ABSTRACT Title of Dissertation: FUNCTIONAL PARTICLE GENERATION BY AEROSOL-ASSISTED PROCESSES Yujia Liang, Doctor of Philosophy, 2018 Dissertation directed by: Sheryl H. Ehrman Department of Chemical and Biomolecular Engineering Aerosol-assisted processes are continuous with short residence times, simple operating procedures, and facile equipment requirements. They are scalable and promising for fabrication of functional particles as conductive pastes in solar cell metallization and interference packaging, electrode materials in energy storage devices, and photocatalysts in energy conversion. Although aerosol-assisted processes have long been used in manufacturing and their fundamentals have been intensively explored, further investigation is still required to better understand the particle formation mechanisms of different aerosol-assisted processes. In this dissertation, three different aerosol-assisted processes are investigated, spray pyrolysis, colloidal spray pyrolysis (CSP), and spray drying. These processes can be conducted under mild reaction conditions with simple operation procedures. The product particles are controllable. The effects of process variables on the product particles are studied. Furthermore, the prospects of applying these three aerosol-assisted processes in generating functional particles in applications, including solar cell metallization, battery, and photocatalysis are assessed. Part 1) includes Chapters 3-5. I first present Cu-Sn binary particle generation by spray pyrolysis. Through studying the particle oxidation behaviors under different reaction conditions, the Cu-Sn binary particles exhibit high oxidation-resistance. The one-dimensional and two-dimensional structures fabricated by direct printing inks containing Cu-Sn powders display low resistivity. They all suggest that Cu-Sn binary particles produced by spray pyrolysis are promising materials in the inks in printed electronics and in the conductive pastes in solar cell metallization and interference packaging. In Part 2), Chapters 6, a novel aerosol-assisted process, CSP, is developed. This process addresses one restriction of conventional spray pyrolysis which can only be used to fabricate particles from precursor solutions containing high-solubility salts. By applying CSP, tin@carbon (Sn@C) composite particles are produced with controllable interior structures. These composite particles exhibit high-performance as the anode materials for Li-ion and Na-ion batteries. In Part 3), Chapter 7, spray drying is utilized to fabricate photocatalysts from precursor solutions containing SnO2 colloids and edge-oxidized graphene oxide (eo- GO) sheets. The particle morphology, element distribution, and band structures were investigated by various tools. The photocatalytic activity of the composite particles is five times that of commercialized TiO2 (P25) in reducing CO2 into CH4. FUNCTIONAL PARTICLE GENERATION BY AEROSOL-ASSISTED PROCESSES by Yujia Liang 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 Sheryl Ehrman, Chair Dr. Howard Glicksman Professor Chunsheng Wang Professor Dongxia Liu Professor Bao Yang © Copyright by Yujia Liang 2018 Dedication Dedicated to Qingnian Liang, Weiyi Zhang, and Sixing Lu To my 20s. ii Acknowledgement I sincerely want to thank my advisor, Dr. Sheryl Ehrman, for her guidance during the years. Her support assisted me to accomplish this dissertation. She is a model for me to learn as a scholar, advisor, and leader. I would like to thank Dr. Howard Glicksman for his suggestions and advice during my research. I want to express my gratitude to National Science Foundation (CBET-1336581) and DuPont Company for their financial support of this project. I also want to thank Dr. Chunsheng Wang and Dr. Dongxia Liu for their assistance on our collaborated projects and serving on my committee. I am also grateful to Dr. Bao Yang for serving on my committee and providing me with valuable suggestions. I want to thank Dr. Sz-Chian Liou for training me to be an expert in TEM, Dr. Peter Zavalij for instructing me to acquire the knowledge of Rietveld refinement, and Dr. Karen Gaskell for XPS data analysis. I want to thank lab mates, Dr. Kai Zhong, Dr. Alex Langrock, Dr. Tim Vinciguerra, and Trang Nguyen. I would also like to thank my collaborators, Chao Luo, Fudong Han, Fei Wang, Yangang Liang, Huajun Tian, Yong Yang, Sucheun Oh, Ji Chen, Wei Wu, Peng Wang, Huilong Hou, and all other people who have ever helped me during my research. At last, I want to send my special acknowledgment to my parents in China, Qingnian Liang and Weiyi Zhang. It is always your love that helps and supports me to go through ups and downs in my life. For my wife, Dr. Sixing Lu, I owe you a lot during the years for your endless patience and encouragement. The spatial distance (between College Park, MD and Tucson, AZ) strengthened our love and forced us concentrate on our research. We both become PhDs in less than 5 years. iii Table of Contents Dedication ....................................................................................................................................... ii Acknowledgement ......................................................................................................................... iii Table of Contents .......................................................................................................................... iv List of Tables ................................................................................................................................. vii List of Figures ............................................................................................................................... viii Chapter 1: Introduction and overview......................................................................................... 1 1.1 Motivations ..................................................................................................................... 1 1.2 Research strategies and objectives ............................................................................... 4 Chapter 2: Background information............................................................................................ 7 2.1 Developing alternatives to Ag particles in applications of printed electronics, interference packaging, and solar cell metallization ............................................................... 7 2.2 Approaches to fabricate high-performance anode materials ................................... 10 2.3 Methods to fabricate photocatalysts ........................................................................... 13 Chapter 3: Cu-Sn binary particle generation by spray pyrolysis ............................................ 17 3.1 Abstract ......................................................................................................................... 17 3.2 Introduction .................................................................................................................. 17 3.3 Experimental ................................................................................................................ 20 3.3.1 Preparation of precursor ............................................................................................ 20 3.3.2 Spray pyrolysis procedures ........................................................................................ 21 3.3.3 Mechanism exploration .............................................................................................. 23 3.3.4 Characterization and data analysis ........................................................................... 24 3.4 Results and discussion ................................................................................................. 24 3.4.1 Reaction temperature ................................................................................................. 24 3.4.2 Ethylene glycol volume percentage in the precursor ............................................... 36 3.4.3 Residence time ............................................................................................................. 40 3.4.4 Effects of quench ......................................................................................................... 46 3.4.5 Mechanism ................................................................................................................... 48 3.5 Summary of this chapter ............................................................................................. 56 Chapter 4: Oxidation behavior of Cu-Sn binary particle ........................................................ 58 4.1 Abstract ......................................................................................................................... 58 iv 4.2 Introduction .................................................................................................................. 58 4.3 Experimental ................................................................................................................ 60 4.3.1 CuSny particle preparation ........................................................................................ 60 4.3.2 CuSny powder characterization ................................................................................. 61 4.3.3 Oxidation kinetics investigation ................................................................................. 62 4.4 Results and discussion ................................................................................................. 63 4.4.1 CuSny particle preparation by spray pyrolysis ........................................................ 63 4.4.2 Oxidation behavior of CuSny particles ...................................................................... 72 4.4.3 Oxidation kinetics of CuSny powders ........................................................................ 78 4.5 Summary of this chapter ............................................................................................. 91 Chapter 5: Conductive one-dimensional and two-dimensional structures fabricated by Cu- Sn binary particles ....................................................................................................................... 92 5.1 Abstract ......................................................................................................................... 92 5.2 Introduction .................................................................................................................. 92 5.3 Experimental ................................................................................................................ 94 5.4 Results and discussions ................................................................................................ 96 5.5 Summary of this chapter ........................................................................................... 109 Chapter 6: Colloidal spray pyrolysis approach to engineer hierarchical tin@carbon particles as anode materials ...................................................................................................... 110 6.1 Abstract ....................................................................................................................... 110 6.2 Introduction ................................................................................................................ 111 6.3 Experimental .............................................................................................................. 115 6.3.1 Sn@C fabrication by CSP ........................................................................................ 115 6.3.2 Characterization ....................................................................................................... 116 6.3.3 Electrode preparation and electrochemical measurement .................................... 116 6.4 Results and discussion ............................................................................................... 117 6.4.1 Aerosol dynamics and reaction kinetics in colloidal spray pyrolysis ................... 117 6.4.2 Application of colloidal spray pyrolysis in fabricating Sn@C particles. ............. 120 6.4.3 Electrochemical performance of Sn@C anodes in LIBs and SIBs. ...................... 137 6.5 Summary of this chapter ........................................................................................... 145 Chapter 7: Fabrication of nanocomposite particles containing SnO2 colloids and edge- oxidized graphene oxide sheets as photocatalysts ................................................................... 147 7.1 Abstract ....................................................................................................................... 147 7.2 Introduction ................................................................................................................ 147 v 7.3 Experimental .............................................................................................................. 151 7.3.1 Materials .................................................................................................................... 151 7.3.2 Photocatalyst fabrication by spray drying .............................................................. 152 7.3.3 Materials characterization ....................................................................................... 154 7.3.4 Photoreduction of CO2 .............................................................................................. 154 7.4 Results and discussions .............................................................................................. 155 7.5 Summary of this chapter ........................................................................................... 172 Chapter 8: Conclusions and future work ................................................................................ 173 8.1 Conclusions ....................................................................................................................... 173 8.2 Future work ...................................................................................................................... 176 8.2.1 Alternative materials in applications, including printed electronics .................... 176 8.2.2 Further investigation on CSP................................................................................... 177 Appendices .................................................................................................................................. 180 A1. Detailed procedures and equations to understand oxidation of Cu-Sn particles by shrinking-core model (SCM) in Chapter 4 .......................................................................... 180 A2. Modeling on the colloidal spray pyrolysis process in Chapter 6 ................................. 189 A2.1 General equations for CSP modeling ...................................................................... 189 A2.2 Detailed procedures to calculate the time of SnO2 colloid reduction ................... 191 A3. Detailed calculation procedures to obtain the colloid concentration in Chapters 6.. 194 References ................................................................................................................................... 196 vi List of Tables Table 2.1. Salt solubilities of corresponding metals as high-performance battery materials listed in (Goriparti et al. 2014)……………………………………………………………10 Table 3.1. Vapor pressure of water and EG………………………………………………36 Table 4.1. Summary of the configurations of CuSny particles (y≤0.1) based on mass balance calculation and experimental conditions to obtain these particles………………………...70 Table 5.1. Dimensions of the wires fabricated by direct printing………………………101 Table 6.1. Precursors to fabricate different structures of Sn@C particles……………….114 Table 6.2. Precursor to obtain Sn@C intermediate particles…………………………….130 Table 6.3. The half-time of inter-particle collision between Sn@C particles……………134 Table 6.4. Comparison of electrochemical performance of our Sn@C anodes with references……………………………………………………………………………….140 Table 7.1. Composition of precursor solutions for fabrication of SnO2/eo-GO photocatalysts…………………………………………………………………………...153 Table 7.2. D and G band positions and D/G ratio of Raman spectra……………………..158 Table 7.3. Band structures of eo-GO, TinGO17, and SnO2 as photocatalysts…………...167 Table 7.4. Detailed yields and selectivities of SnO2/eo-GO photocatalysts……………..168 Table A.1.1. Variables used in modeling with their definitions and values……………...183 Table A.1.2. Parameters used in modeling with definitions……………………………..183 Table A.1.3. Simulation results of Cu-Sn particle oxidation at 300 oC in a packed-bed reactor by SCM…………………………………………………………………………185 Table A.2.1. Parameters used in the modeling in Section A.2…………………………...192 vii List of Figures Figure 3.1 23 Temperature profile of a double furnace reactor of 81.3 cm in length with (a) fixed CGFR at 3 L/min and various temperatures; (b) fixed temperature at 750 oC and different CGFRs; (c) estimated residence time. (d) and (e) The heating rates in the reactor with respect to the conditions in (a) and (b), respectively. Figure 3.2 25 XRD results of powders generated from systematic conditions. The precursor contained 1 M Cu(NO3)2, 0.1 M SnCl2, 0.1 M HNO3, and 4.8 M EG. The CGFR was fixed at 3 L/min. The powders were generated at 500 oC, 750 oC, and 1000 oC. The major peaks are from Cu13.7Sn (ICDD with PDF No. 03-065-6821) and Cu (ICDD PDF No. 01-070-3038). Figure 3.3. SEM images of powders generated from the precursor solution with 1 26 M Cu(NO3)2, 0.1 M SnCl2, 0.1 M HNO3, and 4.8 M EG. The reaction temperatures were 500 oC (a), 750 oC (b), and 1000 oC (c). The CGFR was fixed at 3 L/min. Insets are higher magnification images of selected regions marked by solid arrows. The scale bars in all figures are 1 µm. (d) The summary of the diameter of the particles marked as 1, 2, and 3 in (a)–(c). Detailed size distributions for each condition can be found in Figure 3.6. Figure 3.4 27 Samples generated at 750 oC with EG concentration of 3.6 M (a), 4.8 M (b)-(c), and 5.9 M (d)-(e). The concentrations of Cu(NO3)2, SnCl2, and HNO3 were 1 M, 0.1 M, and 0.1 M, respectively. The scale bars of the SEM images are 1 μm. Figure 3.5 27 TEM images of particles fabricated by precursors with 1 M Cu(NO3)2, 0.1 M SnCl2, and 0.1 M HNO3. The CGFR was fixed at 3 L/min. (a)–(b) Particles generated at 500 oC, with 5.9 M EG. (c)–(d) Particles obtained at 750 oC, with 4.8 M EG. (e)–(f) Particles obtained at 1000 oC, with 4.8 M EG. Figure 3.6 28 Diameter distributions of the products generated at systematic conditions. The concentrations of Cu(NO3)2, SnCl2, and HNO3 were 1 M, 0.1 M, and 0.1 M, respectively. Figure 3.7 29 SEM images of powders generated from systematic conditions. EG concentration was tuned at 3.6 M, 4.8 M, and 5.9 M at 500 oC, 750 oC, and 1000 oC, respectively. The concentrations of Cu(NO3)2, SnCl2, and HNO3 in the precursor were 1 M, 0.1 M, and 0.1 M, respectively. The CGFR was fixed at 3 L/min. Insets highlighted by yellow squares are higher magnification images of selected regions marked by yellow arrows. The scale bars in all figures are 1 μm. viii Figure 3.8 33 The Cu/Sn ratios summarized from XPS results. The precursor contained 1 M Cu(NO3)2, 0.1 M SnCl2, 0.1 M HNO3, and 4.8 M EG. The CGFR was set at 3 L/min. Sputtering was conducted in vacuum by argon ions for 20 min. Figure 3.9 35 XPS signals collected at different temperatures. The concentrations of Cu(NO3)2, SnCl2, and HNO3 were 1 M, 0.1 M, and 0.1 M, respectively. The EG concentration was fixed at 4.8 M. The CGFR was set at 3 L/min. The temperature was set at 1000 oC (a)-(c); 750 oC (d)-(f); 500 oC (g)-(i); (a), (d), and (g) represent the data from Sn 3d electrons. (b), (e), and (h) illustrate the data from Cu 2p electrons. (c), (f), and (i) are the data from Cu LMM Auger electrons. Figure 3.10 38 (a) Densities and viscosities of precursors with 3.6 M, 4.8 M, and 5.9 M EG. The concentrations of Cu(NO3)2, SnCl2, and HNO3 were 1 M, 0.1 M, and 0.1 M, respectively. (b) Densities and viscosities of a precursor with 1 M Cu(NO3)2, 0.1 M SnCl2, 0.1 M HNO3, and 5.9 M EG during the spray pyrolysis. Figure 3.11 39 XRD results of powders generated from systematic conditions. The concentrations of Cu(NO3)2, SnCl2, and HNO3 in the precursor were 1 M, 0.1 M, and 0.1 M, respectively. The CGFR was fixed at 3 L/min. The powders were generated at 500 oC (black), 750 oC (red), and 1000 oC (blue), with EG concentration varied at 3.6 M, 4.8 M, and 5.9 M (from bottom to top), respectively. Figure 3.12 42 SEM images of powders obtained at 750 oC from precursors with 1 M Cu(NO3)2, 0.1 M SnCl2, 0.1 M HNO3, and 4.8 M EG. The residence times are: 4.5 s (a), 1.5 s (b), and 0.92 s (c). The insets of each figure are higher magnification images of the area marked by arrows. Arrows show irregular structures. (d) The XRD results of the powders obtained at the conditions of (a)–(c). The peaks are attributed to Cu13.7Sn (ICDD with PDF No. 03-065-6821). All scale bars in the SEM images are 1 µm. Figure 3.13 43 Images of as-sprayed filters generated from precursors with Cu(NO3)2, SnCl2, and HNO3 of 1 M, 0.1 M, and 0.1 M, respectively. The EG concentration was 4.8 M. The temperature was set at 750 oC. The diameters of the filters are ~ 9.5 cm. Figure 3.14. 45 XPS signals collected at different CGFRs. The concentrations of Cu(NO3)2, SnCl2, and HNO3 were 1 M, 0.1 M, and 0.1 M, respectively. The EG concentration was fixed at 4.8 M. The temperature was set at 750 oC. The CGFR was set at 5 L/min (a)-(c); 3 L/min (d)-(f); 5 L/min (g)-(i); (a), (d), and (g) are the data from Sn 3d ix electrons. (b), (e), and (h) are the data from Cu 2p electrons. (c), (f), and (i) are the data from Cu LMM Auger electrons. Figure 3.15. 47 The effect of quench gas on the morphologies and crystallinities of the powders obtained at 1000 oC. The precursors contained 1 M Cu(NO3)2, 0.1 M HNO3, and 4.8 M EG. The SnCl2 concentration was 0.05 M (a)–(b) and 0.1 M (c)–(d). SEM images of the prepared powders are shown in (a), (c) with quench gas, and (b), (d) without quench gas. (e) The XRD results of the generated powders at the four conditions. Standard peaks from Cu13.7Sn, (Cu32Sn)0.12, and Cu are displayed below the XRD results for reference. Scale bars in the SEM images are 1 µm. Figure 3.16 49 XRD diagram of powders fabricated from precursors containing 1 M Cu(NO3)2, 0.1 M SnCl2, 0.1 M HNO3, and 5.9 M EG in a total volume of 150 ml aqueous solution. Spray pyrolysis was conducted at 300 oC with CGFR = 5 L/min and a quench gas velocity of ~ 5.2 L/min. Figure 3.17 50 XRD diagram of powders fabricated from precursors containing 1 M Cu(NO3)2, 0.1 M SnCl2, 0.1 M HNO3, and 5.9 M EG. Spray pyrolysis was conducted at 500 oC with CGFR = 10 L/min and a quench gas velocity of ~ 10.8 L/min. Figure 3.18 53 Schematic illustration of the proposed Cu-Sn binary particle formation process. The circle in Step (1) represents the precursor droplet. The magnified image of the Step (1) product illustrate the Cu-Sn solid solution and the metallic copper, respectively. The small circles are the metal particles formed by gas-to-particle conversion. Figure 4.1 63 SEM images of Cu (a), CuSn0.01 (b), CuSn0.05 (c), and CuSn0.1 (d) particles. The insets in each panel are the high magnification images of the selected regions marked by dashed squares. The scale bars are 1 μm for all the images. Figure 4.2 64 TEM images of Cu particle (a) with corresponding HRTEM image (b) of selected area in (a). The inset of (b) is the fast Fourier transform (FFT) of the HRTEM. TEM image of CuSn0.01 particles (c) with HAADF image and line scan results of Cu and Sn (d), elemental mapping of Cu (e) and Sn (f), and overlap mapping (g). Figure 4.3 66 XRD diagrams of CuSny solid particles at room temperature (bottom) and after being heated at 150 oC for 10 min in ambient air (top). The positions marked by circles and squares are standard peaks of Cu13.7Sn (circles, PDF No. 03-065-6821) and Cu (squares, PDF No. 01-070-3038), respectively. x Figure 4.4 67 TEM images of CuSn0.05 (a), CuSn0.1 (b), CuSn0.2 (c), CuSn0.4 (d), CuSn0.8 (e), and CuSn (f) products. Figure 4.5 68 SEM images with the corresponding high magnification images of the areas marked by dashed boxes of CuSn0.2 (a), CuSn0.4 (b), CuSn0.8 (c), and CuSn (d). The scale bars in all images are 1 μm. Figure 4.6 70 The Sn/Cu atomic ratios summarized from XPS results (Figure 4.7). The inset represents the regime where Sn/Cu ratio is below 0.1. Sputtering was conducted in vacuum by argon ions for 20 min. Figure 4.7 71 XPS signals of Cu (a-c), CuSn0.01 (d-f), CuSn0.05 (g-i), CuSn0.1 (j-l), CuSn0.2 (m- o), CuSn0.4 (p-r), CuSn0.8 (s-u), and CuSn (v-x) particles before and after sputtering. Images in the left column represent the spectra from Sn 3d electrons. Images in the middle column illustrate the spectra from Cu 2p electrons. Images in the right column are the spectra from Cu LMM Auger electrons. Figure 4.8 72 Line scans of Cu and Sn elemental profiles of CuSn0.05 (a), CuSn0.1 (b), CuSn0.2 (c), CuSn0.4 (d), and CuSn0.8 (e). Figure 4.9 73 TEM images of the Cu (a-b) and CuSn0.1 (c-d) particles after being oxidized in ambient for 60 min at 100 oC. Oxide layers were formed on the particle surfaces as shown by the lattice fringe in (d), which is attributed to the (110) plane of Cu2O (PDF No. 01-071-3645). Figure 4.10 75 XRD patterns (Yobs, black dots) of Cu (a), CuSn0.01 (b), CuSn0.05 (c), and CuSn0.1 (d) powders, after they were oxidized at 300 oC for 10 min in ambient air analyzed by Rietveld refinement. XRD patterns were analyzed by Rietveld refinement (Ycalc, red curves) with the difference between the observed and calculated values (Ycalc-obs, blue curves). Tick marks represent the phases utilized in refinements, including Cu (black, PDF No. 00-004-0836), Cu2O (red, PDF No. 01-071-3645), CuO (blue, PDF No. 01-073-6023), and Cu-Sn solid solutions. Cu-Sn solid solutions are Cu (black) in (b), (Cu32Sn)0.12 (green, PDF No. 01-077-7742) in (c), and (Cu13.7Sn)0.068 (yellow, PDF No. 01-071-7874) in (d). (e) Detailed results of the weight percent of crystal phases in post-oxidation products. (f) The evolution of lattice parameter from Cu in (a) to Cu-Sn solid solutions in (b-d). The Cu-Sn solid solution phases and Cu phase share the same space group of Fm-3m. Figure 4.11 77 xi In-situ XRD measurements of pure Cu, CuSn0.05, and CuSn0.1 powders in ambient air at increasing temperature from 30 oC to 400 oC (a) with corresponding temperature profile during the oxidation (b). The red dots correspond to initiation of XRD scans. Every XRD measurement lasts 15 min. Figure 4.12 78 In-situ XRD measurements of pure Cu (a), CuSn0.05 (b), and CuSn0.1 (c) when the sample temperature was held at 200 oC in ambient air subject to heating. (d) The temperature profile of the sample during the oxidation. In (d), the red dots correspond to initiation of XRD scans. Figure 4.13 79 O2 partial pressures in the product streams of the packed-bed reactor when the volume percents of O2 in feed stream are 1.4 % (a) and 10 % (b). Figure 4.14 80 (a) O2 flow rates in the product streams of the packed-bed reactors when the volume percent of O2 in feed stream is 1.4 %. (b) and (d) The weight gains of the powder during the oxidation at 300 oC (b) and 500 oC (d). They are calculated by applying the trapezoidal method on the results as shown in (a), for example the area marked by “1” for the CuSn o0.1 sample at 300 C. (c) The initial 20 min of the oxidation process at 300 oC. Figure 4.15 81 XRD patterns of powders after being oxidized in a packed-bed reactor at 300 oC (bottom) and 500 oC (top). The volume percent of O2 in the feed stream is 1.4 %. Figure 4.16 83 TEM images of Cu (a)-(b), CuSn0.05 (c)-(d), and CuSn0.1 (e)-(f) particles after being oxidized in packed-bed reactor under 300 oC. The volume percent of O2 in feed stream is 1.4 %. (b), (d), and (f) are the higher magnification images of the areas marked by red boxes in (a), (c), and (e), respectively. Figure 4.17 87 TEM images of Cu (a)-(c), CuSn0.05 (d)-(f), and CuSn0.1 (g)-(i) particles after being oxidized in packed-bed reactor under 500 oC. The volume percent of O2 in feed stream is 1.4 %. (b), (e), and (h) are the higher magnification images of the areas marked by red boxes in (a), (d), and (g), respectively. (c), (f), and (i) are the selected area electron diffraction (SAED) patterns of areas highlighted by yellow boxes in (b), (e), and (h), respectively. SAED images are obtained by performing the fast Fourier transform (FFT). The identified phases are CuO (PDF No. 01-073- 6023) and SnO2 (PDF No. 01-072-1147). Figure 4.18 89 xii (a) O2 flow rate in the product stream of the packed-bed reactor when the volume percent of O2 in feed stream is 10 %. (b) The relative weight gains of the powder during the oxidation. The weight gain is calculated by applying the trapezoidal method on the results shown in (a). (c) XRD pattern of powders after being oxidized in a packed-bed reactor at 500 oC under 10 vol % O2 stream. Figure 4.19 90 TGA (left axis) and DSC (right axis) measurements on Cu particles (black curves) and CuSn0.1 particles (green curves). Figure 5.1 96 Powders used in the inks. (a-b) SEM images of Cu (a) and CuSn0.1 (b) particles. (c-d) TEM images of Cu (c) and CuSn0.1 (d) particles. (e) XRD diagram of Cu, CuSn0.05, and CuSn0.1 particles. Solid black circles and squares represent Cu13.7Sn (PDF No. 03-065-6821) and Cu (PDF No. 01-070-3038), respectively. Figure 5.2 97 SEM image of CuSn0.1 particles. Figure 5.3 98 (a) Measured viscosity of H2O-EG mixture solutions with varying volume percent of EG determined using a viscometer. (b) Vapor pressures of H2O and EG as a function of temperature, adapted from (Yaws 1999). Figure 5.4 99 Microstructural evolution caused by increasing sintering temperature. (a) Schematic of direct printing ink on a 5 mm×5 mm quartz substrate. (b) An image of the wire after being sintered. (c-e) SEM images of wires after being sintered at 573 K (c), 673 K (d), and 773 K (e). (f) Schematic summary of microstructural evolution during sintering (f). Figure 5.5 100 In-situ resistivity measurements of the wires from 2 K to 400 K. (a) Resistivities of CuSn0.1 wires being sintered at 573 K (red) and 673 K (black). Inset is the image of the experimental setup. (b) Resistivity-temperature profiles of Cu (black), CuSn0.05 (blue), and CuSn0.1 (red) wires. These wires were sintered at 773 K after direct printing. CuSn0.1 wires were fabricated from different batches. (c) Thermogravimetric analysis (TGA) of CuSn0.1 wire. Figure 5.6 101 Typical cross-section profile of wire after sintering. Figure 5.7 102 Temperature dependent resistivities of Cu, CuSn0.05, and CuSn0.1 wires with y-axis in linear scale. Inset is the resistivities when the measurement temperature was below 40 K. xiii Figure 5.8 104 Temperature dependent resistances obtained by measuring the same wires twice. (a) Three CuSn0.1 wires. (b) Cu, CuSn0.05, and CuSn0.1 wires. All the wires were sintered at 773 K. Figure 5.9 105 SEM images of CuSn0.1 wire after being sintered at 773 K (a) with EDS mappings of overlap mapping (b), Cu (c) and Sn (d) elements. Figure 5.10 106 XRD pattern of the CuSn0.1 film after being sintered at 773 K in H2-N2 gas for 2 h. The peaks are attributed to Cu13.7Sn phase with PDF No. 03-065-6821. Figure 5.11 107 Resistivity ratios of film resistivities at given time (t) during oxidation at 573 K in ambient air to the resistivities before the oxidation. The films were sintered at 773 K after direct printing. The inset is the schematic illustration of the measurement. Figure 5.12 107 In-situ resistance measurements of Cu, CuSn0.05, and CuSn0.1 films during oxidation at 573 K in ambient air. Figure 6.1 113 Schematic drawings of conventional spray pyrolysis and colloidal spray pyrolysis. (a) A comparison of conventional spray pyrolysis (Route I) and our strategy (Route II) with their advantages (in blue) and disadvantages (in red). (b) Schematic of colloidal spray pyrolysis to fabricate pitaya-structured, pomegranate-structured, and chestnut-structured particles. Figure 6.2 121 Composition control of the product particles by colloidal spray pyrolysis. (a-b) TEM image of the SnO2 colloids used in the precursor (a) with HRTEM image (b). (c) Diameter distribution of SnO2 colloids with mean diameter, standard deviation, and sample size. (d) Schematic of parameters used in calculating the time of colloid reduction in colloidal spray pyrolysis. (e) Time required to reduce a SnO2 colloid as a function of initial colloid diameter, based on reaction-controlled model (left y-axis) and diffusion-controlled model (right y-axis). (f) XRD diagram of powders obtained at 750 oC when the residence time of the process is 4.5 s (black curve), 1.5 s (red curve), and 0.92 s (blue curve). Figure 6.3 122 (a) is a beaker containing an aqueous solution with 3.8×10-2 M SnO2 and 4.8 M EG. After adding 0.11 M KCl into (a), the solution was stirred for 1 h (b), 2 h (c), and 3 h (d) consecutively. (e) is a beaker containing the precursor for the generation xiv of Sn@C pomegranate particles. The precursor composed of 0.2 M sucrose, 3.8×10-2 M SnO2, and 4.8 M EG. After adding 0.11 M KCl into (e), the solution was stirred for 1 h (f), 2 h (g), and 3 h (h) consecutively. (j) is the image of (e) after being ultrasonically sprayed for 2 h. After adding 0.11 M KCl into (j), the solution was stirred for 1 h shown in (k). Density measurements were taken for the precursor before (e) and after (j) the reaction. When taking the photos, beakers were placed on the marked area of paper with letters of “UMD”. Without adding sucrose, SnO2 colloids precipitated within 1 h after adding KCl into the solution. As shown in (b), “UMD” was hard to see. With sucrose, the SnO2 colloids remained stable and dispersed in the precursor shown in (f)-(h). After adding KCl into the solution (f), “UMD” can still be seen at least for 3 h as shown in (h). For the precursor after being sonicated for 2 h (j), “UMD” can be seen after adding KCl and stirring for 1 h as illustrated in (k). Since the size of SnO2 colloids is small (mean diameter = 4.4 nm), we cannot separate them from precursor via filtration and high-speed centrifuging. This is the best method to demonstrate the high uniformity and the high stability of our precursor. (l) is an aqueous solution with 0.1 M SnCl2, 0.4 M sucrose, and 4.8 M EG. The molar ratio of sucrose to SnCl2 is 4. “UMD” is hard to be observed due to the hydrolysis of SnCl2. (m) is an image of the solution in (l) after being stirred for 1 h. The solution is cloudy. We cannot see the “UMD” through the solution. (n) shows the solution in (m) after being placed for 1 h without stirring. The precipitates concentrate at the bottom of the beaker. Figure 6.4 125 TEM images of powders obtained at 750 oC with the same precursor solution used to generate Sn@C pomegranate particles. The residence times for the fabricated process are 4.5 s (a), 1.5 s (b), and 0.92 s (c)-(d). Figure 6.5 126 SEM images of powders obtained at 750 oC with the same precursor solution used to generate Sn@C pomegranate particles. The residence times for the fabrication process are 4.5 s (a), 1.5 s (b), and 0.92 s (c). The insets are the high magnification images of the corresponding areas marked by dashed boxes. Fragmented particles are marked by red arrows in (c). Figure 6.6 127 Structure and composition control of the product particles by colloidal spray pyrolysis. (a-c) TEM images of pitaya-structured (a), pomegranate-structured (b), and chestnut-structured (c) particles. (d-n) STEM-HAADF image and EELS elemental mappings of C signal, Sn signal, and overlap result of Sn@C pitaya (d- f, respectively), pomegranate (h-k, respectively), and chestnut (l-o, respectively) particles. The scale bars in STEM-HAADF images are 50 nm. (p-t) Statistical summaries of Sn@C particle and interior Sn bead diameters of pitaya-structured (p,s, respectively), pomegranate-structured (q,t, respectively), and chestnut- structured (r) particles with mean values, standard deviations, and sample sizes. (u-v) Box plots to summarize the size distributions of Sn@C particles (u) and Sn xv beads (v). (w) Calculations of the colloid reduction time on conditions to generate Sn@C pitaya, pomegranate, and chestnut particles. (x) XRD patterns of the generated Sn@C pitaya, pomegranate, and chestnut particles. Figure 6.7 128 SEM images of Sn@C pitaya (a), pomegranate (b), and chestnut (c) particles. Figure 6.8 128 TGA results of the Sn@C pitaya (black), pomegranate (red), and chestnut (blue) powders. Figure 6.9 130 SEM (a) and TEM (b) images of Sn@C intermediate powders fabricated from a precursor with sucrose/SnO2 (mass ratio) = 6.1, which is between that of Sn@C pomegranate and chestnut particles. Figure 6.10 131 XRD diagram of Sn@C intermediate powders fabricated from a precursor with [sucrose]/[SnO2] = 2.7, which is between that of Sn@C pomegranate and chestnut particles. All the peaks are attributed to the Sn phase (PDF No. 01-075-9188). Figure 6.11 132 XRD diagram of powders fabricated with the same [sucrose]/[SnO2], reaction temperature, and residence time to obtain Sn@C pomegranate structured particles without the addition of EG to the precursor. Figure 6.12 135 Carbon frames in Sn@C particles. (a-h) TEM images of chestnut-structured (a), pomegranate-structured (d), and pitaya-structured (g) particles with its corresponding HRTEM images (b)-(c), (e)-(f), and (h), respectively. The HRTEM images were taken from the areas marked by red color boxes. (c,f) Higher magnification images of areas in (b) and (e), respectively. (i) STEM-HAADF images (left) and EELS line scan results (right) of chestnut-, pomegranate-, and pitaya-structured particles (from top to bottom). The scale bars in HADDF images are 100 nm. (j) EELS spectra of C K-edge extracted from position A (left) and B (right) of (i). Figure 6.13 137 Raman spectra of the Sn@C pitaya (blue), pomegranate (red), and chestnut (black) powders. It was recorded by a Raman spectrometer, equipped with a 532 nm laser. Figure 6.14 138 (a) Long-term cycling stability of the Sn@C electrodes at a current density of 1/10 C for the initial three cycles and at 1C for the following cycles. The inset is cycling performance of the Sn@C anodes for the initial 100 cycles at 1C. (b-k) Schematic xvi summaries of structural evolution during lithiation/delithiation and TEM images of pitaya-structured (b,e-f), pomegranate-structured (c,g-h), and chestnut- structured (d,i-k) particles in anodes after 100 cycles in rate performance test. (l) Rate capability of Sn@C pomegranate anode in SIB. Electrochemical experiments in (a) and (l) are conducted by collaborator (Huajun Tian). Figure 6.15 140 Comparison of electrochemical performance of our Sn@C anodes with references in Table 6.4. Figure 6.16 142 Post-cycling characterization of Sn@C pitaya (a), pomegranate (b), and chestnut (c) anodes. Figure 6.17 144 Post-cycling characterization of Sn@C intermediate anode. Figure 7.1 151 Particles synthesized by spray drying. (a) Schematic illustration of spray drying process. (b) XRD patterns of TinGOx powders obtained through spray drying. (c) Raman spectra of TinGOx powders. Figure 7.2 152 Molecular structure of 4-10% edge-oxidized graphene oxide sheet received from Sigma-Aldrich. Figure 7.3 157 Image of GO (a) and EELS core loss spectra of graphene (black) and GO (red) (b). (c) is the molecular structure of the GO sheet. The red groups in (c) highlight the difference from eo-GO sheet, as shown in Figure 7.2. Figure 7.4 158 Typical morphologies and structures of SnO2/eo-GO powders. (a-b) TEM image of a TinGO17 particle (a) with corresponding HRTEM image (b). (c-f) HAADF image (c) together with C (d), O (e), and Sn (f) elemental mappings. (g) SEM image of the TinGO17 particles with a statistical summary of the particle size distribution. The number mean diameter, standard deviation, and sample size are presented. (h) Box plots of size distributions of TinGOx powders. (i) EELS spectra of C K-edge (left) and Sn M4,5-edge (right) in core-loss level region. (j) Low-loss EELS spectrum of TinGO17 particles. Figure 7.5 161 SEM images and size distributions of particles fabricated from different precursor solutions. SEM images of particles produced from precursor solutions with SnO2/eo-GO ratios (mass) are 10 (a), 34 (c), and SnO2 only (e). Particle size- xvii distributions when SnO2/eo-GO ratios (mass) are 10 (b), 34 (d), and SnO2 only (f) in precursor solutions. Statistical summaries include number means, standard deviations, and sample sizes. Figure 7.6 163 EELS core loss spectra of graphite (blue) and monolayer graphene (black). Figure 7.7 165 (a) C 1s and K 2p XPS spectra of TinGO17 and eo-GO. (b) Sn 3d XPS spectra of TinGO17 and SnO2. Valence band spectra (c) and Tauc plots (d) of SnO2, TinGO17, and eo-GO photocatalysts. (e) Schematic illustration of photocatalytic reactions and reduction potentials of possible CO2 reduction pathways. (f) CH4 yields over TinGOx photocatalysts under irradiation of a Xe lamp for 4 h. Figure 7.8 166 C 1s and K 2p XPS spectra of SnO2. Figure 7.9 166 Diffuse reflectance spectra (DRS) of eo-GO, TinGO10, TinGO14, TinGO17, TinGO21, TinGO34, and SnO2. Figure 7.10 168 Absorption edges of TinGOx photocatalysts, eo-GO, and SnO2 determined by Tauc plots. Figure 7.11 170 N2 adsorption-desorption isotherm of TinGO10 (a) and TinGO17 (c) particles. (b) and (d) Pore size distributions derived from N2 desorption branches of (a) and (c) according to the BJH method. Figure 8.1 178 Schematic illustration of colloidal spray pyrolysis at high temperature. Figure A.1.1 181 SEM images of particles after being oxidized in a packed-bed reactor under 300 oC (a)-(c) and 500 oC (d)-(f). The volume percent of O2 in the feed stream is 1.4 %. The scale bars in SEM images are 2 μm. Figure A.1.2 188 Oxidation time (t) vs x of Cu particles with the assumption that the product of Cu oxidation is Cu2O (a) and CuO (b) at 300 oC. (c)-(d) Plots of oxidation time (t) vs x of CuSn0.05 (c) and CuSn0.1 (d) particles. xviii Chapter 1: Introduction and overview This dissertation is focused on developing aerosol-assisted processes to generate functional particles used in conductive pastes in solar cell metallization and interference packaging, anode and cathode materials for energy storage devices, and photocatalysts for energy conversion. In this chapter, motivations, strategies, and significant results of this work are outlined. 1.1 Motivations Aerosol-assisted approaches have long been recognized as promising for mass production of functional particles at an industrial-scale, because they are continuous one- step processes with facile equipment requirements and minimal waste production (Glicksman et al. 1996; Gurav et al. 1993; Nandiyanto and Okuyama 2011; Song et al. 2004; Zhong et al. 2013a). Therefore, aerosol-assisted methods have been intensively investigated to produce materials that can be used in inks in printed electronics and in the conductive pastes used in solar cell metallization and interference packaging (Comiskey et al. 1998; Faddoul et al. 2012; Glicksman et al. 1997; Glicksman et al. 1995; Hsien-Hsueh et al. 2005; Kim et al. 2004; Leenen et al. 2009; Li et al. 2005; Ma et al. 2014; Matienzo et al. 2008; Perelaer et al. 2010; Songping and Shuyuan 2006; Wang et al. 2007a; Wu et al. 2016; Yiwei et al. 2007; Zhong et al. 2012), anode and cathode materials in energy storage (Luo et al. 2017; Strobel and Pratsinis 2007; Xu et al. 2013a; Zhu et al. 2017), and photocatalysts (Nie et al. 2016; Strobel et al. 2006; Wang et al. 2011). However, several obstacles still exist and restrict the further application of aerosol-assisted methods in these fields. Although the formation mechanisms of various aerosol processes have been researched (Gurav et al. 1993; Nandiyanto and Okuyama 2011), the particle structure and 1 composition are difficult to design in advance (Shen et al. 2014b). Other issues are also remained to be addressed to promote the applications of aerosol-assisted methods in these applications. I will discuss these issues in this section. Spray pyrolysis has long been proven as an effective approach to fabricate particles used in the conductive pastes in solar cell metallization and interference packaging. The most common material used in these applications is silver (Ag). Ag powders can be facilely prepared by spray pyrolysis (Glicksman and Yang 2004; Kodas et al. 1995; Pingali et al. 2005; Pluym et al. 1993). However, in 2016, 562 million ounces of silver were used globally for industrial applications including electronics, photovoltaic metallization, and solders (O’Connell et al. 2017). Considering the high price of silver (18 USD/oz, (Bloomberg 2017)), replacing it with inexpensive materials, such as Cu (0.18 USD/oz, (Bloomberg 2017)) can greatly reduce manufacturing costs. To facilitate the scale up in industry, Cu particles have been successfully prepared by a continuous process by using a high flash-point co-solvent (ethylene glycol, EG) without direct addition of H2 gas (Zhong et al. 2012). Nevertheless, several challenges, such as the ease of oxidation of copper materials and the high mobility of copper in silicon (a common substrate in the electronics industry), have to be addressed before further application of copper-based powders (Istratov et al. 2000; Kim et al. 2004). For the applications of aerosol-assisted approaches in producing anode and cathode materials in energy storage, spray pyrolysis (Luo et al. 2017; Xu et al. 2013a; Zhu et al. 2017) and spray drying (Jung et al. 2013; Ma et al. 2015) have been successfully utilized. However, one prerequisite of aerosol-assisted methods to achieve uniform particle-to- particle composition and structure is a stable precursor solution, restricting their usage to 2 highly soluble salts for spray pyrolysis and reactions in single phase for spray drying. Unfortunately, for some promising anode materials, such as Sn (theoretical specific capacity: 993 mAhg-1 for Li-ion batteries (LIBs)) (Huang et al. 2015; Winter and Besenhard 1999), it is difficult to obtain stable and uniform aqueous solution containing tin salts, because tin salts are easily hydrolyzed in aqueous solutions (Perry 2011). For high-performance photocatalysts, titania (TiO2) has been intensively studied as the catalyst for the artificial photosynthesis since the pioneering work of Inoue et al. (Inoue et al. 1979). However, it exhibits low catalytic efficiency that can be ascribed to three factors: 1) the low CO2 adsorption on the surface, 2) rapid recombination of the photo-generated electrons and holes, and 3) catalyst deactivation due to adsorption of less reactive intermediates (Chang et al. 2016; Habisreutinger et al. 2013). On the other hand, the innovation of catalyst normally requires investment on the equipment. The fabrication processes are complex. For the photocatalyst fabrication, as the widespread approach, hydrothermal process endures the long processing time, safety concerns raised by the accumulation pressure in the reactor, and sophisticated post-production treatment (Hsieh et al. 2014; Ide et al. 2016; Sun et al. 2015). Alternative approaches have also been developed, including molecular beam epitaxy (MBE), sol-gel, and ion-exchange (AlOtaibi et al. 2015; Li et al. 2015a; Tahir and Amin 2015). Nevertheless, MBE requires significant equipment investment. Sol-gel and ion-exchange approaches have sophisticated operating procedures and long processing time. 3 1.2 Research strategies and objectives To address the aforementioned issues in Section 1.1, three aerosol-assisted processes will be developed to fabricate functional particles as conductive pastes in solar cell metallization and interference packaging, anode and cathode materials in energy storage devices, and photocatalysts in CO2 reduction. I will concentrate on process development, study of underlying mechanisms, and materials characterizations. In Chapter 2, I review approaches to develop oxidation-resistant Cu powders as alternative materials to Ag in applications including printed electronics. Then an overall description of the methods to fabricate anode materials is given. In the end, the reported strategies of preparing photocatalysts are provided. This chapter introduces the specific motivations of developing three different aerosol-assisted processes. In Chapter 3, a method will be presented to generate micron-sized Cu-Sn binary particles via spray pyrolysis. Copper nitrate and tin chloride are selected as the metal sources. Water and EG are used as solvent and co-solvent, respectively. The effects of experimental conditions (reaction temperature, EG concentration, residence time, and quench gas flow rate) on the final particle properties (morphologies, crystallinity, and metal distribution within the particles) are investigated. As a next step, in Chapter 4, we will present a detailed investigation of the oxidation behavior of CuSny (y is the atomic ratio of Sn/Cu in the product) solid particles fabricated by spray pyrolysis. A quantitative understanding of the phase evolution of CuSny particles during oxidation is obtained through Rietveld refinement analysis of the X-ray diffraction (XRD) patterns. To better understand the oxidation behavior of CuSny particles under high temperature, in-situ XRD measurements are conducted in ambient air. Kinetics 4 experiments are performed to determine the oxidation kinetics of CuSny particles and the oxidation mechanisms at different temperatures. Together, these results are used to identify the CuSny particle composition that has optimal oxidation resistance and spherical morphology In Chapter 5, I will assess the prospect of applying Cu-Sn powders in the inks of printed electronics by in-situ measurement of the resistivities of one-dimensional and two- dimensional CuSny structures. These structures are fabricated by direct printing inks containing CuSny powders. The resistivity measurements are conducted under various conditions including continuously changing the measurement temperature from 2 K to 400 K in oxygen-free environment and heating CuSny patterns at 573 K in ambient air for 90 min, enabling a systematic study of the effects of temperature and oxidation on the resistivities of CuSny structures. In Chapter 6, I will present colloidal spray pyrolysis (CSP), a process that can generate functional particles with uniform particle-to-particle composition and structure from stable multiphase precursor solutions, which could be a promising strategy to prepare uniform electrode materials in LIBs and sodium-ion batteries (SIBs). The structure and composition of particles are pre-designable based on our models. Several distinguishable advantages of CSP make it unique when compared to conventional aerosol processes. Unlike spray pyrolysis, CSP can process low solubility salts, expanding the application of spray pyrolysis to a wider range of functional materials. CSP uses an aqueous precursor solution, with no direct addition of H2 gas or pure organic solvents, which makes CSP safer and simpler for scale-up manufacturing. CSP is also different from multi-step spray drying, where colloids are used as templates to be etched away in the end (Boissiere et al. 2011; 5 Jung et al. 2013). CSP is a one-step process with a short residence time where colloids are involved in in-situ solid-state reactions and structural evolution. A novel photocatalyst engineering process will be developed in Chapter 7 to produce efficient nanocomposites as catalysts for artificial photosynthesis. Spray drying, a continuous and scalable process, is applied here to fabricate nanocomposites, comprising of eo-GO sheets and nano-sized SnO2 colloids. The obtained powders are termed TinGOx, where x is the mass ratio of SnO2 to eo-GO in the precursor. The eo-GO sheets have large intact sp2 carbon network to deliver large surface area, excellent electron transfer ability, sufficient active sites of negatively charged O atoms, and narrow band gap. The utilization of nano-sized SnO2 can efficiently trap the photogenerated electron and create a defect state to narrow its wide bandgap. State-of-art techniques are utilized to reveal the configuration of nanocomposites, including EELS in low-loss and core-level regions. A detailed investigation on the photoreduction of CO2 is conducted to understand the photocatalytic activity, optimal SnO2/eo-GO mass ratio, and reaction mechanism. The best TinGOx photocatalyst exhibits superior performance in photoreducing CO2 into CH4 with high yield and selectivity compared to bare eo-GO, SnO2, and the commercial TiO2 (P25). Based on the results and discussions in the previous chapters, conclusions are given for the process developments of aerosol-assisted approaches to fabricate functional particles in Chapter 8. Cu-Sn binary particles have been successfully produced by spray pyrolysis as conductive pastes in solar cell metallization and interference packaging. CSP has been utilized in generating Sn@C particles as anode materials. TinGOx photocatalysts have been fabricated by spray drying. Future work is also provided in this chapter to discuss possible projects that can be conducted based on the results presented in this dissertation. 6 Chapter 2: Background information In this chapter, a review will be presented on reported strategies to address issues in 1) developing alternatives to Ag particles in applications such as printed electronics, 2) generating high-performance anode materials, and 3) photocatalyst fabrication. 2.1 Developing alternatives to Ag particles in applications of printed electronics, interference packaging, and solar cell metallization Metal particles have been widely applied in printed electronics (Comiskey et al. 1998; Li et al. 2005), interference packaging (Faddoul et al. 2012; Wang et al. 2007a), and solar cell metallization (Mokkapati et al. 2009; Temple and Bagnall 2011). Gold and silver are by far the most commonly used materials in these fields due to their high electrical conductivity and oxidation resistance (Comiskey et al. 1998; Faddoul et al. 2012; Li et al. 2005; Mokkapati et al. 2009; R et al. 2015; Temple and Bagnall 2011; Wang et al. 2007a). However, the high cost of these noble metals has increased interest in developing alternative cost-efficient materials such as copper (Cu) (Huang and Sheen 1997; Songping and Shuyuan 2006; Yonezawa et al. 2015a; Yonezawa et al. 2015b; Zhong et al. 2012) for these applications. However, copper particles are easily oxidized even at room temperature, which could increase the resistivity (Kim et al. 2004). Particle engineering with an oxidation-resistant layer on a Cu core has been proposed as an effective configuration. The oxidation preventing layer can be secondary metals, for example, Au (Xu et al. 2012), Ag (Hai et al. 2013; Jung et al. 2011; Li et al. 2016a), Ni (Sharma et al. 2014), and Al (Noor et al. 2015). To synthesize Cu-Au nanoparticles, HAuCl4 and Cu(acac)2 were reduced in 1,2-hexadecanediol and oleylamine (Xu et al. 7 2012). The solution was then heated at 160 oC for 2 h. The Cu-Au nanoparticles were collected after washing the precipitates with ethanol and centrifugation. Hai et al. reported a technique to coat the Cu powder with an Ag layer (Hai et al. 2013). 5-μm-sized Cu particles were dispersed in distilled water under vigorous agitation. The surface oxide layers of Cu particles were removed by adding a solution comprising of ammonium hydroxide (NH4OH) and ammonium sulfate ((NH4)2SO4) into the powders. Then potassium tartrate (C4H4O6K2) was added to the solution. In the end, an aqueous solution containing AgNO3 and NH4OH was added dropwise to obtain the product. The particles were collected by centrifugation. All the procedures were completed in a closed system bubbled with Ar gas. Jung et al. used spray pyrolysis to fabricated Ag-coated Cu powders (Jung et al. 2011). An aqueous precursor solution containing AgNO3 and Cu(NO3)2 was atomized and carried into a reactor by H2 (5 vol.%) and Ar mixture gas. The reaction temperature was 900 oC and the residence time was 2.1 s in generating the Cu-Ag particles. Sharma et al. fabricated Cu-Ni nanopowders by flame-driven aerosol method (Sharma et al. 2014). Precursors containing copper nitrate and nickel nitrate were delivered into the throat of the nozzle. The high-velocity hot gas stream atomizes the precursor solution and assists the formation of Cu-Ni nanopowders. Cu-Al particles were synthesized by an improved electric explosion of wire (EEW) method (Noor et al. 2015). Al and Cu were twisted in wires and placed in the production chamber, which was filled with Ar gas. A high voltage power unit applied a high current in the order of 105 A/mm2 to the wires. The mixtures of superheated vapor and boiling droplets of the exploding wires were cooled down in Ar gas. Based on the above discussion, spray pyrolysis has its unique advantages in fabricating metal powders with simple operating procedures, short processing time, and 8 minimal waster production. Unfortunately, adding the reported secondary metals will increase the cost of the Cu particle manufacturing because of their high materials costs (Bloomberg 2017). Organic compounds have also been proposed as the passivation layers to protect Cu from being oxidized (Jianfeng et al. 2011; Yonezawa et al. 2015a). Jianfeng et al. utilized a polyol method to synthesize Cu nanoparticles (Jianfeng et al. 2011). A mixture of polyvinylpyrrolidone (PVP), sodium hypophosphite, and EG was heated at 90 oC. A solution of copper nitrate in EG liquid was then added into the previous mixture. After the formation of PVP-coated Cu nanoparticles, they were centrifuged and washed several times. Yonezawa et al. mixed gelatin and CuO microparticles in hot water. An aqueous ammonia solution was introduced to adjust the pH to 11. After the formation of fine particles, the pH of the dispersion was reduced to 8.5 by adding saturated aqueous citric acid into the mixture. Then gelatin-coated Cu powders were collected by decantation. Before using these particles, the organic coating needs to be removed because the organic layer on the Cu particle surface could result in a significant increase of the resistivity. Therefore, there is a need to explore other strategies from a comprehensive view of cost and electrical conductivity. Jeong et al. reported that by tuning the molecular weight of poly(N-vinylpyrrolidone), which is the capping molecules, the Cu nanoparticles fabricated from polyol method can have minimal surface oxidation (Jeong et al. 2008). The surface oxidized layer can prevent the further oxidation of Cu particle (Jeong et al. 2008). Although this particle configuration has been proven can fabricate low-resistivity patterns, the sophisticated particle generation process (polyol method) requires other particle fabrication process. Based on this work, I 9 propose that utilizing a reactive metal to form the sacrificial oxide layer can effectively increase the oxidation resistance of the particles. One obstacle in this strategy is to find a reactive metal, whose oxide has relatively high conductivity. Based on these ideas, I developed a spray pyrolysis process to fabricate Cu-Sn powders with inexpensive raw materials, one-step operating, and enhanced oxidation-resistance. Details will be presented in Chapters 3-5. 2.2 Approaches to fabricate high-performance anode materials As discussed by Goriparti et al., alloys and metals (such as Sn, Ge, and Sb) are promising anode materials in LIBs with high capacity density (Goriparti et al. 2014). Nevertheless, the salts of these metals can easily hydrolyze in aqueous solutions (Table 2.1). The hydrolysis of these salts increases the instability of the precursor solutions, resulting in the early precipitation in the precursor solutions. This could lead to the uncontrollable particle-to-particle composition and structure. Table 2.1. Salt solubilities of corresponding metals as high-performance battery materials listed in (Goriparti et al. 2014) Metal salts Solubility Si(OC2H5)4 (TEOS) Reacts with H2O GeCl4 Hydrolysis H2O AlCl3 Dissolve in H2O evolving HCl SnCl4 Hydrolysis H2O SbCl2 9881. g/ 100 g H2O *The solubility data are summarized from (Perry 2011). 10 Reactions between phases are more difficult and require significantly longer processing time than those within single phases because of interphase mass transport resistance. It will be significantly more difficult for a continuous and scalable process to produce materials from multiphase reaction system than single phase reaction system. Several techniques have been reported to produce the high-capacity materials. Here, I use Sn as an example, since these fabrication techniques can also be used to produce other metals. The hydrothermal approach has been reported to synthesize Sn/C composite particles (Youn et al. 2016). SnCl4 and citric acid monohydrate were mixed and dissolved in water and stirred for 3 h. Then the mixture was heated at 300 oC in a muffle furnace for 1.5 h, following by calcining the precipitates under flowing Ar at 650 oC for 3h. Wet chemistry method has also been applied to produce Sn-based materials for batteries (Derrien et al. 2007; Eom et al. 2015; Zhu et al. 2014a). SnCl4 and trisodium citrate dehydrate were dissolved in EG, then NaBH4 was added into the solution to reduce SnCl4 (Eom et al. 2015). Methanol and deionized water were used to wash the products. After the product was washed and centrifuged for 5 times, it was collected and dried at 120 oC for 24 h under vacuum. Nano-Sn powders were then formed. In another method, SnCl2 and salenH2 ligand were dissolved into ethanol (Zhu et al. 2014a). Triethylamine was added to the solution, then the mixture was stirred at 85 oC for 24 h. The product was washed and filtered by ethanol and then sintered at 650 oC for 2 h in Ar. Derrien et al. mixed resorcinol, sodium carbonate, and formaldehyde in water for 2 h at 70 oC (Derrien et al. 2007). The product was washed with water and then immersed in tertiobutanol and filtered for several times. Then it was stirred in tributylphenyltin (TBPT) and tertiobutanol overnight. In the 11 end, the mixture was placed in Ar flux at 700 oC for 2 h to form the nanostructured Sn-C composite. Alternative approaches have also been proposed to fabricate Sn-based materials. For example, a solution containing citric acid, SnCl2, NaCl, and H2O was stored at -20 °C for 24 h (Qin et al. 2014). Then freeze-drying technology was used to remove the H2O. The dry gel was then ground to powders and annealed at 750 oC for 2 h under H2. After washing the sample several times with water, 3D Sn@G-PGNWs were obtained. Tao et al added SnCl4·5H2O and s. platensis in absolute ethanol to form precursor solution (Tao et al. 2014). Then, the solution was filtered and sintered at 80 oC. The powders were later mixed with a solution prepared by dissolving polystyrene (PS) in dimethylformamide. After calcining the sample at 600 °C in N2 for 3 h, anode materials were prepared. Based on the above discussion, it can be seen the reported methods have sophisticated experimental procedures and long operating time. Therefore, spray pyrolysis has been utilized to facilitate the potential scale-up production (Hong and Kang 2015; Xu et al. 2013a; Zhang et al. 2014a). Xu et al. dissolved SnCl2 and PVP into ethanol (Xu et al. 2013a). The precursor solution was ultrasonically atomized into droplets and transported into furnaces by Ar-H2 mixture gas flow. The reaction temperature is 900 oC. The process of spray pyrolysis reported by Zhang et al. is complicated (Zhang et al. 2014a). The resorcinol and formaldehyde were polymerized to form resorcinol formaldehyde. It was mixed with SnCl2 in ethanol to form the precursor solution. The precursor solution was atomized and carried by Ar gas into the furnace with the set point of 800 oC. Hong et al. prepared the precursor solution by dissolving Sn (II) oxalate (C2O4Sn) and PVP (Kanto, 12 Mw: 40,000) into water (Hong and Kang 2015). The spray pyrolysis was conducted at 650 oC with N2 as the carrier gas. As we can find in the above discussions, the reduction of tin salts to tin is challenging. To ensure the single-phase reaction, pure organic solvents are commonly utilized to dissolve the tin salts. In addition, the operating procedures are complicated and time- consuming. Even though spray pyrolysis has been applied to fabricate Sn-based materials, pure organic solvents are still utilized. Moreover, H2 gas is added directly into the reaction system to guarantee the complete reduction of tin salts to tin. All of these issues need to be addressed to promote the mass-production of Sn-based anode materials. In that case, a new process, CSP, will be discussed in Chapter 6, to continuously generate the Sn-based composite particles with mild reaction condition. 2.3 Methods to fabricate photocatalysts Up to now, the hydrothermal approach has been the most commonly utilized to fabricate photocatalysts (Hsieh et al. 2014; Ide et al. 2016; Sun et al. 2015). Hsieh et al. firstly made a solution containing TaCl5, iced H2O, KOH, and methanol (Hsieh et al. 2014). Then this solution was mixed with InCl3 and hydrothermally heated at 150 oC for 4 h. After the reaction, the products were washed and centrifuged several times and dried at 90 oC to obtain tantalum-based pyrochlore and indium hydroxide. Sun et al. prepared a solution containing CdCl2, NH2CSNH2 and ethylenediamine (Sun et al. 2015). The solution was hydrothermally heated in a Teflon-lined, stainless-steel autoclave at 160 oC for 48 h. Then, the products were mixed with Ni(NO3)2, yellow phosphorus, and ethylenediamine to make a solution, which is hydrothermally heated in Teflon-lined, stainless-steel autoclave for the 13 second time at 140 oC for 12 h. The powders were dried in vacuum overnight at 60 oC to obtain Ni2P/CdS nanorods. Ide et al. hydrothermally heated a solution containing TiO2 (P25), tetrapropylammonium hydroxide (TPA), H2O, and NH4F in a Teflon-lined stainless- steel autoclave 170 oC for 7 days (Ide et al. 2016). The products were then washed and centrifuged several times to obtain the zeolite-TiO2. Alternative approaches have also been reported to fabricate Ti-based photocatalysts. Li et al rapidly heated the iced aqueous solution containing TiCl4 to 100 oC within 5 min (Li et al. 2015b). Then the solution was dried in air for 24 h at 80 oC. The product was calcined in a muffle furnace at 400 oC for 2h to form the sub-10 nm rutile titanium dioxide nanoparticles. Tan et al. heated the solutions containing graphene oxide (GO), TiO2, and H2O at 80 oC for 2h (Tan et al. 2015). Then the product was dried in air overnight to form the oxygen-rich TiO2/GO. Long et al. stirred the solution containing TiO2 nanosheets, PVP, ascorbic acid (AA), and KBr in an oil bath at 80 oC for 10 min (Long et al. 2017). Then, K o2PdCl4 and CuCl2∙were added into the previous solutions and heated in air at 80 C for 3 h. PdxCu1-TiO2 hybrid was obtained after washing and centrifugation several times of the previous solids. Tahir et al. used the sol-gel method to fabricate In-doped TiO2 (Tahir and Amin 2015). A solution containing Ti(C4H9O)4, C2H5OH, CH3COOH, isopropanol, and acetic acid was hydrolyzed at 30 oC for 24 h. Then indium nitrate was added to the previous solution and stirred for 12 h. After that, the powder was dried in an oven in air at 80 oC for 12 h. In the end, the product was calcined in a muffle furnace at 500 oC for 5h. Other photocatalysts have also been fabricated by non-hydrothermal methods. Li et al. dried a mixture at 80 oC overnight (Li et al. 2016b). The mixture contained containing urea, Ce(NO3)3, SiO2 particles, and water. The powders were calcined at 550 oC for 2h. Then, 14 the product was reacted with NH4HF2 to form CeO2/g-C3N4. Shown et al. used the microwave approach to fabricate Cu/GO hybrid (Shown et al. 2014). Suspensions containing GO, Cu(NO3)2, and EG was microwaved for 180 s under 350 W. EG was removed by centrifugation. Then, the solids were dried under vacuum overnight to form the final product. In preparation of BiVO4/V2O5, the purchased NH4BEA zeolite was firstly calcined in air at 500 oC for 6 h (Murcia-López et al. 2017). Then the powders were dissolved into a solution containing VOSO4 and water. After that, the solution was dried overnight at 80 oC and calcined at 500 oC for 6 h in dry air flow. The solids were dissolved into a solution containing Bi(NO3)3 and water. The solution was dried overnight at 80 oC, calcined at 500 oC for 6 h in dry air flow to form the final product. To fabricate SrTiO3:La, Rh/Au/BiVO4:Mo photocatalyst sheet, Wang et al. suspended SrTiO3:La (La-doped SrTiO3), BiVO4:Mo (Mo-doped BiVO4), and Rh in isopropanol and drop-cast on a glass substrate (Wang et al. 2016). After drying at room temperature, a thin gold layer was deposited by thermal vacuum evaporation at 400 oC with a rate of 20 nm/s under a base pressure of 2.6×10-3 Pa. The final product was obtained by heating the sample at 200 oC to 500 oC for 10 to 30 min. AlOtaibi et al. used the MBE method to grow Rh/Cr2O3 on GaN nanowires (AlOtaibi et al. 2015). The Si wafer was firstly cleaned by acetone, methanol, and HF. Then GaN nanowire was grown by MBE on Si substrate. Precursor solutions containing Na3RhC6, CH3OH, K2CrO4, chloroplatinic acid, CH3OH, and water were used to photodeposit Rh/Cr2O3 on GaN nanowires under irradiation of 300 W xenon lamp to obtained the desired product. The hydrothermal method normally results in accumulation of pressure in the reactor, resulting in the potential safety concern. Other approaches exhibit complicated 15 experimental procedures, time-consuming process, and huge instrumental investment. Therefore, in Chapter 7, a spray drying process was developed to fabricate high- performance photocatalysts by one-step operating. 16 Chapter 3: Cu-Sn binary particle generation by spray pyrolysis 3.1 Abstract Cu-Sn binary particles were generated via spray pyrolysis from metal salt precursors with ethylene glycol as the co-solvent and reducing agent. The morphology, crystallinity, and elemental distribution of particles were tunable by changing the reaction temperature, residence time, and quench gas flow rate. Hollow porous particles were fabricated with a higher Sn concentration on the particle surface when the furnace set point was 500 oC, while solid particles with a lower surface Sn concentration were generated when the furnace set point was 1000 oC. Particles with spherical morphologies were obtained at long residence time conditions (4.5 s). Cu-Sn binary particles with irregular structures (e.g., pores on the particle surface, fragmented spherical particles, and lamellar fragments) were formed at short residence time conditions (0.92 s). A possible spray pyrolysis mechanism was proposed that incorporates chemical reaction steps and structural progression. By this mechanism, the metal salts are believed to sequentially undergo hydrolysis to metal hydroxides, decomposition to metal oxides, reduction to metals, and finally diffusion of Sn into the Cu matrix to generate the Cu-Sn solid solution. 3.2 Introduction Metal particles have attracted a large amount of interest due to their wide application in the fields of electronics, catalysis, and plasmonics (Kim et al. 2002; Li et al. 2005; Talley et al. 2005). It is of great importance to develop scalable processes for the fabrication of metal particles with controllable morphologies, crystal phases, and composition. Plenty of methods have been reported to solve this problem, e.g., seed-growth procedures (Cho et al. 17 2010; Xie et al. 2014), polyol processes (Fievet et al. 1993; Park et al. 2007), sonochemical treatments (Dhas et al. 1998; Fujimoto et al. 2001), and spray pyrolysis (Kim et al. 2006; Pluym et al. 1993; Xia et al. 2001a). Spray pyrolysis is advantageous because it is a continuous one-step process with facile equipment requirements and less waste production (Song et al. 2004; Zhong et al. 2013a). Spray pyrolysis consists of three steps: (i) atomization of a precursor into droplets; (ii) reaction and solvent evaporation in the isolated droplets surrounded by carrier gas; and iii) final product collection. As the precursor droplets have the same droplet-to-droplet composition, spray pyrolysis can be used to obtain relatively uniform powders containing single or multiple components (Lenggoro et al. 2000). Various metals, e.g., Ag (Pingali et al. 2005), Ni (Xia et al. 2000; 2001a), Co (Gürmen et al. 2006), and Cu (Kim et al. 2003; Zhong et al. 2012), have been synthesized by this technique. To produce the metals from metal salts, H2 was first introduced with carrier gas to generate oxide-free metal powders (Gurav et al. 1993). However, addition of H2 above the lower explosion limit was required (Kim et al. 2003; Zhong et al. 2012). To solve this problem, ammonium ion has been used as a reducing agent to obtain metal particles (Xia et al. 2000). Nevertheless, this method required the precise control of the molar ratio of ammonium to metal salt in the precursor and the removal of residual ammonium salts from the product (Xia et al. 2000; 2001a). Co-solvents, e.g., formic acid and ethanol have also been successfully utilized to fabricate the metals (Kim et al. 2003; Xia et al. 2001b). However, the low flash points of formic acid (69 oC) and ethanol (13 oC) have led to exploration of alternative co-solvents (Yaws 1999). Recently, EG has attracted attention 18 for its higher flash point (111 oC) and reducing ability in metallic powder fabrication (Yaws 1999; Zhong et al. 2013a). Copper particles, especially solid and spherical particles with narrow size distribution, have potential for applications in conductive pastes, integrated materials, and interference packaging (Huang and Sheen 1997; Youngil et al. 2008; Zhong et al. 2012). They are favored because of the relatively inexpensive raw materials, high solubility salts (Cu(NO3) o 2∙3H2O: 145 g/100 g H2O, 25 C) (Lide 2007-2008), and low resistivity (nanocrystalline Cu, 1.8×10-7 Ω∙m, 300 K) (Lu et al. 2004). However, some obstacles still need to be addressed before their further application: (i) copper particles will oxidize at room temperature once exposed to air, increasing the resistivity (Kim et al. 2004), and (ii) high copper mobility in silicon could reduce device efficiency (copper can diffuse through a standard 0.625 mm thick intrinsic silicon wafer in about 3 h at room temperature) (Istratov et al. 2000; Morinaga et al. 1996). The diffusivity of Cu in Si is 4×10-2 cm2/s at 1073 K, and the diffusivity of Si in Cu is 3.7×10-2 cm2/s at the same temperature (Degang et al. 1998). For reference, the thickness of a 4-inch Si wafer, widely used in the electronics industry, is only 0.525 mm (semiwafer.com). To remedy these challenges, CucoreAgshell particles were fabricated to prevent copper oxidation while maintaining high conductivity (Grouchko et al. 2009). However, there is still a need to explore other methods utilizing less expensive materials. Tin was selected as the secondary material in copper-based materials because of its higher reaction activity towards oxidation than copper. It has been shown that the formation of sacrificial tin oxides can protect copper from being oxidized while maintaining relatively high conductivity in the case of Cu-Sn binary nanowires (Chen et al. 2014). Another benefit 19 of doping with tin is the low diffusivity of copper in tin. Tin and tin oxides have been reported as effective copper diffusion barrier materials (Liu et al. 2005; Liu et al. 2007a; Mei et al. 1992). However, reported methods for Cu-Sn material synthesis focus on producing nano-sized particles (Cao et al. 2014; Jo et al. 2011) and thin films (Kumar et al. 2011; Polat et al. 2014). Therefore, it is desirable to develop a continuous and potentially scalable process to fabricate micron-sized Cu-Sn powders, because they are favored for applications such as conductive pastes and interference packaging (Deshpande et al. 2005; Wu et al. 2009). Here, a method is presented to generate micron-sized Cu-Sn binary particles via spray pyrolysis. In this work, Cu(NO3)2 and SnCl2 were selected as the metal sources. Water and EG were used as solvent and co-solvent, respectively. The effects of experimental conditions (reaction temperature, EG concentration, residence time, and quench gas flow rate) on the final particle properties (morphologies, crystallinity, and metal distribution within the particles) were investigated. 3.3 Experimental 3.3.1 Preparation of precursor All chemicals were purchased from the companies listed and used without further purification. Cu(NO3)2∙3H2O (99.5% purity, Strem Chemicals, Newburyport, MA, USA) was first dissolved into distilled water. SnCl2∙2H2O (Fisher Scientific, Waltham, MA, USA) was added into EG (≥99%, Sigma Aldrich, St. Louis, MO, USA). The two solutions were mixed to make a solution of 1 M Cu(NO3)2, 0.1 M SnCl2, and 3.6 M, 4.8 M, or 5.9 M EG based on different experiments. To ease the hydrolysis of SnCl2, HNO3 (SIGMA- ALDRICH, 70%) was added with a final concentration of 0.1 M. 20 3.3.2 Spray pyrolysis procedures The setup of the spray pyrolysis is the same as reported in previous work (Zhong et al. 2012). In brief, precursor was transported into the atomizer (with a 1.7 MHz ultrasonic generator) by a peristaltic pump. The atomized droplets of precursor had a volume average diameter ~5 µm (Zhong et al. 2012). Once the droplets were formed, they were transported by carrier gas (N2, 99.5%, Arigas, Rador, PA, USA) into a quartz tube (inner diameter = 19 mm), which was heated by two furnaces in series (total heated length = 81.3 cm). The carrier gas flow rate (CGFR) was varied from 1 L/min to 5 L/min based on the experimental design with the residence times shown in Figure 3.1c. At the end of the tube, a polytetrafluoroethylene (PTFE) filter (with a diameter of ~9.5 cm) was utilized to collect the powders. Quench gas (N2, AIRGAS, 99.5%) was introduced at ~8 L/min to reduce the temperature at the filter. To better understand the heating conditions in the system, the temperature along the centerline of the quartz tube and immediately downstream of the filter was measured within the heated range (81.3 cm) by a K-type thermocouple. Distilled water was atomized as a surrogate for the precursor during the measurement. N2 gas was used as the carrier gas, the same as during the actual particle generation process. To estimate the residence time of the double furnace reactor of 81.3 cm length, N = 32 calculation elements were selected. It was assumed that the temperature was uniform within each element. Thus, the average time required for each droplet (or particle) to traverse the element can be calculated using a plug-flow reactor approach. The reactor can be divided into a number of sub-volumes along the axial direction. In each sub-volume the 21 reaction extent is treated as spatially uniform (Fogler 1999). Therefore, residence time per sub-volume can be calculated as: 𝐴×𝑙 𝑇 𝑡 = × 𝑟𝑜𝑜𝑚𝑖 (3.1) 𝐹 𝑇𝑖 Here, A is the cross sectional area of the tube, l is the length of the calculation element, F is the CGFR, Troom is the room temperature, and Ti is the temperature in each element which can be found in Figure 3.1a-b. Thus, the total residence time (τ) can be estimated as: 𝜏 = ∑𝑁1 𝑡𝑖 (3.2) 22 Figure 3.1. Temperature profile of a double furnace reactor of 81.3 cm in length with (a) fixed CGFR at 3 L/min and various temperatures; (b) fixed temperature at 750 oC and different CGFRs; (c) estimated residence time. (d) and (e) The heating rates in the reactor with respect to the conditions in (a) and (b), respectively. 3.3.3 Mechanism exploration Conditions of high temperatures (500 oC, 750 oC, and 1000 oC) and long residence times (~0.92 s) were utilized to fabricate the Cu-Sn binary powders, as discussed in the previous section. To investigate the mechanism of the particle formation process, 23 experiments were conducted at lower temperatures and shorter residence times to identify reaction intermediates. The precursor contained 1 M Cu(NO3)2, 0.1 M SnCl2, and 0.1 M HNO3. EG was added with a concentration of 5.9 M. The spray pyrolysis was conducted at i) 300 oC with CGFR = 5 L/min and quench gas at ~5.2 L/min; and ii) 500 oC with CGFR = 10 L/min and quench gas at ~10.8 L/min. During those two experiments, only a single furnace was used, and the residence time was less than 0.74 s in both cases. 3.3.4 Characterization and data analysis Scanning electron microscopy (SEM, Hitachi SU-70, Hitachi, Tokyo, Japan) was used to observe the particle morphologies. Transmission electron microscopy (TEM, Jeol Jem 2100 LaB6, Jeol Jem, Tokyo, Japan) was utilized to understand the details of the particle structures. Before conducting the SEM and TEM observations, powders were suspended in ethanol. After sonication for ~1 min, the mixture was added dropwise onto a silicon wafer or TEM grid, and dried under ambient conditions. X-ray diffractometer (XRD, Bruker D8 advance, Bruker, Billerica, MA, USA) was used to obtain crystallographic information of powders. Particles were also examined by X-ray photoelectron spectroscopy (XPS, Kratos Axis 165, Kratos Axis, Manchester, UK) to determine the elemental distribution on the powder surface. In addition, the powders were sputtered in vacuum by argon ions for 20 min to collect the sub-surface concentration profile. The sputtering gun was set at 4 kV and 5 mA. 3.4 Results and discussion 3.4.1 Reaction temperature To identify the composition of the products, powder XRD with Cu Ka radiation was utilized, as shown in Figure 3.2. When the set point of the furnaces was 500 oC, Cu and 24 Cu13.7Sn were the main phases. The residence time of the process was 2 s. When the temperature increased to 750 oC, the residence time of the process was 1.5 s. Powders displayed higher quality crystallinity with sharp peaks of Cu13.7Sn without Cu peaks. However, there was one peak at 32.6o, which could not be identified as Cu or Cu13.7Sn. It may be attributed to intermediate salt phase(s) of small quantity, with low peak intensity compared to the major Cu13.7Sn peaks. The increase of temperature could promote the forming of a Cu-Sn solid solution, which is thermodynamically favorable. A similar phenomenon has also been reported in other alloy particles that a particle composed of a single alloy phase will have lower free energy compared to a particle containing multiple phases (Zhong et al. 2013a). When the temperature was increased to 1000 oC, the residence time was 1.2 s. The intensity of the unidentified peak (32.6o) was negligible compared to the strong peaks of Cu13.7Sn. The XRD diagram exhibited high quality crystallinity with Fm-3m symmetry, implying complete reaction to generate Cu13.7Sn particles. Figure 3.2. XRD results of powders generated from systematic conditions. The precursor contained 1 M Cu(NO3)2, 0.1 M SnCl2, 0.1 M HNO3, and 4.8 M EG. The CGFR was fixed at 3 L/min. The powders were generated at 500 oC, 750 oC, and 1000 oC. The major peaks are from Cu13.7Sn (ICDD with PDF No. 03-065-6821) and Cu (ICDD PDF No. 01-070- 3038). 25 The morphologies of particles generated at different reaction temperatures with 4.8 M EG in the precursor are shown in Figure 3.3 with additional images in Figure 3.4. When the temperature was set at 500 oC, hollow porous particles could be seen with primary particles (marked by solid line circle) on the surface, as shown in Figure 3.3a. The number mean diameter of the hollow porous particles and the primary particles are 1.5 µm and 150 nm, respectively (Figure 3.3d). The TEM observations of these particles further validate the porous hollow structure with primary particles scattered within the shell, as shown in Figure 3.5a. Figure 3.3. SEM images of powders generated from the precursor solution with 1 M Cu(NO3)2, 0.1 M SnCl2, 0.1 M HNO3, and 4.8 M EG. The reaction temperatures were 500 oC (a), 750 oC (b), and 1000 oC (c). The CGFR was fixed at 3 L/min. Insets are higher magnification images of selected regions marked by solid arrows. The scale bars in all figures are 1 µm. (d) The summary of the diameter of the particles marked as 1, 2, and 3 in (a)–(c). Detailed size distributions for each condition can be found in Figure 3.6. 26 Figure 3.4. Samples generated at 750 oC with EG concentration of 3.6 M (a), 4.8 M (b)- (c), and 5.9 M (d)-(e). The concentrations of Cu(NO3)2, SnCl2, and HNO3 were 1 M, 0.1 M, and 0.1 M, respectively. The scale bars of the SEM images are 1 μm. Figure 3.5. TEM images of particles fabricated by precursors with 1 M Cu(NO3)2, 0.1 M SnCl2, and 0.1 M HNO3. The CGFR was fixed at 3 L/min. (a)–(b) Particles generated at 27 500 oC, with 5.9 M EG. (c)–(d) Particles obtained at 750 oC, with 4.8 M EG. (e)–(f) Particles obtained at 1000 oC, with 4.8 M EG. Figure 3.6. Diameter distributions of the products generated at systematic conditions. The concentrations of Cu(NO3)2, SnCl2, and HNO3 were 1 M, 0.1 M, and 0.1 M, respectively. 28 According to reported results, the hollow structures were caused by a solute concentration gradient within the droplets which led to solute precipitation on the surface of the droplets (Gurav et al. 1993). The granular precipitates within the precursor droplets could concentrate on the droplet surface for the maximum structural stability during the solvent evaporation (Nandiyanto and Okuyama 2011). After the solvents vaporized and solutes reduced to metals, hollow porous particles were formed with primary particles comprising the shell (morphologies shown in Figure 3.3a). Pressure may accumulate within the shell during evaporation of the solvent, leading the droplets to burst (Gurav et al. 1993). Thus, semispherical particles can be observed, as shown in the inset of Figure 3.3a. Figure 3.7. SEM images of powders generated from systematic conditions. EG concentration was tuned at 3.6 M, 4.8 M, and 5.9 M at 500 oC, 750 oC, and 1000 oC, respectively. The concentrations of Cu(NO3)2, SnCl2, and HNO3 in the precursor were 1 29 M, 0.1 M, and 0.1 M, respectively. The CGFR was fixed at 3 L/min. Insets highlighted by yellow squares are higher magnification images of selected regions marked by yellow arrows. The scale bars in all figures are 1 μm. When the furnace set point temperature was increased to 750 oC, the amount of primary particles was significantly decreased. Instead, hollow particles with a crust on the surface can be seen in Figure 3.3b. The particles in the TEM image (Figure 3.5c) also exhibit the hollow structure, with the dark part representing the crust and bright area (marked by white arrow) representing the hollow core. The number mean diameter of these particles is 0.68 µm (Figure 3.3d), which is smaller than that of the hollow porous particles generated at the 500 oC. The decrease of the number mean diameter supports the hypothesis that the hollow crusted particles in Figure 3.3b are caused by the coagulation of the primary particles and shrinking of the hollow porous structure in Figure 3.3a. Size can affect the melting point of Cu and Sn (Ju et al. 2015; Karabacak et al. 2006). The reported melting point of bulk tin is ~230 oC (Ju et al. 2015; Wronski 1967). The melting of bulk Cu-Sn alloy ranges from 1085 oC (Sn at % = 0) to 945 oC (Sn at % = 10%) (Saunders and Miodownik 1990). Copper nanorods with a length of 2.3 µm and a diameter of 100 nm have also been found to melt below 550 oC (Karabacak et al. 2006). The fragmented spherical particles may stem from the melting of the fragmented hollow particles (inset of Figure 3.3a) obtained at 500 oC. Thus, other morphologies, e.g., spherical particles with pores (inset of Figure 3.7d) and fragments (marked by a dashed arrow in Figure 3.3b, and circled by dashed lines in Figure 3.4c) were seen at this temperature, which indicates these particles may stem from the coagulation of the primary particles in Figure 3.3a. Based on Figures 3.3b,3.4, it can be concluded that the morphologies of particles at 750 oC showed the evolution from hollow 30 to solid. The hollow to solid transition occurred between 500 oC and 1000 oC. In spray pyrolysis, reaction temperature is a critical factor influencing the particle structure. With the increase of temperature, particle structure could evolve either from hollow to solid (Ortega et al. 1991) or from solid to hollow (Eslamian and Ashgriz 2006) based on the chemical properties and precipitation rates of the salts in the precursor (Eslamian and Ashgriz 2006). In our system, the evaporation of HNO3 (Pvap, HNO3 = 172 kPa at 373 K) from precursor droplet into the gas phase promotes hydrolysis of SnCl2 (Yaws 1995). Based on the prior reports, SnCl2 prefers to hydrolyze to form hydroxides and decompose to SnO, even at room temperature (Cohen and West 1972). SnO is likely formed at the droplet surface. Thus, with the increase of reaction temperature, the Cu-Sn particles shifted from hollow to solid due to the melting of the primary particles and densification of the Cu-Sn particles. This mechanism may also be applicable to other systems with rapidly precipitating components. Small particles can be seen on the surface of the hollow crusted particles at 750 oC (inset of Figure 3.3b, solid line circle). The small particles, with number mean diameter of 0.13 µm (Figure 3.3d), were believed to form by gas-to-particle conversion, which has been reported in spray pyrolysis (Fotou et al. 2000; Schulz et al. 2005; Tsai et al. 2004). One possible formation mechanism is that the volatized Cu(NO3)2 vaporizes into the gas once it has precipitated from the solvent. The vapor pressure of Cu(NO3)2 at 495.8 K is 480 Pa (Addison and Hathaway 1958). Cu(NO3)2 may vaporize below 500 K and decompose to CuO, NO2, and O2 above 500 K (Stern 1972). CuO may nucleate into small oxide particles in the gas phase. With the reducing environment created by EG, it is rational to believe the CuO will be reduced to Cu, and then the small particles, as shown in Figure 31 3.3b,c, were formed. For reference, the vapor pressures of CuO and SnO at 1000 oC are P -2vap,CuO = 1.1×10 Pa (Mack et al. 1923) and Pvap,SnO = 370 Pa (Platteeuw and Meyer 1956) and the vapor pressures of the pure metals are even lower, Pvap,Sn = 1.5×10 -2 Pa and Pvap,Cu = 6.0×10-3 Pa, at 1000 oC (Smithells et al. 2004). Another possible mechanism is that after the solvent in the precursor droplets evaporated, the Cu and Sn may be reduced by the EG. The metals in the particles would vaporize, nucleate, and grow into particles with a diameter less than 100 nm in the gas phase. A similar mechanism has also been reported by other researchers (Tsai et al. 2004). In the TEM observation, the small particles were found to be pure Cu particles (Figure 3.5d), which further validated the hypothesis that the small particles are likely generated by gas-to-particle conversion. When the temperature was increased to 1000 oC, the particles underwent further densification to form spherical solid particles (Figure 3.3c) with a number mean diameter of 0.71 µm (Figure 3.3d). The TEM image (Figure 3.5e) also provides evidence of this conclusion. Small particles are still seen with number mean diameter of 0.13 µm on the particle surfaces, which are also believed to be formed by gas-to-particle conversion. The SnO2 seen in Figure 3.5f is believed to be caused by the oxidation of the Sn on the particle surface. To further understand the elemental distributions and chemical states on and beneath the Cu-Sn particle surface, XPS analysis was utilized together with ion-beam sputtering. As shown in Figure 3.8, the Cu/Sn ratio was lower on the surface (before sputtering) than under the surface (after sputtering). Both cases exhibited Sn enrichment compared to the Cu/Sn ratios in the precursor and the major crystal phase (Cu13.7Sn). In addition, the low temperature led to a lower Cu/Sn ratio than the high temperature. This Sn enrichment on 32 the particle surface is likely responsible for the difference in the Cu/Sn ratio observed from XRD measurements (13.7) compared to that in the precursor (10). A mass balance calculation was made assuming that the Cu-Sn particles were encapsulated by an amorphous Sn layer that would not appear in the XRD. From this calculation, the overall Cu/Sn ratio in the product particles is maintained at 10, if the ratio of the calculated thickness of Sn crust to the radius of Cu-Sn core (with Cu/Sn ratio = 13.7) is 0.008. Therefore, for particles obtained at 1000 oC, with a mean diameter of 710 nm (Figure 3.6), a thickness of 3 nm Sn enrichment layer would explain the Cu/Sn ratio observed with XRD. Thus, both the calculated and XPS results indicates that Sn accumulates on the particle surface (Figure 3.8). Figure 3.8. The Cu/Sn ratios summarized from XPS results. The precursor contained 1 M Cu(NO3)2, 0.1 M SnCl2, 0.1 M HNO3, and 4.8 M EG. The CGFR was set at 3 L/min. Sputtering was conducted in vacuum by argon ions for 20 min. The dominant tin species on the particles surface can be identified as SnO2 (Figure 3.9a,d,g, before sputtering). After sputtering the XPS peaks representing metallic tin increased significantly in all tested samples. For copper, the signal from Cu 2p is shown in 33 Figure 3.9b,e,h (after sputtering). From the reference, metallic copper has a major peak at 932.4 eV (Fleisch and Mains 1982), which is marked by a grey line in Figure 3.9b,e,h. It can be seen that the powders contained metallic copper on and beneath the particle surface at all three temperatures. To better understand the copper states on and beneath the sample surface, the signals from Cu LMM auger electrons were collected. The shoulder peaks around 565 eV, as shown in Figure 3.9c,f,i, are strong evidence for the existence of metallic copper (Fleisch and Mains 1982), However, the peaks of the 1000 oC sample were weak around 565 eV (Figure 3.9c, before sputtering), which means the metallic copper presents a small amount on the surface. Peaks around 570 eV are evidence for the existence of copper oxides. Unfortunately, as the shape and peak position of CuO and Cu2O are similar (Fleisch and Mains 1982), it is hard to distinguish between them in all samples. Results shown here can only confirm that there are copper oxides on the surface. However, a qualitative conclusion still can be made based on the relative intensity of the peak around 570 eV to the peak around 565 eV. The amount of copper oxides increased with temperature. The Cu/Sn ratio increased on the Cu-Sn binary particle surface with the increase of temperature. Therefore, Cu was more readily oxidized at high temperatures because of insufficient Sn to form sacrificial oxides, which may protect Cu from being oxidized. After sputtering, the intensity of the peak around 570 eV decreased, making it negligible compared to the peaks around 565 eV (Figure 3.9c,f,i, after sputtering). This indicates the existence of metallic copper as the main component beneath the particle surface. 34 Figure 3.9. XPS signals collected at different temperatures. The concentrations of Cu(NO3)2, SnCl2, and HNO3 were 1 M, 0.1 M, and 0.1 M, respectively. The EG concentration was fixed at 4.8 M. The CGFR was set at 3 L/min. The temperature was set at 1000 oC (a)-(c); 750 oC (d)-(f); 500 oC (g)-(i); (a), (d), and (g) represent the data from Sn 3d electrons. (b), (e), and (h) illustrate the data from Cu 2p electrons. (c), (f), and (i) are the data from Cu LMM Auger electrons. For a mixture of H2O and EG, the evaporation of H2O is faster than EG, because the vapor pressure of H2O is higher than EG (Table 3.1). Solutes would be expected to precipitate on the droplet surface with EG remaining as the inner core. During this process, a fast formation of SnO is possible because of the evaporation of HNO3 and hydrolysis of SnCl2 as discussed above. With the further evaporation of the solvent, copper salts would reach the saturation point and precipitate. Therefore, the Cu/Sn ratio was lower on the 35 particle surface than beneath the surface, as shown in Figure 3.8. It led to the concentration difference of Cu and Sn between the particle surface and inner core, which could provide the driving force for the diffusion of Cu towards the particle surface (D -11Cu-in-Sn: 7.57×10 m2/s, 220 oC; D : 5.09×10-24 2Sn-in-Cu m /s, 220 oC) (Mei et al. 1992). Therefore, the increase of reaction temperature could promote the diffusion of Cu towards the particle surface, causing the increasing Cu/Sn ratio on the particle surface, as seen in the XPS results in Figure 3.8. Table 3.1. Vapor pressure of water and EG Temperature (oC) Pwater (mm Hg) (Speight 2004) PEG (mm Hg) (Jordan 1954) 60 149.38 1.6 70 233.7 3 80 355.1 5.3 90 525.76 9 100 760 15 110 1074.56 25 120 1489.14 40 3.4.2 Ethylene glycol volume percentage in the precursor The co-solvent, acting as the solute carrier, and reducing agent, is one of the most critical factors affecting the precursor properties. Before EG was identified as an effective co-solvent to generate oxide-free copper particles (Zhong et al. 2012), ethanol was applied in spray pyrolysis as the co-solvent (Kim et al. 2003). As reported, 10 vol % of ethanol in aqueous solution can promote the formation of phase pure copper powder from copper 36 nitrate precursor at temperatures above 600 oC (Kim et al. 2003). This was attributed to the high H2 yield of ethanol decomposition during spray pyrolysis. Experimental measurements showed that 2.2 moles of H2 could be generated from every mole of ethanol (Zhong et al. 2013b). It has been experimentally proven that every mole of EG can generate 0.81 mol of H2 during spray pyrolysis (Zhong et al. 2013b). Therefore, metal salts can be reduced to metal particles by H2, the product of ethanol and EG decomposition. Compared to ethanol as the co-solvent, the consumption of EG would be greater to reduce the same amount of metal salt. Though ethanol can create a better reducing atmosphere during spray pyrolysis, EG is still favored due to its higher flash point (Zhong et al. 2012). As reported, the decomposition of EG to H2 has several different mechanisms (Yue et al. 2012): 𝐶2𝐻6𝑂2 → 2𝐶𝑂 + 3𝐻2 (3.3) 𝑥 𝐶2𝐻6𝑂2 → 𝐶2𝐻6−𝑥𝑂2 + 𝐻2 (3.4) 2 here, Formula (3.3) is the complete decomposition of EG, and Formula (3.4) is the general dehydrogenation of EG. During the experiments, the relative EG amount needs to be controlled. High EG volume percentages in the precursor can lead to high H2 production in the reactor, which can promote the generation of oxide-free Cu-Sn binary particles. However, excessive EG will result in a precursor solution too viscous to be atomized and H2 above the lower flammability limit. The dynamic viscosities and densities of the precursors with various EG concentrations (Figure 3.10) were measured because they have been reported as two of the factors affecting the precursor droplet size (Avvaru et al. 2006; Ramisetty et al. 2013; Tsai et al. 37 2009). As shown in Figure 3.10, the viscosity of the precursor increased rapidly with EG concentration compared to density. For the conditions with EG concentration > 5.9 M (viscosity > 3 mPa∙s), it was difficult to atomize the precursor in the experiments because of the high viscosity. Similar results were observed and discussed by other researchers with similar transducers (1.7 MHz), who found the precursor cannot be atomized when the viscosity exceeded 2 mPa∙s in their experiments (Wang et al. 2008). Figure 3.10. (a) Densities and viscosities of precursors with 3.6 M, 4.8 M, and 5.9 M EG. The concentrations of Cu(NO3)2, SnCl2, and HNO3 were 1 M, 0.1 M, and 0.1 M, respectively. (b) Densities and viscosities of precursor with 1 M Cu(NO3)2, 0.1 M SnCl2, 0.1 M HNO3, and 5.9 M EG during the spray pyrolysis. To further understand the effect of EG concentration on the product crystallinity and morphology, XRD and SEM images of the powders obtained from the precursor solutions with different EG concentrations (i.e., 3.6 M, 4.8 M, and 5.9 M) at different temperatures (500 oC, 750 oC, and 1000 oC) was taken as shown in Figures 3.7,3.11. There was no significant morphological difference between the tested EG concentrations. However, the morphologies and number mean diameters of particles varied significantly between 38 different temperatures, consistent with the discussion in Section 3.4.1. Generally, the particles were within the range of 0.5 to 2.5 μm, which is consistent with previous results (Zhong et al. 2012). Figure 3.11. XRD results of powders generated from systematic conditions. The concentrations of Cu(NO3)2, SnCl2, and HNO3 in the precursor were 1 M, 0.1 M, and 0.1 M, respectively. The CGFR was fixed at 3 L/min. The powders were generated at 500 oC (black), 750 oC (red), and 1000 oC (blue), with EG concentration varied at 3.6 M, 4.8 M, and 5.9 M (from bottom to top), respectively. The major peaks are from Cu13.7Sn (ICDD with PDF No. 03-065-6821) and Cu (ICDD PDF No. 01-070-3038). XRD measurements were conducted to analyze the detailed crystal information of these samples. As shown in Figure 3.11, no significant difference can be seen between the samples generated at various EG volume percentages with the same temperature, which is consistent with the SEM results. Based on the results, it can be concluded that the EG can effectively reduce the metal salts at concentrations as low as 3.6 M in the precursor. 39 However, the EG concentration had no significant effect on the particle morphology and crystallinity in the tested EG volume percentages. One important thing needs to mention here is the composition change during the spray pyrolysis. It is difficult to understand the composition of precursor solution directly because there are various variables to be determined, for example, the concentrations of Cu(NO3)2, SnCl2, HNO3, and EG. I still attempted to understand this issue by monitoring the density and viscosity changes of the precursor solution during the spray pyrolysis. As shown in Figure 3.10b, the densities and viscosities of precursor solution which contains 1 M Cu(NO3)2, 0.1 M SnCl2, 0.1 M HNO3, and 5.9 M EG were measured every 20 min during the spray pyrolysis. Although fluctuations can be observed in the density and viscosity measurements, densities and viscosities of the precursor solution varied within narrow ranges. This indicates that the composition of the precursor can remain within a narrow range during the spray pyrolysis. 3.4.3 Residence time To explore the effect of residence time on powder formation, spray pyrolysis was conducted at 750 oC with various carrier gas flow rates (CGFRs). 750 oC was chosen because it represented the transition conditions (as discussed in Section 3.4.1), highlighting an observable difference in the morphology. The residence times were 4.5 s, 1.5 s, and 0.92 s for CGFR = 1, 3, and 5 L/min, respectively (Figure 3.1). EG concentration was 4.8 M. As shown in Figure 3.12, the morphologies of the particles varied under different residence times. The long residence time (4.5 s) promoted the generation of spherical particles with smooth surfaces (Figure 3.12a), which are at the lowest surface energy state. Few small particles were observed under this experimental condition. Short residence time (0.92 s) 40 favored the formation of irregular shapes (e.g., pores on the particle surface, fragmented spherical particles, and lamellar fragments, marked by dashed arrows in Figure 3.12c). The longer residence time may have led to the loss of small particles by thermophoretic deposition on the tube walls (Nguyen et al. 2002). Thus, small particles could not be seen on the filter (Figure 3.12a). The temperature difference between the centerline and wall position has been reported for similar gas flow rates, reactor tube dimensions, and furnace set point temperatures to range from 150 oC to 200 oC (Damour et al. 2005). Considering the mean particle sizes at 750 oC (small particles: 79 nm, big particles: 630 nm, Figure 3.6), the system is in the continuum regime and the thermophoretic velocity (ct) can be calculated as (Talbot et al. 2006): 𝑘𝑔 2𝑙𝑝 𝑑𝑇/𝑑𝑟 2𝐶𝑠𝑣( +𝐶 )𝐶𝑘 𝑡 𝑐 = 𝑝 𝑑𝑝 𝑇0 𝑡 2𝑙 𝑘 2𝑙 (3.5) 𝑝 𝑔 𝑝 (1+3𝐶𝑚 )(1+2 +2𝐶𝑑 𝑘 𝑡 ) 𝑝 𝑝 𝑑𝑝 where Cs (thermal slip coefficient), Ct (a numerical factor of order unity), and Cm (momentum exchange coefficient) are suggested to be 1.2, 2.2, and 1.1, respectively (Loyalka 1968; Talbot et al. 2006). For the radial temperature gradient, dT/dr = 16,000 K/m, when the temperature difference between the center line and tube wall was assumed to be 150 oC. Other parameters of v (kinematic viscosity), lp (mean free path), and kg/kp (thermal conductivity ratio of gas to particle) are 3.7×10-5 m2/s, 0.14 nm, and 2×10-4, respectively (Yaws 1999). Based on the mean diameter of particles and small particles at 750 oC, C (slip correction factor) are 1.3 and 5.0, respectively in each case (Friedlander 2000). By Equation (3.5), the thermophoretic velocities of large particles and small particles are 4.0×10-3 mm/s and 0.15 mm/s, respectively. Therefore, the difference between 41 the thermophoretic velocities of big and small particles could cause the results observed in Figure 3.12. For the situation where residence time was short, particles had insufficient time to form spheres. Lamellar structures (inset of Figure 3.12c) can be seen, which may have originated from metal hydroxide intermediates (Zhong et al. 2012). Pores on the surface of the particles were formed as a result of the rapid evaporation caused by the short residence time. The powders generated at medium residence time (1.5 s) exhibited an intermediate morphology between the long (4.5 s) and short (0.92 s) residence time conditions. To further understand the effect of residence time on the crystallinity of Cu-Sn binary particles, XRD measurement was performed. As shown in Figure 3.12d, there was no significant difference in crystal structure among particles fabricated at different residence times. All samples exhibited high quality crystallinity with sharp Cu13.7Sn peaks, similar to the powders generated at other conditions. The small peaks around 32.6o are still believed to be unidentified intermediate(s) as discussed previously. Figure 3.12. SEM images of powders obtained at 750 oC from precursors with 1 M Cu(NO3)2, 0.1 M SnCl2, 0.1 M HNO3, and 4.8 M EG. The residence times are: 4.5 s (a), 42 1.5 s (b), and 0.92 s (c). The insets of each figure are higher magnification images of the area marked by arrows. Arrows show irregular structures. (d) The XRD results of the powders obtained at the conditions of (a)–(c). The peaks are attributed to Cu13.7Sn (ICDD with PDF No. 03-065-6821). All scale bars in the SEM images are 1 µm. When the set point of the furnaces were 750 oC, the residence times for each condition are 4.5 s (CGFR=1 L/min), 1.5 s (CGFR=3 L/min), and 0.92 s (CGFR=5 L/min). Long residence time (4.5 s) would cause a large quantity loss of products. Particles would more readily stick to the tube walls due to the thermophoretic force. Thus, the production rate was low. Furthermore, it became difficult to collect enough powder for future characterization within 2 h. The color map of the as-sprayed filters for different residence times are shown in Figure 3.13. For the 4.5 s condition, as the production rate was low, powders could be seen only in a circular area at the filter center. The powder production rate increased significantly under the other two conditions when the residence time was decreased to 1.5 s and 0.92 s. Figure 3.13. Images of as-sprayed filters generated from precursors with Cu(NO3)2, SnCl2, and HNO3 of 1 M, 0.1 M, and 0.1 M, respectively. The EG concentration was 4.8 M. The temperature was set at 750 oC. The diameters of the filters are ~ 9.5 cm. 43 As shown in Figure 3.1, low CGFR led to a long residence time, which may increase the possibility that tin would diffuse into the center of the particles, consequently reducing the tin concentration on the particle surface. For the particles fabricated with residence time was 4.5 s, the Cu/Sn = 7.14 on the particle surfaces before sputtering. Therefore, with such a low concentration of Sn, metallic tin was not observed on the Cu-Sn binary particle surfaces (Figure 3.14g, before sputtering) for particles generated at 1 L/min. The major peak (around 932.4 eV) in the powder prior to sputtering at 500 oC (Figure 3.14h) was attributed to metallic copper (Fleisch and Mains 1982). The absence of peaks around 565 eV, and existence of a peak around 570 eV in the Cu LMM Auger diagram (Figure 3.14i) were strong evidence that demonstrated copper oxides were the major component on the particle surface (Sen et al. 2009). When the residence time was 1.5 s, the Cu/Sn ratio decreased to 2.5 on the particle surface prior to sputtering. With such a high tin concentration on the particle surface, tin oxides were observed together with metallic tin (Figure 3.14d, before sputtering). For the copper with the protection by tin, metallic copper existed on the particle surface before sputtering, which can be validated by the peaks around 932.4 eV in Figure 3.14e, and 565 eV in Figure 3.14f (Fleisch and Mains 1982). However, peaks belonging to copper oxides were still seen around 570 eV in Figure 3.14f, before sputtering (Sen et al. 2009). For particles generated with a residence time of 0.92 s, irregular shapes were seen in Figure 3.12c, with pores and holes on the particle surface which increased the surface area. Therefore, a higher possibility for the metal oxidation was expected. Before sputtering, the intensity of the metallic tin peaks were low and no metallic copper peaks can be found around 565 eV (Figure 3.14a,c, respectively) (Fleisch and Mains 1982). After sputtering, all samples generated at 4.5 s, 1.5 s, and 0.92 s displayed 44 a Cu/Sn ratio around 10, which was the input Cu/Sn ratio in the precursor solutions. In addition, metallic tin peaks exhibited higher intensities compared to the tin oxides in Figure 3.14a,d,g, after sputtering. The existence of metallic copper was validated by the peaks at 932.4 eV (Figure 3.14b,e,h, after sputtering) and peaks at 565 eV (Figure 3.14c,f,i, after sputtering) (Fleisch and Mains 1982). The amount of copper oxides was negligible compared to the metallic copper after sputtering, which can be proven by the low intensities of the peaks at 570 eV in Figure 3.14c,f,i (Fleisch and Mains 1982). Figure 3.14. XPS signals collected at different CGFRs. The concentrations of Cu(NO3)2, SnCl2, and HNO3 were 1 M, 0.1 M, and 0.1 M, respectively. The EG concentration was fixed at 4.8 M. The temperature was set at 750 oC. The CGFR was set at 5 L/min (a)-(c); 3 L/min (d)-(f); 5 L/min (g)-(i); (a), (d), and (g) are the data from Sn 3d electrons. (b), (e), and (h) are the data from Cu 2p electrons. (c), (f), and (i) are the data from Cu LMM Auger electrons. 45 3.4.4 Effects of quench As discussed previously, Cu and Sn were thought to evaporate into the carrier gas in the heated zone. Gas-to-particle conversion may lead to small particles after the carrier gas was cooled by the quench gas. To validate this hypothesis, a control experiment was designed, and conducted at the same conditions following the same procedures, except no quench gas was introduced. Thus, the temperature at the filter increased from 105 oC (with quench) to 175 oC (without quench). The cooling rate decreased from 162 oC/s (with quench) to 66 oC/s (without quench). To better investigate the effect of quench on the products, SnCl2 concentration was set to 0.05 and 0.1 M in the precursor to investigate the origin of the small particles. The concentrations of Cu(NO3)2 and EG were still kept at 1 M and 4.8 M in both cases, respectively. During spray pyrolysis, the set point of the furnaces was 750 oC. The residence time was 1.5 s. As shown in Figure 3.15a,c, powders generated with quench gas displayed a large quantity of small particles (marked by arrows). Without quench gas, the amount of small particles decreased significantly (Figure 3.15b,d). This observation supports the hypothesis that small particles originated from gas-to-particle conversion, since increase the cooling rate (by introducing quench) would enhance the homogeneous and heterogeneous nucleation (Schulz et al. 2005). In addition, sharp edges can be observed on the particle surface, as shown in Figure 3.15b,d, marked by arrows. These flat structures were attributed to the (111) planes of copper, indicating the doping of other elements (Chatain et al. 2004). From Figure 3.15a,c, the amount of small particles did not change significantly. Together with Figure 3.5d, it could be concluded that the copper species may be the main source for the small particles instead of tin species. 46 The powders generated from the four experimental conditions mentioned above were also characterized by XRD to explore the effect of quench on the final particle crystallinity. The products fabricated by precursors with 0.1 M SnCl2 exhibited Cu13.7Sn as the major peaks (Figure 3.15e). For products fabricated from precursors of 0.05 M SnCl2, the peaks were also sharp (Figure 3.15e) with positions close to those of Cu13.7Sn. However, the peaks were also close to other two standard peaks in the ICDD database. The standard peaks of the possible candidates are presented below the XRD diagram in Figure 3.15e. It seems that the different input Cu/Sn ratio in the precursor may lead to the formation of solid solutions other than Cu13.7Sn in the products. Therefore, intermediate products, i.e., (Cu32Sn)0.12 and Cu matrix were seen in the XRD results. In addition, there was no significant difference between the XRD results of the samples generated at the same SnCl2 concentration with or without quench. Figure 3.15. The effect of quench gas on the morphologies and crystallinities of the powders obtained at 1000 oC. The precursors contained 1 M Cu(NO3)2, 0.1 M HNO3, and 4.8 M EG. The SnCl2 concentration was 0.05 M (a)–(b) and 0.1 M (c)–(d). SEM images of the prepared powders are shown in (a), (c) with quench gas, and (b), (d) without quench 47 gas. (e) The XRD results of the generated powders at the four conditions. Standard peaks from Cu13.7Sn, (Cu32Sn)0.12, and Cu are displayed below the XRD results for reference. Scale bars in the SEM images are 1 µm. 3.4.5 Mechanism After understanding how temperature, residence time, quench gas, and EG concentration affect the powder morphology, crystallinity, and elemental distribution, the mechanism of Cu-Sn binary particle formation was investigated from two perspectives: chemical reaction steps and structural progression. To explore the chemical process of Cu- Sn binary particles generation from the precursor, two conditions were chosen at low temperature and low residence time to identify reaction intermediates. 300 oC was chosen because the initial decomposition temperature of EG is ~240 oC (Yue et al. 2012). As the droplets have not fully evaporated in this condition, the crystallinity of the powder was poor with an obvious amorphous background in the XRD diagram (Figure 3.16), compared to previous XRD results. The peaks in the Figure 3.16 were attributed to metal hydroxide salts (Cu2Cl(OH)3 and SnCl2(H2O)2) and metal (Sn). 48 Figure 3.16. XRD diagram of powders fabricated from precursors containing 1 M Cu(NO3)2, 0.1 M SnCl2, 0.1 M HNO3, and 5.9 M EG in a total volume of 150 ml aqueous solution. Spray pyrolysis was conducted at 300 oC with CGFR = 5 L/min and a quench gas velocity of ~ 5.2 L/min. Possible phases are Cu2Cl(OH)3 (ICDD, PDF No. 01-070-0821), SnCl2(H2O)2 (ICDD, PDF No. 01-075-2033), and Sn (ICDD, PDF No. 00-005-0390). To further explore the particle formation process, the temperature was increased to 500 oC and the CGFR was increased to 10 L/min to keep the residence time less than 0.74 s. As shown in Figure 3.17, the samples generated from 500 oC consisted of metal hydroxide salts (Cu2Cl(OH)3 and Cu2(NO3)(OH)3), metal oxides (Cu2O and CuO), and solid solutions (Cu13.7Sn). Based on the discussion above, it is proposed that the spray pyrolysis process of converting precursor to Cu-Sn binary particles is composed of the following chemical reaction steps. 49 Figure 3.17. XRD diagram of powders fabricated from precursors containing 1 M Cu(NO3)2, 0.1 M SnCl2, 0.1 M HNO3, and 5.9 M EG. Spray pyrolysis was conducted at 500 oC with CGFR = 10 L/min and a quench gas velocity of ~ 10.8 L/min. Possible phases are: Cu2Cl(OH)3 (ICDD, PDF No. 01-070-0821), Cu2(NO3)(OH)3 (ICDD, PDF No. 00- 045-0594), Cu2O (ICDD, 01-080-7711), CuO (ICDD, PDF No. 01-080-0076), and Cu13.7Sn (ICDD, 03-065-6821). i) Hydrolysis of the salts With the increase of temperature within the droplets (Figure 3.1d,e) and the relative solute concentration (caused by the solvent evaporation), metal salts gradually hydrolyze 50 to hydroxides (Cohen and West 1972; Munnik et al. 2011). Thus, the possible reactions are: 2𝐶𝑢(𝑁𝑂3)2 + 3𝐻2𝑂 → 𝐶𝑢2(𝑁𝑂3)(𝑂𝐻)3 + 3𝐻𝑁𝑂3 (3.6) 2𝑆𝑛𝐶𝑙2 + 3𝐻2𝑂 → 𝑆𝑛2𝐶𝑙(𝑂𝐻)3 + 3𝐻𝐶𝑙 (3.7) These products were observed in Figures 3.16,3.17. CuCl2 may be formed from ion exchange, which is possible for the lower solubility of CuCl2 (43.8 g in 100 g H2O) compared to Cu(NO3)2 (60.1 g in 100 g H2O) (Perry 2011). CuCl2 may also hydrolyze in a similar way. 2𝐶𝑢𝐶𝑙2 + 3𝐻2𝑂 → 𝐶𝑢2𝐶𝑙(𝑂𝐻)3 + 3𝐻𝐶𝑙 (3.8) This product phase was observed in Figure 3.16. ii) Formation of metal oxides With the further increase of temperature, the hydroxides begin to decompose to oxides: 𝐶𝑢2(𝑁𝑂3)(𝑂𝐻)3 → 2𝐶𝑢𝑂 + 𝐻𝑁𝑂3 + 𝐻2𝑂 (3.9) 𝐶𝑢2𝐶𝑙(𝑂𝐻)3 → 2𝐶𝑢𝑂 + 𝐻𝐶𝑙 + 𝐻2𝑂 (3.10) 𝑆𝑛2𝐶𝑙(𝑂𝐻)3 → 2𝑆𝑛𝑂 + 𝐻𝐶𝑙 + 𝐻2𝑂 (3.11) Similar results have also been reported (Cohen and West 1972; Munnik et al. 2011). CuO was detected in XRD measurements (Figure 3.17). iii) Reduction of the metal oxides to metals 51 As EG can create a reducing environment (Yue et al. 2012), metal oxides will finally be reduced to Cu and Sn. The evidence for the existence of copper and tin can be seen in the Figures 3.2,3.16, respectively. Cu2O was detected in the XRD (Figure 3.17). iv) Diffusion of Sn to form a solid solution Finally, the Sn will dissolve into the Cu crystal matrix to form the Cu-Sn solid solution. However, the solid solution formation would not be completed at low temperature such as 500 oC. Therefore, both Cu and Cu13.7Sn phases were observed in the XRD diagram (Figure 3.2, black peaks). Based on Cu-Sn phase diagram, the phases in the Cu-Sn binary system with same Cu/Sn ratio at 500 oC should be Cu and Cu3.73Sn (Miettinen 2008). The formation of Cu13.7Sn could be attributed to three causes: i) The cooling effect caused by introducing quench gas could cause the product to be trapped in a nonequilibrium state; ii) Cu and Cu13.7Sn share the same Fm-3m point group symmetry, with close lattice spacing (Cu: 3.625 Å; Cu13.7Sn: 3.688 Å) (Andresen 1958; Srinivasan and Anantharaman 1963); iii) The accumulation of Sn on the particle surface may increase the Cu/Sn ratio in the bulk compared to the Cu/Sn ratio in the precursor (10). Cu13.7Sn is a stable phase with Cu/Sn ratio higher and close to 10 (Chatterjee and Gupta 1975). When the set point of the furnaces was 750 oC, the increased temperature would promote the formation of Cu-Sn solid solution. Thus, only Cu13.7Sn was observed in Figure 3.2 (red peaks). When the temperature was further increased to 1000 oC, the high temperature would lead to the system in the liquid state (Booth et al. 1977), implying facile diffusion of tin into copper matrix to form the Cu13.7Sn phases. Therefore, Cu13.7Sn was found in Figure 3.2 (blue peaks). Peaks in Figure 3.12d were also belonged to Cu13.7Sn for the same reasons. 52 A general mechanism has been proposed to explain the particle formation in spray pyrolysis using multi-source precursors (Gurav et al. 1993). Applying this mechanism to our specific Cu-Sn system, hydrolysis of SnCl2 may lead to the precipitation of SnO in the droplets before precipitation of Cu(NO3)2. The formation of Cu-Sn particles involves a structural evolution from granular precipitates to a dense solid solution of Cu-Sn. The detailed process can be summarized as follows, together with the schematic illustration presented in Figure 3.18: Figure 3.18. Schematic illustration of the proposed Cu-Sn binary particle formation process. The circle in Step (1) represents the precursor droplet. The magnified image of the 53 Step (1) product illustrate the Cu-Sn solid solution and the metallic copper, respectively. The small circles are the metal particles formed by gas-to-particle conversion. a) Precipitation on the droplet surfaces After the precursor droplets were carried into the furnaces, rapid evaporation of HNO3 and solvents, particularly H2O (due to a higher vapor pressure than EG) (Speight 2004; Yue et al. 2012), would occur. It could assist the hydrolysis of tin salts, forming the precipitates (Perry 2011), as well as generate a concentration gradient of the dissolved solutes within the droplets. Precipitation will readily occur when the solute concentration reaches the saturation point. Thus, droplets with primary particle shells and liquid EG cores would take shape. The reduction of precipitates to metals occurs by reaction with H2 which is produced from EG decomposition. In the magnified image of the Step 1 product (Figure 3.18), the brown part represents the Cu-Sn solid solution, while the red part illustrates the metallic copper. These substances were both found in XRD results at 500 oC (Figure 3.2). Once the solvents evaporate, hollow structures would be formed with primary particles comprising the shell (Step 2, Figure 3.18). If the process is ceased here, porous hollow particles would be generated (Step 3’, Figure 3.18). The hollow porous structure has been observed in Figure 3.3a and 3.5a. In some cases, bursting of the droplets may occur due to the buildup of pressure caused by solvent evaporation (Gurav et al. 1993). Then fragmented, hollow, and porous particles can be found, as shown in the insets of Figure 3.3a. b) Melting and shrinking of the hollow structure 54 After the hollow porous structure is generated, the primary particles will begin to sinter at the high temperatures, possibly caused by melting. In the meantime, Sn may diffuse into the Cu matrix to form a solid solution (Step 3, Figure 3.18). Hollow structures with rough crusted shells would be fabricated if the process stops here, as shown in Figure 3.5c. Other structures e.g., fragmented spherical particles (Figure 3.3b, marked by red arrow), could also be regarded as products of the interconnection of primary particles seen in the powders at 500 oC (inset of Figure 3.3a). In addition, the evaporation of solvent may cause pore formation on the particle surfaces (Figure 3.7d). Therefore, morphologies as shown in Figures 3.3b,3.7d-f can be seen once the process is stopped after the gas-to-particle conversion (step 4’, Figure 3.18). c) Solidification In the final step, solidification would be achieved once the residence time is long enough and temperature is high enough (step 4, Figure 3.18). After being cooled by the quench, small particles could be generated by gas-to-particle conversion (step 5, Figure 3.18). The detailed gas-to-particle conversion is presented in Section 3.4.1. On the filter, the small particles tend to attach on the particles instead of coating on the surface of big particles. Thus, morphologies as shown in Figures 3.3c,3.5e could be observed. Micron-sized solid particles are favored for applications including conductive thick film pastes used for printing conductive lines and interference packaging (Huang and Sheen 1997; Wu et al. 2009). Based on our experimental results, high temperature (1000 oC) and long residence time (4.5 s) promotes the formation of micron-sized solid Cu-Sn particles, a potential condition for the further real application. The accumulation layer of Sn on the particle surface may protect the particle from being oxidized. The exploration of 55 Cu/Sn oxidation resistance and conductivity as a function of composition and processing environment is the subject of following chapters. 3.5 Summary of this chapter Cu-Sn binary particles were fabricated by spray pyrolysis with EG as the co-solvent and reducing agent. By tuning the temperature, residence time, and quench gas flow rate, particles could be generated with controllable morphologies, crystallinities, and elemental distributions, which may broaden the further application of Cu-Sn bimetallic particles. Hollow porous particles were generated at 500 oC, with primary particles comprising the surface. The existence of Cu and Cu13.7Sn indicated that the generation of Cu-Sn binary particles may be incomplete at 500 oC. Cu13.7Sn phase was formed because of the cooling by introducing quench, the increased Cu/Sn ratio in the bulk caused by Sn enrichment on the particle surfaces, and the thermodynamic stability of Cu13.7Sn phase. With the increase of temperature, a transition occurred, causing the particle morphology to shift from hollow to solid, which has been validated by samples generated at 750 oC. At 1000 oC, solid particles could be generated with high quality crystallinity, and only Cu13.7Sn peaks were present in the XRD diagram. The Cu/Sn ratio was further increased on the particle surface at 1000 oC. The small particles were likely formed likely via gas-to-particle conversion. The residence time and quench gas flow rate mainly contributed to the morphological differences. At the condition where the residence time was 4.5 s, spherical particles with smooth surfaces were obtained. The increase of the CGFR decreased the residence time, leading to insufficient time for the thermophoretic deposition of small particles on the tube walls, which could significantly reduce the loss. Thus, small particles became observable 56 when the residence time was 1.5 s. Particles with irregular morphologies were synthesized at the condition where residence time was 0.92 s. However, high quality crystallinity of Cu13.7Sn could be seen in all samples at 750 oC regardless of the CGFR changes. From the experimental results, the mechanism of Cu-Sn binary particle generation can be explained through two perspectives: chemical reaction steps and structural progression. Substances in the precursor were first hydrolyzed into hydroxides. Then the hydroxides decomposed into metal oxides. Subsequently, the metal oxides were reduced to metals, and a solid solution formed. For the structural evolution, precipitation of the solutes occurred during solvent evaporation of the precursor droplets. Thus, hollow structures could be generated. Finally, the shrinking of the hollow particles continued until solid particles were formed. The accumulation of Sn on the particle surfaces could protect Cu from being oxidized. In addition, the method of generating Cu-Sn binary particles combined with the investigation of the underlying mechanism provides a foundation for fabricating other bimetallic particles. 57 Chapter 4: Oxidation behavior of Cu-Sn binary particle 4.1 Abstract Micron-sized solid Cu-Sn particles have been considered as a replacement for more expensive materials (e.g., Ag, Pd, and Au) in conductive pastes used in printed electronics, solar cell metallization, and interference packaging. With formation of a tin oxide layer, Cu-Sn particles could combine relatively low electrical resistivity with high oxidation resistance. However the oxidation behavior of this system is not well understood. In this chapter, the oxidation of CuSny solid particles, fabricated by spray pyrolysis without direct addition of H2, was investigated. Our experimental results and theoretical analysis suggest that at a low oxidation temperature (300 oC), the migration of O2- through the oxide layer controls the oxidation. At high temperature (500 oC), the grain growth of the oxide layer is believed to be the rate-limiting step. Among the CuSny particles tested, CuSn0.1 powders exhibited the best particle structure (solid and spherical) and highest oxidation resistance. 4.2 Introduction Metal particles have been widely applied in printed electronics (Comiskey et al. 1998; Li et al. 2005), interference packaging (Faddoul et al. 2012; Wang et al. 2007a), and solar cell metallization (Mokkapati et al. 2009; Temple and Bagnall 2011). Gold and silver are by far the most commonly used materials in these fields due to their high electrical conductivity and oxidation resistance (Comiskey et al. 1998; Faddoul et al. 2012; Li et al. 2005; Mokkapati et al. 2009; R et al. 2015; Temple and Bagnall 2011; Wang et al. 2007a). However, the high cost of these noble metals has increased interest in developing alternative cost-efficient materials such as copper (Cu) (Abhinav K et al. 2015; Yonezawa 58 et al. 2015a; Yonezawa et al. 2015b) for these applications. Several challenges, such as the ease of oxidation of copper materials and the high mobility of copper in silicon (a common substrate in the electronics industry), have to be addressed before widespread application of copper-based powders (Istratov et al. 2000; Kim et al. 2004). The oxidation behavior of Cu is known to be complex since many processes, including electron migration across the metal/oxide interface, recrystallization of the pseudomorphic oxide layer, external/internal diffusion of O2 gas, and chemical reactions, can happen simultaneously (Cabrera and Mott 1949; Feng et al. 2003; Fujita et al. 2013; Levenspiel 1999). Particle engineering with an oxidation-resistant layer on a Cu core has been proposed as an effective configuration. The oxidation-passivation layer can be metals (e.g., Ag, Zn, Ni, and Au) or organic compounds (Jianfeng et al. 2011; Jung et al. 2011; Kim et al. 2013; Li et al. 2016a; Sharma et al. 2014; Xu et al. 2012; Yonezawa et al. 2015a; Yuan et al. 2013). Nevertheless, there is a need to explore other strategies from a comprehensive view of cost and electrical conductivity. Tin (Sn) has been reported as an effective protector against copper oxidation, since tin is more reactive towards oxygen than copper and tin oxides maintain relatively high conductivity (Liu et al. 2005; Liu et al. 2007a). In addition, tin is a diffusion barrier between copper and silicon, which provides extra benefit for practical applications, in particular, in solar cell metallization (Liu et al. 2005; Liu et al. 2007a). Recently, the electrical resistivity of CucoreSnshell nanowires fabricated by electroplating method was found to remain at the same order of magnitude after being oxidized for 25 h, while the resistivity of Cu nanowires increased by 107 in 15 min under the same oxidation conditions (Chen et al. 2014). However, rather than nanowires, micron-sized solid particles are favored by industry for applications such as conductive pastes and interference packaging (Deshpande et al. 2005; 59 Wu et al. 2009). To address this issue, Cu-Sn binary powders with Sn-enriched surface layer have been produced by a scalable process, spray pyrolysis, without the direct addition of H2 (Chapter 3). As a next step, here, we present a detailed investigation of the oxidation behavior of CuSny (y is the atomic ratio of Sn/Cu in the product) solid particles fabricated by spray pyrolysis. A quantitative understanding of the phase evolution of CuSny particles during oxidation was obtained through Rietveld refinement analysis of the X-ray diffraction (XRD) patterns. To better understand the oxidation behavior of CuSny particles under high temperature, in-situ XRD measurements were conducted in ambient air. Kinetics experiments were performed to determine the oxidation kinetics of CuSny particles and the oxidation mechanisms at different temperatures. Together, these results were used to identify the CuSny particle composition that has optimal oxidation resistance and spherical morphology. 4.3 Experimental 4.3.1 CuSny particle preparation We synthesized the CuSny particles following procedures discussed in Chapter 3. In brief, a 150 mL aqueous precursor solution that contained 1 M Cu(NO3)2 (99.5%, STREM CHEMICALS), 0.1 M HNO3 (70%, SIGMA-ALDRICH), and 4.8 M ethylene glycol (EG, 98% SIGMA-ALDRICH) was firstly prepared. SnCl2 (≥98% FISHER SCIENTIFIC) was used as the tin source and then added to the aqueous precursor solution with a concentration based on the targeted Sn/Cu atomic ratio (from 0 to 1) in the product particles. The precursor solution was atomized by a 1.7 MHz ultrasonic generator into droplets with 60 volume mean dimeter of 5 μm, measured experimentally under similar experimental conditions (Zhong et al. 2012). The droplets were carried by N2 (99.5%, AIRGAS) into a quartz tube reactor heated by two furnaces in series both set at 1000 oC. A polytetrafluoroethylene (PTFE) filter was used to collect the products at the end of the quartz tube after the powders were cooled down by N2 gas. The residence time of the process was estimated to be 1.2 s, measured and calculated by the same procedures reported in Chapter 3. 4.3.2 CuSny powder characterization Scanning electron microscopy (SEM, HITACHI SU-70) and transmission electron microscopy (TEM, JEOL 2100 FEG) with energy dispersive X-ray spectroscopy (EDS) were utilized to understand the particle morphology, structure, and elemental distributions. CuSny powders were suspended and sonicated in ethanol for 1 min before being added dropwise onto silicon wafers for SEM and TEM grids for TEM. X-ray photoelectron spectroscopy (XPS, Kratos Axis 165) was performed to understand the element distributions on the powder surface and sub-surface (sputtered in vacuum by argon ions for 20 min with the sputtering gun set at 4 kV and 5 mA). X-ray diffractometry (Bruker D8 advance) was used to obtain crystallographic information of the CuSny powders. Rietveld refinement was conducted by Topas-5 software from Bruker to determine the phase composition of the powders. In-situ XRD (Bruker C2 Discover) was performed to continuously monitor the phase evolutions of the CuSny powders during oxidation. Powders were placed on a Pt plate and heated by a programmed hot stage in ambient air. When the powders reached the targeted temperature, the hot stage was kept at that temperature for 15 min for the XRD measurements. After the 61 XRD measurements were completed, the temperature of the Pt plate was raised to the next target. Thermogravimetric (TG) and differential scanning calorimetry (DSC) results were measured by a SDT Q600 (TA Instruments) to understand the weight gain and heat flow during the oxidation. Alumina pans containing ~2 mg CuSny powders were heated from room temperature to 600 °C at a rate of 5 oC/min under 100 mL/min of compressed O2 flow. 4.3.3 Oxidation kinetics investigation The oxidation kinetics of CuSny particles was studied by performing the reaction in a U-shape tubular quartz reactor with a 10 mm inner diameter. Typically, 0.5 g of the particle powder was loaded into the quartz reactor, and the reactor was placed inside a furnace (National Electric Furnace FA120 type) controlled by a Watlow Controller (96 series). A K-type thermocouple was placed inside the furnace to monitor the particle temperature by attaching the tip of the thermocouple to the outer wall of the powder bed. The particle powder was heated to the desired temperature in He atmosphere (30 mL/min) at a ramp rate of 5 oC/min. Afterwards, O2 (1 mL/min, 99.9993% purity, Airgas) diluted in N2 (as internal standard) (15 mL/min, 99.95% purity, AIRGAS) and He (as balance gas) (35 mL/min, 99.9993% purity, AIRGAS) was introduced to the reactor through heated transfer lines held at 70 oC. The O2 concertation in the reaction system was quantified using a gas chromatograph (Agilent Technologies, 6890N) equipped with a ShinCarbon ST packed column connected to a thermal conductivity detector (TCD). The volume percent of O2 in the feed stream was set to be 1.4 % and 10 %, respectively, by changing the flow rate of O2 and balancing the total flow using He. 62 4.4 Results and discussion 4.4.1 CuSny particle preparation by spray pyrolysis The CuSny particles were obtained by reducing the corresponding metal salts during spray pyrolysis. After the precursor was atomized into droplets, the droplets were transported by carrier gas into the quartz tube reactor. In the reactor, Cu(NO3)2 and SnCl2 were hydrolyzed into hydroxides. Subsequently, the hydroxides were decomposed into metal oxides, which were then reduced to metals by the reducing gases (H2 and CO) generated through the decomposition of EG (Yue et al. 2012; Zhong et al. 2012). At the same time, a series of processes occurred, including solvent evaporation, solute precipitation, chemical reaction, nucleation, and crystal growth, leading to the formation of particles. Based on our previous findings in Chapter 3, high temperatures and long residence times promote the formation of solid particles. Herein, the CuSny particles were produced at 1000 oC with a residence time of 1.2 s, based on the experimental results presented in Chapter 3. Figure 4.1. SEM images of Cu (a), CuSn0.01 (b), CuSn0.05 (c), and CuSn0.1 (d) particles. The insets in each panel are the high magnification images of the selected regions marked by dashed squares. The scale bars are 1 μm for all the images. 63 Cu particles display solid and spherical structures with smooth surfaces, as demonstrated by SEM (Figure 4.1a) and TEM (Figure 4.2a) images. The lattice fringes in Figure 4.2b represent the (111) plane of Cu (PDF No. 01-070-3038) with a d-spacing of 2.1 Å. The corresponding diffraction pattern attributed to the Cu (111) plane (inset of Figure 4.2b) is obtained by performing fast Fourier transform (FFT), suggesting a high degree of crystallinity. This hypothesis is also validated by the pronounced Cu peaks in XRD spectrum (Figure 4.3). Figure 4.2. TEM images of Cu particle (a) with corresponding HRTEM image (b) of selected area in (a). The inset of (b) is the fast Fourier transform (FFT) of the HRTEM. TEM image of CuSn0.01 particles (c) with HAADF image and line scan results of Cu and Sn (d), elemental mapping of Cu (e) and Sn (f), and overlap mapping (g). With increasing Sn concentration, the CuSny particles retain spherical and solid structure when y≤0.1, as shown in SEM (Figure 4.1b-d) and TEM (Figure 4.2c,4.4a-b) images. The number mean diameter of the CuSny particles is 710 nm, same as the results 64 in Chapter 3. Small particles are observable on the surface of CuSny (y≤0.1) sample. These small particles are mainly composed of Sn, validated by EDS mapping of Cu and Sn elements (Figure 4.2e-g). We believe these small Sn particles originate from evaporation and gas-to-particle conversion. The vapor pressures of SnO and Sn at 1000 oC are 370 Pa and 1.5×10-2 Pa, respectively (Gale and Totemeier 2003; Platteeuw and Meyer 1956). The evaporated SnO may be reduced to Sn by the reducing gases from EG decomposition and nucleate into small particles. With the increase of the relative concentration of Sn in the powder, Sn began to dissolve into Cu matrix to form a Cu-Sn solid solution, resulting in a blue shift of the Cu peaks to Cu13.7Sn (PDF No. 03-065-6821) peaks, as shown in Figure 4.3. For the CuSn0.1 powders, the existence of Cu13.7Sn phase is attributed to i) the similar crystal structure and same Fm-3m point group symmetry of Cu and Cu13.7Sn (Saunders and Miodownik 1990); ii) the lower Sn/Cu ratio (less than that of precursor=0.1) in the particle center due to the Sn enrichment on the particle surface; iii) stable phase of Cu13.7Sn when Sn/Cu atomic ratio is close to 0.1 (Chatterjee and Gupta 1975). 65 Figure 4.3. XRD diagrams of CuSny solid particles at room temperature (bottom) and after being heated at 150 oC for 10 min in ambient air (top). The positions marked by circles and squares are standard peaks of Cu13.7Sn (circles, PDF No. 03-065-6821) and Cu (squares, PDF No. 01-070-3038), respectively. 66 Figure 4.4. TEM images of CuSn0.05 (a), CuSn0.1 (b), CuSn0.2 (c), CuSn0.4 (d), CuSn0.8 (e), and CuSn (f) products. For CuSny (y≥0.1) particles, the small Sn particles on the particle surface tend to coagulate into bigger structures (Figures 4.4c-f,4.5). Since the set points of the furnaces (1000 oC) are higher than the melting point of bulk Sn (230 oC) (Ju et al. 2015), melted Sn may have formed the interconnection between adjacent particles when y = 0.2 and 0.4 (highlighted by dashed circles in Figure 4.5a-b) and even led to the coalescence of CuSn0.8 and CuSn particles at the surface into islands (Figures 4.4e-f,4.5c-d). The melting points of Cu (1085 oC) and Cu-Sn solid solutions (from 1085 oC (Sn at % = 0) to 945 oC (Sn at % = 10%)) are much higher than that of Sn (Saunders and Miodownik 1990). Thus, the coalescence of the CuSny also indicates that a Sn-enriched layer is on the particle surface. 67 Figure 4.5. SEM images with the corresponding high magnification images of the areas marked by dashed boxes of CuSn0.2 (a), CuSn0.4 (b), CuSn0.8 (c), and CuSn (d). The scale bars in all images are 1 μm. To test this hypothesis, XPS measurements and EDS line scan characterizations were performed on the particle samples. Based on the XPS results shown in Figure 4.6, the Sn/Cu atomic ratio is significantly higher on the particle surface (before sputtering) than beneath the surface (after sputtering) or in the precursor. In Figure 4.7, for Cu 2p electrons, metallic copper has a major peak at 932.4 eV (Fleisch and Mains 1982). For Cu LMM Auger electrons, shoulder peaks around 565 eV are attributed to metallic copper, while peaks around 570 eV are evidence for the existence of copper oxides (Fleisch and Mains 1982; Sen et al. 2009). The EDS line scan also validates that hypothesis. The relative Sn/Cu 68 atomic ratio is higher on the particle surface than in the core (Figures 4.2d,4.8). This difference of Sn and Cu distribution along particle radius may stem from the different precipitation rates between the salts and subsequent mass diffusion (Gurav et al. 1993). The reported diffusivities are DCu-in-Sn = 7.57×10 -11 m2/s, DSn-in-Cn = 5.09×10 -24 m2/s, and DCu-in-Cu = 6.22×10 -26 m2/s at 220 oC (Mei et al. 1992). In spray pyrolysis, the components with lower solubility precipitate earlier than those with higher solubility during the evaporation of droplet solvent (Gurav et al. 1993). With the evaporation of HNO3 during the solvent evaporation, the hydrolysis of SnCl2 may accelerate the precipitation of SnO, while Cu(NO3)2 still remains dissolved in the droplet solvent (Perry 2011). Together with the aforementioned small Sn particles depositing on the surface of CuSny particles, attributed to gas-to-particle conversion, a Sn-enriched layer is formed on the particle surface. The fluctuations of Sn/Cu atomic ratio (before sputtering, Figure 4.6) are possibly caused by the irregular structures as shown in Figures 4.4,4.5,4.8. Based on the mass balance calculations between the precursors and the product particles (mean diameter: 710 nm), the CuSn0.1 particle is composed of a Cu13.7Sn core and a 3 nm-Sn layer as the shell (discussed in Chapter 3). For CuSn0.05 particle, a 2.1 nm Sn-enriched layer is estimated on a Cu32Sn core. For CuSn0.01 particles, the Cu core is encapsulated by a 1.2 nm Sn-enriched layer. Although only for a single particle, the EDS mapping results are consistent with this calculated Sn shell thickness. A summary of the experimental conditions to obtain the spherical CuSny (y≤0.1) particles and their particle configuration is shown in Table 4.1. The previously reported particle engineering methods mainly focus on the core-shell configuration to improve oxidation resistance (Grouchko et al. 2009; Jianfeng et al. 2011; Kim et al. 2013; Kobayashi and Sakuraba 2008; Yamauchi et al. 2010). However, core- 69 shell particle production may incorporate sophisticated operating procedures, restricting the future scale-up production. Our CuSny particles with Sn enrichment on surface, produced from one-step process, can protect Cu by forming a sacrificial oxide layer. As solid spherical particles are favored in the fields of solar cell metallization and interference packaging, the oxidation behavior of CuSny (y≤0.1) powders was investigated next. Figure 4.6. The Sn/Cu atomic ratios summarized from XPS results (Figure 4.7). The inset represents the regime where Sn/Cu ratio is below 0.1. Sputtering was conducted in vacuum by argon ions for 20 min. Table 4.1. Summary of the configurations of CuSny particles (y≤0.1) based on mass balance calculation and experimental conditions to obtain these particles Cu CuSn0.01 CuSn0.05 CuSn0.1 Core Cu Cu Cu32Sn Cu13.7Sn Particle configuration Shell 1.2 nm Sn 2.1 nm Sn 3 nm Sn [SnCl2] in precursor solution (mol/L) 0 0.01 0.05 0.1 [Cu(NO3)2] in precursor solution 1 (mol/L) Reaction temperature (oC) 1000 Residence time (s) 1.2 70 Figure 4.7. XPS signals of Cu (a-c), CuSn0.01 (d-f), CuSn0.05 (g-i), CuSn0.1 (j-l), CuSn0.2 (m-o), CuSn0.4 (p-r), CuSn0.8 (s-u), and CuSn (v-x) particles before and after sputtering. Images in the left column represent the spectra from Sn 3d electrons. Images in the middle column illustrate the spectra from Cu 2p electrons. Images in the right column are the spectra from Cu LMM Auger electrons. 71 Figure 4.8. Line scans of Cu and Sn elemental profiles of CuSn0.05 (a), CuSn0.1 (b), CuSn0.2 (c), CuSn0.4 (d), and CuSn0.8 (e). 4.4.2 Oxidation behavior of CuSny particles After the as-prepared CuSny (y≤0.1) powders were oxidized at 150 oC for 10 min in ambient air, no detectable oxide peaks can be seen in the XRD patterns (Figure 4.3). Since the low temperature oxidation normally results in the formation of amorphous oxides or oxide crystals with low crystallinity (Kim et al. 2004), peaks from the oxides are negligible, compared to those from metals. We also observed the oxide layers at the edges of both Cu and CuSn0.1 particles (Figure 4.9) after powder was oxidized at 100 oC for 60 min. However, the low degree of crystallinity of the oxide layer resulted in negligible peaks in the XRD spectrum. Therefore, the oxidation of CuSny particles at this conditions is insignificant. 72 Figure 4.9. TEM images of the Cu (a-b) and CuSn0.1 (c-d) particles after being oxidized in ambient for 60 min at 100 oC. Oxide layers were formed on the particle surfaces as shown by the lattice fringe in (d), which is attributed to the (110) plane of Cu2O (PDF No. 01- 071-3645). More pronounced peaks attributed to oxides began to appear in Cu particles, after the as-prepared powders were heated at 300 oC for 10 min in ambient air, while observable peaks belonging to Cu2O were exhibited in CuSn0.01, CuSn0.05, and CuSn0.1 particles (Figure 4.10a-d). Therefore, Cu, CuSn0.01, CuSn0.05, and CuSn0.1 powders are oxidized more significantly at 300 oC than at 150 oC. To understand the phase composition in these powders, Rietveld refinement was applied to analyze the XRD patterns. The Cu particles exhibit the lowest oxidation resistance at 300 oC with only 18 wt % of post-oxidation 73 powders remaining as Cu (Figure 4.10e). CuSn0.01 powders display a sharp increase in the relative metallic amount (including Cu and Cu-Sn solid solution) in the post-oxidation products, from 18 wt % to 72 wt %, compared to that of Cu particles. However, CuO is still observable in the post-oxidation product of both Cu and CuSn0.01 powders. Further increase of the Sn loading in the particles led to a higher relative amount of metallic products after oxidation, namely 92 wt % in CuSn0.05 and 93 wt % in CuSn0.1 powders. The oxidation resistance of CuSn0.01 particles is close to that of Cu powders and significantly lower than CuSn0.05 and CuSn0.1 powders. This is possibly caused by the CuSn0.01 particles not having enough Sn compared to the CuSn0.05 and CuSn0.1 powders. As shown in Figure 4.10f, the lattice parameter of Cu-Sn solid solution phases in the post-oxidation products of CuSn0.01, CuSn0.05, and CuSn0.1 particles increases with the relative amount of Sn in the as-prepared particles, and larger than the lattice parameter of Cu. This indicates the increasing Sn relative amount in the post-oxidation products, since dissolving Sn into Cu matrix can increase the lattice parameter of Cu-Sn solid solution (Saunders and Miodownik 1990), which is consistent with the results in Figure 4.3. The lattice parameter of Cu-Sn solution in the post-oxidation product of CuSn0.01 powders is close to that of Cu phase in the post-oxidation product of Cu particles. This further verifies that the insufficient Sn loading in the as-prepared CuSn0.01 powders results in the similar oxidation resistance as that of Cu particles. Thus, the oxidation behavior of CuSn0.05 and CuSn0.1 particles was further investigated to find the optimal oxidation resistant composition, while Cu particles were utilized as the reference. 74 Figure 4.10. XRD patterns (Yobs, black dots) of Cu (a), CuSn0.01 (b), CuSn0.05 (c), and CuSn0.1 (d) powders, after they were oxidized at 300 oC for 10 min in ambient air analyzed by Rietveld refinement. XRD patterns were analyzed by Rietveld refinement (Ycalc, red curves) with the difference between the observed and calculated values (Ycalc-obs, blue curves). Tick marks represent the phases utilized in refinements, including Cu (black, PDF No. 00-004-0836), Cu2O (red, PDF No. 01-071-3645), CuO (blue, PDF No. 01-073-6023), and Cu-Sn solid solutions. Cu-Sn solid solutions are Cu (black) in (b), (Cu32Sn)0.12 (green, PDF No. 01-077-7742) in (c), and (Cu13.7Sn)0.068 (yellow, PDF No. 01-071-7874) in (d). (e) Detailed results of the weight percent of crystal phases in post-oxidation products. (f) The evolution of lattice parameter from Cu in (a) to Cu-Sn solid solutions in (b-d). The Cu- Sn solid solution phases and Cu phase share the same space group of Fm-3m. 75 To further explore the phase evolution of the CuSny powders, in-situ XRD measurements during oxidation were conducted with the temperature profile illustrated in Figure 4.11b. When temperature is below 200 oC and the oxidation duration is 62 min, no significant oxide peaks can be seen in all XRD patterns of the CuSny (y = 0, 0.05, and 0.1) powders (Figure 4.11a). Peaks attributed to Cu2O appear in all samples after being oxidized at 300 oC and the oxidation duration is 87 min. CuO phase is observed in CuSny powders when the temperature is increased to 400 oC and the oxidation duration is up to 112 min. Therefore, the onset temperature of significant oxidation of CuSny (y = 0, 0.05, and 0.1) powders is between 200 oC and 300 oC. To test this hypothesis, an oxidation experiment of long-time duration (up to 242 min) was conducted, keeping the temperature of powders at 200 oC during the in-situ XRD measurements, as shown in Figure 4.12. The CuO phase does not appear even after the CuSny powders are heated for 227 min at 200 oC. Moreover, peaks of Cu2O are observed after the powders are heated for 17 min, though the intensities of its peaks were low. The continuous oxidation of Cu, CuSn0.05, and CuSn0.1 powders leads to the increasing intensities of Cu2O peaks. However, their intensities are still not comparable with those of metal peaks, even after being oxidized for 227 min, especially for the oxidation of CuSn0.05 and CuSn0.1. These results show that the CuSn0.05 and CuSn0.1 powders are more oxidation resistant than Cu particles. 76 Figure 4.11. In-situ XRD measurements of pure Cu, CuSn0.05, and CuSn0.1 powders in ambient air at increasing temperature from 30 oC to 400 oC (a) with corresponding temperature profile during the oxidation (b). The red dots correspond to initiation of XRD scans. Every XRD measurement lasts 15 min. 77 Figure 4.12. In-situ XRD measurements of pure Cu (a), CuSn0.05 (b), and CuSn0.1 (c) when the sample temperature was held at 200 oC in ambient air subject to heating. (d) The temperature profile of the sample during the oxidation. In (d), the red dots correspond to initiation of XRD scans. 4.4.3 Oxidation kinetics of CuSny powders To quantitatively understand the oxidation kinetics of CuSny particles, the oxidation experiments were conducted in a packed-bed reactor by flowing a diluted O2 gas through the CuSny particles at different reaction temperatures. The difference of O2 flow rates 78 between the feed and product streams represents the O2 gas consumption due to the oxidation of CuSny particles in the packed-bed reactor. The CuSny powders (y = 0, 0.05, and 0.1) were oxidized at 300 oC and 500 oC, respectively. Both oxidation temperatures were chosen based on the melting point of Sn (230 oC) (Ju et al. 2015), onset temperature of significant oxidation of CuSny powders (200 – 300 oC, discussed in above), and the oxidation rate. The O2 partial pressure was measured in the product stream (Figure 4.13) and then converted into O2 gas flow rate (Figure 4.14a). To better understand the oxidation extent of the powders with time, we assume all the consumption of O2 contributes to the weight gain of the powders. Therefore, the area enclosed by the line that represents the O2 flow rate in the feed stream at the top and the line which represents the O2 flow rate in the product stream at the bottom (Figure 4.14a) is regarded as the consumption of O2 in the packed-bed or the weight gain of the powders. For example, the weight gain of CuSn0.1 particles being oxidized at 300 oC can be demonstrated by the area marked as “1” in Figure 4.14a. By applying the trapezoidal rule, the relationships between the relative weight gain (∆m/m) and the oxidation time (t) at 300 oC and 500 oC are shown in Figure 4.14b,d, respectively. Figure 4.13. O2 partial pressures in the product streams of the packed-bed reactor when the volume percents of O2 in feed stream are 1.4 % (a) and 10 % (b). 79 Figure 4.14. (a) O2 flow rates in the product streams of the packed-bed reactors when the volume percent of O2 in feed stream is 1.4 %. (b) and (d) The weight gains of the powder during the oxidation at 300 oC (b) and 500 oC (d). They are calculated by applying the trapezoidal method on the results as shown in (a), for example the area marked by “1” for the CuSn o0.1 sample at 300 C. (c) The initial 20 min of the oxidation process at 300 oC. Low temperature oxidation mechanism At 300 oC, O2 was observed earlier in product stream of CuSn0.05 and CuSn0.1 particle- packed reactors than Cu particle-packed reactor. The subsequent plateaus of the O2 flow rate in the product streams suggest the slow oxidation rates after the significant oxidation in the initial 20 min. The oxidation of Cu particles lasts longer than CuSn0.05 and CuSn0.1 powders. As shown in Figure 4.15, after being oxidized, a trace amount of Cu remained in 80 the Cu powders with low intensity peaks. However, peaks attributed to CuO and Cu2O phases are pronounced in the post-oxidation powders of Cu, which is consistent with the results in Figure 4.10 and the reported oxidation products of 1 μm Cu particles at 300 oC (Feng et al. 2003). For the post-oxidation powders of CuSn0.05 and CuSn0.1, distinct peaks belonging to metals (Cu and Cu13.7Sn) are observed, compared to those attributed to Cu2O and SnO2 phases. SnO2 instead of SnO is seen because SnO2 is more stable than SnO in the presence of O2 gas (Dai et al. 2002; Duan 2008). The CuO phase is negligible in the post-oxidation powders of CuSn0.05 and CuSn0.1. Figure 4.15. XRD patterns of powders after being oxidized in a packed-bed reactor at 300 oC (bottom) and 500 oC (top). The volume percent of O2 in the feed stream is 1.4 %. 81 The kinetic equations to describe the relative weight gains of Cu, CuSn0.05, and CuSn0.1 powders with the oxidation time are obtained by fitting the curves in Figure 4.14b. CuSny (y = 0, 0.05, and 0.1) powders underwent rapid oxidation until 20 min. The rapidly formed oxide layer, consisting of both copper and tin oxides, may reduce the oxidation rates of these particles after 20 min. Thus, the relative weight gains with oxidation time can be described by logarithmic equations. The higher reactivity of Sn towards O2 than Cu causes the faster formation of oxide layer (Liu et al. 2005; Liu et al. 2007a), which may lead to the relative weight gains of powders in the following sequence: Cu>CuSn0.05>CuSn0.1. As shown in Figure 4.14c, in the first 20 min of oxidation, the oxidation rates of these particles are observed to obey the linear equations. The apparent rate constants of the oxidation rate in the initial 20 min can be obtained by: ∆𝑚 𝑑( ) 𝑘 𝑚𝑎𝑝𝑝 = (4.1) 𝑑𝑡 kapp,Cu = 2.2×10 -3 min-1 and k -3 -1app,CuSn0.05 = 2.0×10 min are significantly higher than k -4 -1app,CuSn0.1 = 9.1×10 min . Thus, higher reactivity of Sn towards O2 than Cu may contribute to the faster formation of oxide layer on the particle surface and lower oxidation rates of CuSn0.1 and CuSn0.05 than Cu particles in the initial 20 min. According to classic oxidation theory, the fast formation of the oxide layer in the initial oxidation step will set up a strong electric field in the oxide layer, creating an obstacle to diffusion of O2-, and resulting in the logarithmic growth of oxide layer (Cabrera and Mott 1949). Therefore, we believe that the oxidation of Cu, CuSn0.05, and CuSn0.1 powders at 300 oC is controlled by the migration of absorbed O2- through the oxide layer. 82 Figure 4.16 TEM images of Cu (a)-(b), CuSn0.05 (c)-(d), and CuSn0.1 (e)-(f) particles after being oxidized in packed-bed reactor under 300 oC. The volume percent of O2 in feed stream is 1.4 %. (b), (d), and (f) are the higher magnification images of the areas marked by red boxes in (a), (c), and (e), respectively. Some researchers also claimed that the oxidation of Cu-based composite particles was controlled by the internal diffusion of O2 through the oxide ash layer on the particle surface 83 by analyzing their experimental results using a shrinking-core model (SCM) (Chuang et al. 2010). As detailed in Section A1 of Appendices, SCM was used to analyze the experimental data in Figure 4.14b. This model does not clearly explain the experimental observations for the following reasons: 1) the diffusivities of O2 through oxide ash layer either cannot be obtained (for Cu particles) or are 104 higher than the reported values for the diffusion coefficients at the same temperature (for CuSn0.05 and CuSn0.1 particles), as shown in Figure A.1.2 (Appendices). This phenomena may be attributed to the different experimental conditions between the reported work and ours. In the reported work, the oxidation of Cu-Al2O3 composite particles was conducted in a fluidized-bed (Chuang et al. 2010), which has different mass transport phenomena than that in a packed-bed of CuSny solid particles. 2) the key assumption in the SCM of an oxide layer of uniform thickness (Levenspiel 1999) may not be appropriate for CuSny powders, which have oxide layers that vary in thickness as shown in Figure 4.16. Therefore, we believe the O2- migration through the oxide layer instead of the internal diffusion of O2 through the oxide ash layer controls the oxidation of CuSny particles at 300 oC. High temperature oxidation mechanism O2 appeared later in the product streams at 500 oC than 300 oC, implying longer and more significant oxidation of the CuSny powders at 500 oC than that at 300 oC (Figure 4.14a). In addition, the O2 concentration in the product stream of CuSn0.1 powders reached a plateau faster in than those in CuSn0.05 and Cu particles. The disappearance of peaks attributed to metals in the post-oxidation particles of Cu and CuSn0.05 also indicates the higher oxidation extent at 500 oC than 300 oC (Figure 4.15). CuO is the main oxidation product of Cu and CuSn0.05 powders. Cu2O phase is still observable in the post-oxidation 84 products of CuSn0.1 than Cu or CuSn0.05 powders. This result implies that the existence of sufficient Sn increases the particle oxidation resistance. SnO2 is the product of the Sn oxidation in both CuSn0.05 and CuSn0.1 powders. The kinetic equations to describe the weight gains of Cu, CuSn0.05, and CuSn0.1 powders with the oxidation time are obtained by fitting the plots in Figure 4.14d. The relative weight gains of these particles with the oxidation time can be described by parabolic formula. The oxidation mechanism could change dramatically by elevating the oxidation temperature and this results in the change in oxidation behavior of the materials, as evidenced by the shifting of oxidation equations from logarithmic to parabolic or cubic (Cabrera and Mott 1949; Feng et al. 2003; Fujita et al. 2013). For CuSn0.1 particles, the sufficient Sn loading may greatly increase the their oxidation resistance by fast formation of a self-limiting oxide layer in the initial stage of oxidation, due to the higher activity of Sn towards O2 than Cu (Cabrera and Mott 1949; Feng et al. 2003; Liu et al. 2005; Liu et al. 2007a). The rapid growth of the SnO2 layer has been reported that it can block further incorporation of oxygen into the bulk (Korber et al. 2011), which can prevent further bulk oxidation. In addition, grain growth of oxides is observed in the post-oxidation powders of CuSn0.1 at 500 oC. The full width at half maximum (FWHM) of the SnO2 peak at 26.7 o decreases from 1.056 (oxidized at 300 oC) to 0.441 (oxidized at 500 oC), as shown in Figure 4.15. The FWHM of the Cu2O peak at 36.6 o decreases from 0.611 (oxidized at 300 oC) to 0.254 (oxidized at 500 oC). The CuO phase does not appear in the post-oxidation sample at 300 oC. In the sample at 500 oC, the FWHM of the CuO peak at 35.8o is 0.312. Therefore, after being oxidized at 500 oC, the oxides in CuSn0.1 powders display a higher degree of crystallinity, because the smaller FWHM 85 represents larger size of crystalline domains based on Scherrer equation. Therefore, the decrease of FWHM of XRD peaks implies the grain growth in the powders. Grain growth was also observed in the post-oxidation products of Cu and CuSn0.05 particles. The FWHMs of the CuO peak at 35.8o are 0.233 for CuSn0.05 particles and 0.308 for Cu particles at 500 oC. Furthermore, the TEM images of the post-oxidation products of 500 oC also display higher degree of crystallinity than at 300 oC (Figures 4.17, 4.16, respectively). Thus, compared to the post-oxidation sample of 300 oC, results suggest the oxides undergo grain growth during the oxidation of 500 oC. The grain growth of the oxide may be the rate- limiting step for the oxidation of Cu, CuSn0.05, and CuSn0.1 particles at 500 oC, with the relative weight gain with oxidation time following a parabolic equation, which is consistent with similar phenomena observed by other researchers in Cu oxidation (Fujita et al. 2013; Matsnnaga and Homma 1976). 86 Figure 4.17. TEM images of Cu (a)-(c), CuSn0.05 (d)-(f), and CuSn0.1 (g)-(i) particles after being oxidized in packed-bed reactor under 500 oC. The volume percent of O2 in feed stream is 1.4 %. (b), (e), and (h) are the higher magnification images of the areas marked by red boxes in (a), (d), and (g), respectively. (c), (f), and (i) are the selected area electron diffraction (SAED) patterns of areas highlighted by yellow boxes in (b), (e), and (h), respectively. SAED images are obtained by performing the fast Fourier transform (FFT). The identified phases are CuO (PDF No. 01-073-6023) and SnO2 (PDF No. 01-072-1147). The relative weight gain (∆m/m) of Cu particles also approaches the theoretical limit given by complete oxidation of Cu to CuO, which is 0.25. Based on the theoretical 87 calculation, the weight gains of CuSny (y ≤ 0.1) particles given by complete oxidation to CuO and SnO2 are also 0.25. Therefore, the oxidation extent of CuSn0.1 particles within 100 min during the packed-bed reaction, which is 0.19 (Figure 4.14d), is lower than the theoretical limit and the oxidation extent of Cu particles. This further validates that CuSn0.1 powders have higher oxidation resistance than Cu particles. To further examine the oxidation resistance of CuSn0.1 powders, O2 concentration in the feed stream was increased from 1.4 vol % to 10 vol %. As shown in Figure 4.18, even though the O2 concentration in the feed stream increases by ~ 7 times, the relative weight gain of CuSn0.1 powders only increases by 1.6 times. This suggests that the O2 gas concentration may not be the rate-limiting step of the particle oxidation at 500 oC. The relative weight gain of CuSn0.1 powders during the oxidation under 10 vol % O2 is close to that of Cu particles during the oxidation under 1.4 vol % O o2 at 500 C. This also implies the existence of Sn-enriched layer on the CuSn0.1 particles greatly improves the oxidation resistance compared to Cu particles. In TGA measurements, significant weight gain can be observed from 133 oC for Cu particles, which is 44 oC lower than that of CuSn0.1 particles (Figure 4.19). In addition, the relative weight gain of Cu is 4 % higher than that of CuSn0.1 at 233 oC. 88 Figure 4.18. (a) O2 flow rate in the product stream of the packed-bed reactor when the volume percent of O2 in feed stream is 10 %. (b) The relative weight gains of the powder during the oxidation. The weight gain is calculated by applying the trapezoidal method on the results shown in (a). (c) XRD pattern of powders after being oxidized in a packed-bed reactor at 500 oC under 10 vol % O2 stream. 89 Figure 4.19. TGA (left axis) and DSC (right axis) measurements on Cu particles (black curves) and CuSn0.1 particles (green curves). The initial oxidation temperature of CuSn0.1 (vertical green dashed line) is ~ 44 oC higher than Cu (black dashed line). There are two distinct exotherms in both heat flow curves at 380 oC (Cu) and 365 oC (CuSn0.1) corresponding to the weight losses. As discussed, spherical solid CuSny particles are generated when the Sn/Cu atom ratio is below 0.1. The thickness of the Sn-enriched layer on the spherical particle surface increases with Sn loading in the particles, which is supported by SEM (Figure 4.1), TEM with elemental mappings (Figure 4.2), XPS (Figure 4.6), and mass balance calculations (Table 4.1). Our engineering of CuSny particles with a Sn-enriched layer on the surface can increase the oxidation resistance of powders, as shown in the XRD with refinement (Figure 4.10), in-situ XRD (Figure 4.11), and packed-bed oxidation (Figure 4.14). Furthermore. CuSn0.1 particles exhibited highest oxidation resistance among all the CuSny (y≤0.1) particles with desired spherical morphology at both 300 oC and 500 oC. 90 4.5 Summary of this chapter Spray pyrolysis has been utilized to fabricated CuSny products with Sn/Cu atomic ratio ranging from 0 to 1. Spherical micron-sized CuSny solid particles are obtained when the Sn/Cu atomic ratio is below 0.1. CuSny (y≤0.1) particles with a Sn-enriched layer on the particle surface exhibited better oxidation resistance than Cu particles, as confirmed by XRD with Rietveld refinement analysis, in-situ XRD, and oxidation kinetics studies. Based on theoretical analysis of the oxidation, the migration of O2- through the oxide layer is a possible rate-limiting step when CuSny powders are oxidized at 300 oC. For oxidation at 500 oC, grain growth of the oxide likely controls the oxidation process. Micron-sized CuSny solid powders with high oxidation resistance and conductivity are a promising alternative material to replace noble metal-based conductive powders, such as Ag and Au, in the fields of printed electronics, interference packaging, and solar cells metallization. CuSn0.1 powders displayed the highest oxidation resistance with best spherical morphology among the CuSny powders. In addition, the scalable particle fabrication process provides an extra benefit for industrial-scale manufacturing. However, to evaluate the utility of these powders for applications requires high conductivity. Further characterization of their electrical properties is still demanded, and is the subject of ongoing investigation in our group. 91 Chapter 5: Conductive one-dimensional and two-dimensional structures fabricated by Cu-Sn binary particles 5.1 Abstract Cu-Sn powders are promising alternatives to Ag and Au in applications including printed electronics because of their low cost and high oxidation-resistance. Further development requires the knowledge of the conductivity of their corresponding one- dimensional and two-dimensional structures. Herein, CuSny (y=atom ratio of Sn/Cu) wires and films were produced by direct printing. In-situ measurements of structural resistivities while varying the temperature from 2 K to 400 K in oxygen-free conditions revealed that CuSn0.1 wires have comparable resistivities to Cu wires. Furthermore, CuSn0.1 films exhibited significantly lower resistivity increase after being heated at 573 K in ambient air, compared with Cu films, indicating the CuSn0.1 film are higher oxidation-resistant than Cu film. 5.2 Introduction Cu particles are potential alternatives to replace Ag and Au powders in printed electronics, solar cell metallization, and interference packaging, because of their low cost and relatively low resistivity (Jeong et al. 2008; Li et al. 2017; Liu et al. 2016; Wu et al. 2016). However, Cu particles are easily oxidized even at room temperature (Jeong et al. 2013; Kim et al. 2004), resulting in increased resistivity, which greatly restricts their application. To prevent the formation of the oxide layer, approaches such as introducing secondary metals (Au and Ag) or organic molecules to coat the surface of Cu particle have 92 been proposed (Jianfeng et al. 2011; Jung et al. 2011; Sharma et al. 2014). However, neither noble metals nor organic molecules are ideal because of their high material cost or high resistivity, respectively. In Chapter 3, micron-sized solid CuSny particles with a Sn- enriched surface layer have been developed to address these issues. The Sn-enriched layer can form a sacrificial oxide on the surface to prevent further oxidation of the particles, as discussed in Chapter 4. Although CuSny particles are still expected to maintain low resistivity as a result of the relatively low resistivities of tin and tin oxides, introducing Sn into Cu solid solutions can still deteriorate particle conductivity since tin and tin oxides have higher resistivities than copper (Liu et al. 2007a; Shirai et al. 2016; Smithells et al. 2004). Therefore, further investigation is required on the resistivities of structures fabricated using oxidation-resistant CuSny particles to evaluate the prospect of applying these particles. Among the methods for converting metal particles into electronic patterns, the direct printing process has unique advantages, such as low cost, scalability to mass-production, and flexibility in designing patterns (Cao et al. 2008; Shen et al. 2014a). This process has been successfully applied to print conductive wires from inks containing Ag, Au, or Cu particles (Moonen et al. 2013; Schirmer et al. 2011; Shen et al. 2014a). These wires displayed resistivities comparable to their corresponding bulk materials. However, the previous investigations mainly focused on measuring the resistivities at/above the room temperature (Jeong et al. 2013; Jeong et al. 2008; Li et al. 2017; Liu et al. 2016; Wu et al. 2016). Moreover, for Cu-based materials, the influence of oxidation on the pattern resistivity is a crucial factor compared to Ag and Au-based materials. 93 Herein, we assessed the prospect of applying Cu-Sn powders in the aforementioned areas by in-situ measurement of the resistivities of one-dimensional and two-dimensional CuSny structures fabricated by direct printing inks containing CuSny powders. The resistivity measurements were conducted under various conditions including continuously changing the measurement temperature from 2 K to 400 K in oxygen-free environment and heating CuSny patterns at 573 K in ambient air for 90 min, enabling a systematic study of the effects of temperature and oxidation on the resistivities of CuSny structures. 5.3 Experimental Materials Ethylene glycol (EG, ≥99%) and nitric acid (HNO3, 70%) were purchased from Sigma- Aldrich. Cu(NO3)2∙3H2O (99.5% purity) was purchased from Strem Chemicals. SnCl2∙2H2O (>95%) was bought from Fisher Scientific. All chemicals were used without further purification. N2 gas (99.5%) and mixture gas (5% H2-95% N2) were obtained from Airgas. Quartz substrates with dimensions of 5 mm×5 mm×1 mm were purchased from MTI Corporation. Cu and CuSny particles synthesis Cu and CuSny particles were produced by spray pyrolysis with a 1.7 MHz ultrasonic atomizer (Chapters 3-4). Aqueous precursor solutions contained 1.0 M Cu(NO3)2, y M SnCl2 (y based on the designed atomic ratio of Sn/Cu in CuSny particles), 0.1 M HNO3, and 4.8 M EG. After the precursors were atomized into droplets, the droplets were carried by N2 gas through two furnaces (both set at 1000 oC) in series. The residence time of the 94 process is 1.2 s, which was calculated based on the plug-flow reactor with the temperature profile measured along the axial direction (Chapter 3). The product particles were then cooled by quench gas (N2) and collected by a polytetrafluoroethylene (PTFE) filter. One-dimensional and two-dimensional structures fabrication by direct printing Ink with a concentration of (1 g powder)/(1 ml EG) was prepared by dispersing metal particles into EG and then sonicating the mixture for 1 h. A 5-ml syringe with needle (ID=0.2 mm) was utilized to print liquid lines on the quartz substrates. One drop of the ink (~0.025 ml) was added onto the substrate through pipette to form a liquid film. Structures were sintered at 573 K, 673 K, or 773 K in H2-N2 mixture gas for 2 h to form wires and films. Characterization Particle morphology and structure were determined by scanning electron microscopy (SEM, Hitachi SU-70) with energy-dispersive X-ray spectroscopy (EDS) and transmission electron microscopy (TEM, Jeol Jem 2100 LaB6). Powders were sonicated in ethanol for 1 min and then added onto silicon wafers (for SEM) and TEM grids (for TEM). X-ray diffractometer (XRD, Bruker D8 advance) was applied to understand the crystal phases in the samples. The temperature-dependent resistances of wires were measured by four-point- probe method from 2 K to 400 K with 0.01 K step in helium. The setup of the experiments is shown in the inset of Figure 5.5a. The area of the cross-section of wire was obtained by profilometry (New View 6000), as shown in Figure 5.6. The effect of oxidation on the film resistance was determined using the four-point-probe method in a probe station (WL250, Signatone, Inc) equipped with a hot chuck system. The hot chuck system was stabilized at 95 a set temperature within ±4 K. The oxidation was operated at 573 K in ambient air. It took 5 min for the chuck to reach 573 K from room temperature. 5.4 Results and discussions The one-dimensional and two-dimensional structures were fabricated by direct printing inks that are suspensions of CuSny powders in ethylene glycol (EG). The metal powders were obtained by spray pyrolysis. Aqueous precursor solutions containing Cu(NO3)2, SnCl2, and EG were ultrasonically atomized into droplets. The droplets were then transported by the carrier gas (N2) into a heating unit, where a series of processes happened, including droplet solvent evaporation, chemical reaction, nucleation, and crystal growth. In the end of this continuous process, the powders were collected by filtration. 96 Figure 5.1. Powders used in the inks. (a-b) SEM images of Cu (a) and CuSn0.1 (b) particles. (c-d) TEM images of Cu (c) and CuSn0.1 (d) particles. (e) XRD diagram of Cu, CuSn0.05, and CuSn0.1 particles. Solid black circles and squares represent Cu13.7Sn (PDF No. 03-065- 6821) and Cu (PDF No. 01-070-3038), respectively. Figure 5.2. SEM image of CuSn0.1 particles. The small particles on the particle surface, indicated by the red arrow, are believed to be Sn particles. The prepared Cu particles exhibit solid and spherical morphology with a smooth particle surface, as shown in scanning electron microscopy (SEM) and transmission electron microscopy (TEM) images (Figure 5.1a,c), while small particles are observable on the surface of CuSn0.1 particles (Figures 5.1b,d and 5.2). These small particles (denoted by red arrows in Figures 5.1d,5.2) are believed to be Sn particles, attributed to evaporation of Sn and subsequent gas-to-particle conversion during spray pyrolysis, as discussed in Chapter 4. In addition, SnCl2 is easily hydrolyzed in aqueous solution (Perry 2011). During the solvent evaporation, granular precipitates could concentrate on the droplet surface, where the droplet is supersaturated, promoting the formation of Sn-enriched surface layers on the CuSny particles. The as-prepared metal particles show a high degree of crystallinity 97 with pronounced peaks in X-ray diffractometer (XRD) diagram without observation of oxide phases (Figure 5.1e). A blue shift of the peaks with the increase of Sn concentration in the product is caused by dissolving Sn atoms into Cu matrix, same phenomenon was observed in Chapter 4. The full width at half maximum (FWHM) of each XRD curve is 0.125 at 43.4o for Cu, 0.203 at 42.9o for CuSn0.05, and 0.378 at 42.5 o for CuSn0.1. The increasing FWHM indicates the decreasing grain size based on Scherrer equation. The smaller grain size results in a higher fraction of atoms on the grain boundaries (Schiotz et al. 1998), which may lead to an increased electrical resistivity after converting these powders into electronic patterns (Mayadas and Shatzkes 1970; Sharma et al. 2014). The CuSny particles have a number mean diameter of 710 nm with a standard deviation of 300 nm, consistent with Chapter 3. Based on mass balance calculations between the precursors and product particles presented in Chapter 4, the configuration of CuSn0.1 particles is 3- nm-Sn surface layers on the Cu13.7Sn cores assuming the Sn enriched layer are pure Sn and no quantity loss of Cu and Sn caused by the evaporation during the spray pyrolysis. For CuSn0.05 particles, the Cu32Sn cores are estimated covered by Sn layers of 2.1 nm. Figure 5.3. (a) Measured viscosity of H2O-EG mixture solutions with varying volume percent of EG determined using a viscometer. (b) Vapor pressures of H2O and EG as a function of temperature, adapted from (Yaws 1999). 98 Figure 5.4. Microstructural evolution caused by increasing sintering temperature. (a) Schematic of direct printing ink on a 5 mm×5 mm quartz substrate. (b) An image of the wire after being sintered. (c-e) SEM images of wires after being sintered at 573 K (c), 673 K (d), and 773 K (e). (f) Schematic summary of microstructural evolution during sintering (f). EG was used as the ink solvent because its high viscosity and low vapor pressure (Figure 5.3) can prevent the formation of a ring stain on the contact line (Talbot et al. 2014), contributing to a relatively uniform particle distribution after the solvent dries. The inks were sonicated for 1 h, then printed on 5 mm×5 mm quartz substrates, as shown in Figure 5.4a-b. The printed patterns were sintered under H2-N2 gas at 573 K, 673 K, and 773 K for 2 h. The minimum sintering temperature (573 K) was selected based on the initial decomposition temperature of EG which is 513 K (Yue et al. 2012). The post-printing 99 process is important for removing the solvent and decreasing the porosity of wires in order to decrease the resistivity (Ma et al. 2014). At low sintering temperature (573 K), the CuSn0.1 particles largely retained spherical and intact morphology (Figure 5.4c), similar to that of freshly fabricated particles (Figure 5.1b,d). This phenomenon could be ascribed to the low input of thermal energy. Increasing the sintering temperature to 673 K leads to significant inter-particle necking, as shown in Figure 5.4d, contributing to the formation of conduction pathways. When the sintering temperature is 773 K, it is hard to observe distinct particle shapes in the microstructure of wire (Figure 5.4e). The microstructural evolution of the wires with the increasing sintering temperature is summarized in Figure 5.4f, which is consistent with the reported phenomena of sintering Cu and Ag patterns (Jeong et al. 2013; Shen et al. 2014a; Yong et al. 2015). Large quantity of conduction pathways are created at elevated temperature as a result of high input of thermal energy (Jeong et al. 2013; Shen et al. 2014a; Yong et al. 2015), which may lead to a low pattern resistivity. 100 Figure 5.5. In-situ resistivity measurements of the wires from 2 K to 400 K. (a) Resistivities of CuSn0.1 wires being sintered at 573 K (red) and 673 K (black). Inset is the image of the experimental setup. (b) Resistivity-temperature profiles of Cu (black), CuSn0.05 (blue), and CuSn0.1 (red) wires. These wires were sintered at 773 K after direct printing. CuSn0.1 wires were fabricated from different batches. (c) Thermogravimetric analysis (TGA) of CuSn0.1 wire. Table 5.1. Dimensions of the wires fabricated by direct printing CuSn0.1 CuSn0.05 Cu Sintering temperature 773 673 573 773 773 (K) Average cross- 1.2×10-4 2.7×10-4 1.7×10-4 4.0×10-4 1.1×10-4 2.2×10-4 8.0×10-5 1.2×10-4 section area (cm2) Length (cm) 0.45 0.46 0.42 0.45 0.42 0.41 0.46 0.43 Figure 5.6. Typical cross-section profile of wire after sintering. Based on the average dimensions of the wire (Table 5.1), the expected maximum errors are 133 nm in width and 101 1 nm in height. Therefore, the maximum deviation of the measured resistivity caused by the uncertainty of profilometry is 2×10-4 of the measured values. To validate this hypothesis, the resistances of these wires were measured by the four- point-probe method from 2 K to 400 K with 0.01 K steps in an oxygen-free environment. The four-point-probe method has been widely utilized as a tool to measure the resistivities of materials in various geometries (Smits 1958). The experimental setup is shown in the inset of Figure 5.5a. The obtained resistance (R) was converted into resistivity (ρ) by: 𝐴 𝜌 = 𝑅 (5.1) 𝑙 where A is the area of the cross-section of wire, which is obtained by profilometry, as shown in Figure 5.6. l is the length of the wire. The detailed dimensions of each wire are presented in Table 5.1. On average, A = 2×10-4 cm2, l = 0.44 cm. Figure 5.7. Temperature dependent resistivities of Cu, CuSn0.05, and CuSn0.1 wires with y- axis in linear scale. Inset is the resistivities when the measurement temperature was below 40 K. 102 The low sintering temperature leads to high porosity, resulting in a high resistivity (Jeong et al. 2013; Shen et al. 2014a; Yong et al. 2015). The fabricated structures may be regarded as insulators because of the limited conductive paths. The CuSn0.1 wire sintered at 573 K exhibits a resistivity around 103 Ω∙cm when the measurement temperature (Tm) is above 390 K (Figure 5.5a), which is remarkably higher than that of bulk Cu (1.7×10-6 Ω∙cm) and Sn (1.0×10-5 Ω∙cm) (Smithells et al. 2004). In addition, when Tm < 390 K, negligible current goes through the electrodes. This is believed to be caused by poor inter- particle connections when sintered at 573 K. Therefore, as the sintering temperature increases to 673 K, the intensified coagulation between particles creates paths to transfer electrons, contributing to the reduction of resistivities (Figure 5.5a). However, the measured resistivities over 2 K ≤ Tm ≤ 400 K are higher than 2.6×10 -3 Ω∙cm, implying that further increase of the sintering temperature is still required. When sintered at 773 K, the significant fusing between the CuSn0.1 particles contributes to the lowest resistivity (around 10-4 Ω∙cm, 2 K ≤ Tm ≤ 400 K, Figure 5.5b, red curves) among the examined sintering temperatures. In addition, the resistivity-temperature curves (Figures 5.5b,5.7) are observed to follow the classic resistivity-temperature relationship, which can be described as (Kittel 2005), 𝜌 = 𝜌0 + 𝜌(𝑇𝑚) (5.2) where ρ0 is the resistivity ascribed to the scattering of the electron waves, which is caused by the static defects in the lattice. ρ0 is independent of temperature. ρ(Tm) represents the resistivity resulted from the thermal photons. Consequently, ρ(Tm) is temperature- dependent and vanishes when Tm approaches 0 K (Kittel 2005). At low temperatures, the resistivity is dominated by the scattering of the electron waves, supported by the plateaus 103 in Figure 5.7 (Tm ≤ 30 K). ρ(Tm) becomes pronounced at high temperatures because the rate of collision between the electron and thermal photons/electrons is intensified. The collision rate is proportional to thermal photons and can be assumed to be proportional to temperature at high temperatures (Kittel 2005). Therefore, the resistivity increases linearly with temperature when the temperature exceeds a threshold (Tm ≥ 50 K in Figure 5.7). Figure 5.8. Temperature dependent resistances obtained by measuring the same wires twice. (a) Three CuSn0.1 wires. (b) Cu, CuSn0.05, and CuSn0.1 wires. All the wires were sintered at 773 K. CuSn0.1 particles produced from different batches were used to fabricate four CuSn0.1 wires for the resistivity measurements to understand the repeatability of this method. The fabricated CuSn0.1 wire exhibits a weight gain of 23 % (Figure 5.5c), validating the 104 composition of wire is CuSn0.1. As shown in Figure 5.5b (red curves), the resistivities of CuSn0.1 wires are all smaller than 7.5×10 -4 Ω∙cm with a narrow deviation over the wide temperature range (2 K ≤ Tm ≤ 400 K). When Tm = 2 K, the mean value and standard deviation are 1.3×10-4 Ω∙cm and 9.7×10-5 Ω∙cm, respectively. When Tm = 400 K, the CuSn0.1 wires have a mean resistivity of 3.9×10 -4 Ω∙cm and a standard deviation of 2.7×10- 4 Ω∙cm. Moreover, the resistivity of each wire was measured twice to further understand the possible uncertainty. As shown in Figure 5.8, the resistivities of each wire show consistency without significant deviation. Figure 5.9. SEM images of CuSn0.1 wire after being sintered at 773 K (a) with EDS mappings of overlap mapping (b), Cu (c) and Sn (d) elements. Introducing Sn into Cu particles can increase the oxidation resistance of Cu particles as well as the electrical resistivity, attributed to the higher resistivity of tin than copper and possible grain boundaries in solid solutions (Mayadas and Shatzkes 1970; Sharma et al. 2014; Smithells et al. 2004). As shown in Figure 5.9, strong signals from both Cu and Sn elemental mapping validate the coexistence of Cu and Sn in the wire after being sintered. The resistivity of Cu wire ranges from 2.1×10-5 Ω∙cm (Tm = 2 K) to 7.9×10 -5 Ω∙cm (Tm = 105 400 K), as shown in Figure 5.5b (black curve). Cu wire has lower resistivities than CuSn0.05 (blue curve) and CuSn0.1 (red curves) wires. However, CuSn0.1, CuSn0.05, and Cu wires still display low resistivities. For comparison, Cu-Au wires fabricated from similar sized CucoreAgshell particles (1.7 µm) by the same method (direct printing) show a resistivity of ~5×10-3 Ω∙cm at room temperature (Li et al. 2016c), which is higher than the resistivities of our CuSny wires. Figure 5.10. XRD pattern of the CuSn0.1 film after being sintered at 773 K in H2-N2 gas for 2 h. The peaks are attributed to Cu13.7Sn phase with PDF No. 03-065-6821. The Sn-enriched surface layer of CuSny particles can prevent particle oxidation by forming a sacrificial oxide layer, as discussed in Chapter 4. Thus, we investigated the effect of oxidation on the structural resistivity by in-situ monitoring the relative resistivity increases of CuSny and Cu films at 573 K in ambient air. The metal films were also fabricated by direct printing inks on 5 mm×5 mm quartz substrates and subsequent sintering at 773 K in H2-N2 gas for 2 h. The as-prepared metal films are oxide-free, supported by the XRD results. For example, only Cu13.7Sn phase are observed in CuSn0.1 film (Figure 5.10), same as the XRD pattern of CuSn0.1 powders (Figure 5.1e). The 106 resistances of the films were also measured by the four-point-probe method, as shown in the inset of Figure 5.11. The resistances of the metal films range from 10-3 to 10-2 Ω before oxidation (Figure 5.12). However, the resistance of Cu film increases remarkably after the onset of oxidation, while the resistances of CuSny films remain below 0.1 Ω, even after being oxidized for 90 min. Figure 5.11. Resistivity ratios of film resistivities at given time (t) during oxidation at 573 K in ambient air to the resistivities before the oxidation. The films were sintered at 773 K after direct printing. The inset is the schematic illustration of the measurement. Figure 5.12. In-situ resistance measurements of Cu, CuSn0.05, and CuSn0.1 films during oxidation at 573 K in ambient air. 107 To demonstrate the oxidation effect on the film resistivity, Figure 5.11 presents the ratio of resistivity at any given time to the resistivity before oxidation (ρ(t)/ρ(0)) with the oxidation time. The resistivity ratios (ρ(t)/ρ(0)) of Cu and CuSny films with the oxidation time at 573 K are similar to the reported weight gains (Δm/m) of Cu and CuSny particles with oxidization time at 573 K, presented in Chapter 4. The resistivity ratios increase rapidly for Cu and CuSny films in the initial 3 min. After that, the resistivities of CuSn0.05 and CuSn0.1 films are less than two times of their resistivities before oxidation. In the meantime, ρ(t)/ρ(0) of Cu film continuously increases to 6×104 until being oxidized for 40 min, and then remains around that value. As reported, a 1.1-nm-thick oxide layer can be expected to form on the surface of 140-nm-diameter Cu particles for one year in ambient air at room temperature (Kim et al. 2004), resulting in an increased resistivity of Cu particles. Similar to the particle oxidation, the oxidation of Cu-based films also progresses from the surface towards the bulk. The formation of a sacrificial tin oxide layer has been reported to block the further incorporation of oxygen into the bulk (Korber et al. 2011), preventing the further oxidation of CuSny particles. Thus, the conductive pathways beneath the surface of CuSny wires are still able to transfer electrons even though oxidation happens on the surface. Contrarily, the significant oxidation of Cu results in remarkably less conductive paths, causing the highest increase of relative resistivity. Thus, the higher oxidation-resistance of CuSny particles than Cu particles is substantiated again by in-situ resistance measurements of their corresponding films. This is consistent with the results obtained from measuring the O2 consumptions during oxidizing Cu and CuSny particles (Chapter 4). 108 5.5 Summary of this chapter We successfully fabricated CuSny wires and films via direct printing inks containing oxidation-resistant CuSny particles and ethylene glycol. Regardless of higher resistivity of tin than copper, the CuSn0.05 and CuSn0.1 wires exhibited low resistivities ~10 -4 Ω∙cm (2 K ≤ Tm ≤ 400 K), which is comparable with that of Cu wires. CuSn0.1 wires displayed a lower resistivity than the reported resistivity of wires produced from CucoreAgshell powders. CuSn0.05 and CuSn0.1 films have significantly less resistivity increases than Cu films when being oxidized at 573 K in ambient air. Together with our reported discovery in Chapter 4 that CuSn0.1 powders have the best overall properties, including spherical structure and high oxidation resistance among the CuSny particles, this work further demonstrates that one-dimensional and two-dimensional structures fabricated from CuSn0.1 particles have low resistivities and high oxidation-resistance. Therefore, CuSn0.1 particles are promising materials to replace Au and Ag in applications, including printed electronics, solar cell metallization, and interference packaging. 109 Chapter 6: Colloidal spray pyrolysis approach to engineer hierarchical tin@carbon particles as anode materials 6.1 Abstract Spray pyrolysis is a scalable process to fabricate functional particles as cathode/anode materials in rechargeable batteries from precursor solutions. However, one prerequisite of spray pyrolysis to achieve uniform particle-to-particle composition and structure is a stable precursor solution, restricting its usage to highly soluble salts. Otherwise, extremely acidic precursors are necessary to ease the uncontrollable hydrolysis of the salts and the subsequent precipitation. Moreover, strong reduction agents such as H2 are also needed for complete solid-state reactions, introducing potential safety concerns. Herein, we develop a novel process, colloidal spray pyrolysis (CSP), which can eliminate all the prerequisites simultaneously. Our process can generate particles directly from a multiphase precursor in mild processing conditions through in-situ solid-state reactions. The product structure and composition can be precisely designed based on aerosol dynamics and reaction kinetics. By applying CSP, Sn@C particles with three distinct interior nanostructures have been synthesized and evaluated as anodes for lithium-ion batteries (LIBs) and sodium-ion batteries (SIBs). The best performing Sn@C anode delivers 627.9 mAh/g at 2C with capacity retention of 88.5% after 1500 cycles in LIBs and demonstrates superior rate capability for SIBs. This novel CSP process is promising in preparing electrode materials in LIBs and SIBs for future practical applications. 110 6.2 Introduction Spray pyrolysis is a scalable process with low operating cost, high process throughput, and minimal waste production (Jung et al. 2014; Lin et al. 2016; Zhu et al. 2017). It has been widely used to fabricate functional particles for many applications, including catalysts (Koirala et al. 2016), lithium-ion batteries (LIBs) and sodium-ion batteries (SIBs) (Choi et al. 2015; Reddy et al. 2013), and superconductors (Bang and Suslick 2010). A stable precursor solution with all salts dissolved is a necessity, or the precipitation in the precursor solution during spray pyrolysis can lead to non-uniform droplet-to-droplet concentration. This could cause product particles with non-uniform particle-to-particle composition and structure (Gurav et al. 1993). Therefore, it is challenging to synthesize functional particles from aqueous solutions containing poorly-soluble or easily-hydrolyzed salts. To resist precipitation, metal salts with strong acids and pure organic solvents are normally used to form stable precursor solutions (Route (I), Figure 6.1a) (Xu et al. 2013a; Zhang et al. 2014a). H2 is typically introduced into the carrier gas to promote the complete reduction of the salts (Xu et al. 2013a), or residual oxides could be formed in the product (Hong and Kang 2015; Joshi et al. 2017). However, potential safety concerns accompanied with the usage of low flash point organic solvents (e.g., ethanol, 13 oC), an acidic precursor, and explosive gas (H2) make industrial scale production difficult (Yaws 1999). Here, we present colloidal spray pyrolysis (CSP) for the first time, a process that can generate functional particles with uniform particle-to-particle composition and structure from stable multiphase precursor solutions (Route (II), Figure 6.1a), which could be a promising strategy to prepare uniform electrode materials in LIBs and SIBs. The structure and composition of particles are pre-designable based on our models. Several 111 distinguishable advantages of CSP make it unique when compared to conventional aerosol processes. Unlike spray pyrolysis, CSP can process low solubility salts, expanding the application of spray pyrolysis to a wider range of functional materials. CSP uses an aqueous precursor solution, with no direct addition of H2 gas or pure organic solvents, which makes CSP safer and simpler for scale-up manufacturing. CSP is also different from multi-step spray drying, where colloids are used as templates to be etched away in the end (Boissiere et al. 2011; Jung et al. 2013). CSP is a one-step process with a short residence time where colloids are involved in in-situ solid-state reactions and structural evolution. 112 Figure 6.1. Schematic drawings of conventional spray pyrolysis and colloidal spray pyrolysis. (a) A comparison of conventional spray pyrolysis (Route I) and our strategy (Route II) with their advantages (in blue) and disadvantages (in red). (b) Schematic of colloidal spray pyrolysis to fabricate pitaya-structured, pomegranate-structured, and chestnut-structured particles. 113 To evaluate the robustness of CSP, we targeted tin@carbon (Sn@C) nanocomposite particles that are difficult to synthesize by conventional spray pyrolysis because Sn salts are easily hydrolyzed in aqueous solutions (Perry 2011). Sn@C nanoparticles are also promising anode materials for LIBs and SIBs due to their high theoretical capacities (993 mAh/g for Li4.4Sn and 847 mAh/g for Na15Sn4) and suitable working potentials (<0.5V) (Chen and Sieradzki 2013; Im et al. 2013; Tian et al. 2015; Xu et al. 2013a). Guided by theoretical analysis of the aerosol dynamics and reaction kinetics, Sn@C with different interior nanostructures of Sn in carbon matrices, similar to pitayas, pomegranates, and chestnuts, were synthesized by tuning the process parameters (Figure 6.1b, Table 6.1). The pomegranate-structured Sn@C powders exhibit superior electrochemical performance as anode materials in LIBs and SIBs. CSP with in-situ solid-state reactions opens up new opportunities to produce uniform composition particles with complex structures from precursors without solubility limitations. Table 6.1. Precursors to fabricate different structures of Sn@C particles Pitaya Pomegranate Chestnut [Sucrose] (mol/L) 2.0×10-1 2.0×10-1 1.0×10-2 Sucrose Mass loading 7.0×10-2 7.0×10-2 3.5×10-3 (kg/L) [SnO2] (mol/L) 2.3×10 -2 3.8×10-2 7.6×10-2 Mass loading 3.5×10-3 5.8×10-3 1.2×10-2 SnO2 (kg/L) [SnO2 colloid] 1.1×1019 1.9×1019 3.8×1019 (#/L) Molar ratio 8.9 5.2 0.13 Sucrose/SnO2 Mass ratio 20.2 12.1 0.3 Viscosity of precursor (mPa∙s) 2.24 2.14 1.63 Concentration of ethylene glycol 4.8 (M) 114 6.3 Experimental 6.3.1 Sn@C fabrication by CSP In the precursor, SnO2 colloids (Nyacol, SN15ES) were added to an aqueous solution containing sucrose (FISHER CHEMICAL), and then mixed with ethylene glycol (EG, SIGMA-ALDRICH, ≥ 99 %) to form a 150-ml-precursor solution. The mass ratio of sucrose to SnO2 colloids in the precursor was 20.2 (Sn@C pitaya), 12.1 (Sn@C pomegranate), and 0.3 (Sn@C chestnut), as shown in Table 6.1. Detailed calculation procedures to obtain the colloid concentration are presented in Section A3 of Appendices. The prepared precursor was then atomized into droplets by a home built ultrasonic atomizer with a 1.7 MHz transducer. The droplets were transported by N2 gas through a quartz tube heated by a double furnace system (two furnaces connected in series, 81.3 cm in length). The set points of both furnaces were 750 oC. The residence time of the reactor was 1.5 s for experiments to obtain Sn@C pitaya particles and 4.5 s for the fabrication of Sn@C pomegranate and chestnut particles. The residence time was tuned by changing the carrier gas flow rate into the system. At the end of the tube, the powders were cooled to 57 oC by quench gas (N2) and collected on a polytetrafluoroethylene (PTFE) filter (with a diameter of ~ 9.5 cm). The residence time of our process was calculated by assuming our reactor is a plug- flow reactor, consisting 32 sub-units of identical length. By measuring the temperature in each sub-unit, the residence time of the process can be calculated as: 𝐴×𝑙 𝑇 𝜏 = ∑32 × 𝑟𝑜𝑜𝑚1 (6.1) 𝐹 𝑇𝑖 115 Here, A is the cross sectional area of the tube, l is the length of the calculation element, F is the carrier gas flow rate, Troom is the room temperature, and Ti is the temperature in each element which has been presented in Chapter 3. 6.3.2 Characterization TEM was carried out using a JEOL 2100 LaB6 TEM and a JEOL 2100F FEG TEM equipped with Gatan image filter (GIF, Tridiem 863). The collected powders were dispersed into ethanol and sonicated for ~ 1 min. Then the mixture was added dropwise onto a lacey carbon film coated Cu grid in order to avoid the interference of C signal from the carbon film itself. SEM images were taken by a Hitachi SU-70 SEM. Similar to the TEM sample preparation, the ethanol particle mixture was added dropwise onto a silicon wafer. In addition, a BRUKER D8 advance XRD was used to determine crystallographic information. 6.3.3 Electrode preparation and electrochemical measurement Electrochemical measurements were performed using coin cells (CR2032) assembled in an argon-filled glove box (<1 ppm of water and oxygen). To prepare the Sn@C electrode, Sn@C powders, carboxyl methyl cellulose (CMC), and carbon black with a mass ratio of 70: 20: 10 were magnetically stirred for over 5 h to ensure thorough mixing. The slurry was cast onto a thin copper foil and dried. Prior to cell fabrication, the electrodes were dried overnight at 80 °C in a vacuum. The coin cells were assembled with lithium foil as the counter electrode, 1.0 M LiPF6 solution in a mixture of dimethyl carbonate (DMC) and ethyl methyl carbonate (EMC) (1:1 volume ratio) with 10% fluoroethylene carbonate (FEC) for LIBs or 1.0 M NaClO4 solution in a mixture of ethylene carbonate/dimethyl carbonate (EC/DMC, 1:1 by volume) for SIBs as the electrolyte, and a polypropylene film 116 (Celgard3501, LLC Corp., USA) as the separator. The electrochemical properties of the Sn@C particles were studied with a multichannel battery-testing system (Arbin Instruments, TX, US), using charge/discharge galvanostatic cycling from 2.0 V to 0.01 V with the loading mass of ~ 1.0 mg∙cm-2. The specific capacity was calculated based on the mass of Sn@C powders. The electrochemical impedance spectrum and cyclic voltammetry measurements were carried out on an electrochemistry workstation (Solartron 1287/1260) at room temperature. 6.4 Results and discussion 6.4.1 Aerosol dynamics and reaction kinetics in colloidal spray pyrolysis In CSP, the aqueous precursor solution containing colloids, co-solvent, and stabilizer for colloids is first atomized into droplets. The droplets go through a series of processes including solvent evaporation, colloid reduction, and inter-colloid collision, forming product particles. The particle engineering of CSP involves both the oxide-colloid reduction and inter-colloid collision/coagulation processes (Figure 6.1b). The product composition is controlled by colloid reduction. The inter-colloid collision/coagulation determines the interior nanostructure of the product particles. Although both processes happen simultaneously during CSP, we can still model CSP in the ideal case in which the two processes are isolated. For the colloid reduction, the co-solvent is the main reducing agent in CSP during its decomposition. In a multiphase system, the diffusion of reactant gases through the colloids (Figure 6.1b) may be difficult because of the mass transport resistance. For example, Dgas- in-solid/Dliquid-in-liquid (diffusivity ratio) is as low as 10 -5 at 25 oC (Deen 1998; Nowick 2012). 117 In many cases, the densities of metals and their oxides are comparable. Therefore, the colloid reduction process can be simulated by the shrinking-core model (Levenspiel 1999). In CSP, H2 is mainly generated from co-solvent, which are uniformly dispersed in the droplets. The concentration of H2 (Cs,H2) was assumed to be uniform. The reduction process consists of mass transport of reactants and the chemical reaction. Corresponding equations to calculate the time of the colloid reduction process were derived with detail procedures in A2.1 (Appendices). The chemical formula of the oxide (MOx) reduction into metal (M) can be described as: 𝑥𝐻2 + 𝑀𝑂𝑥 ⇌ 𝑀 + 𝑥𝐻2𝑂 (6.2) When the reduction is controlled by diffusion of H2 through the metal ash outside the oxide core, the colloid reduction time is: 𝜌 2 𝜏 = 𝑀𝑂𝑥 𝑅 (diffusion controls) (6.3) 6𝑥𝐷𝑒𝐶𝑠,𝐻2 If chemical reaction between H2 and the oxide core controls the CSP process, the colloid reduction time can be calculated as: 𝜌 𝜏 = 𝑀𝑂𝑥 𝑅 𝑛 (reaction controls) (6.4) 𝑥𝑘𝑟𝐶𝑠,𝐻2 Here, De is diffusivity of H2 in the metal layer. kr is rate constant. n is reaction order. ρMOx is density of metal oxides. R is colloid radius. Based on Equations (6.3,6.4), the composition of product particles can be controlled. When the residence time of CSP process is larger than the time required for colloid reduction, the oxide colloids can be reduced completely into metal. 118 In aerosol dynamics, the colloid collision frequency within individual droplets is a critical factor influencing the interior structure of the product particles (Nandiyanto and Okuyama 2011). The collision frequency (f) of colloids in the droplets can be expressed as (Levich 1962): 𝑓 = 8𝜋𝐷𝑅𝑁∞(𝑡) 2 (6.5) where D is the viscosity of the media which can be estimated by Stokes-Einstein formula (Equation 6.6). Equation 6.5 can be expanded into Equation 6.7: 𝑘𝑇 𝐷 = (6.6) 6𝜋𝑅𝜇 4 𝑘𝑇 𝑓 = ∙ ∙ 𝑁 (𝑡)2∞ (6.7) 3 𝜇 k is Boltzmann constant. T is temperature. μ is the viscosity of the precursor. t is the time elapsed after the droplets enter the reactor. The collision frequency of colloids significantly depends on the initial colloid loading in the droplet (N∞(0)). A high colloid loading results in a high collision frequency, promoting the formation of large interior structures. Thus, precursor solutions with higher initial colloid loading favor denser particles, while lower colloid loading results in more porous particles composed of smaller beads as the interior structures (Bahadur et al. 2010; Iskandar et al. 2009; Wang et al. 2007b). Based on the analysis and Equations (6.5-6.7), we predict that the increasing colloid loading in the precursor solution may result in a growth of interior structure of the product particles, from dispersed small beads to large core, making the structure of product particles similar to pitaya, pomegranate, and chestnut in sequence, as illustrated in Figure 6.1b. 119 6.4.2 Application of colloidal spray pyrolysis in fabricating Sn@C particles. To test the feasibility of CSP, it was applied to produce Sn@C particles from an aqueous precursor solution, containing SnO2 colloids, ethylene glycol (EG) as the co- solvent and sucrose as the stabilizer (Figure 6.1b). The decomposition of EG can generate H2 and CO gases during the processing (Yue et al. 2012; Zhong et al. 2013b). Sucrose acts as a stabilizer for colloids in precursor, carbon source, and reducing agent during its decomposition to carbon (Cho et al. 2013; Wang et al. 2009). Therefore, both EG and sucrose provide H2. The precursor was atomized into 5 μm droplets (a volume mean diameter) by a 1.7 MHz ultrasonic generator (Zhong et al. 2013b). The droplets were then transported by N2 gas into two tube furnaces in series, both set at 750 oC. The SnO2 colloids used in precursor solution have high crystallinity with lattice fringes attributed to the (200) plane of SnO2 and narrow size distribution (number mean = 4.4 nm; standard deviation = 1.3 nm, Figure 6.2a-c). The SnO2 colloids-sucrose-EG aqueous precursor displayed stability with respect to precipitation for at least 2 h (Figure 6.3), which is the prerequisite condition for uniform particle-to-particle composition and structure in products (Gurav et al. 1993). 120 Figure 6.2. Composition control of the product particles by colloidal spray pyrolysis. (a-b) TEM image of the SnO2 colloids used in the precursor (a) with HRTEM image (b). (c) Diameter distribution of SnO2 colloids with mean diameter, standard deviation, and sample size. (d) Schematic of parameters used in calculating the time of colloid reduction in colloidal spray pyrolysis. (e) Time required to reduce a SnO2 colloid as a function of initial colloid diameter, based on reaction-controlled model (left y-axis) and diffusion-controlled model (right y-axis). (f) XRD diagram of powders obtained at 750 oC when the residence time of the process is 4.5 s (black curve), 1.5 s (red curve), and 0.92 s (blue curve). 121 Figure 6.3. (a) is a beaker containing an aqueous solution with 3.8×10-2 M SnO2 and 4.8 M EG. After adding 0.11 M KCl into (a), the solution was stirred for 1 h (b), 2 h (c), and 3 h (d) consecutively. (e) is a beaker containing the precursor for the generation of Sn@C pomegranate particles. The precursor composed of 0.2 M sucrose, 3.8×10-2 M SnO2, and 4.8 M EG. After adding 0.11 M KCl into (e), the solution was stirred for 1 h (f), 2 h (g), and 3 h (h) consecutively. (j) is the image of (e) after being ultrasonically sprayed for 2 h. After adding 0.11 M KCl into (j), the solution was stirred for 1 h shown in (k). Density measurements were taken for the precursor before (e) and after (j) the reaction. When taking the photos, beakers were placed on the marked area of paper with letters of “UMD”. 122 Without adding sucrose, SnO2 colloids precipitated within 1 h after adding KCl into the solution. As shown in (b), “UMD” was hard to see. With sucrose, the SnO2 colloids remained stable and dispersed in the precursor shown in (f)-(h). After adding KCl into the solution (f), “UMD” can still be seen at least for 3 h as shown in (h). For the precursor after being sonicated for 2 h (j), “UMD” can be seen after adding KCl and stirring for 1 h as illustrated in (k). Since the size of SnO2 colloids is small (mean diameter = 4.4 nm), we cannot separate them from precursor via filtration and high-speed centrifuging. This is the best method to demonstrate the high uniformity and the high stability of our precursor. (l) is an aqueous solution with 0.1 M SnCl2, 0.4 M sucrose, and 4.8 M EG. The molar ratio of sucrose to SnCl2 is 4. “UMD” is hard to be observed due to the hydrolysis of SnCl2. (m) is an image of the solution in (l) after being stirred for 1 h. The solution is cloudy. We cannot see the “UMD” through the solution. (n) shows the solution in (m) after being placed for 1 h without stirring. The precipitates concentrate at the bottom of the beaker. Firstly, the reduction time of SnO2 colloids was calculated to understand the practicability of this process. The densities of Sn (ρSn=7.31 g/cm 3, 20 oC) and SnO2 (ρSnO2=6.95 g/cm 3, 20 oC) are comparable (Perry 2011). Hence, the reduction time was determined based on Equations (6.3-6.4) using shrinking-core model, as shown in Figure 6.2d. The detailed simulation procedures are presented in A2.2 (Appendices). The precursor composition is listed in the Column Pomegranate of Table 6.1. The calculated results are summarized and plotted in Figure 6.2e. The ratio of calculated colloid reduction time based on diffusion-controlled model to that based on reaction-controlled model is ~10- 7, when colloid radius is smaller than 75 nm. Therefore, chemical reaction is determined to 123 be the rate-limiting step. In addition, when the residence times are 4.5 s, 1.5 s, or 0.92 s, the maximum radii of the SnO2 colloids that can be completely reduced are 70 nm, 23 nm or 14 nm, respectively (Figure 6.2e). These estimated radii are all larger than that of the SnO2 colloids we used (2.2 nm), which means theoretically SnO2 colloids can be completely converted to Sn by CSP. Experimentally, the SnO2 colloids have been successfully reduced in CSP with the conditions we used in the simulation mentioned above, as shown in Figure 6.2f. All the peaks in the XRD diagram are pronounced and attributed to Sn (PDF No. 01-075-9188) without the existence of oxide phases. When the process residence time is 4.5 s, the product particles exhibit pomegranate-like structure with beads inside the particles (Figures 6.4- 6.5). The product particles change from spherical to fragmental structures with decreasing residence time, which is commonly observed in products by tuning the residence time of aerosol processes (Nandiyanto and Okuyama 2011). 124 Figure 6.4. TEM images of powders obtained at 750 oC with the same precursor solution used to generate Sn@C pomegranate particles. The residence times for the fabricated process are 4.5 s (a), 1.5 s (b), and 0.92 s (c)-(d). The fragmented particles shown in (d) together with Figure 6.5c are believed to stem from the burst of the droplets caused by build-up pressure inside the droplet during solvent evaporation, as discussed in Chapter 3. 125 Figure 6.5. SEM images of powders obtained at 750 oC with the same precursor solution used to generate Sn@C pomegranate particles. The residence times for the fabrication process are 4.5 s (a), 1.5 s (b), and 0.92 s (c). The insets are the high magnification images of the corresponding areas marked by dashed boxes. Fragmented particles are marked by red arrows in (c). The fragmented particles shown in (c) are believed to stem from the burst of the droplets caused by build-up pressure inside the droplet during solvent evaporation, as discussed in Chapter 3. Guided by aerosol dynamics and Equations (6.5-6.7), we have also successfully controlled the interior structure of Sn@C particles by adjusting the initial colloid loading in precursor (Table 6.1). Consistent with the theoretical prediction, when the initial SnO2- colloid loading increases from 1.1×1019 to 3.8×1019 colloids/L, the interior nanostructure in the carbon matrix shifts from small beads (pitaya-structure in Figure 6.6a,d-g), through more coagulated beads (pomegranate-structure in Figure 6.6b,h-k), to large cores (chestnut-structure in Figure 6.6c,l-o), as clearly demonstrated in HRTEM images and their corresponding elemental mappings. From both TEM and SEM images (Figure 6.7), particle size distributions are summarized in Figure 6.6p-v. The average size of Sn beads in C matrix are 16 nm (pitaya-structured), 24 nm (pomegranate-structured), and 139 nm (chestnut-structured). The increasing average size of Sn beads with colloid loading 126 confirms that inter-colloid collision frequency increased with initial colloid loading. Sn@C particles show spherical morphology with average particle diameter decreasing from 340 nm (pitaya-structured), through 301 nm (pomegranate-structured), to 139 nm (chestnut- structured). The mass fraction of Sn in the Sn@C particles increases from 77% (pitaya- structured), through 83% (pomegranate-structured), to 89% (chestnut-structured) with decreasing mass ratio of sucrose to SnO2 in precursor, as shown in Figure 6.8. 127 Figure 6.6. Structure and composition control of the product particles by colloidal spray pyrolysis. (a-c) TEM images of pitaya-structured (a), pomegranate-structured (b), and chestnut-structured (c) particles. (d-n) STEM-HAADF image and EELS elemental mappings of C signal, Sn signal, and overlap result of Sn@C pitaya (d-f, respectively), pomegranate (h-k, respectively), and chestnut (l-o, respectively) particles. The scale bars in STEM-HAADF images are 50 nm. (p-t) Statistical summaries of Sn@C particle and interior Sn bead diameters of pitaya-structured (p,s, respectively), pomegranate-structured (q,t, respectively), and chestnut-structured (r) particles with mean values, standard deviations, and sample sizes. (u-v) Box plots to summarize the size distributions of Sn@C particles (u) and Sn beads (v). (w) Calculations of the colloid reduction time on conditions to generate Sn@C pitaya, pomegranate, and chestnut particles. (x) XRD patterns of the generated Sn@C pitaya, pomegranate, and chestnut particles. Figure 6.7. SEM images of Sn@C pitaya (a), pomegranate (b), and chestnut (c) particles. 128 Figure 6.8. TGA results of the Sn@C pitaya (black), pomegranate (red), and chestnut (blue) powders. The relative weight percentages of Sn were calculated based on following equation (Xu et al. 2013a): 𝑊𝐹𝑖𝑛𝑎𝑙,𝑆𝑛𝑂2 ∙𝑀𝑤 𝑀𝑤 𝑆𝑛 𝑆𝑛 𝑤𝑡 % = 𝑆𝑛𝑂2 ∙ 100 % (6.8) 𝑊𝐼𝑛𝑖𝑡𝑖𝑎𝑙,𝑆𝑛@𝐶 The formation of crystal Sn and amorphous carbon in all three types of Sn@C particles is supported by the theoretical analysis on the colloid-reduction and confirmed by the XRD measurement (Figure 6.6w-x). Although the colloid loading varies between the three conditions to obtain different structured Sn@C particles, as shown in Table 6.1, the colloid reduction time is also controlled by chemical reaction based on the calculated results (Figure 6.6w). The theoretical colloid reduction times are all smaller than the process residence times, suggesting the colloids can be completely reduced by the CSP. The XRD patterns further validate this hypothesis with noticeable peaks from Sn phase (PDF No. 01- 075-9188) and the absent of oxide phases. The full-width at half-maximum (FWHM) of the main peak at 30.8o decreased from 0.27 (pitaya-structured), through 0.25 (pomegranate- structured), to 0.18 (chestnut-structured), indicating that the Sn@C chestnut particles have the largest Sn crystalline domain based on Scherrer equation. Sn@C chestnut has the greatest degree of crystallinity, which is attributed to the fastest nucleation and crystal growth among the Sn@C particle production processes. This is also consistent with the aerosol dynamic analysis that the high initial colloid loading contributes to the high frequency of inter-colloid collision. To validate this aerosol dynamic analysis, we further changed the initial SnO2 colloid loading to 6.1 (mass ratio of sucrose to SnO2 colloids, Table 6.2), a condition between 129 those to obtain Sn@C pomegranate (12.1) and chestnut (0.3) particles. The Sn@C particles fabricated at this condition (termed as intermediate) display a mix of pomegranate- structured and chestnut-structured particles (Figure 6.9), which is consistent with our theoretical analysis. In the XRD pattern of Sn@C intermediate particles (Figure 6.10), the FWHM of the main peak at 30.8o is 0.24, which is also between those of the Sn@C pomegranate and chestnut powders, consistent with the SnO2 loading in the precursor. Table 6.2. Precursor to obtain Sn@C intermediate particles Intermediate [Sucrose] (mol/L) 2.0×10-1 Sucrose Mass loading (kg/L) 7.0×10-2 [SnO2] (mol/L) 7.6×10 -2 SnO2 Mass loading (kg/L) 1.2×10 -2 [SnO 192 colloid] (#/L) 3.8×10 Molar ratio 2.7 Sucrose/SnO2 Mass ratio 6.1 Concentration of ethylene glycol (M) 4.8 Figure 6.9. SEM (a) and TEM (b) images of Sn@C intermediate powders fabricated from a precursor with sucrose/SnO2 (mass ratio) = 6.1, which is between that of Sn@C pomegranate and chestnut particles. 130 Figure 6.10. XRD diagram of Sn@C intermediate powders fabricated from a precursor with [sucrose]/[SnO2] = 2.7, which is between that of Sn@C pomegranate and chestnut particles. All the peaks are attributed to the Sn phase (PDF No. 01-075-9188). It is worth mentioning that the model in the above discussion is based on the ideal situation where the colloid reduction and inter-colloid collision/coagulation are isolated procedures. In reality, these steps happen simultaneously, which means the size of colloids is increasing due to coagulation during the reduction. Therefore, the actual radius of the colloids that needs to be reduced is larger than the radius of the colloids added into the precursors. A precursor solution was also processed at the same conditions to obtain Sn@C pomegranate-structured particles but without the addition of EG (Cs,H2=1.5 M). Based on the reaction kinetics, the colloid reduction time is 0.3 s, which is smaller than the process residence time (4.5 s). Thus, theoretically, SnO2 colloids should be completely reduced. However, residual SnO2 exists in Sn@C particles synthesized from the EG-free precursor, as demonstrated by a weak SnO2 peak in the XRD diagram (Figure 6.11). This 131 phenomenon implies that EG is the main reducing agent in the CSP process and is crucial in tuning the composition of the products. Furthermore, the coagulation of the colloids increases the difficulty of complete reduction of the colloids. However, our model can still be applied to screen the colloids that can be processed by CSP and pre-design product structures. Figure 6.11. XRD diagram of powders fabricated with the same [sucrose]/[SnO2], reaction temperature, and residence time to obtain Sn@C pomegranate structured particles without the addition of EG to the precursor. The peaks are indexed to SnO2 (PDF No. 00-041-1445) and Sn (PDF No. 01-075-9188). The possibility of collision between the product particles should also be considered. Both conventional spray pyrolysis and spray drying have long been considered as one- droplet-one-particle processes (Gürmen et al. 2006; Lengsfeld et al. 2000; Zhao et al. 1998). Therefore, we believe the multiphase spray pyrolysis also shares the same basis. To verify this hypothesis, the collision half-life of the Sn@C particles, defined as the time 132 necessary for the number concentration to fall to one half the value because of collisions, was calculated. Since the average diameters of all the three kinds of the Sn@C particles are less than 340 nm, the collision between the particles depends on the Brownian motion (Friedlander 2000). With the initial colloid loading (Np,∞(0)), viscosity (μ), temperature (T) of the system, and target particle concentration (Np,∞), the characteristic time (t) can be calculated as (Friedlander 2000): 𝑁𝑝,∞(0) −1 𝑁 𝑡 = 𝑝,∞ (6.9) 8𝑘𝑇 𝑁𝑝,∞(0) ∙ 3𝜇 2 Here, k is the Boltzmann constant. To compare the collision time in a quantitative manner, we calculate the half-life of the colloids during the process, where Np,∞(0)/Np,∞ = 2. Then, 3𝜇 𝜏1/2 = (6.10) 4𝑘𝑇∙𝑁𝑝,∞(0) To obtain the collision time by Equations (6.9-6.10), μ was calculated as the viscosity of N2 gas. The viscosity of the N o 2 at 750 C can be obtained as (Yaws 1999): 𝜇 = 42.606 + 0.475𝑇 − 9.88 × 10−5𝑇2 [𝜇𝑃] (6.11) At 750 oC, μ = 425 μP, which is 4.25×10-5 Pa∙s. The Np,∞(0) was obtained by the assumption that our process is a plug-flow reactor. Therefore, collisions could only happen laterally, and not along the flow path (Levenspiel 1999). The average consumption rate of the precursors was determined experimentally as 5.5×10-3 ml/s. Accounting for the average diameter of our droplets (5 μm) (Zhong et al. 2013b), reactor volume (3.0×10-4 m3), and 133 residence times, Np,∞(0) and τ1/2 were calculated and listed in Table 6.3. Therefore, considering the residence times of the conditions, no inter-particle collision would be expected, which validates the one-droplet-one-particle mechanism. Table 6.3. The half-time of inter-particle collision between Sn@C particles Sn@C pitaya Sn@C pomegranate Sn@C chestnut Residence time (s) 1.5 4.5 4.5 N (0) (#/m3) 4.2×1011 1.3×1012 1.3×1012p,∞ τ 31/2 (s) 5.4×10 1.8×10 3 1.8×103 The collision rates between Sn@C particles in all three scenarios were also calculated and shown in Table 6.3. Significantly longer collision/coagulation half-lives than process residence times indicate negligible collision between Sn@C particles, verifying the one- droplet-one-particle model (Lengyel et al. 2014). 134 Figure 6.12. Carbon frames in Sn@C particles. (a-h) TEM images of chestnut-structured (a), pomegranate-structured (d), and pitaya-structured (g) particles with its corresponding HRTEM images (b)-(c), (e)-(f), and (h), respectively. The HRTEM images were taken from the areas marked by red color boxes. (c,f) Higher magnification images of areas in (b) and (e), respectively. (i) STEM-HAADF images (left) and EELS line scan results (right) of chestnut-, pomegranate-, and pitaya-structured particles (from top to bottom). The scale bars in HADDF images are 100 nm. (j) EELS spectra of C K-edge extracted from position A (left) and B (right) of (i). In addition to the interior nanostructure and composition of Sn@C particles, carbon can also be controlled by varying the process parameters. As discussed in above, sucrose produces reducing gases during its decomposition and also results in the formation of 135 carbon (Cho et al. 2013), which serves as frame in Sn@C particles. As shown in Figure 6.12a-h, the edge of the chestnut-structured particles is encapsulated by crystalline carbon layers, revealing long-range ordered lattice fringes with spacing of 3.1~3.3 Å (Figure 6.12a-c). It is consistent with the results that the inter-layer distance in the stacked carbon layers is around 3.3 Å (de Andres et al. 2008; Lee et al. 2013). The strong C signal in electron energy loss spectroscopy (EELS) spectra (Figure 6.12i-j) validates the SncoreCshell structure of chestnut-like particles. The EELS spectral feature of C K-edge from both the edge and center of the chestnut particle displays clear π* peaks at ~ 284 eV (Figure 6.12i- j), which are excited from C atoms with the sp2 coordination environment (Suenaga and Koshino 2010). The nature of π bonding prohibits its formation on structures with sharp arches or small radii, because the p-orbital electrons can only delocalize sideways in a planar formation. Therefore, neither long-range stacking of crystalline carbon layers is observed in TEM images (Figure 6.12g-h) nor significant π* peaks appear in the EELS spectra (Figure 6.12i-j) in Sn@C pitaya particles, because the diameters of the Sn beads of pitaya-structured particles are remarkably smaller than those of Sn@C chestnut particles (Figure 6.6). For the Sn@C pomegranate, the stacking of C layers can only be found in the localized region (Figure 6.12d-f). Therefore, the difference of crystallinity of carbon also verifies the size difference of Sn beads in carbon frame. Raman spectra of Sn@C pitaya, pomegranate, and chestnut are consistent with TEM and EELS observations with Sn@C chestnut showing the highest D/G peak ratio and most pronounced 2D peak (Figure 6.13). Contrarily, Sn@C pitaya exhibits the lowest D/G peak ratio and negligible 2D peak. 136 Figure 6.13. Raman spectra of the Sn@C pitaya (blue), pomegranate (red), and chestnut (black) powders. It was recorded by a Raman spectrometer, equipped with a 532 nm laser. Compared to reported Sn-based functional particle engineering, CSP exhibits distinguished advantages including simple operating procedures, short processing time, mild reducing conditions, and better control of product structure and composition (The reported methods are summarized in Section 2.2 in Chapter 2). The increasing concentration of co-solvent in the precursor solution can tune the product from oxides to metals. Moreover, the desired product structure (pitaya-like, pomegranate-like, or chestnut-like) can be achieved by adjusting the colloid loading in precursor solution. 6.4.3 Electrochemical performance of Sn@C anodes in LIBs and SIBs. The Sn@C particles are promising anode materials for LIBs because Sn is an electrochemically active component with a high capacity (The theoretical specific capacity of Sn is 993 mAhg-1 for LIBs) (Huang et al. 2015; Winter and Besenhard 1999) and the elastic carbon matrix can effectively act as a sponge to alleviate the mechanical stress accompanying the volume change of Sn, which is a critical factor to influence the 137 electrochemical performance of Sn-based anodes (Huang et al. 2015; Kaspar et al. 2014; Ke et al. 2013; Ni et al. 2014). Figure 6.14 (a) Long-term cycling stability of the Sn@C electrodes at a current density of 1/10 C for the initial three cycles and at 1C for the following cycles. The inset is cycling performance of the Sn@C anodes for the initial 100 cycles at 1C. (b-k) Schematic summaries of structural evolution during lithiation/delithiation and TEM images of pitaya- 138 structured (b,e-f), pomegranate-structured (c,g-h), and chestnut-structured (d,i-k) particles in anodes after 100 cycles in rate performance test. (l) Rate capability of Sn@C pomegranate anode in SIB. Electrochemical experiments in (a) and (l) are conducted by collaborator (Huajun Tian). The long-term cycling stability of three Sn@C anodes was evaluated at 1C after activation for the initial 3 cycles at 1/10C (inset of Figure 6.14a). The capacity of Sn@C chestnut anode endured the most significant capacity decay at the initial 100 cycles. In contrast, the Sn@C pitaya and pomegranate anodes retained highly stable electrochemical performance after 100 charge/discharge cycles. Similarly, the Sn@C pitaya also exhibited good cycling stability due to its nanostructured Sn and elastic carbon framework, which can support the large volume change during lithiation process, after capacity decay in first 5 cycles. Long-term cycling performance (Figure 6.14a) of the Sn@C electrodes were evaluated at a high rate of 1C and 2C. The capacity retention of Sn@C pomegranate anode can be maintained at 627.9 mAhg-1 even after 1500 cycles, corresponding to 88.5% capacity retention at 2C, which is one of the best performance in Sn@C anode reported to date (Detailed comparison See Table 6.4 and Figure 6.15). The Sn@C pitaya anode also displayed a stable cycling performance, which could be maintained at 400 mAhg-1 after 500 cycles, about 72.8% of initial capacity at 1C. Contrarily, for Sn@C chestnut anode, the capacity dropped rapidly after only 200 cycles to less than 200 mAhg-1. Therefore, the electrochemical performance of Sn@C in LIBs could be effectively controlled by CSP. 139 Figure 6.15 Comparison of electrochemical performance of our Sn@C anodes with references in Table 6.4. Table 6.4. Comparison of electrochemical performance of our Sn@C anodes with references Par Initial Highest Properties Sn/C composite ticle Coulombic capacitances Ref. -1 degradation size efficiency obtained (mAh g ) ~30 88.5% capacity 952mAh/g @0.2Ag- Sn/C composite 0 74.3% 1 retention over 1500 This work nm cycles Mic 786.4 mAhg- ~96.3% capacity 1 -1 (Huang et al. Sn/C composite ro ~56% @0.06Ag retention over 1000 2015) size (0.01~2.5V) cycles Sn quantum ~21 1176 mAh/g 75% capacity (Im et al. dots in N-doped 0 71% @0.1Ag-1 retention over 200 2013) CNFs nm (0.01~3V) cycles 12 µm 769 mAh/g (Liao and TiO2-Sn/C core- 459 mAh/g after 16 ± ~60% @0.0335Ag-1 -1 Manthiram shell cycles at 335 mAg 0.5 (0.01~3V) 2014) µm ˃10 1029 ~100% capacity nano-Sn/C (Xu et al. 0 69% mAh/g@0.2Ag-1 retention over 130 composite 2013b) nm (0.02~3V) cycles ~40 ˃ 800 mAh/g Sn in Carbon ˃ 550 mAh/g over (Zhang et al. 0 <70% @0.2Ag-1 Spheres 100 cycles 2008) nm (0.005~3V) 100 865.3 porous Sn@C ~ (Zhang et al. 67.3% mAh/g@0.2Ag-1 N.A nanocomposite 300 2015) (0.01~3V) nm 140 Sn in N-doped ~50 ~1490mAh/g@~0.2 630 mAh/g after 400 (Youn et al. 48% Carbon µm Ag-1 (0.01~3V) cycles 2016) ~40 0 nm yolk-shell (ave 979mAh/g@0.05Ag ˃ 500 mAh/g after (Ni et al. nanostructures rage ~70% -1(0.01~2V) 10 cycles 2014) Sn/C dia met er) ~21 Nano-Sn in 0 650 mAh/g @ ~ 600 mAh/g after (Yan et al. expanded ~55% nm( 0.13Ag-1 (0.01~2V) 350 cycles 2016) graphite Sn) 40 nm in dia met 850 mAh/g Sn-carbon er @0.06Ag-1 612 mAh/g after 10 (Liu et al. nanotube and ~75% (0.01~2.5V) cycles 2014) nanocapsules 200- 300 nm in leng th 200 nm (a 1795 mAh/g @ 650 mAh/g after 100 (Li et al. Sn/graphene ~65% side 0.1Ag-1 (0.005~3V) cycles 2014) leng th) 8 nm Sn@C and 1007.1 mAh/g @ 410 mAh/g after (Zhang et al. 63.3% nanocomposites 40n 0.2Ag-1 1000 cycles 2014b) m (Sn) ~5 71% capacity Nano Sn in N- 1014 mAh/g @ (Zhu et al. nm ~75% -1 retention over 200 doped Carbon 0.2Ag 2014b) (Sn) cycles As reported, the volume expansion of Sn could be ~260 % (Huang et al. 2015; Kaspar et al. 2014; Ke et al. 2013; Ni et al. 2014), thus the superior cycling and rate performance of Sn@C pomegranate anodes are attributed to the excellent structural stability, originated from the well distributed Sn beads within relative small Sn@C particles. To effectively 141 accommodate for the volume change in Sn@C, both diameters of Sn@C particle and interior Sn bead should be small, while the Sn@C particle size is more crucial in cycling stability (Figure 6.14b-d). As demonstrated in Figure 6.6, the pomegranate-structured particles have smaller Sn@C particle size but larger interior Sn size (301 nm and 24 nm, respectively) than pitaya-structured particles (340 nm and 16 nm, respectively). The smaller pomegranate-structured particle in Sn@C pomegranate anode ensures better cycling stability than Sn@C pitaya anode. Although chestnut-structured particles have the smallest Sn@C particle size (139 nm), the significantly large Sn core (close to 139 nm) can seriously damage the Sn@C composite particles (Figure 6.14d), resulting in quick capacity decay. Therefore, the specifically designed structure of Sn@C pomegranate particles has sufficient and well-defined internal space to tolerate the expansion of Sn beads in the elastic carbon frames without deteriorating the structural integrity of the anode (Figure 6.14c). To test this hypothesis, the particle structures of Sn@C anodes were characterized by TEM (Figure 6.14) and SEM (Figure 6.16) after 100 cycles in rate performance tests. The Sn@C pitaya and pomegranate anodes maintained stable structures with spherical shape of Sn@C particles, which are similar to those before the electrochemical test (Figure 6.14e-f and g-h, respectively). However, cracking or fracture in the Sn@C chestnut anode was observed as shown in Figure 6.14i-k. 142 Figure 6.16. Post-cycling characterization of Sn@C pitaya (a), pomegranate (b), and chestnut (c) anodes. To further validate our theory, the Sn@C intermediate anode was tested for electrochemical performance. Since Sn@C intermediate particles are comprised of structures between pomegranate-structured and chestnut-structured particles, Sn@C intermediate anode in theory should have a structural and cycling stability between Sn@C pomegranate and chestnut anodes, which was verified experimentally in the electrochemical performance testing (Figure 6.14a) and the structure characterization of the anode after the electrochemical test (Figure 6.17). 143 Figure 6.17. Post-cycling characterization of Sn@C intermediate anode. Although the volume expansion of Sn-based anodes in SIBs is 420% (Zhu et al. 2013), the Sn@C pomegranate anodes for SIBs present excellent electrochemical performance, as shown in Figure 5.14l. Sn@C pomegranate anode in SIBs was galvanostatically discharged 144 and charged at different current densities from 0.05C to 5C. The cell could deliver 487.3, 378.0, 293, 237.6, 197.2, 120.4 mAhg-1 at 0.1C, 0.2C, 0.5C, 1C, 2C and 5C, respectively. In particular, the capacity could be recovered to its original value when the rates were set back to 0.1 C, 0.2C and even 1C after 100 cycles, respectively. This clearly substantiates that the sodiation reaction in Sn@C pomegranate anode is highly reversible. 6.5 Summary of this chapter We have successfully developed a general strategy to synthesize multicomponent particles by CSP. This process is fundamentally different from the conventional aerosol processes where only highly soluble salts, pure organic solvents, and extreme reducing conditions are favored. By applying CSP, a wider range of functional materials can be produced at low cost, with safer process conditions, and facile operating procedures. A specific example of the application of CSP was presented to generate Sn@C particles for LIBs and SIBs anodes, which is difficult to obtain by conventional methods. Although this work only utilized Sn as the example, CSP is applicable to other elements. Furthermore, the model we developed provides a better understanding of the particle formation mechanism in the aerosol processes. Meanwhile, this model can also be utilized to prescreen the colloids that can be processed by CSP and design the process variables to obtain desired products. With the guidance of theoretical analysis of the colloid-reduction kinetics and aerosol dynamics, Sn@C particles with controllable interior structures (pitaya, pomegranate and chestnut-structured) were produced through complete reduction of the SnO2 colloids by in-situ solid-state reaction. The Sn@C pomegranate anodes displayed excellent battery performance with 88.5% (1500 cycles; 2C) capacity retention for LIBs 145 and exhibited superior rate capability for SIBs, which are among the best performance for Sn-based anodes reported to date. 146 Chapter 7: Fabrication of nanocomposite particles containing SnO2 colloids and edge-oxidized graphene oxide sheets as photocatalysts 7.1 Abstract Artificial photosynthesis requires the usage of catalysts to cleave the C=O bond of CO2 and convert it into valuable hydrocarbon products. However, this technology is strongly restricted by two issues: insufficient catalytic efficiency and complicated processes of catalyst fabrication. Herein, a novel photocatalyst engineering process, spray drying, has been developed to address these two issues. Through one-step spray drying with a residence time of 1.5 s, nanocomposites containing tin oxide (SnO2) nanoparticles and edge-oxidized graphene oxide (eo-GO) sheets were fabricated without post-treatment. After 4 h irradiation of simulated sunlight, the nanocomposites exhibited 28-fold and 5- fold enhancements in photocatalytic efficiency in CO2 reduction compared to SnO2 and commercialized TiO2 (P25), respectively. Our approach is scalable with short residence time and simple equipment required, which can promote the practical application of artificial photosynthesis by realizing mass-production of efficient photocatalysts. 7.2 Introduction Advancement of society has greatly relied on the consumption of fossil fuels, which results in the accumulation of greenhouse gases such as carbon dioxide (CO2) and the depletion of non-renewable fossil resources. A promising strategy to address these concerns simultaneously is artificial photosynthesis, in which CO2 is converted into hydrocarbons with the assistance of biomimetic catalysts. In consequence of the large dissociation energy of C=O bond (750 kJ/mol) (Chang et al. 2016), advanced catalysts are 147 required to enhance the efficiency in artificial photosynthesis. Unfortunately, photocatalyst engineering is greatly restricted by the complexity in catalyst development and fabrication. It is even more challenging to develop high-performance photocatalysts that can be produced via a simple, scalable, and commercially applicable approach. Titania (TiO2) has been intensively studied as a catalyst for artificial photosynthesis since the pioneering work of Inoue et al. (Inoue et al. 1979). However, it exhibits low catalytic efficiency that can be ascribed to three factors: 1) the low CO2 adsorption on the surface, 2) rapid recombination of the photo-generated electrons and holes, and 3) catalyst deactivation due to adsorption of less reactive intermediates (Chang et al. 2016; Habisreutinger et al. 2013). Various strategies have been attempted to resolves these issues including modification of TiO2 and development of alternative photocatalysts. For example, TiO2 with particle sizes in the nanometer regime shows enhanced CO2 adsorption due to increased surface area (Fang et al. 2014). The inclusion of noble metals as the electron sink in TiO2 retards the recombination between electrons and holes (Jiao et al. 2017). Alternative photocatalysts with narrow band gap and better separation of the charges, such as CeO2 (Li et al. 2016b; Ye et al. 2017), C3N4 (Li et al. 2015a; Li et al. 2016b), BiVO4 (Mao et al. 2012; Murcia-López et al. 2017), and graphene oxide (GO) (Hsu et al. 2013; Lin et al. 2017; Shown et al. 2014) can outperform TiO2 in artificial photosynthesis applications. Among the alternative photocatalysts that have been developed to date, GO is an efficient hole-collector with semiconductor-like band gap, large specific surface area, and negatively charged O atoms to adsorb and activate CO2 in artificial photosynthesis (Hsu et al. 2013; Lin et al. 2017; Shown et al. 2014). However, the wide band gap and the 148 interruptions in π bond in the aromatic rings can deteriorate light absorption and electron transfer capability of GO (Hsu et al. 2013; Zerjav et al. 2017). On the other hand, an electron trapper together with GO is favored to further retard the recombination of electrons and holes. SnO2 is a better electron acceptor than TiO2, because of its more positive conduction band (CB) edge. Moreover, SnO2 is an effective photocatalyst to degrade dye molecules (Liu et al. 2007b). Nevertheless, the wide bandgap of SnO2 (3.8 eV) (Inoue et al. 1979) is an obstacle to promote its application in artificial photosynthesis. The combination of GO and SiO2 to develop a composite may result in a more efficient catalyst that can outperform each of the individual components in CO2 reduction in artificial photosynthesis. The synthesis of photocatalysts often involves multiple synthesis steps or sophisticate equipment (see Section 2.3 of Chapter 2). This has constrained the scale-up of the photocatalyst synthesis for widespread applications. For example, hydrothermal process has been often employed for preparation of a variety of photocatalysts, but unfortunately, it requires long processing time and sophisticated post-treatment steps, and it raises safety concerns because of the autogenous pressure in the reactor (Hsieh et al. 2014; Ide et al. 2016; Sun et al. 2015). Approaches such as molecular beam epitaxy (MBE), sol-gel, and ion-exchange, etc. have also been developed (AlOtaibi et al. 2015; Li et al. 2015a; Tahir and Amin 2015). Nevertheless, MBE requires significant equipment investment, while sol- gel or ion-exchange approaches have sophisticated operating procedures and long processing times. A simple, short processing time, low cost and scalable approach is desired to promote the practical application of artificial photosynthesis. 149 Herein, a novel photocatalyst engineering process has been developed to produce efficient SnO2-GO nanocomposites as catalysts for artificial photosynthesis. Spray drying, a continuous and scalable process, is applied to fabricate nanocomposites that comprise of edge-oxidized graphene oxide (eo-GO) sheets and nano-sized SnO2 colloids, as shown in Figure 7.1a. The obtained powders are termed TinGOx, where x equals to mass ratio of SnO2 to eo-GO in the precursor mixture. The eo-GO sheets have large intact sp 2 carbon network to deliver large surface area, excellent electron transfer ability, sufficient active sites of negatively charged O atoms, and narrow band gap. The utilization of nano-sized SnO2 can efficiently trap the photogenerated electron and create a defect state to narrow its wide bandgap. State-of-art techniques have been utilized to reveal the configuration of nanocomposites, including electron energy loss spectroscopy (EELS) in low-loss and core- level regions. A detailed investigation of the photoreduction of CO2 has been conducted to understand the photocatalytic activity, optimal SnO2/eo-GO mass ratio, and reaction mechanism. The best TinGOx photocatalyst exhibits superior performance in photoreducing CO2 into CH4 with high yield and selectivity compared to bare eo-GO, SnO2, and the commercial TiO2 (P25). 150 Figure 7.1. Particles synthesized by spray drying. (a) Schematic illustration of spray drying process. (b) XRD patterns of TinGOx powders obtained through spray drying. (c) Raman spectra of TinGOx powders. 7.3 Experimental 7.3.1 Materials The SnO2 colloid with particle diameters of 10-15 nm was received from NYACOL (SN15). The counter ion is K+. The specific gravity (SG) is 1.15. The 4-10% edge-oxidized eo-GO powder was purchased from Sigma-Aldrich (product number: 796034, molecular weight = ~4200 g/mol, bulk density = ~1.8 g/cm3). The molecular structure of eo-GO is schematically shown in Figure 7.2. It is noted that the eo-GO sheets are dispersible in water or other polar solvents. Ethanol (ethyl alcohol, 99.98 %) was bought from Pharmco-Aaper. 151 Figure 7.2. Molecular structure of 4-10% edge-oxidized graphene oxide sheet received from Sigma-Aldrich. 7.3.2 Photocatalyst fabrication by spray drying A 150-mL-aqueous precursor containing SnO2 colloids, eo-GO sheets, and ethanol were utilized to fabricate TinGOx phototcatalysts, where x is the mass ratio of SnO2 to eo- GO in the precursor mixture. The concentration of SnO2 colloids can be calculated as: 𝑣∙𝑆𝐺∙𝑤∙𝜌𝐻 𝑂 𝑛 2𝑆𝑛𝑂 = ∙ 10 −3 [𝑚𝑜𝑙] (7.1) 2 𝑀𝑤 𝑛𝑆𝑛𝑂 𝐶 2𝑆𝑛𝑂 = ∙ 1000 [𝑚𝑜𝑙/𝐿] (7.2) 2 𝑉 𝑣∙𝑆𝐺∙𝑤∙𝜌𝐻2𝑂𝑁 21𝑐𝑜𝑙𝑙𝑜𝑖𝑑,𝑆𝑛𝑂 = 1 ∙ 10 (7.3) 2 𝜋∙𝑑30 ∙𝜌6 𝑆𝑛𝑂2 𝛿𝑁𝑐𝑜𝑙𝑙𝑜𝑖𝑑,𝑆𝑛𝑂 𝑁2 𝑐𝑜𝑙𝑙𝑜𝑖𝑑,𝑆𝑛𝑂= 2 ∙ 1000 [#/𝐿] (7.4) 𝛿𝑉 𝑉 where v [ml] is the volume of extracted SnO2 colloid raw solution, specific gravity (SG) = 1.15 is the specific gravity of SnO2 colloid solution at 25 oC, w (=0.15) is the mass fraction of SnO2 in its colloid solution, Mw [g/mol] is the molecular weight of SnO2, V =150 ml is 152 the volume of precursor, d0 [nm] is particle size or particle diameter (=15 nm), ρH2O [kg/m 3] is the density of bulk H 32O (1.0×10 ), and ρ 3 SnO2 [kg/m ] is the density of bulk SnO2 (6.85×103), respectively. Based on Equations (7.1-7.4), the concentration of SnO2 colloids and eo-GO in each experimental condition are summarized in Table 7.1. Table 7.1. Composition of precursor solutions for fabrication of SnO2/eo-GO photocatalysts. Product particle TinGO10 TinGO14 TinGO17 TinGO21 TinGO34 SnO2 [SnO2] (kg/L) 5.2×10 -4 6.9×10-4 8.6×10-4 1.0×10-3 1.7×10-3 1.7×10-3 [SnO2] (mol/L) 0.022 0.030 0.037 0.044 0.072 0.072 SnO 16 16 162 colloid quantity 4.3×10 5.7×10 7.1×10 8.6×10 16 1.4×1017 1.4×1017 (#) [SnO2 colloids] (#/L) 2.8×10 17 3.7×1017 4.6×1017 5.5×1017 8.9×1017 8.9×1017 eo-GO (kg/L) 3.3×10-4 3.2×10-4 3.2×10-4 3.2×10-4 3.1×10-4 0 SnO2/GO (mass ratio) 10 14 17 21 34 ∞ [Ethanol] (mol/L) 6.7 6.7 6.6 6.6 6.4 0 During the spray drying, the precursors were atomized by a 1.7 MHz transducer into droplets with a volume mean diameter of 5 μm (Zhong et al. 2012). The droplets were transported by the N2 gas (99.5%, Airgas) into two furnaces connected in series, as shown in Figure 1a of main text. The furnaces were set at 500 oC and the product particles were collected on a PTFE filter. The residence time was controlled at ~1.5 s, which was calculated by using a plug-flow model, consisting 32 sub-units of identical length. Detailed calculation procedures are reported in Chapter 3. Ethanol was used to ease the oxidation of eo-GO sheets during the spray drying. 153 7.3.3 Materials characterization Scanning electron microscopy (SEM, HITACHI SU-70) and transmission electron microscopy (TEM, JEOL 2100F field emission gun) with in Gatan image filter (GIF, Tridiem 863) were utilized to understand the particle morphology, structure, and chemical distributions. X-ray diffractometer (XRD, Bruker D8 advance) was used to determine the crystallographic information of the TinGOx powders. X-ray photoelectron spectroscopy (XPS, Kratos Axis 165) was performed to understand the valence band and chemical status of the elements. Raman signals were collected using a Horiba Jobin Yvon LabRAM Raman microscope with a 532 nm laser. N2 adsorption-desorption isotherms of the samples were measured using an Autosorb-iQ analyzer (Quantachrome Instruments) at -196 oC. The samples were outgassed at 250 oC for 8 h and 1 mm Hg prior to measurements. Brunauer- Emmett-Teller (BET) method was used to determine the specific surface area of the sample. Barret-Hoyner-Halenda (BJH) method was utilized to analyze the pore size distribution in the particles. The UV-Vis-NIR spectrophotometer (Lambda 1050, PerkinElmer) equipped with an integrating sphere (Labsphere Model No. 150MM RSA ASSY) and an InGaAs detector was used for the measurement of the diffuse reflectance spectra (DRS) on the fabricated photocatalyst samples. 7.3.4 Photoreduction of CO2 The photoreduction of CO2 was performed in a cylindrical batch reactor under ambient condition. The batch reactor was custom-built with a stainless steel body and two quartz windows of 1.4 cm in diameter and 0.5 cm thick. The inner cavity of the batch reactor is 3.0 cm in diameter and 2.0 cm in depth. A xenon (Xe) arc lamp (450 W) was placed on the top of the reactor. The batch reactor is connected by two gas lines for inlet and outlet gases 154 of the reactor on the perimeter wall of the cylindrical reactor body. The outlet gas goes through an automated gas valve and a gas-chromatograph (Agilent Technologies, 6890N) in series for gaseous composition determination. The gas-chromatograph is equipped with a 3.0 m×0.0254 m packed column (Agilent HAYESEP DB) and a 30 m×0.25 mm capillary column (Supelco SP-2330), which are connected to a thermal conductivity detector (TCD), a flame ionization detector (FID), respectively. Prior to each reaction, 20 mg catalyst and 0.05 mL water were manually loaded into the batch reactor. The reactor was then purged with a mixture of CO2 (99.99%, Airgas, Inc.) and Argon (99.99%, Airgas, Inc.) at 42 mL/min and 9 mL/min, respectively, for 50 min. Then the batch reactor was enclosed by turning off gas valves next to the reactor body and the Xe lamp was turned on. After 4 h irradiation, the Xe lamp was turned off. CO2 gas was introduced at 3 mL/min to the rector to flush the product gases into the gas- chromatograph for composition analysis. 7.4 Results and discussions The nanocomposites were fabricated by a one-step and one-pot process, i.e., spray drying. As illustrated in Figure 7.1a, aqueous precursor containing SnO2 colloids (particle diameter: 10-15 nm) and eo-GO sheets was firstly ultrasonically atomized into droplets. Then the droplets were transported by the N2 gas into the furnaces, in which solvent evaporation and structural progression lead to the formation of product particles. At the end of this continuous process, nanocomposite particles were collected on a polytetrafluoroethylene (PTFE) filter and can be immediately used as photocatalysts without any post-production treatments. As SnO2 colloids are uniformly and stably 155 dispersed in the aqueous precursor and eo-GO is soluble in polarized solvents, the product particles have a uniform particle-to-particle composition and structure, as discussed in Chapter 3. By tuning the mass ratio of SnO2 to eo-GO in the precursor (Table 7.1), TinGOx particles have been prepared over a range of 10 to 34. Compared with reported photocatalyst fabrication processes, including hydrothermal, MBE, sol-gel, and ion- exchange methods, our process is scalable and continuous with short processing time, facile equipment requirement, and simple operating procedures (detailed comparison is presented in Section 2.3 of Chapter 2). All these merits can promote future photocatalyst manufacturing at an industrial-scale. In the TinGOx powders, SnO2 is a main component, verified by XRD characterization (Figure 7.1b). The broad SnO2 peaks indicate small particle sizes in the nanocomposites based on the Scherrer equation. Additionally, the short residence time, low processing temperature, and existence of eo-GO all inhibit the significant coagulation between SnO2 colloids during spray drying. The existence of eo-GO in the nanocomposites is validated by the Raman spectroscopic study with three distinct peaks from the eo-GO in the spectrum (Figure 7.1c). The D band at ~1345 cm-1 represents the disorder in carbon materials while G band at ~1585 cm-1 is indicative of sp2 graphitic carbon lattice. Another peak at ~2700 cm-1 is the second disorder-induced Raman signature, or 2D band (Eigler et al. 2012). The ratio of D band to G band peak areas (D/G ratio) is an indicator of the defect density (Eigler et al. 2012). Since the D/G ratios in Figure 7.1c (including pure eo-GO) are all less than 1, the oxidized (or disorder) extent of eo-GO is less than GO (Eigler et al. 2012). The eo-GO is 4-10% edge oxidized, as shown in Figure 7.2, compared with GO (Figure 7.3c). The Raman signals from TinGO17 powder exhibit the highest D/G ratio among all the TinGOx 156 nanocomposite particles. Together with a red shift of D band and a blue shift of G band in Raman spectrum of TinGO17 (Figure 7.1c and Table 7.2), the carbon materials in TinGO17 particles have the highest defect density. This could be attributed to the modification of planer structured eo-GO by spherical structured SnO2 colloids. Similar phenomena have also been observed by Shown et al in the system of Cu nanoparticles decorated GO sheets (Shown et al. 2014). The highest defect density of eo-GO in TinGO17 can promote the collection of photogenerated holes (Yang et al. 2017), increasing the efficiency in CO2 reduction. Figure 7.3. Image of GO (a) and EELS core loss spectra of graphene (black) and GO (red) (b). (c) is the molecular structure of the GO sheet. The red groups in (c) highlight the difference from eo-GO sheet, as shown in Figure 7.2. 157 Table 7.2. D and G band positions and D/G ratio of Raman spectra D band (cm-1) G band (cm-1) D/G eo-GO 1337.27 1564.35 0.4 TinGO10 1336.33 1564.63 0.4 TinGO17 1341.85 1569.46 0.6 TinGO21 1337.03 1567.39 0.5 TinGO34 1339.09 1565.32 0.3 Figure 7.4. Typical morphologies and structures of SnO2/eo-GO powders. (a-b) TEM image of a TinGO17 particle (a) with corresponding HRTEM image (b). (c-f) HAADF 158 image (c) together with C (d), O (e), and Sn (f) elemental mappings. (g) SEM image of the TinGO17 particles with a statistical summary of the particle size distribution. The number mean diameter, standard deviation, and sample size are presented. (h) Box plots of size distributions of TinGOx powders. (i) EELS spectra of C K-edge (left) and Sn M4,5-edge (right) in core-loss level region. (j) Low-loss EELS spectrum of TinGO17 particles. A representative TEM image of TinGOx (e.g. TinGO17 sample) is shown in Figure 7.4a with particle-like morphology. An enlarged high resolution TEM (HRTEM) image of the red squared area in Figure 7.4a is shown in Figure 7.4b. Distinct lattice fringes indexed to the [200] and [111] orientations of SnO2 can be observed through the incident e-beam direction of [01-1], substantiating that SnO2 is one constituent component. SnO2 and eo- GO distribute uniformly over the product particles, validated by the EELS elemental mapping of C K-edge, O K-edge, and Sn M4,5-edge of TinGO17 particles (Figure 7.4d-f, respectively). During the spray drying process, eo-GO sheets in the vicinity of SnO2 colloids preferentially adsorb on the colloid surface, because the oxygenated groups in eo- GO have high affinity to the SnO2 colloids through the hydrogen bonding and electrostatic attraction (Song et al. 2011). The full coverage of SnO2 colloids by eo-GO sheets has been estimated on the basis of simple calculation by considering an extreme scenario (TinGO34 particles), in which eo-GO relative amount is the lowest in the TinGOx particles. Based on Figures 7.1,7.4, we assume the configuration of TinGOx particles are eo-GO covered on the surface of SnO2 colloids. In one extreme condition, the SnO2 colloids are assumed to be isolated without coagulation in the product particles to deliver the largest surface area of SnO2 colloids. 159 The atomized droplets have a mean volume diameter of 5 μm (Zhong et al. 2013b). Therefore, based on the one-droplet-one particle model, the TinGO34 powders contain many SnO2 colloids in one single particle, which is N=5.8×10 4. By assuming the thickness of eo-GO layer on SnO2 particle surface is t (nm), the mass of eo-GO in one single TinGO34 particle (mGO,req) is required to be: 4 𝑑 3 𝑑 𝑚𝐺𝑂,𝑟𝑒𝑞 = 𝑁 × 𝜋 [( 0 + 𝑡) − ( 0)3] 𝜌 −24 3 2 2 𝐺𝑂 × 10 [𝑘𝑔] (7.5) where ρGO is the density of eo-GO, which is 1.8 g/cm 3. The mass of eo-GO in each product particle is 2.0×10-17 kg, which equals mGO,req. Thus, t is 132 nm, significantly thicker than the thickness of single sheet of eo-GO. As other TinGOx particles have less SnO2 colloids but same amount of eo-GO as TinGO34 powders, our theoretical analysis indicates the SnO2 colloids are covered by eo-GO in the product powders. This implies excellent affiliation could be realized between the SnO2 and eo-GO components. SEM observation of the TinGOx powders further verifies their spherical structure, as shown in Figure 7.4g for TinGO17 and Figure 7.5 for all other TinGOx samples. The number mean diameter of TinGO17 particles is 325 nm with a standard deviation of 172 nm, statistically summarized from 294 particles, as depicted in the inset of Figure 7.4g. Same statistical analysis has also been conducted on other TinGOx powders, as shown in Figure 7.5. Apparently, the increasing SnO2 amount in the precursor solution leads to an increased size of product particles (Figure 7.4h), which is consistent with reported results in other aerosol process (Nandiyanto and Okuyama 2011). 160 Figure 7.5. SEM images and size distributions of particles fabricated from different precursor solutions. SEM images of particles produced from precursor solutions with SnO2/eo-GO ratios (mass) are 10 (a), 34 (c), and SnO2 only (e). Particle size-distributions when SnO2/eo-GO ratios (mass) are 10 (b), 34 (d), and SnO2 only (f) in precursor solutions. Statistical summaries include number means, standard deviations, and sample sizes. To better understand the chemical bonding, electronic excitations between valence band (VB) and CB, EELS was performed on TinGO17 particles, and results are shown in 161 Figure 7.4i-j. The EELS spectra of C K-edge and O K-edge with background subtracted are shown in Figure 7.4i. The spectral peaks at ~ 285 eV and ~ 300 eV in C K-edge spectrum represent the sp2 (π* bond) and the sp3 (σ* bond) in carbon, respectively, verifying the existence of eo-GO (Mkhoyan et al. 2009). The shoulder spectral features (~515 eV and 524 eV) and spectral peak (~532 eV) correspond to Sn M4,5-edge and O K- edge of SnO2, respectively (Huang et al. 2010). The spectral peak at ~538 eV is O K-edge ascribed to both eo-GO and SnO2 (Huang et al. 2010; Mkhoyan et al. 2009). For comparison, the core-level EELS spectra of graphite, graphene, and GO are depicted in Figures 7.3,7.6. Only in the C K-edge spectrum of graphite, there are three spectral peaks between 300 to 310 eV (Figure 7.6). In addition, the spectral peak at ~538 eV can be found in the O K-edge spectra of eo-GO and GO (Figure 7.3i,7.4) instead of graphene (Figures 7.3,7.6). In the low-loss EELS spectrum (Figure 7.4j), the spectral peaks at 13 eV, 20 eV and 32 eV are the surface plasmon, volume plasmon, and Sn M4,5-edge of SnO2 (Huang et al. 2010). For the eo-GO, the π+σ plasmon peak occurs at ~19 eV, similar to the low-loss spectrum of GO (Johari and Shenoy 2011; Mkhoyan et al. 2009). Weak peaks at ~5 eV and 6 eV are observable possible because the scattering vector q is parallel to the basal plane of eo-GO (q||a) (Johari and Shenoy 2011). In addition, low-loss EELS observation indicates the absorption edge of TinGO17 is ~3.7 eV, substantiating the semiconductor- like property of eo-GO. 162 Figure 7.6. EELS core loss spectra of graphite (blue) and monolayer graphene (black). The wide-range XPS measurements were performed to further understand the electronic structures of SnO2, TinGO17, and eo-GO. As shown in Figure 7.7a-b, C and Sn are the main components of the TinGO17 powders, which are stemmed from eo-GO and SnO2, respectively. The C 1s spectrum of TinGO17 displays three features of eo-GO at 284.4, 286.0, 288.4 eV attributed to the sp2 C (C-C/C=C) in the aromatic rings, sp3 C, and O-C=O, respectively. For reference, the XPS C 1s spectrum of eo-GO is also presented, as shown in the bottom of Figure 7.7a. It can be deconvoluted into five peaks at 284.4, 286.1, 287.3, 288.5, and 290.5 eV, corresponding to the sp2 C, C-O, C=O, O-C=O, and π-π plasmonic, respectively. This is consistent with the molecular structure of eo-GO (Figure 7.2). The pronounced sp2 C peak in C 1s spectrum of eo-GO supports that the defect density of eo-GO sheets is low, in agreement with D/G ratio in Raman observation. However, when eo-GO sheets couple with SnO2 colloids to form TinGO17, the small radii of SnO2 colloids (10-15 nm) may bend the aromatic ring of eo-GO sheets, resulting in more discorded 163 structures, which is consistent with our Raman observation of TinGO17 and the reported phenomena in GO-nanoparticle system (Shown et al. 2014). Therefore, the intensities of O-C=O peak and sp3 C increase significantly in C 1s spectrum of TinGO17 compared with eo-GO. The K 2p peaks are observed in both TinGO17 and SnO2 spectra (Figures 7.7a,7.8, respectively), substantiating that K+ is the counter-ion in the received SnO2 colloid dispersion solution. In Figure 7.7b, two peaks are displayed in the Sn 3d XPS spectrum of TinGO17 particles at 486.2 and 494.7 eV, corresponding to the 3d5/2 and 3d3/2 levels, respectively. The area ratio of 3d5/2 peak to 3d3/2 peak (~1.5) and their peak positions all validate the existence of SnO2 in TinGO17 particles according to our XPS characterization on SnO2 (bottom of Figure 7.7b) and reported Sn 3d XPS spectrum of SnO2 (Song et al. 2011). The VB maxima of SnO2, eo-GO, and TinGO17 were also investigated by XPS and determined by extrapolating the peaks to baseline, as shown by inset plots in Figure 7.7c. SnO2 has the band edge position of 3.2 eV. The band edge position of the eo-GO is 1.5 eV, which is significantly greater than the graphene (theoretical VB maxima=0 eV) but smaller than the reduced GO (3.05 eV) (Zerjav et al. 2017). When the SnO2 colloids couple with the eo-GO sheets, the VB maxima of TinGO17 exhibits a binding energy of 2.7 eV. 164 Figure 7.7. (a) C 1s and K 2p XPS spectra of TinGO17 and eo-GO. (b) Sn 3d XPS spectra of TinGO17 and SnO2. Valence band spectra (c) and Tauc plots (d) of SnO2, TinGO17, and eo-GO photocatalysts. (e) Schematic illustration of photocatalytic reactions and reduction potentials of possible CO2 reduction pathways. (f) CH4 yields over TinGOx photocatalysts under irradiation of a Xe lamp for 4 h. 165 Figure 7.8. C 1s and K 2p XPS spectra of SnO2. Figure 7.9. Diffuse reflectance spectra (DRS) of eo-GO, TinGO10, TinGO14, TinGO17, TinGO21, TinGO34, and SnO2. To better understand the band structures of the photocatalysts, the band gaps of SnO2, TinGO17, and eo-GO are calculated using Tauc equations. The band gaps are determined in the Tauc plot of linear extrapolation of the edge to the x-axis (Figure 7.7d). The Tauc 166 plot is obtained from analyzing the DRS spectrum (Figure 7.9) through Kubelka–Munk function followed by the procedures provided in these references.(Ehsan and He 2015; Kumar et al. 2016) 2 1−𝑅(𝛼ℎ𝜈) = [( )ℎ𝜈]2 (7.6) 2𝑅 𝑐 ℎ𝜈 = ℎ (7.7) 𝜆 where, R is the relative reflectance obtained from DRS (Figure 7.9), h is the Planck constant, which is 6.63×10-34 m2∙kg/s, c is the speed of light (=3×108 m/s), and λ is the corresponding wavelength to detect R in DRS. SnO2 displays an Eg of 3.69 – 3.87 eV, which is similar to the results reported by Inoue et al.(Inoue et al. 1979). The Eg of eo-GO is 2.34 – 3.19 eV, narrower than those of GO sheets (Hsu et al. 2013). The TinGO17 photocatalysts have an absorption edge of 3.40 – 3.70 eV, between those of eo-GO and SnO2 and consistent with the low-loss EELS result. A summary of the band structures of these photocatalysts is presented in Figure 7.7e and Table 7.3. Moreover, the absorption edges of TinGOx were calculated by converting the DRS spectra (Figure 7.9) into Tauc plots, as shown in Figure 7.10. The increasing SnO2 concentration in TinGOx increases the absorption edge. All the results obtained from EELS, DRS, and XPS validate that TinGOx particles are promising photocatalyst with suitable electronic band structure and materials construction. Table 7.3. Band structures of eo-GO, TinGO17, and SnO2 as photocatalysts Band gap energies (eV) VB (eV) CB (eV) eo-GO 2.34 – 3.19 1.5 -0.84 – -1.69 TinGO17 3.40 – 3.70 2.7 -0.70 – -1.00 SnO2 3.69 – 3.87 3.2 -0.49 – -0.67 167 Figure 7.10 Absorption edges of TinGOx photocatalysts, eo-GO, and SnO2 determined by Tauc plots. Table 7.4. Detailed yields and selectivities of SnO2/eo-GO photocatalysts Yield (nmol∙h-1∙(g catalyst)-1) Selectivity Formic Methane Formaldehyde Methanol Ethanol (%) acid eo-GO 0.88 0.0065 0.04 0 0.22 76 TinGO10 10 0.073 0.068 0.028 0.32 95 TinGO14 17 0.065 0.058 0.018 0.36 97 TinGO17 85 0.25 0.038 0.04 0.53 99 TinGO21 28 0.045 0.058 0.010 0.33 98 TinGO34 14 0 0 0 0 100 SnO2 3 0.023 0.065 0 0.55 82 TiO2 17 0 0 0 0 100 (P25) The photocatalysts were irradiated by Xe lamp for 4 h. 168 Photocatalysts The photocatalytic activities of the TinGOx, eo-GO sheets, and SnO2 in reducing gaseous CO2 were investigated under irradiation of a xenon (Xe) lamp for 4 h. The main product from this reaction was CH4. Trace amounts of CHO, CH3OH, C2H5OH, and HCOOH were also identified, as shown in Table 7.4. The yield of CH4 is remarkably enhanced over TinGOx than either eo-GO or SnO2 (Figure 7.7f). After 4 h reaction, the yield and selectivity of CH4 over eo-GO photocatalyst are 3.5 nmol/(g catalyst) and 76%, respectively. The low photocatalytic conversion of CO2 into CH4 by eo-GO is possibly caused by the poor light absorption in the wavelength range of 222-316 nm, corresponding to photons with energies of 5.6-3.9 eV, as shown in Figure 7.9. The yield and selectivity of CH4 are slightly increased to 57 nmol/(g catalyst) and 82%, respectively over SnO2. However, the wide Eg deteriorates its efficiency in CO2 photoreduction. Among the TinGOx photocatalysts, TinGO17 achieves the highest CH4 yield (mean and maximum values of three experiments: 260 and 340 nmol/(g catalyst), respectively, Figure 7.7f), which is 5 times that of P25, 97 times that of eo-GO, and 28 times that of SnO2 under the same experimental condition. Moreover, the TinGO14 and TinGO21 also depict higher CH4 yields compared with P25, although less than that of TinGO17. 169 Figure 7.11. N2 adsorption-desorption isotherm of TinGO10 (a) and TinGO17 (c) particles. The surface areas are 106 m2/g (TinGO10) and 77 m2/g (TinGO17) based on slit/cylindrical NLDFT model. (b) and (d) Pore size distributions derived from N2 desorption branches of (a) and (c) according to the BJH method. The average pore sizes are 3.46 nm (TinGO10) and 3.27 nm (TinGO17), implying the average distance between the SnO2 colloids. By our particle engineering, the photogenerated holes and electrons could migrate independently and spatially separated, because eo-GO and SnO2 are efficient collectors for holes and electrons, respectively. The strong affinities of SnO2 to electrons and eo-GO to holes could further alleviate the recombination of electrons and holes. The utilization of eo-GO can also enhance the adsorption of CO2 molecules on the photocatalysts, since GO is reported to have large surface area (Lin et al. 2017). Experimentally, both TinGO10 and TinGO17 photocatalyst have high surface areas, supported by the BET results (Figure 170 7.11). TinGO10 powders display larger surface area and inner pore size than TinGO17 due to the higher relative eo-GO amount. Moreover, the negatively charged O atoms in eo-GO can perform as the active sites during the catalysis (Yang et al. 2017), facilitating the CO2 reduction. The higher VB maxima of TinGO17 than TiO2 (=-0.5 V) implies a higher activity of the photogenerated holes in oxidation, because the greater overpotential between the VB maxima and the CO2 reduction potentials is believed can enhance the photocatalytic efficiency (Habisreutinger et al. 2013). Therefore, the reductions mainly happen on SnO2 with lower reduction potential and oxidation will proceed on eo-GO with lower oxidation potential (Low et al. 2017). This evidence validates that the coupling between SnO2 colloids and eo-GO sheets is the key factor to improve the catalytic efficiency and explain the highest CH4 yield and selectivity achieved over TinGO17 photocatalyst. Theoretically, the CO2 reduction over TinGOx photocatalysts can be completed in a single step by involving multiple electrons, since the lower edge of CB of TinGO17 is still higher than the reduction potentials of possible reduction products according to (Hong et al. 2013), as shown in Figure 7.7e. Unfortunately, limited evidence has been reported to validate that mechanism (Habisreutinger et al. 2013). Instead, the reduction of CO2 over TinGOx photocatalysts is believed to follow the formaldehyde pathway by a series of one- electron steps (Habisreutinger et al. 2013; Koci et al. 2010). This is supported by the various reduction products (HCOOH, HCHO, CH3OH, and CH4) in our experiments, as presented in Table 7.4. In addition, the geometries of CO2 molecule adsorption on the photocatalyst surface can also influence the chemical pathways (Tran et al. 2012). The anchoring of the C atom in the CO2 molecule on the photocatalyst surface favors the formation of CO. The produced CO can be the final products or the intermediate for further 171 hydrogenation to various hydrocarbon products, including CH4 (Tran et al. 2012). Based on these findings, the configuration of CO2 molecular adsorption on the TinGOx photocatalyst surface is believed also through its C atom. 7.5 Summary of this chapter Through fabrication of TinGOx nanocomposites by a spray drying process, innovations have been made to the artificial photosynthesis regarding both photocatalyst development and fabrication. The TinGOx photocatalysts exhibit a suitable electronic band structure and materials construction with SnO2 colloids and eo-GO sheets uniformly distributed over the nanocomposites. The utilization of nano-sized SnO2 colloids and eo-GO guarantees the TinGOx photocatalysts can efficiently photogenerate and separate the electrons and holes. The high surface area attributed to the eo-GO also promotes the adsorption of CO2 gases. Among the TinGOx photocatalysts, TinGO17 displays the best catalytic performance. It delivers a CH4 yield of 85 nmol∙h -1∙(g catalyst)-1 under irradiation of full-spectrum light for 4 h, which is five times that of P25 and 28 times that of SnO2 under same reaction condition. Our catalyst engineering for artificial photosynthesis has successfully increased the efficiencies in both catalysis reaction and catalyst fabrication, which is scalable for the future mass-production. 172 Chapter 8: Conclusions and future work 8.1 Conclusions Three different aerosol-assisted processes (spray pyrolysis, colloidal spray pyrolysis, and spray drying) were developed in this dissertation to fabricate functional particles as conductive pastes in solar cell metallization and interference packaging, anode and cathode materials in energy storage devices, and photocatalysts in energy conversion. A spray pyrolysis process was developed to fabricate Cu-Sn particles from aqueous precursor solutions. The particle configuration is a Sn-enriched surface layer on the Cu-Sn solid solution core. In the precursor solution, EG was used as the co-solvent, because of its higher flash point (111 oC) than conventional reducing agents (formic acid: 69 oC and ethanol: 13 oC) (Yaws 1999; Zhong et al. 2013a; Zhong et al. 2012). This addresses the safety concerns during the potential scale-up manufacturing in the future. The formation mechanism of Cu-Sn binary particles was also studied. Substances in the precursor were first hydrolyzed into hydroxides. Then the hydroxides decomposed into metal oxides. Subsequently, the metal oxides were reduced to metals, and a solid solution formed. For the structural evolution, precipitation of the solutes occurred during solvent evaporation of the precursor droplets. Thus, hollow structures were also generated. Finally, the shrinking of the hollow particles continued until solid particles were formed. Therefore, high reaction temperature (in our experiments: 1000 oC) promoted the formation of solid dense particles of Cu-Sn particles. Low reaction temperature, 500 oC in our experiments, led to the formation of hollow porous particles of Cu-Sn. Further decreasing the reaction temperature (to 300 oC in our experiments), resulted in generating intermediate products. The residence time also exhibited significant influence on the product morphology. With low residence 173 time (0.92 s, 750 oC), fragmented particles or irregular morphologies were observed in the products. With long residence time (4.5 s, 750 oC), spherical particles with smooth surfaces were obtained. With intermediate residence time (1.5 s, 750 oC), small particles became distinct on the surface of product particles. After I understood the formation mechanism of Cu-Sn particles in spray pyrolysis and successfully obtained solid dense particles, I investigated their oxidation resistance. Micron-sized solid particles are favored by industry for applications such as conductive pastes and interference packaging (Deshpande et al. 2005; Wu et al. 2009). Therefore, Cu- Sn solid particles fabricated at 1000 oC were used. The Sn-enriched surface layer guarantees the Cu-Sn particles exhibit better oxidation resistance than Cu particles, as confirmed by XRD with Rietveld refinement analysis, in situ XRD, and oxidation kinetics studies. Among the CuSny particles, CuSn0.1 powders displayed the highest oxidation resistance with best spherical morphology among the CuSny powders. Therefore, CuSn0.1 particles are promising alternative materials to replace noble metal-based conductive powders, such as Ag and Au, in the fields of printed electronics, interference packaging, and solar cells metallization. Moreover, based on theoretical analysis of the oxidation, the migration of O2- through the oxide layer is a possible rate-limiting step when CuSny powders are oxidized at 300 oC. For oxidation at 500 oC, grain growth of the oxide likely controls the oxidation process. Next, investigation was performed on the resistivities of structures fabricated using oxidation-resistant CuSny particles to evaluate the prospect of applying these particles. We successfully fabricated CuSny wires and films via the direct printing of inks containing oxidation-resistant CuSny particles and EG. The CuSn0.1 wires exhibited comparable 174 resistivities with those of Cu wires, and lower than the reported resistivities of wires produced from CucoreAgshell powders. In addition, CuSn0.1 films have significantly less resistivity increases than Cu films when being oxidized at 573 K in ambient air. Therefore, based on the above discussions, CuSn0.1 particles are promising materials to replace Au and Ag in applications, including printed electronics, solar cell metallization, and interference packaging. Colloidal spray pyrolysis has also been developed to addresses issues of conventional spray pyrolysis. For example, conventional spray pyrolysis requires a stable precursor solution to achieve uniform particle-to-particle composition and structure, restricting its usage to highly soluble salts. Moreover, extremely acidic precursors are necessary to ease the uncontrollable hydrolysis of the salts and the subsequent precipitation. By using colloidal spray pyrolysis, all the prerequisites are eliminated simultaneously. A specific example of the application of colloidal spray pyrolysis was presented to generate Sn@C particles for anodes of LIBs and SIBs, which is difficult to obtain by conventional methods. With the guidance of theoretical analysis of the colloid-reduction kinetics and aerosol dynamics, Sn@C particles with controllable interior structures (pitaya, pomegranate and chestnut-structured) were produced. The Sn@C pomegranate anodes displayed excellent battery performance with 88.5% (1500 cycles; 2C) capacity retention for LIBs and superior rate capability for SIBs. Spray drying process was developed to fabricate nanocomposites (TinGOx) comprising of tin oxide (SnO2) nanoparticles and edge-oxidized graphene oxide (eo-GO) sheets as photocatalysts in CO2 reduction. This approach addressed the issues in conventional methods for generating photocatalysts, such as sophisticated experimental 175 procedures, long processing time, and high investment on instruments. TinGO17 displays the best catalytic performance. It delivers a CH4 yield of 85 nmol∙h -1∙(g catalyst)-1 under irradiation of full-spectrum light for 4 h, which is 5 times that of P25 and 28 times that of SnO2 under same reaction condition. State-of-art tools, including EELS, have been utilized to validate the TinGOx photocatalysts have a suitable electronic band structure and particle configuration with SnO2 colloids and eo-GO sheets uniformly distributed over the nanocomposites. 8.2 Future work This dissertation not only develops three specific processes and applies them in fabrication various functional particles, but also investigates the underlying mechanisms, for example, those controlling the CSP process. Based on the experimental and theoretical analyzing results, I will propose several research projects that can be conducted in the future. 8.2.1 Alternative materials in applications, including printed electronics In this dissertation, we investigated the prospect of applying Cu-Sn particles as the alternative materials in applications, including printed electronics. In one research article published by (Chen et al. 2014), Cu wires were coated by second metals (Sn, Zn, and In) by electroplating. All these Cu-based wires exhibited significantly higher oxidation- resistance than Cu wire. I have validated that Cu-Sn particles are promising materials. For the other two metals, they are also worth exploring. 176 Challenges also exist in these two approaches. For example, the reduction of Zn and In ions to metals are more difficult than that of Sn ions, based on the Ellingham diagrams and reported article (Ellingham 1944; Schoeller and Cho 2011). Oxides are expected in the product particles. Moreover, it will be difficult to obtain spherical solid particles because the oxides and metals are not miscible to form solid solutions. Phase segregation may be observed in the product particles, in which Cu will expose directly to air without the protection from the secondary materials (Zn and In). Another obstacle may be expected is during converting the Cu-Zn and Cu-In particles into one- and two-dimensional structures. Since Zn and In ions are difficult to be reduced, their oxides may remain in the product particles. After the one- and two-dimensional structures are printed by inks containing these particles, more reductive condition is required during the sintering step than that to fabricate Cu-Sn structures. Therefore, higher sintering temperature, longer sintering time, and higher H2 concentration are expected to form the metallic Cu-Zn and Cu-In structures, which may bring additional safety concerns. 8.2.2 Further investigation on CSP In Chapter 6, I have mentioned that the inter-colloid collision and colloid reduction happen simultaneously during the CSP. I have explored the particle formation at 750 oC. However, when the reaction temperature further increased to 1000 oC, the inter-colloid coagulation is expected to be completed faster than colloid reduction, as shown in Figure 8.1. One major obstacle in this process is the difficulty in completely reducing colloids. As the inter-colloid coagulation is expected to be completed faster than colloid reduction, the 177 radius of actual colloids need to be reduced is significantly larger than that of the input colloids. Therefore, oxides or reaction intermediates may remain in the product particles. For anode materials, the metals and alloy are more favorable than the oxides due to higher capacities (Goriparti et al. 2014). On the other hand, the fabricated particles are expected to have core-shell structures, since the faster inter-colloid coagulation has high possibility of resulting in a core covered by another component. The core-shell structure may be promising to be applied as battery materials because it can tolerate large volume expansion. Figure 8.1 Schematic illustration of colloidal spray pyrolysis at high temperature. Another possible project that worth investigating is applying CSP to produce other battery materials listed in Figure 2.1 and Table 2.1. Based on the procedures presented in Chapter 6 and in Appendices, the possible materials that can be obtained from aqueous 178 solutions containing their corresponding colloids can be determined. This can further expand the application of CSP. One difficulty in this approach is to find the values of variables in the equations shown in Chapter 6 and Appendices. For example, the kr in Equation (A.2.19) is approximated by the equation to calculate the kr of CuO reduction. In this dissertation, this approximation is valid because our experimental results are consistent with the theoretical analysis. Nevertheless, for other metal oxide colloids, the approximation may not be valid. In the worst case, the values of other variables may have not be reported and are difficult to be determined by experiments. For example, Equations (6.6-6.7) contain µ, which is the viscosity of the precursor solution. In this dissertation, it is measured by using a viscometer with a capillary tube. Since the diameter of the SnO2 colloids used is 4.4 nm, the viscosity of precursor solution is still measurable by this viscometer. Unfortunately, for other colloids with larger size, clots may be formed in the capillary tube during the viscosity measurement. Then alternative approaches may be required. Nevertheless, the procedures I presented in this dissertation are still applicable to determine the possible metal colloids that can be processed by CSP. 179 Appendices A1. Detailed procedures and equations to understand oxidation of Cu-Sn particles by shrinking-core model (SCM) in Chapter 4 SCM was utilized to simulate the oxidation process and to investigate the chemical kinetics and material transport (Chuang et al. 2010; Levenspiel 1999). The densities of Cu (8.9 g/cm3) and Sn (7.3 g/cm3) are comparable to their corresponding oxides, such as Cu2O (6.0 g/cm3), CuO (6.3 g/cm3), SnO (6.5 g/cm3), and SnO2 (7.0 g/cm 3) (Perry 2011). Therefore, the particle diameter was assumed to be constant during the oxidation process at 300 oC, which is also validated by the SEM images of the post-oxidation particles, as shown in Figure A.1.1a-c. For oxidation at 500 oC, SCM is not applicable because of significant coagulation between particles and the resulting particle size change (Figure A.1.1d-f). The oxidation process can be divided into three steps: (i) migration of O2 gas through the gas film around the particle to the particle surface; (ii) diffusion of O2 gas through the porous oxide ash layer to the inner metal core; (iii) reaction of O2 gas with metal core (Chuang et al. 2010; Levenspiel 1999). If Step (i) controls the process the radius of the unreacted core, rc, can be related to the oxidation time through (Levenspiel 1999): 𝜌𝑝𝑅 𝑟 𝑡 = [1 − ( 𝑐)3] (A.1.1) 3𝑏𝑘𝑔𝐶𝑂2,𝑔 𝑅 𝜌𝑝𝑅 𝜏 = (A.1.2) 3𝑏𝑘𝑔𝐶𝑂2,𝑔 where t is the oxidation time. τ is the time for a complete oxidation. ρp is the density of the particle. R is the radius of the particle, which is assumed to be 355 nm. b is the 180 stoichiometric coefficient in the oxidation. kg is the mass transfer coefficient. CO2,g is the O2 gas concentration in the main stream. Figure A.1.1. SEM images of particles after being oxidized in a packed-bed reactor under 300 oC (a)-(c) and 500 oC (d)-(f). The volume percent of O2 in the feed stream is 1.4 %. The scale bars in SEM images are 2 μm. If Step (ii) controls the process (Levenspiel 1999): 𝜌𝑝𝑅 2 𝑟 𝑡 = [1 − 3( 𝑐)2 𝑟 + 2( 𝑐)3] (A.1.3) 6𝑏𝐷𝑒𝐶𝑂2,𝑔 𝑅 𝑅 𝜌𝑝𝑅 2 𝜏 = (A.1.4) 6𝑏𝐷𝑒𝐶𝑂2,𝑔 181 where De is the effective diffusion coefficient of O2 in the ash layer. If Step (iii) controls the process (Levenspiel 1999): 𝜌𝑝 𝑡 = (𝑅 − 𝑟𝑐) (A.1.5) 𝑏𝑘𝑣𝐶𝑂2,𝑔 𝜌𝑝 𝜏 = 𝑅 (A.1.6) 𝑏𝑘𝑣𝐶𝑂2,𝑔 where kv is the first-order rate constant of Cu oxidation. kg in Equations (A.1.1-A.1.2) can be obtained as follows (Fogler 1999): 1−∅ 𝐷 𝑘𝑔 = ( 𝐴𝐵)Sh (A.1.7) ∅ 𝑑𝑝 𝑆ℎ = 𝑅𝑒1/2𝑆𝑐1/3 (A.1.8) 𝑈𝑑𝑝 𝑅𝑒 = (A.1.9) (1−∅)𝜈 𝑢 𝑈 = 0 (A.1.10) 𝐴𝑐 𝜈 𝑆𝑐 = (A.1.11) 𝐷𝐴𝐵 1 1 𝑇1.75( + )1/2 𝐷 = 10−3 𝑀𝑂2 𝑀𝑁2 𝐴𝐵 1 1 (A.1.12) 𝑃[(𝑉𝑂2)3+(𝑉𝑁2)3]2 All dimensionless numbers, Sh, Re, Sc, and symbols are defined in Tables A.1.1-A.1.2. The diffusivity of O2 gas through the oxide ash layer (De) is first assumed to be 10 -17 m2/s at 300 oC based on the reported O2 gas diffusion coefficient in Cu2O (Li and Mayer 1992). After that, the actual De in our situation will be approximated by fitting to our experimental data. 182 Table A.1.1. Variables used in modeling with their definitions and values Variables Definition Value 8.9×103 kg/m3 ρCu Density of Cu 1.4×105 mol/m3 [(Perry 2011)] 7.3×103 kg/m3 ρSn Density of Cu 6.1×104 mol/m3 [(Perry 2011)] 6.0×103 kg/m3 ρCu2O Density of Cu2O 4.2×104 mol/m3 [(Perry 2011)] R Mean radius of particles 355 nm [(Chapter 3)] Mean diameter of dp 710 nm [(Chapter 3)] particles Φ Porosity 0.4 [(Chuang et al. 2010)] U Superficial velocity 6.5×10-3 m/s u0 inlet gas flow rate 50 ml/min 3.6×10-4 m2/s at 300 oC ν kinematic viscosity 6.1×10-4 m2/s at 500 oC [(Yaws 1999)] 0.65 cm2/s at 300 oC DAB Gas-phase diffusivity 1.1 cm2/s at 300 oC Cross-sectional area of A 1.3×10-4c m 2 pipe MO2 Molecular weight of O2 32 g/mol MN2 Molecular weight of N2 28 g/mol P Pressure 1 atm VO2 Diffusion volume for O2 16.6 [(Cussler 2009)] VN2 Diffusion volume for N2 17.9 [(Cussler 2009)] Table A.1.2. Parameters used in modeling with definitions Parameter Definition Unit t Oxidation time s or min τ Time for complete oxidation s b Stoichiometry of particle oxidation dimensionless Mass transfer coefficient between fluid and kg m/s particle CO2,g O2 gas concentration in the main stream mol/m 3 rc Radius of metal core nm Effective diffusion coefficient of O2 in the ash De m 2/s layer kv The first-order rate constant s -1 Sh Sherwood number dimensionless Re Reynolds number dimensionless Sc Schmidt number dimensionless kapp First order apparent rate coefficient min -1 183 For Equations (A.1.5-A.1.6), as the products of Cu oxidation are Cu2O and CuO, the rate constants representing the processes of converting Cu to Cu2O (kv,1) and Cu2O to CuO (kv,2) are described as (Chuang et al. 2010): 𝑘𝐽 −44±15 𝑘𝑣,1 = 8.2 × 10 6 exp ( 𝑚𝑜𝑙) (𝑠−1)(A.1.13) 𝑅𝑇 𝑘𝐽 −55±15 𝑘 7 𝑚𝑜𝑙𝑣,2 = 1.0 × 10 exp ( ) (𝑠 −1)(A.1.14) 𝑅𝑇 The carrier gas was assumed to be pure N2 gas when calculating the kinematic viscosity (ν), which is 4.8×10-5 m2/s based on the dynamic viscosity of N2 gas at 300 oC and 1.01×105 Pa (Yaws 1999). For Cu particles, the oxidation products include CuO and Cu2O (Figure 4.15 in Chapter 4). As shown in Table A.1.3, Step (ii), internal diffusion of O2 gas through porous oxide ash layer controls the process in both scenarios. For CuSn0.05 particles, the oxidation products are Cu2O and SnO2 (Figure 4.15). Step (ii) is the rate-limiting step. Detailed results are shown in Table A.1.3. For CuSn0.1 particles, the oxidation products are Cu2O and SnO2 (Figure 4.15). Step (ii) controls the process. Detailed results are shown in Table A.1.3. Therefore, the oxidation of Cu, CuSn0.05, and CuSn0.1 particles are all controlled by internal diffusion of O2 gas through the oxide ash layer based on SCM simulation. 184 Table A.1.3. Simulation results of Cu-Sn particle oxidation at 300 oC in a packed-bed reactor by SCM. The diffusion coefficients listed here are obtained by fitting the experimental results (Figure 4.16). The reported diffusion coefficients are 10-17 m2/s for both situations of O2 through Cu2O and O2 through SnO2 and at 300 oC (Cussler 2009; Li and Mayer 1992). Detailed procedures to obtain the De In the SCM simulation, we utilized a diffusion coefficient of O2 gas through Cu2O layer at 300 oC obtained from the literature, 10-17 m2/s at 300 oC (Li and Mayer 1992), which may not represent the internal diffusion of O2 in our system. Thus, the actual diffusion coefficients in Equations (A.1.3-A.1.4) were fitted. To simplify the fitting model, 𝑟 𝑟 𝑥 = [1 − 3( 𝑐)2 + 2( 𝑐)3] (A.1.15) 𝑅 𝑅 Then Equation (A.1.3) becomes: 𝜌 2𝑝𝑅 𝑡 = 𝑥 (A.1.16) 6𝑏𝐷𝑒𝐶𝑂2,𝑔 To obtain the actual De, the results in Figure 4.14 need to be converted from (∆m/m vs. t) to (t vs. x). The rc/R in Equation (A.1.3) can be obtained by assuming the thickness of 185 the oxide layer was uniform on the particle surface (Kim et al. 2004). Therefore, the volume of oxide layer can be calculated as: 4 𝑉 = 𝜋(𝑅3 − 𝑟3𝑐 ) (A.1.17) 3 The relative weight gain caused by oxidation is: 𝑉𝜌𝑝 2𝑀𝑤𝑂 ∆𝑚 ∙ = 𝑀𝑤 𝑏 (A.1.18) 𝑚 𝑉𝜌𝑝 Cu particles: To convert the results in Figure 4.14 (∆m/m vs. t) into t vs. x, equations are also derived based on the two situations when Cu2O and CuO were the oxidation products. If Cu is oxidized to Cu2O, b=4. Then Equation (A.1.18) can be converted into: 4 𝜋(𝑅3−𝑟3)𝜌 3 𝑐 𝐶𝑢 𝑀𝑤 ∆𝑚 ∙ 𝑂 𝑀𝑤𝐶𝑢 2 𝑀𝑤 𝑟= = 𝑂4 [1 − ( 𝑐)3](A.1.19) 𝑚 𝜋𝑅3𝜌𝐶𝑢 2𝑀𝑤 𝑅3 𝐶𝑢 When Cu is assumed to be directly oxidized into CuO, b=2. Then: 4 𝜋(𝑅3−𝑟3 3 𝑐 )𝜌𝐶𝑢 ∆𝑚 𝑀𝑤𝑀𝑤 𝑂 = 𝐶𝑢 𝑀𝑤 𝑟 4 = 𝑂 [1 − ( 𝑐)3](A.1.20) 𝑚 𝜋𝑅3𝜌 3 𝐶𝑢 𝑀𝑤𝐶𝑢 𝑅 The curves representing the t vs. x are presented in Figure A.1.2a-b. In the case of Cu2O as the main product, the results in Figure A.1.2a are not well explained by SCM. In the SCM, Equation (A.1.16) is linear and its slope is a monotonic function of De. If the oxidation product is CuO, the curve is still not linear (Figure A.1.2b). However, De is 2.2×10-13 m2/s, if we fit the non-linear curve to a linear equation. CuSn0.05 particles: 186 Equation (A.1.18) can be transformed into the following equation, because the oxidation products are Cu2O and SnO2 (Figure 4.15): 4 𝜋(𝑅3−𝑟3 3 𝑐 )𝜌𝐶𝑢𝑆𝑛0.05 𝑀𝑤 ∙( 𝑂∆𝑚 +0.1𝑀𝑤 )𝑀𝑤𝐶𝑢𝑆𝑛0.05 2 𝑂 0.6𝑀𝑤 = 𝑂 𝑟 4 = [1 − ( 𝑐)3] (A.1.21) 𝑚 𝜋𝑅3𝜌 3 𝐶𝑢𝑆𝑛0.05 𝑀𝑤𝐶𝑢𝑆𝑛0.05 𝑅 By applying Equations (A.1.16,A.1.21), curves in Figure 4.14 (∆m/m vs. t) can be converted into t vs. x. The result is shown in Figure A.1.2c. For CuSn0.1 particles: Equation (A.1.18) can be transformed into following equation, because the oxidation products are Cu2O and SnO2 (Figure 4.15): 4 𝜋(𝑅3−𝑟3𝑐)𝜌3 𝐶𝑢𝑆𝑛0.1 𝑀𝑤∙( 𝑂∆𝑚 +0.2𝑀𝑤𝑀𝑤 2 𝑂 ) 𝐶𝑢𝑆𝑛0.1 0.7𝑀𝑤 𝑟= 4 = 𝑂 [1 − ( 𝑐)3] (A.1.22) 𝑚 𝜋𝑅3𝜌𝐶𝑢𝑆𝑛0.1 𝑀𝑤3 𝐶𝑢𝑆𝑛0.1 𝑅 By applying Equations (A.1.16,A.1.22), curves in Figure 4.14 (∆m/m vs. t) can be converted into t vs. x. The result is illustrated in Figure A.1.2d. 187 Figure A.1.2. Oxidation time (t) vs x of Cu particles with the assumption that the product of Cu oxidation is Cu2O (a) and CuO (b) at 300 oC. (c)-(d) Plots of oxidation time (t) vs x of CuSn0.05 (c) and CuSn0.1 (d) particles. x is a variable defined by Equation (A.1.15). The internal diffusivity of O2 through oxide ash layer (De) is obtained by fitting the curves in (a-d) by Equation (A.1.16). The oxidation is conducted in O2-He flow with 1.4 vol % O2. The diffusion coefficients obtained are 103 – 104 higher than reported values for the diffusion coefficients, which are 10-17 m2/s for both situations of O2 through Cu2O and O2 through SnO2 and at 300 oC (Hernandez-Ramirez et al. 2008; Li and Mayer 1992). 188 A2. Modeling on the colloidal spray pyrolysis process in Chapter 6 A2.1 General equations for CSP modeling In CSP, the oxide colloids were reduced by reducing gases as discussed in the main text. Here, we calculated the time scale of colloid reduction, as explained in Formula 6.2 (Chapter 6). The oxide-colloid reduction by H2 can be modeled by two senarios (Levenspiel 1999), as shown in the schematic drawing in Figures 6.1b,6.2d. i) Diffusion of gaseous H2 though a layer of metal ash towards the interface of oxide/metal. ii) Reaction of the gaseous H2 with oxide at the interface. For a general approach, we analyzed the colloid reduction process in two situations, where the rate-limiting step is i) the diffusion of H2 through the metal ash and ii) chemical reaction between the H2 and oxide core. In addition, the process was assumed to be pseudo- steady state and the colloid radius was assumed to remain constant during the process (discussed in Chapter 6). i) Diffusion of gaseous H2 though the metal ash layer controls Based on the model presented in Figures 6.1b,6.2d, we can build a material balance with respect to H2: 𝑑𝑁 − 𝐻2 = 4𝜋𝑟2𝑄 = 4𝜋𝑟2𝑄 = 𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡 [𝑚𝑜𝑙 ∙ 𝑠−1𝐻2 𝑐 𝑐,𝐻2 ] (A.2.1) 𝑑𝑡 Here, QH2 represents the flux of H2 through surface of any radius r. Qc,H2 is the flux of H2 to the reaction surface. 𝑑𝐶 𝑄 𝐻2𝐻2 = 𝐷𝑒 [𝑚𝑜𝑙 ∙ 𝑚 −2 ∙ 𝑠−1] (A.2.2) 𝑑𝑟 189 𝑑𝑁 − 𝐻2 2 𝑑𝐶 = 4𝜋𝑟 𝐷 𝐻2𝑒 = 𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡 (A.2.3) 𝑑𝑡 𝑑𝑟 To integrate Equation A.2.3, we can use the following boundary conditions: r = R CH2 = Cs,H2 r = rc CH2 = 0 R is the radius of the colloid. Cs,H2 is the H2 concentration on the colloid surface. Then we can have: 𝑑𝑁𝐻2 1 1− ( − ) = 4𝜋𝐷𝑒𝐶𝑠,𝐻2 [𝑚𝑜𝑙 ∙ 𝑚 −1 ∙ 𝑠−1] (A.2.4) 𝑑𝑡 𝑟𝑐 𝑅 From stoichiometry, Equation A.2.4 can be extended as: 4 3 𝑑𝑁 𝑑𝑁 𝑑( 𝜋𝑟 ) −𝑥 𝐻2 = − 𝑀𝑂𝑥 𝑑𝑉 3 𝑐 2 𝑑𝑟= −𝜌 = −𝜌 𝑐 𝑑𝑡 𝑑𝑡 𝑀𝑂𝑥 𝑑𝑡 𝑀𝑂𝑥 = −4𝜋𝜌𝑀𝑂𝑥𝑟𝑐 (A.2.5) 𝑑𝑡 𝑑𝑡 Therefore, 1 1 𝑑𝑟 −𝜌 𝑟2 ( − ) 𝑐 = 𝑥𝐷 −1 −1𝑀𝑂𝑥 𝑐 𝑒𝐶𝑠,𝐻2 [𝑚𝑜𝑙 ∙ 𝑚 ∙ 𝑠 ] (A.2.6) 𝑟𝑐 𝑅 𝑑𝑡 Integrating Equation A.2.6 with the following boundary conditions, rc = R t = 0 rc = rc t = t we obtain: 𝜌 2 2 𝑡 = 𝑀𝑂𝑥 𝑅 𝑟 [1 − 3 ( 𝑐 𝑟 ) + 2( 𝑐)3] (A.2.7) 6𝑥𝐷𝑒𝐶𝑠,𝐻2 𝑅 𝑅 The time scale for colloid reduction can be obtained by setting rc = 0 in Equation A.2.7, 𝜌 𝑅2 𝜏 = 𝑀𝑂𝑥 (A.2.8) 6𝑥𝐷𝑒𝐶𝑠,𝐻2 ii) Reaction controls 190 When colloid reduction is controlled by chemical reactions, the material balance can be analyzed at the interface between oxide/metal. Unlike the case in which diffusion is the rate-controlling step, the interface between the core (oxide) and ash (metal) is not stationary. The rate expression for oxide reduction by H2 with respect to H2 is kr and the reaction order is n. Therefore, by building the material balance of H2 on the interface and considering the stoichiometry, we have: 𝑥 𝑑𝑁𝐻2 1 𝑑𝑁− 2 = − 𝑀𝑂𝑥 2 = 𝑥𝑘𝑟𝐶 𝑛 𝑠,𝐻2 (A.2.9) 4𝜋𝑟𝑐 𝑑𝑡 4𝜋𝑟𝑐 𝑑𝑡 In addition, based on Equation A.2.5, we can turn Equation A.2.9 into: 1 − 4𝜋𝜌 2 𝑑𝑟𝑐 𝑑𝑟𝑐 𝑛 2 𝑀𝑂𝑥𝑟𝑐 = −𝜌𝑀𝑂𝑥 = 𝑥𝑘𝑟𝐶𝑠,𝐻2 (A.2.10) 4𝜋𝑟𝑐 𝑑𝑡 𝑑𝑡 Integrating the above equation with the following boundary conditions, rc = R t = 0 rc = rc t = t we obtain: 𝜌 𝑡 = 𝑀𝑂𝑥𝑛 (𝑅 − 𝑟𝑐) (A.2.11) 𝑥𝑘𝑟𝐶𝑠,𝐻2 Setting rc = 0, the time scale for reduction is 𝜌𝑀𝑂𝑥𝑅𝜏 = (A.2.12) 𝑥𝑘 𝑛𝑟𝐶𝑠,𝐻2 A2.2 Detailed procedures to calculate the time of SnO2 colloid reduction For the reduction of SnO2 colloid, the chemical formula is: 2𝐻2 + 𝑆𝑛𝑂2 ⇌ 𝑆𝑛 + 2𝐻2𝑂 (A.2.13) 191 The volume average droplet diameter produced by the atomizer is 5 μm (Zhong et al. 2013b). To calculate the H2 concentration on the colloid surface (Cs,H2), the SnO2 colloid loadings for each condition are listed in Table 6.1. Every mole of EG will produce 0.81 mol H2 under reaction conditions similar to this work, based upon experimental results (Zhong et al. 2013b). In addition, each mole of sucrose produces 7.48 mol H2 (Cho et al. 2013). With this information, the Cs,H2 was calculated and shown in Table A.2.1. Table A.2.1. Parameters used in the modeling in Section A.2 Sn@C particle type Pitaya Pomegranate chestnut Average number of SnO2 colloids per droplet 7.4×10 5 1.2×106 2.5×106 Average quantity of EG account for each colloid 4.2×10- 2.5×10-19 1.3×10-19 (mol) 19 Average quantity of sucrose account for each 1.8×10- 1.1×10-20 2.7×10-22 colloid (mol) 20 Cs,H2 (mol/L) 5.4 5.4 4.0 i) Diffusion of gaseous H2 though the Sn ash layer controls Equations A.2.5-A.2.8 can be extended as: 4 3 𝑑𝑁 𝑑𝑁 𝑑( 𝜋𝑟 ) −2 𝐻2 = − 𝑆𝑛𝑂2 𝑑𝑉 3 𝑐 2 𝑑𝑟= −𝜌 = −𝜌 = −4𝜋𝜌 𝑟 𝑐 (A.2.14) 𝑑𝑡 𝑑𝑡 𝑠𝑛𝑂2 𝑑𝑡 𝑠𝑛𝑂2 𝑑𝑡 𝑠𝑛𝑂2 𝑐 𝑑𝑡 2 1 1 𝑑𝑟−𝜌𝑠𝑛𝑂2𝑟𝑐 ( − ) 𝑐 = 2𝐷𝑒𝐶𝑠,𝐻2 [𝑚𝑜𝑙 ∙ 𝑚 −1 ∙ 𝑠−1] (A.2.15) 𝑟𝑐 𝑅 𝑑𝑡 𝜌𝑆𝑛𝑂2𝑅 2 𝑟 2 𝑟 𝑡 = [1 − 3 ( 𝑐) + 2( 𝑐)3] (A.2.16) 12𝐷𝑒𝐶𝑠,𝐻2 𝑅 𝑅 𝜌 2 𝜏 = 𝑆𝑛𝑂2 𝑅 (A.2.17) 12𝐷𝑒𝐶𝑠,𝐻2 Here, ρSnO2 is 4.6×10 4 mol/m3 (Perry 2011). At our experimental conditions, the set point of the furnaces was 750 oC. Thus, it is rational to assume that Sn was liquid, since the metaling point of bulk Sn is ~230 oC (Wronski 1967). The melting point of bulk SnO2 192 is believed to be 1630 oC (Sun et al. 2004). Some researchers even reported that it could be up to 2100 oC (Galazka et al. 2014). Therefore, De represents the H2 diffusivity in liquid tin. Unfortunately, the De of H2 diffusion through Sn is hard to obtain, because of the large uncertainty in measuring the solubility of H2 in liquid Sn (Sacris and Parlee 1970). Thus, we used the Stokes-Einstein equation to approximate the De: 𝑘𝑇 𝐷𝑒 = (A.2.18) 6𝜋𝑟𝐻2𝜇𝑆𝑛 Here, k is the Boltzmann constant (=1.38×10-23 J∙K-1). rH2 is the radius of H2 molecule (2.05×10-10 m) (Schlapbach and Zuttel 2001). μSn is the viscosity of liquid Sn at 750 oC, which is 0.83 mPa∙s (Assael et al. 2010). By Equation A.2.18, De is 6.6×10-9 m2∙s-1 at 750 oC. Based on Equation A.2.17, when the diffusion of H2 through the Sn ash controls the process, the time scales of SnO2 colloid reduction in generating Sn@C particles are 7.8×10 - 10 s (pitaya-structured), 7.8×10-10 s (pomegranate-structured), and 1.1×10-9 s (chestnut- structured). ii) Reaction controls As it is difficult to find the exact rate expression for SnO2 reduction by H2, we used the kinetic equation representing CuO reduction by H2 as an alternative approach. The reaction order is 0.6 and the rate constant expression with respect to H2 is (García-Labiano et al. 2004): −33000 𝑘𝑟 = 1.0 × 10 −4 exp ( ) [𝑚𝑜𝑙0.4𝑚−0.2𝑠−1] (A.2.19) 8.314×𝑇 Therefore, Equations A.2.9-A.2.12 become: 193 2 𝑑𝑁 − 𝐻2 1 𝑑𝑁𝑆𝑛𝑂2 2 = − = 2𝑘 0.6 4𝜋𝑟 𝑑𝑡 4𝜋𝑟2 𝑑𝑡 𝑟 𝐶𝑠,𝐻2 (A.2.20) 𝑐 𝑐 1 2 𝑑𝑟𝑐 𝑑𝑟− 2 4𝜋𝜌 𝑐 0.6 4𝜋𝑟 𝑠𝑛𝑂2 𝑟𝑐 = −𝜌𝑠𝑛𝑂2 = 2𝑘𝑟𝐶𝑠,𝐻2 (A.2.21) 𝑐 𝑑𝑡 𝑑𝑡 𝜌 𝑡 = 𝑠𝑛𝑂20.6 (𝑅 − 𝑟𝑐) (A.2.22) 2𝑘𝑟𝐶𝑠,𝐻2 𝜌 𝜏 = 𝑠𝑛𝑂2 𝑅 0.6 (A.2.23) 2𝑘𝑟𝐶𝑠,𝐻2 Thus, when chemical reaction controls the SnO2 colloid reduction, reduction times are 0.14 s (pitaya-structured), 0.14 s (pomegranate-structured), and 0.17 s (chestnut-structured) by using Equation A.2.23. These data are all much larger than the calculation results obtained based on the assumption that the rate-controlling step is the diffusion. Therefore, chemical reaction controls colloid reduction in our situations. A3. Detailed calculation procedures to obtain the colloid concentration in Chapters 6. 𝑣∙𝑆𝐺∙𝑤∙𝜌𝐻 𝑂 𝑛𝑆𝑛𝑂 = 2 ∙ 10−3 [𝑚𝑜𝑙] (A.2.27) 2 𝑀𝑤 𝑛𝑆𝑛𝑂 𝐶 = 2𝑆𝑛𝑂 ∙ 1000 [𝑚𝑜𝑙/𝐿] (A.2.28) 2 𝑉 𝑣∙𝑆𝐺∙𝑤∙𝜌𝐻 𝑁 2 𝑂 21 𝑐𝑜𝑙𝑙𝑜𝑖𝑑,𝑆𝑛𝑂 = 2 1 ∙ 10 (A.2.29) 𝜋∙𝑑3∙𝜌 6 𝑆𝑛𝑂2 𝛿𝑁𝑐𝑜𝑙𝑙𝑜𝑖𝑑,𝑆𝑛𝑂2 𝑁𝑐𝑜𝑙𝑙𝑜𝑖𝑑,𝑆𝑛𝑂= 2 ∙ 1000 [#/𝐿] (A.2.30) 𝛿𝑉 𝑉 v [ml] is the volume of extracted SnO2 colloid raw solution. SG is the specific gravity of SnO2 colloid solution (SN15ES) from Nyacol TM at 25 oC, which is 1.15. w is the mass fraction of SnO TM2 colloid solution (SN15ES) from Nyacol , which is 0.15. Mw [g/mol] is molecular weight of SnO2. V [ml] is the volume of precursor. d [nm] is particle size or 194 particle diameter. ρ 3H2O [kg/m ] is the density of bulk H2O. ρ 3 SnO2 [kg/m ] is the density of bulk SnO2. 195 References Abhinav K, V., Rao R, V. K., Karthik, P. S., Singh, S. P. (2015). Copper conductive inks: synthesis and utilization in flexible electronics. RSC Advances 5:63985-64030. Addison, C. C. and Hathaway, B. J. (1958). 628. The vapour pressure of anhydrous copper nitrate, and its molecular weight in the vapour state. Journal of the Chemical Society (Resumed):3099-3106. AlOtaibi, B., Fan, S., Wang, D., Ye, J., Mi, Z. (2015). Wafer-Level Artificial Photosynthesis for CO2 Reduction into CH4 and CO Using GaN Nanowires. ACS Catalysis 5:5342-5348. Andresen, A. F. (1958). THE EFFECT OF DISSOLVED TIN ON THE LATTICE PARAMETER OF COPPER. Transactions of the American Institute of Mining and Metallurgical Engineers 212:259-260. Assael, M. J., Kalyva, A. E., Antoniadis, K. D., Michael Banish, R., Egry, I., Wu, J., Kaschnitz, E., Wakeham, W. A. (2010). Reference Data for the Density and Viscosity of Liquid Copper and Liquid Tin. Journal of Physical and Chemical Reference Data 39:033105. Avvaru, B., Patil, M. N., Gogate, P. R., Pandit, A. B. (2006). Ultrasonic atomization: Effect of liquid phase properties. Ultrasonics 44:146-158. Bahadur, J., Sen, D., Mazumder, S., Paul, B., Khan, A., Ghosh, G. (2010). Evaporation- induced self assembly of nanoparticles in non-buckling regime: Volume fraction dependent packing. Journal of Colloid and Interface Science 351:357-364. Bang, J. H. and Suslick, K. S. (2010). Applications of Ultrasound to the Synthesis of Nanostructured Materials. Advanced Materials 22:1039-1059. Bloomberg (2017). Precious and Industrial Metals, Bloomberg Markets, Bloomberg Markets. Boissiere, C., Grosso, D., Chaumonnot, A., Nicole, L., Sanchez, C. (2011). Aerosol Route to Functional Nanostructured Inorganic and Hybrid Porous Materials. Advanced Materials 23:599-623. 196 Booth, M. H., Brandon, J. K., Brizard, R. Y., Chieh, C., Pearson, W. B. (1977). GAMMA-BRASSES WITH F CELLS. Acta Crystallographica Section B-Structural Science 33:30-36. Cabrera, N. and Mott, N. F. (1949). Theory of the oxidation of metals. Reports on Progress in Physics 12:163-184. Cao, Q., Kim, H.-s., Pimparkar, N., Kulkarni, J. P., Wang, C., Shim, M., Roy, K., Alam, M. A., Rogers, J. A. (2008). Medium-Scale Carbon Nanotube Thin-Film Integrated Circuits on Flexible Plastic Substrates. Nature 454:495-500. Cao, W., Li, W., Yin, R., Zhou, W. (2014). Controlled fabrication of Cu–Sn core–shell nanoparticles via displacement reaction. Colloids and Surfaces A: Physicochemical and Engineering Aspects 453:37-43. Chang, X., Wang, T., Gong, J. (2016). CO2 photo-reduction: insights into CO2 activation and reaction on surfaces of photocatalysts. Energy & Environmental Science 9:2177- 2196. Chatain, D., Ghetta, V., Wynblatt, P. (2004). Equilibrium Shape of Copper Crystals Grown on Sapphire. Interface Science 12:7-18. Chatterjee, S. and Gupta, M. (1975). Lattice Parameters of Some Binary and Ternary Copper Alloys. Journal of Applied Crystallography 8:492-493. Chen, Q. and Sieradzki, K. (2013). Spontaneous evolution of bicontinuous nanostructures in dealloyed Li-based systems. Nat Mater 12:1102-1106. Chen, Z., Ye, S., Stewart, I. E., Wiley, B. J. (2014). Copper Nanowire Networks with Transparent Oxide Shells That Prevent Oxidation without Reducing Transmittance. ACS Nano 8:9673-9679. Cho, E. C., Camargo, P. H. C., Xia, Y. (2010). Synthesis and Characterization of Noble- Metal Nanostructures Containing Gold Nanorods in the Center. Advanced Materials 22:744-748. Cho, Y. K., Jung, K. Y., Lee, H., Heo, S. (2013). Synthesis and characterization of C/Cu core–shell particles by hydrogen-free spray pyrolysis assisted with citric acid and sucrose. Materials Research Bulletin 48:3424-3430. 197 Choi, S. H., Ko, Y. N., Lee, J. K., Kang, Y. C. (2015). 3D MoS2-Graphene Microspheres Consisting of Multiple Nanospheres with Superior Sodium Ion Storage Properties. Adv Funct Mater 25:1780-1788. Chuang, S. Y., Dennis, J. S., Hayhurst, A. N., Scott, S. A. (2010). Kinetics of the Oxidation of a Co-precipitated Mixture of Cu and Al2O3 by O2 for Chemical-Looping Combustion. Energy & Fuels 24:3917-3927. Cohen, R. L. and West, K. W. (1972). Solution Chemistry and Colloid Formation in the Tin Chloride Sensitizing Process. Journal of The Electrochemical Society 119:433-438. Comiskey, B., Albert, J. D., Yoshizawa, H., Jacobson, J. (1998). An electrophoretic ink for all-printed reflective electronic displays. Nature 394:253-255. Cussler, E. L. (2009). Diffusion: Mass Transfer in Fluid Systems. Cambridge university press. Dai, Z. R., Pan, Z. W., Wang, Z. L. (2002). Growth and Structure Evolution of Novel Tin Oxide Diskettes. Journal of the American Chemical Society 124:8673-8680. Damour, T. M., Ehrman, S. H., Karlsson, M. N. A., Karlsson, L. S., Deppert, K. (2005). Experimental Evidence for Nonuniform Flow in a Horizontal Evaporation/Condensation Aerosol Generator. Aerosol Science and Technology 39:444-451. de Andres, P. L., Ramírez, R., Vergés, J. A. (2008). Strong covalent bonding between two graphene layers. Physical Review B 77:045403. Deen, W. M. (1998). Analysis of Transport Phenomena (Topics in Chemical Engineering). Oxford University Press, New York. Degang, C., Yoshifumi, O., Hirotaka, H., Hiroaki, S., Toshio, O., Seiichi, T., Hiroshi, K. (1998). Surface Protrusions of Chemical Vapor Deposited TiN Films Caused by Cu Contamination of Silicon Substrates. Japanese Journal of Applied Physics 37:L607. Derrien, G., Hassoun, J., Panero, S., Scrosati, B. (2007). Nanostructured Sn–C Composite as an Advanced Anode Material in High-Performance Lithium-Ion Batteries. Advanced Materials 19:2336-2340. 198 Deshpande, V., Kshirsagar, A., Rane, S., Seth, T., Phatak, G. J., Mulik, U. P., Amalnerkar, D. P. (2005). Properties of lead-free conductive thick films of co- precipitated silver–palladium powders. Materials Chemistry and Physics 93:320-324. Dhas, N. A., Raj, C. P., Gedanken, A. (1998). Synthesis, Characterization, and Properties of Metallic Copper Nanoparticles. Chemistry of Materials 10:1446-1452. Duan, Y. (2008). Electronic properties and stabilities of bulk and low-index surfaces of SnO in comparison with SnO2: A first-principles density functional approach with an empirical correction of van der Waals interactions. Physical Review B 77:045332. Ehsan, M. F. and He, T. (2015). In situ synthesis of ZnO/ZnTe common cation heterostructure and its visible-light photocatalytic reduction of CO2 into CH4. Applied Catalysis B: Environmental 166–167:345-352. Eigler, S., Dotzer, C., Hirsch, A. (2012). Visualization of defect densities in reduced graphene oxide. Carbon 50:3666-3673. Ellingham, H. J. T. (1944). Transactions and Communications. Journal of the Society of Chemical Industry 63:125-160. Eom, K., Jung, J., Lee, J. T., Lair, V., Joshi, T., Lee, S. W., Lin, Z., Fuller, T. F. (2015). Improved stability of nano-Sn electrode with high-quality nano-SEI formation for lithium ion battery. Nano Energy 12:314-321. Eslamian, M. and Ashgriz, N. (2006). Effect of precursor, ambient pressure, and temperature on the morphology, crystallinity, and decomposition of powders prepared by spray pyrolysis and drying. Powder Technology 167:149-159. Faddoul, R., Reverdy-Bruas, N., Blayo, A. (2012). Formulation and screen printing of water based conductive flake silver pastes onto green ceramic tapes for electronic applications. Materials Science and Engineering: B 177:1053-1066. Fang, B., Bonakdarpour, A., Reilly, K., Xing, Y., Taghipour, F., Wilkinson, D. P. (2014). Large-Scale Synthesis of TiO2 Microspheres with Hierarchical Nanostructure for Highly Efficient Photodriven Reduction of CO2 to CH4. ACS Applied Materials & Interfaces 6:15488-15498. Feng, Z., Marks, C. R., Barkatt, A. (2003). Oxidation-Rate Excursions During the Oxidation of Copper in Gaseous Environments at Moderate Temperatures. Oxidation of Metals 60:393-408. 199 Fievet, F., Fievet-Vincent, F., Lagier, J.-P., Dumont, B., Figlarz, M. (1993). Controlled nucleation and growth of micrometre-size copper particles prepared by the polyol process. Journal of Materials Chemistry 3:627-632. Fleisch, T. H. and Mains, G. J. (1982). Reduction of copper oxides by UV radiation and atomic hydrogen studied by XPS. Applications of Surface Science 10:51-62. Fogler, H. S. (1999). Elements of chemical reaction engineering. Prentice-Hall International, Inc. Fotou, G. P., Kodas, T. T., Anderson, B. (2000). Coating Titania Aerosol Particles with ZrO2, Al2O3/ZrO2 and SiO2/ZrO2 in a Gas-Phase Process. Aerosol Science and Technology 33:557-571. Friedlander, S. K. (2000). Smoke, Dust, and Haze. Oxford university press New York. Fujimoto, T., Terauchi, S.-y., Umehara, H., Kojima, I., Henderson, W. (2001). Sonochemical Preparation of Single-Dispersion Metal Nanoparticles from Metal Salts. Chemistry of Materials 13:1057-1060. Fujita, K., Ando, D., Uchikoshi, M., Mimura, K., Isshiki, M. (2013). New model for low- temperature oxidation of copper single crystal. Applied Surface Science 276:347-358. Galazka, Z., Uecker, R., Klimm, D., Irmscher, K., Pietsch, M., Schewski, R., Albrecht, M., Kwasniewski, A., Ganschow, S., Schulz, D., Guguschev, C., Bertram, R., Bickermann, M., Fornari, R. (2014). Growth, characterization, and properties of bulk SnO2 single crystals. physica status solidi (a) 211:66-73. Gale, W. F. and Totemeier, T. C. (2003). Smithells Metals Reference Book. Butterworth- Heinemann. García-Labiano, F., de Diego, L. F., Adánez, J., Abad, A., Gayán, P. (2004). Reduction and Oxidation Kinetics of a Copper-Based Oxygen Carrier Prepared by Impregnation for Chemical-Looping Combustion. Industrial & Engineering Chemistry Research 43:8168- 8177. Glicksman, H. D., Kodas, T. T., Majumdar, D. (1996). Method for making copper (I) oxide powders by aerosol decomposition, Google Patents. 200 Glicksman, H. D., Kodas, T. T., Majumdar, D. (1997). Method for making gold powders by aerosol decomposition, Google Patents. Glicksman, H. D., Kodas, T. T., Pluym, T. C. (1995). Method for making silver- palladium alloy powders by aerosol decomposition, Google Patents. Glicksman, H. D. and Yang, H. (2004). Aqueous developable photoimageable thick film compositions with photospeed enhancer, Google Patents. Goriparti, S., Miele, E., De Angelis, F., Di Fabrizio, E., Proietti Zaccaria, R., Capiglia, C. (2014). Review on recent progress of nanostructured anode materials for Li-ion batteries. Journal of Power Sources 257:421-443. Grouchko, M., Kamyshny, A., Magdassi, S. (2009). Formation of air-stable copper-silver core-shell nanoparticles for inkjet printing. Journal of Materials Chemistry 19:3057- 3062. Gurav, A., Kodas, T., Pluym, T., Xiong, Y. (1993). Aerosol Processing of Materials. Aerosol Science and Technology 19:411-452. Gürmen, S., Stopić, S., Friedrich, B. (2006). Synthesis of nanosized spherical cobalt powder by ultrasonic spray pyrolysis. Materials Research Bulletin 41:1882-1890. Habisreutinger, S. N., Schmidt-Mende, L., Stolarczyk, J. K. (2013). Photocatalytic Reduction of CO2 on TiO2 and Other Semiconductors. Angewandte Chemie International Edition 52:7372-7408. Hai, H. T., Takamura, H., Koike, J. (2013). Oxidation behavior of Cu–Ag core–shell particles for solar cell applications. Journal of Alloys and Compounds 564:71-77. Hernandez-Ramirez, F., Prades, J. D., Tarancon, A., Barth, S., Casals, O., Jimenez-Diaz, R., Pellicer, E., Rodriguez, J., Morante, J. R., Juli, M. A., Mathur, S., Romano- Rodriguez, A. (2008). Insight into the Role of Oxygen Diffusion in the Sensing Mechanisms of SnO2 Nanowires. Advanced Functional Materials 18:2990-2994. Hong, J., Zhang, W., Ren, J., Xu, R. (2013). Photocatalytic reduction of CO2: a brief review on product analysis and systematic methods. Analytical Methods 5:1086-1097. 201 Hong, Y. J. and Kang, Y. C. (2015). One-pot synthesis of core–shell-structured tin oxide– carbon composite powders by spray pyrolysis for use as anode materials in Li-ion batteries. Carbon 88:262-269. Hsieh, M.-C., Wu, G.-C., Liu, W.-G., Goddard, W. A., Yang, C.-M. (2014). Nanocomposites of Tantalum-Based Pyrochlore and Indium Hydroxide Showing High and Stable Photocatalytic Activities for Overall Water Splitting and Carbon Dioxide Reduction. Angewandte Chemie International Edition 53:14216-14220. Hsien-Hsueh, L., Kan-Sen, C., Kuo-Cheng, H. (2005). Inkjet printing of nanosized silver colloids. Nanotechnology 16:2436. Hsu, H.-C., Shown, I., Wei, H.-Y., Chang, Y.-C., Du, H.-Y., Lin, Y.-G., Tseng, C.-A., Wang, C.-H., Chen, L.-C., Lin, Y.-C., Chen, K.-H. (2013). Graphene oxide as a promising photocatalyst for CO2 to methanol conversion. Nanoscale 5:262-268. Huang, C.-Y. and Sheen, S. R. (1997). Synthesis of nanocrystalline and monodispersed copper particles of uniform spherical shape. Materials Letters 30:357-361. Huang, J. Y., Zhong, L., Wang, C. M., Sullivan, J. P., Xu, W., Zhang, L. Q., Mao, S. X., Hudak, N. S., Liu, X. H., Subramanian, A., Fan, H., Qi, L., Kushima, A., Li, J. (2010). In Situ Observation of the Electrochemical Lithiation of a Single SnO2 Nanowire Electrode. Science 330:1515-1520. Huang, X. K., Cui, S. M., Chang, J. B., Hallac, P. B., Fell, C. R., Luo, Y. T., Metz, B., Jiang, J. W., Hurley, P. T., Chen, J. H. (2015). A Hierarchical Tin/Carbon Composite as an Anode for Lithium-Ion Batteries with a Long Cycle Life. Angew Chem Int Edit 54:1490-1493. Ide, Y., Inami, N., Hattori, H., Saito, K., Sohmiya, M., Tsunoji, N., Komaguchi, K., Sano, T., Bando, Y., Golberg, D., Sugahara, Y. (2016). Remarkable Charge Separation and Photocatalytic Efficiency Enhancement through Interconnection of TiO2 Nanoparticles by Hydrothermal Treatment. Angewandte Chemie International Edition 55:3600-3605. Im, H. S., Cho, Y. J., Lim, Y. R., Jung, C. S., Jang, D. M., Park, J., Shojaei, F., Kang, H. S. (2013). Phase Evolution of Tin Nanocrystals in Lithium Ion Batteries. Acs Nano 7:11103-11111. Inoue, T., Fujishima, A., Konishi, S., Honda, K. (1979). Photoelectrocatalytic reduction of carbon dioxide in aqueous suspensions of semiconductor powders. Nature 277:637- 638. 202 Iskandar, F., Kim, S.-G., Nandiyanto, A. B. D., Kaihatsu, Y., Ogi, T., Okuyama, K. (2009). Direct synthesis of hBN/MWCNT composite particles using spray pyrolysis. Journal of Alloys and Compounds 471:166-171. Istratov, A. A., Flink, C., Hieslmair, H., McHugo, S. A., Weber, E. R. (2000). Diffusion, solubility and gettering of copper in silicon. Materials Science and Engineering: B 72:99- 104. Jeong, S., Lee, S. H., Jo, Y., Lee, S. S., Seo, Y.-H., Ahn, B. W., Kim, G., Jang, G.-E., Park, J.-U., Ryu, B.-H., Choi, Y. (2013). Air-Stable, Surface-Oxide Free Cu Nanoparticles for Highly Conductive Cu Ink and Their Application to Printed Graphene Transistors. Journal of Materials Chemistry C 1:2704-2710. Jeong, S., Woo, K., Kim, D., Lim, S., Kim, J. S., Shin, H., Xia, Y., Moon, J. (2008). Controlling the Thickness of the Surface Oxide Layer on Cu Nanoparticles for the Fabrication of Conductive Structures by Ink-Jet Printing. Advanced Functional Materials 18:679-686. Jianfeng, Y., Guisheng, Z., Anming, H., Zhou, Y. N. (2011). Preparation of PVP Coated Cu NPs and the Application for Low-Temperature Bonding. Journal of Materials Chemistry 21:15981-15986. Jiao, J., Wei, Y., Zhao, Y., Zhao, Z., Duan, A., Liu, J., Pang, Y., Li, J., Jiang, G., Wang, Y. (2017). AuPd/3DOM-TiO2 catalysts for photocatalytic reduction of CO2: High efficient separation of photogenerated charge carriers. Applied Catalysis B: Environmental 209:228-239. Jo, Y. H., Park, J. C., Bang, J. U., Song, H., Lee, H. M. (2011). New Synthesis Approach for Low Temperature Bimetallic Nanoparticles: Size and Composition Controlled Sn-Cu Nanoparticles. Journal of Nanoscience and Nanotechnology 11:1037-1041. Johari, P. and Shenoy, V. B. (2011). Modulating Optical Properties of Graphene Oxide: Role of Prominent Functional Groups. ACS Nano 5:7640-7647. Jordan, T. E. (1954). Vapor Pressure of Organic Compounds. Interscience Publishers. Joshi, B. N., An, S., Kim, Y. I., Samuel, E. P., Song, K. Y., Seong, I. W., Al-Deyab, S. S., Swihart, M. T., Yoon, W. Y., Yoon, S. S. (2017). Flexible freestanding Fe2O3-SnOx- carbon nanofiber composites for Li ion battery anodes. Journal of Alloys and Compounds 700:259-266. 203 Ju, Y., Tasaka, T., Yamauchi, H., Nakagawa, T. (2015). Synthesis of Sn nanoparticles and their size effect on the melting point. Microsystem Technologies 21:1849-1854. Jung, D. S., Hwang, T. H., Lee, J. H., Koo, H. Y., Shakoor, R. A., Kahraman, R., Jo, Y. N., Park, M.-S., Choi, J. W. (2014). Hierarchical Porous Carbon by Ultrasonic Spray Pyrolysis Yields Stable Cycling in Lithium–Sulfur Battery. Nano Letters 14:4418-4425. Jung, D. S., Hwang, T. H., Park, S. B., Choi, J. W. (2013). Spray Drying Method for Large-Scale and High-Performance Silicon Negative Electrodes in Li-Ion Batteries. Nano Letters 13:2092-2097. Jung, D. S., Lee, H. M., Kang, Y. C., Park, S. B. (2011). Air-Stable Silver-Coated Copper Particles of Sub-Micrometer Size. Journal of Colloid and Interface Science 364:574-581. Karabacak, T., DeLuca, J. S., Wang, P.-I., Ten Eyck, G. A., Ye, D., Wang, G.-C., Lu, T.- M. (2006). Low temperature melting of copper nanorod arrays. Journal of Applied Physics 99:064304. Kaspar, J., Terzioglu, C., Ionescu, E., Graczyk-Zajac, M., Hapis, S., Kleebe, H. J., Riedel, R. (2014). Stable SiOC/Sn Nanocomposite Anodes for Lithium-Ion Batteries with Outstanding Cycling Stability. Adv Funct Mater 24:4097-4104. Ke, F. S., Huang, L., Jamison, L., Xue, L. J., Wei, G. Z., Li, J. T., Zhou, X. D., Sun, S. G. (2013). Nanoscale tin-based intermetallic electrodes encapsulated in microporous copper substrate as the negative electrode with a high rate capacity and a long cycleability for lithium-ion batteries. Nano Energy 2:595-603. Kim, H.-S., Lee, K.-H., Kim, S.-G. (2006). Growth of Monodisperse Silver Nanoparticles in Polymer Matrix by Spray Pyrolysis. Aerosol Science and Technology 40:536-544. Kim, I., Kim, Y., Woo, K., Ryu, E.-H., Yon, K.-Y., Cao, G., Moon, J. (2013). Synthesis of oxidation-resistant core-shell copper nanoparticles. RSC Advances 3:15169-15177. Kim, J. H., Babushok, V. I., Germer, T. A., Mulholland, G. W., Ehrman, S. H. (2003). Cosolvent-assisted spray pyrolysis for the generation of metal particles. Journal of Materials Research 18:1614-1622. Kim, J. H., Ehrman, S. H., Germer, T. A. (2004). Influence of Particle Oxide Coating on Light Scattering by Submicron Metal Particles on Silicon Wafers. Applied Physics Letters 84:1278-1280. 204 Kim, S.-W., Kim, M., Lee, W. Y., Hyeon, T. (2002). Fabrication of Hollow Palladium Spheres and Their Successful Application to the Recyclable Heterogeneous Catalyst for Suzuki Coupling Reactions. Journal of the American Chemical Society 124:7642-7643. Kittel, C. (2005). Introduction to Solid State Physics. Wiley, Hoboken, NJ :. Kobayashi, Y. and Sakuraba, T. (2008). Silica-coating of metallic copper nanoparticles in aqueous solution. Colloids and Surfaces A: Physicochemical and Engineering Aspects 317:756-759. Koci, K., Obalova, L., Solcova, O. (2010). KINETIC STUDY OF PHOTOCATALYTIC REDUCTION OF CO2 OVER TIO2. Chemical and Process Engineering-Inzynieria Chemiczna I Procesowa 31:395-407. Kodas, T. T., Ward, T. L., Glicksman, H. D. (1995). Method for making silver powder by aerosol decomposition, Google Patents. Koirala, R., Pratsinis, S. E., Baiker, A. (2016). Synthesis of catalytic materials in flames: opportunities and challenges. Chemical Society Reviews 45:3053-3068. Korber, C., Wachau, A., Agoston, P., Albe, K., Klein, A. (2011). Self-Limited Oxygen Exchange Kinetics at SnO2 Surfaces. Physical Chemistry Chemical Physics 13:3223- 3226. Kumar, A., Rout, L., Achary, L. S. K., Mohanty, A., Dhaka, R. S., Dash, P. (2016). An investigation into the solar light-driven enhanced photocatalytic properties of a graphene oxide-SnO2-TiO2 ternary nanocomposite. RSC Advances 6:32074-32088. Kumar, K. S., Reinbold, L., Bower, A. F., Chason, E. (2011). Plastic deformation processes in Cu/Sn bimetallic films. Journal of Materials Research 23:2916-2934. Lee, J. U., Lee, W., Yi, J. W., Yoon, S. S., Lee, S. B., Jung, B. M., Kim, B. S., Byun, J. H. (2013). Preparation of highly stacked graphene papers via site-selective functionalization of graphene oxide. Journal of Materials Chemistry A 1:12893-12899. Leenen, M. A. M., Arning, V., Thiem, H., Steiger, J., Anselmann, R. (2009). Printable electronics: flexibility for the future. physica status solidi (a) 206:588-597. 205 Lenggoro, I. W., Hata, T., Iskandar, F., Lunden, M. M., Okuyama, K. (2000). An experimental and modeling investigation of particle production by spray pyrolysis using a laminar flow aerosol reactor. Journal of Materials Research 15:733-743. Lengsfeld, C. S., Delplanque, J. P., Barocas, V. H., Randolph, T. W. (2000). Mechanism Governing Microparticle Morphology during Precipitation by a Compressed Antisolvent:  Atomization vs Nucleation and Growth. The Journal of Physical Chemistry B 104:2725- 2735. Lengyel, M., Elhassid, D., Atlas, G., Moller, W. T., Axelbaum, R. L. (2014). Development of a scalable spray pyrolysis process for the production of non-hollow battery materials. Journal of Power Sources 266:175-178. Levenspiel, O. (1999). Chemical reaction engineering. Wiley. Levich, V. G. (1962). Physicochemical Hydrodynamics. Prentice Hall. Li, H., Gan, S., Wang, H., Han, D., Niu, L. (2015a). Intercorrelated Superhybrid of AgBr Supported on Graphitic-C3N4-Decorated Nitrogen-Doped Graphene: High Engineering Photocatalytic Activities for Water Purification and CO2 Reduction. Advanced Materials 27:6906-6913. Li, J., Li, Y., Wang, Z., Bian, H., Hou, Y., Wang, F., Xu, G., Liu, B., Liu, Y. (2016a). Ultrahigh Oxidation Resistance and High Electrical Conductivity in Copper-Silver Powder. Scientific Reports 6:39650. Li, J. and Mayer, J. W. (1992). Oxidation and reduction of copper oxide thin films. Materials Chemistry and Physics 32:1-24. Li, L., Yan, J., Wang, T., Zhao, Z.-J., Zhang, J., Gong, J., Guan, N. (2015b). Sub-10 nm rutile titanium dioxide nanoparticles for efficient visible-light-driven photocatalytic hydrogen production. Nature Communications 6:5881. Li, M., Zhang, L., Wu, M., Du, Y., Fan, X., Wang, M., Zhang, L., Kong, Q., Shi, J. (2016b). Mesostructured CeO2/g-C3N4 nanocomposites: Remarkably enhanced photocatalytic activity for CO2 reduction by mutual component activations. Nano Energy 19:145-155. Li, W., Wang, Y., Wang, M., Li, W., Tan, J., You, C., Chen, M. (2016c). Synthesis of Stable CucoreAgshell&Ag Particles for Direct Writing Flexible Paper-Based Electronics. RSC Advances 6:62236-62243. 206 Li, W., Zhang, H., Gao, Y., Jiu, J., Li, C.-F., Chen, C., Hu, D., Goya, Y., Wang, Y., Koga, H., Nagao, S., Suganuma, K. (2017). Highly Reliable and Highly Conductive Submicron Cu Particle Patterns Fabricated by Low Temperature Heat-Welding and Subsequent Flash Light Sinter-Reinforcement. Journal of Materials Chemistry C. Li, Y., Wu, Y., Ong, B. S. (2005). Facile Synthesis of Silver Nanoparticles Useful for Fabrication of High-Conductivity Elements for Printed Electronics. Journal of the American Chemical Society 127:3266-3267. Li, Z. J., Lv, W., Zhang, C., Qin, J. W., Wei, W., Shao, J. J., Wang, D. W., Li, B. H., Kang, F. Y., Yang, Q. H. (2014). Nanospace-confined formation of flattened Sn sheets in pre-seeded graphenes for lithium ion batteries. Nanoscale 6:9554-9558. Liao, J. Y. and Manthiram, A. (2014). Mesoporous TiO2-Sn/C Core-Shell Nanowire Arrays as High-Performance 3D Anodes for Li-Ion Batteries. Adv Energy Mater 4. Lide, D. (2007-2008). CRC Handbook of Chemistry and Physics. New York: CRC Press. Lin, F., Nordlund, D., Li, Y., Quan, M. K., Cheng, L., Weng, T.-C., Liu, Y., Xin, H. L., Doeff, M. M. (2016). Metal segregation in hierarchically structured cathode materials for high-energy lithium batteries. Nature Energy 1:15004. Lin, L.-Y., Nie, Y., Kavadiya, S., Soundappan, T., Biswas, P. (2017). N-doped reduced graphene oxide promoted nano TiO2 as a bifunctional adsorbent/photocatalyst for CO2 photoreduction: Effect of N species. Chemical Engineering Journal 316:449-460. Liu, C. J., Huang, H., Cao, G. Z., Xue, F. H., Camacho, R. A. P., Dong, X. L. (2014). Enhanced Electrochemical Stability of Sn-Carbon Nanotube Nanocapsules as Lithium- Ion Battery Anode. Electrochim Acta 144:376-382. Liu, C. M., Liu, W. L., Chen, W. J., Hsieh, S. H., Tsai, T. K., Yang, L. C. (2005). ITO as a Diffusion Barrier Between Si and Cu. Journal of The Electrochemical Society 152:G234-G239. Liu, J., Chen, H., Ji, H., Li, M. (2016). Highly Conductive Cu–Cu Joint Formation by Low-Temperature Sintering of Formic Acid-Treated Cu Nanoparticles. ACS Applied Materials & Interfaces 8:33289-33298. 207 Liu, W. L., Chen, W. J., Tsai, T. K., Hsieh, S. H., Liu, C. M. (2007a). Effect of Tin- Doped Indium Oxide Film as Capping Layer on the Agglomeration of Copper Film and the Appearance of Copper Silicide. Applied Surface Science 253:5516-5520. Liu, Z., Sun, D. D., Guo, P., Leckie, J. O. (2007b). An Efficient Bicomponent TiO2/SnO2 Nanofiber Photocatalyst Fabricated by Electrospinning with a Side-by-Side Dual Spinneret Method. Nano Letters 7:1081-1085. Long, R., Li, Y., Liu, Y., Chen, S., Zheng, X., Gao, C., He, C., Chen, N., Qi, Z., Song, L., Jiang, J., Zhu, J., Xiong, Y. (2017). Isolation of Cu Atoms in Pd Lattice: Forming Highly Selective Sites for Photocatalytic Conversion of CO2 to CH4. Journal of the American Chemical Society 139:4486-4492. Low, J., Yu, J., Jaroniec, M., Wageh, S., Al-Ghamdi, A. A. (2017). Heterojunction Photocatalysts. Advanced Materials 29:1601694. Loyalka, S. K. (1968). Momentum and Temperature‐Slip Coefficients with Arbitrary Accommodation at the Surface. The Journal of Chemical Physics 48:5432-5436. Lu, L., Shen, Y., Chen, X., Qian, L., Lu, K. (2004). Ultrahigh Strength and High Electrical Conductivity in Copper. Science 304:422-426. Luo, C., Langrock, A., Fan, X., Liang, Y., Wang, C. (2017). P2-type transition metal oxides for high performance Na-ion battery cathodes. Journal of Materials Chemistry A 5:18214-18220. Ma, J., Fang, Z., Yan, Y., Yang, Z., Gu, L., Hu, Y.-S., Li, H., Wang, Z., Huang, X. (2015). Novel Large-Scale Synthesis of a C/S Nanocomposite with Mixed Conducting Networks through a Spray Drying Approach for Li–S Batteries. Advanced Energy Materials 5:1500046. Ma, S., Liu, L., Bromberg, V., Singler, T. J. (2014). Fabrication of Highly Electrically Conducting Fine Patterns via Substrate-Independent Inkjet Printing of Mussel-Inspired Organic Nano-Material. Journal of Materials Chemistry C 2:3885-3889. Mack, E., Osterhof, G. G., Kraner, H. M. (1923). VAPOR PRESSURE OF COPPER OXIDE AND OF COPPER. Journal of the American Chemical Society 45:617-623. Mao, J., Peng, T., Zhang, X., Li, K., Zan, L. (2012). Selective methanol production from photocatalytic reduction of CO2 on BiVO4 under visible light irradiation. Catalysis Communications 28:38-41. 208 Matienzo, L. J., Das, R. N., Egitto, F. D. (2008). Electrically Conductive Adhesives for Electronic Packaging and Assembly Applications. Journal of Adhesion Science and Technology 22:853-869. Matsnnaga, S. and Homma, T. (1976). Influence on the oxidation kinetics of metals by control of the structure of oxide scales. Oxidation of Metals 10:361-376. Mayadas, A. F. and Shatzkes, M. (1970). Electrical-Resistivity Model for Polycrystalline Films: the Case of Arbitrary Reflection at External Surfaces. Physical Review B 1:1382- 1389. Mei, Z., Sunwoo, A. J., Morris, J. W. (1992). Analysis of low-temperature intermetallic growth in copper-tin diffusion couples. Metallurgical Transactions A 23:857-864. Miettinen, J. (2008). Thermodynamic description of the Cu–Fe–Sn system at the Cu–Fe side. Calphad 32:500-505. Mkhoyan, K. A., Contryman, A. W., Silcox, J., Stewart, D. A., Eda, G., Mattevi, C., Miller, S., Chhowalla, M. (2009). Atomic and Electronic Structure of Graphene-Oxide. Nano Letters 9:1058-1063. Mokkapati, S., Beck, F. J., Polman, A., Catchpole, K. R. (2009). Designing periodic arrays of metal nanoparticles for light-trapping applications in solar cells. Applied Physics Letters 95:053115. Moonen, P. F., Bat, E., Voorthuijzen, W. P., Huskens, J. (2013). Soft-Lithographic Patterning of Room Temperature-Sintering Ag Nanoparticles on Foil. RSC Advances 3:18498-18505. Morinaga, H., Suyama, M., Nose, M., Verhaverbeke, S., Ohmi, T. (1996). A model for the electrochemical deposition and removal of metallic impurities on Si surfaces. IEICE Transactions on Electronics E79C:343-362. Munnik, P., Wolters, M., Gabrielsson, A., Pollington, S. D., Headdock, G., Bitter, J. H., de Jongh, P. E., de Jong, K. P. (2011). Copper Nitrate Redispersion To Arrive at Highly Active Silica-Supported Copper Catalysts. The Journal of Physical Chemistry C 115:14698-14706. 209 Murcia-López, S., Bacariza, M. C., Villa, K., Lopes, J. M., Henriques, C., Morante, J. R., Andreu, T. (2017). Controlled Photocatalytic Oxidation of Methane to Methanol through Surface Modification of Beta Zeolites. ACS Catalysis 7:2878-2885. Nandiyanto, A. B. D. and Okuyama, K. (2011). Progress in developing spray-drying methods for the production of controlled morphology particles: From the nanometer to submicrometer size ranges. Advanced Powder Technology 22:1-19. Nguyen, Q. T., Kidder Jr, J. N., Ehrman, S. H. (2002). Hybrid gas-to-particle conversion and chemical vapor deposition for the production of porous alumina films. Thin Solid Films 410:42-52. Ni, W., Cheng, J. L., Shi, L. Y., Li, X. D., Wang, B., Guan, Q., Huang, L., Gu, G. F., Li, H. (2014). Integration of Sn/C yolk-shell nanostructures into free-standing conductive networks as hierarchical composite 3D electrodes and the Li-ion insertion/extraction properties in a gel-type lithium-ion battery thereof. J Mater Chem A 2:19122-19130. Nie, Y., Wang, W.-N., Jiang, Y., Fortner, J., Biswas, P. (2016). Crumpled reduced graphene oxide-amine-titanium dioxide nanocomposites for simultaneous carbon dioxide adsorption and photoreduction. Catalysis Science & Technology 6:6187-6196. Noor, F., Vorozhtsov, A., Lerner, M., Bandarra Filho, E. P., Wen, D. (2015). Thermal- Chemical Characteristics of Al–Cu Alloy Nanoparticles. The Journal of Physical Chemistry C 119:14001-14009. Nowick, A. S. (2012). Diffusion in solids: recent developments. Elsevier. O’Connell, R., Alexander, C., Strachan, R., Alway, B., Nambiath, S., Wiebe, J., Wong, L., Rannestad, E., Li, S., Aranda, D., Scott-Gray, N. (2017). World Silver Survey 2017, The Silver Institute. Ortega, J., Kodas, T. T., Chadda, S., Smith, D. M., Ciftcioglu, M., Brennan, J. E. (1991). Formation of dense barium calcium titanate (Ba0.86Ca0.14TiO3) particles by aerosol decomposition. Chemistry of Materials 3:746-751. Park, B. K., Jeong, S., Kim, D., Moon, J., Lim, S., Kim, J. S. (2007). Synthesis and size control of monodisperse copper nanoparticles by polyol method. Journal of Colloid and Interface Science 311:417-424. Perelaer, J., Smith, P. J., Mager, D., Soltman, D., Volkman, S. K., Subramanian, V., Korvink, J. G., Schubert, U. S. (2010). Printed electronics: the challenges involved in 210 printing devices, interconnects, and contacts based on inorganic materials. Journal of Materials Chemistry 20:8446-8453. Perry, D. L. (2011). Handbook of Inorganic Compounds. CRC press. Pingali, K. C., Rockstraw, D. A., Deng, S. (2005). Silver Nanoparticles from Ultrasonic Spray Pyrolysis of Aqueous Silver Nitrate. Aerosol Science and Technology 39:1010- 1014. Platteeuw, J. C. and Meyer, G. (1956). The system tin + oxygen. Transactions of the Faraday Society 52:1066-1073. Pluym, T. C., Powell, Q. H., Gurav, A. S., Ward, T. L., Kodas, T. T., Wang, L. M., Glicksman, H. D. (1993). Solid silver particle production by spray pyrolysis. Journal of Aerosol Science 24:383-392. Polat, D. B., Lu, J., Abouimrane, A., Keles, O., Amine, K. (2014). Nanocolumnar Structured Porous Cu-Sn Thin Film as Anode Material for Lithium-Ion Batteries. ACS Applied Materials & Interfaces 6:10877-10885. Qin, J., He, C., Zhao, N., Wang, Z., Shi, C., Liu, E.-Z., Li, J. (2014). Graphene Networks Anchored with Sn@Graphene as Lithium Ion Battery Anode. ACS Nano 8:1728-1738. R, V. K. R., K, V. A., P. S, K., Singh, S. P. (2015). Conductive silver inks and their applications in printed and flexible electronics. RSC Advances 5:77760-77790. Ramisetty, K. A., Pandit, A. B., Gogate, P. R. (2013). Investigations into ultrasound induced atomization. Ultrasonics Sonochemistry 20:254-264. Reddy, M. V., Subba Rao, G. V., Chowdari, B. V. R. (2013). Metal Oxides and Oxysalts as Anode Materials for Li Ion Batteries. Chemical Reviews 113:5364-5457. Sacris, E. M. and Parlee, N. A. D. (1970). The diffusion of hydrogen in liquid Ni, Cu, Ag, and Sn. Metallurgical Transactions 1:3377-3382. Saunders, N. and Miodownik, A. P. (1990). The Cu-Sn (Copper-Tin) system. Bulletin of Alloy Phase Diagrams 11:278-287. Schiotz, J., Di Tolla, F. D., Jacobsen, K. W. (1998). Softening of Nanocrystalline Metals at Very Small Grain Sizes. Nature 391:561-563. 211 Schirmer, N. C., Ströhle, S., Tiwari, M. K., Poulikakos, D. (2011). On the Principles of Printing Sub-micrometer 3D Structures from Dielectric-Liquid-Based Colloids. Advanced Functional Materials 21:388-395. Schlapbach, L. and Zuttel, A. (2001). Hydrogen-storage materials for mobile applications. Nature 414:353-358. Schoeller, H. and Cho, J. (2011). Oxidation and reduction behavior of pure indium. Journal of Materials Research 24:386-393. Schulz, H., Mädler, L., Strobel, R., Jossen, R., Pratsinis, S. E., Johannessen, T. (2005). Independent Control of Metal Cluster and Ceramic Particle Characteristics During One- step Synthesis of Pt/TiO2. Journal of Materials Research 20:2568-2577. semiwafer.com Silicon Wafer. Sen, D., Mazumder, S., Melo, J. S., Khan, A., Bhattyacharya, S., D'Souza, S. F. (2009). Evaporation Driven Self-Assembly of a Colloidal Dispersion during Spray Drying: Volume Fraction Dependent Morphological Transition. Langmuir 25:6690-6695. Sharma, M. K., Qi, D., Buchner, R. D., Scharmach, W. J., Papavassiliou, V., Swihart, M. T. (2014). Flame-driven Aerosol Synthesis of Copper–Nickel Nanopowders and Conductive Nanoparticle Films. ACS Applied Materials & Interfaces 6:13542-13551. Shen, W., Zhang, X., Huang, Q., Xu, Q., Song, W. (2014a). Preparation of Solid Silver Nanoparticles for Inkjet Printed Flexible Electronics with High Conductivity. Nanoscale 6:1622-1628. Shen, Y., Søndergaard, M., Christensen, M., Birgisson, S., Iversen, B. B. (2014b). Solid State Formation Mechanism of Li4Ti5O12 from an Anatase TiO2 Source. Chemistry of Materials 26:3679-3686. Shirai, H., Nguyen, M. T., Ishida, Y., Yonezawa, T. (2016). A New Approach for Additive-Free Room Temperature Sintering of Conductive Patterns Using Polymer- Stabilized Sn Nanoparticles. Journal of Materials Chemistry C 4:2228-2234. Shown, I., Hsu, H.-C., Chang, Y.-C., Lin, C.-H., Roy, P. K., Ganguly, A., Wang, C.-H., Chang, J.-K., Wu, C.-I., Chen, L.-C., Chen, K.-H. (2014). Highly Efficient Visible Light Photocatalytic Reduction of CO2 to Hydrocarbon Fuels by Cu-Nanoparticle Decorated Graphene Oxide. Nano Letters 14:6097-6103. 212 Smithells, C. J., Gale, W. F., Totemeier, T. C. (2004). Smithells Metals Reference Book. Elsevier Butterworth-Heinemann, Amsterdam ;. Smits, F. M. (1958). Measurement of Sheet Resistivities with the Four-Point Probe. Bell System Technical Journal 37:711-718. Song, H., Zhang, L., He, C., Qu, Y., Tian, Y., Lv, Y. (2011). Graphene sheets decorated with SnO2 nanoparticles: in situ synthesis and highly efficient materials for cataluminescence gas sensors. Journal of Materials Chemistry 21:5972-5977. Song, Y. L., Tsai, S. C., Chen, C. Y., Tseng, T. K., Tsai, C. S., Chen, J. W., Yao, Y. D. (2004). Ultrasonic Spray Pyrolysis for Synthesis of Spherical Zirconia Particles. Journal of the American Ceramic Society 87:1864-1871. Songping, W. and Shuyuan, M. (2006). Preparation of micron size copper powder with chemical reduction method. Materials Letters 60:2438-2442. Speight, J. (2004). Lange's Handbook of Chemistry. McGraw-Hill Professional, New York. Srinivasan, S. and Anantharaman, T. R. (1963). ACCURATE EVALUATION OF LATTICE PARAMETERS OF ALPHA-BRASSES. Current Science 32:262-263. Stern, K. H. (1972). High Temperature Properties and Decomposition of Inorganic Salts Part 3, Nitrates and Nitrites. Journal of Physical and Chemical Reference Data 1:747- 772. Strobel, R., Alfons, A., Pratsinis, S. E. (2006). Aerosol flame synthesis of catalysts. Advanced Powder Technology 17:457-480. Strobel, R. and Pratsinis, S. E. (2007). Flame aerosol synthesis of smart nanostructured materials. Journal of Materials Chemistry 17:4743-4756. Suenaga, K. and Koshino, M. (2010). Atom-by-atom spectroscopy at graphene edge. Nature 468:1088-1090. Sun, S. H., Meng, G. W., Zhang, M. G., An, X. H., Wu, G. S., Zhang, L. D. (2004). Synthesis of SnO 2 nanostructures by carbothermal reduction of SnO 2 powder. Journal of Physics D: Applied Physics 37:409. 213 Sun, Z., Zheng, H., Li, J., Du, P. (2015). Extraordinarily efficient photocatalytic hydrogen evolution in water using semiconductor nanorods integrated with crystalline Ni2P cocatalysts. Energy & Environmental Science 8:2668-2676. Tahir, M. and Amin, N. S. (2015). Indium-doped TiO2 nanoparticles for photocatalytic CO2 reduction with H2O vapors to CH4. Applied Catalysis B: Environmental 162:98- 109. Talbot, E. L., Yang, L., Berson, A., Bain, C. D. (2014). Control of the Particle Distribution in Inkjet Printing through an Evaporation-Driven Sol–Gel Transition. ACS Applied Materials & Interfaces 6:9572-9583. Talbot, L., Cheng, R. K., Schefer, R. W., Willis, D. R. (2006). Thermophoresis of particles in a heated boundary layer. Journal of Fluid Mechanics 101:737-758. Talley, C. E., Jackson, J. B., Oubre, C., Grady, N. K., Hollars, C. W., Lane, S. M., Huser, T. R., Nordlander, P., Halas, N. J. (2005). Surface-Enhanced Raman Scattering from Individual Au Nanoparticles and Nanoparticle Dimer Substrates. Nano Letters 5:1569- 1574. Tan, L.-L., Ong, W.-J., Chai, S.-P., Goh, B. T., Mohamed, A. R. (2015). Visible-light- active oxygen-rich TiO2 decorated 2D graphene oxide with enhanced photocatalytic activity toward carbon dioxide reduction. Applied Catalysis B: Environmental 179:160- 170. Tao, X., Wu, R., Xia, Y., Huang, H., Chai, W., Feng, T., Gan, Y., Zhang, W. (2014). Biotemplated Fabrication of Sn@C Anode Materials Based on the Unique Metal Biosorption Behavior of Microalgae. ACS Applied Materials & Interfaces 6:3696-3702. Temple, T. L. and Bagnall, D. M. (2011). Optical properties of gold and aluminium nanoparticles for silicon solar cell applications. Journal of Applied Physics 109:084343. Tian, H., Xin, F., Wang, X., He, W., Han, W. (2015). High capacity group-IV elements (Si, Ge, Sn) based anodes for lithium-ion batteries. Journal of Materiomics 1:153-169. Tran, P. D., Wong, L. H., Barber, J., Loo, J. S. C. (2012). Recent advances in hybrid photocatalysts for solar fuel production. Energy & Environmental Science 5:5902-5918. Tsai, S. C., Cheng, C. H., Ning, W., Song, Y. L., Lee, C. T., Tsai, C. S. (2009). Silicon- based megahertz ultrasonic nozzles for production of monodisperse micrometer-sized 214 droplets. Ultrasonics, Ferroelectrics and Frequency Control, IEEE Transactions on 56:1968-1979. Tsai, S. C., Song, Y. L., Tsai, C. S., Yang, C. C., Chiu, W. Y., Lin, H. M. (2004). Ultrasonic spray pyrolysis for nanoparticles synthesis. Journal of Materials Science 39:3647-3657. Wang, K., Cai, R., Yuan, T., Yu, X., Ran, R., Shao, Z. (2009). Process investigation, electrochemical characterization and optimization of LiFePO4/C composite from mechanical activation using sucrose as carbon source. Electrochimica Acta 54:2861- 2868. Wang, Q., Hisatomi, T., Jia, Q., Tokudome, H., Zhong, M., Wang, C., Pan, Z., Takata, T., Nakabayashi, M., Shibata, N., Li, Y., Sharp, I. D., Kudo, A., Yamada, T., Domen, K. (2016). Scalable water splitting on particulate photocatalyst sheets with a solar-to- hydrogen energy conversion efficiency exceeding 1%. Nature Materials 15:611-615. Wang, T., Chen, X., Lu, G.-Q., Lei, G.-Y. (2007a). Low-Temperature Sintering with Nano-Silver Paste in Die-Attached Interconnection. Journal of Electronic Materials 36:1333-1340. Wang, W.-N., Park, J., Biswas, P. (2011). Rapid synthesis of nanostructured Cu-TiO2- SiO2 composites for CO2 photoreduction by evaporation driven self-assembly. Catalysis Science & Technology 1:593-600. Wang, W.-N., Purwanto, A., Lenggoro, I. W., Okuyama, K., Chang, H., Jang, H. D. (2008). Investigation on the Correlations between Droplet and Particle Size Distribution in Ultrasonic Spray Pyrolysis. Industrial & Engineering Chemistry Research 47:1650- 1659. Wang, W.-N., Widiyastuti, W., Wuled Lenggoro, I., Oh Kim, T., Okuyama, K. (2007b). Photoluminescence Optimization of Luminescent Nanocomposites Fabricated by Spray Pyrolysis of a Colloid-Solution Precursor. Journal of The Electrochemical Society 154:J121-J128. Winter, M. and Besenhard, J. O. (1999). Electrochemical lithiation of tin and tin-based intermetallics and composites. Electrochim Acta 45:31-50. Wronski, C. R. M. (1967). The size dependence of the melting point of small particles of tin. British Journal of Applied Physics 18:1731. 215 Wu, C.-J., Sheng, Y.-J., Tsao, H.-K. (2016). Copper Conductive Lines on Flexible Substrates Fabricated at Room Temperature. Journal of Materials Chemistry C 4:3274- 3280. Wu, S. P., Gao, R. Y., Xu, L. H. (2009). Preparation of micron-sized flake copper powder for base-metal-electrode multi-layer ceramic capacitor. Journal of Materials Processing Technology 209:1129-1133. Xia, B., Lenggoro, I. W., Okuyama, K. (2000). The roles of ammonia and ammonium bicarbonate in the preparation of nickel particles from nickel chloride. Journal of Materials Research 15:2157-2166. Xia, B., Lenggoro, I. W., Okuyama, K. (2001a). Preparation of Ni particles by ultrasonic spray pyrolysis of NiCl2·6H2O precursor containing ammonia. Journal of Materials Science 36:1701-1705. Xia, B., Lenggoro, I. W., Okuyama, K. (2001b). Preparation of Nickel Powders by Spray Pyrolysis of Nickel Formate. Journal of the American Ceramic Society 84:1425-1432. Xie, Y., Liang, Y., Chen, D., Wu, X., Dai, L., Liu, Q. (2014). Vortical superlattices in a gold nanorods' self-assembled monolayer. Nanoscale 6:3064-3068. Xu, Y., Liu, Q., Zhu, Y., Liu, Y., Langrock, A., Zachariah, M. R., Wang, C. (2013a). Uniform Nano-Sn/C Composite Anodes for Lithium Ion Batteries. Nano Letters 13:470- 474. Xu, Y. H., Liu, Q., Zhu, Y. J., Liu, Y. H., Langrock, A., Zachariah, M. R., Wang, C. S. (2013b). Uniform Nano-Sn/C Composite Anodes for Lithium Ion Batteries. Nano Lett 13:470-474. Xu, Z., Lai, E., Shao-Horn, Y., Hamad-Schifferli, K. (2012). Compositional dependence of the stability of AuCu alloy nanoparticles. Chemical Communications 48:5626-5628. Yamauchi, T., Tsukahara, Y., Sakata, T., Mori, H., Yanagida, T., Kawai, T., Wada, Y. (2010). Magnetic Cu-Ni (core-shell) nanoparticles in a one-pot reaction under microwave irradiation. Nanoscale 2:515-523. Yan, Y., Ben, L. B., Zhan, Y. J., Huang, X. J. (2016). Nano-Sn embedded in expanded graphite as anode for lithium ion batteries with improved low temperature electrochemical performance. Electrochim Acta 187:186-192. 216 Yang, Y.-C., Huang, W.-Q., Xu, L., Hu, W., Peng, P., Huang, G.-F. (2017). Hybrid TiO2/graphene derivatives nanocomposites: is functionalized graphene better than pristine graphene for enhanced photocatalytic activity? Catalysis Science & Technology 7:1423-1432. Yaws, C. L. (1995). Handbook of Vapor Pressure: Volume 4:: Inorganic Compounds and Elements. Gulf Professional Publishing. Yaws, C. L. (1999). Chemical Properties Handbook. McGraw-Hill, New York. Ye, T., Huang, W., Zeng, L., Li, M., Shi, J. (2017). CeO2-x platelet from monometallic cerium layered double hydroxides and its photocatalytic reduction of CO2. Applied Catalysis B: Environmental 210:141-148. Yiwei, A., Yunxia, Y., Shuanglong, Y., Lihua, D., Guorong, C. (2007). Preparation of spherical silver particles for solar cell electronic paste with gelatin protection. Materials Chemistry and Physics 104:158-161. Yonezawa, T., Tsukamoto, H., Matsubara, M. (2015a). Low-temperature nanoredox two- step sintering of gelatin nanoskin-stabilized submicrometer-sized copper fine particles for preparing highly conductive layers. RSC Advances 5:61290-61297. Yonezawa, T., Uchida, Y., Tsukamoto, H. (2015b). X-ray diffraction and high-resolution TEM observations of biopolymer nanoskin-covered metallic copper fine particles: preparative conditions and surface oxidation states. Physical Chemistry Chemical Physics 17:32511-32516. Yong, Y., Yonezawa, T., Matsubara, M., Tsukamoto, H. (2015). The Mechanism of Alkylamine-Stabilized Copper Fine Particles Towards Improving the Electrical Conductivity of Copper Films at Low Sintering Temperature. Journal of Materials Chemistry C 3:5890-5895. Youn, D. H., Heller, A., Mullins, C. B. (2016). Simple Synthesis of Nanostructured Sn/Nitrogen-Doped Carbon Composite Using Nitrilotriacetic Acid as Lithium Ion Battery Anode. Chemistry of Materials 28:1343-1347. Youngil, L., Jun-rak, C., Kwi Jong, L., Nathan, E. S., Donghoon, K. (2008). Large-scale synthesis of copper nanoparticles by chemically controlled reduction for applications of inkjet-printed electronics. Nanotechnology 19:415604. 217 Yuan, L., Wang, C., Cai, R., Wang, Y., Zhou, G. (2013). Spontaneous ZnO nanowire formation during oxidation of Cu-Zn alloy. Journal of Applied Physics 114:023512. Yue, H., Zhao, Y., Ma, X., Gong, J. (2012). Ethylene Glycol: Properties, Synthesis, and Applications. Chemical Society Reviews 41:4218-4244. Zerjav, G., Arshad, M. S., Djinovic, P., Junkar, I., Kovac, J., Zavasnik, J., Pintar, A. (2017). Improved electron-hole separation and migration in anatase TiO2 nanorod/reduced graphene oxide composites and their influence on photocatalytic performance. Nanoscale 9:4578-4592. Zhang, N., Zhao, Q., Han, X., Yang, J., Chen, J. (2014a). Pitaya-like Sn@C nanocomposites as high-rate and long-life anode for lithium-ion batteries. Nanoscale 6:2827-2832. Zhang, N., Zhao, Q., Han, X. P., Yang, J. G., Chen, J. (2014b). Pitaya-like Sn@C nanocomposites as high-rate and long-life anode for lithium-ion batteries. Nanoscale 6:2827-2832. Zhang, W. M., Hu, J. S., Guo, Y. G., Zheng, S. F., Zhong, L. S., Song, W. G., Wan, L. J. (2008). Tin-nanoparticles encapsulated in elastic hollow carbon spheres for high- performance anode material in lithium-ion batteries. Adv Mater 20:1160-+. Zhang, Y. J., Jiang, L., Wang, C. R. (2015). Preparation of a porous Sn@C nanocomposite as a high-performance anode material for lithium-ion batteries. Nanoscale 7:11940-11944. Zhao, X., Zheng, B., Li, C., Gu, H. (1998). Acetate-derived ZnO ultrafine particles synthesized by spray pyrolysis. Powder Technology 100:20-23. Zhong, K., Peabody, G., Blankenhorn, E., Glicksman, H., Ehrman, S. (2013a). A Spray Pyrolysis Approach for the Generation of Patchy Particles. Aerosol Science and Technology 47:i-v. Zhong, K., Peabody, G., Blankenhorn, E., Glicksman, H., Ehrman, S. (2013b). Spray pyrolysis of phase pure AgCu particles using organic cosolvents. Journal of Materials Research 28:2753-2761. Zhong, K., Peabody, G., Glicksman, H., Ehrman, S. (2012). Particle generation by cosolvent spray pyrolysis: Effects of ethanol and ethylene glycol. Journal of Materials Research 27:2540-2550. 218 Zhu, H. L., Jia, Z., Chen, Y. C., Weadock, N., Wan, J. Y., Vaaland, O., Han, X. G., Li, T., Hu, L. B. (2013). Tin Anode for Sodium-Ion Batteries Using Natural Wood Fiber as a Mechanical Buffer and Electrolyte Reservoir. Nano Lett 13:3093-3100. Zhu, Y., Choi, S. H., Fan, X., Shin, J., Ma, Z., Zachariah, M. R., Choi, J. W., Wang, C. (2017). Recent Progress on Spray Pyrolysis for High Performance Electrode Materials in Lithium and Sodium Rechargeable Batteries. Advanced Energy Materials 7:1601578. Zhu, Z., Wang, S., Du, J., Jin, Q., Zhang, T., Cheng, F., Chen, J. (2014a). Ultrasmall Sn Nanoparticles Embedded in Nitrogen-Doped Porous Carbon As High-Performance Anode for Lithium-Ion Batteries. Nano Letters 14:153-157. Zhu, Z. Q., Wang, S. W., Du, J., Jin, Q., Zhang, T. R., Cheng, F. Y., Chen, J. (2014b). Ultrasmall Sn Nanoparticles Embedded in Nitrogen-Doped Porous Carbon As High- Performance Anode for Lithium-Ion Batteries. Nano Lett 14:153-157. 219 Conference presentations 1. Oral presentation, Photoreduction of CO2 by SnO2/graphene oxide composite particles, CATL 405, 254th ACS National Meeting in Washington, DC, August, 2017 2. Oral presentation, A Novel Spray Process Integrates the Features of Spray Pyrolysis and Drying for Battery Material Generation, NM6.9.02, 2016 MRS Fall Meeting, Boston, November-December, 2016 3. Oral presentation, Investigation of the Oxidation Behavior of Cu/Sn and Cu/C Composite Particles Fabricated Via Spray Pyrolysis, 197f, 2016 AIChE Annual Meeting, San Francisco, November, 2016 4. Oral presentation, Copper (Cu) -Tin (Sn) Binary Particles Fabrication Via Spray Pyrolysis and Their Oxidation Behavior, 14f, 2015 AIChE Annual Meeting, Salt Lake City, November, 2015 Selected publications related to this dissertation 1. Y. Liang, W. Wu, P. Wang, S-C. Liou, D. Liu, S. Ehrman, Scalable fabrication of SnO2/eo-GO nanocomposites for CO2 photoreduction to CH4, Nano Research, 2018, DOI: 10.1007/s12274-018-1988-x 2. Y. Liang, H. Tian, J. Repac, S-C. Liou, W. Han, C. Wang, S. Ehrman, Colloidal spray pyrolysis: A new fabrication technology for nanostructured energy storage materials, Energy Storage Materials, 2018, 13, 8-18 3. Y. Liang, H. Hou, Y. Yang, H. Glicksman, S. Ehrman, Conductive One- and Two- Dimensional Structures Fabricated Using Oxidation-Resistant Cu-Sn Particles, ACS Applied Materials & Interfaces, 2017, 9, 34587–34591 220 4. Y. Liang, S. Oh, X. Wang, H. Glicksman, D. Liu, S. Ehrman, Oxidation-Resistant Micron-Sized Cu-Sn Solid Particles Fabricated by a One-Step and Scalable Method, RSC Advances, 2017, 7, 23468-23477 5. Y. Liang, R. Felix, H. Glicksman, S. Ehrman, Cu-Sn binary metal particle generation by spray pyrolysis, Aerosol Science and Technology, 2017, 51, 430-442 221