Diglyme as an electrolyte solvent for sodium-ion batteries Deciphering degradation mechanisms and redox behaviour Master’s thesis in Applied Physics KASPER WESTMAN Department of Physics CHALMERS UNIVERSITY OF TECHNOLOGY Gothenburg, Sweden 2016 Master’s thesis 2016 Diglyme as an electrolyte solvent for sodium-ion batteries Deciphering degradation mechanisms and redox behaviour KASPER WESTMAN Department of Physics Division of Condensed Matter Physics Chalmers University of Technology Gothenburg, Sweden 2016 Diglyme as an electrolyte solvent for sodium-ion batteries Deciphering degradation mechanisms and redox behaviour KASPER WESTMAN © KASPER WESTMAN, 2016 Supervisor: Romain Dugas, Chimie du solide et énergie, Collège de France, Paris Examiner: Patrik Johansson, Department of Physics, Chalmers University of Technology, Gothenburg Department of Physics Division of Condensed Matter Physics Chalmers University of Technology SE-412 96 Gothenburg Telephone +46 31 772 1000 Typeset in LATEX Gothenburg, Sweden 2016 iv Diglyme as electrolyte solvent for sodium-ion batteries Deciphering degradation mechanisms and redox behaviour Kasper Westman Department of Physics Chalmers University of Technology Abstract For storing energy in future sustainable energy systems, sodium-ion batteries (SIB) have emerged as an alternative to the current state-of-the-art lithium-ion batteries (LIBs), since SIBs are potentially cheaper. For LIBs and SIBs alike, the development of stable systems depends on employing stable or metastable electrolytes, forming a SEI. Ether compounds have earlier been investigated for use in electrolytes, due to a presumed high reductive stability. In this project diglyme, an ether solvent, mixed with 1 M NaPF6 has been evaluated for use with Na-metal reference electrodes and a SIB electrode platform com- prising Na3V2(PO4)2F3 (NVPF) and hard carbon (HC), and with a model system using Na3V2(PO4)3 (NVP) at both electrodes. Furthermore, this electrolyte has been investi- gated for electrochemical stability, and for possible degradation mechanisms. Techniques used are galvanostatic cycling (GCPL), cyclic voltammetry (CV) and electrical impedance spectroscopy (EIS), coupling these to ab-initio calculations and physico-chemical charac- terisation methods. Using these techniques, Na-metal is shown to provide a stable reference electrode. In spite of the electrochemical stability of 1 M NaPF6 in diglyme appearing promising when conducting CV, cells of NVPF|HC are shown to exhibit a ηc of 99.13 %, a high initial irreversibility (30 %), and an inferior capacity retention, when compared to equivalent cells using a EC50:DMC50-based electrolyte. NVP|NVP cells, on the other hand, show outstanding capacity retention and low initial irreversibility. The problems experi- enced in NVPF|HC cells are proposed to arise due to a set of different mechanisms: binder degradation, vanadium dissolution-deposition, and possibly reduction of the electrolyte. No coherent indications are given for the formation of an SEI. However, reduction schemes drawn still allow for undetectable reduction products. By independently exploring differ- ent voltage regions of the active materials the oxidative stability of 1 M NaPF6 in diglyme is put to question despite earlier indications of stability. Future studies should aim to change the type of binder to avoid losses. NVP should be tested separately with both the active materials NVPF/HC to determine where most of the loss occurs along with further investigations to confirm if there is a reduction of the electrolyte at sodiated HC, since any lack thereof could point towards a future SEI-free system. Keywords: Sodium-ion Batteries, Electrolytes, Diglyme, Electrolyte Degradation, Galvanostatic Cycling, Electrochemistry v Acknowledgements Firstly: my heartfelt thanks to Prof. Jean-Marie Tarascon and the group of Chimie du solide et énergie at Collége de France, both for accepting me in the lab, and for providing a warmhearted social context away from home. Relating to this, I am especially grateful for the invaluable support of Romain Dugas for being my supervisor and for teaching me the ways of battery research. Without his patience and pedagogy, I would not have been able to conduct this project. Yet another person deserving of special mention is my examiner Patrik Johansson, who took time to arrange this scientific exchange and who has given me continuous feedback on how to improve my work. Furthermore, I would like to acknowledge Rosa Palacín, Alexandre Ponrouch and Enrique Jiménez for valuable discussions and for taking care of me in Barcelona. I direct a special thanks to Enrique for helping me conduct FTIR-specatroscopy and to Alexander for helping me with the ionic conductivity-measurements. I gratefully wish to acknowledge Grégory Gacuhot for helping me conduct GS/MS analysis and for teaching me about the basics of electrolyte reduction. I also wish to acknowledge Piotr Jankowski for providing me with DFT-calacualtion data regarding solvent reduction. Finally, I am thankful for the economical support received from Chalmers Master-Card Foundation, helping me cover my living-costs while in Paris. Kasper Westman, Gothenburg, June 2016 vi Contents List of Figures xiii List of Tables xiv List of Abbreviations xv 1 Introduction 1 1.1 Importance of Stable SIBs for a Future Energy System . . . . . . . . . . . . 2 1.2 Aim . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 1.3 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 2 Batteries and SIBs 5 2.1 Batteries–Primary and Secondary Galvanic Cells . . . . . . . . . . . . . . . 6 2.2 Electrolyte Fundamentals - Salts and Solvents . . . . . . . . . . . . . . . . . 9 2.3 Electrode-Electrolyte Interfaces and Interphases . . . . . . . . . . . . . . . . 10 2.4 Sodium-ion Batteries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 2.4.1 SIB Anode Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 2.4.2 SIB Cathode Materials . . . . . . . . . . . . . . . . . . . . . . . . . . 12 2.4.3 SIB Binders and Current Collectors . . . . . . . . . . . . . . . . . . 14 2.4.4 SIB Electrolytes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 2.5 The Glyme Family of Solvents . . . . . . . . . . . . . . . . . . . . . . . . . . 16 2.5.1 Metal Complexing Properties of Glymes . . . . . . . . . . . . . . . . 17 2.5.2 Electrochemistry of Glymes . . . . . . . . . . . . . . . . . . . . . . . 17 2.5.3 Toxicity and Cost of Glymes . . . . . . . . . . . . . . . . . . . . . . 18 3 Description of Experiments 19 3.1 Theory of Main Methods Employed . . . . . . . . . . . . . . . . . . . . . . . 20 3.1.1 Galvanostatic Cycling with Potential Limitation-GCPL . . . . . . . 20 3.1.2 Cyclic Voltammetry–CV . . . . . . . . . . . . . . . . . . . . . . . . . 21 3.1.3 Electrical Impedance spectroscopy–EIS . . . . . . . . . . . . . . . . 21 3.1.4 Gas Chromatography, Mass spectrometry–GC/MS . . . . . . . . . . 21 3.1.5 FTIR Spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 3.2 Experimental Techniques–General Considerations . . . . . . . . . . . . . . . 22 3.2.1 Choice of Electrode Materials . . . . . . . . . . . . . . . . . . . . . . 22 3.2.2 Tape-Cast Electrodes . . . . . . . . . . . . . . . . . . . . . . . . . . 23 3.2.3 Powdered Electrodes . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 3.2.4 Electrolyte Preparation . . . . . . . . . . . . . . . . . . . . . . . . . 24 3.2.5 Cell Assembly . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 3.2.6 Coin Cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 vii Contents 3.2.7 Swagelok Cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 3.3 Electrolyte Stability with Na-metal . . . . . . . . . . . . . . . . . . . . . . . 26 3.3.1 Investigating Film Growth on Na Surfaces . . . . . . . . . . . . . . . 26 3.3.2 Investigating Gas Release from Electrolyte Decomposition . . . . . . 26 3.3.3 Storage Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 3.4 Ionic Conductivity and Viscosity . . . . . . . . . . . . . . . . . . . . . . . . 26 3.5 Electrolyte Stability Window . . . . . . . . . . . . . . . . . . . . . . . . . . 27 3.6 Galvanostatic Cycling Experiments . . . . . . . . . . . . . . . . . . . . . . . 28 3.6.1 Half-cell Capacity Tests of Diglyme Electrolytes . . . . . . . . . . . . 28 3.6.2 Full Cell Capacity Tests of Diglyme Electrolyte - Galvanostatic Cycling 28 3.6.3 Cycling Different Voltage-regions of the Active Materials . . . . . . . 29 3.6.4 Cycling of NVP/NVP Cells . . . . . . . . . . . . . . . . . . . . . . . 29 3.7 Physico-chemical Characterisation of Capacity Fade . . . . . . . . . . . . . 30 3.7.1 GC/MS of Cycled Cell Separators . . . . . . . . . . . . . . . . . . . 30 3.7.2 SEM and EDX Analysis of Cycled Anodes . . . . . . . . . . . . . . . 30 3.7.3 FTIR-Spectroscopy of Cycled Cells . . . . . . . . . . . . . . . . . . . 31 3.7.4 Binder Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 4 Results and Discussion 33 4.1 Stability of Na Electrodes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 4.1.1 Gas Evolution from Na Surfaces . . . . . . . . . . . . . . . . . . . . 34 4.1.2 Impedance Evolution of Na Electrodes diglyme electrolyte . . . . . . 34 4.1.3 Storing Na Metal in Electrolyte . . . . . . . . . . . . . . . . . . . . . 35 4.2 Ionic Conductivity of Diglyme-based Electrolyte . . . . . . . . . . . . . . . 36 4.3 ESW of the Diglyme-based Electrolyte . . . . . . . . . . . . . . . . . . . . . 36 4.3.1 Oxidative Stability of Diglyme-based Electrolyte . . . . . . . . . . . 37 4.3.2 Reductive Stability of Diglyme Electrolyte . . . . . . . . . . . . . . . 38 4.4 Characterising Loss and Performance . . . . . . . . . . . . . . . . . . . . . . 39 4.4.1 Half Cell Capacity/Voltage Profiles . . . . . . . . . . . . . . . . . . . 40 4.4.2 Full Cell Capacity/Voltage Profiles . . . . . . . . . . . . . . . . . . . 40 4.4.3 Comparing Full and Half Cell Capacity Retention . . . . . . . . . . 42 4.4.4 Comparing Efficiency and Irreversibility of Full and Half Cells . . . . 44 4.4.5 Performance at Different Voltage Regions of the Cathode and Anode 45 4.4.6 Another voltage range–NVP|NVP-cells . . . . . . . . . . . . . . . . . 46 4.5 Investigating Degradation Mechanisms . . . . . . . . . . . . . . . . . . . . . 48 4.5.1 GC/MS Analysis of Degradation Products in Cycled Cells . . . . . . 48 4.5.2 Reduction Schemes for Diglyme . . . . . . . . . . . . . . . . . . . . . 49 4.5.3 FTIR of Cycled HC Anodes . . . . . . . . . . . . . . . . . . . . . . . 50 4.5.4 Surface Structure of Pristine and Cycled HC . . . . . . . . . . . . . 52 4.5.5 EDX of Anodes from cycled Full Cells . . . . . . . . . . . . . . . . . 53 4.6 Test of Binder Defluorination . . . . . . . . . . . . . . . . . . . . . . . . . . 55 4.7 Visual Summary of Degradation Mechanisms . . . . . . . . . . . . . . . . . 56 5 Conclusions and Outlook 57 Bibliography 59 A Industrial Uses of Glymes With R=Me and n=1-4 I B Details of CVs From Half Cells III viii Contents C Full Cell Reconstructions IV D Reduction Schemes for PVDF Binder V E Ionic Conductivity and Viscosity Data VI F Images of Swagelok Insulating Polymers After Cycling VII ix Contents x List of Figures 1.1 Map of world’s most Li producing countries and the total mining output during 2015 [1]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 2.1 The fundamental principles of a galvanic cell. . . . . . . . . . . . . . . . . . 6 2.2 Conventional components of an electrode coated on a metal current collector. 6 2.3 Molecular structure of some common solvents used in alkaline metal batteries. 10 2.4 Decomposition scheme taking a range of different Li-ion battery processes into account when using carbonate electrolytes. Reprinted with permission from [2]. Copyright 2011 American Chemical Society. . . . . . . . . . . . . 11 2.5 Intercalation of Na-atoms in turbostratic graphitic regions of HC and of Li-atoms between ordered layers of graphite . . . . . . . . . . . . . . . . . . 12 2.6 The polyanionic NVPF. Each vanadium atom is in the middle of an octa- hedra bonding PO4 3− and F– speices. In the interlayer Na is intercalated. Rendered using VESTA [3]. . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 2.7 Schematic of the NVPF Na extraction mechanism looking down into the intercalation plane of Na : Showing 1) Extraction of the first sodium, 2) internal reorganisation into a 3) meta-stable phase and finally 4) extraction of the second reversible sodium. For each sodium extraction a vanadium changes oxidation state. Associated potentials of extraction are given vs. Na+/Na◦. Based on [4]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 2.8 Molecular structure of binders used in SIB electrodes. . . . . . . . . . . . . 14 2.9 Number of publications until 2015 of the different parts of a SIB. Light green bar is projection for 2015. Reproduced under the CC-BY licence from [5]. . 15 2.10 Molecular structure of some salts used in SIBs. . . . . . . . . . . . . . . . . 16 2.11 General structure of glymes. "n" denotes chain length. . . . . . . . . . . . . 16 3.1 Principle of a three electrode setup. A current is measured between the counter and working electrodes. Voltages are, instead measured against a third, reference electrode. Image adopted with permission from [6]. . . . . 20 3.2 Principle of a GC/MS setup. . . . . . . . . . . . . . . . . . . . . . . . . . . 22 3.3 Sequence of electrode fabrication procedure featuring, a), tape cast pre- cursor on both sides of Al current collector, b), scraped precursor and, c), punched electrodes with a set of current-collectors prepared for weighting . 23 3.4 Schematic of Coin cell assembly . . . . . . . . . . . . . . . . . . . . . . . . . 25 3.5 Schematic of Swagelock cell assembly . . . . . . . . . . . . . . . . . . . . . . 25 3.6 Layout of pressure cell used for determining gas evolution from Na in contact with electrolyte. Based on illustration from [7]. . . . . . . . . . . . . . . . 27 3.7 Experimental procedure used to compare degradation of binders by Na in EC:DMC- and diglyme-based electrolytes. . . . . . . . . . . . . . . . . . . . 32 xi List of Figures 4.1 Pressure evolution at 25 ◦ C from Na metal stored, A), in 1 M NaPF6 in diglyme and ,B),in 1 M NaPF6 in EC50:DMC50. At the left y-axis and at the right y-axis the total pressure and amount of gas related to the total number of moles of Na are shown, respectively. . . . . . . . . . . . . . . . . 34 4.2 Impedance spectra at different times for a symmetric Na|Na Swagelok cell left at OCV for 90 hours. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 4.3 A Na sphere stored with 1 M NaPF6 in diglyme . . . . . . . . . . . . . . . . 35 4.4 Walden plots for a NaPF6 diglyme electrolyte at three different concentrations. 36 4.5 Linear sweep voltammogram for 1 M NaPF6 in diglyme (Blue) and 1 M NaPF6 in EC50:DMC50 (Orange), respectively. . . . . . . . . . . . . . . . . 37 4.6 Cyclic voltammograms showing 5 cycles from a half cell of NVPF using 1 M NaPF6 in diglyme. Inset: Magnification at 3.4 V to visualise small features with related surface charge densities listed in figure header. . . . . . . . . . 37 4.7 Linear voltammograms indicating reductive behaviour of diglyme and EC50:DMC50 electrolytes with 1 M NaPF6 on Cu-foil. Inset: magnified part of the curve. 38 4.8 Typical cell profiles of half cells cycled in a diglyme 1 M NaPF6 in diglyme 40 4.9 Differential capacity curves of half cells cycled in a diglyme 1 M NaPF6 in diglyme. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 4.10 Typical cell profile in a cycled full cell using 1 M NaPF6 in diglyme with corresponding differential capacities. Inset: visualisations of small features. The capacities are plotted in terms of the weight of cathode material, mak- ing comparison easier between half and full cells. . . . . . . . . . . . . . . . 41 4.11 Capacity retention and coulombic efficiency per cycle of full cells, using 1 M NaPF6 in diglyme (A) or 1 M NaPF6 in EC50:DMC50 (B) . . . . . . . . 42 4.12 A), total specific capacities and, B), specific discharge capacities normalised by discharge capacity on first cycle for full and half cells of NVPF and HC. Setup using a 1 M NaPF6 diglyme electrolyte. . . . . . . . . . . . . . . . . . 43 4.13 Evolution of the two plateaus in the full cell with cycling. Inset: Legend to what parts of the full cell are denoted by upper plateau, lower plateau and full cell. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 4.14 Actual and recombined efficiencies and accumulated irreversibilities for half and full cell. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 4.15 Convention when naming the different voltage regions for a fully balanced system. Based on half cell profiles. . . . . . . . . . . . . . . . . . . . . . . . 45 4.16 Specific capacity retention when exploring different regions of the electrode materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46 4.17 Specific capacity retention for a NVP|NVP cell as compared to the NVPF|HC full cell. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 4.18 Gas chromatograms from experiment aiming to analyse soluble non-charged degradation products from cycled cell separators. . . . . . . . . . . . . . . . 48 4.19 Reduction scheme based on the breaking of the C1-O1 bond in the diglyme molecule. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 4.20 Reduction scheme based on the breaking of the O1-C2 bond in the diglyme molecule assuming that no environmental hydrogen is available. . . . . . . . 50 4.21 Reduction scheme based on the breaking of the O1-C2 bond in the diglyme molecule assuming that environmental hydrogen is available. Crossed out boxes relate to schemes deemed impossible according to the discussion in the text. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 xii List of Figures 4.22 FTIR spectrum of a HC surface cycled 2 times compared to a pristine electrode sokaed in electrolyte and dried for 12 hours. . . . . . . . . . . . . 51 4.23 SEM micrographs of pristine HC. Inset: Magnifications of one part of the pristine HC surface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 4.24 SEM - micrographs of a pristine HC electrode cycled in 1 M NaPF6 in diglyme for 50 cycles. Upper and lower insets: Magnifications of lighter part and darker part of surface, respectively. . . . . . . . . . . . . . . . . . . 53 4.25 EDX spectrum aqcuired from a HC electrode cycled 1.5 times in full cell setup using a diglyme 1 M NaPF6 electrolyte. . . . . . . . . . . . . . . . . . 53 4.26 Photo taken of two electrodes cycled in full cell setups showing a bluish tone as compared to the pristine electrodes. . . . . . . . . . . . . . . . . . . 53 4.27 Pourbaix diagram for Vanadium species in aqueous solutions. Printed with permission from [8]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54 4.28 Results from experiment studying binder degradation. PVDF binder expe- riences severe reaction in when exposed to Na in diglyme based electrolyte but seems to be stable in EC:DMC based counterpart. . . . . . . . . . . . . 55 4.29 Separator showing HC disintegration after cycling. . . . . . . . . . . . . . . 55 4.30 Different mechanisms occurring sequentially from right to left, top to bot- tom, that are contributing to system capacity fade. . . . . . . . . . . . . . . 56 B.1 Cyclic voltammogram from cycling an EC50:DMC50/NaPF6 electrolyte be- yond the third peak five times . . . . . . . . . . . . . . . . . . . . . . . . . . III C.1 Capacity curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . IV D.1 Reduction scheme for PVDF binder. . . . . . . . . . . . . . . . . . . . . . . V E.1 Measured temperature dependence of Ionic conductivity and Viscosity for a G2 NaPF6 electrolyte. In the case of ionic conductivity the 1 M electrolyte has only been measured once due to lack of avalibility. . . . . . . . . . . . . VI F.1 Different types of polymers used as insualtionin Swagelok cells comprising NVPF and HC with a G2 1M NaPF6 electrolyte. Films are recovered under glovebox conditions after cycling. . . . . . . . . . . . . . . . . . . . . . . . . VII xiii List of Tables 2.1 Characteristics of two NASICON-type materials, NVPF and NVP. . . . . . 12 2.2 Chain length, common name, acronym, and physical properties of the first four glymes in the series. Data for EC and DMC are provided as a reference. Data adapted from [9] unless other reference explicitly stated. . . . . . . . . 16 2.3 Cost of diglyme per unit weight as compared to EC and DMC. . . . . . . . 18 3.1 Electrode area, mean weight of current collector and standard deviation of current collector (C.C.) weight for two different sizes of electrodes used. . . 23 3.2 Cut-off voltages for cycling of materials in half-cells, using Na metal as counter electrode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 3.3 Cut-off voltages for cycling of materials in full cells . . . . . . . . . . . . . . 29 3.4 Regions cycled and their approximate voltage intervals for the anode and cathode as approximated by half-cell reconstructions. masses are values used for one experiment. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 3.5 Cycling scheme for symmetrical NVP/NVP cell . . . . . . . . . . . . . . . . 30 3.6 Cut-off voltages and cycling scheme for further exploring possible vanadium extraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 4.1 Reduction potentials and activation energies, Ea, for diglyme at different levels of complexation with Na+. . . . . . . . . . . . . . . . . . . . . . . . . 39 4.2 Initial irreversibillity (Iinitial) and coulombic efficiencies (ηc) when cycling different regions of HC and NVPF in a 1 M NaPF6 diglyme electrolyte using a full cell setup. The potential ranges have been estimated using half- cell reconstruction upon first charge. ?: Not comparable since cell is not discharged fully. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 A.1 Industrial uses of Mono- and diglyme. Directly adapted from [9] . . . . . . I A.2 Industrial uses of glymes. Directly adapted from [9] . . . . . . . . . . . . . . II xiv List of Abbreviations List of Abbreviations Abbreviation Explanation CV Cyclic voltammetry An electrochemical technique where the cur- rent response of a setup is recorded as the function of applied voltage. ESW Electrochemical stability window The range of potentials at which a substance can exist without oxidising or reducing. EIS Electrical impedance spec- troscopy Technique where the frequency dependent ac- impedance of a system is investigated to mon- itor changes in bulk or surface morphology. Cs Specific capacity The capacity per unit mass of an electrode material. Measured in mAh g−1. Es Specific energy The energy per unit mass of total electrode material or whole unit a complete cell or bat- tery. Measured in Wh kg−1. GCPL Galvanostatic cycling with potential limitation Method for testing battery performance in which a constant current is drawn from the cell studied while the voltage is tracked. GC Gas chromatography An analytical technique used to separate gases with respect to their masses and ten- dency to flow through a gas column. LIB Lithium-ion Battery xv List of Abbreviations Abbreviation Explanation LSV Linear sweep voltammetry As cyclic voltammetry but only one voltage sweep is conducted either towards higher or towards lower potentials. MS Mass spectrometry An analytical technique where the ionised fragments of a substance are analysed. Cou- pled to GC it becomes a very sensitive analy- sis technique to precisely identify compounds in the analyte. NASICON Na-ion super ionic con- ductor A class of materials with good solid state ion conductive properties. Used as cathodes in SIBs. OCV Open circuit voltage The voltage between two battery terminal when no current is flowing. SHE Standard hydrogen eectrode A measurement standard in electrochemistry, toward which different substances and pro- cesses are often referenced. SEI Solid electrolyte interphase The interphase formed upon degradation of electrolyte components on the negative elec- trode. SIB Sodium-ion Battery xvi 1 Introduction Modern applied research is to a higher and higher de- gree focusing on solving the problems of a future sus- tainable energy system. Many efforts strive to find convenient, cheap ways of storing energy. Batteries surely belong to this category of solutions and has seriously been considered for use in larger scale appli- cations since the advent of the Li-ion battery (LIB) during the 1990:s. Room temperature sodium-ion batteries (SIB), on the other hand, is a new cate- gory of systems that have recently appeared and been considered for future use. The thesis at hand inves- tigates one type of SIB using a somewhat unconven- tional electrolyte. In this chapter a background to the research conducted is given along with the aim and limitations. 1 1. Introduction 1.1 Importance of Stable SIBs for a Future Energy System The signing of the climate treaty in Paris in 2015 was seen as a success from many per- spectives. All members of the United Nations agreed upon collectively, but according to their ability, to reduce their greenhouse gas emissions, to keep global warming below 2 ◦ C. However, to still keep our current way of consuming energy, large structural reforms of each nation’s energy systems are needed. Batteries and their development may have a major role to play when it comes to such challenges. Not only are they a key element for storing energy in tomorrows fossil-free electric cars, but they are also candidates for load balancing in a power grid with more renewable, intermittent, sources of energy, such as wind and solar power [10]. The current state-of-the-art batteries for consumer use excelling in energy density at room temperature are all based on lithium-ion technology. However, stakeholders for Li are not limited to battery manufacturers. At the same time Li production and resources are limited and centred to a few certain regions (Figure 1.1). This means that both geopolitical and supply reasons, motivates finding an alternative to Li in battery technologies could prove to be crucial if a development and expansion in the sustainable energy sector is not to be halted [1, 11]. In recent years, Na has risen as a potential candidate to fill the shoes of its alkali metal cousin. A major reason is that Na is approximately three orders of magnitude more abundant in the earths crust and reserves are abundant nearly all over the globe [12]. If only considering the price and raw-material availability as the main attributes on how to judge a battery technology’s future viability, Na has a great advantage. Figure 1.1: Map of world’s most Li producing countries and the total mining output during 2015 [1]. Even though theoretical energy density is lower for Na metal than for Li metal, pure metals are seldom used in real battery systems, due to safety reasons. This also gives SIBs a possibility of having a better performance than LIBs–if the SIB materials could be designed smart enough [13]. Such outlooks have spurred a revival of the SIB field, turning dormant after the commercialisation of LIB cells in the 1990:s. Many projects have started worldwide, connecting industry and academia to promote the development of commercially viable SIB solutions. One such endeavour is the ”Naiades” consortium 1, 1http://www.naiades.eu/ 2 http://www.naiades.eu/ 1. Introduction funded by the EU Horizon 2020 programme. As for now, a good candidate system for electrodes in a future commercial SIB consists of a cathode made of either the polyanionic framework Na3V2(PO4)2F3 (NVPF) or the polyanionic framework Na3V2(PO4)3 (NVP), and an anode made of hard carbon (HC). Electrolytes for SIBs have so far not been studied as extensively as the different materials for anodes and cathodes but are also very important, since they often determine cycling stability an safety of a battery [14]. Such properties are equally important for commercial- isation of any SIB system. Even though much know-how on is thought to be transferable from the LIB side, each electrolyte still has to be rigorously tested for use in these new systems. In the end it is the system which has to show good performance and not only the individual components. The scene is thus open for improvements in electrolytes that stabilise prospective SIB systems. Any electrolyte living up to such criteria should be considered for use, since the correct choice might vary among different applications–a long term energy storage will not demand the same characteristics as a battery in a cell-phone or electric toothbrush. There is also a need to properly understand any capacity fade mechanism for a future SIB since it could help in the development of mitigation strategies. Electrolytes based on diethylene glycgol dimethyl ether, also known as diglyme or ”G2”, have been considered for use in batteries due to this solvent’s good ionic conductivity, favourable viscosity, thermal stability and high donor number, having the ability to chelate small cations [15, 9]. Also, since diglyme is an ether and is already more reduced than any carbonate solvent, it is possibly more stable against further reduction at the anode. Despite this, no study has gone into detail of neither its electrochemical stability of diglyme based electrolytes, their performance in SIBs or evaluated their suitability in a prospective SIB systems. 3 1. Introduction 1.2 Aim This master’s thesis project has aimed to evaluate the possibilities of using 1 M NaPF6 in diglyme with future SIBs and to understand the electrolyte’s stability and degradation mechanisms. Even though the main focus has been but on the electrolyte, the intricate nature of any battery system has required the work to also encompass studies of the interplay between all the system components. Apart from studying systems behaviour, the work has further aimed to evaluate the use of Na-metal electrodes as references using the mentioned electrolyte. The aim of the initial literature study has been for the reader to internalise knowledge of battery systems and to summaries the research relevant for anyone desiring to considering to use diglyme, also summarising such aspects as toxicity and cost. 1.3 Limitations No focus has been put on changing the salt into other types than NaPF6, even though this might provide fruitful in terms of cell performance. Nor has any other solvent been tested. Instead previous data of the same system, using Ethylene Carbonate (EC) and Dimethyl Carbonate (DMC) in a 1:1 ratio as a solvent, has been used as a benchmark when considered necessary. The cell setups have in these cases comprised the electrodes: NVPF, HC and NVP. No further studies of the electrodes used were done, even though there it is possible for such change being able to lead to capacity or stability improvement. The main task has been not to optimise all cell components, but to characterise the presumptive degradation and performance of the electrolyte at hand. Furthermore, the work has solely been a liter- ature survey complemented by experimental work. Where modelling and computational methods have been needed, it has been provided by a third party. Even though, future commercialisation of SIBs would require an in-depth study of possible applications and a market analysis, no attempt has been made to do so here. Also for commercialisation, larger systems comprising many cell elements would have to be investigated to ensure proper performance. This, as well has not been considered and is left for future work. 4 2 Batteries and SIBs In the following sections, the theory relevant for this project is presented. First, the general concepts of batteries will be presented. Second, a section intro- ducing the reader to electrolyte fundamentals is laid out. Third, comes a section presenting the background to electrolyte decomposition and the importance of formation of stable surface films. Fourth, a section ex- plaining the SIB with its components, their challenges, and future prospects is presented. Finally comes a sec- tion featuring a short literature review of the glyme family of solvents–the category that diglyme belongs to. An attempt is made both to cover the electro- chemistry relevant for batteries and the more practical aspects, such as cost and toxicity. 5 2. Batteries and SIBs 2.1 Batteries–Primary and Secondary Galvanic Cells Galvanic cells are defined as devices able to store electrical energy as chemical energy. In daily life, we usually talk about ”batteries” but strictly speaking a battery is a device comprising many galvanic cells. Since the main part of this thesis will not be dealing with such large systems, the words ”battery”, ”galvanic cell”, or simply ”cell” will be used interchangeably, unless indicated otherwise. In the model battery, two electrodes of dif- ferent electrochemical potential are kept apart by a separator, preventing the electrodes from touching each-other (Figure 2.1). The electrodes are often made up of an active ma- terial, a conductive matrix, such as carbon black, serving to improve electrical connection among active-material grains, and a binder that keeps everything together (Figure 2.2) [5]. The electrode is, further on, usually coated on top of a current collector made of metal providing mechanical stability and an evenly distributed surface current [16]. Figure 2.1: The fundamental principles of a galvanic cell. Figure 2.2: Conventional components of an electrode coated on a metal current collector. 6 2. Batteries and SIBs Apart from preventing electrodes from touching each-other and short circuiting, the sep- arator is responsible for hosting the electrolyte. Mostly consisting of a salt containing the ion of interest that is dissolved in an organic solvent, the purpose of the electrolyte is to conduct ions inside the cell [17]. Optimally the electrolyte should be stable at the redox potentials of the electrodes to not degrade and cause unwanted side-reactions. When discharging a galvanic cell, the electrodes are connected by an external load. Be- cause of a difference in redox potential associated with the anode and cathode, electrons will start flowing through the outer circuit [16]. At the same time ions will flow in the electrolyte to maintain charge neutrality. The flow of electrons and ions is sustained by a continuous oxidation of anode species, giving up electrons and ion, and a continuous reduction of cathode species, absorbing electrons and ions. A current supported by a change in oxidation state of participating species is denoted as a ”Faradic current”. In a battery, such a redox process gradually lowers the total internal energy of the system, the current being sustained until the full capacity of either of the active materials has been depleted. The discharge capacity is the the total number of redox conversions, providing a charge to the external circuit, that can occur reversibly, and is dependent on the type and amount of active material used. The potential at which extraction or insertion of ions and electrons takes place will instead be dependent on the material structure and instantaneous composition, making the fraction of inserted/extracted ionic charge carriers a factor determining the cell potential. When charging a galvanic cell, if that is possible, the opposite process takes place: Energy is supplied to the system and stored by reversing the flow of ions and electrons, causing the internal energy of the system to increase. The separator and electrolyte can not be electrically conducting, since this would result in a self-discharge of the cell–i.e. electrons and ions would both propagate in the electrolyte, thermodynamically equilibrating the system. It is the energy of the flowing electrons in an outer circuit that can be harnessed by a battery-powered devices, such as a car or a laptop [18]. Features mentioned so far are common for all batteries. Making an initial distinction, galvanic cells can be divided into two main categories: primary and secondary cells. Primary cells are driven by chemical reactions that are irreversible, meaning that the cells cannot be recharged after the first cycle [16]. Cells are assembled in the charged state, discharge by the end consumer and then discarded–hopefully for material recycling. Thus, total energy density, cost and environmentally friendliness become important design parameters for primary cells. A common example of a primary cell is the alkaline batteries used by low demanding consumer electronics. The chemistry involved during discharge is given by: Anode : Zn(s) + 2 OH−(aq) −−→ ZnO(s) + H2O(l) + 2 e− [e◦= −1.28V] Cathode : 2 MnO2(s) + H2O(l) + 2 e− −−→ Mn2O3(s) + 2 OH−(aq) [e◦= 0.15V] Total : 2 MnO2(s) + Zn(s) −−→ Mn2O3(s) + ZnO(s) [Vcell= 1.43V][16] Secondary cells, on the other hand, are designed upon reactions that are reversible, yielding batteries that can be recharged [19]. With rechargeable cells other aspects of cell development than energy density and cost becomes important. A rechargeable bat- tery must show good cyclabillity to not loose its capacity when charged and discharged. This is also referred to as showing a low irreversibillity. Another requirement is that the amount of polarisation experienced during charge is low, polarisation meaning the 7 2. Batteries and SIBs difference in cell voltage between charge and discharge at the same level of charge. Polari- sation, thus, determines the energy efficiency 1 of storage in the battery. Furthermore, the systems must be safe with a margin to not risk to damage of persons or property, even if the batteries are slightly mishandled by the user. The charge rate is expected to be an increasingly important factor end-users of electrical vehicle [20] since people are expected to prefer rapid charge of their electrical vehicles. In many cases secondary batteries are built upon intercalation electrode materials, where the ion responsible for shuttling the charge takes up a vacant space in the anode or cathode material matrix that keeps its structural integrity upon insertion and extraction [5]. Structures featuring these properties are often denoted ”Topotactic materials” [21, p.9]. One of the most common chemistries employed in early high performance batteries showing reasonable performance in terms of both total capacity and cyclabillity, is the lithium ion rocking chair battery, commercialised by Sony in 1991 [22], which utilises a metal oxide cathode and a graphite anode. When charging, the half-cell reactions are given by: Anode : Li+ + C6 + 1 e− −−→ LiC6 Cathode : LiCoIIIO2 −−→ Li+ + CoIVO2 + e− and similarly upon discharge the reactions are given by Anode : LiC6 −−→ Li+ + e− + C6 Cathode : CoIVO2 + Li+ + e− −−→ LiCoIIIO2 [23]. Upon charge (CoIII) is oxidised. Li+ ions will at the same time leave the LiCoO2 to compensate for the lost electron’s charge and intercalate inside the graphite. During intercalation Li+ is recombined with the electron and reduced to its metallic state (Li). Upon discharge, the Li is instead oxidised to an ion (Li+) while the other specie (CoIV ) is reduced. The ion is then incorporated into the lattice of the cathode material, to maintain charge neutrality [16]. The reason for using intercalation materials such as graphite for the anode, instead of pure lithium metal, which would yield a higher energy density, is that alkali metal electrodes have shown to experience unwanted dendrite formation. The dendrites are structures that successively grow upon cycling because of kinetic factors. They have the unwanted ability of penetrating the separator and finally short circuiting the battery. A short circuit, in turn, may cause dangerous overheating and explosion, which could endanger the end user [19]. 1Not to be confused with coulombic efficiency, determining the ratio of charges shuttled at charge and discharge. 8 2. Batteries and SIBs 2.2 Electrolyte Fundamentals - Salts and Solvents The most commonly used electrolytes reported for alkali metal ion batteries are made up of two components: a salt and an organic, non-aqueous, solvent. The salt is a source of the ions to be transported during charge and discharge and the solvent provides a matrix and a means of dissociating the salts. The ability of an electrolyte to conduct ions is called its ionic conductivity, σ, measured in mScm−1. Ionic conductivity is the ion analogue to conductivity (reciprocal resistivity) for electrons in an electrical circuit; it determines how much loss is associated with the process of moving ions back and forth in the electrolyte. Being related to a liquid behaviour, the ionic conductivity can be modelled by a empiric Vogel-Tamman-Fulcher expression [17, 24]: σ = σ0e −B (T −T0) , (2.1) where T is the absolute temperature, T0 is a temperature related to the glass transition temperature, B is an empiric parameter and σ0 is the ionic conductivity at T0. Also, remembering that the viscosity, η, can be modelled using the similar functional relationship [24]: η = η0e B (T −T0) , (2.2) where η0 is the viscosity at T0, we see that σ0 and η0 have a reciprocal relationship according to: ησ = constant, (2.3) known as the "Walden rule" [25]. By plotting 1 η against σ, experimentally it is possible to determine if the electrolyte conforms to equation (2.3) and a vehicular conduction mechanism of ions. The ionic conductivity of an electrolyte is dependent on both the salt and solvent used. Fig- ure 2.3 shows some common reported solvent molecules for both SIBs and LIBs [17]. Apart from providing a good conduction and electrochemical stability, which will be touched upon in Section 2.3, solvents have to fulfil other criteria [13]. The liquidus range of the electrolyte has to be broad; the electrolyte should neither freeze, nor boil, even if the am- bient temperature drops or increases several tenths of degrees. Thermal stability (which is not the same as liquidus range) is also necessary to consider as it tells us how a solvent may respond to a unintended increase in temperature. In these cases the solvent should not degrade exothermally–causing thermal runaway; the final battery should be as unlikely to catch fire or even explode, as possible. Furthermore, viscosity of the electrolyte should be low enough to allow for proper soaking of the separator and electrodes [13]. Since no single solvent is usually able meet all of the criteria above, practical electrolytes many times consist of mixtures of several solvents with one salt. One such composition, com- monly used for Li ion batteries, is the LP 30 electrolyte–a 1:1 binary mixture of ethylene carbonate (EC) and dimethyl carbonate (DMC) with 1 M LiPF6 salt [13]. Salts have the same requirements as solvents when it comes to thermal stability [14]. Another important property is that the salt anion should not cause any unwanted side reaction with any part of the cell. Further characteristics of a good salt include easiness to remove residual waters, low cost, toxicity and a high degree of dissociation – i.e. to which degree the salt provides mobile charge carriers in the electrolyte [13]. The degree of dissociation is not only a salt dependent factor, but also depends on how strongly the specific solvents act to dissociate the salt – a property approximately proportional to the solvents dielectric constant. Salt dissociation is thus both a question of salt, solvent and how they match. One salt that has been shown to have good or decent performance with respect to these aspects for Li-ion batteries is LiPF6–a component in the LP30 electrolyte. 9 2. Batteries and SIBs O O O DMC–Dimethyl carbonate O O O DEC–Diethyl carbonate O O O EMC–Ethyl methyl carbonate O O G1–Monoglyme Linear Cyclic O O O PC–Propylene carbonate O O O EC–Ethylene carbonate Figure 2.3: Molecular structure of some common solvents used in alkaline metal batteries. 2.3 Electrode-Electrolyte Interfaces and Interphases Constructing stable, safe, alkali metal batteries with long cycling life does not only include developing efficient electrode materials. The interface between the electrolyte and the electrodes is also important. Many conventional solvents used in LIBs and SIBs have a to low reductive stability for the potentials reached at the cell anode, meaning that ionic charge carriers would get lost continuously if there was no balancing mechanism. The solution to the problem comes from the kinetics of electrolyte decomposition [17]. As predicted by reduction potentials, there is an initial solvent decomposition. However, instead of continuing the reaction, the decomposition products serve to render a film and passivate the surface. The film formed, commonly called the Solid Electrolyte Interphase (SEI), is formed during the first charge and then stabilised during subsequent cycles [17]. Petr Novák et al. have described properties of an ideal SEI, noting that it should be electrically insulating, leading to prevention of further electrolyte breakdown, but a good ion-conductor, leaving kinetic performance and polarisation of the cell unaffected [26]. Another quality would be that they are insoluble in the solvent used for the electrolyte, leaving a stable film. Attempts to probe the SEI have been made in a number of studies for lithium ion batteries [26]. For SIBs results are scarce. Weadock et al. used colloidal probe microscopy on metal surfaces cycled against sodium metal in a 1 M NaPF6 1:1 ethylene carbonate and diethyl carbonate (EC:DEC) electrolyte [27]. By mapping out the Young’s modulus of the surface at different cycles, he showed that the SEI seemed to consist of a bi-layered structure that had stabilised after the first cycles. The proposed structure corresponds to a model proposed by Doron Aurbach investigating SEI growth on Li metal anodes [28]. According to his study, the SEI in carbonate solutions forms a multi-layered structure relating to the difference in reduction potential of the possible species in the electrolyte solution; all components with a reduction potential higher than that of lithium will be reduced initially. Thereafter more selective growth of the films take place only featuring reduction products with the highest reduction potential. 10 2. Batteries and SIBs In the same study the film formation on graphitic carbons was investigated. Since the carbons were cycled galvanostatically, the film growth is even more selective with only those components in solution that have a higher reduction potential reducing first. The films formed initially may thus be so resistive that solvent components with lower reduction potentials are never decomposed. Another feature of intercalation anode materials is that they show some degree of volume expansion when sodiated/lithiated. The expansion causes a cracking of the surface film and further decomposition of the electrolyte when the small cracks are healed. The cracking-healing is continued until a film with the right viscoelastic properties is formed [28]. Figure 2.4: Decomposition scheme taking a range of different Li-ion battery processes into account when using carbonate electrolytes. Reprinted with permission from [2]. Copy- right 2011 American Chemical Society. Other studies have studied the chemical composition of the SEI. G. Gachot et al. used Gas Chromatography and Mass Spectrometry coupled to FTIR to investigate primary and secondary electrolyte decomposition products of LiPF6 in cyclic and linear carbonates, such as EC-DMC (Figure 2.4) [2, 29]. When conducting a similar study for SIBs it was found that the higher redox potential of Na limited the rate of further reduction of decomposition products, such as diethylene carbonate. The lack of reduction indicates that the SEI would have a different character and different (worse) performance in terms of stability [30, 31]. 2.4 Sodium-ion Batteries The materials and concepts for possible future rechargeable SIBs are in many ways similar to the rocking chair lithium ion batteries described in Section 2.1. This is much because the chemistry of the two alkali metals are similar. Even though the research in room temperature SIBs is lagging behind its Li counterpart, the prospect of cheaper cells using a nearly unlimited sodium resource, has encouraged a revival of the field, as can be be found in many recent reviews [23, 13, 5, 32, 33, 34]. In this section, a summary is given of the different materials and electrolytes currently used in SIBs. 11 2. Batteries and SIBs 2.4.1 SIB Anode Materials In the case of sodium, as compared to lithium, graphite is not an optimal insertion material [21, p.9]. In graphite, the chemical potential for Na+ intercalation is higher than that for plating and thus the Na prefers to electroplate instead of intercalating [35]. Instead, hard carbon (HC), synthesised directly from pyrolysis of sugar, can be used, providing a specific capacity of more than 300 mAhg−1 [36, 37]. As elaborated upon by Stevens and Dahn, sodium inserts in two distinct ways into hard carbon resulting in two different insertion- voltage regions[35]: Nanoporous, turbostratic, graphite-like regions result in a sloping insertion-voltage region, which is followed by insertion into mesoporous regions, providing a steady voltage plateau at 0.05 V vs. Na+/Na◦. The difference between intercalation in graphite-like regions and pure graphite is illustrated in Figure 2.5 [35]. Hard carbon is, thus, an attractive alternative as a future anode material. Unfortunately, it still shows a large initial irreversibility when cycled in conventional electrolytes. To tackle this it has been proposed to use extra sodium sources [31]. Figure 2.5: Intercalation of Na-atoms in turbostratic graphitic regions of HC and of Li-atoms between ordered layers of graphite 2.4.2 SIB Cathode Materials The transition metal oxides used in lithium ion batteries have, in some cases, proved to be unsuitable as cathode materials due to large polarisation [38]. Also, the materials show less capacity than they theoretically could have [32, 5]. With this being said, the metal oxides could still very well prove to be usable. Instead of transition metal oxide a parallel class of polyanionic materials have been found to work [39]. One subclass of such compounds is called NASICONs (Na-ion super ionic conductors) and was initially used for solid electrolytes. These intercalation frameworks have been found to show good performance and specific capacities of up to 128 mAh g−1 [40]. Two NASICON type materials are Na3V2(PO4)3 (NVP) [4, 40] and Na3V2(PO4)2F3 (NVPF) [41, 42]. Their general electrochemical properties are outlined in Table 2.1. Insertion potential vs Na+/Na◦ Theoretical specific Capacity [mAh g−1] NVPF 3.7, 4.1 128 NVP 1.6, 3.4 118 Table 2.1: Characteristics of two NASICON-type materials, NVPF and NVP. For NVPF, each sodium atom is stored in a matrix of edge-sharing octahedra centred around V 3+ species. Octahedral corners consists of either PO4 3− anions or F– anions. In the inter-layer spacing sodium ions are intercalated (Figure 2.6). The insertion and extraction of sodium from NVPF happens at four distinct voltages distributed over the two main plateaus (Table 2.1), elaborated upon by Park et al. [41]. Out of three adjacent sites for the alkali metal, two can be emptied without collapsing the NVPF structure. 12 2. Batteries and SIBs Figure 2.6: The polyanionic NVPF. Each vanadium atom is in the middle of an octahedra bonding PO4 3− and F– speices. In the interlayer Na is intercalated. Rendered using VESTA [3]. Upon each extraction of a sodium-ion, a vanadium atom changes oxidation state from V + 3 to V + 4 according to: Na3V2(PO4)2F3 −−→ Na3−xV2(PO4)2F3 + xe− + xNa+(In electrolyte). Among the adjacent sites, the first sodium ion is easiest to extract as it leads to a metastable trans-configuration of the material, minimising the sodium ion-ion repulsion [41]. For both the first and second extraction of sodium from the lattice the associated potentials vs. Na+/Na◦ change (Figure 2.7). Figure 2.7: Schematic of the NVPF Na extraction mechanism looking down into the intercalation plane of Na : Showing 1) Extraction of the first sodium, 2) internal reorgani- sation into a 3) meta-stable phase and finally 4) extraction of the second reversible sodium. For each sodium extraction a vanadium changes oxidation state. Associated potentials of extraction are given vs. Na+/Na◦. Based on [4]. 13 2. Batteries and SIBs NVP, on the other hand features two voltage plateaus of asymetrical capacity (50 mAh g−1, 118 mAh g−1) [4][40] but upon synthesis one intercalation site, at the lower potential of 1.6, is empty and one, at the upper potential of 3.4, is filled. This means that pristine NVP can be used both as a low capacity anode or as a high capacity cathode. The insertion potentials are, however, lower than for NVPF resulting in a lower specific energy [4]. Returning to the NVPF, the specific energy density of a system featuring a hard carbon anode and a NVPF cathode can be calculated. By assuming that the mean voltage of the slope and plateau of the hard carbon is 0.25 V vs Na+/Na◦ [43], the energy density of a perfectly balanced system becomes 284 kWh kg−1 . Such an energy density is comparable to the first generations of Li-ion batteries given that the assembled system, with electrolyte, has a good capacity retention. 2.4.3 SIB Binders and Current Collectors Binders are often polymers serving to hold the active material of the electrodes together [5]. The molecular structure for some binders reported for SIB:s are shown in Figure 2.8. Among these, the fluorinated polymer PVDF, very similar in structure to Teflon®(PTFE), is the most commonly used in literature [5]. However, as reported in some recent studies , PVDF seems to suffer from gradual defluorination at low potentials when switching from a lithium to a sodium chemistry [5, 44, 45, 46]. Since the defluorination relies on the consumption of sodium, this is a process which inevitably contributes to capacity fade in the battery. Instead, carboxymethyl cellulose (CMC), has been suggested as a replacement [45]. This is an organic polymer sometimes found as a food additive [47]. When used as a binder for SIB applications the side-group, R, is chosen to be sodium. A third binder reported in literature is sodium alginate, a polymer extracted from brown algae [44]. Both CMC and sodium alginate have been reported to result in a slower capacity loss compared to PVDF when used as a anode material binder [44]. Figure 2.8: Molecular structure of binders used in SIB electrodes. When it comes to current collectors, one of the major selling points when it comes to SIB-technology has been that cathodic current collectors can be switched from copper to aluminium since sodium does not alloy with the latter, while lithium does. The change of metals makes the cell setup both cheaper and lighter. As a figure of merit, the price for aluminium as of the end of March 2016 was 1 550 USD per ton while copper cost approximately 5000 USD per ton [48]. At the same time, the specific density of aluminium is 2700 kg m−3 while that of copper is 8920 kg m−3 [49]. The tensile strength of aluminium is about 60 % of that of copper, possibly forcing a manufacturer to use a current collector 14 2. Batteries and SIBs with approximately the double thickness [50]. Judging from the physical characteristics, aluminium is still ca. 40 % lighter and 40 % cheaper than copper. 2.4.4 SIB Electrolytes For SIBs, the electrolyte is the one component that has received the least attention in literature (Figure 2.9 ). The reason is not necessarily that electrolytes deserve less at- tention but perhaps that many groups, during the time which the SIB concept boomed, have so far focused their efforts on developing those components setting the baseline for the capacity in SIBs. As described in the Section 2.3 electrolytes and their interplay with electrodes are very important in determining charge kinetics and stability of any battery. Figure 2.9: Number of publications until 2015 of the different parts of a SIB. Light green bar is projection for 2015. Reproduced under the CC-BY licence from [5]. Solvents reported for sodium-ion electrolytes are mainly those described in Section 2.2. Some common salts reported in literature for sodium-ion electrolytes can be seen in Figure 2.10. Eshetu et al. reported on the stability of these salts with respect to Al current- collectors [30]. They concluded that NaPF6 is the most stable candidate and attributed this stability to the formation of AlF3 species which protect the current collectors. The stability changes drastically if there is much residual water in the electrolyte, since NaPF6 risks breaking down according to the following mechanism: LiPF6 → PF5 + LiF(s) PF5 + H2OResidual → 2HF + PF3O(g)[29], (2.4) producing HF, which in turn can serve to degrade other parts of the system. Taking the stability of NaPF6 into account, and assuming the amount of residual water is in the lower ppm range, many studies in SIBs have utilised the same LP 30 (1M LiPF6 in EC50:DMC50) electrolyte for investigations, only changing the ion from lithium to sodium [51]. 15 2. Batteries and SIBs Name of Na salt Anion structure Name of Na salt Anion structure NaPF6  F P F F F F F  − NaClO4 O Cl O O− O NaTFSI F F F S O O N− S O O F F F NaFTFSI F F F S O O N− S O O F NaFSI F S O O N− S O O F Figure 2.10: Molecular structure of some salts used in SIBs. 2.5 The Glyme Family of Solvents Another set of organic molecules, than those described in Section 2.2, that have been considered for use in SIB electrolytes is the glyme family of solvents (Figure 2.11) [52]. For the sake of briefness only those members with n:s ranging from 1 to 4 and featuring R = CH3 will be reviewed here. The glymes are also abbreviated as ”G(n)”–G1 being monoglyme, G2 being diglyme, and so on. The names of the compounds studied are presented in Table 2.2, together with physical properties relevant for battery research. All the members of the glyme family are stable, amphiphilic, solvents, meaning that they to some degree show mixing both in polar and non-polar liquids. Another common feature of the glymes is that many of them are able to complex with metal ions to some degree, a property also denoted as ”Chelation” in literature [53]. Figure 2.11: General structure of glymes. "n" denotes chain length. n Common name Acronym Density at 25 ◦C (g cm−3) Melting point (◦C) Boiling point (◦C) Dielectric con- stant, ε at 25 ◦C - Ethylene Car- bonate EC 1.321 [54] 36 [54] 248 [54] 89.8 [17] - Dimethyl Carbonate DMC 1.064 [54] 5.0 [54] 90 [54] 3.1 [17] 1 Monoglyme G1 0.859-0.864 -69 84.5-85.2 7.3 [55] 2 Diglyme G2 0.938-0.939 (-70)–(-64) 162 7.2 [55] 3 Triglyme G3 0.980-0.986 (-45)–(-40) 216-220 7.5 [56] 4 Tetraglyme G4 0.861-0.868 (-30)–(-29) 275 7.7 [56] Table 2.2: Chain length, common name, acronym, and physical properties of the first four glymes in the series. Data for EC and DMC are provided as a reference. Data adapted from [9] unless other reference explicitly stated. 16 2. Batteries and SIBs The glymes have, partly because of their stable solvent properties and their ability of chela- tion, been found useful in a range of different applications, both in industry and academia, as summarised by Tang and Zhao [9]. Main fields in academia include electrochemistry, organic synthesis, bio-catalysis, in material synthesis and for use in chromatography and NMR-techniques. Industrial applications involve uses as solvents in inks, components of refrigeration fluids, production of Active Pharmaceutical Ingredients (API). A number of industrial uses of G1-G4 are listed in Table A.2 in Appendix A. One noteworthy applica- tion to highlight is the use of glymes in products dedicated towards etching of fluorinated polymers–relevant to the implications of using fluorinated binders in an electrode of a battery featuring a glyme-based electrolyte. In the subsections below, first the metal complexing properties of the glymes will be presented. Then, the main areas of usage of glymes in electrochemistry in battery research will be given a brief treatment. Finally, the findings of recent years’, regarding the toxicity of glymes will be highlighted to point out possible difficulty and moral implications of using glymes in commercial application. 2.5.1 Metal Complexing Properties of Glymes The G1 to G4 glymes all have ether oxygens interconnected by flexible alkoxy chains. Sol- vation of cations can thus occur by pairing the polar oxygens to the ion, rendering effective complexing agents [9]. For example, the coordination number diglyme with lithium is two, while the total coordination number of lithium is four, meaning that two of the oxygens of two molecules will coordinate with the cation [9]. In the case of diglyme, Jache et al. sug- gested that the same was true for sodium and that two diglyme molecules tightly embrace the Na-ions within the solvent [52]. Computational studies have indicated that the chela- tion mechanisms lead to a large amount of free ions in the solution, lowering the amount of ion pairing [9]. One explanation provided for the effective complex formation was that the loss in entropy associated with forming a glyme-ion complex is small [9]. Other ab- initio calculations have indicated that the high chain flexibility of the glymes allows many metastable structures for the metal-glyme complex featuring different geometries [57]. 2.5.2 Electrochemistry of Glymes The glymes’ complexing properties have lead them to be used in a range of electrochemistry applications. Xia et al. showed the importance of the phase transfer catalysis role of the glymes when used in Na-Air battery concepts [58]. Here, the glymes play both the role of ion conduction and of facilitating NaO2 super-oxide growth because of their phase-transfer properties. For conventional battery chemistries, Aurbach and Granot pointed out how glyme-based solvents were unsuitable for use with Li-metal anodes. This was because of the growth of a rough morphology, resembling dendrite formation, occurring during dissolution-deposition. They also suggested an anodic electrochemical instability [59]. The resulting anodic decomposition was mainly observed for ethyl glyme (R = CH2CH3, n=1) and it was concluded that in spite of decomposition, the stability was much better for the glymes than for comparable cyclic ethers, esters or carbonates. Cui et al. conducted a smaller study where the authors claimed to observe anodic ether reduction products in XPS-spectra of cycled surfaces. The reduction was proposed to yield a very thin layer of ROCH2−Na species on top of a layer of inorganic NaF and NaO2 species. They furthermore demonstrated how a diglyme-based electrolyte leads to a very efficient plating-stripping behaviour which they attributed to the stability and mechanical properties of the aforementioned film [60]. In two subsequent studies Jache et al. showed how graphite could be activated as an anode material when used in electrolytes based on 17 2. Batteries and SIBs mono-, di- and tetraglyme [61, 52]. They attributed this to the co-intercalation, readily supported by XRD data. The co-intercalation seems to allow the Na to form so-called ternary graphite-intercalation compounds (t-GIC) featuring a site in the active material, a sodium atom and a solvent molecule supporting the intercalation. Comparing the results to similar experiments carried out with tetraglyme, they did not see any co-intercalation, suggesting that the coordination of triglyme with Na ions was unsuitable t-GICs to form. Furthermore, the study showed how the t-GIC were supposedly very stable and associated with a high coulombic efficiency and good capacity retention. Common for all of the studies using diglyme, is that they ascribe stable reductive behaviour to the solvent molecules or the films the solvents form upon reduction. No studies found in this literature review have focused solely on determining the oxidative and reductive limits of the G1-G4 molecules when used as a components in electrochemical systems. However, the general belief seems to be that they are stable against reduction when used with alkali metals or at least with sodium, either because a good SEI is formed or because no reduction takes place at all. 2.5.3 Toxicity and Cost of Glymes The reason for deciding if a specific electrolyte is suitable for use in a future battery systems should not rely solely in its electrochemical performance. Two other notable factors are cost and toxicity. In Table 2.3 the costs of diglyme, EC and DMC are presented. The costs are based on the vendor price for purchasing solvent at 99.5 % purity in the larges possible volume [62]. Where volume based prices have been converted to weight based ones, the lowest value for density tabulated in 2.2 has been used. Chemical name Cost (EUR kg−1) EC 51.5 DMC 104 Diglyme (G2) 303 Table 2.3: Cost of diglyme per unit weight as compared to EC and DMC. Regarding the toxicity, a few studies have been conducted on the glymes–much due to the rise of the solvents’ use during the advent of semi-conductor and IT-industries [9]. With its base in animal trials on rats the toxicity of some members of the glyme family have been evaluated and concluded to show a ”moderate acute toxicity ” [9]. The LD50 dose for all the glymes listed in Table 2.2 is more than 2900 mg kg−1, with diglyme showing a LD50 dose of 3779 mg kg−1. Concerns regarding glymes have instead been directed towards the long- term reproductive effects. McGregor et al. reported the increase in sperm abnormalities detected in rats exposed to 250 - 1000 ppm concentrations of diglyme. Furthermore, doses of 684 mg kg−1 suggested a pathological change in the testicles of exposed rats. In 2011, the U.S Environmental protection agency announced that three glymes pose a high concern to the public because of these indications of reproductive toxicity [9]. In relation to this, the safety data sheet of the glymes in Table 2.2 , carry the phrases ”May impair fertility” and ”May harm the unborn child”. When it comes to impact on the natural environment, the same substances are mentioned as benign, with low bio- accumulation factors [9]. Searching for the Mono-Tetraglymes in the European REACH database, screening tests indicates inherent biodegradability in water for di-, tri- and tetraglyme, but not for monoglyme [63]. 18 3 Description of Experiments Battery research spans a whole ensemble of measurement techniques and instruments. This project has focused on a few of those methods, for which the underlying theory is described in the first section of this chapter. The second section presents some general considerations and methodology used and also motivates the choice of electrode materials used. The third section presents experimental meth- ods carried out to investigate the suitability of Na metal as a reference electrode in a diglyme-based electrolyte. Section four presents the techniques used for viscosity and ionic conductivity measurements. This is followed by section five, where experiments for stability of a future SIB using diglyme are presented. The sixth and final section presents experiments targeting identification of possible degradation mechanisms. 19 3. Description of Experiments 3.1 Theory of Main Methods Employed The battery field, being multidisciplinary features the use of many different techniques. Only for studying SEI properties Pallavi Verma et al. listed no less than 22 different techniques [26]. In this section, the theory behind the methods mainly used in this project are elaborated on. Other techniques not used to a large extent or not being the main techniques for myself, are not treated herein. Some examples of such techniques with data nevertheless reported in this thesis are: • Rolling ball viscosimetry to obtain electrolyte viscosity. • SEM to study surface microstructure of electrodes. • EDX to analyse elemental composition of cycled electrode surfaces. • EIS to measure electrolyte ionic conductivity. 3.1.1 Galvanostatic Cycling with Potential Limitation-GCPL One common method to test electrode materials and cell setups is through galvanostatic cycling. By connecting the cells to a potentiostat that can track and regulate the current and voltage applied to the cell very precisely, arbitrary working conditions for a cell can be simulated. Experiments are often run in either three or two electrode setups. In the prior case, two of the electrodes–denoted the counter- and working electrode–are constantly referenced towards a third, reference electrode (Figure 3.1). With such a setup, the electrochemical processes taking place at the anode and cathode in a battery setup can be properly sepa- rated. In a two electrode setup, one of the electrodes will instead serve as both reference and counter electrode. In this case, the proper working electrode voltage can only be controlled if the electrochemical potential is constant at the counter/reference electrode, which is sometimes the case when using alkali metal counter electrodes. Figure 3.1: Principle of a three electrode setup. A current is measured between the counter and working electrodes. Voltages are, instead measured against a third, reference electrode. Image adopted with permission from [6]. In galvanostatic cycling, the current is set to some finite constant value by the user. Mostly, a potential limit is also set for both charge and discharge of the cell. The setup tested can 20 3. Description of Experiments then be cycled between the upper and lower voltage cut-off limit for any number of times at the programmed current, while the total charge that can be harnessed upon charge and discharge is tracked along with the voltage at accumulated value of charge extracted per charge/discharge cycle. It is common practice to define the amount of current drawn in terms of C-rates, express- ing the number of times the theoretical capacity of the active material is passed in one hour. For example, a rate of C/10 equals a current that will extract/insert the complete theoretical capacity from the electrode material in 10 hours. A rate of 10C instead indi- cates that all theoretical charge is extracted/inserted 10 times in one tenth of an hour. Since the C rate is dependent on the total amount of theoretical charge available, the current drawn during an experiment will vary with both the type and amount of active material used. 3.1.2 Cyclic Voltammetry–CV In cyclic voltammetry (CV), the voltage applied to the cell, instead of the current, is controlled and the current is recorded. When interpreting cyclic voltammetry data, the current, I, detected at each voltage, V, is plotted in a so called voltammogram. Each peak in the voltammogram will correspond to a specific electrochemical process taking place [17]. 3.1.3 Electrical Impedance spectroscopy–EIS The different mechanisms and regions that are present in any electrochemical system can be modelled as an equivalent circuit featuring a set of resistive and capacitive elements. As with every equivalent circuit featuring a network of passive components, it can be reduced to a single, frequency dependent, equivalent impedance. For example, the multilayered structure of the SEI described in Section 2.3 will contribute with a certain phase shift and damping of any time dependent electrical signal. So will the transfer of ions in the bulk of the electrolyte and the transfer of electron in the bulk of the electrode. By applying a small amplitude sinusoidal voltage at varying frequencies and tracking the current response, the equivalent impedance of a system can be probed at different frequencies and plotted on a Bode or Nyqvist form. From the shape of the Nyqvist curve, it is later possible to extract parameters for previously designed models, giving information about system kinetics and charge transfer [64]. However, it is also possible to use the techniques to simply detect any change taking place in the SEI structure or bulk of the electrode, as this would change the EIS spectrum. 3.1.4 Gas Chromatography, Mass spectrometry–GC/MS The principle of coupled gas chromatography/mass spectrometry setup is illustrated in Figure 3.2. For the gas chromatography (GC) part, an analyte in gas phase is injected into the column where a steady stream of inert gas is flowing. The gas flow then separates the different components of the analyte based on their tendency to be retained inside the column. Retention will in turn depend on the amount of interaction the component experiences with the column walls, which are often coated with some kind of polymer to improve separation. In case the analyte was a liquid, it would first have to be heated for evaporation, thus limiting the detection to that of substances able to evaporate. 21 3. Description of Experiments Figure 3.2: Principle of a GC/MS setup. After the GC-column, the components separated will reach a mass spectrometer (MS) at different times. The mass spectrometer works by first ionising the analyte component and then separating the resulting fragments by their mass/charge ratio in the analyser. Each substance emerging from the gas column produces a cascade of ionised fragments serving as a fingerprint for the substance. In this way the contents of the analyte can be determined one-by-one [65, 66]. 3.1.5 FTIR Spectroscopy Fourier Transform Infrared Spectrosocpy (FTIR) involves illuminating a sample with broadband infrared radiation and recording reflected or scattered light. Because some frequencies of the incoming IR-radiation will match vibrational transitions in the sample, the response will be a modulated signal containing information about the frequencies that were absorbed. Illuminating a sample at an angle and taking advantage of the IR radiation being heavily absorbed in the bulk of the material sample studied, the technique can be used to probe surface layers. SEIs from cycled surfaces in batteries could thus be analysed if they remain intact [67]. 3.2 Experimental Techniques–General Considerations Laying the foundation for the description of the further experiments described in this chapter, the following section aims to summarise all those methods that were employed repeatedly throughout the project. Such experimental considerations include the choice of electrode materials, the way electrodes were fabricated, how electrolytes were prepared and how cells were assembled. If not noted otherwise, the methods and choices of materials outlined here will be assumed in the following sections. 3.2.1 Choice of Electrode Materials Even though the main focus has been to study degradation mechanism and performance of employing diglyme as an electrolyte solvent, it is impossible to overlook the role of the electrodes in any battery system. A future SIB could comprise many different electrode materials. However not all are as well documented and practical for use in large batches, as is desired when testing electrolytes. Furthermore, not all electrodes are stable and undergo irreversible phase transformations upon cycling. One SIB candidate can be built using a NVPF cathode and a HC anode. These materials have shown stable cycling behaviour and high specific capacities. Combining the electrodes the theoretical energy density becomes 22 3. Description of Experiments 284 Wh kg−1, when only considering the active materials. Another stable material that could be used is NVP, which can act as both an anode and a cathode. NVPF, NVP and HC were chosen for this work, because they have a previously documented stable cycling performance. Also they have well documented capacity/voltage profiles, with major parts of the profiles being distinct voltage plateaus, mitigating identification of anomalies in a system. A third reason for using NVPF and HC, was that the materials were accessible as large batches of tape cast products–improving reproducibility in experiments. 3.2.2 Tape-Cast Electrodes Electrodes of HC and NVPF were manufactured using a pre-fabricated, tape cast product (CEA, Grenoble), on two sides of a aluminium current-collector. To make electrodes, one side of the collector tape was scraped using small amounts of ethanol and a scalpel (Figure 3.3). Figure 3.3: Sequence of electrode fabrication procedure featuring, a), tape cast precursor on both sides of Al current collector, b), scraped precursor and, c), punched electrodes with a set of current-collectors prepared for weighting The scraped tape was left to dry over night in room temperature. After drying, electrodes of desired size were punched from the electrode tape using a punching-tool. Also, 10 pieces of aluminium foil of the same thickness, size and composition as that used for the tape electrode were punched using the same size of punching-tool. Foil pieces were produced to be able to estimate the average weight of a electrode current-collector and make an estimate of the error associated with spread in the weight distribution of the current collectors. The resulting characteristics of two sizes of electrodes used are shown in Table 3.1. After punching, each electrode was weighed and labelled, and then dried in a Büchi-oven for 24 hours at 80 ◦ C and brought into the glovebox, while remaining under vacuum in the Büchi tube. Electrode area [cm2] Mean weight of Al- current collector [mg] C.C. weight standard deviation [mg] 0.62 3.57 0.073 0.95 5.19 0.040 Table 3.1: Electrode area, mean weight of current collector and standard deviation of current collector (C.C.) weight for two different sizes of electrodes used. 23 3. Description of Experiments 3.2.3 Powdered Electrodes In a handful of measurements it was not practically possible to use the tape cast electrodes. Here the experiments had to fall back on powdered electrodes of either NVPF or NVP. For the powders, the pristine active material was ball milled for 15 minutes with Super Carbon to form a 80:20 mixture by weight. This was done to ensure proper mixing and good conduction among the different active material grains. An important feature of powdered electrodes is that they do not contain any binder, making the system less complex in terms of components, but also less stable in terms of mechanical rigidity. Less rigidity makes them more susceptible to disconnection of active material grains. 3.2.4 Electrolyte Preparation The main component investigated – the 1 M NaPF6 in diglyme electrolyte– for use in a future full cell system was prepared by mixing diglyme (Sigma-Aldrich, Anhydrous, 99.5 % purity) and NaPF6 salt (Stella-Chemifa). Solutions were prepared with molarities of 1 M, 0.75 M or 0.5M, removing any residual water using molecular sieves. Water content was verified to be less than 20 ppm using Karl-Fischer titration. After half of the project the Karl-Fischer apparatus broke down. Subsequent preparations of electrolytes were instead left to dry according to previous know how, drying for at least 7 days with the molecular sieves before usage, to be completely sure the water content was low. In experiments where other solvents were used for bench-marking, the same procedure as outlined above was employed. 3.2.5 Cell Assembly For the experiments where performance and degradation of systems using a diglyme based electrolyte was studied, cells were built according to two layouts - coin cells and Swagelok cells. Because of the air-sensitive nature of the SIB-system and salt, all cells were assembled in a inert Ar-atmosphere inside a glovebox (MBRAUNMB200-MOD) with the surrounding oxygen and water content being less than 10 ppm. In this section follows the assembly description for the two types of cells used. 3.2.6 Coin Cells Coin cells made out of stainless steel were assembled according to Figure 3.4. Electrodes were placed on both sides of a boro-silicate glass-fiber separator (Whatman GF/D). Elec- trode size was varied due to the specific balancing required for each setup. In case of a NVPF positive electrode, a piece of Al-foil was cut and placed to completely cover the the cathode part of the casing to avoid corrosion of the stainless steel, which had been indicated by earlier tests. In the case of a Na-metal anode, the stainless steel disk in the anode side was coated with a thin layer of Na that was rolled out with a test-tube that was separated from the Na by a Mylar®, ensuring that the metal covered the whole electrode and that the surface was as flat as possible. Surface flatness was desired to not introduce any roughness-related effects in the system that would not be reproducible. After putting the electrodes and foil in place, 14 drops of the desired electrolyte were added to the separator using a Pasteur pipette to ensure proper soaking of the separator. The cell was sealed through clamping and then brought to testing within 30 minutes after assembly. 3.2.7 Swagelok Cells Swagelok cells were assembled according to the illustration 3.5. All cell components, apart from electrodes and separator, were first washed and scrubbed with a Scotchbrite sponge, water and dish-washing liquid. Then the plungers were polished with a grade 1000 24 3. Description of Experiments Figure 3.4: Schematic of Coin cell assembly sandpaper to remove any surface roughness and surface contaminants (such as surface oxides). Finally the components were sonicated in ethanol and then in acetone for 15 minutes each. After sonication the components were dried in a oven at 55 ◦ C for at least 3 hours. The cell components were then pre-assembled as far as was possible for the cell type to be built. In the case of EC-DMC being used as electrolyte a PTFE tape was used to insulate the plungers from the cell body. In case of diglyme being used as an electrolyte solvent, instead a Mylar® foil was used and cut to precisely cover the inner part of the cell body. Mylar was used because of earlier experiments indicating that PTFE was degraded upon cycling in Swagelok cells. When it turned out that the Mylar® foil might suffer from a similar problem, two other insulation materials were tested. One was Kapton® film. The other was a Poly-amide 6,12 laminate, commonly employed in food-packaging industry, which was provided by AlfaPac AB, Sweden. These materials were tested to finally fall back to the coin-cell concept when cycling cells. The resulting polymer films after cycling experiments in Swagelok cells are presented in Appendix F. Figure 3.5: Schematic of Swagelock cell assembly 25 3. Description of Experiments 3.3 Electrolyte Stability with Na-metal To be able to proceed with any further testing of full cell systems employing 1 M NaPF6 in diglyme, the initial part of the project focused on determining the stability of pure Na electrodes. Earlier test in EC:DMC based electrolytes [68] had shown that stability of the alkali metal electrodes is not necessarily the case and that continuous reduction of the electrolyte seems to be occurring at a Na metal electrode. Since such a process would interfere with further measurements, reactivity of Na electrodes was determined by studying gas evolution, visual inspection and by impedance spectroscopy of symmetric Na|Na cells. 3.3.1 Investigating Film Growth on Na Surfaces To judge whether a presumed electrolyte decomposition resulted in a growing resistive SEI on the Na, an EIS study was made in Swagelok cell using symmetric Na|Na electrodes and Mylar film for insulation. The spectroscopy measurement was preformed using BIO-logic VMP-3 potentiostat recording the impedance each 6 hour for 96 hours applying a 10 mV sinusoidal voltage at frequencies ranging from 0.02 Hz to 200 kHz with 10 points sampled per frequency decade. In each 6-hour interval the cell was left at OCV to let the growth of any SEI continue. 3.3.2 Investigating Gas Release from Electrolyte Decomposition To study if a possible decomposition of diglyme electrolyte on Na metal surfaces resulted in any continuous gas release, a pressure-cell featuring a Swagelok body, a column and a pressure sensor was assembled according to Figure 3.6. On the bottom part of the cell a 5.52 mg disk with radius 4 mm of Na metal was coated. 250 µL of 1 M NaPF6 in diglyme electrolyte was added to the cell, whereupon the cell was sealed directly. The pressure evolution in the cell was tracked for more than 9 days. 3.3.3 Storage Experiment A final experiment was carried out to detect any film formation arising from reduction of diglyme on a Na metal surface. This was done by submerging a sphere of Na metal in a vial of the 1 M NaPF6 diglyme electrolyte. The colour change of the disk was then recorded over 30 days. 3.4 Ionic Conductivity and Viscosity An electrolyte suitable for a future industry application needs to have a high ionic conduc- tivity, since this property impacts the potential charging rate of a battery. Furthermore, the temperature dependence of such a property would also be important as a future bat- tery would be assumed to operate at a range of different temperatures without losing too much performance. Aiming to further characterise if the diglyme electrolyte had practical use in a future SIB, the ionic conductivity was measured at different concentrations of NaPF6. A further reason to do so was to understand if the special, chelating, properties of the solvent could have any immediate impact on the ionic conduction in the electrolyte and if the Walden rule was obeyed. The ionic conductivity was tested for 0.5M, 0.75M and 1 M solutions of NaPF6 using a Biologic, MCS 10 temperature controlled frequency response analyser at temperatures between 10◦C and 80◦ C, sampling at steps of 5 ◦ C. For the 0.75M an 0.5M concentrations, two simultaneous measurements were carried out. For the 1 M solution the amount of electrolyte prepared was not enough, leading to only one measurement being possible for 26 3. Description of Experiments Figure 3.6: Layout of pressure cell used for determining gas evolution from Na in contact with electrolyte. Based on illustration from [7]. the conductivity measurement. Viscosity was measured for the same set of concentrations using a Anton Paar, Lovis 2000 M/ME rolling-ball microviscosimeter with a capillary tube featuring a 1.59 mm diameter. Temperatures were sampled at 5 ◦ C steps between 10◦C and 60◦C. 3.5 Electrolyte Stability Window Guided by the indication of being able to use Na as a stable reference and realising that ionic conductivity was satisfactory, further work was carried out to characterise the oxidative and reductive stability of the electrolyte 1 M NaPF6 in diglyme, thus determining the electrochemical stability window. The oxidative potential was studied using linear sweep and cyclic voltammetry in Na Swagelok half-cells using working electrode of NVPF to maintain the cathode present in future full cells. The cell was then cycled between 3 and 5 V vs Na+/Na◦ at a rate of 0.1 mVs−1 using a BIO-logic VMP3 potentiostat. As an electrolyte, both 1 M NaPF6 in diglyme and a 1 M NaPF6 in EC50DMC50 were tested – the EC50DMC50 being used as a reference to detect any solvent-independent electrochemical processes. The the onset potential was defined to occur when the current density of the cell reached 50 µ A cm−1, the area of the electrode being 0.95 cm2. Such a condition was consistent with previous work on electrocatalysis [69]. After determining the oxidation potential at the NVPF electrode, an equal setup was constructed to cycle a NVPF half- cell 5 times below the oxidative stability limit. The cycling was done to further study the insertion/extraction characteristics of the electrode at hand. Reductive stability for diglyme electrolyte at the anode could not be tested with the same demand for realistic condition, since hard carbon shows insertion of Na starting for 27 3. Description of Experiments relatively high potentials continuing almost to 0V vs Na+/Na◦, thus masking any reduction of the electrolyte. Instead, the reductive potential was studied using a Cu-foil working electrode, since previous studies had indicated excellent plating/stripping efficiencies on such foils when using diglyme as a solvent [60]. The foil had previously been roughened with a grade 300 sandpaper, to increase the available surface area. The cells were cycled to 0V vs Na+/Na◦ at a rate of 0.1 mVs−1 using a BIO-logic VMP3 potentiostat and any reductive peaks were recorded in the voltammogram. The same measurement was later repeated using a EC50:DMC50 1 M NaPF6 electrolyte for comparison. The results acquired from the reduction experiment were later compared to calculations of reduction potential made by Piotr Jankowski at Chalmers University of Technology. Here density functional theory calculations werre used, including an implicit solvation by C-PCM for diglyme in solution at different levels of complexation with an Na+ cation. 3.6 Galvanostatic Cycling Experiments Noting that Na-metal electrodes seemed stable in the electrolyte and that the ESW of the 1 M NaPF6 in diglyme electrolyte seemed broad, further steps were taken to characterise how individual HC, NVPF and NVP electrodes performed in full and half cells. The reason doing so was that any electrolyte for future commercial cells, a stable cycling behaviour must first be verified. In this section, various cycling experiments conducted to characterise the considered electrolyte are presented. All capacity/voltage profiles for the described experiments were acquired at C/10 using a BIO-Logic VMP3 potentiostat at room temperature. The C rate was calculated based on the extraction of 2 Na from NVPF, making rates consistent between half cells and full cells. For NVP, instead, the C rate was based on the mass of anode. 3.6.1 Half-cell Capacity Tests of Diglyme Electrolytes To study capacity fading and performance of the two different electrodes in a diglyme electrolyte, half-cell galvanostatic cycling was performed in coin cells using Na-metal as an anode. The cells were cycled 15 times using the cut-off potentials given in table 3.2 and then stopped based on the stable capacity retention seen. Based on the results from the half-cell, a full cell reconstruction was done by subtracting the half-cell voltages corresponding to when the electrodes at each level of extraction/insertion of ionic charge carriers. Working electrode Lower cut-off potential (V vs Na+/Na◦) Upper cut-off potential (V vs Na+/Na◦) HC 0 2 NVPF 2 4.3 Table 3.2: Cut-off voltages for cycling of materials in half-cells, using Na metal as counter electrode When doing the reconstructions, further presented in appendix C, irreversibility was not accounted for, passing from one cycle to the next, with the purpose of saving time. Instead the reconstructions from only the first cycle were used to approximate the potential vs. Na of each electrode in a full cell. 3.6.2 Full Cell Capacity Tests of Diglyme Electrolyte - Galvanostatic Cycling Recording a good cycling behaviour in half-cells, the focus was shifted towards full cells of HC/NVPF–the type of setup which would have a chance of being commercialised. 28 3. Description of Experiments Cycling was done as in using 1 M NaPF6 in diglyme in coin cell, using one HC anode and one NVPF cathode and employing the cut-off scheme in table 3.3. In the full cells, the electrodes were not completely balanced in capacity because of difficulties doing so with tape cast electrodes. Instead, the capacity ratio between carbon an NVPF was approximately NVPF:HC 1.0:1.1 based on charge capacity of the first charge in half-cells. Since HC capacity was in excess, the mass of active NVPF was used to set the C-rate. Electrode setup Lower cut-off potential (Vcell) Upper cut-off potential (Vcell) NVPF/HC 2 4.3 Table 3.3: Cut-off voltages for cycling of materials in full cells Swagelok cells were also built to test full cell behaviour, but because of problems with stability and reproducibility, this concept was abandoned for the more stable coin-cell setup. 3.6.3 Cycling Different Voltage-regions of the Active Materials Wanting to understand the voltage dependence of the capacity fade observed in full cells the different potential-regions of the electrode materials were explored by changing the balancing of the electrodes and cycling them in coin-cells. The different schemes employed are listed in 3.4. Here, also the regular full cell is included as a reference. The electrode masses listed are typical for one repetition of the experiment. Were the experiments to be repeated, the exact mass of one electrode would most likely be different, forcing the mass of the other electrode to be changed in order to maintain the balancing. Region cycled Lower cut- off voltage [Vcell] Upper cut- off voltage [Vcell] Cathode active ma- terial mass [mg] Anode active ma- terial mass [mg] Comment NVPF plateaus, HC slope and plateau 2.0 4.3 9.50 4.89 - Upper plateau of NVPF only, HC Slope and Plateau 3.8 4.3 9.64 4.68 Mostly HC plateau is used, NVPF com- pletely discharged on first cycle to reach second plateau Lower plateau of NVPF only, HC slope and plateau 2 3.8 12.85 3.28 - Both NVPF plateaus, HC slope only 2.0 4.3 6.259 4.74 - Both NVPF plateaus, HC slope only 2.0 4.0 6.259 4.74 Upper cut-off volt- age in previous scheme was lowered after 16 cycles Table 3.4: Regions cycled and their approximate voltage intervals for the anode and cathode as approximated by half-cell reconstructions. masses are values used for one experiment. 3.6.4 Cycling of NVP/NVP Cells Further investigating the potential dependence of the capacity loss seen in full cells, a NVP/NVP coin cell was assembled using NVP powder electrodes as both anode and cathode. The reason for doing so was to be able to explore plateaus featuring yet another set of voltage plateaus to decide whether the diglyme 1 M NaPF6 was stable at those, 29 3. Description of Experiments well-defined, potentials. The cycling was performed in coin-cells with a scheme that can be seen in Table 3.5. Worth noting is that the upper cut-off was lowered after 10 cycles because it was feared that some of the conductive carbon sp was also used to intercalate Na at the end of charge, when the anode reaches its lowest potential. By lowering the cut-off the risk of doing so could be mitigated. Region cycled Lower cut- off voltage [Vcell] Upper cut- off voltage [Vcell] Cathode active ma- terial mass [mg] Anode active ma- terial mass [mg] Comment NVP upper plateau, NVP lower plateau 0.0 2.8 3.23 7.81 mg Capacity for the two plateaus of NVP are not symmetrical NVP upper plateau, NVP lower plateau 0.0 2.0 3.23 7.81 mg Same setup as previous row, upper cutoff low- ered after 10 cycles. Table 3.5: Cycling scheme for symmetrical NVP/NVP cell 3.7 Physico-chemical Characterisation of Capacity Fade The final part of the project aimed towards characterising cycled cell components and to further investigate the capacity fade seen in full and half cells. In this section methods used in trying to do so are presented. The techniques employed vary from quite complicated (GC/MS) to simple storage experiments to detect colour changes. 3.7.1 GC/MS of Cycled Cell Separators Observing a large initial irreversibility from hard carbon half-cells, an attempt was made to identify possible decomposition products using coupled liquid injection-GC/MS and the chromatograms to reduction schemes of the electrolyte. The separators from full cells that had been cycled 50 times and half-cells that had been cycled 10 times were recovered and soaked in 1 mL of dry acetonitrile (H2O<0.001%). The separators were then removed from the soaking liquid and discarded. The remaining blend of acetonitrile and soluble compounds from the separator was diluted 100 times before being injected into the GS/MS setup, consisting of a trace 1300 series GC ultra-gas chromatograph coupled to an ISQ mass spectrometer. The chromatographic separation was performed by using a “BPX70” cyanopropylpolysilene-siloxane- based capillary column (30 m 0.25 mm i.d., 0.25 mm) from SGE. 3.7.2 SEM and EDX Analysis of Cycled Anodes Complementing the GS/MS analysis to detect any solid degradation products in HC half cells using 1 M in NaPF6 diglyme, SEM micrographs of cycled cells were recorded using a Hitachi S-3400N Microscope. Before imaging, HC anodes from full cells had been cycled 1.5, 2 and 50 times in 1 M NaPF6 diglyme electrolyte using the same cut-off voltages as in Section 3.6.2. Because of limitations in the movement of the stage of the microscope, the samples had to be transferred in air before being mounted inside the microscope. While inside the microscope, an EDX analysis was carried out on the samples to identify any elemental contamination. After detecting vanadium in the EDX spectra it was decided to investigate if the metal came from soluble species forming at the cathode and if it was only deposited during the first cycle. This was done by cycling a NVPF cathode in 30 3. Description of Experiments half cell and HC anode in half cell according to the scheme in table 3.6. The half cells were then disassembled and the electrodes recovered. The NVPF was then washed in fresh electrolyte, with the purpose of removing any soluble species. The washed NVPF electrode and the extracted HC electrode were then used to assemble a new coin cell which was cycled 10 times according to the scheme on the third line of table 3.6. After cycling the full cell was disassembled and the anode analysed using EDX. Electrode setup Lower cut-off po- tential Upper cut-off po- tential Cycles Comment NVPF/Na 3.5 V vs Na+/Na◦ 4.3 V vs Na+/Na◦ 1 HC/Na 0 V vs Na+/Na◦ 1.4 V vs Na+/Na◦ 1 NVPF/HC 2 (Vcell) 4.3 (Vcell) 10 Assembled using the active materi- als from previous 2 rows after wash- ing the NVPF. Table 3.6: Cut-off voltages and cycling scheme for further exploring possible vanadium extraction 3.7.3 FTIR-Spectroscopy of Cycled Cells To further determine if there was any SEI film formation on the surface of cycled hard carbon anode, FTIR was used to study electrodes cycled 2 and 1.5 times in full cells using the same cut-off voltages as in Section 3.6.2. The reason for not examining electrodes cycled more times was that any film formation was thought to happen mainly at the first and second cycles, since these cycles were associated with the largest irreversibilities. For the FTIR analysis, cycled coin cells were disassembled under argon glove-box conditions and the carbon anodes allowed to dry in the glove-box atmosphere for more than 24 hours. The electrodes were then transported in sealed coffee-bags to a nitrogen-filled