusion of oxygen from the particle core to the surface is the actual rate limiting step. Based off these results, LSCF is a good candidate for surface modification to improve dissociative adsorption, and LSM could be improved by increasing the oxygen diffusivity. ( en )
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
Includes bibliographical references.
Source of Description:
Description based on online resource; title from PDF title page.
Source of Description:
This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Thesis (Ph.D.)--University of Florida, 2009.
Adviser: Wachsman, Eric D.
Statement of Responsibility:
by Cynthia Kan.

Record Information

Source Institution:
Rights Management:
Copyright Kan, Cynthia. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Resource Identifier:
665095429 ( OCLC )
LD1780 2009 ( lcc )


This item has the following downloads:

Full Text




2009 Cynthia C. Kan 2


To Heywood 3


ACKNOWLEDGMENTS First, I thank my parents for their guidance and for raising me in an environment which nurtured my ambitions. My family has been a great source of happiness for me and I am very lucky to have such loving relatives. My s econd acknowledgement goes to Heywood for helping me adapt to graduate school and form a strong work ethic. He went to great lengths to ensure my happiness and seeing him was the best part of the day. During days when I just wanted to defenestrate my entire project, I am grateful fo r my friends, especially Ken and Christine, who were willing to listen to me rant. Without my strong support system, I would not have the confidence nor drive to pursue grea t challenges (like getting a PhD). Being surrounded by people who were intellec tually curious and motivated was an inspiration to me. I owe a great deal of gratitude to the wonderful staff of the Particle Science and Engineering Center and the Major Analytical Instrumentation Center for their patience and for teaching me how to use various instruments. I also thank my group members, for providing thoughtful discussion, assistance, their friendship and hours of en tertainment! It was a joy working with them all. In particular, I want to acknowledge Bryan, Eric and Martin for their considerable contribution to my laboratory knowledge. Another member who made my life much easier during the past four years is Je nnifer Tucker, who has done a tremendous job helping our group secure resources. I also thank Dr. Wachsman for providing this opportunity for me to learn and contribute to the collective knowledge of mankind. 4


TABLE OF CONTENTS page ACKNOWLEDGMENTS ............................................................................................................... 4LIST OF TABLES ...........................................................................................................................7LIST OF FIGURES .........................................................................................................................8CHAPTER 1 INTRODUCTION ................................................................................................................ ..132 BACKGROUND ....................................................................................................................182.1Solid Oxide Fuel Cell Overview ...................................................................................182.2The Cathode Reaction ...................................................................................................192.3Perovskite structured cathode materials ........................................................................202.3.1Lanthanum Strontium Manganite .....................................................................212.3.2Lanthanum Strontium Cobalt Ferrite ................................................................222.4Oxygen transport parameters D and k ...........................................................................242.5Gas phase isotopic oxygen exchange ............................................................................252.6 Proposed Mechanism of Oxygen Exchange ................................................................293 TEMPERATURE PROGRAMMED EVALUATI ON OF CATALYTIC PROPERTIES ....323.1Introduction .................................................................................................................. .323.2Experimental .................................................................................................................3 33.2.1Powder Materials ..............................................................................................333.2.2Catalysis System ...............................................................................................333.2.3Temperature Programmed Reaction, Desorption and Exchange ......................353.3Results and Discussion ..................................................................................................353.3.1Oxygen Non-Stoichiometry from TPR .............................................................353.3.2Temperature Programmed Isotopic Exchange ..................................................373.3.3Temperature Programmed Desorption after Isotopic Exchange .......................383.4Conclusion .................................................................................................................... 394 ISOTHERMAL ISOTOPIC SWITCHING ............................................................................414.1Introduction .................................................................................................................. .414.2Theory ........................................................................................................................ ...424.3Experimental .................................................................................................................4 54.3.1Isothermal Isotopic Switching...........................................................................454.4 Results and Discussion .................................................................................................474.4.1(La0.8Sr0.2)0.98MnO3 ........................................................................................474.4.2La0.6Sr0.4Co0.2Fe0.8O3.......................................................................................534.4.3Trends in LSM and LSCF .................................................................................564.4.3.1 Degradation of LSCF ............................................................................63 5


4.4.4Gas Phase Behavior Simulation ........................................................................644.5Conclusion .................................................................................................................... 745 IDENTIFYING TRENDS IN CATALY TIC ACTIVITY WITH SYSTEMATIC SURFACE MODIFICATION ................................................................................................775.1Introduction .................................................................................................................. .775.2Experimental .................................................................................................................7 85.2.1Materials ............................................................................................................785.2.2Powder Characterization ...................................................................................785.2.3Catalysis System Configuration ........................................................................795.2.4Temperature Programmed Isotopic Exchange ..................................................795.3Results and Discussion ..................................................................................................804.3.1X-Ray Photoelectron Spectroscopy ..................................................................824.3.2Temperature Programmed Isotopic Exchange ..................................................874.4Conclusion .................................................................................................................... 996 CONCLUSION .................................................................................................................. ...101APPENDIX A DERIVATION OF MODEL ................................................................................................104B PARAMETER FITTING SCRIPT .......................................................................................106LIST OF REFERENCES .............................................................................................................108BIOGRAPHICAL SKETCH .......................................................................................................115 6


LIST OF TABLES Table page 3-1 Temperature programmed conditions ................................................................................354-1 Exchange parameters obtained from simulation for LSM at 700 C .................................684-2 Exchange parameters obtained from simulation for LSM at 2% O2 .................................694-3 Exchange parameters for LSCF at 700 C, at various pO2 ................................................724-4 Exchange parameters for LSCF at 1% O2 40 sccm at 400-500 C ....................................724-5 Exchange parameters for va rying sample sizes of LSCF ..................................................725-1 Specific surface areas of infiltrated powders .....................................................................80 7


LIST OF FIGURES Figure page 1-1 Cell voltage and polarization losses ( U) as a function of operating temperature............142-1 Schematic of the triple phase boundary .............................................................................192-2 Oxygen reduction pathways for pure elec tronic, composite (electronic + ionic conducting phases) and mixed ioni c electronic conducting cathodes ...............................202-3 Unit cell of cubic perovskite structure ...............................................................................212-4 Oxygen exchange showing possible species formed .........................................................282-5 Schematic showing reactor feed and products ...................................................................303-1 Catalytic testing system comprised of tw o separate feed streams, a microreactor within a furnace and quadrupole mass spectromete ..........................................................333-2 Quartz micro-reactor schematic .........................................................................................343-3 La0.6Sr0.4Co0.2Fe0.8O3Oxygen non-stoichiometry (3) vs. temperature .........................363-4 Temperature programmed exchange prof ile for A) LSM and B) LSCF ......................373-5 Temperature programmed desorption af ter TPX for A) LSM B) LSCF ......................394-1 Oxygen isotopic switch profile th rough a room temperature reactor ...............................464-2 (La0.8Sr0.2)0.98MnO3 isothermal isotopic exchange under 1% O2 ...................................484-3 Argon Tracer Step-Profile ..................................................................................................494-4 Initial rates of exchange vs. oxygen partial pressure for LSM at 600 and 800 C ..............514-5 Arrhenius-like plot showing activ ation energy of oxygen exchange on LSM ..................524-6 Isothermal exchange of LSCF at 650 C and 10,000 ppm O2 ............................................534-7 Initial rates of exchange vs. oxygen partial pressure for LSCF at 600 C ..........................544-8 Arrhenius plot showing activati on energy of oxygen exchange on LSCF ........................554-9 Isothermal isotopic switch profile at 800 C for A) LSM and B) LSCF ..................564-10 Integration of amount of 16O from the gas stream .............................................................574-11 Fraction of oxygen in solid converted from 16O to 18O vs. time .......................................58 8


4-12 Reaction rate versus square root of fr ee stream velocity divided by pellet size ................594-13 Conversion of LSCF in 1% O2...........................................................................................604-14 Switching profiles for LSCF A) 300 C B) 400 C and C) 500 C ...............................634-15 Conversion of LSCF vs. time at 500 C as a function of sample size ...............................634-16 Exchange with fresh LSCF and LSCF after many cycles ..................................................644-17 Numerical solution for A) Dissociative adsorption limite d case B) Incorporation limited case .................................................................................................................. ......654-18 Value of parameters A and B as a function of iteration number for LSM .........................664-19 Profile of 16fads and 16fbulk obtained from LSM with model superimposed ........................674-20 Normalized 16O2, 18O2 and 16O18O signals and predicted profile for LSM ........................684-21 Value of parameters A and C as a function of iteration number for LSCF .......................704-22 Profile of 16fads and 16fbulk obtained from LSCF with model superimposed .......................704-23 Normalized 16O2, 18O2 and 16O18O data and predicted profile for LSCF ...........................714-24 Normalized data and predicted profile for LSCF, calculated by fitting the 18fads data instead of the 16fads data ......................................................................................................715-1 Powder morphology captured with SEM, taken at 20,000x ..............................................815-2 Diffraction patterns of LSM, LSM + 20 wt% Co, Co3O4 ..................................................815-3 Survey spectra (XPS) for LSM and LSCF taken at 45 angle ...........................................825-4 (La0.8Sr0.2)0.98MnO3 O1s peaks A) before heating B) after heating ............................835-5 O1s spectra for LSM infiltrated with 5 wt% A) La, Sr B) Fe, Co .................................835-6 O1s spectra for LSCF and LSCF + 5 wt% Mn ..................................................................845-7 La 3d5/2 peaks for LSM and LSM + 5 wt% La ..................................................................855-8 Sr 3d peaks for LSM and LSM + 5 wt% Sr .......................................................................855-9 Co 2p3/2 profile of LSM + 5wt% Co and LSCF .................................................................865-10 Fe 2p3/2 profile of LSM + 5wt% Fe and LSCF ..................................................................875-11 Temperature programmed exchange profile of infiltrated samples ...................................89 9


5-12 Temperature programmed exchange prof ile of LSM + 5 wt% Co plotted with statistical equilibrium model ..............................................................................................915-13 Net oxygen desorption from temperatur e programmed reaction (TPR) profile for LSM and LSCF ..................................................................................................................925-14 Temperature programmed exchange profile for LSCF + 5 wt% Mn.................................955-15 Temperature programmed exch ange profile for 40 mg Co3O4 ..........................................965-16 Conversion (1-normalized 18O2) vs. temperature ...............................................................985-17 Formation of 16O18O (normalized by total oxygen) as a function of temperature .............99 10


Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy USE OF OXYGEN ISOTOPIC EXCHANGE TO EXPLORE CATALYTIC ACTIVITY AND THE MECHANISM OF OXYGE N REDUCTION ON OXIDES By Cynthia Kan May 2009 Chair: Eric D. Wachsman Major: Materials Scie nce and Engineering The electrochemical performance of intermed iate temperature solid oxide fuel cells is limited by high polarization losses incurred from the oxygen reduction reaction at the cathode. The mechanism of oxygen reduction as well as the key characteristics responsible for high oxygen reactivity are currently not well understo od and represent a hurdle in the targeted development of electro-catalytic ally active cathode materials. These properties were studied using heterogeneous catalysis techniques coupl ed with labeled oxygen. The primary materials studied were (La0.8Sr0.2)0.98MnO3 (LSM) and La0.6Sr0.4Co0.2Fe0.8O3(LSCF). Temperature programmed isotopic exchange indicated that LSM was less active for surface exchange than LSCF, in agreement with previous results from the literature. This confirms the validity of isotopic exchange as a means to gaug e the activity for oxygen reduction. Samples of LSM and LSCF were infiltrated with metal oxides to modify the surface properties and tested with this technique to identify trends in catalytic activity. Of the samples examined, LSCF was the most active. Addition of cobalt to LSM improved the activity while iron infiltration was detrimental. According to XPS, the iron on the surface of LSCF and LSM were of different oxidation states, whic h could explain why it caused a ne gative effect in LSM. These 11


12 results indicate that optimization of catalytic activity is possible th rough surface modification; however, it is not only the surface composition, but also the electronic properties of the surface which determine activity. A two-step mechanism consisting of dissocia tive adsorption of mo lecular oxygen followed by incorporation into th e lattice was proposed and used to model the behavior obtained from isothermal isotopic switching. Effective rate cons tants for each step were obtained by fitting the model to the gas phase isotopologue distribution. These parameter va lues indicate which step is rate-limiting. LSCF was shown to be limited by dissociative adsorption. The reaction for LSM was slower than predicted by the model, suggesti ng that the diffusion of oxygen from the particle core to the surface is the actual rate limiting step. Based off these results, LSCF would be a good candidate for surface modification to improve dissociative adsorption and LSM could be improved by increasing the oxygen diffusivity.


CHAPTER 1 INTRODUCTION The relatively high standard of living enjoyed in industrialized countries today is afforded by the immense consumption of energy and resources [1]. With developing countries eager to catch up and benefit from the amenities of a first world lifestyle, their energy needs will increase and put an additional strain on the limited supply of fossil fuels. The growing demand for energy must continue to be met or else depressed economic growth and societal instability will ensue [1]. This presents a great technological challe nge as fossil fuels such as coal, petroleum and natural gas provided over 80% of the worlds energy in 2004 [2]. This heavy dependence on non-renewable energy sources cannot be sustained indefinitely. Therefore, a comprehensive portfolio of renewable energy so lutions must be developed to meet present and future energy needs. Some of the energy technologies in development today are so lar cells, wind turbines, and biofuels. These sources are great for electricity production but are not convenient for energy storage, with the exception of biofuels. One issue concerning biofuels is th at they are subject to many of the same problems facing other car bon based fuels. Harnessing energy through combustion, while economical, is inefficient and a major contributor to environmental pollution. Incomplete hydrocarbon combustion produces harmful compounds such as soot, carbon monoxide, and nitrogen oxides. A cleaner and mo re efficient method of releasing the energy stored in chemical fuel is through electroc hemical conversion by fuel cells. As long as a continuous supply of air and fuel are introduced into the fuel cell, electricity is produced. In particular, the solid oxide fuel cell (SOFC) is especially attractive due to its high energy efficiency and fuel flexibility. The worldwide av erage efficiency for electricity produced from combustion of fossil fuels is a mere 36% [3] whereas SOFCs can reach 60% stand-alone 13


efficiency and up to 85% in a combined heat and pow er system [4]. All fuel cells are able to run off of hydrogen, but SOFCs operate at high temperature (1000 C), which allows various hydrocarbon fuels to be reformed internally. Gi ven the remarkable bene fits of this technology, developers are excited about commercializing SOFCs for widespread use. However, there are currently several significant barriers to bringing SOFCs to market. One consequence of operating at such high te mperatures is that it requires the use of expensive and specialized materials in the cell a nd in the balance of plant [5]. Decreasing the operating temperature to the in termediate range of 600-800 C would allow more economic materials, such as stainless steel, to be us ed for interconnects and w ould decrease insulation requirements. It would also reduce the start-up time as well as degradation rates. All these benefits, however, come at the cost of performanc e [6,7]. It is shown in Figure 1-1 that as the Figure 1-1. Cell voltage a nd polarization losses ( U) at constant current density of 100mA/cm2 as a function of operating temperature [6]. Reprinted from Journal of the European Ceramic Society 21, E. Ivers-Tiffe, A. Weber, D. Herbstritt, Materials and technologies for SOFC-components, 18051811, Copyright (2001), with permission from Elsevier 14


temperature decreases, losses from the cathode become the largest component. In order to preserve adequate performance while extending the operating temperature into the intermediate range, the kinetics of the heterogeneous electrode reaction must be improved by designing better catalysts. Before targeted development can occu r, however, a fundamental understanding of the oxygen reduction processes occurring on the cathode surface must be obtained. Currently, the mechanism and rate laws gove rning surface reactions on cathode materials are not well established [8-11]. Advancements in the state of know ledge of oxygen surface exchange have been made using methods such as conductivity-relaxati on [12-16], temperature programmed reaction [17-20], isotopic excha nge depth profiling [21,22], and impedance spectroscopy [8,23,24]. However, it is difficult to get mechanistic information from these approaches. Heterogeneous catal ysis techniques such as temperature programmed and steady state isothermal isotopic exchange hold much pr omise in shedding light on the cathode reaction kinetics, and are explored in depth in this dissertation. The use of isot opically labeled oxygen gas is critical due to the la rge normal oxygen background in the oxide materials examined. Interpretation of results thr oughout this dissertation is through the framework introduced in Chapter 2, where a two-step mechanism of dissoci ative adsorption followed by incorporation is proposed. The gas phase molecular oxygen isotopol ogue distribution is determined statistically by the isotopic composition of surface oxygen adatoms. This key concept is used in Chapters 4 and 5 to demonstrate the validity of the model. The organization of the dissertation is as follo ws. Chapter 2 is a literature overview of the cathode reaction, material properties and theory of oxygen transport and surface exchange. The purpose of Chapter 3 is to introduce the experimental system and show th at the results obtained from our experimental system are comparable wi th those from the literature. Additionally, the 15


temperature programmed isotopic exchange behavi or of the two main cathode materials is discussed in detail and serves as a reference po int for analysis in later chapters. The use of isothermal isotopic exchange, where the feed stream oxygen is otopic labeling is switched abruptly, and the ensuing analysis is presented in Chapter 4. This method is used to identify trends in the time dependent behavior of the di fferent cathode materials. An extensive section on extracting the effective rate constants from para meter fitting is included at the end. These rate constants, coupled with the observed behavior, ar e used to determine the rate limiting step of oxygen exchange. Chapter 5 covers trends in cata lytic activity from isotopic exchange of cathode materials infiltrated with various metal oxides. Before this work, the surface exchange r eaction was characterized mainly through mass transport coefficients [25] which were not mechanistic in nature. The application of isotopic exchange to determine oxide reactivity was interpreted, as recently as 20 05 [26], using the same one-step mechanisms used in the 1960s [27] which took into account the exchange of zero, one or two atoms between the gas and solid phases. These reactions do not constitute a mechanism and therefore cannot be used to determ ine the rate limiting proce ss. Instead of merely applying the technique of isotopi c exchange to cathode materials, and then using the deeply rooted and stale analysis from the literature, th e better approach was to start from scratch and develop my own framework for interpreting result s from isothermal isotopic exchange. As a result, the following contributions have been made: Developed a comprehensive framework for oxygen isotopic exchange analysis based off of a two-step mechanism of dissociative adsorption followed by incorporation 16


17 Modeled the gas phase and sample surf ace/bulk oxygen compositions as a function of time for two SOFC cathode materials Created a custom program to extract mech anistic rate constants from parameter fitting isothermal isotopic exchange profiles. Determined that the catalytic activity of an oxide material can be affected through surface modification without altering the apparent mechanism of oxygen exchange


CHAPTER 2 BACKGROUND 2.1 Solid Oxide Fuel Cell Overview A fuel cell is an electrochemical device wh ich transforms chemical energy stored in gasified fuel into electricity through a redox reaction. Like batteries, fuel cells have three main components: the anode, electrolyte and cathode. Interconnects are used to join the devices in series to produce the required el ectrical output. The electrolyte, which is the component used to classify the various types of fu el cells, ideally has zero electr onic conductivity, which forces the electrons liberated by the electrode reactions into the external circ uit to perform useful work. For solid oxide fuel cells (SOFCs), a thin dense layer of oxygen ion conducting ceramic, such as doped ZrO2 or CeO2, functions as the electrol yte. Sandwiching the electro lyte on either side are the electrodes where the half cell reactions o ccur. On the cathode, oxygen is reduced to O2and incorporated into the electrol yte (Eqn. 2-1). After passing through the electrolyte, the oxygen ion combines with hydrogen on the anode side to form water and electrons (Eqn. 2-2). Cathode O2 (g) + 2e O22-1 Anode H2 (g) + O2H2O (g) + 2e 2-2 Overall H2 + O2 H2O 2-3 As long as fuel and oxidant are supplied, th e reaction will conti nue to produce power. Characteristics of a good electrode are (1 ) high electrical conduc tivity (2) long term stability (3) fast gas transport and (4) high catalytic activity for redox reactions. The state-of-theart anode material is currently accepted to be nickel/yttria-stabilized zirconia (YSZ) cermet for its affordability, good electro-mechan ical properties and catalytic activity [28]. In the case of the cathode, however, there is no material which adequately satisfies all the above criteria in the intermediate temperature range. 18


2.2 The Cathode Reaction The cathode half cell reaction, written in KrgerVink notation, appears to be quite simple: O2 (g) + 2e + Vo Oo x 2-4 However, this reaction actually encompasses a series of several elementary processes ranging from diffusion to dissociat ive adsorption and charge transfer, finally resulting in incorporation into the solid [11,25,29-32]. For mu lti-step chemical reactions, it is common for one elementary reaction to be significantly slower than the ot hers. It is this step which determines the overall rate of the reaction and is known as the rate determining step [30,33]. Once it is understood what step is limiting the overall reaction, catal ysts can then be engineered with the proper functionality. The reduction reaction pathway is expected to vary depending on the cathode material. For a purely electronically conducting cathode, such as platinum, the oxygen reduction reaction is thought to be restricted to the triple phas e boundary (TPB) where the electronically conducting, oxygen ion conducting and gas phases converge (Figure 2-1). This is because the reactants Figure 2-1. Schematic of the triple phase boundary showing the electron ically conducting phase ( ), gas phase ( ), and ionically conducting phase ( ) and their roles in oxygen reduction 19


(electrons, vacancies and oxygen) are unique to a particular component, and can only participate in oxygen reduction when all three ar e available. To increase the reaction area from a 2D line to a 3D surface, researchers turned to mixed ioni c/electronic conductors (MIECs) so that oxygen could be exchanged throughout the entire cathode/el ectrolyte interface (Fi gure 2-2). Because of this difference in pathway between the electronic and mixed conductors, it is expected that the mechanism of oxygen exchange will also be differe nt. Thus, one material of each type has been chosen for examination using isotopic exchange. Figure 2-2. Oxygen reduction pathways for pure electronic, composite (electronic + ionic conducting phases) and mixed ionic electronic conducting cathodes [34] Reprinted from Journal of Power Sources 127, A. We ber, E. Ivers-Tiffe, Materials and concepts for solid oxide fuel cells (SOFCs ) in stationary and mobile applications, 273, Copyright (2004), with permission from Elsevier 2.3 Perovskite structured cathode materials As mentioned previously, cathode materials mu st fulfill numerous requirements [35], not the least of which is affordability. Inexpensive repl acements for porous platinum cathodes used in the past are oxides which crystallize in the ABO3 perovskite structure (Fig ure 2-3). This material system, commonly with a rare earth metal in the A site and a transition metal in the B site, is 20


very accommodating and allows for the introduction of various defects through doping on the A or B site. Through controlled defe ct engineering, va rious properties, such as electronic conductivity or thermal expansion, can be optimized. The specific materials chosen for study in this work are perovskites which have s hown good performance and are well studied [10,14,3639]. Both are available commercially. Figure 2-3. Unit cell of cubi c perovskite structure 2.3.1 Lanthanum Strontium Manganite Strontium doped lanthanum manganite (LSM) is the most studied and prevalent SOFC cathode material [22,29,30,32,37,40] due to its long term stability and the fact that its thermal expansion coefficient is very similar to that of the electrolyte material YSZ. Although LSM performs well at high temperatures, cathodic polar ization becomes significa nt at intermediate temperatures, limiting its usefulness as an inte rmediate temperature SOFC cathode material. 21


Therefore, it is of great interest to understa nd the oxygen reduction pro cess on LSM so that the catalytic activity may be improved. The undoped base material, lanthanum mangan ite, is p-type with a conductivity of approximately 75 S/cm at 1000 C [41]. Strontium substitution of the A site lanthanum enhances the electrical conductivity and reaches a maximum at around 15-20% (150 S/cm [41,42]). The addition of a divalent ion results in the Sr2+ on a La3+ site being charge compensated through the formation of Mn4+ (Eqn. 2-5). This enhances the sma ll polaron conduction mechanism where charges hop between Mn3+ and adjat [43] n acenMn4+2,4. 2-5 In addition to the electrical c onductivity, the oxygen non-stoichio metry is also of interest. Defects, such as oxygen vacancies, in ionic so lids are in equilibrium with the external environment at sufficiently high temperatures and doped lanthanum ma nganite is no exception. This work focuses on the specific composition (La0.8Sr0.2)0.98MnO3 (LSM), and as the indicates, the material can be superand sub-stoichiometric [38,42]. At 1000 C, the material does not begin to become sub-stoichiometr ic through loss of oxygen until the oxygen partial pressure (pO2) drops below 10-10 atm. The lack of oxygen vacancies translates to a low ionic conductivity for LSM, with values between 10-7 and 10-6 S/cm being calculated from diffusivity data at 900-1000 C [37,44]. Given that the di fference between the electronic and ionic conductivity is over 8 orders of magnitude, it is appropriate to classify LSM as a purely electronic conductor. 2.3.2 Lanthanum Strontium Cobalt Ferrite More recently, interest in developing MIEC cathodes has directed res earch efforts toward the La1-xSrxCo1-yFeyO3system [45,46]. These materials ar e double substituted perovskites with elements of similar ionic radii occupying the same site (La and Sr on A site, Co and Fe on B site) 22


[30]. The substitution of Sr2+ on the La3+ sublattice requires that elect roneutrality (Eqn. 2-6) be maintained by either oxidizing the B site element from +3 to +4 or through formation of oxygen vacancies [47]. 2-6 At low temperatures, significant electroni c compensation occurs, thought to be from preferential formation of Fe4+ [39]. At high temperature, unlik e its effect in LSM, strontium substitution on the lanthanum site results in ionic compensation via formation of oxygen vacancies [47]. This lead uc tncentration of BB s to a re dtion inhe co 2-7 This explains the materials decrease in electronic conductivity a nd increase in ionic conductivity at high temperatures corresponding to reaction 27. In addition, the electrical conductivity is also affected by the B site elemen t. For a given A site composition, the electrical conductivity can be increased by simply increasing the B site cobalt content [47,48]. However, the addition of cobalt also ha s the unintended effect of in creasing the thermal expansion coefficient (TEC); therefore a careful balan ce between conductivity and TEC matching with the electrolyte must be achieved. The addition of ir on to the B site decrea ses electronic conductivity but also restricts thermal expans ion. Due to insulating zirconate phase formation at the interface [49], cobalt containing cathodes are not compatible with YSZ, and mu st be interfaced with ceria based electrolytes for which th ere are no known reactions [46,48]. It is generally accepted that the composition La0.6Sr0.4Co0.2Fe0.8O3(LSCF) is well optimized and mechanically compatible (17.5 x 10-6 K-1 TEC [50,51]) with gadolinium doped ceria (GDC), which has a TEC around 12.5 x 10-6 K-1 [51]. At 900 C, LSCF has an ionic conductivity of 0.23 S/cm and an electronic conductivity of 252 S/cm [47]; the corresponding 23


ionic transport number is 9 x 10-4, which indicates the majority of the conductivity is electronic. Compared to the ionic conductivity of a material like LSM, ho wever, LSCF is orders of magnitude more conductive, and it is even more ionically conductive than GDC electrolyte material at the same temperature [52]. According to the Nernst-Einstein relationship, the ionic conductivity is proportional to the concentration [37,47] and diffusivity of oxygen vacancies. At room temperature, LSCF is stoichiometric with respect to oxygen in air [47] and becomes non-stoichiometric at temperatures above 600C and at a pO2 below 1.3 x 10-3 atm [53,54]. 2.4 Oxygen transport parameters D and k Oxygen transport through MIECs via defects is generally well understood [9], but surface processes remain unclear due to experimental limitations and variations. In the literature, transport between the gas/solid interface is commonly described by two parameters, the bulk diffusivity (D) and surface exchange coefficient [14,16,55] (kse). D and kse values are usually found experimentally by applying one of the foll owing driving forces to the material under study: electrical potentia l gradient, isotopic trac er concentration gradient [40,56] or chemical potential gradient [14,57]. However, kse values found from differen t techniques are usually not comparable for reasons given by Maier [58]. A ttempts have been made to correlate the kse values to the mechanism of oxygen reduction using ch emical kinetics [25,58-60], but a consistent expression for a given mechanism has yet to emerge. One technique used to get D and kse is called conductivity relaxation, which utilizes an abrupt change in the pO2 to generate a change in the conduc tivity of the sample. Additionally, another difficulty of studying oxygen reduction on oxides is that the large oxygen background from the lattice makes it difficult to study the re acting oxygen on the surface, therefore the use of isotopically labeled oxygen, 18O2, is necessary. More recently, D and kse values have been 24


obtained from isotope exchange depth profiles (IE DP), which are generated from secondary ion mass spectrometry on samples that have been annealed in isotopica lly enriched oxygen [21,22,55]. kse and D values obtained using an isotopi c tracer are designated k* and D*. At 800 C in 1 atm of O2, k* for LSM and LSCF is 6.3 x 10-9 and 1.0 x 10-5 cm/s, respectively [36,40]. D* is 4.0 x 10-15 cm2/s for LSM [40] and 2 x 10-8 cm2/s for LSCF [36]. These values show that oxygen transport through LSCF is clear ly more facile than in LSM. There is evidence that, in general, IEDP is much more sensitive to D than kse. A couple of studies [56,61] show excellent agreement between the fitted and measured depth profile for the bulk of the material, but not within the first 10-20 nm, which is where kse is most significant. The work by De Souza [25] show s large error bars for kse but not for D. Lane [36] finds good agreement between D values from IEDP and conductivity relaxation but finds no apparent relationship between the measured kse values. An important sample property is the characteristic length, Lc, which is defined by the quantity D/kse [62]. For sample geometries much larger than Lc, oxygen transport is limited by diffusion through the bulk while samples much smaller than Lc are kinetically limited by reactions occurring on the surface. This concept was explored in Lanes conductivity relaxation study [14] on bars (1.5 x 103 m thick) of LSCF, which were found to have a characteristic length of 103 m at 800 C; the characteristic length fo r LSM is much smaller and, from IEDP studies, appears to be 8 x 10-3 m at 900 C (2 x 103 m thick disk) [22]. Sample dimensions in these experiments are typically similar to or much larger than Lc, which means that the measurements are significantly aff ected by bulk transport kinetics. 2.5 Gas phase isotopic oxygen exchange One way to ensure that surface kinetics is the dominating process is to use powder samples. Isotopic exchange techniques have b een used since the 1940 s to study the surface 25


oxygen exchange rate of various oxides [27,63-67]. In these experi ments, the powder sample was confined in a closed system sometimes with a gas recirc ulating design, and allowed to equilibrate with normal oxygen at a chosen te mperature and oxygen partial pressure. Then, isotopically labeled oxygen gas was introduced and the gas phase composition was monitored for the three oxygen isotopologues (molecules that differ only in isotopic composition) using a mass spectrometer. The analysis in these early studi es was overly simplistic and often neglected key steps such as adsorption, which may in fact dominate these surface processes. It was found [68] that these earlier models were quite limited and a more rigorous approach to find the mechanism and rate limiting step of surf ace exchange is preferred. The main focus of this body of work is usi ng the more modern tec hnique of steady state isotopic transient kinetic analysis (SSITKA) [69] to determine kinetic information for cathode materials and using temperature programmed is otopic exchange to meas ure catalytic activity. With SSITKA, a plug flow reactor is used in plac e of the recirculating system, resulting in more sensitive and rapid gas composition detection, an d a constant feed stream composition that makes data analysis more straightforw ard. A mass spectrometer located immediately downstream from the reactor continuously monito rs the gas phase composition. The experiment is based on the rapid switching of the reactor feed line between streams containing different isotopes of oxygen. Ideally the only change i nduced by the switch is a step in oxygen labeling. Temperature, pressure, flow rate and oxygen pa rtial pressure should remain undisturbed. The following analysis is then honed for reversible re actions such as gas phase oxygen exchange with the solid. In contrast to previously mentione d techniques (IEDP, conductivity relaxation) which yield overall mass transport information (such as surface exchange coefficient and diffusion 26


constants), this method has the potential to re veal not only the mechanism of oxygen exchange, but also the kinetic rate constants wh ich determine the ove rall reaction rate. Another technique used in this dissertati on is called temperature programmed reaction, which quickly surveys a materials behavior at different temperatures. These experiments involve flowing gas of a known composition into th e reactor while increasing the temperature at a constant rate and monitoring the reactor effluent in real time. The relative catalytic activity of two different materials can be compared by determining the onset temperature for reaction, desorption or isotopic exchange. Because of the apparent similarity betw een the IEDP technique [22,37,44,55,56] and this work, it is important to first establish the concep tual similarities and differences between these two methods. Both techniques assu me that the interaction of 18O2 with the sample material is identical to that of 16O2, and that the flux of oxygen entering the material is equal to that leaving. It is not possible to study any ch arge transfer step(s) from eith er technique. The main difference is the SSITKA based method is a more catalysi s based approach and aims to determine the mechanism of oxygen incorporation in to the solid using chemical kine tics and the concept of rate coefficients whereas the IEDP method extracts the D and kse values ex situ by fitting the isotopic penetration depth profile in the solid to the Crank solution to the diffusion equation. As mentioned before, it is difficult to conclude what the kinetic mechanism is from kse. Using oxygen isotope 18O2 as a tracer to differentiate from the 16O found in the solid, it is possible to follow the oxygen exchange process in real time and elucidate the kinetic dependence on temperature and pO2 for these materials through temp erature programmed and isothermal exchange methods. By tracking the rise and fall of different oxygen species (seen as masses 32, 34 and 36 from the mass spectrometer), the up take and release of oxygen can be better 27


understood (Figure 2-4) Mass 32 represents 16O2 coming from the bulk of the material while mass 36 is the labeled oxygen 18O2 from the gas stream. Mass 34 represents the scrambled product 16O18O that results from heteroexchange [ 70] between the gas and solid phases. Because oxygen is a relatively stable molecu le, exchange between oxygen molecules in the gas phase (homoexchange) without interaction with those from th e solid is not significant at temperatures below 1000 C and will not be considered in the analysis [70]. Furthermore, verification that homoexchange did not occur in a blank reactor was done by flowing a mixture of 18O2 and 16O2 and monitoring for any changes while increasing the temperature from 50800 C. Figure 2-4. Oxygen exchange showi ng possible species formed [71] Each materials active temperature range is first identified from temperature programmed experiments, and then isothermal isotopic exchange is used to obtain kine tic parameters that can be compared to a theoretical model, discussed in the next section. From a simple two step mechanism of dissociative adsorption follo wed by incorporation, oxygen partial pressure dependencies, which are determined by the ra te limiting step, are extracted. The extracted 28


parameter values are used to reconstruct the time dependent gas behavior and the extent of exchange. 2.6 Proposed Mechanism of Oxygen Exchange The actual mechanism for oxygen exchange may be quite complex, involving a multitude of possible steps [72]. However, as a starting po int, a simple 2-step model (equation 2-8) is proposed [71]. Charge transfer is not considered a separate step in th is analysis because no electrical measurements were made. The first step is the dissociative adsorption of a molecule of oxygen on the surface. This is followed by the inco rporation of the adsorbed oxygen atom into the solid. The (g) indicates the gas phase, S indicates a surface site and Vo is a vacancy at the surface of the material. Thedr adsorbed or bulk species. as o bulk subs cripts indicate n a 2-8 The ks are the forward and backward reaction rate constants for each step. In this work, the measured quantity is the concentration of O2 (g) from the first step. Introducing an isotopic tracer into the system allows for the formation of three different oxygen species 16O2, 18O16O, and 18O2, which are produced on the surface according to the equations in 2-9. These equations represent the backward reacti the f on 2-8). on ofi rst step in the model (equati 2-9 From the assumptions that the surface and ga s are in equilibrium with each other, the composition of the gas phase oxygen species can be used to back derive the statistical distribution of oxygen isotope 16O or 18O adsorbed at the surface. This serves as the starting point 29


for much of the isotopic excha nge (both temperature programmed and isothermal) analysis to follow. Overall, the gas interacts with the oxygen adsorbed at the surface which then interacts with the oxygen in the sample la ttice. All three phases (gas, surface and bulk) can be modeled using this relatively simple two step process. All that remains is to link the measured qua ntities to the rates of production. Illustrated next is a schematic of the reacto r (figure 2-5) with the flow of feed and reaction products. The reaction is assumed to occur at feed conditions as the reactants are quickl y swept away. There is also no accumulation within the reactor since the residence time of the gas is fairly small. A constant flux, F, of 18O2 enters the reactor and reacts wi th the oxide sample to form 18O2,16O2 and 16O18O. The fluxes of the products, F, are monitored by the mass spectrometer. Figure 2-5. Schematic showing reactor feed and products [71] From mass balance (equation 2-10), the followi ng expressions for the production of the various oxygen species are derived. SA is the sample surface area. For the last two equations Accumulation = 0 = Fin-Fout + rate produced rate consumed 2-10 18 1 2 18 1 2 182 2 18 2 18][ ][O adsPkOkFF d t Od SAO O 2-11 ]][[ ][16 18 1 18161816ads adsOOkF d t OO SAOO 2-12 30


31 2 16 1 2 16][ ][2 16adsOkF d t Od SAO 2-13 (2-12 and 2-13), it is apparent that the quantity measured by th e mass spectrometer is directly proportional to the rate of production. Using this link between the measured and theoretical rate of production, a simulation of the exchange react ion is possible and used to deduce the most likely mechanism of isotopic exchange.


CHAPTER 3 TEMPERATURE PROGRAMMED EVALUATI ON OF CATALYTIC PROPERTIES 3.1 Introduction Temperature programmed reaction (TPR) is a ch aracterization technique used in the field of heterogeneous catalysis. The pr ocess involves flowing reactant gases through a bed of catalyst and measuring the composition of the effluent as the temperature is increased. Using TPR, the oxygen non-stoichiometry of (La0.8Sr0.2)0.98MnO3 (LSM) and La0.6Sr0.4Co0.2Fe0.8O3(LSCF) can be measured under different temperature and pO2 conditions, and compared to that in the literature as a means to validate this technique. A variation of TPR, temperature programmed desorption (TPD), is used for investigating the desorption of oxygen from the surface. First, adsorbates are adsorbed onto the catalyst by flowing the gas of interest th rough a cooling catalyst bed. On ce the starting temperature is reached, an inert gas such as nitrogen or helium is used for the feed stream to carry off any excess reactant gas [73]. Then th e temperature is linearly ramped and the desorbing species is detected with a mass spectrometer. TPD can yield information on the surface coverage and differentiate between differ ent binding sites. Another variation is temperature programmed isotopic exchange (TPX), which sheds light on the interaction between the oxyge n in the gas and solid phase. First the oxide catalyst is annealed in a stream containing normal 16O2 and cooled back down to room temperature to ensure that the oxygen in the solid is of norma l isotopic labeling. Then the oxygen in the feed stream is replaced with labeled oxygen (18O2). The different oxygen species (18O2, 18O16O, and 16O2) in the gas are then monitored as the reactor is heated [18]. The temperature of exchange onset is then an indicator of the catalytic activity. Depending on the evolution of the oxygen species relative to each other, differences in mechanism can be detected. 32


3.2 Experimental 3.2.1 Powder Materials (La0.8Sr0.2)0.98MnO3powder from Nextech (5.73 m2/g) and La0.6Sr0.4Co0.2Fe0.8O3powder from Praxair (6.53 m2/g) were used without additional modification. By testing each cathode material individually, in teraction between dissimilar materials (such as secondary phase formation between electrolyte and cathode) can be avoided. 3.2.2 Catalysis System The continuous flow system (Figure 3-1) consis ts of two separate ga s lines connected to a switching valve with one line exit ing to vent and the other le ading to a small quartz microreactor. Inside the reactor, gas flows through a be d containing a small quantity of powder sample supported by quartz frit (Figure 3-2). The reactor effluents flow to a quadrupole mass spectrometer (Extrel QMS), where the compositi on is analyzed and recorded. A temperature controller (Eurotherm) connected to a type-K thermocouple in the reactor controls furnace temperature. Switching Valve Stream 2: Oxygen 18 Vent Microreactor and Furnace Temperature Controller Stream 1: Oxygen 16 Vent Mass Spectrometer Data Acquisition Figure 3-1. Catalytic testing system comprised of two separate feed streams, a microreactor within a furnace and quadrupole mass spectrometer [71] 33


Figure 3-2. Quartz micro-reactor schematic [74] Each gas line contains multiple mass flow controllers connected to oxygen gas and research grade helium tanks (Airgas). All gas mi xtures were balanced in helium. Stream 1 contained oxygen isotope 16O2 while stream 2 contained oxygen isotope 18O2 (Cambridge Isotope, 95% pure) and a small amount of argon (Airgas, 0.1%) used as an inert tracer. Mass flow controllers (MKS & Alicat) and exhaust valves were adjusted so that the two lines were flow and pressure matched. After each experi ment, the mass spectrometer was calibrated by flowing gases of a known composition through a r oom temperature reactor. Flow rates were verified using a bubble-meter connected to on e of the exhausts. Masses 16, 18, 28, 32, 34, 36 and 40 were monitored during every run. Mass 28 corresponds to either N2 or CO and mass 40 represents the inert argon tracer. Because the upta ke and release of oxygen is repeatable in LSCF [19], this fact was used to calculate relative sensitivity factors (RSF) for mass 32, 34 and 36 from a series of TPR experiments. 34


3.2.3 Temperature Programmed Reaction, Desorption and Exchange The sample was loaded into the quartz r eactor and pretreated for 10 minutes at 650 C. After pretreatment, the reactor wa s cooled to room temperature in the pretreatment gases and then the reacting gases were flowed through the r eactor (Table 3-1). Once steady state conditions were reached (approximately 30 minutes), the temperature was ramped up to between 650 C and 800 C at a rate of 30 C/minute while the composition of the reactor effluent was recorded. For TPD, the sample was cooled in an oxygen atmos phere and heated in an inert helium stream. The TPR involved heating the sample in an oxygen containing feed stream, while the TPX is nearly identical to the TPR except the oxygen in th e feed stream was is otopically labeled. Table 3-1. Temperature programmed conditions [71] Temperature Programmed Experiment Sample Size (mg) Flow Rate (ml/min) Pretreatment Gas Reaction Gas Feed Stream O2 Content (ppm) Reaction (TPR) 15-LSCF 17-LSM 20 1% 16O2 16O2, He 2,500; 5,000; 10,000 Exchange (TPX) 40 25 1% 16O2 18O2, He 2,500 Desorption (TPD) 40 25 1% 16O2, 2,500ppm 18O2 He <10 3.3 Results and Discussion 3.3.1 Oxygen Non-Stoichiometry from TPR In order to determine the validity of th is technique, the oxyge n non-stoichiometry of LSM and LSCF was found using our quantitative mass spectrometry method and compared to published results. LSM was heated in 2,500 and 5,000 ppm oxygen. At no time did the oxygen signal increase noticeably beyond the baseline, indi cating that the LSM oxygen stoichiometry is constant throughout the temperature range 50-700 C. This is in agreement with results obtained by Kuo [38], which indicate that LSM has few vacancies under these conditions. 35


The oxygen vacancy concentration of La0.6Sr0.4Co0.2Fe0.8O3calculated from a series of TPRs under various oxygen partial pressures is shown in Figure 3-3. As expected, the oxide deviates from stoichiometry ( =0) as the temperature was increased or the oxygen partial pressure was decreased. (3) values at 800 C are similar to those interpolated from Hartleys study on oxygen non-stoichiometry [54], although the data from this study indi cates that LSCF is more sensitive to the oxygen partial pressure. Inte rpolated data from Hart ley appear as dots. Due to the smaller sample size in our study (15 mg), calibration errors and differences in detection method could have contributed to the discrepancy. The sample size was weighed out to yield a total surface area of approximately 0.1m2 for each material: 17 mg of LSM powder and15 mg of LSCF powder. 2.92 2.94 2.96 2.98 3 100200300400500600700800900 2,510 5,027 10,000 3-deltaTemperature (Celsius)Oxygen Concentration (parts per million) Data from Hartley (2000) Figure 3-3. La0.6Sr0.4Co0.2Fe0.8O3Oxygen non-stoichiometry (3) vs. temperature for various oxygen partial pressures. The circles at 800 C represent points interpolated from Hartlys work [54,71]. 36


3.3.2 Temperature Programmed Isotopic Exchange Heating the sample at a consta nt rate in isotopically en riched oxygen gas revealed the formation of 16O18O, as well as the exchange of gas phase oxygen (18O2) with that in the bulk (16Oo). Figure 3-4 shows the exchange behavior of LSM and LSCF. Onset of oxygen exchange in LSCF began around 200 C, while it did not begin until 350 C for LSM, indicating LSCF is more catalytically active. The isotopic exchange sample size is larger than the TPR sample size to maximize detection of isotopic exchange. 0 1000 2000 3000 4000 150 300 450 600Total OxygenConcentration (ppm)Temperature (Celsius)18O2 16O2 16O18OA 0 1000 2000 3000 4000 150 300 450 600Concentration (ppm)Temperature (Celsius)16O2 16O18OTotal Oxygen18O2B Figure 3-4. Temperature programme d exchange profile for A) LSM and B) LSCF [71] LSM shows extensive 16O18O formation and little bulk oxygen desorption up to 600 C, which suggests a more rapid dissociative adsorption step relative to the in corporation step. Much of the 18O from the gas phase 18O2 dissociates on the LSM surfa ce and, rather than becoming incorporated into the lattice, de sorbs back into the gas phase as 16O18O. This is consistent with LSMs behavior as an electronic conductor with few vacancies and low oxygen diffusivity. The decrease in 16O18O starting at 620 C suggests either a change in the reduction mechanism or the exhaustion of 16Oads species at that point. The total oxygen does increase slightly around 500 C, which was not observed in earlier TPR experiments performed with regular 16O2. This may be 37


due to calibration error or it could also indicate that the material is slightly sub-stoichiometric under testing conditions at higher temperature. Although some surface scrambling (formation of 16O18O) takes place on LSCF (Figure 24b), the majority of the gas phase oxygen (18O2) exchanges by switching pl aces with the original bulk oxygen. Overall, more oxygen was released than absorbed by the oxide, and the oxygen ejected was primarily from the bulk. Because th ere was relatively little surface mixing, this indicates that step 2 is overall very fast compared to step 1. These results are consistent with those obtained from Boukamps study [75] on La1-xSrxCoO3 based perovskites. Boukamp and coworkers also observed relatively little cross product formation in their sample while observing that the labeled oxygen exchanged with the solid by replacing the bulk oxygen 16O. 3.3.3 Temperature Programmed Desorpti on after Isotopic Exchange Running a TPD after the TPX on both materials revealed that the oxygen lattice was of mixed isotope. TPD spectra obtain ed after adsorption in regular 16O2 follow the same shape and magnitude of total oxygen desorbed. The TPD for LSM (Figure 3-5a) shows two notew orthy characteristics. First, the presence of 16O18O indicates that some exchange did occur in LSM, and sec ond, the amount of oxygen given off by the sample is very small compared to LSCF. This profile is similar to those presented in previous works [20,76], which also show the base of a desorption peak starting near500 C. Although the peaks at 175 C and 350 C were not seen in th e earlier studies [20,76], they are very small and can arise from differences in catalyst pretreatment. The profile for LSCF is shown in Figure 3-5b, with one view zoomed in to show contributions to the low temperature peak near 230 C. This peak was also seen at the same temperature in another study [19]. Overall, the majority of the desorbed oxygen consisted of the non-exchanged species 16O from the lattice. Upon closer inspection of the low temperature peak, 38


0 20 40 60 80 100 150300450600Concentration (ppm)Temperature (Celsius)Total Oxygen18O2 18O16O16O2 A 0 200 400 600 800 100200300400500600Concentration (ppm)Temperature (Celsius)Total Oxygen16O2 18O16O18O2 0 15 30 45 60 100150200250300350B Figure 3-5. Temperature programmed desorption af ter TPX for A) LSM B) LSCF, broken down by isotopic labeling [71] called the -peak [17,73], it is apparent that there may be an oxygen surface state at room temperature. It is likely that th e labeled species originates as oxyge n either in the surface layer or as an adsorbate and as the temperature increases more bulk oxygen is expelled from the lattice as the material equilibrates with the environment [73]. Because the 16O18O is twice the value of the 18O2, it is evidence that the low temperatur e surface state of oxygen is atomic. During pretreatment in the labeled gas, the isotopically labeled oxygen accumulates near the surface due to slow diffusion as the sample cools; this ex plains the corresponding low temperature peak in 18O2. It should be mentioned that the total oxygen TPD profile is reproducible regardless if it is performed with regular or labeled oxygen adsorbate. 3.4 Conclusion The oxygen non-stoichiometry of LSM and LSCF obtained by temperature programmed reaction is comparable to values reported in the literature. Using an isotopic tracer, it was shown that oxygen surface exchange occurs in LSCF above 200 C and in the electronic conductor LSM above 350 C. At all temperatures tested, LSCF is more catalytically active than LSM with regard to exchange. The amount of scrambled surface species formation is indicative of the nature of the 39


40 oxygen exchange reaction; LSM shows difficulties with incorporation while LSCF appears to be limited by the dissociativ e adsorption step. Desorption studies performed after the isotopic exchange experiment shows that there is a low temperature (100-300 C) peak for both LSM and LSCF. The release of mixed product is further proof that the exchange between the gas and solid oxygen sp ecies does occur. The magnitude of desorption is much smaller for LS M than with LSCF. It is thought that the high temperature desorption peak originates from the oxide bulk rather than from the desorption of oxygen adsorbates.


CHAPTER 4 ISOTHERMAL ISOTOPIC SWITCHING 4.1 Introduction Once the active temperature range for (La0.8Sr0.2)0.98MnO3 (LSM) and La0.6Sr0.4Co0.2Fe0.8O3(LSCF) are identified from temperature programmed experiments, isothermal isotopic exchange can then be performed to obtain an exchange profile, which is compared to the theoretical model. This method is also called the steady state isotopic transient kinetic analysis technique, described in detail by Shannon and Goodwin [69]. This technique involves feeding a gas mixture of 16O2 of a known concentration into a reactor containing the cathode pow der at constant temperature a nd abruptly switching the gas feed stream to one containing the same concentration of 18O2. The oxygen isotope distribution in the reactor effluent versus time is then mon itored by mass spectrometry to generate exchange profiles. Ideally the only change induced by the switch is a step in oxygen labeling. Temperature, pressure, flow rate and oxygen partial pressure should remain undisturbed. It is generally believed that the surface reaction process can be broken down into several steps: first gas diffuses to the surface, then it adsorbs and reacts on the surface, and finally is followed by desorption away from the surface. The mechanism of oxygen reduction on (La0.8Sr0.2)0.98MnO3 (LSM) and La0.6Sr0.4Co0.2Fe0.8O3(LSCF) are thought to be different. Since LSM is an electronic conductor with poor ionic conductivity, it is thought that the oxygen reduction reaction is limited to the triple phase boundary where gas, electrode and electrolyte meet. With LSCF, which is a mixed ionic elec tronic conductor, diffusion through the bulk oxide is thought to be facile and as a result, the whole surface is active for oxygen reduction. From a simple two step mechanism, the oxygen partial pressure dependencies can be extracted, which are determined by the rate limiting step. Also, the kinetic parameters are 41


extracted and used to reconstruc t the time dependent gas behavior and the extent of exchange. Many parallels can be made between this studys modeling approach and an earlier Monte Carlo simulation [77] of the gas phase composition applied to fluorite structured oxides. In addition to simulating the evolution of the gas phase oxygen composition with time, this work also models the exchange behavior of the bulk from the measured gas phase reaction. 4.2 Theory Starting with the 2 step mechanism mentioned in Chapter 2, equation 2-8, the overall rate is defined by the extent of oxygen incorporation into the solid (second st ep). Although the total oxygen concentration is nearly cons tant during an isotopic switch expe riment, there is an isotopic driving force and eventually labeled oxygen 18O from the gas phase replaces lattice 16O. Therefore it is possible to track the overall reaction rate as d[18Oo]/dt using isotope mass balance. Because the relative differences in mass and bonding energy are small amongst diatomic oxygen species, isotopic effects are ignored for the sake of simplicity [69]. In the limit that one of the tw o steps is rate determining, tw o different behaviors can be predicted. If one step is assumed to be in qua si-equilibrium with respect to the labeled oxygen concentration, a simplified expression can be written using the e quilibrium constants K1=k1/k-1 and K2=k2/k-2 to yield the initia l oxygen dependence of the overall rate of 18O incorporation. In the case of step 2 being in quasi-equilibrium, at time t 0, [18Oads]=[18Oo] 0. 2 18]][[2 ][ ][ ][][12 0 18 2 18 O o t ads o oPSVkK dt Od S V K dt Od rate 4-1 However this is not the case when step 1 is in quasi-isotopic equilibrium. 5.0 5.0 1 2 182 18]][[ ][OPKSVk dt Od rateo o 4-2 42


The initial reaction orde r with respect to O2 is then predicted to be either 1 or 0.5 depending on whether dissociative adsorption or incor poration, respectively, is rate limiting. In this study, the measured quant ity is the concentration of O2 (g) from the first step. Introducing an isotopic tracer into the system allows for the formation of three different oxygen species 16O2, 18O16O, and 18O2, which are produced on the surface according to equations 4-3 through 4-5 (the extra in equation 4-4 arises from the reaction multiplicity). These equations represent the backward reaction of the first step in the model (equation 2-8). Equation 4-6 shows the normalization relationship for oxygen adatoms of a particular is otopic labeling, assuming that the total oxygen adsorbate concentration is c onstant during the switc h due to steady state conditions. 4-3 4-4 4-5 4-6 If the oxygen on the surface and in the bulk is considered to be thoroughly mixed, and assuming that the total adsorbed and bulk oxygen concentration remains constant throughout the exchange, then the oxygen isotope distribution on th e surface and in the soli d can be represented in terms of fractions of one isotope (either 16 or 18). This si mplification is also discussed in previous work [71]. The convention used here will focus on the fraction of adsorbed oxygen species on the material surface (16fads) and in the lattice (16fbulk) which are of normal isotopic labeling (16O) because this is the isotope used for calibrating the mass spectrometer signal. Using equation 4-3, the 16fads at a given time is obtai ned from the instantaneous 16O2 gas phase signal, 43


and serves as the starting point for much of the an alysis. This allows us to use the measured gas phase composition during the exchange to predict the real time isotope di stribution on and in the sample. 16fbulk can be obtained by integrating the amount of 16O exchanged (from 16O2 and 16O18O) versus time and then normalizing to the sample size. The time dependence of the system can be mode led with a system of nonlinear differential equations describing the isotopic composition of the surface and bulk (Figure 2-4), which depend on kinetic parameters and a corr ection to account for the purity of the isotopically labeled feed stream. The gas purity is taken into account by the following: 18O2 + 16O2 + 16O 18O = Total O2 4-7 Where is the fraction of 18O2, is the fraction of 16O2 and is the fraction of 16O 18O in the gas phase. Inserting these corrections in to the derivation (see Appendix A) [71], the followingte nbine. sysm of equatios is otad 4-8 ) 4-9 From here, the following parameters are defined: A = k-1[Oads]; B = k2[V0]; C = [Oads]/[Obulk]. The parameters A and B are representativ e of the kinetics that describe the oxygen exchange reaction and are the eff ective rate constants; the smalle r of the two rate constant will indicate if the reaction is dissociative adsorption limited (smaller A) or incorporation limited (smaller B). C is the ratio of adsorbed oxygen to oxygen in the bulk. Note that extracting the parameters A, B and C from the experimental data will require solving the system of nonlinear equations 4-8 and 4-9 using numerical methods. 44


This method of evaluating the isotopic composition of the surface and bulk, as predicted by the measured gas composition, will a llow the validity of this mode l, along with its assumptions, to be thoroughly tested. 4.3 Experimental The materials (LSM and LSCF) and experiment al setup are the same as described in sections 3.2.1 and 3.2.2. Previously, the sample was h eated at a constant ra te while the feed gas was held constant. In this chapte r, the temperature is held consta nt and the gas stream is varied. 4.3.1 Isothermal Isotopic Switching Switching experiments consist of isothermally and isobarically (with respect to total oxygen pressure) switching the gas f eed stream into the reactor fr om a line containing a mix of normal 16O2 and helium to a secondary line containing 18O2 (Cambridge Isotope, 97% enriched) and helium mixed to the same concentration. A small amount of argon (1,000 ppm) is included in this second line as an inert tracer. The change in isotope concentrati on is recorded using mass spectrometry (Extrel). The common gases are metered with MKS mass flow controllers while an Alicat Scientific mass flow controller is used with the isotopically labeled oxygen. Flow rates [20 standard cubic centimeters pe r minute (sccm) total unless stated otherwise] were calibrated using a bubblemeter and mass spectrometer signals were calibrated by flowing known concentrations of oxygen balanced in helium thr ough the reactor. A typical room temperature gas switch profile, without reaction, is shown in Figure 4-1. Within 20 seconds, switch has reached new equilibrium with the 18O2 enriched oxygen feed stream, in terms of both 18O2 and the inert Argon tracer. After the sample was loaded into the reactor, the temperature and 16O2 partial pressure was adjusted to the desired values in f eed stream 1. In feed stream 2, the 18O2 partial pressure, flow rate (20ml/min) and pressure (15-18 PSIA) we re adjusted to match the conditions in the 16O2 45


0 5000 10000 15000 20000 0 200 400 600 800 1000 1200 00.511.522.533.5Concentration (ppm)Time (minutes) 16O2 18O2 Argon Tracer 18O16OTotal Oxygen Switch Figure 4-1. Oxygen isotopic sw itch profile through a r oom temperature reactor line. Once steady state was reached, a rapid switching of the feed lines quickly changed the isotopic labeling of the reacting gases and the reaction was allowed to proceed for up to 5 minutes or whenever steady state was reached, whic hever came first. Different combinations of temperature and oxygen concentration were systema tically tested for each material. In order to extract the pO2 reaction order at 600 C and 800 C, the temperature was held constant while the oxygen concentration was changed from run to run. To obtain the activ ation energies, the 18O2 concentration was held at 20,000 parts per million (ppm) as the temperature changed from run to run. Because the reaction was assumed to be lim ited by surface kinetics, the sample size was normalized by surface area, which was approximately 0.1 m2 for each material (17 mg of LSM powder and15 mg of LSCF powder). A full exchange spectrum was taken at 650 C in 10,000ppm 18O2 because the TPX spectrum shows extensive activity at this temperature for both materials. A full exchange spectrum entailed waiting until the mass spectrometer signals for masses 32, 34 and 36 have reached steady st ate and no more reaction was observed. Additional isotopic switching experiments i nvolved varying the temperature or oxygen concentration conditions of the test for the full duration of the exchange. Before switching tests, fresh powder was heated to 800 C in approximately 20% 16O2 for an hour to remove 46


contaminants. After testing, the sample is again heated to 800 C in 20% 16O2 for an hour to reset the lattice with 16O. Tests on both materials were perf ormed at 700 C while varying the partial pressure of oxygen (pO2) from run to run (0.5%, 1% and 2%). Then the pO2 was kept at 2% while the temperature was held at 600, 700 and 800 C. The oxygen partial pressures were chosen to avoid saturating the mass spectrometer detector. Also, the mass flow controller metering the isotopic oxygen has a maximum flow rate of 0.5 sccm, allowing for precise control but limiting the range of pO2 achievable. Upon finishing the set of switc hing experiments, it was noted that LSCF appeared to be gas diffusion limited due to its sample temperature insensitivity. In other words, the surface reaction is occurring more rapidly than the gas phase diffusion of reactants to the LSCF surface. This effect is not observed with LSM because its surface reaction occurs more slowly than diffusion and is thus rate limiting. Because it is the surface kinetics which are of interest, it is necessary to either slow down the reaction on LSCF or decrease the boundary layer thickness such that gas phase diffusion is no longer limiting. The first approach taken was to double the flow rate to 40 sccm and then tests on LSCF were conducted at 600, 700 and 800 C in 1% O2. Then the temperature was decreased to 300, 400 and 500 C while holding the flow rate at 40 sccm and 1% O2. Some degradation was noticed so a new sample was used for the 40 sccm tests. Also, the sample size of LSCF was varied from between 5, 10 and 15 mg and tested under 40 sccm 1%O2 at 500 C. 4.4 Results and Discussion 4.4.1 (La0.8Sr0.2)0.98MnO3 Figure 4-2 shows the evolution of various oxyge n species with time immediately after the switch in feed stream oxygen labeling (16O 18O) for LSM at 650 and 800 C under 10,000 ppm 47


of labeled oxygen. Time=0 was taken to be the time at which the inert argon signal was first detected and any data before that point was disregarded (Figure 4-3). 0 2000 4000 6000 8000 10000 12000 02468Concentration (ppm)Inert Argon Tracer16O18OMinutes10 18O2 16O2Total Oxygen A 0 2000 4000 6000 8000 10000 010203040Minutes16O18O16O2Total Oxygen18O2Concentration (ppm)B50 Figure 4-2. (La0.8Sr0.2)0.98MnO3 isothermal isotopic exchange under 1% O2 at A) 650 C and B) 800 C [71] Comparing the time scale of the reaction in the profile from the switch in 1% O2 (shown in Figure 4-2 for LSM) to that in Figure 4-1, the decay in the 16O2 signal cannot be due to the switch itself and is attributed to the heterogeneous exchange with the 16O containing sample. 48


These profiles show that the total oxygen concentrati on during the experiment is very stable, with the exception of the discontinuity near th e switch (due to pressure imbalance between the two lines). Total oxygen was reasonably constant th roughout the experiment; the switch does introduce an initial discontinuity because the lines were closely, but not pe rfectly, flow, pressure and concentration balanced. 16O2 falls off monotonically while 18O2 increases with time as expected in both Figure 4-2a and b. 16O18O increases and reaches a ma ximum before gradually Concentration Time Experiment Start time = 0 Actual Switch Figure 4-3. Argon Tracer Step-Profile [71] dropping off. If exchange of any type had not occurred, there would be no formation of 16O18O and the time to switch isotopic labeling from 16O2 to 18O2 would be comparable to the argon switch time. Although the y-axis of the figure is labeled co ncentration, it can also be considered the instantaneous rate of producti on of the three oxygen isotopologues. It is possible to know if exchange with the solid is actually occurring or not by comparing the amount of oxygen exchanged with the amount of oxygen present in th e original sample. If exchange with the bulk occurs, the conversion will approach 100% as th e experiment proceeds. If only the adsorbed oxygen species were exchanged, then this reactio n will occur very quickly and the amount of oxygen exchanged will fall short of the amount need ed to convert the oxy gen in the lattice to 49


100% labeled oxygen. Therefore, to get the amount of oxygen exchanged, it is necessary to integrate the area under the 16O2 curve plus half of the area under the 16O18O curve. To put into perspective the amount of time it takes for a full exchange to occur, under typical experimental conditions for our sample mass, it takes at least 20 minutes to flow enough labeled oxygen through the system to replace each oxygen in the sample lattice. The prob ability of replacing a 16O in the lattice with an 18O from the gas phase decreases with time which leads to long exchange times if 100% conversion is desired. Also, close to 100% conversion is not possible since our gas is enriched to only 97% 18O (95% 18O2). Although the TPX (Figure 3-4a) would s uggest that LSM is reactive at 350 C, that is merely the lower temperature limit for this material s reactivity. The sample size is larger and the oxygen partial pressure is lower in the TPX speci fically so that the r eactivity limit can be studied. Significant exchange from isothermal switching was not observed until the temperature increased past 600 C. At first glance at the time scales in Figure 4-2, it may appear that LSM reacts much faster at 650 C than at 800 C, however this is not really the case. The rapid production of the asymmetric molecule is primaril y due to mixing with surface species. There is much more oxygen in the lattice that has not come out yet and it would take an extremely long time for the material to fully exchange. After increasing the temperature to 800 C, LSM shows characteristics of both surface and bulk exchange (Figure 4-2b) and subsequently takes much longer to approach steady state. If the initial oxygen partial pr essure dependence of the overall reaction is assumed to follow a power law ( ), it can be written in a linea rized form (equation 4-10). Slopes 2 18ln ln) ][ ln(18 O effective oPmk dt Od 4-10 50


were calculated from the data at t = 15 sec onds. The slope taken any earlier would incorporate switching nonidealities and any later would invalidate the assumption that the 18O content in the surface and solid is zero. The pO2 dependence for LSM at 600 and 800 C is shown in Figure 4-4. The switching nonideality is expected to cause gr eater error in the lower temperature LSM runs due to the more rapid attainment of steady state. The initial reaction orde r with respect to oxygen partial pressure for LSM is 0.67 at 600 C and 0.82 at 800 C, indicating that the rate determining step in LSM changes with temperature. -12.6 -11.9 -11.2 -10.5 -6 -5.25 -4.5 -3.75ln(pO2/atm)ln(d[18Oo]/dt) (moles/min)600oC 800oC 0.82 0.67 Figure 4-4. Initial rates of exchange vs. oxyge n partial pressure for LSM at 600 and 800 C [71] The change in pO2 dependence between 600 and 800 C signifies that a change in the mechanism occurs somewhere between those two te mperatures. To investigate further, a series of isothermal exchange experiments wa s conducted under 20,000 ppm oxygen at various temperatures. The first series of experiment s investigated the beha vior of LSM over the temperature range 600-800 C. By plotting the overall rate an d temperature (Figure 4-5) on an 51


Arrhenius-like plot, the apparent activation energy was extracted from the different slopes. There was a transition region between 650 C and 700 C and a second series was conducted spaced 15 C apart between 650 C and 710 C (Figure 4-5 inset). The higher temperature slope corresponded to an activation en ergy of 0.90 eV while the lower temperature slope indicated a much lower activation energy of 0.41 eV. This re inforces the hypothesis that only adsorbed surface species were exchanging at temperatures lower than 650 C, since exchanging surface -13 -12 -11Ea=0.41 eV E a =0.90 eV800C 750C 700C 650C 600C 1.02 1.04 1.06 1.08 650C 665C 680C 695C 710C -11.7 -11.9 -12.11.15 0.9 0.95 11.05 K-11/T x 1031.1 Figure 4-5. Arrhenius-like plot showing act ivation energy of oxygen exchange on LSM under 20,000 ppm 18O2 [71] species is a lower energy process. In order fo r surface adsorbate exchange to occur (step 1), oxygen only needs to adsorb and dissociate ont o the surface; for actua l exchange with the oxygen in the lattice to occur, additional energy n eeds to be spent on processes such as vacancy diffusion and creation. The cause of this change in mechanism may be a phase transition [78]. Previously reported values for the activation en ergy (from IEDP surface exchange coefficients) 52


in LSM was significantly higher: 1.3 3 eV at temperatures over 700 C [79]. This discrepancy may arise from differences in pO2 testing conditions if there is a different mechanism (100% O2 vs. 2% in this study) or sample configuration (as mentioned in the introduction, a compact body ma be less sensitive to surface reactions compared to a powder). 4.4.2 La0.6Sr0.4Co0.2Fe0.8O3y ended isothermal exchange behavior for LSCF at 650 C and 10,000 f the sample surface, shows Figure 4-6 shows the ext ppm oxygen concentration. Alt hough Figure 4-6 is data from a specific combination o temperature and oxygen partial pressure, the trends remain the same for most temperatures and oxygen concentrations tested. At 400 C exchange occurs but not at 200 C, which is consistent with the TPX results (Figure 3-4b). At 200 C, the TPX shows that the exchange is rather slow and would not react fast enough to be seen in the switching experiments. The lag in production of 16O18O, which is assumed to occur only on that the labeled oxygen does not displace th e original surface specie s right away, but must gradually accumulate. This is consistent with a slow dissociation step a nd fast incorporation. 0 2000 4000 6000 8000 10000 12000 020406080Total Oxygen18O2 16O18OMinutesConcentration (ppm)16O2Inert Argon Tracer Figure 4-6. Isothermal exchange of LSCF at 650 C and 10,000 ppm O2 [71] 53


Applying equation 4-10 to the series of LSCF isotopic exchange, the curve fit shows that the slope (Figure 4-7) is 0.97, i ndicating a nearly first order dependence on the isotopic oxygen partial pressure at 600 C. This result is consistent with the derived expression (equation 4-1) for dissociative adsorption limited kinetics. The slope at 800 C is 0.91; the small decrease in pO2 -12.5 -12 -11.5 -11 -10.5 -6-5.5-5-4.5-4ln(d[18Oo]/dt) (moles/min)ln(pO2/atm) 0.97 Figure 4-7. Initial rates of exchange vs oxygen partial pressure for LSCF at 600 C [71] dependence indicates that the reac tion is slightly less limited by dissociative adsorption relative to incorporation. The slope is an indication of the mechanism only. The onset temperature of oxygen isotope exchange and the amount of time it takes for a material to exchange a certain amount of oxygen are better indicators of a materials catalytic activity. Figure 4-8 shows the activation energy fo r exchange on LSCF between 400 and 700 C. The activation energies for LSCF are mu ch lower than those for LSM. Above 475 C, the activation energy is 0.10 eV, whic h is very low considering kbT at 700 C is about 0.08 eV. Below 475 C, it increases to 0.29 eV. From the LSCF TPX (Figure 3-4), below 450 C, LSCF is unable to react with all the 18O2 in the environment. The increa se in activation energy below 54


475 C is consistent with the results from the TPX and is thought to occur due to a decrease in vacancy concentration (see trend in Figure 3-3). An activation energy of 1.09 eV wa s found from IEDP data above 650 C [80]. Again, the IEDP value is much higher than the activation en ergy reported here, but both sets of data are consistent with LSM having the higher activa tion energy. It is unknown if the reason for the -11.6 -11.2 -10.8 K-1700Cln(rate) (d[mol]/d[min])600C 500C 450C 400CEa=0.29 eV Ea=0.10 eV Figure 4-8. Arrhenius plot showing activati on energy of oxygen exchange on LSCF under 20,000 ppm 18O2 [71] large discrepancy is due to trying to compare two fundamentally different rate constants (kse vs. k1, k2, etc) or to the different experimental parameters. Original IEDP data independent from the wo rk of Benson et al. [81] and Benson [82] could not be found to further compare the activat ion energy for this particular composition. kse values from conductivity relaxati on indicate the apparent activati on energy is 1.88 eV [15], but this is quite close to the activation energy of 1.75.16 eV for D found by the same method [83]. This, in addition to the observation of an empirical correlation between D and kse [79], indicates 55


that there may be some difficulty in separating surface and bulk transport processes from each other with the recently me ntioned techniques [84]. 4.4.3 Trends in LSM and LSCF Profiles from the switch at 800 C in 20 sccm of 2% O2 are shown in Figure 4-9 for LSM and LSCF. Compared to the time scale in the profile from the switch at room temperature 0 5000 10000 15000 20000 010203040506Concentration (ppm)Minutes16O2 Total Oxygen 18O2 18O16OArgon Tracer 0 A 0 5000 10000 15000 20000 010203040506Concentration (ppm)Minutes18O2 18O16OTotal Oxygen 16O2 Argon Tracer 0 B Figure 4-9. Isothermal isotopic switch profile at 800 C in 20sccm 2% O2 for A) LSM and B) LSCF 56


(shown in Figure 4-1), the decay in the 16O2 signal cannot be due to the switch itself and is attributed to the heterogeneous exchange with the 16O containing sample. These profiles show that the total oxygen concentration during the experiment is very st able, with the exception of the discontinuity near the switch (due to pressure imbalance between the two lines). The argon tracer in the effluent also maintains a constant concen tration during the exchange As discussed in the theory section, the time dependent con centration of the gaseous oxygen species (18O2, 16O18O and 16O2) reflects changes in the oxygen isotope distribution at the surf ace. The analysis of these curves can be approached in a variety of wa ys. The main assumptions made are that any 16O detected above the gas phase baseline from either 16O2 or 16O18O species originates from the sample and that the total oxygen c ontent of the solid is constant By integrating the amount of 16O exchanged, as shown in Figure 4-10, and comp aring it with the amount expected in the stoichiometric sample, the percentage of the lattice (1-16fbulk) which has been exchanged versus time can be obtained, as shown in Figure 411. Under current testing conditions, LSM is Figure 4-10. Integrat ion of amount of 16O from the gas stream involves counting the 16O2 signal twice and the 16O18O signal once 57


insensitive to changes in pO2 (Figure 4-11a) while LSCF is te mperature insensitive (Figure 411d). Because the rate of exchange is independent of the oxygen partial pressure (a reactant in step 1) and strongly dependent on the temperature of the solid (Figure 4-11c), LSM must be limited by the incorporation step. With regard to LSCF, its behavior is strongly dependent on the oxygen partial pressure (Figure 4-11b), indicating th at dissociative adsorption is the rate limiting step. Oxygen is a stable molecule and its dissociation is expected to be highly activated [25,29]. However, the observed temperature insensitive behavior (Figure 4-11d) is strong evidence that the rate limiting step is actually gas phase diffusion of reactants to the sample surface. 0 0.2 0.4 0.6 0.8 1 010203040506 0A LSM 700 C 0 0.2 0.4 0.6 0.8 1 01020304050618O Conversion FractionMinutes0.5% O21% O22% O2B LSCF 700 C 0.5% O21% O2 2% O218O Conversion FractionMinutes 0 0 0.2 0.4 0.6 0.8 1 010203040506 0C LSM 2% O218O Conversion FractionMinutes600 C 700 C 800 C 0 0.2 0.4 0.6 0.8 1 010203040506 0 600 C 700 C 800 C18O Conversion FractionMinutesD LSCF 2% O2 Figure 4-11. Fraction of oxygen in solid converted from 16O to 18O vs. time for A) LSM 700 C B) LSCF 700 C C) LSM in 2% O2 D) LSCF in 2% O2 58


Previously, it was noted th at LSM has an initial pO2 dependence of between 0.67 and 0.82, which can also be seen near t = 0 in Figure 4-11 a. It needs to be emphasized that the overall pO2 -dependence is what is being discussed here, and it appears to be independent of changes in pO2. A first order dependence of the total reaction rate on pO2, for LSCF, was observed earlier and was initially interpreted as an indication of dissociative adsorption being the rate limiting step [71], but the strong dependence on oxygen concentr ation (Figure 4-11b) can also arise from gas diffusion limitations since that is also a first order process according to Ficks first law. The concept of surface reaction and gas phase diffusion limitation is discussed by Fogler in his textbook [85], where the react ion rate is written to incl ude both gas phase diffusion of reactants and the surface reaction. Using Frsslings correl ation for mass transfer around a spherical particle, the reaction rate (rA) is related to the free stream velocity (U) and particle size (dp), and results in the relations hip shown in Figure 4-12. When determining kinetic information -rA(U/dp)1/2Reaction Limited Diffusion LImited Figure 4-12. Reaction rate versus square root of free stream velocity divided by pellet size; illustrates diffusion limited and reaction limited operating regimes about a surface reaction, the system should be oper ating in the reaction limited regime and not in the diffusion limited region. This can be achieved by reducing the particle size (not an option in this study), reducing the reaction rate by reducing temperature, or increasing space velocity. To move LSCF into the reaction li mited region, the flow rate was doubled to 40 sccm and then the 59


isotopic switching experiments were performed again. The range of temperatures examined was extended down to 300 C. Present equipment lim itations prevent testing under conditions with higher flow rates and concentrations. Figure 4-13a shows the normalized convers ion of LSCF under 40 sccm of 1% O2 and for 20 sccm at 1% O2 and 700 C. The sample size for the 40 sc cm tests is believed to be slightly larger due to weighing errors and this is take n into account by adjusti ng the amount of oxygen to which the sample is normalized. The assumption for the 20 sccm set is that each sample contains approximately 2.17 x 10-4 moles of oxygen while the 40 sccm sample contains, and is normalized to, 2.20 x 10-4 moles of oxygen. This is calculated by integrating up to the point at which the number of moles converted reaches a plateau (see Figure 4-11d for example). Two 0 0.2 0.4 0.6 0.8 1 051015202530 600 C & 40 sccm 700 C & 40 sccm 800 C & 40 sccm 700 C & 20sccmMinutes18O Conversion FractionA 0 0.2 0.4 0.6 0.8 1 051015202530 300 C 400 C 500 C 800 C18O Conversion FractionMinutesB Figure 4-13. Conversion of LSCF in 1% O2 A) at various temperatures and 2 flow rates and B) at 40 sccm flow rate and lower temperature range observations indicate that the system is not yet operating completely in the surface reaction limited regime even under 40 sccm flow rate: (1) the rate of reaction is dependent upon the flow rate and (2) the weak dependen ce on temperature again indicate s gas diffusion to be the limiting step. The isotope exchange was performed at lower temperatures to slow down the reaction rate (Figure 4-13b) and it becomes evident that th e reaction rate begins to show a temperature dependence between 300 and 500 C. 60


Gas phase profiles from isotopic exchange (1% O2, 40 sccm) at 300, 400 and 500 C are shown in Figure 4-14. No reaction is detected at 300 C ; as was also observed at room temperature (Figure 4-1). This is contrary to the TPX prof ile (Figure 3-4) which doe s show some exchange occurring around that temperature. The explanation for this is that the sample size is larger and the pO2 is smaller for more sensitivity in Figu re 3-4. It is possibl e with the higher pO2 during the isothermal isotopic switching, that the reaction rate is not detectab le. The exchange at 500 C is identical (within error) to that at 800 C (not shown), and the absence of 18O2 in the effluent at the beginning of the switch is evidence that all of the feed stream molecules are impinging upon the sample surface and reacting (100% conversio n). The profile obtained at 400 C possesses characteristics similar to that obtained at both 300 and 500 C. The discontinuous jump at the beginning of the switch indicates that some of the feed gas does not react on the sample surface. Assuming that the number of activ e sites remains the same regardless of temperature, it can be inferred that the turnover frequency for each site is reduced by half since only 50% conversion is achieved. By subtracting the non-exchanged baseline from the 18O2 signal and halving the scale of the y-axis, the resu lting profile (Figure 4-14b inset) once again resembles the behavior observed at higher temperatures. 16O18O reaches a maximum when the fractions of 18O and 16O on the surface are equal (at 16fads = 50%). The effect of varying the LSCF sample size while holding th e temperature, pO2, and flow rate constants are displayed in Figure 4-15. As e xpected, the conversion with time is more rapid for smaller samples since there is less oxygen for exchange. Residence time of the gas through the catalyst bed is estimated (from flow rate a nd LSCF density of 6.39 g/ml) to be at least 3.16 x 10-7, 6.10 x 10-7, and 1.11 x 10-6 seconds for corresponding masses of 4.8, 9.4 and 17.1 mg of LSCF. These masses were back calculated from the amount of oxygen exchanged; due to 61


weighing errors they are very close but not exac tly at the intended sample sizes of 5, 10 and 15 mg. 0 2500 5000 7500 10000 051015202530Concentration (ppm)Time (minutes) 18O2Total Oxygen16O2 Argon Tracer16O18O A 300 C 0 2500 5000 7500 10000 051015202530Concentration (ppm)Time (minutes) 18O2Total Oxygen16O2 Argon Tracer16O18OB 400 C 0 2500 5000 01 02 03 0Concentration (ppm)Time (minutes)18O2 16O2 16O18O 0 2500 5000 7500 10000 051015202530Concentration (ppm)Time (minutes) 18O2Total Oxygen16O2Argon Tracer16O18OC 500 CSwitch 62


Figure 4-14. Switching profiles for LSCF in 1% O2 & 40 sccm flow rate at A) 300 C B) 400 C [insert shows excess reactant (18O2) removed] and C) 500 C 0 0.2 0.4 0.6 0.8 1 05101520253018O Conversion FractionTime17.1 mg 9.4 mg 4.8 mg Figure 4-15. Conversion of LSCF vs. Time under 40 sccm of 1% O2 at 500 C as a function of sample size Degradation of LSCF While testing for repeatability, it was noticed th at a sample of LSCF began displaying signs of degradation. This is shown for an exchange pe rformed at 500 C (Figure 4-16). The sample of LSCF which had been cycled many times (~20) showed a discontinuous jump in the initial 18O2 and 16O2 signal, much like that observed at 400 C in Figure 4-14b, whereas this was not observed in the fresh sample. The maximum 16O18O was smaller in the degraded sample, also indicating that a significant num ber of oxygen molecules passed through the catalyst bed without participating in exchange. Becau se the surface is thought to be more active and more defective than the bulk [86], it is not surprising that the powder sample degraded relatively quickly. Also note that the degradation manifests as a reduction in apparent turnover frequency and not as a change in mechanism. This is a strong indi cation that there are r eactive sites on the surface 63


which have yet to be identified and which become blocked as segregation and other degradation mechanisms start to occur. If a standard is es tablished, then it would be possible to use isotopic exchange to determine if degradation had occurred in a batch of powder. 0 2000 4000 6000 8000 10000 02468Concentration (ppm)Time (minutes) 18O2 16O2 18O16O Figure 4-16. Exchange (1% O2, 40 sccm flow rate, 500 C) with fresh LSCF (dashed) and LSCF after many cycles (solid) 4.4.4 Gas Phase Behavior Simulation The behavior predicted by the model for di ssociative adsorption limited and incorporation limited scenarios is presented in Figure 4-17 for three values of the surf ace to bulk oxygen ratio C. As discussed in section 4.2, the effective rate constant characteristic of dissociative adsorption is A, for incorporation it is B. Therefore, by adjusting the relative values of A and B, the behavior of the two limiting cases can be explored. For simulating dissociative adsorption limited reactions (Figure 4-17a), A is set to be two orders of magnitude smaller than B. With incorporation limited reactions (Figure 4-17b), the opposite is true and B is much smaller than A. Open symbols represent the surface 16O fraction (equation 4-8) while solid symbols correspond to the bulk 16O fraction (equation 4-9). Figure 4-17 a shows that the surface and bulk isotope compositions are very similar at any given time for dissociative adsorption limited reactions. This is expected because the first step (of equation 2-8) is ra te limiting and the second 64


step, incorporation, is rapid enough to keep up with the firs t step. When the relative amount of adsorbed oxygen is small (meaning small C), the reaction proceeds very slowly. The larger the 0 0.2 0.4 0.6 0.8 1 0102030405060Dissociative Adsorption Limited A = 0.01 B = 1Fraction of 16OTime (minutes)C = 0.01 C = 1 C = 10 A 0 0.2 0.4 0.6 0.8 1 0102030405060Incorporation Limited A = 1 B = 0.01Fraction of 16OTime (minutes)C = 0.01 C = 1 C = 10SurfaceB Figure 4-17. Numerical solution fo r A) Dissociative adsorption limited case with A = 0.01 and B = 1 while varying C B) Incorpor ation limited case with A = 1 and B = 0.01 while varying C. Open symbols represent surface behavior (16fads) and solid symbols represent bulk behavior (16fbulk) value of C, the more rapidly the reaction pr oceeds (toward zero in the case of exchanging 16O for 18O in the sample). In Figure 4-17b, this behavior is not observed; instead, there is a dramatic difference between the isotope composition of th e surface and bulk. The su rface appears to react in the same manner regardless of the value of C. The bulk rate of exchange, however, increases with C. From a fundamental perspec tive, the meaning of C is the relative amount of adsorbed oxygen to oxygen in the bulk. If there is very little oxygen on the surface to participate in the incorporation step, then the reaction rate (=k2[Oads][Vo]) will be very slow, all 65


other factors remaining equal. Overall, the in corporation limited reaction will appear, based on the surface isotope composition, to have obtained st eady state more rapidly than the dissociative adsorption limited case. This incorporation limited surface behavior can be misleading, and requires comprehensive analysis si nce it is the behavior of the bulk which determines the real extent of reaction. To model the behavior of LSM under 2% O2 at 800 C in 20 sccm flow rate (Figure 4-9a), a Matlab m-file (See Appendix B for code) was wr itten to extract the parameters A, B and C from the time dependent behavior of 16fads obtained from the experi mental data via equation 4-3. Only the first 30 minutes worth of data was fitted in order to reduce computation time. First, an initial guess for A, B and C is used to generate a numerical solution to equations 4-8 and 4-9 using ODE45, and then the program a pplies the Gauss-Newton algorithm for least squares fitting of the numerical solution to the 16fads data and recalculates the parameters. This process is iterated (150 times to within a tolerance of 5 x 10-5 %) to ensure the solution converges (Figure 4-18). A good initial guess for A, B and C is required for stability. 0.4 0.405 0.41 0.415 0.42 0.16 0.17 0.18 0.19 0.2 04080120160ABIteration Figure 4-18. Value of parameters A and B as a function of iteration num ber for LSM at 800 C under 2% O2 The resulting profiles for 16fads and 16fbulk, which describe the time varying isotopic composition of adsorbates and lattice oxygen in Figure 2-4, are shown in Figure 4-19 with final parameter values. It is consistent with the incorporat ion limited behavior in Figure 4-17b. The parameter 66


values also indicate the rate lim iting step is step 2, incorporati on. Although the fit for the surface oxygen composition has an R2 value of 0.995, the bulk model indi cates a much higher degree of exchange activity than actually calculated. Ta king into consideration LSMs rather low oxygen diffusivity [37,40] it is possible that the assu mption made earlier about the oxygen isotopes being well mixed in the solid is not valid. We envision a spherical particle of LSM where only the outermost shell of material initially participates in the exchange reaction and then diffusion of the normal oxygen isotope from the core to the su rface constitutes an additional, rate limiting, 0 0.2 0.4 0.6 0.8 1 01020304050Fraction of 16OTime (minutes)A = 0.4155 B = 0.1616 C = 0.9068 Beta = 0.03Bulk Surface60 Figure 4-19. Profile of 16fads ( ) and 16fbulk ( ) obtained from LSM exchange profile at 800 C with model (solid lines) superimposed. Final R2 = 0.99548 (16fads only). process. This is consistent with other works which also concluded that bulk diffusion is the slowest step [37,87]. It is possibl e that the thin outermost shell is well represented by the model; however, we did not verify this. As for the gas phase behavior, Figure 4-20 shows the model superimposed on the normalized exchange data. The fit is very good for 18O2 and 16O2 but overestimates 16O18O. Small changes in the 16O2 data profile and feed gas purity (2 + ) can result in large deviations in the predicted behavior. Table 4-1 lists the different exchange para meters produced by the computer fitting program for LSM at 700 C in 20 sccm flow rate at various oxygen concentrations. Although the model may not completely capture the bulk behavior of LSM, it is impor tant to compare the 67


0 0.2 0.4 0.6 0.8 1 01020304050Normalized Gas Phase SignalTime (minutes)18O2 16O18O16O2A = 0.4155 B = 0.1616 C = 0.9068 Beta = 0.0360 Figure 4-20. Normalized 16O2, 18O2 and 16O18O signals (open points) a nd predicted profile from the model (solid lines) for LSM at 800 C behavior obtained from the simulation with the isotopic exchange reactio ns observed earlier in the study. Regarding the concentration of adsorb ed to bulk oxygen, C, there seems to be a general trend that the value increases with increasing pO2 or decreasing temperature. This is an expected adsorption behavior. The parameter char acteristic of the dissociative adsorption step, A, scales proportionally with th e oxygen partial pressure. This is not the case for the parameter corresponding to the incorporation step, B and it is also insensitive to the pO2, as observed in Figure 11a. Since incorporation appe ars to be rate limiting, the eff ective rate constant, calculated with an assumed value of the oxygen vacancy con centration, is given in the last column and appears to be constant within the error of th e measurement and assumption. Table 2 shows that the effective rate constant is dependent on te mperature and corresponds to an activation energy of 6.48 kJ/mole calculated between 700 and 800 C, and an activation energy of 76.8 kJ/mole from 600-700 C. The change in ac tivation energy at 700 C was previously observed; however, Table 4-1. Exchange parameters obtained from simulation for LSM at 700 C, 20 sccm flow rate pO2 A = k 1[Oads] B =k2[Vo] C = [Oads]/[Obulk] Feed impurity [Vo] (1/m3) k2 2% 1.381 0.150 1.701 0.03 1.16 x 1022 1.29 x 10 23 1% 0.640 0.126 0.940 0.03 1.16 x 1022 1.08 x 10 23 0.5% 0.315 0.144 0.779 0.03 1.16 x 1022 1.24 x 10 23 Value from [37] 68


Table 4-2. Exchange parameters obtaine d from simulation for LSM at 2% O2, 20 s flow rate ccm Temperature (C) A = k 1[Oads] B =k2[Vo] C = [Oads]/[Obulk] Feed impurity [Vo] (1/m3) k2 600 2.157 0.050 7.174 0.03 1.16 x 1022 4.35 x 10 24 700 1.381 0.150 1.701 0.03 1.16 x 1022 1.29 x 10 23 800 0.416 0.162 0.907 0.03 1.16 x 1022 1.39 x 10 23 the activation energy decreases with increasing temperature and is opposite to what was previously calculated using a different analysis technique consisting of using the initial slope of the overall reaction rate [71]. Previously, the smaller activation energy below 700 C was attributed to the possibility that the initial reaction rate calculation was onl y taking into account the more facile exchange of adsorbed species on the surface without ex change with the bulk oxygen [71]. Now, using the simulation and parame ter fitting method in this study, we can better capture the real behavior of LSM, where exchan ge with the bulk is more highly activated at lower temperatures. These calculated values are all lower than that reported by De Souza et al, who obtained an activation energy for the surface exchange coefficient (ks*) of 128 kJ/mole from the isotope exchange depth profile (IEDP) method performed in the temper ature range of 700-1000 C [40]. This could be due to exchange between the powder and gas being much more facile than exchange with a dense IEDP solid sample. It is not known if these values are directly comparable or if it is more like comparing apples to squa res. One represents a ma ss transfer coefficient (ks*), while the values presented here are mechanistic rate constants with uni que units as defined by equation 2-8. When modeling LSCF, the program would ofte n become unstable due to the value of B continuously growing unbounded. To stabilize the computation, the program was only permitted to fit the value of A and C over 150 it erations to a tolerance of 5 x 10-6 % (Figure 4-21). Figure 4-22 shows the data with superimposed model as a function of time. The experimental data 69


0.15 0.175 0.2 0.225 0.25 0.9 0.95 1 1.05 1.1 020406080100120140160ACIteration Figure 4-21. Value of parameters A and C as a function of iteration num ber for LSCF at 800 C under 2% O2. 0 0.2 0.4 0.6 0.8 1 01020304050 60 Surface (data) Bulk (data) Surface (model) Bulk (model)Fraction of 16OTime (minutes)A = 0.199 B = 20.000 C = 1.072 Feed impurity = 0.01 Figure 4-22. Profile of 16fads ( ) and 16fbulk ( ) obtained from LSCF 2% O2 exchange profile at 800 C, with model (lines ) superimposed. Final R2 = 0.995 (16fads only). indicates that the oxygen isotope composition at the surface and in the bulk mirror each other closely, which is consistent with a very fast incorporation step (large B). The bulk model follows the surface model, which has a correlation of R2 = 0.995 with the surface data. (2 + ), the fraction of 16O in the isotopically labeled stream, fo r LSCF is smaller than that for LSM because the source of isotopic oxygen was changed dur ing the course of this study with one of a slightly higher purity. The gas pha se profile (Figure 4-23) shows th at the model can also predict the behavior of LSCF fairly well. To illustrate the system reciprocality referenced in section 4.2, the same exchange behavior was fitted using the 18fads data rather than 16fads; results are presented in Figure 4-24. The gas feed impurity is no longer the 16O in the feed stream but rather 18O. 70


Notice that the values of A and C (B is again held constant), are sim ilar to those obtained by fitting 16fads. The R2 value is slightly lower at 0.988 versus 0.995 when 16fads is used. Because of this, we have greater confidence using 16fads to fit the data, although theoretically it should not matter. We acknowledge there coul d be errors in the calibration of the labeled oxygen signal and therefore prefer to use the directly calibrated oxygen signal for analysis. 0 0.2 0.4 0.6 0.8 1 01020304050Normalized SignalTime (minutes)18O2 16O18O16O2A = 0.199 B = 20.000 C = 1.072 Feed impurity = 0.0160 Figure 4-23. Normalized 16O2, 18O2 and 16O18O data (open points) and predicted profile from the model (solid lines) for LSCF 2% O2 at 800 C 0 0.2 0.4 0.6 0.8 1 01020304050Normalized SignalTime (minutes)18O2 16O18O16O2A = 0.124 B = 20.000 C = 1.003 Feed impurity = 0.9960 Figure 4-24. Normalized 16O2, 18O2 and 16O18O data (open points) and predicted profile from the model (solid lines) for LSCF 2% O2 at 800 C, calculated by fitting the 18fads data instead of the 16fads data. R2 = 0.988 71


Alternatively, a simplification was made to require only one parameter, A, be estimated to solve for the time dependent behavior for LSCF and then, with various assumptions, the effective rate constant was calculated under differe nt scenarios. The simplification is simply to set 16fbulk = 16fads since the bulk follows th e surface closely, and upon in sertion into equation 4-8, the second term drops out. This leaves only A and does not require any guessing of B and C values. Results are given in Table 4-3 and Table 4-4 for 15 mg sample size in all cases. Table 4-5 shows the variation with LSCF sample size at 50 0 C. The surface coverage is estimated from the 400 C profile (Figure 4-14b) by setting the maxi mum rate of production equal to the amount of oxygen exchanged at the knee of the figure, representing the oxygen which is actually exchanging and not that just due to the switc h. This number is approximately 4734 ppm of O2 exchanged initially when 16fads=1. From the concentration, sample size and flow rate, the moles Table 4-3. Exchanges for LSCF at 700 C, 20 sccm total flow rate at various pO parameter2 pO2 A = k 1[Oads] Feed impurity [Oads] ( mole O/m2)k1 = A[Oads]/pO2 2% 0.217 0.01 11.3 123 1% 0.125 0.01 11.3 141 0.5% 0.059 0.01 11.3 133 Table 4-4. Exchange paramSCF at 1% O2 40 sccm at 400-500 C eters for L Temperature (C) A = k 1[Oads] Feed impurity [Oads] ( mole O/m2) k1 = A[Oads]/pO2 500(100% Conversion) 0.247 0.01 11.3 278 400(47% Conversion) 0.114 0.01 11.3 128 Table 4-5. arameters for varying sample sizes of LSCF (1% O 40 sccm, 500 C) Exchange p2 Sample Weight (mg) A = k 1[Oads] Total Surface Area (m2) Feed impurity [Oads] ( mole O/m2) k1 = A[Oads]/pO2 T otal Rate A x k1 x pO2 ( ) S 4.8 0.566 0.03 0.01 11.3 639 0.19 9.4 0.329 0.06 0.01 11.3 371 0.22 17.1 0.240 0.11 0.01 11.3 270 0.30 72


of oxygen reacted per unit area can be estimated. Se tting this equal to the expression in equation 4-3 and using the fitted parameter for A, the concentration of Oads can be back-calculated. The value of 11.3 mole O/m2 obtained is reasonable when compared to the value of 4 mole O2/m2 8 mole O/m2 used in earlier analysis by Teraoka [17] and 13.7 mole O2/m2 used by Joly [88]. Using the assumption that th e overall oxygen exchange reaction is in equilibrium, the value of the rate constant k1 is calculated from parameter A using equation A6 in the appendix (the concentration of surface sites is ignor ed due to lack of information). Table 4-3 shows a direct rela tionship between the paramete r A and the oxygen partial pressure. However, the resulti ng calculated rate constant k1 is essentially unaffected by the partial pressure after dividing by the pO2. This is to be expected fo r a mechanistic rate constant, which should not be pO2 dependent and is thus in contrast to surface exchange coefficients obtained by conductivity relaxation or IEDP methods [14,15,36,40,57]. The A value for LSCF under 20 sccm flow rate at 800 C is calculated to be 0.199 (Figure 4-23) when computed with both B and C values. A is calculated to be 0.217 for LSCF at 700 C when the approximation is made and only A is fitted (Table 4-3). As in dicated previously in Figure 4-11d, the rate of oxygen exchange with LSCF is essentially te mperature independent above 600 C in the pO2 range investigated, so there is no difference in th e profiles at 700 C and 800 C. The similarity of these results show s the validity of the approximation. Table 4-4 shows that the valu e of A has a strong depende nce on the temperature between 400-500 C, in agreement with behavior seen in Figure 4-13b and in Figure 4-14b. Note that the conversion in 5-14b is approxima tely 47%; the fact that k1 = 128 at 400 C is very close to 47% x (k1 at 500C) = 131 (Table 4-4), seems more than coincidence. The calcu lated activation energy 73


of k1 is 33.6 kJ/mole, again much lower than that reported for the surface exchange coefficient from IEDP in this temperatur e range of 105.1 kJ /mole [80]. Table 4-5 contains the dissocia tive adsorption parameter, A, for three different sample sizes of LSCF at 1%O2 and 40 sccm, 500 C. Without taking into account the sample size, the smaller sample will appear to react more rapidly (k1) since there is less oxygen in the sample available for exchange. When the surface area is factored in to produce the total rate, the total rate is proportional to the sample surface area, as shown in the last column. As the surface area increases, the total amount of oxygen exchange increases, which s upports previous findings that higher cathode surface area is benefici al for SOFC performance [89]. Our present equipment configuration prev ents testing under higher flow rates and concentrations; however there is much potential to continue this wo rk with other cathode materials. Isotopic exchange is sufficient for determining which of the two steps discussed is rate limiting, but the reality is that the oxygen reduc tion reaction is accompanied by charge transfer as well. The two steps can be furt her divided into steps containing a net transfer of two electrons for ever oxygen which is reacted at the cathode. This presents the need to combine this method with additional electrical measurements to trul y understand all the processes occurring on the surface. 4.5 Conclusion Several oxygen surface exchange characteristic s for LSM and LSCF were determined from the isothermal isotopic switching te chnique as interpreted by behavi or predicted from a two step model consisting of dissociative adsorption of an oxygen molecule followed by incorporation into the solid. From the initial reaction rate dependence on pO2, the rate limiting step in LSCF, as given by the derivation, is dissociativ e adsorption between 600 and 800 C. In the case of LSM, both 74


steps are significant at 600 and 800 C. The activation energy for LSM (0.90 eV above 700 C) is significantly higher than that of LSCF (0.10 eV above 475 C) at all temperatures tested. An activation energy of 0.41eV was measured fo r LSM below 650 C. This lower value is interpreted as the activation energy for oxygen diss ociation on the surface, step 1 of the reaction only, and suggests that there is a large barrier to oxygen diffusi on through the LSM bulk. It is therefore advisable to operate SOFCs with LSM cathodes at temperatures no lower than 700 C so that some oxygen diffusion through the bulk can take place. For SOFCs with LSCF cathodes, the lower limit to the operating temperature should be 500 C since below that limit, the activation energy increases to 0.29 eV and incorporation becomes more difficult. The time dependent conversion of the sample oxygen lattice from 16O to 18O, under testing conditions, is insensitive to pO2 for LSM and strongly dependent on temperature indicating incorporation limitations. In contrast LSCF is strongly dependent on pO2 and insensitive to temperature. After doubling the flow rate and re-examining LSCF, it was concluded that LSCF was not exclusively limited by the surface reactio n but also somewhat by gas phase diffusion. Computer simulations of the dissociative adsorption limited and in corporation limited scenarios predicted that the time dependent isot opic distribution of the bulk and surface atomic oxygen would follow different behaviors. In the case that dissociative adsorption was limiting, the surface and bulk conversion from 16Oads to 18Oads would be nearly identical. When incorporation is rate limiting, the surface appears to reach steady state far more quickly than the bulk. The rate of conversion of the bulk does not appear to affect the surface behavior in the incorporation limited scenario. Another program was used to extract kineti c parameters from the surface oxygen isotope composition profile and simulate gas phase profil es. The fit for LSM was good with regard to the 75


76 surface composition, but did not ade quately model the overall bulk behavior. It is therefore likely that LSM is not limited by the surface reaction but rather by diffusion through the solid. There is no consideration for diffusion from the core to the surface in the model, which could explain why it did not fully describe th e exchange kinetics of LSM. The model considered in this study does a good job capturing the behavior of the LSCF exchange reaction. The isotopic composition of th e surface and bulk mirror each other closely, in agreement with a relatively fast incorporation st ep. A rate constant for dissociative adsorption, the rate limiting step, was obtained from fitting the data to the model. After adjusting for sample mass, the overall reaction rate was f ound to be proportional to surface area. This technique was also used to detect degr adation in a sample of LSCF which had been subjected to multiple switching cycles. The reacti vity of the cycled sample was significantly less than that of a fresh sample. By establishing a standard, isotopic methods can be used to easily and quickly detect degr adation in powders. Now that a comprehensive framework has been provided for interpreting the mechanism of oxygen surface exchange, the next step is to stud y how properties of the surface affect catalytic activity. This is the focus of th e next chapter, where infiltrated samples of LSM and LSCF are produced and compared using isotopic exchange.


CHAPTER 5 IDENTIFYING TRENDS IN CATALYTIC AC TIVITY WITH SYSTEMATIC SURFACE MODIFICATION 5.1 Introduction With the goal of developing better intermediate temperature solid oxide fuel cells (SOFC), the principal focus of SOFC research has been on improving the functi onality of the cathode component. (La0.8Sr0.2)0.98MnO3 (LSM) and La0.6Sr0.4Co0.2Fe0.8O3(LSCF) are two high performing and well studied SOFC cathode materi als [10,47,90-94]. Both mate rials belong to the perovskite family and are commercially avai lable, however there is one major difference between the two: LSM is prim arily an electronic conductor whereas LSCF is a mixed ionicelectronic conductor (MIEC). In the previous chapters [ 71], isotopically labeled oxygen exchange was used to show that LSM is not as catalytically active as LS CF. Other studies have also shown that SOFCs with LSCF cathodes outperform those with LSM (and even LSM-YSZ composite cathodes) at intermediate temperatur es [48,95]. However, LSM fares better in terms of long term stability [10,95,96], which is why it remains favored by industry. In order to combine the best attributes of both materials into one cathode, this chapter focuses on the surface modification of a stable LSM base material. The idea of surface modification has been previously studied [5,97,98] for an actual cathode on a SOFC button cell; these studies focused on the apparent overall elec trochemical performance. In contrast, this study focuses only on the catalytic properties of the cathode material to isolate the effect of surface modification. First the most catalytically active surface component of LSCF needs to be isolated by systematically studying how each constituent affects surface reactivity Therefore, after depositing La, Sr, Co or Fe oxide onto the su rface of LSM powder, temperature programmed isotope exchange (TPX) experiments [71] were pe rformed to test for any changes in catalytic activity toward oxygen exchange. These experiments involve linearly heating the material in an 77


isotopically labeled oxygen stream while monitoring changes in signal from the different oxygen isotopologues 16O2, 16O18O and 18O2. Because of the large regular oxygen background in the materials of interest, the use of isotopically labeled oxygen is re quired to distinguish betwee n oxygen originating within the oxide lattice and that from the feed stream. Without using heavy oxygen, only the total oxidation or reduction behavior of these oxides can be observed [19,20,99]. The use of isotopically labeled oxygen in the gas phase adds an additional dimens ion to the reaction and makes the interaction between gas and solid much more transparent. 5.2 Experimental 5.2.1 Materials Unmodified LSM (Nextech), LSCF (Praxair) and Co3O4 powder (Alfa Aesar) were used as reference materials. To make the infiltrated powde rs, five different metal nitrate solutions were first made by combining metal nitrate salts with deionized water. The metal nitrates used were lanthanum, strontium, cobalt, iron (Spectrum Ch emicals) and manganese (Sigma Chemicals). Solution concentrations were verified using i nductively coupled plasma spectroscopy. For each infiltrated sample, a few gram s of LSM powder (or LSCF for manganese infiltration) was weighed out and combined with enough metal ni trate solution to produce 5 wt% metals loading. The mixture was ultrasonicated for a few minutes and then dried on a hot plate for several hours. The dried material was then calcined at 800 C for 2 hours in air and then hand ground to break up agglomerates. Additional infiltrated samples with 10 wt% and 20 wt% cobalt were created in the same manner. 5.2.2 Powder Characterization Brunauer-Emmett-Teller (BET) surface area an alysis was performed on the resulting powders using a Quantachrome NOVA 1200. X-ray diffraction (XRD, Philips APD 3720) and x78


ray photoelectron spectroscopy (XPS, Perk in-Elmer PHI 5100 ESCA, Photon Energy = 1253.6eV) were used to investigate phase purit y and surface composition of the impregnated powders. XPS samples were made by mixing the powders in methanol and depositing the mixture onto 1 cm x 1.5 cm silicon pieces. Spectra were taken after drying and also after heating the samples at 300 C in air for a couple hours to remove adsorbates. Each spectra was the result of 20 or more individual scans. Transmission electron microscopy (TEM) coupled with energy dispersive X-ray spectroscopy (EDS) and diffrac tion were employed to fu rther characterize the particles on the LSM or LSCF base material. Particle morphology was captured with SEM. 5.2.3 Catalysis System Configuration Additional information regarding the experime ntal setup can be found in Chapter 3 and elsewhere [71]. A schematic of the micro-reacto r (Figure 3-2) within the custom built furnace shows the powder sample resting on quartz frit with the thermocouple to uching the sample. The feed stream, containing a mix of oxygen and helium, flows over the powder and exits the reactor base to be sampled by the mass spectrometer. 5.2.4 Temperature Programmed Isotopic Exchange Using surface areas obtained from BET, each temperature programmed exchange (TPX) sample was weighed to yield a total surf ace area of 0.2 m2. The powder was placed in a quartz micro-reactor and then heated to 800 C under 4% 16O2 (balance helium) and allowed to slowly cool to room temperature. Once cooled, a gas mixture of approximately 2,500 ppm 18O2 (95%, Cambridge Isotope) balanced in he lium was fed into the reactor at a flow rate of 20 cc/minute. Then the temperature was ramped at 30 C/minute as the mass spectrometer recorded the oxygen isotope composition of the effluent from 25 C to 800 C in real time. This was repeated twice for each powder. The calibration of the mass spectrome ter signal is done through flowing gases of known composition past the inlet and is describe d in detail elsewhere (C hapter 3 and [71]). 79


5.3 Results and Discussion Table 5-1 shows the results from 6 point B ET analysis for each powder. The instrument (Quantachrome NOVA 1200) acquire s 6 data points and obtains the surface area from curve fitting (R2=0.99 or greater). It is believed that the resulting infiltra ted powders consist of a LSM (or LSCF) base covered in metal oxide particle s. While it was straightforward to detect the presence of Fe, Co on LSM and Mn on LSCF using TEM-EDS and XPS, quantifying the additional presence of La or Sr over the backgr ound signal presented a ch allenge. Results from TEM-EDS and TEM diffraction were inconclusive due to the beam spot size and drift from charging of the oxide materials. Table 5-1. Specific surface areas of infiltrated powders [74] Material Specific Surface Area (m2/gram) LSCF 6.42 LSM 5.14 LSM + 5 wt% La 3.76 LSM + 5 wt% Sr 3.53 LSM + 5 wt% Co 4.73 LSM + 5 wt% Fe 7.23 LSCF + 5 wt% Mn 6.67 LSM + 10 wt% Co 4.57 LSM + 20 wt% Co 5.00 SEM images of the base LSM and two infiltrated LSM powders are shown in Figure 5-1. There appear to be more small particles on the infiltrated powders than on the plain LSM, but there is no way to tell from the phase contrast if the tiny particles ar e bits of LSM or the deposited metal oxides. The other infiltrated powders not shown in the figure had similar morphology. Next, the powders were examined with XRD. The 5 wt% iron and manganese infiltrated powders only showed LSM or LSCF peaks, how ever strontium, cobalt and lanthanum 5 wt% infiltrated powders showed additional peaks from a second phase. These peaks were small and 80


A B C Figure 5-1. Powder morphology captured with SE M, taken at 20,000x A) LSM powder, and LSM infiltrated with B) 5wt% Sr C) 5wt% Co [74] difficult to distinguish from the background, possibl y due to fine dispersion of the deposited particles. In the case of the stron tium infiltrated LSM, the presence of -Sr2MnO4 peaks indicate a reaction occurred between the base material and strontium nitrate to create a layered perovskite [100,101] (K2NiF4 type) and possibly other secondary phases. As expected, lanthanum infiltrated LSM displayed La2O3 peaks from the decomposition of la nthanum nitrate [102]. The 5 and 20 wt% cobalt infiltrated sample showed Co3O4 peaks along with the LSM ba se peaks (Figure 5-2). 20 30 40 50 60 70Intensity (A.U.)2 ThetaLSM + 20 wt% Co LSM Co3O4110 014 012 022 006 024 122 116 124 018 028 220 113 022 222 004 224 115 044 Figure 5-2. Diffraction patterns of LSM, LSM + 20 wt% Co, Co3O4 [74] 81


In order to improve secondary phase detection, 20 wt% iron infiltrated LSM was produced in the same manner as described above and studied using XRD. The 20 wt% iron infiltrated XRD displayed small peaks which looked much like alpha-Fe2O3, the expected decomposition product from iron nitrate [103]. Twenty wt% manganese in filtrated LSCF was also produced and studied with XRD, but did not produce a ny detectable secondary peaks. 5.3.1 X-Ray Photoelectron Spectroscopy The XPS survey for LSM and LSCF is shown in Figure 5-3 with the C 1s reference peak located at 284.8 eV. The expected elements are pr esent in each sample. The profile for the O 1s peak of LSM before and after heating is shown in Figure 5-4. After heating in air at 300 C, the 0 200 400 600 800 1000LSCF LSMIntensity (A.U.)Binding Energy (eV)La3d Co2p OKV Fe2p O1s Sr3pSr3d La4d Mn2p Figure 5-3. Survey spectra (XPS) for LS M and LSCF taken at 45 angle [74] high energy shoulder at 532.9 eV is gr eatly reduced; therefore this peak is attributed to surface contamination (probably from water) [104-106]. The peak at 530.9 eV is assigned to either adsorbed surface oxygen species or to oxygen in the lattice with semicovalent character [107,108]. Finally, the low energy peak at 529.0 eV is assigned to oxyge n in the lattice with very ionic character [73,106,107,109]. The peak at 530.9 eV is noticeably broader than the peak at 529.0 eV. Yamazoe et al also noticed similar results [73] in a co mparable perovskite material (La1-xSrxCoO3) and attributed this broadness to the assort ment of surface states that are available 82


to the adsorbed oxygen. Alternatively, in doped pe rovskites, the coordination of the oxygen ion between different cations can cause polarizati on in the oxygens valence shell [107]. For LSM, the A site is statistically occupied by either a La or a Sr and this may be responsible for the broad peak at 530.9 eV. For consistency, XPS profiles af ter this point are from samples after heating (to remove surface water). 526 528 530 532 534 Binding Energy (eV)Intensity (AU)AContamination Lattice Oxygen Adsorbed or Semicovalent Oxygen 526 528 530 532 534Intensity (AU)Binding Energy (eV) 529.0 eV 530.9 eV 532.9 eV B Figure 5-4. (La0.8Sr0.2)0.98MnO3 O1s peaks A) before heating B) after heating. Data is displayed as points, individual peaks ar e dashed. The summation of the resolved peaks is displayed as a thin solid line [74] Normalized spectra for infiltrated LSM samples against the base LSM O1s profile are shown in Figure 5-5. It is apparent that La a nd Sr infiltration did not change the oxygen surface bonding properties by much, especially for Sr. However, Co and Fe infiltration caused the O1s spectra originally at 529.0 eV to shift to highe r binding energies (by approximately 0.37 eV), and 526 528 530 532 534 LSM LSM + 5 wt% La LSM + 5 wt% SrBinding energy (eV)Intensity (A.U.)A 526 528 530 532 534 536 LSM LSM + 5 wt% Fe LSM + 5 wt% CoBinding energy (eV)Intensity (A.U.)B 0.37 eV Figure 5-5. O1s spectra for LSM infiltrated with 5 wt% A) La, Sr B) Fe, Co [74] 83


also reduced the adsorbed oxygen component near 530.9 eV. If the oxygen adsorption sites are assumed to be metal sites, then the latter observation could indicate more open metal sites at the surface relative to the base LS M. A similar comparison is made (Figure 5-6) with the O1s spectra of LSCF and the LSCF + 5 wt% Mn samp le. The O1s spectrum is different in that the manganese sample is missing the peak at 532.9 eV and the adsorbed oxygen peak at 531.1 eV is slightly smaller. The peaks at 528.6 eV overlap cl osely, which indicate that the surface lattice oxygen species are similar in these two samples. 525 528 531 534 537 540Binding Energy (eV)Intensity (AU)LSCF + 5 wt% Mn LSCF 528.6 eV 531.1 eV Figure 5-6. O1s spectra for LSCF and LSCF + 5 wt% Mn [74] Figure 5-7 shows the La 3d5/2 peaks of the LSM + 5wt% La compared to the base LSM. The similarity in the relative peak intensities a nd positions indicates that the oxidation state of the La on the surface of the infiltra ted material is probably the same as that on the surface of the base material. The 4 eV gap is characteristic of La2O3 [110] and is similar to reported spectra [109,111] for the La3+ oxidation state. The binding energy of the La 3d5/2 peaks agrees with values reported from Wu [109] for La0.5Sr0.5MnO3 pellets with a main line at 834.1eV. 84


830 832 834 836 838 840 842 844Binding Energy (eV)LSM + 5 wt% La LSM 4 eV 834.1 eVIntensity (AU)SatelliteFigure 5-7. La 3d5/2 peaks for LSM and LSM + 5 wt% La [74] Sr 3d peaks for LSM + 5 wt% Sr is shown w ith those from LSM in Figure 5-7. These two profiles are almost identical, which indicate Sr in the infiltrated sample is comparable in oxidation state to Sr in the base LSM. In this case, the Sr 3d peak separation matches those reported by Wu and Kumigashir a [109,112], who studied LSM with differing compositions than the one chosen for this study. Th ey attributed the signal to Sr2+ from comparison to similar compounds. A SrO-like peak is observed around 135.4 eV [113], and this is larger in the case of the strontium infiltrated LSM. Very little data on -Sr2MnO4, including XPS work, could be found for comparison. 128 130 132 134 136 138 140Intensity (AU) 1.7 eV LSM + 5 wt% Sr LSMBinding Energy (eV) SrO Figure 5-8. Sr 3d peaks for LS M and LSM + 5 wt% Sr [74] Figure 5-9 shows the comparison between the LSM + 5wt% Co and LSCF XPS Co 2p3/2 spectrum. The C1s signals of the two compounds had the same binding energy, so no correction was needed. The binding energy is characteristic of c obalt oxides [114], but si nce the tabulated 85


776 780 784 788 LSM + 5 wt% Co LSCFIntensity (AU)Binding Energy (eV) Figure 5-9. Co 2p3/2 profile of LSM + 5wt% Co and LSCF [74] values are very close, it is difficult to say exactly which oxide contributed to this signal. Fortunately, the XRD data have verified that Co3O4 is on the surface of th e infiltrated LSM and it is deduce that the oxidation state of cobalt at the LSCF surface is mixed. The fact that the spectra overlap perfectly indicates the electronic state of cobalt on the surface of LSM is identical to that of the cobalt at the surf ace of the LSCF lattice. Next, Figure 5-10 shows the difference between the Fe 2p3/2 spectrum of LSM + 5wt% Fe and LSCF. Fujii [115] reported binding energies of 710.7 and 710.9 eV for and Fe2O3, respectively, so it appears the iron infiltrated LSM has Fe3+ at the surface. Brundle [116] found that Fe3+ and Fe2+ could be distinguished using XPS and they had binding energies of approximately 711.2 and 709.7 eV, respectively. The LSCF Fe 2p3/2 binding energy is significantly different from the infiltra ted LSM and fits better with the Fe2+ oxidation state. The Mn 2p3/2 peak from manganese infiltrated LS CF showed a 0.3 eV shift to higher energy when compared to the 2p3/2 peak from LSM centered at 641.7 eV. It is difficult to determine the oxidation state from the Mn 2p sp ectra because the peaks for several manganese oxide compounds have very similar binding energies. 86


708 710 712 714 716 LSM + 5 wt% Fe LSCF Binding Energy (eV)Intensity (AU)710.8 eV 710.0 eV Figure 5-10. Fe 2p3/2 profile of LSM + 5wt% Fe and LSCF [74] 5.3.2 Temperature Programmed Isotopic Exchange The use of 18O2 to investigate catalytic activity was di scussed previously [71]. Winter [64] and Boreskov [65] were some of the first re searchers to use this technique to study oxide reactivity. The basic idea is to differentiate between oxygen from the gas phase and oxygen from the solid lattice. Without using isotopically en riched oxygen, it would not be possible to observe the exchange reaction occurring below the total o xygen curve in the profiles from Figure 5-11. If no reaction occurs, then the signal for each of the oxygen species would remain constant; this is not the case. A decrease in the 18O2 line means that species are be ing taken out of the gas phase and either going into the lattice or recombining with a 16O at the powder surface to form the mixed product 16O18O. The increase in 16O2 means that the oxygen originally from the lattice is now being released into the gas phase. 87


0 1000 2000 3000 100200300400500600700800Concentration (ppm)Temperature (C)16O18O16O2 18O2Total O2ALSCF 0 1000 2000 3000 100200300400500600700800BConcentration (ppm)Temperature (C)16O18O16O2 18O2Total O2LSM 0 1000 2000 3000 100200300400500600700800Concentration (ppm)Temperature (C)C16O18O16O2Total O2LSM + 5 wt% La 88


0 1000 2000 3000 100200300400500600700800Total O2DConcentration (ppm)16O18O16O2 18O2Temperature (C)LSM + 5wt% Sr 0 1000 2000 3000 100200300400500600700800E Total O2 18O2 16O18O16O2Temperature (C)Concentration (ppm)LSM + 5 wt% Co 0 1000 2000 3000 100200300400500600700800Concentration (ppm)Temperature (C)16O18O 16O2 18O2Total O2FLSM + 5 wt% Fe Figure 5-11. Temperature programmed exchange profile for A) LSCF B) LSM C) LSM + 5 wt% La D) LSM + 5 wt% Sr E) LSM + 5 wt% Co F) LSM + 5wt% Fe [74] 89


According to Boreskov [70], the gas-phase recombination reaction (equation 4-1) is insignificant below 1000 C. Since conditions in this study do not exceed 800 C, the formation of 16O18O above the baseline is assumed to occur only on the surface with the participation of adsorbed oxygen atoms according to the model. 16O2 above the baseline is primarily due to the oxygen within the lattice ginh p oing to te gashase. 4-1 As discussed in Chapter 2, the 2 step model (repr inted here as equation 2-8) consists of the dissociative adsorption of a mol ecule of oxygen on the surface fo llowed by the incorporation of the adsorbed oxygen atomtos in the olid. 2-8 The implication of step 1 is that the different oxygen species will eventua lly become statistically distributed during reaction accordi ng to the isotopic composition of the adsorbed oxygen. This is shown in Figure 5-12 for LSM + 5 wt% Co. The square root of the normalized 16O2 signal (Figure 5-12a) is taken to find the corresponding fraction (16fads) of 16Oads on the surface. Then the distribution of 18O2 and 16O18O is reconstructed using (1-16fads)2 and 2*(16fads)(1-16fads) respectively (Figure 5-12b); th e 2 in front of the term (16fads)(1-16fads) arises from the reaction multiplicity. Initially, due to kinetic limitations at low temperature, there is no correlation for 18O2 and 18O16O, but as the surface temperature and the reaction rate increase, the model falls on top of the data showing statistical equilibrium has been obtained for all species. 90


0 0.2 0.4 0.6 0.8 1Normalized SignalNormalized 16O2 signal 16fadsA 0 500 1000 1500 2000 2500 100200300400500600700800Concentration (ppm)Temperature (C)16O2 18O2 16O18OB Figure 5-12. Temperature programmed excha nge profile of LSM + 5 wt% Co A) 16O2 signal normalized and raised to power of 0.5 to yield theoretical 16Oads fraction on surface and B) data (points) plotted with sta tistical equilibrium model (solid lines) [74] Furthermore, it can be assumed that the sample is the source of 16O throughout the experiment because conversion of the sample oxygen lattice (with changes in oxygen stoichiometry factored in) from 16O to 18O is calculated to be onl y 10% for LSCF (the most active material) and less for the others. This means that back reaction of the 18O out of the solid is minimal, and it is assumed that the 18O2 signal originates from the gas phase and not from saturation of the surface with 18O. If saturation were to occur, th en the signal would rise back up with increasing time/temperature, which is not observed. Another feature present in all profiles from Figure 5-11 is th e increase in total oxygen with increasing temperature. This ne t release of oxygen from the solid indicates the formation of oxygen vacancies, with LSCF showing a much la rger non-stoichiometry change than LSM and 91


LSM based samples. The amount of oxygen releas ed, as shown in Figure 5-13, is the same regardless of the isotopic labe ling of the gas phase during the temperature programmed reaction. 0 200 400 600 800 1000 160240320400480560640720 Oxygen from 16O2 TPR Total Oxygen from 18O2 TPRNet Oxygen Desorbed (ppm)Temperature (C)LSCF LSM Figure 5-13. Net oxygen desorption from temper ature programmed reaction (TPR) profile for LSM and LSCF. Performed in approximately 2,500 ppm O2 using both normal and isotopically enriched oxygen gas [74] The exchange profile for LSCF and LSM are shown in Figure 5-11a-b. It was found that the rate determining step in LSM up to 700-750 C is incorporation (step 2) while the rate determining step in LSCF is dissociative adsorpti on (step 1) [71]. The onset of exchange occurs at a significantly lower temperature in LSCF (260 C) than in LSM (375 C). Previous studies have also shown that LSCF is more catalyt ically active toward oxygen than LSM [10,48,71]. LSM not only activates at a higher temperature, it also appears to have a different mechanism from LSCF up until 700 C. Most of the 18O2 in LSCF becomes incorporated into the lattice upon adsorption, with relatively little remaining on the surface and desorbing as 16O18O. However, the 16O18O peak in LSM is significantly larger, indicating that after an 18O2 dissociates on the surface, there is a tendency for the 18Oads to recombine and desorb from the surface instead of becoming incorporated into the lattice. This is consistent with the fact that LSM has poor oxygen diffusivity [37,40]. Above 700 C, 16O18O decreases due to a change in mechanism [71], since the incorporation of 18O into the latti ce is no longer as highly activated. 92


The effects of infiltration in LSM can be seen in Figure 5-11c-f. Iron infiltration appears to reduce the onset temperature of exchange. A lthough the cobalt infiltr ated sample was an improvement over the base LSM, LSCF is stil l more catalytically active. Results from Yamaharas study on cobalt infiltrated LSM cat hodes also showed improvement over an unmodified cathode [5]. The au thors hypothesized that the en hancement was due to surface processes such as increased number of reaction sites and/or increased triple phase boundary length, which lead to a reduction in the charge transfer resistance region of the AC impedance response. However, their work relied on impedance spectroscopy for analysis. This technique gives overall polarization changes between meas ured electrodes but can not by itself directly deconvolute whether the observed changes in polari zation are due to microstructural or specific reaction rate contributions. In contrast, our appro ach directly measures th e specific reaction rate, independent of cathode microstructure, and changes in catalytic mechanism upon addition/modification of reactive surface sites. In perovskite materials with the general ABO3 formula, researchers have found that it is usually the B site that is more strongly associated with the ca talytic activity [17,117]. With this idea in mind, it is not surprising th at the addition of La or Sr (A site elements) did not produce a large change in the catalytic activity. The addition of B site elements Fe and Co produced a more noticeable change. In their book, Biela ski and Harber [72] discu ss the correlation between the first row transition metal oxide-oxygen surface b ond strength and isotopic exchange activities found by Boreskov [65,118] and Halpern [119]. Co3O4 had a higher specific activity for exchange than Fe2O3 and a correspondingly lower oxygen-surface bond strength. MnO2 fell in between Co3O4 and Fe2O3 in terms of activity and surface bond strength. Previously, XPS work on LSM with a slightly different composition showed it was possi ble to engineer a LSM thin 93


film to terminate in a MnO2 layer [112]. If the LSM in this study was also manganese oxide terminated, it could explain why cobalt infilt ration improved LSM surface activity while iron impaired it. One observation that doe s not appear to fit this theory, at least initially, is that LSCF contains more iron than cobalt on the B site and is the most ac tive. However, the iron 2p peaks (Figure 5-10) from LSCF and LSM + 5 wt% Fe do not match up, indicating that the states of the two surfaces are different. This is evidence that th e activity is determined by more than just the surface composition; it is dependent on the specific active site. The TPX profile for LSCF infiltrated with manganese is shown in Figure 5-14. Adding manganese to the surface of LSCF suppresses th e exchange onset temperature from 230 C to 310 C, but does not prevent the labeled oxygen fr om going into the lattice above 310 C, as evidenced by the relatively small concentration of 18O2 and 16O18O in the gas stream at high temperatures. This is consistent with the reactiv ity trends previously me ntioned indicating that manganese is not as active as cobalt for surface ex change. It is significant to note that the Mn infiltrated LSCF still performs better than even LSM + 5wt% Co. For LSCF, the dissociative adsorption step is thought to be limiting and therefore the surface modification produced a notable change in the onset temperature while not producing a change in the mechanism. It appears that improving the reac tivity of LSCF can be accomplis hed through cobalt infiltration. Tentative results from a conductivity-relaxation st udy [120] on cobalt infiltrated bars of LSCF show it is possible to obtain faster surface kinetic s with the modified bars over plain bars of LSCF. 94


0 1000 2000 3000 100200300400500600700800Concentration (ppm)Temperature (C)Total O2 16O2 16O18O18O2LSCF + 5 wt% Mn Figure 5-14. Temperature programmed excha nge profile for LSCF + 5 wt% Mn [74] Next, the relationship between activity and th e amount of cobalt infiltrated into LSM was examined. 5, 10 and 20 wt% infiltrated LSM samples were tested and compared to each other. Little difference in catalytic activity was found between th e powders. All temperature programmed exchange profiles showed exchange onset near 300 C. It is unclear if this is due to the low oxygen concentration (2,500 ppm) of the TPX testing conditions. Al so, the particle size of the Co3O4 may be changing such that the coverage remains about the same in each sample. Since cobalt oxide infiltration improved LS Ms catalytic activity for oxygen exchange, we decided to see if pure Co3O4 would be even better and the results are shown in Figure 5-15. It is interesting to see a transition region o ccur between 630-710 C. The phase transition from Co3O4 to CoO at an oxygen partial pressure of 9.9 x 10-10 atm was 350 C, as observed by Oku [121], and between 800-900 C in air (pO2 = 2 x 10-1 atm) [122,123]. Since the conditions for this study were under 2.5 x 10-3 atm O2, it is reasonable that the transition could occur at a temperature range somewhere in between, which ma y explain the equilibrium region in the plot. Above 710 C, the oxide appears to be releasing a small amount of O2, as seen in the rise of the total oxygen line, which is consiste nt with the decomposition of Co3O4 into CoO. 95


0 1000 2000 3000 100200300400500600700800Total O2 16O18O16O2 18O2Concentration (ppm)Temperature (C) Transition Region Figure 5-15. Temperature programmed exchange profile for 40 mg Co3O4 [74] Comparing the catalytic activity of Co3O4, LSM and LSM+ 5wt % Co samples to each other, it is apparent that the infiltrated sample is better than the LSM, which in turn is better than Co3O4. LSCF is still best, and we can draw some conclusions as to wh at properties improve catalytic activity based on the different surfaces of this study. First, we believe LSM is more active than cobalt oxide due to th e lack of oxygen vacancies in th e cobalt oxide. Both materials do not have many vacancies to begin with, but Co3O4 has an excess of oxygen at lower temperatures such that the actual composition should be written as Co3O4+ [72]. The material does not become stoichiometric until betw een 770 and 850 C in air depending on the preparation method [122]. This point is probably lower in this study because the pO2 is two orders of magnitude smaller and could occur right before the oxide enters the phase transition region around 600 C. LSM on the other hand is n early stoichiometric at 1000 C under 2.5 x 103 atm O2 [38,92]. The release of oxygen from Co3O4 is attributed to the decomposition reaction forming CoO while the oxygen desorption in LS M is attributed to vacancy formation. Discrepancies in oxygen stoichiometry alone are not enough to e xplain the difference between the LSM and the cobalt infiltrated LSM. If only oxygen vacancies were responsible for 96


the activity, then the LSM would perform as well as, if not better than the LSM infiltrated with cobalt. The cobalt infiltration a llows oxygen dissociation to occur at lower temperatures, which indicates that the cobalt oxide at the surface may act as an adsorption/active site on the infiltrated surface. Co3O4 itself is an important industrial catalyst because of its ability to easily transition between multiple valance states [124]. C obalt oxide exists as a mixture of Co+3 and Co+2 states in order to preserve charge neutrality. Although the surf ace of cobalt oxide may change valence and stoichiometry with reducing/oxidi zing conditions, in general, th e presence of redox centers on the surface allows it to a ccept or donate extra electrons. Low oxidation state Co2+ functions as an electron donor center through the following reaction: Co+2 Co+3 + e-. If this is the function of the cobalt oxide on LSM, then th e dissociative adsorption step could become less activated due to reduced charge transfer resistance, a ssuming that oxygen dissociation is accompanied by reduction. To effectively compare the activity for oxyge n exchange, the conversion (1normalized 18O2 signal) versus temperature is displayed for all samples in Figure 5-16. The change in 18O2 signal was used as a measure of reactivity since the other oxygen species are reaction by products. LSCF is more active than all other sa mples and is followed, in activity, by the LSCF infiltrated with Mn. The reactivity trends for LS M follow a definite pattern where the addition of A site elements La or Sr results in similar be havior and the B site elements follow the trend predicted by earlier isotopic exchange studies. Iron infiltration significan tly increased the onset temperature of exchange. Royer et al also obs erved a connection between iron contamination and lower activities [26]. All other infiltration elements appeared to improve the catalytic activity of LSM, with the addition of cobalt the most be neficial (30% conversi on vs. 13% for LSM at 97


400 C). However, pure Co3O4 had approximately the same activity as LSM infiltrated with iron; these two samples had the worst exchange properties of all the samples tested. 0 0.2 0.4 0.6 0.8 1 200250300350400450500550600 LSCF LSCF + Mn LSM + Co LSM + La LSM LSM + Sr LSM + Fe Co3O4ConversionTemperature (C) Figure 5-16. Conversion (1-normalized 18O2) vs. temperature, points av eraged over 2 runs [74] All LSM based samples show significant formation of 16O18O compared to LSCF based samples (Figure 5-17), which is consistent with the incorporation limited mechanism of oxygen exchange in LSM. The addition of Co, Sr a nd La to LSM improves the surface recombination reaction, which means that the dissociative ad sorption step is improved. The simultaneous appearance of 18O16O as the 18O2 signal decreases suggests that dissociation of the oxygen at the surface (step 1) is more facile in the infiltrated sample, and the limited incorporation step results in desorption of surface scrambled 18O16O rather than 16O2 from the bulk. Co3O4 is more active than LSM + Fe with regard to surface recombination, yet its overall activity as measured by the removal of 18O2 (Figure 5-16) is the same. This is eviden ce that LSM + Fe is more facile with regard to incorporation than Co3O4 since some of the labeled oxygen is becoming incorporated into the bulk. 98


0 0.1 0.2 0.3 0.4 0.5 0.6 20030040050060070080016O18O / Total OxygenTemperature (C) LSCF LSCF + Mn LSM + Fe LSM + Co LSM Co3O4LSM + La LSM + Sr Figure 5-17. Formation of 16O18O (signal normalized by total oxygen) as a function of temperature [74] As a final note, one of the more important concerns is the behavior of these cathode materials once integrated into an actual device. For example, LSM is typically mixed with an ionic conductor such as YSZ to produce a comp osite electrode; the a ddition of the oxygen ion conducting phase could drastically change the oxygen reduction behavior. Logically, the next application for this technique should be the extension to composite cathodes. 5.4 Conclusion Temperature programmed isotopic exchange is a fast and simple method for surveying oxygen exchange properties of individual oxide materials. Not only does it identify the active temperature range, this technique can also cast light on the mechan ism of oxygen exchange (adsorption or incorporation lim ited). This study used isotopic ex change to identify trends in reactivity. Surface modification through infilt ration of LSM and LSCF powders altered the materials catalytic activity toward isotopic surface exchange but did not affect the overall mechanism of surface exchange. Infiltrated LSM powders were still observed to be limited by oxygen incorporation into the lattice just like the base LSM. A lthough the cobalt in filtrated LSM had 99

PAGE 100

100 improved exchange properties compared to plai n LSM, its overall activ ity was not as high as LSCF. The addition of A site elements La and Sr onto the LSM surface did not produce a significant change in isotopic exchange. Cobalt infiltration improved the performance of LSM the most while iron reduced the apparent activity; this can be attributed to the properties of the oxides formed on the LSM surface, which determ ine how strongly oxygen binds to the surface and subsequently the reactivity. Overall, plain LSCF had the best reactivity. The addition of manganese to the surface of LSCF caused the exch ange to occur at a higher temperature, and is in agreement with previously observed reactivity trends where cobalt oxide is more active than manganese oxide which is more active than iron oxide. Two attributes from this study which are associated with catalytic activity are ionic conductivity and active cobalt (or other surface) sites. The extent to which each of these properties affects cata lytic activity is unclear and needs to be further studied. The issue of l ong term stability also needs to be addressed, and perhaps the improvement in the dissociation st ep achieved with cobalt infiltrate d LSM can be translated into a solution with a low degradation rate.

PAGE 101

CHAPTER 6 CONCLUSION The development of high performance solid oxide fuel cells is dependent upon the fundamental understanding of the oxygen reduction process at the cathode surface. Isotopic exchange is a powerful heterogeneous catalysis tool for comparing catalytic activity and mechanism amongst different oxides. Properties, such as oxygen non-stoichiometry, obtained from this method were comparable to results from the literature. During temperature programmed desorption and reaction, the same total oxygen vs. temperature behavior was obtained using oxygen isotope 16 and isot ope 18 enriched feed streams. Temperature programmed oxygen isotopic exchan ge is a rapid and sensitive method for evaluating catalytic activity and is used to id entify trends in catal ytic activity. It was demonstrated that LSM is not as active as LSCF and that the mechanism of exchange was different for these materials based upon th e relative amounts of each oxygen isotopologues produced. As the temperature increased, the composition of the gas phase would become statistically distributed as predicted by the relative amount of each isotope on the materials surface. This statistical equilibrium between th e surface and gas phases is required for further mechanistic analysis of steady state isotopic exchange. Under testing conditions for isothermal isot opic switching, it was determined that the LSCF surface reaction was occurring so rapidly that the limiting step was influenced by gas phase diffusion. This was not observed in LSM due to its slower reaction rate, which is insensitive to changes in pO2. A two step mechanism consis ting of dissociative adsorption followed by incorporation of the atomic oxygen into the lattice, was used to model not only the gas phase evolution with time, but also the c onversion of the adsorbed oxygen and lattice oxygen from one isotopic labeling to another duri ng the switch. The model predicted greater bulk 101

PAGE 102

oxygen exchange conversion in LSM than back calculated from the profile. This indicates that the LSM could in fact be limited by bulk diffusion and not by the surface reaction as described by the model. In the case of LSCF, inputting a relatively large value for the parameter characteristic of the second step of the model predicted the ga s phase isotopologue distribution and bulk conversion very well, indicating that between the two step s of the model, the dissociative adsorption step is rate limiting. A dditionally, the curve fit ting produced effective rate constants for both LSM and LSCF. The values reported in this dissertation represent one of the first attempts to obtain this information usi ng the two step model. However, the calculated rate constants rely on certain assumed values (i.e. oxygen coverage) and could be improved through actual measuremen t of those quantities. A systematic approach to understanding how the surface composition affects catalytic activity was done through depositing various me tal oxides onto the surf ace of LSM and LSCF. The study found that cobalt infiltrated LSM was more active than plain LSM and that iron infiltrated LSM decreased the activity, implying that the active site is the surface B site atom and not oxygen vacancies. Although the composition of LSCF examined contains more iron than cobalt, it was found to be the most active material of the study. XPS indicated that the oxidation state of the iron at the LSCF surface was different from that on the infiltrated LSM surface. This is evidence that the catalytic activity is not just a st raightforward function of composition, but the electronic surface properties must also be considered when developing cathode materials. Therefore, the use of surface and electronic sensi tive tools, such as a Kelvin Probe, should be the next step in cathode development. Increasing the cobalt content did not produ ce further improvements in the catalytic activity. Pure Co3O4, the oxide deposited on the cobalt inf iltrated LSM, performed worse than 102

PAGE 103

103 plain LSM and was about as active as the LSM infiltrated with iron. Infiltrating LSCF with manganese, the B site atom for LSM, was detrimen tal to the onset temperature of exchange, but it was more active than the LSM infiltrated with cobalt. The apparent mechanism did not change with infiltration for either LSM or LSCF. Give n these observations, it is possible that LSCFs activity is based off an optimi zation of the oxygen non-stoichiome try, cobalt or active redox sites and other unidentified factors. The ability to test cathode mate rials independently from the complete SOFC system is a strength and weakness of the isotopic exchange techniques described in this body of work. Great care must be taken when applying the results si nce the cathode behavior may vary with the electrolyte material and could change under an electrochemical gradient during SOFC operation. Effects from the electrode micros tructure, an important feature wh ich affects performance, also cannot be easily examined. One way to bridge this gap is to extend the application of isotopic exchange to composite cathodes, such as LSM/ YSZ or LSCF/GDC powders, since the important function of the cathode is to assist with the transfer of oxygen from the gas phase to the electrolyte lattice.

PAGE 104

APPENDIX A DERIVATION OF MODEL Starting with Equation 2-8, the introduction of a second isotop ic label into the system produces five distiea atw for the second. nct rc tions, three for the first step nd o (A1) (A2) (A3) (A4) (A5) The intermediate in this 2 st ep model is the adsorbed oxygen 16Oads or 18Oads; it determines the production of the different isotopic species 16O2, 18O2, and 16O18O in the gas phase as well as the incorporation as bulk oxygen. Ther efore, the evolution of the adsorbed oxygen concentration belonging to a particular isotopic labeling is of interest and is derived by balancing the rate of production and consumption. Picking 16Oads for the example, it is seen that this species is produced by the forward step of A1 and consumed by the backward step. It is also produced and consumed in equations A3 and A4, respectively. Another assumption is that both steps in equation 2-8 are in equilibrium with respect to the total oxygen in the system, and the forward and backward rates are equal. (A7) (A6) 104

PAGE 105

105 Equations A6-A7 show that it is po ssible to write an expression for 16Oads with either k1 or k-1, k2 or k-2. The end result is equation 4-8, which has been normalized. Then the process is repeated for 16Obulk which yields equation 4-9.

PAGE 106

APPENDIX B PARAMETER FITTING SCRIPT Main M-File: exchange.m % Extracts parameters A, B and C from fitting differential equations to data % 1. Use initial guesses for A, B and C, then generate numerical solution with ODE45 % 2. Compare the guessed solution to data and use Gauss Newton method to optimize the fit % 3. After some iterations, spit out the optimized solution with data, parameters and error % Cynthia Kan 2009, with Gauss Newton help from Sean Bishop clear, clc tic % Computation time load LSM.dat; % Load data and process into lists oldlist=LSM; rows=size(oldlist,1); n=1; for k=1:2:rows %% data reduction to reduce computation time newlist(n,:)=oldlist(k,:); n=n+1; end t=newlist(:,1); % time in minutes DATA ydata=newlist(:,2); % y values DATA A=1; B=.1; C=1; %starting guesses beta=.015; % Gas impurity dA=0.0001; dB=0.0001; dC=0.0001; %guesses for cha nge in A B and C, B can get extremely large without bound for LSCF Y0=[1 1]; %initial c onditions, 100% oxygen 16 on surface and bulk for j=1:150; %number of program iterations [time,soln] = ode45('rxn',t,Y0,[ ],A,B,C,beta); % calculates the soln to rxn equations, soln is two column vector, column 1=y1, column2=y2 ymod=soln(:,1); %takes solution to 16fads and saves as the current model solution [time,soln] = ode45('r xn',t,Y0,[ ],A+ dA,B,C,beta); Z(:,1)=(soln(:,1)-ymod)/dA; %computes derivative of function for each change in A B or C [time,soln] = ode45('r xn',t,Y0,[ ],A, B+dB,C,beta); Z(:,2)=(soln(:,1)-ymod)/dB; [time,soln] = ode45('r xn',t,Y0,[ ],A, B,C+dC,beta); Z(:,3)=(soln(:,1)-ymod)/dC; ZTZ=(Z'*Z); ZTZinv=inv(ZTZ); Diff=ydata-ymod; ZTD=Z'*Diff; d=ZTZinv*ZTD; dA=d(1)*.1; %computed change in A B and C dB=d(2)*.1; 106

PAGE 107

107 dC=d(3)*.1; A=A+dA; %computed A B and C coefficinets B=B+dB; C=C+dC; abc(:,j)=[A, B, C]'; j; dlog(:,j)=d; %data log of A B an d C changes to see where program died. %%% computes the regression R^2 = 1-SSerr/SStot ydataave=mean(ydata); %average y data values [time,soln] = ode45('rxn',t,Y0,[ ],A,B,C,be ta); %Final computa tion of model values ymod=soln(:,1); %final y values for modeling (f16ads) SStotads=sum((ydata-ydataave).^2); % computes the sum of squares SSerrads=sum((ydata-ymod).^2); % co mputes the summation (ydata-ymodel)^2 Rsquared(j)=1-SSerrads/ SStotads; %R^2 value for fit end figure(1) plot(t,ydata,'o', time,ymod,'--') figure(2) plot(1:j,abc(1,:)) toc % End of Computation time tf=linspace(0,120,2000); [time,soln] = ode45('rxn',tf,Y0,[ ],A,B,C,beta ); %Final computat ion of model values ymod=soln(:,1); %final y values for modeling (f16ads) ymod2=soln(:,2); % final f16bulk m34=ymod.*(1-ymod)*2; m32=ymod.^2; m36=(1-ymod).^2; %%% DATA OUTPUT %%% mydata has 4 columns, first column is time mydata(:,1)=time; mydata(:,2)=m34; mydata(: ,3)=m32; mydata(:,4)=m36; mydata(:,5)=ymod; mydata(:,6)=ymod2; mydata2(:,1)=[1:j]'; mydata2(:,2)=a bc(1,:)'; mydata2(:,3)=abc(2,:)'; save my_data.out mydata -ASCII -tabs save my_data2.out mydata2 -ASCII tabs ODE45 Function File: rxn.m function dydt = rxn(t,y,flag,A,B,C,beta) % solve the system of equations for 16Fads dydt=zeros(size(y)); Y1=y(1); Y2=y(2); % RHS expression dydt(1)=-A*Y1^2-(A+B)*Y1+B*Y2+A*2*beta; dydt(2)=B*C*(Y1-Y2); % dydt=[ dydt(1) dydt(2)]

PAGE 108

LIST OF REFERENCES [1] A. D. Pasternak, "Global Energy Future s and Human Development: A Framework for Analysis," Study done in partnership with US Department of Energy and Lawrence Livermore National Laboratory, Un iversity of California, 2000 [2] "World Consumption of Primary Energy by Energy Type and Selected Country Groups, 1980-2004." Energy Information Administrati on, U.S. Department of Energy, 2006. [3] "Worldwide Trends in Energy Use a nd Efficiency Key Insights from IEA." International Energy Agency, 2008. [4] M. C. Williams, J. P. Strakey, W. A. Surdoval, J. Power Sources 143 (2005) 191-196. [5] K. Yamahara, C. P. Jacobson, S. J. Visco, X.-F. Zhang, L. C. De Jonghe, Solid State Ionics 176 (2005) 275-279. [6] E. Ivers-Tiffe, A. Weber, D. Herbstritt, J. Eur. Ceram. Soc. 21 (2001) 1805-1811. [7] M. Dokiya, Solid State Ionics 152-153 (2002) 383-392. [8] S. B. Adler, J. A. Lane, B. C. H. Steele, J. Electrochem. Soc. 143 (1996) 3554-3564. [9] S. B. Adler, X. Y. Chen, J. R. Wilson, J. Catal. 245 (2007) 91-109. [10] S. P. Jiang, Solid State Ionics 146 (2002) 1-22. [11] J. W. Erning, T. Hauber, U. Stimming, K. Wippermann, J. Power Sources 61 (1996) 205211. [12] L. M. van der Haar, M. W. den Otter, M. Mo rskate, H. J. M. Bouwmeester, H. Verweij, J. Electrochem. Soc. 149 (2002) J41-J46. [13] X. Chen, S. Wang, Y. L. Yang, L. Smith, N. J. Wu, B. I. Kim, S. S. Perry, A. J. Jacobson, A. Ignatiev, Solid State Ionics 146 (2002) 405-413. [14] J. A. Lane, J. A. Kilner, So lid State Ionics 136-137 (2000) 997-1001. [15] H. J. M. Bouwmeester, M. W. Den Otter, B. A. Boukamp, J. Solid State Electrochem. 8 (2004) 599-605. [16] R. Ganeshananthan, A. V. Virkar, J. Electrochem. Soc. 152 (2005) A1620-A1628. [17] Y. Teraoka, H. M. Zhang, N. Yamazoe, Chem. Lett. 14 (1985) 1367-1370. [18] D. Martin, D. Duprez, J. Phys. Chem. 100 (1996) 9429-9438. 108

PAGE 109

[19] A. Mai, F. Tietz, D. Stver, Solid State Ionics 173 (2004) 35-40. [20] R. J. Bell, G. J. Millar, J. Dr ennan, Solid State Ionics 131 (2000) 211-220. [21] R. A. De Souza, J. Zehnpfenning, M. Ma rtin, J. Maier, Solid State Ionics 176 (2005) 1465-1471. [22] S. Carter, A. Selcuk, R. J. Chater, J. Kajda, J. A. Kilner, B. C. H. Steele, Solid State Ionics 53-56 (1992) 597-605. [23] F. S. Baumann, J. Fleig, H.-U. Habermeier J. Maier, Solid State Ionics 177 (2006) 10711081. [24] J. R. Wilson, D. T. Schwartz, S. B. Adler, Electrochim. Acta 51 (2006) 1389-1402. [25] R. A. De Souza, Phys. Chem. Chem. Phys. 8 (2006) 890-897. [26] S. Royer, D. Duprez, S. Ka liaguine, J. Catal. 234 (2005) 364-375. [27] K. Klier, J. Novakova, P. Jiru, J. Catal. 2 (1963) 479-484. [28] R. J. Gorte, J. M. Vohs, J. Catal. 216 (2003) 477-486. [29] S. B. Adler, Chem. Rev. 104 (2004) 4791-4844. [30] F. S. Baumann, 2006. Oxygen reduction kinetics on mixed conducting SOFC model cathodes, Ph.D. thesis, University of Stuttgart, Germany [31] A. C. Co, V. I. Birss, The Journa l of Physical Chemistry B 110 (2006) 11299-11309. [32] S. P. Jiang, J. Power Sources 124 (2003) 390-402. [33] D. W. Oxtoby, H. P. Gillis, N. H. N achtrieb, "Principles of Modern Chemistry, 4th ed. Saunders College Publishing. (1999). [34] A. Weber, E. Ivers-Tiffe, J. Power Sources 127 (2004) 273-283. [35] S. C. Singhal, Solid State Ionics 135 (2000) 305-313. [36] J. A. Lane, S. J. Benson, D. Waller, J. A. Kilner, Solid Stat e Ionics 121 (1999) 201-208. [37] I. Yasuda, K. Ogasawara, M. Hishinuma, T. Kawada, M. Dokiya, Solid State Ionics 8688 (1996) 1197-1201. [38] J. H. Kuo, H. U. Anderson, D. M. Sparlin, J. Solid State Chem. 83 (1989) 52-60. 109

PAGE 110

[39] L. W. Tai, M. M. Nasrallah, H. U. Ande rson, D. M. Sparlin, S. R. Sehlin, Solid State Ionics 76 (1995) 273-283. [40] R. A. De Souza, J. A. Kilner, J. F. Walker, Mater. Lett. 43 (2000) 43-52. [41] J. H. Kuo, H. U. Anderson, D. M. Sparlin, J. Solid State Chem. 87 (1990) 55-63. [42] K. Katayama, T. Ishihara, H. Ohta, T. S-J, Y. Esaki, E. Inukai, J. Ceram. Soc. Jpn. 97 (1989) 1327-1333. [43] R. M. Ormerod, Chem. Soc. Rev. 32 (2003) 17-28. [44] R. A. De Souza, J. A. Kilner Solid State Ionics 106 (1998) 175-187. [45] S. Wang, T. Kato, S. Nagata, T. Honda, T. Kaneko, N. Iwashita, M. Dokiya, Solid State Ionics 146 (2002) 203-210. [46] A. Esquirol, N. P. Brandon, J. A. Kilner, M. Mogensen, J. Electrochem. Soc. 151 (2004) A1847-A1855. [47] J. W. Stevenson, T. R. Armstrong, R. D. Carneim, L. R. Pederson, W. J. Weber, J. Electrochem. Soc. 143 (1996) 2722-2729. [48] E. Perry Murray, M. J. Sever, S. A. Barnett, Solid State Ionics 148 (2002) 27-34. [49] F. M. Figueiredo, J. A. Labrincha, J. R. Fr ade, F. M. B. Marques, Solid State Ionics 101103 (1997) 343-349. [50] A. Petric, P. Huang, F. Tietz, Solid State Ionics 135 (2000) 719-725. [51] W. Zajac, K. Swierczek, J. Molenda, J. Power Sources 173 (2007) 675-680. [52] H. Inaba, H. Tagawa, Solid State Ionics 83 (1996) 1-16. [53] D. Mantzavinos, A. Hartley, I. S. Metcalfe M. Sahibzada, Solid State Ionics 134 (2000) 103-109. [54] A. Hartley, M. Sahibzada, M. Weston, I. S. Metcalfe, D. Mantzavinos, Catal. Today 55 (2000) 197-204. [55] J. A. Kilner, R. A. De Souza, I. C. Fullarton, Solid State I onics 86-88 (1996) 703-709. [56] A. Esquirol, J. Kilner, N. Bra ndon, Solid State Ionics 175 (2004) 63-67. [57] J. E. t. Elshof, M. H. R. Lankhorst, H. J. M. Bouwmeester, J. Electrochem. Soc. 144 (1997) 1060-1067. 110

PAGE 111

[58] J. Maier, Solid Stat e Ionics 112 (1998) 197-228. [59] T. Ishihara, J. A. Kilner, M. Honda, N. Sa kai, H. Yokokawa, Y. Takita, Solid State Ionics 113-115 (1998) 593-600. [60] R. E. van Doorn, I. C. Fullarton, R. A. de S ouza, J. A. Kilner, H. J. M. Bouwmeester, A. J. Burggraaf, Solid Stat e Ionics 96 (1997) 1-7. [61] S. Fearn, J. Rossiny, J. Kilner Solid State Ionics 179 (2008) 811-815. [62] B. C. H. Steele, Solid State Ionics 75 (1995) 157-165. [63] N. Morita, T. Titani, Bull. Chem. Soc. Jpn. 13 (1938) 601-607. [64] E. Winter, Adv. in Catal. 10 (1958) 196. [65] G. Boreskov, Adv. in Catal. 15 (1964) 285. [66] V. S. Muzykantov, G. I. Pa nov, Kinet. Katal. 13 (1972) 350-357. [67] E. R. S. Winter, Journal of the Chemical Society (Resumed) 1950 (1950) 1170-1175. [68] B. A. Boukamp, H. J. M. Bouwmeester, A. J. Burggraaf, Proceedings of the Second Int. Symp. on Ionic and Mixed Conducting Ce ramics San Francisco, CA, 1994, pp. 141 [69] S. L. Shannon, J. G. Goodwin, Chem. Rev. 95 (1995) 677-695. [70] G. Boreskov, "Heterogeneous Catalysis. Hauppauge, New York Nova Science Publishers, Inc. (2003). [71] C. C. Kan, H. H. Kan, F. M. Van Assche IV, E. N. Armstrong, E. D. Wachsman, J. Electrochem. Soc. 155 (2008) B985-B993. [72] A. Biela ski, J. Haber, "Oxygen in Catalysis. New York, New York Marcel Dekker, Inc. (1991). [73] N. Yamazoe, Y. Teraoka, T. Seiyama, Chem. Lett. 10 (1981) 1767. [74] C. Kan, E. D. Wachsman, J. Electrochem. Soc. Submitted (2009). [75] B. A. Boukamp, H. J. M. Bouwmeester, H. Verweij, A. J. Burggraaf, Mater. Res. Soc. Symp. Proc. 293 (1993) 361-366. [76] Y. Teraoka, M. Yoshimatsu, N. Yamazo e, T. Seiyama, Chem. Lett. 13 (1984) 893-896. 111

PAGE 112

[77] D. Efremov, L. Pinaeva, V. Sadykov, C. Mirodatos, Solid State Ionics 179 (2008) 847850. [78] S. Otoshi, H. Sasaki, H. Ohnishi, M. Ha se, K. Ishimaru, M. Ippo mmatsu, T. Higuchi, M. Miyayama, H. Yanagida, J. El ectrochem. Soc. 138 (1991) 1519-1523. [79] R. A. De Souza, J. A. Kilner Solid State Ionics 126 (1999) 153-161. [80] B. C. H. Steele, J.-M. Bae, Solid State Ionics 106 (1998) 255-261. [81] S. J. Benson, R. J. Chater, J. A. K ilner, Proceedings of th e Third International Symposium on Ionic and Mixe d Conducting Ceramics, 1997, pp. 596 [82] S. J. Benson, 1999. Ph.D. thesis, Imperial College, London [83] S. Wang, P. A. W. van der Heide, C. Chav ez, A. J. Jacobson, S. B. Adler, Solid State Ionics 156 (2003) 201-208. [84] B. T. Dalslet, M. Soegaard, P. V. Hendriksen, J. Electrochem. Soc. 154 (2007) B1276B1287. [85] H. S. Fogler, S. H. Fogler, "Ele ments of chemical reaction engineering," Prentice-Hall Englewood Cliffs, NJ. (1992). [86] S. R. Bishop, E. D. Wa chsman, Unpublished (2009). [87] A. Endo, H. Fukunaga, C. Wen, K. Ya mada, Solid State Ionics 135 (2000) 353-358. [88] J. P. Joly, D. Klavana, J. Kirchnero va, React. Kinet. Catal. Lett. 68 (1999) 249-255. [89] D. Gostovic, J. R. Smith, D. P. Kundinger, K. S. Jones, E. D. Wachsman, Electrochem. Solid-State Lett. 10 (2007) B214-B217. [90] N. Q. Minh, J. Am. Ceram. Soc. 76 (1993) 563-588. [91] J. A. M. van Roosmalen, P. van Vlaanderen, E. H. P. Cordfunke, W. L. Ijdo, D. J. W. Ijdo, J. Solid State Chem. 114 (1995) 516-523. [92] F. W. Poulsen, Solid St ate Ionics 129 (2000) 145-162. [93] V. A. Cherepanov, L. Y. Barkhatova, V. I. Voronin, J. Solid State Chem. 134 (1997) 3844. [94] L. W. Tai, M. M. Nasrallah, H. U. Ande rson, D. M. Sparlin, S. R. Sehlin, Solid State Ionics 76 (1995) 259-271. 112

PAGE 113

[95] A. Mai, V. A. C. Haanappel, S. Uhlenbruc k, F. Tietz, D. Stover, Solid State Ionics 176 (2005) 1341-1350. [96] F. Tietz, V. A. C. Haanappel, A. Mai, J. Mertens, D. Stover, J. Power Sources 156 (2006) 20-22. [97] H. Uchida, M. Yoshida, M. Watana be, J. Electrochem. Soc. 146 (1999) 1-7. [98] V. A. C. Haanappel, D. Rutenbeck, A. Ma i, S. Uhlenbruck, D. Sebold, H. Wesemeyer, B. Rwekamp, C. Tropartz, F. Tietz, J. Power Sources 130 (2004) 119-128. [99] N. Caillol, M. Pijolat, E. Si ebert, Appl. Surf. Sci. 253 (2007) 4641-4648. [100] A. N. Grundy, B. Hallstedt, L. J. Gauc kler, J. Phase Equilib. Diffus. 25 (2004) 311-319. [101] K. Tezuka, M. Inamura, Y. Hinatsu, Y. Shimojo, Y. Morii, J. Solid State Chem. 145 (1999) 705-710. [102] A.-E. Gobichon, J.-P. Auffredic, D. Louer, Solid State Ionics 93 (1996) 51-64. [103] G. Ketteler, Weiss, W., Ranke, W., Schlgl, R. Phys. Chem. Chem. Phys 3 (2001) 1114 1122. [104] J. T. Kloprogge, L. V. Duong, B. J. Wood, R. L. Frost, J. Colloid Interface Sci. 296 (2006) 572-576. [105] P. A. W. v. d. Heide, Surf Interface Anal. 33 (2002) 414-425. [106] N. A. Merino, B. P. Barbero, P. Eloy, L. E. Cadus, Appl. Surf. Sci. 253 (2006) 14891493. [107] T. L. Barr, "Modern ESCA: The prin ciples and practice of X-ray photoelectron spectroscopy, CRC press. (1994). [108] D. A. Pawlak, M. Ito, M. Oku, K. Shim amura, T. Fukuda, J. Phys. Chem. B 106 (2002) 504-507. [109] Q.-H. Wu, M. Liu, W. Jaeger mann, Mater. Lett. 59 (2005) 1980-1983. [110] W. M. R. C.D.Wagner, L.E.Davis, J.F.Moulder, and G.E.Mullenberg, "Handbook of Xray Photoelectron Spectroscopy. Eden Prairie, MN, USA Perkin-Elmer Corp. (1979). [111] Y. Uwamino, T. Ishizuka, H. Yamatera, J. Electron. Spectrosc. Relat. Phenom. 34 (1984) 67-78. 113

PAGE 114

114 [112] H. Kumigashira, K. Horiba, H. Ohgu chi, K. Ono, M. Oshima, N. Nakagawa, M. Lippmaa, M. Kawasaki, H. Koinuma, Appl. Phys. Lett. 82 (2003) 3430-3432. [113] H. Van Doveren, J. A. T. H. Verhoeven, J. Electron. Spectrosc. Relat. Phenom. 21 (1980) 265-273. [114] S. C. Petitto, E. M. Marsh, G. A. Carson, M. A. Langell, J. Mol. Catal. A: Chem. 281 (2008) 49-58. [115] T. Fujii, F. M. F. de Groot, G. A. Sawa tzky, F. C. Voogt, T. Hibma, K. Okada, Phys. Rev. B: Condens. Matter Mater. Phys. 59 (1999) 3195. [116] C. R. Brundle, T. J. Chuang, K. Wandelt, Surf. Sci. 68 (1977) 459-468. [117] T. Seiyama, Cat. Rev. Sci. Eng. 34 (1992) 281. [118] G. K. Boreskov, V. V. Popvskii, V. A. Sazonov, Proceedings of the Fourth International Congress on Catalysis 1 (1968). [119] B. Halpern, J. Germain, J. Catal. 37 (1975) 44-56. [120] D. J. Oh, "Conductivity rela xation of La0.6Sr0.4Co0.2Fe0.8O3-d." 2008. [121] M. Oku, Y. Sato, Appl Surf. Sci. 55 (1992) 37-41. [122] R. Garavaglia, C. M. Mari, S. Trasat ti, C. de Asmundis, Surf. Technol. 19 (1983) 197215. [123] K. Koumoto, H. Yanagida, Jpn. J. Appl. Phys. 20 (1981) 445-446. [124] N. Bahlawane, E. Fischer Rivera, K. Kohse-Hinghaus, A. Br echling, U. Kleineberg, Applied Catalysis B, Environmental 53 (2004) 245-255.

PAGE 115

BIOGRAPHICAL SKETCH Cynthia Kan was born in Philadelphia, the same year that her father received his PhD from the University of Pennsylvania. The family moved to Jacksonville Florida a year later and then relocated to the San Francisco Bay Area when sh e was in the third grade. She graduated from Mission San Jose High School and went on to attend the University of California. There, she met her husband and accompanied him to the University of Florida for graduate school. She graduated with a masters degree and went on to attain a PhD. During her time at UF, she served as a gra duate senator in Student Government and was promoted to Chair of the Senate ad hoc Sustainability Committee. She has written numerous bills and resolutions, the most significan t being the allocation of reserv es toward an energy efficient lighting project which would save up to $44,000 a year on energy costs. Working with the committee and the Student Government Environmen tal Affairs Secretary, she ended the term by consolidating the ad hoc sustainability committee and secretary positions into a permanent agency, enabling Student Governments co mmitment to sustainability on campus. Cynthias experiences serving the student body has translated in a desire to pursue a career in public service. She hopes to work on clean en ergy and become an influential policy maker.