ABSTRACT Title of Dissertation: LATTICE MODIFICATIONS OF SOLID STATE ELECTROLYTES FOR THE OPTIMIZATION OF ION TRANSPORT Adam Garrett Jolley, Doctor of Philosophy, 2018 Dissertation directed by: Professor, Eric D. Wachsman, Materials Science and Engineering A solid state electrolyte is one of the primary components of both a solid oxide fuel cells (SOFC) and an all-solid-state sodium battery. In both cases, the ionic conductivity of the electrolyte has a major impact on the performance of the electrochemical cell. For SOFCs, the conductivity of traditional electrolytes is not high enough for sufficient performance at intermediate and low temperature operation. Therefore, novel bismuth oxide compositions were developed to achieve higher conductivity. The conductivity of Bi2O3 was improved by reducing the total dopant concentration required to stabilize the highly conductive cubic phase. This strategy lead to the development of a Bi2O3 electrolyte (La7Zr3) with the highest oxygen ion conductivity to date. Unfortunately, at temperatures below 600°C the conductivity of the cubic phase was unstable. Therefore, rhombohedral bismuth oxide was investigated for low temperature SOFC operation due to its stability. For the first time, a dopant concentration less than 10% was used to stabilize the rhombohedral phase of Bi2O3. Furthermore, a novel phase diagram was constructed for the low dopant regime of the rhombohedral phase. Ultimately, the double doped bismuth oxide material (La5.1Y1.4) developed here was among the highest and most stable oxygen ion conductors below 600°C. Performance of an SOFC with a La5.1Y1.4 electrolyte verified that it is a promising material for low temperature SOFCs. A similar strategy of doping an electrolyte material to increase ionic conductivity was carried out on NASICON (Na3Zr2Si2PO12). NASICON is a promising electrolyte for room temperature sodium batteries, but traditionally it does not exhibit high enough conductivity to garner high performance. For the first time, the mechanism driving the phase transition in NASICON was determined and mapped out. Mitigation of the phase transition in the material was established to lower the activation energy barrier for sodium ion transport. Additionally, divalent cations were substituted into the NASICON lattice to generate an increase in sodium ion conductivity. Ultimately the phase and dopant concentration was optimized to deliver a material that is among the best sodium ion conducting ceramics to date (20% Zn-doped NASICON). LATTICE MODIFICATIONS ON SOLID STATE ELECTROLYTES FOR THE OPTIMIZATION OF ION TRANSPORT by Adam Garrett Jolley Dissertation submitted to the Faculty of the Graduate School of the University of Maryland, College Park, in partial fulfillment of the requirements for the degree of Doctor of Philosophy 2018 Advisory Committee: Professor Eric D. Wachsman, Chair Professor Liangbing Hu Professor Bryan Eichhorn Professor Isabel K. Lloyd Professor Yifei Mo © Copyright by Adam Garrett Jolley 2018 Acknowledgements First, I would like to thank my advisor Dr. Eric Wachsman for all his guidance, knowledge and support. You have helped me enjoy my time as a graduate student and succeed in a research environment. Next, I would to thank all the members of the Wachsman lab for the comradery and intellectual discussions they provided. The members of the lab really made me feel at home and have a great time in the graduate program. Specifically, I would like to thank Dr. Mohammed Hussain for his help constructing and testing a full SOFC, as well as his general SOFC expertise. Ian Robinson for his help with SEM imaging. Tom Hays for his general advice and know how. Chris Pellegrinelli for his help with all electronic things in the lab, as well as his assistance on ionic transference number measurement and calculations. Yi-Lin Huang for his knowledge of composite cathodes and synthesis abilities. Finally, Greg Hitz for originally bringing me up to speed in the lab and helping me make sense of the NASICON materials. I want to acknowledge Dr. Peter Zavalij for his expertise in X-ray diffraction, as well as Dan Taylor and Rishvi Jayathilake for their help in obtaining and fitting X- ray diffraction patterns. Also, I would like to acknowledge all my financial support. The US Department of Energy's Office of Electricity Delivery & Energy Reliability and Sandia National Laboratories, contract # 1161548. Also, Redox Power Systems and funding from the U.S. Department of Energy, Advanced Research Projects Agency - Energy (ARPA-E) contract number DE-AR0000494 ii Finally, I would also like to thank my friends and family, for their love and support. Without you I would not be where I am today. iii Table of Contents Acknowledgements ....................................................................................................... ii Table of Contents ......................................................................................................... iv List of Tables ................................................................................................................ v List of Figures .............................................................................................................. vi List of Abbreviations ................................................................................................... ix Chapter 1: Introduction ................................................................................................. 1 1.1 General Electrochemistry Introduction ............................................................... 1 1.2 Fundamentals of Solid Oxide Fuel Cells ............................................................ 2 1.3 Bismuth Oxide as an SOFC Electrolyte.............................................................. 5 1.4 Fundamentals of Solid State Sodium Batteries ................................................. 15 1.5 NASICON as a Solid State Sodium Battery Electrolyte .................................. 17 Chapter 2: Experimental Procedures .......................................................................... 22 2.1 SOFC Electrolyte Bismuth Oxide Synthesis .................................................... 22 2.2 SOFC Symmetric Cell Fabrication and Cathode Synthesis .............................. 23 2.3 SOFC Full Cell Fabrication .............................................................................. 23 2.4 Sodium Battery NASICON Electrolyte Synthesis ............................................ 25 2.5 Electrochemical Impedance Spectroscopy ....................................................... 26 2.6 Transference Number Measurements ............................................................... 29 2.7 SOFC Button Cell Testing ................................................................................ 30 2.8 X-ray Diffraction .............................................................................................. 31 2.9 Microscopy and Elemental Analysis ................................................................ 31 2.10 Differential Scanning Calorimetry .................................................................. 32 Chapter 3: Doping Bi2O3 to Improve Conductivity and Stability............................... 33 3.1 Motivation ......................................................................................................... 33 3.2 Effects of Lanthanide Substitution on Cubic Bi2O3 ......................................... 33 3.3 Optimizing the Rhombohedral Bi2O3 Lattice for Maximum Conductivity ...... 48 Chapter 4: Doping NASICON to Increase Conductivity ............................................ 67 4.1 Motivation ......................................................................................................... 67 4.2 The Effect of Aliovalent Cation Substitution on NASICON Structure ............ 67 4.3 The effect aliovalent cation substitution on NASICON conductivity .............. 80 4.4 Optimizing the Bulk Conductivity of NASICON with Aliovalent Doping ...... 94 Chapter 5: Electrochemical Cell Development and Testing .................................... 103 5.1 Composite Cathodes and Bilayer SOFCs Using Bi2O3 .................................. 103 5.2 Fabricating a NASICON Scaffold for a Solid State Sodium Battery ............. 111 Bibliography ............................................................................................................. 115 iv List of Tables 1.3.1 Key pre-exponential factor contributions for the temperature dependence of conductivity on disordered and ordered cubic Bi2O3. Table Adapted from1. 2.5.1 Capacitance values measured by EIS and the possible interpretation. Table adapted from2. 3.3.1 Voltage measurement, bulk resistance, total resistance and calculated transference number (Eq. 3.3.1) for the La5Y2 sample over a range of PO2 values. 3.3.2 Phase, average dopant ionic radius, and room temperature lattice parameters of many Bi2O3 samples. 4.2.1 Room temperature lattice parameters and actual stoichiometry of all NASICON samples. 4.2.2 Phase transition temperature of all NASICON samples determined from the endothermic peak (Tc) using DSC. 4.3.1 Sintering temperature and resulting density of all NASICON samples. 4.3.2 Fitting parameters and percent error generated by the equivalent circuit fit of the EIS data in Z-view for all NASICON samples. 4.3.3 Bulk Conductivity and total conductivity given for all samples at room temperature. 4.4.1 Sintering temperature for a range of NASICON samples. v List of Figures 1.2.1 Diagram detailing the operation of a solid oxide fuel cell with gas and components labeled. 1.3.1 a Phase diagram and conductivity of pure Bi2O3 as a function of temperature. Fig. is taken from3 1.3.1 b The δ-cubic Bi2O3 lattice with the cations, anions, and vacancies labeled on the diagram. Fig is taken from4. 1.3.2 a Minimum dopant required to stabilize the FCC phase of Bi2O3 vs. the ionic radius of the dopant. Fig. is taken from5 1.3.2 b Conductivity of cubic Bi2O3 as a function of dopant and ionic radius. Fig. is taken from6. 1.3.3 Arrhenius plot of multiple oxygen ion conducting electrolytes. Fig. is taken from7. 1.3.4 Conductivity as a function of time for multiple bismuth oxide samples at 500°C. Fig. is taken from8. 1.3.5 Conductivity vs temperature for 25% Dy-doped Bi 52O3. Fig. taken from . 1.3.6 Time dependence of mean-square displacement (MSD) of oxygen diffusion for both polarizable and non-polarizable δ-cubic Bi2O3 systems. Fig. taken from9. 1.3.7 The effect of relative thickness and PO2 on ESB stability in an ESB/GDC electrolyte bilayer design. Fig. taken from10. 1.4.1 Diagram of a solid state sodium battery during operation. 1.5.1 Schematic of the Na3Zr2Si2PO12 structure. Fig. taken from 11. 1.5.2 a SEM images of undoped NASICON and SEM images of NASICON doped with 3% yttria in the form of tetragonal polycrystalline zirconia (TZP). 1.5.2 b EIS plot of undoped and TZP doped NASICON. Fig taken from12. 1.5.3 Unit cell of NASICON. 1.5.4 Arrhenius behavior of NASICON over a large temperature region. Fig. taken from13. 2.3.1 Schematic showing the cross-section of an SOFC button cell with a bi-layer electrolyte design. 3.2.1 Arrhenius behavior of D8N4 and DWSB. 3.2.2 Conductivity vs. time for D8N4 and DWSB at 500°C. 3.2.3 Change in conductivity vs. average dopant radius of multiple cubic Bi2O3 samples. 3.2.4 X-ray diffraction patterns of four different double doped Bi2O3 samples. 3.2.5 EIS plot for La1.5Y8.5 from 700°C-500°C. 3.2.6 Arrhenius plot of three cubic bismuth oxide samples. 3.2.7 Conductivity as function of aging time for La1.5Y8.5, DWSB and ESB at 500°C. 3.2.8 Room temperature X-ray diffraction pattern of La1.5Y8.5 before and after aging at 500°C for 60 hours. 3.2.9 Room temperature X-ray diffraction patterns of 10% La and Zr doped cubic Bi2O3 materials. The La2Zr2O7 impurity phase is indicated with triangles. vi 3.2.10 Conductivity as function of aging time for La7Zr3 and DWSB at 650°C. The DWSB is plotted as a reference cubic bismuth oxide material14. 3.2.11 Conductivity as function of aging time for La5.1Y1.4, DWSB and ESB at 500°C. The DWSB and ESB samples are plotted as reference cubic oxide materials14. 3.2.12 Room temperature X-ray diffraction pattern of cubic La7Zr3 before and after aging at 500°C for 100 hours. The black line is the initial XRD plot and the dashed line is after aging. The intensity of the secondary phase La2Zr2O7 peaks grew after aging. 3.3.1 Conductivity as function of aging time for four bismuth oxide samples at 500°C. 3.3.2 Room temperature X-ray diffraction pattern of La3.6Y6.4 before and after aging at 500°C for 60 hours. 3.3.3 Decay in conductivity (Eq. 3.2.1) vs. average dopant ionic radius for 10% La and Y doped Bi2O3 at 500°C. Plotted on the second axis is the rhombohedral/cubic phase percentage determined by room temperature XRD and Rietveld refinement. 3.3.4 Conductivity vs. relative dopant ionic radius in Bi2O3. 3.3.5 Conductivity vs. total dopant concentration in rhombohedral Bi2O3. 3.3.6 High temperature X-ray diffraction of La5.1Y1.4. 3.3.7 Conductivity as a function of aging time for the La5.1Y1.4 rhombohedral sample. 3.3.8 Conductivity of multiple rhombohedral Bi2O3 samples measured at 500°C. 3.3.9 a Voltage as function of “anode” PO2 supplied to the La5.1Y1.4 symmetric cell at 500°C. 3.3.9 b EIS plot of the La5.1Y1.4 symmetric cell at 500°C. 3.3.10 Arrhenius plot of multiple SOFC electrolytes from 550°C to 400°C. Adapted from7. 3.3.11 Conductivity as function of aging time for La5.1Y1.4, ESB and DWSB as a function of aging time at 500°C. 3.3.12 Approximate phase stability window of rhombohedral Bi2O3 (dotted lines) given as function of total dopant concentration and average dopant ionic radius. 4.2.1 Room temperature XRD pattern of all 10% doped NASICON samples. 4.2.2 a SEM images of Al-doped NASICON 4.2.2 b SEM images of Fe-doped NASICON 4.2.2 c SEM images of Y-doped NASICON 4.2.2 d SEM images of Co-doped NASICON 4.2.2 e SEM images of Ni-doped NASICON 4.2.2 f SEM images of Zn-doped NASICON 4.2.3 XRD patterns of Na3Zr2Si2PO12 from 30°C up to 300°C. 4.2.4 Geometric representation of the monoclinic unit cell aligning with the rhombohedral unit cell. 4.2.5 a Distortion of the a-base observed from the 001 direction. 4.2.5 b Shear deformation of the unit cell observed from the 010 direction. vii 4.2.6 a Shear deformation (Eq. 4.2.1) plotted as a function of temperature for all doped and undoped NASICON. 4.2.6 b Distortion of the unit cell a-base (Eq. 4.2.2) plotted as a function of temperature for all doped and undoped NASICON. 4.3.1 a EIS plot of undoped and divalent substituted NASICON at 25°C. 4.3.1 b EIS plot of undoped and trivalent substituted NASICON at 25°C. 4.3.2 EIS plot of Al-doped NASICON at selected temperatures indicated on the right. 4.3.3 a Real part of the complex impedance for all doped NASICON at 25°C. 4.3.3 b Imaginary part of the complex impedance for all doped NASICON at 25°C. 4.3.4 Real and imaginary parts of the complex impedance for Al-doped NASICON at selected temperatures. 4.3.5 DC polarization of Co-doped NASICON at 25°C represented as current vs. time. 4.3.6 Arrhenius plot of all NASICON samples. 4.3.7 Bulk conductivity of doped NASICON vs. the ionic radius of the dopant. 4.4.1 Room temperature X-ray diffraction patterns for Zn-doped NASICON. 4.4.2 Room temperature bulk conductivity as a function of dopant concentration for Zn, Co, and Ni-doped NASICON. 4.4.3 Arrhenius plot of bulk conductivity for 10% Zn-doped, 20% Zn-doped, and undoped NASICON. 4.4.4 Shear deformation of the rhombohedral unit cell of NASICON (Eq. 4.2.2) as function of temperature for multiple Zn-doped samples. 4.4.5 Room temperature conductivity of 20% Zn-doped NASICON as a function of sintering time. 4.4.6 Conductivity vs. temperature of multiple sodium ion conductors. Adapted from15. 5.1.1 EIS plot of the La5.1Y1.4 symmetric cell at 500°C. 5.1.2 a ASR of a La5.1Y1.4-LSM cathode and an ESB-LSM cathode as a function of aging time at 500°C on a symmetric cell. 5.1.2 b The total change in ASR of the La5.1Y1.4-LSM cathode as a function of aging time at 500°C. 5.1.3 a ASR of a La5.1Y1.4-LSM composite cathode at 500°C as a function of the volume% of LSM in the cathode. 5.1.3 b ASR of the La5.1Y1.4-LSM cathode as a function of aging time at 500°C. 5.1.4 SEM image of the La5.1Y1.4 symmetric cell. 5.1.5 a Ohmic and non-ohmic ASR as a function of aging time at 500°C. 5.1.5 b Peak power density and open circuit voltage as a function of aging time at 500°C. 5.2.1 SEM image of the 20% Zn-doped NASICON triple layer. 5.2.2 a Nyquist plot of 20% Zn-doped NASICON and undoped NASICON measured at room temperature. 5.2.2 b The ASR of 20% Zn-doped NASICON and undoped NASICON determined by EIS fitting. viii List of Abbreviations AC alternating current AFL anode functional layer ASL anode support layer ASR area specific resistance BBP butyl benzyl phthalate BCC body centered cubic DSC differential scanning calorimetry DWSB 8% Dy-doped 4% Dy-doped Bi2O3 EIS Electrochemical Impedance Spectroscopy ESB 20% Er-doped Bi2O3 FCC face centered cubic FRA frequency response analyzer GDC 10% Gd-doped Ceria LSM La0.8Sr0.2MnO3 MD molecular dynamics OCP open circuit potential OCV open circuit voltage PVB polyvinyl butyral SCCM standard cubic centimeters per second SEM scanning electron microscope SOFC solid oxide fuel cell Tc critical temperature TZP tetragonal polycrystalline zirconia WDS wavelength-dispersive spectroscopy YSZ yttria stabilized zirconia XRD X-ray diffraction ix Chapter 1: Introduction 1.1 General Electrochemistry Introduction While there is still a dependence on fossil fuels for global energy requirements, a shift in recent years has increased the demand for alternative energy sources. With this transition, electrochemical devices have carved out a vital role in the changing energy landscape. In particular, rechargeable batteries have become a viable option for grid-scale storage and this promotes the practicality of both wind and solar energy. Additionally, solid oxide fuel cells (SOFCs) present a path toward clean hydrogen energy, while still retaining functionality with traditional fossil fuels such as natural gas. Advances in SOFC and battery technology will be very significant as the energy landscape continues to evolve and grow. Even though different mechanisms drive the operation of batteries and SOFCs, there are key fundamentals between the two electrochemical devices that are the same. The oxidation reaction occurs at the anode and provides electrons for the external circuit, ions travel across the electron blocking electrolyte between electrodes, and the electrons reconvene at the cathode where the reduction reaction occurs. An SOFC operates fundamentally different than a battery because it does not act as an energy storage device, but rather converts incoming fuel directly into electrical power. Also for an SOFC to function it must operate at high temperatures, whereas many batteries operate at room temperature. The other key difference between the two electrochemical devices is that gaseous fuel is used to generate power in an SOFC, while a battery relies 1 on the oxidation of a self-contained anode to provide electoral output. Ultimately both the cathode and anode of an SOFC must have the ability to catalyze an incoming gas into its component ions and oxygen ions are conducted through the solid electrolyte to balance the electrons traveling through the external circuit. Battery electrodes however, provide all of the power within the packaging and don’t need to rely on gas catalysis or incoming fuel. Furthermore, there are a host of battery chemistries and designs that prevail, as well as both liquid and solid electrolytes. 1.2 Fundamentals of Solid Oxide Fuel Cells SOFCs are a promising technology that provide high energy density and a low overall environmental impact16–18. Furthermore high temperature SOFCs provide higher operating efficiencies than combustion engines19,20. A solid oxide fuel cell is comprised of three main components. The anode catalyzes the incoming hydrogen gas and the oxidation reaction involves the formation of water and release of electrons as described in Eq. 1.2.1 H −22 + O  H O + 2e' 2 (Eq. 1.2.1) The cathode is the site of oxygen reduction and is described by Eq. 1.2.2 ½ O + 2e' O−2 2  (Eq. 1.2.2) The last main component of a solid oxide fuel cells is the electrolyte. The electrolyte transports the oxygen ions from the cathode to the anode where they combine with hydrogen to form water. Electrons flow from anode to cathode along an external circuit. The whole cell can be described by Eq. 1.2.3 H + ½ O H O 2 2  2 (Eq. 1.2.3) 2 The diagram detailing a working SOFC is illustrated in Fig. 1.2.1. O2 in Cathode e' flow O-2 -2 Electrolyte O Anode H2 in H2O out H2 in outin]]out Fig. 1.2.1. Diagram detailing the operation of a solid oxide fuel cell with gas and components labeled. The fundamental performance of a solid oxide fuel cell is also dependent on the open circuit voltage (OCV) achieved by the cell. The maximum theoretical OCV is determined by the Nernst equation, which is expressed in Eq. 1.2.4 𝑅𝑇 (𝑃𝐻 𝑂) 𝐸 = 𝐸° − ln ( 2 1 ) (Eq. 1.2.4) 2𝐹 (𝑃𝐻2) • (𝑃𝑂2)2 E° is the standard cell potential, R is the gas constant, T is the temperature, F is faradays constant, PH O is the partial pressure of water vapor on the anode side, PH is the partial 2 2 pressure of hydrogen on the anode side, and PO is the partial pressure of oxygen on the 2 cathode side. Although it is possible to reach the theoretical voltage when the cell is not operating under a load, the actual performance of the cell is dictated by polarization losses. The three main overpotential losses are defined as activation polarization, ohmic 3 polarization and concentration polarization. Activation polarization is related to the electrode response and is a result of decreased charge transfer. Ohmic polarization is mainly caused by resistive loses in the electrolyte layer, but is also caused by contact resistance and resistances in the electrodes. Finally, concentration polarization is related to a reduced transport of gaseous elements through the porous electrodes. Overall the actual voltage of an operating cell is detailed in Eq. 1.2.5. V = Vo – ηactivation – ηohmic – ηconcentration (Eq. 1.2.5) Vo is the theoretical voltage, ηactivation is the activation polarization, ηohmic is the ohmic polarization, and ηconcentration is the concentration polarization. While SOFCs are a promising technology, there are still issues with the system that hold it back from being a mainstream energy conversion device. One of the main limitations associated with traditional SOFCs is the high temperature of operation. Lowering the operation temperature to approximately 600°C would help reduce costs and improve overall efficiency10,21. Ceramic interconnects and insulation is needed for elevated temperatures, which greatly increases the cost over interconnects made of stainless steel18,22. Decreasing the operating temperature of solid oxide fuels hinges on electrolyte materials with higher conductivity. Yttria stabilized zirconia (YSZ) is a traditional electrolyte material that does not provide high enough ionic conductivity at intermediate temperature to allow for high performance in SOFC23–25. It is possible to reduce the thickness of the YSZ layer, but processing temperatures and lower economy of scale manufacturing techniques are required. Ultimately electrolyte materials with higher ionic conductivity are necessary for lowering the operating temperature of SOFCs. 4 1.3 Bismuth Oxide as an SOFC Electrolyte At elevated temperatures, Bi2O3 exhibits the highest known oxygen ion conductivity of any material3,26. From approximately 730°C up to its melting point, bismuth oxide exists in the δ-cubic phase. The δ-cubic phase of Bi2O3 has a defect fluorite structure with 25% of its oxygen sites are vacant. The unit cell of δ-cubic Bi2O3 is face centered cubic (FCC) with cations at the FCC sites and anions at the tetrahedral positions. The intrinsic vacancies of the material, the loose bonding of bismuth with oxygen, and the overall disorder of the structure allows for high oxygen ion mobility. Below 730°C pure Bi2O3 exists as the monoclinic α-phase. Cooling the δ-cubic Bi2O3 below 730°C can cause a transition to the metastable tetragonal β-phase or body centered cubic (BCC) γ-phase. The monoclinic phase, tetragonal phase, and BCC phase all have much lower conductivity than the δ-cubic phase of Bi2O3 as illustrated in Fig. 1.3.1a. The δ-cubic lattice is presented in Fig. 1.3.1b. a. b. Fig 1.3.1 a. Phase diagram and conductivity of pure Bi2O3 as a function of temperature. Fig. is taken from3. b. The δ-cubic Bi2O3 lattice with the cations, anions, and vacancies labeled on the diagram. Fig is taken from4. 5 Although δ-cubic Bi2O3 exhibits exceptional conductivity at elevated temperature, the transition to lower symmetry phases makes pure bismuth oxide unsuitable for use as a solid electrolyte. The conductivity of the monoclinic, tetragonal, and BCC phase of bismuth oxide is too low to achieve high performance in an SOFC. Thus stabilizing the highly conductive δ-cubic Bi2O3 to lower temperatures has been a priority for many researchers. It has been found that doping the bismuth oxide lattice with a sufficient amount of lanthanide cations actually preserves the cubic phase down to room temperature5,6,27. Ultimately it was found that the highest conductivity of doped bismuth oxide corresponded to the minimum required cation substitution for stabilizing the highly conductive FCC phase. The relationship between minimum required doping and the conductivity of doped bismuth oxide is presented in Fig 1.3.2 a. and b. a. b. Fig 1.3.2 a. Minimum dopant required to stabilize the FCC phase of Bi2O3 vs. the ionic radius of the dopant. Fig. is taken from5. b. Conductivity of cubic Bi2O3 as a function of dopant and ionic radius. Fig. is taken from6. 6 Overall Er+3 provided the minimum necessary dopant concentration at 20% to stabilize the cubic phase of Bi2O3, and consequently exhibited the highest conductivity. As the dopant ionic radius diverged from approximately 1 Å, the amount of cation substitution required to stabilize the cubic phase increased. Furthermore, as the amount of dopant concentration increased the resulting conductivity of the bismuth oxide material decreased. Ultimately a reduction in the dopant concentration in bismuth oxide leads to less defect association within the lattice as well as a higher lattice polarizability, ultimately generating higher oxygen ion mobility. While 20% Er-doped Bi2O3 (ESB) exhibited the highest conductivity of singly doped bismuth oxide, recent strategies have been employed to increase the conductivity even further. It has been found that doping the Bi2O3 lattice with two different cations provides a synergistic effect that allows for a lower overall dopant concentration to effectively stabilize the cubic phase8,28. By incorporating two different dopants into the lattice, the entropy of the system increases due to an increase in entropy from mixing. Wachsman et al. successfully stabilized the cubic phase of bismuth oxide with a total of 12% cation substitution, thereby creating a material with the highest oxygen ion conductivity to date8,28. The highly conductive material, 8% Dy-doped and 4% W- doped Bi2O3 (DWSB), is shown on an Arrhenius plot versus other electrolyte materials in Fig. 1.3.3. The higher conductivity of double doped DWSB over singly doped ESB is an outcome of the decreased amount of dopant. Overall, reducing the cation substitution in Bi2O3 will bring the lattice closer to the pure δ-cubic phase and improve the ionic conductivity. 7 Fig. 1.3.3. Arrhenius plot of multiple oxygen ion conducting electrolytes. Fig. is taken from20. While the FCC cubic phase can be stabilized and conductivity maximized in both single doped and double doped bismuth oxide systems, there is still issues with these materials at intermediate temperatures. When aged below approximately 600°C the cubic bismuth oxide undergoes a process of conductivity degradation. The degradation is general caused by two separate mechanisms. One cause of conductivity decay in cubic Bi2O3 is a result of a phase transition from the cubic phase to the rhombohedral phase. It has been observed that yttria stabilized δ-cubic Bi2O3 as well as other rare earth stabilized Bi2O2 undergo a phase transformation to the rhombohedral phase when annealed for a period of time below 700°C29,30. The transformation from the highly conductive δ-cubic phase to the rhombohedral phase 8 results in a drop in conductivity. Therefore, some compositions of cubic bismuth oxide are only metastable below approximately 700°C. Another cause of conductivity degradation in δ-cubic phase of Bi2O3 below approximately 600°C is an ordering phenomenon in the oxygen ion sublattice. Many methods of characterization, including X-ray diffraction (XRD) and neutron diffraction have indicated that there is long range ordering of the oxygen vacancies1,31,32. In particular the oxygen vacancies align along the <111> direction with a displacement of the oxygen ions from the normal 8c position to the 32f sites. This long-range ordering of the oxygen vacancies causes a large drop in conductivity as a result of fewer jump directions for the oxygen ions. Table 1.3.1 describes the difference between oxygen ions in the ordered and disordered cubic Bi2O3 structure in terms of the pre-exponential factor in the conductivity equation. Pre-exponential Factor Disordered Bi2O3 Ordered Bi2O3 Contributions n 6 1 ze 1.5e 2e ci 1 0.33 Table 1.3.1. Key pre-exponential factor contributions for the temperature dependence of conductivity on disordered and ordered cubic Bi2O3. Table Adapted from 1. The number of equivalent jump directions (n) is much less for the ordered structure than the disordered structure, which greatly reduces the contribution of pre- exponential factor in the conductivity equation. The equation governing the pre- exponential factor is detailed in Eq. 1.3.1 9 vc A = nλ2 (ze)2 i (Eq. 1.3.1) 6vk For this equation, n is equivalent jump directions for mobility, λ is jump distance, ze is effective charge of the charge carrier, ci is concentration of charge carriers, and v is the volume of the unit cell. The pre-exponential factor fits into the temperature dependent conductivity equation as expressed in Eq. 1.3.2 𝐸 𝜎𝑇 = 𝐴 exp (− 𝐴) (Eq.1.3.2) 𝑘𝑇 For this equation, σ is conductivity, T is temperature, A is the pre-exponential factor, Ea is activation energy, and k is the Boltzmann constant. Although the ordering phenomena can result in over an order of magnitude of conductivity degradation in the cubic bismuth oxide system, there have been techniques for decreasing the degree of degradation. For example, increasing the dopant concentration in bismuth oxide improves the stability of conductivity. The ordering phenomena that is responsible for the degradation is suppressed in higher dopant regimes. Fig. 1.3.4 illustrates the aging properties of a range of bismuth oxide samples at 500°C. From the figure it is clear that the Bi2O3 sample with greater substitution displayed less overall degradation. The sample with 21% cation substitution (14D7WSB) had approximately an order of magnitude decrease in conductivity while the 12% cation substitution sample (8W4WSB) underwent a decay of almost two orders of magnitude. While increasing the dopant concentration suppresses the conductivity decay, it ultimately drops the overall conductivity. The benefit of δ-cubic Bi2O3 is the superior conductivity to other materials, and thus maintaining high conductivity should be paramount. 10 Fig. 1.3.4. Conductivity as a function of time for multiple bismuth oxide samples at 500°C. The number represents the amount of cation substitution where D is Dysprosium, W is tungsten, and E is erbium. Fig. is taken from8. As discussed earlier, the tetragonal phase, BCC, and monoclinic phase are all natural polymorphs of pure bismuth oxide that exhibit much lower conductivity than the FCC cubic phase. The rhombohedral phase was not presented in this list because only doped bismuth oxide forms this phase. In general, there has not been extensive research into the rhombohedral phase. The lack of investigation into rhombohedral Bi2O3 is most likely as a result of its lower conductivity when compared to the cubic phase. Comparatively the rhombohedral phase is about ten times less conductive than the δ-cubic phase of bismuth oxide5. Fig. 1.3.5 illustrates the higher conductivity of the cubic phase compared to the rhombohedral phase for the same sample. 11 Fig. 1.3.5. Conductivity vs temperature for 25% Dy-doped Bi2O3. The cubic phase is represented by the closed circles and the rhombohedral phase is the open circles. Fig. taken from5. Besides having a distinct lattice, the rhombohedral phase is also different from the cubic phase in how it is produced. In general, the rhombohedral phase of bismuth oxide is created by doping the Bi2O3 lattice with large trivalent cations such as Nd, and La, while the cubic phase is made by doping the lattice with smaller cations such Er and Y. The rhombohedral phase of Bi2O3 has been mapped out by previous authors, but has generally been limited to larger dopant radii33–35. Investigations into the lower dopant regime could improve the conductivity of rhombohedral bismuth oxide and make it viable electrolyte for SOFCs at intermediate temperatures. Although the cubic phase has higher conductivity than other polymorphs of bismuth oxide, the conductivity of Bi2O3 decreases with increasing cation substitution. Since the cubic phase can only be stabilized by doping the bismuth site, it is important 12 to understand the role that the dopant plays on the conductivity. It has been shown that the polarizability of the substituted cation plays a major role in the conductivity and the stability of δ-cubic Bi2O3 material 1,36. As the polarizability of the dopant cation increases, the time constant for the disorder-order transition also increases. This means that increasing the polarizability of the lattice promotes less degradation of conductivity in bismuth oxide materials. Furthermore, the polarizability of the Bi2O3 lattice effects the oxygen ion mobility. Using molecular-dynamics (MD) simulation, Aidhy et. al. demonstrated that oxygen ion diffusion is dramatically reduced in a non-polarizable system9. The effect of polarizability on the MD simulation of oxygen diffusion in Bi2O3 is shown in Fig. 1.3.6. Fig. 1.3.6. Time dependence of mean-square displacement (MSD) of oxygen diffusion for both polarizable and non-polarizable δ-cubic Bi2O3 systems. Fig. taken from 9. As illustrated by the figure, there is almost no oxygen diffusion in the bismuth oxide lattice when the Bi+3 cation is non-polarizable. Because bismuth is highly 13 polarizable, it greatly adds to the oxygen transport. The lone pair of electrons on the Bi+3 cation is displaced from the O−2 anion causing a charge separation that decreases Coulombic repulsion and enhances diffusion. Thus, for all bismuth oxide systems, the conductivity can be maximized when the concentration of the bismuth cation is maximized due to the increase in polarizability. Although the conductivity of Bi2O3 is exceptional, it cannot be used as a single electrolyte in solid oxide fuel cells because of decomposition at low PO2 ranges 37–39. The bismuth oxide material is therefore used in bilayer design with a YSZ or GDC electrolyte to not only protect the Bi2O3 from decomposition, but also boost the performance of the cell10,37,40,41. A stable electrolyte layer such as YSZ or GDC on the fuel side of an SOFC will incur a voltage drop across it, thus lowering the effective PO2 on the bismuth oxide portion. Fig. 1.3.7 depicts a GDC layer protecting an ESB electrolyte from a low PO2 environment. Fig. 1.3.7. The effect of relative thickness and PO2 on ESB stability in an ESB/GDC electrolyte bilayer design. Fig. taken from10. 14 The bilayer electrolyte design is interdependent. The GDC layer protects the bismuth oxide layer from decomposing, and the bismuth oxide electrolyte increases performance. With higher conductivity, the Bi2O3 layer can be increased while the traditional electrolyte layer shortened, ultimately decreasing the ohmic resistance of the cell and increasing the power density. Furthermore, the bismuth oxide layer can act as a blocking layer. GDC is not a pure ionic conductor at reducing conditions, and thus bismuth oxide can block leakage current and increase the operating voltage of an SOFC. Overall a bismuth oxide electrolyte must have high ionic conductivity, low electronic conductivity, and good structural and ionic stability at intermediate temperatures to be effective in a SOFC bilayer design. 1.4 Fundamentals of Solid State Sodium Batteries Inexpensive and reliable batteries for grid-scale storage is vital to facilitating the continued use of clean renewable energy. Alternative energy sources such as wind and solar require large energy storage devices to continue delivering electricity when the wind is not blowing, or the sun goes down. Utilizing batteries for energy storage at the grid level can ultimately improve reliability and stimulate renewable energy42,43 Specifically Sodium-ion batteries offer a viable, lower cost alternative to lithium-ion batteries for grid-level storage44–46. While sodium-ion batteries have shown promise, it is the all-solid-state sodium battery that offers the best of all worlds. The solid state battery is a safer, cheaper, and more reliable alternative to other battery chemistries47,48. While the energy density of an all-solid-state sodium battery is sufficient for many applications, the performance is largely dependent on the anode and cathode chemistry that is applied. Authors have 15 demonstrated rechargeable solid state sodium batteries operating with a sodium metal anode49–51 or a Na3V2(PO4)3 anode 47,48,52. Furthermore different cathode materials have been tested in solid state sodium batteries including sulfur49,50, Na3V2(PO4) 47,48,51,52 3 , and Na0.7CoO 53 2 . No matter the chemistry of a solid state sodium battery, the fundamental elements that govern its operation will be the same. Fig. 1.4.1 is a diagram that details the operation of an all-solid-state sodium battery. Cathode e' flow + Na+ Electrolyte Na Sodium Containing Anode Fig. 1.4.1. Diagram of a solid state sodium battery during operation. Flow of sodium ions and electrons are labeled. Inside the battery, the sodium anode will oxidize to sodium ions and the resulting electrons will flow through the external circuit. Simultaneously the sodium ions will diffuse through the electron blocking electrolyte to recombine with the electrons at the cathode. Overall the electrolyte is a very important component of the solid state battery. It must have high ionic conductivity, low electronic conductivity, and chemical stability with sodium. Ultimately increasing the ionic conductivity of a solid state electrolyte is especially important if the battery is going to be run at room temperature. 16 1.5 NASICON as a Solid State Sodium Battery Electrolyte One of the major hurdles in creating a room temperature solid state sodium battery is finding a solid electrolyte with high enough conductivity. NASICON (Na3Zr2Si2PO12) was selected as a blueprint for investigation by many authors due to its high conductivity50,54–56. While it has relatively high conductivity for a solid state electrolyte, NASICON still has lower conductivity compared to many liquid electrolyte counterparts57–60. Ultimately modifications to the NASICON structure could lead to an increase in the room temperature sodium ion conductivity, and make it a promising candidate for solid state sodium batteries. Na3Zr2Si2PO12 is a three-dimensional solid network composed of sodium sites nestled between interlocking silica and phosphorus tetrahedrons and zirconia octahedrons, and was originally developed by Hong and Goodenough as a fast sodium ion conductor61,62. The NASICON structure is depicted in Fig. 1.5.1. Fig. 1.5.1. Schematic of the Na3Zr2Si2PO12 structure. This includes the red spheres as oxygen, a mixture of PO4 and SiO4 tetrahedra, ZrO6 octahedra, and the transport path of Na+ indicated. Fig. taken from11. 17 The NASICON structure can vary greatly since it is comprised of four sodium sites that can have between 25% and 100% occupancy as well as interchangeable silicon and phosphorus sites. Therefore while the NASICON compound Na1+xZr2SixP3- xO12 (x=2) is regarded as having the highest conductivity, the lattice parameters and conductivity change significantly from x=0 to x=362,63. Not only can the sodium concentration as well as the Si to P ratio vary in NASICON, but it is also possible dope the structure. Some doping studies have been carried out to decrease the grain boundary resistance and thus boost the total conductivity of NASICON. It has been demonstrated that doping the zirconia site of Na3Zr2Si2PO12 with yttrium can improve the microstructure and density and consequently increase the total conductivity12,64. The improvement to the microstructure and conductivity of NASICON due to yttrium substitution is illustrated in Fig. 1.5.2 a., b., and c. a. NASICON c. b. NASICON TZP Fig. 1.5.2 a. SEM images of undoped NASICON. b. SEM images of NASICON doped with 3% Y2O3 using tetragonal polycrystalline zirconia (TZP). c. EIS plot of undoped and TZP doped NASICON. Fig. is taken from12. 18 It is clear from the figure that the NASICON-TZP has better necking between particles. This is particularly important for mitigating grain boundary resistance to improve the total conductivity. The EIS data of the undoped NASICON illustrates that the majority of the resistance across the sample was caused by the grain boundary component. Doping the lattice improved the microstructure, and thus the resistance of the NASICON was cut in half. Since much of the conductivity depends on the grain boundary resistance, it follows that many authors would focus on reducing the grain boundary contribution. Nevertheless, understanding the effect of doping on the bulk properties of NASICON is also a primary component of increasing the conductivity. Other authors have studied the impact of doping on the bulk properties of NASICON, but there has not been a general consensus on what component is most influential on the conductivity. It has been demonstrated that doping the zirconium site with larger radius trivalent cations in the NASICON structure NaZr2P3O12 produced an increase in conductivity65,66. This trend appears to be limited to the silicon free version of NASICON though. There is less consensus on the effectiveness of doping the zirconium site of the Na3Zr2Si2PO12 compound. The dopant cations radius and valence do not have a distinct impact on the NASICON conductivity. Depending on the substituted cation, both higher and lower valance can either increase or decrease the conductivity of NASICON while the ionic radius does not have a clear effect on the conductivity67,68. More recently though, it has been demonstrated that decreasing the ionic radius of an aliovalent dopant or increasing the concentration of sodium in the lattice can increase the conductivity of NASICON50,69. In both cases the increase of conductivity resulted from an opening of the conductivity pathway. The decreased 19 ionic radius of the substituted divalent cation increased the opening of the bottleneck that the sodium ion must pass through. Excess sodium also expanded the pathway that sodium ion is transported through. The bottleneck opening of NASICON is illustrated in Fig. 1.5.3. Fig. 1.5.3. Unit cell of NASICON on the left. It is comprised of green ZrO6 tetrahedra, blue SiO4 tetrahedra, purple SiO4 tetrahedra, red oxygen spheres, and purple, yellow, and white sodium ions. The bottleneck that sodium ions travel through is indicated on the right by triangle T1. Fig. is taken from69. Understanding the effect of doping on the phase of the material is also an important angle of study. In general the highly conductive Na3Zr2Si2PO12 solid solution has the rhombohedral R3c space group above approximately 160°C-167°C, but transitions to the monoclinic C2/c space group below this temperature13,70. The phase transition has a large impact on the mobility of sodium ions in NASICON. The higher symmetry rhombohedral phase has a lower activation energy barrier for sodium ion transport. Ultimately NASICON would exhibit higher conductivity at room temperature if the phase change could be bypassed due to the lower activation energy of the rhombohedral phase. The Arrhenius behavior of the rhombohedral and monoclinic phase of NASICON is illustrated in Fig. 1.5.3. 20 Fig. 1.5.4. Arrhenius behavior of NASICON over a large temperature region. The dotted line indicates where the phase transitions from rhombohedral to monoclinic. Fig. taken from13. Ultimately the phase transition in NASICON is both temperature dependent and reversible. In contrast, the Na3Zr2Si2PO12 crystal structure will permanently transform if the sodium, silicon and phosphorus stoichiometry is altered. The end members of Na1+xZr2SixP3-xO12 (i.e. x=0 and x=4) are rhombohedral at room temperature, but their lattice parameters are observed to change significantly as x approaches 2 with the lattice distorting back to the monoclinic crystal structure71. In fact, Colomban et al. demonstrated that incrementally increasing sodium by increasing the silicon to phosphorus ratio in Na1+xZr2SixP3-xO12 caused the critical temperature (Tc) of the phase transition to be continuously lowered until the rhombohedral phase was stabilized at room temperature72. While silicon and phosphorus substitution has been studied widely, zirconium substituted NASICON is notably less understood. 21 Chapter 2: Experimental Procedures 2.1 SOFC Electrolyte Bismuth Oxide Synthesis All bismuth oxide electrolyte materials were synthesized using a standard solid state synthesis route. There were a number of bismuth oxide materials produced by mixing raw materials together in a stoichiometric ratio. In all cases the Bi+3 cation in Bi2O3 was substituted with a dopant cation. Raw materials included Bi2O3 (Alfa Aesar 99.99%), La2O3 (Alfa Aesar 99.9%), Nd2O3 (Sigma Aldrich 99.9%), Gd2O3 (Alfa Aesar 99.9%), Dy2O3 (Alfa Aesar 99.9%), Er2O3 (Alfa Aesar 99.9%), Sm2O3 (Alfa Aesar 99.9%), Y2O3 (Alfa Aesar 99.9%), CaCO3 (Alfa Aesar 99.5%), and SrCO3 (Sigma Aldrich 99.9%). The raw materials were mixed together with ethanol and YSZ media to provide mixing and grinding. The raw materials were ball milled for twenty-four hours in ethanol. After twenty-four hours of ball milling, the slurry was passed through a sieve and into a metal pan where it was dried at 100°C. The resulting powder was placed in an alumina crucible and calcined at 800°C for 16 hours with a heating and cooling rate of 5°C per minute. After calcining was complete, the samples were removed from the crucibles and hand ground with a mortar and pestle. The hand ground powder was then ball milled in ethanol a second time for 24 hours. The slurry was dried again at 100°C and the resulting powder was pressed into pellets. The pellets were pressed in 10 mm steel die and a carver press. The disks were generally sintered at 800°C for 16 hours with a heating and cooling rate of 5°C per minute unless otherwise stated. 22 2.2 SOFC Symmetric Cell Fabrication and Cathode Synthesis Symmetric cells were fabricated to determine the effectiveness of bismuth oxide samples in a composite cathode design. The composite cathode was made by mixing Bi2O3 electrolytes produced by solid state reactions with an La0.8Sr0.2MnO3 (LSM, Fuel Cell Materials) cathode powder in ethanol. The ratio of cathode to electrolyte powder was varied and will be discussed later. The slurry was ball milled for twenty-four hours prior to making the cathode paste. After ball milling, the slurry was blended in a Thinky centrifugal mixer and the ethanol was slowly replaced with a vehicle for the composite cathode powder (ESL type 441 thinner). The cathode paste was ready for deposition after it reached an optimal consistency, which occurred when the ESL thinner and cathode reached approximately equal weight and all the ethanol was removed. The cathode paste was blade coated on to the face of a dense Bi2O3 electrolyte pellet and dried at 120°C for one hour. After drying was complete, the other pellet face was coated with cathode paste and dried at 120°C for an additional hour. The symmetric cell was fired under different conditions and will be discussed in more depth in Chapter 5. 2.3 SOFC Full Cell Fabrication Full SOFC button cells with a bilayer design were fabricated for testing. The cell is comprised of a porous Ni-GDC anode support layer (ASL), a Ni-GDC anode functional layer (AFL), a dense GDC electrolyte, a dense Bi2O3 electrolyte, and finally a Bi2O3-LSM composite cathode. A schematic of the full SOFC with approximately layer thicknesses is illustrated in Fig. 2.3.1. 23 20 µm Bi2O3-LSM Cathode 5 µm Bi2O3 Electrolyte 20 µm GDC Electrolyte Ni-GDC AFL 500µm Porous Ni-GDC ASL Fig. 2.3.1. Schematic showing the cross-section of an SOFC button cell with a bi-layer electrolyte design. The left side details the approximate thickness of each layer (not drawn to scale). The anode, AFL, and GDC electrolyte were all fabricated using a tape casting method. The raw materials consisted for these layers consisted of NiO (99.9% Alfa Aesar) and Ce0.9Gd0.1O2 (JT Baker). The appropriate amount of raw materials were ball milled for twenty four hours with toluene and ethanol as solvents, as well as menhaden fish oil (Tape Casting Warehouse). Next, a binder of polyvinyl butyral (PVB) and the plasticizer butyl benzyl phthalate (BBP) was added to the slurry and milled another twenty four hours. After milling, the slurry was de-aired in a vacuum chamber with additional ethanol to prevent a skin formation on the surface. The slurry was then poured on a Mylar sheet, which was pulled through a doctor blade set on a tape caster bed (Procast). Once this process was complete, the tape dried on the bed set at 120°C. Finally the complete tapes of the ASL, AFL, and GDC electrolyte were laminated together with a Carver hot press at 180°C and 180 psi. The resulting tape was presintered at 900°C for 2 hours and fully sintered at 1450°C for 4 hours. 24 After completing the anode half-cell, the Bi2O3 electrolyte and composite cathode layer needed to be added. The bismuth oxide electrolyte and composite cathode layer were deposited using a blade coat method rather than tape casting. The Bi2O3 electrolyte ink for blade coating was fabricated following the same steps for making the composite cathode ink that was described earlier. The electrolyte ink was blade coated on the anode half-cell and fired at 775°C for 2 hours to make a dense layer. The composite cathode ink was then blade coated on to the Bi2O3 layer and fired at 725°C for 2 hours. After the cathode finished sintering, the final SOFC button cell was ready for testing. 2.4 Sodium Battery NASICON Electrolyte Synthesis All NASICON samples were synthesized using conventional solid state procedures and consisted of an aliovalent cation substitution of zirconium. The nominal compositions were Na3Zr2Si2PO12, Na 3+ 3+ 3.2Zr1.8M0.2Si2PO12 (M=Al , Fe , Y3+), and Na3.4Zr1.8M0.2Si2PO12 (M=Co 2+, Ni2+, Zn2+). Additionally, the amount of divalent cation substitution was varied for each sample and had the general formula Na 2+ 2+ 2+3+2xZr2-xMxSi2PO12 (for x=0.1 0.2, 0.3, 0.4, 0.5 and M=Co , Ni , Zn ). The raw materials used in the synthesis included Na2CO3 (Sigma-Aldrich, 99.95-100.05%), nanocrystalline ZrO2 (Inframat Advanced Materials, 99.9%) amorphous SiO2 (Sigma-Aldrich, 99.80%), and Na2HPO4·7H2O (Sigma-Aldrich, 98.0-102.0%), were mixed with 2% excess phosphorus and sodium to compensate for volatility. The dopant raw materials were added to NASICON raw materials in a stoichiometric ratio and included Al2O3 (Inframat Advanced Materials, 99.99%), Fe2O3 (Sigma-Aldrich, 25 ≥99%), Y2O3 (Alfa Aesar, 99.9%), Co(NO3)2·6H2O (Carolina, 98.0-102.0%), NiO (J.T. Baker, ≥99%), and ZnO (J.T. Baker, ≥99%). The raw materials were ball milled with 8 mm YSZ media in isopropanol for twenty-four hours. After ball milling the slurry was sieved into a metal pan and dried at 100°C. The resulting powder was placed in an alumina crucible and calcined at 700°C for 12 hours. The calcined powders were ground with an agate mortar and pestle and ball milled in isopropanol for twenty-four hours. The slurry was sieved and dried again and a second calcination of the powder occurred at 900°C for 12 hours. A final grinding of the calcined powder with a mortar as well as 24-hour ball milling in isopropanol was carried out. After sieving and drying of the slurry, the calcined powder was pressed into pellets using a 10mm steel die and a metric ton of uniaxial force from a carver press. The pellets were sintered for 12 hours in alumina crucibles at a range of temperatures that will be discussed later. The heating and cooling rates for all calcining and sintering were 5°C/minute. 2.5 Electrochemical Impedance Spectroscopy Electrochemical Impedance Spectroscopy (EIS) is an important technique for measuring the performance of an electrochemical cell. Specifically, EIS was used to measure the impedance of the cell which was then translated to the conductivity. The impedance of a system is determined by Eq. 2.5.1 𝑉(𝑡) 𝑍 (𝜔) = (Eq. 2.5.1) 𝐼(𝑡) For this equation V is the AC voltage and I is the current. The impedance Z has a real (Z') and imaginary (Z'') component that are related by Eq. 2.5.2. 26 𝑍 (𝜔) = 𝑍′ + 𝑖𝑍′′ (Eq. 2.5.2) The real and imaginary parts of the impedance can be plotted on a Nyquist plot so that Z' is plotted on the x-axis and Z'' is on the y-axis. The Nyquist plot is composed of a high frequency capacitance arc related to the electrolyte and a low capacitance arc related to the electrodes. The real portion of the impedance Z' can be related to the resistance, R, by Eq. 2.5.3. 𝑍′ = 𝑅 (Eq. 2.5.3) Ultimately the resistance of the arc can be translated to electrical conductivity, σ, with Eq. 2.5.4. 𝐿 𝜎 = (Eq. 2.5.4) 𝐴 • 𝑅 In this equation, L is the total thickness of the species, A is the cross-sectional area, and R is the resistance measured by EIS. The conductivity of the electrolyte be expressed as S/cm while electrodes or full cells will be in terms of area specific resistance (ASR) or ohm/cm2. In practice it can be difficult to interpret raw EIS data. Non-perfect semicircles, test setup inductance, and wire resistances all add to potential reading error. Furthermore, the raw data itself can be difficult to model and interpret. Thus, all EIS data was modeled with equivalent circuit models and fit with Scribner Z-view software to obtain the best possible fits. The fit data was then compared to literature vales to confirm the components of the EIS arcs. In general, the calculated capacitance of the arcs could help identify the different regions of the EIS plot. Table 2.5.1 lists EIS phenomena associated with different capacitance values. 27 Capacitance (F) Phenomenon Responsible 10−12 Bulk 10−11 Minor, secondary phase 10−11 – 10−8 Grain boundary 10−9 – 10−7 Surface layer 10−7 – 10−5 Sample-electrode interface 10−4 Electrochemical reactions Table 2.5.1. Capacitance values measured by EIS and the possible interpretation. Table adapted from2. The conductivity of all electrolyte samples and the ASR of symmetric cells were all determined by EIS. A similar setup was used for the measurement of all the samples. An alumina tube with four bored holes was outfitted with four separate silver wires. Two of the wires were attached to silver mesh and acted as the working electrode and working sense. The other two wires were attached to silver and mesh and acted as the counter electrode and counters sense. The faces of the sample were covered with a conductive paste to act as a current collector and placed between the pieces of mesh. The alumina tube and sample were placed in a tube furnace and the thermocouple controlling the furnace was placed directly above the sample. The exact procedure was different for each sample. The Bi2O3 electrolyte was coated on both faces with Heraeus gold paste to act as a current collector. The gold paste was not fired at high temperature but rather sintered in situ at the maximum temperature that the electrolyte conductivity was measured (generally 700°C or 550°C). Resistance of the samples was determined from 28 fitting the impedance plot with Z-view software and the conductivity was calculated using Eq. 2.5.4. A Solartron 1260A frequency response analyzer (FRA) was used to measure the impedance of electrolyte samples under an AC voltage of 50 mV and a frequency range of 1 Mhz to 1 Hz. For aging of the electrolyte samples, a Solartron 1260A and a Solartron 1470E that acted as a multiplexer was employed to take hourly EIS measurements of samples as they aged. The bismuth oxide symmetric cells were measured much the same way that the electrolyte samples were measured. Heraeus gold paste was painted on both faces of the symmetric cell to act as a current collector and the resistance was measured with a Solartron 1260A FRA and Solartron 1470E multiplexer. The resistance was divided by the area of the two electrode areas to determine ASR. As opposed to the electrolyte, only the low frequency, non-ohmic portion of the EIS plot was used in ASR calculation. The electrode response is located at low frequency and low capacitances. The faces of the NASICON electrolytes were painted with platinum paste and fired at 700°C for one hour. The impedance sweep was from 5 Mhz to 1 Hz using 250 mV AC voltage. The bulk and total conductivity were deciphered from the EIS plot using Z-view software. 2.6 Transference Number Measurements The transference number of Bi2O3 was calculated using voltage measurements. The bismuth oxide sample was sealed between two gaseous environments with different oxygen concentrations. The open circuit potential (OCP) generated by the oxygen activity was measured and compared to the theoretical Nernst potential. To measure the OCP, silver paste was applied to both faces of a sintered Bi2O3 disk and 29 acted as electrodes for the measurements. The sample was sealed between two gas lines with a compressive seal made from Thermicullite 866 O-rings. The cell was tightened down with bolts to facilitate the seal around the O-rings. One gas line supplied air, while the other gas line supplied gas mixtures with a range of PO2 values. An oxygen sensor was used downstream to measure the exact PO2 being supplied to the sample. The OCP was measured using a Solartron 1455A FRA connected to a 1470E potentiostat. The transference number of NASICON was calculated using a DC polarization technique. DC measurements were performed on sintered pellets in air at 25°C. Lead wires from a Keithley 2400 Broad Purpose SourceMeter provided current to the blocking platinum electrodes on the sintered disks. Measurements were run until the current stabilized, and the electronic conductivity was calculated by normalizing the resistance to pellet thickness and area. 2.7 SOFC Button Cell Testing A Solartron 1400 CellTest system was used to perform galvanostatic measurements of cell OCP and power density, as well as EIS measurements for ASR. These measurements were performed on an SOFC button cell using hydrogen fuel on the anode side and air on the cathode side. The gas was controlled with mass flow controllers and set to 200 SCCM. The button cell was sealed on an alumina tube that supplied the hydrogen to isolate the two gas streams using a ceramic sealant (Ceramabond). The anode side was connected to the working and sense electrodes using silver wires that bonded to the cell with silver paste (ESL). The cathode side was connected to the counter and reference electrodes with silver wires adhered to the cell 30 with gold paste (Heraeus). Once the button cell was connected and sealed with Ceramabond, it was dried for 4 hours at room temperature. The setup was then moved to furnace for a final curing procedure of 94°C for 2 hours and 265°C for 2 hours. Finally, the cell was heated up to the maximum temperature of testing where the anode could be reduced. 2.8 X-ray Diffraction X-ray diffraction (XRD) was performed on crushed and ground powder. For all room temperature XRD, sintered pellets were ground with an agate mortar and pestle to obtain fine powder. The room temperature XRD was collected with a Bruker D8 X- ray Diffractometer using Cu Kα radiation. High temperature XRD studies were performed with a Bruker C2 Discover X-ray Diffractometer with Cu Kα radiation and 2D area detector. An Anton Paar DHS 1100 hot stage was used for sample heating. Approximately 50mg of each powdered sample was pressed into a 6mm pellet with 1,000 lbs. of uniaxial force and placed on a Pt foil for heating. Each sample was heated at a rate of 100 C/min and held for 1 minute at the destination temperature before the pattern was collected. The high temperature XRD data was fit using the Le Bail method with the software program TOPAS (Bruker AXS, Karlsruhe, Germany) to obtain lattice parameters. The room temperature lattice parameters were extracted from room temperature XRD patterns using Rietveld refinement on TOPAS. 2.9 Microscopy and Elemental Analysis A Hitachi SU-70 scanning electron microscope (SEM) was used for the imaging of all materials and the accelerating voltage was generally set at 2.0 keV. For imaging, 31 samples were placed on double sided carbon tape as a way to minimize charging. Elemental analysis of NASICON samples was performed using a JEOL 8900 electron probe microanalyzer that employed standard wavelength-dispersive spectroscopy (WDS). The microprobe settings included a beam diameter of 30 µm, an accelerating voltage of 15 kV with current of 1 nA, and the absorbed current was monitored to detect sample degradation. Above sample currents of 1 nA or a beam diameter below 30 µm, electron densities would stimulate sodium migration and generate inaccurate elemental readings. ZAF correction procedures corrected raw x-ray counts, and determined the composition. Oxygen content was not directly measured. The oxygen stoichiometry of undoped NASICON was applied to all samples and was included in the ZAF corrections. 2.10 Differential Scanning Calorimetry Differential Scanning calorimetry (DSC) measurements were performed on NASICON samples using a Perkin Elmer Pyris 1. Measurements were taken in the temperature range 100°C to 200°C and a ramp rate of 5°C per minute was used. Approximately 15 mg of powder for each sample was analyzed, and the critical temperature (Tc) corresponded to the endothermic peak of the measurements and was determined by the Pyris Thermal Analysis Ver. 7 software. 32 Chapter 3: Doping Bi2O3 to Improve Conductivity and Stability 3.1 Motivation Decreasing the operating temperature of an SOFC below 600°C is the main objective of this study. Because the conductivity of traditional electrolytes is not high enough to achieve high performance in cells operating at low temperatures, electrolytes with higher oxygen ion conductivities were considered. As discussed earlier, doped cubic bismuth oxide has very high ionic conductivity, but rather poor stability below approximately 600°C. Thus, doped bismuth oxide materials generally have sufficient conductivity at reduced temperatures, but experience a large decay in conductivity that drives cell performance down. Overall the objective of this work is to develop a Bi2O3 material that is stable in phase and conductivity below 600°C. Initially the cubic phase of bismuth oxide was studied and methods of reducing the degradation in conductivity was explored. Ultimately the rhombohedral phase of Bi2O3 was selected for its high stability and modifications of the material were examined to sufficiently improve the ionic conductivity. 3.2 Effects of Lanthanide Substitution on Cubic Bi2O3 As previous authors have shown, a double doping strategy was employed to minimize the amount of cation substitution required to stabilize the cubic phase. The dopant level was set at 12% and a range of cations were tested. For the initial study, a secondary cation with a smaller ionic radius was used at a concentration of 4% for 33 every sample. Niobium was selected as the secondary dopant due to the ionic radius being 0.74 Å. The primary dopant concentration was set at 8% and covered a range of rare earth elements including Dy, Gd, Sm, Nd, and La. This study was established to determine if there is an effect of dopant ionic radius on the aging rate of cubic bismuth oxide. The list of samples that were tested include Bi0.88Dy0.08Nb0.04O1.5 (D8N4), Bi0.88Gd0.08Nb0.04O1.5 (G8N4), Bi0.88Sm0.08Nb0.04O1.5 (S8N4), Bi0.88Nd0.08Nb0.04O1.5 (N8N4), and Bi0.88La0.08Nb0.04O1.5 (L8N4). All of these samples existed in the cubic phase at room temperature based on X-ray diffraction. The D8N4 sample showed particularly good conductivity over the range of temperatures tested as seen in Fig. 3.2.1. 0.0 D8N4 -0.5 0.691 eV -1.0 DWSB -1.5 1.445 eV -2.0 -2.5 -3.0 1.0 1.1 1.2 1.3 1.4 1.5 1.6 1000/T (1/K) Fig.3.2.1. Arrhenius behavior of D8N4 and DWSB. The activation energy of D8N4 is provided on the plot. DWSB is a reference material18. 34 log σ (S/cm) Like other doped cubic Bi2O3 electrolytes, the D84N4 sample had a high temperature and low temperature region present in the Arrhenius behavior. The high temperature region was associated with a disordered oxygen ion sublattice, while the low temperature region was associated with ordered state. The ordered state of cubic bismuth oxide has many fewer equivalent jump directions for oxygen ion transport, which greatly reduced the pre-exponential factor and thus activation energy of the mobile ions. The main issue attributed with the low temperature ordered state of D8N4 was the degradation in conductivity at 500°C. There is a large drop as the disordered state transitions to the ordered state. Fig. 3.2.2 illustrates the conductivity decay of D8N4 as a function of aging time at 500°C. -0.5 DWSB -1.0 -1.5 D8N4 -2.0 -2.5 -3.0 0 10 20 30 40 50 60 70 80 Time (hours) Fig. 3.2.2. Conductivity vs. time for D8N4 and DWSB at 500°C. DWSB is a reference material18. 35 log σ (S/cm) The other cubic bismuth oxide samples tested all showed similar behavior to the D8N4 sample. A high temperature activation energy attributed to the disordered state and a low temperature activation energy attributed to the ordered state was seen in all cases. Furthermore, the ordering of the oxygen vacancies caused a huge decay in conductivity as the samples were aged for an extended period of time at 500°C. Eq. 3.2.1 models the decay of conductivity. Δ log(σ) = log(σi) − log(σf) (Eq. 3.2.1) For this equation σi is initial conductivity and σf is conductivity measured after aging. The log of initial conductivity minus the log of final conductivity (conductivity after 70 hours of aging at 500°C) is graphed as a function of dopant ionic radius in Fig. 3.2.3. 1.8 1.7 1.6 1.5 1.4 1.3 1.2 1.1 1.0 0.92 0.94 0.96 0.98 1.00 1.02 1.04 1.06 Average Dopant Ionic Radius (Å) Fig. 3.2.3. Change in conductivity vs. average dopant radius of multiple cubic Bi2O3 samples. Change in conductivity is determined by Eq. 3.2.1 after seventy hours of aging at 500°C. 36 Δ log σ (S/cm) The cubic Bi2O3 samples in Fig. 3.2.3 include D8N4, G8N4, S8N4, N8N4, and L8N4 described earlier as well as Bi0.88Gd0.08Y0.04O1.5 (G8Y4), which was used to bump the dopant ionic radius above 1.04 Å. The trend in the data suggests that as the dopant ionic radius increases in cubic bismuth oxide, the stability of the conductivity also increases. The larger dopant ionic radius works to decreases the overall ordering effect and lessens the degradation of conductivity. The ionic radius was investigated further with a second system. Again, a co-doping strategy was utilized to study the effects of the dopant ionic radius on the conductivity of bismuth oxide. The dopant concentration was set at 10% to decrease the amount of substitution and obtain higher conductivity. For this study, lanthanum and yttrium were selected as dopants due to La+3 and Bi+3 having similar ionic radii, while Y+3 was selected due to its relatively small size. Rather than utilizing different cations to alter the ionic radius, the ratio of La and Y was varied to increase or decrease the average dopant radius. The relative ionic radius was an average of the concentration and ionic radius of La+3 = 1.16Å and Y+3 = 1.019Å over the 10% total dopant content. The ionic radius was taken from Shannon radii for a coordination eight cation. Two alternative highly conductive bismuth oxide samples were used to compare the performance of materials synthesized here. Both single doped Bi0.8Er0.2O1.5 (ESB) and double doped Bi0.88Dy0.08W0.04O1.5 (DWSB) were used as reference materials. Initially, four different samples were synthesized: Bi0.9La0.015Y0.085O1.5 (La1.5Y8.5), Bi0.9La0.022Y0.078O1.5 (La2.2Y7.8), Bi0.9La0.029Y0.071O1.5 (La2.9Y7.1), and Bi0.9La0.036Y0.064O1.5 (La3.6Y6.4). This naming scheme will be applied to all subsequent sample. The letters signify the dopant and the numbers signify the 37 percentage of cation substituted for Bi+3. The X-ray diffraction patterns of the doped samples are laid out in Fig. 3.2.4. Cubic La1.5Y8.5BO La2.2Y7.8BO La2.9Y7.1BO Rhombohedral La3.6Y6.4BO BO 20 25 30 35 40 45 50 55 60 2θ (Degrees) Fig. 3.2.4. X-ray diffraction patterns of four different double doped Bi2O3 samples. La1.5Y8.5 was purely cubic while La3.6Y6.4 was entirely rhombohedral. It was observed that as the relative ionic radius increased, the amount of rhombohedral phase present in the sample increased. La1.5Y8.5 had an average dopant radius of 1.04Å and existed in the cubic phase, while the La3.6Y6.4 sample was entirely rhombohedral with an average dopant radius of 1.07Å. La2.2Y7.8 had a very tiny impurity peak attributed to the rhombohedral phase while La2.9Y7.1 was clear mixture of the cubic and rhombohedral phase. 38 Intensity (A.U.) The Arrhenius behavior and stability of the cubic La1.5Y8.5 sample were examined first. Fig. 3.2.5 is the EIS data obtained from the La1.5Y8.5 sample over a range of temperatures. 15 700°C 650°C 600°C 10 550°C 500°C 5 0 -5 0 5 10 15 20 Z' (Ohm•cm) Fig. 3.2.5. EIS plot for La1.5Y8.5 from 700°C-500°C. The black lines are the fits of the impedance data. An equivalent circuit model used to fit the impedance arcs. The equivalent circuit was composed of a resistor in series with two separate resistor and CPE elements in parallel. For temperatures ≤500°C, an inductor was also used to model the EIS data, since inductive effects from the setup became prominent. The high frequency intercept corresponded to the total conductivity of the material with no resolvable grain boundary resistance detected. Meanwhile, the low frequency arcs were attributed to oxidation reduction reactions taking place at the gold current collectors/electrodes. The 39 -Z" (Ohm•cm) conductivity of La1.5Y8.5 at multiple temperatures was determined from the EIS arcs and plotted in Fig 3.2.6. Temperature (°C) 700 650 600 550 500 450 400 0.0 La1.5Y8.5 -0.5 -1.0 ESB DWSB -1.5 -2.0 -2.5 -3.0 1.0 1.1 1.2 1.3 1.4 1.5 1000 / T (1/K) Fig 3.2.6. Arrhenius plot of three cubic bismuth oxide samples. The 10% doped La1.5Y8.5 sample is compared to DWSB and ESB for reference. The DWSB and ESB data is taken from18. The high temperature conductivity of La1.5Y8.5 is higher than both the previously measured D8N4 and DWSB. This is particularly notable since DWSB has one of the highest known oxygen conductivities below 700°C. The higher conductivity of La1.5Y8.5 is most likely a result of the lowering the overall dopant concentration while still maintaining the cubic phase. Decreasing the amount of bismuth substitution in Bi2O3 increases the polarizability of the cations and thus the mobility of oxygen ions9. The down side of decreasing the dopant concentration is a notable increase in the 40 log σ (S/cm) rate of ordering. While the previous D8N4 sample exhibited a kink in the activation energy going from the high temperature region to the low temperature region, the La8.5Y1.5 sample shows a large drop in conductivity. The drop off in conductivity below 500°C is most likely a result of the anion sublattice transitioning from the disordered state to the ordered state. The disorder-order transition occurs so rapidly that there is large decay in conductivity just from short period of time the sample was cooled below 600°C. The stability at 500°C was also investigated. Fig 3.2.7 is a plot of La8.5Y1.5, ESB, and DWSB conductivity as a function of aging time at 500°C. Like both ESB and DWSB, La1.5Y8.5 exhibited considerable conductivity degradation after just sixty hours of aging at 500°C. -0.5 DWSB -1.0 -1.5 ESB -2.0 La1.5Y8.5 -2.5 -3.0 0 10 20 30 40 50 60 Time (hours) Fig. 3.2.7. Conductivity as function of aging time for La1.5Y8.5, DWSB and ESB at 500°C. The DWSB and ESB data is used as a reference and is taken from18. 41 log σ (S/cm) While the huge drop in the Arrhenius behavior and the rapid decay in aging conductivity could signify a phase change had occurred in the material, the XRD confirmed the cause was the ordering of oxygen sublattice. Fig. 3.2.8 is the room temperature X-ray diffraction pattern of the La1.5Y8.5 sample prior to aging and post aging at 500°C for sixty hours. 20 25 30 35 40 45 50 55 60 2θ (Degrees) Fig. 3.2.8. Room temperature X-ray diffraction pattern of La1.5Y8.5 before and after aging at 500°C for 60 hours. The black line is the initial XRD plot and the dashed line is after aging. Because the cubic phase was retained after aging, the transition from the disordered to ordered state was the primary cause of conductivity degradation in La1.5Y8.5. The drop in conductivity was more abrupt in the aging of La1.5Y8.5 as a result to faster kinetics in the disorder-order transition. As discussed earlier, the lower dopant concentration allows for more rapid ordering of the oxygen vacancies and a 42 Intensity (A.U) hence a more rapid degradation of conductivity. Ultimately the cubic phase of bismuth oxide could not be stabilized with cations large enough to stabilize the conductivity at lower temperatures. As the average dopant ionic radius increased, the rhombohedral phase in bismuth oxide began to form. Although stabilizing the cubic phase of bismuth oxide at high dopant regimes was not possible, the rhombohedral phase could be stabilized. Also, Zr+4 was also explored as a dopant due to its small size compared to Bi+3. In order to stabilize the cubic phase, the large La+3 cation and the small Zr+4 cation were both substituted into the Bi2O3 lattice. The room temperature XRD patterns of two La and Zr double doped Bi2O3 samples are given in Fig. 3.2.9. 20 25 30 35 40 45 50 55 60 2θ (Degrees) Fig. 3.2.9. Room temperature X-ray diffraction patterns of 10% La and Zr doped cubic Bi2O3 materials. The La2Zr2O7 impurity phase is indicated with triangles. The La7Zr3 composition possessed a relative ionic radius of 1.065Å and the 43 Intensity (A.U.) La5Zr5 composition had an ionic radius of 1.00Å. While there was a large variation in average dopant ionic radius, the La and Zr doped compounds produced the cubic phase. The main difference in phase, was at higher zirconium concentrations, there was a marked increase in La2Zr2O7 impurity phase formation. Consequently, a substantial secondary phase is formed within the La5Zr5 composition. Also the conductivity of both samples are measured. Overall, the La7Zr3 compound exhibited the highest conductivity of all Bi2O3 materials synthesized in this investigation. The Arrhenius behavior of the cubic La7Zr3 is compared to other oxygen ion conductors in Fig. 3.2.10. Temperature (°C) 700 650 600 550 500 450 400 7 6 5 4 3 2 1 0 0.0010 0.0011 0.0012 0.0013 0.0014 0.0015 1/T (K-1) Fig. 3.2.10. Arrhenius plot of multiple SOFC electrolytes from 700°C to 400°C. The high temperature activation energy of the cubic La7Zr3 sample is 0.520 eV and the low temperature activation energy is 0.847 eV. Plot adapted from 44 Ln [σT (S/cm∙K)] The conductivity of La7Zr3 compound exceeds all other oxygen ion conducting materials developed to date. The conductivity of La7Zr3 is over 30% higher than DWSB at 650°C, which previously held the record for highest oxygen ion conductivity8. In fact, the cubic La7Zr3 material is the fastest oxygen ion conducting material ever developed. The superior conductivity of the La7Zr3 electrolyte would permit further increase in Bi2O3 layer thickness and reduction of the less conductive electrolyte layer (GDC, SNDC, etc.) to minimize the ohmic ASR of an SOFC without sacrificing OCV73. Like DWSB, the high conductivity in La7Zr3 was an outcome of lowering the total dopant concentration in Bi2O3, while also retaining the highly conductive cubic phase. The conductivity of La7Zr3 was improved beyond DWSB by reducing the total dopant concentration to 10%, providing a larger concentration of highly polarizable Bi+3 in the lattice. Because the polarizability of bismuth is higher than the dopant cations, decreasing the total cation substitution increased the lattice polarizability, consequently allowing for greater oxygen ion mobility and facilitated anion diffusion9,74. Higher polarizability equates to larger cation charge separation and improved anion diffusion due to the reduction in coulombic repulsion that oxygen ions experience when moving through the saddle point of the conductivity pathway. The lanthanum cation has the largest ionic radius of all lanthanide series cations, and since the ionic radius is linearly related to polarizability, the large La+3 cation is more polarizable than the other rare earth elements9. The La7Zr3 compound contained over four times more lanthanum than the La1.5Y8.5 sample for an equivalent dopant ionic radii. 45 The La7Zr3 composition was also aged at multiple temperatures to determine its stability. First, La7Zr3 was aged at intermediate temperatures (650°C) due to the high stability of cubic Bi O at this temperature142 3 . The aging performance of La7Zr3 at 650°C (Fig. 3.2.11) illustrates the high stability of the electrolyte at intermediate temperature. 0.0 650°C La7Zr3 DWSB -0.5 -1.0 0 10 20 30 40 50 60 70 80 90 100 Time (hours) Fig. 3.2.11. Conductivity as function of aging time for La7Zr3 and DWSB at 650°C. The DWSB is plotted as a reference cubic bismuth oxide material14. At lower temperatures (<600°C), the conductivity of La7Zr3 is not as stable with time. While cubic Bi2O3’s generally display stable performance at intermediate temperature, degradation at lower temperatures is a problem for SOFCs operating below 650°C. In particular, anion ordering is most rapid in cubic Bi2O3 at 500°C 14. To ensure the Bi2O3 electrolytes were capable of operation at low temperatures, the sample was aged for one hundred hours at 500°C. Both ESB and DWSB exhibit a sharp decay 46 log σ (S/cm) in conductivity, due to the ordering of the oxygen ion sublattice14. The La7Zr3 compound meanwhile, had a much more gradual decay in conductivity as it was aged at 500°C (Fig 3.2.11). -0.5 500°C -1.0 La7Zr3 -1.5 -2.0 -2.5 DWSB ESB -3.0 0 10 20 30 40 50 60 70 80 90 100 Time (hours) Fig 3.2.11 Conductivity as function of aging time for La7Zr3, DWSB and ESB at 500°C. The DWSB and ESB samples are plotted as reference cubic oxide materials14. It has been reported that the disorder-order transition in Bi2O3 can be suppressed with aliovalent doping of Zr+4 cations75. Thus, La7Zr3 may undergo a much slower aging phenomenon than other cubic bismuth oxide materials doped with rare earth cations. The decline in conductivity though, was likely exacerbated by phase instability at 500°C. After a 100 hour hold at 500°C, there was a growth in a La2Zr2O7 secondary phase, which is illustrated by the room temperature XRD pattern of La7Zr3 before and after aging given in the supplemental section (Fig 3.2.12). The lower conductivity after 47 log σ (S/cm) aging can in part be attributed to La and Zr dopants leaving the Bi2O3 solid solution to form the La2Zr2O7 impurity phase. 20 25 30 35 40 45 50 55 60 2θ (Degrees) Fig 3.2.12. Room temperature X-ray diffraction pattern of cubic La7Zr3 before and after aging at 500°C for 100 hours. The black line is the initial XRD plot and the dashed line is after aging. The intensity of the secondary phase La2Zr2O7 peaks grew after aging. 3.3 Optimizing the Rhombohedral Bi2O3 Lattice for Maximum Conductivity As the La and Y doped Bi2O3 system was examined further, it became clear that the rhombohedral phase was a very promising electrolyte material for low temperature SOFCs. At a dopant concentration rate of 10%, the rhombohedral phase of bismuth oxide grew rapidly after the dopant radius was increased above approximately 1.05 Å. At a La and Y cation substitution rate of 3.6% and 6.4% respectively (La3.6Y6.4) the average dopant ionic radius was equal to approximately 1.07Å. At a dopant ionic radius of 1.07Å, XRD confirmed that the pure rhombohedral phase of Bi2O3 was present as 48 Intensity (A.U) shown in the previous Fig. 3.2.3. As opposed to the cubic phase, the rhombohedral phase of bismuth oxide does not undergo an ordering of the oxygen vacancies at lower temperatures. By avoiding the ordering phenomena, the rhombohedral structure does not exhibit a large drop in conductivity. The stable conductivity of the rhombohedral phase is demonstrated in Fig. 3.3.1 which is the aging of La3.6Y6.4 for sixty hours at 500°C compared to other cubic Bi2O3 samples. -0.5 DWSB -1.0 -1.5 La3.6Y6.4 -2.0 La1.5Y8.5 -2.5 ESB -3.0 0 10 20 30 40 50 60 Time (hours) Fig. 3.3.1. Conductivity as function of aging time for four bismuth oxide samples at 500°C. The DWSB and ESB data is used as a reference and is taken from18. It is clear that conductivity of the rhombohedral La3.6Y6.4 sample is much more stable than the cubic La1.5Y8.5, ESB, and DWSB samples. Since it does not have a disorder-order transition, the rhombohedral phase does not incur a drop in conductivity when aged at 500°C. Not only is the conductivity of La3.6Y6.4 stable, but the rhombohedral phase is also thermodynamically stable. After sixty hours of aging at 49 log σ (S/cm) 500°C, the rhombohedral phase of La3.6Y6.4 does not break down. Furthermore, there is no growth of secondary phases that occur after aging. Fig. 3.3.2 is the XRD pattern of La3.6Y6.4 before and after aging at 500°C for sixty hours. It is clear that the there is almost no change in the phase after aging. 20 25 30 35 40 45 50 55 60 2θ (Degrees) Fig. 3.3.2. Room temperature X-ray diffraction pattern of La3.6Y6.4 before and after aging at 500°C for 60 hours. The black line is the initial XRD plot and the dashed line is after aging. Overall the phase of Bi2O3 was highly dependent on the ionic radius of the dopant. As the ionic radius increased, the amount of rhombohedral phase also increased. Furthermore, as the amount of rhombohedral phase increased, the stability of conductivity at 500°C increased. The correlation between dopant ionic radius, the phase of the material, and conductivity degradation after aging was determined by Rietveld refinement and Eq. 3.2.1. Rietveld refinement was carried out to determine the proportion of cubic phase and rhombohedral phase in each sample. Additionally, 50 Intensity (A.U.) Eq. 3.2.1 was used to evaluate the amount of conductivity degradation occurred after aging the samples for sixty hours at 500°C. The relationship between ionic radius, phase, and stability of conductivity for Bi2O3 is plotted in Fig. 3.3.3. 1.8 100 1.6 1.4 80 1.2 60 1.0 0.8 40 0.6 0.4 20 0.2 0 0.0 1.04 1.05 1.06 1.07 Average Dopant Ionic Radius (Å) Fig. 3.3.3. Decay in conductivity (Eq. 3.2.1) vs. average dopant ionic radius for 10% La and Y doped Bi2O3 at 500°C. Plotted on the second axis is the rhombohedral/cubic phase percentage determined by room temperature XRD and Rietveld refinement. The rhombohedral phase exhibited much better stability, but lower initial conductivity when compared to the cubic phase. Therefore, increasing the conductivity of the rhombohedral phase would make it even more viable as an SOFC electrolyte. As with the La and Y double doped cubic bismuth oxide study, the ratio of La to Y was varied to study the effect of average dopant radius on the conductivity. With a relative dopant ionic radius of 1.07Å, La3.6Y6.4 had the smallest ratio of lanthanum to yttrium that could stabilize the rhombohedral phase. Increasing the amount of La+3 substitution and decreasing the amount of Y+3 substitution tested the upper limits of the 51 Δ log σ (S/cm) Percent Rhombohedral (%) rhombohedral phase stability. For a dopant concentration held constant at 10%, a relative dopant ionic radius from 1.07Å up to 1.14Å resulted in the formation of the rhombohedral phase of Bi2O3. Above 1.14Å, secondary phase formation occurred and caused a drop in conductivity. Below 1.07Å the formation of the cubic resulted in lower conductivity due to the ordering of the secondary cubic phase. Like La3.6Y6.4, all the La and Y double doped Bi2O3 samples in the rhombohedral phase had very stable conductivity for over sixty hours of aging time at 500°C. Overall the relative dopant ionic radius, and hence the lattice parameters of the rhombohedral Bi2O3 unit cell had very little impact on the conductivity of the material. The average dopant ionic radius and rhombohedral lattice volume vs conductivity is plotted in Fig. 3.3.4. -1.5 384 -1.6 383 -1.7 382 -1.8 381 -1.9 380 -2.0 379 1.06 1.08 1.10 1.12 1.14 Average Dopant Ionic Radius (Å) Fig. 3.3.4. Conductivity vs. relative dopant ionic radius in Bi2O3. Conductivity was measured after a two hour hold at 500°C. The dotted lines represent approximate stability window of the rhombohedral phase. The secondary vertical axis identifies the volume of the rhombohedral unit cell determined by Rietveld refinement. 52 log σ (S/cm) Rhombohedral Cell volume (Å3) While the lattice volume scales with dopant radius, the conductivity of the material is relatively constant over the entire stability window of the rhombohedral phase. Since the ionic radius of the dopant had very little effect on the conductivity of rhombohedral bismuth oxide, the total dopant concentration was lowered to improve performance. Like previous authors’ work to decrease bismuth substitution in cubic Bi2O3 to increase conductivity, the amount of dopant in the rhombohedral phase was similarly carried out. Fig. 3.3.5 illustrates the conductivity as a function of total dopant concentration in the rhombohedral Bi2O3 lattice. -1.5 -1.6 -1.7 -1.8 -1.9 5 6 7 8 9 10 11 Dopant Percentage (%) Fig. 3.3.5. Conductivity vs. total dopant concentration in rhombohedral Bi2O3. Conductivity was measured after a two hour hold at 500°C. The dashed line represents the approximate rhombohedral phase boundary determined by room temperature XRD. The solid line represents the linear fit of the data. Average dopant ionic radius is 1.12Å. From the figure, it is clear the conductivity of the La and Y double doped bismuth oxide material increased as the dopant concentration decreased. The linear 53 log σ (S/cm) trend of increasing conductivity with decreasing dopant concentration held true all the way up to the rhombohedral phase boundary. When the dopant level was dropped below approximately 6.5%, the tetragonal secondary phase began to form for the La and Y double doped Bi2O3 system with a relative ionic radius of 1.12 Å. The secondary phase formation resulted in a decrease of conductivity in the material. Thus, a maximum conductivity was recorded at a substitution rate of 6.5% which corresponded to the sample La4.7Y1.8. Less substitution of trivalent cations into the Bi2O3 network increased the concentration of the highly polarizable Bi+3 cation. The improvement in conductivity therefore resulted from the increased polarizability of the cation network. The lone pair of electrons on the Bi+3 cation in conjunction with its loose bonding with surrounding disordered oxygen ions makes a particularly favorable environment for fast O−2 transport. With the phase stability window mapped out for the low dopant regime of La and Y doped Bi2O3 rhombohedral phase, it was necessary to also examine the phase stability at elevated temperatures. High temperature X-ray diffraction was performed on La5.1Y1.4 to determine the phase stability of the rhombohedral phase at elevated temperatures. La5.1Y1.4 was chosen because of its high conductivity. The X-ray diffraction patterns were recorded from room temperature up to 700°C to determine if there were any phase transitions that occurred in the material. The plot of XRD patterns taken every ten degrees from 550°C to 600°C is laid out in Fig. 3.3.6. From this figure there is a sharp peak shift between 580°C and 590°C. By fitting the XRD patterns it was determined that the shift in peaks was a result of the low temperature rhombohedral phase transforming to the cubic phase. 54 600°C 590°C 580°C 570°C 560°C 550°C 26 27 28 29 30 31 32 2θ (Degrees) Fig. 3.3.6. High temperature X-ray diffraction of La5.1Y1.4. The temperature of each scan is indicated on the plot. The phase transition can be observed above 580°C. From room temperature up to approximately 580°C the La5.1Y1.4 sample exhibits the pure rhombohedral phase. Above 580°C the cubic phase forms and grows while the rhombohedral phase shrinks. Once the temperature reaches 700°C, the La5.1Y1.4 sample is entirely cubic. The phase transformation at high temperature is problematic for the low dopant rhombohedral phase. Since the cubic phase has issues with ordering, operating a fuel cell above the phase transition of the rhombohedral to cubic phase could cause instability. Since the phase transition occurs around 580°C, the rhombohedral phase should be maintained at or below 550°C. The conductivity of the La5.1Y1.4 sample was measured as a function of time for multiple temperatures and is plotted in Fig. 3.3.7. 55 Intensity (A.U.) 0.0 -0.5 600°C -1.0 550°C -1.5 500°C -2.0 400°C -2.5 0 10 20 30 40 50 60 70 80 90 100 Time (hours) Fig. 3.3.7. Conductivity as a function of aging time for the La5.1Y1.4 rhombohedral sample. The aging temperature is listed on the plot. While the conductivity of La5.1Y1.4 sample was stable from 400°C through 550°C, at 600°C there was a large decay in conductivity. The initial conductivity at 600°C was relatively high, but it quickly fell off. Because the cubic phase forms at 600°C, it boosts the initial conductivity of the material, but also causes degradation. Like the La1.5Y8.5 sample that was discussed earlier, the cubic phase present at 600°C in the La5.1Y1.4 undergoes an ordering of the oxygen ion sublattice and subsequent decay in conductivity. The low overall dopant concentration of La5.1Y1.4 leads to faster kinetics of ordering and thus a more rapid decay in conductivity than would be observed in ESB. Since the phase stability window of the rhombohedral phase has been mapped out in terms of dopant ionic radius, the total dopant concentration and the 56 log σ (S/Cm) temperature, it was next important to identify what type of cation substitution maximized the conductivity. While minimizing total dopant concentration is a key component of increasing ionic conductivity of rhombohedral Bi2O3, the type of substituted cation is also an important consideration. A variety of other dopants were selected to study their effect on the conductivity of rhombohedral bismuth oxide. For this investigation, the average dopant radius was set between 1.095Å and 1.12Å to remain sufficiently within the rhombohedral phase boundary. For this investigation a single cation was selected if it was between these bounds, otherwise yttrium was used as a secondary dopant to decrease the relative dopant radius. Furthermore, the dopant concentration was held constant at 8% to ensure high conductivity. Fig. 3.3.8 depicts the conductivity of the rhombohedral bismuth oxide material as a function of primary cation dopant. It can be seen from the figure that substitution with La+3 provides a peak in conductivity. The two lanthanum double doped materials exhibited higher conductivity than the calcium, strontium, neodymium, or samarium doping. The peak in conductivity is likely a product of lanthanum possessing the largest ionic radius of the trivalent cations. Other authors have observed that substituting larger cations in to the rhombohedral bismuth oxide framework increases conductivity33,35. The larger ionic radius widens the conduction pathway and allows for greater oxygen ion mobility. The investigation also demonstrated that trivalent cation substituted Bi2O3 samples had greater conductivity than the divalent cation substitutions. The lower conductivity of the divalent substitution may have been caused by alterations in the defect structure. 57 -1.4 Ca+2 Sr+2 La+3 Nd+3 Sm+3 -1.6 La5.7Y2.3 Nd8 La6Er2 Nd4Sm4 -1.8 Sr3.2Y4.8 -2.0 Ca8 -2.2 Increasing +2 Increasing +3 Cation Ionic Cation Ionic Radius Radius Fig. 3.3.8. Conductivity of multiple rhombohedral Bi2O3 samples measured at 500°C. Divalent cation substituted Bi2O3 samples are located on the left side and trivalent cations substitution is on the right. The ionic radius of the primary dopant increases toward the center. Each sample is labeled on the plot. Doping Ca+2 in to the Bi+3 site would result in additional negative charge from the aliovalent substitution. The charge imbalance would need to be compensated with additional positive charge, such as decreased oxygen vacancies. A reduction in oxygen vacancies would bring about fewer jumping sites for oxygen ions and cause a decrease in ionic conductivity. Ultimately using lanthanum in a double doping scheme allowed for higher conductivity in a low dopant regime. With an ionic radius of 1.16Å, La+3 was too large to stabilize the rhombohedral phase at low dopant regimes. Substituting a smaller secondary cation into the lanthanum doped Bi2O3 lattice resulted in a lower relative dopant ionic radius. Thus, a La-doped rhombohedral phase could form at lower 58 log σ (S/cm) dopant concentrations than would be possible otherwise due to double doping. Thus, the conductivity of rhombohedral bismuth oxide was maximized by decreasing the total cation substitution as well as widening the conduction pathways by doping large trivalent cations. Ultimately the La5.1Y1.4 sample exhibited the highest conductivity of any rhombohedral Bi2O3 sample. The La5.1Y1.4 contained both the large cation La+3 and a low total dopant concentration. The La and Y double doped Bi2O3 samples exhibited the highest conductivity of all rhombohedral samples tested. To ensure this material would be an appropriate ion conducting electrolyte, the ratio of the ionic conduction to electronic conduction was measured. The electrical potential generated from an oxygen ion conductor in a two gas environment was used to determine the ionic transference number. The La5.1Y1.4 sample was made into a disc with silver electrodes and sealed between two different PO2 environments. The symmetric cell was held for an hour at each condition before the OCP was measured and EIS was performed. The ionic transference number was calculated using Eq. 3.3.1, as detailed by other authors76. R V T b OCi = 1 − (1 − ) (Eq. 3.3.1) RT VN In this equation Rb = bulk resistance, RT = total resistance, VOC = open circuit potential, and VN = Nernst potential. The voltage measurement as a function of PO2 for La5.1Y1.4 is given in Fig. 3.3.9 a and the impedance of the cell taken at different PO2 values is plotted in Fig. 3.3.9 b. All measurements were recorded at 500°C. Overall the La5.1Y1.4 sample exhibited an average ionic transference number greater than 0.997 across the entire PO2 range that was measured. 59 120 a. 100 80 60 40 20 0 -20 -40 -3.5 -3.0 -2.5 -2.0 -1.5 -1.0 -0.5 0.0 0.5 Log PO2 (atm) 70 b. 60 50 40 30 20 0.0664% O2 10 3.017% O2 0 12.074% O2 0.384% O2 99.75% O2 -10 0 10 20 30 40 50 60 70 80 Z' (Ohm) Fig. 3.3.9 a. Voltage as function of “anode” PO2 supplied to the La5.1Y1.4 symmetric cell at 500°C. The “cathode” side was steadily supplied with air. b. EIS plot of the La5.1Y1.4 symmetric cell at 500°C. The atmosphere suppled to the “anode” side is indicated by the oxygen concentration on the plot. The “cathode” side was steadily supplied with air. 60 -Z" (Ohm) Voltage (mV) The measured voltage, the bulk resistance, total resistance and calculated ionic transference number for different PO2 values is tabulated in Table 3.3.1. PO2 (atm.) Voltage (mV) Rb (ohm) Rt (ohm) Transference number 0.9975 -25.464 4.70 15.16 0.9942 0.12074 9.159 4.75 19.72 0.9985 0.030168 32.266 4.78 25.40 0.9997 0.0038394 65.419 4.80 68.03 0.9987 0.0006636 92.262 4.90 77.02 0.9976 Table 3.3.1. Voltage measurement, bulk resistance, total resistance and calculated transference number (Eq. 3.3.1) for the La5.1Y1.4 sample over a range of PO2 values. The PO2 value refers to the atmosphere on the “anode” of the La5.1Y1.4 sample. The “cathode” side is exposed to air. To be a suitable electrolyte in an SOFC, an oxygen conductor must have high and stable ionic conductivity, but it also must be able to block electrons. If the electrolyte conducted electrons, the OCV and operating voltage would both be low and the cell would have poor power density. There was some small variability in the transference number measurements over the entire PO2 range, thought it was likely due to small experimental errors. Overall the transference number approached unity, confirming that the La and Y double doped rhombohedral Bi2O3 is a pure ion conductor. After verifying the ionic conduction of the La and Y doped rhombohedral bismuth oxide, it was necessary to compare its conductivity to other prominent SOFC electrolytes. The Arrhenius behavior compared to other prominent oxygen ion conductors and bismuth oxide electrolytes is presented in Fig. 3.3.10. 61 Temperature (°C) 550 500 450 400 5 4 3 2 1 0 -1 0.0012 0.0013 0.0014 0.0015 1/T (K-1) Fig. 3.3.10. Arrhenius plot of multiple SOFC electrolytes from 550°C to 400°C. The activation energy of the La5.1Y1.4 sample is 0.867 eV. Adapted from7. It is clear that ionic conductivity of La5.1Y1.4 is higher than traditional electrolytes such as GDC and YSZ, and comparable to ESB over the stability range of the rhombohedral phase. Only DWSB has higher conductivity over the entire temperature range. The rhombohedral Bi2O3 though has much lower activation energy when compared to the cubic Bi2O3 samples. Since the ESB and DWSB are in the ordered state in this temperature range, the mobility of the oxygen ions is much lower. What makes the La5.1Y1.4 sample a superior and more reliable oxygen ion conducting material at intermediate temperatures though, is its stable performance. Fig. 3.3.11 62 Ln [σT (S/cm∙K)] illustrates the degradation of conductivity in La5.1Y1.4, ESB, and DWSB samples as a function of aging time at 500°C. -1.0 La5.1Y1.4 -1.5 -2.0 DWSB -2.5 ESB -3.0 0 10 20 30 40 50 60 70 80 90 100 Time (hours) Fig. 3.3.11. Conductivity as function of aging time for La5.1Y1.4, ESB and DWSB as a function of aging time at 500°C. The DWSB and ESB data is used as a reference and is taken from18. The La5.1Y1.4 sample exhibited no appreciable decay for the entire hold at 500°C, while both ESB and DWSB experienced over an order of magnitude drop in conductivity. With high ionic conductivity, low electronic conductivity, and stable performance, the rhombohedral La and Y doped Bi2O3 is a good SOFC electrolyte candidate. Furthermore, the rhombohedral phase of bismuth oxide in general is a promising SOFC electrolyte due to its flexibility. The rhombohedral phase of Bi2O3 can be formed using a variety of different dopants. With a large range of potential 63 log σ (S/Cm) chemistries, rhombohedral bismuth oxide can be optimized and tailored for specific traits. The La5.1Y1.4 sample was optimized to produce the highest conductivity at 500°C, but other attributes could be examined further. With a large phase boundary, there are many options to explore. Utilizing either divalent or trivalent cations with sufficient ionic radius should be sufficient in stabilizing the rhombohedral phase. Fig. 3.3.12 is a partial phase diagram of bismuth oxide that highlights the stability range of the rhombohedral phase. 6.0 6.5 7.0 7.5 8.0 8.5 9.0 9.5 10.0 1.06 1.08 1.10 1.12 1.14 Average Dopant Ionic Radius (Å) Fig. 3.3.12. Approximate phase stability window of rhombohedral Bi2O3 (dotted lines) given as function of total dopant concentration and average dopant ionic radius. The plus symbols indicate Bi2O3 compositions that are entirely rhombohedral, triangles are a rhombohedral/cubic mixture, squares indicate a rhombohedral/tetragonal mixture, and circles represent a rhombohedral/monoclinic mixture. The rhombohedral phase of bismuth oxide included a large possible array of dopant ionic radius and dopant concentrations. While an exhaustive set of samples were not synthesized and characterized to construct and exact phase diagram of Bi2O3, the 64 Dopant Concentration (%) approximate stability window of the rhombohedral phase is helpful for crafting the most highly conductive electrolytes. At lower dopant concentrations, bismuth oxide electrolytes are more conductive. Therefore, rhombohedral bismuth oxide samples that exist within the top of the plot should exhibit the highest conductivity. Ultimately using a trivalent cation substitution rate of approximately 6.5% and an average dopant ionic radius of around 1.125 Å should provide the highest conductivity among rhombohedral bismuth oxide. The Rietveld refinement and conductivity of all samples in the phase diagram are listed in Table 3.3.2. Dopant Lattice Lattice Ionic Secondary Name Composition Parameter Parameter c Radius Phase a (Å) (Å) (Å) La2.9Y7.1 Bi0.9La0.029Y0.071O1.5 1.06 Cubic 3.95902(17) 27.90979(63) La3.6Y6.4 Bi0.9La0.036Y0.064O1.5 1.07 None 3.96255(13) 27.9143(10) La4.3Y5.7 Bi0.9La0.043Y0.057O1.5 1.08 None 3.96423(17) 27.9131(14) La5Y5 Bi0.9La0.05Y0.05O1.5 1.09 None 3.96677(11) 27.9314(9) La5.7Y4.3 Bi0.9La0.057Y0.043O1.5 1.10 None 3.96890(11) 28.0581(8) La6.5Y3.5 Bi0.9La0.065Y0.035O1.5 1.11 None 3.96929(12) 27.9932(9) La7.2Y2.8 Bi0.9La0.072Y0.028O1.5 1.12 None 3.96901(19) 28.0366(15) La7.9Y2.1 Bi0.9La0.079Y0.021O1.5 1.13 None 3.9686(2) 28.0575(17) La8.6Y1.4 Bi0.9La0.086Y0.014O1.5 1.14 None 3.97037(21) 28.0810(51) La9.3Y0.7 Bi0.9La0.093Y0.007O1.5 1.15 Monoclinic 3.97095(21) 28.0810(15) La6.4Y2.6 Bi0.91La0.064Y0.026O1.5 1.12 None 3.97017(13) 28.0401(10) La5.7Y2.3 Bi0.92La0.057Y0.023O1.5 1.12 None 3.96875(12) 28.0575(10) La5Y2 Bi0.93La0.05Y0.02O1.5 1.12 None 3.96515(13) 28.1061(10) La4.7Y1.8 Bi0.935La0.047Y0.018O1.5 1.12 None 3.96712 28.0789 La5.1Y1.4 Bi0.935La0.051Y0.014O1.5 1.13 None 3.9672(2) 28.1539(17) La4.3Y1.7 Bi0.94La0.043Y0.017O1.5 1.13 Tetragonal 3.96743(10) 28.1680(8) La6Y1 Bi0.93La0.06Y0.01O1.5 1.14 None 3.96902(12) 28.1238(9) 65 La6Er2 Bi0.92La0.06Er0.02O1.5 1.12 None 3.96810(19) 28.0633(14) Nd9 Bi0.91Nd0.09O1.5 1.11 None 3.9675(1) 27.964(1) Nd8 Bi0.92Nd0.08O1.5 1.11 None 3.9662(2) 28.010(1) Nd7.5 Bi0.925Nd0.075O1.5 1.11 None 3.9649(2) 28.044(1) Nd7 Bi0.93Nd0.07O1.5 1.11 None 3.9648(2) 28.055(1) Nd6.5 Bi0.935Nd0.065O1.5 1.11 Tetragonal 3.9631(2) 28.070(1) Nd6 Bi0.94Nd0.06O1.5 1.11 Tetragonal 3.9622(2) 28.0901(9) Nd4Sm4 Bi0.92Nd0.04Sm0.04O1.5 1.095 None 3.96375(10) 27.9842(7) Nd4Gd4 Bi0.92Nd0.04Gd0.04O1.5 1.08 Tetragonal 3.9648(4) 28.011(3) Ca8 Bi0.92Ca0.08O1.5 1.12 None 3.95211(13) 27.94928(79) Sr3.2Y4.8 Bi0.92Sr0.032Y0.048O1.5 1.115 None 3.95718(12) 28.03506(94) Table 3.3.2. Rhombohedral Bi2O3 compositions, average dopant ionic radius, secondary phase, and lattice parameters based on Rietveld refinement and the Le Bail method. Lattice parameters relate to the R-3m space group of the rhombohedral lattice. The lattice parameters of the secondary phase is not included. Overall the rhombohedral lattice showed promise as an SOFC electrolyte for intermediate temperature operation. It had high ionic conductivity, low electronic conductivity, and good stability at 500°C. While the La5.1Y1.4 sample exhibited the highest stable conductivity at 500°C, the partial phase diagram of bismuth oxide and the host of rhombohedral samples synthesized demonstrates that a large variation in chemistry is possible. Ultimately the La5.1Y1.4 sample is a promising intermediate temperature SOFC electrolyte due to its high and stable conductivity at 500°C. 66 Chapter 4: Doping NASICON to Increase Conductivity 4.1 Motivation When considering batteries for grid-level storage, sodium-ion batteries have been shown to have a more viable and lower cost alternative to lithium-ion batteries44– 46. Specifically using a ceramic Na3Zr2Si2PO12 (NASICON) electrolyte in an all solid- state battery design has been demonstrated to be a safer, less expensive, and more reliable than other sodium battery chemistries47,48. Although NASICON is safer than liquid electrolytes, its ionic conductivity is an order of magnitude lower than many organic electrolytes and is thus less appealing as an electrolyte57. The main objective of this research is to enhance the room temperature conductivity of NASICON sufficiently to operate a high-performance room temperature solid-state battery. To increase the conductivity of NASICON, modifications of the structure will be carried out and interpreted to ultimately optimize the material properties and improve Na+ transport through the lattice. Much of the analysis focused on understanding the effect of doping on the bulk properties of NASICON rather than its effect on microstructure. 4.2 The Effect of Aliovalent Cation Substitution on NASICON Structure As discussed previously, structural investigations on doping the NASICON lattice have been done, but most of the work was focused on the microstructure. The work described here delves more into how the lattice is modified by lattice substitution on Na3Zr2Si2PO12 rather than the microstructural changes that occur. Although the 67 microstructure was detailed, the effect of doping on the phase transition from rhombohedral to monoclinic and changes in NASICON composition was the primary focus. For this study, a host of divalent and trivalent cations were substituted for the Zr+4 cation in the Na3Zr2Si2PO12 structure. The objective was to determine the effect of dopant ionic radius and valance on the phase of NASICON. To begin with, the zirconia octahedrons in NASICON have a coordination number of six, so at the very least all the potential dopant cations must support a six-fold coordination. Furthermore, the lower valance of the dopants generates a charge imbalance that must be compensated. Since there are vacant sodium sites in the NASICON structure, charge neutrality was obtained by filling the vacant sites with additional sodium ions. Overall the dopants were selected by their oxidation state and ionic radius to ensure a spectrum of ionic radii and overall lattice sodium content. The smallest coordination six ionic radius dopant was Al+3 (0.535 Å) and largest ionic radius dopant was Y+3 (0.90 Å). Ultimately the nominal stoichiometry of the NASICON materials had the following compositions: Na3.2Zr1.8M0.2Si2PO12 (M=Al 3+, Fe3+, Y3+), and Na Zr 2+3.4 1.8M0.2Si2PO12 (M=Co , Ni 2+, Zn2+). After all the materials were synthesized, room temperature x-ray diffraction was carried out to determine purity and lattice parameters of each sample. The room temperature XRD patterns of each sample is given in Fig. 4.2.1. The Rietveld fit confirmed that all samples exhibited greater than 90% purity. Monoclinic ZrO2 was an impurity observed in all prepared samples, but this is a common secondary phase found 68 in all types of NASICON synthesis70,77,78. The absence of a dopant containing secondary phase reinforces a successful substitution for the zirconium site. Fig. 4.2.1. Room temperature XRD pattern of all 10% doped NASICON samples. Rectangles, triangles, and diamonds representing Na3PO4, monoclinic ZrO2, and ZrSiO4 secondary phases respectively. All unlabeled peaks represent monoclinic Na3Zr2Si2PO12. Furthermore, WDS was performed on all the NASICON samples to provide an elemental configuration. In general, the actual composition of each sample was close to the nominal stoichiometry. The elemental analysis of all samples as well as room temperature lattice parameters determined by Rietveld refrainment are provided in Table 4.2.1. Of all the samples, only the Y doped sample showed noticeably lower dopant levels than the intended stoichiometry. Y2O3 is generally stable at elevated temperatures so volatilization of the species is unlikely. Therefore, the lower amount 69 of yttrium could be an artifact of the peak overlap present between phosphorus and yttrium in the WDS. The phosphorus content of the Y-doped NASICON is much higher than the rest of the samples, reinforcing the likelihood of peak overlap causing an inaccurate reading of both species. Sample a (Å) b (Å) c (Å) beta (°) Stoichiometry Undoped 15.63638 9.048573 9.220053 123.6849 Na2.99Zr2.03Si2.07P0.92O12 Al-doped 15.66292 9.0589967 9.216331 123.8453 Na3.38Zr1.80Al0.26Si2.06P0.88O12 Fe-doped 15.6505 9.050462 9.219722 123.7619 Na3.13Zr1.94Fe0.21Si2.12P0.85O12 Y-doped 15.70341 9.0802053 9.222833 124.0603 Na3.24Zr1.91Y0.09Si1.93P1.02O12 Co-doped 15.6755 9.069542 9.214981 123.9256 Na3.47Zr1.93Co0.18Si1.97P0.91O12 Ni-doped 15.68352 9.073429 9.211291 123.9757 Na3.59Zr1.92Ni0.21Si1.93P0.91O12 Zn-doped 15.68515 9.075624 9.212081 123.9897 Na3.43Zr1.83Zn0.22Si1.93P1.02O12 Table 4.2.1. Room temperature lattice parameters and actual stoichiometry of all NASICON samples. Rietveld refinement was used to determine lattice parameters. Stoichiometry was determined from WDS and normalized by 12 oxygen atoms per molecular formula of NASICON. Overall elemental analysis confirmed there was an increased amount of sodium in aliovalent substituted NASICON. Furthermore, divalent substituted NASICON contained a higher sodium content than trivalent doped NASICON as expected. The microstructure of the samples was also investigated. SEM images of all the NASICON samples are given in Fig. 4.2.2. In general, the particle size is in the submicron range and the grains are well connected. While most NASICON samples were defined by cubic grains with an average particle size smaller than 1 µm, Y-doped NASICON 70 differed in that there was greater necking between particles, smaller average particle size, and much more rounded grains. Furthermore Al-doped NASICON and Fe-doped NASICON had an inhomogeneous grain shape, displaying cubic particles, triangular particles, and irregular particles. a. d. b. e. c. f. Fig. 4.2.2. SEM images of a. Al-doped NASICON b. Fe-doped NASICON c. Y-doped NASICON d. Co-doped NASICON e. Ni-doped NASICON and f. Zn-doped NASICON 71 The microstructure of Ni-NASICON, Zn-NASICON, and Co-NASICON were the most consistent, demonstrating that NASICON doped with the +2 valent cations had the most uniform cubic particles. Additionally, there was no indication from the SEM photographs or from energy dispersive spectroscopy (EDS) that there was significant glassy phase in any sample. The microstructure of the samples could probably be further improved, as optimization of the fabrication process and sintering conditions was not the focus of this investigation. As indicated earlier, the structure and phase of NASICON is the primary concern of the investigation. Namely, understanding the effect of aliovalent doping on the rhombohedral-monoclinic phase transition. To accomplish this, in-situ X-ray diffraction was run from 30°C up to 300°C to monitor the NASICON structure over the phase transition. Fig. 4.2.3 is a plot of the high temperature XRD scans for the undoped NASICON. Fig. 4.2.3. XRD patterns of Na3Zr2Si2PO12 from 30°C up to 300°C. The high temperature XRD scan is zoomed in on the 2theta region from 18° to 28°. 72 The plot clearly illustrates a change from lower symmetry phase at 30°C to higher symmetry phase at 300°C, due to the increased number of peaks at lower temperature. The XRD patterns were fit at each temperature and lattice parameters were extracted. The refined lattice parameters at each temperature as well as the geometric relationships between the monoclinic and rhombohedral unit cells were used to model the phase transition. The phase transition from rhombohedral to monoclinic was modeled by equating the lattice parameters geometrically. The low temperature (monoclinic) and high temperature (rhombohedral) phases of NASICON are geometrically related as illustrated in Fig. 4.2.4. Fig. 4.2.4. Geometric representation of the monoclinic unit cell aligning with the rhombohedral unit cell. The orange base represents the rhombohedral phase and the blue base represents monoclinic. The hexagonal lattice system is used as a reference. 73 Above the phase transition the two unit cells fit together as illustrated in the figure and the following relationships are true: 𝑎𝑀 = 𝑎𝑅√3, 𝑏𝑀 = 𝑎𝑅, 𝑐𝑀 = 1⁄3 √3𝑎 2 2𝑅 + 𝑐𝑅 , with aM, bM, cM, and 𝛽𝑀 representing the monoclinic lattice parameters, and aR and cR for the rhombohedral lattice parameters. As the lattice cools below the phase transition temperature, the rhombohedral lattice distorts to the monoclinic cell. The distortion of the rhombohedral cell to the less symmetric monoclinic cell can be modeled by two different mechanisms (Eq. 4.2.1 and Eq. 4.2.2). 1 − 𝑎𝑀⁄(𝑏𝑀√3) (Eq. 4.2.1) 1 − (3𝑐𝑀 cos(180 − 𝛽))⁄𝑎𝑀 (Eq. 4.2.2) Eq. 4.2.1 models the distortion of the rhombohedral a-base and is illustrated by the graphic in Fig. 4.2.5 a. Eq. 4.2.2 models the shear distortion of rhombohedral unit cell and is represented in Fig. 4.2.5 b. a. b. Fig. 4.2.5 a. Distortion of the a-base observed from the 001 direction. b. Shear deformation of the unit cell observed from the 010 direction. The subscript R indicates rhombohedral and R’ indicates non-rhombohedral. 74 Above the phase transition, Eq. 4.2.1 and Eq. 4.2.2 will nominally equal zero, indicating that the unit cell is entirely rhombohedral. Below the phase transition the two equations should diverge from zero. Thus, the mechanism of the phase transition could be determined by plotting these equations against temperature. Fig. 4.2.6 a. and Fig. 4.2.6 b. are the plots of Eq. 4.2.1 and Eq. 4.2.2 versus temperature respectively for all NASICON samples. While the data in Fig. 4.2.6 a. remains relatively constant over the entire temperature region, Fig. 4.2.6 b. can be seen to change significantly with temperature. The discontinuous change in Eq. 4.2.2 is indicative of the temperature where the phase change occurred. Furthermore, the dependence of Eq. 4.2.2 on temperature indicates that shear distortion of the rhombohedral unit cell is responsible for the phase transition. Meanwhile Eq. 4.2.1 independence on the temperature reveals that the a-base of the unit cell remains relatively constant through the phase transition. Finally, despite the robustness of the Le Bail method, there is a small systematic error that prevents Eq. 4.2.1 and Eq. 4.2.2 from ever truly reaching zero. This error is a result of the limited number of observable peaks within the 2θ range covered by the detector and the high degree of peak overlap between the high and low temperature phases. Nevertheless, the discontinuous change in the value of Eq. 4.2.2 is undeniable proof of the phase transition. From the data it is clear the deformation of the unit cell is dependent on the dopant. Yttrium doped NASICON exhibits approximately 50% less unit cell distortion than undoped NASICON at 100°C as well as smaller overall distortion of the lattice through the phase transition. 75 a. b. Fig. 4.2.6 a. Shear deformation (Eq. 4.2.1) plotted as a function of temperature for all doped and undoped NASICON. b. Distortion of the unit cell a-base (Eq. 4.2.2) plotted as a function of temperature for all doped and undoped NASICON. 76 In fact all aliovalent doping of the zirconium site resulted in a reduction of shear deformation of the rhombohedral unit cell below the phase transition. DSC measurements also confirmed that doping NASICON reduced the transition temperature of the phase change. The endothermic peak was determined by the Perkin- Elmer software and represents the critical temperature (Tc) in the phase transition. The phase transition temperature determined from DSC is presented for all samples in Table 4.2.2. DSC confirms that the phase transition temperature of all doped NASICON in this investigation was lower than that of base NASICON. Sample Tc (°C) Undoped NASICON 157.7 Al-doped NASICON 151.3 Fe-doped NASICON 156.5 Y-doped NASICON 139.5 Co-doped NASICON 148.0 Ni-doped NASICON 151.2 Zn-doped NASICON 151.2 Table 4.2.2. Phase transition temperature of all NASICON samples determined from the endothermic peak (Tc) using DSC. The reduction of the monoclinic-rhombohedral phase transition in the NASICON samples is a result of doping the zirconium site with lower valent cations. However, vacant sodium sites are filled to compensate for the charge imbalance of +3 and +2 valent dopants in the +4 zirconium site, and this could also have a stabilizing 77 effect that lowers the phase transition temperature. The reduction in the phase transition temperature as a result of aliovalent substitution for zirconium is analogous to the phenomena observed in aliovalent phosphorus substitution in NASICON72. Colomban observed that there was lower static disorder in Na1+xZr2SixP3-xO12 for 1.8 86. Fu, K. (Kelvin) et al. Three-dimensional bilayer garnet solid electrolyte based high energy density lithium metal–sulfur batteries. Energy Environ. Sci. 10, 1568–1575 (2017). 87. Han, X. et al. Negating interfacial impedance in garnet-based solid-state Li metal batteries. Nat. Mater. 16, 572–579 (2016). 88. Zhou, W., Li, Y., Xin, S. & Goodenough, J. B. Rechargeable Sodium All- Solid-State Battery. ACS Cent. Sci. 3, 52–57 (2017). 119