Dual electrolyte design for lithium-ion batteries Extended electrochemical stability window by immiscible electrolytes Master’s thesis in Physics SOFIA REINER DEPARTMENT OF PHYSICS CHALMERS UNIVERSITY OF TECHNOLOGY Gothenburg, Sweden 2023 www.chalmers.se www.chalmers.se Master’s thesis 2023 Dual electrolyte design for lithium-ion batteries Extended electrochemical stability window by immiscible electrolytes SOFIA REINER Department of Physics Division of Material Physics Chalmers University of Technology Gothenburg, Sweden 2023 Dual electrolyte design for lithium-ion batteries Extended electrochemical stability window by immiscible electrolytes SOFIA REINER © SOFIA REINER, 2023. Supervisor: Dr. Clément Pechberty, Department of Physics Examiner: Prof. Patrik Johansson, Department of Physics Master’s Thesis 2023 Department of Physics Division of Material Physics Chalmers University of Technology SE-412 96 Gothenburg Telephone +46 31 772 1000 Cover: Schematic of a LNMO/graphite intercalating lithium-ion battery with two immiscible electrolytes. Typeset in LATEX Printed by Chalmers Reproservice Gothenburg, Sweden 2023 iv Dual electrolyte design for lithium-ion batteries Extended electrochemical stability window by immiscible electrolytes SOFIA REINER Department of Physics Chalmers University of Technology Abstract The lithium-ion battery is the fastest growing energy storage technology, and dom- inates in demand for smaller portable applications and electric vehicles due to its high energy density, small weight, and long lifespan. New developments in the ma- terials used as positive electrodes enable the battery cells to operate under higher voltages than their predecessors, increasing the maximal energy contained. How- ever, the voltage range of these cells extends beyond the reduction and oxidation limits of conventional electrolyte solvents, causing degradation and subsequent fail- ure. In this work, a dual electrolyte concept is proposed and evaluated, in which each electrode has its own catered electrolyte, reducing the voltage window each electrolyte must be electrochemically stable in. Pairs of immiscible electrolytes con- sisting of both organic solvents and room temperature ionic liquids were evaluated and the diffusion between them in a cell setting was studied by Raman confocal spectroscopy, showing little diffusion of higher viscosity ionic liquids. Furthermore, the electrochemical properties of the electrolyte pairs were tested by galvanostatic cycling, using lithium manganese nickel oxide (LNMO) and graphite as the cathode and anode respectively. An interface between two immiscible electrolytes, consti- tuted by an ionic liquid and an organosulfur compound on one side, and an ether and a fluorinated carbonate ester on the other, was shown to allow for adequate mobility of ions in order to cycle a cell. Keywords: lithium-ion batteries, high voltage electrodes, LNMO, dual electrolytes, ionic liquids v Acknowledgements First and foremost I would like to thank my supervisor Dr. Clément Pechberty for his guidance and patience in the lab. It has been a true pleasure working with him and getting to learn the ins and outs of battery research and cell characterisation. I would also like to thank my examinator Prof. Patrik Johansson for giving me the chance to work with his group and for introducing me to the world of batteries. There are many I would like to thank at the Material Physics division who gave me the warmest welcome and support during my thesis work. Sharing the lab, office, and especially fika with all of you wonderful people has been lovely. Thank you for showing me a supporting work environment and for your openness to discuss all my questions, whether it be on how to best approach my Raman measurements, how to navigate life in academia, and on which coffee machine to avoid in the lunch room. I would also like to add a special thank you to Morrow Batteries for graciously supplying the LNMO electrodes used in this work. Finally, I would like to thank my family and friends for their endless support during my time at Chalmers. Sofia Reiner, Gothenburg, June 2023 vii List of Acronyms The acronyms that have been used in this thesis are here listed in alphabetical order. ATR-FTIR Attenuated Total Reflectance Fourier-Transform Infrared Spectroscopy CCD Charged Coupled Device CEI Cathode Electrolyte Interphase DBE Dibutyl Ether DEE Diethyl Ether DMC Dimethyl Carbonate DME Dimethoxyethane DMF Dimethyl Formamide DMSO Dimethyl Sulfoxide EC Ethylene Carbonate Emim 1-Ethyl-3-methylimidazolium ESW Electrochemical Stability Window FEC Fluoroethylene Carbonate FSI Bis(fluorosulfonyl)imide FWHM Full Width at Half Maximum GC Galvanostatic Cycling IL Ionic Liquid Im Imidazolium IRC Irreversible Capacity LFP Lithium Iron Phosphate LIB Lithium-Ion Battery LNMO Lithium Nickel Manganese Oxide LSV Linear Sweep Voltammetry LTO Lithium Titanate PC Polypropylene Carbonate PYR Pyrrolidinium PYR14 1-Butyl-1-methylpyrrolidinium RTIL Room Temperature Ionic Liquids SEI Solid Electrolyte Interphase TFSI Bis(trifluoromethanesulfonyl)imide TMS Sulfolane ix Contents List of Acronyms ix 1 Introduction 1 1.1 Principle of lithium-ion batteries . . . . . . . . . . . . . . . . . . . . . 1 1.2 Electrochemical stability window of electrolytes . . . . . . . . . . . . 3 1.3 Dual electrolyte concepts for graphite/LNMO cells . . . . . . . . . . 3 1.4 Aim and scope of thesis . . . . . . . . . . . . . . . . . . . . . . . . . 4 1.5 Thesis outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 2 Lithium-ion batteries 7 2.1 Redox reactions and electrochemistry of LIBs . . . . . . . . . . . . . 7 2.2 Mass transport and polarisation . . . . . . . . . . . . . . . . . . . . . 10 2.3 Formation of solid electrolyte interphases . . . . . . . . . . . . . . . . 10 2.4 Electrode materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 2.4.1 High voltage LNMO electrodes . . . . . . . . . . . . . . . . . 13 2.4.2 Graphite anodes . . . . . . . . . . . . . . . . . . . . . . . . . 13 2.4.3 Lithium metal half-cell configuration . . . . . . . . . . . . . . 14 2.5 Electrolytes for LIBs . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 2.5.1 Lithium salts . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 2.5.2 Non-aqueous organic solvents . . . . . . . . . . . . . . . . . . 16 2.5.3 Ionic liquids . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 2.5.3.1 Cations . . . . . . . . . . . . . . . . . . . . . . . . . 18 2.5.3.2 Anions . . . . . . . . . . . . . . . . . . . . . . . . . . 19 2.6 Separators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 2.7 Research cell configurations . . . . . . . . . . . . . . . . . . . . . . . 20 2.7.1 Coin cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 2.7.2 Swagelok cells . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 3 Immiscibility and diffusion of electrolytes 23 3.1 Immiscible electrolytes . . . . . . . . . . . . . . . . . . . . . . . . . . 24 3.1.1 Polarity of solvents . . . . . . . . . . . . . . . . . . . . . . . . 24 3.1.2 Immiscible pairs . . . . . . . . . . . . . . . . . . . . . . . . . . 24 3.1.3 Wetting of Celgard and glass fibre separators . . . . . . . . . . 26 3.2 Raman spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 3.2.1 Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 3.2.2 Raman penetration depth in separators . . . . . . . . . . . . . 29 xi Contents 3.3 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 3.3.1 Preparation of the electrolytes and separators . . . . . . . . . 31 3.3.2 In situ Raman cell . . . . . . . . . . . . . . . . . . . . . . . . 31 3.3.3 Instrumental setup, calibration and focusing . . . . . . . . . . 33 3.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 3.4.1 Baseline fit . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 3.4.2 [Emim][TFSI] and [P66614][TFSI] . . . . . . . . . . . . . . . . . 35 3.4.3 [PYR14][TFSI] and DBE . . . . . . . . . . . . . . . . . . . . . 37 3.4.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 4 Dual electrolyte cell characterisation 41 4.1 Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 4.1.1 Voltage response . . . . . . . . . . . . . . . . . . . . . . . . . 41 4.1.2 Characterisation of cells . . . . . . . . . . . . . . . . . . . . . 42 4.2 Cycling of cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 4.2.1 Linear sweep voltammetry . . . . . . . . . . . . . . . . . . . . 43 4.2.2 Galvanostatic cycling . . . . . . . . . . . . . . . . . . . . . . . 44 4.3 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 4.3.1 Preparation of materials and cells . . . . . . . . . . . . . . . . 45 4.3.2 Cycling protocol . . . . . . . . . . . . . . . . . . . . . . . . . 46 4.3.3 Cycling equipment . . . . . . . . . . . . . . . . . . . . . . . . 46 4.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 4.4.1 Example with benchmark materials . . . . . . . . . . . . . . . 47 4.4.2 Single electrolyte cells . . . . . . . . . . . . . . . . . . . . . . 48 4.4.3 Dual electrolytes . . . . . . . . . . . . . . . . . . . . . . . . . 51 4.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 5 Final remarks 55 Bibliography 57 A Additional cell cycling data I xii 1 Introduction The demand for batteries has increased rapidly, with forecasts predicting a yearly in- crease of 25-30% up until 2030 [1][2]. Significant efforts are made in both academia and industry to match the rapid growth of applications needing battery storage. Challenges include both grid applications and smaller appliances as well as vehicles. To enable the shift from fossil fuels to renewable energy sources, there is a need to develop higher energy density, longer-lasting, and more cost-effective battery tech- nologies. Parallel to the development of new designs, new materials are progressively added to the EU critical raw materials list, signifying a high risk that these materials will be hard to come by in the near future. In 2023, nickel joined lithium and cobalt on the list while also being metals that constitute the basis of materials used in bat- teries [3]. Consequently, care needs to be taken when developing new materials and technologies that enable the batteries to have long lifecycles while being sustainable in the long run, both in terms of criticality and environmental impact. The first versions of the lithium-ion battery (LIB) resembling the current battery setup, with a metal oxide cathode and carbon-based anode, were introduced in the early 1990s. In comparison to the already established lead-acid, nickel-cadmium and nickel-metal hydride batteries, the LIBs had higher energy density as well as a long cycle life [4][5]. The high specific energy has made them especially suitable for portable applications which also has helped propel the rapid development of LIB technologies. Today, LIBs dominate in demand in the automobile industry and this industry is expected to dictate in which direction battery research is headed [6]. 1.1 Principle of lithium-ion batteries Batteries enable the storage of chemical energy and consist of two or more electro- chemical cells. The term battery is often used to describe the full layout of assembled cells with additional circuitry or components, while the cell describes the unit where electrochemical reactions take place. A first distinction can be made between bat- teries available in the market. Batteries that can only be discharged once are called primary batteries, while rechargeable batteries are called secondary batteries. A bat- tery or cell can be charged by letting non-spontaneous chemical reactions take place when applying a voltage across the cell. Through spontaneous reactions the stored energy can later be extracted and used for applications connected as a load [5]. A simple electrochemical cell consists of two electrodes, a liquid electrolyte, a separator and two metallic conductive terminals, or collectors, as shown in a simple layout in Fig. 1.1. The chemical reactions that drive all cells are the reduction and oxidation, 1 1. Introduction or redox reactions, of different species. In LIBs, the charging process is driven by the oxidation of lithium stored in one electrode, the migration of free lithium ions through the electrolyte and the subsequent reduction of the ions into lithium at the opposite electrode. During discharge the opposite process takes place, and the lithium is once again returned to the original electrode, a lower energy state for the cell. The electrons freed in the oxidation reactions contribute to a current in the external circuit which can be utilised by a connected application. An electrode will either act as a cathode or anode depending on if the cell is charged or discharged. However by convention, the reducing electrode during discharge will always be called cathode, or alternatively the positive electrode. Figure 1.1: A simplified model of a typical intercalating LIB cell with a graphite anode and a transition metal oxide cathode, where X could for example be nickel, cobalt or manganese. Both the cathode and anode need to be able to store and release lithium atoms. While there are multiple materials that have been found suitable as a cathode/anode pair for LIBs, the state-of-the-art cells consist of transition metal oxides as cathodes and graphite as an anode. They store the lithium atoms in interstitial sites in the atomic structure of the materials. In graphite, the lithium is intercalated between the sheets of graphene while the cathode materials have slightly different structures and interstitial sites depending on the chemical composition [5]. To ensure sepa- ration of ions and electrons and avoid unintentional cell discharge, the electrolyte must have high ionic conductivity with close to no electric conductivity. Commer- cial electrolytes are almost exclusively solid lithium salts dissolved in organic-based carbonate solvents [6][7]. However, other non-aqueous liquid electrolytes such as room temperature ionic liquids (RTIL), which are salts that are liquid at tempera- tures below 100◦ C, solid ceramics or semi-solid gel polymers are being developed and tested. The separator holding the electrolyte is a porous membrane that sepa- rates the electrodes, avoiding short circuiting. It is most often made of electrically 2 1. Introduction insulating materials such as polymers or glass fibres. 1.2 Electrochemical stability window of electrolytes To properly intercalate the ions into the atomic structure of the electrodes, a high enough voltage needs to be applied between cathode and anode. Furthermore, this requires the electrolyte to be electrochemically stable against the electrodes of the batteries as well as on its own when placed under the operating voltage range of the battery. Outside of the electrochemical stability window (ESW), the electrolyte will start to oxidise or reduce and to decompose into new products. The decomposition would result in a decrease of lithium-ions taking part in the cycling process, leading to loss of cell capacity. Additionally, the properties of the product may not be desirable and may hinder the mobility of ions, subsequently leading to the failure of the cell. Considerable effort is therefore taken in battery research to develop new electrolyte solutions that extend the ESW or better caters to the operating voltage of the electrodes. As an example, the standard operating voltage of the well-established cathode material lithium iron phosphate LiFePO4 (LFP) is approximately 3.4 V, measured against a pure lithium metal counter electrode (vs. Li/Li+) [8]. LFP is a commer- cially available material for both research and industry, and is especially favoured in the vehicle industry due to it being relatively inexpensive and lacking cobalt, a critical material that also has been connected to unethical mining [6][9]. The typi- cal carbonate-based solvents used in electrolytes such as, ethylene carbonate (EC), propylene carbonate (PC) and dimethyl carbonate (DMC) have an oxidation limit >4.3 V vs. Li/Li+, which well covers the operation voltage of LFP [10][11]. 1.3 Dual electrolyte concepts for graphite/LNMO cells Since the energy density of a cell is determined by the product of specific capacity and potential across the cell, high voltage positive electrodes are especially sought after and researched in industry and academia [10]. A promising cathode material currently in development is lithium-nickel-manganese-oxide LiNi0.5Mn1.5O4 (LNMO) [6][10]. LNMO has a theoretical operating voltage of 4.7 V vs. Li/Li+, well over the voltage of LFP, but unfortunately outside of the ESW of conventional electrolytes. As of the writing of this thesis, no commercial batteries based on LNMO material exist due to its poor cycling performance. It is however considered to be a part of the next generation of battery technologies [6]. Finding one single electrolyte with a wide enough ESW to accommodate both LNMO and the graphite anode has posed a challenge, leading to new battery con- cepts being investigated. The use of two different electrolytes in the same cell, catered to an electrode each and with an overlapping ESW to encompass the full voltage range of the cell, has been investigated in a few articles. To guarantee sta- bility at each electrode the electrolytes must not mix and only be in contact with its 3 1. Introduction designated electrode. A proposed concept is the usage of an ion conductive mem- brane to avoid mixing but to let lithium ions pass through [12][13], which has been realised for a metal oxide cathode versus a silicon/graphite anode with Nafion as a membrane [12]. Another concept that has been realised for lithium-sulfur batteries is the use of dual electrolytes separated purely by the immiscibility of the liquids, with the goal of both extending the ESW and preventing certain compounds from crossing the cell [14][15]. However, to the best of the author’s knowledge, no article to date has proposed a concept with dual electrolytes separated purely by chemical immiscibility for LIBs. A proposed layout is presented with two separators, each soaked with a different electrolyte (Fig. 1.2). Assuming capillary forces are not enough to keep the elec- trolytes in the porous separators, the proposed layout demands that the cell is kept in a vertical position as well as the highest density electrolyte be on the bottom. Basic requirements for the dual electrolytes include their compatibility and stability with their respective electrode as well as having a good ionic conductivity both in the bulk but also across the electrolyte/electrolyte interface. Figure 1.2: In the dual electrolyte concept two electrolytes would be used to cover the potential window and reach the operating voltage of the LNMO electrode. In this configuration, the electrolyte with the highest density should be placed on the bottom. 1.4 Aim and scope of thesis The aim of this thesis is to investigate a dual electrolyte concept based on im- miscible electrolytes and to explore the feasibility of creating a cell that can be reliably cycled. The diffusion of the electrolytes in the separators will be studied with Raman spectroscopy and the electrochemical properties by assembling pro- 4 1. Introduction totype batteries at the Material Physics department at Chalmers. With the goal being to find an electrolyte solution suitable for high voltage electrodes, the thesis will limit itself to working with LNMO and graphite as the cathode/anode pair. Furthermore, since the research concerning LIB electrolytes is such a vast area on its own and slightly out of the scope of this thesis, the focus will be placed on finding potential immiscible pairs comprised of electrolytes already well documented within LIB research and which are easily accessible. Given the contents of previously pub- lished work, this thesis will only include liquid non-aqueous electrolytes, including but not limited to, lithium salt in organic solvents, highly concentrated electrolytes and room-temperature ionic liquids or a combination of them. Where possible, ex- tra care will be taken to choose sustainable and non-toxic materials and chemicals. Equally important is to choose electrolytes bearing the ecological effects in mind, e.g avoiding chemicals that could be harmful if accidentally released, to the greatest extent possible. 1.5 Thesis outline Chapter 2 of this thesis gives a more thorough theoretical background to the chemi- cal reactions taking place inside a lithium-ion cell as well as introducing the different categories of materials in LIBs and their properties. It concludes with the introduc- tion of different cell configurations and how cells are tested within research. Next in Chapter 3, the immiscibility study of electrolytes is presented, commencing with a literary review of suitable electrolytes. There, the relevant theory of Raman spec- troscopy is also presented, the chosen characterisation technique for investigating diffusion in the cell. The chapter ends with the results from a diffusion study per- formed on the separators and dual electrolyte pairs. Chapter 4 explains the basics of battery cell characterisation and battery cycling before discussing the electrochem- ical performance of the electrolytes in cells. Finally, in a finishing chapter, the dual electrolyte design is evaluated and discussed, while challenges and future outlooks on the design are presented. 5 1. Introduction 6 2 Lithium-ion batteries Initial interest in lithium as a material for battery applications is mainly attributed to its high reducing capability, its small ion/atom size and its low weight. The first batteries containing lithium can be found as far back as the 1910s, while the first cells based on intercalation mechanisms were introduced in the late 1970s. Finally, in the 1990s the modern LIB, with a lithium cobalt oxide cathode and a carbon anode, was released onto the commercial market fuelling the development of small portable devices due to its high energy density and relatively low price [4][16]. However, even though LIBs are increasingly gaining ground, some obstacles still stand in the way of them dominating the market of electric vehicles and grid applications in the same way as for portable electronics. The main issues are cost and availability of lithium, cobalt, and other materials in LIBs and not the theoretical limitations of the material [4][6]. Therefore, research into new cell concepts and optimisation of materials are as relevant as ever. In this chapter, the basic electrochemistry of the LIB cell is presented along with the fundamentals of cell cycling and the underlying kinetics at the interfaces of the cell. The building blocks, the electrodes, electrolyte, and separator are thereafter introduced together with commonly used materials and their strengths and limita- tions. Finally, some experimental cell setups used in research for characterisation of battery materials are presented. 2.1 Redox reactions and electrochemistry of LIBs As previously touched upon in the introduction, the chemical reactions occurring in a LIB cell are the reduction and oxidation of species at the interface between electrolyte and electrodes. In the case of the graphite/LNMO electrodes system, the reactions taking place during discharge at the two electrodes will be: Anode: LiC6 ↔ C6 + Li+ + e− (2.1) Cathode: Ni0.5Mn1.5O + Li+ + e− ↔ LiNi0.5 Mn1.5O (2.2) Each of these reactions takes place between one of the electrodes and the electrolyte and are denoted as half-reactions. Together, these half-reactions give the total re- action of the full cell: LiC6 + Ni0.5Mn1.5 ↔ C6 + LiNi0.5 Mn1.5O (2.3) 7 2. Lithium-ion batteries The direction of this chemical reaction is driven by the system’s tendency to always reduce the Gibbs free energy. This energy is related to the redox potential, or electromotive force, of the full cell E◦ by ∆G = −nFE◦, (2.4) where n is the number of electrons transferred between electrodes in moles and F is Faraday’s constant, which gives the charge per mole. Intuitively, a larger potential difference across the cell will result in a higher thermodynamic driving force. The redox potential is also the difference between the individual reduction and oxidation potentials of the two electrodes measured in standard conditions. These voltages would be measured against a reference, typically the standard hydrogen electrode, in room temperature and with a standard concentration and pressure [17]. In a practical cell however, the conditions are rarely standard so adjustments to the calculations of the cell voltage must be done. When the cell is in equilibrium, meaning no direction is favoured in Eq. 2.3 and the net current is zero across the cell, the voltage can be described by the Nernst equation, E = E◦ − RT nF ln [C6]A [LiNi0.5Mn1.5O4]C [LiC6]A [Ni0.5Mn1.5O4]C (2.5) where the brackets and subscripts indicate the activities of the species at either the anode or cathode, R is the gas constant and T is the temperature [17][18]. The activity of a species is proportional to the concentration of the species but corrects for the fact that particles of the species in a mixture do not behave like in an ideal mixture, meaning the mixture has the behaviour of an ideal gas. The correction is called the activity coefficient and will be different for each of the species inside the brackets in Equation 2.5. The equation makes it possible to approximate the redox potential of an assembled cell by using standard electrode potentials from data sheets and taking the concentrations in place of the activities [18]. A cell reaches its equilibrium state soon after assembly. Initially one direction of the reactions in Eq. 2.1 and 2.2 will be favoured and have a higher reaction rate. This leads to polarisation at the electrode/electrolyte interface due to the accumulation of charged particles involved in the reaction. This polarisation will work opposite to the initial potential difference, favouring the opposite reaction direction until equilibrium is reached. Finally, the stabilisation results in a joint equilibrium potential Ee that will be dependent on electrode material, choice of electrolyte and temperature and that serves as an energy barrier to each of the reactions [18]. What is even more relevant for cells and batteries is the current through the cell when an external voltage is applied, bringing the cell out of equilibrium. The voltage applied would change the energy barriers at the electrode meaning either oxidation or reduction would be favoured over the other. For example, a positive applied voltage to cathode, would raise the Gibbs free energy barrier at the cathode meaning it is energetically more favourable for the system to oxidise LiNi0.5Mn1.5O4 than to reduce Li, leading to a positive net current away from the electrode. A negative applied voltage would naturally favour the opposite reaction and a net current to the electrode [18]. 8 2. Lithium-ion batteries The Butler-Volmer equation describes the relationship between the currents and the applied voltage. It can be derived by applying the law off mass action, where the rate of a reaction is proportional to the activities of the reactants. Assuming that the reaction rate constants are exponentially proportional to the applied voltage: ired = i0 red exp (−αFη RT ) (2.6) iox = −i0 ox exp ( (1 − α)Fη RT ) (2.7) The ired and iox are the currents stemming from the reduction and oxidation re- actions respectively. The voltage η is equal to the difference between the applied voltage and the equilibrium potential E − Ee and is called the overvoltage. When the applied voltage is equal to the equilibrium potential, η = 0 and the currents will be given by i0 red and i0 ox, and since the net current ired + iox should be zero at equilibrium, they must be equal [17]. Finally, the transfer coefficient α is a measure of how large part of the electrode potential will drive the reduction process and tell if one of the reaction directions is hindered more than the other due to obstacles in the cell [18]. The currents according to the Butler-Volmer equation can be seen in Fig. 2.1. As previously mentioned, at η = 0 the net current is zero and the system is in equi- librium as described by the Nernst equation. For small overvoltages the net current will increase linearly with increased voltage, and the cell will show Ohmic behaviour with a fixed resistance across the interface. For higher voltages the behaviour turns exponential which is also called Tafel behaviour [18]. Figure 2.1: The oxidation and reduction current at an electrode/electrolyte inter- face as a function of the overvoltage in accordance with the Butler-Volmer equation. The net current is seen passing through the origin in accordance with the Nernst equation. 9 2. Lithium-ion batteries 2.2 Mass transport and polarisation To ensure fast reaction rates and good battery performance, the transport of charged species to and from the electrodes and reaction sites needs to be efficient. Since the movement of ions is slower than for electrons, mobility of ions is of particular importance [19]. There are multiple mechanisms for mass transport through the electrolyte: convection, electric migration due to a potential gradient, and diffusion due to a concentration gradient. For lithium-ion cells the latter two mechanisms are the most relevant [17][19]. The mobility of ions is decisive in how fast the cell can be charged and discharged. The redox reactions are fast, and if ions cannot be supplied to, or removed from, the interface fast enough, there will eventually be an accumulation of positively charged ions on the oxidising side of the cell and a lack of ions on the reducing side. This will create a polarisation in the cell that could hinder the mobility of ions [17]. The flux caused by the diffusion of ions is described by Fick’s first law of diffusion, q = D δC δx (2.8) where D is the diffusion constant and δC/δx is the concentration gradient. Since current is charge over time per unit surface and the change in concentration over a small distance can be approximate to be linear, the current just before the interface can be approximated by, i = nF DA (CB − CE) ∆ (2.9) Here, A is the surface area of the electrode exposed to the electrolyte, CB and CE the concentration of lithium ions in the bulk and at the electrode respectively and ∆ the width of the layer where a gradient in concentration exists [17]. This ex- pression gives the current supplied to the reaction sites at the electrode until the concentration at the electrode goes to zero and no redox reactions can occur. Natu- rally, to increase the supplied current without causing a destructive polarisation, the diffusion constant needs to be high. Approximating the ions as spherical particles with radius r, the Stokes-Einstein equation gives the inverse relationship between the diffusion coefficient and the viscosity µ: D = kBT 6πµr (2.10) This means low viscosity liquids are more suitable as solvents in electrolytes [20][21]. 2.3 Formation of solid electrolyte interphases When batteries containing alkali metals, such as lithium, were first cycled together with non-aqueous electrolytes, a thin layer was discovered to appear on the anode surface. The layer, which covered the exposed surface, consisted of decomposed electrolyte. Furthermore, the layer prevented further decomposition of the anode, in the same way rust protects metal, while still letting through lithium ions. The 10 2. Lithium-ion batteries passivating surface was named the solid electrolyte interphase (SEI) and was deemed to have a large influence on the performance during cycling [22]. In 1979 a model for the formation of the SEI and the kinetics of the interphase was proposed that further improved the simpler Butler-Volmer equation [23]. LIBs are high voltage cells that use low potential anodes, such as graphite. The SEI layer is a key element in the development of LIBs, since without it the cell would be unstable, and the decomposition of material would eventually lead to cell breakdown. The SEI extends the stability window of the electrolyte, making cycling with graphite and other alkali metals such as pure lithium as anode possible [24]. Similar passivation layer formation can also be found to occur on the cathode side, called the cathode electrolyte interphase (CEI). The CEI also aids in protecting the cathode and electrolyte from decomposition but has not been discussed or researched to the same extent as SEI [25]. However, alongside finding electrolytes with a larger ESW to accommodate high-voltage electrodes such as LNMO, another strategy, while not discussed in this thesis, is to improve the CEI on the cathodes. A more compact and passivating layer on the LNMO surface, while still being permeable to lithium ions, could also ensure the stability of the cell [10]. The composition of the SEI and its microstructure, and therefore its properties, is dependent on the type of electrolyte and electrode used. Additionally, the re- duction of the lithium salt and the solvent occurs separately and gives different products. The anion of the salt reduces and forms inorganic products, without car- bon and hydrogen bonds, that stick to the anode surface. Meanwhile, the solvent which is often composed of carbonates or other organic compounds, form polymers or semi-carbonates [26]. The conduction of lithium ions through the SEI is due to the properties of the organic components, while the non-reactive properties of the inorganic compounds help protect the surface from further decomposition [22]. Since the formation of the SEI demands the consumption of lithium ions that will subsequently not be able to take part in the intercalation process, there will be an inevitable loss of capacity during the first cycles of the battery. This irreversible loss of capacity (IRC) cannot be prevented fully, but efforts are being made to minimise it while still allow the formation of a protective SEI layer [5]. 2.4 Electrode materials The critical properties of electrode materials include their ability to efficiently host and intercalate/de-intercalate lithium into its structure. Furthermore, they should be good electrical conductors with little internal resistance. Preferably, they should also be able to hold as much charge as possible per unit weight and volume. In lithium-ion cells, transition metal oxides with the property to host lithium atoms as interstitials have been used as cathode active material, while variants of carbon such as graphite or hard carbon have been used in the anode. While other electrode mate- rials are consistently being researched and developed, the beforementioned materials have been dominating both the commercial market and research topics [27][28]. The transition metal oxides can be divided into groups according to their structure as well as in which directions the lithium can be intercalated. The layered structure oxides can hold lithium in between atomic layers meaning the lithium can diffuse 11 2. Lithium-ion batteries freely in two dimensions. Spinel structures (such as LNMO) and olivine structures instead host the lithium in interstitial sites forming tunnels with the former allowing lithium movement in two directions and the latter in only one direction [5][27]. For the electrolyte to be in contact with the largest surface area possible and therefore allow for quick diffusion of ions to reaction sites, the microstructure of the electrode is porous [17]. There are multiple ways to prepare electrodes but usually it involves sputtering, spreading, or spraying of small particles of active material onto a current collector foil, leaving a thin layer of material fully stuck. The particles are mixed with a binder that helps form the film and stick the particles to the electrode. The electrode can also be flattened between heated cylinders to increase the volumetric energy density and improve the contact between its constituents [29]. In high-voltage electrodes sometimes a thin carbon coating is added prior to the active material to prevent corrosion of the collector. From the foil, which usually has a thickness of 10 to 100 µm, electrodes can be stamped out into wanted shape and size. Figure 2.2: Layout and microstructure of an LNMO electrode. The LNMO active material has been applied to a carbon coated aluminium current collector. By adding a carbon coating between the LNMO and aluminium, adhesion is improved and corrosion of the foil may be prevented. It is the material properties and quantity of the electrode materials used in cells that dictate the theoretical capacity of a full battery. The capacity is measured in either ampere-hours (Ah) or coulombs (C) and is dependent on the maximum number of electrons that can be exchanged in the redox reactions at the interface. By considering the molecular weight M of the active material, the material of the electrode crucial for the reactions, and the quantity of it in the cell, a specific theoretical capacity for the electrode can be calculated, C = nF M C/g = nF M 1000 3600 mAh/g (2.11) In the case of lithium, only one electron is exchanged in the redox reactions, resulting in n = 1 [30]. The theoretical values are, as is usual in real life applications, not often reached. The intercalation of ions could for example cause structural instability in the electrode and a passivating SEI could hinder reactions, both resulting in loss of capacity [27]. In Table 2.1 the theoretical (CT ) and practical (CP ) specific capacities of a few common anode and cathode active materials are presented. 12 2. Lithium-ion batteries 2.4.1 High voltage LNMO electrodes Compared to commercially available cathode active materials, such as LCO and NMC, LNMO has a higher operating voltage of 4.7 V vs. Li/Li+, making it a promising cathode for applications where a high power output is desired. This has been attributed to the reduction and oxidation of the Ni2+/Ni4+ pair (by Ni2+/Ni3+ and Ni3+/Ni4+) at 4.7 V [27]. Despite it having a slightly lower theoretical capacity than other commercial electrodes, it has been shown to have a better thermal sta- bility than other cathodes and due to its lack of cobalt it is also a relatively cheap and environmentally friendly material [31][32]. Additionally, LNMO has a spinel structure with interstitial sites forming tunnels in the crystal structure, hosting the reduced lithium atoms. LNMO has not yet been commercially implemented due to its cycling issues. In addition to the lack of electrolytes covering the ESW, LNMO cells with graphite as an anode have shown a gradual loss of capacity after several cycles. The capacity loss implies either a loss of lithium ions participating in the reactions or some sort of deterioration of the SEI or electrode. The cause is believed to be the release of manganese ions into the electrolyte that then diffuse into the bulk and deposit on the opposite graphite. This further encourages the decomposition of other species at the interface, creating a thick impenetrable SEI that also consumes the active Li+. Consequently, LNMO has also been cycled with lithium titanate (LTO) with better results, since the same deposition of the transition metal ions does not occur. However, since the voltage of LTO against lithium is 1.55 V compared to the 0.1 V of graphite, it means the advantage of LNMO as a high voltage electrode is lost [31][33]. 2.4.2 Graphite anodes Graphite has been the preferred anode in commercial batteries since the end of the 1980s due to its high specific capacity and thermal stability. Before graphite, hard carbon or coke, two types of carbon with a more compact structure, were also used. However, due to graphite having a flatter discharge curve and higher specific capacity, the voltage output was more stable than the other options, leading to graphite taking over the market [5][28]. Like the cathode materials, the graphite is organised in smaller, 10-20 µm particles which are usually deposited on a copper collector. Due to the tendency of lithium to corrode aluminium and form alloys on its surface at low potentials, aluminium current collectors are rarely used on the anode. The size of the particles, impurities in the graphite, and if a protective coating has been applied to the particles are all properties that change between producers and will naturally have a significant effect on the performance of the anode [5]. The intercalation of lithium into the graphite layers is only possible if the layers are correctly aligned. Furthermore, the intercalation will not be homogenous, but the lithium will form clusters in the layers. Depending on the fraction of misaligned layers in the particles, the specific capacity of the anode will be diminished. Fur- thermore, the intercalation of lithium will cause the expansion of layers, resulting in a volume increase of about 10% from fully discharged to charged [4]. This makes the expansion of the anode an important parameter in the construction of a full-scale 13 2. Lithium-ion batteries battery and the metal casings which hold the cell. Table 2.1: Data for a few active materials used in LIBs. Name Structure Formula CT (mAh/g) CP (mAh/g) Voltage vs. Li/Li+ (V) Cathodes LCO Layered LiCO2 274 [5] 155-185 [5] 3.9 [5] NMC622 Layered LiNi0.6Mn0.2 Co0.2O2 275 [34] 140-190 [5] 3.8 [5] LMO Spinel LiMn2O2 148 [25] 120 [5] 4.05 [5] LFP Olivine LiFePO4 170 [35] 120-160 [8] [36] 3.45 [5] LNMO Spinel LiNi0.5Mn1.5O4 147 [25] 134 [31], 110 [32] 4.7 [11] Anodes Graphite Layered LiC6 372 [25] 355 [36] 0.1 [5] LTO Spinel Li3/4Ti5/3O4 170 [5] 160 [5] 1.55 [5] Lithium BCC Li 3860 [37] - 0 2.4.3 Lithium metal half-cell configuration Historically, lithium metal has also been used as an anode although not at the same industrial scale as for graphite and alloys. Lithium metal does not intercalate the ions but instead plates the lithium on its surface during charging and strips the lithium during discharge. The volume increase after charging is higher than for graphite, which may cause issues, but lithium also has a significantly larger theo- retical capacity [5][38]. However, lithium metal anodes have not been implemented commercially with liquid electrolytes due to the issue of dendrite formation. The lithium is not plated uniformly on the surface of the lithium metal anode but can during multiple cycles instead form spiky structures that poke into the electrolyte. If the dendrites reach the cathode, the cell will be short circuited which constitutes a severe safety hazard in applications [38]. In research however, lithium anodes have found another purpose and are com- monly used as a standard electrode against both anode and cathode materials to isolate the performance of only one electrode. Lithium is a stable reference poten- tial, and its high capacity also ensures that it does not set the capacity limit of the cell setup. This setup is often called a half-cell as it only focuses on the behaviour and characteristics of one of the electrodes and the electrolyte [39]. Since the cells in research often are not cycled as many times as full batteries and are not as sensitive to short circuiting, the lithium metal anode is often utilised and is an important tool in the characterisation of electrochemical cells. 14 2. Lithium-ion batteries 2.5 Electrolytes for LIBs The role of the electrolyte in an electrochemical cell is to separate positive and negative charge and to transport the positive ions between electrodes during charge and discharge. Furthermore, it must do this reliably for many cycles. A high ionic conductivity, low electronic conductivity and good electrochemical stability are therefore the key properties of a functioning electrolyte. The source of lithium ions in the LIB electrolyte comes from the solvation of lithium metal salts in some type of solvent. This also requires the electrolyte to have good salt solubility properties to solve enough salt for the operation of the cell and not form precipitates [7]. This can be measured in permittivity, as a solvent with a high dielectric constant will solve ionic compounds more easily. In Section 2.2, it was shown that low viscosity materials facilitate the movement of ions through the electrolyte. Unfortunately for many commercial electrolytes, this property rarely goes hand in hand with a high dielectric constant. As a consequence, they are almost exclusively mixtures between solvents of varying properties [18]. Furthermore, the electrolyte needs to be able to form a thin but protective layer of SEI together with the electrode to allow the permeation of ions but simultaneously protect against degradation [7]. The benchmark electrolyte used in LIB applications is lithium hexafluorophos- phate, LiPF6, in a mixture of carbonate esters such as ethylene carbonate (EC), dimethyl carbonate (DMC) or diethyl carbonate (DEC). This class of electrolytes is the one used almost exclusively in commercial applications and has been shown to form a good SEI while being stable in the ESW set by commercial anodes such as LCO or LFP [7][16]. EC has a high dielectric constant but is a solid at room temperature, while DMC and DEC are of low viscosity with high ionic mobility [18]. Beyond these electrolytes, there is a vast number of solvents and salts all with greatly varying properties. In addition to organic solvents, ionic liquids (IL), salts that are liquid below 100◦C, have become popular both as solvents as well as addi- tives due to their large ESW as well as their thermal stability [10]. The tuning of salt concentration and composition creates a near infinite selection of electrolytes. In this section some of the most common salts, solvents, and ionic liquids, as well as ones relevant for the scope of this thesis, are presented. 2.5.1 Lithium salts The most important properties of the lithium salt are dissociation degree, solubility in the solvent, and both chemical and thermal stability. Additionally, together with the solvent, it should help form a good SEI. In almost all commercial applications, lithium hexafluorophosphate (LiPF6) is the salt used due to its high ionic conduc- tivity and its compatibility with graphite anodes [7]. The anion is constituted by a central phosphorous atom, bound to six surrounding fluoride atoms in an octa- hedron shape. The P-F bonds are, however, relatively weak and in the presence of water and high temperatures, the anion can undergo hydrolysis creating highly toxic hydrogen fluoride gas. In the case of battery malfunctions where water could penetrate the cell this is a concern, especially in the case of high voltage electrodes and in electric vehicles where the temperatures are heightened [18]. 15 2. Lithium-ion batteries Other salts, known for their superior thermal stability as well as their stability against hydrolysis, are lithium bis(fluorosulfonyl)imide (Li[N(SO2F)2]) and lithium bis(trifluoromethanesulfonyl)imide (Li[N(SO2CF3)2]), more commonly known as LiFSI and LiTFSI, respectively. Since the S-F bond in FSI is weaker than the C-F bond in TFSI, the latter anion has a slightly higher thermal and electrochemical stability. However, electrolytes of both show slightly lower ionic conductivities than when us- ing LiPF6 [7][40]. Also, important to mention is the tendency of LiTFSI and LiFSI to corrode the aluminium foil current collector at the cathode. Curiously enough, an exception to this is ionic liquids with TFSI or electrolytes with a high concentration of the salt [7]. Figure 2.3: The chemical structures of two common lithium salts. LiTFSI is a solid crystalline powder at room temperature. Since the P-F bond of the LiPF6 is less stable than the C-F bonds, LiTFSI is more suitable for cells with a high voltage cathode. Models made with MolView. Another area of research is high concentration or solvent-in-salt electrolytes where the salt concentration is raised from the conventional 1-2 M (mol/dm3) to up to 1:1 molar ratio. Since the ratio of salt is so high, there are few available free solvent molecules. The ESWs are wide, and the electrolyte becomes less volatile. The downsides of these types of electrolytes are their high viscosities, reducing the ionic mobility. Additionally, the high viscosity can cause slow or bad wetting of separators and electrodes, leading to poor contact and slow reactions in the battery [10][41]. Finally, in addition to the already mentioned, there is a plethora of salts being researched and developed. Furthermore, adding a level of complexity is the mixing of two or more salts to fine-tune the desired properties [42]. 2.5.2 Non-aqueous organic solvents Organic solvents are the most common medium for the dissolution of lithium salts in liquid electrolytes. Aqueous LIBs with water as a solvent also exist, with for example spinel oxides as cathodes, although with different anode materials. However, due to the narrow ESW of the aqueous electrolytes of about 1.23 V, they are effectively 16 2. Lithium-ion batteries eliminated from being used in an oxide/graphite cell [43][44]. Instead, carbonate esters, ethers and other organic compounds are the basic solvents found in LIBs. A common property for most of the organic solvents is their flammability, which has been and still is a challenge in LIB research, although a selection of additives to the solvents has been shown to significantly improve battery safety [42]. As previously mentioned in the benchmark example at the beginning of section 2.5, the carbonate esters are the most established solvents due to their low viscosity, good ionic conductivity and high dielectric constant, as well as their sufficient elec- trochemical stability with commercial electrodes [10][38]. The carbonate esters can be divided into linear carbonates (DEC, DMC, EMC) and cyclic carbonates (EC, PC) depending on the chemical structure, which also gives them different properties. DEC has linear chains which lowers the viscosity compared to the ring-shaped EC, which instead has a higher dielectric constant [38]. The carbonate esters also form a very stable SEI making them suitable against graphite anodes. The downside of the carbonates however is the limited ESW in the case of high voltage electrodes. Ethers have not been utilised to the same extent as carbonates as they have shown to not be stable against cathodes operating higher than 4 V. However, in a dual electrolyte setup the ether could potentially work against the anode side. Moreover, they are less flammable than the carbonates making them relevant as an additive [44]. The most common ether in connection to LIBs is dimethoxyethane (DME), also known as glyme which has been paired together with PC and EC to lower the viscosity and create non-flammable electrolytes [45][46]. To improve the properties of both carbonate and ether solvents, fluorine atoms can be added to the compounds. It has been proved to both increase the stabil- ity range as well as making the solvent less flammable [7][10]. One example is the addition of fluoride to EC, resulting in fluoroethylene carbonate (FEC), that has been shown to form a superior SEI on both anode and cathode in high voltage sys- tems [5][36]. Furthermore, organic compounds containing sulphur such as dimethyl sulfoxide (DMSO) and sulfolane (TMS) have also been utilised as solvents due to their low flammability and high dielectric constant. In comparison to ethers, they are more stable against various cathodes but have issues forming proper SEIs due to their high viscosity [38]. Table 2.2: Viscosity and ionic conductivity of some organic solvents at room tem- perature (293 K). Solvent Type Viscosity (mPa·s) Ionic conductivity, (mS/cm), 1 M LiPF6 EC Cyclic carbonate ester 1.9 (40◦C) [7] 6.9 [5] PC Cyclic carbonate ester 2.5 [7] 5.2 [5] DMC Linear carbonate ester 0.59 [7] 6.5 [5] DEE Ether 0.288 [47] - DBE Ether 0.414 [48] - DMSO Organosulfur compound 1.99 [7] ∼11 [49] TMS Organosulfur compound 10.28 [7] 0.2 (in 1 M LiTFSI) [50] FEC Fluorinated ether ∼4 [7] 5 [7] 17 2. Lithium-ion batteries Figure 2.4: Chemical structures of some organic solvents. Models made with MolView. 2.5.3 Ionic liquids Ionic liquids (ILs), salts that are liquid below 100◦C, and most often are also room- temperature ILs, were first brought to the attention of LIB researchers when new cation-anion pairs were developed which were water-insensitive and resistant to hy- drolysis. The ILs have for some time been considered promising due to their high thermal stability as well as their expanded ESW compared to organic solvents [50]. At high temperatures their ionic conductivity is relatively high, but they often show high viscosity at room temperature, potentially limiting diffusion of lithium ions. Additionally, due to the large selection of both anions and cations and the various pair they form, ILs can be considered “designer solvents” where there is a large spread of properties available. The desired properties of ionic liquids are the same as for organic solvents, namely good salt solubility and good SEI formation and they are simultaneously being researched as a stand-alone solvent and as an additive. Is- sues with ILs include their viscosity, leading to bad wetting of the separator and electrode, as well as problems with forming SEIs and CEIs against various electrodes that cycle well [7][45][51]. 2.5.3.1 Cations Generally, the positively charged cations of ILs are organic compounds in either ring formations such as the pyrrolidinium (PYR) and imidazolium (Im) or an inorganic atom such as phosphorous, aluminium or sulphur with chains of organic groups, such as alkyls, attached [51]. The imidazolium ILs were the first to be researched and showed better ionic conductivity than organic solvents. The Im-cations include two highly reactive acidic hydrogen groups, lowering the upper cathodic limit of the ESW of ILs with the cation [10]. Furthermore, on low voltage anodes such as graphite, ILs with Im-cations would reduce and not form a proper SEI, although adding small amounts of carbonate solvents was shown to help with the formation [38]. PYR-based ILs exhibit higher cathodic stability than Im, even up to 5.5 V against lithium for some anions, while simultaneously having lower viscosity. Like most ILs, 18 2. Lithium-ion batteries it forms an improved SEI with the addition of organic solvents and therefore has a better anodic stability, but with the drawback of a reduced ESW [51]. ILs based on aluminium and phosphonium atoms have been researched less due to their high melting temperature and high viscosity. However, they have shown to be stable against several cathode materials as well as pure lithium anodes [52]. 2.5.3.2 Anions The anions of ILs relevant for LIBs are primarily the same anions as for the lithium salts, such as [PF6]−, [BF4]−, [TFSI]− and [FSI]−. Moreover, it is believed that the anion is the determining factor when it comes to anodic stability [38]. A larger anion causes the ionic bond between anion and cation to be weaker, meaning they dissociate ions better leading to better mobility. For example, [FSI]− has showed better ionic conductivity and lower viscosity than [TFSI]− due to the molecule’s smaller radius [52]. Table 2.3: Some properties of a selection of ionic liquids. Trivial name Cation/ Anion ESW (V) Viscosity (mPa/s) PYR14TFSI 1-butyl-1-methylpyrrolidinium/TFSI 5.88 [53] 84.33 [54] EmimTFSI 1-ethyl-3-methylimidazolium/TFSI 5.5 [55] 35.55 [56] P66614TFSI trihexyltetradecylphosphonium/TFSI - 304 [57] BmimBF4 1-butyl-3-methylimidazolium/BF− 4 5.3 [58] 72.51 [59] 2.6 Separators Without an insulating separator between electrodes, the cell runs the risk of short circuiting. Since the separator must allow ion permeation, the separators are made porous in order to soak up the electrolyte. This presumes that the electrolyte is able to wet the material. Additionally, the materials need to be durable and stable when put under elevated temperatures and charge [5]. Traditionally, materials such as glass fibre and polymers have been used. An example is Celgard, a membrane made from polypropylene and polyethylene which is often used in battery research. The thickness of these separators is typically 10-25 µm with a pore size of less than 1 µm. In the case of a dual electrolyte setup, the capillary forces of the separators when soaked with different electrolytes are crucial in preventing mixing of the electrolyte. A system where the electrolytes completely wet their separators and the separators are compressed minimally after wetting would be ideal. Furthermore, there should be good contact between electrolytes, and preferably the capillary forces should contain the electrolytes fully. 19 2. Lithium-ion batteries 2.7 Research cell configurations Naturally, there is no point in building a full-size commercial battery when testing new materials for LIB applications, as it is neither economical nor resourceful. In- stead, smaller setups are used to characterise the cells, designed after the intended characterisation method. To evaluate the performance of the batteries, the batteries are cycled by varying the applied current or voltage across the cell while measuring and recording the resulting voltage, current or impedance. To further understand the mechanisms of the cell, other characterisation techniques such as optical spec- troscopy or tomography can be used to study the materials during cycling. In these cases, special cells that expose the area of interest can be constructed, an example being the in situ Raman spectroscopy cell used in this work, which will be detailed further on. To avoid contamination and oxidation of the cell and prevent moisture in non- aqueous electrolytes, the cells are assembled in an inert environment, most often in an argon-filled glovebox. While some gloveboxes allow cycling of batteries inside, the cell materials are most often placed in a container that can be closed off and sealed off from air and moisture and then taken out of the glovebox to be cycled. Finally, important to note is while these cells are intended to emulate the performance of industrial battery stacks on a smaller scale, there will always be limitations. Ex- amples include differences in electrolyte and active material ratios, heat generation and dissipation as well as available surfaces for reactions [60]. Nevertheless, they are important tools in research and in the development of new materials. 2.7.1 Coin cells The most common and simple cell used for electrochemical characterisation is the coin cell, a small flat cylindrical cell which is available in multiple sizes and dimen- sions. The standardised CR2032 coin cell is often used and has a diameter of 20 mm and height of 3.2 mm when assembled. Fig. 2.5 shows the components of a conventional coin cell stack before assembly. One electrode is placed in the case with the current collector side of the electrode towards the metal, while the other electrode will be in contact with the spacer. The role of the spacer and spring is to assert pressure on the stack to ensure good connectivity between the parts of the battery. Different thickness of spacer can be used to better control the pressure inside the cell. The gasket is a plastic ring which prevents the cell from short circuiting by separating the case and cap and also makes the cell airtight. Likewise, the separator is cut slightly bigger than the electrodes to avoid accidental contact between them. The contents and metal casing of the cell are simply stacked and are thereafter placed in a closing device which crimps the edges, sealing the cell from the environment. When cycled, the two poles are connected to the flat top and bottom. After cycling, the cell can be opened by bending the case open, but not without some disturbance of the cell stack. 20 2. Lithium-ion batteries Figure 2.5: Components of a coin cell battery. The cathode and anode are attached to the metal current collectors as previously described in Fig. 2.2. 2.7.2 Swagelok cells Another configuration, more easily taken apart, is the Swagelok cell. The electrodes and separator are placed in a hollow tube made of fluorinated polymer and com- pressed by two metal cylinders on each side. Next, a plastic casing is screwed on the tube, with packings ensuring isolation from the environment. Furthermore, pistons inside the tube makes it possible to control the pressure asserted on the cell stack. The cylinders are in contact with the electrodes and by inserting pins in the holes at the ends of the cylinders, the cell stack can be connected to an external circuit. The result is a cell that is easily assembled and disassembled, making it possible to remove the electrodes and separator for further characterisation such as spectroscopy or microscopy. Another advantage of the Swagelok is the possibility to cycle the cell at higher temperatures. The plastic gasket of the coin cell is at risk of deforming which may cause the cell to leak while the fluorinated polymer casing is less sensitive to heat. Figure 2.6: The Swagelok cell and its inner components. 21 2. Lithium-ion batteries 22 3 Immiscibility and diffusion of electrolytes To realise a dual electrolyte concept based on the immiscibility of electrolytes, the first natural step is to find appropriate liquids that in theory fulfil the requirements of a LIB electrolyte. This chapter presents the results of a literature review where electrolyte pairs have been found and subsequentely tested for their immiscibility. The compatibility of the electrolytes with common separators were also investigated since proper wetting is crucial to prevent diffusion between the pairs. Multiple techniques were considered to study the diffusion in the separators. Since the most crucial component of the dual electrolyte concept is to prevent the contact between one electrode-catered electrolyte and the other electrode, diffusion close to the electrolyte/electrolyte interface was deemed not to be the primary issue at this stage. Instead, the question asked was if the emulsion region of mixed electrolytes was thin enough to ensure that the electrolytes did not seep through to the opposite electrode. Furthermore, as the assembly of a cell stack involve applying pressure to guarantee good contact, this element should also be included. Optical spectroscopy techniques to detect the presence of electrolyte on the separators were deemed to be the best approach, due to the possibility of distinguishing between molecular species and their non-destructive nature, ensuring that the probing of the material would not interfere with the diffusion or cause mixing. Additionally, a number of spectroscopy instruments were readily available at the Material Physics division at Chalmers. A first attempt at characterisation of the separator surface was made using a Swagelok cell and attenuated total reflectance Fourier-transform infrared spec- troscopy (ATR-FTIR). Two separately wetted separators containing the electrolytes were placed in the cell together, compressed and left for 24 hours, replicating the assembly and environment of a real cell but without electrodes. ATR-FTIR is a technique where infrared laser light is directed onto either a solid or liquid sam- ple placed on a crystal, making fast identifications of materials at a depth of 1-5 µm by analysing the reflected spectra. Another advantage of using the ATR-FTIR spectrometer was its placement inside the glovebox, making measurements possible without having to remove the cell and the separators from the inert environment. Af- ter resting, the cell was disassembled and both sides of each separator were measured to check for the presence of the opposing electrolyte. However, the measurements of solid materials involved pressing the separator against the crystal with a clamp to get a good signal, which probably interfered with the diffusion. Moreover, it was also difficult to assess how the removal of the separators from the Swagelok tube 23 3. Immiscibility and diffusion of electrolytes affected the measurements, making the results unreliable. A more uninvasive technique was developed after taking inspiration from Raman spectroscopy in situ measurements of graphite electrodes. A special cell, developed specifically for Raman measurements, was used to make measurements of one side of the separator without removing it from the other. The cell also provided an isolat- ing environment, preventing evaporation and contamination, as well as offering the possibility to assert pressure on the separator stack without opening the cell. This chapter presents the theory behind Raman spectroscopy, a technique that like ATR- FTIR can identify the presence of species on the surface of materials by analysing the chemical structure. It is followed by a discussion on the penetration depths of transparent or semi-transparent materials, like the glass fibre or Celgard separators and their significance on the measurements. Finally, the chapter concludes with the results of diffusion measurements of two of the pairs, one IL/IL and one solvent/IL. 3.1 Immiscible electrolytes The concept of immiscible electrolytes that can carry charges across their interface is not new. For instance, the interface between two immiscible electrolyte solutions (ITIES) has for example been used since the early 1900s as a physical model to study the characteristics of biological membranes that also transfer charged particles across an interface [53]. Other practical applications include sensors for electrochemistry measurements and in pharmaceutical studies in connection to the delivery of drugs [61]. Although not discussed in detail in this thesis, which mainly focuses on the diffusion, the charge transfer across the interface is significant for the ionic mobility and therefore also for the operation of the cell. 3.1.1 Polarity of solvents Liquids that fully mix do so because the total internal energy for the mixture is lower than that of the liquids separately. It is thermodynamically favourable to break the interactions between the same molecular species and form new in the mixture. Liquids with similar interaction strength and polarity are more likely to mix and form a miscible solution than pairs for which it differs, a simple example being organic chains like hexane not mixing with water, a high polarity molecule. In this work, this was taken as a rule of thumb when searching for solvents, while keeping in mind that other factors may affect solubility. Nevertheless, solvents used in LIBs need also to be able to solve the polar bonds of the lithium salts, meaning low polar solvents such as hexane are excluded, at least as a stand-alone solvent [62]. The targeted solvents were therefore in the rough range of slightly polar to polar. 3.1.2 Immiscible pairs In the early stages of the review, no solvents or ILs were excluded based on their documented cycling performance in LIB cells, as the main interest lay in the study of diffusion. However, the electrolytes tested for their immiscibility and thereafter selected for the measurements were in part decided by the availability at the Material 24 3. Immiscibility and diffusion of electrolytes Physics laboratory, as the lead time for some chemicals, especially the ILs, did not match the time plan. While a number of immiscible pairs were found, unfortunately no solvents used in commercial electrolytes, such as PC and EC, were found to be immiscible. The carbonate esters were instead found to be miscible with a wide range of solvents only excluding non-polar liquids such as gasoline and turpentine. Solubility charts from chemical and pharmaceutical company Merck indicated the immiscibility between the polar solvent dimethyl sulfoxide (DMSO) with several non-polar solvents such as diethyl ether (DEE), dimethyl formamide (DMF) and dimethoxyethane (DME). The downside of DMF and DME is their high toxicity ([63][64]) making them less than ideal when handling them in a diffusion study, outside of a glove box. Furthermore, the boiling point of DEE is only 34.6◦C making it highly volatile and unsuitable in battery applications. Instead, dibutyl ether (DBE) was considered due to its slightly longer alkyl chains with four alkyl groups on each side of an oxygen atom instead of two. The longer chains result in stronger intermolecular bonds, raising the boiling point to 142◦C. In a glovebox, DMSO and DBE were added to a small glass vial, shaken, and left to stabilise with the higher density DMSO on the bottom. The result can be seen in Fig. 3.1, showing a small visible meniscus and proving immiscibility. Figure 3.1: Mixed electrolytes. a) DBE and DMSO, b) DBE and Sulfolane, c) DBE and [PYR14][TFSI], d)DBE, [PYR14][TFSI] and TMS, e)[PYR14][TFSI] and [BmimBF4][TFSI], f)[P66614][TFSI] and [Emim][TFSI]. Continuing, datasheets from Solvionic, a chemistry company specialising in ILs, stated that several PYR-ILs and Im-ILs were immiscible in DEE and other non- polar solvents such as hexane and toluene. This could be expected given the po- larity of the freely moving anions and cations in the IL. The well documented 1- butyl-1-methylpyrrolidinium bis(trifluoromethylsulfonyl)imide, [PYR14][TFSI], was selected, and tested together with DBE, also showing a clear meniscus, although hard to see on the picture. Compared to the organic solvents, [PYR14][TFSI] had a higher viscosity and the two phases, with [PYR14][TFSI] at the bottom, took longer to stabilise after the vial was shaken. One of the articles presenting a dual 25 3. Immiscibility and diffusion of electrolytes electrolyte configuration for a lithium-sulfur battery, by Ren and Mantiram [14], demonstrated the immiscibility between DBE and TMS. The less viscous mixture of [PYR14][TFSI] and TMS in 1:1 volume ratio was mixed with DBE, and also proved to be immiscible. Moreover, a few IL/IL pairs were also tested. It was reported in literature that imide-based ILs, with anions such as [TFSI]−, were immiscible in water while borate- based ILs like [BF4]− were not. Therefore two ILs with these anions were tested, paired with two different cations, [PYR14][TFSI] and 1-butyl-3-methylimidazolium tetrafluoroborate [Bmim][BF4]. In this case, the liquids were fully miscible, even after leaving it to stabilise for several weeks. Additionally, it would have been in- teresting to test these anions with the same cation, but due to [PYR14][BF4] being crystalline at room temperature, this was not done. Finally, a 2006 article by Arce et al. ([65]) documented the immiscibility between the two ILs tetratrihexylte- tradecylphosphonium TFSI, [P66614][TFSI], and 1-ethyl-3-methylimidazolium TFSI, [Emim][TFSI]. This pair was therefore also tested, and their immiscibility proved, with [Emim][TFSI] settling on the bottom. Out of the pairs, [P66614][TFSI] and [Emim][TFSI] exhibited the highest viscosities. 3.1.3 Wetting of Celgard and glass fibre separators Proper wetting of the separators, meaning all porosities in the material are filled with electrolyte, is important for the performance of LIBs since empty space in the cell would mean higher resistance and lower ionic conductivity across the cell. Fur- thermore, in a dual electrolyte and separator cell the improper wetting of one of the sides could lead to the diffusion of the other electrolyte into the opposite sepa- rator, filling up the empty porosities. In this thesis, glass fibre separators made of borosilicate glass and Celgard 2400, made of polypropylene, were used. All solvents in the immiscible pairs mentioned previously, were able to wet the glass fibre sepa- rators relatively well which could be seen by the separator turning semi-transparent when filled. However, higher viscosity electrolytes, namely the ILs [P66614][TFSI], [Emim][TFSI] and [PYR14][TFSI] were all slow to fill the glass fibre, taking up to a couple of minutes to properly diffuse into a 10 mm diameter, 260 µm thick separator. While DBE easily wet Celgard, turning the separator almost entirely transparent, [PYR14][TFSI] first formed a bead on the surface before slowly diffusing into the material. The process could be sped up slightly by wetting the Celgard from the side, where presumably the fibres were aligned differently. [P66614][TFSI] also wet the separator rather slowly but did not form an initial bead. The two remaining electrolytes, DMSO and [Emim][TFSI], would not wet Celgard at all. Separators were placed in vials and completely covered in the electrolytes, but even after being left for weeks, the separators floated to the top of liquid or stuck to the walls of the vials without being wetted at all. This has likely to do with the long organic chains of the polymer and how it differs from the short aromatic ring of [Emim]+ and the short polar molecule of DMSO. In comparison, DBE, [PYR14]+ and [P66614]+ all have slightly longer alkyl chains connected to an oxygen atom, pyrrolidinium group and phosphonium atom respectively. The difference makes it unfavorable to form intermolecular bonds. 26 3. Immiscibility and diffusion of electrolytes However, the non-wettability of some solvents might not be an issue at all since it could be used to further ensure that no diffusion takes place between separators. Using glass fibre for [Emim][TFSI] and Celgard for [P66614][TFSI] would make it theo- retically impossible for [Emim][TFSI] to diffuse through the separator. Nonetheless, under enough pressure in the cell the liquids could still escape around the separators to the opposite electrodes. Furthermore, two different separators could potentially hinder the mobility of ions if it creates an isolated interface without an intermedi- ate emulsion layer. In the Raman measurements, both pairings of separators were tested. 3.2 Raman spectroscopy Raman spectroscopy is a noninvasive spectroscopy method capable of identifying materials by analysing inelastically scattered photons off the sample. The technique probes the vibrational modes of the molecules, revealing information such as chem- ical structure, molecular orientation, and strength of bonds without damaging the sample. It has since the 1960s been an important characterisation method, that can both probe solids, liquids and, in the right setup, even gases. Furthermore, it is quick while still making it possible to detect specific compounds on a micrometer scale [66]. In this section, the underlying interactions between vibrational energy levels in the molecules, resulting in identifiable spectrums, are explained. Moreover, the Raman spectrometer is introduced together with the Confocal spatial filter and it is explained how the latter makes it possible to achieve even more precise mea- surements, as well as measurements along the vertical axis. 3.2.1 Theory A simple model of the energy states of a material can be seen in Fig. 3.2, with E0 being the ground vibrational state and ν1 the first vibrational energy state. A monochromatic beam, for instance a laser, is directed onto the sample and a photon is absorbed, leaving the molecule in a virtual energy state. The interaction between the absorbed photon can either be elastic, namely Rayleigh scattering, or inelastic, Stokes and Anti-Stokes. In the case of the former, a photon will be emitted with the same frequency as the absorbed one, leaving the molecule in the initial ground state. If the molecule falls back to a different vibrational state, the emitted photon will experience a change in energy equal to the distance between ground state and the first vibrational state. This is called the Raman shift, and since different molecules have different energy configurations, the photons from the shift can be measured to characterise the material. If the molecule is already in a higher vibrational state, it can, after absorbing a photon, fall back to the ground state and emit a photon of higher frequency than the absorbed photon. This is called Anti-Stokes scattering which can also be measured, although it is not used as frequently for characterisation as Stokes scattering. The scattered photons are counted as a function of their frequency, creating a spectrum that can be used to identify the molecule. Instead of frequency, the intensity is plotted against the Raman shift, measured in cm−1 which is the 27 3. Immiscibility and diffusion of electrolytes difference in wavelength between the energy states. Naturally the peak formed by the Rayleigh scattered photons will be located at 0 cm−1 [66] A challenge when performing spectroscopic measurements is the fluorescence of the material, which is in many cases inevitable. Fluorescence is the result of transi- tions between electronic energy levels, in contrast to the vibrational levels responsible for the Raman shift. The fluorescence does not form peaks on the spectrum but instead leaves a baseline, as is visible in the spectrum presented in Section 3.4.1. In addition to data processing and background fitting of the fluorescence, changing the wavelength of the laser beam may help reduce fluorescence since photons with longer wavelengths will not have enough energy to excite the photons to a higher electronic energy level [66]. Figure 3.2: Energy states of a molecule absorbing a photon. The scattering can either be elastic (Rayleigh) or inelastic (Stokes and Anti-Stokes) with the latter emitting a photon of a different wavelength A simplified instrumental setup of a Raman spectrometer can be seen in Fig. 3.3. This instrument is equipped with a confocal filter which gives a better spatial resolution, both vertically and horizontally in the sample. By changing the pinhole of the confocal filter, only scattered light from a specific area will reach the detector, making it possible to perform detailed profiling of the sample along different axes, even into the sample. However, less light will be let through to the detector, which will give a weaker signal. In addition to the pinhole, the objective will also affect the resolution since it determines the area of illumination as well as the numerical aperture and the maximum angle at which the reflected light will be collected. A higher magnification on the objective will focus the laser beam in a smaller area, reducing the sampling area and improving the spatial resolution. Furthermore, inside the spectrometer the scattered photons are spatially dis- persed by a diffraction grating and counted when hitting the charge coupled device (CCD) detector. Changing the diffraction grating to a higher groove, broadens the dispersion of photons on the detector, which in turn increases the energy resolution 28 3. Immiscibility and diffusion of electrolytes of the spectrum. However, it comes at the cost of longer measurement times, as the entire energy range will not fit on the CCD detector. Instead, multiple measure- ments will need to be done to cover the full range. The period of time reflected light is allowed through to the detector is called the acquisition time. Longer acquisition times will give more counts on the detector and higher intensity but will not neces- sarily give less noise. Instead, the same acquisition can be performed a number of accumulations, which will thereafter be averaged, which may give a cleaner signal. Figure 3.3: A Raman spectroscopy instrumental setup. With a confocal filter, the spatial resolution of the spectrum as only light from the focal plane will be able to pass through the pinhole. Decreasing the size of the pinhole, will increase the resolution at the cost of a weaker signal. 3.2.2 Raman penetration depth in separators In order to make reliable measurements of the diffusion in the separators, the pene- tration depth needs to be considered. Since the thickness of the separators is small, and given the separators turn semi-transparent when wet, there is a risk of the pen- etration depth of the photons being as deep as the separators are thick, resulting in the Raman spectrum covering the entire material. The theoretical maximum depth resolution set only by the objective can be approximated with ∆z ≈ 2λ (NA)2 , (3.1) where ∆z is the full width at half maximum (FWHM) of the point spread function along the optical axis, λ the wavelength of the laser, and NA the numerical aperture of the objective [67][68]. This value will give the maximum probing depth of the laser, meaning that outside of this range the intensity of the reflected light will be weak. Furthermore, the limit of lateral resolution will be determined by the spotsize of the laser on the focal plane given by ∆x ≈ 1.22λ NA (3.2) 29 3. Immiscibility and diffusion of electrolytes This will give the smallest possible distance two points on the sample can be distin- guished from each other. For NA=0.25, the smallest aperture of the objectives used in this work, and lambda=633 nm (red laser), ∆z = 20 µm and ∆x = 3 µm , and for NA=0.5, ∆z = 5 µm and ∆x = 1.5 µm, Given that the thickness of the Celgard and glass fibre used in this work is 25 µm and 260 µm respectively this is an issue, at least for Celgard measurements. Furthermore, this is the theoretical limit. In practice, the material will further diffuse the laser beam, lowering both the lateral and depth resolution, especially in the case of transparent materials. However, with a confocal spatial filter the depth resolution can be improved and measurements inside transparent samples can be made. By calculating the spread of the reflected light on the back focal plane an intensity ratio between the Raman- scattered light transmitted through the pinhole and the initial light intensity can be calculated. The point spread function will have the shape of a Lorenztian curve centred around the focus point of the laser in the material and the width of the curve will be dependant on the numerical aperture, pinhole diameter and the magnification of the objective. A geometrical model to calculate this function was introduced and verified experimentally by Tabaksblat et. al. [69]. By measuring on a 2 µm thick polyethylene film on top of a thicker polypropylene film they could measure at which instrumental settings polypropylene showed up on the spectrum. For a objective of 50x and a pinhole of 300 µm, the depth resolution (defined as the FWHM of the point spread function) was deemed to be 3 µm, and 1.5 µm for a 100 µm pinhole which was in line with their model. This is more promising for measurements of the separators. There are additional effects that should be taken into consideration. Tabaksblat’s model does not take into the account the effects of refraction at the sample inter- face. A high refractive index will shift the focus point of the laser further down in the sample, altering the depth resolution [70]. The measurements by Tabaksblat et. al were on dry polymers, while the separators are porous and will be wetted. This will change the refractive index, but it is unsure by how much. Furthermore, the electrolytes will possibly have slightly varying refractive index depending on their viscosity. Everall ([70]) proposed a correction to the depth resolution taking into account the refractive index, however only for homogenous samples. Finally, FWHM approximation of the point spread function is somewhat simplified. While the majority of the signal will come from within this depth profile range, it still as a tail that could interfere with measurements. Interactions with the sample outside of the range might very well show up on the spectrum, especially if the material is more prone to Raman interactions than the material at the focus point. 3.3 Method The diffusion of two immiscible electrolytes in separators was investigated with Ra- man spectroscopy. Out of the immiscible pairs found in literature and tested positive for their immiscibility, two were selected for the Raman measurements, DBE and [PYR14][TFSI] as well as [P66614][TFSI] and [Emim][TFSI]. Since [Emim][TFSI] does not wet Celgard it was decided that a setup with two glass fibres and a second setup with one glass fibre and one Celgard would be tested. For DBE and [PYR14][TFSI], 30 3. Immiscibility and diffusion of electrolytes one setup with two glass fibre separators and a second with two Celgard separators were measured. A specific in situ Raman cell, meant for measurements on cells while cycling was repurposed and used with only separators. The separators were filled up with the different electrolytes and measured under various conditions, such as after pressured was applied and after some resting time. The following sections detail the method and practicals of the experiment, such as preparation of the samples and settings of the Raman spectrometer. 3.3.1 Preparation of the electrolytes and separators The electrolytes were prepared in the same way as they would be for cycling purposes and all handling of the salt and solvents took place in an argon-filled glovebox until the sealed cell was taken out for measurements. Therefore, both solvents and salts were dried to remove moisture that could hinder the performance of the cell. LiTFSI was chosen as the lithium salt, partly due to it being well documented in literature in relation to the chosen solvents, but also due to the usage of [TFSI]− as an anion in the ILs. The salt was dried in a Büchi oven under vacuum and at 120◦C, for 24 hours. Likewise, the solvents were dried in a Büchi oven but at slightly varying temperatures. [PYR14][TFSI] (From Solvionic), DBE (Sigma-Aldrich) and [Emim][TFSI] (Solvionic) were all dried at 100◦C for 24 hours. [P66614][TFSI] (Sigma- Aldrich), had a documented flash point of 52◦C and was dried at 50◦C for 24 hours. It is however unclear whether such a low temperature as 50◦C managed to evaporate any moisture as the boiling temperature of water is slightly below 100◦C. Finally, the electrolytes were mixed to a concentration of 1 M LiTFSI and stirred on a magnetic stirrer overnight. All solvents except for [P66614][TFSI]completely solved their salt. In [P66614][TFSI], small crystals could still be seen even after additional stirring under heat for a number of hours. The Celgard was of model 2400, with a thickness of 25 µm and 40% porosities. The glass fibre (Hartman GF/C) had a thickness of 260 µm and pore size of 1.2 µm. They were cut into 10 mm rounds to fit the metal piston. 3.3.2 In situ Raman cell Performing Raman measurements on the materials of a cell during cycling, in situ, makes it possible to study the evolution of the electrode materials during intercala- tion and deintercalation. For instance, the insertion of lithium in graphite causes the distance between the graphene sheets to increase. This expansion and subsequent change in bond strength between the sheets can be seen as a shift of the peaks in the Raman spectrum. To gain access to the graphite surface, in situ Raman cells such as the one in Fig. 3.4 have been developed. The cell stack is built in a cylindrical compartment with a metal piston at the bottom and a transparent glass pane at the top and is sealed by screwing a metal top to the rest of the cell. Contrary to regular coin cells, where the copper current collector is attached to the graphite, the in situ cell utilises a copper mesh between the top glass window and the graphite, ensuring good connectivity to the external circuit but still allowing the laser through the holes of the mesh. The cell has three 31 3. Immiscibility and diffusion of electrolytes pins that can be screwed into the stack, connecting the electrodes to the external circuit. For the diffusion measurements the electrodes were removed as well as the copper mesh and an additional separator was added. A spring can be added to the bottom of the cell and screwed in to push the cell stack platform against the glass. This could be done without exposing the cell to the environment as the piston had an internal packing sealing the cell. To prevent leakage, vacuum grease was applied to both the O-ring by the glass and the packing of the piston. The diameter of the piston holding the stack is 10 mm, and the separators were cut to match. Approximately 15 µl of electrolyte was added to wet the Celgard while 20 µl was added to the glass fibre. The first separator was placed in the cell dry and then wetted, while the second separator was wetted before being added on top. Figure 3.4: Left: Raman in situ cell used for cycling measurements. Right: The setup for the Raman diffusion measurements of this work. To properly evaluate the diffusion, individual spectra of the electrolytes was taken to be used for comparison. The same cell was used for these measurements to have the same environment and parameters. Without the spring inserted in the cell, the piston could be pulled down creating a cylindrical well to hold the liquids. To prevent spillage and use less electrolyte a Teflon spacer with a small hole was added on top of the piston, creating an even smaller well to hold the electrolyte. 32 3. Immiscibility and diffusion of electrolytes Figure 3.5: Setup for measuring the Raman spectrum of the individual electrolytes. The white Teflon spacer holds the liquid and prevents spillage. 3.3.3 Instrumental setup, calibration and focusing A LabRAM HR Evolution Confocal Raman Microscope from Horiba was used, with a HeNe laser of wavelength 633 nm (red HeNe) from REO. During the first trial measurements a blue laser of wavelength 488 nm from Coherent was used to in- vestigate if the fluorescence from the sample was less. Although the fluorescence was reduced slightly, the signal was weaker due the less energetic photons, making it necessary to take longer measurements which, compared to the red laser, was noisier. Furthermore, two diffraction gratings of 300 lines/mm and 1800 lines/mm were tested, before choosing the 300 grating due to faster measurements, and rela- tively similar energy resolution. The size of the pinhole was set to 199 µm, while objectives with magnifications 10x (NA 0.25) and 50x (NA 0.50) were used. Before measurements the spectrometer was calibrated with a 100x objective using a silicon sample which shows a clear recognisable peak at 520 cm−1. The calibration was done automatically by the LabSpec 6 software used with the LabRAM. The cell was placed in plastic holder on a movable platform below the objective and the laser beam was focused on the surface of the separators by moving the platform and cell along the three axes. A camera behind the objective fed image to the LabSpec 6 and making it possible to see where the focus of the laser was. Given the transparent nature of the Celgard it was possible to focus both on the top surface of the Celgard and then adjust the focus further down. Images from the spectrometer camera can be seen in Fig. 3.6. The texture of the Celgard was finer with smaller fibres while the glass fibre had a shinier surface with larger fibres. The wet glass fibre was however not transparent enough to find a focus further down. In the case of two stacked, wetted, Celgards the metal surface of the Raman cell could be seen through the separators. Before pressing the stack of Celgard, both surfaces could be distinguished and focused on. When pressed, they stuck together and to the bottom of the glass pane, making focusing harder. As a solution, the focus was placed just below the bottom of the glass which had a distinct structure making it easy to locate. The measurements were taken with the doors of the spectrometer closed. 33 3. Immiscibility and diffusion of electrolytes Figure 3.6: The measurement platform with a plastic holder for the cell. Camera images from the spectrometer of Celgard on glass fibre, both separators wetted. 3.4 Results Four setups were measured, a glass fibre/glass fibre and Celgard/Celgard stack for DBE and [PYR14][TFSI], and a glass fibre/glass fibre and Celgard/glass fibre stack for [P66614][TFSI] and [Emim][TFSI]. All spectra were processed before being plotted. A background function was fitted individually to each spectrum and the intensity of the peaks scaled the same for easy comparison, while small lateral shifts were added to distinguish between spectra more easily. The relative intensity of the peaks is dependent on the concentration as well as the scattering properties of the molecules [66]. LiTFSI was present in all electrolytes although sometimes emitted in the captions or legends due to lack of space. Moreover, areas of interest in the spectra are the locations of peaks that only appear for one of the electrolytes as that will make it possible to differentiate between the molecules present. Some peaks the electrolytes will share due to it being a bond shared by molecules in the solvents, such as methyl (-CH3), or the functional groups of the LiTFSI. Therefore, the plotted graphs are split into smaller ranges, excluding areas where no peaks are visible, or where the peaks are located at the same shift. 3.4.1 Baseline fit Due to fluorescence and the inelastic scattering being a relatively weak signal, there is often a significant baseline in Raman spectra. However, since this background is added to the signals from the Raman scattering, it does not hide the peaks but only adds to them. Therefore, it is relatively simple to remove the background and make the peaks more visible. In this work a 4th degree Huber function was fitted to the background using the MATLAB function backcor.m [71]. The chosen function gave the closest fit when compared to other functions and degrees. A baseline fit of a measurement of 1 M LiTFSI in [P66614][TFSI] can be seen in Fig. 3.7 34 3. Immiscibility and diffusion of electrolytes Figure 3.7: Correction of the measured Raman spectra of 1 M LiTFSI in [P66614][TFSI] by fitting a 4th degree Huber function to the background. 3.4.2 [Emim][TFSI] and [P66614][TFSI] The individual electrolytes were measured with the 50x objective as it was easier to focus on the bulk liquids. It proved difficult however to find and focus on the surfaces of the separators, leading to the 10x objective being used for all surface measurements. The spectra of measurements made of the glass fibre stack can be seen in Fig. 3.8. Measurements of an unwetted glass fibre was also taken but practically no peaks were visible, hence it is not included. Given that [Emim][TFSI] has a higher density, it was applied to the bottom glass fibre. Without pressing, the measured spectrum looks practically identical to [P66614][TFSI], indicating no immediate diffusion of [Emim][TFSI] to the top. Likewise, after inserting the spring and focusing on the glass fibre surface pressed against the glass pane, no traces of [Emim][TFSI] could be detected. The cell was thereafter left for 24 hours before repeating the last measurement. At 600 cm−1 and 1025 cm−1 there might be some small traces of [Emim][TFSI] peaks, although not clear enough to be certain it is peaks forming. When the separators were soaked the high viscosity electrolytes took a long time to reach the edges of the glass fibre. During assembly it was suspected that the [P66614][TFSI], did not entirely reach the edges of the separator before it was placed on top of the [Emim][TFSI]. In consideration of this, a final measurement of the edge of the separator was done. Contrary to the other measurements, here the presence of [Emim][TFSI] is obvious as well as the presence of [P66614][TFSI]. Based on the improper wetting during the assembly, the most probable cause would be the diffusion of lower viscosity [Emim][TFSI] up into the empty pores at the edge. 35 3. Immiscibility and diffusion of electrolytes Figure 3.8: Raman spectra of a 1 M LiTFSI in [P66614][TFSI]-soaked glass fibre separator in contact with a 1 M LiTFSI in [Emim][TFSI]-soaked glass fibre separator, with the former being on top. As for the previous separator setup, the 10x objective was used to measure the [P66614][TFSI]-soaked Celgard against the [Emim][TFSI] glass fibre separators (Fig. 3.9). Contrary to glass fibre, Celgard has a distinct Raman spectrum with few peaks coinciding with [Emim][TFSI] or [P66614][TFSI]. At 740 cm−1 and 1250 cm−1 two clear peaks are visible, belonging to LiTFSI. Given that [Emim][TFSI] cannot wet the Celgard on its own, the diffusion into the top separator is presumed unlikely. Since the Celgard became close to transparent when wetted, focus was placed on both the Celgard and glass fibre surface. However, these two measurements, taken before pressing the separator have an almost identical appearance. Furthermore, the peaks both coincide with the peaks for Celgard and [Emim][TFSI], but not [P66614][TFSI], which is unexpected. The occurrence of [Emim][TFSI] in both these measurements strongly suggests that the depth resolution in wetted Celgard is wider than the thickness of the separator. The measurement after pressing only brought the stack closer, resulting in this spectrum being identical to the two measurements of the unpressed stack. The hypothesis that it is the transparency of the Celgard which is responsible for the visibility of [Emim][TFSI] and not diffusion, is seemingly confirmed by the measured spectrum of the Celgard separator on its own. After opening the cell under a fume hood, carefully removing the bottom glass fiber and placing the Celgard back in the cell without the spring, a final measurement was made. Evaporation of the electrolyte was not deemed an issue simply due to the low vapour pressure of ILs, 36 3. Immiscibility and diffusion of electrolytes meaning the probability of molecules escaping the liquid is very small. This spectrum showed almost exclusively Celgard peaks while there were no traces of [Emim][TFSI] since the 600 cm−1 and 1425 cm−1 peaks disappeared after the removal of the glass fibre. Instead, some additional peaks appeared, for instance at 1440 cm-1 and 1300 cm-1 that does not seem to belong to Celgard but [P66614][TFSI]. However, the origin of peak at 1160 cm−1, is still unclear. Possibly, the metal piston visible through the electrolyte could be the cause. Figure 3.9: Raman spectra of a 1 M in LiTFSI in [P66614][TFSI]-soaked Celgard separator in contact with a 1 M LiTFSI in [Emim][TFSI]-soaked glass fibre separator, with the former being on top. 3.4.3 [PYR14][TFSI] and DBE Analogous to the first electrolyte pairs, the Raman cell was assembled by first wet- ting a glass fibre placed on the metal piston with [PYR14][TFSI]. Thereafter another glass fibre was entirely wetted with DBE before being placed on top of the first. As for the other pair, all surface measurements were made with the 10x objective and the electrolytes alone with the 50x. In comparison with the ILs however, the spectra for DBE as well as the spectra containing it, appeared noisier. Without pressing, the measured spectra seem to only contain DBE. The [PYR14][TFSI] peak at 900 cm−1 is missing while the DBE peaks at 850 cm−1, 1300 cm−1 and 2900 cm−1 are clearly visible. However, after pressing the separators these three peaks immediately disap- pear and th