In-Depth Analysis of Irreversible Processes in Lithium Ion Batteries, That Is, the Dependence of Irreversible Charge Losses on the Depth Inside the Electrode

Home , Harry Kroto

IN-DEPTH ANALYSIS OF IRREVERSIBLE PROCESSES

IN LITHIUM ION BATTERIES

DISSERTATION

SUBMITTED FOR THE DEGREE OF

DOCTOR OF NATURAL SCIENCES (DR. RER. NAT.)

FAKULTÄT FÜR CHEMIE UND BIOCHEMIE

ANALYTISCHE CHEMIE – ELEKTROANALYTIK & SENSORIK

BOCHUM, MAY 2013

STEFAN KLINK

The work presented in this thesis was carried out during my doctoral studies from August 2009 to April 2013 in the group of Prof. Dr. Wolfgang Schuhmann, Analytical Chemistry, Ruhr-University Bochum.

Gas phase oxidation of carbon nanofibres incl. FTIR and XPS analysis was performed by Dr. Wei Xia in the group of Prof. Dr. Martin Muhler, Chair of Technical Chemistry. FEM simulations were performed together with Dr. Daniel Höche at the Helmholtz-Zentrum Geesthacht. CAD reproduction and fine- mechanical works were performed by Armin Lindner and his team. Parts of this thesis have been published as research articles or are currently in the process of being published.

Date of submission 06.05.2013

Date of defence 05.07.2013

Chair of examination board Prof. Dr. M. Wark

First supervisor Prof. Dr. W. Schuhmann

Second supervisor Prof. Dr. M. Muhler

ABSTRACT

The current interest in electric vehicles and systems for stationary energy storage imposes high requirements on lithium ion batteries (LIB) since they are the most promising battery system in terms of energy density and cycle life. Fundamental understanding of electrochemical processes such as irreversible charge losses (ICL) is necessary to guide materials research and develop the next generation of batteries. In my PhD work, I studied the development of ICL, a key factor for a long cycle life of LIB, in conventional graphite electrodes as well as in carbon nanofibres (CNF) as innovative battery materials. By oxidizing CNF in the gas phase the amount of ICL was significantly decreased due to a change of the surface oxygen chemistry (chapter 4.1).

Comprehensive understanding of ICL, however, is only possible with advanced methods of electrochemical analysis. Electrochemical Impedance Spectroscopy (EIS) is a powerful tool for the investigation of electrochemical systems, but EI spectra recorded in the commonly used three- electrode test cells are heavily distorted due to geometrical and electrochemical asymmetries. As a solution to this drawback, a new coaxial impedance cell (CIC) was developed, in which the distribution of current lines was optimized by moving the reference electrode to a coaxial position and precisely aligning the electrodes (chapter 4.2). As supported by simulations based on the Finite Element Method (FEM, chapter 4.3), EIS in these cells yields reliable spectra up to frequencies of 50 kHz, regardless of the electrode configuration.

Electrochemical reactions in porous electrodes are characterized by mass-transport. Using a multi- layered electrochemical cell (MLC), the vertical distribution of ICL within a porous graphite battery electrode was investigated (chapter 4.4). It was shown for the first time that, due to mass-transport limitations, solvent reduction and electro-polymerization of the SEI forming additives during the formation of the solid-electrolyte interphase (SEI) on the graphite surface occurs vertically distributed and more pronounced in the surface layer than in the lower layers.

ABOUT THE COVER PICTURE

The scheme is a depiction of the vertical distribution of ionic current over time during electrochemical lithiation of a graphite negative electrode used in lithium ion batteries. The lines correspond to the relative current density distribution within a 6-layered multiple working electrode as measured by six current followers during galvanostatic cycling as described in chapter 4.4 of this thesis.

THANK YOU!

“It's a dangerous business, Frodo, going out your door. You step onto the road, and if you don't keep your feet, there's no knowing where you might be swept off to.”

[1] - BILBO BAGGINS, The Lord of the Rings

I would like to express my gratitude to Prof. Dr. Wolfgang Schuhmann for accompanying me on a journey that started seven years ago, spanning from sensors to fuel cell catalysis to batteries. During this time, Wolfgang gave me the freedom to give many things a try and taught me to not care about what he thinks. His unconventional approach to things, the consequent living of flat hierarchies and his motivation to take over responsibility is what made this journey a most enjoyable time. I am deeply indebted to him for placing so much confidence in me, but most important, for conveying the passion for science.

I also would like to sincerely thank Prof. Dr. Martin Muhler of the Laboratory of Technical Chemistry for kindly taking over the co-supervision of my work for the past four years, for valuable discussions and advice and encouraging me to do research with members of his research group. I am particularly grateful to Dr. Wei Xia for the good collaboration that was always marked by kindliness and respect.

Without Dr. Fabio La Mantia and Dr. Edgar Ventosa, this thesis would not have been possible in its present form. I am deeply grateful for receiving their invaluable advice in the last four years and for all the fun we had together.

Cordial thanks go to Armin Lindner and his team of the fine mechanical workshop for the friendly collaboration and the skilful and swift manufacture of all these little gadgets.

My most profound thanks go to all members of the SPP1473, especially to Prof. Dr. Alfred Ludwig and Sara Borhani-Haghighi for the good and fruitful collaboration. Many thanks also to Dr. Daniel Höche for the good and friendly teamwork and for simulating dozens of impedance spectra.

I like to thank the INTERNATIONAL ASSOCIATION FOR THE

EXCHANGE OF STUDENTS FOR TECHNICAL EXPERIENCE (IAESTE) and the IAN WARK RESEARCH INSTITUTE for funding my research stay at the University of South Australia, the GESELLSCHAFT DEUTSCHER CHEMIKER (GDCH) for several travel grants, and the RUHR-UNIVERSITY RESEARCH SCHOOL for admission to their stipend program and long lasting care of my PhD education including travel grants to excellent conferences and in- house workshops “outside the box”.

III

Special thanks go out to all the people who helped me with correcting, grammar checking, spell- checking, galley-proofing and generally improving the quality of journal articles, talks and of course this thesis: Alberto Battistel, Marie-Soleil Beaudoin, Jillian Healey, Dr. Manuela Jüstel, Dr. Björn Klink, Dr. Fabio La Mantia, Dr. Justus Masa, Nora Poschmann, Dr. Rosalba Rincon-Ovalles, Briana Shymkus, Wibke Wesener and Dr. Edgar Ventosa.

During the last four years, I had the opportunity to meet and work with many good people and have inspiring discussions. I would like to thank Prof. Dr. Martin Winter, Dr. Miriam Kunze and Christian Dippel for introducing me to lithium ion batteries, for friendly discussions and for the opportunity to look over their shoulders. I am grateful to Stefanie Grützke for Raman measurements & experiments. Furthermore I would like to thank Hervé Bonin for interesting help and discussions about electrochemical instrumentation. Special thanks go to Manuela Jüstel for pleasant discussions and sharing the joy of science, to Ou Jin and Dr. Kathrin Eckhardt for opening doors, and Martin Tagoe for providing a humanistic view on things. Finally, I like to send warmest regards to Prof. Dr. John Ralston and Dr. Craig Priest for hosting me in Adelaide.

My cordial thanks go to all ELANos, those who made my every day with coffee, cake, kicker, football and more, especially Alberto Battistel, Andrea Contin, Lisa Feldhaus, Stefanie Grützke, Ramona Gutkowski, Dr. Edyta Madej, Dr. Justus Masa, Dr. Michaela Nebel, Dr. Andrea Puschhof, Lutz Stratmann, Giorgia Zampardi and Dr. Aleksandar Zeradjanin. I cannot praise too highly the good souls of Sandra Schmidt, Bettina Stetzka and Dr. Thomas Erichsen for their long-lasting and amicable support in so many aspects. Thank you, Sascha, for covering fire, many good times with science, Mensa, coffee and kicker and company during our PhD.

The last lines shall be reserved for the most important people in my life. I dedicate this success to you those who provided me with faith, scolding, good will, love, money, friendship and the trust there is always someone I can lean on to. My friends, my parents Heinz and Jeanette, my sisters Denise and Sylvia, my brother Björn and Wibke.

TABLE OF CONTENTS

ABSTRACT ...... II

ABBREVIATIONS AND GLOSSARY ...... 3

1 INTRODUCTION ...... 8

2 STATE OF THE ART AND PROBLEM IDENTIFICATION ...... 10

2.1 FUNDAMENTALS OF ELECTROCHEMICAL SYSTEMS ...... 10 2.1.1 Flux of charge and transport processes ...... 10 2.1.2 Kinetic control – the overpotential ...... 12 2.2 LITHIUM ION BATTERIES ...... 13 2.2.1 History of lithium ion batteries ...... 15 2.2.2 Structure and components ...... 16 2.2.3 Electrochemical analysis of lithium ion batteries ...... 24 2.2.4 Thermodynamic and kinetic aspects ...... 26 2.3 DISTRIBUTION OF CURRENT AND CHARGE IN POROUS ELECTRODES ...... 31 2.4 CARBON NANOTUBES AND NANOFIBRES ...... 34 2.4.1 Application to lithium ion batteries ...... 35 2.4.2 Characterization methods ...... 38 2.5 ELECTROCHEMICAL IMPEDANCE SPECTROSCOPY ...... 40 2.5.1 Introduction to impedance spectroscopy ...... 40 2.5.2 Impedance spectroscopy for electrochemical systems ...... 46 2.5.3 Electrochemical impedance spectroscopy of lithium ion batteries ...... 53 2.5.4 Experimental considerations ...... 58

3 AIM OF THIS STUDY ...... 63

4 RESULTS AND DISCUSSION ...... 65

4.1 IRREVERSIBLE CHARGE LOSSES IN OXIDISED CARBON NANOFIBRES ...... 65 4.1.1 Morphology and degree of graphitization ...... 66 4.1.2 Oxygen content and surface concentration ...... 68 4.1.3 Irreversible charge losses...... 71 4.1.4 Reversible capacity ...... 74 4.1.5 Discussion ...... 75 4.2 DEVELOPMENT OF A COAXIAL IMPEDANCE CELL FOR LITHIUM ION BATTERIES ...... 77 4.2.1 Initial situation ...... 77 4.2.2 Cell development ...... 78 4.2.3 Discussion ...... 83 4.3 RELIABILITY LIMITS OF COAXIAL IMPEDANCE CELLS FROM FEM MODELLING ...... 85 4.3.1 Development of the FEM model ...... 86 4.3.2 Identification of critical parameters ...... 90 4.3.3 Discussion ...... 96 4.4 VERTICAL DISTRIBUTION OF CURRENT, CHARGE AND CHARGE LOSSES ...... 97 4.4.1 Development of a multiple working electrode ...... 99 4.4.2 Development of a multi-layered electrochemical cell ...... 100 4.4.3 Current distribution in graphite electrodes ...... 102 4.4.4 SEI formation and irreversible charge losses ...... 107 4.4.5 Towards vertically distributed impedance spectroscopy ...... 111

5 CONCLUSIONS AND OUTLOOK ...... 115

6 EXPERIMENTAL SECTION ...... 119

6.1 CELL MANUFACTURE AND ASSEMBLY ...... 119 6.1.1 Three-electrode Swagelok T-cell ...... 119 6.1.2 Coaxial impedance cell ...... 120 6.1.3 Multi-layered electrochemical cell...... 122 6.2 CARBON NANOFIBRES ...... 124 6.2.1 Oxidation of CNF ...... 124 6.2.2 Instrumental analysis ...... 125 6.3 ELECTRODE PREPARATION...... 126 6.3.1 Carbon nanofibre electrodes ...... 127

6.3.2 LiFePO4 electrodes ...... 128 6.3.3 Graphite mesh electrodes ...... 129

6.3.4 Li0.5FePO4 reference electrodes ...... 130 6.4 ELECTROCHEMISTRY ...... 131 6.4.1 Electrochemical charge/discharge of CNF electrodes ...... 131

6.4.2 EIS on LiFePO4 electrodes ...... 131 6.4.3 Charge/discharge of multiple working electrodes ...... 132 6.5 SOFTWARE...... 133

7 BIBLIOGRAPHY ...... 134

7.1 SOME NOTEWORTHY PUBLICATIONS AND RESOURCES ...... 134 7.2 LITERATURE ...... 135

8 APPENDIX – TECHNICAL DRAWINGS ...... 152

8.1 SWAGELOK THREE-ELECTRODE T-CELL ...... 152 8.2 COAXIAL IMPEDANCE CELL ...... 155 8.3 MULTI-LAYERED ELECTROCHEMICAL CELL ...... 163

ABBREVIATIONS AND GLOSSARY

GENERAL TERMS

3C Computer, communications and consumer electronics a major market for small lithium ion batteries ac Alternating current AFM Atomic force microscopy uses a tip to scan the topology of samples surfaces in the nm-range AN Anode an electrode, where oxidation takes place, i.e. the flux of electrons from the reactant into the electrode[3] in batteries, the negative electrode is usually called anode, which is formally true during discharge ASA Active surface area 2 -1 [m g ] measured by O2-chemisorption Battery formally: a stack of galvanic cells primary battery: a galvanic cell, which transforms chemical energy into electrical energy once secondary battery: an electrochemical system, which reversibly stores and releases electrical energy BET Brunauer-Emmett-Teller surface area 2 -1 [m g ] surface area measured by N2-physisorption C65 Conductive carbon (Super P™) a battery grade, nano-sized conductive carbon black (“soot”) from Timcal[4] CA Cathode electrode, where reduction takes place, i.e. the flux of electrons from the electrode into the reactant[5] in batteries, the positive electrode is usually called cathode CE Counter electrode the second electrode in any electrochemical cell, providing or accepting electrons for the WE CNF Carbon nanofibres bamboo-like →MWCNT with tilted graphene cups, typically slightly larger than MWCNT CNLS Complex non-linear least squares a mathematical fitting routine for →EIS CNT Carbon nanotubes a “one-dimensional” allotrope of carbon, consists of graphene layers rolled up in tubes CPE, Q Constant phase element a frequency-dispersive capacitor used for →EIS simulations C-rate Charge rate [C/x] current at which a battery (electrode) (dis)charges in 1/x hours CV Cyclic voltammetry a potentiodynamic method similar to LSV; the potential sweep is reversed after reaching the set potential CVD Chemical vapour deposition a chemical process in which volatile precursors react on a substrate surface to form a solid material DEC Diethyl carbonate a low viscosity solvent used in LIB electrolytes E1, E2 Electrode 1, Electrode 2 EC Ethylene carbonate a cyclic, high dielectricity solvent used in LIB electrolytes, solid at room temperature EIS Electrochemical impedance spectroscopy electroanalytical tool to study the ac resistance of an electrochemical system over a range of frequencies Electrolytic cell consumes work to drive electrochemical reaction, the opposite of a galvanic cell EQC Equivalent circuit an electric circuit which yields the same impedance of a real system at every frequency, used for fitting (H)EV (Hybrid) electric vehicle a vehicle that is (partially) powered by electricity

FEM Finite element method numerical technique to solve partial differential equations in a three-dimensional space FRA Frequency response analyser compares the sine response ̃ of a system with two reference signals ̃ and ̃ ) to calculate Galvanic cell an electrochemical cell which produces work, the opposite of an →electrolytic cell GEIS Galvanostatic electrochemical impedance spectroscopy EIS technique; perturbs the electrochemical system by a sine current and measures the voltage response GCPL Galvanostatic cycling with potential limitation constant current potential-time-curves used to determine charge/discharge capacities of LIB ICL Irreversible charge losses [mAh/g] charge losses in a LIB, e.g. due to surface group reduction, SEI formation or exfoliation Li Lithium the lightest metal in the periodic system with the lowest electrochemical potential LIB Lithium ion battery a secondary battery which consists of two electrodes with lithium ion redox-couples LFP Lithium iron phosphate LiFePO4 low voltage (ca. 3.4 V) but inexpensive positive electrode material for LIB with olivine structure LP40 Merck Selectilyte LP40 [6] a ready-to-use electrolyte for LIB; consists of 1 M LiPF6 in EC:DEC 1:1 wt.-% LPF Lithium hexafluorophosphate LiPF6 inorganic lithium salt used as ionic charge carrier in LIB electrolytes LSV Linear sweep voltammetry a potentiodynamic method; records current during linear potential ramp of the electrode over time M Warburg element for finite diffusion MLC Multi-layered electrochemical cell developed in this work to investigate the vertical distribution of currents, charge and →ICL in →LIB MWCNT Multi-walled carbon nanotubes →CNT which sidewalls consist of several layers of graphene MWE Multiple working electrode a stack of thin, mesh supported working electrodes, electrically isolated by a thin polymer separator NMP N-methyl pyrrolidone Typical solvent for →PVdF used as binder in →LIB electrode slurries OCP Open circuit potential [V] potential of an electrode, when no current is flowing ppm Parts per million measure for the purity of solvents, atmospheres etc. PC Propylene carbonate high dielectricity solvent used in →LIB electrolytes; has a strong tendency to exfoliate graphite PCGA Potentiostatic cycling with galvanostatic acceleration a step potential sweep in which the next step is applied as soon as the current has dropped, →PITT PE Polyethylene used in LIB separators, “shuts down” during thermal runaway PEEK Polyether ether ketone an organic, thermoplastic polymer with excellent chemical resistance that can be used for LIB test cells PEIS Potentiostatic electrochemical impedance spectroscopy perturbs the electrochemical system by a sine voltage and measures the current response PITT Potential intermittent titration technique charges a lithium ion battery by small charge increments and records the →OCP Polarization potential drop of an electrode due to the passage of current[7] PP Polypropylene used in LIB separators, no shrinkage compared to PE PVdF Polyvinylidene difluoride chemically and electrochemically resistant binder used in LIB electrodes

Randles circuit a simple equivalent circuit used to describe the impedance behaviour of an electrode R Resistor used in →equivalent circuits to model electrolyte or charge-transfer resistance RE Reference electrode electrode with a well-defined and stable potential RT Room temperature typically 20 °C SEM Scanning electron microscopy uses an electron beam with a wavelength shorter than light for increased resolution (Abbe limit) S5130 Solef S5130 a commercial →PVdF binder used in LIB SFG6 Graphite natural flake graphite from Timcal; particle size below 6 µm; used as LIB anode material SFG44 Graphite natural flake graphite from Timcal; particle size below 44 µm; used as LIB anode material SEI Solid-electrolyte interphase passivating layer due to electrolyte reduction formed on graphite surfaces SoC State of charge [%] fraction of charge available for discharge of an electrode SHE Standard hydrogen electrode standard reference electrode based on the redox potential of hydrogen at 1 atm H2; pH 0 and RT TLM Transmission line model →EQC first applied to telegraph transmission lines and adopted to many fields incl. →LIB, Fig. 2-13, p. 32 V4A Stainless steel VC Vinylene carbonate an electro-polymerisable electrolyte additive used in →LIB electrolytes as →SEI forming agent W Warburg element a →CPE with a 45° phase angle, used to simulate diffusion in →EQC WE Working electrode the electrode of interest in an electrochemical cell XPS X-ray photoelectron spectroscopy detects energy of emitted electrons after x-ray irradiation to investigate surface group binding energies ZRA Zero resistance amperometry applies zero volt between WE and CE and measures current and potential; used as “current follower”

SYMBOLS

C Electrical capacity [F=As V-1] C0 Double-layer capacitance of current collector [F=As V-1] CF Faradaic charge capacity -1 [F=As V ] used in equivalent circuits to describe the limited lithium intercalation capacity of electrode, also: Qrev CDL Capacity of the electrochemical double-layer -1 [F=As V ] used in equivalent circuits; in parallel to the charge-transfer resistance Rct C eff Coulombic efficiency [%] the quotient of lithium extraction to lithium insertion, ideally reaches 100 % after a few cycles (∇)ci Concentration (gradient) of species i [m2 s-1] de Thickness of electrode [m] Di Diffusion coefficient of species i [m2 s-1]

E Electrode potential [V] if not stated otherwise, all potentials are reported vs. Li+/Li pseudo-reference electrode Dielectric constant [dim. less ] defines the strength of a solvent to solvate ions η Total overpotential [V] Magnitude of potential drop due to the passage of current ηs Surface overpotential [V] potential drop of an electrochemical reaction caused by its kinetics of charge-transfer ηc Concentration overpotential [V] combination of change in OCP and ohmic drop induced by concentration gradient in the electrolyte F Faraday constant -1 [C mol ] from NIST[8] fr Relaxation frequency [Hz] characteristic frequency of a →Randles circuit, highest point of the semicircle i Current density of electrons [A cm-2] ̃ Transform of current [A] itotal Total current density [mA g-1] of the →MWE as defined by the master galvanostat iL1 – iL6 Cross current density [mA g-1] of electrochemical reaction in each layer of the →MWE j Ionic current density [A cm-2] Flux of charge, i.e. flow rate of positive charge per unit area perpendicular to flow direction[7] Ni Ionic conductivity [mol cm-2s-1] Inverse of the resistivity, typical value for a battery electrolyte is ~1 S m-1 (R=10 Ω; A=1 cm²; d=100 µm) Ndiff Flux density of species i [mol cm-2s-1] Vector quantity of species flux per units area perpendicular to flux direction Nconvect Flux density due to diffusion [mol cm-2s-1] Nmigration Flux density due to convection [mol cm-2s-1] (∇)Φ Potential (gradient) [V] ΔΦohm, Ohmic drop or ηohm[V] Potential drop due to current passing through an electrochemical system QICL Charge capacity of a battery [mAh g-1] the physically correct term, however, would be charge, as capacity is defined as dQ/dE. 1 mAh= 3.6 As Qrev Irreversible charge losses [mAh g-1] charge that is consumed due to irreversible reactions such as the →SEI formation or exfoliation RCT Reversible charge [Ω] charge of reversible lithium insertion and extraction Re Charge-transfer resistance of an electrochemical reaction [Ω] Rion,p Electric resistance of the solid phase in in a porous electrode [Ω] Rion,s Ionic resistance of pores in a porous electrode [Ω] Rion,CE Ionic resistance of electrolyte in the →MWE separators [Ω] E Ionic resistance of separator close to the CE [V]

Re,cc Electric resistance of the current collector [Ω] RL Charge-transfer resistance of leakage current (side reactions) [Ω] R Gas constant -1 -1 [J mol K ] from NIST[9] Rct Charge-transfer resistance [Ω] of an electrochemical reaction κ or σel Ionic conductivity [S m-1] Inverse of the resistivity, typical value for a battery electrolyte is ~1 S/m (R=10 Ω; A=1 cm²; d=100 µm) T Temperature [K] [°C] ui Migration mobility [cm2∙mol/J∙s] Mobility of charged species in an electric field ̃ Transform of voltage [V] Wssd Warburg element for solid state diffusion

Δx Electrode shift in x direction [m] zi Charge number of the species i

Zs Impedance of electrode surface

ZF Impedance of faradaic reactions

ZW Warburg impedance Impedance contribution due to limited diffusion of educts to the electrode

7 1 INTRODUCTION

1 INTRODUCTION

"Electricity," said the old gentleman, sagely, "is destined to become the motive power of the world. The future advance of civilization will be along electrical lines. "[…] "And in the meantime," said the mother, despairingly, "we shall all be electrocuted, or the house burned down by crossed wires, or we shall be blown into eternity an explosion of chemicals!"

- L. FRANK BAUM (1901), The Master Key[10]

L. FRANK BAUM was writing these lines at a time when CAMILLE JENATZKI drove his electric LA JAMAIS

[12] CONTENTE to new world records , and the first, battery powered, WIRELESS TELEGRAPH OUTFITS were rolled out at a price of 8.50 $ (an equivalent value of about 200 $ now)[13]. Today, “portable telegraph outfits” have evolved significantly, but the motive power of most vehicles is still (or again?) based on the combustion of fossil fuels.

Nevertheless, electric vehicles finally offer, more than a century later, a practical solution for sustainable mobility. The economic success of electro-mobility is strongly linked to the performance, reliability and cost of the storage battery. The most promising candidate for this purpose are

LITHIUM ION BATTERIES (LIB), which outperform Ni-Cd and NiMH batteries in terms of energy density and cycle life, with energy densities approaching 200 Wh kg-1, and a life expectancy of up to thousands of charge/discharge cycles. FIG. 1-1 1888 FLOCKEN ELEKTROWAGEN, THE These numbers are the result of four decades of FIRST ELECTRIC VEHICLE THAT BECAME APPARENT IN [11] GERMANY (RECONSTRUCTION BY FRANZ HAAG) interdisciplinary research efforts by electrochemists, materials scientists and engineers alike.

The rise of electric vehicles and the need for green energy storage, however, impose even higher requirements on batteries, which can only be fulfilled by new materials, new electrolytes and

[14] optimized electrode architectures. As MAX PLANCK said: “insight must precede application” ; a fundamental understanding of crucial electrochemical processes, especially at the interface between the active material and electrolyte, is necessary to guide materials research towards next generation batteries. Recognition and assessment of aging and failure mechanisms, though, is only possible with appropriate analysis of physical and electrochemical processes associated with the formation of the

SOLID-ELECTROLYTE INTERPHASE (SEI).

8 1 INTRODUCTION

These processes can be distributed both in time (due to their specific relaxation time constants) and space (due to mass transport phenomena in the porous electrode). The development and application of electroanalytical methods for in-depth-analysis of irreversible processes in lithium ion battery materials in both dimensions is the objective of this study. Knowing where charge losses occur can help to understand aging and to engineer better battery electrode architectures with longer lifetime and less capacity fading.

9 2 STATE OF THE ART

2 STATE OF THE ART AND PROBLEM IDENTIFICATION

2.1 FUNDAMENTALS OF ELECTROCHEMICAL SYSTEMS

An ELECTROCHEMICAL CELL is a system in which electrical energy is converted to chemical energy

(ELECTROLYTIC CELL) or, vice versa, chemical energy is converted to electrical energy (GALVANIC CELL). It contains two ELECTRODES which are ionically connected by an ELECTROLYTE and electrically connected by an external conductor. An electrode is an electron conductor which is in contact with an electrolyte (i.e. an ion conducting phase). In particular, an electrode participates in electrochemical reactions, i.e. a CHARGE-TRANSFER between an electrode and an electrolyte. In most cases, the electrode is a solid while the electrolyte is liquid. Exceptions however exist in both cases such as liquid mercury electrodes and solid electrolytes. The distinct feature of an electrochemical reaction is, unlike a chemical redox-reaction that reduction and oxidation reactions are spatially separated from each other. Oxidation takes places at the ANODE and reduction takes place at the CATHODE. Since the electrolyte allows movement of ions but is non-conductive to electrons, electrons are required to move through an external circuit under consumption or provision of work. The direction of current flow is defined by the THERMODYNAMIC properties of the electrode, expressed as ELECTROCHEMICAL

POTENTIAL. The CURRENT DENSITY, i.e. the flux of charge, is defined by the KINETICS of the electrochemical system. In particular, these include the kinetics of CHARGE-TRANSFER (surface kinetics) and MASS-

TRANSPORT (MIGRATION, DIFFUSION and CONVECTION). OVERPOTENTIAL is the potential drop of an electrode potential due to the passage of current[7,15].

2.1.1 FLUX OF CHARGE AND TRANSPORT PROCESSES

In an electron conductor, the flux of electrons (current density i) is equal to

(Ohm’s law) EQ. 2-1 where is the conductivity of the electron conductor and ∇ is the potential gradient. In the electrolyte, charge is not carried by electrons but by the flux of charged species (i.e. ions):

∑ EQ. 2-2

where zi is the charge number of species i, F is the Faraday constant and Ni is the flux density of species i. The movement of charged species in the electrolyte is subjective to different driving forces. These are i) convection; ii) migration and iii) diffusion.

10 2 STATE OF THE ART

CONVECTION is the movement of the bulk liquid carrying charged species with it. It can occur due to physical effects related to the electrochemical reaction (e.g. by change of electrolyte density) or it is forced externally, e.g. by rotating electrodes. The flux by convection Nconvect is therefore defined as the product of concentration ci and bulk velocity 휈:

EQ. 2-3

MIGRATION describes the movement of a charged species due to an electric field. Because the velocity of migration υi depends on the force of the electric field ∇ϕ, the charge of species zi and the ion mobility ui, the flux of migration of species i is defined as:

EQ. 2-4

Summing up the migration fluxes of all species in the electrolyte, the IONIC CONDUCTIVITY κ can be expressed according to an electron conductor:

EQ. 2-5

The current of charged species under an electric field is similar to Ohm’s law for electric currents:

(Ohm’s law) EQ. 2-6

DIFFUSION is the movement of species i due to a concentration gradient ∇ci:

EQ. 2-7 where Ndiff is the diffusional flux and Di is the diffusion coefficient of species i.

Depending on the electrochemical system, the vector quantities of migration and diffusion can point in the same or in opposed directions. The total flux density of migration, diffusion and convection is then given by:

EQ. 2-8

The overall current density of the charge transported via ions is expressed as:

∑ EQ. 2-9

11 2 STATE OF THE ART

2.1.2 KINETIC CONTROL – THE OVERPOTENTIAL

While the overall open circuit voltage U, i.e. the potential difference between two electrodes when no current is flowing, is determined by the thermodynamics of each electrode, the passage of current will induce a potential drop or overpotential η that leads to the apparent cell voltage V:

EQ. 2-10

The TOTAL OVERPOTENTIAL η is a combination of several contributions:

EQ. 2-11 with s = surface overpotential; s = concentration overpotential; ∆ ohm = ohmic drop.

The SURFACE OVERPOTENTIAL or ACTIVATION OVERPOTENTIAL ηs of the charge-transfer reaction relates to the kinetics of the electrode reaction. It is described by the BUTLER-VOLMER EQUATION, which relates the electric current density to the electrode potential:

{ [ ] [ ]} EQ. 2-12

with i0 = EXCHANGE CURRENT DENSITY; a = anodic charge-transfer coefficient; c = cathodic charge- transfer coefficient; R = universal gas constant; F = Faradaic constant; T = temperature.

At low surface overpotentials, it simplifies to:

EQ. 2-13

Thus, at low currents, the surface overpotential increases linearly with current:

EQ. 2-14

where Rct is the CHARGE-TRANSFER RESISTANCE of the electrochemical reaction.

The OHMIC DROP in the electrolyte ΔΦohm depends on the ionic conductivity of the electrolyte as defined in Eq. 2-6. The combined CONCENTRATION OVERPOTENTIAL ηc is induced by a concentration gradient, causing a change in the equilibrium potential (like in a concentration cell) as well as an additional ohmic drop due to the change in ionic conductivity.

CONCENTRATION GRADIENTS can build up in the solid phase of a lithium ion intercalation electrode as well as in the liquid electrolyte of a battery.

12 2 STATE OF THE ART

2.2 LITHIUM ION BATTERIES

“The storage battery is, in my opinion, a catch penny, a sensation, a mechanism for swindling the public by stock companies.”

- THOMAS ALVA EDISON (1883)[13]

PRIMARY BATTERIES are galvanic cells, which transform chemical energy into electrical energy. They are not rechargeable.

SECONDARY BATTERIES are devices for the reversible storage of electric energy as chemical energy. They are galvanic cells during discharge and electrolytic cells during charge. In a typical LITHIUM ION BATTERY (LIB), the chemical energy is stored in two lithium INTERCALATION electrodes such as GRAPHITE and

LITHIUM COBALT OXIDE (LCO). Conduction of lithium ions is realised by an electrolyte consisting of a lithium salt (typically

LITHIUM HEXAFLUOROPHOSPHATE) dissolved in a mixture of organic carbonates such as DIETHYL CARBONATE (DEC) or

ETHYLENE CARBONATE (EC). To avoid electric short circuits, a FIG. 2-1 CUTAWAY VIEW OF A COMMERCIAL porous polyolefin membrane SEPARATOR is placed in between LITHIUM ION BATTERY (COURTESY OF JEWO the electrodes. During charge, the lithium cobalt oxide is BATTERIETECHNIK GMBH) oxidised, and lithium ions are released. Simultaneously, the graphite is reduced and lithium ions are intercalated into the graphite lattice (FIG. 2-2). During discharge, the opposite occurs, and lithium ions are shuttled back to the lithium cobalt oxide. For this reason, LIB are also called ROCKING CHAIR

[16,17] BATTERIES . A cutaway view of a modern laminated cell can be seen in FIG. 2-1.

13 2 STATE OF THE ART

FIG. 2-2 THE ARCHETYPE OF A LITHIUM ION BATTERY: GRAPHITE ANODE AND LICOO2 CATHODE

ANODE OR CATHODE? An anode is an electrode at which oxidation takes place[3]. Thus, the negative electrode (the one with the more negative potential) is an anode only during discharge. Nevertheless, the terms anode and cathode are used among battery scientists for the negative electrode and positive electrode, respectively, independent of the direction of current flow. Unless otherwise stated, this convention will be kept throughout this study.

The ENERGY DENSITY of a battery is defined by its VOLTAGE, i.e. the difference between electrode potentials and by its CHARGE CAPACITY, i.e. the amount of electrical charge which it can supply for doing work. It is usually expressed as Ah or Wh per cell, Wh kg-1 (specific energy density) or Wh L-1 (volumetric energy density). Charge capacities typically range from about 1.5 to 2 Ah for a 18650 cylindrical cell and up to 50 Ah at 3.7 V for a large capacity pouch cell. The specific energy, depending on the cell chemistry and design, ranges between about 70 and 200 Wh kg-1. For an overview of electrode potentials and charge capacities of active materials, refer to 2.2.2.1, p. 17).

The 18650 cell is the most prominent of cylindrical cells and is used for example in laptop batteries. It has a diameter of Ø18 mm and is 65 mm long.

-1 -1 The POWER DENSITY of batteries is often expressed as W kg or W L . It can be divided in constant power and burst power. On the cell level, however, the C-RATE is more informative. It describes how fast a battery is charged or discharged. At C/n, the battery is charged in n hours. The C-RATE CAPABILITY is a function of maximum available capacity versus the C-rate.

14 2 STATE OF THE ART

2.2.1 HISTORY OF LITHIUM ION BATTERIES

Since lithium is the lightest metal in the periodic table with the lowest electrode potential (-3.045 V vs. NHE), its use for batteries was extensively explored during the 1970s. The first primary (i.e. non- rechargeable) lithium batteries that entered the market were lithium-iodine batteries. Those were used for pace makers, outperforming previous state-of-the-art zinc-mercury oxide batteries by four to five times. For consumer electronics, lithium-manganese dioxide batteries entered the market soon after[18]. Around this time fell also the (serendipitous) invention of the lithium thionyl chloride battery (Li-SOCl2(C)) in the group of A. HELLER. First explored for its use as solvent for liquid lasers, it soon revealed that lithium thionyl chloride can be used as liquid cathode[19]. This was possible due to the formation of a SOLID-ELECTROLYTE INTERPHASE (SEI), which was explained in detail in 1979 by

[20] E. PELED (see chapter 2.2.4.3, p. 27). Even today, the Li-SOCl2(C) battery remains the power source with the highest known energy density (760 Wh kg-1, 1420 Wh L-1) with the widest operating temperature range (-50 °C to 70 °C), the longest shelf-life and constant operating voltage[21].

Soon after, it was tried to develop secondary batteries based on lithium. Attempts to recharge the lithium manganese oxide battery, however, resulted in fire hazards and poor cycle life[18]. This motivated researchers to develop stable INTERCALATION cathodes such as LITHIUM COBALT OXIDE (LICOO2, [22] [23] LCO) , which was suggested as electrode material in the group of J.B. GOODENOUGH :

EQ. 2-15

It was only after several fire incidents which occurred even with intercalation cathodes that actually the anode and not the cathode could be identified as source of hazard[18]. The lithium, which was initially believed to redeposit evenly on the anode during charge turned out to form needle-shaped lithium dendrites growing from the anode towards the cathode, eventually piercing through the polyolefin separator and causing local short-circuits. These short circuits lead to sudden high discharge currents accompanied by heat generation and eventual ignition of the organic electrolyte.

This phenomenon is commonly known as THERMAL RUNAWAY. These safety issues could first be diminished by using GRAPHITE as lithium intercalation host in the negative electrode. Graphite reversibly intercalates and deintercalates lithium at potentials just slightly above the redox potential of elemental lithium[18]:

EQ. 2-16

15 2 STATE OF THE ART

CHARGE CAPACITY OF GRAPHITE -1 M(LiC6)= 72 g mol F = 96,485 A s mol-1 Q = F/M = 372 mAh g-1

Accordingly, the first successfully commercialised lithium ion battery (SONY, 1990) utilised a graphite

[24,23] anode and a LCO cathode , see FIG. 2-2.

2.2.2 STRUCTURE AND COMPONENTS

Electrodes for LIB require for the charge and discharge reactions the influx and removal of electrons and lithium ions to and from the active material particles. LIB are therefore made from porous electrodes (see chapter 2.3, p. 31). For the production of LIB electrodes, electrode slurries are prepared by dispersing the particulate active material together with conductive additives in a

[25] solution of polymeric binder . POLYVINYLIDENE DIFLUORIDE (PVDF) copolymers dissolved in N-METHYL

PYRROLIDONE (NMP) are typically used as binder solutions, but environmentally friendly (and cheap!) aqueous binders such as CARBOXYMETHYL CELLULOSE (CMC) or STYRENE BUTADIENE RUBBER (SBR) are currently gradually replacing PVdF binders[27–36]. The slurry preparation itself is very complex and has a large influence on the electrode architecture, distribution of carbon additives, porosity et cetera[30,32,35,37–40]. These slurries are subsequently coated on a current collector foil or mesh and dried.

ALUMINIUM is an inexpensive and light metal that can be used as CURRENT COLLECTOR for cathodes. Oxidative corrosion is sufficiently inhibited by the formation of a passivation layer. At low potentials, however, it alloys with lithium, which precludes its use as negative electrode current collector.

COPPER is used instead for negative electrode current collectors. After the paste has dried on the current collector, a calendering step can be used to compact the electrode paste for better adhesion and conductivity. Cut-to-size anodes, cathodes and separators are then stacked or wound, followed by canning (cylindrical cells) or lamination (pouch cells). FIG. 2-3 illustrates the full process of battery manufacturing including crucial obstacles[29].

16 2 STATE OF THE ART

FIG. 2-3 SCHEME OF BATTERY MANUFACTURING, [29] INCLUDING CRUCIAL PROCESS PARAMETERS, BASED ON

2.2.2.1 ACTIVE MATERIALS

In the past 40 years, extensive research on lithium ion batteries revealed a wide range of possible active materials. The mechanisms behind lithium uptake and release include: i) ELECTROPLATING

[41] (metallic lithium) ii) INTERCALATION into a crystalline host (such as in graphite or lithium cobalt

[22,23,42] [43] [44] oxide ), iii) ALLOYING with metals (such as with tin or silicon ), iv) CONVERSION REACTIONS and v) other types of insertion such as in amorphous carbon.

Intercalation is the insertion of a guest into a crystalline host.

FIG. 2-4 MATERIALS FOR LITHIUM ION BATTERIES

From the many materials discovered, about a dozen have been already commercialized or are in the stage of development. FIG. 2-4 visualises the electrochemical potential and charge capacity of the

17 2 STATE OF THE ART

most important active materials under investigation.

The highest energy densities can be achieved by combining anodes from the lower right corner with cathodes from the upper right corner of the diagram. Typical batteries for the computer, communications and consumer electronics market (known as 3C) therefore still contain

GRAPHITE anodes as they exhibit the lowest intercalation potential of all materials (ca. 0.1 V vs. Li+/Li) and a quite high specific capacity (372 mAh g-1) as well. Emphasis on power density, safety, price, ecology or cycle life, however, may result in other combinations than the

archetype depicted in FIG. 2-2. For reasons of cost, lithium cobalt oxide is often replaced by ternary transition metal oxides of the composition

LiNixCoyMnzO2 with x, y and z close to 0.33. These NICKEL-COBALT-

MANGANESE (NCM) spinels have slightly less energy but superior power and cycle life times compared to LCO[45]. The properties of NCM can be

fine-tuned by changing the composition as depicted in FIG. 2-6.

FIG. 2-6 FINE-TUNING OF NICKEL-COBALT-MANGANESE

LITHIUM IRON PHOSPHATE (LIFEPO4, LFP) suggested by A.K. PADHI and J.B. [46] GOODENOUGH and LITHIUM MANGANESE OXIDE (LIMN2O4, LMO) suggested [47,48] by M.M. THACKERAY and J.B. GOODENOUGH are based on inexpensive

raw materials which renders LFP and LMO batteries attractive for electric vehicle batteries. Batteries using LITHIUM TITANATE (LI TI O , LTO) FIG. 2-5 SPIDER-WEB DIAGRAMS 4 5 12 OF SEVERAL COMMERCIALIZED anodes have relatively low energy densities but operate in ACTIVE MATERIALS electrochemical windows in which the organic electrolyte is perfectly stable (no formation of SEI). For example, batteries made from LTO anodes and LFP cathodes show low volumetric expansion and almost no irreversible charge losses during lithium intercalation, which enables for a cycle life well beyond 10,000 cycles. These properties perfectly match for stationary

18 2 STATE OF THE ART

[49,50] storage applications . FIG. 2-5 visualizes the most important properties of some commercialized active materials. A short overview of important requirements for several markets is given in TAB. 2-1.

TAB. 2-1 REQUIREMENTS OF DIFFERENT MARKETS PUT ON BATTERIES Market C-rate capability Energy density Cycle life 3C electronics ~1 C very important 1000 cycles, 1-2 years Electric vehicles >5 C (HEV) most important > 5.000 cycles, >2 C (EV) >10 years Stationary storage < C/3 almost irrelevant > 10.000 cycles, (home appliance) (1-2 cycles per day) >10 years Stationary storage >1 C, depending on almost irrelevant > 10.000 cycles, (peak shaving) application >10 years

2.2.2.2 CARBON CONDUCTING ADDITIVES

While graphite is sufficiently conductive, literally all cathode materials require CARBON CONDUCTIVE

[51–53] ADDITIVES to provide sufficient electron transport to the site of lithium intercalation . The relation of bulk electrode resistivity versus the mass fraction of conductive carbon follows a sigmoidal shape

(FIG. 2-7). First, the addition of carbon has little effect. When the mass fraction is sufficiently high and

[54] exceeds the PERCOLATION THRESHOLD , there is enough carbon to form a connected, percolating network throughout the electrode. As a consequence, the bulk electrode resistivity decreases rapidly until it slowly reaches the resistivity of the pure carbon (ultimate resistivity of the electrode)[53,55–57].

[53] FIG. 2-7 LEFT: ELECTRODE BULK RESISTIVITY DEPENDENCE ON ADDITIVE FRACTION (REPRODUCED FROM ); RIGHT: PERCOLATION THRESHOLD AS FUNCTION OF ASPECT RATIO; SKETCH OF THE CRITICAL CONCENTRATION [54] FOR THE FORMATION OF A PERCOLATION NETWORK AS FUNCTION OF THE ASPECT RATIO (BASED ON ).

19 2 STATE OF THE ART

Graphite particles guarantee conductivity throughout TAB. 2-2 EFFECT OF CARBON ADDITIVES ON ELECTRODE [4] the whole electrode without too many inter-particle PROPERTIES (BASED ON ) contacts, while nano scale CARBON BLACK (“soot”) is Parameters Graphite Carbon black more suitable for contacting single particles[53]. Electrical Conductivity Inter-particle Graphite and carbon black therefore have conductivity through contact complementary purposes, as summarized in TAB. 2-2 electrode [4] . As depicted on the right of FIG. 2-7, particles with a Ionic Governs Better conductivity porosity absorption of high length-to-diameter ratio need a lower mass electrolyte fraction to form a percolating network[54,58]. Under this Manufacturing Stabilisation Low viscosity aspect, the use of fibrous carbon such as CARBON process of dispersion NANOFIBRES or CARBON NANOTUBES (see chapter 2.4, p. 34) as conductive additive is beneficial and has been commercialised for batteries, e.g. by Showa Denko[59]. As side effect, carbon nanofibres increase the mechanical stability of the electrode[51].

2.2.2.3 ELECTROLYTE AND ADDITIVES

The electrolyte is the medium of ionic conduction. As such, it has a huge impact on battery properties. The rate capability, stability, coulombic efficiency, temperature range and safety of any lithium ion battery are in most cases limited by the composition of the electrolyte. The two main components of liquid electrolytes are i) the salt and ii) the solvent. The SALT supplies the charge carriers needed to cycle lithium ions in between anode and cathode. It needs to dissolve (and dissociate!) in the solvent and should have a high ion mobility to increase its conductivity. The anion should be stable against oxidation and inert versus the solvents and other cell components. For commercial cells, non-toxicity and thermal stability are essential factors. Several salts such as LiPF6,

LiClO4, LiAsF6 and LiBF4 have been explored for batteries, but the aforementioned criteria narrowed down the choice to LiPF6 for commercial systems (see TAB. 2-3). On the laboratory scale, LiClO4 can still often be found as a convenient salt, as it is more robust towards hydrolysis[52,60,61].

20 2 STATE OF THE ART

[60] TAB. 2-3 SOME POSSIBLE INORGANIC SALTS FOR LIB ELECTROLYTES, BASED ON

Structure Name σ [mS cm-1] Comments

1 M, EC:DMC

LiPF Lithium hexafluoro- 10.7 forms non-conducting LiF in SEI, moisture 6 phosphate sensitive, passivates aluminium, only commercially used salt in LIB

LiBF Lithium tetrafluoro- 4.9 high mobility, but small dissociation, low 4 borate conductivity, has been reinvestigated recently for new electrode chemistries

LIClO Lithium perchlorate 8.4 stable towards moisture, 4 highly oxidative laboratory use, not commercialized

The SOLVENT needs to dissolve the salt in high amounts (i.e. have a high dielectric constant) and be fluid over a wide range of temperatures to allow for rapid ion transfer. At the same time, it should remain inert (or sufficiently passivating) towards the electrodes and all other cell components. Other requirements include safety, economy and toxicity[60,52]. The large potential window of battery electrode reactions (ca. -3 to 1 V vs. NHE) prevents the use of any protic solvents, since protons would be readily reduced at the negative electrode while the corresponding anion would be oxidised at the positive electrode[60]. The other solvents with a high dielectric constant include ethers, alkyl carbonates, esters and some solvents with nitrile and sulfonyl groups[62]. Ethers such as diethyl ether or tetrahydrofuran[63] were considered as possible solvent candidates for quite some time. However, oxidative breakdown limits their cycling stability[60].

A more general review on the history and properties of non-aqueous electrolytes used in electrochemistry can be found in [64–66].

The first commercial LIB were realised by using mixtures of ALKYL CARBONATES. Special attention was paid to ETHYLENE CARBONATE (EC), which has a dielectric constant (~89) that even exceeds that of water

(~79). Although solid at room temperature, the dissociation of LiPF6 in EC produces a liquid solution at room temperature. Its viscosity could be further lowered by adding small portions of PROPYLENE

CARBONATE (PC). Propylene carbonate, however, did not prove to be satisfactory for rechargeable batteries, mostly due to its lack of ability to from a stable SEI (see chapter 2.2.4.3, p. 27)[67]. Instead, linear carbonates such as DIEMETHYL (DMC) or DIETHYL CARBONATE (DEC) first described by D. GUYOMARD

[68–70] and J.-M. TARASCON were used as alternative . Mixtures of EC and linear alkyl carbonates

21 2 STATE OF THE ART

synergistically use the properties of cyclic HIGH DIELECTRICITY SOLVENTS (for the dissociation of salt) and linear LOW VISCOSITY SOLVENTS (for high ion mobility), see TAB 2-4. Typical ready-to-use electrolytes

[6] contain ~1 M LiPF6 salt dissolved in mixtures of EC (30-50 wt.-%) with DEC, DMC or EMC .

ELECTROLYTE ADDITIVES are added only in small amounts (typically around 2-5 %), yet they are the most effective and economic means to improve the performance, cycle life and safety of lithium ion batteries. Additives can serve a certain purpose or several purposes at once. ANODE ADDITIVES include reagents which facilitate the formation of a stable solid-electrolyte interphase and reduce gas formation and irreversible charge losses. The most prominent member of this family is probably

[71] VINYLENE CARBONATE (VC) . VC can be polymerized electrochemically at potentials at about 1.4 V vs.

+ Li /Li (FIG. 2-8), which is slightly above the reduction potential of the bulk solvents. The resulting SEI is more stable than the one formed by other solvents due to the polymeric contents. BULK ADDITIVES improve physical properties of the electrolyte bulk for safety and performance. This includes for example flame retardants, lithium salt stabilizers, wetting agents, plasticisers and viscosity diluters.

CATHODE ADDITIVES include overcharge protection agents and corrosion inhibitors. A comprehensive overview of additives can be found in [60,72,73].

FIG. 2-8 ELECTRO-POLYMERIZATION OF VINYLENE CARBONATE

22 2 STATE OF THE ART

TAB. 2-4 PHYSICAL PROPERTIES OF (CO-)SOLVENTS USED IN LIB ELECTROLYTES, [60] [52] COMPILED FROM AND , AT 25°C

Structure Abb. Name Viscosity dielectric E Wettability Comments reduction constant Ƞ [cP] + vs. Li /Li ε

EC Ethylen 1.85 (40 °C) 89.78 1.36 Poor solid at room

ity carbonate 1.90 (40 °C) temperature

PC Propylene 2.53 64.92 1-1.60 Poor graphite dielectric

2.53 (30 °C) exfoliation igh

carbonate h

DMC Dimethyl 0.59 (20 °C) 3.107 1.32 Good

carbonate 0.78 (30 °C)

EMC Ethylmethyl 0.65 2.96

carbonate

ow viscosity viscosity ow DEC Diethyl 0.75 3.12 1.32 Good l

carbonate 0.585

VC Vinylene 1.40 SEI forming

carbonate additive SEI

2.2.2.4 SEPARATORS

SEPARATORS are necessary to avoid electrical short circuits between anode and cathode. Commercial separators consist of polyolefin membranes such as POLYETHYLENE (PE) or POLYPROPYLENE (PP). Tri-layer membranes consisting of PP/PE/PP bonds combine the advantages of better oxidation stability and chemical resistivity of PP with the “shutdown mechanism” of PE. If a thermal runaway occurs, the PE melts and blocks further ionic current[74,75]. A ceramic separator for better mechanical and thermal stability was recently developed by EVONIK DEGUSSA and commercialised under the trade name

[76,77] Separion . For laboratory, glass fibre membranes such as WHATMAN GF/D are often used due to theirs ease of handling.

This chapter only gives a broad overview of “classical lithium batteries”, which is by itself a very large field of research and development. Significant technologies, however, still emerge. For example, at the time of publishing, around 2000 electric vehicles (the Bolloré BlueCar) can be found throughout France, using a battery with lithium metal anode, polymer electrolyte and vanadium oxide cathode.

23 2 STATE OF THE ART

2.2.3 ELECTROCHEMICAL ANALYSIS OF LITHIUM ION BATTERIES

Of the many electroanalytical techniques developed to investigate electrochemical systems (for a detailed review refer e.g. to [15]), the most prevalent and helpful techniques used in battery research will be shortly introduced here. Polarization under potentiodynamic control such as LINEAR SWEEP

VOLTAMMETRY (LSV) and CYCLIC VOLTAMMETRY (CV) are a quick and easy way to investigate unknown systems. They can for example be used to determine potentials of lithium insertion and deinsertion, cycle electrodes with unknown mass, or investigate the electrochemical stability window of electrolytes. During LSV, the potential of the WE vs. the RE is ramped linearly with time and the current between WE and CE is recorded. In CV, the potential-time diagram has a triangular shape; start and end-potential are usually identical. LSV and CV can be performed both in two-electrode (control of cell voltage) or three-electrode setup (observation of WE and CE potential). The advantage of CV is a good control over the WE potential irrespective of the mass of active material. The timespan of experiments is well defined by the scan rate of the potential ramp. For charge/discharge cycling of battery electrodes, though, it has also certain disadvantages. Since many battery materials have a very narrow range of lithium insertion potentials, currents recorded in CV can be small over a wide range of potential and very large during lithium insertion. To avoid large overpotentials caused by ohmic drop and concentration polarization, small potential scan rates (10 to 500 µV s-1) have to be chosen. This results in time-consuming experiments, especially in experiments with large potential windows.

As an alternative, the potential scan rate can be designed dynamic depending on the measured

[78] current. In the POTENTIODYNAMIC CYCLING WITH GALVANOSTATIC ACCELERATION (PCGA) protocol , a small potential step is applied upon the open circuit potential of the WE, followed by a chronoamperometric measurement period. Once the current has dropped below a pre-defined value, the next potential step is applied. This technique can be applied to perform the POTENTIOSTATIC

[79] INTERMITTENT TITRATION TECHNIQUE (PITT) . In PITT, each potential step is followed by an OCP determination period.

A similar method can also be applied in the galvanostatic mode, commonly known under the term

GALVANOSTATIC CYCLING WITH POTENTIAL LIMITATION (GCPL). Here, a constant current is imposed on the working electrode until a certain potential limit has been reached. At the potential cut-off, the

[80] current is reversed (or stopped) . This method can also be used to perform the GAVANOSTATIC

[81] INTERMITTENT TITRATION TECHNIQUE (GITT) . For this purpose, small charging increments are alternated with potential relaxation periods. The advantage of the GCPL technique compared to CV or PCGA is a perfectly constant current with consistent ohmic drops. In GCPL potential profiles,

24 2 STATE OF THE ART

lithium insertion in two-phase solids (such as LiFePO4 / FePO4) appears as potential plateaus (corresponding to sharp current peaks in the CV), while lithium insertion into a solid solution yields a

[22] continuous dependence of the insertion potential versus composition . CHARGE DERIVATIVE PLOTS are common means to visualize and identify redox potentials of electrochemical reactions. Drawing the derivative of charge versus the electrode potential yields a diagram which resembles that of a cyclic voltammogram: sharp peaks indicate potential plateaus from intercalation reactions while broader peaks indicate reactions with distributed potential or very slow reactions.

In ELECTROCHEMICAL IMPEDANCE SPECTROSCOPY (EIS), a sinusoidal current (GEIS) or voltage (PEIS) is superimposed on the working electrode under steady-state conditions, yielding a sinusoidal voltage (respectively current) answer depending on the perturbation frequency. Physical processes such as ion conduction, charge-transfer reactions, charging of the electrochemical double layer or solid state diffusion can then be separated via the difference of their specific relaxation time constants. A full explanation of EIS is given in chapter 2.5, p. 40.

On a lab-bench scale, LIB components are often tested in pouch cells, coin cells or two- or three- electrode SWAGELOK CELLS (FIG. 2-9). The latter are alienated tube fittings, in which the disc-shaped electrode stack is pressed together (and contacted) by solid plungers (instead of tubes) and a spring.

FIG. 2-9 THE STANDARD THREE-ELECTRODE SWAGELOK CELL, COMMONLY USED IN BATTERY RESEARCH A) DRAWING B) EXPLOSION PHOTOGRAPH OF DISASSEMBLED CELL AND SKETCH OF ELECTRODES. ASSEMBLY ORDER IS CE(RIGHT) - WE(LEFT) - RE(TOP)

25 2 STATE OF THE ART

2.2.4 THERMODYNAMIC AND KINETIC ASPECTS

2.2.4.1 CONDUCTION IN LITHIUM ION BATTERIES

As it was discussed in section 2.1, the overpotential of an electrochemical reaction depends on several processes, and all of them can be rate limiting steps during charge and discharge of a LIB electrode. In summary, these include: i) SOLID STATE DIFFUSION of lithium towards the particle surface

(or into the particle), ii) SURFACE KINETICS of lithium (de-)intercalation, iii) ELECTRIC CONDUCTION through the conductive carbon network, iv) IONIC CONDUCTION (migration) and v) DIFFUSION through the liquid phase (electrolyte) in the electrode pores and through the separator, see FIG. 2-10. For the production of high power electrodes, the contribution of conduction processes can be kept small by i) increasing the amount of conductive carbon, ii) reducing the size of active particles and iii) using thinner electrode films and iv) utilizing highly conducting, low viscosity electrolytes. The question, which type of conduction is most rate limiting, is often difficult to answer. Advanced electroanalytical techniques such as CURRENT INTERRUPT (CI) or ELECTROCHEMICAL IMPEDANCE SPECTROSCOPY (EIS) are therefore required to determine the INNER RESISTANCE of the battery. A full review on conduction phenomena in LIB was

[52] recently given by PARK ET AL .

FIG. 2-10 SEM CROSS-SECTIONAL IMAGE OF A POROUS LIFEPO4 ELECTRODE

26 2 STATE OF THE ART

2.2.4.2 THE STAGING EFFECT IN GRAPHITE INTERCALATION ELECTRODES

As can be seen in FIG. 2-11, lithium intercalation into graphite takes place around 250 to 80 mV vs. Li+/Li. In graphite particles with high crystallinity, the first cycle potential profile during galvanostatic cycling (see section 2.2.3) consists of one irreversible process (see chapter 2.2.4.3) and three reversible intercalation processes with well-defined potential plateaus. These plateaus correspond to three phase-transformations to lithium intercalation compounds with the following formal compositions[41,82,83]:

EQ. 2-17

ST FIG. 2-11 TYPICAL 1 CHARGE POTENTIAL PROFILE OF A GRAPHITE ELECTRODE INCL. SEI FORMATION AND 3 DISTINCT POTENTIAL PLATEAUS DUE TO THE STAGING EFFECT

2.2.4.3 IRREVERSIBLE CHARGE LOSSES IN GRAPHITE ELECTRODES

Since the intercalation potentials of lithium into graphite is beyond the electrochemical stability window of organic electrolytes, the electrolyte is reduced on any surface of the electrode in contact with the electrolyte[84]. This reaction would continuously consume electrolyte, if not a passivating

[85,20] layer, the so called SOLID-ELECTROLYTE INTERPHASE (SEI) was formed. This passivation process results in a significant amount of IRREVERSIBLE CHARGE LOSSES (ICL) and can be observed in the first charge/discharge cycle as potential shoulder around 0.8 V (see FIG. 2-11). The SEI is a passivating layer composed of both organic and inorganic reductive decomposition products of the electrolyte

[86] such as Li2CO3, LiF, ROCO2Li, polymeric carbonates .

27 2 STATE OF THE ART

“It is suggested that in practical non-aqueous battery systems the alkali and alkaline earth metals are always covered by a surface layer which is instantly formed by the reaction of the metal with the electrolyte. This layer, which acts as an interphase between the metal and the solution, has the properties of a solid electrolyte. The corrosion rate of the metal, the mechanism of the deposition- dissolution process, the kinetic parameters, the quality of the metal deposit, and the half-cell potential depend on the character of the solid-electrolyte interphase (SEI).”

- E. PELED (1979)[20]

While the precise composition of the SEI is still a matter of debate, the general function is nowadays well understood. Depending on the type of electrolyte, electrolyte components are electrochemically reduced around 1.5 to 0.7 V vs. Li+/Li, and either precipitate in form of the SEI or dissolve into the electrolyte. The onset potential for SEI formation depends on many factors such as the active material, solvents, salt and additives in the electrolyte and scan rate. Values reported in the literature range are as high as 2 V, but potentials between 1.2 V and 0.8 V are most widely found values[87,60]. The remaining SEI layer is electronically insulating, but ionically conducting. Therefore, the graphite surface is passivated and further reduction of the electrolyte does not take place.

The SEI is an interphase between electrode and electrolyte. It effectively blocks electrons and conducts lithium ions. It is a true SOLID ELECTROLYTE.

In the ideal case, solvated lithium ions will migrate towards the SEI surface, strip off their solvation shell, diffuse through the SEI layer as unsolvated ions and intercalate in between the graphene layers. Depending on the type and structure of the solvent and the corresponding SEI (if it is already built), lithium ions may not completely lose their solvation shell when entering the graphite lattice. Especially when the reduction of lithium-solvent complexes is kinetically hindered, as it is the case for lithium ions solvated by propylene carbonate, solvent molecules can co-intercalate into graphite and be reduced deep inside the particle. This reaction takes place already at about 0.7 to 0.3 V vs. Li+/Li[88–93]. The reduction to gaseous compounds such as ethylene or propylene inside the graphite lattice results in a rise of internal pressure, ultimately followed by exfoliation of the graphite particle. Exfoliation does not only reduce the overall charge capacity of the material but also generates fresh surfaces for further SEI formation. In contrast to PC, the reduction of lithium ion-ethylene carbonate complexes is not kinetically hindered. Electrolytes containing EC therefore form a stable SEI before solvent co-intercalation followed by exfoliation takes places. The SEI can so to say be called the boon and bane of the battery. On one hand, it consumes lithium and electrolyte and results in significant

28 2 STATE OF THE ART

loss of available charge. It also adds a considerable part of internal resistance to the cell. On the other hand it prevents further electrolyte reduction or exfoliation. Only the application of a good knowledge of the SEI and the use of purpose-made electrolytes could minimise solvent co- intercalation and exfoliation and thus enabled the production of commercial batteries with graphite

-1[69,85,87] electrodes exceeding a charge capacity of 300 mAh g . FIG. 2-12 visualizes the formation of SEI and exfoliation on graphite surfaces.

FIG. 2-12 MAIN IRREVERSIBLE CHARGE LOSS REACTIONS ON GRAPHITE

The SEI formation was also found to heavily depend on the surface properties of the graphite.

R. FONG from J.R. DAHN’S group found a correlation of the amount of ICL associated with SEI formation

[87,94] with the BRUNAUER-EMMET-TELLER SURFACE AREA (BET, measured by nitrogen physisorption) . In

1963, N. R. LAINE ET. AL. demonstrated the usefulness of the ACTIVE SURFACE AREA (ASA), which is

[95] determined by O2 chemisorption . The ASA is ascribed to the amount of surface defects like kinks, steps and free valences on the surface. W. P. HOFFMAN further developed the concept to explain the

[96] kinetics of different carbons for heterogeneous reactions . F. BEGUIN ET AL. and others successfully applied this concept for the analysis of SEI formation[97–99]. Strongly related to the ASA is the amount and type of surface groups on the carbons, which was investigated by D. AURBACH, M. SPAHR and

[91,100–104] P. NOVÁK . The knowledge on the importance of surface chemistry also led to several attempts to deliberately alter the surface properties of carbons, for example by oxidation[105,106]. The most successful way of improving SEI formation in terms of coulombic efficiency and cycle stability, however, was by careful choice of the electrolyte and use of certain SEI forming additives[71,107]. A

29 2 STATE OF THE ART

review on electrolyte additives can be found in [72]. Several reviews on formation, properties and different characterization methods of the SEI have been published[108,85,86].

Even before SEI formation, surface groups on the graphite may be reversibly or irreversibly reduced during first cycling of the electrode. Since the negative electrode potential is always held below 1 V vs. lithium in order to avoid re-oxidation of the SEI, the reduction of surface groups is considered small, but irreversible. Finally, metallic lithium may be deposited when the electrode potential is not monitored well. Since lithium is highly reactive towards the electrolyte, its surface will immediately form an SEI layer. Irreversible oxidative charge losses can also occur on the cathode. These include i) current collector oxidation, ii) electrolyte oxidation, followed by the formation of a CATHODE

ELECTROLYTE INTERPHASE (CEI) and iii) oxidation of active material in combination with oxygen evolution. A recent review on surface reactions in cathodes can be found in ref. [109].

30 2 STATE OF THE ART

2.3 DISTRIBUTION OF CURRENT AND CHARGE IN POROUS ELECTRODES

A POROUS ELECTRODE is a three-dimensional, porous solid with an electrochemically active surface much greater than its geometric area. Porous electrodes are of high relevance in industry. The large surface area enhances the kinetics of electrochemical reactions, e.g. for fuel cells or for efficient metal remediation from wastewater. Under corrosion, metal surfaces can turn into unintended porous electrodes[7,110].

Porous electrodes are fundamentally different from electrodes with porous layers. In the first case the pore walls are electrochemically active, while in the latter case the pore walls are inactive.

The most important characteristic of porous electrodes is the NON-UNIFORM DISTRIBUTION OF CURRENT under polarization. This phenomenon was first observed and qualitatively described by K. FISCHBECK

[111,112] [113] and E. EINECKE in the 1920s and later by J. J. COLEMAN in 1946 , who observed a non-uniform colour change during reduction of porous battery electrodes. The first theoretical explanation, however, was not done before V. S. DANIEL-BEKH used sliding probes to measure the potential distribution inside the electrolyte. Double integration of the potential yields the current distribution[114]:

EQ. 2-18

where q is the total diameter of pores within 1 cm² apparent electrode surface, f the area of the inner surface within the volume and is the specific resistance of the electrolyte.

[115] A first direct measurement of the current distribution was made by J. J. COLEMAN . He sliced a 3 cm thick porous pyrolusite (MnO2) electrode into three slabs of 1 cm each and separated those electrically by electrolyte wetted paper. Electric current of each electrode layer under polarization was then measured from carbon rods pricked into each slab. Although the current distribution in one of three lamellar is still far from being constantly distributed, Coleman was the first to directly investigate current distributions by experiment.

[116– Theoretical treatment of this discretised approach was done by J. EULER and W. NONNENMACHER 118]. They considered the porous electrode to consist of a discrete amount of thin layers, in which the current distribution is believed to be homogeneous. As depicted in FIG. 2-13, they accordingly used a

TRANSMISSION LINE MODEL (TLM) to describe the porous electrode as a discretized ladder network of linear resistances. The first linear chain describes the electric resistance of the solid phase Re and the second chain describes the ionic resistance in the pores Rion,p, followed by the ionic resistance of the

31 2 STATE OF THE ART

separator Rion,s. Each layer is described by one resistor for the solid phase Re and one resistor for the ionic phase Rion,p and converts a fraction of electric current into ionic current through a non-linear polarization resistance interconnect Z between the two chains. The polarization resistance in each layer was assumed to be equal, when the electrodes were rested and ion depletion is negligible. In summary, the mathematical solution of the ladder network yields two opposed hyperbolic cosine functions for the fraction of the cross-current diL/itotal:

[ ( )] ( ) EQ. 2-19

where a and b describe the electric and ionic resistance, c is the polarization resistance, dx is the length of one layer, L is the thickness of the porous electrode and α2 = (a+b)/c. In the case where the ionic resistance is much higher than the electric resistance, the current distribution is governed by the second cosh term and the current is highest close to the electrolyte.

FIG. 2-13 THE TRANSMISSION LINE MODEL (TLM) FOR POROUS ELECTRODES AS APPLIED FOR DC [116] CONDITIONS BY EULER AND NONNENMACHER , CONSISTING OF THREE EXEMPLARY LAYERS. ELECTRIC CURRENT ENTERS FROM THE LEFT CURRENT COLLECTOR WHILE IONIC CURRENT ENTERS FROM THE RIGHT. IN THE BACKGROUND: CROSS-SECTION SEM IMAGE OF A POROUS LIFEPO4 ELECTRODE USED IN LIB. A SIMILAR [119] MODEL WAS LATER ADOPTED FOR AC CONDITIONS BY R. DE LEVIE

Transmission line models were originally used to describe the behaviour of long telegraph transmission lines and later applied to other phenomena such as corrosion[120] and Warburg impedance (section 2.5.2.2, [121])

[122] J. NEWMAN further extended the model of Euler and Nonnenmacher by introducing Tafel-type polarization and mass transport phenomena such as concentration variations caused by the charge- transfer reactions. Newman’s work, based on the foundation from Euler and Nonnenmacher, laid the fundaments for many of today’s battery modelling and simulation approaches[123–127].

For ac treatment of porous electrodes, see chapter 2.5.2.3, p. 52.

32 2 STATE OF THE ART

EXPERIMENTAL DETERMINATION OF CHARGE OR CURRENT DISTRIBUTION in lithium ion batteries has been quite scarce. Imaging of LIB porous electrode structures (e.g. by FIB/SEM) was performed in order to provide information for theoretical studies under the aspect of porosity and tortuosity[128–131].

SYNCHROTRON X-RAY MICRO-DIFFRACTION was used to visualise charge distribution with a lateral resolution of several micrometers[132]. The applicability, however, is limited to crystalline materials.

NEUTRON IMAGING allows for the in-situ determination of lithium concentration gradients within a pouch cell. This includes lithium concentration in the electrolyte bulk as well as lithium stored in the particles. The time-frame of the experiments, however, permits only steady-state conditions[133].

K. NISHIKAWA measured the REFRACTIVE INDEX PROFILE caused by a lithium ion concentration profile in bulk electrolyte by holographic interferometry[134]. So far, the lithium depletion inside a porous electrode was not investigated with this method. A very elegant OPTICAL METHOD for the

[135] determination of lateral charge distribution was developed by P. MAIRE in the group of P. NOVÁK .

They utilized the colour change of graphite intercalation compounds for LATERALLY RESOLVED determination of the STATE-OF-CHARGE (SOC). The idea of colorimetric determination of SoC was pushed further by S. J. HARRIS, who visualised in-situ the VERTICAL CHARGE DISTRIBUTION (i.e. perpendicular to the counter electrode surface) over time by imaging an appropriate electrode configuration through a quartz window[136]. A video is available at [137].

In summary, the information provided by the aforementioned techniques was in all cases rather qualitative and limited to the distribution of charge. Direct or indirect determination of the current distribution was not done. K. C. HESS ET AL. rediscovered the technique of potential distribution measurements developed by V. S. DANIEL-BEKH. They used several sensing layers to investigate current distributions in a column of an aqueous electrochemical double-layer capacitance

[138] electrode . Following the approach of J. J. COLEMAN, the experimental equivalent of the discretised transmission line model has recently been rediscovered and applied to lithium ion batteries by F. LA

[139,140] MANTIA and S.-H. NG . They used a MULTIPLE WORKING ELECTRODE (MWE) to follow the distribution of charge and current during lithium intercalation into three graphite paste coated mesh electrodes separated by polyolefin membranes in-situ. Currents were recorded by current follower ammeters, one for each layer. Very recently, a similar approach was made by G. ZHANG ET AL. who used a MULTI-

[141] SEGMENTED ELECTRODE to investigate the distribution of current and charge in the lateral plane .

33 2 STATE OF THE ART

2.4 CARBON NANOTUBES AND NANOFIBRES

[142] FIG. 2-14 A MULTI-WALLED ARMCHAIR CARBON NANOTUBE

CARBON NANOTUBES (CNT) are allotropes of carbon and belong to the family of fullerenes. A CNT is a macromolecular structure consisting of sp²-hybridised graphene cylinders. Depending on the direction in which the graphene is “rolled,” one can distinguish between the zigzag tubes, armchair tubes and mixed forms that exist in these two vectors. MULTI-WALLED CNT (MWCNT) consist of several single-walled CNT (SWCNT) with different diameters stacked within another.

Unless stated otherwise, the term CNT refers to MWCNT.

On the lab scale, CNT can be produced via the catalytic chemical vapour deposition (CCVD), the arc discharge process or by laser ablation. Since it is the only process currently available for low-cost scale-up, commercial CNT are produced solely via the CCVD technique. This process nearly exclusively yields MWCNT with diameters between several nanometres to hundreds of nanometres (5-200 nm, generally 10-60 nm) and lengths between several to hundreds of micrometres, depending on process parameters and the type of catalyst[143,144].

In 1996 HARRY KROTO, ROBERT CURL and RICHARD SMALLEY received the Noble Prize for their seminal discovery of fullerenes in 1985[145]. In 1991, it was SUMIO IIJIMA, who presented well resolved TEM images of MWCNT[146]. Nevertheless, reports on filamentous carbon date back much earlier, ranging from T. V. HUGHES and C. R. CHAMBERS in 1889 to [147] L. V. RADUSHKEVICH and V. M. LUKYANOVICH in 1952 to A. OBERLIN and [148] M. ENDO in 1976 .

VAPOUR-GROWN CARBON FIBRES (VGCF) and their smaller pendant VAPOUR-GROWN CARBON NANOFIBRES

(VGCNF OR CNF) are structurally very similar to CNT. While CNT form perfect cylinders, CNF have a structure of stacked graphitic cones, cups or plates. As depicted in FIG. 2-15, the inner part can thus be completely closed or hollow similar to CNT. The terminology, however, is ambiguous in literature.

34 2 STATE OF THE ART

All types of CNF are therefore also sometimes referred to as stacked-cup CNT (SCCNT), bamboo or herringbone CNT[149,150,143]. Both CNT and CNF are essentially mesoporous, whereas CNF usually have slightly larger diameters than CNT (10-500; typically 50-100 nm). The BET surface area is usually higher for CNT (150-450 m²/g) than for CNF (10-250 m²/g) [143].

The terminology and classification of CNT, MWCNT and CNF etc. can be ambiguous and varies among authors.

FIG. 2-15 MULTIWALLED CNT, STACKED-CUP CNF (OR CNT) AND HERRINGBONE CNF

CNT and CNF have impressive mechanical stabilities and are electrically conductive which offers a manifold of potential applications, such as for polymer compounds[151], photochemistry, catalysis, or electrochemistry. Their application in electrochemistry includes electrocatalysis (as catalyst or as catalyst support), photo electrochemistry, sensors and electrochemical storage devices. VGCF and

[152] VGCNF were first commercialised by NIKOSSO in 1991 as GRASKER . BAYER MATERIAL SCIENCE produces

[153] MWCNT under the trade name BAYTUBES .

2.4.1 APPLICATION TO LITHIUM ION BATTERIES

CNF and CNT are very interesting materials which can serve as conductive additives due to their high aspect ratio, i.e. length to diameter ratio, of up to thousands. This aspect ratio allows for a very low conductive additive percolation mass threshold and thus a reduction of necessary mass of additive

(see p. 19, FIG. 2-7). CNT and CNF are also called NANOWIRES. CNF for LIB are commercially available and are already used in batteries[59,149]. A comprehensive review of publications and patents on CNF

[51] and CNT as conductive additives was recently presented by Q. ZHANG . In addition to their electronic conductivity, CNF and CNT can also provide structural stability to electrode architectures, and thus, increase lifetime and cycle stability, for positive electrode materials as well as negative electrode materials such as silicon[154–156].

Since they consist of sp² carbon, both CNF[157–159] and CNT[160–163] were investigated with regard to their potential use as active material. For CNT, both first principle calculations[164] and

35 2 STATE OF THE ART

experiments[165–167] have shown that the hollow structure of CNT allows for the insertion of large amounts of lithium. Like graphite, they can intercalate lithium between their graphene sidewalls and insert lithium into their central canal. It is also possible to produce conductive paper-like self- supporting electrodes (also known as buckypapers), which circumvent the use of a binder, conducting additive and current collector[161,168,169].

Nevertheless, two major obstacles remain for CNT to be used as an active material: i) the high surface area and mesoporous volume, which, due to the formation of the solid-electrolyte interphase, exfoliation and other side reactions, leads to an increased amount of irreversible charge losses during first discharge[170] and ii) the fact that pristine CNT suffer from reversible capacities that are much lower than expected. Lithium diffusion can only occur through sidewall defects and opened tube ends[164]. Therefore, a treatment of pristine CNT, like high-energy ball milling[171,172] or chemical[173–176,176–184] or electrochemical[185–187] oxidation is necessary to create defects and gain access to the inner core of CNT. Oxidation has three different effects on CNT. First, oxidizing acids dissolves impurities, such as metallic catalyst residues. Second, the treatment partly oxidises CNT, leading to the removal of end caps and fragmentation. Third, the surface is oxidised, forming various types of surface-bound oxygen. Some fragments are oxidised to polycyclic decomposition products (denoted as fulvic acids) and strongly adsorb on the surface. These compounds can significantly contribute to the total amount of functional groups on the surface[188,189]. Both types of surface modification help disperse the insoluble CNT into polar solvents, which is a major factor in the fabrication of CNT electrodes.

Several researchers reported positive effects regarding reversible capacity and the ICL of the CNT pre-treatment. An overview of results reported in literature is provided in TAB. 2-5. H. SHIMODA ET AL. applied an ultrasound treatment with a mixture of sulphuric acid and nitric acid to etch single walled

[184] CNT samples to enhance their reversible capacity from approximately LiC6 to LiC3 . H. C. SHIN ET AL. could slightly increase the reversible capacity of MWCNT, from 189 mAh g-1 to 223 mAh g-1, with a similar ultrasound treatment in fuming sulphuric acid, but only at the expense of an increase in irreversible capacity. They related this observation to the increase of surface area and the degree of

[190] graphitization . J. EOM ET AL. successfully applied high-energy ball-milling to decrease irreversible capacity of MWCNT, from 1012 mAh g-1 to 518 mAh g-1 while reversible capacity increased from

-1 -1[172] 351 mAh g to 641 mAh g . S. YANG ET AL. suggested that the number of active functional groups is responsible for electrolyte decomposition and formation of SEI[191]. The influence of the electrolyte on electrode performance was also investigated. B. LANDI ET AL. found out that the lithium ion capacity of both SWCNT and MWCNT shows high dependency on the composition of the electrolyte.

36 2 STATE OF THE ART

For SWCNTs, the highest reversible capacity (625 mAh g-1) has been achieved with EC:PC:DMC (1:1:2 v/v) and the addition of 5 vol.-% of vinylene carbonate as an SEI forming additive. At 100°C and EC:PC:DPC:VC (1:1:2:0.05v/v) capacities reached more than 1000 mAh g-1. According to their findings,

[160] all experiments to date show higher capacities with higher alkyl-chains of the co-solvents .

TAB. 2-5 SOME VALUES FOR THE REVERSIBLE CAPACITY AND IRREVERSIBLE CHARGE LOSSES OF CARBON NANOTUBES USED FOR LITHIUM INSERTION AS REPORTED IN LITERATURE author year details capacity ICL eff. ref. [mAh/g] [mAh/g]

[170] BEGUIN 1999 MWCNT refluxed in nitric acid 780 900 ~47 %

[184] SHIMODA 2002 single-walled CNT, sonicated incr. from not stated n/a

in 3:1 H2SO4:HNO3 10-24 h ~400 to ~700 removal of functional groups by annealing in 500 °C

[190] SHIN 2004 sonication of MWCNT in incr. from incr. from from 29 % fuming sulphuric acid for 24 h 455 to 486 189 to 223 to 28 %

[172] EOM 2006 ball-milled MWCNT from 351 from 1012 from 25 % to 641 to 518 to 50 %

[192] EOM 2008 ball-milled SWCNT 988 845 54 %

[193] FU 2009 HNO3 treated MWCNT 368 870 <50 %

in 1M LiPF6/EC:EMC:DEC 1:1:1

[160] LANDI 2009 MWCNT paper 150 400-210

in EC:PC:DMC 1:1:2 in LiPF6 MWCNT paper 210 190 [160]

in EC:PC:DEC1:1:2 in LiPF6 MWCNT paper from pyridine 340 300 [160] precursor in EC:PC:DEC

In summary, the reversible capacity has been attributed to the degree of graphitization and the accessibility to the inner core, leading to several potential regions for intercalation (between the sidewalls) and insertion (in the inner tube). The SEI is not discussed in detail in any of the aforementioned publications, but it is generally believed to form similarly on CNF/CNT as it does on graphite. Like it was described in chapter 2.2.4.3, several surface related parameters governing the

SEI formation were already identified for graphite, including the BRUNAUER-EMMET-TELLER SURFACE AREA

37 2 STATE OF THE ART

[87,94] (BET, measured by nitrogen physisorption) , and, perhaps more appropriately, the ACTIVE SURFACE

[97] [91,99] AREA (ASA, measured by oxygen chemisorption) and the amount of surface oxygen . It is reasonable to assume that these parameters affect the SEI formation on CNT in a similar way as on graphite, and it is also quite evident that the ASA is much larger for nano-structured materials, such as CNT, than it is for graphite, which already explains the higher degree of SEI formation.

At this point, the intrinsic differences and similarities between CNF and CNT shall be highlighted once again. Research was focused more on CNT than CNF as active material, although CNF as additives were commercially introduced earlier and to a wider extend than CNT[51,59]. On one hand, the tilt of graphene layers towards the fibre axis should allow for easier and more efficient lithium intercalation. On the other hand, this tilt turns the smooth outer tube wall, which is dominated by basal planes, into a roughened surface rich in prismatic edges. In yet another aspect, the larger tube diameter of hollow CNF effectively decreases the ASA in comparison to CNT.

Nevertheless, several questions remain unclear for both CNT and CNF. Although some relationships between CNT properties and ICL were discussed in previous publications, results remain incomparable due to the different types of CNT and different methods of functionalization used in these works. Thus, a conclusive knowledge about important parameters is still missing. More explicitly, the role of oxygen remains unclear; sophisticated methods of surface analysis should be employed to gain as much information as possible.

2.4.2 CHARACTERIZATION METHODS

SCANNING ELECTRON MICROSCOPY (SEM) is a convenient tool for quick and qualitative investigation of morphology. Statistical information on length or diameter distribution can be collected if a sufficiently large number of tubes is manually counted. For the resolution of sidewalls, it is necessary to apply TRANSMISSION ELECTRON MICROSCOPY (TEM). TEM also allows for a qualitative assessment of the degree of graphitization.

ELEMENTAL ANALYSIS, that is in most cases CHNS combustion analysis by atomic absorption spectroscopy, is useful to determine the total content of oxygen, nitrogen and other heteroatoms[194,195].

RAMAN SPECTROSCOPY has become an important tool in the investigation of carbon. The inelastic scattering of photons (i.e., the emitting or absorption of lattice phonons during the scattering process), depends on the properties of the sp²-hybridised carbon layers and thus yields valuable

38 2 STATE OF THE ART

information about the crystallinity of the material. In graphite, the most prominent Raman features are i) the only first-order Raman peak (the G band), which is approximately around 1580 cm-1 and results from the E2g modes in a plane graphene sheet and ii) the disorder-induced, second order mode (D-band) which is around 1350 cm-1 and originates from defects like sp3-hybridised carbon in the graphene layer and at the edges[196]. In CNT and CNF, the sidewalls consist of rolled-up layers of graphene, hence, they have spectroscopic features that resemble the modes of graphite[197,198].

X-RAY PHOTOELECTRON SPECTROSCOPY (XPS) is a quantitative analysis of speed and number of electrons that are emitted from the first 1-10 nm of a sample surface under irradiation with X-rays under ultrahigh vacuum conditions. Since the energy of an X-ray photon is determined by its wavelength, the binding energy of surface atoms can be calculated from the X-ray wavelength, the work function and the kinetic energy of the emitted electrons. In theory, any atom larger than helium can be detected via XPS, and different functional groups can be distinguished due to the different binding energies of atoms depending on their chemical state. This makes XPS especially useful for the analysis of oxygen groups[199,150,200]. In particular, XPS can yield information on the surface concentration and type of oxygen groups on the surface.

ACID-BASE TITRATIONS can be used to determinate the acidic groups on the surface of carbons. First developed by H. P. BOEHM ET AL., this technique of NaOH/HCl back-titration has commonly been

[201,202] known as Boehm titration . In 2005, C. G. SALZMANN ET AL. showed that highly oxidised carbon fragments adsorbed on SWCNT after nitric acid treatment are the sole contributors to acidic functional groups[188]. After washing with 8 M NaOH, no carboxyl groups could be detected by infrared spectroscopy. WANG ET AL. extended this study to MWCNT. They showed that these fragments also appear on MWCNT after nitric acid oxidation. Although sidewall modification does take place in MWCNT, there is still a considerable amount of surface adsorbed species, which can be identified as CNT-derived fulvic acids and account for up to 43 % of the surface acidity. Interestingly weak acid functionalities such as phenol groups could be associated with phenolic groups from surface adsorbed fulvic acids, while the MWCNT after fulvic acid removal only exhibited strong acid functions[189].

39 2 STATE OF THE ART

2.5 ELECTROCHEMICAL IMPEDANCE SPECTROSCOPY

“If you want to find the secrets of the universe, think in terms of energy, frequency and vibration”

- NIKOLA TESLA

2.5.1 INTRODUCTION TO IMPEDANCE SPECTROSCOPY

IMPEDANCE is the response of a system to an alternating signal perturbation. The system can be a simple passive electric element, a circuit or an electrochemical system. IMPEDANCE SPECTROSCOPY denotes the analysis of a system by resolving the system’s response in the frequency domain.

In this chapter, the basics of impedance spectroscopy will first be introduced for classical electric circuits and later extended to electrochemical systems in general (chapter 2.5.2) and to lithium ion

[203] batteries in particular (chapter 2.5.3). For further reading, the books from E. BARSOUKOV and

[204] [205] M. ORAZEM as well as several online resources such as those from R. RODGERS and colleagues or

[206] E. BARSOUKOV may be considered.

The term impedance SPECTROSCOPY may be misleading since no absorption of electromagnetic waves is involved. However, the image of a frequency-dependent response of a system to a sine perturbation may justify the use of the term in this work.

2.5.1.1 PASSIVE ELEMENTS AND THEIR RESPONSE TO A SINUSOIDAL SIGNAL

Passive elements (i.e. elements not generating a potential) can be distinguished depending on the fundamental relationship between current and voltage. If a passive element is perturbed by a sinusoidal voltage (current), it will respond with a sinusoidal current (voltage) with the same frequency, eventually shifted in PHASE and scaled in AMPLITUDE. The IMPEDANCE is defined as the transfer function of the system to the perturbation signal ,̃ thus it can be defined in the Fourier domain as:

̃ EQ. 2-20 ̃

The ADMITTANCE is defined as the transfer function of the system to the perturbation signal ̃:

̃ EQ. 2-21 ̃

In the case of a pure resistor perturbed by a sinusoidal voltage signal ,

40 2 STATE OF THE ART

EQ. 2-22 the resistor will respond with a current of the same frequency and in phase with the perturbation voltage:

EQ. 2-23

Since the impedance of the resistor is independent of ω, it is analogue to Ohm’s law, and Z simplifies to:

EQ. 2-24 Other systems, however, will respond with a current of the same frequency, but shifted by the phase angle ϕ:

EQ. 2-25

Thus, is a frequency-dependent vector, represented by its magnitude | | and phase shift ϕ in polar coordinates. Alternatively, it can be represented as a complex number in the Cartesian plot,

EQ. 2-26

with

Q | | E . 2-27

Q | | E . 2-28

Z is a complex number and can be represented either by its magnitude and phase shift or by its real and imaginary part.

FIG. 2-16 shows such a vector resulting from the combination of a RESISTOR R, a CAPACITOR C and an

INDUCTOR L in series. The REAL PART of impedance relates to the resistance to movement of charged species (resistors), while its IMAGINARY PART is inversely related to the accumulation of charged species in the system. The current flowing through a capacitor upon a voltage signal follows with a phase shift of , its impedance decreases with higher frequencies:

EQ. 2-29

The impedance of an INDUCTOR follows Eq. 2-30 with a phase shift of -

41 2 STATE OF THE ART

EQ. 2-30

FIG. 2-16 COMPLEX PLANE ILLUSTRATION OF THE IMPEDANCE VECTOR Z

TAB. 2-6 summarizes the impedance response of three passive elements. For elements in series, the impedance of each element is added, while for elements in parallel, the admittance, i.e. the inverse FIG. 2-17 A RANDLES CIRCUIT CONSISTING OF A of the impedance is cumulative. For example, the RESISTOR IN PARALLEL WITH A CAPACITOR, FOLLOWED BY [207] circuit depicted in FIG. 2-17 has an impedance of: A SECOND RESISTOR IN SERIES

EQ. 2-31

For elements in series, the impedance is additive, while for elements in parallel the admittance is additive.

TAB. 2-6: BASIC, PASSIVE ELEMENTS OF AN ELECTRIC CIRCUIT Element Symbol Fundamental Boundary conditions equation steady-state, high frequency,

Resistor R current proportional to potential Inductor behaves like short circuit behaves like open circuit

Capacitor behaves like open circuit behaves like short circuit

42 2 STATE OF THE ART

2.5.1.2 DATA ACQUISITION

In a typical potentiostatic EIS experiment, a small alternating voltage signal is superimposed on a constant electrode potential, ideally in equilibrium conditions. The first

impedance measurement device was the IMPEDANCE BRIDGE,

based on the simple WHEATSTONE’S bridge circuit (FIG. 2-18).

The unknown resistance Rx is determined by balancing the

two arms of the circuit ADC and ABC with known resistors R1 FIG. 2-18 WHEATSTONE’S BRIDGE CIRCUIT. THE UNKNOWN RESISTANCE IS DETERMINED BY to R3, so that no current is flowing through the galvanometer BALANCING BOTH ARMS OF THE BRIDGE SO THAT [204] NO CURRENT FLOW IS DETECTED BETWEEN between D and B . [208] THEM Since the impedance bridge requires a lot of manual work

(one balancing for each frequency), soon LOCK-IN AMPLIFIERS and FREQUENCY RESPONSE ANALYSERS (FRA) were developed for EIS. A lock-in amplifier multiplies an unknown input wave signal with a reference signal and integrates it in a low pass filter. Over a time much longer than the wave period, the integration signal is zero unless the two signals are of the same frequency (the reference signal is “locked” on the input signal). In this case, the integration signal output is different from zero and dependent on the phase shift.

Nowadays, literally all impedance analysers are FRA. A FRA is based on the Fourier analysis of a periodic signal. A FRA determines the impedance of the system by correlating the system’s response with two reference signals, one is the perturbation signal and the other, the perturbation signal shifted in phase by 90°. While older FRAs used to be independent instruments attached to a potentiostat, modern commercial potentiostats can usually be purchased with an integrated FRA device. They conveniently cover frequency ranges from ~100 MHz to ~10 µHz. A comprehensive introduction to FRA and other impedance measuring techniques can be found in [209].

2.5.1.3 DATA PRESENTATION AND ANALYSIS

Since impedance data points are triples, there are several ways to plot them. Plotting the phase angle and magnitude vs. the frequency yields the so-called BODE PLOT. Plotting the imaginary part of the impedance –Zim versus the real party Zre at any given frequency yields the NYQUIST PLOT. This is a rather common representation, due to its intuitive and easy way of interpretation. FIG. 2-19 gives an example of an electric circuit and the corresponding 3D, Nyquist and Bode plots. In the Nyquist plot, each data point represents one frequency and data points result in distinctive and easily recognizable patterns. Since impedance experiments are usually performed in a known range of frequencies with

43 2 STATE OF THE ART

a fixed number of data points (usually around 50-60 data points), it is usually enough to label just few points (like characteristic frequencies, see below) with a frequency value.

FIG. 2-19 DIFFERENT WAYS OF REPRESENTING THE IMPEDANCE OF THE CIRCUIT DEPICTED IN FIG. 2-17 A) 3D PLOT OF REAL PART, IMAGINARY PART AND FREQUENCY, B) NYQUIST PLOT C) REAL AND IMAGINARY PART VS. FREQUENCY D) BODE PLOT

FIG. 2-20 gives an overview of some simple circuit elements and circuits and their corresponding impedance response as Nyquist plot. The shown Nyquist plots were simulated using the ZSim software[210] in the frequency range of 100 kHz to 100 mHz, with 10 points per decade in logarithmic spacing. Increasing frequency is indicated by the black arrow. For a resistor, all frequency points fall together on one point on the x-axis (real part of impedance), since the resistance is frequency- independent. For capacitors, the impedance is infinite for dc current and approaches zero with increasing frequencies, while for inductors, the opposite is the case: the impedance decreases with decreasing frequency.

44 2 STATE OF THE ART

The connection of circuit elements in series or parallel results in similarly distinctive Nyquist plots. For example, connecting a capacitor C in parallel to a resistor R yields a semicircle with a diameter of

R and the CHARACTERISTIC FREQUENCY fc (i.e. the frequency with the highest value of impedance), which can be determined as:

EQ. 2-32

FIG. 2-20 IMPEDANCE RESPONSE OF SEVERAL ELECTRIC ELEMENTS AND SIMPLE ELECTRIC CIRCUITS REPRESENTED AS NYQUIST DIAGRAMS

45 2 STATE OF THE ART

2.5.2 IMPEDANCE SPECTROSCOPY FOR ELECTROCHEMICAL SYSTEMS

ELECTROCHEMICAL IMPEDANCE SPECTROSCOPY (EIS) utilises the frequency dependent current response of an electrochemical system following a sinusoidal voltage perturbation (or vice versa) to separate physical processes via their specific relaxation time constants over a wide range of values (from 10-5 to 103 s). EIS is a very powerful tool for the investigation of electrochemical systems. Typically, characteristics such as: i) resistance of electrolyte solution, ii) capacitance of double layer, iii) charge- transfer resistance and iv) transport of educts and products can be observed in the impedance spectrum.

In many cases, the impedance response of an electrochemical system results in an impedance spectrum quite similar to simple electrical circuits. For example, the impedance spectrum of a metal electrode in an ion conducting electrolyte can often also be simulated by an EQUIVALENT CIRCUIT (EQC) consisting of a resistor and a capacitor in parallel, followed by a second resistor in series. The values of these circuit elements can then be easily determined either geometrically or by regression analysis using the corresponding complex impedance equations. A common regression analysis for this purpose is the method of COMPLEX NONLINEAR LEAST SQUARES (CNLS), which is commonly implemented in modern impedance analysis software.

An ideal equivalent circuit gives at every frequency the same impedance response like the real system.

For our example, electrons from the metal electrode can either accumulate on the electrode surface

(charging of the electrochemical double-layer with capacitance Cdl) or, instead, be involved in a charge-transfer reaction across the electrode-electrolyte interface (with the charge-transfer resistance Rct), followed by diffusion through the electrolyte (electrolyte resistance Rel). Thus, the so- called RANDLES CIRCUIT, depicted in FIG. 2-21, is not only a FORMAL MODEL suitable for a mathematical fit

[207] but also qualifies as simplified PHYSICO-CHEMICAL MODEL for certain cases .

The Randles circuit yields quite accurate results for certain conditions (no diffusion limitation, no dispersion in time constants etc.). Depending on the electrochemical system, however, it may be necessary to introduce additional elements for meaningful regression analysis. In this context, it may be appropriate to refer to the dilemma of model sophistication (FIG. 2-22). A higher degree of sophistication does not imply that an impedance model is physically more accurate than a simple one. The question is, if additional EQC elements can be justified from a physico-chemical point of view or if they are pure mathematical means to optimise the fit of an observed EI spectrum.

46 2 STATE OF THE ART

FIG. 2-21 RANDLES CIRCUIT, A SIMPLE EQUIVALENT CIRCUIT FOR A METAL ELECTRODE IN SOLUTION

WARBURG ELEMENT OMITTED; REL, RCT AND CDL CAN BE EASILY DETERMINED BY GRAPHICAL OR REGRESSION ANALYSIS

Equivalent circuit models are not unique. A good fit of a measured impedance spectrum to an assumed EQC never validates physical processes suggested by the EQC. They should always be used based on physical reactions and be validated by additional experiments.

FINJA, 4 years FINJA, 6 years

FIG. 2-22 THE DILEMMA OF MODEL SOPHISTICATION: TWO SELF-PORTRAITS OF THE SAME GIRL; WHILE THE ONE ON THE RIGHT APPEARS TO BE MORE ACCURATE IN CERTAIN PARTS, IT ALSO INCLUDES AN INDUCTIVE (=IMAGINARY) TAIL, WHICH HAS NO PHYSICAL MEANING

2.5.2.1 DEVIATION FROM IDEAL BEHAVIOUR - THE CONSTANT PHASE ELEMENT

Until now, only electrode processes with a single time-constant were considered. In many cases, however, a dispersion of time constant may be observed. In practice, this may be observed as the semicircle in an EI spectrum appearing “depressed”, i.e. with the centre located below the real axis.

47 2 STATE OF THE ART

Several hypothesizes have been suggested for this behaviour, e.g. local properties of the electrode, like grain boundaries or inhomogeneous coating thicknesses, non-uniform current distributions etc. Speaking of a corresponding EQC, this behaviour can often be empirically described by replacing the capacitor in the Randles circuit by a CONSTANT PHASE ELEMENT (CPE or Q), which was first described by

[211] COLE and COLE . It can be written as:

EQ.

2-33 with . For , the CPE behaves as an ideal capacitor; while for the CPE behaves as pure resistor. For , the CPE behaves as inductor.

The use of a CPE often leads to improved regression of EI spectra. The physical system, however, may not follow the specific time constant distribution implied by the CPE and the interpretation of the CPE is not straightforward. It is therefore only an empirical approach and should be used with caution.

a) b)

THE CPE (ONE OF MANY SYMBOLS)

FIG. 2-23 THE CONSTANT PHASE ELEMENT (CPE), A) TYPICAL CIRCUIT ELEMENT USED IN EIS MODELLING; B) TYPICALLY DEPRESSED SEMICIRCLE OF A RANDLES CIRCUIT CONTAINING A CPE

48 2 STATE OF THE ART

2.5.2.2 DIFFUSION AND THE WARBURG ELEMENT

Diffusion is a common rate-limiting process associated with electrochemical reactions. With EIS, diffusion is not observed directly; instead, the decrease of educt concentration at the interface

[121] influences the impedance of the electrode. It was first described by E. WARBURG as early as 1899 .

The WARBURG ELEMENT ZW is a common circuit element used to describe semi-infinite diffusion for linear diffusion to a large electrode and can be written as[204]:

EQ. 2-34 √ √

Its magnitude is inversely proportional to the square root of frequency.

√ | | EQ. 2-35 √

The Warburg coefficient is defined as

EQ. 2-36 √ √ √

with

σ can therefore be used to evaluate diffusion constants of the electrochemically active species. The Warburg element actually is a special form of a CPE with a value for α of 0.5 (see EQ. 2-33). The

Warburg diagram can also be represented using the TRANSMISSION LINE MODEL (TLM), in which the chain of capacitors represents the system’s possibility to store charge carriers (i.e. ions) along the horizontal axis in form of the concentration gradient (FIG. 2-24).

49 2 STATE OF THE ART

FIG. 2-24 TRANSMISSION LINE MODEL (TLM) FOR (SEMI-) INFINITE DIFFUSION AND FINITE DIFFUSION (INCL. TERMINUS)

The Warburg impedance can be recognised with EIS at low frequencies as a 45° line in the Nyquist diagram, see FIG. 2-25. It is usually associated with Faradaic reactions, i.e. it is observed in series with the charge-transfer resistance and in parallel with the double layer capacitance. For that reason, it may not be identified easily at medium frequencies. The full derivation of the Warburg impedance can for example be found in[212].

A Warburg element may not always be easily recognised. It is, however, present in many cases.

a) b) c)

WARBURG ELEMENT WITH THE WARBURG ELEMENT LIMITED (LINEAR) DIFFUSION

√

FIG. 2-25 THE WARBURG ELEMENT A) PURE WARBURG ELEMENT B) RANDLES CIRCUIT CONTAINING A WARBURG ELEMENT C) WARBURG ELEMENT FOR FINITE PLANAR DIFFUSION D) RANDLES CIRCUIT CONTAINING LIMITED DIFFUSION

50 2 STATE OF THE ART

For many problems such as in thin electrolytes, semi-infinite diffusion is no longer guaranteed, and the governing impedance equation has to be altered to take into consideration the blocking nature of the finite diffuse layer[213]:

√ EQ. 2-37

√ the limiting resistance RDl can be determined by

EQ. 2-38

For practical analysis and simulation of impedance spectra, Eq. 2-37 can be rearranged to yield[210]

√ EQ. 2-39 √

and and are convenient fitting constants, from which the limiting resistance and limiting frequency can be derived (also see FIG. 2-25):

EQ. 2-40

EQ. 2-41

The imaginary part of finite-length impedance approaches infinity and the real part of finite-length impedance approaches the limiting diffusion resistance RD. Since the imaginary part at the zero frequency limit approaches that of a capacitor, it can be denoted as[213]:

EQ. 2-42 where dE/dc is the change in electrode potential with concentration. For an ideal solution this is[212]:

EQ. 2-43

and the impedance spectrum of finite diffusion at lower frequencies approaches that of the limiting diffusion resistance in series with the diffusion capacitance (FIG. 2-25C).

51 2 STATE OF THE ART

2.5.2.3 IMPEDANCE OF POROUS ELECTRODES

The behaviour of porous electrodes under direct current conditions has already been approached in chapter 2.3. R. DE LEVIE used the TRANSMISSION LINE MODEL (FIG. 2-26) already established for dc

[114] [116] conditions by V. S. DANIEL-BEKH and J. EULER and W. NONNENMACHER and for diffusion by

E. WARBURG (previous chapter) to discretise the impedance response within the depth of an electrode pore in a platinum wire brush electrode[214,119]. Following de Levie, the total impedance of a simplified single pore Zp follows:

√ EQ. 2-44 √ with

EQ. 2-45

and limiting conditions

EQ. 2-46 √

In practice, the impedance of pores is observed, similarly to the Warburg impedance, as a 45° line in the Nyquist plot, followed by a capacitive line in the case of ideally polarizable electrodes and a semicircle in the case of an electrode with faradaic behaviour. The model was further developed by A. LASIA to include the influence of concentration gradients and potential

[215,216] FIG. 2-26 TRANSMISSION LINE MODEL BY DE LEVIE drops in the pores . FOR A SINGLE ELECTRODE PORE, DEVELOPED FOR A [214] “POROUS” PLATINUM WIRE BRUSH ELECTRODE

The mathematical treatment of the Warburg impedance and the impedance for porous electrodes are similar and hence deliver the same solution. Nevertheless, their physical meaning is completely different.

52 2 STATE OF THE ART

2.5.3 ELECTROCHEMICAL IMPEDANCE SPECTROSCOPY OF LITHIUM ION BATTERIES

2.5.3.1 GENERAL

Batteries are complex systems, and analysis of their impedance requires the application of several concepts of physical electrochemistry concurrently. The most important properties that have to be taken into account are

The impedance of the battery reaction itself, incl. charge-transfer and storage The impedance of irreversible reactions and resulting insulating layers The impedance related to transport of electrons and ions Dispersion of impedance due to the vertical dimension and porosity The influence of the measurement system, i.e. the electrochemical cell, type of CE and RE and geometrical parameters as summarised in 0 and further developed in chapter 4.2.

Since batteries generally consist of the same elements, it is of no surprise that impedance models first developed for other battery systems such as lead acid batteries were later on adapted for lithium ion batteries. EIS can serve several purposes, such as fundamental research, quality control, determination of STATE-OF-CHARGE (SOC) or determination of STATE OF HEALTH (SOH) for life cycle assessment. Accordingly, several physicochemical, semi-empirical or purely empirical models emerged to image electrode or full battery processes in a more or less physical meaningful

[127] equivalent circuit . D. AURBACH ET AL. used EIS to characterise the behaviour of lithium metal electrodes[217–219] and graphite electrodes[220–222]. They suggested a semi-empirical model which describes the electrolyte resistance, lithium migration through the SEI, charge-transfer reaction and finally diffusion and accumulation in the bulk of the active particle (FIG. 2-27).

Models for lithium ion batteries can be sorted into physico-chemical models, semi-empirical and purely empirical models.

53 2 STATE OF THE ART

FIG. 2-27 SEMI-EMPIRICAL EQC FOR A GRAPHITE ELECTRODE [220] AS SUGGESTED BY LEVI AND AURBACH , WITH RION = ELECTROLYTE RESISTANCE, RSEI = MIGRATION

RESISTANCE THROUGH SEI, RCT = CHARGE TRANSFER RESISTANCE, CDL = DOUBLE-LAYER CAPACITANCE OF PARTICLE SURFACE, ZW = WARBURG IMPEDANCE FOR SOLID-STATE DIFFUSION AND CF = CHARGE CAPACITY OF LI INTERCALATION

2.5.3.2 ELECTROLYTE CONDUCTIVITY

The first resistance of the AURBACH model (FIG. 2-27) is the electrolyte resistance observed in every electrochemical experiment. It is composed of the bulk electrolyte and membrane impedance between both electrodes. It is generally assumed that the high salt content in battery electrolytes classifies as supporting electrolyte and therefore the migration term of conductivity can be omitted. In practice it means, that only the diffusion part is considered and the resistance can be considered constant. If the salt content changes significantly due to diffusion control, it would be necessary to introduce a finite length Warburg impedance element with transmission boundary, such as in the treatment of electrolytes in fuel cells[223].

Usually, the electrolyte resistance is treated as constant resistor in series with the electrode impedance. This is only valid under equilibrium conditions.

2.5.3.3 DIFFUSION AND MIGRATION THROUGH THE SEI LAYER

The formation of (porous) surface films is a common phenomenon for many electrochemical reactions such as corrosion, and is also observed in LIB. Passivating layers form, depending on the electrode potential, on any surface exposed to the electrolyte. For graphite particles, the formation of the solid-electrolyte interphase (SEI, see section 2.2.4.3) can be readily observed with impedance spectroscopy.

[220–222] [224] The growth of the SEI layer was observed using impedance by D. AURBACH and E. BARSOUKOV . The influence of different solvents on the SEI thickness and conductivity was also investigated[225,226].

54 2 STATE OF THE ART

Following the theory of passivating surface films on electrodes as depicted in FIG. 2-27, the impedance of the SEI is implemented by the use of one (or several) Voigt elements, i.e. C/R circuit in series with the electrode surface. RSEI is assumed to be a constant migration resistance of lithium ions through the SEI film and CSEI accounts for the accumulation of charge in the film. D. AURBACH ET AL. attributed as many as 4 Voigt elements to different surface layers. More recent works, however, used one element containing a resistor in parallel with a CPE to account for dispersion of the time constant of the SEI impedance[226].

2.5.3.4 FARADAIC IMPEDANCE: SURFACE IMPEDANCE AND SOLID-STATE DIFFUSION

The indisputably most important process in a LIB electrode is the charge-transfer associated with lithium insertion (or deinsertion) in parallel with electrochemical double-layer charging at the particle surface. The use of a Randles circuit[207], in most cases in combination with a CPE, is therefore the starting point for all modelling attempts of LIB electrodes.

Lithium intercalation into an active mass particle is naturally limited by the charge capacity of the given material. The intercalation electrode therefore behaves as a BLOCKING ELECTRODE under dc conditions and the diffusion of lithium into the particle contributes to the overall impedance. A FINITE LENGTH WARBURG ELEMENT M with blocking boundary accounts for an EQC element for solid state diffusion into the centre of the active mass FIG. 2-28 IMPEDANCE OF AN LIB ACTIVE PARTICLE DESCRIBED BY TWO POSSIBLE EQC: A WARBURG particle. This element is located in series with the charge- ELEMENT AND A CAPACITOR OR A LIMITED DIFFUSION WARBURG ELEMENT transfer resistance of lithium intercalation. The rule of conservation of charge, however, demands that the Warburg element is located in parallel to the double layer capacitance of the particle as shown in FIG. 2-28. The importance of diffusion can be seen when the limiting resistance of the Warburg element under dc conditions is compared to the charge-transfer resistance of the electrode reaction[213,204].

2.5.3.5 OTHER REACTIONS

Depending on the electrode potential, parasitic reactions such as surface group reduction or solvent reduction in combination with SEI formation may occur as well. If parasitic reactions significantly contribute to the overall current, another resistor should be included in parallel to the electrode surface reactions representing the CHARGE-TRANSFER RESISTANCE FOR LEAKAGE CURRENTS RL. Furthermore,

55 2 STATE OF THE ART

the current collector and conductive additives can also be involved in double-layer charging, electrode dissolution or electrolyte decomposition reactions. A special form of storing charge on the surface particles is PSEUDO-CAPACITANCE, which is, like the double-layer capacitance correlated which the surface area of particle. Unlike the double-layer, however, it includes a charge-transfer process[227].

2.5.3.6 INDUCTIVE COMPONENTS

Inductive contributions to battery impedance can have several reasons. The most natural one is of course the presence of coils. In a full (winded) battery such as in prismatic or cylindrical cells, the leads and current collectors usually require to include an inductance for impedance modelling. In the case of laboratory cells, coiled springs used for compression can have a similar effect if they are not short-circuited. The inductance is usually observed at frequencies above 10 kHz.

There is no direct electrochemical equivalent of inductive behaviour. Nevertheless, inductive bends or even full inductive loops may be observed in certain setups even without coils in the connections, especially when a reference electrode is used. These have been observed e.g. in Swagelok test cells and attributed to a combination of geometrical and electrochemical asymmetries[228]. These asymmetries lead to a frequency dependent distribution of equipotential lines, which may result in an inductive like impedance response. Also stray capacitances between RE, WE and CE may play a role[229,230]. A comprehensive explanation of inductive artefacts due to the cell setup is given in chapters 2.5.4, p. 57 and 4.3, p. 85.

Inductive loops can arise from real inductances (i.e. coils) from the cell or cell connections, from stray capacitances[229] and three-electrode setups[230] or from asymmetries of electrodes (chapter 2.5.4).

2.5.3.7 SUMMARY: THE FULL MODEL

The previous chapters have shown, which EQC elements can be used for the projection of processes on the active particle surface. A possible appropriate EQC for the surface impedance Zs is depicted in

FIG. 2-29A. If it assumed that no lateral distribution of impedance is present, this EQC can be implemented in the transmission line model as discussed in chapter 2.3, p. 31 and 2.5.2.3, p. 52[231,223,213].

Just like electrochemical processes, the impedance response and applied equivalent circuits depend on the electrode potential.

56 2 STATE OF THE ART

A sophisticated mathematical model describing the impedance response of porous electrodes with spherical particles of different particle size (including surface films) was developed by J. P. MEYERS in

[232] the group of J. NEWMAN . They show, how the distribution of ohmic and ionic resistances through the porous electrode yields a distributed impedance of 45 ° in the high-frequency limit of the Nyquist plot, unlike the expected capacitive behaviour with 90 ° angle. Furthermore they scrutinise the usefulness of impedance spectra for the determination of solid-phase diffusion coefficients in the light of particle size distribution. An efficient way of eliminating the influence of porosity by using

[220] very thin, single-particle electrodes was suggested by M. D. LEVI and D. AURBACH . N. OGIHARA recently demonstrated how the contribution of pore resistance can be isolated from the electrolyte bulk resistance and charge-transfer resistance by comparing symmetric electrodes at different State- of-Charge[233]. While the charge-transfer resistance is dependent on the SoC, the ionic resistance in the pores is not. Hence, the ionic resistance in the pores can be determined from the 45 ° line in the high frequency region which is left unaltered by the SoC.

A good model for fundamental research should always aim at describing a real system with sufficient accuracy, not at best fitting a curve, if the elements itself or the element parameters used have no physical meaning.

a) b)

FIG. 2-29 A) A POSSIBLE EQC FOR A SINGLE LAYER IN A LIB ELECTRODE (SEI OMITTED) B)TRANSMISSION LINE MODEL

FOR A POROUS ELECTRODE; INCLUDING SURFACE IMPEDANCE ZS, DOUBLE-LAYER CAPACITANCE CDL, IMPEDANCE OF FARADAIC PROCESSES ZF, CHARGE-TRANSFER AND SEI IMPEDANCE RCT, WARBURG ELEMENT OF SOLID-STATE DIFFUSION

WSSD, FARADAUC CHARGE CAPACITANCE CF, CHARGE-TRANSFER RESISTANCE OF LEAKAGE CURRENTS RL, ELECTRIC

RESISTANCE OF PARTICLES RE, ELECTRIC RESISTANCE BETWEEN PARTICLES AND CURRENT COLLECTORS RE,CC, IONIC

RESISTANCE OF THE PORES RION,P, IONIC RESISTANCE OF THE SEPARATOR RION,S, DOUBLE-LAYER CAPACITY OF THE CURRENT COLLECTOR C0

57 2 STATE OF THE ART

2.5.4 EXPERIMENTAL CONSIDERATIONS

2.5.4.1 GENERAL

For reliable impedance measurements, certain experimental requirements are necessary; the most important are i) LINEARITY, ii) STATIONARITY OF THE SYSTEM and iii) A VALID CELL CONFIGURATION WITH

REFERENCE ELECTRODE. First, the current response to a voltage perturbation needs to be linear. While this is true for a passive element (e.g. a resistor used as equivalent of the charge-transfer resistance), the current response of an electrochemical reaction shows a non-linear behaviour following the

BUTLER-VOLMER EQUATION (EQ. 2-12), see FIG. 2-30. If the amplitude of the voltage perturbation is chosen small enough, it can be considered quasi-linear. If it is chosen too small, the quality of the EI spectrum is impaired by noise. Usually, a voltage amplitude of 1 to 10 mV can be considered sufficiently small to yield a quasi-linear current response.

Second, especially at low frequencies, stationarity may become an issue. If we consider for example the dissolution or deposition of a metal, the electrode may completely change its surface (and thus its impedance) over the course of the impedance experiment. A good example is the FIG. 2-30 DIAGRAM OF THE BUTLER-VOLMER EQUATION: CURRENT DENSITY preferred formation of dendrites AS A FUNCTION OF OVERPOTENTIAL η; THE CURRENT DENSITY IS THE SUM OF THE ANODIC CURRENT DENSITY IA AND THE CATHODIC CURRENT DENSITY IK (TAKEN [234] under mass-transfer controlled AND MODIFIED FROM ) deposition of lithium, copper and many other metals. In how far non-stationarity may be an issue, strongly depends on the type of system and the frequency range of the experiment.

The third and probably most difficult requirement to fulfil is the electrode setup. In general, impedance measurements can be performed using a two-electrode setup or a three-electrode setup. In a two-electrode setup, the impedance contribution of both working and counter electrodes cannot be distinguished and the impedance response is a sum of both contributions. This might be acceptable in the case when the impedance of the CE is either small compared to the WE or when it is sufficiently known, e.g. in the case of two symmetric electrodes. In all other cases it is necessary to separate the impedance of the WE from the CE by means of a third, or REFERENCE ELECTRODE (FIG.

2-31A).

58 2 STATE OF THE ART

a) b) c)

FIG. 2-31 A) SIMPLIFIED RANDLES CIRCUIT FOR A THREE-ELECTRODE ELECTROCHEMICAL CELL; B) HABER- LUGGIN CAPILLARY C) SOLID-STATE ANALOGA OF A LUGGIN CAPILLARY

However, the use of three-electrode systems in impedance spectroscopy can be troublesome. In

1984, H. GOEHR pointed out that stray capacitances can be the reason for inductive behaviour that cannot be solely attributed to the inductances of cell cables[229]. He concluded that stray capacitance induced inductive loops can be avoided by, among other measures, appropriate shielding and keeping the current density at the WE as uniform as possible. In a more recent, theoretical work,

S. FLETCHER mathematically derived a two-terminal equivalent circuit from a simplified multi-node network of a three-terminal electrochemical cell containing resistances and stray capacitances of WE, RE and CE. This derivation unavoidably resulted in the presence of an inductive artefact in series and a capacitive artefact in parallel to the impedance of the working electrode[230]. These artefacts are present in every three-electrode measurement. The magnitude of these artefacts depends on the value of the other circuit elements. He concluded that everything that reduces the resistance of CE and RE will diminish this inductance (metal RE, large CE, well conducting solutions, wide bore Luggin capillary, no separator).

Even more important, the reference electrode itself can be the source of significant errors. A suitable RE must be stable in time and non-polarizable (i.e. involve a redox couple with high exchange current density). Its potential should return to its initial value after small currents have passed through[15,235]. The RE should act as a non-interfering spectator that observes one point of an equipotential plane close to the working electrode to minimise the uncompensated potential drop that appears once current is passing the WE. In aqueous system this is usually achieved using a metal / metal salt electrode with a HABER-LUGGIN CAPILLARY(FIG. 2-31B). To not influence the current lines of the ionic current to or from the working electrode, the Luggin capillary should be placed not closer than four times its own radius to the WE surface to avoid any influence on the ionic current lines[15,235,236]. It

59 2 STATE OF THE ART

was also shown that the position of the reference electrode has a strong influence on the electrochemical impedance[237].

For flat electrode assemblies and electrolytes with low conductivities (as in fuel cells or batteries) it is experimentally quite challenging to fulfil the aforementioned requirements. Distortions observed in the EI spectra are therefore the consequence of bad cell designs. G. HSIEH and co-workers provided a comprehensive overview of common and alternative geometric electrode setups used for EIS in solid electrolytes[238–241]. They concluded that symmetric geometries gave most reliable results, because no sharp drops in current lines are observed[241]. Some solid-state analogues of Luggin capillaries are presented in FIG. 2-31C. More recently, M. CIMENTI ET AL. summarised the sources of distortion for impedance spectra in solid electrolyte fuel cells[242,243]. According to Cimenti, distortions are mainly caused by i) variation of the electrolyte resistance with frequency[242], ii) cross-contamination of electrode impedances and iii) electric (i.e. electrochemical) asymmetries[243]. Geometric asymmetries (e.g. misalignments) lead to non-uniform current and potential distributions that change with frequency and, hence, a variation of electrolyte resistance, which is then erroneously attributed to the WE impedance[244]. Cross-contamination occurs, when the reference electrode fails to separate both impedances. Electrochemical asymmetries arise from a mismatch in polarization resistance or relaxation time constant. Depending on the applied frequency, the working and counter electrode can be dominated by either capacitive charging or charge-transfer processes. Different electrode activities can then lead to different current distributions.

Experimental findings can be well supported and generalised by theoretical calculations.

Computational methods such as the FINITE ELEMENT METHOD (FEM) can help to visualise current distributions during EIS. FEM is a numerical technique to approximately solve partial differential equations in a three-dimensional space by breaking down the space in a finite number of elements. The advantage of FEM is that certain parameters such as electrolyte thickness or conductivity can be easily changed to investigate their impact on the EI spectra. Numerous experiments can thus be

[241] avoided. FEM was already applied e.g. by G. HSIEH and co-workers to solid state electrolytes and

[245,246] further developed by S. B. ADLER . More recently, FEM was applied to understand distortions in

[228] commonly used Swagelok three-electrode cells . TAB 2-7 summarizes causes of and possible solutions to various distortions in impedance spectra.

60 2 STATE OF THE ART

TAB. 2-7 A SUMMARY OF REASONS FOR AND SOLUTIONS [243,242] TO DISTORTIONS IN IMPEDANCE SPECTRA, BASED ON

parameter problem solution

impedance of Mismatch of polarisation use low impedance CE

WE & CE resistance (impedance of CE is such as mirrored WE or dendritic lithium too high), cross contamination use large CE compared to WE

impedance of voltage divider effect reduce impedance of RE asymmetries Electrochemical Electrochemical RE use high impedance analyzer[236] size of RE RE probes several use smaller RE, place RE away from sharp equipotential lines electrolyte potential drops

position of RE distortion of current lines use Luggin capillary

y different size non-uniform current optimise geometry to achieve uniform of WE and CE distribution current distribution (current density may

geometr differ); use same size for WE & CE misalignment variation of electrolyte align WE & CE, change position of RE to resistance WE/RE with minimise influence of geometry frequency

2.5.4.2 IMPEDANCE CELLS FOR LITHIUM ION BATTERY RESEARCH

While the influence of electrode geometry and position of reference electrode has been well studied for solid electrolytes and fuel cells, little of this knowledge has been applied for studies in lithium ion batteries. This gap may also exist due to experimental difficulties. In commonly used three-electrode Swagelok cells[247], the electrodes are not well aligned and consequently suffer from EIS distortions such as scaling of spectra (misalignment of electrodes), high or low frequency inductive loops (electrochemical asymmetry) or even artificial capacitive loops (combination of both)[248,228]. Alternative cell geometries that satisfy the aforementioned criteria are, however, difficult to realise, since most of the cell geometries developed for solid state electrolytes and fuel cells are not suitable for very thin electrolytes as needed in batteries.

Possible experimental solutions include the use of lithium micro-reference electrodes in between

[224,249,250] two separators between WE and CE as e.g. suggested by E. BARSOUKOV . Another way of realizing a three-electrode setup which is both suitable for reliable cycling and impedance is by placing a small reference electrode behind a centre hole in the CE[213] to avoid blocking the current lines. Alternatively the reference can be placed around the WE in form of a ring[251]. If a porous

61 2 STATE OF THE ART

working electrode with mesh current collector is used, the RE can also be placed behind the WE[252]. To avoid cross-contamination caused by the lithium CE, its impedance can be reduced by cycling the lithium to grow dendrites prior to EIS[139]. An elegant way to completely avoid the influence of CE and RE is using a two-electrode symmetric cell consisting of two similar electrodes at the same State-of-

[253] Charge, as suggested by T. S. ONG ET AL. . On the downside, these cells are not convenient for repeated cycling and impedance analysis, since they need to be re-assembled from half cells after each cycling step.

The reliable validation of a given impedance cell is of vital importance. In literature, authors usually

[254] refer to testing the KRAMERS-KRONIG relationship and further, to the sum of impedances.

Following M. DOLLÉ ET AL., the cell geometry is said to be appropriate if the sum of the impedance of WE and CE vs. RE is equal to the impedance of the total cell, i.e. WE vs. CE[249]. This (incorrect) assumption will be further discussed in section 4.2.

62 3 AIM OF THIS STUDY

3 AIM OF THIS STUDY

"Insight must precede application."

- Max Planck[14]

Research on lithium ion batteries has gained a tremendous impetus due to the recent demand from electronics, electro-mobility and stationary energy storage sector. On the cell level, most efforts are undertaken on materials science and safety aspects. Targeted research, however, requires an understanding of the seminal principles behind each battery process.

Ionic and electronic conduction are pivotal processes governing the power density of LIB. An equally important aspect is surface reactivity which influences coulombic efficiency and cycle stability. New materials are often involved in the interplay of these two facets. carbon nanofibres, for example, are already commercially applied to increase electric conductivity and mechanical stability of LIB positive electrodes. The high amount of ICL, however, still limits their use in negative electrodes.

The first part of this thesis aimed at investigating the influence of surface properties on the behaviour of commercially available CNF at low potentials (3 V to 0.01 V vs. Li+/Li). As described in chapter 2.4, the important factors are the surface area and the surface chemistry. Since the amount and type of surface oxygen groups on CNF surfaces is significant, but has not been investigated conclusively yet, the first part of this work will focus on the influence of amount and type of oxygen groups on the amount and type of irreversible charge losses of CNF used in negative electrodes (chapter 4.1).

Each technique for surface analysis has its limitation and a proper investigation demands the combination of several techniques to obtain conclusive knowledge of surface properties. Elemental Analysis, SEM, Raman and XPS contribute with complementary valuable information concerning the elemental composition, morphology, degree of graphitization and nature of surface functional groups. These techniques are, however, mostly limited to ex-situ analysis and do not gain information on electrochemical reactions. On the contrary, electrochemical impedance spectroscopy (chapter 2.5) is a powerful tool for the investigations of electrochemical processes. Unfortunately, its application to lithium ion batteries has been limited, since severe distortions in impedance spectra are commonly observed in routinely used laboratory Swagelok cells.

The second aim of this thesis was the development of an electrochemical cell, which allows for reliable and routine acquisition of electrochemical impedance spectra during standard cycling protocols of LIB electrodes. This part will be discussed in chapters 4.2 and 4.3.

63 3 AIM OF THIS STUDY

While the theoretical energy density of lithium ion batteries is determined by the thermodynamics of the system, kinetic effects govern most of the apparent characteristics, such as internal resistance, capacity retention at high charge/discharge rates and even the amount of irreversible charge losses. To understand the electrochemical behaviour of porous electrodes, the recognition of mass transport is of vital importance. As it has been outlined in section 2.3, an ideal porous electrode can be considered as a 1-dimensional transmission line, in which an electric current is converted into an ionic current. If electronic conductivity throughout the electrode thickness is ensured by sufficient amounts of conductive additives, the performance of a lithium ion battery is governed by ionic conductivity and mass transport. It is, however, commonly neglected that the performance of an electrode is in many cases not limited by solid state diffusion but limited by diffusion in the electrolyte. Concentration profiles can quickly build up in organic electrolytes and create significant diffusion overpotentials, resulting in “layered charging” of active mass particles[140,136,132]. It is also known that the coulombic efficiency of a graphite negative electrode is dependent on its mass loading per electrode area[139]. It is, however, an unsolved question, if the irreversible charge losses of graphite electrodes can likewise be linked to phenomena of ion conduction, and thus, a distribution pattern of irreversible charge losses throughout the electrode thickness exists. The

[115] [139] generic approach of J. J. COLEMAN , which was applied to lithium ion batteries by F. LA MANTIA is potentially useful to answer this question. Several questions concerning layer resolution and reproducibility, however, have to be addressed.

The third aim of this thesis is the investigation of charge and current distributions throughout the depth of an electrode and the in-depth analysis of irreversible processes in lithium ion batteries, that is, the dependence of irreversible charge losses on the depth inside the electrode. For this purpose, a multiple working electrode (MWE) and a multi-layered electrochemical cell (MLC) were developed for high-resolution analysis of current distributions and studies including EIS, charge/discharge cycling and determination of irreversible charge losses (chapter 4.4).

64 4 RESULTS AND DISCUSSION

4 RESULTS AND DISCUSSION

4.1 IRREVERSIBLE CHARGE LOSSES IN OXIDISED CARBON NANOFIBRES

The methods and results discussed in this chapter were previously published in a research article[255]. Figures and discussions are therefore extended, yet analogous to main conclusions. Highlights

Carbon nanofibres exhibit high irreversible charge losses (ICL) during first lithium insertion. Reactions leading to ICL depend on the reduction potential. ICL reactions include surface group reduction, SEI formation and exfoliation. Nitric acid gas phase oxidation leads to different oxygen chemistry from liquid oxidation. Carbonyl surface groups prevent exfoliation better than single-bonded oxygen groups.

Disambiguation: The CNF used in this study were previously denoted as (stacked-cup) MWCNT[149,255]. For clarity and consistency within this thesis, however, the more precise term CNF will be used to avoid confusion with perfectly smooth tubular MWCNT.

To investigate the influence of oxygen chemistry of MWCNT and CNF on their performance in LIB, it is first necessary to investigate the amount and the type of oxygen groups. Therefore, commercially available CNF were oxidised by different treatments and analysed considering their morphology, degree of graphitization, amount of total oxygen, surface concentration of oxygen and type of oxygen groups.

As model system, hollow CNF (Pyrograf III, PR-19 PS) from Applied Sciences Inc., Ohio, U.S.A. were chosen. These CNF have a diameter of 20-50 nm and an outer diameter of 70-200 nm, resulting in a tube wall thickness of approximately 30-40 sidewalls[149]. The CNF are pyrolytically stripped, i.e. polyaromatic hydrocarbons were removed from the surface, but CVD-deposited non-graphitised carbon is still present on the surface. The bamboo-like structure results in more sites for possible oxygen functionalization compared to classical MWCNT. Since the majority of MWCNT reported in literature were oxidised with nitric acids or mixtures with nitric acids, one set of oxidised CNF was prepared in a similar way to compare the influence of oxidation time on the nature of the oxygen groups and ICL. Oxidised CNF were prepared by refluxing as-received CNF in concentrated nitric acid for different times (1.5; 3; 6, 12 and 24 h), followed by washing and drying. CNF obtained by this

LIQUID PHASE OXIDATION method were denoted as l-CNT(1.5 h, …24 h).

W. XIA ET AL. recently developed a method for GAS PHASE OXIDATION of CNT/CNF. They designed a fluidised bed reactor in which heated nitric acid vapours are used to oxidise CNT and CNF. This

65 4 RESULTS AND DISCUSSION method allows raising the reaction temperature from 120-125 °C (boiling point of nitric acid) to significantly higher temperatures (typically about 200-250 °C). As a consequence, oxidation products differ substantially from those obtained by liquid oxidation in terms of morphology and surface chemistry[256,257]. A clear advantage lies in the ease of processing. The oxidation in vapours effectively circumvents the need to filter, wash and dry the CNF. Since oxidation in the gas phase takes considerably longer than in liquid phase, CNF were oxidised with this method for 24 h and 72 h, denoted as g-CNF(24 h) and g-CNF(72 h), respectively[256].

In both cases, the total amount of oxygen groups continuously increases during oxidation. At the same time, however, degradation may also occur, leading to shorter tubes and thus an increase of active surface area per CNF mass. It is therefore of special importance to investigate structural changes and changes in graphitization alongside with total oxygen and surface oxygen analysis, before electrochemical analysis is performed during lithium insertion.

4.1.1 MORPHOLOGY AND DEGREE OF GRAPHITIZATION

FIG. 4-1 LIQUID OXIDATION FRAGMENTS CNF; SEM IMAGES OF CNF AFTER 90 MIN. (LEFT), 6 H (MIDDLE) AND 24 H OF LIQUID OXIDATION IN CONC. NITRIC ACID, 120 °C

FIG. 4-2 LENGTH OF CNF IS NOT MUCH AFFECTED BY GAS PHASE OXIDATION; SEM IMAGES OF CNF AS- RECEIVED (LEFT), 24H GAS-OXIDISED, 72H GAS-OXIDISED, 225 °C CONC. NITRIC ACID VAPOUR

66 4 RESULTS AND DISCUSSION

SEM imaging revealed that the CNF were stepwise fragmented during reflux in liquid nitric acid. After

90 min, the average length of l-CNF was around 5 µm. As can be seen from FIG. 4-2, fragments of 1 µm or less appeared in the SEM sample upon prolonged oxidation. After 24 h non-stop oxidation, almost all CNF were apparently shortened to 1 µm or less, since fragments significantly larger than 1 µm could not be observed anymore. For gas phase oxidation, the average length was less affected, and a large amount of g-CNF remained longer than 5 µm. This was also previously reported[256]. The average CNF length is a critical parameter and is important for the fabrication of strong free-standing electrodes[167] as well as for conductive additives, since the longer the tube is, the lower is the percolation threshold.

The breakdown of CNF was also accompanied by a significant weight loss. After wet filtration and drying of the oxidised CNF, the weight loss was steadily increasing and exceeded 25 % for l-CNF(24 h), see FIG. 4-3. In the case of g-CNF, the weight loss generally progressed slower and remained below 15 % after 24 h. The weight loss suggests that either a significant amount of CNF fragments was completely solubilised to e.g. fulvic acids or that small CNF fragments simply pass through the filter pores. As was already shown by SEM, the amount of short fragments is significantly smaller with gas phase oxidation. FIG. 4-3 TOTAL WEIGHT LOSS OF CNF AFTER OXIDATION AND DRYING PROCESS In addition, g-CNF did not require a filtering step at all, thus preventing filtration losses.

As described in chapter 2.4.2, p. 38, RAMAN SPECTROSCOPY can be used to determine the DEGREE OF

GRAPHITIZATION in CNF. For this purpose, Raman spectra were recorded in the range of 1250 to 1700 cm-1. The as-received sample exhibits a strong peak at 1580 cm-1, corresponding to the E2g modes of the bent graphene sheets (commonly referred to as G band) with a strong shoulder at 1555 cm-1 (unknown experimental error, possibly due to CNF burning) and a less pronounced shoulder at about 1610 cm-1 (D’ line). This shoulder was previously ascribed to structural defects[258]. The second peak at 1345 cm-1 can be assigned to the D band similar to disordered graphite and

[259] originates from defects in the CNF sidewalls such as prismatic edges of graphene layers . In FIG. 4-4, one can see how G- and D-band are sharpened after liquid or gas oxidation. The D/G ratio increases, indicating an increase in disordered sites with functional groups. It can also be seen that l-CNF(1.5 h) and g-CNF(24 h) have very similar Raman spectra.

67 4 RESULTS AND DISCUSSION

14000 as-received l-CNF(1.5 h) 12000 l-CNF(24 h) g-CNF(24 h) g-CNF(72 h)

10000 D'

[a. u.] [a.

8000 Abs.

6000 D band G band

1300 1400 1500 1600 1700 Raman shift [cm-1]

FIG. 4-4 RAMAN SPECTRA OF AS-RECEIVED CNF, L-CNF(1,5H) AND G-CNF(24H), EXCITATION WAVELENGTH 532 NM

4.1.2 OXYGEN CONTENT AND SURFACE CONCENTRATION

ELEMENTAL ANALYSIS revealed a steady increase in the total amount of oxygen after oxidation in the liquid phase. After 24 h, it slowly reaches saturation at around 15 %. The hydrogen content also increased slightly from 0.6 to 0.9 wt.-%, while the total amount of nitrogen remained unaffected. These contents are possibly due to residual nitric acid or nitrogen containing groups on the surface. The increase of oxygen effectively reduced the carbon content to <85 %. Despite the much longer oxidation time, gas phase oxidation for 24 hours FIG. 4-5 TOTAL H, N, C ELEMENTAL ANALYSIS OF CNF leads to total oxygen content of only ~5 %, which L-CNF (1.5 TO 24 H) AND G-CNF(24 H). OXYGEN CONTENT WAS CALCULATED AS RESIDUAL PERCENTAGE; DATA FOR G- is comparable to l-CNF (1.5 h). [260] CNF TAKEN FROM

68 4 RESULTS AND DISCUSSION

It has been shown that the SEI formation is strongly dependent on the surface on which it is formed[100,97,104,99,98,91,86]. Hence, it is of special importance to gain information about surface properties of oxidized CNF as well. As described in chapter 2.4.2, p. 38, different surface oxygen groups can essentially be identified via the vibrational modes of the bonds. C-O bond vibrations feature Raman shifts of 1260-1300 cm-1, while Raman shifts of around 1575 cm-1 may originate from carboxyl groups[262]. However, differentiation of oxygen groups or even quantitative assignment of single oxygen groups was not possible for the given samples, as seen from FIG. 4-4. Alternatively, FTIR, in principal, can qualitatively resolve C=O (1620-1720 cm-1) and C-O groups (1103- 1242 cm-1)[263]. Nevertheless, it was not possible to obtain meaningful data with the available FTIR instruments due to the high background, even under UHV conditions (data not shown).

Apart from vibrational spectroscopy, X-RAY

PHOTOELECTRON SPECTROSCOPY (XPS) was used to further investigate the surface chemistry, since it is known that it can be used to determine surface oxygen concentrations and clearly distinguish different oxygen groups by the decomposition of C1s and O1s regions in the spectrum.

Deconvolution of the C1S REGION allows FIG. 4-6 C1S REGION OF L-CNF(1.5H) AND G-CNF(24H) XP SPECTRA, INCLUDING BINDING ENERGIES OF C-O, C=O AND differentiating different oxygen-carbon groups due COO GROUPS FOLLOWING GENERALLY ACCEPTED ASSIGNMENTS [257,261] to the difference in carbon-oxygen binding . A MEANINGFUL DECONVOLUTION, HOWEVER, WAS NOT POSSIBLE DUE TO THE AMOUNT AND POSITION OF PEAKS energies. In particular, these are C-O groups (phenol, pyran, ether, hydroxyl) at 286.1 eV, C=O groups (carbonyl, quinone) at 287.5 eV and COO groups (carboxyl, carboxyl anhydride, ester) at 288.3 eV[257]. Nevertheless, quantitative evaluation of

[150] oxygen groups using XP spectra remains challenging. G. WILDGOOSE ET AL. correctly pointed out that deconvolution of the C1s region is highly subjected to the number of peaks which are chosen for the fitting. At least five peaks should be taken into account, including graphitic carbon at 284.5 eV and π-π* transformations at 290.5 eV. Indeed this high number of peaks, combined with the relatively low intensity of C-O bonds compared to graphitic carbon at 284.5 eV, introduces a considerable source for errors and only XP spectrometers with a resolution high enough to resolve small chemical shifts in the binding energies can yield meaningful quantitative values. As one can see from FIG. 4-6, the peaks of interest (C=O at 287.5 eV, C-O at 286.1 eV and COO at 288.3 eV) in the XP spectra of l-CNF(1.5 h) and g-CNF(24 h) are completely dominated by the graphite peak around 284.5 eV. Attempts to deconvolute the available spectra did not result in any meaningful quantitative

69 4 RESULTS AND DISCUSSION analysis of carbon-oxygen groups.

On the opposite, the O1S REGION can be easily deconvoluted into two peaks with a chemical shift of 1.6 eV, since they can be sufficiently separated by the instrument (~0.5 eV resolution). Qualitative comparison is even possible without deconvolution. One can observe two overlying peaks at 533.4 and 531.6 eV binding energy, which can be attributed to oxygen doubly bound to carbon (O=C) as in quinones, ketones and aldehydes and oxygen singly bound to carbon (O-C) as in ethers, hydroxyls and phenols, respectively. Oxygen atoms in esters, carboxyls, anhydrides and pyrones will contribute

[257] to both O 1s peaks, as they possess both single and double bonds , see FIG. 4-7.

FIG. 4-7 DIFFERENT OXYGEN CONTAINING GROUPS IN GRAPHITIC MATERIALS

FIG. 4-8 shows the O1s regions of l-CNF (1.5 h) and g-CNF (24 h). In accordance to elemental analysis, both samples have a comparable surface oxygen concentration of 16.6 % and 16.1 %, respectively. Nevertheless, a significant difference in the type of oxygen bonds could be observed. After deconvolution to 533.3 and 531.6 eV, one can see by the intensity of peaks that gas vapour treatment preferably leads to the formation of singly bonded oxygen species. After deconvolution, the C=O to C-O ratio can be determined to 0.69 in the case of l-CNF, while g-CNF have a C=O/C-O ratio of 1.68. This preference of bond types is presumably due to a strong dependence on thermal conditions. The oxidation in nitric acid is limited to its relatively low boiling point of 122 °C and thus allows for the formation of thermally less stable groups like hydroxyl and carboxyl groups. On the contrary, treatment in the vapour phase allows for much higher temperatures (here 200 °C) and favours thermally more stable double bonded species.

70 4 RESULTS AND DISCUSSION

FIG. 4-8 XPS O1S REGION OF TWO COMPARABLE LIQUID AND GAS PHASE OXIDISED CNF, DECONVOLUTION OF PEAKS REVEALS A LARGER PROPORTION OF C=O GROUPS FOR G-CNF

4.1.3 IRREVERSIBLE CHARGE LOSSES

During electrochemical discharge, the potential curve of CNF electrodes quickly drops from an OCP of about 3.2 V and proceeds, unlike graphite, with a continuous slope during both reduction and oxidation. This behaviour suggests the presence of many, thermodynamically different electrochemical reactions. Neither could discrete potential plateaus nor shoulders due to irreversible charge losses, although unambiguously present, be resolved, see FIG. 4-9A. Therefore, a data analysis previously developed for the investigation of ICL in carbon cloth was applied to better identify the potential areas in which irreversible charge losses occur or reversible lithium insertion takes place[264]. Therefore, the two first reduction curves were expressed as charge vs. potential curves and the second discharge curve was subtracted from the first one. The resulting difference curve is a measure of ICL and allows for a quantitative comparison of ICL in different potential regions (FIG.

4-9B). This definition of first to second cycle loss differs slightly from the definition usually used for the ICL, i.e. first reduction charge minus first oxidation charge. Nevertheless, this treatment is sufficiently accurate for an immediate visualization of different contributions to the total ICL. The maximum amount of ICL reaches that of the classic definition when the coulomb efficiency approaches 100 %.

71 4 RESULTS AND DISCUSSION

FIG. 4-9 A) FIRST AND SECOND CHARGE-DISCHARGE CYCLE OF G-CNF(24 H); THE CONTINUOUS DECREASE AND INCREASE OF INSERTION AND DEINSERTION POTENTIALS SUGGESTS A MULTITUDE OF DIFFERENT INSERTION SITES FOR LITHIUM; B) CHARGE CURVE VS. POTENTIAL AND ICL AS FIRST TO SECOND CYCLE CHARGE LOSS

The first order derivative of the first to second cycle charge loss curve reveals three main regions, which can be ascribed to several irreversible processes during first discharge, see FIG. 4-10. The highest potential region was previously attributed to the reduction of surface groups and/or the binding of lithium to surface groups. In particular, the charge derivative plot reveals a peak at approx. 2.8 V for g-CNF and around 2.3 V for l-CNF. These reduction processes most likely originate from the reduction of functional surface groups and/or the irreversible formation of lithium bound to oxygen groups. Surface group reduction has also been observed with CNT in aqueous media at around 0.4 to 0.55 V vs. NHE, where a quasi-redox type behaviour of quinone groups was suggested[265]. Examples of redox active oxygen groups include (in the order from high to low potentials) phenols, quinones, ethers, esters, aldehydes and ketones[266]. Other reactions may include the stripping of residual water or oxygen, which was reported to occur at around 1.6 V vs. Li+/Li. A clear derivative peak can also be observed between 0.9 and 0.8 V. Since the electrolyte is thermodynamically unstable at these potentials, it is widely accepted to ascribe the charge in this potential region to electrolyte reduction and formation of the SEI[84–86], preferentially formed on the defect-rich prismatic planes of carbons.

It is well-known for graphite that if a solvated lithium ion is able to penetrate the SEI layer, the solvation gets reduced and forms, depending on the solvent,

ST ND FIG. 4-10 CHARGE DERIVATIVE PLOT FROM 1 TO 2 REDUCTION gaseous compounds such as ethylene or CHARGE LOSS FOR TWO EXEMPLARY SAMPLES

72 4 RESULTS AND DISCUSSION propylene. These gases cause a volume expansion, resulting in exfoliation of the graphene layers. For graphite, exfoliation was observed at around 350 to 450 mV[89,90,92] and is expected to commence at similar potentials for CNF. Irreversible lithium entrapment without exfoliation, however, may also occur. These reactions form the third group of ICL in CNF. In the current sample, exfoliation and/or entrapment manifests as continuously shoulder below 0.5 V with a superimposed sharp peak at 0.2- 0.3 V.

To examine, in how far the functionalization of CNF influences these three different ICL, the potential curve was arbitrarily divided into three different potential regions according to surface group reduction (3.0-1.5 V), SEI formation (1.5-0.5 V) and exfoliation and lithium entrapment (0.5-0.01 V).

FIG. 4-11 compares the first to second cycle charge loss of l-CNF (1.5h to 24 h) and g-CNF (24h and 72h) samples in these potential areas. L-CNF, irrespective of oxidation time, suffer from a substantially higher total ICL than g-CNF. The overall charge loss shows a slight tendency to decrease with oxidation time, from ~280 mAh/g (1.5 h) to 220 mAh/g (24 h). Both g-CNF(24 h) and g-CNF(72 h) exhibit significantly lower charge losses. Interestingly, the ICL between 3 and 0.5 V are quite similar for both g-CNF and l-CNF within experimental errors, suggesting a similar ASA and amount of SEI. The exfoliation charge loss, however, is significantly smaller after gas phase oxidation, which is 60 mAh/g for g-CNF(24 h) and 33 mAh/g for g-CNF(24 h), compared with 146 mAh/g and 107 mAh/g for l-CNF (1.5 h and 24 h, respectively). The decrease amount of ICL for g-CNF therefore solely originates from a decrease in exfoliation.

1,5h liq. 311

300 6h liq. 283

12h liq. 273 24h liq. 250

24h gas 232 72h gas

200 200

172

148 146

/ [mAh/g] 150

123

119

107

103

101

98 98 -Q

100 94

42 41

50 39

33 27

0 0-0.5 0.5-1.5 1.5-3.0 total irreversible charge loss potentials

ST ND FIG. 4-11 DISTRIBUTION OF ICL AS 1 TO 2 REDUCTION CHARGE LOSS, BINNED IN 3 POTENTIAL REGIONS: IRREVERSIBLE SURFACE GROUP REDUCTION (3.0-1.5 V), SEI FORMATION (1.5-0.5 V), EXFOLIATION AND LITHIUM ENTRAPMENT (0.5-0.05 V)

73 4 RESULTS AND DISCUSSION

4.1.4 REVERSIBLE CAPACITY

The almost linear increase in oxidation potential during lithium extraction, which can be seen from

FIG. 4-9A suggests a progressive release of lithium from a variety of binding sites. Three reversible oxidation processes could be identified analysing the charge derivative plot depicted in FIG. 4-12A. In the first region 0.01 V to approximately 0.6 V, lithium deintercalates from in between the graphene sidewalls. The second, broader region between 0.6 and 1.9 V, has been attributed for CNT to deinsertion from the inner core[172].The highest potential region 1.9-3 V results from extraction of lithium intercalated adjacent to defect sites and faradaic (pseudocapacitive) surface reactions such as extraction of lithium bound to surface groups or a lithium-independent reversible redox-behaviour of quinone groups.

As it can be seen in FIG. 4-12, the sidewall deintercalation charge of l-CNF continuously decreases with increasing oxidation time, from 174 mAh/g for l-CNF(1.5 h) to 149 mAh/g (24 h). This trend can be simply explained by the successive oxygenation of CNF effectively leading to a decrease of graphitic carbon (from 93 wt.% to 83 wt.%) and an increase of carbon bound to oxygen. Indeed, a strong increase of capacity at potentials above 1.9 V (reversible reactions directly or indirectly associated with surface groups) can be observed, rising from 87 mAh/g (1.5 h) to 231 mAh/g (24 h). The charge derivative plot for g-CNF shows similar features, however, the deintercalation charge is surprisingly lower than for l-CNF (135 mAh/g and 98 mAh/g compared to 174-135 mAh/g), even though elemental analysis and Raman have shown that g-CNF(24 h) have a content of graphitised carbon similar to l-CNF(1.5 h). This observation suggests that fewer sites, even though present, are accessible for lithium intercalation.

FIG. 4-12 REVERSIBLE CAPACITY OF G-CNF AND L-CNF, DIVIDED INTO THE 3 POTENTIAL AREAS, I.E. GRAPHENE LAYER DEINTERCALATION (0.01-0.6 V), INNER CORE DEINSERTION (0.6-1.9 V) AND EXTRACTION FROM FUNCTIONAL GROUPS (1.9-3.0 V); DROP-COATED SAMPLES ON DENDRITIC COPPER FOIL, ~0.5MG/CM², 1M LIPF6 IN 1:1 EC:DEC, 50 MAH/G

74 4 RESULTS AND DISCUSSION

4.1.5 DISCUSSION

Carbon nanofibres were oxidised by two different oxidation methods, liquid and gas phase oxidation in nitric acid. From SEM imaging (tube length), Raman spectroscopy (degree of graphitization), elemental analysis (total oxygen content) and XPS (concentration of surface oxygen) it was concluded that l-CNF(1.5 h) and g-CNF(24 h) are very similar in terms of tube length, degree of graphitization, total oxygen content and concentration of surface oxygen. Nevertheless, while C1s analysis was not successful, deconvolution of the O1s region in XPS revealed a difference in the type of oxygen bonds, favouring formation of double-bonded oxygen during oxidation in the gas phase. Both types of CNF were subsequently examined regarding irreversible charge losses and reversible charge capacity in three different potential regions. The amount of ICL was divided into the contribution of irreversible surface reactions, SEI formation and exfoliation. It was observed that exfoliation of CNF is decreased for higher amount of total oxygen and furthermore, significantly minimised in the case of g-CNF. Considering that l-CNF(1.5 h) and g-CNF(24 h) only differ in the type of oxygen groups, there is a good case to believe that this difference is the reason for the significantly decreased exfoliation observed with g-CNF.

Until now, a good relationship between the irreversible capacity of carbonaceous electrodes and the active surface area (ASA)[85,86] was found. In our case, a similar charge amount used for SEI formation (0.5 V to 1.5 V) suggests a similar value of ASA for both samples. The exfoliation, however, differs substantially. Accordingly, the exfoliation charge loss (<0.5 V) should be reviewed independently of charge losses at other potentials. The influence of ASA and surface oxygen groups naturally occurring on graphite particles has recently been investigated by Spahr et al. It was found out that the removal of oxygen groups by heat treatment in inert atmosphere leads to a heavily promoted exfoliation of graphite particles. Curing the reduced samples in air restored oxygen groups and prevented exfoliation again[102,91,92,98]. This effect was explained by i) the necessity of a certain amount of ASA and ii) the necessity of oxygen groups to form a SEI that prevents exfoliation[99]. In the light of this work, provided that similar SEI formation and exfoliation mechanisms apply for CNF, it appears that not only the amount of total surface oxygen but, additionally, also the type (i.e. the ratio double- bonded to single-bonded oxygen) determines the extent of exfoliation.

Single-bonded oxygen groups like phenols can provide electron density to a π-conjugated system via mesomeric effects (+M effect). On the contrary, aldehyde or carboxyl groups withdraw electron density from aromatic systems due to mesomeric and inductive effects (−M/−I effect)[267]. It is therefore suggested that the aromatic electron density near the surface is important for the affinity of graphene layers to positively charged, solvated lithium ions. The effect of electron providing or

75 4 RESULTS AND DISCUSSION withdrawing groups on the cation-pi interaction of lithium with aromatic systems is believed to be mostly electrostatic, purely resonance seems not to substantively contribute to cation-π

[268,269] binding . Nevertheless, C. H. SURESH and S. R. GADRE investigated the influence of several substituents by using molecular electrostatic potential minima as measure of substituents effects and found that "[…] in the case of systems with electron-donating substituents Li+ binds more strongly than benzene, whereas in the case of halogens and electron-withdrawing systems, the binding of Li+ is weaker than that found in benzene”[270]. The binding energies decrease with increasing solvent polarity[271]. Hence, the right choice of electron-withdrawing substituents on the graphite prismatic surface may sufficiently increase the activation energy for solvent co-intercalation and thus hinder exfoliation. The lower value for the reversible capacity of g-CNF compared to l-CNF supports this idea, since non-solvated lithium ions are similarly repelled, in the bottom line leading to a lower amount of intercalation. Nevertheless, further investigations are necessary to see, if this theory can be supported. Electrochemical impedance spectroscopy may reveal, if for example the charge- transfer resistance of solvent co-intercalation is increased by the surface treatment.

Extensive studies on the correlation of surface properties with the exfoliation of graphite reported in literature served as starting point for the investigation of similar processes in CNF. Looking at it the other way around, CNF also are an interesting model materials to further understand the role of oxygen, since their remarkably high surface area allows for a much higher concentration of oxygen in a sample, effectively simplifying the examination of surface effects. An obvious question concerning graphite indeed is, whether natural (atmospheric) oxidation of graphite already forms an oxygen terminated surface which is best suitable for avoiding subsequent exfoliation, or, if a tailored oxygen chemistry can likewise decrease tendency of graphite to exfoliate.

76 4 RESULTS AND DISCUSSION

4.2 DEVELOPMENT OF A COAXIAL IMPEDANCE CELL FOR LITHIUM ION BATTERIES

The methods and results discussed in this chapter were previously published in a research article[248]. Figures and discussions are therefore extended, yet analogous to main conclusions.

CHAPTER OUTLINE In three-electrode Swagelok-type test cells, EIS suffers from distortions in the Nyquist plot. Distortions can include inductive loops, scaling effects and artificial capacitive loops. Reducing the lithium RE diameter alone does not improve EIS quality. RE in coaxial position, plus electrode alignment removes scaling effects. Inductive loops in the high-frequency domain may appear if edge effects are present. Inductive loops are limited to above 50 kHz if edge effect are avoided.

4.2.1 INITIAL SITUATION

As it was explained in 2.5.4, p. 58, electrochemical impedance spectroscopy for lithium ion batteries faces several experimental difficulties. On one hand, reliable EIS has certain requirements concerning the size, shape, characteristics and position of WE, CE and RE. A good summary for requirements in

[243,242] solid electrolytes was recently given by M. CIMENTI . In short, Cimenti gave the following suggestions to obtain reliable EI spectra.

The electrolyte should be thick compared to the WE/CE size and misalignment. The electrolyte should be well conductive. The CE should be of same size or larger than the WE, depending on the setup. The impedances of WE and CE should match each other. The RE should be small enough to probe a single equipotential plane. The RE should not disturb the current lines between WE and CE. The RE should be placed close to or far away from WE/CE, depending on the setup.

On the other hand, the investigation of LIB electrodes in research and development requires laboratory scale test cells, which are as close as possible to real systems. These cells should be reproducible regarding geometry, electrolyte thickness and mechanical pressure on the electrodes. If a reference electrode is used, it should not interfere with the experiment; be long-term stable and not deteriorate or poison the electrolyte. Many of the suggestions for proper impedance experiments provided for solid electrolytes by Cimenti are therefore not applicable to LIB test cells.

As consequence, a standard Swagelok three-electrode T-cell (FIG. 2-9, p. 25) was chosen as the starting point for the development of an impedance test cell suitable for LIB. In the Swagelok cell, two disc electrodes, commonly the electrode of interest (a porous electrode on a current collector

77 4 RESULTS AND DISCUSSION foil), and a lithium metal foil (for unlimited supply of lithium), are separated by a thin polyolefin or glass fibre separator soaked in electrolyte. The reference electrode is a small piece of lithium (Ø 5 mm) pressed onto the perimeter of the electrode stack.

E1-R-E2 denominates the EI spectrum of E1 against a RE in T position, with E2 as CE. R-E1-E2 denominates the EI spectrum of E1 vs. a coaxial RE inside E1. E2-E1-R denominates the EI spectrum of E2 vs. RE inside E1.

It is known from both experiment[248] and FEM simulation[228] that the three-electrode Swagelok cell commonly used for charge/discharge experiments in LIB research (referred to as CELL 1) is not suitable for EIS experiments. FIG. 4-13 illustrates what these distortions can look like. Depending on electrodes and assembly of the cell, distortions such as inductive loops, scaling effects or even artificial “ghost” semicircles can be observed. Scaling of spectra was ascribed to improper electrode alignment, resulting the RE probing several equipotential surfaces between WE and CE. Inductive loops were ascribed to electrochemical asymmetries. A combination of both may result in ghost semicircles. A full explanation of these effects can be found in chapter 2.5.4, p. 57.

200 LFP -R-LFP LFP-R -Li 6 1 2 Li LFP2-R-LFP1 Li-RLi-LFP LFP +LFP (full cell) 1 2 LFP-Li (full cell) 150 100 (LFP1+LFP2)/2 (calc.)

]



[ [

100 [

im 50 im

2 -Z

-Z -Z

0 0

0 6 8 10 12 14 0 50 100 150 200 0 50 100 150

Z [] Z [] Zreal [] real real

FIG. 4-13 EXEMPLARY EI SPECTRA ILLUSTRATING DIFFERENT DISTORTIONS OBSERVED IN SWAGELOK THREE- ELECTRODE CELLS; LEFT: INDUCTIVE LOOPS IN THE NYQUIST PLOT OF A TIO2 ELECTRODE AT 0 % SOC AGAINST 5 MM LITHIUM RE (LITHIUM CE); MIDDLE: SCALING EFFECT IN THE NYQUIST PLOT OF TWO IDENTICAL LFP ELECTRODES AT 0 % SOC, THE DASHED LINE IS THE CALCULATED HALF IMPEDANCE OF THE FULL-CELL AND INDICATES THE “TRUE” IMPEDANCE OF BOTH ELECTRODES; RIGHT: THE NYQUIST PLOT OF A LFP ELECTRODE AGAINST LI RE WITH LITHIUM CE EXHIBITS A SECOND “GHOST” SEMICIRCLE (~50 % SOC, OCP ~3.44 V)

4.2.2 CELL DEVELOPMENT

In order to overcome the aforementioned distortions, several alternatives to cell 1 were designed and investigated. As a model system, an LFP electrode at 0 % or 50 % SOC was chosen in order to observe the influence of the cell geometry on its EI spectrum. First, the reference electrode was decreased in diameter, so a smaller electrolyte volume is probed (FIG. 2-9B). This is achieved by introducing a 0.5 mm extrusion hole, through which the metallic lithium can be squeezed through

78 4 RESULTS AND DISCUSSION and subsequently cut off to create a 0.5 mm disc electrode with a fresh surface. As can be seen in

FIG. 4-14, the EI spectrum the LFP electrode did not change much, regardless of whether a 5 mm or 0.5 mm lithium electrode was used as RE. Both spectra exhibited a second semicircle in the Nyquist plot which was not expected.

As a consequence, a second generation cell (CELL 2) was conceived in order to reduce the frequency- dependent distribution of the equipotential surfaces. The symmetry was increased by moving the RE from its perpendicular “T” position to a coaxial position in plane with the WE. For this purpose, the shapes of WE and CE were changed to an annular ring by adding a 2 mm hole to the disc-shaped electrodes. The electrode plungers were exchanged by a new plunger which had an outer V4A steel contact sleeve and an inner non-conducting cylinder made from PEEK. In this inner part, a stepped surface (dia. 2 mm) with a small bore (0.5 mm), similar to the micro-RE from FIG. 4-14 was made. This bore, in combination with a thread and a set screw, served as an extrusion hole for the new metallic lithium RE. The design of CELL 2 is depicted in FIG. 4-15. Compared to CELL 1, EIS in CELL 2 lead to a better semicircle resolution of two symmetric LFP electrodes. Nevertheless, FIG. 4-15 also illustrates a still-present 50 % scaling effect. Interestingly, both spectra seem to be independent of the RE position. Both spectra R-LFP1-LFP2 and LFP1-LFP2-R are practically the same. This observation fits to the findings of M. ENDER ET AL., who ascribed scaling effects to the misalignments of WE and CE, which is still possible in this cell[228].

LFP-RE-Li (5 mm RE)

100 LFP-RE-Li (0.5 mm RE)

]



[

im 50 -Z

0 0 50 100 Z [] real

FIG. 4-14 OPTIMIZATION OF THE REFERENCE ELECTRODE DIAMETER: A) STANDARD 5 MM LITHIUM RE USED IN THE SWAGELOK CELL (CELL 1) AND AN ALTERNATIVE 0.5 MM LITHIUM DISC MICRO-RE MADE BY EXTRUSION TO PROBE A SMALLER ELECTROLYTE VOLUME IN CELL 1; B) INFLUENCE OF RE DIAMETER ON THE NYQUIST PLOT OF AN LFP ELECTRODE AT 50 % SOC, LI RE AND LI CE; IN BOTH CASES, THE SECOND SEMICIRCLE IS STILL PRESENT

79 4 RESULTS AND DISCUSSION

R-LFP -LFP 140 1 2

LFP1-LFP2-R

120 R-LFP2-LFP1

LFP2-LFP1-R

100 LFP1+LFP2 (full cell)

] 80



[

im 60 -Z

0 0 20 40 60 80 100 120 140 Z [] real

FIG. 4-15 (CELL2) A) COAXIAL CELL DEVELOPED TO FIT IN½ IN. (12 MM) SWAGELOK CELL; TWO DISK REFERENCE ELECTRODES (EXTRUDED LITHIUM, DIA. 0.5 MM) ARE LOCATED IN-PLANE IN THE CENTRE OF THE ANNULAR ELECTRODES (I. D. 2 MM); MISALIGNMENT IS STILL POSSIBLE B) IMPEDANCE SPECTRA OF TWO SIMILAR LFP ELECTRODES (0% SOC); VERSUS 0.5 MM COAXIAL LITHIUM RE AND AS FULL-CELL

To reduce the apparent misalignment of electrodes, a second generation coaxial cell (CELL 3) was finally designed as depicted in FIG. 4-16. One electrode plunger was replaced by a cell base with a stepped well which fits and centres the first annular electrode. The RE is realised in the same way like cell 2. A second step fits separators with a diameter of 14 mm. To align the second electrode, a PEEK sleeve (o.d. 14 mm, i.d. 12 mm) is then inserted in the cell base. The misalignment of electrodes with this sleeve is believed to be less than 0.1 mm, and highly depends on the precision of electrode cutting. A plunger closes the cell and a steel spring at fixed pitch applies a defined and reproducible pressure on the cell stack. These two additional modifications do not directly influence the EIS measurement itself; however, the option to use a larger separator is beneficial because a thin separator, such as a Celgard foil separator, can be used without the risk of short circuits between WE and CE. The constant pressure was also a prior issue with the Swagelok cell, and now allows for more reproducible experimental conditions.

80 4 RESULTS AND DISCUSSION

electrode LiFePO4 on Al foil composition LFP-C65-S5130 85–10–5 wt.-% load ~5 mg cm-2 SOC 0 % (a); 50 % (b,c) experiment Pstat EIS at OCP, 10 mV amplitude, 100 kHz to 10 mHz in logarithmic spacing,

FIG. 4-16 (CELL 3) A) IMPROVED COAXIAL CELL WITH PRECISE ELECTRODE ALIGNMENT AND DEFINED SPRING PRESSURE SUITABLE FOR THIN SEPARATORS LIKE CELGARD®; FITTED TO ¾ IN. SWAGELOK CELLS; 12 MM ELECTRODES, I. D. 2 MM; 14 MM SEPARATOR; CELL IS ASSEMBLED FROM RIGHT TO LEFT

For the best comparison, the same electrodes, LFP1 and LFP2 from the previous experiment, were taken from CELL 2 and transferred to CELL 3. As can be seen in FIG. 4-17A, both spectra now match each other and no longer show scaling effects.

Kirchhoff rules 1. The sum of all currents at any node of a network is zero 2. The sum of all potential differences around a loop is zero

The addition of the impedances of both electrodes at any frequency is equal to the impedance of the full-cell in any geometry. Yet, it should be kept in mind that judging the quality of the impedance spectra by an overlay comparison of the measured and the expected two electrode full-cell spectrum (by addition of the single spectra), as suggested earlier[249], also holds true for scaled, but wrong spectra (see FIG. 4-13). It is merely a consequence of KIRCHHOFF’S 2ND RULE.

Kirchhoff’s 2nd rule does not validate impedance results.

Nevertheless, it is not yet proven that the geometry of CELL 3 yields valid results in the case of electrochemically asymmetric cells as it is the case in almost all battery cycling experiments. The most common configuration is the half-cell using a lithium metal as CE. One LFP electrode was therefore taken from cell 3 and each LFP1 and LFP2 electrode was measured using a lithium CE. The shape of the Nyquist plot was equal for both electrodes, however, the value for charge-transfer increased from approximately 90 Ω to 110 Ω (see FIG. 4-17B). It should be kept in mind, though, that reassembling electrodes may slightly alter the surface of the electrodes and, since the OCP at 0 % SOC is determined by surface groups, the impedance spectra at this SOC may be very sensitive.

81 4 RESULTS AND DISCUSSION

Therefore, both electrodes were charged to 50 %, which resulted in a charge-transfer resistance of

128 Ω for LFP1 and 126 Ω for LFP2 (see FIG. 4-17C). Not only are these two values similar to each other, but when LFP1 and LFP2 are reassembled to the symmetric setup, they yielded very similar charge-transfer resistances (approximately 121 Ω for LFP1 and 134 Ω for LFP2), corresponding to an error of about 5% between the electrochemically symmetric and asymmetric setup and an error of less than 10 % between the two electrodes. The scaling effect could thus be successfully removed through precise alignment of electrodes.

a) 250

R-LFP1-LFP2 b)

LFP2-LFP1-R LFP +LFP (full cell) 1 2 150 200 R-LFP1-LFP2 LFP1+LFP2 (calc.)

(LFP1+LFP2) /2 (calc.) R-LFP1-Li LFP -LFP -R

150 2 1 ] ] 100 LFP -Li-R

 2



[

100 im

-Z -Z

50 50

0 0 0 50 100 150 200 250 0 50 100 150 Z [] Z [] real real

c) d) R-LFP -Li (flat) 14 1 10 Hz 250 R-LFP1-Li (flat) R-LFP1-LFP2 R-LFP1-LFP2 12 LFP2-Li-R (step) LFP -Li-R (step) 200 2 LFP2-LFP1-R 10

LFP2-LFP1-R ]

150 ] 8



[

im 6 -Z -Z 100 20 Hz 4 50 50 Hz 2

0 0 0 50 100 150 200 250 0 2 4 6 8 10 12 14 Z [] Z [] real real

FIG. 4-17 NYQUIST PLOTS OF TWO SIMILAR LFP ELECTRODES IN CELL 3: A) NYQUIST PLOT OF TWO SIMILAR LFP ELECTRODES AT 0 % SOC SHOWS A PERFECT MATCH OF IMPEDANCES, WHEN ELECTRODES FROM CELL 2 ARE TRANSFERRED INTO CELL 3; B) NYQUIST PLOTS OF SAME ELECTRODES, BUT SEPARATED AND ASSEMBLED WITH LITHIUM CE ALSO MATCHES EACH OTHER; C) SAME ELECTRODES, CYCLED TO 50 % SOC, FIRST WITH LITHIUM CE AND THEN REASSEMBLED TO THE SYMMETRIC SETUP ; D) MAGNIFICATION OF FIGURE C REVEALS INDUCTIVE BENDS ABOVE 10 KHZ

82 4 RESULTS AND DISCUSSION

Although the scaling effect was removed, small inductive bends could be observed in the high frequency part for some cell configurations. They appear in the order of 1-5 Ω and hence mainly influence the determination of electrolyte resistance between WE and RE. As can be seen in FIG.

4-17D, the inductive bend is most pronounced for the configuration R–LFP1–Li, starting from 10 kHz, but it is almost absent in the case of LFP2–Li–R. This behaviour can probably be explained by differences in the cell base used for the EIS experiments. In the first case, the inner edges of the electrodes were in contact with the electrolyte, while in the latter case, the inner edge of the lithium foil was shielded by an additional small step of the thickness of the lithium foil (0.4 mm) around the RE, which levelled it in-plane with the Li CE. Apparently, the inner edges of the electrodes develop an

EDGE EFFECT which causes or promotes inductive loops. The edge effect also seems to impinge the impedance differently, depending on which side the RE is located. Interestingly, these inductive distortions are dependent on a geometric effect, rather than the electrochemical asymmetries[228] to which they were previously attributed. However, in this cell geometry, electrochemical asymmetries do not seem to play a major role, as long as the geometry is controlled at the same time. It should also be pointed out that the inductive distortions observed here seem minor. Nevertheless, they can be significant when materials with a much smaller charge-transfer resistance are investigated, and this newly observed edge effect is not taken care of.

4.2.3 DISCUSSION

In order to optimise the current line distribution and electrolyte potential measured by the RE, the geometry of a common Swagelok cell was altered by i) decreasing the size of RE by a factor of 10, ii) moving the RE to a coaxial position inside the WE or CE and iii) precise alignment of WE and CE. Neither decreasing the RE nor changing the RE position alone yielded reliable EI spectra. In combination with electrode alignment, however, scaling effects could be removed within experimental errors. As a side benefit, the newly developed cell allows for thin separators without the risk of short circuiting and a defined pressure on the electrode stack. A second, geometry-based source of possible distortions in the high frequency range was identified as the EDGE EFFECT. It seems to be worst when the edges of thick electrodes, such as lithium foil, are in contact with the electrolyte. To avoid this effect, it is beneficial to use stepped electrode sockets.

In all three cells, the sum of impedances of LFP1 and LFP2 equals the measured full-cell impedance, according to KIRCHHOFF’S 2ND RULE. Thus, this criterion does not prove the validity of any measured spectrum, since it is only in cell 3 that the calculated half of the measures full-cell spectrum (LFP1- LFP2)/2 also exactly matches the impedance of both LFP1 and LFP2, validating the geometry. This is especially important in cases, when electrodes are tested in asymmetric half-cells and scaling effects

83 4 RESULTS AND DISCUSSION are not obvious. Hence, the only way to validate impedance spectra from a given cell geometry for a given electrode is to compare the EI spectrum of both WE and CE in a symmetric configuration (i.e. WE and CE are equal) with that of the asymmetric cell (see chapter 4.2). While impedance spectra could be obtained with sufficient accuracy for the electrodes tested in this study, it is, of course, of interest how much reliability is limited with the developed cell, such as when different electrodes are used or where distortions exist.

84 4 RESULTS AND DISCUSSION

4.3 RELIABILITY LIMITS OF COAXIAL IMPEDANCE CELLS FROM FEM MODELLING

The methods and results discussed in this chapter were previously published in a research article[272]. Figures and discussions are therefore extended, yet analogous to main conclusions.

CHAPTER OUTLINE Frequency dependent current line distributions can lead to distortions in EI spectra. Finite Element Method (FEM) visualises the current and potential distribution during EIS. FEM modelling thus helps to evaluate the suitability of test cells for EIS experiments. Limiting conditions in coaxial impedance cells could be identified. Distortions in EI spectra can be avoided by electrode alignment and thick separators.

Swagelok three-electrode test cells commonly used in lithium ion battery research are not suitable for impedance spectroscopy, for the reasons discussed in chapters 2.5.4. In short, asymmetric current line distributions cause severe distortions of impedance spectra. This includes scaling, inductive loops and even artificial “ghost” semicircles. In consequence, a new and easy to assemble, three-electrode test cell with coaxially oriented reference electrode was developed to expel these distortions (chapter 4.2, p. 77[248]). The cell was tested with LFP electrodes as a model system and it proved to yield reliable results with these electrodes. It remained, however, an open question, if the CIC yields reliable and reproducible spectra for all relevant types of electrodes, provided that a certain degree of error can occur during electrode cutting and assembly.

This chapter only deals with distortions caused by electrode and cell design. Impedance distortions caused by imperfect instrumentation was not topic of this work.

Instead of performing numerous electrochemical experiments, a FINITE ELEMENT METHOD (FEM) simulation was developed instead. FEM is a numerical technique used to approximately solve partial differential equations (in this case the Laplace equations) in a three-dimensional space by breaking down the space in a finite number of elements. It has been previously used to visualise the current distribution at different frequencies during impedance spectroscopy in solid electrolytes and fuel cells[244–246,243,242]. It was also recently adopted to understand EI distortions observed in Swagelok three-electrode cells for lithium ion batteries[228].

A main issue of interpreting EIS spectra is the difficulty to experimentally validate electrochemical parameters such as the charge-transfer resistance. Distortions due to geometric effects are hard to separate from those due to distribution of reaction rates or equilibrium potentials (i.e. energetic distributions). With FEM - once a sufficiently accurate model of the system of interest has been identified - relevant parameters such as geometric or electrochemical parameters can be

85 4 RESULTS AND DISCUSSION systematically changed. The calculated EI spectra obtained from this three-dimensional model can then be compared to those from the corresponding, dimension-less equivalent circuit. In the latter, the impedance response is completely independent from current line distributions. As such, FEM is practically the only way to visualise current distributions and assess the influence of geometry and non-uniform current distributions on the measured EI spectra.

For this purpose, several sets of geometric and electrochemical parameters were developed to simulate the influence of: i) GEOMETRIC ASYMMETRIES (i.e. misalignments and/or edge effects), ii)

ELECTROCHEMICAL ASYMMETRIES (e.g. varying charge-transfer resistances or double-layer capacitances) and iii) a COMBINATION OF BOTH ASYMMETRIES on the impedance spectra obtained from the coaxial impedance cell. Furthermore, iv) the INFLUENCE OF THE EQC used for the electrode boundary condition will be discussed.

4.3.1 DEVELOPMENT OF THE FEM MODEL

FIG. 4-18 GEOMETRY OF COAXIAL IMPEDANCE CELL AS USED FOR FEM SIMULATION; [272] GRID LINES REPRESENT FEM TETRAHEDRAL MESH

All FEM simulations were conducted using Comsol V 4.2a[273]. The electrode geometry used for FEM simulations was based on the coaxial impedance cell developed in chapter 4.2, p. 77[248]. It consists of two opposing ring electrodes E1 and E2, an electrolyte disc, and a reference electrode RE, see FIG. 4-18. Both E1 and E2 are 12 mm in diameter and have a centre hole with 2 mm diameter. The electrodes are separated by the electrolyte with a diameter of 14 mm and a thickness de arbitrarily set to i) 400 µm or ii) 25 µm. These thicknesses correspond to i) a glass fibre separator frequently used in laboratory test cells and ii) a PE/PP film separator used in commercial lithium ion batteries.

-1 The electrolyte conductivity σel was set to 0.2 S m , which is comparable to a separator (40 %

86 4 RESULTS AND DISCUSSION

porosity) soaked with a rather poorly conducting electrolyte such as 1 M LiPF6 in EC:DEC 1:1 wt.-% electrolyte below 10 °C[6]. In more conductive electrolytes, the effects of asymmetries are likely to be less pronounced[228]. Only the surfaces of E1, E2 facing each other share an interface with the electrolyte. Hence, the thickness of the electrodes (400 µm) was irrelevant for the simulation. The reference electrode was implemented as a disc, which is 0.5 mm in diameter. It was located in a coaxial position within and in plane with E1, sharing only the disc-shaped surface with the electrolyte.

Electrochemical reactions in a porous battery electrode are difficult to constrain in a mathematical model (see chapter 2.5). The simulation of impedance spectroscopy in this case, however, may be conducted by using a much less sophisticated model, since anticipated distortions can be ascribed to the change of current line distributions in the separator (compare [228]). Therefore, the electrochemical behaviour of E1 and E2 was chosen to be implemented by a simplified Randles circuit, consisting of a resistor Rct (charge-transfer resistance of lithium intercalation) in parallel to a capacitor Cdl (capacitance of the electrochemical double-layer) as boundary condition for the interface between E1/E2 and the electrolyte (see FIG. 2-21, p.47 and FIG. 4-19A). For sake of simplicity, diffusion and inductive elements were not considered in the model. Any inductive behaviour observed in the impedance spectra are therefore distortions which originate from frequency dependent current line distributions.

FIG. 4-19 THE TWO EQUIVALENT CIRCUITS USED IN FEM SIMULATIONS A) SIMPLIFIED RANDLES CIRCUIT USED FOR SETS 1-5; B) RANDLES CIRCUIT EXTENDED BY FARADAIC CAPACITANCE (SET 6)

The main goal of the FEM simulations performed in this study is the identification and evaluation of critical parameters that lead to significant distortions in EI spectra. For this, several parameter sets were developed. First, impedance spectra were simulated for a set of two perfectly aligned and electrochemically identical electrodes. These served as reference point for the subsequent introduction of asymmetries (SET 1, FIG. 4-20). Second, the simulation was repeated for the same two, but misaligned electrodes, (SET 2, FIG. 4-21 and FIG. 4-22). For this purpose, E1 was shifted by Δx=200 µm in the x-axis to mimic a geometric asymmetry, for example caused by errors in electrode cutting or improper cell assembly. Third, EI spectra were calculated for two perfectly aligned, but electrochemically asymmetric electrodes (SET 3, FIG. 4-23).

87 4 RESULTS AND DISCUSSION

Geometric asymmetry can be caused by an asymmetric reference electrode position[228] or by a misalignment of electrodes[248]. Electrochemical asymmetry describes two electrodes which differ in their electrochemical characteristics (charge-transfer resistance, double layer capacitance or relaxation frequency).

Fourth, both asymmetries were combined to visualise expected distortions in a cell configuration close to “real” conditions (SET 4, FIG. 4-24). The aforementioned parameter sets were complemented by the introduction of an EDGE EFFECT (SET 5, FIG. 4-25). In this case the electrolyte was expanded to fill the inner volume of E1. Thus, an additional electrode/electrolyte boundary at the inner edge of electrode E1 was introduced. The RE was accordingly moved backward by 400 µm. Finally, the influence of limited charge capacity of the active material particle on the EI spectra was investigated

(SET 6, FIG. 4-26). For this purpose, an additional capacitor CF (Faradaic capacitance, i.e. a pseudo- capacitance) was implemented in the Randles circuit in series with Rct to investigate the influence of interface reactivity on the EI spectra (FIG. 4-19B). An overview of all parameter sets is given in TAB.

4-1 and TAB. 4-2.

Potential and current distributions were computed using a free tetrahedral mesh with a fine mesh around the reference electrode and the inner electrode edges (see FIG. 4-18). Impedance spectra were calculated in the frequency spectrum from 100 kHz to 100 mHz, with 10 frequency points per decade in logarithmic spacing. The spectra of E1 and E2 were calculated using the reference electrode in a three-electrode setup, denominated as R-E1-E2 and E2-E1-R, respectively. The full-cell impedance spectrum was calculated as a two-electrode setup. As reference, impedance spectra were also calculated purely based on the corresponding EQC using the ZSim module of Bio-Logic EC-lab

V10.22, as if the current line distribution were ideally independent from frequency. In this case, Rel was arbitrarily set to fit the FEM simulation.

TAB. 4-1 GENERAL PARAMETERS FOR THE FEM MODEL parameter symbol unit value temperature T K 293 geometric area A m² 1.10E-04

-1 electrolyte conductivity σel S m 0.2 electrode conductivity S m-1 1.00E+09

4 RESULTS AND DISCUSSION

100

R/C

1500

0.96

yes

400

0.2

200

3000

9.65

(R+C)/C

100

R/C

1500

0.96

yes

400

0.2

3000

9.65

(R+C)/C

Set6

100

R/C

1500

0.96

400

0.2

200

3000

9.65

(R+C)/C

100

R/C

1500

0.96

400

0.2

3000

9.65

(R+C)/C

yes

400

0.2

R/C

3000

9.65

Set5

E1=E2

EDANCE CELL EDANCE

150

100

9.65

200

0.2

R/C

3000

9.65

Set4

150

100

9.65

200

400

0.2

R/C

3000

9.65

MODEL OF COAXIAL IMP COAXIAL OF MODEL

FEM

100

1500

0.96

400

0.2

R/C

1500

19.30

6000

4.82

400

0.2

R/C

Set3

PARAMETER SETS FOR SETS PARAMETER

1500

19.30

150

100

9.65

400

0.2

R/C

3000

9.65

200

0.2

R/C

3000

9.65

E1=E2

Set2

200

400

0.2

R/C

3000

9.65

E1=E2

400

0.2

R/C

3000

9.65

Set1

E1=E2

[Ω]

[µm]

[Hz]

S/m

unit

[µF/cm²]

[mF/cm²]

Δx

Rct

Cdl

σel

EQC edge

Parameter

89 4 RESULTS AND DISCUSSION

4.3.2 IDENTIFICATION OF CRITICAL PARAMETERS

4.3.2.1 GEOMETRICAL ASYMMETRIES: ALIGNMENT AND ELECTROLYTE THICKNESS

One major reason, why Swagelok cells cannot be used for reliable EIS, is the missing alignment of electrodes, which leads to severe scaling effects (compare FIG. 4-13, p. 78). This issue has been addressed by developing a coaxial impedance cell with a positioning ring, as described in chapter 4.2. With FEM, the distribution of potential between two same and symmetric electrodes E1 and E2 under ac perturbation of E1 can be visualised (SET 1). As one can see from FIG. 4-20A, the potential in the electrolyte between E1 and E2 distributes evenly, and the area around RE is left sufficiently free from changes in the electrolyte potential in both high frequency and low frequency conditions. Calculation of the impedance response in this configuration yields EI spectra that are very close to the one calculated from the pure equivalent circuit, i.e. perfect semi-circles with a diameter of Rct

(FIG. 4-20B). Under such condition, the influence of the RE position is negligible, i.e. both spectra RE- E1-E2 and E2-E1-RE yield almost identical results. Both ionic current and electrolyte potential can be visualised. Since equipotential lines are perpendicular towards current lines, the information gained is redundant. For qualitative analysis, however, the potential distribution is more illustrative.

FIG. 4-20 A) FEM SIMULATION OF THE POTENTIAL DISTRIBUTION IN THE ELECTROLYTE AT HIGH FREQUENCY ALONG THE CROSS-SECTION OF TWO PERFECTLY ALIGNED, SYMMETRIC ELECTRODES (SET 1); B) CALCULATED IMPEDANCE SPECTRA FOR E1 AND E2 IN COMPARISON TO THE CALCULATED SPECTRUM OBTAINED FROM THE

EQUIVALENT CIRCUIT (ELECTROLYTE RESISTANCE ARBITRARILY SET TO 8.5 Ω)

Unfortunately, real systems are always prone to slight shifts and even in systems with positioning aids, electrodes may shift in the order of 100 µm, for example when electrodes were cut smaller than the positioning ring. It can be seen from FIG. 4-21A that a shift of 200 µm for E1 generates a wider distribution of electrolyte potentials in the proximity of the reference electrode. In the case of thick electrolytes such as a glass fibre separator, the shape of simulated impedance spectra remains

90 4 RESULTS AND DISCUSSION

mostly unaffected (below the experimental error of <1 %, see FIG. 4-21B), resulting only in a shift of electrolyte resistance which is decreasing for E1 and increasing for E2.

FIG. 4-21 SET 2A: INFLUENCE OF ELECTRODE SHIFT ON IMPEDANCE SPECTRA (E1 SHIFTED 200 µM) A) FEM SIMULATION OF POTENTIAL DISTRIBUTION AT HIGH FREQUENCY; B) CALCULATED EI SPECTRA OF E1 AND E2 When the thickness of the electrolyte is decreased to values similar to commercial separators (25

μm, FIG. 4-22), the effect of misalignment on the current density distribution becomes stronger; it results in a scaling effect. Rct of E1 appears to be ca. 0.5 Ω (i.e. 10 %) smaller than its theoretical value, while E2 appears to be 0.5 Ω larger. This effect has been previously described for Swagelok

[228] cells , in which Rct of the electrode closer to the RE (here E1) appears smaller, while the other one appears larger. The scaling is approximately 10 % of the real value, which shows a certain reliability of the cell even with misalignment.

FIG. 4-22 SET 2B: A) POTENTIAL DISTRIBUTION AND B) CALCULATED IMPEDANCE SPECTRA AT HIGH FREQUENCY OF TWO MISALIGNED ELECTRODES (200 µM SHIFT) USING A THIN (25 µM) SEPARATOR

91 4 RESULTS AND DISCUSSION

4.3.2.2 ELECTROCHEMICAL ASYMMETRIES: INFLUENCE OF RCT, FR AND CDL

The second major issue in test cells for lithium ion batteries is the obvious difference in the electrochemical behaviour of electrodes. The most common arrangement is probably the half-cell setup, in which a porous WE composed of negative or positive active materials with addition of conductive additives and binder is combined with a metallic lithium CE serving as infinite source of lithium ions. Nevertheless, this combination is probably one of the most difficult in terms of EIS, since the metallic lithium is known to have a large impedance compared to porous electrodes[249]. In addition to this, its surface morphology and surface area is constantly changing depending on formation of dendrites, dissolution and aging. To investigate the influence of the electrochemical parameters, three parameter sets were conceived (SETS 3A-C) keeping the thickness of the electrolyte

(400 μm) constant. In SET 3A, Cdl was kept constant and Rct was changed. In SET 3B, Rct was kept constant and Cdl was decreased. Finally, in SET 3C fr was fixed and Rct and Cdl were changed accordingly. FIG. 4-23A shows an example of potential distribution in the electrode for SET 3B. From figure FIG. 4-23B it can be deduced that the parameters influencing the distortion of the semicircle in

E1 are due to differences in Cdl, which generate the inductive loop and differences in fr from which the scaling effect derive. The most common case that appears by using metallic lithium as CE, however, is the third one, in which the relaxation frequency does not change too much (resistance increases and capacitance decreases), and therefore inductive loops at high frequency are typically observed in real half-cell setups.

FIG. 4-23 SET3A: INFLUENCE OF ELECTROCHEMICAL ASYMMETRIES: A) EXEMPLARY POTENTIAL DISTRIBUTION FOR ALIGNED BUT ELECTROCHEMICALLY ASYMMETRIC ELECTRODES (SET 3A): THE ELECTROLYTE CLOSE TO THE RE IS HOMOGENEOUS, BUT DOMINATED BY E1; B) CALCULATED IMPEDANCE SPECTRA OF THREE DIFFERENT

ASYMMETRIES: DIFFERENT RCT, FR AND/OR CDL

92 4 RESULTS AND DISCUSSION

4.3.2.3 COMBINATION OF GEOMETRICAL AND ELECTROCHEMICAL ASYMMETRIES

The previous case studies anticipated EI distortions caused by geometric asymmetries expressed as x-shift of electrode E1 and electrochemical asymmetries caused by differences in Rct, Cdl and fr. The most realistic scenario, however, is a combination of both types of asymmetries. In the worst case, the electrode of interest has a small Rct compared to the CE. This is probably the most common case, when lithium foil is used as CE.

In this case, the calculated EI spectrum highly depends on the electrolyte thickness. For rather thick electrolytes (SET 4A, FIG. 4-24), the impedance appears about 0.5 Ω smaller compared to the reference spectrum and therefore yields still acceptable results. For thin electrolytes (SET 4B), however, a large difference can be observed: the charge-transfer process seems to be composed of two semicircles, giving a highly distorted form of the process, and an absolute value of the charge- transfer that is more than 10 Ω (double as large as expected). In real experiments, this is often attributed in to energetic distributions.

FIG. 4-24 SET 4A: COMBINED GEOMETRIC AND ELECTRICAL ASYMMETRIES: A) POTENTIAL DISTRIBUTION

AND B) SIMULATED IMPEDANCE SPECTRA OF (SAME CDL, DIFFERENT RCT AND FR) IN THICK AND THIN ELECTROLYTES

4.3.2.4 EDGE EFFECTS

Another effect is observed in the real experiment when a rather thick metallic lithium foil is used as CE. In this case, the inner edge of the lithium CE is likely to be in contact with the electrolyte and distortions (inductive loops or bends) are observed in the high frequency region (above 10 kHz). Instead, when the inner edge of the CE was covered with an insulating material (realised through a small step in the cell base), these distortions were almost vanishing (see FIG. 4-17, p. 82). The latter case was assumed in the previous simulations, that is, it was assumed that the electrode edges

93 4 RESULTS AND DISCUSSION would not share a boundary with the electrolyte.

To investigate the first case (edge effect present), the FEM model was adapted to the new geometry with SET 5 (see FIG. 4-25). The edge effect can appear in three different ways, depending on the position of the reference electrode with respect to the working electrode. For E1 (RE on same side), the edge effects appears as an inductive bend of the semicircle in the high frequency region; the semicircle also appeared smaller than expected. For E2 (RE on opposite side), the electrolyte resistance appeared higher and the semicircle appeared larger. Additionally, a capacitive bend (i.e. the opposite of an inductive bend) can be observed (see FIG. 4-25B).

FIG. 4-25 INFLUENCE OF EDGE EFFECTS, I.E. WHEN THE ELECTROLYTE IS COVERING THE INNER EDGE OF ELECTRODES. A) EXAMPLE OF ELECTROLYTE POTENTIAL DISTRIBUTION AND B) SIMULATED IMPEDANCE SPECTRA FOR PERFECTLY SYMMETRIC ELECTRODES

4.3.2.5 INFLUENCE OF BLOCKING ELECTRODES

Active materials for lithium-ion batteries have a limited amount of storage capacity. This fact can be represented in the EQC by a faradaic capacitance (i.e. a pseudo-capacitance), Cf, in series with the faradaic Rct (see FIG. 4-19B). Taking into account that most of the tests performed in research and development are performed against a metallic lithium foil CE (half-cell configuration), it becomes evident, how important it is to anticipate possible distortions due to the coupling of two electrochemically different electrodes, i.e. one with pseudo-capacitive behaviour and one with purely faradaic behaviour. As discussed previously, the use of a thin separator amplifies all distortions rising from misalignment and edge effects. Thus, it should be avoided in the study of the fundamentals of the reaction mechanism. Only thick separators have been taken into consideration in this paragraph.

As can be seen for SET 6 in FIG. 4-26A, an inductive loop is observed in the high frequency region in

94 4 RESULTS AND DISCUSSION

absence of any edge effect (FIG. 4-26A). This is due to the strong difference in the Cdl between the two electrodes as discussed in section 4.3.2.2, p. 92. Interestingly, when there is no misalignment between the two electrodes, the EI spectrum shows a slight bending of the capacitive behaviour Cf towards right, resembling a constant phase element and an ideal behaviour of the capacitances at the lowest frequencies. This distortion is due to the strong difference in relaxation frequency between E1 and E2. In presence of the edge effect (FIG. 4-26B), the inductive effect remains and additionally the scaling effect is observed with a reduction of the apparent Rct of ca. 20 %. It should be mentioned that in the medium frequency range the EI spectrum does not rise straight up but bends towards left. This effect is from time to time observed in experimental data, but was not discussed previously due to the fact that it is restricted to a small frequency range, after which the classic capacitive behaviour is observed (FIG. 4-27). Although the frequency range, in which this effect appears, is small, it is very important because it seems to be indicative for the edge effect in the most commonly used three-electrode configuration.

FIG. 4-26 ELECTRODE IMPEDANCE WITH PSEUDO-CAPACITANCE. A) WITHOUT, B) WITH EDGE EFFECT

95 4 RESULTS AND DISCUSSION

4.3.3 DISCUSSION

The influence of geometric and electrochemical asymmetries on EI spectra was investigated using FEM simulations. Under the assumption of physically reasonable asymmetries, it was anticipated that calculated impedance spectra show only small deviations (less than 10 %), when a 400 μm thick electrolyte is used. Nevertheless, more severe distortions were observed when a thin electrolyte of 25 µm (such as a Celgard separator) was used for simulation. The appearance of electrochemical asymmetry depends on the type of FIG. 4-27 EI SPECTRUM OF AN LFP ELECTRODE (82 % LFP, 1 % CMC; 2 % SBR, 10 % C65, 5 % + asymmetry. Inductive loops could be observed, when a CE GRAPHITE), RECORDED AT OCP (~3.4 V VS. LI /LI), AFTER ASSEMBLY AND EQUILIBRATION; NOTE: THE X- with low double layer capacitance Cdl was used. However, AXIS WAS STRETCHED TO ILLUSTRATE THE DISTORTED these distortions were rather small. Scaling effects were CAPACITIVE BRANCH observed from coupling electrodes with very different relaxation frequencies. When the electrodes are not well embedded in the insulation and the edges of electrodes are wetted by the electrolyte, the so-called edge effect is observed. This is causing severe distortions in terms of scaling effects and inductive loops at high frequency. Finally, when the finite capacity of the active material was taken into account, the capacitive branch in the medium to low frequency range bent towards left when an edge effect was present. This simulation may partially explain the observation sometimes made in experimental test cells. The fact that “left-bends” are typically (if at all), observed with electrodes before they are cycled for the very first time (i.e. at fully intercalated state), may speak for an appearance only with very small pseudo-capacitances.

96 4 RESULTS AND DISCUSSION

4.4 VERTICAL DISTRIBUTION OF CURRENT, CHARGE AND CHARGE LOSSES

The methods and results discussed in this chapter are currently being prepared to be published in a research article[274]. Figures and discussions are extended, yet analogous to the main conclusions.

“The language of experiment is more authoritative than any reasoning: facts can destroy our ratiocination – not vice versa.”

- Alessandro Volta[13]

CHAPTER OUTLINE In a multiple working electrode (MWE), layers are in ionic contact but electrically isolated. MWEs allow for the discrete vertical analysis of current distributions in porous electrodes. Graphite negative electrodes get lithiated stage-by-stage and layer-by-layer. Several mass-transport as well as non-mass-transport limited processes could be identified. Local current densities can be double the average, especially at the beginning of each stage. Diffusion of additive and lithium lead to significantly different SEI on the electrode top layer.

As it was described in chapter 2.3, porous electrodes are characterized by vertical (i.e. perpendicular to the electrode surface) distribution of current. In an appropriate simplification, this current distribution can be simulated in good approximation by a transmission line model consisting of small,

[114] [116– discretized electrode layers as developed by V. S. DANIEL-BEKH , E. EULER and W. NONNENMACHER

118,275,276] [214,119] [122] , and further developed by R. DE LEVIE , J. S. NEWMAN and C. W. TOBIAS (see FIG. 2-13, p. 32). The method of direct current distribution analysis, first published by J. J. COLEMAN, actually utilizes the experimental equivalent of electrode layer discretization[115]. A “thick” electrode is sliced into a finite amount of “thin” layers which are believed to have a sufficiently homogeneous current distribution inside. Each “thin” layer is electrically insulated by porous separators. The cross current in each layer under polarization of the whole MULTI-WORKING ELECTRODE (MWE) is then measured by a zero resistance ammeter that solely acts as a current follower. The ladder-type connection of each MWE layer effectively reduces the electric resistance of the solid phase. Nevertheless, since this work will concentrate on graphite electrodes, it is anticipated that electric conductivity will be substantially higher than the ionic conductivity. Hence, the current distribution is not much influenced by the additional electric connection. The ionic resistance is, in the ideal case, only negligibly increased by the addition of a thin separator.

J. J. COLEMAN’s idea has been recently applied to lithium ion batteries by F. LA MANTIA and S. H.

[139,140] NG . A lithium ion battery electrode has a thickness in the range of 50-200 µm and a mass loading of 10-20 mg cm-2. Electrodes can therefore not be physically split into several layers. Instead, several “thin” porous electrode layers have to be prepared individually. FIG. 4-28 depicts a schematic

97 4 RESULTS AND DISCUSSION of a MWE in comparison to a “thick” classical electrode. The virtual splitting of film electrodes, however, comprises several challenges on each WE layer. The resolution of current distribution is specified by the amount of layers of which the MWE stack is composed of. Considering an amount of at least 5 layers, each WE layer would be 10-40 µm thick, which is in the range of thicknesses of standard current collectors alone. Hence, alternative electrode architectures need to be developed with an overall electrochemical behaviour of the MWE equivalent to that of a comparable “thick” normal electrode with the same mass loading and thickness. The multi-layer cell itself needs to provide an environment similar to classical test cells such as Swagelok or pouch cells. Issues that should be approached here include alignment of the MWE stack and CE, electric connections, supply of electrolyte and application of a constant and even pressure.

In the following sections, a multiple working electrode stack (chapter 4.4.1) and a multi-layered electrochemical cell (chapter 4.4.2) with a high vertical resolution suitable for precise charge/discharge experiments will be described. Furthermore, the distribution of current during galvanostatic charge and discharge of SFG6 and SFG44 graphite electrodes will be described in detail, including insights into the distribution of charging currents (chapter 4.4.3) and irreversible charge losses (chapter 4.4.4). Finally, the analysis of electrolyte concentration in-situ during charge/discharge cycling will be discussed and prerequisites for experimental determination identified (chapter 4.4.5).

FIG. 4-28 NORMAL AND MULTIPLE WORKING ELECTRODE A) SCHEMATIC OF A “THICK” ELECTRODE COATED ON COPPER FOIL. B) SCHEME OF A MWE STACK CONSISTING OF THREE “THIN” ELECTRODES COATED ON COPPER MESH ELECTRODES

98 4 RESULTS AND DISCUSSION

4.4.1 DEVELOPMENT OF A MULTIPLE WORKING ELECTRODE

The MWE stack design used in this study was developed based on the earlier report of F. LA MANTIA ET

[139,140] AL. . Due to the drawbacks of his experimental design already discussed in chapter 2.3, including alignment of the electrodes and limited vertical resolution, a different current collector was chosen. Each layer should be sufficiently thin to develop a stack of layered electrodes, which resemble a real “thick” electrode. The current collector needs to be as thin as possible but supply sufficient mechanical support for the electrode. It must provide electrons in the lateral direction and, at the same time, have a high porosity so that the ionic conductivity of the electrode is not limited by the current collector. Thin, electro-deposited copper meshes were therefore chosen as substrate for the preparation of electrode layers. The thickness of these meshes is stated to be 0.00065” (16.5 µm), which is three times thinner than commercially available expanded copper grid[277]. With 250 wires per inch, the open space between two wires is 85 µm, resulting in an open mesh coverage of 70 %.

As active material, commercial graphites, TIMCAL SFG6 and SFG44 were chosen. Both are natural graphites; SFG6 has a size distribution of 5.5-7.5 µm (D90)[278], while SFG44 has a size distribution of 44-53 µm (D90)[279]. Given the square shape of the openings, the length of the electric pathway from the copper grid to about 90 % of all graphite particles is less than around 30 µm, i.e. about six to eight particle-particle contacts for SFG6 and two or three contacts for SFG44. The electrode slurry and coating was prepared similar to the one used for standard electrodes (graphite:C65:S5130 of 90:5:5, see chapter 6.3.3, p. 129). The resulting electrode had a thickness of about 33 µm, corresponding to several monolayers of graphite platelets. FIG. 4-29 shows an SEM cross-sectional view of such an SFG44 electrode as well as a top view by SEM and optical microscopy.

FIG. 4-29 A MWE LAYER UNDER THE MICROSCOPE: SEM CROSS-SECTION AND OVERVIEW AND OPTICAL IMAGE OF MC33 ELECTROLYTIC COPPER MESH COATED WITH SFG44 GRAPHITE PARTICLES. THE THICKNESS OF ONE LAYER IS 33 4 µM.

99 4 RESULTS AND DISCUSSION

4.4.2 DEVELOPMENT OF A MULTI-LAYERED ELECTROCHEMICAL CELL

FIG. 4-30 SCHEME OF THE MLC IN THE ASSEMBLED STATE, IT HOLDS 1 CE AND UP TO 6 WE OR RE

The MLC developed in this study was, unlike the pouch cell used by La Mantia, made of rigid parts similar to those used e.g. in a Swagelok cell. V4A stainless steel was used for all conducting parts and PEEK was used for all insulating parts. CAD drawings and fabrication was done by the team of

A. LINDER (central workshop RUB). Detailed description of the cell can be found in section 6.1.3, p. 122 and technical drawings can be found in the appendix. The MWE stack consists of up to six layers of 18 mm working electrodes. Each WE was electrically contacted by a thin strip (3x10 mm) of expanded copper grid underneath the electrode covering less than 10 % of the electrode area. For electric separation of the MWE, 22 mm Celgard polyolefin separators were used. A photo of the

MWE stack placed in the cell base can be seen in FIG. 4-31. The MWE was fixed on its outer rim by the fixation ring. A lithium CE was placed on the electrode stack, separated by an additional GF/D glass filter to block possible growth of lithium dendrites from the CE. The WE contact tabs were then clamped by six V4A setscrews in the fixation ring. These setscrews were contacted from above by gold-plated contact springs located in the upper body of the cell. Each contact feed-through was hot- pressed into the upper body to guarantee air-tightness. A spring served both for a reproducible load on the electrode stack as well as for electric contact to the CE.

100 4 RESULTS AND DISCUSSION

FIG. 4-31 PHOTOGRAPH OF A MWE STACK CONSISTING OF 6 EQUAL SFG6 GRAPHITE MESH ELECTRODES

Fig. 4-32 deteriorated copper plating In the beginning, gold and copper-plated contact springs were used for directly contacting the current collectors. Unfortunately, small amounts of electrolyte crept over the contact strips and created an ionic contact. Lithium alloying with gold and formation of SEI created significant contact resistance and distortion of electrochemical results. Lithium plating was also observed at high resistance contact points.

For the operation of the MLC, a Bio-Logic VMP-3 multi-channel potentiostat was used since it allows for easy grouping of several independent channels connected to the same counter electrode. A master channel was used for galvanostatic (or potentiostatic) control of the MWE stack. The cross- current through each MWE layer was monitored via a current follower in between electrode and master channel (see FIG. 4-28). For this purpose, all current followers were ‘grouped’ (CE to ground) using the zero resistance amperometry (ZRA) technique. For data analysis the same data acquisition rate (10 s) was used. The detailed wiring is described in chapter 6.4.3.

The ZRA technique is similar to a standard chronoamperometry (CA) at 0 V between WE and CE/RE. In CA, however, significant background currents were measured.

101 4 RESULTS AND DISCUSSION

4.4.3 CURRENT DISTRIBUTION IN GRAPHITE ELECTRODES

The most striking difference of the MWE stack compared to a normal electrode is apparently the presence of the polyolefin separators in between the layers, which adds the ionic resistance of the separator Rion,s to the ionic resistance of the pores Rion,p (see FIG. 4-33). The electric resistance of the solid phase is also different compared to a normal electrode. If the resistance of the contact wires is neglected, the total electric resistance of the MWE is de facto reduced to the electric resistance of a single layer, since each layer is contacted in parallel. This fundamental change of wiring certainly influences the current distribution of each electrode by effectively eliminating the electric contribution of the double cosh-function to the cross-current distribution. Nevertheless, since graphite is a very good conductor (below 0.1 Ω cm compared to ca. 100 Ω cm for the electrolyte[53,6]), the influence of electric conductance to the cross-current distribution in a graphite porous electrode is very low anyway. It can therefore be anticipated that the cross-current distribution is almost exclusively governed by the ionic conductivity.

Furthermore, the addition of separators increases the electrolyte bulk volume. On the macro-scale, this effectively “expands” the electrode: the porosity of the MWE as well as the thickness appears increased compared to the MWE without separators. As a result, it can be anticipated that the current density within the MWE stack will be somewhat more condensed towards the counter electrode. Analysis therefore has to be done carefully. Nevertheless, the presence of separators is not expected to change general trends in current distribution, since no lithium is consumed within the separator.

FIG. 4-33 EQUIVALENT CIRCUIT DIAGRAM OF A THREE-LAYERED MWE INCL. THE POLARIZATION IMPEDANCE

OF THE LAYERS ZL1-L3 AND THE IONIC RESISTANCE OF THE PORES RION,P, THE SEPARATOR RION,S, AND THE BULK

ELECTROLYTE CLOSE TO THE CE RION,CE; ELECTRIC RESISTANCE OMITTED

102 4 RESULTS AND DISCUSSION

Disambiguation - cross current density and ionic current density: In any electrochemical reaction, the electric current itotal is converted throughout the electrode to ionic current of the same value. The ionic current density accordingly increases continuously towards the electrode surface. The term current density in the text refers to the measured cross-current density iL1-L6 in each lamellar.

Impedance spectroscopy is a convenient tool to investigate the electrolyte resistance. As described in chapter 2.5, EIS can be used to determine the sum of electric (neglected) and ionic (Rion,p and Rion,s) resistances. The sum can be read in the Nyquist plot as the high frequency intercept of the spectrum with the x-axis. FIG. 4-34 shows that the ionic resistance in relation to the CE linearly increases with layer depth. The slope (expressed as Ω per layer) is higher than for a real electrode since it consists of two contributions. In this cell configuration, Rion,p and Rion,s cannot be separated. They can, however, be determined either by multiple experiments with increasing numbers of separators (or electrode layers) or estimated by the electrolyte conductivity, thickness and porosity of the electrode or separator, respectively. Extrapolation of the measured electrolyte resistance to zero layers yields the electrolyte resistance of the counter electrode separator, which was a glass filter in this case. Values of the x-intercept and slopes vary with experimental conditions and are prone to changes in the electrolyte composition, temperature, single layer thickness, layer depth, and porosity.

FIG. 4-34 A) ELECTROCHEMICAL IMPEDANCE SPECTRA OF 5 SFG6 LAYERS VS. LITHIUM CE/RE AFTER ASSEMBLY AT OCP (~3.4 V) B) SUM OF IONIC RESISTANCE VS. ELECTRODE LAYER DEPTH DETERMINED FROM THE REAL PART INTERCEPT AT HIGH FREQUENCIES. THE IONIC RESISTANCES INCREASES LINEARLY; THE INTERCEPT RESULTS FROM THE GLASS FIBRE FLEECE SEPARATING THE COUNTER ELECTRODE

To investigate the current distribution in graphite electrodes, two sets of experiments were designed. In the first set, SFG6 was cycled in a ready-to-use LP40 electrolyte. To reduce irreversible charge losses and to investigate the influence of vinylene carbonate as SEI forming agent, the second

103 4 RESULTS AND DISCUSSION set of experiments was performed with SFG44 electrodes and an LPP40 electrolyte incl. 2-wt.-% of VC. Galvanostatic Cycling with Potential Limitation (GCPL) was chosen as charging protocol. The SFG6 and SFG44 electrodes were cycled galvanostatically at a total current density of C/35 (10 mAh/g) and C/10 (35 mAh/g), respectively, in a window of 1 V to 20 mV vs. Li+/Li. These rates are assumed to be sufficiently low to allow for even formation of the SEI layer. A potentiostatic step was included at both potential ends to maintain the potential until the total current had dropped to one quarter of its original value.

It had to be validated that the electrochemical behaviour of the MWE setup was similar to that of a standard “thick” electrode. Therefore, a SFG6 electrode with a mass loading of ca. 12 mg cm-2 was prepared by the standard slurry technique and cycled with the same parameters as the MWE with a

-2 total mass loading of about 9 mg cm . As it can be seen from FIG. 4-35, the overall shape of the GCPL potential profile of the 5x MWE stack is comparable to the thick electrode. A certain difference in charge capacity and charge losses can be observed; the overall shape of the potential profile, however, is very similar. The MWE gets, similar to normal electrodes, lithiated STAGE-BY-STAGE.

FIG. 4-35 A “THICK” SFG6 ELECTRODE COMPARED TO A MWE, SHOWING THE FIRST TWO CHARGING CYCLES OF A SINGLE THICK ELECTRODE 12 MG/CM² COATED ON DENDRITIC COPPER FOIL AND A 5-LAYERED -2 MWE WITH A TOTAL MASS LOADING OF 9 MG CM ; C-RATE WAS C/35 (10 MA/G), LI CE/RE

The first cycle of lithium intercalation in graphite is significantly different from the subsequent cycles, due to solvent reduction and formation of the SEI layer. The analysis of current distribution during the stage of SEI formation is therefore separately treated in chapter 4.4.4.

During electrochemical reduction of the SFG44 MWE, the total current defined by the master galvanostat first distributes evenly throughout the six electrode layers. The current density for each

104 4 RESULTS AND DISCUSSION layer in the potential range from 1 V to about 200 mV is close to the median current density of the

-1 whole stack (that is 35 mA g , see FIG. 4-36). Within this part of the (sloping) potential profile, SEI formation and exfoliation (if not finished in the first cycle, see the next chapter) may continue and, especially below 400 mV, disordered nucleation of lithium intercalation compounds takes place[280].

When a sufficient amount of lithium ions has intercalated, the first phase formation (STAGE 3) with a composition of LiC18 can take place, and the typical potential plateau of a two-phase intercalation reaction can be observed at ca. 200 mV. Following the porous electrode theory, the distribution of cross-current densities does not only depend on the ionic resistance of the electrode but also on the polarization behaviour of the electrode in each layer. When the kinetics of faradaic reactions is the limiting factor, as it is obviously the case in this region, the current densities will distribute evenly.

Just at the beginning of the first potential plateau, the current densities in the layers start to diverge. In the first layer L1 (closest to the CE), a steep increase of current to a maximum value of about 75 mA g-1 was observed within a few minutes. This current density was more than twice the average current density of the MWE (that is 35 mA g-1). Also the second layer L2 observed, slightly later, an increased current, peaking at 48 mA g-1. Since the total current was set constant by the master galvanostat, the currents in the other layers L3-L6 decline. The lowest layer L6 showed the smallest proportional current; its value decreases to about one half of its original value. After reaching their peak current, the currents in the upper layers became smaller while the currents in the lower layers became larger. Within one stage, the MWE is lithiated LAYER-BY-LAYER. The current densities of the lowest layers iL5 and iL6 peaked, when the first layers almost reached the maximum charge of the first stage. This behaviour can be well explained by MASS-TRANSPORT LIMITATION. After quick depletion of lithium ions in the pores, the cross-current density in each layer depends on the flux of lithium ions by migration and diffusion. Since the kinetics of lithium intercalation is fast, a large fraction of the lithium ion flux from the counter electrode is consumed within the upper layers, effectively creating a significant concentration gradient. Just when the upper layers are completely charged and stop consuming lithium, the concentration gradient decreases and the concentration overpotential for the lower layers becomes smaller.

Towards the end of the first plateau, the current densities distribute evenly again. Unlike the previous potential drop, the potential drop towards the second plateau ( ) is accompanied by a (very small) divergence of current densities (35+/- 5 mA/g), suggesting slightly faster kinetics of crystallite growth. On the second and third plateau, the divergence of currents is repeated. The concentration overpotential of the lower levels, however, is apparently so high that lithium intercalation is not yet finished, when the cut-off potential of the MWE (20 mV) is reached. More

105 4 RESULTS AND DISCUSSION

than two hours were needed to sufficiently fill the lower layers. As can be seen in FIG. 4-37, a similar trend in current distribution during oxidation of the MWE can be observed. The deintercalation also proceeds stage-by-stage and layer-by-layer, although the current peaks appear less pronounced. The last current peaks before full delithiation, however, show a larger overlap. As a result, all layers are delithiated at the same time and basically no potential hold is needed for full delithiation in contrast to the lithiation process.

ND FIG. 4-36 POTENTIAL PROFILE AND CURRENT DISTRIBUTION DURING 2 CYCLE INTERCALATION -1 IN A SIX-LAYERED MWE, SFG44 GRAPHITE CHARGING CURRENT C/10 (35 MA G ), LI CE/RE

ND FIG. 4-37 POTENTIAL PROFILE AND CURRENT DISTRIBUTION DURING 2 CYCLE OXIDATION -1 IN A SIX-LAYERED MWE, SFG44 GRAPHITE, CHARGING CURRENT C/10 (35 MA G ), LI CE/RE

106 4 RESULTS AND DISCUSSION

4.4.4 SEI FORMATION AND IRREVERSIBLE CHARGE LOSSES

The first cycle for electrochemical reduction of graphite electrodes differs significantly from the subsequent cycles. The potential profile quickly drops from about 3.3-3.4 V vs. Li+/Li (typical open circuit potential due to redox behaviour of surface groups) to below 2 V. As can be seen from FIG. 4-38, the most distinct feature in the first cycle potential profile is a shoulder at around 0.7 V, which appears for both types of graphite. It is commonly attributed to solvent reduction and formation of the SEI (see section 2.2.4.3, p. 27). Furthermore, two additional reductive processes above 1 V can be identified in the charge derivative plot of the SFG44 electrode (FIG. 4-39): the first reduction takes place at around 1.3 V (onset 1.4 V) and the second takes place at 1.0 V (onset 1.15 V). These

[60] potentials match the onset potentials of solvent reduction as reported by K. XU . The reduction potentials of ethylene carbonate and diethyl carbonate, the two solvents used in the electrolyte, are about 1.36 and 1.32 V, respectively (see, p. 2.2.4.3, p. 27). Vinylene carbonate (VC), which was added in some experiments as an SEI forming additive (2 wt.-%), has an onset potential of 1.40 V, and should therefore be reduced at potentials slightly above those of the solvents.

Interestingly, the analysis of current distribution in the 6 MWE layers reveals that the current in the first layer was occasionally significantly higher than in all the other layers; at times it was twice as large as in the other layers. In the case of SFG44 (2 wt.-% VC, C/10), two reductions waves could be resolved for potentials above 0.6 V (end of the “SEI plateau”), whereas the first one (above 1 V) may be composed of several reduction processes (see FIG. 4-38). Below 0.6 V, after a total charge consumption of about 16-17 mAh g-1, the currents distributed equally again. Until this point, the charge consumed in layer 1 was about 3 mAh g-1 higher than it was for the other layers. For SFG6 (no VC added, C/35), the current distribution was similar, however, only one distinct peak centred at around 1 V could be identified in the first layer.

107 4 RESULTS AND DISCUSSION

FIG. 4-38 FIRST CYCLE CHARGE LOSSES IN SFG44, MAGNIFICATION OF FIG. 4-36, FIRST CYCLE POTENTIAL PROFILE AND VERTICAL CURRENT DISTRIBUTION DURING GALVANOSTATIC CYCLING OF A SIX-LAYERED MWE, -1 SFG44 GRAPHITE, CHARGING CURRENT C/10 (35 MA G ), LI CE/RE

FIG. 4-39 CHARGE DERIVATIVE PLOTS OF THE TOTAL CURRENT DURING THE FIRST AND SECOND -2 -1 GALVANOSTATIC REDUCTION CYCLE OF A MULTI-LAYERED SFG44 ELECTRODE; 6X 2 MG CM , 10 MA G , LI CE/RE, CELGARD SEPARATOR. IN THE FIRST CYCLE, 3 DERIVATIVE PEAKS CAN BE IDENTIFIED BEFORE LITHIUM INTERCALATION, CORRESPONDING TO REDUCTION PROCESSES AT CA. 1.3 V (ONSET 1.4 V), 1.0 V (ONSET 1.15 V) AND 0.6 V (ONSET 0.7 V).

Several significant conclusions can be drawn from these observations. First, the reduction reactions above 0.6 V are apparently controlled by mass-transport, since the reduction current is higher in the first layer due to an additional influx of species from the electrolyte bulk located in the pores of the glass fibre separator between MWE and CE. One MWE layer (18 mm diameter, thickness ca. 35 µm) at ca. 40 % porosity holds an electrolyte volume of roughly 3.5 µL. Furthermore, the electrolyte in the separator (ca. 23 µm thick) holds roughly another 2 µL and the electrolyte in the whole MWE stack incl. separators has a volume of about 30 µL. At an additive concentration of 2 wt.-% VC, roughly

108 4 RESULTS AND DISCUSSION

0.009 mmol VC is readily available for reduction in the pores of the MWE. At a total current rate of C/10 (ca. 1 mA), the molar conversion rate of any reduction with one-electron-transfer is about 0.036 mol h-1:

∆ ⁄ EQ. 4-1 ⁄

Since the second layer already shows similar currents compared to the deeper layers 3 to 6, all additional VC diffusing from the CE is apparently already reduced within the pores of layer 1. The number of 3 mAh g-1 may appear small in comparison to the total amount of ICL. It should be noted, though, that VC can be easily electrochemically polymerized by a radical chain reaction. Only the initiation of the electro-polymerization (see FIG. 2-8, p. 22) involves the transfer of electrons. As a consequence, small amounts of charge can induce the polymerization of large amounts of VC, and the consumption of VC can be significantly larger in the uppermost layer compared to the lower layers. This should be addressed with special care considering the stability of the SEI towards exfoliation as well as prolonged cycling.

Note: The electrolyte volume in the separator is significantly less in “real batteries” since no thick glass fibre separator is used. Nevertheless, instead, the porosity of the “real”, porous cathode, allows for extra electrolyte volume.

The second conclusion relates to the appearance of the current peak in the first layer of the SFG6 electrode (see FIG. 4-40), and the second reduction peak in the SFG44 electrode (FIG. 4-38), respectively. Since the SFG6 electrode did not contain VC as additive, only EC or DEC can be reduced. However, since both components are used as solvents, mass transport limitation seems unlikely for these reactants. Linear carbonates such as DEC are generally believed to be reduced to organic lithium salts via a one-electron mechanism, while EC, as cyclic carbonate, is believed to be either reduced to lithium carbonate via a two electron-transfer or to form dimers and oligomers after one- electron reduction[60]. Therefore a depletion of lithium ions may explain the favoured current distribution into the first layer instead. Assuming again an electrolyte volume of about 30 µL in the MWE, about 0.03 mmol of lithium ions are readily available for counterbalancing the charge of the solvent decomposition products. The total current of about 1 mA (C/10, SFG44) and 225 µA (C/35, 10 mA g-1, SFG6) equals a molar depletion of lithium salt of roughly 0.036 and 0.008 mmol per hour: Hence, a complete turnover of lithium ions takes place in less than one hour for a C-rate of C/10 and in 3 h for a C-rate of C/35. Of course, these values are rough estimates. Nevertheless, this example anticipates how much lithium is depleted during reduction.

109 4 RESULTS AND DISCUSSION

FIG. 4-40 ICL IN A SFG6 MWE, FIRST CYCLE POTENTIAL PROFILE AND VERTICAL CURRENT DISTRIBUTION -2 DURING GALVANOSTATIC CYCLING OF A 5-LAYERED SFG6 MWE; CA. 9 MG CM ; LP40 ELECTROLYTE, NO -1 VC ADDED, 10 MA G (C/35)

In the potential region below 0.7 V, down to about 130 mV, the reduction current is evenly distributed again; no mass-transfer control was observed anymore. There are two possibilities for explanation. Either, the reduction does not involve any reactants that are significantly limited by mass transport. This would be the case for the release (or dissolution) of negatively charged species into the electrolyte such as anionic organic compounds. Or, mass transport is not the rate-limiting step in the reaction and the polarization resistance of the electrolyte decomposition is significantly higher than the ionic resistance. The reasons for increased resistance include: i) an increase of the charge-transfer resistance for further reduction of the SEI, ii) a hindrance for lithium ions to migrate through the pre-form of the SEI or iii) following the Besenhard model of SEI formation: slow kinetics of solvent co-intercalation and gas formation as necessary steps for SEI formation[88]. La Mantia applied differential electrochemical mass spectroscopy to investigate the formation of gaseous compounds during SEI formation and found that ethylene gas evolves in the potential region between 700 to 200 mV[139].

Rate-limiting reduction reactions above the potential of lithium intercalation, however, do not necessarily need to be related to SEI formation. FIG. 4-39 shows that the potential of stage 3 lithium intercalation (i.e. a two-phase conversion) is 130 mV in the first layer, but 180 mV in the subsequent cycles. It can therefore be expected that the first nucleation of diluted graphite intercalation compounds (GIC), which is a necessary step to enable facile lithium intercalation, proceeds as rate-

110 4 RESULTS AND DISCUSSION limiting single phase reactions. As soon as the growth of GIC nuclei has sufficiently started, lithium intercalation is facilitated and the charge-transfer resistance of lithium intercalation becomes very small. Again, mass transport becomes limiting and layer-by-layer charging according to the diffusion overpotential is observed.

FIG. 4-41 DISTRIBUTION OF IRREVERSIBLE CHARGE LOSSES WITHIN THE FIRST THREE CYCLES OF A 6-LAYERED SFG44 MWE, LP40 ELECTROLYTE WITH 2 WT.-% VC, CHARGE RATE C/10

4.4.5 TOWARDS VERTICALLY DISTRIBUTED IMPEDANCE SPECTROSCOPY

It was shown in the previous chapter that the SEI formation of the first layer of the MWE is significantly different than in the subsequent layers. A larger consumption of vinylene carbonate suggests a different (or thicker) SEI. Impedance spectroscopy may be useful to further investigate the influence of this extra charge, since it was already successfully applied to investigate the resistance of lithium transport through the SEI layer[222,281,282]. Although a comprehensive study of vertically distributed impedance is beyond the scope of this work, some comments will be given on the use of a reference electrode.

4.4.5.1 POSITION AND CHOICE OF REFERENCE ELECTRODE

As it was explained in chapter 2.5, reasonable electrochemical impedance spectroscopy requires the use of a three-electrode cell to split the impedance of the working electrode from the one of the counter electrode. The choice of the reference electrode, however, is limited by the geometry of the MLC. At first, a second, solid lithium metal electrode was placed behind the last layer of the MWE and used as reference electrode. The galvanostatic chronopotentiogram of the master channel suggested a similar behaviour like for a two electrode system. The layer-dependent chronoamperograms, however, exhibited a remarkable difference compared to the ones obtained in

111 4 RESULTS AND DISCUSSION a two-electrode setup. Sequential intercalation could still be observed for layers 1 to 4. Layer 5, however, showed a contradicting behaviour and was charged at the same time as layer 1 (data not shown).

There are two reasons behind this behaviour. First, the separator of the reference electrode created an electrolyte reservoir which provided additional lithium ions to diffuse from the back into the layers. To understand the second reason, it may be helpful to investigate the influence of RE position on the potential drop due to a flowing current. When current flows through L1, a potential drop

ΔΦCE due to the ionic resistance in the pores and the separator Rion,S1 between WE and CE can be observed. If most of the current is flowing through the lower layers L2 and L3, this ohmic potential drop will increase due to the added ionic resistances Rion,S2 and Rion,S3. In the case that the reference electrode is located behind the MWE (see FIG. 4-42), and current flows through L1, the potential drop towards the reference electrode ΔΦback will probably be smaller than ΔΦCE. When current flows through L5, however, ΔΦback is much lower than ΔΦCE.

FIG. 4-42 TRANSMISSION LINE MODEL OF A THREE-LAYERED MWE INCL. A REFERENCE ELECTRODE IN THE “BACK” POSITION

4.4.5.2 DEVELOPMENT OF A POROUS LI0.5FEPO4 REFERENCE ELECTRODE

The need to use a reference electrode in front of the MWE comes along with the requirement that it needs not to influence the current line distribution between WE and CE. This can for example be achieved by a Luggin-capillary analogue for thin film electrodes as described in chapter 4.2. Another option is the use of a porous reference electrode placed between WE and CE, which will be presented here.

Lithium iron phosphate (LFP), a natural mineral of the olivine group was suggested as intercalation host for lithium by J. B. GOODENOUGH in 1996 and quickly gained popularity for being a low-cost, environmentally friendly alternative to lithium cobalt oxide[283]. It shows a perfectly flat potential plateau due to a clean two-phase intercalation reaction with a lithium rich LiFePO4 phase and a lithium-deficient FePO4 phase.

112 4 RESULTS AND DISCUSSION

The notation LixFePO4 is an abbreviation and actually describes the two phases x LiFePO4 + (1-x) FePO4.

F. LA MANTIA has recently shown its applicability as reference electrode for lithium ion batteries when it is charged to about 50 % SoC[284]. The charge of LFP to 50 % SoC, however, requires another step in the electrode preparation. As alternative, it was shown that LFP can also be delithiated chemically by

+ + [285] oxidizing agents such as NO2BF4 (redox potential of NO2 /NO2 is ca. 5.1 V vs. Li /Li) :

EQ. 4-2

Li0.5FePO4 was prepared by oxidation with NO2BF4 solution dropped into a stirred suspension of commercially available LFP in acetonitrile. The washed and dried powder was then directly used to prepare paste electrodes. As can be seen in

FIG. 4-43, the success of chemical delithiation was tested on one electrode by electrochemical charge/discharge. The open circuit potential of the freshly prepared Li0.5FePO4 was neither stable nor reproducible due to the contribution of FIG. 4-43 POTENTIAL PROFILE OF A LIFE0.4FEPO4 ELECTRODE; ST C-RATE CA. C/5; 1 REDUCTION WAS 41 % OF THE TOTAL surface groups to the OCP. Nevertheless, after a CAPACITY, PROVING THAT 59 % LITHIUM WAS REMOVED DURING CHEMICAL OXIDATION; 80-8-5-7 LFP-S5130-SFG6-C65; MASS full charge/discharge cycle, followed by reduction -2 LOAD CA. 1.5 MG CM to 50 % SoC, the OCP stabilized at 3.426 V vs.

+ Li /Li (see FIG. 4-44).

113 4 RESULTS AND DISCUSSION

FIG. 4-44 POTENTIAL PROFILE OF AN LFP ELECTRODE DURING FULL REDUCTION/OXIDATION AND HALF CHARGE REDUCTION AT C/5 TO 50 % SOC, FOLLOWED BY A 12 H REST PERIOD; THE OCP STABILIZES AT + AROUND 3.428 V VS. LI /LI.

Since the aim was to avoid electrochemical cycling, attempt was made to achieve a similar stability without potentiostatic control. As can be seen in FIG. 4-45, a similarly stable potential could indeed be achieved by introduction of a short conditioning step, in which the LFP electrode was oxidized and reduced with a small charge (ca. 5 % of total charge capacity). For illustration, the open circuit potentials of two Li0.5FePO4 electrodes were recorded versus a lithium reference electrode. The apparent simultaneous drop in OCP of both LFP electrodes (red and blue) was actually due to a non- stable potential of the lithium reference electrode. The voltage difference between the two LFP electrodes, however, approached zero volts, especially after the short conditioning time. Since there was no net consumption of charge, this conditioning step can be performed in-situ without potentiostatic control by a second (lithium) reference electrode.

FIG. 4-45 RECORD OF THE POTENTIAL DIFFERENCE (BLACK) OF TWO SYMMETRIC LI0.5FEPO4 ELECTRODES; THE RED AND BLUE CURVES SHOWS THE LFP POTENTIAL AGAINST A LITHIUM RE. OBVIOUSLY, THE OBSERVED POTENTIAL FADING OF BOTH ELECTRODES IN THIS PARTICULAR CASE WAS DUE TO ACCIDENTAL DETERIORATION OF THE LITHIUM REFERENCE ELECTRODE WHILE THE LFP POTENTIAL REMAINED STABLE.

114 5 CONCLUSIONS AND OUTLOOK

5 CONCLUSIONS AND OUTLOOK

The importance of surface oxygen groups of graphite during SEI formation was recently discussed in literature. For carbon nanotubes and carbon nanofibres, however, experimental results concerning the relationship between morphology, surface groups and irreversible charge losses (ICL) remained incomparable and conclusive knowledge about these crucial parameters is still missing. Therefore, a method for identification of the different mechanisms contributing to the total amount of ICL was developed and applied to various CNF electrodes with different amounts and types of surface oxygen groups.

Gas phase oxidised CNF proved to be more suitable for battery-related applications than liquid oxidised CNF due to an easier way of processing, smaller degree of mechanical degradation and, most important, a lower specific charge loss during first charging when used as negative electrode material. Usually, oxidation is poorly selective and results in a complex mixture of functional groups. However, due to the higher possible reaction temperature of 200 °C in the gas phase oxidation reactor, the formation of surface oxygen groups could be driven from single-bonded towards double- bonded oxygen groups at similar total oxygen contents and surface oxygen concentration[257]. The formation of DOUBLE-BONDED SURFACE OXYGEN GROUPS effectively reduced the amount of total ICL from 283 mAh g-1 to 173 mAh g-1 by supressing the charge losses spent in the potential region of 10 to 500 mV[255]. This is an important step towards understanding and utilizing CNF as active material, as conductive additive or as matrix for other active materials in negative electrodes for lithium ion batteries with high energy densities. Optimization of the electrode architecture and use of SEI forming additives are likely to further increase the coulombic efficiency.

CNF are also an interesting model material to understand the role of oxygen in other carbons such as graphite, because their remarkably high surface area allows for a much higher concentration of oxygen in a sample, thus providing a higher precision in the quantification of surface effects. A possible mechanism for the surface-dependent solvent co-intercalation was already proposed in section 4.1.5, p.74. To confirm this hypothesis, electrochemical impedance spectroscopy should be used for a further investigation of the effects of surface oxidation on the SEI. Especially a systematic investigation of the charge-transfer resistance during solvent co-intercalation would provide useful insights[222,220]. The basis for reliable investigations of the SEI is now provided by the successful development of the coaxial impedance cell[248].

While the influence of electrode geometry and position of the reference electrode has been well studied for solid electrolytes and fuel cells, little of this knowledge has been consistently applied for

115 5 CONCLUSIONS AND OUTLOOK studies in lithium ion batteries. Three-electrode Swagelok cells used by many researchers in the field of battery science, are especially characterised by imprecise electrode positioning and asymmetric reference geometry and consequently suffer from EIS distortions such as scaling of spectra (misalignment of electrodes), high or low frequency inductive loops (electrochemical asymmetry) or even artificial capacitive loops (combination of both)[228,248].

In this work, a new test cell was developed, which is superior to, but still comparable with standard Swagelok cells (chapter 4.2, p. 77)[248]. The new cell design features a coaxially oriented reference electrode and precisely aligned electrodes. These measures prevented scaling effects as well as inductive loops in the high frequency region up to about 50 kHz. Furthermore, by applying finite element method (FEM) simulations, it was possible to identify the range of parameters for which the electrochemical and geometric asymmetries in the cell do not substantially affect the experimental measurements (chapter 4.3, p. 85)[272]. Impedance spectra were simulated with physically extreme parameters to anticipate possible distortions in the “worst case”. Real parameters are expected to be less extreme, that means differences in electrochemical parameters and the degree of misalignment are expected to be less and the electrolyte is expected to be more conductive.

In summary, the FEM simulations underlined the reliability of the coaxial impedance cell for impedance measurements. At the same time, guidelines for cell assembly could be given. To steer clear from impedance distortions, misalignments and edge effects have to be avoided and good care of the electrode cutting and insulation has to be taken of. Very thin film electrolytes such as foil separators should be avoided where possible.

A multiple working electrode (MWE) for lithium ion batteries was developed by coating electrode slurries on thin electroformed meshes as current collector. Using this technique, a standard graphite electrode with a mass load of about 12 mg cm-2 could be virtually split into six independently addressable layers, with a mass load of about 2 mg cm-2 each. Considering the thickness of the current collector and the graphite particles, even splitting to thinner layers, such as 1 mg cm-2, would be possible. Electrochemical charge/discharge cycling of the MWE with the galvanostatic cycling under potential limitation protocol was possible in the corresponding gas-tight multi-layered cell (MLC) by using a combination of a master galvanostat with six current followers. The analysis of current distribution revealed that layers are charged and discharged stage-by-stage and layer-by- layer. At times, current densities within individual layers were found to be twice as large as the average current densities. Since the resolution of current density is limited by the thickness of individual layers, the local current densities in electrode planes thinner than one layer of the MWE may be even larger. Local current densities were observed to be highest for the first electrode layer

116 5 CONCLUSIONS AND OUTLOOK at the beginning of each stage. In cases where high current densities pose the risk of undesired side reactions such as metallic lithium plating or other aging mechanisms of the battery, it may be beneficial to avoid high currents, especially at the beginning of potential plateaus.

Graphite electrodes are characterized by irreversible charge losses such as the formation of the solid- electrolyte interphase. It turned out that the SEI formation of the first layer of the MWE was significantly different from the case for the subsequent layers. It could be shown that at potential above 1 V, a higher amount of charge was consumed in the first layer of the MWE. Mass transport effectively governs the current distribution even at very low currents. Since the reduction of vinylene carbonate as SEI forming additive as well as the solvent reduction itself takes place in this potential region, it is suggested that the SEI close to the electrode surface is significantly different and thicker than in the other layers. Surprisingly, no difference between the second and subsequent layers was observed. Apparently, vinylene carbonate diffusing from the CE is preferentially reduced within a thin surface layer of the negative electrode which is less than 30 µm thick. This observation is in good agreement with a theoretical work which predicted a non-uniform distribution of irreversible charge losses[293]. If it can be assumed that an even distribution of SEI is beneficial, e.g. for longer cycle life, it is suggested to use a pulsed current profile with long pauses (for diffusional redistribution) during the first cycle SEI formation to allow for the VC to be evenly deposited throughout the vertical axis.

A critical point about the introduction of additional separators in the MWE stack is that they significantly increase the diffusion length as well as the ionic resistance within the MWE. This is of special importance in cases where the ionic resistance and the diffusion length of the separator are larger than those of the pores of the electrode, since it bloats the diffusion overpotential in relation to the contributions of surface overpotential. Nevertheless, since no charge is consumed within the separators, the general conclusions concerning the current distribution remains valid and in good agreement with theoretical predictions[7] as well as other experimental works[139,140,136]. An additional experiment that can be conducted to complement the information obtained with the setup in this study would omit the separators and perform charge/discharge tests with several layers in direct (i.e. ionic and electric) contact. Redistribution of charge is a very slow process[139], meaning that the SoC of each layer will not be significantly changed after stopping the charging current. The MLC could therefore be easily disassembled and the layers of the separator-free MWE easily delaminated from each other ex-situ, which offers a “smorgasbord” of opportunities for further investigations. Characterization may include re-assembly for electrochemical determination of SoC, electrochemical impedance spectroscopy, SEM, TEM, spectroscopy or others. These techniques will provide further evidence for the uneven distribution of the SEI.

117 5 CONCLUSIONS AND OUTLOOK

This work delivered important information on the vertical distribution of current, charge and charge losses in graphite electrodes. The electrolyte resistance depending on the layer depth was investigated by electrochemical impedance spectroscopy in a two-electrode setup and a ready-to- use porous electrode for three-electrode EIS measurements was successfully developed. The multi- layered cell and the multiple working electrode provide the basis for vertically resolved dc and ac analysis in-situ as well as vertically resolved analysis ex-situ with any other desired technique.

118 6 EXPERIMENTAL SECTION

6 EXPERIMENTAL SECTION

6.1 CELL MANUFACTURE AND ASSEMBLY

6.1.1 THREE-ELECTRODE SWAGELOK T-CELL

Swagelok cells were machined in the fine mechanical workshop of the Ruhr-University Bochum under the guidance of A. LINDNER following a template kindly provided by the group of Prof. M. WINTER,

MEET, Münster (see. FIG. 2-9, p. 25). A standard Swagelok three-way tube connector was drilled out to a bore diameter of 13 mm and served as gastight housing. The electrode stack, consisting of 12 mm disc shaped electrodes (see chapter 6.3) with one or two 12 mm glass fibre separators in between was compressed by V4A stainless steel plungers with a diameter of 12.6 mm. A gas-tight enclosure was ensured by Teflon compression rings substituting the standard steel cutting rings in combination with swivel nuts. To minimise reactions with the electrolyte the steel parts in contact with electrolyte were equipped with a PEEK sleeve. A variable pressure was applied by a steel spring. To avoid electric short circuits, the electrode plungers were isolated from the tube walls by a 42x39 mm sheet of Mylar foil with a hole for the reference electrode. The RE was realised by a small piece of lithium pressed onto a concave PEEK/V4A holder placed in the T-opening of the cell. Detailed technical drawings of all parts can be found in the appendix. Prior to assembly, all cell parts were kept in an oven at 110 °C for at least 1 h and inserted into the glove box while hot. Assembly of the cell was performed in the order i) fixed plunger; ii) electrode stack (CE, 1 or 2 glass fibre separators and WE); iii) plungers with spring; iv) electrolyte filling through RE bore; v) RE and plunger.

119 6 EXPERIMENTAL SECTION

TAB. 6-1 COMPONENTS USED FOR SWAGELOK THREE-ELECTRODE T-CELLS Component Specification Supplier Swagelok three-way tube SS-12M0-3 B.E.S.T. Fluidsysteme GmbH Düsseldorf connector (12 mm) D-41468 Neuss, Germany Swagelok Teflon compression rings front T-12M3-1 B.E.S.T. Fluidsysteme GmbH Düsseldorf (12 mm) back T-12M4-1 D-41468 Neuss, Germany PEEK and V4A parts home-made fine mechanical workshop, Ruhr-University Bochum, Germany Steel spring RDF-1581 Schweizer Federntechnik D-72766 Reutlingen, Germany Mylar foil 0.10 mm, isolation class B, Dr. D. Mueller GmbH width 42 +/-0.2 mm, D-26197 Ahlhorn, Germany Glass fibre separator Whatman GF/D GE Healthcare Lithium ribbon # 265985 99.9% (trace metals) Sigma-Aldrich Chemie GmbH thickness W0.38 mm × 23 mm Munich, Germany Electrolyte Merck Selectilyte LP40 VWR International GmbH

(1 M LiPF6 in EC:DEC 1:1) D-64295 Darmstadt, Germany

6.1.2 COAXIAL IMPEDANCE CELL

Coaxial impedance cells were machined in a similar manner as explained in the previous chapter. As can be seen from FIG. 4-16, p. 81, the final design developed in this work has an inner part, consisting of cell base hosting the electrode assembly, a positioning sleeve and an electrode plunger. For exclusion of air, this inner part was placed inside a drilled out Swagelok tube connector equipped with two PEEK plungers and Teflon compression rings. Electric contact was made by V4A steel cylinders hot-pressed in the PEEK plungers in combination with gold coated contact springs. Technical drawings can be found in the appendix.

Both cell base and electrode plunger have a centre bore for the extrusion of a lithium reference electrode. A small piece of lithium was placed in the thread holes of the cell base and plunger and extruded through the 0.5 mm reference bore by a set screw. A clean disc shaped lithium surface was then created by cutting away excess lithium. For the counter electrode, lithium metal ribbon was cleaned prior use by scraping of the first layer with a scalpel. A 12 mm o. d., 2 mm i. d. annular lithium disc was cut out with a commercial double hole punch and placed inside the stepped cell base, flattened out, and covered with one or two layers of 14 mm diameter glass fibre separator. A 12 mm (i. d.) positioning sleeve inserted in the cell base served as guidance for the working electrode placed on top of the separator. Electrolyte could then easily be added through the WE centre hole

120 6 EXPERIMENTAL SECTION before closing the cell with the electrode plunger.

TAB. 6-2 COMPONENTS USED FOR SWAGELOK THREE-ELECTRODE T-CELLS Component Specification Supplier Swagelok tube connector SS-1210-6 B.E.S.T. Fluidsysteme GmbH Düsseldorf (¾ inch) D-41468 Neuss, Germany Swagelok Teflon compression rings front T-1213-1 B.E.S.T. Fluidsysteme GmbH Düsseldorf (¾ inch) back T-1214-1 D-41468 Neuss, Germany PEEK and V4A parts home-made fine mechanical workshop, Ruhr-University Bochum, Germany Steel spring RDF-1586 Schweizer Federntechnik D-72766 Reutlingen, Germany Contact springs for cell base 12H1381 Bürklin OHG TK0054B.05.1,30.C.200.A D-82041 Oberhaching, Germany Contact springs for plunger 11H5558 Bürklin OHG F111.09 D-82041 Oberhaching, Germany Glass fibre separator Whatman GF/D GE Healthcare Lithium ribbon # 265985 Sigma-Aldrich Chemie GmbH thickness W0.38 mm × Munich, Germany 23 mm, 99.9% trace metals basis Electrolyte Merck Selectilyte LP40 VWR International GmbH

(1 M LiPF6 in EC:DEC 1:1) D-64295 Darmstadt, Germany

121 6 EXPERIMENTAL SECTION

6.1.3 MULTI-LAYERED ELECTROCHEMICAL CELL

FIG. 6-1 CROSS-SECTIONAL VIEW OF THE MULTI-LAYERED ELECTROCHEMICAL CELL (FINAL DESIGN)

The MLC was made from PEEK for all insulating parts and V4A stainless steel for all conducting parts. For technical drawings, refer to the appendix. A detailed explanation of all parts is also given in chapter 4.4. An isometric cross-sectional view of the assembled cell can be found in FIG. 4-30, p. 100.

TAB. 6-3 COMPONENTS OF THE MULTI-LAYER CELL Component Specification Supplier Expanded copper grid 2Cu4-100-FA Dexmet Corp. 2Cu6-100-FA Wallingford, CT 06492, U.S.A. Separator polyolefin membrane Celgard ca. 23 µm thick Charlotte, NC 28273, U.S.A. Contact springs 5110/G-D-1.5 N-Au-2.3 C Bürklin OHG D-82041 Oberhaching, Germany Pressure spring RDF-1586 Schweizer Federntechnik D-72766 Reutlingen, Germany Glass fibre separator Whatman GF/D GE Healthcare Electrolyte Merck Selectilyte LP40 VWR International GmbH

(1 M LiPF6 in EC:DEC 1:1) D-64295 Darmstadt, Germany Vinylene Carbonate (VC) 97 %, stabilised by ≤2% BHT Sigma-Aldrich Chemie GmbH (electrolyte additive) [872-36-6] Munich, Germany

122 6 EXPERIMENTAL SECTION

Before introduction to the glove box, all parts were dried for at least 1 h in 110 °C and inserted while being hot. A PEEK cylinder with a stepped well (for the electrodes) and six radial grooves (for connection of contact tabs) served as cell base for the multiple working electrode stack. A short piece of expanded copper mesh was placed in one of the six grooves on the outer perimeter so about 2- 3 mm of the copper mesh was lying inside the inner area. One Ø18 mm graphite electrode layer supported by electrolytic mesh (see chapters 4.4.2 and 6.3.3) was then placed in the centre of the cell and soaked with about 40 µL of electrolyte. A dry Celgard polyolefin separator was placed on top of the electrode, followed by the next set of copper grid lead, electrode layer and separator. After six electrode layers were stacked like this, the whole electrode stack was evenly fixed by a PEEK sleeve and the copper grid leads contact by gently tightening the six setscrews located in the fixation ring. A slightly oversized lithium disc (Ø 20 mm) served as counter / reference electrode and was placed inside the fixation ring, separated by a second separator made from glass fibres to avoid eventual lithium dendrite formation. An extra volume of about 200 µL LP40 electrolyte (in the experiments using SFG44 graphite including 2 wt.-% VC) was added to the glass fibre separator. After finishing the inner cell with the stainless steel plunger, it was placed inside the outer housing cup and closed by the upper body including the spring contacts for all WE. The pressure on the electrode stack can be easily altered by using springs with different spring constants.

typical electrolyte uptake porous electrode 30-50 µL, depending on thickness Celgard Separator sufficiently wetted by electrode glass fibre separator 100-150 µL

After cell assembly, the MLC was left overnight with each WE layer short-circuited for equilibration and soaking of separators. For quality control of the cell assembly and measurement of the electrolyte resistance, an impedance spectrum was recorded at OCP (usually around 3.2-3.4 V vs. Li/Li+ CE/RE) for each layer separately.

123 6 EXPERIMENTAL SECTION

6.2 CARBON NANOFIBRES

6.2.1 OXIDATION OF CNF

Multi-walled hollow carbon nanofibres were purchased from Applied Sciences and oxidised by liquid or gaseous nitric acid, yielding oxidised CNF denoted as l-CNF and g-CNF, respectively. L-CNF were oxidised by dispersing 1 g of CNF in 200 mL 65 % HNO3 in a round bottom flask with magnetic stir bar and heating under reflux (110 °C) for 1.5; 3; 6, 12 and 24 h. The CNF were cooled to about 50 °C, filtered through fine filter paper with the aid of a suction flask, and repeatedly washed with ultrapure water until the effluent was clear and had neutral pH. The CNF were subsequently dried for several hours at 90 °C in a convection oven and scraped off the filter. Weight losses were determined by comparing the weight of CNF after filtration and drying with the initially weighed portion.

FIG. 6-2 SCHEME OF THE REACTOR USED FOR GAS PHASE OXIDATION OF CNF

For the preparation of g-CNF, CNF were placed in a reactor recently developed for the gas phase oxidation of CNF[256]. Nitric acid vapours were generated from boiling conc. nitric acid and passed over a crucible with 200 mg of CNF sample in a reaction chamber heated with a resistance heating coil to 200 °C for 24 or 72 h. After oxidation, the flow of gas phase nitric acid was stopped and the temperature in the reaction chamber held at 110 °C for 2 h to dry the g-CNF sample.

124 6 EXPERIMENTAL SECTION

TAB. 6-4 COMPONENTS USED FOR CNF OXIDATION Component Specification Supplier Multi-walled carbon nanofibres Pyrograf III PR-19 PS Applied Sciences i. d. 20-50 nm Ohio, U.S.A. o. d. 70-200 nm 30-40 sidewalls; ~90 % purity conc. nitric acid [7697-37-2] 65 % Mallinckrodt Baker B.V. 7400 AA Deventer filter paper MN617 Macherey-Nagel, Germany Ultra-pure water Ultra-clear UV Siemens Water Technologies 0.055 µS/cm, TOC < 1 ppb D-22885 Barsbüttel, Germany

6.2.2 INSTRUMENTAL ANALYSIS

SCANNING ELECTRON MICROSCOPE images were taken with a FEI Quanta 3D FEG which was operated at 20 kV with a 10 mm working distance in the secondary electron imaging mode. For this purpose, a small amount of the oxidised CNF was dispersed in Ethanol by ultrasound agitation, diluted and drop- coated on a piece of silicon wafer. Differences in average CNF (fragment) length were qualitatively assessed on different positions of the SEM sample.

H, C, N ELEMENTAL ANALYSIS was performed by standard combustion analysis performed by in-house services. Oxygen content was calculated as residual percentage because the content of other heteroatoms was negligible.

RAMAN SPECTROSCOPY was performed together with S. GRÜTZKE using a Horiba Jobin Yvon iHR 550 spectrometer, equipped with a green Helium-Neon Laser (532 nm), operated at 90 % maximum amplitude (~80 mW). The laser beam was focused on slightly compacted CNF powder samples. Spectra were then recorded in the range of 1250 to 1700 cm-1 with an integration time of 500 s and 20 spectral frame accumulations.

X-RAY PHOTOELECTRON SPECTROSCOPY measurements were carried out by DR. W. XIA in an ultrahigh vacuum (2x10-10 mbar) setup with a Gammadata-Scienta SES 2002 analyser. Incident radiation came from an Al Kα source (1486.6 eV; 14 kV; 55 mA). A quantitative study of the oxygen species was performed by peak analysis in the O 1s region based on the deconvolution of the two overlapping O 1s peaks, using the CASA XPS program with a Gaussian-Lorentzian mix function and Shirley background subtraction as has been reported before[257].

125 6 EXPERIMENTAL SECTION

6.3 ELECTRODE PREPARATION Most electrodes used in this study were prepared using the “slurry technique” in which all electrode components (active material, conductive additive and binder) are dispersed in a suitable solvent (such as N-methyl pyrrolidone) and applied on a current collector foil[294]. Binder stock solutions were prepared by stirring PVdF binder overnight in NMP, with typical concentrations of 20-30 mg/mL. The conductive additive was then dispersed in the binder solution by means of a rotor-stator disperser. The active material was then added to the suspension, topped up with an appropriate amount of extra solvent and carefully dispersed with the other components. Quality was controlled by eye and usually a sufficiently good dispersion with honey-like viscosity was achieved after at least 30 min of mixing.

An electrode paste needs to be as liquid as necessary to flow onto the current collector during the coating process, but as thick as possible to avoid settling and segregation of particles during the drying process.

For electrode coating, a suitable current collector foil (copper or aluminium) was placed on a glass plate and wiped with an ethanol soaked fuzz-free paper cloth to clean it as well as smoothly attach it to the glass surface. The ethanol also serves as an adherent to keep the foil wrinkle free on the glass plate.

TAB. 6-5 GENERAL EQUIPMENT USED FOR ELECTRODE PREPARATION Component Specification Supplier Rotor-stator disperser Ultra-Turrax T-10 basic IKA®-Werke GmbH & CO. KG Tool S 10 N - 5 G D-79219 Staufen, Germany Ultrasound bath Sonorex RK 100 Bandelin electronic GmbH & Co. KG D-12207 Berlin, Germany Ultrasound agitation probe Sonopuls HD 3100 Bandelin electronic GmbH & Co. KG MS 73 microtip 3 mm D-12207 Berlin, Germany Blade coater 0-380 µm slit #UO-2325; w = 51 mm, Olbrich know how by micro meter calliper (10:1 leverage) D-58675 Hemer, Germany Convection oven n/a Binder GmbH D-78532 Tuttlingen, Germany Hole punch # 832100 Hoffmann Tools Munich, Germany Vacuum oven B-585 Drying Büchi Labortechnik GmbH D-45127 Essen, Germany

126 6 EXPERIMENTAL SECTION

Vacuum pump for oven Laboport N86KT18 KNF Neuberger GmbH 79112 Freiburg, Germany Ball calliper Holex 421505 Hoffmann Tools (for thickness measurements) 0-25 mm; +/- 0.001 mm Munich, Germany Glovebox Jacomex, Argon filled GS Glovebox Systemtechnik GmbH

O2, H20 < 2 ppm D-76316 Malsch, Germany Multi-channel potentiostat VMP-3 EIS Bio-Logic, Claix, France

6.3.1 CARBON NANOFIBRE ELECTRODES

For the preparation of BINDER-FREE CNF ELECTRODES, g-CNF or l-CNF powder was dispersed in ethanol (5 mg/mL) by ultrasonic agitation (ultrasound bath) for 1 h and drop coated on 12 mm discs of dendritic copper foil. The entangled CNF sufficiently attached to the rough copper surface so no additional binder or conducting additive was needed. The final drop-cast weight of the electrodes was 0.5 and 1 mg/cm². The electrodes were dried several hours at 120 °C in a vacuum oven.

PVDF-BOUND CNF electrodes were prepared by stepwise adding 50-100 mg l-CNT(1.5 h), l-CNT(6 h) or g-CNT(24 h) in portions of about 10 mg to a 25 mg/mL S5130 PVdF binder solution, followed by rotor- stator dispersion at ~5000 rpm (speed ‘3’) until a sufficiently homogeneous dispersion was achieved before the next portion was added. The mass fraction in the slurry was about 100 mg per 1 mL solvent and the dry composition of CNT:PVdF was 95:5. Electrodes were then coated as aforementioned with a slit width of 200 µm.

TAB. 6-6 COMPONENTS USED FOR CNF ELECTRODES Component Specification Supplier L-CNT Pyrograf III PR-19 PS Applied Sciences, Ohio, U.S.A.

oxidised in 65 % HNO3 treatment in-house G-CNT Pyrograf III PR-19 PS Applied Sciences, Ohio, U.S.A.

oxidised in gas phase HNO3 treatment in-house Ethanol abs.; 99.8 % (GC) Sigma-Aldrich Chemie GmbH

Munich, Germany N-Methyl pyrrolidone ReagentPlus®, 99% Sigma-Aldrich Chemie GmbH Munich, Germany Polyvinyl di-fluoride binder (PVdF) Solef S5130 Solvay Plastics Dendritic copper foil SE-Cu58, t=10 µm Schlenk Metallfolien GmbH & Co. KG treated on both sides 91154 Roth

127 6 EXPERIMENTAL SECTION

6.3.2 LIFEPO4 ELECTRODES

LiFePO4 slurries with a final composition of LFP:C65:S5130 of 85:10:5 wt.-% were prepared by pouring 2 mL of 25 mg mL-1 PVdF stock solution onto 100 mg C65 carbon black and rotor-stator mixing for 10 min at speed “3” (ca. 5000 rpm). 850 mg LFP were subsequently added and topped up with 500 µL of NMP for viscosity adjustment. The slurry was dispersed for another 20 min at 5000 rpm. Final mass loading of the slurry was ca. 0.4 mg mL-1.

A 15 μm aluminium foil was used as current collector. To remove surface oxides and increase contact, its surface was etched by wetting the foil for 30 s with conc. potassium hydroxide, followed by intensively rinsing with ultrapure water. The slurry was then spread onto the foil using a blade coater with a slit width of 220 μm, resulting in a dry mass loading of about 5 mg cm−2. The electrode foil was pre-dried at 80 °C without convection for 2-3 hours. 12 mm disk or annular disk electrodes with a 2 mm hole were punched out and dried over night at 120 °C in a vacuum oven.

TAB. 6-7 COMPONENTS USED FOR LFP ELECTRODES Component Specification Supplier Aluminium foil battery grade, t=15 µm MTI Corp, Richmond, U.S.A. Potassium hydroxide techn. grade Mallinckrodt Baker B.V. 7400 AA Deventer Polyvinyl di-fluoride binder (PVdF) Solef S5130 Solvay Plastics Lithium iron phosphate battery grade MTI Corp, Richmond, U.S.A. Conducting additive Super P C65 Timcal CH-6743 Bodio, Switzerland N-Methyl pyrrolidone ReagentPlus®, 99% Sigma-Aldrich Chemie GmbH

Munich, Germany

128 6 EXPERIMENTAL SECTION

6.3.3 GRAPHITE MESH ELECTRODES

For graphite mesh electrodes, slurries with a dry weight composition of 90:5:5 (graphite:C65:S5130) and an overall solid mass fraction of about 0.357 mg mL-1 was used. For this purpose, 35 mg of Super P C65 conductive carbon was weighed in a 3 mL Eppendorf vial and dispersed in 1.4 mL of 25 mg mL-1 PVdF binder solution in NMP using the Ultra-Turrax disperser for 20 min at pace 3 (ca. 5 000 rpm). 630 mg graphite and 560 µL NMP for viscosity adjustment was subsequently added and dispersed two times for 10 min at pace 1. In the case of SFG6 graphite, the procedure was similar; the mass fraction, however, was reduced to about 0.26 mg mL-1.

Thin electrolytic copper mesh was straightened out on a glass plate covered by a Teflon adhesive tape. The paste was spread on the mesh and tape casted with a blade coater at a slit width of 90 µm. After drying for 2 h at 60 °C, the coated mesh could be carefully pulled off the Teflon tape. 18 mm electrode discs were punched by a commercial hole puncher.

To straighten out the copper mesh, it is helpful to dab the mesh on the Teflon tape with an ethanol wetted fuzz free cloth and gently skim the mesh in one direction. No part of the blade coater should ever touch the mesh.

TAB. 6-8 COMPONENTS USED IN MANUFACTURE OF MESH ELECTRODES Component Specification Supplier Conducting additive Super P C65 Timcal CH-6743 Bodio, Switzerland N-Methyl pyrrolidone ReagentPlus®, 99% Sigma-Aldrich Chemie GmbH

Munich, Germany Binder Solef 5130 Solvay Solexis Electrolytic copper meshes MC33, 333 lines / inch Precision Eforming 16,5 µm 70 % open Cortland, NY 13045, U.S.A. MC49, 117.6 lines / inch t=12,7 µm, 88,6 % open Graphites Timrex SFG6 Timcal Timrex SFG44 CH-6743 Bodio, Switzerland

129 6 EXPERIMENTAL SECTION

6.3.4 LI0.5FEPO4 REFERENCE ELECTRODES

1,579 g (10 mmol) carbon coated LFP (MTI, Richmond, USA) was dried in vacuum and dispersed in

10 mL dry acetonitrile in a Schlenk tube by magnetic stirring. 10 mL of a 0.5 M solution of NO2BF4 in dry acetonitrile was drop-wise added to the LFP suspension through a septum, which resulted in immediate gas evolution. The suspension was stirred for one hour, filtered through a polypropylene filter disc and washed 3 times with acetonitrile. The powder was subsequently dried in vacuum at 110 °C.

Electrode slurries were prepared by dispersing chemically oxidised LxFP with C65 carbon black and SFG6 graphite (Timcal, Bodio, Switzerland) as conductive additive in a polyvinylidene difluoride (PVdF) binder solution (Solef S5130, Solvay) in N-methyl pyrrolidone (NMP) (Sigma-Aldrich). After dispersing for 30 min at 5000 rpm by means of an Ultra-Turrax, the resulting slurry was spread onto a 15 µm aluminium foil or on a thin nickel mesh (as described in chapter 6.3.3) using a blade coater with a slid width of 100 µm. The resulting electrode had a dry composition of 80-7-5-8 wt.-% of LFP- C65-SFG6-PVdF. 12 mm annular disc electrodes with a 2 mm hole or 18 mm disc electrodes were punched out using a hole punch and dried overnight at 120 °C under vacuum.

TAB. 6-9 COMPONENTS USED FOR SYNTHESIS OF LI0.5PO4 Component Specification Supplier Acetonitrile dried in-house Sigma-Aldrich Chemie GmbH Munich, Germany Aluminium foil battery grade, t=15 µm MTI Corp, Richmond, U.S.A. Binder Solef 5130 Solvay Solexis Conducting additive Super P C65 Timcal CH-6743 Bodio, Switzerland Electrolytic nickel mesh MN33, 333 lines / inch Precision Eforming 16,5 µm 70 % open Cortland, NY 13045, U.S.A. Lithium iron phosphate battery grade MTI Corp, Richmond, U.S.A. N-Methyl pyrrolidone ReagentPlus®, 99% Sigma-Aldrich Chemie GmbH

Munich, Germany Nitronium tetrafluoroborate >95 % Sigma-Aldrich Chemie GmbH Munich, Germany Polypropylene filter disc Supelco Analytical Sigma-Aldrich Chemie GmbH # 57181 Munich, Germany

130 6 EXPERIMENTAL SECTION

6.4 ELECTROCHEMISTRY

6.4.1 ELECTROCHEMICAL CHARGE/DISCHARGE OF CNF ELECTRODES

Cycling of CNF electrodes was performed in a standard Swagelok three-electrode T-cell assembled in an Argon-filled glove box (O2 and H2O content < 2 ppm). Merck Selectilyte LP40 (1 M LiPF6 in EC:DEC 1:1) was used as received as electrolyte. The cells were left overnight for soaking of electrolyte and equilibration. Charge/discharge experiments of CNF electrodes were conducted with a Bio-Logic VMP-3 Potentiostat/Galvanostat using the EC-lab software. The charge rate was set to 50 mA/g (approx. C/10) in a potential window of 0.01 to 3 V (vs. Li+/Li) using the GCPL protocol. For capacity derivative plots etc., the raw data was smoothed by adjacent averaging (typically 7-9 points) and interpolated to 1000 evenly spaced potential values from 10 mV to 3 V. Interpolation was important to allow for determination of 1st to 2nd reduction capacity losses depending on the reaction potential. All data treatment was done in OriginPro.

6.4.2 EIS ON LIFEPO4 ELECTRODES

Three-electrode electrochemical test cells with LFP paste electrodes as WE and lithium foil or LFP paste electrodes as CE were assembled using either the classical Swagelok cell with RE in T-position (cell 1, chapter 6.1.1), or inside the developed coaxial cells (cell 2 and cell 3,chapter 4.2). For the coaxial cells, the reference electrode was located either inside the working electrode (denoted as R- LFP-CE) or inside the counter electrode (denoted as LFP-CE-R). Ready to use Selectilyte LP40 electrolyte (1 M LiPF6 in EC:DEC 1:1) from Merck was used as-received as electrolyte. Each cell was soaked and equilibrated overnight before electrochemical experiments were performed.

Electrochemical impedance spectra were recorded using the frequency response analyser (FRA) of the VMP-3 multipotentiostat. EIS was performed at 0 % SoC (i.e. after assembly) by applying a sine voltage perturbation (PEIS) of 10 mV around the open circuit potential (OCP). The frequency range was 100 kHz to 100 mHz, with 10 points per decade in logarithmic spacing and an averaging of 4 measures per frequency. For PEIS measurements at 50 % SoC, the LFP electrodes were charged (i.e. oxidised) by applying a constant current (GCPL) of about 20 mA g-1 (C-rate C/6) for 3 h, based on the practical capacity estimated from 2 full charge/discharge cycles.

For evaluation of charge-transfer resistances, the EI spectra were fitted with the Biologic Zfit tool, assuming an R1+(R2/Q1) equivalent circuit (i.e. a resistor in series with a pair of resistors and a constant phase element, see chapter 2.5.2). Single electrode impedances were obtained from three- electrode setups, For full-cell measurements, the RE lead of the potentiostat was hooked up to the

131 6 EXPERIMENTAL SECTION counter electrode (two-electrode measurement). Two-electrode full-cells were denoted as LFPx+LFPy. All spectra were subsequently recorded by rewiring. Where applicable, the calculated full- cell response was obtained by adding both single electrode impedances (from the three‐electrode cell), while the calculated single electrode response was obtained by dividing the measured full-cell impedance (from the two‐electrode cell) by two.

6.4.3 CHARGE/DISCHARGE OF MULTIPLE WORKING ELECTRODES

Before electrochemical charge/discharge cycling, the assembled MWE was left at least 12 h for equilibration and sufficient wetting of separators. A constant negative current was applied by a master galvanostat using the GCPL protocol. Assuming a practical capacity of 350 mAh g-1, the MWE was cycled with C-rates of C/35 (10 mA g-1) or C/10, C/5, C/2, 1C, 2C and 3C. When the MWE potential reached 0.02 V vs. Li+/Li, the current was reversed until a cut-off potential of 1 V was reached. In some sets, a potentiostatic regime was included in the GCPL protocol, holding the potential at its cut-off value until the current dropped below one quarter of its original value. The ionic current through each MWE layer was monitored using the ZRA protocol of EC-lab (see FIG. 4-28, p. 98). For this purpose, each layer was connected to the CE/REF cable of one out of six grouped “current follower” (“slave”) channels (CE to ground). The WE cable of each current follower was connected to the CE/REF of the MLC. The sum of all ZRA currents recorded was equal to the total current applied via the master channel. Data points were taken every 10 s. The wiring can be summarised as follows:

1. in EC-lab: group all slaves (CE to GND) 2. isolate and discard all slave WE/CA2(pink) 3. connect all slave CE/CA1&REF3(blue) to each cell layer electrode 4. connect all slave GND(black) & Ref1(red) & Ref2(white) with master WE/CA2(pink) & Ref1(red) 5. connect master CA/CA1(blue) & Ref3(blue) to cell CE 6. connect master Ref2(white) to cell RE 7. start grouped ZRA recording 8. start master galvanostat

132 6 EXPERIMENTAL SECTION

6.5 SOFTWARE

TAB. 6-10 SOFTWARE PACKAGES USED IN THE COURSE OF THIS WORK Program Version Publisher Chem3D pro V 12.00 Perkin-Elmer Waltham, USA ACD/ChemSketch V 10.00 Advanced Chemistry Development (ACD/Labs), Frankfurt, Freeware Germany Citavi V 3.4.0 Swiss Academic Software GmbH, Wädenswil, Schweiz Comsol V 4.2a COMSOL Multiphysics, Göttingen, Germany EC-lab V 10.23 Bio-Logic Science Instruments, Claix, France OriginPro V 9.0G OriginLab Corporation, Northampton, MA Microsoft Office V14.0 Microsoft Corporation Professional TinyCAD Freeware V 2.80.03 Matt Pyne

133 7 BIBLIOGRAPHY

7 BIBLIOGRAPHY

7.1 SOME NOTEWORTHY PUBLICATIONS AND RESOURCES

ABOUT ELECTROCHEMISTRY & LITHIUM ION BATTERIES

What Are Batteries, Fuel Cells, and Supercapacitors?[17] a compact introduction Lithium-Ion Batteries – Science and Technologies[295] a very thorough introduction to major aspects of LIB www.batteryuniversity.com[45] basic introduction to lithium ion batteries The thermodynamic origin of hysteresis in insertion batteries [296]

HISTORY OF ELECTROCHEMISTRY

Über das Verhalten sogenannter unpolarisierbarer Elektroden gegen den Wechselstrom[121] the first application of impedance spectroscopy Studies in electrode polarisation. Part IV. The automatic control of the potential of a working electrode[297] the invention of the potentiostat The Battery – how portable power sparked a technological revolution [13] the development of electricity from a little bit different perspective The Thomas Edison papers[298] a huge editorial effort to chronicle the life of Thomas Alva Edison

SOMETHING TO THINK ABOUT

The seven sins in Academic behavior in the natural sciences[299] by Wilfred F. van Gunsteren living healthy on a dying planet[300] by Daniel G. Nocera

134 7 BIBLIOGRAPHY

7.2 LITERATURE [1] J. R. R. Tolkien, The lord of the rings, Grafton, London, 1991. [2] M. Nič, J. Jirát, B. Košata, A. Jenkins, A. McNaught (Eds.) IUPAC Compendium of Chemical Terminology, IUPAC, Research Triagle Park, NC, 2009. [3] "anode" in IUPAC Compendium of Chemical Terminology (Eds.: M. Nič, J. Jirát, B. Košata, A. Jenkins, A. McNaught), IUPAC, Research Triagle Park, NC, 2009. [4] Timcal Ltd.; "Carbon Powders for Lithium Ion Battery Systems", can be found under www.timcal.com, 2012; 2012-07-31. [5] "cathode" in IUPAC Compendium of Chemical Terminology (Eds.: M. Nič, J. Jirát, B. Košata, A. Jenkins, A. McNaught), IUPAC, Research Triagle Park, NC, 2009. [6] BASF; "Selectilyte™ LP - Ready-to-use Electrolyte Solutions. BF-9822 USL 05/12", can be found under http://www.catalysts.pro/downloads/public/pdfs/Battery%20Materials/52012BF- 9822_US_SelectilyteLP_Datasheet-051512.pdf; 2012-12-02. [7] J. S. Newman, K. E. Thomas-Alyea, Electrochemical systems, J. Wiley, Hoboken, N.J, 2004. [8] US National Institute of Standards and Technology; "Fundamental Physical Constants: Faraday Constant. 2010 CODATA recommended values", can be found under http://physics.nist.gov/cgi-bin/cuu/Value?f; 2012-11-25. [9] US National Institute of Standards and Technology; "Fundamental Physical Constant: Molar Gas Constant. 2010 CODATA recommended values", can be found under http://physics.nist.gov/cgi-bin/cuu/Value?r; 2012-11-25. [10] L. F. Baum, The Master Key. An electrical fairy tale founded upon the mysteries of electricity and the optimism of its devotees … Illustrations, etc, B. F. Stevens & Brown, London, 1901. [11] Franz Haag; "1888 Flocken Elektrowagen (reconstruction)", can be found under http://commons.wikimedia.org/wiki/File:1888_Flocken_Elektrowagen_sw.jpg; 2013-04-11. [12] K. Möser, Geschichte des Autos, Campus-Verl., Frankfurt/Main, 2002. [13] H. R. Schlesinger, The battery. How portable power sparked a technological revolution, Smithsonian Books, Washington, DC, 2010. [14] Max-Planck-Gesellschaft; "A short portrait of the Max Planck Society", can be found under http://www.mpg.de/183251/portrait; 2013-04-11. [15] A. J. Bard, L. R. Faulkner, Electrochemical methods. Fundamentals and applications, Wiley, New York, 2001. [16] M. S. Whittingham, R. F. Savinell, T. Zawodzinski; "Introduction: Batteries and fuel cells", Chemical Reviews 2004, 104, 4243–4244. [17] M. Winter, R. J. Brodd; "What Are Batteries, Fuel Cells, and Supercapacitors?", Chemical Reviews 2004, 104, 4245–4270. [18] B. Scrosati; "History of lithium batteries", Journal of Solid State Electrochemistry 2011, 15, 1623–1630. [19] J. J. Auborn, K. W. French, S. I. Lieberman, V. K. Shah, A. Heller; "Lithium Anode Cells Operating at Room Temperature in Inorganic Electrolytic Solutions", Journal of the Electrochemical Society 1973, 120, 1613. [20] E. Peled; "The Electrochemical Behavior of Alkali and Alkaline Earth Metals in Nonaqueous Battery Systems---The Solid Electrolyte Interphase Model", Journal of the Electrochemical Society 1979, 126, 2047–2051.

135 7 BIBLIOGRAPHY

[21] A. Heller, Searching for Nature’s Truths and Life-Improving Products, Bochum, 2013. [22] M. S. Whittingham; "Chemistry of intercalation compounds: Metal guests in chalcogenide hosts", Progress in Solid State Chemistry 1978, 12, 41–99. [23] K. Mizushima, P. C. Jones, P. J. Wiseman, J. B. Goodenough; "LixCoO2 (o less-than x less-than- or-equal-to 1) - A new cathode material for batteries of high-energy density", Materials Research Bulletin 1980, 15, 783–789. [24] H. Ikeda, K. Narukawa, H. Nakashima, Jap. Patent #1769661, 1981. [25] B. Lestriez; "Functions of polymers in composite electrodes of lithium ion batteries", Comptes Rendus Chimie 2010, 13, 1341–1350. [26] M. Yoshio, R. J. Brodd, A. Kozawa (Eds.) Lithium-ion batteries. Science and technologies, Springer, New York, NY, 2009. [27] A. Nagai; "Applications of PVdF-Related Materials for Lithium-Ion Batteries" in Lithium-ion batteries. Science and technologies (Eds.: M. Yoshio, R. J. Brodd, A. Kozawa), Springer, New York, NY, 2009. [28] H. Yamamoto, H. Mori; "SBR Binder (for Negative Electrode) and ACM Binder (for Positive Electrode)" in Lithium-ion batteries. Science and technologies (Eds.: M. Yoshio, R. J. Brodd, A. Kozawa), Springer, New York, NY, 2009. [29] K. Tagawa, R. J. Brodd; "Production Processes for Fabrication of Lithium-Ion Batteries" in Lithium-ion batteries. Science and technologies (Eds.: R. J. Brodd, A. Kozawa, M. Yoshio), Springer, New York, NY, 2010. [30] J.-H. Lee, U. Paik, V. A. Hackley, Y.-M. Choi; "Effect of Carboxymethyl Cellulose on Aqueous Processing of Natural Graphite Negative Electrodes and their Electrochemical Performance for Lithium Batteries", Journal of the Electrochemical Society 2005, 152, A1763. [31] J. H. Lee, H. H. Kim, S. B. Wee, U. Paik; "Effect of Additives on the Dispersion Properties of Aqueous Based C/LiFePO4 paste and its Impact on Lithium Ion Battery High Power Properties", Kona Powder and Particle Journal 2009, 239–245. [32] W. Porcher, B. Lestriez, S. Jouanneau, D. Guyomard; "Optimizing the surfactant for the aqueous processing of LiFePO4 composite electrodes", Journal of Power Sources 2010, 195, 2835–2843. [33] L. El Ouatani, R. Dedryvère, J.-B. Ledeuil, C. Siret, P. Biensan, J. Desbrières, D. Gonbeau; "Surface film formation on a carbonaceous electrode: Influence of the binder chemistry", Journal of Power Sources 2009, 189, 72–80. [34] C. Daniel; "Materials and processing for lithium-ion batteries", Journal of the Minerals Metals and Materials Society 2008, 60, 43–48. [35] J. Li, C. Daniel, D. Wood; "Materials processing for lithium-ion batteries", Journal of Power Sources 2011, 196, 2452–2460. [36] S. F. Lux, F. Schappacher, A. Balducci, S. Passerini, M. Winter; "Low Cost, Environmentally Benign Binders for Lithium-Ion Batteries", Journal of the Electrochemical Society 2010, 157, A320. [37] A. Hintennach, P. Novák; "Influence of surfactants and viscosity in the preparation process of battery electrodes containing nanoparticles", Physical Chemistry Chemical Physics 2009, 11, 9484. [38] T. Patey, A. Hintennach, F. La Mantia, P. Novák; "Electrode engineering of nanoparticles for lithium-ion batteries—Role of dispersion technique", Journal of Power Sources 2009, 189, 590– 593.

136 7 BIBLIOGRAPHY

[39] K. M. Kim, W. S. Jeon, I. J. Chung, S. H. Chang; "Effect of mixing sequences on the electrode characteristics of lithium-ion rechargeable batteries", Journal of Power Sources 1999, 83, 108– 113. [40] M. Yoo, C. W. Frank, S. Mori, S. Yamaguchi; "Interaction of Poly(vinylidene fluoride) with Graphite Particles. 2. Effect of Solvent Evaporation Kinetics and Chemical Properties of PVDF on the Surface Morphology of a Composite Film and Its Relation to Electrochemical Performance", Chemistry of Materials 2004, 16, 1945–1953. [41] J. R. Dahn, T. Zheng, Y. H. Liu, J. S. Xue; "Mechanisms for lithium insertion in carbonaceous materials", Science 1995, 270, 590–593. [42] M. S. Whittingham; "Lithium Batteries and Cathode Materials", Chemical Reviews 2004, 104, 4271–4302. [43] C. S. Wang, A. J. Appleby, F. E. Little; "Electrochemical study on nano-Sn, Li4.4Sn and AlSi0.1 powders used as secondary lithium battery anodes", Journal of Power Sources 2001, 93, 174– 185. [44] J. Cabana, L. Monconduit, D. Larcher, M. R. Palacín; "Beyond Intercalation-Based Li-Ion Batteries: The State of the Art and Challenges of Electrode Materials Reacting Through Conversion Reactions", Avanced Materials 2010, 22, E170. [45] I. Buchmann; "Battery University. Types of Lithium-ion", can be found under http://batteryuniversity.com/learn/article/types_of_lithium_ion; 2013-04-12. [46] A. K. Padhi; "Effect of Structure on the Fe[sup 3+]∕Fe[sup 2+] Redox Couple in Iron Phosphates", Journal of the Electrochemical Society 1997, 144, 1609. [47] M. M. Thackeray, W. David, P. Bruce, J. Goodenough; "Lithium insertion into manganese spinels", Materials Research Bulletin 1983, 18, 461–472. [48] M. M. Thackeray, P. J. Johnson, La Depicciotto, P. G. Bruce, J. B. Goodenough; "Electrochemical Extraction of Lithium from LiMn2O4", Materials Research Bulletin 1984, 19, 179–187. [49] T. Xu, W. Wang, M. L. Gordin, D. Wang, D. Choi; "Lithium-ion Batteries for Stationary Energy Storage", Journal of the Minerals Metals and Materials Society 2010, 62, 24–30. [50] D. Doughty, P. Butler, A. Akhil, N. Clark, J. Boyes; "Batteries for large-scale stationary electrical energy storage", Electrochemical Society Interface 2010, 19, 49–53. [51] Q. Zhang, Z. Yu, P. Du, C. Su; "Carbon Nanomaterials Used as Conductive Additives in Lithium Ion Batteries", Recent Patents on Nanotechnology 2010, 4, 100–110. [52] M. Park, X. Zhang, M. Chung, G. B. Less, A. M. Sastry; "A review of conduction phenomena in Li-ion batteries", Journal of Power Sources 2010, 195, 7904–7929. [53] M. E. Spahr; "Carbon-Conductive Additives for Lithium-Ion Batteries" in Lithium-ion batteries. Science and technologies (Eds.: M. Yoshio, R. J. Brodd, A. Kozawa), Springer, New York, NY, 2009. [54] S. H. Munson-McGee; "Estimation of the critical concentration in an anisotropic percolation network" 1991, 43, 3331–3336. [55] K.-J. Euler; "The conductivity of compressed powders. A review", Journal of Power Sources 1978, 3, 117–136. [56] A. Espinola, P. M. Miguel, M. R. Salles, A. R. Pinto; "Electrical properties of carbons—resistance of powder materials", Carbon 1986, 24, 337–341. [57] K.-J. Euler, R. Kirchhof, H. Metzendorf; "Electronic conductivity of battery active masses — a limiting factor?", Journal of Power Sources 1980, 5, 255–262.

137 7 BIBLIOGRAPHY

[58] A. Celzard, E. McRae, C. Deleuze, M. Dufort, G. Furdin, J. F. Mareche; "Critical concentration in percolating systems containing a high-aspect-ratio filler", Physical Review B: Condensed Matter and Materials Physics 1996, 53, 6209–6214. [59] Showa Denko, Carbon Nanofiber VGCF-S, can be found under http://www.sdk.co.jp/english/products/126/132/2094/12067/12068.html [60] K. Xu; "Nonaqueous liquid electrolytes for lithium-based rechargeable batteries", Chemical Reviews 2004, 104, 4303–4417. [61] G.-A. Nazri; "Liquid Electrolytes: Some Theoretical and Practical Aspects" in Lithium Batteries: Science and Technology (Eds.: G.-A. Nazri, G. Pistoia), Springer, 2008. [62] D. Aurbach, Nonaqueous electrochemistry, Marcel Dekker, New York, 1999. [63] V. R. Koch; "The Stability of the Secondary Lithium Electrode in Tetrahydrofuran-Based Electrolytes", Journal of the Electrochemical Society 1978, 125, 1371. [64] D. Aurbach; "Review of selected electrode-solution interactions which determine the performance of Li and Li ion batteries", Journal of Power Sources 2000, 89, 206–218. [65] K. Izutsu, Electrochemistry in nonaqueous solutions, Wiley-VCH, Weinheim, 2009. [66] K. Izutsu; "History of the use of nonaqueous media in electrochemistry", Journal of Solid State Electrochemistry 2011, 15, 1719–1731. [67] A. N. Dey, B. P. Sullivan; "Electrochemical Decomposition of Propylene Carbonate on Graphite", Journal of the Electrochemical Society 1970, 117, 222. [68] D. Guyomard; "Rechargeable Li1+xMn2O4∕Carbon Cells with a New Electrolyte Composition", Journal of the Electrochemical Society 1993, 140, 3071. [69] J.-M. Tarascon, D. Guyomard; "New electrolyte compositions stable over the 0 to 5 V voltage range and compatible with the Li1+xMn2O4/carbon Li-ion cells", Solid State Ionics 1994, 69, 293–305. [70] D. Aurbach, Y. Ein-Eli, B. Markovsky, A. Zaban, S. Luski, Y. Carmeli, H. Yamin; "The study of electrolyte-solutions based on ethylene and diethyl carbonates for rechargeable Li batteries 2. graphite electrodes", Journal of the Electrochemical Society 1995, 142, 2882–2890. [71] D. Aurbach, K. Gamolsky, B. Markovsky, Y. Gofer, M. Schmidt, U. Heider; "On the use of vinylene carbonate (VC) as an additive to electrolyte solutions for Li-ion batteries", Electrochimica Acta 2002, 47, 1423–1439. [72] S. Zhang; "A review on electrolyte additives for lithium-ion batteries", Journal of Power Sources 2006, 162, 1379–1394. [73] M. Ue; "Role-Assigned Electrolytes: Additives" in Lithium-ion batteries. Science and technologies (Eds.: M. Yoshio, R. J. Brodd, A. Kozawa), Springer, New York, NY, 2009. [74] P. Arora, Z. M. Zhang; "Battery separators", Chemical Reviews 2004, 104, 4419–4462. [75] L. L. Celgard; "Lithium Battery Separators", can be found under www.celgard.com; 2013-03-23. [76] V. Hennige, G. Hörpel, C. Hying, M. Pascaly, M. Schuster, Herstellung und verwendung keramischer kompositmaterialien basierend auf polymer-trägerfolie, EP2425476 A1. [77] Evonik Industries; "Verbesserte Eigenschaften für Lithium-Ionen-Batterien", can be found under http://nano.evonik.de/sites/nanotechnology/de/technologie/ anwendung/separion/pages/default.aspx; 2013-03-23. [78] Bio-Logic Science Instruments; "Protocols for studying intercalation electrodes materials: Part II: Potentiodynamic Cycling with Galvanostatic Acceleration (PCGA)", can be found under www.bio-logic.info, 2013.

138 7 BIBLIOGRAPHY

[79] C. J. Wen; "Thermodynamic and Mass Transport Properties of “LiAl”", Journal of the Electrochemical Society 1979, 126, 2258. [80] Bio-Logic Science Instruments; "Protocols for studying intercalation electrodes materials: Part I: Galvanostatic cycling with potential limitation (GCPL)", can be found under www.biologic.info. [81] W. Weppner; "Determination of the Kinetic Parameters of Mixed-Conducting Electrodes and Application to the System Li[sub 3]Sb", Journal of the Electrochemical Society 1977, 124, 1569. [82] D. Guerard, A. Herold; "Intercalation of lithium into graphite and other carbons", Carbon 1975, 13, 337–345. [83] J. R. Dahn; "PHASE-DIAGRAM OF LIXC6", Physical Review B: Condensed Matter and Materials Physics 1991, 44, 9170–9177. [84] X. R. Zhang, R. Kostecki, T. J. Richardson, J. K. Pugh, P. N. Ross; "Electrochemical and infrared studies of the reduction of organic carbonates", Journal of the Electrochemical Society 2001, 148, A1341-A1345. [85] M. Winter; "The Solid Electrolyte Interphase - The Most Important and the Least Understood Solid Electrolyte in Rechargeable Li Batteries", Zeitschrift für Physikalische Chemie 2009, 223, 1395–1406. [86] P. Verma, P. Maire, P. Novák; "A review of the features and analyses of the solid electrolyte interphase in Li-ion batteries", Electrochimica Acta 2010, 55, 6332–6341. [87] R. Fong, U. von Sacken, J. R. Dahn; "Studies of Lithium Intercalation into Carbons Using Nonaqueous Electrochemical Cells", Journal of the Electrochemical Society 1990, 137, 2009– 2013. [88] J. O. Besenhard, M. Winter, J. Yang, W. Biberacher; "Filming mechanism of lithium-carbon anodes in organic and inorganic electrolytes", Journal of Power Sources 1995, 54, 228–231. [89] G. C. Chung, H. J. Kim, S. I. Yu, S. H. Jun, J. W. Choi, M. H. Kim; "Origin of graphite exfoliation - An investigation of the important role of solvent cointercalation", Journal of the Electrochemical Society 2000, 147, 4391–4398. [90] H. Buqa, A. Würsig, D. Goers, L. J. Hardwick, M. Holzapfel, P. Novák, F. Krumeich, M. E. Spahr; "Behaviour of highly crystalline graphites in lithium-ion cells with propylene carbonate containing electrolytes", Journal of Power Sources 2005, 146, 134–141. [91] M. E. Spahr, H. Wilhelm, T. Palladino, N. Dupont-Pavlovsky, D. Goers, F. Joho, P. Novák; "The role of graphite surface group chemistry on graphite exfoliation during electrochemical lithium insertion", Journal of Power Sources 2003, 119, 543–549. [92] M. E. Spahr, T. Palladino, H. Wilhelm, A. Würsig, D. Goers, H. Buqa, M. Holzapfel, P. Novák; "Exfoliation of Graphite during Electrochemical Lithium Insertion in Ethylene Carbonate- Containing Electrolytes", Journal of the Electrochemical Society 2004, 151, A1383. [93] D. Goers, M. E. Spahr, A. Leone, W. Märkle, P. Novák; "The influence of the local current density on the electrochemical exfoliation of graphite in lithium-ion battery negative electrodes", Electrochimica Acta 2011, 56, 3799–3808. [94] M. Winter, P. Novák, A. Monnier; "Graphites for lithium-ion cells: the correlation of the first- cycle charge loss with the Brunauer-Emmett-Teller surface area", Journal of the Electrochemical Society 1998, 145, 428–436. [95] N. R. Laine, F. J. Vastola, P. L. Walker Jr.; "The importance of active surface area in the carbon- oxygen reaction", Journal of Physical Chemistry 1963, 67, 2030-&.

139 7 BIBLIOGRAPHY

[96] W. P. Hoffman; "The Importance of Active Surface Area in the Heterogeneous Reactions of Carbon", Carbon 1991, 29, 769–776. [97] F. Béguin, F. Chevallier, C. Vix-Guterl, S. Saadallah, V. Bertagna, J. N. Rouzaud, E. Frackowiak; "Correlation of the irreversible lithium capacity with the active surface area of modified carbons", Carbon 2005, 43, 2160–2167. [98] P. Novák, J. Ufheil, H. Buqa, F. Krumeich, M. E. Spahr, D. Goers, H. Wilhelm, J. Dentzer, R. Gadiou, C. Vix-Guterl; "The importance of the active surface area of graphite materials in the first lithium intercalation", Journal of Power Sources 2007, 174, 1082–1085. [99] S.-H. Ng, C. Vix-Guterl, P. Bernardo, N. Tran, J. Ufheil, H. Buqa, J. Dentzer, R. Gadiou, M. E. Spahr, D. Goers et al.; "Correlations between surface properties of graphite and the first cycle specific charge loss in lithium-ion batteries", Carbon 2009, 47, 705–712. [100] D. Aurbach, Y. Ein-Eli, O. Chusid, Y. Carmeli, M. Babai, H. Yamin; "The correlation between the surface-chemistry and the performance of Li-carbon intercalation anodes for rechargeable rocking-chair type batteries", Journal of the Electrochemical Society 1994, 141, 603–611. [101] D. Aurbach, B. Markovsky, I. Weissman, E. Levi, Y. Ein-Eli; "On the correlation between surface chemistry and performance of graphite negative electrodes for Li ion batteries", Electrochimica Acta 1999, 45, 67–86. [102] M. E. Spahr, H. Wilhelm, F. Joho, J. C. Panitz, J. Wambach, P. Novák, N. Dupont-Pavlovsky; "Purely hexagonal graphite and the influence of surface modifications on its electrochemical lithium insertion properties", Journal of the Electrochemical Society 2002, 149, A960-A966. [103] M. E. Spahr, D. Goers, W. Märkle, J. Dentzer, A. Würsig, H. Buqa, C. Vix-Guterl, P. Novák; "Overpotentials and solid electrolyte interphase formation at porous graphite electrodes in mixed ethylene carbonate–propylene carbonate electrolyte systems", Electrochimica Acta 2010, 55, 8928–8937. [104] P. Bernardo, J. Dentzer, R. Gadiou, W. Märkle, D. Goers, P. Novák, M. E. Spahr, C. Vix-Guterl; "Influence of graphite surface properties on the first electrochemical lithium intercalation", Carbon 2011, 49, 4867–4876. [105] Z.-Q. Shi, C.-y. Wang, L.-p. Fan, F. He; "Study on natural graphite modified by H2O2 as anode material for lithium ion battery", Dianyuan Jishu 2004, 28, 609-611, 617. [106] P. Verma, T. Sasaki, P. Novák; "Chemical surface treatments for decreasing irreversible charge loss and preventing exfoliation of graphite in Li-ion batteries", Electrochimica Acta 2012, 82, 233–242. [107] H. J. Santner, C. Korepp, M. Winter, J. O. Besenhard, K. C. Moller; "In-situ FTIR investigations on the reduction of vinylene electrolyte additives suitable for use in lithium-ion batteries", Analytical and Bioanalytical Chemistry 2004, 379, 266–271. [108] P. Novák, D. Goers, L. J. Hardwick, M. Holzapfel, W. Scheifele, J. Ufheil, A. Würsig; "Advanced in situ characterization methods applied to carbonaceous materials", Journal of Power Sources 2005, 146, 15–20. [109] S. K. Martha, E. Markevich, V. Burgel, G. Salitra, E. Zinigrad, B. Markovsky, H. Sclar, Z. Pramovich, O. Heik, D. Aurbach et al.; "A short review on surface chemical aspects of Li batteries: A key for a good performance", Journal of Power Sources 2009, 189, 288–296. [110] A. J. Bard, M. Stratmann (Eds.) Encyclopedia of electrochemistry. Electrochemical Engineering, Wiley-VCH, Weinheim ;, [Cambridge], 2002-2004. [111] K. Fischbeck; "Ueber den Verlauf der elektrochemischen Reduktion fester Elektroden", Zeitschrift fuer anorganische und allgemeine Chemie 1925, 148, 97–129.

140 7 BIBLIOGRAPHY

[112] K. Fischbeck, E. Einecke; "Elektrochemische Reduktion fester Elektroden. II. Mitteilung", Zeitschrift fuer anorganische und allgemeine Chemie 1927, 167, 21–39. [113] J. J. Coleman; "Dry Cell Dynamics: The Bobbin", Transactions of The Electrochemical Society 1946, 90, 545. [114] V. S. Daniel-Bekh; "polarization of porous electrodes: about distribution of current and potential within the electrode", Zhurnal Fizicheskoi Khimii 1948, 22, 697. [115] J. J. Coleman; "Distribution of Current in Porous Electrodes", Journal of the Electrochemical Society 1951, 98, 26. [116] J. Euler, W. Nonnenmacher; "Stromverteilung in porösen Elektroden. Current distribution in porous electrodes", Electrochimica Acta 1960, 2, 268–286. [117] J. Euler, K.-N. Müller; "Über die Genauigkeit mit der sich Stromverteilungen in porösen Elektroden ermitteln lassen", Electrochimica Acta 1962, 8, 949–959. [118] J. Euler; "Zur Stromverteilung in porösen Braunsteinelektroden. Diffusion zwischen den einzelnen Lamellen", Electrochimica Acta 1963, 8, 409–417. [119] R. de Levie; "On porous electrodes in electrolyte solutions-IV", Electrochimica Acta 1964, 1231. [120] A. N. Frumkin; "Distribution of the corrosion process along a pipe", Zhurnal Fizicheskoi Khimii 1949, 23, 1477–1482. [121] E. Warburg; "Über das Verhalten sogenannter unpolarisierbarer Elektroden gegen den Wechselstrom", Annalen der Physik und Chemie 1899, 3, 493–499. [122] J. S. Newman, C. W. Tobias; "Theoretical analysis of current distribution in porous electrodes", Journal of the Electrochemical Society 1962, 109, 1183–91. [123] T. F. Fuller; "Simulation and Optimization of the Dual Lithium Ion Insertion Cell", Journal of the Electrochemical Society 1994, 141, 1. [124] G. G. Botte, V. R. Subramanian, R. E. White; "Mathematical modeling of secondary lithium batteries", Electrochimica Acta 2000, 45, 2595–2609. [125] V. R. Subramanian, V. Boovaragavan, V. D. Diwakar; "Toward Real-Time Simulation of Physics Based Lithium-Ion Battery Models", Electrochemical and Solid-State Letters 2007, 10, A255. [126] A. Nyman; "An Experimental and Theoretical Study of the Mass Transport in Lithium‐Ion Battery Electrolytes", Dissertation, Kungliga Tekniska Högskolan, Stockholm, 2011. [127] T. K. Dong, A. Kirchev, F. Mattera, J. Kowal, Y. Bultel; "Dynamic Modeling of Li-Ion Batteries Using an Equivalent Electrical Circuit", Journal of the Electrochemical Society 2011, 158, A326. [128] A. M. Stux, E. Gorzkowski, D. Rowenhorst, D. Stephenson, D. R. Wheeler; "Imaging and Modeling for Engineering the Li-ion Battery Electrode" in 214th ECS Meeting. [129] P. Shearing, L. Howard, P. Jørgensen, N. Brandon, S. J. Harris; "Characterization of the 3- dimensional microstructure of a graphite negative electrode from a Li-ion battery", Electrochemistry Communications 2010, 12, 374–377. [130] D. E. Stephenson, B. C. Walker, C. B. Skelton, E. P. Gorzkowski, D. J. Rowenhorst, D. R. Wheeler; "Modeling 3D Microstructure and Ion Transport in Porous Li-Ion Battery Electrodes", Journal of the Electrochemical Society 2011, 158, A781. [131] N. A. Zacharias, D. R. Nevers, C. Skelton, K. Knackstedt, D. E. Stephenson, D. R. Wheeler; "Direct Measurements of Effective Ionic Transport in Porous Li-Ion Electrodes", Journal of the Electrochemical Society 2012, 160, A306. [132] J. Liu, M. Kunz, K. Chen, N. Tamura, T. J. Richardson; "Visualization of Charge Distribution in a Lithium Battery Electrode", Journal of Physical Chemistry Letters 2010, 1, 2120–2123.

141 7 BIBLIOGRAPHY

[133] J. B. Siegel, X. F. Lin, A. G. Stefanopoulou, D. S. Hussey, D. L. Jacobson, D. Gorsich; "Neutron Imaging of Lithium Concentration in LFP Pouch Cell Battery", Journal of the Electrochemical Society 2011, 158, A523-A529. [134] K. Nishikawa, Y. Fukunaka, Sakka T., Y. H. Ogata, J. R. Selman; "Measurement of concentration profile during charging of Li battery anode materials in LiClO4-PC electrolyte", Electrochimica Acta 2007, 53, 218–223. [135] P. Maire, A. Evans, H. Kaiser, W. Scheifele, P. Novák; "Colorimetric Determination of Lithium Content in Electrodes of Lithium-Ion Batteries", Journal of the Electrochemical Society 2008, 155, A862. [136] S. J. Harris, A. Timmons, D. R. Baker, C. Monroe; "Direct in situ measurements of Li transport in Li-ion battery negative electrodes", Chemical Physics Letters 2010, 485, 265–274. [137] S. J. Harris; "Li Ion Battery Aging, Degradation, and Failure", can be found under http://lithiumbatteryresearch.com/; 2013-03-24. [138] K. C. Hess, J. F. Whitacre, S. Litster; "In Situ Measurements of Potential, Current and Charging Current across an EDL Capacitance Anode for an Aqueous Sodium Hybrid Battery", Journal of the Electrochemical Society 2012, 159, A1351. [139] F. La Mantia; "Characterization of Electrodes for Lithium-Ion Batteries through Electrochemical Impedance Spectroscopy and Mass Spectrometry", Dissertation, ETH, Zürich, 2008. [140] S.-H. Ng, F. La Mantia, P. Novák; "A Multiple Working Electrode for Electrochemical Cells: A Tool for Current Density Distribution Studies", Angewandte Chemie International Edition 2009, 48, 528–532. [141] G. Zhang, C. E. Shaffer, C.-Y. Wang, C. D. Rahn; "In-Situ Measurement of Current Distribution in a Li-Ion Cell", Journal of the Electrochemical Society 2013, 160, A610. [142] Eric Wieser; "A multi-walled armchair carbon nanotube. rendered in POVRay", can be found under http://en.wikipedia.org/wiki/File:Multi-walled_Carbon_Nanotube.png; 2013-01-28. [143] M. Rümmeli, P. Ayala, T. Pichler; "Production and Formation" in Carbon nanotubes and related structures. Synthesis, characterization, functionalization, and applications (Eds.: D. M. Guldi, N. Martin), Wiley-VCH, Weinheim, 2010. [144] Roempp online; "Kohlenstoffnanoröhren", can be found under www.roempp.com; 2010-08- 02. [145] H. Kroto, J. Heath, S. O'Brien, R. Curl, R. E. Smalley; "C-60 - Buckminsterfullerene", Nature 1985, 318, 162–163. [146] S. IIjima; "Helical Microtubules of Graphitic Carbon", Nature 1991, 354, 56–58. [147] L. V. Radushkevich, V. M. Lukyanovich; "Structure of the carbon produced in the thermal decomposition of carbon monoxide on an iron catalyst", Zhurnal Fizicheskoi Khimii 1952, 26, 88–95. [148] A. Oberlin, M. Endo, T. Koyama; "Filamentous growth of carbon through benzene decomposition", Journal of Crystal Growth 1976, 32, 335. [149] I. Applied Sciences; "Pyrograf III. carbon nanofibers", can be found under http://apsci.com/?page_id=19; 2013-02-02. [150] G. G. Wildgoose, P. Abiman, R. G. Compton; "Characterising chemical functionality on carbon surfaces", J. Mater. Chem 2009, 19, 4875. [151] J. Sandler, J. Kirk, I. Kinloch, M. Shaffer, A. Windle; "Ultra-low electrical percolation threshold in carbon-nanotube-epoxy composites", Polymer 2003, 44, 5893–5899.

142 7 BIBLIOGRAPHY

[152] P. Morgan, Carbon fibers and their composites, Taylor & Francis, Boca Raton, 2005. [153] Bayer MaterialScience AG, Baytubes. The Future is Nano. [154] Y. K. Zhou, J. Wang, Y. Y. Hu, R. O'Hayre, Z. P. Shao; "A porous LiFePO4 and carbon nanotube composite", Chemistry Communications 2010, 46, 7151–7153. [155] J. Lee, J. Bae, J. Heo, I. T. Han, S. N. Cha; "Effect of Randomly Networked Carbon Nanotubes in Silicon-Based Anodes for Lithium-Ion Batteries", Journal of the Electrochemical Society 2009, 156, A905-A910. [156] X.-M. Liu, Z. d. Huang, S. w. Oh, B. Zhang, P.-C. Ma, M. M. Yuen, J.-K. Kim; "Carbon nanotube (CNT)-based composites as electrode material for rechargeable Li-ion batteries: A review", Composites Science and Technology 2012, 72, 121–144. [157] T. Doi, A. Fukuda, Y. Iriyama, T. Abe, Z. Ogumi, K. Nakagawa, T. Ando; "Low-temperature synthesis of graphitized nanofibers for reversible lithium-ion insertion/extraction", Electrochemistry Communications 2005, 7, 10–13. [158] V. Subramanian, H. W. Zhu, B. Q. Wei; "High rate reversibility anode materials of lithium batteries from vapor-grown carbon nanofibers", Journal of Physical Chemistry B 2006, 110, 7178–7183. [159] L. Ji, X. Zhang; "Fabrication of porous carbon nanofibers and their application as anode materials for rechargeable lithium-ion batteries", Nanotechnology 2009, 20, 155705-15572. [160] B. J. Landi, R. A. Dileo, C. M. Schauerman, C. D. Cress, M. J. Ganter, R. P. Raffaelle, R. A. Dileo; "Multi-walled carbon nanotube paper anodes for lithium ion batteries", Journal of Nanoscience and Nanotechnology 2009, 9, 3406–3410. [161] S.-H. Ng, J. Wang, Z. P. Guo, J. Chen, G. X. Wang, H. K. Liu; "Single wall carbon nanotube paper as anode for lithium-ion battery", Electrochimica Acta 2005, 51, 23–28. [162] S. Y. Chew, S.-H. Ng, J. Wang, P. Novák, F. Krumeich, S. L. Chou, J. Chen, H. K. Liu; "Flexible free- standing carbon nanotube films for model lithium-ion batteries", Carbon 2009, 47, 2976–2983. [163] L. Hu, H. Wu, F. La Mantia, Y. Yang, Y. Cui; "Thin, Flexible Secondary Li-Ion Paper Batteries", ACS Nano 2010, 4, 5843–5848. [164] V. Meunier, J. Kephart, C. Roland, J. Bernholc; "Ab initio investigations of lithium diffusion in carbon nanotube systems", Physical Review Letters 2002, 88, 75506. [165] G. Maurin; "Electrochemical lithium intercalation into multiwall carbon nanotubes: a micro- Raman study", Solid State Ionics 2000, 136, 1295–1299. [166] Z. H. Yang; "Electrochemical intercalation of lithium into carbon nanotubes", Solid State Ionics 2001, 143, 173–180. [167] B. J. Landi, M. J. Ganter, C. D. Cress, R. A. Dileo, R. P. Raffaelle; "Carbon nanotubes for lithium ion batteries", Energy & Environmental Science 2009, 2, 638. [168] B. J. Landi, M. J. Ganter, C. M. Schauerman, C. D. Cress, R. P. Raffaelle; "Lithium ion capacity of single wall carbon nanotube paper electrodes", Journal of Physical Chemistry C 2008, 112, 7509–7515. [169] R. L. D. Whitby, T. Fukuda, T. Maekawa, S. L. James, S. V. Mikhalovsky; "Geometric control and tuneable pore size distribution of buckypaper and buckydiscs", Carbon 2008, 46, 949–956. [170] F. Béguin, K. Metenier, R. Pellenq, S. Bonnamy, E. Frackowiak; "Lithium Insertion in Carbon Nanotubes", Molecular Crystals and Liquid Crystals 2000, 340, 547–552. [171] J. H. Ahn, H. S. Shin, Y. J. Kim, H. Chung; "Structural modification of carbon nanotubes by various ball milling", Journal of Alloys and Compounds 2007, 428-432, 428–432.

143 7 BIBLIOGRAPHY

[172] J. Eom, D. Kim, H. Kwon; "Effects of ball-milling on lithium insertion into multi-walled carbon nanotubes synthesized by thermal chemical vapour deposition", Journal of Power Sources 2006, 157, 507–514. [173] K. L. Klein, A. V. Melechko, T. E. McKnight, S. T. Retterer, P. D. Rack, J. D. Fowlkes, D. C. Joy, M. L. Simpson; "Surface characterization and functionalization of carbon nanofibers", Journal of Applied Physics 2008, 103, 61301. [174] N. Karousis, N. Tagmatarchis, D. Tasis; "Current Progress on the Chemical Modification of Carbon Nanotubes", Chemical Reviews 2010, 110, 5366–5397. [175] B. C. Satishkumar, A. Govindaraj, J. Mofokeng, G. N. Subbanna, C. N. Rao; "Novel experiments with carbon nanotubes: Opening, filling, closing and functionalizing nanotubes", Journal of Physics B 1996, 29, 4925–4934. [176] G. J. Yu, J. L. Gong, S. Wang, D. Z. Zhu, S. X. He, Z. Y. Zhu; "Etching effects of ethanol on multi- walled carbon nanotubes", Carbon 2006, 44, 1218–1224. [177] K. Balasubramanian, M. Burghard; "Chemically Functionalized Carbon Nanotubes", small 2005, 1, 180–192. [178] Y.-C. Chiang, W.-H. Lin, Y.-C. Chang; "The influence of treatment duration on multi-walled carbon nanotubes functionalized by H2SO4/HNO3 oxidation", Applied Surface Science 2011, 257, 2401–2410. [179] V. Datsyuk, M. Kalyva, K. Papagelis, J. Parthenios, D. Tasis, A. Siokou, I. Kallitsis, C. Galiotis; "Chemical oxidation of multiwalled carbon nanotubes", Carbon 2008, 46, 833–840. [180] P. V. Lakshminarayanan, H. Toghiani, C. U. Pittman Jr.; "Nitric acid oxidation of vapor grown carbon nanofibers", Carbon 2004, 42, 2433–2442. [181] M. T. Martinez, M. A. Callejas, Am Benito, M. Cochet, T. Seeger, A. Anson, J. Schreiber, C. Gordon, C. Marhic, O. Chauvet et al.; "Sensitivity of single wall carbon nanotubes to oxidative processing: structural modification, intercalation and functionalisation", Carbon 2003, 41, 2247–2256. [182] P. M. Ajayan, T. W. Ebbesen, T. Ichihashi, S. IIjima, K. Tanigaki, H. Hiura; "Opening carbon nanotubes with oxygen and implications for filling", Nature 1993, 362, 522–525. [183] M. L. Toebes, J. M. van Heeswijk, J. H. Bitter, A. van Jos Dillen, K. P. de Jong; "The influence of oxidation on the texture and the number of oxygen-containing surface groups of carbon nanofibers", Carbon 2004, 42, 307–315. [184] H. Shimoda, B. Gao, X. P. Tang, A. Kleinhammes, L. Fleming, Y. Wu, O. Zhou; "Lithium intercalation into opened single-wall carbon nanotubes: Storage capacity and electronic properties", Physical Review Letters 2001, 88, 015502-1-015502-4. [185] G. Moraitis, Z. Špitalský, F. Ravani, A. Siokou, C. Galiotis; "Electrochemical oxidation of multiwall carbon nanotubes", Carbon 2011, 49, 2702–2708. [186] K. Balasubramanian, M. Burghard; "Electrochemically functionalized carbon nanotubes for device applications", Journal of Materials Chemistry 2008, 18, 3071–3083. [187] M. Burghard, A. Maroto, K. Balasubramanian, T. Assmus, A. Forment-Aliaga; "Electrochemically modified single-walled carbon nanotubes", physica status solidi (b) 2007, 244, 4021–4025. [188] C. G. Salzmann, S. A. Llewellyn, G. Tobias, M. A. Ward, Y. Huh, M. L. Green; "The role of carboxylated carbonaceous fragments in the functionalization and spectroscopy of a single- walled carbon-nanotube material", Avanced Materials 2007, 19, 883.

144 7 BIBLIOGRAPHY

[189] Z. Wang, M. D. Shirley, S. T. Meikle, R. L. D. Whitby, S. V. Mikhalovsky; "The surface acidity of acid oxidised multi-walled carbon nanotubes and the influence of in-situ generated fulvic acids on their stability in aqueous dispersions", Carbon 2009, 47, 73–79. [190] H. C. Shin, M. L. Liu, B. Sadanadan, Am Rao; "Lithium insertion into chemically etched multi- walled carbon nanotubes", Journal of Solid State Electrochemistry 2004, 8, 908–913. [191] S. Yang, H. Song, X. Chen, A. V. Okotrub, L. G. Bulusheva; "Electrochemical performance of arc- produced carbon nanotubes as anode material for lithium-ion batteries", Electrochimica Acta 2007, 52, 5286–5293. [192] J. Eom, H. Kwon; "Improved lithium insertion/extraction properties of single-walled carbon nanotubes by high-energy ball milling", Journal of Materials Research 2008, 23, 2458–2466. [193] Y. B. Fu, R. B. Ma, Y. M. Chen, D. D. Jiang, Q. Y. Zhang, X. H. Ma; "The effect of acidic treatment on the lithium storage capacity of multi-walled carbon nanotubes", Journal of Materials Science - Materials in Electronics 2009, 20, 709–712. [194] H. A. Strobel, Chemical instrumentation. A systematic approach, Addison-Wesley Pub. Co., Reading, Mass, 1973. [195] B. Welz, Atomic absorption spectrometry, Wiley-VCH, Weinheim ;, New York, 1999. [196] F. Tuinstra, J. L. Koenig; "Raman Spectrum of Graphite", Journal of Chemical Physics 1970, 53, 1126-&. [197] M. S. Dresselhaus, G. Dresselhaus, R. Saito, A. Jorio; "Raman spectroscopy of carbon nanotubes", PHYS REP 2005, 409, 47–99. [198] H. Hiura, T. W. Ebbesen, K. Tanigaki, H. TAKAHASHI; "Raman studies of carbon nanotubes", Chemical Physics Letters 1993, 202, 509–512. [199] G. E. McGuire, J. Fuchs, P. Han, J. G. Kushmerick, P. S. Weiss, S. J. Simko, R. J. Nemanich, D. R. Chopra; "Surface characterization", Analytical Chemistry 1999, 71, 373R-388R. [200] L. A. Langley, D. E. Villanueva, D. H. Fairbrother; "Quantification of Surface Oxides on Carbonaceous Materials", Chemistry of Materials 2006, 18, 169–178. [201] H. P. Boehm; "Some Aspects of the Surface-Chemistry of Carbon-blacks and other Carbons", Carbon 1994, 32, 759–769. [202] H. P. Boehm; "Surface oxides on carbon and their analysis: a critical assessment", Carbon 2002, 40, 145–149. [203] E. Barsoukov, J. R. Macdonald (Eds.) Impedance spectroscopy. Theory, experiment, and applications, Wiley-Interscience, Hoboken, N.J, 2005. [204] M. E. Orazem, B. Tribollet, Electrochemical Impedance Spectroscopy, John Wiley & Sons, Inc., Hoboken, N.J, 2008. [205] R. S. Rodgers; "Resources for Electrochemistry", can be found under http://www.consultrsr.com/resources/eis/index.htm; 2012-01-14. [206] E. Barsoukov; "Evgenij Barsoukov's collection of impedance spectroscopy resources", can be found under http://impedance0.tripod.com/; 2012-01-14. [207] J. E. B. Randles; "Kinetics of rapid electrode reactions", Discussions of the Faraday Society 1947, 1, 11. [208] Wikipedia; "Wheatstone's bridge circuit diagram", can be found under http://en.wikipedia.org/wiki/File:Wheatstonebridge.svg.

145 7 BIBLIOGRAPHY

[209] M. C. H. McKubre, D. D. Macdonald; "Measuring Techniques and Data Analysis" in Impedance spectroscopy. Theory, experiment, and applications (Eds.: E. Barsoukov, J. R. Macdonald), Wiley-Interscience, Hoboken, N.J, 2005. [210] Bio-Logic Science Instruments, EC-Lab, Bio-Logic Science Instruments, Claix. [211] K. S. Cole, R. H. Cole; "Dispersion and Absorption in Dielectrics I. Alternating Current Characteristics", Journal of Chemical Physics 1941, 9, 341. [212] I. D. Raistrick, D. R. Franceschetti, J. R. Macdonald; "Mass and Charge Transport in the Presence of Concentration Gradients" in Impedance spectroscopy. Theory, experiment, and applications (Eds.: E. Barsoukov, J. R. Macdonald), Wiley-Interscience, Hoboken, N.J, 2005. [213] E. Barsoukov; "Special Aspects of Impedance Modeling of Power Sources" in Impedance spectroscopy. Theory, experiment, and applications (Eds.: E. Barsoukov, J. R. Macdonald), Wiley-Interscience, Hoboken, N.J, 2005. [214] R. de Levie; "Electrochemical response of porous and rough electrodes" in Advances in Electrochemistry and Electrochemical Engineering (Ed.: P. Delahay), Interscience, New York. [215] A. Lasia; "Impedance of Porous-Electrodes", Journal of Electroanalytical Chemistry 1995, 397, 27–33. [216] A. Lasia; "Porous electrodes in the presence of a concentration gradient", Journal of Electroanalytical Chemistry 1997, 428, 155–164. [217] D. Aurbach, A. Zaban; "Impedance spectroscopy of lithium electrodes. Part 1. General behavior in propylene carbonate solutions and the correlation to surface chemistry and cycling efficiency", Journal of Electroanalytical Chemistry. [218] D. Aurbach, A. Zaban; "Impedance Spectroscopy of Lithium Electrodes. Part 2. The Behavior in Propylene Carbonate Solutions - The Significance of the Data Obtained", Journal of Electroanalytical Chemistry 1994, 367, 15–25. [219] D. Aurbach, A. Zaban; "Impedance spectroscopy of lithium electrodes. Part 3. The importance of Li electrode surface preparation", Journal of Electroanalytical Chemistry. [220] M. D. Levi, D. Aurbach; "Simultaneous measurements and modeling of the electrochemical impedance and the cyclic voltammetric characteristics of graphite electrodes doped with lithium", Journal of Physical Chemistry B 1997, 101, 4630–4640. [221] M. D. Levi, D. Aurbach; "Diffusion coefficients of lithium ions during intercalation into graphite derived from the simultaneous measurements and modeling of electrochemical impedance and potentiostatic intermittent titration characteristics of thin graphite electrodes", Journal of Physical Chemistry B 1997, 101, 4641–4647. [222] D. Aurbach, M. D. Levi, E. Levi, H. Teller, B. Markovsky, G. Salitra, U. Heider, L. Heider; "Common electroanalytical behavior of Li intercalation processes into graphite and transition metal oxides", Journal of the Electrochemical Society 1998, 145, 3024–3034. [223] E. Barsoukov; "Generic Approach to Battery Impedance Modeling" in Impedance spectroscopy. Theory, experiment, and applications (Eds.: E. Barsoukov, J. R. Macdonald), Wiley-Interscience, Hoboken, N.J, 2005. [224] E. Barsoukov, J. H. Kim, C. O. Yoon, H. Lee; "Effect of low-temperature conditions on passive layer growth on Li intercalation materials - In situ impedance study", Journal of the Electrochemical Society 1998, 145, 2711–2717. [225] S. Zhang, K. Xu, T. Jow; "EIS study on the formation of solid electrolyte interface in Li-ion battery", Electrochimica Acta 2006, 51, 1636–1640.

146 7 BIBLIOGRAPHY

[226] M. Itagaki, S. Yotsuda, N. Kobari, K. Watanabe, S. Kinoshita, M. Ue; "Electrochemical impedance of electrolyte/electrode interfaces of lithium-ion rechargeable batteries", Electrochimica Acta 2006, 51, 1629–1635. [227] J. Wang, J. Polleux, J. Lim, B. Dunn; "Pseudocapacitive Contributions to Electrochemical Energy Storage in TiO2 (Anatase) Nanoparticles", Journal of Physical Chemistry C 2007, 111, 14925– 14931. [228] M. Ender, A. Weber, E. Ivers-Tiffée; "Analysis of Three-Electrode Setups for AC-Impedance Measurements on Lithium-Ion Cells by FEM simulations", Journal of the Electrochemical Society 2012, 159, A128. [229] H. Goehr, M. Mirnik, C. Schiller; "Distortions of high frequency electrode impedance. their causes and how to avoid them", J. ElectroAnalytical Chemistry 1984, 180, 273–285. [230] S. Fletcher; "The two-terminal equivalent network of a three-terminal electrochemical cell", ElectroChemistry Communications 2001. [231] F. La Mantia, J. Vetter, P. Novák; "Impedance spectroscopy on porous materials: A general model and application to graphite electrodes of lithium-ion batteries", Electrochimica Acta 2008, 53, 4109–4121. [232] J. P. Meyers, M. Doyle, R. M. Darling, J. Newman; "The Impedance Response of a Porous Electrode Composed of Intercalation Particles", Journal of the Electrochemical Society 2000, 147, 2930. [233] N. Ogihara, S. Kawauchi, C. Okuda, Y. Itou, Y. Takeuchi, Y. Ukyo; "Theoretical and Experimental Analysis of Porous Electrodes for Lithium-Ion Batteries by Electrochemical Impedance Spectroscopy Using a Symmetric Cell", Journal of the Electrochemical Society 2012, 159, A1034-A1039. [234] Wikipedia; "Butler–Volmer equation graph", can be found under http://upload.wikimedia.org/wikipedia/commons/f/fe/Butler_volmer_equation_graph_nl.png; 2012-01-14. [235] R. G. Compton, G. H. W. Sanders, Electrode potentials, Oxford University Press, Oxford, 1996. [236] S. Barnartt; "Primary current distribution around capillary tips used in the measurement of electrolytic polarization", Journal of the Electrochemical Society 1952, 99, 549–553. [237] C. Fiaud, M. Keddam, A. Kadri, H. Takenouti; "Electrochemical impedance in a thin surface electrolyte layer. Influence of the potential probe location", Electrochimica Acta 1987, 32, 445–448. [238] G. Hsieh, S. J. Ford, T. O. Mason, L. R. Pederson; "Experimental limitations in impedance spectroscopy .1. Simulation of reference electrode artifacts in three-point measurements", Solid State Ionics 1996, 91, 191–201. [239] G. Hsieh, T. O. Mason, L. R. Pederson; "Experimental limitations in impedance spectroscopy .2. Electrode artifacts in three-point measurements on Pt/YSZ", Solid State Ionics 1996, 91, 203– 212. [240] G. Hsieh, S. J. Ford, T. O. Mason, L. R. Pederson; "Experimental limitations in impedance spectroscopy .6. Four-point measurements of solid materials systems", Solid State Ionics 1997, 100, 297–311. [241] G. Hsieh, T. O. Mason, E. J. Garboczi, L. R. Pederson; "Experimental limitations in impedance spectroscopy .3. Effect of reference electrode geometry/position", Solid State Ionics 1997, 96, 153–172.

147 7 BIBLIOGRAPHY

[242] M. Cimenti, A. C. Co, V. I. Birss, J. M. Hill; "Distortions in Electrochemical Impedance Spectroscopy Measurements Using 3-Electrode Methods in SOFC. I – Effect of Cell Geometry", Fuel Cells 2007, 7, 364–376. [243] M. Cimenti, V. I. Birss, J. M. Hill; "Distortions in Electrochemical Impedance Spectroscopy Measurements Using 3-Electrode Methods in SOFC. II. Effect of Electrode Activity and Relaxation Times", Fuel Cells 2007, 7, 377–391. [244] J. Winkler, P. V. Hendriksen, N. Bonanos, M. Mogensen; "Geometric requirements of solid electrolyte cells with a reference electrode", Journal of the Electrochemical Society 1998, 145, 1184–1192. [245] S. B. Adler, B. T. Henderson, M. A. Wilson, D. M. Taylor, R. E. Richards; "Reference electrode placement and seals in electrochemical oxygen generators", Solid State Ionics 2000, 134, 35– 42. [246] S. B. Adler; "Reference Electrode Placement in Thin Solid Electrolytes", Journal of the Electrochemical Society 2002, 149, E166. [247] V. R. Koch, C. Nanjundiah, G. B. Appetecchi, B. Scrosati; "The Interfacial Stability of Li with 2 New Solvent - Free Ionic Liquids - 1,2-Dimethyl-3-Propylimidazolium Imide and Methide", Journal of the Electrochemical Society 1995, 142, L116-L118. [248] S. Klink, E. Madej, E. Ventosa, A. Lindner, W. Schuhmann, F. La Mantia; "The importance of cell geometry for electrochemical impedance spectroscopy in three-electrode lithium ion battery test cells", Electrochemistry Communications 2012, 22, 120–123. [249] M. Dolle, F. Orsini, A. S. Gozdz, J.-M. Tarascon; "Development of reliable three-electrode impedance measurements in plastic Li-ion batteries", Journal of the Electrochemical Society 2001, 148, A851-A857. [250] J. Zhou, P. H. L. Notten; "Development of Reliable Lithium Microreference Electrodes for Long- Term In Situ Studies of Lithium-Based Battery Systems", Journal of the Electrochemical Society 2004, 151, A2173. [251] H. Duncan, Y. Abu-Lebdeh, I. J. Davidson; "Study of the Cathode–Electrolyte Interface of LiMn[sub 1.5]Ni[sub 0.5]O[sub 4] Synthesized by a Sol–Gel Method for Li-Ion Batteries", Journal of the Electrochemical Society 2010, 157, A528. [252] C. Wang, A. J. Appleby, F. E. Little; "Criteria for reliable electrochemical impedance measurements on Li-ion battery anodes", Journal of the Electrochemical Society 2003, 150, A143-A148. [253] T. S. Ong, H. Yang; "Symmetrical cell for electrochemical AC impedance studies of lithium intercalation into graphite", Electrochemical and Solid-State Letters 2001, 4, A89-A92. [254] D. D. Macdonald; "Application of Kramers-Kronig Transforms in the Analysis of Electrochemical Systems", Journal of the Electrochemical Society 1985, 132, 2316. [255] S. Klink, E. Ventosa, W. Xia, F. La Mantia, M. Muhler, W. Schuhmann; "Tailoring of CNT surface oxygen groups by gas-phase oxidation and its implications for lithium ion batteries", ElectroChemistry Communications 2012, 15, 10–13. [256] W. Xia, C. Jin, S. Kundu, M. Muhler; "A highly efficient gas-phase route for the oxygen functionalization of carbon nanotubes based on nitric acid vapor", Carbon 2009, 47, 919–922. [257] S. Kundu, Y. Wang, W. Xia, M. Muhler; "Thermal Stability and Reducibility of Oxygen- Containing Functional Groups on Multiwalled Carbon Nanotube Surfaces: A Quantitative High- Resolution XPS and TPD/TPR Study", Journal of Physical Chemistry C 2008, 112, 16869–16878.

148 7 BIBLIOGRAPHY

[258] W. S. Bacsa; "High-Resolution Electron-Microscopy and Inelastic Light-Scattering of Purified Multishelled Carbon Nanotubes", Physical Review B: Condensed Matter and Materials Physics 1994, 50, 15473–15476. [259] W. Z. Li; "Raman characterization of aligned carbon nanotubes produced by thermal decomposition of hydrocarbon vapor", Applied Physics Letters 1997, 70, 2684–2686. [260] W. Xia, weight loss vs. time during gas-phase oxidation of CNF, Bochum. [261] T. Okpalugo, P. Papakonstantinou, H. Murphy, J. McLaughlin, N. Brown; "High resolution XPS characterization of chemical functionalised MWCNTs and SWCNTs", Carbon 2005, 43, 153– 161. [262] G. Socrates, Infrared and raman characteristic group frequencies. Tables and charts, John Wiley & Sons, West Sussex, 2007. [263] A. F. Holloway, G. G. Wildgoose, R. G. Compton, L. Shao, M. L. H. Green; "The influence of edge-plane defects and oxygen-containing surface groups on the voltammetry of acid-treated, annealed and “super-annealed” multiwalled carbon nanotubes", Journal of Solid State Electrochemistry 2008, 12, 1337–1348. [264] E. Ventosa, W. Xia, S. Klink, F. La Mantia, M. Muhler, W. Schuhmann; "Influence of surface functional groups on lithium ion intercalation of carbon cloth", Electrochimica Acta 2012. [265] T. Schilling, M. Bron; "Oxygen reduction at Fe-N-modified multi-walled carbon nanotubes in acidic electrolyte", Electrochimica Acta 2008, 53, 5379–5385. [266] C. G. Zoski, Handbook of Electrochemistry, Elsevier, Amsterdam ;, Boston, 2007. [267] J. March, Advanced Organic Chemistry, Wiley, 1992. [268] R. K. Raju, J. W. G. Bloom, Y. An, S. E. Wheeler; "Substituent Effects on Non-Covalent Interactions with Aromatic Rings: Insights from Computational Chemistry", ChemPhysChem 2011, 12, 3116–3130. [269] S. Mecozzi, A. P. West, D. A. Dougherty; "Cation−π Interactions in Simple Aromatics: Electrostatics Provide a Predictive Tool", Journal of the American Chemical Society 1996, 118, 2307–2308. [270] C. H. Suresh, S. R. Gadre; "Electrostatic Potential Minimum of the Aromatic Ring as a Measure of Substituent Constant", Journal of Physical Chemistry A 2007, 111, 710–714. [271] J. P. Gallivan, D. A. Dougherty; "A Computational Study of Cation−π Interactions vs Salt Bridges in Aqueous Media: Implications for Protein Engineering", Journal of the American Chemical Society 2000, 122, 870–874. [272] S. Klink, D. Höche, W. Schuhmann, F. La Mantia; "FEM modelling of a coaxial 3-electrode test cell for electrochemical impedance spectroscopy in lithium ion batteries", Journal of Power Sources 2013, 273-280. [273] COMSOL Multiphysics GmbH, Comsol, Göttingen, 2012. [274] S. Klink, W. Schuhmann, F. La Mantia; "Vertical Distribution of Charge and Charge Losses in Lithium Ion Batteries, Energy & Environmental Science, in preparation 2013. [275] K. J. Euler; "Ueber die Verteilung des Innenwiderstandes in galvanischen Primaerzellen", Archiv fuer Elektrotechnik 1966, 51, 123–130. [276] K. Euler; "Die Änderung der Stromverteilung in porösen positiven Elektroden von Akkumulatoren und galvanischen Primärzellen während Ladung und Entladung", Electrochimica Acta 1968, 13, 1533–1549.

149 7 BIBLIOGRAPHY

[277] Dexmet Corporation; "MicroGrid® Metals", can be found under http://www.dexmet.com/Expanded-Metals.html; 2013-02-16. [278] Timcal Ltd., Timrex SFG6, Bodio, Switzerland, 2010. [279] Timcal Ltd., Timrex SFG44, Bodio, Switzerland, 2010. [280] M. Winter, J. O. Besenhard, M. E. Spahr, P. Novák; "Insertion Electrode Materials for Rechargeable Lithium Batteries", Avanced Materials 1998, 10, 725–763. [281] C. S. Wang, A. J. Appleby, F. E. Little; "Electrochemical impedance study of initial lithium ion intercalation into graphite powders", Electrochimica Acta 2001, 46, 1793–1813. [282] C. Wang, I. Kakwan, A. J. Appleby, F. E. Little; "In situ investigation of electrochemical lithium intercalation into graphite powder", Journal of Electroanalytical Chemistry 2000, 489, 55–67. [283] A. K. Padhi, K. S. Nanjundaswamy, J. B. Goodenough; "Phospho-olivines as Positive-Electrode Materials for Rechargeable Lithium Batteries", Journal of the Electrochemical Society 1997, 144, 1188. [284] F. La Mantia, C. Wessells, H. Deshazer, Y. Cui; "Reliable reference electrodes for lithium-ion batteries", Electrochemistry Communications 2013. [285] X.-J. Wang, H.-Y. Chen, X. Yu, L. Wu, K.-W. Nam, J. Bai, H. Li, X. Huang, X.-Q. Yang; "A new in situ synchrotron X-ray diffraction technique to study the chemical delithiation of LiFePO4", Chemistry Communications 2011, 47, 7170. [286] F. Béguin, F. Chevallier, C. Vix, S. Saadallah, J. N. Rouzaud, E. Frackowiak; "A better understanding of the irreversible lithium insertion mechanisms in disordered carbons", Journal of Physics and Chemistry of Solids 2004, 65, 211–217. [287] C. Menachem, E. Peled, L. Burstein; "Improvement of graphite anode for lithium ion battery by mild oxidation", Proceedings of the Power Sources Conference 1996, 37th, 208–211. [288] B. Gao, A. Kleinhammes, X. P. Tang, C. Bower, L. Fleming, Y. Wu, O. Zhou; "Electrochemical intercalation of single-walled carbon nanotubes with lithium", Chemical Physics Letters 1999, 307, 153–157. [289] Z. H. Yang, S. B. Sang, Y. Feng, Y. H. Zhou, K. L. Huang, H. Q. Wu; "Electrochemical impedance study of Li-ion insertion into the raw acid-oxidized carbon nanotubes", Diamond Related Materials 2005, 14, 1302–1310. [290] B. Gao, C. Bower, J. D. Lorentzen, L. Fleming, A. Kleinhammes, X. P. Tang; "Enhanced saturation lithium composition in ball-milled single-walled carbon nanotubes", Chemical Physics Letters 2000, 327, 69. [291] S. Kawasaki, Y. Iwai, M. Hirose; "Electrochemical lithium ion storage properties of single-walled carbon nanotubes containing organic molecules", Carbon 2009, 47, 1081–1086. [292] S. W. Lee, N. Yabuuchi, B. M. Gallant, S. Chen, B.-S. Kim, P. T. Hammond, Y. Shao-Horn; "High- power lithium batteries from functionalized carbon-nanotube electrodes", Nature Nanotechnology 2010, 5, 531–537. [293] H. Jannesari, M. Emami, C. Ziegler; "Effect of electrolyte transport properties and variations in the morphological parameters on the variation of side reaction rate across the anode electrode and the aging of lithium ion batteries", Journal of Power Sources 2011, 196, 9654– 9664. [294] T. Marks, S. Trussler, A. J. Smith, D. Xiong, J. R. Dahn; "A Guide to Li-Ion Coin-Cell Electrode Making for Academic Researchers", Journal of the Electrochemical Society 2011, 158, A51. [295] R. J. Brodd, A. Kozawa, M. Yoshio (Eds.) Lithium-ion batteries. Science and technologies, Springer, New York, NY, 2010.

150 7 BIBLIOGRAPHY

[296] W. Dreyer, J. Jamnik, C. Guhlke, R. Huth, J. Moškon, M. Gaberšček; "The thermodynamic origin of hysteresis in insertion batteries", Nature Materials 2010, 9, 448–453. [297] A. Hickling; "Studies in electrode polarisation. Part IV. The automatic control of the potential of a working electrode", Transactions of the Faraday Society 1942, 38, 27. [298] Paul Israel; "The Thomas A. Edison Papers Project", can be found under http://edison.rutgers.edu/index.htm; 2013-04-13. [299] W. F. van Gunsteren; "The Seven Sins in Academic Behavior in the Natural Sciences", Angewandte Chemie International Edition 2013, 52, 118–122. [300] D. G. Nocera; "Living healthy on a dying planet", Chemical Society Reviews 2009, 38, 13–15.

151 PUBLICATIONS AND TALKS

8 APPENDIX – TECHNICAL DRAWINGS

8.1 SWAGELOK THREE-ELECTRODE T-CELL

152 PUBLICATIONS AND TALKS

153 PUBLICATIONS AND TALKS

154 PUBLICATIONS AND TALKS

8.2 COAXIAL IMPEDANCE CELL

155 PUBLICATIONS AND TALKS

156 PUBLICATIONS AND TALKS

157 PUBLICATIONS AND TALKS

158 PUBLICATIONS AND TALKS

159 PUBLICATIONS AND TALKS

160 PUBLICATIONS AND TALKS

161 PUBLICATIONS AND TALKS

162 PUBLICATIONS AND TALKS

8.3 MULTI-LAYERED ELECTROCHEMICAL CELL

163 PUBLICATIONS AND TALKS

164 PUBLICATIONS AND TALKS

165 PUBLICATIONS AND TALKS

166 PUBLICATIONS AND TALKS

167