Oriented Epitaxial Bismuth Ferrite Thin-Films with Robust Ferroelectric Properties

Development of Chemical Solution Deposition Derived (001)-Oriented Epitaxial Bismuth Ferrite Thin-Films with Robust Ferroelectric Properties

Qi Zhang

A thesis in fulfilment of the requirements for the degree of Doctor of Philosophy

School of Materials Science and Engineering Faculty of Science The University of New South Wales

April 2015

Originality Statement

‘I hereby declare that this submission is my own work and to the best of my knowledge it contains no materials previously published or written by another person, or substantial proportions of material which have been accepted for the award of any other degree or diploma at UNSW or any other educational institution, except where due acknowledgement is made in the thesis. Any contribution made to the research by others, with whom I have worked at UNSW or elsewhere, is explicitly acknowledged in the thesis. I also declare that the intellectual content of this thesis is the product of my own work, except to the extent that assistance from others in the project’s design and conception or in style, presentation and linguistic expression is acknowledged.’

Signed......

Date......

Acknowledgement

First, I wish to thank my supervisor, Dr Owen Standard, for his patient guidance and insightful advice during my thesis. His critical thinking and professional character regarding experiment design, data analysis and thesis writing have impacted immensely on my research training. I also would like to express my appreciation to my co- supervisor, Prof. Nagarajan Valanoor for the fabulous opportunity to be a part of his research group. His professional and enthusiastic guidance, as well as friendly personality, have made a positive influence on me as well as the whole group.

Next, I express my gratitude to many people who helped me over the course of the research work. In particular, I would like to thank Prof. Chris Sorell for using his spin coater, Dr Yu Wang for his help on XRD analysis, Dr Chris Marjo and Dr Anne

Rich for their expert opinions on FTIR and Raman analysis, Dr Donald Thomas for his time and effort on NMR analysis, Dr Nadia Court and Dr Fay Hudson from ANFF for their help on device preparation, Mr. Bill Joe for his support on ferroelectric testing system maintenance, Dr Rahmat Kartono for his furnace maintenance, Dr George Yang for his assistance in microstructural sample preparation, Mr Anthony Zhang and Ms Soo

Chong for their lab assistance, and Ms Jane Gao and Mr Danny Kim for their IT service support.

I also would like to thank my group members and friends who supported me for my project. Miss Hsin-Hui Huang and Miss Yanyu Zhou for their help on TEM, Mr

Jeffrey Cheung for his patient technical support on thin film characterization, Mr Ruoyu

Li for his help on data analysis, Mr Rui Ding for his help on film property measurement,

Mr Guangqing Liu for his help on film buffer layer optimization and deposition, Dr

Dewei Chu for his help and advice on film resistivity measurement, Miss Xuan Cheng

for her help on my thesis and paper proof reading, Miss Cong Chen for the happiness she brings to my research life as the best neighbour. I also would like to thank Ruoming

Tian, J.J.Lee, Xing Xing, Qinghua Cao, for their help, support and encouragement over the past four-year of my PhD study.

Last but not least, I am deeply grateful to my husband, Mr Ming Xu for his trust and encouragement of me all these years, to my parents for their understanding and support on my decision in pursuing PhD degree abroad, and to all my friends who encouraged and supported me. Without them, I would have gone insane on this seemingly endless endeavour.

iii

Abstract

Bismuth ferrite (BiFeO3, BFO) has attracted recent attention due to its multifunctional properties, including multiferroism, resistive switching and photovoltaic effects. In particular, epitaxial BFO has been shown to demonstrate giant polarization, polarization-mediated resistive switching and unique magnetic properties. Until now, the most popular methods to obtain epitaxial BFO films with robust properties have been pulsed laser deposition (PLD) and radio frequency (RF) sputtering. Films made using these methods have been reported to have a high spontaneous polarization of up to 130 µC/cm2 and a switchable diode effect.

Chemical deposition techniques, such as chemical solution deposition (CSD), have attracted recent interest for the preparation of BFO films and owing to them offering a cost-effective and more convenient manufacturing method compared with

PLD and RF sputtering, an aspect of particular importance in an industrial context.

However, the large scale adoption of CSD-derived BFO thin films for a variety of applications has been stymied by a number of significant limitations and challenges including: (1) the imprecision of the starting chemical composition and the subsequent volatilisation of Bi during the annealing step leading to the formation of secondary phases and/or highly conductive films with very poor leakage resistance; (2) variable sintering and densification behaviour leading to films having porosity and poor microstructures; and, (3) limited epitaxy between the film and substrate. Collectively, these dramatically impair the structural and electromechanical properties of the BFO films rendering them unsuitable for practical application. Thus, there is the important need to optimize the CSD preparation process for obtaining pure-phase epitaxial BFO.

In this thesis, a non-aqueous CSD route was developed and studied with the aim to optimise it for the preparation of epitaxial (001) BFO thin films with robust (square) polarization hysteresis loops, high dielectric constant, strong piezoelectric response and distinct diode behavior.

Molecular changes in the organic precursors on heating (determined by NMR and FTIR) and the effects of gelation temperature–time and thickness on film morphology were studied to develop an optimal deposition–gelation process for the synthesis of homogenous, defect-free gel films suitable for subsequent crystallization.

The key to obtaining a homogenous gel was control of the delicate balance between gelation and salt (metal nitrate) precipitation through solvent evaporation. The optimized synthesis route consists of spin-coating 0.25 M precursor on 70°C preheated substrate at 3000 rpm for 30 seconds then gelating at 90°C then drying at 270°C.

The crystallization of optimized gel films was studied as a function of Bi/Fe concentration and stoichiometry in the precursor solution, film thickness and single versus multiple depositions, crystallization temperature and atmosphere. Oxygen atmosphere was found to be essential for suppression of Bi volatilization and promotion of film epitaxial orientation. Pure-phase, epitaxial BFO thin film on (001)-strontium titanate (STO) substrate was obtained by rapidly heating the thin film to 650°C in an oxygen atmosphere and holding at the temperature for 30 minutes. A multi-layer deposition process for fabrication of films of various thicknesses was optimised by study of the deposition-heating sequence.

The ferroelectric properties of pure-phase, epitaxial BFO thin films on lanthanum strontium manganite buffered (001)-STO substrates were studied as a function of thickness (40, 70, and 150 nm). The 70 and 150 nm films exhibited

exhibited square hysteresis loops at room temperature with high remanent polarization

2 (2Pr up to 100 μC/cm ), low coercive field (2Ec down to 193 kV/cm), and high relative dielectric constant (up to 613). High-cycle fatigue tests showed that these films are resistant to polarization fatigue (up to 108 cycles). All thicknesses showed resistive switching behaviour and a polarization-mediated diode effect both of which became more pronounced with decreasing thickness. The CSD technique developed in this work yielded high-quality BFO thin films and offers a viable low-cost alternative to current

BFO deposition techniques.

List of Publications

1. Q. Zhang, V. Nagarajan, and O. Standard, Chemical solution deposition derived (001)- oriented epitaxial BiFeO3 thin films with robust ferroelectric properties using stoichiometric precursors (invited manuscript), Journal of Applied Physics, 2014, 116, 066810.

2. A. Rana, H. Lu, K. Bogle, Q. Zhang, R. Vasudevan, V. Thakare, A. Gruverman, S. Ogale and V Nagarajan, Scaling behavior of the resistive switching in epitaxial bismuth ferrite heterostructures, Advanced Functional Materials, 2014, 24 (25), 3962.

3. Q. Zhang, V.Nagarajan, and O.Standard, Epitaxial (001) BiFeO3 thin-films with excellent ferroelectric properties by chemical solution deposition-the role of gelation, Journal of Materials Chemstry C, 2015,3, 582

vii

Table of Contents

Originality statement ...... i

Acknowledgement ...... ii

Abstract ...... iv

List of Publications ...... vii

Table of Contents ...... viii

List of Figures ...... xii

List of Tables ...... xx

Chapter 1...... 1 1.1 Motivations ...... 1 1.2 Objectives ...... 2 1.3 Overview ...... 5

Chapter 2...... 7

2.1 Piezoelectricity and Ferroelectricity ...... 7 2.1.1 Piezoelectricity ...... 7 2.1.2 Ferroelectricity ...... 10 2.1.3 Applications ...... 16 2.1.4 Materials ...... 18 2.2 Bismuth Ferrite ...... 20 2.2.1 Crystallographic Structure...... 20 2.2.2 Phase Composition ...... 22 2.2.3 Characteristics of BFO ...... 27 2.3 Fabrication Processes for Bismuth Ferrite Thin-Film ...... 31 2.3.1 Film Fabrication Processing ...... 31 2.3.2 Thin Film Preparation Process using CSD Technique ...... 33 2.4 Summary ...... 44

viii

Chapter 3...... 46 3.1 Bismuth Ferrite Thin film Preparation Process ...... 46 3.1.1 Precursor Preparation ...... 46 3.1.2 Film Deposition ...... 48 3.1.3 High-Temperature Heat Treatment ...... 49 3.1.4 Device Preparation ...... 50 3.2 Analytical Equipment and Methods ...... 54 3.2.1 Chemical Analysis ...... 54 3.2.2 Phase Composition and Microstructure analysis...... 57 3.2.3 Scanning Probe Microscopy...... 59 3.2.4 Ferroelectric Test System ...... 70 3.2.5 Macro Scale I-V Tester ...... 72 3.2.6 Impedance Analysis ...... 73 3.3 Summary ...... 73

Chapter 4...... 75 4.1 Introduction ...... 75 4.2 Experimental Procedure ...... 79 4.2.1 Investigation of Gelation Chemistry ...... 79 4.2.2 Investigation of Gelation Conditions and the Effect of Film Thickness ...... 81 4.2.3 Investigation of Film Crystallization and Microstructure ...... 84 4.3 Results and Discussion ...... 85 4.3.1 Investigation of Gelation Chemistry ...... 85 4.3.2 Investigation of Gelation Conditions and the Effect of Film Thickness ...... 100 4.3.3 Investigation of Film Crystallization and Microstructure ...... 118 4.4 Summary ...... 125

Chapter 5...... 127 5.1 Introduction ...... 127 5.2 Experimental Procedure ...... 129 5.3 Results and Discussion ...... 132 5.3.1 Phase Evolution and Crystal Structure Investigation ...... 132 5.3.2 Film Cross-Section Microstructure Investigation ...... 138 5.3.3 Surface Microstructure and Domain Structure Investigation ...... 141 5.3.4 Micro and Macro Ferroelectric Properties ...... 144 5.4 Summary ...... 146

Chapter 6...... 148

6.1 Introduction ...... 148 6.2 Experimental Procedure ...... 151 6.2.1 Differential Scanning Calorimetric Testing ...... 151 6.2.2 BFO Thin Film Preparation...... 151 6.3 Results and Discussion ...... 153 6.3.1 Precursor Thermal Analysis ...... 153 6.3.2 Phase Composition ...... 154 6.3.3 Microstructure and Domain Structure ...... 156 6.3.4 Ferroelectric Property Investigation ...... 159 6.3.5 Current-Voltage Behaviour ...... 161 6.3.6 Discussion ...... 163 6.4 Summary ...... 171

Chapter 7...... 173

7.1 Introduction ...... 173 7.2 Experimental Section ...... 174 7.2.1 Thin Film Preparation ...... 174 7.2.2 Composition Evolution and Structural Characterization ...... 175 7.2.3 Macro-Scale Ferroelectric and Dielectric Characterizations ...... 176 7.2.4 Micro-Scale Ferroelectric Characterizations ...... 177 7.3 Results and Discussion ...... 178 7.3.1 Composition Evolution and Structural Characteristics ...... 178 7.3.2 Macroscale Ferroelectric Properties ...... 184 7.3.3 Local Electromechanical and Diode Behaviour Investigations via Atomic Force Microscopy ...... 193 7.4 Summary ...... 199

Chapter 8...... 200

8.1 Thesis Conclusions ...... 200 8.2 Future Work ...... 205

8.2.1 Preparation and Characterization of CSD-Derived Mixed-Phase (001)-BiFeO3

Thin-Film on (001)-LaAlO3 Substrate ...... 205 8.2.2 Study of La-doped BFO system and Selected Properties ...... 207

Appendices ...... 209 A1. Study on the Gelation Chemistry and Gelation Condition using

Stoichiometric Precursor ...... 209 A1.1 Precursor Formulations...... 209 A1.2 FTIR Analysis ...... 210 A1.2.1 Precursor AA ...... 210 A1.2.2 Precursor BB ...... 213 A1.2 Gelation Condition ...... 215 A1.3.1 Heating Condition for Gelation ...... 216 A1.3.2 Effect of Preheating on gelation ...... 218

A2. Frequency dependency of P-E hysteresis loop ...... 222

References ...... 224

List of Figures

Figure 2-1. Illustration of direct and converse piezoelectric effect ...... 7 Figure 2-2. Illustration of the relationship between crystal symmetry and polarization ((a) (b) symmetric structure; (c)(d) asymmetric structure) ...... 8 Figure 2-3. Notation of polarization axes ...... 9

Figure 2-4. Illustration of the perovskite structure ABO3 ...... 11 Figure 2-5. 90° and 180° domain walls in tetragonal perovskite ferroelectric materials (all symbols are explained in the text) ...... 13 Figure 2-6. 71°, 109° and 180° domain walls in rhombohedral perovskite ferroelectric materials ...... 13 Figure 2-7. Typical hysteresis loop of the ferroelectric materials ...... 14

Figure 2-8. Structure of BiFeO3 ...... 21

Figure 2-9. Phase diagram of Bi2O3-Fe2O3 ...... 23 Figure 2-10. Flow chart of sol-gel process ...... 33 Figure 2-11. Pseudotetragonal or pseudocubic a-axis lattice constants (in angstroms) of some ferroelectric perovskites of current interest ...... 38 Figure 2-12. Spin-coating process ...... 40 Figure 2-13. Schematic diagram of the free energies of a sol-gel derived amorphous film compared with the corresponding equilibrium liquid and a crystal ... 43 Figure 3-1. Illustration of BFO precursor preparation ...... 47 Figure 3-2. Schematics of BFO thin film device ...... 50 Figure 3-3. Schematic diagram of the photolithographic process ...... 52 Figure 3-4. Schematics of the metal thermal evaporation process ...... 53 Figure 3-5. Schematic diagrams depicting (a) the optical lever detection system used in AFM and (b) the photodiode detector ...... 60 Figure 3-6. AFM image of bismuth ferrite thin film with secondary phase (tapping mode) ...... 62 Figure 3-7. Principle of PFM measurement ...... 63 Figure 3-8. PFM amplitude (a) and phase (b) image of polycrystalline BFO thin film

on Pt/Ti/SiO2/Si substrate ...... 63

xii

Figure 3-9. Amplitude(a)(b) and phase(c)(d) images of VPFM(a)(c) and LPFM(b)(d)

of PbTiO3 thin film ...... 64 Figure 3-10. Principle of Dual AC resonance tracking PFM ...... 65 Figure 3-11. Domain switching by writing with DC bias ((a) amplitude and (b) phase images of (001)BFO thin films ) ...... 66 Figure 3-12. (a)Waveform for PFM spectroscopy hysteresis loop and (b) PFM phase and amplitude hysteresis loops of BFO/LSMO/STO(001) thin films ...... 67 Figure 3-13. Principle of CAFM measurement ...... 68 Figure 3-14. (a) Topography, (b) PFM phase and (c) current map acquired from the same region (1.5 µm × 1.5 µm) of BFO/LSMO/STO thin film ...... 69 Figure 3-15. I-V curve of BFO thin film measured by CAFM ...... 69 Figure 3-16. (a) Triangular drive voltage sweep for ferroelectric polarization measurement and (b) example of ferroelectric polarization hysteresis loop obtained for BFO/ LSMO/ STO thin films ...... 71 Figure 3-17. Pulses drive voltage for PUND measurement ...... 72 Figure 4-1. Illustration of Chapter 4 research route ...... 79 Figure 4-2. Examples of dry gel films with different thicknesses used in experiments ...... 82 Figure 4-3. (a) 1H NMR spectra of Precursor A1 solution; (b) FTIR spectra of raw materials and as-prepared precursor A (2000 cm-1-780 cm-1) ...... 86 Figure 4-4. FTIR analysis of bismuth ferrite Precursor A showing: (a) FTIR patterns of bismuth ferrite starting precursor before heating and after heating for various durations at 90°C (2000-780 cm-1); (b) FTIR patterns of 2-MOE and bismuth ferrite starting Precursor A before heating and after heating for 75 s (1200-950 cm-1) ...... 88 Figure 4-5. FTIR patterns of 2-MOE and metal nitrates in 2-MOE precursor after heating at 90°C for 75 s (1200-900 cm-1) ...... 91 Figure 4-6. (a) 1H NMR spectrum of Precursor B1; (b) FTIR spectra of raw materials and as-prepared Precursors A and B (2000 cm-1-780 cm-1) ...... 93 Figure 4-7. FTIR spectra of 2-MOE, acetic anhydride, and mixture of 2-MOE and acetic anhydride before heating (1900-1650cm-1) ...... 95

xiii

Figure 4-8. FTIR analysis of bismuth ferrite precursor in 2-MOE and acetic anhydride solvent showing: (a) FTIR patterns of bismuth ferrite starting precursor before and after heating for various durations at 90°C (2000-650 cm-1); (b) FTIR patterns of bismuth ferrite starting precursor before heating and after heating for 90 s (1200-950 cm-1); and (c) FTIR patterns of bismuth ferrite starting precursor before heating and after heating for 90 s (1850-1650 cm-1) ...... 97 Figure 4-9. Microscopic structures of thin films after heating at various temperatures for 10 minutes, followed by drying at room temperature in an open ambient air: (a) No heat treatment; (b) 50°C;(c) 60°C; (d) 70°C; (e) 80°C; (f) 90°C ...... 102 Figure 4-10. Microscopic structures of medium thickness films after heating at various temperatures for 10 minutes, followed by drying at room temperature for 2 hours in open ambient air: (a) No heat treatment; (b) 50°C;(c) 60°C; (d) 70°C; (e) 80°C; (f) 90°C ...... 103 Figure 4-11. Microscopic structures of thick films after heating at various temperatures for 10 minutes, followed by drying at room temperature for 2 hours in an open ambient air: (a) No heat treatment; (b) 50°C;(c) 60°C; (d) 70°C; (e) 80°C; (f) 90°C ...... 105 Figure 4-12. Microscopic structures of thin films after heating at 70°C for various durations, followed by drying at room temperature in an open ambient air: (a) 5 minutes; (b) 10 minutes;(c) 20 minutes; (d) 60 minutes; (e) 120 minutes ...... 107 Figure 4-13. Microscopic structures of medium-thick films after heating at 70°C for various durations, followed by drying at room temperature in an open ambient air: (a) 5 minutes; (b) 10 minutes;(c) 20 minutes; (d) 60 minutes; (e) 120 minutes ...... 108 Figure 4-14. Microscopic structures of thick films after heating at 70°C for various durations, followed by drying at room temperature in an open ambient air: (a) 5 minutes; (b) 10 minutes;(c) 20 minutes; (d) 60 minutes; (e) 120 minutes ...... 109

xiv

Figure 4-15. Microscopic structures of films with various thicknesses((a)(b)(c)-thin film;(d)(e)(f)-medium thick film; (g)(h)(i)-thick film) after heating at 70°C for 10 minutes, followed by drying under various post-gelation drying processes ((a)(d)(g)-drying at room temperature in an open ambiance; (b)(e)(h)-drying at 40°C for 10 minutes in an oven; (c)(f)(i)- drying at 40°C for 60 minutes in an oven) ...... 111 Figure 4-16. Microscopic structures of thin films heated at (a) 60°C for 10 mins and (b) 80°C for 10 mins, annealed at 750°C for 30 minutes. Insets show the corresponding optical microscopic images of gelated and dried glass substrates prepared using the same gelation conditions...... 120 Figure 4-17. EDS elements scan of films with various microstructures ((a) film with crystals on surface; (b) homogenous film) ...... 121 Figure 4-18. Surface microstructures of single-layer bismuth ferrite thin films prepared on STO(100) substrate by precursors with different concentrations ((a) 0.25 M; (b) 0.35 M; (c) 0.45 M) crystallized at 750°C ...... 122 Figure 4-19. XRD patterns of single-layer bismuth ferrite thin films derived from precursor with various concentrations crystallized at 750°C ...... 123 Figure 4-20. Surface ((a) and (b)) and cross sectional ((c) and (d)) microstructures of 10-layer bismuth ferrite thin films prepared on (100)STO substrate from 0.25 M and 0.35 M precursors and crystallized at 750°C ...... 124 Figure 5-1. Illustration of Chapter 5 research route ...... 129 Figure 5-2. XRD patterns BFO/STO(001) thin films derived from: (a) Bi-excess precursor annealed in air; (b) Bi-excess precursor heated in oxygen; and (c) stoichiometric precursor annealed in oxygen. The temperatures used (450 to 850°C) are indicated in the patterns and the time at temperature was 30 minutes ...... 133 Figure 5-3. (a) XRD patterns of BFO/STO(001) and BFO/LSMO/STO(001) thin film derived from stoichiometric precursor after heating at 650°C in oxygen atmosphere for 30 min; (b) Phi scan of the (110) plane of (001)-oriented BFO/LSMO/STO(001) thin film and the STO substrate ...... 137 Figure 5-4. Interface of a 40 nm BFO/STO (001) thin film showing: (a) TEM image; (b) HRTEM image; and (c) SAED diffraction pattern ...... 139

Figure 5-5. (a) TEM and (b) SAED diffraction pattern of 40 nm BFO/LSMO/STO(001) thin film ...... 141 Figure 5-6. AFM image of 40 nm BFO/LSMO/STO(001) thin film surface in a size of (a) 3 µm×3 µm and (b) 1.5 µm× 1.5 µm; Vertical PFM (c) amplitude and (d) phase images of virgin BFO domain structures in a size of 1.5 µm× 1.5 µm; Vertical PFM (e) phase image (5 µm×5 µm) of BFO /LSMO /STO(001) thin film after applying +6 V (3 µm×3 µm) and -6 V (1 µm × 1 µm) ...... 142 Figure 5-7. Room-temperature (a) Piezoresponse force spectroscopy amplitude and phase hysteresis curves and (b) Polarization hysteresis loop of 40 nm BFO/LSMO/STO (001) thin film ...... 145 Figure 6-1. Illustration of Chapter 6 research route ...... 150 Figure 6-2. DSC analysis of powder derived from stoichiometric precursor heated in air ...... Figure 6-3. XRD of stoichiometric BFO thin films prepared by various deposition- heating processes after annealing at 650°C in oxygen atmosphere for 30 minutes ...... 156 Figure 6-4. (a) Topographies, (b) PFM amplitude, and (c) phase images of BFO thin films prepared by various heat treatment sequences ((i) Sample A-90; (ii) Sample B-270; (iii) Sample C-450; (iv) Sample D-650) (each image size is 1.5 µm x 1.5 µm) ...... 157 Figure 6-5. (a)P-E loops and (b)-(e) local PFM amplitude and phase curves of bismuth ferrite thin films prepared by different deposition and heat treatment ((b) A-90, (c) B-270, (d) C-450 and (e) D-650) ...... 160 Figure 6-6. I-V curves of BFO thin films of various thicknesses (A-90, B-270, C-450, D-650) ...... 162 Figure 6-7. Schematic diagram of showing the stages of deposition, drying, pyrolysis and sintering process of 2-layer thin film for the A-90 heat-treatment sequence ...... 164 Figure 6-8. Schematic diagram of showing the stages of deposition, drying, pyrolysis and sintering process of 2-layer thin film for the B-270 heat-treatment sequence ...... 166 Figure 7-1. Illustration of Chapter 7 research route ...... 174

xvi

Figure 7-2. XRD patterns of 40nm, 70nm and 150 nm BFO/LSMO/STO(001) thin films derived from stoichiometric precursor after annealing at 650°C in an oxygen atmosphere for 30 minutes, insets show the BFO (003) peaks (left) and Phi scan of (110) peaks (right) of the three corresponding thin films ...... 180 Figure 7-3. (a) Cross-section microstructures and (b) corresponding SAED patterns of films with various thicknesses of 1-layer, 2-layer and 3-layer BFO thin films observed by TEM ...... 181 Figure 7-4. AFM (a) (3 µm × 3 µm ), (b) (1.5 µm × 1.5 µm); out-of–plane PFM (c) amplitude(1.5 µm × 1.5 µm) and (d) phase (1.5 µm × 1.5 µm) images of BiFeO3 films on LSMO/STO(001) substrate (I - 40 nm; II - 70 nm; III - 150 nm) ...... 184 Figure 7-5. Ferroelectric polarization hysteresis loops of 40 nm, 70 nm and 150 nm BFO thin films, measured at room temperature at 10 kHz; ...... 185 Figure 7-6 . Room temperature pulsed polarization of 70 nm and 150 nm BFO thin films as a function of electric field ...... 186 Figure 7-7. Room temperature pulsed polarization of 70 nm and 150 nm BFO thin films as a function of pulse width ...... 187 Figure 7-8. Pulsed polarization fatigue behaviour for the 70 nm and 150 nm BFO thin films at room temperature ...... 188 Figure 7-9. Dielectric constant and loss tangent curves of 70 nm and 150 nm BFO thin films, measured at room temperature at 1 MHz ...... 189 Figure 7-10. PFM amplitude hysteresis loops of BFO thin films measured through Au/Ti top electrodes ...... 190 Figure 7-11. Typical I-V curves of the BFO/LSMO/STO thin films with thickness of (a) 40 nm, (b) 70 nm and (c) 150 nm ...... 192 Figure 7-12. Local piezoresponse force microscope phase (a) and amplitude (b) hysteresis loops of 40 nm, 70 nm and 150 nm BFO/LSMO/STO(001) thin films ...... 194 Figure 7-13. (a) PFM image of natural domain patterns of 40 nm BFO thin film (purple regions – polarization switched upward; yellow regions – polarization switched downward); (b) Current map obtained in the same region as in (a) (bright region – higher current) ...... 196

xvii

Figure 7-14. (a) PFM image (5 µm × 5 µm) of a polarization pattern of 40 nm (I), 70 nm (II) and 150 nm (III) BFO thin film after writing by DC bias with -6 V/-9 V (3 µm × 3 µm) and +6 V/+9 V (1.5 µm × 1.5 µm ) (yellow regions – polarization switched upward; purple regions – polarization switched downward); (b) Current maps obtained in the same regions as in (a) (bright regions – higher current) ...... 197 Figure 7-15. I-V curve for two opposite polarization directions of 40 nm, 70 nm and 150 nm BFO thin films measured by CAFM ...... 198 Figure 8-1. XRD patterns of BFO/LAO thin-films after heating at various temperatures (BFO-T: BFO T phase; BFO-R: BFO R phase; O: sample stage from XRD equipment) ...... 206 Figure 8-2. AFM topographic images of BFO/LAO thin-films after heating at various temperatures ...... 206

Figure 8-3. XRD patterns of Bi0.9La0.1FeO3 / STO(001) thin-films after heating at various temperatures ...... 208 Figure A1-1. FT-IR spectra of stoichiometric raw materials and as-prepared precursors, (2000-780 cm-1) ...... 210 Figure A1-2. FTIR patterns of bismuth ferrite stoichiometric Precursor AA in 2-MOE solvent showing: (a) Precursor AA before heating and after heating for various durations at 90°C, 2000-780 cm-1; (b) 2-MOE and bismuth ferrite starting precursor before heating and after heating for 75 s (1200-950 cm-1) ...... 211 Figure A1-3. FTIR patterns of 2-MOE and metal nitrates in stoichiometric 2-MOE based precursor (Precursor AA) after heating at 90°C for 75 s (1200- 900 cm-1) ...... 212 Figure A1-4. FTIR spectra of raw materials and as-prepared precursors (2000-780 cm-1) ...... 213 Figure A1-5. FTIR analysis of stoichiometric bismuth ferrite precursor BB: (a) FTIR patterns of bismuth ferrite starting precursor before and after heating for various times at 90°C (2000-650 cm-1); (b) FTIR patterns of bismuth ferrite starting precursor before heating and after heating for 90 s (1200- 950 cm-1); (c) FTIR patterns of bismuth ferrite starting precursor before heating and after heating for 90 s (1850-1650 cm-1)...... 214

xviii

Figure A1-6. Transmission optical microscope images of thin films derived from different metal salt concentrations (0.25 M, 0.35 M, 0.45 M) after heating at various temperatures (RT, 50°C, 60°C, 70°C, 80°C, 90°C) for 10 minutes, followed by drying at room temperature in open ambient air for 12 hours ...... 217 Figure A1-7. AFM topographic images of BFO/STO(001) thin films prepared by stoichiometric precursor B with various metal salts concentrations (0.25 M, 0.35 M, 0.45 M) after crystallized at 650°C in oxygen atmosphere ...... 219 Figure A1-8. XRD patterns of BFO thin films prepared with and without preheating process ...... 220 Figure A2-1. Frequency dependency of (a) ferroelectric polarization hysteresis loop and (b) coercive field ...... 222

xix

List of Tables

Table 2-1. Summary of PZT and lead-free piezoelectric ceramics ...... 19 Table 2-2. Summary of Ferroelectric properties of BFO thin films using different deposition processes ...... 29 Table 3-1. Chemicals for precursor preparation ...... 46 Table 3-2. Substrate Materials ...... 48 Table 3-3. PLD deposition conditions of LSMO on (001)-STO substrate ...... 51 Table 3-4. Specification and application of AFM/PFM probes ...... 61 Table 4-1. Composition of precursor solutions ...... 80 Table 4-2. Experimental design for film gelation investigation ...... 83 Table 4-3. Composition of precursor solutions with various metal salts concentrations ...... 84 Table 4-4. Summary of FTIR patterns of 2-MOE, Precursor A before and after heating for 75 s ...... 90 Table 4-5. Summary of FTIR patterns of precursor B after heating for 90 s ...... 98 Table 4-6. Summary of the reaction sequence of Precursor B for the precursor preparation and gelation processes ...... 100 Table 4-7. Summary of effect of gelation heating conditions on film microstructure ...... 112 Table 4-8. Mass concentration and vapour pressure of precursor with various metal nitrate concentrations ...... 116 Table 5-1. Composition of 0.25 M precursors with and without 10% excess Bi...... 130 Table 6-1. Sample preparation on BFO/LSMO/STO(001) by various heating processes ...... 152 Table 6-2. Summary of macro and micro-scale ferroelectric properties of BFO/LSMO/STO(001) thin films ...... 161 Table 8-1. Summary of the objective and outcomes of the overall thesis ...... 204 Table A1-1. Composition of starting materials of precursor solutions ...... 210 Table A1-2. Summary of FTIR patterns of 2-MOE, stoichiometric Precursor AA before and after heating for 75 s ...... 212 Table A1-3. Summary of FTIR patterns of precursor BB after heating for 90 s ...... 215 Table A1-4. Composition of Precursor BB solutions with various metal salts concentrations ...... 215

Chapter 1.

Introduction

1.1 Motivations

Piezoelectric oxides materials have been studied intensively over the past decades because of their wide applications in transferring mechanical energy into electrical energy and vice versa. Ferroelectric materials, which show both piezoelectric and ferroelectric properties, present reversible spontaneous polarization. For example, ferroelectric materials can be used to produce the random-access memory device

(FeRAM) to achieve non-volatility and fast write performance[1, 2]. Lead zirconium titanate (PZT) has been found with extraordinary electromechanical and ferroelectric properties, making it one of the most important piezoelectric/ferroelectric oxides in various applications[3, 4]. However, owing to environmental and health issues, there is a strong motivation to find “green” Pb-free substitutions for PZT. One candidate, multiferroic material bismuth ferrite (BiFeO3, BFO) shows both a spontaneous polarization and anti-ferromagnetism at room temperature[5]. In particular, BFO thin films with epitaxial structure have been found to have excellent high remanent polarization and polarization mediated resistive switching behaviour[6], which make them particularly attractive as a new generation ferroelectric oxide[5]. Nevertheless, leakage current due to secondary phases, oxygen vacancies or poor density due to porosity often restrains the potential applications of BFO thin films[7, 8].

Therefore, to reduce the film leakage current and maximise the ferroelectric performance of BFO thin films, it is important to develop the BFO thin film deposition process to achieve dense structure, pure phase, and epitaxial crystallographic structure.

To date, the best quality epitaxial BFO thin films have been achieved usually by physical vapor deposition (PVD) methods, such as pulsed laser deposition (PLD)[6, 9, 10] or radio-frequency (RF) sputtering[11, 12], with such films typically demonstrating high spontaneous polarization[13] and resistance switching effects[6, 14, 15]. However, these methods are either time or cost consuming and are not amenable for the industrial mass manufacturing. On the other hand, the chemical solution deposition (CSD) technique is of particular industrial interest due to it being low-cost and offering accurate control of the precursor composition, as well as enabling processing ease for large-area wafers[16].

In spite of these attractions, CSD-derived BFO thin films with properties comparable to

PVD BFO remain a challenge. The typical difficulties associated with CSD in obtaining

BFO thin films with robust ferroelectric properties include high leakage current[7, 17], formation of secondary phases[7, 8], porous microstructures[16], or crystallographic defects[17]. Thus, there is still a gap between the CSD process and other film deposition processes regarding the quality and ferroelectric properties of BFO thin films.

1.2 Objectives

The main objective of the overall thesis is to develop CSD technique to be able to fabricate the high-quality epitaxial BFO thin films with robust ferroelectric properties, which are comparable with their PLD counterparts. The specific objectives are summarized as below.

(1) One main difference between CSD and PVD is that the deposition and preparation of homogenous gel films are prior to the high temperature crystallization and includes complex chemical reactions in the starting precursor when using CSD methods, while in a typical PLD process, the deposition of the precursor (often the same stoichiometry as the end material) and the high-temperature crystallization occur nearly simultaneously. Thus, the process of the homogenous gel films formation becomes extremely important in CSD process regarding the final BFO thin-film microstructure, phase composition and ferroelectric properties.

(a) In spite of the efforts and attempts from literature on the fabrication of BFO or other perovskite thin-films using CSD method, a comprehensive understanding of the chemistry occurred during gelation of initial precursor is still lacking. A systematic investigation on the gelation process will be conducted first in this thesis.

(b) Once understand what happened during gelation, the next objective is to optimize the gel-film preparation condition in order to obtain the homogeneous

BFO thin films. This study is also underdeveloped from the literatures, and it will be investigated regarding the factors of heating temperature, heating duration and precursor concentration in this thesis.

(2) Compared with polycrystalline BFO thin-film, epitaxial BFO thin films show excellent and particular properties. The key process in epitaxial structure preparation is the high-temperature crystallization process which makes the amorphous dried gel film into the crystallized BFO thin film. A systematic study of the crystallization temperature and atmosphere will be carried out to understand the evolution of BFO thin film crystallographic structure and phase composition when heating at various conditions, and thus to obtain the pure phase BFO thin films with epitaxial structure.

(3) The thickness control of CSD technique is through the multi-layer deposition.

The whole preparation process includes deposition, heating for gelation, heating for film drying, pyrolysis and crystallization. The film microstructure is different after each heating process, thus at which heating process (and how) to conduct the second deposition remains a question in literatures. Thus, it is important to understand the effect of the sequence of heating and multi-layer deposition on the multi-layer thin- film microstructure, phase composition and properties.

(4) According to the literatures, the epitaxial BFO thin films prepared by CSD technique often show poor room-temperature ferroelectric properties due to the high leakage current. This strongly impedes the application of CSD derived epitaxial BFO thin films. Thus, the last objective of this thesis is to successfully prepare epitaxial

BFO thin-films with robust room-temperature ferroelectric properties using CSD techniques.

(a) First, it is necessary to fabricate epitaxial BFO thin films with various thicknesses using the above optimized precursor, gelation condition, crystallization condition and multi-layer deposition processes.

(b) Then the thickness dependency of ferroelectric properties of the CSD derived epitaxial BFO thin films will be investigated.

1.3 Overview

The above objectives of this thesis are achieved through systematic step-wise investigations.

Chapter Two gives the in-depth literature survey on the whole thesis. It starts with the definition of ferroelectricity and piezoelectricity and their typical characteristics, followed by the introduction of their applications and the candidate materials. Then, as the research subject in this thesis, the nature of BFO is discussed regarding its phase diagram, crystallographic structure and characteristics. In the last part of this chapter, the CSD fabrication processes of BFO thin film, including precursor preparation, film deposition and crystallization, are introduced in detail.

Chapter Three introduces the experimental procedures over this thesis. It starts from the thin film and device preparation processes, which include precursor preparation, thin film deposition and heat-treatment, preparation of bottom electrodes and top electrodes. The second part of this chapter introduces the equipment and techniques used in this thesis for precursor molecule structure investigation and BFO thin film characterization

Chapter Four aims to prepare the defect-free homogenous gel and crystallized films. The first part of this chapter investigates the gelation chemistry of the non- aqueous precursor. More specifically, it studies the chemical structure change of organic molecules of precursor during gelation. The second part optimizes the heating condition for obtaining homogenous gel film. After that, the optimal precursor recipe and gelation conditions for obtaining defect-free crystallized BFO thin film is investigated in the third part of this chapter.

In Chapter Five, the thin film crystallization process is then optimized by studying the effects of precursor Bi/Fe ratio, heating temperature and heating atmosphere on phase purity and crystallographic orientation of the BFO thin films. The outcome of this chapter provides the optimal conditions to prepare single-layer BFO thin film that are pure phase and have (001)-oriented epitaxial structure.

In Chapter Six, the multi-layer BFO thin films are prepared by repetitive deposition and heat-treatment processes. The effect of the deposition-heating sequence on the microstructure and ferroelectric properties of the multi-layer thin films is studied.

In this chapter, the optimal deposition and heating processes to prepare pure phase epitaxial BFO thin film with various thicknesses were determined.

In Chapter Seven, BFO thin films of three different thicknesses are prepared.

The characteristics of these thin films, including ferroelectric, piezoelectric, dielectric properties and ferroelectric resistive switching, are investigated carefully.

Chapter Eight summarizes all the experiment results and present the major conclusions of the work. Recommended future experimental work is also discussed.

The outcome of this thesis provides an in-depth insight into the optimization of

CSD process for the preparation of epitaxial BFO thin films with robust ferroelectric properties. By this study, it is promising to fill the gap of ferroelectric properties between CSD and other developed processes and thus promote the application of BFO thin films for industrial manufacturing.

Chapter 2.

Literature Review

2.1 Piezoelectricity and Ferroelectricity

2.1.1 Piezoelectricity

2.1.1.1 Definition

The definition of piezoelectricity is introduced first for a thorough understanding of ferroelectricity. Piezoelectricity is defined as the ability of a material to develop an electric charge proportional to an external mechanical stress or to present a geometric strain (deformation) proportional to an applied voltage[18-22]. These two phenomena are described as the direct piezoelectric effect and converse piezoelectric effect, respectively, and are as shown schematically in Figure 2-1.

Figure 2-1. Illustration of direct and converse piezoelectric effect [23]

As shown in Figure 2-2, in a material having an isotropic structure, no dipole will be present under an external mechanical stress resulting in a zero net charge (P=0) across the material. In a material with an anisotropic structure, under an external force, the deviation of the centres of positive and negative charges will lead to a net charge and thus a dipole is produced (P≠0). Thus the material is polarized due to the net charge difference due to the existence of the dipoles in the material[19].

Figure 2-2. Illustration of the relationship between crystal symmetry and polarization ((a) (b) symmetric structure; (c)(d) asymmetric structure)

2.1.1.2 Piezoelectric Constant

In piezoelectric materials, the external mechanical stress is proportional to the produced net charge but it is of opposite sign for compression and tension[21]. The direct piezoelectric constant (d1) can be expressed by the relationship between dielectric displacement (D) and external stress (T), assuming a constant electric field strength (E), as:

d1=D/T (E constant) Equation 2-1

where d1 is the direct piezoelectric constant (in C/N). In the converse effect of piezoelectricity, an applied electric field strength (E) produces a proportional strain (S), specifically an expansion or contraction depending on polarity direction. The converse piezoelectric constant (d2) is expressed, assuming a constant external stress, as:

d2=S/E (T constant) Equation 2-2 where d2 is the converse piezoelectric constant (in m/V). Although they have different units, d1 and d2 are dimensionally and numerically equivalent. That is, the general

[19] piezoelectric constant is d=d1=d2 .

For piezoelectric ceramics, the stress not only affects the polarization in the stress direction, but also in the shear direction[24]. As shown in Figure 2-3, the direction of the positive polarization is usually taken as the Z axis of a right-hand orthogonal crystallographic set X, Y, Z. The X, Y, Z directions are represented as 1, 2, 3 respectively as the subscript in the equation below and shear axes are represented as 4, 5,

6, respectively, which are in planes normal to the 1, 2, 3 axes[18].

Figure 2-3. Notation of polarization axes[23]

The relationship between polarization D and applied electric field E are as follows:

D1=ε11E1+ε12 E2+ε13 E3 Equation 2-3

D2=ε21E1+ε22 E2+ε23 E3 Equation 2-4

D3=ε31E1+ε32 E2+ε33 E3 Equation 2-5

where εij gives the component of polarization along i-direction as a result of electric field applied along the j-direction. Thus, when the piezoelectric effect is taken into account, the polarization or strain of the piezoelectric materials can be expressed[18] as:

D1=ε11E1+ε12 E2+ε13 E3 +d11T1+d12T2+d13T3+d14T4+d15T5+d16T6 Equation 2-6

S1=s11T1+s12T2+s13T3+s14T4+s15T5+S16T6+d11E1+d12E2+d13E3+d14E4+d15E5+d16E6

Equation 2-7

In the above equations, the first subscript gives the electrical direction while the second subscript gives the component of mechanical strain or stress. For example, d33 indicates the relationship between the mechanical strain in the z-direction and a z- direction electric field, or between the polarization along the z-axis and an external stress along the z-direction. The individual dij will depend on the geometric symmetry of the crystal — the higher the symmetry, the fewer the number of independent dij. In

[18] case of centro-symmetrical crystal, all the dij are zero .

For polycrystalline materials, the crystal structure of the grains is usually randomly orientated, such that, even if each single grain is of piezoelectric nature, the material as a whole does not show a net piezoelectric property[18]. However, in a given material with a spontaneous polarization (Ps), piezoelectric properties can remain in the polycrystalline ceramic. These materials are called ferroelectric materials which show the property called ferroelectricity — this is examined next.

2.1.2 Ferroelectricity

2.1.2.1 Definition

Ferroelectricity is defined as a property of a material having spontaneous electric polarization which can be reversed by means of an applied electric field[18, 25, 26]. It was first discovered for a single crystal, Rochelle salt[27], and then for many other ceramics.

[25, 28] In particular, materials having perovskite structure (ABO3) , such as BiFeO3 (BFO),

BaTiO3 (BTO), PbTiO3 (PTO) and Pb(ZrxTi1-x)O3 (PZT; 0≤x≤1), are found to show excellent ferroelectric behaviour. The perovskite structure is shown in Figure 2-4. In

this structure, A site ions (Pb) are located in the corners of the lattice while oxygen ions

(O) are located in the centre of each lattice face and form octahedral interstices in which

B site ions (Ti) are located. A cubic phase shows no spontaneous dipole (Ps=0) under an external stress due to its isotropic lattice structure. However, the charge centre in a tetragonal (T) / rhombohedral (R) / monoclinic (M) crystal structure can be shifted off- centre when an external stress is applied to generate a spontaneous electric dipole (Ps≠0).

The crystal structure can transform between cubic phase and T/R/M phase at a particular temperature, called Curie temperature, for a given ferroelectric material.

[19] Figure 2-4. Illustration of the perovskite structure ABO3

Ferroelectric materials have the following characteristics: (1) there is a spontaneous polarization in the material when the applied electric field is zero and this spontaneous polarization can be reversed by an electric field in the direction opposite to the polarization direction[19, 26]; (2) spontaneous polarization only happens in a certain temperature range; it disappears when the temperature is higher than the Curie temperature (Tc), above which the phase of the material transfers into an isotropic structure, such as cubic phase, and thus lose its ferroelectricity[18]; (3) electric domains, which are defined as the areas with the same spontaneous polarization direction, can be observed in ferroelectric materials[18, 29]. All ferroelectric materials are piezoelectric, but

only some piezoelectric materials, having specific asymmetric crystal structures and can be switched by an external electric field, are ferroelectric[19].

2.1.2.2 Polarization Direction and Domain Structure

Depending on the crystal structure, spontaneous polarization in a ferroelectric material will occur in a particular direction. For example, the polarization direction in tetragonal and rhombohedral structures are in [001] and [111] directions, respectively.

Regions with the same oriented polarization are called ferroelectric domains and the boundaries between domains with different polarization directions are called domain walls[19, 30]. The domains form as a result of a thermodynamic need to minimize the electrostatic energy of depolarization field Ed and the elastic strain energy which occurs due to distortion of the unit cell structure[31, 32]. The depolarization field gives surface charges opposite to the polarization and will form whenever there is nonhomogeneous distribution of the spontaneous polarization[32].

In materials with a tetragonal perovskite structure, such as PZT and PbTiO3, there are two types of different domain walls: 90° and 180°[19, 31, 33]. They are determined by the angle between polarization directions of two domains, as shown in

Figure 2-5. Domain walls between two domains with perpendicularly oriented polarization are called 90° domain walls while those between two anti-parallel aligned domains are called 180° domain walls. In a cooling processing, the PZT/PbTiO3 crystal structure transforms from cubic structure into tetragonal structure at the Curie point and the crystal is mechanically compressed along the (100) direction. Thus, the long-axis of the unit cell CT will develop perpendicularly to the stress and polarization in the unstressed part remains parallel to the direction of the stress (short aT axis perpendicular to the stress). Both domain walls may reduce the Ed effects but only 90º domain walls can minimize the elastic energy[19].

Figure 2-5. 90° and 180° domain walls in tetragonal perovskite ferroelectric materials[19](all symbols are explained in the text)

In rhombohedral perovskite materials, such as BFO, polarization occurs in the

<111> family of directions thus giving 8 possible polarization directions with three different domain walls — 71°, 109° and 180°[34-36]. As shown in Figure 2-6, these three domain walls are parallel to the (110) plane (71° domain wall), in the (100) plane (109° domain wall), and on planes containing the [111] polar vector (180° domain wall)[34], respectively.

Figure 2-6. 71°, 109° and 180° domain walls in rhombohedral perovskite ferroelectric materials[34]

2.1.2.3 Hysteresis Loop

According to the definition of ferroelectricity, ferroelectric materials can retain a remanent polarization after removing the external electric field. Polarization-Electric field (P-E) hysteresis loops, are one of the most important defining characteristics of ferroelectric materials[2]. Figure 2-7 shows a typical ferroelectric polarization hysteresis, in which Ec, Pr and Ps are coercive electric field, remanent polarization and spontaneous polarization, respectively.

Figure 2-7. Typical hysteresis loop of the ferroelectric materials[19]

First, the virgin domains at point A are randomly aligned and thus the material shows no net polarization. This domain (or dipole) directions can be well aligned (or switched) under a high enough electric field, which is called poling process. As the poling process starts with an increasing electric field applied to the material, domains start to align towards the poling direction and the net polarization increases dramatically.

Ideally, all the domains are aligned with the electric field (point C) when saturated polarization (Psat) is obtained. The polarization increases linearly with further increase in the electric field strength. Then, as the electric field decreases back to zero (point E), only part of the domains switched back to their original directions and most domains

remain aligned. The polarization when the electric field is returned to zero is called the remanent polarization (Pr). The polarization can be returned to zero only if the electric field keeps reducing to the negative coercive field (-Ec) (point F). Then the saturated polarization will be obtained at point G when the electric field is strong enough to switch back all the domains. This switching ability from one state to another is the basis of many electronic devices, such as ferroelectric non-volatile memories.

Generally, the domain density or domain size influences the domain switching behaviour in different ways. On one hand, smaller domains have smaller depolarization energy[30] and domain switching begins with the growth of residual domains at domain wall. That is, the higher domain density can facilitate a greater number of nucleation sites for switched domain nucleation thereby reducing the coercive field. On the other hand, defects, such as oxygen vacancies or electrons, are often trapped in the domain- wall area and inhibit the domain switching by impeding domain wall movement[19] thereby increasing the coercive field.

Good ferroelectric behaviour for switching application is characterized by a material having a hysteresis loop which has an approximately square loop shape, a high remanent polarization, and a low coercive field. A square loop indicates a small difference between saturated polarization and remanent polarization thereby being able to achieve a higher Pr. Furthermore, a square loop also demonstrates a high-resistive

(low current leakage) nature of the material that it is able to hold and retain charges as a capacitor when an electric field is applied. A high remanent polarization, indicating a high polarization without presence of external electric field, is desired. It relates to a high strain and electromechanical coupling[37] effect due to the piezoelectricity. Further, a low coercive field indicates a smaller energy is needed to switch the domains

(polarization directions) in the ferroelectric material, making it easier to control in an electronic device. On the contrary, due to the loss of charge, a polarization loop with a decreasing curve at a high electric field (e.g. C to D) is often observed to be a high current leakage material which is detrimental to its ferroelectric applications[2].

In certain perovskite materials, a transition phase between T and R structures or a mixed phase with both T and R structures can exist in the material. For example, in

PZT (Pb(ZrxTi1-x)O3) materials, the phase diagram shows a morphotropic phase boundary (MPB) between rhombohedral and tetragonal phases, in which the Zr/Ti molar ratio is around 52/48 at room temperature. Extraordinary properties, including piezoelectric constant, dielectric constant and electrocmechanical coupling factor, have been obtained from the PZT with MPB composition[18]. It is found the high electromechanical properties are related to the phase transition where a monoclinic symmetry is formed. The polarization direction of the phase at MPB is not along [001] or [111] in T or R phase, but along the directions associated with a monoclinic distortion[38]. It is thought that the large piezoelectric constant and dielectric constant result from the instability of the order parameter (degree of order across the boundaries in a phase transition system) [38] and due to the complex polarization axis direction in

MPB region.

2.1.3 Applications

Ferroelectric materials can be used as memory devices because of their switchable polarization. In addition, the sensitivity of ferroelectric/pieozoelectric materials towards external electric field or external stress makes them useful as electrical-mechanical or mechanical-electrical energy transducers. Such devices have

been widely used in various aspects of our lives. Below lists several significant applications of the ferroelectric and piezoelectric materials:

(1) Ferroelectric memory devices

Using the spontaneous polarization of ferroelectric materials, new non-volatile memory devices called ferroelectric random access memory (FeRAM) have been produced and currently under intensive investigation[1]. Compared with traditional storage memory, FeRAM requires a low power usage and such devices are able to achieve a high-speed read/write and store data when power-off, which is a promising technique as a future generation memory device[2, 39].

(2) Piezoelectric devices

For the application of piezoelectric materials, one extensive application is used as SONAR (SOund Navigation And Ranging) mainly for civilian and military applications, such as measuring depth of water, detection of fish, identification of underwater obstructions, and so on[40, 41]. One of the critical components of SONAR is the piezoelectric transducer. An acoustic signal is first generated by the piezoelectric transducer and directed towards a target obstacle. The piezoelectric transducer then receives the reflected wave signals and transfers them into an electrical signal. The interval between the sending and receiving signal is a measure of the intervening distance, often one transducer is used for both transmitting and receiving with these functions being switched electronically. In addition, by transfer the energy between electric and mechanic, piezoelectric materials can also be used as the strain gauges[42, 43], accelerometers[44, 45], gasoline motor ignition and power generation[46, 47].

Depending on the particular applications, ferroelectric materials are fabricated into different configurations, namely, thin films[2, 48], thick films[48] or bulk

components[49, 50]. Bulk ceramics are relatively easy to fabricate and have a relative better mechanical strength, on which large external stress can be applied so as to produce higher electric energy. However, they also have disadvantages such as low fracture toughness, large volume and high inertia, which make it more difficult to work under high-frequency AC voltages to produce a high-frequency vibration. In addition, large bulk volumes make the bulk component unsuitable for most micro-devices used today. In comparison, thin-films, with a smaller volume, often require a smaller space, a low driving voltage and they can work under a high frequency[51], making them highly suitable for microelectronic applications.

2.1.4 Materials

Lead-based ferroelectric ceramics, such as PZT[52, 53], have been widely used for the past ~30 years due to their excellent ferroelectric/piezoelectric properties[3, 4]. As indicated in the PbZrO3-PbTiO3 phase diagram, the crystallographic phases of PZT vary with composition. It is found that the ferroelectric properties peak at the MPB region where a high remanent polarization up to ~40 µC/cm2 and excellent piezoelectric

[54, 55] properties d33 of 165-740 pC/N . In particular, in PZT thin-films, the piezoelectric constant d33 is often restrained to ~87 pm/V due to the substrate clamping effect in which the piezoelectric response of thin films is restrained by the substrate which often shows no piezoelectricity[33]. Nevertheless, high polarization has been obtained in the strained PZT thin films with a remanent polarization of ~75 µC/cm2[33]

PZT contains lead oxide, albeit in solid solution in the PZT. The toxicity of lead oxide has significant health and environment issues and lead is on the list of Restriction of Hazardous Substances Directive (RoHS)[56] which restricts hazardous materials used in electrical and electronic devices. Therefore, it is urgent and necessary to develop

lead-free alternatives having excellent piezoelectric and ferroelectric properties. Many

lead-free materials having the perovskite structure (ABO3) have been developed, such

as BiFeO3 (BFO), BaTiO3 (BTO), Ba0.5Na0.5TiO3 (BNT), K0.5Na0.5NbO3 (KNN), and so

on. The electromechanical properties and main advantages and disadvantages of these

materials are summarized in Table 2-1.

Table 2-1. Summary of PZT and lead-free piezoelectric ceramics

2 Tc (°C) d33 (pC/N) Pr (µC/cm ) Advantages Disadvantages

 High d33 [55] [33, 54]  Toxicity of Pb PZT 280 165-740 ~40-75  Inexpensive  Volatility of Pb  High dielectric constant

 Stable composition  Low Tc [57] [18, 58] BTO ~120 350 ~26-70  High d33  High sintering  High dielectric constant temperature  Good piezoelectric  Volatility of Na and K KNN ~420 ~80-160[59] ~10-20 [59] property  Low density  High Tc  Difficult in poling ~820-  High T  High leakage current BFO ~15-60[34] ~6.1-130[13, 34] c 850  Multiferroic material  Volatility of Bi

BTO, is a well-established piezoelectric material and has been studied for many

years. It has a relatively high sintering temperature, but also has a relatively stable

composition at high temperature such that the loss of any metal element is very low

thereby giving a good chemical stoichiometry and control of the phase composition.

The remanent polarization of bulk BT crystals is ~26 µC/cm2[18] and can be increased to

as high as ~70 µC/cm2 in a BTO thin film sample[58]. In addition, BTO has been

[57] reported to achieve a high d33 value to 350 pC/N . However, a low Curie temperature

of merely 120°C has largely limited its application to devices which operate at room

temperature.

KNN, as another popular piezoelectric material, has a comparatively higher Tc

of 420°C. However, compared with PZT and BTO, it has a lower d33 of ~80-160 pC/N

and lower relative dielectric constant of ~230-475[59]. In addition, KNN shows a weak

ferroelectric property with low remanent polarization of mere ~10-20 µC/cm2 [59].

Further, the volatility of alkali and difficulty in poling process[60] also limit the application and development of KNN materials.

BFO is a multiferroic material, showing simultaneous ferroelectric and antiferromagnetic characteristics[5]. Its high Curie temperature (~820-850ºC) and Neel temperature (~350-380°C)[61, 62] (above which an antiferromagnetic or ferromagnetic material transfers into paramagnetic) makes it an excellent candidate for high temperature applications[63]. Bulk BFO shows a poor remanent polarization of up to only ~6.1 µC/cm2, but it is found that the ferroelectric properties are significantly enhanced in the thin film[5]. A high spontaneous polarization of up to 130 µC/cm2[13] and piezoelectric constant of ~15-60 pm/V[34] have been reported recently for some

BFO thin films. In addition, resistive switching and photovoltaic behaviour[64] have also been observed for BFO. These properties combined with its high ferroelectric polarization and multifunctional properties, make BFO a very promising prospect for various ferroelectric device applications. In this thesis, a major aim was made to prepare

BFO thin-films with robust ferroelectric properties using chemical solution deposition.

The characteristics of BFO are examined in detail in the next section.

2.2 Bismuth Ferrite

2.2.1 Crystallographic Structure

Bulk BFO, as a material with typical perovskite structure, has been shown to present a rhombohedral (R) symmetry with a space group of R3c[65]. The lattice parameters of room temperature bulk BFO are a=b=c=0.396 nm, α=β=γ=89.3° in a pseudocubic lattice. Figure 2-8 shows the crystal structure of BFO. The spontaneous

polarization of rhombohedral BFO is along [111] direction and Bi, Fe and O are displaced relative to one another along this axis[5, 65].

[65] Figure 2-8. Structure of BiFeO3

When a thin film presents a well-defined crystallographic orientation in a specific crystallographic orientation relative to the substrate, it is called epitaxial thin film. In a BFO thin-film with epitaxial structure, tensile and compressive strain can be produced between BFO and substrate due to lattice mismatch[66, 67], which is defined as the in-plane lattice parameter difference between the thin film and substrate. The lattice mismatch (δ ) is quantified by Equation 2-8.

δ=(dβ-dα)/dα Equation 2-8

where dα and dβ are the unstressed inter-planar spacing of substrate (α) and film (β), respectively.

With this compressive (tensile) strain effect, crystalline structure of thin film may change with an in-plane contraction (elongation) and an out-of plane elongation

(contraction). It has been found this epitaxial strain can be used to stabilize a tetragonal- like (T)[68] or monoclinic (M)[69] polymorph of BFO thin film[67]. The polarization

direction can also be changed from [111] in R phase to [001] in T phase. If the strain is controlled to a certain point then both the T (or M) and R phase can co-exist thus establishing the MPB in the thin film which ultimately can produce higher piezoelectric response[67, 70].

In addition to the effect of lattice parameter mismatch, the epitaxial strain can also be controlled via film thickness[67, 70]. Below a critical thickness, films are fully strained and are identified as monoclinic (M) or tetragonal-like(T) phase. This critical thickness for stabilized M or T phase, increase with increasing of substrate-film misfit [71]. The selection of substrate will be discussed in section 2.3.2.2 and the effect of thickness on film crystallographic structure and properties will be investigated in this thesis.

2.2.2 Phase Composition

The phase diagram of Bi2O3-Fe2O3 is shown in Figure 2-9. The pure phase BFO can be synthesized only when the Bi:Fe ratio is equal to 1. Secondary phases, such as

Bi2Fe4O9, Bi25FeO39 or Bi2O3, can easily form due to the nonstoichiometric starting reactants (Bi: Fe ≠ 1) and/or decomposition of BFO at high temperature (typically over

700ºC).

[34] Figure 2-9. Phase diagram of Bi2O3-Fe2O3

2.2.2.1 Secondary Phases and Decomposition

The synthesis kinetics of BFO and its secondary phases are sensitive to the temperature. Morozev et al[72] studied the synthesis kinetics of BFO by mixing equimolar amounts of Bi2O3 and Fe2O3 followed by heating. During the heating process in air, Bi25FeO40 first forms at 250°C and reaches the maximum amount at ~500°C.

BFO and Bi2Fe4O9 start to form at 550°C and 580°C, respectively. However, the formation rate of Bi2Fe4O9 is more than 5 times faster than that of BFO when the temperature is above 650°C. Thus the secondary phases, such as Bi25FeO40 and

[72] Bi2Fe4O9 are always stand on the way of pure phase BFO preparation .

On the other hand, at temperatures over ~700ºC[17], pure BFO tends to decompose into its component metal oxides[34], as shown in Equation 2-9.

2BiFeO3  Fe2O3 + Bi2O3 Equation 2-9

[73] Or into Bi2Fe4O9 and Bi25FeO39 in Equation 2-10.

49BiFeO3  12Bi2Fe4O9 + Bi25FeO39 Equation 2-10

[17] Others have reported that the BFO may also decompose to Bi2Fe4O9 + liquid .

Thus, it is important to restrict the synthesis of BFO below 700°C to avoid the formation of secondary phases. The pure phase of bulk BFO is often sintered at

~800°C[74], but this pure BFO is very unstable during sintering process and usually rapidly decomposes into secondary phases[34]. Compared with bulk material, the synthesis of pure phase BFO thin film is relatively easier because of its lower sintering temperature. Nevertheless, the pure phase BFO thin film preparation is still a challenge due to the complex phase and microstructural evolution during heating.

2.2.2.2 Pure Phase BFO Preparation

The potential applications of BFO thin-film are greatly limited due to current leakage, which is mainly caused by structural defects (e.g. porosity) and/or

[75] [76] nonstoichiometric second phases such as Bi2O3, Bi2Fe4O9 or γ-Fe2O3 . Significant effort has been made to achieve pure-phase BFO thin-films. However, the small synthesis window of pure phase BFO preparation has been demonstrated regarding the annealing temperature, initial Bi/Fe ratio of the starting materials, and annealing atmosphere.

One main concern in BFO preparation is the volatilization of Bi and Bi2O3 due to their low melting points (271.5°C and 817°C respectively) and boiling point (1564°C and 1890°C respectively). To obtain the pure BFO, the most common method is to use

5%-10% mol excess Bi[76, 77] in the initial system before high-temperature annealing in order to compensate the loss of Bi and thus avoid the formation of secondary phases.

However, precise control of the amount of Bi is still difficult because the volatility rate of Bi varies with heating atmosphere, oxygen partial pressure, heating temperature, etc.

Bea et al.[77] first reported the sensitive nature of this system in a series of systematic investigations on BFO thin films deposited by pulsed laser deposition (PLD).

The system is not only sensitive to the temperature but also to the oxygen partial pressure. As a secondary phase, Bi2O3 is formed either at a relatively low temperature

(~200°C) or high oxygen partial pressure. On the contrary, Fe2O3 often forms under heating conditions involving relatively high temperatures or at low partial pressures because a large fraction of the Bi evaporates under this condition[77].

Tyhold et al.[17] and Hardy et al.[78] investigated the effect of crystallization conditions on the material composition synthesized using chemical solution deposition

(CSD) method. Although they use different chemical solution system (metal alkoxides and 2-MOE were used to prepare non-aqueous precursor by Tyhold et al.[17]; metal alkoxides, metal nitrates and citric acid were used to prepare aqueous precursor by

Hardy et al.[78]), similar results were obtained. When heating in oxygen atmosphere, the reduction of Bi3+ to Bi metal was observed at around 200°C[17] [78]. This was attributed to the decomposition and pyrolysis of polymers when the decomposition gas creates a local oxygen deficient atmosphere and thus contributes to the reduction of Bi3+. After the decomposition and pyrolysis, Bi metal was found oxidized into Bi25FeO40 at ~275°C and bismuth oxide at 365°C, followed by the formation of BFO form at 460°C[17] [78].

2.2.2.3 Oxygen Vacancies in BFO Thin Films

The main reason of poor ferroelectric properties of BFO is leakage current. In particular, as discussed in section 2.2.2.1, the presence of secondary phases in BFO may increase the leakage current. In addition, oxygen vacancies in the film bulk or interface can also cause the high conductivity[79, 80]. The formation of oxygen vacancies can be explained using Kroger-Vink Notation in the following formula.

1 2퐹푒 + 3푂 → 2퐹푒′ + 푉.. + 푂 + 2푂 Equation 2-11 퐹푒 푂 퐹푒 푂 2 2 푂

1 2퐹푒 + 푂 → 2퐹푒′ + 푉.. + 푂 Equation 2-12 퐹푒 푂 퐹푒 푂 2 2

Valence of Fe changes from +3 to +2 according to the Equation 2-11 and 2-12 while oxygen vacancies are formed in this process. The equilibrium constant of

Equation 2-11 can be expressed as:

1 ′ 2 .. [퐹푒퐹푒] ∙[푉푂]∙[푃푂 ]2 퐾 = 2 Equation 2-13 [퐹푒퐹푒]∙[푂푂]

′ .. since [퐹푒퐹푒] = [푂푂] = 1 and [퐹푒퐹푒] = 2[푉푂],

1 − [퐹푒′ ] = 2[푉..] = 푐표푛푠푡푎푛푡 ∙ 푃 6 Equation 2-14 퐹푒 푂 푂2

Conductivity (δ) can be expressed by the formula of

δ=n(Ze)μ0 Equation 2-15 where, n = number of charge carriers per unit volume Z = valence charge e =unit charge

μ0 = mobility 1 − From Equation 2-14 and 2-15, it can be derived that δ=constant • 푃 6 . It 푂2 explains the relationship between the oxygen partial pressure and oxygen vacancy derived ionic conductivity of the BFO, in particular, that a higher oxygen partial pressure can decrease the conductivity.

In summary, the BFO thin film synthesis process and its properties are quite sensitive to the heating temperature and atmosphere. To obtain the pure phase BFO with

low leakage current, it is important to identify the optimal heating condition (heating temperature, oxygen partial pressure /annealing atmosphere).

2.2.3 Characteristics of BFO

2.2.3.1 Microstructure and Domain Structure

(1) Microstructure

The microstructure of BFO thin films can vary significantly with the deposition method/condition, annealing temperature/atmosphere, thickness, substrate and film crystallographic orientation. Polycrystalline BFO thin-films typically show equiaxed grains with clearly defined grain boundaries. On the contrary, grains of oriented epitaxial thin-films are anisotropic, often aligned in a particular direction (e.g. perpendicular to the substrate surface)[81]. In epitaxial thin-films, the out-of-plane grain size often increases with increasing film thickness[82, 83], whereas the in-plane grain size varies significantly with the conditions of the deposition and heat-treatment process[83], such as the heating temperature and oxygen pressure[84].

Generally, a smooth and dense microstructure free from pores or cracking is required for a ferroelectric thin-film having good ferroelectric performance. However, this is a challenge for chemical solution deposition-derived thin-films since defects such as pores and cracking are often formed during pyrolysis due to the removal of relatively large amount of polymers. In addition, precipitates or agglomerates formed during precursor preparation and film deposition would make a thin film a rough surface as well as an inhomogeneous chemical composition and usually, an inhomogeneous phase composition. In this thesis, the optimization of chemical solution deposition derived thin film microstructure will be studied.

(2) Domain structure

As discussed in section 2.1.2.2, ferroelectric domains are regions in which electric dipoles are polarized in the same direction. There are two general different domain structures found in BFO thin films: stripe domains and fractal domains[34].

Stripe domains, which show an ordered and oriented stripe shape, are often found in the epitaxial thin films. Domain width of the stripe domains follows the formula given in

Equation 2-16, where w is domain width, A is a constant and d is the film thickness.

w=A•d1/2 Equation 2-16

As the thickness reduces below a critical value, stripe domains tend to break up into fractal domains[69], which are characterized by irregular mosaic shapes of the domain areas[34]. These fractal domains have also been prepared in epitaxial BFO thin films[85] with thicknesses below approximately 100 nm.

2.2.3.2 Ferroelectric Properties of BFO Thin-Films

Ferroelectric property with high remanent polarization of ~50-60 µC/cm2 has been observed from PLD-derived epitaxial BFO thin film[5] at room temperature. Yun et al. successfully prepared 300 nm polycrystalline BFO thin-film with a tetragonal

[86] structure (c/a=1.018) by PLD on Pt/TiO2/SiO2/Si . Their BFO thin-films achieved a

2 high remanent ferroelectric polarization of Pr=100 µC/cm at room temperature and

2 Pr=146 µC/cm at 90 K. The theoretical switched polarization of BFO is expected to be able to achieve ~300±20 µC/cm2 in a pure T phase BFO according to linear extrapolation[13]. Later, high remanent ferroelectric polarization of ~130 ± 5 µC/cm2 was then found at room temperature in the PLD-derived epitaxial BFO thin film, which has a mixed-phase with high T-phase concentration[13]. The ferroelectric properties of BFO thin films prepared by PLD or CSD have been summarized in Table 2-2. Using

chemical solution deposition, polycrystalline BFO thin films having high ferroelectric polarization have been prepared[87-89]. Although epitaxial BFO thin films have been well prepared by chemical solution deposition[90], few has reported high ferroelectric polarization value or piezoelectric properties at room temperature, which largely restrict the application of the chemical route processing.

Table 2-2. Summary of Ferroelectric properties of BFO thin films using different deposition processes

2 Substrate Thickness Process 2Pr (µC/cm ) 2Ec (kV/cm) Ref. (nm) STO(001) 200 PLD ~110(@ R.T.) ~300(@ R.T.) [5] Pt/TiO2/SiO2/Si 300 PLD 200 (@ R.T.) [86] 292 (@ 90 K) 240 (@90 K) LAO (001) 160 PLD ~230 (@ R.T.) 1200(@ R.T.) [13] BFO STO(001) unknown CSD 100 (@ 80 K) ~600(@ 80 K) [90] Pt/TiO2/SiO2/Si 250 CSD 180 (@ R.T.) ~1000(@ R.T.) [88]

Pt/TiO2/SiO2/Si 400 CSD 200(@ 80 K) 1600(@ 80 K) [89]

The out-of-plane piezoresponse of the thin film often decreases with the reduction of film thickness because of the epitaxial strain and substrate clamping effect[5], which makes it difficult to benefit from the mixed phase or MPB effect.

[5, 34] Piezoelectric constant d33 of epitaxial BFO thin film is around 15-70 pm/V which is relatively small compared with other perovskite ferroelectric materials[34].

2.2.3.3 Ferroelectric Resistive Switching Behaviour

As an electric field is applied to BFO, the resistivity of the BFO can be suddenly changed. This resistance change is non-volatile and reversible in BFO material. This phenomenon is called resistive switching, in which resistance random access memory

(ReRAM) devices are produced. In particular, resistive switching controlled by ferroelectric polarization is called ferroelectric resistive switching[14], which is used to

prepare ferroelectric RAM (FeRAM) device. FeRAM has advantages in high-speed writing, high-density memory, and non-volatile storage of electronic RAM devices[91].

The switchable diode effect of bulk BFO material was first reported by Choi et al.[64] and the ferroelectric resistive switching (RS) was subsequently observed in epitaxial BFO thin films[6, 14, 91]. There are two principal mechanisms to explain this phenomenon: a thermal chemical mechanism and a valence-change mechanism. In the former one, the resistive switching is through the conductive filament which is formed in an insulating matrix through electroforming process, and then the local redox reaction in a joule heating caused filament induces the resistive switching behaviour[91]. The latter mechanism mainly results from the electromigration of oxygen vacancies which changes the valence states of cations in a metal oxide (e.g. Equation 2-13), leading to the change of conductivity[91].

In particular, ferroelectric resistive switching can be categorized into two types: a tunnel electroresistance effect and a Schottky diode[91]. The tunnel electroresistance dominated resistive switching behaviour often increases in a thinner films (thickness typically of a few nanometres), but on the other hand, the large leakage current in the ultrathin film makes it impossible to obtain a high polarization. That is, in the tunnelling current mechanism dominated ultrathin films, the better resistive switching is established by sacrificing the ferroelectric properties. A ferroelectric Schottky diode often occurs due to presence of a Schottky barrier at the electrode-thin film interface and it has been found in the thick BFO thin films with thickness above

100 nm[6]. This Schottky barrier is highly resistive at low voltages, which contributes to a high resistive state (HRS). The resistive decreases when the voltage is high enough to exceed the barrier, which leads to a low resistive state (LRS). A ferroelectric Schottky

diode can be changed or controlled by polarization switching [92-94] during the voltage sweep.

Recently, the resistive switching behaviour and Schottky diode effect of thicker

BFO films have been reported by Jiang et al.[6] and Wang et al.[14]. Although the diode current density increases with the decrease of film thickness due to the leakage behaviour, the high ferroelectric polarization and resistive switching are able to coexist in thin films with a thickness ranging from 200 to 500 nm[6] [14]. The elimination of the requirement of ultrathin film on ferroelectric resistive switching and diode effect makes it feasible to use chemical solution deposition technique to prepare multi-functional

BFO thin films.

2.3 Fabrication Processes for Bismuth Ferrite Thin-Film

2.3.1 Film Fabrication Processing

Many fabrication techniques have been developed for preparing ceramic films.

High quality epitaxial BFO thin films are made usually by PLD, as shown in Section

2.2.3.2 and Table 2-2. In comparison, CSD offers an efficient and economic method for film deposition and is under intensive investigate for industrial manufacturing.

(1) PLD

PLD is a well-established thin film deposition technique. A high-power pulsed laser beam is used in a vacuum chamber to strike the target material thus vaporizing it.

Some of the resultant plume is then deposited as a thin film on the substrate[95]. PLD can be used to fabricate ceramic oxide thin films or high-temperature super conductive thin films using the high energy lasers. Moreover, it is easy to prepare multi-layer thin films and is easy to control the thickness of the film. Often, thin films prepared by PLD

present pore-free with smooth surface. Thus, PLD is widely used to prepare ferroelectric epitaxial thin films. In particular, BFO epitaxial thin films prepared by PLD have demonstrated robust ferroelectric properties due to their excellent crystallographic epitaxial structure and nearly crack-free and pore-free microstructure. For example,

BFO ultrathin films[96, 97], BFO thin films with giant polarization[13, 86] and BFO thin films with resistive switching and diode effects have all been well prepared by PLD processes[6, 91].

(2) CSD

Chemical solution deposition (CSD), commonly known as sol-gel processing, was developed and widely used in preparation of ceramics, glasses and composites since the publication of Dislich in 1971[98], and is especially well developed for the commercial manufacture of spinel and perovskite materials[99]. Films made by sol-gel processing typically have good uniformity, especially important for when dopants are introduced in the system. In addition, it is also a relatively simple fabrication technique for low-cost thin films of large area[100]. Unlike PLD and sputtering, the deposition process of CSD is usually carried out at around room temperature and subsequently heated up to a higher temperature for pyrolysis and crystallization. This differences makes it relatively easy to control the deposition process and microstructure of thin film before crystallization[16]. On the other hand, it tends to make the crystallization process less controllable. In particular, preparing epitaxial BFO thin film by CSD is challenging work among worldwide researchers owing to the process having a small synthesis window, as discussed in Section 2.2. In this thesis, this method will be used to prepare epitaxial BFO thin films, with the specific aim to prepare BFO thin-films with high ferroelectric polarization and resistive switching behaviour.

2.3.2 Thin Film Preparation Process using CSD Technique

The fabrication process starts from a precursor chemical solution. A typical precursor includes metal compound (alkoxides, metal chlorides or metal nitrate), a solvent, and a chelation agent. Under certain heating conditions, the precursor solution will transform into a gel[16] which typically, is in a cross-linked polymeric state. The general process of sol-gel processing is illustrated in Figure 2-10.

Figure 2-10. Flow chart of sol-gel process [101]

There are two different sol-gel processes according to their different starting materials. Process (a) describes the process of forming from solution a suspension of colloidal particles (1-1000 nm), called a sol, from which, a particulate gel is formed.

Process (b) describes the process starting of forming from solution, a sol which consists of polymer chains without discrete particles from which a polymeric gel is formed[101].

Thus the polymeric gel at a molecular level can be obtained via the solution sol-gel route (process b), and this has been widely used to fabricate highly homogenous films.

2.3.2.1 Precursor Preparation and Gelation Process

(1) Traditional aqueous sol-gel process

Traditional aqueous sol-gel processes are based on metal alkoxides in water as the starting materials, and which follow a route of hydrolysis-condensation-gelation, have been extensively studied. For this process, several factors, notably temperature, pH value and aging time, exert substantial influence upon the stability of precursor, the gelation process and the quality of the gel film.

Metal alkoxides are widely used to prepare aqueous starting precursors because of their high reactivity. For some highly electropositive metals, the metal alkoxide can be simply considered as the reaction product of a metal and alcohol as shown in

Equation 2-17:

 1  M  zROH  M (OR) z   zH 2 (g) Equation 2-17  2 

where M is the metal ions, R is the alkyl (e.g. –CH3, -CH2-CH3) and z is the valence of metal ions. Usually, the metal alkoxide is purchased and is the starting precursor for the sol-gel process.

(a) Hydrolysis

Hydrolysis is a chemical reaction of water whose molecules are split into hydrogen cations (H+) and hydroxide anions (OH-), as shown in Equation 2-18.

  H 2O  OH  H Equation 2-18

When a metal alkoxide is present in solution, hydrolysis of the metal alkoxide occurs as shown in Equation 2-19.

M (OR) z  H 2O  M (OH )(OR) z1  ROH Equation 2-19

It is important to avoid the formation of insoluble precipitates which may

[16] suppress the polymerization process . Take Al(OR)3 for example, if excess water is involved in the reaction (Equation 2-20), the precipitation may occur as follows:

Al(OR)3  2H 2O  AlO(OH )(s)  3ROH Equation 2-20

(b) Condensation

Condensation is a process of partially hydrolyzed molecules linking and the exclusion of small molecules, such as water or alcohol. For example,

(RO) z1 MOM (OR) z1 is formed in the condensation reaction of metal alkoxide:

M (OH )(OR)  M (OR)  (RO) MOM (OR)  ROH z1 z z1 z1 Equation 2-21 2M (OH )(OR) z1  (RO) z1 MOM (OR) z1 H 2O

In order to keep the condensation forward reactions proceeding, the solution needs to be slightly heated so as to keep removing the H2O by vaporization.

The hydrolysis and condensation reactions can lead to the networking of polymers into a cross-linked system, named a gel. Gels have mechanical properties ranging from soft to hard, weak to strong, and exhibit high viscosity in the steady state

[102]. Following from Equation 2-21, the gelation process of metal alkoxide can be described as:

M (OH )(OR) z1  (RO) z1 MOM (OR) z1  (RO) z1 MOM (OR) z2 OM (OR) z1  ROH

n[M (OH )(OR) z1 ]  (RO)[(RO) z2 MO]n1 M (OR) z2 (OH )  (n 1)ROH

Equation 2-22

Since the concentration of H+ or OH- can control the direction and rate of

Equation 2-19 by adjusting the concentration of product ROH, pH value plays a vital role in controlling the process in forms of the effect of acid/base additive or the amount of water, and both factors need to be considered in the process.

Besides metal alkoxide and water, other additives, such as a stabilizer or chelation agent, are often used to stabilize the solution or to lower the formation temperature of the pure gel phase[75]. The most often used stabilizer is acetylacetone[75, 103-105].

(2) Non-aqueous sol-gel processes

Metal alkoxides are relatively expensive and, as such, can compromise the economic advantage of the chemical solution deposition technique. Instead, precursor derived from metal salts (typically metal nitrates or metal chloride) and organic solvents

(such as 2-methoxyethanol (2-MOE)[103, 106-109] or ethylene glycol[75]) have been developed recently to prepare various perovskite materials.

In particular, metal nitrates, 2-methoxyethanol and acetic anhydride are chosen as the starting materials to prepare bismuth ferrite precursors in this work. In addition to their economic advantage over metal alkoxides, metal nitrates, have a relatively low decomposition temperature which is considered to contribute to the reduction of carbon contamination[17] following pyrolysis. Substituting for water as the solvent, 2- methoxyethanol (2-MOE) is widely used due to it offering good solubility of various of starting reagents[110] (including metal nitrates) as well as it having suitable viscosity and surface tension for spin-coating deposition processes[111], which, in turn, can potentially reduce the occurrence of defects, such as pores or agglomeration. The latter often occurs during deposition due to relatively low or high precursor viscosities or high surface

tension (poor surface wettability). In addition, acetic anhydride is often used to adjust the solution viscosity through the formation of oligomeric structures during film deposition[110].

The development of 2-MOE based precursor system for sol-gel processing was first reported by Gurkovich et. al[112] and Budd et al[113]. They found that alkoxide and acetate in the starting materials convert into 2-methoxyethoxide or 2-methoxyethoxy acetate species. In Schwartz’s review[114], using lead carboxylate and 2-MOE as the starting precursors, it is found that Pb(OOCCH3)(OCH2CH2OCH3) was formed. The 2- methoxyethoxides or 2-methoxyethoxy acetate are then involved in the polymerization process during subsequent gelation and diluted water is often used to facilitate the proceeding of hydrolysis, condensation and gelation[17, 114]. Eventhough, how the gelation process works in non-aqueous BFO precursor using Bi/Fe nitrates, 2-MOE and acetic anhydride is still not clear. Understanding of the role of the organic solvent during gelation process and the corresponding chemistry is one of the major aims of this thesis.

2.3.2.2 Film Deposition

(1) Substrate selection

The substrate plays a key role in the microstructure and properties of epitaxial thin-films because it not only determines the crystallographic orientation of an as- prepared thin-film, but also influences the crystalline structure of the thin-film.

Generally, for perovskite thin-films, similar perovskite materials are employed as the substrates for epitaxial film growth. One important consideration is the crystallographic lattice mismatch between substrate and film, which has been explained in section 2.2.2

(see Equation 2-10). On one hand, the smaller the mismatch is, the better the epitaxial

structure of thin-film grows because a lower energy barrier is present for nucleation in film-substrate interface compared with when there is a higher lattice mis-match between film and substrate crystal[114]. On the other hand, using the substrates with a larger mismatch with BFO generate compressive/tensile strain thereby leading to the formation of T (or M) and R mixed phases in the BFO thin-film thus giving a MPB effect thereby significantly improving BFO thin film properties[115].

Schlom et al.[116] summarized the lattice parameters of some common perovskite materials of thin films and substrates, as shown in Figure 2-11. The most common

[5, 81] substrate used for BFO thin-film is SrTiO3 and this has a pseudocubic a-axis lattice parameter of 0.391 nm and small lattice mismatch of 1.3% with BFO, leading to a weak

[117, 118] [70, 71] compressive strain. Other substrates, such as LaAlO3 , LaSrAlO4 or

[119] [115, 119] NdScO3 are often used to prepare highly strained BFO thin films .

Figure 2-11. Pseudotetragonal or pseudocubic a-axis lattice constants (in angstroms) of some ferroelectric perovskites of current interest[116]

(2) Buffer layer selection

A buffer layer, which is the layer between substrate and thin film, often acts as the bottom electrode (for electromechanical applications) in BFO thin films. It is required that the buffer layer material also have perovskite structure, has close lattice

parameters with both the substrate and BFO, as well as having excellent conductivity.

[120, 121] [122, 123] SrRuO3 (SRO) and La0.67Sr0.33MnO3 (LSMO) are the two most common buffer-layer materials used for epitaxial BFO thin films. SRO has a lattice parameter of

~0.392-0.393 nm which is close to and intermediate of BFO (~0.396 nm) and STO

(~0.391 nm). This small lattice mismatch (~0.75-1% for SRO-BFO and ~0.25-0.5% for

STO-SRO) makes it easier for the epitaxial structural growth of both SRO and BFO on

STO substrate. LSMO, which has a lattice parameter of ~0.386 nm and a lattice mismatch of ~2.5% with BFO, can produce a compressive strain in BFO thin film. In addition, LSMO also shows good magnetic properties[124, 125], which is likely to enhance the magnetic properties of BFO thin film by their coupling effect.

(3) Deposition technique

Spin coating is one of the most common methods to prepare thin films by chemical solution deposition[16, 126]. The deposition process is summarized in Figure 2-

12, which includes four main steps. (1) Deposition: the precursor solution is dropped on the surface of substrate; (2) Spin-up: the substrate starts to spin and the solution flows fast outward of the substrate by the centrifugal force; (3) Spin-off: the substrate spins at a constant velocity and excess precursor is removed by the centrifugal force as the droplet; (4) Evaporation: In the last part of the spinning process, the film becomes thinner due to evaporation[16].

Figure 2-12. Spin-coating process [16]

The thickness of the deposited film prior to evaporation strongly depends on the precursor viscosity (η), precursor density (ρ), spin-coating angular velocity (ω) and spin coating time (t). The relationship between film thickness and above factors can be described as[16]:

1 2 2 2 h(t)  h0 /(1 4 h0 t / 3) Equation 2-23

where, h0 = initial film thickness before spin-up

Equation 2-23 indicates that the thickness of the as-deposited film increases with increasing precursor viscosity, decreasing solution density, decreasing spin-coating time, and decreasing spin-coating velocity. However, thicker films derived from a large precursor viscosity or slow spin coating velocity may introduce defects such as cracking or inhomogeneous film after drying.

For example, a single-layer BFO thin-film is often prepared by spin coating at

3000 rpm to 5000 rpm for 15 to 30 seconds [8, 88, 127-129] to obtain various thickness. The thickness for a single deposition is often thinner than 50 nm [88, 128]. To prepare the

films of larger thicknesses and to accurately control the thickness, the spin-coating deposition can be repeated several times with drying (or pyrolysis/annealing) being done between depositions[90] [127, 130].

2.3.2.3 Heat-Treatment

After spin coating deposition, heat-treatment processes are used to transform the as-deposited thin films into the final crystalline BFO thin film. These heat-treatment processes include drying, pyrolysis, sintering, and crystallization.

(1) Drying

The drying process is to remove any excess solvent from the film after spin- coating and gelation. Drying is done typically by heating at a low temperature (less than

300°C) to facilitate the removal of solvent by evaporation (in ambient or in a drying chamber). Generally, the drying can be divided into three main stages: (1) Film shrinkage which is proportional to the volume of evaporated liquid and the rate of evaporation; (2) The rate of evaporation decreases sharply as the loss of solvent in gel and it is sensitive to the temperature and vapour pressure; (3) Evaporation happens inside the material, the rate of evaporation becomes very slow and it is less sensitive to the temperature or vapour pressure[16]. The film will experience a large shrinkage during stage 1 and stage 2. However, compared to bulk material, thin films have a comparatively large surface area, so the evaporation process for a thin film is often easier and faster. In addition, with the decreasing film thickness, the shrinkage direction of a thin film will be more towards out-of-plane rather than in-plane because of the restraint or clamping effect from substrate, which can reduce the chance of cracking or pores during drying.

The drying temperature depends on the precursor solvent composition, solvent vapour pressure, and the metal alkoxides or metal salts concentrations. For BFO thin- films prepared from metal nitrates and 2-MOE, a drying temperature of approximately

250°C[89, 90, 131] is typically used because it is high enough to remove all excess solvent from the film[17].

(2) Pyrolysis

A pyrolysis step is needed to decompose and oxidise the polymers in the gel film thus converting them into a metal oxide. Differential scanning calorimetry (DSC) technology is often used to investigate the decomposition process and determine the optimal pyrolysis temperature[17, 78] for sol-gel films. Depending on the precursor system, pyrolysis temperatures various reported in the literature range from

350°C[89, 129, 132] to 450°C[88, 133]. After pyrolysis, a film is usually amorphous and subsequent sintering and nucleation-and-growth processes (crystallization)[114] are employed to produce the final film.

(3) Sintering

Ceramic thin films can be heat-treated by one or two distinct methods, namely conventional heating or rapid thermal processing. In conventional heating, the material is heated up slowly to a certain temperature and then kept at this temperature for a certain duration. Rapid thermal processing (RTP) is an increasingly popular heating process which heats films to a high temperature in a short time (typically several seconds) and then holds at this temperature for a set time.

(4) Crystallization

During heat treatments, the nucleation of the amorphous thin films occurs first during the heating process, followed by the grain growth. The driving force of

crystallization is determined by the free energy difference between the amorphous and crystallized material, and the temperature below thin-film material melting point[114], as shown in the Figure 2-13.

Figure 2-13. Schematic diagram of the free energies of a sol-gel derived amorphous film compared with the corresponding equilibrium liquid and a crystal

Nucleation in a thin-film takes place either in the bulk body or the substrate-film interface. Nucleation in the bulk body is called homogeneous nucleation and leads typically to the randomly oriented grains and polycrystalline thin films. Nucleation at the substrate-film interface is called heterogeneous nucleation and often gives nucleation and growth of the oriented columnar grain which match the substrate lattice structure. Generally, nucleation barrier for heterogeneous nucleation is lower than that for homogeneous nucleation. According to the diagram in Figure 2-13, a higher heating temperature could decrease the driving force, which makes it more difficult for the homogenous nucleation due to its higher barrier energy. Therefore, by rapid heating the thin film to a high temperature, the crystallization of BFO thin films could be postponed

to this high temperature, making the heterogeneous nucleation at interface be preferred and thus the epitaxial grain growth is promoted[17, 114].

2.4 Summary

This literature review introduces the fundamental concepts of ferroelectricity of perovskite oxides as well as explains the important applications of these materials. In particular, bismuth ferrite thin films have attracted intensive attention due to its multifunctional properties. The structural and functional properties of bismuth ferrite thin films, and the synthesis strategy employed in this thesis have been highlighted in this literature survey. The key points are summarized as below.

 Perovskite ferroelectric materials, showing both ferroelectricity and

piezoelectricity, are important materials, which can be applied in a wide variety of

applications, such as micro-electromechanical systems, sensors, voltage generator

and memory devices.

 BFO, as a lead-free multiferroic material, exhibits a rhombohedral perovskite

structure. Its high Curie temperature and Neel temperature makes it an excellent

ferroelectric candidate material for high-temperature applications.

 In particular, BFO thin films with epitaxial structure have been demonstrated to

have high polarization and ferroelectric resistive switching. These properties make

it a promising memory device with high-speed writing performance, high-density

memory and non-volatile storage. However, the small synthesis window for pure

phase BFO and its relatively high leakage current have limited the application of

BFO thin films for such devices

 The CSD process is of particular significance from an industrial perspective in

ferroelectric thin-film fabrication due to its high efficiency and low cost. However,

it is still under developed in BFO epitaxial thin film preparation because of the

challenges of Bi/Fe stoichiometry control and obtainment of phase-pure epitaxial

films.

This thesis aims to optimize the CSD process for preparation of epitaxial BFO thin film with robust ferroelectric properties. The preparation process and characterization techniques will be discussed in next chapter.

Chapter 3.

Experimental Procedures

The thesis comprises the optimization of the CSD preparation process for bismuth ferrite thin films (including precursor solution preparation, thin film deposition and thin film crystallization), study of bismuth ferrite gelation chemistry, and electromechanical characterization of CSD bismuth ferrite thin films. This chapter outlines the basic methodologies used to make the thin films (divided into precursor preparation, film deposition, and high-temperature heat treatment), the fabrication route to prepare BFO devices for electromechanical characterization, and the techniques used to characterise the films.

3.1 Bismuth Ferrite Thin film Preparation Process

3.1.1 Precursor Preparation

The specifications of the chemicals used to prepare the precursor solutions for film deposition are listed in Table 3-1. Both bismuth nitrate and iron nitrate were stored in sealed containers in a dry and ventilated cupboard so as to avoid absorption of water into the nitrates.

Table 3-1. Chemicals for precursor preparation

Chemical Name Chemical Formula Specification Supplier

Bismuth nitrate pentahydrate Bi(NO3)3·5H2O ACS 98% Alfa Aesar, UK

Iron nitrate nonahydrate Fe(NO3)3·9H2O ACS 98% Alfa Aesar, UK

2-Methoxyethonal (2-MOE) C3H8O2 ACS 99.3% Alfa Aesar, UK

Acetic anhydride (CH3CO)2O ACS 97% Alfa Aesar, UK

The bismuth nitrate and iron nitrate starting materials provide the source of metal Bi and Fe. The 2-methoxyethanol (2-MOE) is the solvent into which the Bi and

Fe nitrates are dissolved and, relevant to consideration later of the molecular changes in the precursor solution, has the structural formula of CH3-O-CH2-CH2-OH. Acetic anhydride acts as the chelating agent in the solution. The basic preparation route is shown schematically in Figure 3-1. Bismuth nitrate and iron nitrate were added first in requisite amounts to 2-MOE. After stirring for 30 mins at room temperature, acetic anhydride was added under constant stirring to the solution. After stirring for a further

60 mins at room temperature, the concentrations of Bi and Fe were adjusted to the final desired level by the addition of 2-MOE. The whole process was performed in an ambient atmosphere at room temperature.

Bi(NO3)3·5H2O Stirring for Stirring for 30 mins 60 mins

Fe(NO3)3·9H2O Solution 1 Solution 2

Stirring 2-MOE for 120 2-MOE Acetic anhydride mins

Precursor to prepare BFO thin film

Figure 3-1. Illustration of BFO precursor preparation

3.1.2 Film Deposition

Bismuth ferrite gel film was prepared by dropping a small amount of precursor solution onto a substrate followed by spin coating. Relevant details of this process are described below.

3.1.2.1 Substrates

In this thesis, silicate glass substrates were used to study first the chemical mechanism underlying the film gelation process and to ascertain the optimal preparation conditions needed for the formation of homogenous bismuth ferrite gel films. Since these initial studies were focussed primarily on the gel film and were, to the most part, independent of substrate, they did not technically need specific perovskite substrates.

Moreover, perovskite substrates are significantly expensive and a large number of samples were made in the work. Strontium titanate was used to prepare bismuth ferrite thin film for optimization of the crystallization heat-treatment process and the subsequent detail electromechanical characterization. Table 3-2 lists the substrates used in this thesis along with their specification and supplier information. Before deposition, all substrates were cleaned by ultrasonication in iso-propanol, followed by drying with compressed air.

Table 3-2. Substrate Materials

Material Name Chemical Formula Specifications Supplier Silicate Glass N/A* 75 mm x 25 mm x 1.5 mm Sail, China Strontium 5 mm x 5 mm x 0.5 mm (001)-SrTiO (STO) Daheng, China Titanate 3 one side polish, non-stepped Strontium 5 mm x 5 mm x 0.5 mm (001)-SrTiO (STO) Shinkosha, Japan Titanate 3 one side polish, stepped

*chemical composition (wt%) ~72.0% SiO2, 14.5% Na2O, 7.0% CaO, 4.0% MgO, 1.7% Al2O3, 0.3% K2O, <0.1% Fe2O3

3.1.2.2 Spin Coating Process

The principle of spin coating deposition was explained in detail in Chapter 2

(see Section 2.3.2.2). A digitally-controlled spin coater (WS-650-23, Laurell

Technologies Corporation, Pennsylvania, USA) was used in this thesis to prepare bismuth ferrite thin films. The spin coater utilises a vacuum specimen stage to secure the substrate for spinning as well as a programmable control system for spin velocity and duration (these being used to prepare thin films of different thicknesses). After deposition, the sample was removed from the spin coater to a hot plate to dry at a controlled (low) temperature.

3.1.3 High-Temperature Heat Treatment

Crystallization of bismuth ferrite commences at approximately 250°C (see

Section 2.2.2.1) with the final crystallised and sintered BFO phase(s) being obtained by about 650–700°C. In this thesis, the bismuth ferrite thin films were heated by either conventional heating (using a muffle furnace) or rapid thermal processing (using a tube furnace). The thermal cycles used for these heating processes are detailed below.

For conventional heating, the sample was placed into the muffle furnace and heated at a rate of 5°C/min to the desired temperature followed by a hold at this temperature for 30 minutes before then cooling the sample at a rate of 5°C/min back to room temperature.

For rapid thermal processing, the sample was placed onto a small refractory holder. By means of a heat resistant metal rod, the refractory holder was transferred to the hot zone of the tube furnace (the latter already preheated to the desired temperature) thereby rapidly heating the sample to the desired temperature. The heating rate for this process is estimated to be approximately 50-100°C/s via thermal couple measurement.

The sample was then at the particular temperature for 30 minutes then quenched to room temperature by rapidly withdrawing the refractory holder from the hot zone and into the cold end of the tube. Depending upon the experiment, the rapid heating process was conducted in an atmosphere of ambient air or 100% oxygen. For the latter, the tube furnace was pre-flushed by oxygen gas for 60 minutes with a gas flow rate of ~450 ml/min. A gas flow-rate of ~100 ml/min was used during the entire heating process to maintain an oxygen atmosphere of approximately 1 atm in the furnace tube.

3.1.4 Device Preparation

Electrodes are essential components for ferroelectric thin film devices, for electromechanical property measurement and actual device operation The electrode configuration used in this work is shown schematically in Figure 3-2. The bottom electrode was prepared by PLD and top electrodes were prepared by photolithography and metal evaporation. These processes are detailed below.

Figure 3-2. Schematics of BFO thin film device

3.1.4.1 Preparation of Bottom Electrode

Lanthanum strontium manganite (La0.67Sr0.33MnO3, LSMO) was used as the bottom electrode for BFO thin film because of their high electrical conductivity, perovskite structure and close lattice parameter with BFO and STO[134, 135] (see Section

2.3.2.2). The bottom electrode was prepared on the STO substrate by PLD (NEOCERA

Inc, US) prior to deposition of the BFO thin film. The PLD conditions used to deposit the bottom electrode materials on (001) oriented STO substrate are listed in Table 3-3.

The conditions were optimised and the thickness of bottom electrode was controlled to around 20 - 30 nm by using 15 000 - 25 000 pulses for the deposition.

Table 3-3. PLD deposition conditions of LSMO on (001)-STO substrate

Target LSMO Deposition temperature (°C) 700 Deposition oxygen pressure (mTorr) 100 Deposition frequency (Hz) 3 Cooling rate (°C/min) 20 Cooling oxygen pressure (Torr) 450

3.1.4.2 Preparation of Top Electrode

Gold coating was selected as the top electrode material because of its high electrical conductivity and ease of preparation. The top electrodes were prepared on the fabricated

BFO thin film using a UV photolithography technique followed by a metal thermal evaporation deposition. This combination of techniques was used because it is a well- demonstrated, reliable route for preparing firm electrode coatings of small size and controlled pattern. The techniques are detailed below.

(1) Photolithography

Photolithography is a routine technique to prepare micro-sized patterns on thin films and substrates[136]. As shown in Figure 3-3, first a layer of photoresist is deposited on the film surface. A photoresist is an organic material that can become insoluble (negative photoresist) or more soluble (positive photoresist) in a particular solution after exposure to

UV light[136, 137]. The photoresist is then exposed to UV light in a desired pattern (achieved using a mask) and with the exposed areas (for a positive photoresist) or unexposed areas

(for a negative photoresist) being subsequently removed by dissolution in a developer solution.

Figure 3-3. Schematic diagram of the photolithographic process[137]

Prior to the photolithography, the BFO thin film was first cleaned with acetone and subsequently with isopropanol. A layer of negative photoresist (nLOF2020) was then deposited onto the surface of the BFO thin film by spin coating at 3000 rpm for 30 s. The photoresist-coated BFO thin film was then heated on a hot plate at 110°C for 1 minute to remove excess photoresist solvent.

A mask aligner (Quintel Q6000, Neutronix Quintel, US) having a resolution of 1

µm was used for the UV exposure process to prepare the micro size patterns on the film.

The thin film was put on the stage of exposure aligner. In general, the exposure time depends on the reflectivity of the substrate and the roughness of the thin-film; in this work, the optimized exposure time was 5 seconds for BFO thin film on STO. A mask consisting of a regular pattern of squares, each being 23 µm × 23 µm, was applied to the top of the thin film by the mask aligner. The unmasked areas of photoresist on the top of thin-film were then exposed to intense UV light (10 mW/cm2) for 5 seconds. The sample was then heated on a hot plate at 110°C for 1 minute to help crosslink the negative photoresist so as

to improve the adherence of the photoresist on the film surface during subsequent development. The areas of negative photoresist which were not exposed to the UV light

(i.e., the squares) were dissolved by rinsing the thin film in developer (AZ826, TMAH 2.38% in H2O) for 90 seconds. The films were then cleaned by deionized water and dried using high-purity compressed air.

(2) Metal thermal evaporation deposition

Metal thermal evaporation deposition is a common method to prepare metal coatings on substrates. A source metal (e.g. Au, Ti) is heated in a vacuum chamber at a very low pressure (typically ~5×10-6 Torr) to cause vaporization of the metal. The metal vapour then condenses onto the surface of the target substrate positioned above the source material thus producing a metal coating on the substrate. The process is illustrated in Figure

3-4.

Figure 3-4. Schematics of the metal thermal evaporation process

A thermal evaporator (Kurt J. Lesker, Pittsburgh, US) was used in this thesis to prepare the top electrodes of the BFO thin films. A 5-10 nm titanium coating (using high- purity titanium wire as the source) was deposited first in order to strengthen the adhesion of

the gold electrode coating to the thin film. Using high-purity gold wire as the source, the gold coating was then deposited at a thickness of ~60 nm.

(3) Photoresist Lift-off

Obviously, only the areas of film surface free of photoresist (i.e., the squares) were coated in the metal thermal evaporation process; for the remaining areas, the metal electrode coating was deposited on the photoresist. The last step of the top electrode preparation is to remove the photoresist and the unwanted gold coating above it. In this work, this was done by soaking the sample in N-methyl-pyrrolidone (NMP) at 80°C for

30 minutes to dissolve the photoresist. After soaking and removal of unwanted gold coating, the samples were cleaned by iso-propanol followed by drying with compressed air.

3.2 Analytical Equipment and Methods

3.2.1 Chemical Analysis

3.2.1.1 Nuclear Magnetic Resonance (NMR) Spectroscopy

NMR spectroscopy is a powerful technique for the investigation of organic molecules to confirm the identity of a substance based on the substance’s highly unique and analytically tractable spectra [138]. A typical measurement consists of two main steps:

(1) A constant magnetic field H0 is applied to the substance and the polarization of the magnetic nuclear spins in this field. (2) An electro-magnetic wave, such as radio frequency pulse, is then used to perturb the polarization of the nuclear spins. 1H and 13C are two most commonly used nuclei for this measurement. In this measurement, the nuclei in a magnetic field absorb and re-emit electromagnetic radiation which is at a specific resonance frequency corresponding to the isotope. This resonance frequency,

absorption energy and signal intensity are proportional to the magnetic field strength [139].

In the NMR spectrum, protons do not show the same resonant signals at the same frequency because of the varying electronic environments. Thus, the signals are often reported relative to a reference signal and the operation giving a locator number is called “chemical shift” with unit of parts per million (ppm)[140]. In general, chemical shifts for protons are highly predictable and they provide information about molecular structure of the molecules [141]. In the 1H-NMR spectrum of organic molecules, the specific chemical shifts are representative of the main functional groups (such as CH3,

CH2, OH, CH, etc.). Besides the peak shift location, the shapes and area of the peaks also give information about the chemical structure. Generally, the area of the peak is proportional to the quantity of the protons in the group. The same principle applies to analysis of 13C, but the chemical shifts for the heavier nuclei are often strongly influenced by other factors, such as excited states thus rendering identification less definitive.

Nuclear magnetic resonance instrument (NMR, Avance III 400; Bruker,

Germany) is employed in this thesis with a frequency of 400 MHz. The liquid solution to be analysed is put in a thin-walled glass sample tube. A capillary column with deuterated dimethyl sulfoxide (DSMO) solution is then inserted in NMR glass tube to protonate the solution. All the measurement was conducted at room temperature.

3.2.1.2 Fourier Transform Infrared Spectroscopy

Fourier transform infrared spectroscopy (FTIR) is commonly used to analyse molecular structure of organic materials through the spectrum of absorption, emission, and photoconductivity of thin-films, solids or liquid samples [142, 143]. This absorption

spectrum provides information of the extent of light absorption by the sample at each wavelength by shining a light beam containing many frequencies though the sample.

This technique is widely used in sol-gel and other chemical solution based processes to characterize the synthesis chemistry at different steps in the reaction sequence [144, 145].

Infrared radiation is on the low energy side of visible spectrum, with a wavelength range from 2500 to 16000 nm. Vibrational excitation of covalently bonded atoms or groups in molecules can be induced by the photon energies (wavelengths) absorbed from the infrared radiation. The covalent bonds can be stretched or bent under the vibrational excitation thus absorbing the energy of the infrared radiation which inducing those vibrations. FTIR spectrometer produces, a unique absorption spectrum of a measured organic material as a function of wavenumber or frequency, thus identifying characteristic covalently bonded atoms or groups[143].

Generally, the stretching bonds of the functional groups correspond to higher frequencies in the spectrum than their bending counterparts. The absorption band region between 4000 to 1500 cm-1 is called the group frequency region and this gives the stretching bond vibration spectrums of diatomic groups. The region of smaller wavenumber between 1500 to 400 cm-1 is commonly called the finger print region because it includes a very complicated series of absorptions and the unique patterns of specific molecules can often be found in it[146].

FTIR spectrometer (PerkinElmer Spotlight 400, US) was employed in this thesis to study the functional groups in organic molecules of bismuth ferrite precursor and gel materials. The device can either be used for general analysis or for selected zone by using the build-in microscopy, with a measurement range from 4000 to 650 cm-1.

3.2.2 Phase Composition and Microstructure analysis

X-ray diffraction (XRD) and electron microscopy (including scanning electron microscopy and transmission electron microscopy) techniques are used in this thesis to investigate the phase composition, microstructure and crystal structures of BFO thin films.

3.2.2.1 X-Ray Diffraction (XRD) Analysis

XRD is one of the most common and powerful techniques to study the phase composition and crystal structure of the thin films according to Bragg’s Law:

nλ = 2dsinθ Equation 3-1

In this thesis, X’pert Materials Research Diffractometer (PANalytical,

Netherlands) was used for the thin-film phase composition and crystal structure analysis.

The thin-film sample was fixed on a glass slide (75 cm × 25 cm) using liquid glue. XRD analysis of the glass substrates with dried liquid glue was performed before thin film measurement to make sure that no peaks present as the background from either glass slide or liquid glue. Then the thin-film sample along with the glass slide was fixed on the XRD stage. Before measurement, alignments of 2-theta, omega, phi, psi angles, and x/y(in-plane) and z position(out-of-plane) of thin-film were performed in order to accurately position the X-ray beam on the thin-film and maximize the receiving intensity. After alignment, the XRD diffraction patterns of thin-films were acquired using Cu Kα radiation at 45 kV/40 mA over a 2θ angular range of 15–75° at a scanning rate of 5°(2θ)/min.

XRD phi-scan was used to exam the crystallographic structure relationship between the as-received BFO thin film and the substrate. The film was tilt for 45° chi

angle to obtain the X-ray diffraction pattern for the (110) plane phi-scan in a range of

360° at a scanning rate of 60°/min.

3.2.2.2 Scanning Electron Microscopy (SEM)

SEM is widely used for material microstructure analysis. The images from SEM are produced by scanning the sample using a focused electron beam, from which various signals including the information of the sample surface morphology or composition can be produced. These signals are then received by the detector and transformed into images with high resolution (<1 um). Generally, two main electron singles are often used for SEM: secondary electrons (SE) to acquire surface topography information and back-scattered electrons to acquire information of both topography and composition.

Nova NanoSEM230 FESEM (FEI, USA) was used to study the microstructures of the thin film surface. Before SEM analysis, the thin film was coated by a thin chromium conductive layer. Secondary electron source was used to acquire the topography microstructure of thin-films with a high magnification up to 100 000×.

3.2.2.3 Transmission Electron Microscopy (TEM)

TEM technique is used in to achieve a significantly high resolution up to atom level for a better observation of material microstructure and crystal structure. An ultrathin sample is required for the TEM analysis, so that the electron beam is able to transmit through the specimen. The microstructure image and film crystal structure information (crystal diffraction patterns) can be obtained from the interactions of electrons with samples. In particular, bright-field imaging, high-resolution TEM and selected-area electron diffraction techniques are used in this thesis.

(1) Bright field imaging: Microstructural and morphological information are acquired from the transmitted electrons passing through the thin sample. In this mode, the regions with no sample appear bright and contrast can be observed from the regions of sample.

(2) High-resolution TEM (HRTEM): HRTEM is used to acquire the periodic (crystal) structural information of the samples in an atomic scale level. It is often used to investigate the crystal structure in the film/substrate interface or phase boundary (i.e.,

T/R phase boundary) of thin-films.

(3) Selected-area electron diffraction (SAED): This technique is often used in the

TEM to acquire the crystal diffraction patterns of localized region. This is a powerful tool to study the crystal structure or crystal defect of thin-films.

TEM CM200 (Philips, Netherlands) and HRTEM (2200, JEOL, Japan) were used in this thesis for TEM analysis of BFO thin film. BFO/STO(001) and

BFO/LSMO/STO(001) films cross-section samples of approximately 100 nm thick were prepared using focused ion beam (FIB, Auriga, Zeiss, Germany) or manual tripod polish.

The cross section bright field microstructure images and selective area electron diffraction (SAED) were obtained.

By using SEM and TEM, Energy dispersive X-ray spectroscopy (EDS) can be used to identify the element of samples by analysing the energies of characteristic X- rays interacting with regions on the sample surface.

3.2.3 Scanning Probe Microscopy

3.2.3.1 Atomic Force Microscopy (AFM)

AFM is a technique which provides a high-resolution surface microstructural characterisation. A typical AFM device consists of a cantilever with a sharp probe to

scan the sample surface, a laser source, and a quadrant photodiode detector. The laser is focussed to a spot on the top surface of cantilever and it is reflected to the quadrant photodiode detector. During the scanning, the tip is loaded on the sample surface and an elastic force is produced between the probe and the sample according to Hooke’s law[147]. This force leads to a deflection as the probe traverses the surface topography.

The deflection is causes a change in the angle at which the laser spot is reflected off the cantilever and this is detected by the quadrant photodiode detector which, in turn, converts the optical information into digital information. Using this detector, the vertical deflection can be calculated from [(A+B)-(C+D)]/(ABCD) and lateral deflection can be defined as [(B+D)-(A+C)]/(ABCD). The AFM imaging principle is shown in Figure 3-

5(a) and the photodiode detector principle is shown in Figure 3-5(b).

(a) (b)

Figure 3-5. Schematic diagrams depicting (a) the optical lever detection system used in AFM [148] and (b) the photodiode detector

The key component in the AFM measurement is the probe or tip since it has an important role in the AFM image quality. The size of the tip is often of a few nanometres, making it possible to acquire the high-resolution topographical images of microstructure. Silicon is the main material used to prepare AFM cantilever, but various coatings, such as Pt, Au and diamond are often used for various applications, which will be discussed in detail in PFM and CAFM sections. Table 3-4 lists the probes used in

this thesis and their specific applications. For example, AFM is used for acquire film surface topography, PFM and hysteresis loop are used to study the ferroelectric domain structure and domain switching behaviours of thin-film; CAFM is used to study the electrical conductivity of thin films.

Table 3-4. Specification and application of AFM/PFM probes

Resonance Force constant Coating Supplier Application Frequency 2 N/m 70 kHz Ti/Ir Asylum Research PFM, CAFM, hysteresis loop 3 N/m 75 kHz Cr/Pt Budget sensor AFM, PFM, hysteresis loop 80 N/m 400 kHz Diamond Nanoworld PFM, CAFM

Selection of the force constant and resonance frequency of probes is important as it strongly determines scan mode and various applications. There are two scanning modes: Tapping mode and contact mode.

(1) Contact mode

In contact mode AFM, the tip constantly contacts with the sample surface during scanning. Thus, the force between probe and sample surface is repulsive and a constant separation distance is maintained by applying a constant deflection load on the cantilever. Low stiffness cantilevers with low constant force are often used to strengthen the deflection signal.

(2) Tapping mode

Using this mode, driving amplitude of a small sinusoidal voltage is used to oscillate cantilever to oscillate up and down near its resonance frequency. The image is then produced from the intermittent contacts of the tip with sample surface. This mild contact is thought to lessen the damage on the sample surface compared with contact mode. Stiffer tips with higher constant force and higher resonance frequency are often used for tapping mode to prevent the tips becoming "stuck" on the sample surface.

Asylum Cypher AFM instrument (Asylum Research, US) was used in this thesis to observe the surface topography of BFO thin films using tapping mode. Figure 3-6 shows a typical AFM image of a bismuth ferrite thin film, which clearly shows the surface morphology of the thin film in a size of 5 µm × 5 µm. The imaging size can be changed from approximately 100 nm × 100 nm to 20 µm × 20 µm.

Figure 3-6. AFM image of bismuth ferrite thin film with secondary phase (tapping mode)

3.2.3.2 Piezoresponse Force Microscopy (PFM)

PFM is a variant of AFM which is used to image the domain structure of ferroelectric or piezoelectric materials[149]. Different from AFM, tips with conductive coating is required for the PFM measurement and an alternating current (AC) bias is applied to the probe tip. During the scanning, the AC bias excites deformation of ferroelectric/piezoelectric samples by means of the converse piezoelectric effect. The phase and amplitude of the AC response bias of the sample are then demodulated and calculated through a lock-in amplifier. Figure 3-7 shows the principle of PFM measurement.

Figure 3-7. Principle of PFM measurement[150]

Asylum Cypher AFM instrument was used for all PFM measurements in this thesis. Generally, film morphology, phase image and amplitude image can be obtained simultaneously during the scanning. Figure 3-8 shows a typical PFM amplitude and phase image measured by the Asylum Cypher instrument for a polycrystalline BFO thin film prepared in this work.

(a) (b)

Figure 3-8. PFM amplitude (a) and phase (b) image of polycrystalline BFO thin film on Pt/Ti/SiO2/Si substrate

The PFM amplitude indicated in Figure 3-8(a) represents the piezoelectric deformation of BFO thin film under the external voltage. Instead, piezoresponse direction (domain/polarization direction) of the thin-film is given by the PFM phase angle image in Figure 3-8(b). The PFM amplitude is usually zero in the domain wall where is there is no piezoresponse. As explained below, PFM measurements can be made in four distinct configurations to measure different ferroelectric/piezoelectric/ conductivity characteristics of thin films.

(1) Vertical and Lateral PFM (VPFM and LPFM)

In a ferroelectric material, under an external bias, the piezoelectric effect can be observed in both out-of-plane and in-plane directions which give rise to two distinct PFM imaging techniques of vertical PFM (VPFM) and lateral PFM (LPFM). Accordingly,

VPFM and LPFM techniques can be used to detect the deformations in these two directions.

The signal capture and deformation calculation have been explained in section 3.2.2.1.

The Asylum PFM instrument is capable of obtaining the VPFM and LPFM deformations in one scan. The drive frequency of LPFM is around 1 MHz and VPFM is around 300 kHz, varying with different tips. Figure 3-9 shows the typical VPFM and

LPFM images obtained for a PbTiO3 thin film.

(a) (b)

Figure 3-9. Amplitude(a)(b) and phase(c)(d) images of VPFM(a)(c) and LPFM(b)(d) of PbTiO3 thin film

(2) Dual AC Resonance Tracking PFM (DART-PFM) technique

In the traditional PFM measurement, a single AC resonance is used to track the piezoresponse signal. However, as the electromechanical displacement is often as low as a few picometers, the noise-to-signal ratio is comparatively high. The solution of this problem can be either using a large AC voltage or other amplification technique. Using a large voltage easily increases the small response but it also comprises the probe life and may damage the sample as well. To avoid this problem, a technique called Dual

AC Resonance Tracking PFM (DART-PFM) was developed to reduce the noise.

DART-PFM employs the difference between two amplitudes derived from two drive frequencies as the input feedback, rather than using the phase to track the frequency feedback in the traditional PFM measurement, as shown in Figure 3-10. By using

DART, the resonant frequency can be kept stable by controlling the amplitude difference into zero.

Figure 3-10. Principle of Dual AC resonance tracking PFM [151]

(3) DC writing and domain switching

To study the piezoelectric nature and electromechanical characteristics of ferroelectric/piezoelectric samples, domain switching under a high external DC bias

(higher than coercive voltage) can be made using PFM technique. The DC bias is applied to the sample surface by the probe scanning a defined area. Figure 3-11shows a square area of 3 µm × 3µm written on a BFO thin film using a positive DC bias, followed by writing a square area of 1 µm × 1µm using negative DC bias. The domains were first switched downwards in the 3 µm × 3µm square and switched back upwards in the 1 µm × 1 µm square. This operation demonstrates the nature of ferroelectricity and ability of domains to be locally switched by the external DC bias.

(a) (b)

Figure 3-11. Domain switching by writing with DC bias ((a) amplitude and (b) phase images of (001)BFO thin films )

(4) PFM amplitude and phase loop measurement

PFM amplitude and phase loops can be obtained through Asylum PFM by applying a triangle square wave forme bias on the sample. The waveform is shown in

Figure 3-12(a). It includes steps with a ramping maximum value. Two sets of loops are acquired during the measurement. One is acquired when the bias is on, which is called

“applied loop” whereas the other one called “remnant loop” is captured when the bias is

off. An example of a remnant loop obtained for BFO/LSMO/STO(001) is shown in

Figure 3-12(b).

(a)

(b)

Figure 3-12. (a)Waveform for PFM spectroscopy hysteresis loop [152] and (b) PFM phase and amplitude hysteresis loops of BFO/LSMO/STO(001) thin films

3.2.3.3 Conductive Atomic Force Microscopy (C-AFM)

Conductive atomic force microscopy (CAFM) is another variant of AFM which is often used to study the electrical conductivity of a sample. A probe with a

conductive coating is used, allowing current to flow through the tip to the sample.

Figure 3-13 shows the principle of CAFM.

Figure 3-13. Principle of CAFM measurement[153]

Using the Asylum AFM instrument in ORCA mode, a sample bias is applied through bottom electrode to the tip to form an electrical circuit. This enables current mapping and I-V curve measurement to be done on the sample – these are explained next.

(1) Current mapping

Current mapping is the principal imaging mode for CAFM. Similar to AFM and

PFM imaging, a conductive map is acquired by scanning the probe on the sample surface in a small area. During the scan, a DC bias is applied to the sample from bottom electrode with the electrons that conduct through the sample are collected by the tip.

This enables the conductivity of every individual point within the scanning area to be determined thus giving a current map of the area. Figure 3-14 shows an area of a BFO thin film measured by: (a) AFM for surface topography; (b) PFM for phase image; and

higher conductivity. As the current mapping measurement is often conducted under contact mode, the conductive coating on the probe is easy to wear off, which may affect the accuracy of the measurement. To solve this problem, tips with diamond coatings are often used to obtain high quality current mapping images.

(a) (b) (c)

Figure 3-14. (a) Topography, (b) PFM phase and (c) current map acquired from the same region (1.5 µm × 1.5 µm) of BFO/LSMO/STO thin film

(2) I-V curve measurement

I-V curve can be obtained in the spectroscopic mode of CAFM. In this mode, the tip is load on a selected point and a triangular sweep voltage bias is applied through the sample to the tip. The current information is then obtained as a function of the voltage, as shown in Figure 3-15.

Figure 3-15. I-V curve of BFO thin film measured by CAFM

The combined measurements of AFM, PFM and CAFM make it possible to study for thin films the direct relationship between surface topography, domain structure, and material conductivity. In addition, the spectroscopic mode of CAFM enable the study of ferroelectric or electrical characteristics at specific points on the sample surface, such as on a domain wall or grain boundary, thereby enabling the electromechanical behaviour of local features to be elucidated.

3.2.4 Ferroelectric Test System

In this thesis, a ferroelectric testing system (Radiant Technologies; US) equipped with a probe workstation was used for the global ferroelectric polarization measurements. Top electrode pads on the film surface and a bottom electrode (the buffer layer between the film and the substrate) were needed for the global ferroelectric property measurement. The sample is loaded on the probe station and two sharp probes are used to connect the top and bottom electrodes of the samples. Typically, a voltage is applied as the driving voltage to the top electrode through the film and to the bottom electrode. Typically, two different measurements using different drive voltage functions used to obtain the ferroelectric polarization characteristics. These are explained below.

3.2.4.1 Ferroelectric Polarization Hysteresis Loop Measurement

Hysteresis loops of ferroelectric polarization (polarisation vs voltage) are usually determined using a triangular drive voltage function as shown in Figure 3-16(a). The variables of testing frequency, sample thickness and top electrode area are inputted in to the testing program enabling the ferroelectric loop to be determined. A typical loop obtained for BFO/LSMO/STO(001) thin film is shown in Figure 3-16(b).

(a)

(b)

Figure 3-16. (a) Triangular drive voltage sweep for ferroelectric polarization measurement and (b) example of ferroelectric polarization hysteresis loop obtained for BFO/ LSMO/ STO thin films

3.2.4.2 Positive up negative down (PUND)

As shown in Figure 3-17, Positive Up Negative Down (PUND) applications of the drive voltage can be made in which the polarization is produced by the positive and negative voltage pulses. This enables the switched polarization (P*) and non-switched polarization (P^) to be measured. In addition, the pulsed remanent polarization (Pr), which is less affected by the leakage current and non-linear dielectric effects[5], can be calculated by the following formula:

P*-P^ =ΔP= ~2Pr Equation 3-2

Figure 3-17. Pulses drive voltage for PUND measurement[154]

3.2.5 Macro Scale I-V Tester

Macro scale (global) I-V curves or leakage current curves can be obtained by applying a voltage through the top and bottom electrode and measuring the amount of current that flows through the film. Unlike the I-V characteristic acquired through

CAFM, the current is obtained from the top electrode capacitance through the conductive probe, rather than from the sample directly.

In this thesis, a Keithley 2400 SourceMeter (Keithley Instruments, US) was used to acquire the I-V curves of BFO thin films. The same probe workstation used for the ferroelectric property measurement was used here and the voltage is again applied through the top electrode to the bottom electrode. Using this method, the thin film resistance or leakage current can be measured by applying a sweep voltage. In particular, to study the resistive switching behaviour, the high resistance state (HRS) and low resistance state (LRS) can be obtained by applying the sweep voltage from negative to positive and then back to negative.

3.2.6 Impedance Analysis

The measurement of capacitance and dissipation of the BFO thin films made in this work was done using a precision impedance analyzer (Agilent 4294A, US). It is a powerful instrument to study the electrical properties of both bulk materials and thin films. As a macroscale tester, the probe workstation is used in the measurement, connecting with the top and bottom electrodes of thin film samples. Capacitance data is collected under CpD mode (parallel capacitance-dissipation). The capacitance and dissipation data can be collected as a function of frequency or voltage.

3.3 Summary

This chapter briefly outlines the sample preparation process and principal characterisation techniques used in this thesis. The methodologies for the precursor preparation, film deposition, substrate preparation, spin coating process and high- temperature heat treatment are given in detail. The process of electrode deposition

(involving lithography and metal thermal evaporation) for BFO device preparation for electrical characterisation is described. The analytical methods used to characterise coating precursors and fabricated BFO thin films are:

• NMR for investigation of the molecular functional groups in the CSD precursor

solutions.

• FTIR for investigation of the organic species in the CSD precursor solutions and

their subsequent changes during heating and gelation.

• XRD for investigation of film phase composition and crystal structure.

• SEM for investigation of film surface morphology.

• TEM for investigation of microstructure, crystal structure of film cross section.

• AFM for high resolution imaging of surface topography and microstructure of BFO

thin films

• PFM for high resolution imaging of the domain structure in BFO thin films including:

(1) VPFM and LPFM for determination of out-of-plane and in-plane piezoelectric responses (amplitude and phase);

(2) DARTPFM for determination of piezoelectric responses (amplitude and phase);

(3) DC writing and domain switching to determine local domain behaviour;

(4) PFM amplitude and phase loop determination for local hysteresis behaviour.

• CAFM for high resolution imaging of the electrical conductivity of BFO thin films as

well as determination of their local I-V behaviour.

• Ferroelectric testing system to determine:

(1) Global polarisation hysteresis loops;

(2) Switched polarisation, unswitched polarisation and remanent polarisation (by PUND voltage pulsing)

• I-V tester to determine macro scale I-V behaviour

• Impedance analysis to determine the global capacitance and charge dissipation of the

BFO films.

The above techniques were employed in subsequent chapter in the preparation and characterisation of BFO thin films. Unless specifically stated, the techniques were used were those described in this chapter.

Chapter 4.

Investigation of Gelation Chemistry and Gelation Conditions for Preparation of BFO Thin-Films

4.1 Introduction

The fabrication of BFO thin films by chemical solution deposition (CSD) typically includes 4 steps: (1) starting precursor preparation; (2) deposition of precursor solution onto substrate; (3) low-temperature drying and (4) high-temperature crystallization and sintering[130]. Optimization of each of these steps are of critical importance in obtaining BFO thin films of appropriate composition and microstructure and, ultimately, good electromechanical properties. However, defects can occur during each particular step, resulting in a film having a sub-optimal microstructure and properties. Commonly encountered defects include formation of metal salt precipitates, agglomeration of colloids, porosity and cracking on films. Precipitation of the constituent metal salts in the precursor can occur when the deposition temperature and/or solvent volatility are too high such that the precursor dries before substantial gelation can occur[114]. Agglomeration can occur during precursor preparation when the operational temperature is too high to keep colloids in the precursor in a balanced status between Van de Waals attractive force and electrostatic repulsive force[16]. Both precipitation and agglomeration result invariably in significant inhomogeneity in the chemical composition and phase composition of the final heat-treated film, manifesting typically as the formation of secondary phases in the film. The processing stages of

drying (solvent removal), pyrolysis (oxidation of the polymer phase), and high- temperature heat-treatment (crystallization and sintering) collectively result in substantial shrinkage of as-deposited films which, if excessive or uncontrolled, can result in the formation of cracks in the film and/or delamination of the film from the substrate. In both cases, the structural integrity of the film is seriously compromised.

Also, implicit in the sintering process is the removal of porosity from the film which, if incomplete, can result in residual porosity in the final film.

For BFO films in particular, the presence of secondary phases, cracks, and porosity can significantly impair a film's electromechanical properties which, ultimately, diminishes the functionality and properties of the particular device on which the film is based. The first experimental part of this thesis investigates the optimization of the precursor and film gelation process so as to develop a reliable CSD preparation process capable of producing high performance BFO thin films. As explained below, this chapter focuses on the following three aspects of the CSD process: BFO gelation chemistry, starting precursors, and film crystallization.

(1) BFO gelation chemistry

As explained in Chapter 2, metal nitrate and organic solvent (2-methoxyethanol,

2-MOE and acetic anhydride) were chosen in this work as the starting materials to prepare bismuth ferrite instead of the precursors of metal alkoxides and water commonly employed for CSD of oxides ceramic in general. Significantly, the underlying chemistry of gelation processes involving non-aqueous starting precursors is not well understood. For example, how are the solvent (2-MOE and acetic anhydride) chemically involved in the gelation process? How is the gel state, based on the metal- oxygen-metal bond, established between an organic solvent and bismuth/iron nitrates?

What are the reaction products of this system? Such questions are of critical importance in preparing BFO thin films because the chemical composition and uniformity/stability of as-deposited gel films are strongly dependent on the gelation process and involvement of organics in the precursor. Improvements in current understanding of

BFO gelation chemistry should enable the non-aqueous precursor recipe and gelation process to be optimised, thereby improving and controlling gel film deposition and performance.

(2) Conditions of gelation of BFO starting precursor

The gelation of the precursor varies with the starting material, solvent type, and the gelation conditions. For thin and thick films, the gelation process usually involves the simultaneous processes of polymerization and solvent evaporation, both of which are promoted by the heating which is done usually after film deposition. The obtainment of a gel film is dependent on the relative rates of polymerization and solvent evaporation, the latter being used as the practical means of controlling the gelation. If solvent evaporation is too fast then there is insufficient time for polymerization to occur and instead a non-gel precipitated coating is obtained. For BFO in particular, precursors are based on mixed metal salts (Bi and Fe) for which differences in solubilities and gelation rates can contribute to local cationic segregation[17, 155] thus resulting in inhomogeneous compositions and the formation of bismuth-rich or iron-rich phases[17].

A major focus in this chapter is the investigation of the drying/gelation heat treatment conditions on the gel film formation process in respect to the gelation and drying rates.

(3) Crystallization of BFO thin film

Following deposition and gelation, heat-treatment or "annealing" is done to convert the amorphous metal-chelated polymer phase into the final functional ceramic

(usually an oxide) film. This stage of processing involves the overlapping processes of pyrolysis of the polymeric phase, crystallisation of the ceramic phase(s) (and possible subsequent phase transformations of these phase(s)), and sintering and grain growth of resultant ceramic grains. As mentioned previously, these processes invariably result in substantial shrinkage[16] of the film which, if excessive or uncontrolled, can result in the formation of defects such as cracks and pores in the film. For BFO films in particular, such defects, apart from compromising the structural integrity and mechanical reliability of the films, can result in a high leakage current thus greatly impairing the ceramic's electromechanical properties and performance[156]. The occurrence of such defects has, to date, been a major hurdle for the widespread development and application of CSD- produced BFO thin films. Thus, another major focus in this chapter is the investigation of the role of the post-deposition heat treatment conditions on the evolution of the phase composition and microstructure of the crystallized BFO films.

The preceding three aspects of the CSD processing are summarised in Figure 4-1.

This figure also shows the rationale of the experimental work done in this chapter.

Firstly, the initial chemical interactions between 2-MOE, acetic anhydride, and the Bi and Fe nitrates were investigated by nuclear magnetic resonance (NMR) and Fourier transfer infrared (FTIR) spectroscopy. Secondly, the temperature and time of gelation, and post-gelation drying on the efficacy of forming a gel film were studied. Lastly, using drying parameters optimised from the previous work, the effects of metal nitrates concentration were studied in terms of the phase composition and microstructure of fully heated treated BFO thin films. Collectively, the work done in this chapter enabled the precursor recipe and drying conditions to be optimised thus giving the basis to prepare

BFO thin films for study in subsequent chapters.

BFO thin-film Gelation chemistry preparation  Function of 2-MOE  Function of acetic anhydride  Chemical structure of BFO gel Uniform Precursor film gel Film preparation Gelation condition  Heating temperature Gel film  Heating duration  Further drying Condition Crystallized Dense crystallized BFO thin-film BFO thin-film performance film  Post-gelated film Characterization microstructure  Precursor concentration

 Optimized gelation/drying condition  Optimized precursor recipe 

Figure 4-1. Illustration of Chapter 4 research route

4.2 Experimental Procedure

4.2.1 Investigation of Gelation Chemistry

4.2.1.1 Precursor Preparation

Two general precursors (Precursor A/A1 and Precursor B/B1) were prepared to study the gelation process with and without the effect of acetic anhydride. The compositions of the precursors are shown in Table 4-1. BFO precursors A and A1 were prepared by dissolving bismuth nitrate with or without iron nitrate, respectively, in 2- methoxyethanol under constant stirring. Precursor A1 without iron nitrate was prepared for nuclear magnetic resonance (NMR) measurement because no NMR signal could be obtained if iron is present in the solution. Precursors B and B1 were prepared by dissolving bismuth nitrate with or without iron nitrate, respectively, in 2-MOE and acetic anhydride under constant stirring. To compensate for the loss of Bi evaporation

during crystallization, the molar ratio of Bi to Fe was made to 1.1:1 in the precursors A and B initially[76]. The concentration of Fe metal nitrate (or Bi nitrate with 10% excess) content was adjusted to 0.35 M[157] by adding 2-methoxyethanol. This concentration was chosen because, qualitatively at least, it gave a solution of suitable (low) viscosity for subsequent deposition by spin-coating whilst still providing sufficient metal content to yield a coherent BFO coating. The whole preparation process was performed in an ambient atmosphere at room temperature.

Table 4-1. Composition of precursor solutions

Precursor Bi(NO3)3·5H2O Fe(NO3)3·9H2O 2-MOE Acetic anhydride 0.015 mol 0.55 mol Precursor A 0.015 mol - (+ 10% excess) (43 ml) 0.55 mol Precursor A1 0.015 mol - - (43 ml) 0.015 mol 0.29 mol 0.21 mol Precursor B 0.015 mol (+ 10% excess) (23 ml) (20 ml) 0.29 mol 0.21 mol Precursor B1 0.015 mol - (23 ml) (20 ml)

4.2.1.2 Characterisation of Precursors by NMR and FTIR

Precursors A1 and B1 were analyzed by NMR to determine their initial organic molecular structures prior to gelation. 1H NMR spectra were obtained on neat mixtures with deuterated dimethyl sulfoxide (dmso-d6) in a capillary insert. Fourier transform infrared spectroscopy was used to analyze the polymeric functional groups in precursors

A and B and their changes during subsequent gelation. Approximately 0.1 ml of precursor solution was dropped onto a glass substrate (25 mm × 25 mm) and the substrate then inclined to cover the surface with a uniform layer of solution. The coated substrate was placed onto a hot plate heated at 90°C for a set time then FTIR analysis of the film was acquired immediately after heating. This deposition and FTIR measurement was done for separate samples for the following heating times: 0, 15, 30,

45, 60, 75, 90, or 105 seconds. FTIR patterns of the individual starting materials (2-

MOE, acetic anhydride, iron nitrate and bismuth nitrate) also were measured as a reference.

4.2.2 Investigation of Gelation Conditions and the Effect of Film Thickness

4.2.2.1 Film Deposition

Using Precursor A (see Section 4.2.1), three groups of films having different thicknesses were prepared on glass substrates. To enable a better observation of gel film microstructures, films prepared in this experiment were thicker than those used in later work in this chapter (Section 4.2.3) and in later chapters in this thesis. Films of different thicknesses were obtained as follows:

(a) Thin films: Approximately 0.1 ml of precursor solution dropped onto the glass substrate and the substrate then inclined at 90º to cover the surface with a uniform layer of solution and to remove the excess solvent.

(b) Medium thickness films: Approximately 0.1 ml of precursor solution dropped onto one edge of the glass substrate and the substrate then inclined at a steep angle

(~60º) to cover the surface with a uniform layer of solution.

(c) Thick films: Approximately 0.2 ml of precursor solution dropped onto one edge of the glass substrate and the substrate then inclined at a shallow angle (~30º) to cover the surface with a uniform layer of solution.

Using the above three techniques, the thicknesses of obtained films after drying were approximately <1 µm, 5-10 µm and 10-50 µm, respectively, as determined by transmission optical microscopy (E200, Nikon ， Japan) and scanning electron

microscopy. Examples of deposited films (after drying at 70°C for 60 minutes) are shown in Figure 4-2.

(a) Thin Film (<1 µm) (b) Medium Film (~5-10 µm) (c) Thick Film (~10-50 µm)

Figure 4-2. Examples of dry gel films with different thicknesses used in experiments

4.2.2.2 Gelation

The effect of heating conditions (temperature and time) on film gelation was investigated using the experimental design shown in Table 4-2. The red and blue bars indicate the two sets of experiments. Firstly, the effect of temperature on gelation was investigated by heating as-deposited films of each of the 3 thickness groups in air in an electric oven at 50°, 60°, 70°, 80° or 90°C for a set time of 10 minutes. These samples were then taken out of the oven and subsequently left at room temperature in an open ambient air for 2 hours for further drying. For each thickness group, an additional sample was dried at room temperature in open ambient air without any heat-treatment process. This experiment enabled the appropriate gelation temperature (70°C) to be ascertained and this temperature was then used to investigate the effect of gelation time for a further set of as-deposited films: as-deposited films of each of the 3 thickness groups were heated in air in an electric oven at 70°C for 5, 10, 20, 60 or 120 minutes, respectively. Following this heat treatment, the films were removed from the oven and subsequently left at room temperature in an open ambient air for 2 hours for further drying.

The surface morphologies of the as-prepared films were examined by the transmission optical microscopy. The quality of the as-prepared films and the extent of defects were ascertained so as to identify the appropriate gelation conditions.

Table 4-2. Experimental design for film gelation investigation

TEMPERATURE

TIME RT 50°C 60°C 70°C 80°C 90°C 5 min X 10 min X X X X X X 20 min X 60 min X 120 min X

4.2.2.3 Post-Gelation Drying

A post-gelation drying process for removing all remaining solvent in the gel film after gelation was investigated. Using the gelation heating condition of 70°C for 10 minutes (this being considered to be the optimal treatment for gel film formation – see

Section 4.3.2.4(1)) as the basis, three different post-gelation drying conditions were tested: (1) Same as the experiments described in 4.2.2.1 (film deposition), samples were put at room temperature in an open ambient atmosphere; (2) Samples were kept in the oven at 40°C for another 10 minutes before further drying at room temperature in an open ambient atmosphere; (3) Samples were kept in the oven at 40°C for another 60 minutes before further drying at room temperature in an open ambient atmosphere. The as prepared gel films were then examined using a transmission optical microscope.

4.2.3 Investigation of Film Crystallization and Microstructure

Starting precursors of three different metal nitrate concentrations (0.25 M,

0.35 M and 0.45 M) were prepared using Precursor B using the method described in

Section 4.2.2. The recipes of the precursors are summarized in Table 4-3. The amount of

acetic anhydride, which is fixed, is enough to dehydrate the crystal water from metal nitrate according to Equation 4-2 in the discussion section (see Section 4.3.2.4).

Table 4-3. Composition of precursor solutions with various metal salts concentrations

Precursor Bi(NO3)3·5H2O Fe(NO3)3·9H2O 2-MOE Acetic anhydride 0.015 mol 0.50 mol 0.21 mol 0.25 M 0.015 mol (+ 10% excess) (40 ml) (20 ml) 0.015 mol 0.29 mol 0.21 mol 0.35 M 0.015 mol (+ 10% excess) (23 ml) (20 ml) 0.015 mol 0.16 mol 0.21 mol 0.45 M 0.015 mol (+ 10% excess) (13 ml) (20 ml)

(001)-oriented SrTiO3 (STO, Daheng, China) were used as the substrates.

Despite being able to make relatively thick, clear, dry gel coatings on the glass slides in

Section 4.3.2, these coatings tended to suffer extensive cracking and delamination on subsequent high-temperature heat-treatment (results not shown here). Whilst single- layer thin films could be successfully fabricated it was necessary to utilise a multi- deposition process (of which two variations were investigated) for the production of thicker films. Thin and thick films were fabricated as follows:

(a) Thin films: These were produced as a single layer film by dropping a small amount (~0.1 ml) of precursor solution onto 70ºC preheated SrTiO3 substrate followed by spin coating at 3000 rpm for 30 s. Films were then heated in air on a hot plate at 90°C for 1 minute and then heated on a hot plate in air to 270°C for 3 minutes. Finally the film was heated to 750°C for 30 minutes at in a muffle furnace (Woodrow Kilns, Australia) in an ambient atmosphere with a heating/cooling rate of 5ºC/min.

(b) Thick-films: These were produced by depositing 10 layers of thin film. A single-layer film was deposited by spin coating as described above and then dried in air on a hot plate at 90°C for 1 minute and then heated on a hot plate in air to 270°C for 3 minutes. This was repeated 9 times to create a thick coating. The resultant ten-layer

layer film was heated to 750°C for 30 minutes at in a muffle furnace in an ambient atmosphere with a heating/cooling rate of 5ºC/min.

The phase composition and crystallographic orientation of the heat-treated films were analysed by X-ray diffraction using Cu  radiation at 45 kV/40 mA over a 2θ angular range of 10–55° at a scanning rate of 5°(2θ)/min. Scanning electron microscopy was used to study the microstructures of the thin film surface and energy-dispersive X- ray spectroscopy (EDS) was used for elemental analysis.

4.3 Results and Discussion

4.3.1 Investigation of Gelation Chemistry

Owing to it being chemically simpler, it is pertinent to consider first the NMR and FTIR analysis of the solution not containing acetic anhydride.

4.3.1.1 Precursor without Acetic Anhydride

(1) As-prepared precursor

1H NMR analysis and FTIR analysis of the starting precursors were done to analyze the molecular structures prior to any gelation. Figure 4-3(a) shows the 1H NMR

(a)

(b)

Figure 4-3. (a) 1H NMR spectra of Precursor A1 solution; (b) FTIR spectra of raw materials and as-prepared precursor A (2000 cm-1-780 cm-1)

pattern of bismuth nitrate solution in 2-MOE solvent without acetic anhydride (i.e.

Precursor A1). The NMR peaks correlate well with the chemical composition of 2-MOE

(explicitly CH3-O-CH2-CH2-OH), in that its functional groups of CH3 (at ~3.2 ppm),

CH2 (at ~3.3 and ~3.5 ppm), and OH (at ~4.7 ppm) are clearly evident. No peaks of other functional groups containing H are present. The FTIR patterns of the raw materials and as-prepared Precursor A are shown in Figure 4-3(b). The FTIR pattern of

Precursor A includes the same peaks as in pure 2-MOE plus peaks from the metal nitrates. Neither the NMR pattern of Precursor A1 nor the FTIR pattern of Precursor A changed noticeably after initial mixing of the precursors. In particular, the 2-MOE peaks in the FTIR pattern remained unchanged, which suggest that there were no appreciable reactions between the metal nitrates and 2-MOE solvent at room temperature.

(2) Precursor during heating

FTIR analysis of the Precursor A (without acetic anhydride) was done to elucidate the change(s) in chemical bonding during gelation without the complicating presence of acetic anhydride. Figure 4-4(a) shows the FTIR patterns of Precursor A after heating at 90°C for various times. The FTIR patterns remain qualitatively similar for the first ~45 s of heating. By ~60 s, there is a noticeable change of peaks in the ranges of 1200-950 cm-1. With further heating, the remaining FTIR peaks become less pronounced such that by about 90 s virtually all of the original peaks are no longer discernible. The films after this time had a gel-like appearance and were relatively dry.

This correlates well with the loss of FTIR peaks and suggests that appreciable reaction, presumably gelation, has occurred between the starting materials.

(a)

(b)

Figure 4-4. FTIR analysis of bismuth ferrite Precursor A showing: (a) FTIR patterns of bismuth ferrite starting precursor before heating and after heating for various durations at 90°C (2000-780 cm-1); (b) FTIR patterns of 2-MOE and bismuth ferrite starting Precursor A before heating and after heating for 75 s(1200-950 cm-1)

Given that gelation is likely to involve changes primarily in the 2-MOE solvent, better elucidation of the reactions involved in the process is made by closer analysis of the peaks over the 1200-950 cm-1 range. As shown in Figure 4-4(b), the FTIR patternof pure 2-MOE solvent is substantially the same as that of as-prepared Precursor A with both patterns having distinct peaks at ~963, ~1016, ~1059, and ~1119 cm-1. These peaks correspond to specific bonds in specific functional groups (all of which occur in 2-

MOE), namely: the O–H bending bond in –CH2–OH; the C–O stretch bond in general; the C–O stretch bond in –CH2–CH2–OH; and the C–O stretch bond in a generic R–O–R′ structure which, in this case, pertains to –CH2–O–CH3 of 2-MOE. Following heating at

90°C for 75 s, these peaks tended to either disappear or split into lower intensity peaks as follows. The peak at 1119 cm-1 corresponding to C-O stretch bond in R-O-R′ splits into two peaks (1088 cm-1 and 1110 cm-1) which indicates two different structures of

CH3–O–CH2- and -CH2-O-M in the precursor (as shown in Equation 4-1), each having slightly different wavenumbers for the C-O bond[146, 158]. The peak at 1059 cm-1,

[146] representing a typical C-O stretch bond in –CH2-CH2-OH group , disappears after heating, which further indicates the change of -CH2-OH structure in 2-MOE. Further, the C–O stretch bond[146] at 1016 cm-1 splits into two peaks ~1010 and ~1030 cm-1 and

[146] -1 the O–H bending bond in –CH2–OH at 963 cm disappears. These changes suggest that the O-H bond of the 2-MOE has broken on heating and, most likely, the H has been replaced with another element or functional group (in this case specifically, a Bi or Fe cation as explained later in Equation 4-1). Table 4-4 summarizes the functional groups in 2-MOE and in Precursor A before and after heating for 75 s.

Table 4-4. Summary of FTIR patterns of 2-MOE, Precursor A before and after heating for 75 s

Wavenumber (cm-1) C-O in C-O in C-O in O-H in Functional Groups: R-O-R` -CH2-CH2-OH general -CH2-OH 2-MOE 1121 1061 1017 963 Precursor A before heating 1119 1059 1016 963 1088 1011 Precursor A after heating for 75 s - - 1110 1033

To understand if the above changes in the molecular structure of 2-MOE during heating depend on the type of metal in the nitrate, FTIR patterns were acquired for the

2-MOE starting material and gel films derived from solutions of bismuth nitrate in

2-MOE solution, iron nitrate in 2-MOE solution, and bismuth nitrate + iron nitrate in

2-MOE solution (all after heating at 90°C for 75 s). As shown in Figure 4-5, no structural change is observed in pure 2-MOE after heating which indicates that the

2-MOE does not decompose or otherwise react by itself during heating. The FTIR patterns of the solutions containing metal salts are similar to each other but are substantially different to the pattern of 2-MOE. This suggests that, unlike the 2-MOE, these solutions have undergone significant molecular changes during heating. The changes are similar for the Bi solution and Fe solution as well as for the Bi + Fe mixed solution. This result suggests that the Bi and Fe undergo similar reactions with the

2-MOE, which may possibly be due to the close electronegativity values of Bi (2.02) and Fe (1.83).

Figure 4-5. FTIR patterns of 2-MOE and metal nitrates in 2-MOE precursor after heating at 90°C for 75 s (over 1200-900 cm-1)

The similarity of the FTIR patterns of the metal nitrates + 2-MOE solutions and the specific changes in the FTIR peaks of the 2-MOE are attributed possibly to the reaction between 2-MOE and metal nitrate to form metal-2-methoxyethanoxide[17]. As explained previously, the FTIR data suggest that the H cation in OH group of the 2-

MOE is replaced and, in this case, the replacement species is likely to be either a Bi or

Fe cation as shown in Equation 4-1 where M = Bi or Fe:

M(NO3)3 + 3[CH3OCH2CH2OH] ↔ M(-O-CH2CH2-O-CH3)3 + 3HNO3 Equation 4-1

The forward reaction in Equation 4-1 only occurs when HNO3 as the by-product is removed during the reaction, this can be achieved by adding base reagent or promoting volatility or decomposition of HNO3. Critically, little or no gel formed when the precursor coating was heated in a sealed container, indicating that the gelation was

promoted by the removal (evaporation) of the HNO3 reaction product. It is notable that the above process was carried out using a layer on the glass substrate instead of “bulk” solution, which underlines the critical role of the evaporation rate required for the forward reaction of Equation 4-1 to go to completion. The boiling point of pure nitric acid is ~83°C at ambient pressure and the decomposition temperature of liquid phase nitric acid is ~54°-90°C[159]. These temperatures agree well with the onset temperature of gelation (as discussed later in section 4.3.2). Concurrently, the large amount of excess 2-MOE in the precursor (as indicated in Table 4-1) drives the forward reaction in

Equation 4-1 and thus metal-2-methoxyethanoxide is formed.

4.3.1.2 Precursor with Acetic Anhydride

(1) As-prepared precursor

The 1H NMR pattern of the Precursor B1 is shown in Figure 4-6(a). In comparison with the NMR pattern of Precursor A1, the NMR pattern of the Precursor

B1 contains more peaks, which is to be expected since it contains an additional organic species. According to the NMR spectra, besides the peaks corresponding to “CH3-O-

CH2-CH2-” of 2-MOE (peaks 1, 2 and 3), additional peaks corresponding to a “CH3-O-

CH2-CH2-” structure in 2-methoxyethyl acetate (CH3-CO-O-CH2-CH2-O-CH3) (peaks

5, 6 and 7) are also present. This latter structure, as explained later, forms as a result of an esterification reaction between the 2-MOE and acetic anhydride (Equation 4-3 given later). Peaks 9 and 8 around 1.8 ppm indicate H in “CH3-” structures, which suggest the presence of acetic acid (CH3-COOH) and 2-methoxyethyl acetate (CH3-CO-O-CH2-

CH2-O-CH3) (as explained by Equation 4-2 and 4-3 later), respectively. Peak 4 at 5.8 ppm is attributed to “-OH” group which most likely on the 2-MOE.

(a)

(b)

Figure 4-6. (a) 1H NMR spectrum of Precursor B1; (b) FTIR spectra of raw materials and as-prepared Precursors A and B (2000 cm-1-780 cm-1)

The FTIR pattern of Precursor B (iron nitrate, bismuth nitrate, 2-MOE and acetic anhydride) is shown over a wave-number range of 2000-780 cm-1 in Figure 4-6(b) along with the patterns for 2-MOE, acetic anhydride, and Precursor A for the purposes of comparison. It is notable that whilst the FTIR patterns for 2-MOE and Precursor A are very similar to each other, the FTIR pattern of Precursor B is very different to that of

2-MOE as well as to that of Precursor A. Moreover, despite there being an appreciable amount of acetic anhydride in Precursor B, most of the FTIR peaks corresponding to the former are not apparent in the FTIR pattern for Precursor B. In particular, FTIR peaks for the acetic anhydride at 1822 and 1754 cm-1 (corresponding to the C=O stretch bond

[146] in the (CH3CO)2O structure ) disappear completely after mixing with the metal nitrates and 2-MOE to form Precursor B, while a peak at 1740 cm-1 (corresponding tothe C=O stretch bond in ester (RCOOR′) structures[146]) and another at 1719 cm-1

[146] (corresponding to the C=O stretch bond in acetic acid (CH3COOH) ) emerge. These results indicate that significant chemical reactions occurred on the initial mixing of

Precursor B and these are considered in more detail next.

As shown in Figure 4-7, the FTIR pattern of a mixture of 2-MOE and acetic anhydride showed the FTIR peaks corresponding to these two chemicals and it is notable that the FTIR peaks for C=O groups of a generic ester (at ~1730 cm-1) and acetic acid (at ~1715 cm-1) were absent. This indicates that no chemical reaction between 2-MOE and acetic anhydride occurred on mixing these two organic materials at room temperature. However, it should be noted that FTIR patterns revealed ester formation in the mixture after storing in a sealed container at room temperature for 1 week – this suggests that the mixture does react slowly undergoing an esterification reaction. In contrast, the addition of acetic anhydride to a solution of 2-MOE and metal nitrates resulted in a rapid rise in temperature to ~70-90ºC. It is considered that this is

caused by the reaction between acetic anhydride and the water of crystallization (from the metal nitrates) to form acetic acid as follows:

(CH3-CO)2O+H2O↔2CH3COOH Equation 4-2

The reaction in Equation 4-2 is exothermic and it is likely that the heat produced triggers the esterification reaction between 2-MOE and acetic anhydride to form acetic acid and 2-methoxyethyl acetate as shown in Equation 4-3. It is also likely that that metals (Bi or Fe from the starting nitrates) act as catalysts for the reaction given in

Equation 4-3 [160].

(CH3-CO)2O+CH3-O-CH2-CH2-OH↔CH3COOH+CH3-O-CH2-CH2-O-OC-CH3

Equation 4-3

Figure 4-7. FTIR spectra of 2-MOE, acetic anhydride, and mixture of 2-MOE and acetic anhydride before heating (1900-1650cm-1)

This reaction is supported by the NMR pattern for Precursor B1 (Figure 4-6(a)) in which peaks 5, 6 and 7 indicate the formation of a “CH3-O-CH2-CH2-” structure in

2-methoxyethyl acetate (CH3-CO-O-CH2-CH2-O-CH3) in addition to the one already

present in the 2-MOE (peaks 1, 2 and 3). As stated previously, peaks 9 and 8 around 1.8 ppm in the NMR pattern for Precursor B1 indicate “CH3-” structures which correlate with the formation of acetic acid (CH3-COOH) in Equation 4-2 (and 4-3) and 2- methoxyethyl acetate (CH3-CO-O-CH2-CH2-O-CH3) in Equation 4-3. The lack identifiable peaks of acetic anhydride in the FTIR pattern of Precursor B (Figure 4-6(b)) indicates that the acetic anhydride has been completely consumed in the reactions of

Equation 4-2 and 4-3. Lastly, it is likely that the exothermic heat of Equation 4-2 also provides heat for the reaction between the metal nitrates and 2-MOE to form metal-2- methoxyethanoxide (Equation 4-1 as discussed previously).

The above FTIR analysis and Equations 4-1 to 4-3 show that, unlike Precursor A, chemical reactions commence immediately on mixing of the metal nitrates, 2-MOE, and acetic anhydride and that the effective starting composition of Precursor B prior to

deposition is actually a mixture of metal-2-methoxyethanoxide M(CH3OCH2CH2O)3, 2-

MOE (CH3-O-CH2-CH2-OH), acetic acid (CH3COOH) and 2-methoxyethyl acetate

(CH3COOCH2CH2OCH3).

(2) Precursor during heating

Figure 4-8(a) shows the FTIR patterns of Precursor B after mixing ("starting material") and after heating at 90°C for various times up to 105 s. Like Precursor A,

Precursor B also showed a noticeable change in the shape of the FTIR patterns only after ~45 s. Closer inspection of the 1200-950 cm-1 range (Figure 4-8(b)) shows the presence of peaks of the C-O bond in 2-MOE or 2-methoxyethyl acetate

(CH3COOCH2CH2OCH3) in the starting precursor and after 90 s of heating. Similarly,

(a)

(b) (c)

the 1850-1650 cm-1 range (Figure 4-8(c)) shows the presence of peaks of the C=O bond in acetic acid (CH3COOH) and ester, in this case 2-methoxyethyl acetate

(CH3COOCH2CH2OCH3), in the starting precursor and after 90 s of heating. The FTIR peak positions of these functional groups are summarized in Table 4-5. The main change in the FTIR patterns on heating is the loss of the C-O bond in the -CH2-CH2-OH group (at 1052 cm-1). This attributed to the loss of OH groups on the 2-MOE as the solvent reacts with the acetic anhydride (Equation 4-3) and metal nitrates (Equation 4-1).

Table 4-5. Summary of FTIR patterns of precursor B after heating for 90 s

(a) C-O stretch bond

Wavenumber (cm-1) C-O in C-O in C-O Functional groups: R-O-R` -CH2-CH2-OH in general 1125 Precursor B before heating 1052 1015 1100 1125 1007 Precursor B after heating for 90 s - 1085 1029

(b) C=O stretch bond

Wavenumber (cm-1) C=O in C=O in Functional groups RCOOR` CH3COOH Precursor B before heating 1740 1720 Precursor B after heating for 90 s 1740 1708

The formation of acetic acid in Precursor B gives a possible variation to the reaction between the metal nitrates and 2-MOE (Equation 4-1) in that some acetic acid could react with the metal nitrates[161] as indicated in Equation 4-4.

M(NO3)3 + x[CH3OCH2CH2OH] + (3-x)CH3COOH

↔ M(OCH2CH2OCH3)x (OOCCH3)3-x + 3HNO3 where x ≤ 3 Equation 4-4

The FTIR data (Figure 4-8(c) and Table 4-5(b)) indicate a slight shift of the

C=O peak from 1719 cm-1 to 1707 cm-1 and it is possible that this is due to the acetic

acid reacting with metal nitrate[161]. Although this reaction is possible, the FTIR and

NMR data do not indicate the extent of this reaction compared with Equation 4-1.

Precursor B formed an obvious visible gel after ~60 s at 90°C. Gelation is considered to be by primarily the condensation of the metal-2-methoxyethanoxide as follows:

n[M(CH3OCH2CH2O)3] →

(CH3OCH2CH2O-)2M–[O-M(-O- CH2CH2-O-CH3)]n-2-O-M(-O- CH2CH2-O-CH3)2

+ (n-1)[CH3OCH2CH2OCH2CH2OCH3] Equation 4-5

Alternatively, condensation could involve M(CH3OCH2CH2O)x(CH3COO)y, for example:

n[M((-O- CH2CH2-O-CH3)x(OOCCH3)y] →

(CH3OCH2CH2O)x(CH3COO)y-1M–[O-M(CH3COO)]n-2–O-M(-O- CH2CH2-O-CH3)x-1

(OOCCH3)y + (n-1)[CH3COOCH2CH2OCH3] Equation 4-6

Denoting R as either CH3OCH2CH2– or CH3CO–, Equations 4-5 and 4-6 may be written more simply as:

n[M(OR)3] → M(OR)2–[OM(OR)]n-2–OM(OR)2 + (n-1)ROR Equation 4-7

The reaction sequence of the above process is summarized in Table 4-6. In the subsequent pyrolysis process (to 450°C), the organic groups of the gel will be decomposed and removed, leaving -M-O- as the final bond structure in the films.

Provided that the composition is stoichiometrically correct, this oxide structure will be eventually converted to perovskite BFO during subsequent high-temperature crystallization.

Table 4-6. Summary of the reaction sequence of Precursor B for the precursor preparation and gelation processes

Eqn. Description Chemical Reaction Reaction of acetic anhydride with water (CH3CO)2O + H2O → 2CH3COOH 4-2 (the latter from water of crystallization of the starting metal nitrates) Esterification between acetic anhydride (CH CO) O + CH OCH CH OH 4-3 3 2 3 2 2 and 2-MOE solvent → CH3COOH + CH3OCH2CH2OCOCH3 Reaction of metal nitrate with 2-MOE M(NO ) + 3[CH OCH CH OH] 4-1 3 3 3 2 2 solvent ↔ M(OCH2CH2OCH3)3 + 3HNO3 M(NO ) + x[CH OCH CH OH] + (3-x)CH COOH Reaction of metal nitrate with 2-MOE 3 3 3 2 2 3 4-4 ↔ M(OCH CH OCH ) (OOCCH ) + 3HNO solvent and acetic acid 2 2 3 x 3 3-x 3 (where x ≤ 3) Condensation reaction of reaction n[M(OR) ] 4-5 3 product from Eqn. 4-1 or 4-4 → M(OR)2–[OM(OR)]n-2–OM(OR)2 + (n-1)ROR 3+ 3+ where M = Bi or Fe ; R = CH3OCH2CH2– or CH3CO–

4.3.2 Investigation of Gelation Conditions and the Effect of Film Thickness

4.3.2.1 Effect of Heating Temperature

Thin Films

Figure 4-9 shows microstructures obtained by optical transmission light microscopy of the thin films deposited on glass and heated treated for a fixed time of

10 minutes at various temperatures (RT, 50°C, 60°C, 70°C, 80°C, and 90°C) followed by drying at room temperature in open ambient air for a further 2 hours. Films heated at

RT and 50°C contain needle-shaped salt precipitates distributed in a transparent gel matrix. In the RT films, the precipitates are ~200-400 µm in length and ~20 µm in width. In the 50°C films, the precipitates are ~50-100 µm in length and ~20 µm in width. In both the RT and 50°C films, the precipitates appear to be lie in the plane of the film due to the constraint imposed by the thickness of the film. Precipitates are also present in the films heated at 60°C but these are much smaller (~20 µm) and appear to be relatively equiaxed. Notably, the above precipitates did not form during the heat

100

treatment but rather formed during the subsequent 2 hours drying in air at room temperature. No precipitates were observed (up to the maximum magnification of

1000×) in films heated at 70°C, 80°C and 90°C with the films appearing to consist entirely of a uniform transparent gel.

Medium Thickness Films

Figure 4-10 shows microstructures obtained by optical transmission light microscopy of the medium thickness films deposited on glass and heated treated for a fixed time of 10 minutes at various temperatures (RT, 50°C, 60°C, 70°C, 80°C, and

90°C) followed by drying at room temperature in open ambient air for a further 2 hours.

In the films not subjected to any heating (i.e, RT for gelation), granular precipitates

(approximately rectangular and ~10 µm × 20 µm in size) distributed relatively uniformly within a transparent film are observed. For the heating temperature of 50°C, granular (~20 µm), needle-like (~20 µm ×100 µm), and dendritic (~50 µm-100 µm) precipitates are observed. Similarly, granular (~20 µm) and circular-like (~100 µm) precipitates are observed in the thin films heated at 60°C. Transparent and uniform gel films are obtained for films heated at 70°C. At higher temperatures (80°C and 90°C), the coatings were no longer transparent and instead appear to consist of a fine distribution of agglomerated particles.

101

(a) Thin-R.T (b) Thin-50℃

(e) Thin-80℃ (f) Thin-90℃

Figure 4-9. Microscopic structures of thin films after heating at various temperatures for 10 minutes, followed by drying at room temperature in an open ambient air: (a) No heat treatment; (b) 50°C;(c) 60°C; (d) 70°C; (e) 80°C; (f) 90°C

102

(a) Medium-R.T (b) Medium-50℃

(e) Medium-80℃ (f) Medium-90℃

Figure 4-10. Microscopic structures of medium thickness films after heating at various temperatures for 10 minutes, followed by drying at room temperature for 2 hours in open ambient air: (a) No heat treatment; (b) 50°C;(c) 60°C; (d) 70°C; (e) 80°C; (f) 90°C

103

Thick Films

Figure 4-11 shows microstructures obtained by optical transmission light microscopy of the thick films deposited on glass and heated treated for a fixed time of

10 minutes at various temperatures (RT, 50°C, 60°C, 70°C, 80°C, and 90°C) followed by drying at room temperature in open ambient air for a further 2 hours. Translucent films were formed with precipitates distributed on the surface of the films when the heating temperature was 70°C and below. In the films with no heating process (i.e., RT for gelation), precipitates of irregular shape in a size of ~20 µm to 200 µm are observed.

For films heated at 50°C, rectangular precipitates (~20 × 50 µm) and dendritic precipitates (~100 - 200 µm) are observed. Granular (~20 µm) and fan-like (~200 µm in radius) precipitates are observed in the films heated at 60°C. Needle-like precipitates

(~20 × 200 µm) and precipitates with fan-like shapes (radius >200 µm) are observed in the films heated at 70°C. Compared with the thin and medium-thick films, the coverage area of precipitates obtained for lower temperatures (below 70°C) treated films increased with the film thickness increases. In addition, similar to the medium thickness films, translucent thin films with a powdery appearance are observed for samples heated at higher temperatures (80°C and 90°C).

104

(a) Thick-R.T (b) Thick-50℃

(e) Thick-80℃ (f) Thick-90℃

Figure 4-11. Microscopic structures of thick films after heating at various temperatures for 10 minutes, followed by drying at room temperature for 2 hours in an open ambient air: (a) No heat treatment; (b) 50°C;(c) 60°C; (d) 70°C; (e) 80°C; (f) 90°C

105

4.3.2.2 Effect of Heating Duration

The results presented in the above section indicate that for the thin and medium thickness films, uniform and transparent gel films can be obtained by heating at 70°C for 10 minutes, suggesting that 70°C is an appropriate gelation temperature. Thus, in this experiment, all films were heated at 70°C for various durations, followed by drying at room temperature in an open ambient condition for 2 hours. Figures 4-12, 4-13, and

4-14 show the microstructures of thin, medium-thick, and thick films, respectively, heated at 70°C for various times (5, 10, 20, 60, and 120 minutes) followed by drying at room temperature in ambient atmosphere for a further 2 hours. The appearances of each group of films are discussed below.

Thin Films

As shown in Figure 4-12, the thin films heated for 5 minutes contain needle- shaped precipitates (~50 µm in length and ~10 µm in width) distributed in a transparent gel matrix. Again, it was observed that these precipitates formed only during the subsequent drying stage and not during the heat treatment. Films heated for 10 minutes or more consisted entirely of a transparent gel and no precipitates were observed (up to the maximum magnification of 1000×).

Medium-Thickness Films

The medium-thickness films were similar to the thin films at each corresponding time point. Specifically, as shown in Figure 4-13, the medium-thickness films contained "dendritic" precipitates (~50 µm) for the 5 minutes heat treatment but were completely transparent (up to the maximum magnification of 1000×) after heating for

10 minutes or longer.

106

(a) Thin-5 mins (b) Thin-10 mins

(e) Thin-120 mins

Figure 4-12. Microscopic structures of thin films after heating at 70°C for various durations, followed by drying at room temperature in an open ambient air: (a) 5 minutes; (b) 10 minutes;(c) 20 minutes; (d) 60 minutes; (e) 120 minutes

107

(a) Medium-5 mins (b) Medium-10 mins

(e) Medium-120 mins

Figure 4-13. Microscopic structures of medium-thick films after heating at 70°C for various durations, followed by drying at room temperature in an open ambient air: (a) 5 minutes; (b) 10 minutes;(c) 20 minutes; (d) 60 minutes; (e) 120 minutes

108

(a) Thick-5 mins (b) Thick-10 mins

(e) Thick-120 mins

Figure 4-14. Microscopic structures of thick films after heating at 70°C for various durations, followed by drying at room temperature in an open ambient air: (a) 5 minutes; (b) 10 minutes;(c) 20 minutes; (d) 60 minutes; (e) 120 minutes

109

Thick Films

As shown in Figure 4-14, no transparent gel film was obtained for the thick film group for heating of 20 minutes or less. Granular precipitates of ~10 µm, ~20 µm and

~20-100 µm were found in the films heated for 5, 10, and 20 minutes, respectively. The longer heating time of 60 minutes yielded transparent gel films but when the heating time was increased to 120 minutes powdery films with cracks were observed.

4.3.2.3 Effect of Post-Gelation Drying

In this study, films were heated at 70°C for 10 minutes (the conditions which were found previously to give clear transparent dry gel films) before the following different post-gelation drying processes: (1) drying at room temperature in an open ambient condition; (2) drying at 40°C for 10 minutes in an oven; (3) drying at 40°C for

60 minutes in an oven. As shown in Figure 4-15, it is found that no difference was observed in thin film and medium thick film groups between the different drying conditions. Homogenous and transparent gel films were obtained for all three conditions,

(see Figures 4-15(a)-(f)). However, for thick films, granular precipitates (~20-50 µm) and dendritic precipitates (~500 µm) were found in samples dried under ambient conditions or at 40°C for 10 minutes in the oven (see Figures 4-15(g)-(h)). Extending the heating time to 60 minutes in oven was found to be effective in avoiding precipitates in thick film, but cracking was observed (see Figures 4-15(i)).

110

(a) Thin-RT (b) Thin-40°C-10 mins (c) Thin-40°C-60 mins

(d) Medium-RT (e) Medium-40°C-10 mins (f) Medium-40°C-60 mins

(g) Thick-RT (h) Thick-40°C-10 mins (i) Thick-40°C-60 mins

111

4.3.2.4 Discussion

(1) Proposed Mechanism

The results of the effect of gelation heating process and post-gelation drying process on the microstructure of the films are summarized in Table 4-7. From the above descriptions, the dried films can be classified into three distinct morphologies: transparent gel film with distinct precipitates; completely transparent gel film; and powdery film without any apparent gel. The formation of these different morphologies can be interpreted in terms of the processes of film gelation versus film drying and the contributory effects of temperature, time, and film thickness. Film gelation is considered to occur by the chemical reactions described in Section 4.3.1 (i.e., Equations

4-1 to 4-6) whereas film drying is due to the evaporation of the solvent from the film.

The formation of the different film morphologies are considered separately below.

Table 4-7. Summary of effect of gelation heating conditions on film microstructure

(i) Gelation Heating Temperature Thickness Heating Temperature R.T 50°C 60°C 70°C 80°C 90°C Thin Precipitate Precipitate Precipitate Gel Gel Gel Medium Precipitate Precipitate Precipitate Gel Agglomeration Agglomeration Thick Precipitate Precipitate Precipitate Precipitate Agglomeration Agglomeration

(ii) Gelation Heating Duration Thickness Heating Duration 5 mins 10 mins 20 mins 60 mins 120 mins Thin Precipitate Gel Gel Gel Gel Medium Precipitate Gel Gel Gel Gel Thick Precipitate Precipitate Precipitate Gel Agglomeration & Cracking

(iii) Post-gelation Drying Condition Thickness Drying Condition R.T 40°C for 10 mins 40°C for 60 mins Thin Gel Gel Gel Medium Gel Gel Gel Thick Precipitate Precipitate Gel & Cracking

112

(a) Transparent gel film with distinct precipitates: As discussed in Section 4.3.1, the bonds between metal nitrates and organic solvent have not been formed in the as- prepared precursor A. At best, they are partially formed, as in case of the as-prepared precursor B. Considering the effect of temperature first, it is apparent that lower temperatures (≤ 60°C) tended to produce transparent gel films containing distinct precipitates. At these temperatures, both the evaporation rate of the solvent and the gelation rate (controlled by the rate of nitric acid removal in Equation 4-1) would be expected to be relatively slow. Although gelation occurred, the rate of solvent evaporation was sufficiently fast such that the concentration of metal salts in the solvent increased significantly to reach its solubility limit whilst the film was still sufficiently fluid thus resulting in appreciable precipitation.

(b) Completely transparent gel film: With increasing temperature, but significantly below the boiling point of 2-MOE (125°C), the rate of gelation would be expected to increase relative to the rate of solvent evaporation thus resulting in a greater amount of gelation prior to the onset of any precipitation. Also, the gelation process (Equation 4-1,

4-4, 4-5, and 4-6) consumes the cations thus decreasing their concentration in the remaining solvent. This counteracts the increase in cation concentration in the solvent that would otherwise occur due to solvent evaporation and thus helps to maintain the metal nitrate concentration below solution saturation so as to delay the onset of any precipitation. When precipitation does begin, its progression is less favorable because (i) there is a smaller amount of "free" cations available in the solvent for precipitation and

(ii) the formation of the gel phase increases the net viscosity of the solvent-gel system resulting in slower diffusion rates of the cations and anions involved in precipitate growth. For the latter films, it is considered that the gelation reactions involving the metal nitrates and 2-MOE solvent proceed to completion (without the formation of

113

precipitates) with the remaining excess solvent evaporating to yield the final dry transparent gel.

(c) Powdery film without any apparent gel: Films with a powdery appearance tended to form in thicker films under the more “severe” conditions of higher temperature and longer heating duration. Compared with the heating process for thin films, the drying time for thick films is expected to be much longer. Thus, the accumulation of heat energy in the precursor offers extra energy to the stabilized gel to overcome the potential repulsive barrier between each other thereby facilitating agglomeration[16]. In addition, cracking is found in thicker films after heating for a longer time. This attributed to appreciable solvent gradients forming through the thickness by virtue of the increased thickness. This, in turn, would create correspondingly larger differential shrinkage between the top and the bottom, with the top surface being placed into tension thus leading to cracking. Also, unlike thin films which tend to shrink towards the substrate (since the lateral shrinkage is restrained by the adhesion of the film on the substrate), in the thick films there likely to be less lateral constraint which makes the film more prone to cracking, especially at the surface.

Explanation of the effect of time on the film morphology is not as straightforward. Counterintuitively, thin or medium films treated at 70°C for 10 minutes or longer yielded transparent films without precipitates whereas heating for only 5 minutes at this temperature resulted in a significant amount of precipitation. The observation (see Section 4.3.2) that the precipitates formed not during the 5 minutes at

70°C but instead during the subsequent 2 hours under ambient conditions is significant for two reasons. Firstly, holding the film at 70°C would favor gelation over precipitation

114

as already explained for the effect of temperature. However, it is possible that holding the film at this temperature for just 5 minutes resulted in only partial gelation such that on cooling to room temperature there was still an appreciable amount of solvent left.

Secondly, subsequent study on the effect of post-gelation drying conditions regarding the growth of precipitates is in agreement with explanation of the effect of temperature, specifically that at lower temperature the rate of solvent evaporation is fast relative to that of gelation such that the metal salt concentration in the film increases with drying time leading to saturation and thus precipitation. In contrast, it is apparent that holding at 70°C for 10 minutes (and longer) was sufficient for the thin and medium thick films to gelate to such an extent to prevent any subsequent precipitation.

(2) Prevention of Metal Salt Precipitation

Following from the above explanations of the formation of the different film morphologies, it is considered that the precipitation of metal salts due to the onset of saturated mass concentration strongly depends on the solvent vapour pressure and its mass concentration. Thus, prevention of precipitation in the gel films can be realized via two general ways: (1) To decrease the evaporation rate or increase the drying time of precursor during heating (gelation) through changing the drying conditions or optimizing the precursor composition for obtaining a lower mass concentration and a lower vapour pressure; or (2) To increase the saturated solubility of precursor through increasing the precursor temperature. These were investigated by additional experiments as discussed below.

Precursor Concentration

Starting precursors of three different metal nitrate concentrations (nominally

0.25 M, 0.35 M and 0.45 M) were prepared using precursor B to study the effect of

115

precursor concentration and the solvent vapour pressure on the microstructure of crystallized BFO thin-films. To compensate for Bi loss during crystallization, a Bi excess of 10% was used in the precursors. The vapour pressures of the solvent in precursors of different metal nitrates concentrations can be calculated according to

Raoult’s Law in equation 4-8.

퐏풕풐풕 = ∑풊 푷풊 흌풊 Equation 4-8

th where: Ptot is the solvent's vapour pressure, i is the i component in the solvent, and

χ is mole fraction of the ith component in the solvent.

As discussed in Section 4.3.1, in as-prepared precursor B, no acetic anhydride

((CH3-CO)2O) was found in the precursor after stirring and both acetic acid (CH3COOH) and ester (CH3-O-CH2-CH2-O-OC-CH3) were detected suggesting that reactions specified by Equations 4-2 and 4-3 occur. The vapor pressures of 2-MOE (CH3-O-CH2-

CH2-OH), acetic acid (CH3COOH), 2-methoxyethyl acetate (CH3-O-CH2-CH2-O-OC-

CH3) and water at 20°C are 0.82 kPa, 1.53 kPa, 0.27 kPa, and 2.30 kPa, respectively.

The calculated vapour pressures of the precursor solutions (at 20°C) are given in Table

4-8 as a function of metal nitrate mass concentration, which were calculated assuming that Equation 4-2 and Equation 4-3 occur equally (but, in any case, if Equation 4-2 or 4-

3 occur first and to completion, then the vapour pressures of the resultant solutions will decrease or increase, respectively, by only ~2.5% relative to the values stated in Table

4-8 but, notably, the trend remains the same).

Table 4-8. Mass concentration and vapour pressure of precursor with various metal nitrate concentrations

0.25 M 0.35 M 0.45 M Mass Concentration (g/ml) 0.196 0.301 0.427 Vapour pressure (kPa) at 20°C) 1.157 1.287 1.389

116

Thus, as the metal nitrate concentration increases, not only does the effective solids loading increase but also the total vapor pressure of the precursor solution (at

20°C). Owing to the former, the drying rate of the organic solvent would be expected to increase as the mass concentration increases. The accelerated loss of solvent would thus shift the process in favour of precipitation over gelation. On the other hand, a low mass concentration precursor would give defects such as high porosity or cracking, due to insufficient solids material to cover the substrate surface. The microstructure and composition of crystallized BFO thin films derived from precursor of various concentrations will be investigated in Section 4.3.3. The experiments using the stoichiometric precursor (Bi:Fe=1:1) and the optimal gelation and crystallization conditions (which are optimized through the efforts in Chapter 4 and Chapter 5) are shown in appendix A1.2, which confirms that precursor with a lower metal salt concentration (i.e. 0.25 M) contributes to a better film quality regarding microstructure and phase composition.

Precursor Preheating

According to the previous discussion of the gelation mechanism, the precipitation of metal salts due to the onset of saturated mass concentration strongly depends on the solvent vapor pressure and the initial mass concentration of the cation.

Thus, in addition to the control of film gelation heating temperature and precursor composition, it was speculated that the prevention of metal salt precipitates in the thin film before BFO high-temperature crystallization can be achieved also by increasing the precursor solubility through increasing the precursor temperature.

Spin coating is commonly used to fabricate a homogenous thin film for device preparation. The fast spin velocity would increase the air flow rate on the

117

precursor surface during spinning, and thus promotes a fast drying rate of solvent. To reduce metal nitrate precipitation forming during spinning, one effective solution is to slightly increase the precursor temperature before/during spinning in order to increase the metal nitrate solubility in the precursor as well as promoting initial gelation. However, as was discussed above for Table 4-7, agglomerated precipitates start to form at 80ºC in the thicker film due to the accumulation of heat in the precursor before drying. Thus any temperatures higher than 80ºC may cause this sort of precipitation in the precursor. This is overcome by dropping the precursor on a preheated substrate (to 70ºC) immediately before spin-coating. This pre-heating process effectively prevents the formation of precipitate agglomeration in the thinner films during spinning not only because the heating could trigger the gelation before

(during) spinning, but also due to the increase of precursor solubility at a higher temperature (than room temperature). After the optimization of gelation and crystallization conditions in Chapter 4 and Chapter 5, the effect of preheating process has been shown in Appendix A1.2, which confirms that preheating process could effectively prevent the formation of metal salts during spinning and thus contributes to a smooth and defect-free gel and crystallized thin film

4.3.3 Investigation of Film Crystallization and Microstructure

4.3.3.1 Phase Composition of Films with Different Microstructure

The preceding results and discussion identified the occurrence of metal salt precipitation in the gel films prior to the completion of gelation as well as technical approaches to avoid this problem. An additional experiment was done to compare the heat treatment behaviour, in particular crystallisation, between a completely transparent gel film and a transparent gel film containing distinct precipitates. Two thin films were prepared by depositing 0.35 M 10% excess bismuth precursor on

118

STO substrate by spin coating followed by heating the as-deposited films for 10 minutes at either 60°C (to produce a gel film containing precipitates) or 80°C (to produce a completely transparent gel film). The latter employed preheating of the substrate (at 70°C) so as to guarantee a completely clear gel. The films were subsequently heated at 750°C for 30 minutes for crystallization.

As shown in Figure 4-16(a), the precipitate-containing gel film (heated at

60°C) yielded a very non-uniform microstructure after crystallization consisting of a sparse coverage of fine particles on the BFO surface and numerous large globular grains. As shown in Figure 4-17(a), an EDS line scan across the surface revealed the film to have a very non-uniform elemental composition. In particular, the large globular grains were shown to be Bi-rich suggesting that the precipitates present in the initial dried gel films were also Bi-rich. The occurrence of the latter could be due to the excess Bi in the precursor exacerbated by the slow gelation rate (relative to evaporation) for the 60°C gelation treatment. In the subsequent pyrolysis and high- temperature crystallization processes, these precipitates will be oxidized and may act as a nucleation centres for the formation of bismuth oxides or bismuth-rich ferrite phases[161] rather than BFO. XRD patterns (not shown here) of coatings processed at

60°C invariably contain significant peaks corresponding to Bi-rich phases.

In comparison, the gel film produced at 80°C was completely free of precipitates and corresponded to a uniformly distributed thin film on the substrate after high-temperature crystallisation (Figure 4-16(b)). An EDS line scan of the film showed it to have a relatively uniform distribution of Bi. It is noteworthy that despite the initial precursor having 10 % Bi excess, the final crystallised coatings did not contain any observable Bi-rich phases. Furthermore, XRD patterns (not shown here)

119

of coatings processed at 80°C typically show only very minor peaks, if any, corresponding to Bi-rich phases.

The result of this experiment suggests that the tendency of a gel film to form metal precipitates, in this case of Bi, is determined more by the gelation conditions

(i.e., temperature) than by the precursor starting composition. Also, it is likely that having a uniform gel would result in a more uniform volatility of Bi across the film thus making the film less sensitive to any fluctuations in composition during heat treatment thus, ultimately, making the film less prone to the formation of Bi-rich phases.

(a) (b)

Figure 4-16. Microscopic structures of thin films heated at (a) 60°C for 10 mins and (b) 80°C for 10 mins, annealed at 750°C for 30 minutes. Insets show the corresponding optical microscopic images of gelated and dried glass substrates prepared using the same gelation conditions.

Energy-dispersive X-ray spectroscopy (EDS) was used to identify the element distribution over the film surface. As shown in Figure 4-17(a), the precipitates in the fired film in Figure 4-16(a) were identified to be bismuth rich, indicating that the precipitates formed in the post-gelation dried films are bismuth-rich. This could be due to the excessive Bi from the precursor and as-deposited thin film. In the subsequent pyrolysis and high-temperature crystallization processes, these precipitates will be

120

oxidized and act as a nucleation centers for the formation of bismuth oxides or bismuth- rich ferrite phase[161] instead of BFO (in which Bi evaporates homogenously from the bulk). In comparison, a uniform distribution of Bi and Fe was over for the optimized film (Figure 4-17b) was observed after the 750°C annealing.

(b) (a)

Figure 4-17.EDS elements scan of films with various microstructures ((a) film with crystals on surface; (b) homogenous film)

4.3.3.2 Precursor Concentration

Thin films were prepared by using Precursors B of three different mole concentrations (0.25 M, 0.35 M and 0.45 M) followed by crystallization at 750 °C for

30 minutes in muffle furnace. A uniform microstructure of predominantly single- phase BFO is observed for the 0.25 M film, as shown in Figure 4-18(a). A uniform microstructure is also observed for most regions of 0.35 M derived thin films, however, pockets of a bismuth-rich secondary phase can also be observed, as shown in Figure 4-18(b). In comparison, the films obtained from the 0.45 M precursor are non-uniform and porous as shown in inset of Figure 4-18 (c).

121

(a) 0.25 M (b) 0.35 M

(b) 0.45 M

Figure 4-18.Surface microstructures of single-layer bismuth ferrite thin films prepared on STO(100) substrate by precursors with different concentrations ((a) 0.25 M; (b) 0.35 M; (c) 0.45 M) crystallized at 750°C

Figure 4-19 shows the XRD patterns of crystallized BFO thin films derived from precursors of various metal nitrate concentrations. It is found that pure BFO phase was obtained in the film derived from 0.25 M precursors. As the metal nitrate concentration increases to 0.45 M, secondary phases of Bi25FeO40 or Bi2O3 (bismuth rich phases) were observed in the thin film. “Bismuth-rich phase” is used nominally to mean all the secondary phases with a Bi:Fe ratio higher than one, including

Bi25FeO40, Bi24Fe2O39, and Bi2O3. The presence of precipitates results from the high mass concentration precursor (0.35 M and 0.45 M), while 0.25 M is an appropriate

122

concentration to prepare homogenous pure phase BFO thin films. These results agree well with the assumption and discussion on the effect of precursor metal salt concentration (section 4.3.2) on film gelation.

Figure 4-19. XRD patterns of single-layer bismuth ferrite thin films derived from precursor with various concentrations crystallized at 750°C ( STO; ° BFO; * bismuth rich phase)

Figure 4-20(a) and (b) show the microstructures of 10-layer BFO thin film prepared using 0.25 M and 0.35 M precursor solutions. It is found that uniform and dense polycrystalline films with grain size around 200 nm were obtained in both films.

The cross section microstructures of two films are shown in Figure 4-20(c) and (d). The thicknesses of 0.25 M and 0.35 M films are 400 nm, and 430 nm respectively.

Furthermore, grains in thickness direction shows an orderly condition in the film derived from 0.25 M, but grains are more randomly distributed in thickness direction in

0.35 M thin film. Thus, 0.25 M precursor is considered, like the single layer films, to be suitable to prepare multi-layer thin films of various thicknesses with dense and crack- free microstructure. The further optimization on preparation of multi-layer epitaxial

BFO thin films is discussed in Chapters 6 and 7.

123

(a) 0.25 M (b) 0.35 M

(d) 0.35 M (c) 0.25 M

Figure 4-20. Surface ((a) and (b)) and cross sectional ((c) and (d)) microstructures of 10- layer bismuth ferrite thin films prepared on (100)STO substrate from 0.25 M and 0.35 M precursors and crystallized at 750°C

The studies on the gelation chemistry and gelation conditions (to enable the obtainment of high-quality gel films prior to crystallisation) presented in this Chapter were the first major part of experimental work completed in the research project. It is noteworthy that they were completed prior to the optimization of the high-temperature crystallization processes for single layer films (Chapter 5) and multilayer films (Chapter

6). Following the optimization of the crystallisation processes, it was necessary to further refine the gelation chemistry and gelation conditions so as to better match the crystallisation processes. Thus, gelation experiments similar to those presented in this

Chapter were done for the purposes of confirming the applicability, reliability and

124

validity of the refined gelation chemistry and gelation conditions. These additional experiments are presented and discussed in Appendix A1. Although the final optimized precursor composition and gelation conditions are slightly different to those given in this Chapter, the underlying mechanisms and fundamental understanding are unchanged.

4.4 Summary

The precursor for chemical solution deposition of BFO thin films was prepared by mixing the iron nitrate, bismuth nitrate, 2-methoxyethanol and acetic anhydride. The gelation chemistry of this non-aqueous precursor was studied during a low-temperature heating process. The changes in molecular chemical structure and organic functional groups of precursor during the gelation process were investigated by

FTIR. It is found that metal-2-methoxyethanoxide was formed during heating and esters were formed from 2-methoxyethanol and acetic anhydride during precursor preparation with a gel material then forming by the further condensation reactions between the metal-2-methoxyethnoxide and esters.

Following study of the gelation chemistry and the resultant gel chemical structure, the conditions for homogenous gel film formations were optimized by study of the effect of heating temperature, heating duration, and drying conditions. It was found that a competition between gelation and drying occurs during the low- temperature heating process. Uniform gel thin film can be prepared under a moderate heating condition of ~70°C, but precipitates of Bi and Fe salts tended to form in the film at lower temperatures owing to the gelation rate being sufficiently slow so as to permit salt precipitation as salt concentration approached saturation in the solvent. At higher temperatures, the rate of solvent evaporation was significantly increased such that there was insufficient time for gelation to occur prior to solvent loss thus resulting in the

125

formation of a powdery particulate film. In addition, agglomeration was found to occur in the thicker films at or above 80°C or for longer heating time (≥120 mins). This is because the stable polymer colloids would overcome the potential repulsive barrier between stable colloids at a high temperature (≥80°C) thereby facilitating agglomeration in the precursor. To avoid the precipitate and agglomeration in the spin-coating derived thin films, the precursor can be preheated to by preheating the substrate to 70ºC immediately before deposition so at to increase the precursor solubility of metal nitrates.

Homogenous thin films can be prepared by spin coating using 0.25 M precursor, which has a low metal nitrate mass concentration and low vapour pressure. Precipitates can be observed from 0.35 M or 0.45 M precursor due to the faster evaporation of solvent during spinning and following drying processes. In addition, 400 nm thin films with dense and crack-free microstructure can be prepared by multi-layer deposition using 0.25 M, indicating the good performance of 0.25 M precursor in the preparation of multi-layer epitaxial BFO thin films – this is examined further in Chapters 6 and 7.

126

Chapter 5.

Preparation of (001)-Oriented Epitaxial BFO Thin Films by CSD using Stoichiometric Precursor

5.1 Introduction

The preparation of epitaxial BFO thin films by CSD routes is challenging because it requires accurate control of the Bi:Fe stoichiometry[17]. Often, the imprecision of the starting chemical composition and the subsequent volatilisation of Bi during high temperature annealing lead to secondary phases or highly-conductive films with very poor leakage resistance[7, 8]. To counter this problem, excessive Bi is often added to the starting reagents, which in turn, can result in the formation of bismuth-rich phases. These aforementioned problems have stymied the large scale adoption of CSD routes for the fabrication of derived BFO films. Hence, there remains the important need to develop an accurate understanding of the chemical processes such that stoichiometric precursors may be employed.

In the previous chapter, homogeneous gel films were obtained by depositing

0.25 M Bi/Fe precursor solution on substrate and subsequently heating at 90°C and

270°C, successively. This optimized process is used in this chapter to investigate the effect of Bi volatility and annealing conditions on film crystallization and phase

[34] formation. According to the Bi2O3-Fe2O3 phase diagram , BFO has a very small

127

synthesis window and this results in it being very sensitive to annealing temperature and atmosphere. In Tyholdt et al[17]’s report, BFO starts to form at ~550°-600°C and decomposes at ~780°C but these important temperatures vary with changing annealing atmosphere owing to altered synthesis kinetics[78]. For example, the presence of oxygen can enhance the kinetics of metal oxide formation as well as BFO decomposition[78].

The objective of the research work in this Chapter was to systematically investigate the effects of precursor Bi/Fe ratio, heating temperature, and heating atmosphere on the crystallisation of the CSD-derived BFO thin films and the resultant phase purity and crystallographic orientation. The intended practical outcome of this work was to determine the optimal crystallization conditions for obtaining pure-phase

BFO thin films with epitaxial structures on STO substrates. The research route of this

Chapter is illustrated in Figure 5-1.

Using the optimized precursor, deposition, and gelation conditions identified in the previous chapter, gel films having two different Bi:Fe starting ratios (stoichiometric and 10% Bi-excess) were prepared and then crystallised in different atmospheres (air or oxygen) at various temperatures (450°, 550°, 650°, 750° and 850°C). The effects of these processing parameters were investigated in terms of the phase composition, crystallographic alignment, cross-sectional microstructure, surface microstructure, and ferroelectric domain structure of the films as well as the films' micro- and macro- ferroelectric properties.

128

BFO Thin-Film Preparation

Preparation of pure phase (001) Precursor oriented epitaxial BFO thin-film Preparation  Crystallization temperature  Heating treatment atmosphere  Precursor Bi/Fe ratio Gel film

Crystallized BFO BFO Thin-Film Performance  Single phase Thin-Film  Phase composition  Epitaxial Structure  Surface morphology microstructure  Dense film  Ferroelectric polarization hysteresis

Characterization

Optimized crystallization temperature and atmosphere for preparation of pure BFO thin film with epitaxial structure

Figure 5-1. Illustration of Chapter 5 research route

5.2 Experimental Procedure

Strontium titanate (SrTiO3(001), STO, Shinkosha, Japan) with and without a

LSMO buffer layer (depending on the experiment) was used as the substrat.

Stoichiometric (Bi:Fe=1:1) and Bi-excess precursors (Bi:Fe=1.1:1) were prepared and then deposited the STO substrates using the sol-gel process developed Chapter 4.

Bismuth nitrate and iron nitrate were dissolved in 2-methoxyethanol. Acetic anhydride was added under constant stirring to the solution to form a homogeneous 0.25 M

BiFeO3 precursor solution. The compositions of the two precursors are shown in Table

5-1. BFO thin films were then produced by dropping a small amount of precursor solution onto the substrate preheated to 70°C followed by spin coating at 3000 rpm for

30 seconds. As-deposited films were heated in air on a hot plate at 90°C for 1 minute,

129

followed by heating at 270°C for 3 minutes. The films were then rapidly heated by putting the sample into a preheated tube furnace in approximately 5 seconds and then held at the particular temperature (450°, 550°, 650°, 750° and 850°C) for 30 minutes in an ambient air or pure oxygen atmosphere.

Table 5-1. Composition of 0.25 M precursors with and without 10% excess Bi

Bi(NO3)3·5H2O Fe(NO3)3·9H2O 2-MOE Acetic anhydride 0.50 mol 0.21 mol 10% Bi excess 0.0165 mol 0.015 mol (40 ml) (20 ml) 0.50 mol 0.21 mol Stoichiometric 0.015 mol 0.015 mol (40 ml) (20 ml)

The phase composition and crystallographic alignment of the resultant films were analysed by X-ray diffraction using Cu Kα radiation at 45 kV/40 mA over a 2θ angular range of 15–75° at a scanning rate of 5°(2θ)/min. XRD phi-scan was used to examine the crystallographic structure relationship between the deposited BFO thin film and the substrate. The film was tilted at a chi angle of 45° to obtain the X-ray diffraction pattern for the (110) plane phi-scan over a range of 360° at a scanning rate of 60°/min.

Cross-sections of approximately 100 nm thickness were prepared for BFO/STO(001) and BFO/LSMO/STO(001) films using focused ion beam milling. Cross-sectional bright field microstructure images and selective area electron diffraction (SAED) patterns were obtained by transmission electron microscopy. In addition, high resolution transmission electron microscopy was used to observe the cross-sectional microstructure of BFO/STO(001) thin film samples. A commercial scanning probe microscopy system was used for both atomic force microscopy to observe the surface microstructure of films and for piezoelectric force microscopy to study the film domain structures. Conductive Ti-Ir coated silicon cantilevers (ASYELEC-01, Asylum

130

Research, US) were used for AFM/PFM imaging, domain switching, and piezoelectric amplitude/hysteresis loop studies. In particular, for the local ferroelectric domain switching characterization, the PFM tip acted as the top electrode and the voltage was applied through bottom electrode to the tip. To measure global ferroelectric polarization of BFO thin films, bottom and top electrodes were prepared for the device characterization. 20 nm thick lanthanum strontium manganite (La0.67Sr0.33MnO, LSMO) film was deposited by Pulsed Laser Deposition on STO(001) as the bottom electrode prior to the BFO thin film deposition. It was found from XRD that the LSMO films grew epitaxially on the STO substrate. Au/Ti (60 nm/5 nm in thickness) coatings were used as top electrode and they were prepared by photolithography and metal evaporation techniques. The size of each top electrode pad was 23 µm x 23 µm.

Ferroelectric polarization hysteresis loops were acquired by a ferroelectric testing system at room temperature at a frequency of 10 kHz. In this measurement, probes of the ferroelectric tester were connected with LSMO bottom electrode and Au/Ti top electrode and bias was applied through top electrode to the bottom one.

5.3 Results and Discussion

5.3.1 Phase Evolution and Crystal Structure Investigation

The XRD patterns shown in Figure 5-2(a)-(c) compare phase compositions and crystallographic structures of BFO/STO (001) thin films derived from the two different precursors (Bi-excess and stoichiometric) following annealing at various temperatures

(ranging from 450° to 850°C) in different atmospheres (air or oxygen). Bi, as a volatile material, often evaporates from the bulk during high-temperature heat treatment [7, 8],

131

leading to an off-stoichiometric ratio of Bi:Fe (Bi-deficient) in the resultant BFO.

According to others’ reports[76] and the preliminary crystallization results given in

Chapter 4, excess Bi in the precursor can effectively compensate for the loss of Bi during annealing in air. Thus, the consideration of the results in this experiment starts with the precursor with 10% excess Bi.

Figure 5-2(a) shows the effect of heat-treatment temperature on the phase composition of films prepared from Bi-excess precursor heated in air. The XRD pattern of the film heated at 450°C shows only the main peaks of STO (~23° and ~ 47° (2θ)) whilst peaks of BFO are absent. This suggests that BFO has not crystallized to any significant extent at this temperature. As the temperature increases to 550°C, Bi-rich phases and BFO(110) peaks with very weak intensity can be observed at ~35° (2θ) and

31.5° (2θ), respectively. Because BFO(00L) peaks and substrate STO(00L) peaks are close, BFO peaks are not yet evident. Pure-phase BFO is obtained when the film was heated to 650°C and the peak of Bi-rich phases at ~35° (2θ) disappears when heating to this temperature and above. This result indicates the loss of excess Bi has occurred and that the Bi:Fe ratio has decreased close to the stoichiometric ratio of 1 during annealing.

On the other hand, the BFO(110) peak grows stronger as the temperature increases from

132

(a)

(b)

(c)

Figure 5-2. XRD patterns BFO/STO(001) thin films derived from: (a) Bi-excess precursor annealed in air; (b) Bi-excess precursor heated in oxygen; and (c) stoichiometric precursor annealed in oxygen. The temperatures used (450 to 850°C) are indicated in the patterns and the time at temperature was 30 minutes. (O = BiFeO3 (110); * = Bi2Fe4O9 (002); # = Fe2O3; Δ = Bi-rich phase; x = unknown phase)

133

550°C to 750°C, which indicates that the (001) epitaxial orientation of BFO thin film on the (001)STO substrate has been compromised. In addition, at 850°C, bismuth-rich phase was observed again at ~31.2°(2θ) along with Bi2Fe4O9 (002) at ~29.8°(2θ) and

Fe2O3 phase at ~30°(2θ). The emergence of both bismuth-rich and iron-rich phases at high temperature indicates that the decomposition of BFO has occurred, which agrees with others’ reports[73, 162]. The term “bismuth-rich phase” is used thereafter to refer to all secondary phases with a Bi:Fe ratio higher than one, including Bi25FeO40, Bi24Fe2O39, and Bi2O3.

The above experiments highlight the key difficulty in preparing phase-pure epitaxial BFO films by CSD, which is, specifically, BFO has a very small synthesis window with respect to both the optimal Bi/Fe ratio and the heat-treatment temperature.

The results are in agreement with the prior findings of Bea et al.[77], who first reported the sensitive nature of this system in a series of systematic investigations on BFO thin films deposited by PLD. In particular, the Bi/Fe stoichiometric ratio in BFO is strongly dependent on the volatility of bismuth or bismuth oxide[76, 77]. In this work, as shown in

Figure 5-2(a), pure-phase BFO is obtained by annealing the thin film in air at 650°C when 10% excess Bi is added in the precursor, suggesting the loss of Bi during annealing.

One contributor to the volatility of Bi at high temperature is the relatively low oxygen partial pressure (an ambient air atmosphere) such that Bi-O becomes unstable

[77] and decomposes into Bi and O2 . In this process, first oxygen vacancies form when

O2 is released, followed by evaporation of Bi due to its high vapour pressure. Thus, to prevent the breakage of the Bi-O bond and to restrain the formation of oxygen vacancies and the loss of Bi, an oxygen atmosphere can be used in annealing process. The XRD

134

patterns of BFO thin films prepared from Bi-excess precursor heated in oxygen atmosphere are shown in Figure 5-2(b). Similar to the case of heating in air

(see Figure 5-2(a)), BFO starts to crystallize at 550°C with the emergence of the

BFO(110) peak at ~32°(2θ). On increasing the annealing temperature to 650°C, the

BFO(110) peak disappears but the peak of a bismuth-rich phase emerges at ~35°(2θ) and the intensity of this peak increases with further increase in the annealing temperature. By 850°C, there are significant peaks of Bi2Fe4O9 (002) and Fe2O3 present.

Collectively, the presence of peaks of bismuth-rich phase(s) (27.5°(2θ) and ~31.0°(2θ)),

Bi2Fe4O9(002) (29.8°(2θ)) and Fe2O3 (~30°(2θ)) at 750°C suggest that decomposition of

BFO has occurred by this temperature.

Unlike the BFO annealed in air, bismuth-rich phase is formed in the BFO when heated to 650°C and above in oxygen, indicating that the initial excess Bi volatilized to a lesser extent (compared with heating in air) and that this excess Bi remained in the film. In this case, the increase of oxygen partial pressure during crystallization effectively restricts the extent of bismuth volatilisation and, presumably, decreases the oxygen vacancy concentration.

A small BFO (110) peak was observed for BFO films annealed at 550°C in the oxygen atmosphere but this peak was not present in films annealed at 650°C and above.

A BFO thin film of (001) crystallographic orientation is likely to form on annealing because it is both the preferred orientation (to minimise in-plane strain) as well as the epitaxial crystal orientation with respect to the STO (001) substrate. Hence, the BFO

(110) orientation at 550°C is defined as a misorientation. Furthermore, it is obvious that the misorientation is minimized by the higher oxygen partial pressure condition when no Bi is formed from the decomposition of bismuth oxide, because the formation of Bi

135

at higher temperature throughout the film may lead to the nucleation and growth of polycrystalline BFO.

As shown in above experiments, it is not possible to prepare pure-phase BFO with an epitaxial structure on STO when using Bi excess precursor for annealing either in air or in an oxygen atmosphere. However, the XRD patterns shown in Figure 5-2(c) show that it is possible when the BFO thin films are prepared from stoichiometric precursor and annealed in oxygen, provided that the Bi volatility can be restrained to a negligible level. A BFO(110) peak is apparent at ~32°(2θ) for the films annealed at

450°C and 550°C. However, at temperatures of 650°C and higher, this BFO(110) peak disappears. Very significantly, at 650°C only the peaks of BFO(001) are evident and, specifically, no peaks of other BFO crystallographic orientation nor peaks of secondary phases are observed. This indicates that a pure-phase BFO having a (001)-oriented crystal structure has formed. As the annealing temperature increases to 750°C, a weak peak of a bismuth-rich phases emerges at ~35°(2θ) and possibly a very weak peak of

Fe2O3 at ~49°(2θ), these indicating the start of BFO decomposition. By 850°C, relatively high intensity peaks of bismuth-rich phases (~31.0°(2θ) and ~35.0°(2θ)),

Bi2Fe4O9(002) (29.8°(2θ)) and iron oxide (~24°(2θ), ~30°(2θ), ~35°(2θ) and ~49°(2θ)) are present indicating that significant decomposition of the BFO has occurred.

The above results show that pure-phase BFO with (001) oriented epitaxial structure can be obtained from the stoichiometric precursor by annealing at 650°C in an oxygen atmosphere. Having determined the optimal precursor and deposition conditions

(Chapter 4) and now the optimal annealing conditions for BFO/STO, these conditions were employed to prepare BFO epitaxial thin films on a LSMO buffered STO(001) substrate – this was for the eventual purpose of making of actual CSD-derived BFO thin

136

film devices and characterising their ferroelectric characteristics. Figure 5-3(a) shows the XRD patterns of BFO films on STO with and without an intervening LSMO buffer layer. Neither crystallographic misorientation nor secondary phases are observed in the films. Sharp diffraction peaks of STO(00L) (L=1, 2 or 3) can be observed at ~23°, ~ 47° and ~72.5°(2θ). BFO(00L) peaks at ~22°, ~45° and ~70° (2θ) are found on the left shoulder of these STO(00L) peaks, respectively. Since the LSMO peaks are close to

STO peaks, only LSMO (002) and (003) peaks can be observed on the right shoulder of

STO peaks, at ~47.5° and ~73°(2θ) respectively. In addition, as shown in Figure 5-3(b), a 360° phi scan of BFO on the (110) plane gives four set sharp peaks with 90° intervals and these are aligned well with the substrate peaks. This result shows the 4-fold symmetry of BFO thin film crystal structure (which has a same in-plane lattice parameter with substrate STO(001)), confirming the (001)-oriented epitaxial nature of these BFO film samples.

(b) (a)

Figure 5-3. (a) XRD patterns of BFO/STO(001) and BFO/LSMO/STO(001) thin film derived from stoichiometric precursor after heating at 650°C in oxygen atmosphere for 30 min; (b) Phi scan of the (110) plane of (001)-oriented BFO/LSMO/STO(001) thin film and the STO substrate

137

5.3.2 Film Cross-Section Microstructure Investigation

Figure 5-4 shows the TEM and HRTEM cross-sectional images of

BFO/STO(001) thin film prepared from stoichiometric precursor (single layer deposition) and annealing at 650°C in an oxygen atomosphere. The film is approximately 40 nm thick and appears to be fully dense and continuous in the in-plane direction for a distance of least 400 nm (image not shown here). A distinct interface between the BFO thin film and STO(001) substrate can be observed in Figure 5-4(a).

No grain boundaries were apparent in the film cross-section.

The BFO-STO interface is distinct and there are no obvious structural defects

(such as cracks) nor compositional defects (such as interdiffusional phases) along the interface. The HRTEM image (Figure 5-4(b)) shows the BFO/STO(001) cross-section where both BFO and STO structures can be observed. It can be seen that the atoms of the BFO align well with those of the underlying STO substrate. Figure 5-4(c) shows the corresponding selective area electron diffraction (SAED) pattern. The clear diffraction spots indicate the single-crystal nature of the thin film and the STO and BFO planes align with each other along the [0K0] and [00L] directions. Collectively, Figures 5-4(b) and (c) confirm the epitaxial nature of the BFO on the STO(001) substrate.

138

(a) BFO/STO

BFO

STO

(b)

BFO

STO

(c)

Figure 5-4. Interface of a 40 nm BFO/STO (001) thin film showing: (a) TEM image; (b) HRTEM image; and (c) SAED diffraction pattern

139

From the HRTEM image and diffraction patterns, the lattice parameters were determined using the STO substrate as the calibration standard. Bulk BFO exhibits a rhombohedrally (R) distorted perovskite structure, having the space group R3c[163] and a lattice parameter of ~0.396 nm. In this 40 nm BFO/STO(001) thin film, the in-plane lattice parameter (a) is ~0.394 nm which is smaller than that its bulk counterpart constant (~0.396 nm) while the out-of-plane lattice parameter (c) of ~0.400 nm is larger than its bulk counterpart (~0.396 nm). The lattice constant of the cubic STO(001) substrate is ~0.391 nm, showing an in-plane lattice mismatch of 1.46% with bulk BFO.

Thus, the smaller lattice parameter of STO imposes an in-plane compressive strain on the BFO thin film thus causing an in-plane contraction, and out-of-plane elongation, of the BFO thin film crystal. The out-of-plane to in-plane lattice parameter ratio of the

BFO thin film on STO(001) is 1.016 which is smaller than the value of BFO tetragonal

(T) structure (>1.23) reported by others[70]. Therefore, is is possible that this BFO film may show a transition structure between R phase and T phase, this being the monoclinic structure others have reported[5, 69].

Figure 5-5(a) shows the TEM cross-sectional microstructure of a single layer

BFO thin film deposited on STO(001) substrate (using stoichiometric precursor and annealing at 650°C in an oxygen atmosphere) with an intervening 20 nm thick LSMO buffer layer (bottom electrode). Like the BFO/STO(001) thin film described above, the

BFO thin film was ~40 nm thick, was fully dense with no observable pores or other structural defects, and had no observable grain boundaries in the c-axis (thickness) direction. Figure 5-5(b) shows the SAED pattern of BFO/LSMO/STO(001) cross- section. As the LSMO is only 20 nm thick and its lattice parameter is intermediate to that of BFO and STO, only STO and BFO diffraction pattern spots can be clearly observed. Similar to the patterns in Figure 5-4(c), the diffraction spots from the

140

BFO/LSMO/STO(001) align well with each other along the [0K0] and [00L] directions, demonstrating the epitaxial structure of BFO on LSMO buffered STO substrate.

(a) (b)

Figure 5-5. (a) TEM and (b) SAED diffraction pattern of 40 nm BFO/LSMO/STO(001) thin film

5.3.3 Surface Microstructure and Domain Structure Investigation

Figure 5-6 shows the surface morphology of a 40 nm BFO thin film on

LSMO/STO(001) substrate in 3×3 µm2 (Figure 5-6(a)) and 1.5 ×1.5 µm2 area (Figure 5-

6(b)). The film has a linear-intercept grain size of ~150 nm and a RMS surface roughness of ~3.19 nm, demonstrating a relatively smooth surface. Although this roughness value is higher than those typically found for atomically sharp surfaces obtained by PLD technique, there are no particles or open pores over tens of microns.

This indicates that these CSD films have good applicability for device manufacture.

141

(a) (b)

(e)

+6V

-6V

Figure 5-6. AFM image of 40 nm BFO/LSMO/STO(001) thin film surface in a size of (a) 3 µm×3 µm and (b) 1.5 µm× 1.5 µm; Vertical PFM (c) amplitude and (d) phase images of virgin BFO domain structures in a size of 1.5 µm× 1.5 µm; Vertical PFM (e) phase image (5 µm×5 µm) of BFO/LSMO/STO(001) thin film after applying +6 V (3 µm×3 µm) and -6 V (1 µm×1 µm)

142

Figures 5-6(c) and 5-6 (d) show the vertical (out-of-plane) PFM (V-PFM) phase image and amplitude image, respectively, corresponding to the 1.5 ×1.5 µm2 field of view (Figure 5-6 (b)). A fractal-like domain structure with two phase contrasts is observed, indicating a poly-domain state. Two contrasts can be observed in the PFM phase images in Figure 5-6(c), indicating two polarization (domain) directions which are an upwards polarization (purple) and a downwards polarization (yellow). The interfaces between these two regions are the domain walls and these which show a low or no piezoresponse as indicated in the amplitude image in Figure 5-6(d) (the dark colour indicates the domain walls). This virgin domain structure of BFO thin films indicates the existence of ordered domain structures with a tendency to be aligned and it also exhibits a dependence on grain structure.

To observe the domain switching behaviour in the BFO film, +6 V and -6 V DC bias voltage was applied in 3×3 µm2 and 1×1 µm2 square areas, respectively, through

PFM tip to the bottom electrode, effectively as a localised poling process. Compared with the virgin domain region of mixed upwards polarization (purple regions) and downwards polarization (yellow regions) shown in Figure 5-6(c), the V-PFM phase image of the poled region shown in Figure 5-6(e), shows that the virtually all of the domains in the 3×3 µm2 area written with +6 V are switched downwards (a single yellow colour) whereas virtually all of domains in the 1.5×1.5 µm2 area written with

–6 V are switched upwards (a single purple colour). This result demonstrates that the

143

domains in the BFO/LSMO/STO(001) thin films are switchable in macro-scale and can be switched completely under an external voltage of 6 V.

5.3.4 Micro and Macro Ferroelectric Properties

Figure 5-7(a) shows the local piezoelectric force microscope switching loops

(phase and amplitude) of a 40 nm BFO/LSMO/STO (001) thin film. The loop shown here is an average of 3 measurements. As the tip size is approximately 7 to 20 nm in diameter, which is smaller than the BFO film grain size (~150 nm) observed in

Figure 5-6(b), the PFM amplitude and phase loops represent the ferroelectric behaviours of single grains. A square PFM phase loop with a sharp 180° change at ~±3.5 V is observed. At the coercive voltage ~3.5 V, the domains in the BFO thin films are switched upwards or downwards with an opposite polarization direction and thus the piezoresponse phase shows an 180° change before and after this domain switching.

Similarly, a PFM amplitude butterfly loop is observed with a dramatic dip at a coercive voltage of ~±3.5 V. Both loops show typical ferroelectric domain switching characteristics of BFO thin films at the micro scale.

144

(a)

(b)

Figure 5-7. Room-temperature (a) Piezoresponse force spectroscopy amplitude and phase hysteresis curves and (b) Polarization hysteresis loop of 40 nm BFO/LSMO/STO (001) thin film

145

Figure 5-7(b) shows the room temperature macro-scale polarization-electric field

(P-E) hysteresis loops of 40 nm BFO on LSMO buffered STO(001) substrate. This result was repeatable for measurements on 3 different top-electrode pads. Ferroelectric

2 polarization hysteresis loops with low remanent polarization 2Pr of a mere 9.4 µC/cm is obtained at room temperature with a coercive voltage 2Vc of 0.6 V. The shape of the loop is not a saturated square loop and shows a decreasing polarization at the higher voltage, which indicates leakage current. In addition, since this global P-E loop is unsaturated, its coercive voltage is much smaller than that of local piezoresponse loops, which is around 3.5 V. By comparing the micro-scale (local) and macro-scale (global) ferroelectric (piezoelectric) loop, it is considered that the leakage current in the former comes from the grain boundaries where a higher concentration of oxygen vacancies may exist.

5.4 Summary

In this chapter, pure-phase BFO thin films having (001)-oriented epitaxial structure were successfully prepared by the optimized CSD method using stoichiometric precursor. The effects of annealing temperature and atmosphere on film phase composition and crystallographic orientation were studied. It was found annealing at

650°C in an oxygen atmosphere is the optimal condition for the crystallization the BFO thin film because it effectively prevents the loss of Bi and minimises BFO(110) misorientation, thereby enabling phase-pure epitaxial BFO films to be obtained on the

STO(001) substrate. The same result was obtained for BFO coatings on the

LSMO/STO(001) substrate. TEM imaging of the cross-sectional microstructures of both

BFO/STO(001) and BFO/LSMO/STO(001) thin-films showed 40 nm thick BFO thin films which were dense and with a columnar heterogeneous grain structures along the

146

thickness direction. The corresponding SAED patterns of the thin-films' cross-sections confirmed the epitaxial nature of the thin-films with the spots reflections of the STO and BFO planes align with each other. The as-prepared pure phase epitaxial

BFO/LSMO/STO(001) thin-film shows a homogenous surface microstructure and domain structure with roughness of ~3.19 nm. Domain switching behaviour of pure phase epitaxial BFO/LSMO/STO(001) thin-film was observed by PFM when an external voltage of ±6 V was applied on the thin film. Using the PFM technique, the local PFM phase and amplitude loops were are also obtained with a coercive field of

~3.5 V. However, only unsaturated global ferroelectric polarization hysteresis loops

2 with remanent polarization (2Pr) of mere 9.4 µC/cm was obtained due to the leakage current at the local level. To further investigate and improve the ferroelectric properties of the CSD-derived BFO thin films, the preparation and characterization of epitaxial

BFO thin films as a function of thickness will be studied and discussed in next two chapters.

147

Chapter 6.

Multi-layer BFO Thin Film Preparation

6.1 Introduction

The effect of thickness on bismuth ferrite thin film phase composition, microstructure, and ferroelectric properties has been studied recently, but mainly for

PLD coatings[164, 165]. In the chemical solution deposition process, film thickness is dependent largely on precursor viscosity[104], spin-coating velocity[16], spin-coating time[16], and the number of deposition repeats[108]. For single-layer deposition in particular, although film thickness can be increased by increasing precursor viscosity and decreasing spin-coating velocity and time, the resultant coatings typically have significant cracking and pores the due to the large amount of shrinkage that occurs during drying, pyrolysis of the polymer, and high-temperature heat-treatment

(crystallisation and sintering). Thus, the thickness attainable for crack- and pore-free films can be limited. Furthermore, for thinner films, surface coverage of the fired substrate by the film often is incomplete due to the low effective solids loading typically used in CSD precursor formulations. The effective solids loading is restricted by the need to maintain low-enough viscosity of the precursor so as to have satisfactory spin coating behaviour. Thus, it can be difficult to prepare films of varying thicknesses by a single-layer deposition[104]. In comparison, multi-layer deposition is able to produce films ranging in thickness from less than 50 nm to a few microns whilst avoiding microstructural defects such as pores and cracks. As such multi-layer deposition is the principal means of thickness control for CSD coatings [107, 166].

148

Generally, two different deposition-heating processes are utilised in multi-layer deposition: (1) Conventional deposition: deposition of multiple sol-gel layers prior to pyrolysis and crystallization; and (2) Sequential deposition: each deposited layer is pyrolysis and crystallized prior to the deposition of the next layer [167]. In conventional deposition, the pyrolysis and crystallization processes only occur once for the set of multi-layer thin films, thus it is quicker and more convenient. However, it has the disadvantage that the large absolute shrinkage which occurs during pyrolysis and sintering of the collectively thick film may lead to cracking or pores. In comparison, sequential deposition offers the advantage of each thin layer being well crystallized with smaller (absolute) shrinkage of each layer (compared with a thicker multilayer film) which thus helps avoid cracking and porosity. However, it takes significantly longer time to prepare thicker films since each layer needs to be pyrolysed and crystallized prior to the next deposition. Furthermore, the repeated high-temperature heat-treatment which the first few layers experience may cause compositional defects, such as oxygen vacancies or material decomposition, at or near the film-substrate interface. Also, given that each layer is heat treated to a relatively mature extent, planes of compositional difference and/or structural weakness may form at the interfaces between adjacent layers. Lastly, each layer will have received a different net heat-treatment which may result in compositional/structural differences between layers (possibly a graduation from the substrate to the surface) and thus resulting in property differences between the layers.

In this Chapter, using the optimized precursor recipe and deposition process derived in Chapter 3 and the heat treatment conditions determined in Chapter 4, single layer and two-layer films were prepared. For the latter, four types of deposition-heating sequences were investigated: prior to the deposition of the second layer, the first layer

149

was heat treatment at: (1) 90°C (ie., after gelation); (2) 270°C (ie., after drying); 450°C

(ie., after pyrolysis); or 650°C (ie., after crystallization). Following The effects of the deposition-heating sequence on the phase composition, microstructure, and ferroelectric properties of the films were studied accordingly, with particular focus on their uniformity in the overall film. The research route of this Chapter is illustrated in Figure

6-1.

BFO Thin-Film Preparation Deposition-heat treatment sequence Precursor  Preparation of 2-layer thin film preparation deposition

Gel film BFO Thin-Film Performance  Single phase  Phase composition  Epitaxial Structure Crystallized BFO  Surface morphology microstructure  High remanent Thin-Film polarization  Domain structure  Ferroelectric polarization hysteresis  Low leakage  Leakage current current Characterization

Optimized deposition-heat treatment sequence to prepare multi-layer BFO thin-films with various

thicknesses

Figure 6-1. Illustration of Chapter 6 research route

150

6.2 Experimental Procedure

6.2.1 Differential Scanning Calorimetric Testing

Dried gel powder derived from stoichiometric precursor was prepared for differential scanning calorimetric (DSC, Netzsch; Germany) analysis to study the heat flux of BFO gel during heating in order so as to identify the temperatures for polymer decomposition/pyrolysis. DSC testing was conducted in air from room temperature to

500°C using a heating rate of 5°C /min.

6.2.2 BFO Thin Film Preparation

(001)-oriented strontium titanate (SrTiO3, STO(001), Shinkosha, Japan) was used as the substrates. Lanthanum strontium manganite (La0.67Sr0.33MnO, LSMO) was deposited as the bottom electrode on the STO(001) by pulsed laser deposition. 0.25 M stoichiometric precursor was used to prepare BFO thin films for spin coating. 0.015 mol bismuth nitrate and 0.015 mol iron nitrate were dissolved in 40 ml 2-methoxyethanol.

20 ml acetic anhydride was added under constant stirring to the solution to form a homogeneous 0.25 M BiFeO3 precursor solution. Thin films were then produced by dropping a small amount of solution onto the substrate preheated to 70°C followed by spin coating at 3000 rpm for 30 seconds.

For a single layer BFO thin film, the as-deposited film was heated in air on a hot plate at 90°C for 1 minute and 270°C for 3 minutes, successively. The films were put into a 450°C preheated tube furnace in approximately 5 seconds and rapidly heated to

450°C for 30 minutes for pyrolysis followed by 650°C for 30 minutes. Both pyrolysis and crystallization processes occur in an oxygen atmosphere. To study the effect of drying, pyrolysis and annealing process on phase composition and microstructure

151

evolution in 2-layer BFO thin films, the following four different heat-treatment sequences were employed:

1) Sample A-90 was prepared by depositing the second BFO layer after heating the first layer at 90°C;

2) Sample B-270 was prepared by depositing the second BFO layer after heating the first layer at 90 °C and 270 °C;

3) Sample C-450 was prepared by depositing the second BFO layer after heating the first layer at 90 °C, 270 °C and 450 °C;

4) Sample D-650 was prepared by depositing the second BFO layer after heating the first layer at 90 °C, 270 °C, 450 °C and 650 °C.

After the deposition of second layer, the four 2-layer films were successively heated at 90 °C, dried at 270°C in an ambient atmosphere, followed by rapidly heating to 450°C and 650°C in an oxygen atmosphere. Table 6-1 summarises the heating processes used for the 2-layer samples. The red colour highlighted temperatures indicate the last heat-treatment process of first layer before deposition of the second layer.

Table 6-1. Sample preparation on BFO/LSMO/STO(001) by various heating processes

Sample Gelation Drying Pyrolysis Annealing A-90 90* 270°C 450°C 650°C B-270 90 270°C* 450°C 650°C C-450 90 270°C 450°C* 650°C D-650 90 270°C 450°C 650°C* (* indicates the repeat start point of second-layer film preparation)

The phase composition of the films was analysed by X-Ray diffraction using

CuKα radiation at 45 kV/40 mA over a 2θ angular range of 15–50° at a scanning rate of

5°(2θ)/min. The surface microstructures and ferroelectric domain structures of the BFO thin films were examined by atomic force microscopy and piezoelectric force

152

microscopy, respectively, using a commercial scanning probe. Micro-scale (local) ferroelectric domain switching characterization was performed by PFM using conductive Pt coated silicon cantilevers (Multi75E-G, BudgetSensors, Bulgaria) which act as the top electrode with the sweep voltage being applied through bottom electrode to the tip. Macro-scale (global) ferroelectric polarization hysteresis loops were acquired at a frequency of 10 kHz at room temperature by a ferroelectric testing system. For these measurements, top electrodes of Au/Ti (60 nm/5 nm in thickness) were prepared by photolithography and metal evaporation techniques. The size of each top electrode pad is 23 µm x 23 µm. I-V curves of BFO thin films were measured by using a Keithley

2400 SourceMeter. The voltage was applied through the top electrode to the bottom one which was connected to the workstation through two probes.

6.3 Results and Discussion

6.3.1 Precursor Thermal Analysis

Figure 6-2 shows the DSC pattern of stoichiometric precursor derived dry powder upto a temperature of 500°C in air. The first endothermic peak indicates the evaporation of water which could be either from raw material or by-product of polymerization reaction. Subsequent exothermic peaks at 175°C, 232°C and 312°C indicate the decomposition of organic polymers in the film[78, 168]. This suggests that all of the polymers should have been removed by about 400°C. The peaks at 232°C and

312°C may also be result from the reduction of Bi3+ and oxidization of Bi (or formation

[17, 78] of Bi25FeO40) , respectively. In this chapter, gelation was done below 270°C in air but pyrolysis (450°C) and crystallization (650°C) were done in an oxygen atmosphere.

Thus, the heat flow peaks shown in Figure 6-2 (obtained in air), might be slightly different to those above 270°C which would occur the in films heated in oxygen for the

153

actual film preparation. Nevertheless, the oxygen atmosphere would be expected to accelerate the decomposition rate such that all the polymers should have been removed even at a lower temperature compared with pyrolysis done in air. In this thesis, the pyrolysis temperature used was 450°C, as it is a high enough temperature to remove the polymers in an oxygen atmosphere but, on the other hand, is well below the crystallization temperature of BFO according to the discussion in Chapter 5.

Figure 6-2. DSC analysis of powder derived from stoichiometric precursor heated in air

6.3.2 Phase Composition

The effect of heat-treatment sequence on the evolution of phase composition and crystallographic orientation the BFO thin films was determined by XRD analysis of the two-layer BFO thin films after final heating at 650°C. As shown in Figure 6-3, the peaks of STO(00L) (~23° and ~ 47°(2θ)) can be observed and weak LSMO (00L) peaks

(~23.5° and ~47.5°(2θ)) can be observed in the right shoulder of the corresponding STO peaks. BFO(00L) peaks are found on the left side of STO XRD peaks and they shift

154

towards the substrate peaks with the increase of final heat-treatment temperature of the initial layer. In particular, BFO (002) peak shifts from 45.3° in A-90, to 45.5° in B-270, to 45.7° in C-450 and D-650. This indicates a progressive decrease of the out-of-plane lattice parameter from ~0.400 nm in A-90, to ~0.399 nm in B-270, to ~0.397 nm in C-

450 and D-650. Additionally, peaks of bismuth rich phases at ~31°(2θ) are observed in the A-90, B-270 and C-450 BFO thin films. As in previous chapters, the term “bismuth- rich phase” refers to all secondary phases with a Bi:Fe ratio higher than one, including

Bi25FeO40, Bi24Fe2O39, and Bi2O3. BFO(110) peaks at ~32°(2θ) are evident in the A-90,

B-270 and C-450 diffraction patterns. In addition, a BFO(111) diffraction peak is evident at 39.5°(2θ) in the B-270 thin film. As mentioned for BFO(110) in Chapter 5

(and now for BFO(111)), its presence can be interpreted as misorientation relative to the crystallographic orientations of substrate STO (001) and the epitaxial BFO(001) thin film. The emergence of these misoriented BFO crystal structures compromise the epitaxial structure of BFO thin film. Notably, single-phase (001)-oriented epitaxial

BFO thin film was obtained only for sample D-650 of the 2-layer films as evidenced by it showing diffraction peaks of only (001)-oriented BFO and none for any bismuth-rich or other phases.

155

Figure 6-3. XRD of stoichiometric BFO thin films prepared by various deposition-heating processes after annealing at 650°C in oxygen atmosphere for 30 minutes (O-BFO (110) or BFO (111) ; Δ Bismuth rich phase)

6.3.3 Microstructure and Domain Structure

Figure 6-4 shows the AFM surface microstructures and PFM domain structures of each of the 2-layer BFO thin films prepared by the four deposition-heating sequences and after crystallization at 650°C in an oxygen atmosphere. In the AFM images, square- like grains are observed in the A-90 thin film (Figure 6-4(a)) while grains with irregular shapes are observed in the B-270 sample (Figure 6-4(b)). Compared with A-90 and B-

270 thin films, the grain boundaries are less evident for the C-450 ((Figure 6-4(c)) and

D-650 (Figure 6-4(d)) films. An average linear-intercept grain size of ~150 nm is measured for all four samples. In addition, open pores of ~100-200 nm diameter are observed in samples A-90 and B-270. Although root mean square (RMS) roughness of

C-450 and D-650 (~3.0-4.0 nm) is larger than PLD-derived BFO thin films (RMS

156

(a) Morphology (b) Amplitude (c) Phase (i) A-90

(ii) B-270

(iii) C-450

(iv) D-650

Figure 6-4. (a) Topographies, (b) PFM amplitude, and (c) phase images of BFO thin films prepared by various heat treatment sequences ((i) Sample A-90; (ii) Sample B-270; (iii) Sample C-450; (iv) Sample D-650) (each image size is 1.5 µm x 1.5 µm)

157

typically <1.0 nm), there are no open pores observed in the C-450 and D-650 thin films.

Additionally, triangular secondary phases can be observed from the surface of C-450 thin film.

Figure 6-4(b) and (c) show the out-of-plane PFM amplitude and phase images of four thin films, respectively. Fractal-like domain structures with two-phase contrast were observed in all samples from the PFM phase images, indicating a poly-domain state. The two-phase contrasts indicates only two polarization (domain) directions which are, specifically, upwards polarization and downwards polarization. The interfaces between these two regions are domain walls which show a little or no piezoresponse in the amplitude images (domain boundaries are in dark colour). In particular, unlike other three samples whose domain patterns are less dependent on the film grain structure, the grain boundary of B-270 thin film is clearly observable in the out-of-plane PFM amplitude and phase images, as shown in Figure 6-4(b-ii) and (c-ii), thus indicating no piezoresponse in the grain boundary. In addition, the different contrast from B-270 amplitude image demonstrates the piezoresponse of the individual grains. This is attributed to the grains having different crystallographic orientations[169], which is consistent with the XRD result that strong BFO(110) and BFO(111) misoriented peaks were found in this film. The triangular secondary phases in C-450 can also be observed in the PFM amplitude image shown in Figure 6-4(b-iii). The dark contrast those secondary phases indicates that they have no piezoelectric response. For sample D-650, the PFM measurement reveals a homogenous domain structure with a uniform amplitude value, indicating that the crystals are well aligned with preferred orientation (001) which gives the constant piezoresponse amplitude throughout the thin film.

158

6.3.4 Ferroelectric Property Investigation

Figure 6-5(a) shows the macro-scale ferroelectric polarization hysteresis loops of BFO thin films prepared by the four deposition-heating sequences. All the samples were measured at 10 kHz at room temperature. No hysteresis loop is obtained in sample

A-90 because of it having a high leakage current. Hysteresis loops can be observed

2 2 2 with remanent polarization 2Pr of 45.0 µC/cm , 24.0 µC/cm and 73.5 µC/cm for samples B-270, C-450 and D-650, respectively. However, only D-650 shows a saturated square hysteresis loop. It is also evident that the coercive voltage of D-650 (2Vc=2.87 V) is much smaller than the other two samples (2Vc=5.10 V for B-270 and 3.50 V for C-

450). Thus, it follows that the domains are more easily switched in the D-650 thin film compared with the B-270 and C-450 samples. For each heat-treatment sequence condition, these results were statistically repeatable (for replicate samples) and were stable and consistent for 4 or more measurements on 2 different top electrodes of the same sample.

Figures 6-5(b)-(e) show the micro-scale piezoresponse properties measured by

PFM of four samples cooresponding to the four heat-treatment sequences. The loops shown here are averages of 3 measurements for each thin film. Square PFM phase loops with a sharp 180° change at coercive voltages are observed from all four thin films indicating microscale domain-switching. The average 2Vc of four samples are 3.49 V,

8.99 V, 5.25 V, and 5.80 V, respectively. At these coercive voltages, the domains in the

BFO thin films are switched to an opposite polarization direction and thus the piezoresponse phase shows a 180° change before and after this domain switching.

Similarly, the corresponding PFM amplitude butterfly loops show a dramatic dip at the coercive voltage for all four films. Both loops show typical ferroelectric domain switching characteristics of BFO thin films at the micro scale.

159

(a)

(b) (c)

(d) (e)

Figure 6-5. (a)P-E loops and (b)-(e) local PFM amplitude and phase curves of bismuth ferrite thin films prepared by different deposition and heat treatment ((b) A-90, (c) B-270, (d) C-450 and (e) D-650)

160

Table 6-2 summaries the remanent polarization and coercive voltage values obtained from the global and local ferroelectric loops. The differences in Vc between the

P-E hysteresis and PFM derived phase (or amplitude) hysteresis can be explained as the differences between macro- and micro-scale domain switching mechanism[170, 171].

Since the PFM tip size is less than 20 nm, it is likely the new domain nucleation starts near the PFM tip in a single grain in the local micro-scale measurement. However, in the macro polarization hysteresis measurement made through the 23 µm × 23 µm top electrode pads, the domain switching would be expected to begin with the growth of residual domains at the interface, such as domain walls or grain boundaries. In addition, variation of chemical compositions, surface roughness and micro pores within films all would contribute to the differences of domain switching during the measurement made by the two methods.

Table 6-2. Summary of macro and micro-scale ferroelectric properties of BFO/LSMO/STO(001) thin films

A-90 B-270 C-450 D-650

2 Global 2Pr (µC/cm ) - 45.0 24.0 73.5

Global 2Vc (V) - 5.10 3.50 2.87

Local 2Vc (V) 3.49 8.99 5.25 5.80

6.3.5 Current-Voltage Behaviour

Figure 6-6 shows the I-V curves of 2-layer BFO thin films using different heat- treatment deposition sequence. The results are repeatable from 3 measurements on different top electrode pads.

161

Figure 6-6. I-V curves of BFO thin films of various thicknesses (A-90, B-270, C-450, D-650)

It is found that A-90 sample presents a higher leakage current than other three samples (in the measurement range -9V to 9V) with the current in the order of 10-2 A.

This agrees well with the macro-scale ferroelectric polarization hysteresis behaviour(see

Section 6.3.4) that no hysteresis loop can be obtained from A-90 due to the high leakage.

Among the other three samples, below ~4 V, D-650 shows a lowest leakage current

(less than 10-4 A) than B-270 and D-450. However, the resistive switching behaviour is observed in D-650 that the resistance of it changes from high resistance state (HRS) to low resistance state (LRS) at ~4 V. At the voltage above 4V, D-650 shows a similar resistance level with B-270 and C-450 thin films. The RS behaviour of the BFO thin film will be investigated further in Chapter 7 using the conductive AFM technique.

162

6.3.6 Discussion

(1) Microstructure

As was discussed for Figure 6-2, the polymer of the gel coating starts to decompose at around 200°C, based on the DSC patterns, all polymers in the thin film should be removed during pyrolysis by 450°C. The heat treatment process can be divided into three stages[101]: (1) First stage – drying: weight loss without shrinkage occurs and is due solely to the loss of remanent solvent; (2) Second stage – pyrolysis: weight loss with some shrinkage occurs with both being due to the removal of polymers

(since the polymers constitute a significant fraction of the total film volume); (3) Third stage – crystallisation and sintering): shrinkage without weight loss occurs with the shrinkage being due to densification. The initial heat treatment temperature of the first layer plays an important role in the overall heat treatment process because it likely that it results in different microstructures, differing amounts of remaining polymer, and different magnitude of shrinkage of the first layer. The way in which these collectively affect the deposition and heat treatment of the second layer is discussed below for each of the four heating sequences.

(a) A-90

A schematic diagram of deposition, drying, pyrolysis and sintering process of multi-layer bismuth ferrite thin film preparation is given in Figure 6-7. After the deposition of the first layer deposition and the heat treatment at 90°C for gelation, it is likely that only the excess solvent has been removed from the films and large amount of polymers remain, primarily as the gelated phase, in the film. When the second layer is deposited on the top of the first layer and subsequently heated at 90°C for gelation, it is likely that crosslink network would be formed between the two layers by virtue the liquid diffusion of the precursor from the second layer into the first.

163

(a) Deposition of uniform precursor film on STO substrate by spin coating:

(b) Evaporation of solvent at 90°C giving gelation of the first layer (blue squiggle lines indicate the formation of polymer gels in the thin film). The second layer then is deposited onto the first layer:

(c) Diffusion of the precursor of the second layer as a liquid into the gelated first layer as well as evaporation of solvent at 90°C from the second layer giving gelation of the second layer (blue squiggle lines indicate the formation of polymer gels in the thin film):

(d) Removal of polymers by drying (at 270°) and pyrolysis (at 450°C) from the bulk 2- layer thin film resulting in significant (differential shrinkage):

(e) Crystallization and sintering (at 650°C) of the bulk 2-layer thin film effectively preserve the above microstructure and may cause further inhomogeneity by further differential shrinkage (especially in the vertical direction between the top surface and the substrate):

Figure 6-7. Schematic diagram of showing the stages of deposition, drying, pyrolysis and sintering process of 2-layer thin film for the A-90 heat-treatment sequence

164

(b) B-270

After drying the first coating at 90°C and then heating it at 270°C, the solvent can be considered to be completely removed from the deposited film and it is likely that polymer combustion has occurred (the latter suggested by the exothermic peaks at

175°C and 232°C in the DSC pattern – see Figure 6-2). Since 270°C is not high enough to remove much of the polymers in the thin film, and because crystallisation and densification are not likely to have occurred up to this temperature, then the amount of shrinkage undergone by the film would be expected to be low. Ideally, the deposition of the second layer should double the thickness of the thin film when the first layer is completely dried. In the subsequent pyrolysis and crystallization processes, the polymers are removed from the two layers of the thin film which, like sample A-90, could lead to appreciable shrinkage of the film and thus the formation of significant porosity. In a very thin film (eg., a single layer film), the shrinkage is often primarily towards the substrate because the adherence of the film at the substrate surface effectively restrains the film from in-plane shrinkage. This restraining effect decreases, and thus the in-plane shrinkage increases, as the distance from the substrate increases.

Thus, for thicker films (eg., a 2-layer film), increased shrinkage (eg., due to pyrolysis, crystallisation, and sintering) towards the film surface could produce porosity in the film, especially near the film surface. This can also explain the pores shown in surface of the A-90 thin films. Furthermore, for sample B-270, the higher initial heat treatment and whatever loss or breakdown of the polymer film may render the initial film weaker such that it less able to withstand whatever shrinkage (contraction) forces that occur in the of the second layer during its gelation, drying, and pyrolysis (up to 270°C). Also, the higher initial heat treatment temperature of 270°C may also result in the initial layer

165

being considerably transformed from its initial gel state such that there is greater incoherency with the polymer structure of the second layer when the latter is deposited and gelated, thus resulting a definite demarcation between the two layers prior to the final 650°C heat treatment step with this possibly persisting after the heat treatment.

The above description is shown schematically in Figure 6-8.

(1) Deposition of uniform precursor on STO substrate by spin coating followed by gelation at 90°C and drying at 270°C. All of the remanent solvent, and some of the polymer phase, are removed:

90°C 270°C

(2) Deposition of second layer of the precursor followed by gelation at 90°C and drying at 270°C. Again, have removal of all of the remanent solvent, and some of the polymer phase occurs (mainly from the top layer):

90°C 270°C

(3) Pyrolysis (at 450°C) followed by crystallization and sintering (at 650°C) of the bulk 2- layer film resulting in significant (differential) shrinkage and associated formation of porosity

450°C 650°C

Figure 6-8. Schematic diagram of showing the stages of deposition, drying, pyrolysis and sintering process of 2-layer thin film for the B-270 heat-treatment sequence

166

After 450°C heat treatment of the initial deposited layer, it was likely that the polymers would have been completely removed (according to DCS result – see Figure

6.1). Thus, significant shrinkage can be observed from the first layer. According to the single layer BFO thin film microstructure in Chapter 5, it is known that no open pores or cracks has been formed during the heat treatment, indicating that the shrinkage of 1- layer is mainly towards substrate. Similarly, no open pores would be formed in the second layer from the removal of polymers or subsequent steps. In addition, any open pores in the top surface of the initial layer would have been filled/covered by the second deposition of precursor. Thus little porosity forms in the layer (and certainly less than the A-90 and B-270 heat treatments). After that, two-layer thin film is crystallized at

650°C together into the dense BFO thin film, as shown in Figure 6-9.

(1) Deposition of uniform precursor on STO substrate by spin coating followed by gelation at 90°C, drying at 270°C, and pyrolysis at 450°C. All of the remanent solvent, and all of the polymer phase, are removed:

90°C 270°C 450°C

(2) Deposition of second layer of the precursor followed by gelation at 90°C and drying at 270°C (not shown), and pyrolysis at 450°C. Again, all of the remanent solvent and all of the polymer phase are removed (mainly from the top layer). Finally the coating is crystallised and sintered at 650°C:

90°C 450°C 650°C

Figure 6-9. Schematic diagram of showing the stages of deposition, drying, pyrolysis and sintering process of 2-layer thin film for the C-450 heat-treatment sequence

167

(d) D-650

In this sample, the second layer is deposited after the growth and crystallization of the first one. In this sense, the first layer plays the same role as the substrate that it could restrain the shrinkage to the thickness direction and thus minimising differential shrinkage and hence cracking and porosity.

(2) Phase composition

According to the literature[17] and the previous DSC analysis (see Figure 6.1), the reduction of Bi3+ to Bi metal takes place at around 200°C. The Bi metal can act as the nucleation centres for subsequent oxidation of bismuth and the formation of bismuth-rich phases, such as Bi25FeO40. These reactions occur at ~ 250°C prior to BFO formation which commences at ~550°C [17, 72]. Therefore, it is probable that the first layer of the A-90, B-270 and C-450 BFO thin films contains such phases prior to the deposition of the second layer. The second heat treatment at these temperatures then may increase the amount of these phases and/or promote their grain growth, thus leading to the formation of clusters bismuth-rich phase(s) and/or bismuth oxide [17]. In the subsequent crystallization process at 650°C, it is likely that these phases would persist, albeit as different but similar Bi-rich phase(s), thus resulting in compositional and phase inhomogeneity in the final film. In particular, for A-90 thin film, the Bi-rich phase is likely to be due to the formation of the precipitates in the second deposition.

After deposition of the second layer, the solvent from second deposition may dissolve the first layer, making it into a thicker film with a higher metal salts concentration.

According to the discussion in Chapter 4, metal salt precipitates may tend to form in the higher concentration precursor and thicker films because the evaporation of solvent is faster than the completion of gelation process. These metal salt precipitates, having an

168

Bi:Fe off-stoichiometric composition, then convert into secondary phases during crystallization at 650°C. These secondary phase defects, in turn, may lead to the inhomogeneous topography and porosities in A-90 as shown in 6-4.

(3) BFO crystallographic orientation

(110)-misorientation was observed in the XRD patterns obtained for the A-90,

B-270, C-450 thin films and this was attributed to the increasing film thickness of these films before crystallization. Other reasons are given below for the individual heat treatments. For the A-90 thin films, only very weak (110) peaks were observed (See

Figure 6.3). It is possible the "coherency" of the polymer layer is such that the film, after the second deposition and drying, is effectively a single, but thicker, film.

Nucleation and crystallisation of the BFO(001) phase at the film-substrate interface would likely continue throughout the film without much homogeneous nucleation of different BFO orientations. The formation of the small amount of BFO(110) orientation is probably just due to loss of crystallographic coherency with increasing film thickness.

For the B-270 heat treatment sequence, the BFO was largely still in an amorphous state following the heat treatment of the first layer at 270°C. Thus, compared with single-layer thin film, homogeneous nucleation and grain growth may commence in the bulk. Alternatively, any structural irregularities caused by the loss of some polymer as pyrolysis begins, or possibly any structural demarcation between the two layers, may have promoted heterogeneous nucleation of whatever crystallographic orientations (such as BFO(110)) prior to nucleation of the BFO(001) orientation at the film/substrate interface and subsequent grain growth of the out-of-plane BFO(001) columnar grains. Thus, misoriented BFO(110) grains may form within the bulk of the films away from the film-substrate interface[110].

169

Compared with B-270 sample, the BFO(110) peaks are weaker in the C-450 thin films (see Figure 6.3) which suggests that less nucleation/crystallisation of BFO(110) has occurred. It is feasible that by 450°C, nucleation of BFO(001) had commenced at the film-substrate interface and that possibly it has extended to some extent through the thickness effectively giving some "guidance" for the formation of this orientation in the second layer. It may also be possible that the first layer may have a better (001) orientation with respect to the substrate than the second layer. Similar to this, it may be possible that during the main crystallisation stage (at 650°C) that grains nucleate and grow at both the film-substrate interface and whatever film-film interface, thus leading to misorientation of the second layer.

For the D-650 sample, the first layer is completely crystallised (by 650°C) prior to the deposition of the second layer. If the first layer is BFO(001) (which it would be based on the results obtained for the single films – see Chapter 5) then it would act as the aligned substrate for the epitaxial alignment of the second layer thus leading to the second layer also having a 2-layer (001)-oriented epitaxial structure. This result suggests that a relatively thick film CSD-derived film with epitaxial orientation could be made by successive depositions by the D-650 heating sequence, that is, each film is fully crystallised prior to the deposition of the next. The extent of BFO(001) alignment would be a function of the quality of the epitaxy within each successive film.

(4) Ferroelectric property

It is found that the ferroelectric properties of 2-layer BFO thin films varied with the deposition-heating sequence. Owing to the existence of pores in A-90 and B-270

BFO thin films, these films had a lower electrical resistance than the other coatings, as shown in Figure 6-6. Additionally, the A-90, B-270, and C-450 thin films each contained a bismuth-rich secondary phase. Bismuth-rich phases, such as δ-Bi2O3, are

170

typically excellent electrical conductors[77], and so their presence the the BFO films would be expected to decrease resistivity and to diminish the ferroelectric properties.

Furthermore, the amount of the bismuth-rich secondary phase decreased from A-90 to

B-270 to C-450 to D-650 (see Figure 6-3) and this correlated with an increase in electrical resistivity of the films (see Figure 6-6). The D-650 thin films showed no bismuth-rich phases in their XRD spectra and, compared with the A-90 to B-270 to

C-450 samples, they exhibited a low resistance and saturated polarization loops with high remanent polarization. This is attributed to the BFO phase purity and dense

BFO(001) epitaxial structure achieved for the D-650 thin films. In addition, resistive switching behaviour has was observed for the D-650 samples in that the resistance shows a sharp increase when the external voltage increases to 4 V. The ferroelectric properties and resistive switching behaviour of BFO epitaxial thin films with various thicknesses are explored in the next chapter.

6.4 Summary

In this chapter, the effect of deposition-heating sequence on film phase composition, microstructure and ferroelectric properties was studied. In addition, porosity was observed in the films initially heat treated at 90°C or 270°C. The formation of pores in the films was attributed to the in-plane direction shrinkage in the thicker films during rapid heating process for pyrolysis and crystallization/sintering.

Misoriented crystallographic structures (BFO(110)) and secondary phases were observed in films for which the first layer was heat treated up to 90°C, 270°C or 450°C prior to the second deposition. The presence of misorientation is explained by the increasing thickness of BFO thin films before crystallization in that the nucleation and growth of equiaxed grains occurs in the bulk of films (rather than nucleation and growth

171

of BFO(001) at the film-substrate interface). Misorientation was also attributed to the structural inhomogeneities which occur in the films heat treated at 90°C, 270°C or

450°C for the first layer. Secondary phases also tended to occur in the films and this was attributed to the formation of Bi-rich phases during the initial lower temperature heat treatments with the amount of these phases then increasing during the second heat treatment due to continued phase transformation or grain growth. Low remanent polarization was observed in samples initially heat treated to 90°C, 270°C and 450°C and this was generally attributed to the presence of pores, secondary phases, or crystallographic misorientation of the film. Phase-pure, (001)-oriented epitaxial BFO thin films were able to be obtained only for the 2-layer films heat treated at the maximum temperature of 650°C after the initial deposition These films had square

2 hysteresis loops with a high remanent polarization 2Pr of ~73.8 µC/cm .

The D-650 heat treatment sequence was thus consider to be the optimal deposition sequence for preparing multilayer (thick) films. The sequence is restated as follows:

1) After the deposition of initial layer by spin coating, the four 2-layer films were successively heated in an ambient air atmosphere to 90°C (for gelation) and then 270°C (for drying), followed by rapidly heating in an oxygen atmosphere to 450°C (for pyrolysis) and then to 650°C (for crystallisation and sintering).

2) A second deposition is then done and the heat treatment process is repeated.

Further depositions and heat treatment can be then be done to fabricate thicker films. Using this optimal deposition sequence, BFO epitaxial thin-films with 3 different thicknesses were be prepared and characterization as discussed in next Chapter.

172

Chapter 7.

Characterization of CSD Derived Epitaxial BFO Thin-Films

7.1 Introduction

Bismuth ferrite (BiFeO3, BFO) has attracted recent attention due to its multifunctional properties, including multiferroism[5, 34], resistive switching[172] and photovoltaic effects[64]. In particular, epitaxial BFO has been shown to demonstrate giant polarization[13, 115], polarization-mediated resistive switching[6] and unique magnetic properties[68]. Until now, the most popular methods to obtain epitaxial BFO films with robust properties have been pulsed laser deposition[6, 9, 10] and radio frequency (RF) sputtering[11, 12]. Films made using these methods have been reported to have a high spontaneous polarization of up to 130 µC/cm2,[13] and a typical switchable diode effect[6, 14]. However, to date, it has been a challenge to prepare high quality epitaxial BFO thin films through chemical solution deposition (CSD) routes such that few BFO thin films with robust ferroelectric properties has been reported using this technique.

In this chapter, using the optimized precursor and annealing conditions reported in previous chapters, epitaxial BFO thin films with varying thickness were prepared by

1-layer, 2-layer and 3-layer deposition. Stoichiometric precursor derived epitaxial (001)

BFO thin films exhibiting robust (square) polarization hysteresis loops, high dielectric constant, strong piezoelectric response and distinct diode effects were fabricated. The thickness dependence of ferroelectric behaviour and diode effect in BFO thin films was

173

also systematically investigated. The research route of this chapter is illustrated in

Figure 7-1.

BFO Thin-Film Preparation Preparation of epitaxial BFO thin films of three thicknesses  1-layer Precursor  2-layer

preparation  3-layer

Gel film BFO Thin-Film Performance  Phase composition Crystallized BFO  Microstructure Thin-Film  Domain structure  Ferroelectric polarization  Dielectric property Characterization  Resistive switching and diode behaviour

CSD derived BFO thin film Characterization

Figure 7-1. Illustration of Chapter 7 research route

7.2 Experimental Section

7.2.1 Thin Film Preparation

A lanthanum strontium manganite (La0.67Sr0.33MnO3, LSMO) layer of approximately 20-30 nm was deposited by PLD on (001) STO as the bottom electrode.

0.25 M stoichiometric precursor was used to prepare BFO thin films by spin coating.

0.015 mol bismuth nitrate and 0.015 mol iron nitrate were dissolved in 40 ml 2- methoxyethanol. 20 ml acetic anhydride was added under constant stirring to the solution to form a homogeneous 0.25 M BiFeO3 precursor solution. BFO thin films were then produced by dropping a small amount of solution onto the substrate preheated to 70°C followed by spin-coating at 3000 rpm for 30 seconds. As-deposited films were

174

heated in air on a hot plate at 90°C for 1 minute followed by 270°C for 3 minutes. The films were then rapidly heated in approximately 5 seconds by putting the sample into a preheated tube furnace in an oxygen atmosphere and then held at 450°C for 30 minutes and subsequently rapidly heated to 650°C for 30 minutes. The deposition and annealing processes were carried out for 1, 2 and 3 times to produce 1-layer, 2-layer and 3-layer thin-films. As determined from FIB sectioning and examination by SEM, the 1, 2, and

3 layer films corresponded to thicknesses of 40 nm, 70 nm, and 150 nm, respectively, in the crystallized films.

7.2.2 Composition Evolution and Structural Characterization

The phase compositions and crystallographic alignment of the resultant films were analysed by X-ray diffraction using Cu Kα radiation at 45 kV/40 mA over a 2θ angular range of 15–75° at a scanning rate of 5°(2θ)/min. XRD phi-scan was used to examine the crystallographic structure relationship between the heat-treated BFO thin film and the substrate. The film was tilt for 45° chi angle to obtain the X-ray diffraction pattern for the (110) plane phi-scan in a range of 360° at a scanning rate of 60°/min.

Cross-sectional samples, approximately 100 nm thick, were prepared from the heat treated thin films using focused ion beam milling. Cross sectional bright field microstructure images and selective area electron diffraction (SAED) patterns were obtained by transmission electron microscopy.

A commercial scanning probe microscope was used for both atomic force microscopy to observe the films surface microstructures and piezoelectric force microscopy to study the film domain structures. The thin film topography were observed in a region of 3 µm × 3 µm and 1.5 µm × 1.5 µm and the domain structures of corresponding 1.5 µm × 1.5 µm region were observed by PFM. Conductive Ti-Ir coated

175

silicon cantilevers (ASYELEC-01, Asylum Research, US) were used for PFM imaging, polarization switching and piezoelectric amplitude/hysteresis loop studies.

7.2.3 Macro-Scale Ferroelectric and Dielectric Characterizations

Au/Ti (60 nm/5 nm in thickness) was coated on the sample surface as the top electrode through photolithography and metal evaporation. The size of each top electrode pad is 23 µm x 23 µm. In the macro-scale ferroelectric and dielectric measurements, thin-film samples were put on a probe station and probes connecting the testers were connected with LSMO bottom electrode and Au/Ti top electrode. Bias was applied through top electrode to the bottom one via the probes.

Ferroelectric polarization hysteresis loops of 40 nm, 70 nm and 150 nm BFO thin films were acquired by a ferroelectric testing system at room temperature at a frequency of 10 kHz. Pulsed polarization (PUND) measurements of 70 nm and 150 nm

BFO thin films were carried out to determine the pulsed remanent polarization (ΔP = P*

(switched polarization) −P^ (non-switched polarization) =~2Pr) as a function of electric field at a pulse width of 100 µs at room temperature. PUND as a function of pulse width

(from 1 µs to 104 µs) were measured for 70 nm and 150 nm thin films at room temperature using the voltages which give saturated polarization of each film.

Fatigue measurements were carried out on the 70 nm and 150 nm thin films at

100 kHz at room temperature. The applied voltages for obtaining saturated polarizations were used and the ΔP was measured at a pulse width of 100 µs for both samples.

The dielectric constant and dissipation factor of 70 nm and 150 nm BFO thin films were measured as a function of electric field at 1 MHz using a precision impedance analyzer. Both curves were measured by a voltage sweep first from +V bias to -V (down sweep) followed by a voltage sweep in the opposite direction (up sweep)

176

(V=3 V and 5V for 70 nm and 150 nm respectively). Curves of 40 nm BFO thin film were not obtained because of its high leakage current.

Macro-scale PFM amplitude loops of 40 nm, 70 nm and 150 nm BFO thin films were obtained by using scanning probe microscopy. The tip was loaded on the Au/Ti top electrode and the voltage (V=2V, 6V and 9V for 40 nm, 70 nm and 150 nm respectively) was applied through sample to the top electrode. The measurement frequency is 0.2 Hz and a small AC voltage (500 mV) and DC bias sweep of -V to +V and back to -V were applied.

Leakage current curves of BFO thin films through Au/Ti top electrode were measured by using a Keithley 2400 SourceMeter. The curves were measured by a triangular ramp voltage sweep from –V to +V and back to –V (V=5 V, 7V and 10 V for

40 nm, 70 nm and 150 nm BFO thin films respectively).

7.2.4 Micro-Scale Ferroelectric Characterizations

The micro-scale PFM amplitude and phase loops of BFO thin films were obtained by using scanning probe microscope. The PFM tip acted as the top electrode in this measurement and the electric bias was applied through the bottom electrode. The measurement frequency is 0.2 Hz and a small AC voltage (500 mV) and DC bias sweep of -V to +V and back to -V (V=7 V, 7 V and 8 V for 40 nm, 70 nm and 150 nm BFO thin films respectively) were applied to the samples via the PFM tip on BFO thin film.

Current maps of BFO thin films were obtained by using conductive-AFM (C-

AFM) technique using scanning probe microscope. To investigate the polarization- mediated resistive switching and diode behaviour, A 3 µm × 3 µm square area in BFO thin film was switched first downward by applying a DC bias of -6 V, -6 V, -9 V on the

177

40 nm, 70 nm and 150 nm films, respectively. After that, an opposite DC bias of +6 V,

+6 V, +9 V, respectively, was applied in a square region of 1 µm × 1 µm in the centre of the previous 3 µm × 3 µm switched square areas to switch films polarization back upward. The biases used to acquire the current maps were +1.5 V, +2 V and +2 V in 40 nm, 70 nm and 150 nm BFO thin films, respectively. All the current maps were acquired from the third scan 1 hour after the writing to exclude the possibility of domain switching current. Diamond-coated probes (CDT-NCHR, Nanoworld, Switzerland) were used for conductive AFM current mapping and local I-V curve measurement.

7.3 Results and Discussion

7.3.1 Composition Evolution and Structural Characteristics

7.3.1.1 XRD Investigation

Figure 7-2 shows the XRD patterns of 40 nm, 70 nm and 150 nm BFO thin films.

Substrate sharp peaks STO(00L) (L=1, 2 or 3) can be observed at ~22.8°, ~46.5° and

~72.5° (2θ). BFO(00L) peaks at ~22.5°, ~45.0-46.0° and ~70.5-71.5° (2θ) are found on the left shoulder of these STO(00L) peaks. Because the LSMO peaks are close to STO peaks, only LSMO (002) and (003) peaks can be observed from the right shoulder of

STO, at ~47.5° and ~73.5° (2θ) respectively. No misorientation nor secondary phases are observed in the XRD patterns of these three thin films.

As shown in the right inset to Figure 7-2, a 360° phi scan of BFO thin-films on

(110) plane gives four set sharp peaks with 90° intervals and these are aligned well with the substrate peaks. This result shows the 4-fold symmetry of BFO thin film crystal structure and indicates that the BFO film has the same in-plane lattice parameter as the

178

STO(001) substrate, thus indicating the (001)-oriented epitaxial nature of the BFO film samples.

A shift of BFO (00L) peaks towards lower 2 values is also observed as the reduction of film thickness, as shown in Figure 7-2 left inset. The BFO bulk material presents a rhombohedral structure (a=b=c=0.396 nm, α=β=γ=89.3°-89.4°) and pseudocubic crystallographic structure is often used for the lattice parameter calculation for BFO thin film. Thus, according to the 2θ position of BFO(00L) peak and Bragg’s law, the out-of-plane c-lattice parameters of 1, 2 and 3-layer BFO thin films can be obtained from XRD data, which are 0.401 nm, 0.399 nm, and 0.396 nm for the 40 nm,

70 nm, and 150 nm BFO thin films, respectively. Compared with the c-lattice parameter of bulk BFO (~0.396 nm), the 40 nm and 70 nm BFO films show a larger value, indicating an out-of-plane elongation and a corresponding in-plane contraction. This is because the smaller in-plane lattice parameter of substrate STO (a=0.391 nm) and buffer layer LSMO (a=0.385 nm) impose in-plane compressive strain in the thinner BFO thin

[5, 173, 174] films due to the difference of lattice parameters of film and substrate .

179

7.3.1.2 TEM Investigation

The thicknesses of three films are ~40 nm, ~70 nm and ~150 nm for the 1-layer,

2-layer and 3-layer thin films, respectively, as measured from the TEM images of the cross-sectional microstructures shown in Figure 7-3(a). For all three thin films, gold- coating was used to increase the film conductivity for better observation, and this is observed as a dark coating on the top surface of each thin film. Glue used to fix samples can also be observed (as shown in 3-layer sample in Figure 7-3(a)).

180

(a) Cross section (b) Diffraction pattern

1-layer

2-layer

3-layer

Figure 7-3. (a) Cross-section microstructures and (b) corresponding SAED patterns of films with various thicknesses of 1-layer, 2-layer and 3-layer BFO thin films observed by TEM

181

In the cross-sections of 40 and 70 nm thin films, no bright contrast was observed thus indicating the absence of pores in the film. This pore-free microstructure was observed over a linear distance of 500 nm along the film cross-section and no obvious grain boundaries were observed for the 40 nm thin film cross-section. It was concluded from this that the films were of high relative bulk density, close to 100%, The contrast observed from the film, the bright zone of the 40 nm film surface, and the dark zone of

70 nm BFO/LSMO interface were interpreted to be artifacts attributable to the damage made during FIB sample preparation or to the tilt angle during TEM imaging.

In the 150 nm thin film, the grain boundaries can be observed along the out-of plane direction and the surface roughness is larger than the other two films as judged from the cross-section images. Nevertheless, no pores bigger than 5 nm are observed in the film bulk and no laminations can be observed in the multi-layer deposition, indicating the dense nature of the thin film.

Figure 7-3(b) shows corresponding SAED pattern of BFO/LSMO/STO(001) cross-section. As the LSMO is only of 20-30 nm thick and its lattice paramter is close to

BFO and STO, only STO and BFO diffraction pattern spots can be clearly observed.

For all three samples, the spots indicate the single crystal nature of the thin film and the reflections of the STO and BFO planes align with each other along the [0K0] and [00L] shows the epitaxial nature of BFO on STO(001) substrate.

182

7.3.1.3 AFM and PFM Investigation

The AFM and PFM images of BFO thin films of three thicknesses are shown in

Figure 7-4. Figure 7-4(a) and (b) show surface morphologies of 40 nm, 70 nm and 150 nm BFO thin films in a region of 3 µm × 3 µm and 1.5 µm × 1.5 µm, respectively. AFM surface morphologies, along with TEM images of the film cross-sections, show the films to be of relatively high bulk density (no open porosity is observed) and with a linear-intercept grain size of ~150 nm independent of film thickness. The root mean square (RMS) surface roughness increases slightly with film thickness from ~3.2 nm for

40 nm thin film to ~4.2 nm for 150 nm BFO thin films, which agrees with the TEM cross-section observation. Although these roughness values are higher than those found for atomically sharp BFO surfaces obtained by PLD technique, there are no particles or voids over tens of microns as shown in Figure 7-4(c) and (d), which is sufficient for most manufactured devices. Fractal-like domain structures with two-phase contrast are observed by PFM in all three samples, indicating a poly-domain state. The PFM images of 40 nm and 70 nm BFO thin films indicate the existence of ordered domain structures with a tendency to be aligned. As the film thickness increases to 150 nm, the domain structure appears less ordered and exhibits a weaker dependence on grain structure. The average domain size values are acquired from two 5 µm × 5 µm PFM images and are calculated by using auto-correlation transformation[175] from both (110) and ( 11̅0) direction of the out-of-plane PFM phase images. The value first increases from ~39 nm to ~54 nm with the film thickness increase from 40 nm to 70 nm, but it drops to ~28 nm in the 150 nm BFO thin film. This result is different from other reports that domain size decreases with the decrease of epitaxial film thickness[30].

183

(c) PFM amplitude (d) PFM Phase (a) Topography (b) Topography Out-of-plane Out-of-plane 3 µm × 3 µm 1.5 µm × 1.5 µm 1.5 µm × 1.5 µm 1.5 µm × 1.5 µm (I)

40 nm

(II)

70 nm

(III)

150 nm

Figure 7-4. AFM (a) (3 µm × 3 µm ), (b) (1.5 µm × 1.5 µm); out-of–plane PFM (c) amplitude(1.5 µm × 1.5 µm) and (d) phase (1.5 µm × 1.5 µm) images of BiFeO3 films on LSMO/STO(001) substrate (I - 40 nm; II - 70 nm; III - 150 nm)

7.3.2 Macroscale Ferroelectric Properties

7.3.2.1 Ferroelectric Hysteresis Loops

Ferroelectric hysteresis loops acquired for all three sets of samples are shown in

Figure 7-5. Unsaturated and leaky ferroelectric polarization hysteresis loops with low

2 remanent polarization 2Pr of a mere 9.4 µC/cm are obtained for 40 nm thin films at room temperature. In contrast, the 70 nm and 150 nm thin films exhibit well-saturated and square-like ferroelectric polarization loops with no evidence of significant leakage current. The remanent polarization 2Pr and coercive field 2Ec of the 70 nm and 150 nm thin films are 73.5 µC/cm2 and 410 kV/cm, and 99.8 µC/cm2 and 193 kV/cm, respectively. The above results are comparatively stable and independent of samples

184

and top electrodes as confirmed by repeating the measurement on at least 3 different top electrodes on 2 different samples (for each thickness). The high leakage current is only observed for the 40 nm thin film. Such leakage current has been proposed to be a result of the present of oxygen vacancies, which then act as the donor impurity in BFO thin film giving rise to n-type semiconducting behaviour[14, 176]. Thus, only 70 nm and 150 nm thin film are used to subsequently investigate the pulsed polarization, fatigue and dielectric constant characterizations.

Figure 7-5. Ferroelectric polarization hysteresis loops of 40 nm, 70 nm and 150 nm BFO thin films, measured at room temperature at 10 kHz;

In particular, the ferroelectric properties of 150 nm BFO thin-film are comparable with the PLD derived BFO (001)-oriented epitaxial thin film (200 nm in thickness)[5] , which has a remanent polarization 2Pr of approximately 110 µC/cm2 and

ΔP of around 100 µC/cm2 at room temperature. Moreover, our CSD derived BFO thin- film shows a smaller coercive field (2Ec of 193.3 kV/cm) compared with typically PLD derived counterparts (~300 kV/cm).

185

7.3.2.2 PUND

To further investigate the effect of leakage on the polarization behaviour of 70 nm and 150 nm BFO thin films, pulsed polarization (PUND) as a function of electric field was measured. The results are obtained from an average of 3 measurements on 3 different top electrodes. As shown in Figure 7-6, the ΔP of 70 nm and 150 nm BFO thin films start to increase at ~228 kV/cm and ~100 kV/cm and saturate at ~428 kV/cm and

~300 kV/cm, respectively. The saturated ΔP for the 70 nm film is ~56 µC/cm2 which is

2 somewhat lower than the 2Pr value of 73.5 µC/cm derived from the hysteresis loop.

However, the saturated ΔP for the 150 nm film is ~105 µC/cm2 which agrees reasonably

2 well with the 2Pr value of 99.8 µC/cm . These results indicate that the 70 nm BFO thin film has a higher leakage current than the 150 nm film and, more generally, that the effect of leakage on the ferroelectric behaviour becomes weaker as film thickness increases[177]. It is considered that this is because, as the film thickness increases, oxygen vacancies or electrons released from oxygen vacancies are more difficult to move and thus the resistivity of the thin film is enhanced.

Figure 7-6 . Room temperature pulsed polarization of 70 nm and 150 nm BFO thin films as a function of electric field

186

Figure 7-7 shows the room-temperature pulsed remanent polarization (ΔP) value of the 70 nm and 150 nm BFO thin films as a function of pulse width. In this measurement, applied voltages of 3.2V (457 kV/cm) and 4.5V (300 kV/cm) were used for the 70 nm and 150 nm thin films, respectively (to give saturated polarization of each film). It is found that the pulse width of 150 nm thin film is relatively stable with the ΔP decreased by only ~18% from 1000 µs to 1 µs. The pulse width dependence of the 70 nm thin film is more evident in that its ΔP first decreases by ~32% from 1000 µs to 3 µs and then a further decrease of ~59% from 3 µs to 1 µs.

Figure 7-7. Room temperature pulsed polarization of 70 nm and 150 nm BFO thin films as a function of pulse width

187

7.3.2.3 Fatigue

Finally, fatigue measurements were carried out on the 70 nm and 150 nm thin films as shown in Figure 7-8. The results are obtained from the average of 3 measurements on different top electrodes. The 70 nm and 150 nm thin films both demonstrated good high-cycle fatigue resistance that fatigue endurance of both coatings are good up to 106 cycles, but then rapidly decreases to ~65% in ΔP after 107-108 cycles

It should be noted that that the top electrodes were relatively small (23 µm × 23 µm) and there was visual evidence of damage to them after ~106 cycles, and so the degradation in remanent polarization could also be attributed to the fatigue of the top electrodes.

Figure 7-8. Pulsed polarization fatigue behaviour for the 70 nm and 150 nm BFO thin films at room temperature

188

7.3.2.4 Macroscale Dielectric Properties

Figure 7-9 shows the room temperature relative dielectric constant and loss tangent curves of the 70 nm and 150 nm BFO thin films, as a function of electric field.

Curves of 40 nm BFO thin film were not obtained because of its high leakage current.

The relative dielectric constants of 70 nm and 150 nm BFO thin films peak at 314 and

613 respectively during the down sweep at 1 MHz at room temperature. The decrease of the dielectric constant with reducing thickness can be explained either as a size effect intrinsic to the film structure or due to the low dielectric constant of the interface layer which dominates the capacitance property of thinner films[178-180]. Both thin films overall show a low dissipation factor of below 0.05. The 2Ec of 70 nm and 150 nm thin films derived from dielectric constant, which are ~74.3 kV/cm and ~80.0 kV/cm, respectively are in good agreement with dissipation data.

Figure 7-9. Dielectric constant and loss tangent curves of 70 nm and 150 nm BFO thin films, measured at room temperature at 1 MHz

189

7.3.2.5 PFM Amplitude Hysteresis Loop through Top Electrodes

Figure 7-10 shows the macro-scale PFM amplitude loops of 40 nm, 70nm and

150 nm BFO thin films. It is found that the amplitude increases significantly with the film thickness increase, which shows the same tendency as the dielectric constant. The

2Ec of 40 nm, 70 nm and 150 nm thin films are 65.78 kV/cm, 85.56 kV/cm and

89.77 kV/cm, respectively. The coercive fields of 70 nm and 150 nm thin films are close to those derived from dielectric constant and dissipation curves in Figure 7-9, but different from the values in ferroelectric polarization hysteresis in Figure 7-5, which have much higher coercive fields for saturated loops. This inconsistency stems from the different voltage sweep frequencies for each type of measurements. In C-V curve and

PFM amplitude measurement, an AC oscillation of 500 mV is used as a tickle over a

DC bias, while a simple triangle voltage sweep is used for polarization hysteresis measurement.

Figure 7-10. PFM amplitude hysteresis loops of BFO thin films measured through Au/Ti top electrodes

190

7.3.2.6 Current-Voltage Behaviour

Figure 7-11 shows the I-V curve of three samples. All the curves were measured by a triangular ramp voltage sweep from –V to +V and back to –V, as the arrows shown in the figures. The black curves are the original high resistive state (HRS, OFF state) and the red counterparts are the final low resistive state (LRS, ON state). Open hysteresis loops are observed from all three samples, indicating a resistive switching behaviour. The LRS and HRS current shifts at zero bias with ON/OFF current ratios of

332, 146 and 5.2 for 40 nm, 70 nm and 150 nm, respectively. This decreasing tendency

with increase thickness is also observed when |V| less than |Vset|. For example, at 1 V, the ON/OFF current ratios are 38, 11 and 4.5 for 40 nm, 70 nm and 150 nm, respectively.

191

(a)

(b)

(c)

Figure 7-11. Typical I-V curves of the BFO/LSMO/STO thin films with thickness of (a) 40 nm, (b) 70 nm and (c) 150 nm

192

7.3.3 Local Electromechanical and Diode Behaviour Investigations via Atomic Force Microscopy

7.3.3.1 Local PFM Phase and Amplitude Hysteresis Loops

Figure 7-12 (a) and (b) show the local piezoelectric force microscope switching loops (phase and amplitude) of three samples. The PFM tip acted as the top electrode in this measurement and the electric bias was applied through the bottom electrode. A square loop indicating a sharp 180° change in PFM phase and a butterfly loop in PFM amplitude are observed in all three samples, typical of the ferroelectric switching characteristics. The coercive field 2Ec of three films derived from the local piezoresponse loops, are 1724.8 kV/cm, 1126.7 kV/cm and 651.8 kV/cm for 40 nm, 70 nm and 150 nm BFO thin films, respectively.

The coercive fields here are different from that obtained from macro-scale measurements. This can be explained as the differences between macro- and micro- scale domain switching mechanisms[170, 171]. The domain switching begins with the growth of residual domains at interfacial defects in the macro-scale polarization hysteresis measurement while the new domain nucleation starts near the PFM tip in local measurement. Further grain boundaries, variation of chemical compositions, surface roughness and micro pores within films all may contribute to the differences of domain switch during measurement by the two methods. Additionally, it is known that the domain walls influence the domain switching process and hence the measured coercive field. The higher domain density could facilitate a greater number of nucleation sites for switched domain nucleation thereby reducing the coercive field.

This may contribute to the lower coercive field in the saturated global ferroelectric polarization loops of 70 nm and 150 nm thin films (as shown in Figure 7-5), when more domain walls are involved in the switching regions. In particular, the global coercive

193

field 2Ec dereases from 410 kV/cm (70 nm BFO thin film) to 193 kV/cm (150 nm BFO thin film) with the domain size drops from ~54 nm (70 nm BFO thin film) to ~28 nm

(150 nm BFO thin film), which further demonstrates the effect of domain density on the coercive field of our BFO thin films.

(a)

(b)

Figure 7-12. Local piezoresponse force microscope phase (a) and amplitude (b) hysteresis loops of 40 nm, 70 nm and 150 nm BFO/LSMO/STO(001) thin films

194

It was also observed that the piezoresponse amplitude of the BFO thin films increases with an increasing film thickness. This could be an effect of the clamping by the substrate for the thinner films, because of which the out-of-plane displacement along the c-axis direction may be restrained[5, 19, 172]. The thickness dependence of piezoresponse effect corresponds to that of the dielectric constant in Figure 7-9 and the macro-scale PFM amplitude loop in Figure 7-10.

7.3.3.2 Current Map and Local I-V Curves by Conductive AFM

Although the presence of oxygen vacancies in BFO thin films increases the leakage current, it is believed to play a positive role in resistive switching and diode effects in BFO thin films[91, 181]. Macro scale I-V curves in Figure 7-11 have demonstrated the resistive switching in BFO thin films. To further study this behaviour and the diode effect in the CSD derived BFO thin films, current maps of virgin BFO thin films were acquired by a conductive AFM technique. Figure 7-13 show the PFM phase image and corresponding current map of a 40 nm BFO thin film. The yellow regions in the PFM image indicate the downward polarization and these are in good agreement with higher conductivity regions (bright areas) in the current map. This result indicates that there is a strong link between polarization direction and resistivity in the

BFO thin film. However, no such phenomenon was observed in the 70 nm and 150 nm

BFO thin films. Also, this correlates to the polarization hysteresis loops in Figure 3 in which the 40 nm BFO thin film shows a lower resistance compared with 70 nm and

150 nm films.

195

(a) (b)

Figure 7-13. (a) PFM image of natural domain patterns of 40 nm BFO thin film (purple regions – polarization switched upward; yellow regions – polarization switched downward); (b) Current map obtained in the same region as in (a) (bright region – higher current)

To further investigate the polarization-mediated resistive switching and diode behaviour, Figure 7-14(a) shows the resultant PFM images of the 40 nm, 70 nm and 150 nm BFO thin films after this domain switching in a total area of 5 µm × 5 µm. A discernible contrast can be observed between the upward and downward domain switched regions in all three samples. The corresponding current maps of the regions obtained by applying a positive bias through sample to tip using conductive AFM technique are also shown in Figure 7-14(b). Compared to the 40 nm BFO thin film, conductivity of the downwards-polarization regions becomes weaker as film thickness increased to 70 nm and then 150 nm.

196

(a) PFM phase images (b) Current maps 5 µm × 5 µm 5 µm × 5 µm (I) 40 nm -6V

+6V

(II) 70 nm -6 V

+6 V

(III) 150 nm -9 V

+9 V

197

Figure 7-15 compares the local I-V curves of upward and downward poled regions of three BFO thin films. I-V curves of all three samples show the typical diode like characteristics. Current in the poling direction shows a dramatic reduction in resistance. This result well demonstrates that the direction of diode-effect observed in

BFO thin films can be switched and controlled by the ferroelectric polarization directions.

Figure 7-15. I-V curve for two opposite polarization directions of 40 nm, 70 nm and 150 nm BFO thin films measured by C-AFM

The ferroresistive switching behaviour in the BFO thin films prepared in this work can be explained by the modulation of Schottky-like barriers in the film-electrode interfaces via electromigration of oxygen vacancies[6, 14, 91, 182]. In the polarized thin film, oxygen vacancies can release electrons which move to neutralize the positive bound charges in the interface. Oxygen vacancies themselves become positive charged because of loss of electrons and form in the interface with negative bounds charges. The

198

movement of electrons and oxygen vacancies towards charged interface could effectively result in downward and upward bending of band in the upper and lower interfaces thereby reducing the Schottky-like barrier and enhancing the diode effect towards the polarized direction.

7.4 Summary

In this chapter, effect of thickness on film microstructure and ferroelectric properties were investigated. (001)-oriented epitaxial BFO thin films of 3 thicknesses were obtained by multi-layer deposition after annealing at 650°C of previous layers.

Robust ferroelectric properties with square hysteresis loops, low leakage current and high-cycle polarization fatigue resistance for the as-prepared BFO thin films were obtained at room temperature. The best properties were found for the 150 nm thick

2 films, notably, these having values of 2Pr=99.8 µC/cm , peak dielectric constant of 613 and a loss tangent of ~0.045. In contrast, the resistive switching behaviour becomes weaker with increasing film thickness, with the ON/OFF ratio at 1 V reducing from 38 in 40 nm thin film to 4.5 in 150 nm thin film. It is thought the presence of oxygen vacancies contribute to the resistive switching and switchable diode effect, which can be controlled by ferroelectric polarization direction. Having said this, the above electrical properties of the CSD derived BFO thin films developed in this work shows the potential to eliminate the gap between CSD and physical deposition methods, such as

PLD and RF sputtering, thereby promoting the wide applications of BFO for industrial mass manufacture.

199

Chapter 8.

Conclusions and Future Work

8.1 Thesis Conclusions

In this thesis, a chemical solution deposition and heat-treatment method for the synthesis of bismuth ferrite thin films on perovskite substrates for ferroelectric applications was investigated and developed The method was based on a non-aqueous organic precursor system used previously for other ceramics and, specifically for this work, involved Bi and Fe nitrates, 2-methoxyethanol, and acetic anhydride. The principal objectives for the work were to:

(1) Obtain fundamental understanding of the mechanisms underlying the non-aqueous

CSD process as applied to BFO thin films;

(2) Optimise the CSD processing and heat-treatment route to synthesise high-quality

phase-pure epitaxial BFO(001) thin films on perovskite (STO) substrates; and,

(3) Demonstrate the capability of the CSD method for the low-cost commercial

fabrication of high-quality BFO thin-film ferroelectric devices.

The work involved systematic study of the precursor formulation, spin coating deposition conditions, gelation process, pyrolysis-crystallisation-sintering heat treatment conditions, and multi-layer film deposition. The work was fundamental in nature in that it sought to elucidate key understanding of the relevant mechanisms or behaviour underpinning each of these areas. However, the work was also developmental in that it progressively optimised each of these areas so as to produce an

200

accurate and reliable non aqueous route for synthesis of high quality BFO thin films for ferroelectric devices.

Parallel to the study and development of the non-aqueous CSD route, extensive study and characterisation of the microstructural, ferroelectric, and piezoelectric properties of the synthesized BFO thin films was undertaken including surface and cross-sectional topography, surface and cross-sectional phase composition, ferroelectric domain structure, ferroelectric polarization, piezoelectric behaviour, current-voltage behaviour, polarization mediated resistive switching behaviour, and diode effects.

Significantly, the microstructural and electromechanical characteristics of the

BFO thin films made using the optimised CSD process developed in this work were comparable with those of well established (and expensive) physical deposition processes such as pulsed laser deposition and sputter deposition. The main outcomes of this thesis are summarized as follows:

(1) The precursor for chemical solution deposition of BFO thin films is prepared by

mixing the starting materials of bismuth nitrate, iron nitrate, 2-methoxyethanol and

acetic anhydride. During the gelation process, 2-methoxyethanol and metal nitrates

react together to form metal-2-methoxyethanoxides. In addition, esters are formed

from 2-methoxyethanol and acetic anhydride during heating, and these can

effectively influence the viscosity and vapour pressure of the precursor. Following

deposition of the precursor, gelation occurs by the further condensation reactions

between the 2-methoxyethnoxide, metal-2-methoxyethanoxides, and esters. A

competition exists between gelation and solvent evaporation which can, depending

on the temperature and time, lead to either gel film formation, metal salt

precipitation in the gel film, or powdery non-gel coatings. Uniform, transparent

gel thin films can be prepared under a moderate heating condition of ~70-90°C for

201

which the gelation process finishes before the all of the organic solvent has

evaporated. At lower temperatures, crystalline precipitates of metal salts tended to

form in the gelated films. This was attributed to the gelation rate being

significantly slow relative to the evaporation rate such that the loss of solvent

increases the concentration of metal salts to saturation prior to gelation thereby

precipitating metal salts in the film. Such salts promote deviation from BiFeO3

stoichiometry in the final crystallized films and, in particular, lead to the formation

of Bi-rich phases in the microstructure. The formation of such phases was shown

to significantly degrade the ferroelectric properties of the films. This work

suggests that a delicate balance between gelation and evaporation is needed for the

obtainment of transparent, precipitate free gel films. Gelation could be favoured by

decreasing metal salt concentration in the precursor so as to reduce the net vapour

pressure of the precursor and thus reducing the evaporation rate of the precursor.

However, this tended to result in poor surface coverage of the films following

crystallization due to the reduced effective solids loading. A more effective means

avoiding precipitation was to preheat the precursor at the time of deposition so as

to increase both the saturated solubility of metal salts in the precursor and the

gelation rate of the precursor.

(2) Fundamental understanding of the gelation process enabled the following

optimised CSD processing and heat-treatment route for synthesise high-quality

phase-pure epitaxial BFO(001) thin films on perovskite (STO) substrates to be

developed: 0.25 M stoichiometric (Bi:Fe=1:1) precursor solution deposited by spin

coating on BFO(001) substrate preheated at 90°C then held at this temperature for

30 mins in an ambient air atmosphere (for gelation) and then at 270°C (for drying),

followed by rapidly heating in an oxygen atmosphere to 450°C (for pyrolysis) for

202

30 minutes and then to 650°C for 30 minutes (for crystallisation and sintering).

This route produced single-layer thin films having a thickness of the order of

40 nm. Similar high-quality films of greater thickness could be produced by this

method simply by repeating the entire deposition-gelation-heat-treatment route to

obtain the required thickness. Defect free, 3-layer films having satisfactory

ferroelectric properties could be routinely made by this method.

(3) The ferroelectric and electromechanical behaviour of BFO films made by the

optimised route were determined as a function of film thickness. Epitaxial (001)-

oriented BFO thin film of 40 nm, 70 nm and 150 nm thickness obtained by multi-

layer deposition. The as-prepared BFO thin films exhibited robust ferroelectric

properties with square hysteresis loops, low leakage current, and high-cycle

polarization fatigue resistance. The ferroelectric properties improved with film

thickness with the best properties being for the 150 nm thick films which

2 exhibited: 2Pr=99.8 µC/cm , peak dielectric constant of 613, dielectric loss

tangent of ~0.045, and good high-cycle polarisation fatigue resistance of at least

106 cycles. In contrast, the resistive switching behaviour becomes weaker with

increase in film thickness, with the ON/OFF ratio at 1 V reducing from 38 in 40

nm thin films to 4.5 in 150 nm thin films. It is thought the presence of oxygen

vacancies contribute to the resistive switching and switchable diode effect, which

this being controlled by ferroelectric polarization direction.

The objectives and specific achievements of this thesis are summarized in

Table 8-1. The non-aqueous CSD route developed in this work for the synthesis of high-quality phase-pure epitaxial BFO(001) thin films on perovskite (STO) substrates offers an exciting potential as a low-cost, reliable method for the commercial fabrication of high-quality BFO thin-film ferroelectric devices.

203

Table 8-1. Summary of the objective and outcomes of the overall thesis

Objectives Outcomes Chapter 1 To synthesize high-quality Epitaxial BFO thin films with high Overall thesis epitaxial BFO thin films with remanent polarization (up to 2Pr=99.8 robust ferroelectric properties. µC/cm2) and low coercive field (2Ec=193 kV/cm) at room temperature successfully prepared. This result is comparable with PLD derived BFO/LSMO/STO(001) thin films. In addition, an obvious resistive switching and diode effect was achieved in the BFO thin films prepared by the CSD technique. 2 To develop a comprehensive Ester (2-methoxyethyl acetate) is Chapter 4 understanding of the chemistry formed in the precursor before heating. occurring during gelation of initial Gel starts to form during heating when precursor. the bonds between metal nitrates and organic solvents (2-MOE and acetic anhydride) start to form, and followed by condensation and gelation. 3 To obtain homogeneous gel films. Defect-free homogenous gel films Chapter 4 were obtained by depositing 0.25 M precursor sol on 70°C-preheated substrate, followed by spin coating at 3000 rpm for 30 seconds, and heating at 90°C for gelation. 4 To obtain the pure phase BFO thin Pure phase BFO thin film with Chapter 5 films with epitaxial structure. epitaxial structure was obtained by rapidly heating the stoichiometric precursor derived dried gel film to 650°C in oxygen atmosphere for 30 minutes. 5 To understand the effect of the Two-layer epitaxial BFO thin films Chapter 6 sequence of heating and multi- were prepared via multi-layer layer deposition on the multi-layer deposition. The second layer was thin-film microstructure, phase deposited after heating the first layer at composition and properties. 90°C for gelation, 270°C for drying, 450°C for pyrolysis and 650°C for crystallization. Two-layer BFO thin films show high remanent polarization 2Pr of ~73.5 µC/cm2. 6 To fabricate epitaxial BFO thin Epitaxial BFO thin films with 3 Chapter 7 films of various thicknesses. different thicknesses (40, 70 and 150 nm) were prepared using the CSD method. 7 To investigate the thickness Square hysteresis loops with high Chapter 7 dependency of ferroelectric remanent polarization were obtained properties of the CSD derived from 150 nm BFO thin films, while a epitaxial BFO thin films. strong resistive switching and diode effect was observed for thinner BFO thin films (40 nm).

204

8.2 Future Work

The optimized CSD process developed in the thesis for fabricating BFO thin film offers exciting potential for the commercial fabrication of high-quality BFO ferroelectric devices. To further demonstrate its prospective usefulness and applicability, the CSD process was used to synthesise the following important recently developed ferroelectric device materials: (i) mixed-phase BFO thin film on (001)-LaAlO3 substrate; and (ii) La-doped BFO thin films (Bi0.9La0.1FeO3 specifically).

8.2.1 Preparation and Characterization of CSD-Derived Mixed-

Phase (001)-BiFeO3 Thin-Film on (001)-LaAlO3 Substrate

BFO thin films of tetragonal and rhombohedral (T and R) mix phase exhibit excellent electromechanical properties arising from them having a morphotropic phase boundary (MPB) effect. The latter is generated by high strain between the BFO film and the substrate. For example, LaAlO3 (LAO), which has a large mismatch of 4.2% with BFO for the in-plane lattice parameter, is often used as the substrate to produce compressive strain in BFO thin film thereby obtaining T and R mixed phase by controlling the film thickness.

In this work, Precursor B (see Section 4.2.1 for precursor composition) was deposited on 70°C preheated (001)-LAO substrate by spin coating at 3000 rpm for 30 s.

The optimized crystallization process used in Chapter 6 for BFO/STO was used here to prepare a single-layer BFO/LAO(001) thin film. As shown in Figures 8-1 and 8-2, the resultant BFO films showed a T and R mixed-phase composition, were epitaxial ((001)- orientation), and had homogenous microstructures and smooth surfaces. Significantly, these results demonstrate that the optimized CSD process can be used to prepare BFO high-strain, epitaxial thin films on substrates with large lattice mismatch. It offers the

205

exciting possibility to study the effect of substrate/strain on the electromechanical properties of BFO thin films using a relatively simple, cost effective synthesis method.

Figure 8-1. XRD patterns of BFO/LAO thin-films after heating at various temperatures (BFO-T: BFO T phase; BFO-R: BFO R phase; O: sample stage from XRD equipment)

450 °C 550 °C 650 °C 5 µm x 5 µm 5 µm x 5 µm 5 µm x 5 µm

450 °C 550 °C 650 °C 1.5 µm x 1.5 µm 1.5 µm x 1.5 µm 1.5 µm x 1.5 µm

Figure 8-2. AFM topographic images of BFO/LAO thin-films after heating at various temperatures

206

8.2.2 Study of La-doped BFO system and Selected Properties

As discussed earlier in this thesis, the CSD process can give precise control of the precursor and thin film chemical composition and offers particular advantage for

BFO thin films in that it can potentially be used to prepare BFO thin films having highly-controlled concentration dopants, the dopants being employed typically to increase dielectric constant, remanent polarization or saturated magnetization[183].

The optimized CSD process developed in this thesis was used to prepare

Bi0.9La0.1FeO3 (BLFO) thin films on STO(001) substrate. Using the Precursor A recipe as the basis, La(NO3)3, Bi(NO3)3, and Fe(NO3)3 were dissolved in 2-MOE in the ratio of

Bi:La:Fe = 0.1:0.9:1.0. Effectively the composition 10% replacement of Bi with La with (Bi+La):Fe in the molar ratio of 1:1. Single-layer BLFO thin films were made using the optimised deposition and heating process used in Chapter 5 for BFO/STO.

As shown in Figure 8-3, BLFO thin-films were prepared by rapid heating at

650ºC-750ºC in an oxygen atmosphere were pure-phase. Above 750°C, the BLFO starts to decompose. This temperature is significantly higher than the temperature at which pure BFO starts to decompose, indicating the effect of La dopant in the composition during crystallization. AFM images of a BLFO thin film after heating at

650°C for 30 mins are shown in Figure 8-4. The BLFO surface is relatively uniform and dense, although there is a "gully-like" morphology which is not understood at this stage. The linear-intercept grain size of the BLFO thin film is ~70 nm which is less than half of the typical size of BFO thin films (~150 nm). The reduction of grain size is most likely due to the La dopant but is not understood at this stage.

This preliminary study suggests that additional metal nitrates can also be added to the Precursor A formulation so as to prepare doped-BFO thin films of controlled composition and potentially of pure phase and with epitaxial structure.

207

Figure 8-3. XRD patterns of Bi0.9La0.1FeO3 / STO(001) thin-films after heating at various temperatures (O: BiFeO3(110); #: Iron oxide; Δ: Bi-rich phase)

3 µm x 3 µm 1 µm x 1 µm

Figure 8-4.AFM images of Bi0.9La0.1FeO3 / STO(001) thin-films

208

Appendices

A1. Study of the Gelation Chemistry and Conditions Using Stoichiometric Precursor

In Chapter 4, 0.35 M precursors (stoichiometric and 10% excess Bi) were used to study the gelation chemistry and gelation conditions of the thin films. This was the initial study of this thesis and was done for the purpose of preparing a homogenous gel film prior to crystallization. However, the subsequent crystallisation studies (Chapters 5 and 6) determined that the 0.25 M precursor concentration was more appropriate for the overall CSD deposition and heat-treatment process. For the purposes of completeness, and to provide logical consistency to the thesis overall, study of the gelation chemistry and gelation conditions of the thin films was completed for the 0.25 M precursors

(stoichiometric). Similar to the work in Chapter 4, the gelation chemistry of film formation was studied by FTIR, the effect of heating conditions on gelation conditions was studied by transmission optical microscopy, and film crystallization and microstructure was studied by AFM and XRD.

A1.1 Precursor Formulations

Precursors AA and BB having formulations given in Table A1-1 were used in this work. These formulations were equivalent to the 0.35 M formulations of Precursor

A and Precursor B, respectively, used in the main thesis (see Chapter 4).

209

Table A1-1. Composition of starting materials of precursor solutions

Bi(NO3)3·5H2O Fe(NO3)3·9H2O 2-MOE Acetic anhydride Precursor A 0.015 mol 0.015 mol 0.55 mol - Chapter 4 (0.35 M) (+ 10% excess) (43 ml) Precursor AA 0.015 mol 0.015 mol 0.75 mol - Appendix (0.25 M) (60 ml) Precursor B 0.015 mol 0.015 mol 0.29 mol 0.21 mol Chapter 4 (0.35 M) (+ 10% excess) (23 ml) (20 ml) Precursor BB 0.015 mol 0.015 mol 0.50 mol 0.21 mol Appendix (0.25 M) (40 ml) (20 ml)

A1.2 FTIR Analysis

A1.2.1 Precursor AA

The FTIR measurement conditions were the same as those described in

Chapter 4 (see Section 4.2.1). Figure A1-1 shows the FTIR pattern of the raw materials and starting Precursor AA prior to any heat treatment. As expected, the FTIR pattern of

Precursor AA is virtually identical to that of Precursor A reported in Chapter 4 (see

Figure 4-3).

Figure A1-1. FT-IR spectra of stoichiometric raw materials and as-prepared precursors (2000-780 cm-1)

210

Figure A1-2 shows the FTIR patterns of Precursor AA on heating at 90°C as a function of time. Again, the patterns are consistent with the equivalent data reported for

Precursor A in Chapter 4 (see Figure 4-4). In particular, it is confirmed here that gelation starts after 60 seconds of heating. The FTIR peaks measured after 75 s of heating are shown in Figure A1-2 and are listed in Table A1-2. This peaks agree well with those reported for Precursor A in Chapter 4 (see Figure 4-4(b) and Table 4-4, respectively).

(a)

(b)

Figure A1-2. FTIR patterns of bismuth ferrite stoichiometric Precursor AA in 2-MOE solvent showing: (a) Precursor AA before heating and after heating for various durations at 90°C, 2000-780 cm-1; (b) 2-MOE and bismuth ferrite starting precursor before heating and after heating for 75 s (1200-950 cm-1)

211

Table A1-2. Summary of FTIR patterns of 2-MOE, stoichiometric Precursor AA before and after heating for 75 s

Wavenumber (cm-1) Functional groups C-O in C-O in C-O in O-H in R-O-R` -CH2-CH2-OH general -CH2-OH 2-MOE 1121 1061 1017 963 Precursor A 1119 1060 1016 963 Starting material Precursor A 1087 - 1009 - After heating for 75 s 1111 1031

Figure A1-3 shows the FTIR patterns of Precursor AA compared with 2MOE and single metal nitrate precursors after heating at 90°C for 75 s. Again, these patterns also agree well those reported in Chapter 4 for Precursor A (see Figure 4-5), where it is found that no decomposition of the pure 2-MOE occurs on heating to 90°C and that the molecular bond structures which occur in each of the Bi+2-MOE, Fe+2-MOE, and

Bi+Fe+2-MOE solutions on heating (gelation) are similar. Thus, the gelation reaction chemistry for Precursor AA can be consider to be the same in principle as that derived in Chapter 4 for Precursor A (see Section 4.3.1).

Figure A1-3. FTIR patterns of 2-MOE and metal nitrates in stoichiometric 2-MOE based precursor (Precursor AA) after heating at 90°C for 75 s (1200-900 cm-1)

212

A1.2.2 Precursor BB

The FTIR patterns of Precursor BB acetic anhydride and 2-MOE are shown in

Figure A1-4. The pattern of Precursor AA is also shown for the purposes of comparison. Comparison of these patterns with their counterparts in Chapter 4 (see

Figure 4-6(b)) reveal no discernable differences thus indicating that the molecule structures (compositions) of Precursor A and Precursor AA, Precursor B and Precursor

BB are the same.

Figure A1-4. FTIR spectra of raw materials and as-prepared precursors (2000-780 cm-1)

Figure A1-5 shows the FTIR patterns of Precursor BB during heating as a function of time at 90°C. Again, the patterns are consistent with the corresponding data for Precursor B(see Figure 4-8). However, it is apparent that Precursor BB takes longer to finish gelation and drying during heating. The reason for this is not understood but it is most probably owing to the effect of decreased metal concentration. Regardless, the peaks locations of C-O groups and C=O groups in the precursor before and after gelation (shown in Figure A1-5 and Table A1-3) agree well with the initial study in

Chapter 4 (Figure 4-8 and Table 4-5, respectively) and so the molecular changes on

213

heating of Precursor BB are considered to be the same as those of Precursor B. Thus, the gelation reaction chemistry for Precursor BB can be consider to be the same in principle as that derived in Chapter 4 for Precursor B (see Section 4.3.1).

(a)

(b) (c)

Figure A1-5. FTIR analysis of stoichiometric bismuth ferrite precursor BB: (a) FTIR patterns of bismuth ferrite starting precursor before and after heating for various times at 90°C (2000-650 cm-1); (b) FTIR patterns of bismuth ferrite starting precursor before heating and after heating for 90 s (1200-950 cm-1); (c) FTIR patterns of bismuth ferrite -1 starting precursor before heating and after heating for 90 s (1850-1650 cm )

214

Table A1-3. Summary of FTIR patterns of precursor BB after heating for 90 s

(a) C-O stretch bond

Wavenumber (cm-1) Functional groups C-O in C-O in C-O in general R-O-R` -CH2-CH2-OH Precursor B 1125 1052 1016 Starting Material Precursor B 1126 1006 - After heating for 90 s 1090 1046

(b) C=O stretch bond Wavenumber (cm-1) C=O in C=O in Functional group RCOOR` CH3COOH Precursor B 1739 1722 Starting Material Precursor B 1739 1709 After heating for 90 s

A1.3 Gelation Condition

Three precursors with different metal salt concentrations were used in this part to study the effect of heating conditions on the gelation process of these precursors.

These experiments were done to confirm the consistency of gelation temperature as concluded from Chapter 4, that is, the transparent gels form after heating at 70°C for 10 minutes in thin-film. These experiments also enable the effect of metal salt concentration on the gelation process to be better understood. Table A1-4 shows the composition of the precursors we used in this section of work.

Table A1-4. Composition of Precursor BB solutions with various metal salts concentrations

Acetic Bi(NO ) ·5H O Fe(NO ) ·9H O 2-MOE 3 3 2 3 3 2 anhydride 0.50 mol 0.21 mol 0.25 M 0.015 mol 0.015 mol (40 ml) (20 ml) 0.29 mol 0.21 mol 0.35 M 0.015 mol 0.015 mol (23 ml) (20 ml) 0.16 mol 0.21 mol 0.45 M 0.015 mol 0.015 mol (13 ml) (20 ml)

215

A1.3.1 Heating Condition for Gelation

Thin films were prepared on glass substrates using precursor BB of at each of the metal salt concentrations (0.25 M, 0.35 M, and 0.45 M) and heated at various temperatures (RT, 60°, 70°, 80° and 90°C) respectively. The deposition and heating processes are the same as those described in Chapter 4 (see Section 4.2.2).

Figure A2-6 shows the microstructures of deposited thin films obtained by transmission optical microscopy. The appearances and trends of the gel films are similar to those reported for the gel films in Chapter 4. The films either were completely transparent and clear, or were transparent but containing crystals (of metal salts). These results could confirm the validity of the gelation conditions concluded from Chapter 4, specifically, that clear films are obtained for gels heated at >70°C for 10 minutes and that precipitates are form in the films at lower temperatures (60°C or below) in the thin

(and medium thickness) films.

Additionally, this study gives understanding of the role on metal salt concentration on the tendency of metal salts to precipitate in the films. It is apparent from the micrographs shown in Figure A1-6, that clear, homogenous gel films are obtained after heating at 70°C and above for 10 minutes regardless of the metal nitrate concentration. At 60°C or less, metal salts precipitated in the films for each of the precursors (as was found for the 0.35 M precursor with 10% excessive Bi in Chapter 4, see Section 4.3.2). However, qualitatively at least, it appears that the amount of precipitates that form at each time-point tends to increase with increasing metal nitrate concentration. This observation supports the discussion given in Chapter 4 to the effect that high metal concentration precursors will have: (a) higher mass concentrations and therefore higher vapour pressure (as per Raoult's Law), thus giving faster drying rates compared with their lower concentration counterparts; and, (b) a greater amount of precipitates forming during evaporation of solvent because the onset of saturation will be reached sooner. The combined result of effects (a) and (b) is that precursors having

216

higher salt concentrations (eg., 0.45 M here) will tend to undergo a greater extent of precipitation when the external conditions (heating temperature, heating time) are the same (see Section 4.3.2.4).

0.25 M 0.35 M 0.45 M R.T R.T R.T

50°C 50°C 50°C

60°C 60°C 60°C

70°C 70°C 70°C

80°C 80°C 80°C

90°C 90°C 90°C

217

A1.3.2 Effect of Preheating on Gelation

As was discussed in Chapter 4 (see Section 4.3.2.4), preheating the substrates before deposition can effectively prevent the formation of precipitates during spin coating thus improving the homogeneity of gel films and crystallized thin films and their resultant properties. This was re-examined here for Precursor BB.

Using Precursor BB for each of the three metal concentrations (0.25 M, 0.35 M,

0.45 M) thin films were deposited onto STO(001) substrates which were either at room- temperature or preheated at 70°C. The films were immediately spin coated at 3000 rpm for 30 seconds followed by heating at 90°C for gelation then 270°C for drying in air.

The dried films were rapidly heated in oxygen to 450°C for 30 minutes for pyrolysis and subsequently 650°C for 30 minutes for crystallization.

AFM topographic microstructures of selected crystallised BFO thin films are shown in Figure A1-7. The films formed on preheated substrates appeared to be relatively smooth and fine-grained whereas those on the non-preheated substrates appeared to be significantly rougher and coarser grained. For the latter films, there appears to be large particles in the microstructure and these are attributed to the formation of metal salt precipitates during the gelation step. The lowest roughness is obtained for the thin film prepared by 0.25 M precursor with preheating. For the preheated films, the root mean square (RMS) roughness increases with increasing precursor metal nitrate concentration from 1.9 nm (0.25 M) to 4.0 nm (0.35 M) to 4.3

(0.45 M). This result is consistent with the discussion in Section 3.3.2.4 and first part of

Section A1.2.1, in that the drying rate of 0.45 M is higher than that of the 0.25 M precursor due to its higher vapour pressure. In comparison, the thin films without

218

preheating process show a significantly higher RMS roughness of 7.9-12.3 nm which is attributed to the formation of metal salt precipitates formed during gelation.

With preheating process: Without preheating process: 0.25 M 0.25 M

R R

RMS roughness: 1.9 nm RMS roughness: 7.9 nm

R R 0.35 M 0.35 M

R R

RMS roughness: 4.0 nm RMS roughness: 9.2 nm

R R 0.45 M 0.45 M

R R

RMS roughness: 4.3 nm RMS roughness: 12.3 nm

Figure RA1 -7. AFM topographic images of BFO/STO(001)R thin films prepared by stoichiometric precursor B with various metal salts concentrations (0.25 M, 0.35 M, 0.45 M) after crystallized at 650°C in oxygen atmosphere

219

Figure A1-8 shows the XRD of the crystallised films prepared with and without gelation preheating. The preheated substrates result in phase-pure, epitaxial BFO films as indicated by the presence of only BFO peaks of only (00L) orientation. In contrast, the non-preheated substrates exhibited BFO(110) misorientation and a bismuth-rich secondary phase. This is consistent with the gelated films containing metal nitrate precipitates (especially of Bi) in that the precipitates forming during spinning and subsequent gelation would contribute to formation of the bismuth-rich impurities and crystallographic defects (misorientation). Of course, these would significantly impair the ferroelectric properties of the BFO thin films.

Figure A1-8. XRD patterns of BFO thin films prepared with and without preheating process

220

In summary, the results presented in this appendix are similar and completely consistent with the findings in Chapter 4. They confirm the general chemistry and gelation conditions of the precursors and, moreover, confirm the selection of optimal stoichiometric precursor, and optimized gelation and crystallization conditions developed in Chapters 4 and 5).

221

A2. Frequency Dependency of P-E Hysteresis Loop

Figure A2-1(a) shows the P-E loops of 150 nm BFO thin-films as a function of frequency. It is found that the thin films show a good square shape and high polarization

(2Pr =~ 97.48 µC/cm2) even at low frequencies down to 2 kHz.

Figure A2-1. Frequency dependency of (a) ferroelectric polarization hysteresis loop and (b) coercive field

222

It is evident from the linear fitting of the lg(2Ec) vs lg(f) plot (shown in Figure

A2-1(b) that the BFO thin films follows the nucleation dominated switching mechanism according to analysis of Scott[184]. This result agrees with the discussion of differences between macro and micro –scale coercive field in Section 7.3.3.1 to the effect that the coercive voltage of the local (micro-scale) loops (2Vc = 8.5 V) (see Section 7.3.3) is much higher than the macro-scale one (2Vc =2.9-4.8 V). This is due to the high density of domain walls which facilitates the nucleation of new domains during switching. The domain boundaries in the sol-gel derived BFO thin films developed in this work hould thus play a vital role in reducing the macro-scale coercive field.

223

References

[1] Arimoto, Y., et al., Current status of ferroelectric randomm-access memory. Mrs Bulletin, 2004. 29, 823

[2] Dawber, M., et al., Physics of thin-film ferroelectric oxides. Reviews of Modern Physics, 2005. 77, 1083

[3] Muralt, P., PZT thin films for microsensors and actuators: Where do we stand? Ieee Transactions on Ultrasonics Ferroelectrics and Frequency Control, 2000. 47, 903

[4] Moazzami, R., et al., Electrical characteristics of ferroelectric PZT thin-films for dram applications. Ieee Transactions on Electron Devices, 1992. 39, 2044

[5] Wang, J., et al., Epitaxial BiFeO3 Multiferroic Thin Film Heterostructures. Science, 2003. 299,1719

[6] Jiang, A.Q., et al., A resistive memory in demiconducting BiFeO3 thin-film Capacitors. Advanced Materials, 2011. 23, 1277

[7] Valant, M., et al., Peculiarities of a solid-state synthesis of multiferroic polycrystalline BiFeO3. Chemistry of Materials, 2007. 19, 5431

[8] Lakovlev, S., et al., Multiferroic BiFeO3 thin films processed via chemical solution deposition: Structural and electrical characterization. Journal of Applied Physics, 2005. 97, 094901

[9] You, L., et al., Influence of oxygen pressure on the ferroelectric properties of epitaxial BiFeO3 thin films by pulsed laser deposition. Physical Review B, 2009. 80, 024105

[10] Himcinschi, C., et al., Substrate influence on the optical and structural properties of pulsed laser deposited BiFeO3 epitaxial films. Journal of Applied Physics, 2010. 107, 123524

[11] Das, R.R., et al., Synthesis and ferroelectric properties of epitaxial BiFeO3 thin films grown by sputtering. Applied Physics Letters, 2006. 88, 242904

[12] Jang, H.W., et al., Domain engineering for enhanced ferroelectric properties of epitaxial (001) BiFeO thin films. Advanced Materials, 2009. 21, 817

[13] Zhang, J.X., et al., Microscopic origin of the giant ferroelectric polarization in tetragonal-like BiFeO3. Physical Review Letters, 2011. 107, 147602 [14] Wang, C., et al., Switchable diode effect and ferroelectric resistive switching in epitaxial BiFeO3 thin films. Applied Physics Letters, 2011. 98, 192901 [15] Rana, A., et al., Scaling behavior of resistive switching in epitaxial bismuth ferrite heterostructures. Advanced Functional Materials, 2014. 24, 3962

224

[16] Brinker, C.J. and G.W. Scherer, Sol-gel science: the physics and chemistry of sol-gel processing. 1990: Academic Press

[17] Tyholdt, F., et al., Synthesis of oriented BiFeO3 thin films by chemical solution deposition: Phase, texture, and microstructural development. Journal of Materials Research, 2005. 20, 2127

[18] Jaffe, B., et al., Piezoelectric ceramics. 1971: Academic Press London and New York.

[19] Damjanovic, D., Ferroelectric, dielectric and piezoelectric properties of ferroelectric thin films and ceramics. Reports on Progress in Physics, 1998. 61, 1267

[20] Fukada, E., et al., On the piezoelectric effect of bone. Journal of the Physical Society of Japan, 1957. 12, 1158

[21] Damjanovic, D., Stress and frequency dependence of the direct piezoelectric effect in ferroelectric ceramics. Journal of Applied Physics, 1997. 82, 1788

[22] Sirohi, J., et al., Fundamental understanding of piezoelectric strain sensors. Journal of Intelligent Material Systems and Structures, 2000. 11, 246

[23] Van, R. J., Piezoelectric ceramics. 1974.

[24] Omote, K., et al., Temperature dependence of elastic, dielectric, and piezoelectric properties of ''single crystalline'' films of vinylidene fluoride trifluoroethylene copolymer. Journal of Applied Physics, 1997. 81, 2760

[25] Cohen, R.E., Origin of ferroelectricity in perovskite oxides. Nature, 1992. 358, 136

[26] Ok, K.M., et al., Bulk characterization methods for non-centrosymmetric materials: second-harmonic generation, piezoelectricity, pyroelectricity, and ferroelectricity. Chemical Society Reviews, 2006. 35, 710

[27] Valasek, J., Piezo-electric and allied phenomena in Rochelle salt. Physical Review, 1921. 17, 475

[28] Takenaka, T., et al., Current status and prospects of lead-free piezoelectric ceramics. Journal of the European Ceramic Society, 2005, 25, 2693

[29] Tagantsev, A., et al., Domains in ferroic crystals and thin films 2010: Springer Science and Business Media.

[30] Catalan, G., et al., Domain wall nanoelectronics. Reviews of Modern Physics, 2012. 84, 119

[31] Merz, W.J., Domain formation and domain wall motions in ferroelectric BaTiO3 single crystals. Physical Review, 1954. 95, 690

[32] Mehta, R.R., et al., Depolarization fields in thin ferroelectric films. Journal of Applied Physics, 1973. 44, 3379

225

[33] Nagarajan, V., et al., Dynamics of ferroelastic domains in ferroelectric thin films. Nature Materials, 2003. 2, 43

[34] Catalan, G. and J.F. Scott, Physics and applications of bismuth ferrite. Advanced Materials, 2009. 21, 2463

[35] Lubk, A., et al., First-principles study of ferroelectric domain walls in multiferroic bismuth ferrite. Physical Review B, 2009. 80, 104110

[36] Seidel, J., et al., Conduction at domain walls in oxide multiferroics. Nature Materials, 2009. 8, 229

[37] Levinson, L.M., ed. Electronic ceramics: properties, devices, and applications. 1987, CRC press: New York.

[38] Lallart, M., ed. Ferroelectrics - physical effects. 2011, InTech.

[39] Bondurant, D. and F. Gnadinger, Ferroelectric for nonvolatile rams. Ieee Spectrum, 1989. 26, 30

[40] Berlincourt, D., Piezoelectric ceramic compositional development. Journal of the Acoustical Society of America, 1992. 91, 3034

[41] Izyumskaya, N., et al., Processing, structure, properties, and applications of PZT thin films. Critical Reviews in Solid State and Materials Sciences, 2007. 32, 111

[42] Ohji, T., et al., Sensing using electrostatic capacitance. Key Engineering Materials, 2006. 317-318, 865

[43] Yan, T., et al., Development of metallic digital strain gauges. Applied Mechanics and Materials, 2004, 179

[44] Li, D., et al., Magnitude versus frequency performance of vibration acceleration sensor based on cymbal transducer. Key Engineering Material, 2008. 368-372, 226

[45] Marat-Mendes, R., et al., A comparative study of piezoelectric materials using smart angular accelerometers. Key Engineering materials, 2002. 230-232, 181

[46] Koyama, D., et al., Array configurations for higher power generation in piezoelectric energy harvesting,. Japanese Journal of Applied Physics, 2010. 49, 07HD04

[47] Platt, S.R., et al., On low-frequency electric power generation with PZT ceramics. Ieee-Asme Transactions on Mechatronics, 2005. 10, 240

[48] Barrow, D.A., et al., Characterization of thick lead zirconate titanate films fabricated using a new sol gel based process. Journal of Applied Physics, 1997. 81, 876

[49] Haertling, G.H., Ferroelectric ceramics: History and technology. Journal of the American Ceramic Society, 1999. 82, 797

226

[50] Tressler, J.F., et al., Piezoelectric sensors and sensor materials. Journal of Electroceramics, 1998. 2, 257

[51] Fernandes, J. R. A., et al., Optimization of the fabrication parameters of PZT 52/48 thin films by pulsed laser ablation. Materials Science Forum, 2006. 514- 516, 1353

[52] Nagarajan, V., et al., Size effects in ultrathin epitaxial ferroelectric heterostructures. Applied Physics Letters, 2004. 84, 5225

[53] Chen, L., et al., Nonlinear electric field dependence of piezoresponse in epitaxial ferroelectric lead zirconate titanate thin films. Journal of Applied Physics, 2003. 94, 5147

[54] Randall, C.A., et al., Intrinsic and extrinsic size effects in fine-grained morphotropic-phase-boundary lead zirconate titanate ceramics. Journal of the American Ceramic Society, 1998. 81, 677

[55] Piezoelectric Material Specifications Table. 2012; Available from: http:// www.piezotechnologies.com/Documents/120710-Material-Data-Sheet.aspx.

[56] Directive on the restriction of use of certain hazardous substances in electrical and electronic equipment 2002/95/EC.

[57] Takahashi, H., et al., Lead-free barium titanate ceramics with large piezoelectric constant fabricated by microwave sintering. Japanese Journal of Applied Physics Part 2-Letters & Express Letters, 2006. 45, L30

[58] Choi, K.J., et al., Enhancement of ferroelectricity in strained BaTiO3 thin films. Science, 2004. 306, 1005

[59] Rodel, J., et al., Perspective on the development of lead-free piezocermics. Journal of american ceramic society, 2009. 92, 1153

[60] Li, H., et al., Microstructure, crystalline phase and electrical properties of Li0.06(Na0.5K0.5)0.94Nb(1-2x/5)MgxO3 lead-free piezoelectric ceramics. International Journal of Minerals, Metallurgy and Materials. 17, 340

[61] Smolenskii, G.A., et al., New ferroelectrics of complex composition. 4. Soviet Physics-Solid State, 1961. 2, 2651

[62] Michel, C., et al., The atomic structure of BiFeO3. Solid state Commun, 1969. 7, 701

[63] Wu, J., et al., Improved ferroelectric behavior in (110) oriented BiFeO3 thin films. Journal of Applied Physics, 2010. 107, 034103

[64] Choi, T., et al., Switchable ferroelectric diode and photovoltaic effect in BiFeO3. Science, 2009. 324, 63

[65] Neaton, J.B., et al., First-principles study of spontaneous polarization in multiferroic BiFeO3. Physical Review B, 2005. 71, 014113

227

[66] Narayan, J., et al., Domain epitaxy: A unified paradigm for thin film growth. Journal of Applied Physics, 2003. 93, 278

[67] Zeches, R.J., et al., A strain-driven morphotropic phase boundary in BiFeO3. Science, 2009. 326, 977

[68] Cheng, C.J., et al., Thickness-dependent magnetism and spin-glass behaviors in compressively strained BiFeO3 thin films. Applied Physics Letters, 2011. 98, 242502

[69] Chu, Y.H., et al., Ferroelectric size effects in multiferroic BiFeO3 thin films. Applied Physics Letters, 2007. 90, 252906

[70] Chen, Z.H., et al., Low symmetry monoclinic M-C phase in epitaxial BiFeO3 thin films on LaSrAlO4 substrates. Applied Physics Letters, 2010. 97, 242903

[71] Chen, Z.H., et al., Nanoscale domains in strained epitaxial BiFeO3 thin Films on LaSrAlO4 substrate. Applied Physics Letters, 2010. 96, 252903

[72] Morozov, M.I., N.A. Lomanova, and V.V. Gusarov, Specific features of BiFeO3 formation in a mixture of bismuth(III) and iron(III) oxides. Russian Journal of General Chemistry, 2003. 73, 1676

[73] Carvalho, T.T. and P.B. Tavares, Synthesis and thermodynamic stability of multiferroic BiFeO3. Materials Letters, 2008. 62, 3984 [74] Bernardo, M.S., et al., Reaction pathways in the solid state synthesis of multiferroic BiFeO3. Journal of the European Ceramic Society, 2011. 31, 3047

[75] Xu, J., et al., Preparation of BiFeO3 Nanopowders using acetylacetone as stabilizer. Key Engineering Materials, 2010. 434-435, 314

[76] Lahmar, A., et al., Off-stoichiometry effects on BiFeO3 thin films. Solid State Ionics, 2011. 202, 1

[77] Bea, H., et al., Influence of parasitic phases on the properties of BiFeO3 epitaxial thin films. Applied Physics Letters, 2005. 87, 072508

[78] Hardy, A., et al., Effects of precursor chemistry and thermal treatment conditions on obtaining phase pure bismuth ferrite from aqueous gel precursors. Journal of the European Ceramic Society, 2009. 29, 3007

[79] Qi, X.D., et al., Greatly reduced leakage current and conduction mechanism in aliovalent-ion-doped BiFeO3. Applied Physics Letters, 2005. 86, 062903 [80] Wang, Y.P., et al., Room-temperature saturated ferroelectric polarization in BiFeO3 ceramics synthesized by rapid liquid phase sintering. Applied Physics Letters, 2004. 84, 1731

[81] Wu, J. and J. Wang, Orientation dependence of ferroelectric behavior of BiFeO3 thin films. Journal of Applied Physics, 2009. 106, 104111

228

[82] Frost, H.J., Microstructural evolution in thin-films. Materials Characterization, 1994. 32, 257

[83] Thompson, C.V. and R. Carel, Stress and grain growth in thin films. Journal of the Mechanics and Physics of Solids, 1996. 44, 657

[84] Yun, K.Y., et al., Structural and multiferroic properties of BiFeO3 thin films at room temperature. Journal of Applied Physics, 2004. 96, 3399

[85] Catalan, G., et al., Fractal dimension and size scaling of domains in thin films of multiferroic BiFeO3. Physical Review Letters, 2008. 100, 027602 [86] Yun, K.Y., et al., Giant ferroelectric polarization beyond 150 mu C/cm2 in BiFeO3 thin film. Japanese Journal of Applied Physics Part 2-Letters & Express Letters, 2004. 43, L647

[87] Tang, X., et al., Self-limited grain growth, dielectric, leakage and ferroelectric properties of nanocrystalline BiFeO3 thin films by chemical solution deposition. Acta Materialia, 2013. 61, 1739

[88] Nakamura, Y., et al., Improvement of ferroelectric properties of BiFeO3 thin films by postmetallization annealing and electric field application. Journal of Applied Physics, 2009. 105, 061616

[89] Singh, S.K., et al., Enhanced polarization and reduced leakage current in BiFeO3 thin films fabricated by chemical solution deposition. Journal of Applied Physics, 2006. 100, 064102

[90] Singh, S.K., et al., Epitaxial BiFeO3 thin films fabricated by chemical solution deposition. Applied Physics Letters, 2006. 88, 162904

[91] Tsurumaki, A., H. Yamada, and A. Sawa, Impact of Bi deficiencies on ferroelectric resistive switching characteristics observed at p-type schottky-like Pt/Bi1-dFeO3 interfaces. Advanced Functional Materials, 2012. 22, 1040 [92] Blom, P.W.M., et al., Ferroelectric schottky diode. Physical Review Letters, 1994. 73, 2107

[93] Pintilie, L., et al., Ferroelectric Schottky diode behavior from a SrRuO3- Pb(Zr0.2Ti0.8)O3-Ta structure. Physical Review B, 2010. 82, 085319

[94] Schmehl, A., et al., Transport properties of LaTiO3+x films and heterostructures. Applied Physics Letters, 2003. 82, 3077

[95] Chrisey, D. B., et al., Pulsed laser deposition of thin films. 1994: John Wiley & Sons.

[96] Bea, H., et al., Tunnel magnetoresistance and robust room temperature exchange bias with multiferroic BiFeO3 epitaxial thin films. Applied Physics Letters, 2006. 89, 242114

[97] Maksymovych, P., et al., Ultrathin limit and dead-layer effects in local polarization switching of BiFeO3. Physical Review B, 2012. 85, 014119

229

[98] Meier, M., et al., Sol-gel processing under microgravity conditions. Materials Science Forum, 1991. 77, 99

[99] Tulloch, G. E., et al., Commercial development of sol-gel technology in Australia. Key Engineering Materials, 1998. 150, 185

[100] Yu, C. C., et al., Electrical properties and leakage current mechanisms of Bi3.2Gd0.8Ti3O12 thin films prepared by a sol-gel method. Journal of Supercond Nov Magn, 2010. 23, 929

[101] Rahaman, M.N., Ceramic processing and sintering 2003: CRC Press.

[102] Ferry, J.D., Viscoelastic properties of polymers. 1980, New York: Wiley.

[103] Lai, F., et al., Sol-gel processing of lead-free (Na, K)NbO3 ferroelectric films. Journal of Sol-gel Technology, 2007. 42, 287

[104] Zhou, Z., et al., Effect of Sol viscosity on properties of BST thin films prepared by sol-gel method. Advanced Materials Research, 2010. 105-106, 676

[105] Chen, Y. Q., et al., Fabrication of Lead-free (Na0.82K0.18)0.5Bi0.5TiO3 piezoelectric nanofiber by electrospinning. Materials Research Bulletin, 2010. 45, 717

[106] Zhang, Y., et al., The study of BiCrxFe1-xO3 thin films synthesized by sol-gel technique. Journal of the European Ceramic Society, 2010. 30, 271

[107] Lai, F., et al., Preparation of ferroelectric K0.5Na0.5NbO3 films by sol-gel processing. Key Engineering Materials, 2007. 336-338, 203

[108] Tanaka, K., et al., Effect of gel-films-thickness and sintering conditions on the crystal structure and microstructure of Alkoxy-derived BaTiO3 thin films. Key Engineering Materials, 2009. 388, 171

[109] Chowdhury, A., et al., Fundamental issues in the synthesis of ferroelectric Na0.5K0.5NbO3 thin films by sol-gel processing. Chemical materials, 2010. 22, 3862

[110] Carl V, T., et al., Stress and grain growth in thin film. Journal of Mech.Phys. Solids, 1996. 44, 657

[111] Frost, H.J., Microstructural evolution in thin films. materials characterization, 1994. 32, 257

[112] Gurkovich, S.R.., J.B.B., ed. Preparation of monolithic lead-titanate by a sol-gel process. Ultrastructure processing of ceramics, Glasses and composites, ed. D.R.U. L.L.Hench. 1984, Wiley-Interscience: New York. 152.

[113] Budd, K.D., et al., Sol-gel processing of PbTiO3, PbZrO3, PZT, and PLZT thin films. British Ceramic Proceedings, 1985. 36, 21

[114] Schwartz, R.W., Chemical solution deposition of perovskite thin films. Chemistry of Materials, 1997. 9, 2325

230

[115] Ederer, C. and N.A. Spaldin, Effect of epitaxial strain on the spontaneous polarization of thin film ferroelectrics. Physical Review Letters, 2005. 95, 257601

[116] Schlom, D.G., et al., Strain tuning of ferroelectric thin films, in Annual Review of Materials Research. 2007.589

[117] Chen, Z., et al., Low-symmetry monoclinic phases and polarization rotation path mediated by epitaxial strain in multiferroic BiFeO3 thin films. Advanced Functional Materials, 2011. 21, 133

[118] Luo, Z., et al., Periodic elastic nanodomains in ultrathin tetragonal-like BiFeO3 films. Physical Review B, 2013. 88, 064103

[119] Infante, I.C., et al, Bridging multiferroic phase transitions by epitaxial strain in BiFeO3. Physical Review Letters, 2010. 105, 057601

[120] Zheng, R.Y., et al., Multiferroic BiFeO3 thin films deposited on SrRuO3 buffer layer by rf sputtering. Journal of Applied Physics, 2007. 101, 054104

[121] Zheng, R., et al., Multiferroic BiFeO3 thin films buffered by a SrRuO3 layer. Journal of the American Ceramic Society, 2008. 91, 463

[122] Nelson, C.T., et al., Domain dynamics during ferroelectric switching. Science, 2011. 334, 968

[123] Wang, D.H., et al., BiFeO3 film deposited on Si substrate buffered with La0.7Sr0.3MnO3 electrode. Applied Physics Letters, 2006. 89, 182905 [124] Tsui, F., et al., Strain-dependent magnetic phase diagram of epitaxial La0.67Sr0.33MnO3 thin films. Applied Physics Letters, 2000. 76, 2421 [125] Lu, Y., et al., Large magnetotunneling effect at low magnetic fields in micrometer-scale epitaxial La0.67Sr0.33MnO3 tunnel junctions. Physical Review B, 1996. 54, R8357

[126] Natsume, Y. and H. Sakata, Zinc oxide films prepared by sol-gel spin-coating. Thin Solid Films, 2000. 372, 30

[127] Singh, S.K., et al., Dependence of ferroelectric properties on thickness of BiFeO3 thin films fabricated by chemical solution deposition. Japanese Journal of Applied Physics Part 1-Regular Papers Brief Communications & Review Papers, 2005. 44, 8525

[128] Nakamura, Y., S. Nakashima, and M. Okuyama, Influences of surface texture and Bi/Fe ratio on electric properties of BiFeO3 thin films prepared by chemical solution deposition. Japanese Journal of Applied Physics, 2008. 47, 7250

[129] Wang, X.Z., et al., Enhanced ferroelectric properties of Ce-substituted BiFeO3 thin films on LaNiO3/Si substrates prepared by sol-gel process. Journal of the European Ceramic Society, 2009. 29, 1183

231

[130] Fruth, V., et al., Chemical solution deposition and characterization of BiFeO3 thin films. Journal of the European Ceramic Society, 2007. 27, 4417

[131] Nakamura, T., et al., Micro-Raman Study of BiFeO3 Thin Films Fabricated by Chemical Solution Deposition Using Different Bi/Fe Ratio Precursors. Acta Physica Polonica A, 2009. 116, 72

[132] Singh, S.K., et al., Bottom electrodes dependence of ferroelectric properties in epitaxial BiFeO3/SrRuO3/SrTiO3 structures. Integrated Ferroelectrics, 2007. 87, 42

[133] Nakamura, Y., et al., The insertion effect of Bi-excess layers on stoichiometric BiFeO3 thin films prepared by chemical solution deposition. Functional Materials Letters, 2008. 1, 19

[134] Pan, H.C., et al., Low-temperature processing of sol-gel derived La0.5Sr0.5MnO3 buffer electrode and PbZr0.52Ti0.48O3 films using CO2 laser annealing. Applied Physics Letters, 2003. 83, 3156

[135] Izuha, M., et al., Electrical properties and microstructures of Pt/Ba0.5Sr0.5TiO3/SrRuO3 capacitors. Applied Physics Letters, 1997. 70, 1405 [136] Gates, B.D., et al., New approaches to nanofabrication: Molding, printing, and other techniques. Chemical Reviews, 2005. 105, 1171

[137] Wallraff, G.M. and W.D. Hinsberg, Lithographic imaging techniques for the formation of nanoscopic features. Chemical Reviews, 1999. 99, 1801

[138] Hall, L.D., Nuclear magnetic resonance. Advances in Carbohydrate Chemistry, 1964. 19, 51

[139] Bruch, M.D., ed. NMR spectroscopy techniques, Second Editions, Revised and Expanded. 1996, Marcel Dekker: New York.

[140] Keeler, J. NMR and energy levels. 2002; Available from: http://www- keeler.ch.cam.ac.uk/lectures/Irvine/chapter2.pdf.

[141] Wishart, D.S., B.D. Sykes, and F.M. Richards, Relationship between neclear- magnetic-resonance chemical-shift and protein scondary structure. Journal of Molecular Biology, 1991. 222, 311

[142] Levin, I.W. and R. Bhargava, Fourier transform infrared vibrational spectroscopic imaging: Integrating microscopy and molecular recognition, in Annual Review of Physical Chemistry. 2005, Annual Reviews: Palo Alto

[143] Peter R. Griffiths, J.A.d.H., Fourier Transform infrared spectrometry. Chemical Analysis: A series of monographs on Analytical Chemistry and its application, ed. J.D.Winefordner. 2007, New jersey: John Wiley & Sons.

[144] Toyoda, M., et al., Evolution of structure of the precursor during sol-gel processing of ZnO and low temperature formation of thin films. Journal of Sol- Gel Science and Technology, 1999. 16, 93

232

[145] Fe, L., et al., Absorption-reflection infrared spectroscopy studies of sol-gel prepared ferroelectric Pb(Zr,Ti)O3 thin films on Pt electrodes. Journal of Sol- Gel Science and Technology, 2000. 19, 149

[146] Socrates, G., ed. Infrared and Raman characteristic group frequencies: Tables and charts. 2004, Wiley.

[147] Cappella, B. and G. Dietler, Force-distance curves by atomic force microscopy. Surface Science Reports, 1999. 34, 1

[148] Geisse, N. AFM and combined optical techniques. Available from: https://www.asylumresearch.com/Applications/CombinedAFMOptical/Combine dAFMOpticalHR.pdf.

[149] M.Alexe, A.G., ed. Nanoscale characterisation of ferroelectric materials: Scanning probe microscopy approach. Nanoscience and technology, ed. K.V.K. P.Avouris, H.Sakaki, R.Wiesendanger. 2004, Springer: Heidelberg.

[150] Proksch, R. Piezoresponse force microscopy with Asylum Research AFMs. Available from: http:// www. asylumresearch.com/ Applications /PFMAppNote /PFM-ANHR.pdf.

[151] Rodriguez, B.J., et al., Dual-frequency resonance-tracking atomic force microscopy. Nanotechnology, 2007. 18, 475504

[152] Jesse, S., A.P. Baddorf, and S.V. Kalinin, Switching spectroscopy piezoresponse force microscopy of ferroelectric materials. Applied Physics Letters, 2006. 88, 062908

[153] AsylumResearch, Asylum research cypher scanning probe microscope, user guide, version 1025. 2012.

[154] Joe T. Evans, J. The relationship between hysteresis and PUND responses. 2008; Available from: http:// www.ferrodevices.com /1/297/files/ Hysteresis- EqualsPUND.pdf.

[155] Ng, M.F. and M.J. Cima, Heteroepitaxial growth of lanthanum aluminate films derived from mixed metal nitrates. Journal of Materials Research, 1997. 12, 1306

[156] Fan, Q., et al., High density, non-porous anatase titania thin films for device applications. Journal of Physics D-Applied Physics, 2000. 33, 2683

[157] Liu, H.R., et al., Ferroelectric properties of BiFeO3 films grown by sol-gel process. Thin Solid Films, 2006. 500, 105

[158] Laura Fe, G.N., et al., Absorption-reflection infrared sepctroscopy studies of sol- gel prepared ferroelectric Pb(Zr,Ti)O3 thin films on Pt electrodes. Journal of Sol-Gel Science and Technology, 2000. 19, 149

[159] Joe T. Evans, J., Characterizing ferroelectric materials. 2011.

233

[160] Bothwell, J.M., et al., Applications of bismuth(III) compounds in organic synthesis. Chemical Society Reviews, 2011. 40, 4649

[161] Tyholdt, F., et al., Synthesis of oriented BiFeO3 thin films by chemical solution deposition: Phase, testure, and microstructrual development. Journal of Materials Research, 2005. 20, 2127

[162] Selbach, S.M., et al., On the thermodynamic stability of BiFeO3. Chemistry of Materials, 2009. 21, 169

[163] Kubel, F. and H. Schmid, Structure of a ferroelectric and ferroelastic monodomain crystal of the perovskite BiFeO3. Acta Crystallographica Section B-Structural Science, 1990. 46, 698

[164] Damodaran, A.R., et al., Temperature and thickness evolution and epitaxial breakdown in highly strained BiFeO3 thin films. Physical Review B, 2012. 85, 024113

[165] Ma, H., et al., Strain effects and thickness dependence of ferroelectric properties in epitaxial BiFeO3 thin films. Applied Physics Letters, 2008. 92, 182902

[166] Wang, Y., et al., Multiferroic BiFeO3 thin films prepared via a simple sol-gel method. Applied Physics Letters, 2006. 88, 142503

[167] Schwartz, R.W., Chemical solution deposition of perovskite thin films. Chemistry of Materials, 1997. 9, 2325

[168] Ke, H., et al., Factors controlling pure-phase multiferroic BiFeO3 powders synthesized by chemical co-precipitation. Journal of Alloys and Compounds, 2011. 509, 2192

[169] Nagarajan, V., et al., Nanoscale polarization relaxation in a polycrystalline ferroelectric thin film: Role of local environments. Applied Physics Letters, 2005. 86, 262910

[170] Watanabe, T., et al., Probing intrinsic polarization properties in bismuth-layered ferroelectric films. Applied Physics Letters, 2007. 90, 112914

[171] Shvartsman, V.V., et al., Nonlinear local piezoelectric deformation in ferroelectric thin films studied by scanning force microscopy. Journal of Applied Physics, 2005. 97, 104105

[172] Alexe, M., et al., Polarization imprint and size effects in mesoscopic ferroelectric structures. Applied Physics Letters, 2001. 79, 242

[173] Wu, J.G., et al., Ferroelectric behavior in bismuth ferrite thin films of different thickness. Acs Applied Materials & Interfaces, 2011. 3, 3261

[174] Kim, D.H., et al., Effect of epitaxial strain on ferroelectric polarization in multiferroic BiFeO3 films. Applied Physics Letters, 2008. 92, 012911

234

[175] Shvartsman, V.V. and A.L. Kholkin, Evolution of nanodomains in 0.9PbMg1/3Nb2/3O3-0.1PbTiO3 single crystals. Journal of Applied Physics, 2007. 101, 064108

[176] Basu, S.R., et al., Photoconductivity in BiFeO3 thin films. Applied Physics Letters, 2008. 92, 091905

[177] Wu, J. and J. Wang, BiFeO3 thin films of (111)-orientation deposited on SrRuO3 buffered Pt/TiO2/SiO2/Si(100) substrates. Acta Materialia, 2010. 58, 1688 [178] Natori, K., et al., Thickness dependence of the effective dielectric constant in a thin film capacitor. Applied Physics Letters, 1999. 73, 632

[179] Pertsev, N.A., et al., Thickness dependence of intrinsic dielectric response and apparent interfacial capacitance in ferroelectric thin films. Journal of Applied Physics, 2007. 101, 074102

[180] Yamamichi, S., et al., (Ba+Sr)/Ti ratio dependence of the dielectric-properties for (Ba0.5Sr0.5)TiO3 thin-films prepared by ion-beam sputtering. Applied Physics Letters, 1994. 64, 1644

[181] Waser, R., et al., Redox-based resistive switching memories - Nanoionic Mechanisms, Prospects, and Challenges. Advanced Materials, 2009. 21, 2632

[182] Yuan, G.-L. and J. Wang, Evidences for the depletion region induced by the polarization of ferroelectric semiconductors. Applied Physics Letters, 2009. 95, 252904

[183] Lee, Y.H., J.M. Wu, and C.H. Lai, Influence of La doping in multiferroic properties of BiFeO3 thin films. Applied Physics Letters, 2006. 88, 042903 [184] Scott, J.F., Models for the frequency dependence of coercive field and the size dependence of remanent polarization in ferroelectric thin films. Integrated Ferroelectrics, 1996. 12, 71

235