Upload
others
View
6
Download
0
Embed Size (px)
Citation preview
Nanostructured Electrochemically Active
Electrodes for Applications in Energy
Generation and Storage Devices
Peter Lynch
A thesis submitted for the degree of
Doctor of philosophy in physics
Supervised by Prof. Jonathan N. Coleman
Chemical Physics of Low Dimensional Nanostructures Group
School of Physics
Trinity College Dublin
2016
To Mum. Dad and Declan Casey
DECLARATION
I declare that this thesis has not been submitted as an exercise for a degree at this or
any other university and it is entirely my own work.
I agree to deposit this thesis in the University’s open access institutional repository or
allow the library to do so on my behalf, subject to Irish Copyright Legislation and Trinity
College Library conditions of use and acknowledgement.
Elements of this work that have been carried out jointly with others or by
collaborators have been duly acknowledged in the text wherever included.
Peter Lynch
1
Abstract
In this thesis various materials with dimensions in the nanoscale have been
investigated for use in electrochemical systems. For dye-sensitized solar cells the challenge
of replacing the expensive platinum in the catalytic counter electrode with graphene was
investigated. The influence of size of the flakes with a mass of 0.1mg/cm2 with efficiencies as
a percentage of the reference platinum cell going from 15% for large flakes to 35% for the
smallest flakes. This proved to be less effective for improving the efficiency than increasing
the film thickness where the efficiency as a percentage of platinum increase from 45% to
80% as thicknesses increased from 50 to 1000nm. However, creating films with efficiencies
in excess of 80% that were comparable to platinum remained elusive. As such addition of
other materials into the graphene film to bridge the gap between graphene and platinum was
attempted. By enhancing conductivity, particularly in the vertical direction which was limited
due to the anisotropy of charge transport in graphene networks, with carbon nanotubes the
efficiency became comparable to platinum at 96%. The addition of a more catalytic material,
MoS2, also produced similar results (efficiency compared to platinum reference cell of 95%)
with an additional advantage of the material being cheaper due to its presence in nature. On
investigation of the performance of the electrodes using percolation theory it was revealed
that while the edges of MoS2 are more catalytically active the main advantage of using the
MoS2 was that the nanosheets were on average smaller using the same processing conditions.
The smaller sheets in the lateral dimensions allowed for a higher length to area ratio
increasing the percentage mass of the particles contributing to the catalytic activity of the
material. The relative reactivity of MoS2 edges to graphene edges was a factor 1.5.
Supercapacitor electrode materials in the form of PEDOT:PSS films prepared by a
variety of methods and treated with formic acid to increase conductivity were studied as well
in this thesis. Two models were compared to describe the effect of increasing thickness on the
capacitance per unit area. Both models returned capacitances per unit area within 15% of
each other allowing both models to be considered accurate for investigation of that property.
This capacitance represented a capacitance per unit mass of 33-38.5 F/g depending on the
model used. This corresponds to the capacitance of the PEDOT in the film as the PSS
component does not contribute to the capacitance. When looking at the current-voltage
2
characteristics or the time constant however, the Pell & Conway model which accounts for an
initial charge on the electrode performs better than the Higgins & Coleman model which
assumes no charge initially. The role of diffusion affects the capacitance per unit area at
higher scan rates. Neither of the models account for it and from this the diffusion coefficient
was estimated to be sm /103.63.5 210 which is about 30-40% that of the ions in water.
When accounting for time constant and scan rate the effect of the electrical properties of the
film with length and thickness were found to have a greater effect on film performance than
diffusion. For completeness freeze dried foams of the same material were fabricated to both
reduce the effects of diffusion and increase internal surface area of the films. The benefits of
freeze drying are modest (only increasing the capacitance per unit mass from 33 to 38F/g)
and when compared to the disadvantages like additional processing steps and electrode
thickness (which increases by two orders of magnitude) lead to a conclusion of this method
not being viable for this material system.
Optically transparent supercapacitor electrodes represent a solution to energy storage
for transparent electronics. PEDOT:PSS has a combination of good electrical properties at
high transparencies that allow for application as transparent conductors. In addition
PEDOT:PSS also has an appreciable capacitance which led to it being demonstrated as a
transparent supercapacitor electrode material. In an attempt to improve the electrical
properties of the film a 4 layer sheet of graphene was used as a current collector. The effect
on performance of the addition of the graphene was negligible at comparable transparencies
of PEDOT:PSS only electrodes. In order to provide a significant improvement to these
electrodes the conductive layer needs to be at least of the same sheet resistance as the
capacitive layer. While this will lead to a lower capacitance at low rates the performance at
higher rates could exceed the capacitance of the PEDOT;PSS only electrode.
3
Publications
Aspects of this thesis have been published previously in the following publications. Research
containing major contributions from this thesis is indicated with “ * “.
‘Graphene-MoS2 nanosheet composites as electrodes for dye sensitised solar cells’. Peter
Lynch, Umar Khan, Andrew Harvey, Iftikhar Ahmed and Jonathan N Coleman. Materials
Research Express, Volume 3, Number 3.*
‘Liquid exfoliation of solvent-stabilized few-layer black phosphorus for applications beyond
electronics’ Damien Hanlon et al. Nature Communications 6, Article number: 8563
4
Acknowledgements
First and foremost I would like to thank Jonathan Coleman for giving me an opportunity to
pursue my studies. Without his mentorship and support this thesis would not have been
possible.
Thanks to Paul and Tom who helped me with my final year project, providing me
with a taste of the good life as a research student. Paul put up with many a nervous mishap as
I progressed with my graduate studies and Tom was always around to answer my queries on
electrochemistry.
Umar Khan, who served as a bottomless pit of knowledge when it came to producing
dispersions and films and saved me weeks of lost time, is gone but not forgotten. The rest of
the group provided morale support, companionship and occasionally entertainment. The lads
in the office: Peter, Graeme, Conor, Seb and Adam were vital to my sanity. To the rest:
Ronan, Damien, Andrew, Dave, Sonia and JB without whom the labs would have been a
much duller place. Cheers lads for the proof reading and making this into something of
passable readability.
To those outside of Trinity who with or without intending to made these last four
years fly by: Sam, George, James, Eoin. And the housemates who kept me feed and help me
develop my culinary talents: Eoghan, Brian and Harry. My family have been supportive of
me without fail these last four years and never been shy of telling me.
Finally a big thank you is due to Declan Casey. His belief in me from a tender age and
his instruction in maths, physics and applied maths for the leaving cert set me on the path that
resulted in this work but does not end there.
5
Contents Thesis Outline ............................................................................................................................ 8
Dye-Sensitized Solar Cells ........................................................................................................ 9
2.1 Solar Energy......................................................................................................................... 9
2.2 Path to Liquid Junction Solar Cells ................................................................................... 10
2.3 Dye Sensitized Solar Cells (DSSC) – Principles of Operation .......................................... 12
2.4 Working Electrodes ........................................................................................................... 15
2.5 Dyes and Sensitizers ...................................................................................................... 16
2.6 Electrolytes ........................................................................................................................ 17
2.7 Counter Electrodes ........................................................................................................... 18
2.7.1 Platinum Electrodes .................................................................................................... 18
2.7.2 Carbon Electrodes ........................................................................................................ 19
2.7.3 Inorganic Electrodes ................................................................................................... 23
2.7.4 Composite and Hybrid Electrodes .............................................................................. 26
Electrochemical Capacitors ..................................................................................................... 31
3.1 Introduction ........................................................................................................................ 31
3.2 General Characteristics ...................................................................................................... 33
3.3 Non-Faradaic Electrodes ................................................................................................ 35
3.4 Faradaic Electrodes ........................................................................................................ 37
3.5 Electrolytes ........................................................................................................................ 40
3.6 Device Design .................................................................................................................... 41
Methods.................................................................................................................................... 46
4.1 Sample Preparation ............................................................................................................ 46
4.1.1 Liquid Phase Exfoliation by Ultrasonication .............................................................. 46
4.1.2 Stabilisation................................................................................................................. 48
4.1.3 Centrifugation ............................................................................................................. 49
4.1.4 Film Formation ........................................................................................................... 50
4.2 Sample Characterisation .................................................................................................... 53
4.2.1 UV-Vis Specrophotometry .......................................................................................... 53
4.2.2 Profilometry ................................................................................................................. 54
4.2.3 Electron Microscopy .................................................................................................... 54
6
4.2.4 Electrical Charaterisation ............................................................................................. 53
4.2.5 Electrochemcial Charaterisation .................................................................................. 58
4.2.5.1 Voltametry ........................................................................................................... 59
4.2.5.2 Impedance Spectrocopy ........................................................................................ 62
4.2.6 Solar Simulation.......................................................................................................... 63
Materials .................................................................................................................................. 65
5.1 Graphene ............................................................................................................................ 65
5.2 Carbon Nanotubes .............................................................................................................. 68
5.3 MoS2 .................................................................................................................................. 71
5.4 Poly(3,4-ethylenedioxythiophene) ..................................................................................... 72
Graphene Based DSSC Counter Electrodes ............................................................................ 77
6.1 Introduction ........................................................................................................................ 77
6.2 Experimental Procedure ..................................................................................................... 79
6.2.1 Materials ..................................................................................................................... 79
6.2.2 Film Production .......................................................................................................... 79
6.2.3 Electrochemical Charaterisation ................................................................................. 79
6.3 Results and Discussion ...................................................................................................... 82
6.3.1 Graphene Film Thickness Dependence ...................................................................... 82
6.3.2 Addition of Carbon Nanotubes ................................................................................... 88
6.3.3 Addition of MoS2 ........................................................................................................ 90
6.3.4 Graphene Flake Size Dependence .............................................................................. 97
6.4 Conclusions ...................................................................................................................... 100
Thickness Dependence of Capacitance of PEDOT:PSS Supercapacitor Electrodes ............. 101
7.1 Introduction ...................................................................................................................... 101
7.2 Experimental Procedure ................................................................................................... 102
7.2.1 Sample Preparation ................................................................................................... 102
7.2.2 Electrical Charaterisation .......................................................................................... 104
7.2.3 Electrochmical Charaterisation ................................................................................. 104
7.3 Results and Discussion .................................................................................................... 104
7.3.1 Electrical Properties ............................................................................................... 104
7.3.2 Cyclic Voltammetry of Thin Films ....................................................................... 106
7.3.3 Impedance Spectroscopy ....................................................................................... 118
7.3.4 Analysis of Diffusion............................................................................................. 121
7
7.3.5 Limitations due to Electrode Dimensions ............................................................. 124
7.3.5 Freeze Dreied Foams ............................................................................................. 126
7.4 Conclusions ...................................................................................................................... 130
Transparent PEDOT:PSS Supercapacitors with Graphene Current Collectors ..................... 132
8.1 Introduction ...................................................................................................................... 132
8.2 Experimental Procedure ................................................................................................... 133
8.2.1 PEDOT:PSS and PEDOT:PSS on Graphene Film Preparation............................. 133
8.2.2 Optical and Electrical Charaterisation ................................................................... 134
8.2.3 Electrochemical Charaterisation ............................................................................ 134
8.3 Results and Discussion .................................................................................................... 134
8.3.1 Optoelectronic Properties ...................................................................................... 134
8.3.2 Scan Rate Dependence of Capacitance ................................................................. 138
8.3.3 Length Dependence of Capacitance ...................................................................... 141
8.3.4 Impedance Spectroscopy ....................................................................................... 146
8.4 Conclusions ...................................................................................................................... 150
Conclusions and Future Work ............................................................................................... 150
9.1 Conclusions ...................................................................................................................... 150
9.2 Future Work ..................................................................................................................... 151
9.2.1 DSSC Counter Electrodes ..................................................................................... 151
9.2.2 Supercapacitor Electrode Materials ....................................................................... 153
Bibliography .......................................................................................................................... 156
8
1
Thesis Outline
The rise of technology especially personal devices has led to an increased demand for energy.
This leads to a search for alternative energy sources as fossil fuels are not renewable and with
the increasing demand for energy will only become more expensive as time passes. Solar
energy is an abundant energy source waiting to be tapped into on a larger scale. Chapter 2
will introduce the Dye-Sensitized Solar Cell (DSSC) and chapter 6 will explore the
replacement of an expensive component, platinum, with cheaper materials.
These materials are designed to have smaller dimensions to access unique properties
that become available when the particle’s size is reduced. Chapters 4 and 5 will introduce the
materials used in this thesis and how they are processed and tested respectively.
Storing this energy has also become an issue. As devices get more complicated with
more components and better, brighter screens the need to store energy in a compact manner
becomes more pressing. While conventional energy storage uses lithium ion batteries,
supercapacitors with fast charge-discharge times are being used as alternatives or in concert
with them. In this thesis Chapter 7 will look at thicker non-transparent charge storage
materials for more conventional uses. While Chapter 8 will investigate transparent charge
storage materials due to an increase in interest in transparent and flexible devices.
9
2
Dye-Sensitized Solar Cells
2.1 Solar Energy
Due to an increasing global population and the proliferation of technology, the demand for
energy has never been higher and will continue to increase. The decreasing availability of
fossil fuels, combined with the fact that in the process of releasing energy they introduce
carbon dioxide and other toxins into the atmosphere, has made alternative and renewable
energy sources more attractive in recent times.
In Figure 2.1 the graphic shows the relative sizes of the various energy sources, both
renewable and non-renewable, compared to the worldwide energy consumption in 2009 and
the projected energy consumption for 2050. It is clear that the sun is the most abundant
source of renewable energy. The sun provides approximately 1kWh of energy to the surface
of the earth per meter squared on average and capturing even a fraction of a percent of that
would go a long way to catering for mankind’s energy needs.
10
Figure 2.1 Relative sizes in TW of various renewable and non-renewable sources1
Solar energy today is a growing industry with many avenues for generating energy.
Solar water heating is a method of providing hot water and is particularly useful in hot
climates. Concentrating solar light to operate heat engines is another avenue to harvest the
sun’s energy. The main thrust of current research however is in the photocatalytic production
of fuel by water splitting and photovoltaics.
2.2 Path to Liquid Junction Solar Cells
The photovoltaic effect was first observed by Becquerel in 1839. The cell consisted of two
different metal plates separated by an ionic solution, known as an electrolyte. The discovery
of quantum mechanics and the subsequent application to condensed matter theory paved the
way for the first practical solid state solar cell which was announced by Bell Labs in 1954
with an efficiency of about 6%. Research into monocrystalline silicon solar cells has now
advanced to reaching an efficiency of up to 25%2.
11
The original liquid based solar cell took a back seat though a dye sensitized solar cell
had been reported in 1887 by James Moser and in 1964 the enhancement of
photoconductivity of zinc oxide by dye was demonstrated3. In the seventies, interest in
practical liquid junction solar cells began in earnest. These cells consisted of a highly
crystalline semiconductor with a suitable bandgap as a working electrode in an electrolyte
containing a redox couple and a catalyst on the counter electrode to replenish the electrolyte.
Figure 2.3 is a schematic of a liquid junction solar cell showing the various reaction
processes at the electrodes. Incident light is captured by the semiconductor and excites an
electron from the valence band to the conduction band to produce an electron-hole pair. The
hole then moves to the semiconductor-liquid interface where it under goes a reaction to
oxidize the electrolyte molecule as in equation 2.1.
Figure 2.2 Silicon Solar Cell
12
Figure 2.3 Energy Diagram and processes of a Liquid Junction Solar Cell4
Equation 2.1
The electron travels through the circuit and is reintroduced to the electrolyte through a
catalyst, usually platinum or graphite, in the opposite reaction to equation 2.1
Equation 2.2
These solar cells produced reasonably high efficiencies using single crystal semiconductors
of various Transition Metal Dichalcogenides (TMDs)5, Gallium Arsenide
6 and Cadmium
Selenide/Telluride7. The importance of the single crystal was that defects and grain
boundaries acted as electron traps causing recombination losses. Various methods from
pressing8 to chemical treatment
9 of these materials resulted in solar cells with comparable
efficiencies to the single crystal counter parts. However, the base cost of these materials and
the required high purity made them undesirable for commercial application.
2.3 Dye Sensitized Solar Cells (DSSC) – Principles of Operation
A dye sensitized solar cell is essentially a liquid junction solar cell with a different type of
working electrode. Currently, the prevalent working electrode consists of a mesoporous nano-
cystalline titanium dioxide electrode on the order of 10 microns thick covered in a light-
sensitive dye. While the concept had been around for a while in planar oxide electrodes,
13
Gratzel introduced the porous structure to increase dye loading per unit area to produce a
practical solar cell10
.
In Figure 2.4, the schematic of a DSSC is shown with the principle reactions. The dye
captures light in a similar way to the semiconductor in a liquid junction solar cell. In this
case, however, the electron is excited from the highest occupied molecular orbital (HOMO)
to the lowest unoccupied molecular orbital (LUMO) of the dye as in equation 2.3.
Equation 2.3
Where S represents the dye molecule, h is planks constant, ν is frequency and the asterisk
indicates an excited molecule. This is represented by the green line in Figure 4(b). The
electron is then quickly transferred from the excited dye molecule to the titanium dioxide
structure and passed through the circuit. This leaves a positively charged dye molecule in
contact with the electrolyte, usually a tri-iodide couple (I-/I3
-). An ion in the electrolyte loses
an electron to the dye to restore it to the ground state by a reaction similar to equation 2.1.
Equation 2.4
Figure 2.4(a) Schematic of DSSC (b) Energy Level Diagram with electron transfer
processes109.
14
The electron initially produced travels through the circuit and undergoes a similar reaction at
the catalytic counter electrode as in equation 2.2. The transfer of the electrons is indicated by
blue lines in figure 4 (b) and red lines indicate pathways of reverse reactions which result in a
reduction in efficiency analogous to recombination in semiconductor solar cells.
Solar cell efficiency is defined by the short circuit current (Jsc), the open circuit voltage (Voc),
a measure of ideallity known as the fill factor (FF) and the input power (Pin).Efficiency is
given by the following equation:
in
ocsc
P
FFVJ Equation 2.5
To maximize the efficiency of a cell all parameters for a given Pin should be optimized. The
short circuit current density is the current at which there is no applied potential. This is
determined by the incident photon-to-electron conversion efficiency as well as the
recombinations present in the system. The open circuit voltage is the voltage at which no
current flows in the cell. The difference between the Fermi level of the semiconductor and the
redox potential of the electrolyte limits the maximum open circuit potential. The fill factor is
the ratio between the maximum power produced by the cell and the product of the open
circuit voltage and short circuit current.
Currently, the state-of-the-art liquid dye-sensitized solar cell performs with an
efficiency of 12.3%, produced by Yella et al11
. An ideal solid state solar cell has a theoretical
efficiency in excess of 30% however Henry Snaith accounts for losses that occur in the dye-
sensitized solar cell and predicts a maximum theoretical efficiency of 13.4% using the
technologies as of 2010 but describes a pathway to optimizing the maximum efficiency to
20.25%12
. The losses are primarily associated with processes in the working electrode, dye
and electrolyte these include (a) incomplete light harvesting, which can be tackled with
working electrode geometries and dyes with suitable absorption onsets, (b) inefficient photo
induced electron transfer to dye as well as the energetic favourability of such transfers, this
can be tackled by an improvement in the matching of energy levels in the working electrode
oxide and dye, (c) conformational charges in dye due to generation of excitons, (d) and most
importantly the dye regeneration process which is the most energy intensive, due to the over
potential to oxidize the electrolyte and to conversion of the electrolyte species in solution.
15
This can be tackled with alternative electrolyte/dye couples or even replacing the dye with a
hole conductor. (e) Resistive losses throughout the cell also have an impact. The focus of
work in this thesis will be regarding the counter electrode and the resistive losses associated
with the charge transfer process at that electrode.
2.4 Working Electrodes
The design of the working electrode in Grätzel’s seminal paper was 10 microns thick
consisting of 15nm TiO2 particles which were sintered at 450 degrees Celsius to improve the
electrical transport properties of the film. The factors most important in the design of the
working electrode are high surface area for maximum dye adsorption and sufficient porosity
to allow diffusion of the electrolyte.
Since then a significant amount of work has been done to optimise the efficiency of
the cells by adjusting the parameters of the working electrode. For example, increasing
thickness should increase the amount of dye available for light harvesting. However, the
further through the film the light travels, the more light will have already been absorbed by
the previous dye particles. This would lead to a sub-linear relationship with thickness.
Furthermore, optimisation via thickness becomes problematic when one accounts for
transport issues associated with thick films. Optimum thicknesses have been found in the
range of 7 to 14 microns13
.
Particle size is another route towards altering the properties of a working electrode.
Smaller particles have a higher specific surface area allowing for increased loading of dye in
an electrode of similar thickness. A disadvantage of small particles however is the effect on
the electrical properties due to an increased number of inter-particle junctions and as such
must be balanced with the dye loading to produce optimum working electrodes. Work by
Chou et al. demonstrate this effect and show excellent efficiency at 22.5nm14
. This is close to
the commercial 25nm Degussa |TiO2 particles used in multiple studies15–18
. Another aspect of
particle size is scattering effects. Larger particles have found applications as a scattering layer
on top of a transparent layer of smaller particles to improve overall efficiency of the TiO2
working electrodes by passing the light back through the film for further absorption13,19,20
.
16
There are some losses associated with contact between the current collector and the
electrolyte. To prevent this, a thin layer of TiO2, known as a blocking layer, is applied by
various means, the most common being TiCl4 treatment13,18,21
. An added advantage of the
blocking layer is that it improves adhesion of the TiO2 particles to the substrate. As well as
pre-treating the substrate to form a blocking layer, a secondary treatment is carried out after
the film is deposited to improve particle connectivity and surface roughness.
Alternative allotropes of TiO2 have also been considered. One-dimensional nanotubes
have been employed due to the advantage of the path to the current collector being through
one continuous piece of TiO2 as opposed to a series of sintered junctions22,23
. Two-
dimensional titania nanosheets have been utilised also due to the excellent specific surface
area of two-dimensional nanomaterials allowing excellent loading of dye15,17
. They also
demonstrate preferable light scattering properties.
Charge transport improvement has also been investigated through the addition of
conductive particles, especially carbon nanotubes18
and graphene24,25
. These particles provide
a more direct route to the current collector than travelling through the porous TiO2 structure
in addition to having better electrical properties. Ideally, if the sintering process could be
removed due to enhanced conductivity, it would have a considerable impact on the cost
associated with the production of the working electrode.
In addition to TiO2, there are alternative working electrodes. Since TiO2 remains the
state of the art and is the most commonly used working electrode material, the other materials
will only get a brief mention. The n-type semiconductors used are TiO2, ZnO26–28
, SnO228–30
,
Al2O330
and Nb2O531,32
. In addition, there are p-type semiconductors such as NiO33
but these
have more limited applications due to low open circuit voltages.
2.5 Dyes and Sensitizers
The function of the dye in the DSSC is to capture light. The dye is required because the
bandgap of TiO2 is too large so the material does not absorb visible light. The dye initially
used was a ruthenium compound which gave a reddish brown colour. Since then the most
popular dyes that are commercially available are the ruthenium based N3, N719, and ‘black’
dyes shown in figure 2.5.
17
Figure 2.5 molecular structures of common dyes34
The dye needs to satisfy a number of criteria. Firstly, it needs to have an appropriately spaced
HOMO and LUMO to absorb visible light and have a high absorbance coefficient in that
range. It must also adhere suitably to the semiconductor via covalent bonds with the
anchoring groups and be stable in the electrolyte under illumination. The HOMO and LUMO
of the dye molecule must also match the conduction band of the TiO2 to promote efficient
electron transfer while preventing reverse reactions. The same molecular orbital matching is
important with the electrolyte to ensure efficient replenishment of electrons and as such
electrolyte and dye research is sometimes performed in tandem.
Alternative sensitizers are quantum dots and enough research has been performed in
this field to coin the acronym QDSSC or quantum dot sensitized solar cell. In these cells, the
quantum dot replace the dyes. Quantum confinement can alter the optical properties of
particles to produce efficient absorbers in the visible range. CdS, CdSe and PbS have been
popular materials for introduction as quantum dot sensitizers35
.
2.6 Electrolytes
The electrolyte carries the charge transport from the catalytic counter electrode to the
working electrode. It consists of a redox couple in a suitable solvent. The first Grätzel cell
used the iodide/tri-iodide couple and as such it is the most commonly researched couple and
is available commercially.
To ensure high efficiencies, there are numerous criteria for the electrolyte to fulfil.
The redox potential of the redox couple should be negative relative to the oxidation potential
18
of the dye and efficiently regenerate the dye. A high conductivity in the order of 10-3
S/cm is
desirable to ensure good electrical transport between the electrodes. The electrolyte should
not have any adverse reactions with either the dye or the sealant to prevent undesirable losses
due to the electrolyte absorbing light in the visible before it interacts with the dye. Most
importantly, it should be highly stable so as not to lose functionality over time and nor should
it degrade at temperatures below 80⁰C34.
Liquid electrolytes contain an organic solvent, redox couple and occasionally
additives to enhance performance. Currently, the best performing solvent for the electrolyte
has proven to be a mixture of acetonitrile and valeronitrile. As mentioned previously, the
most popular redox couple is the iodide/tri-iodide couple. This is achieved using dissolving
iodine(I2) and a range of iodide salts in the solvent. However others include bromine redox
couples but the state of the art for liquid based DSSCs is with a Cobalt(II/III) couple11
.
Ionic liquids are a way of combining the solvent and redox couple as one component.
The ionic liquids display low vapour pressure leading to better stability when compared to
other electrolytes. Ionic liquids display high viscosities which limits diffusion-assisted
transport however an alternative exchange mechanism allows electron hopping between the
iodide and tri-iodide ions34,36
.
Due to the difficulty working with liquid electrolytes, there has been interest in
developing solid state equivalents of the DSSC. One approach is using additives to gel the
electrolyte. These gelators form a three dimensional matrix in which the liquid electrolyte can
still work as a mobile phase but have a reduced volatility34,37
. Another solid state method
involves using p-type materials to act as the electrolyte and remove the need for a catalytic
counter electrode. This requires a material with a valence band well matched with the HOMO
of the dye molecule. Conducting polymers have been the most promising solid state
electrolytes, with Spiro-OMeTAD showing the highest performance34,38
. Limitations of solid
state electrolytes include poor filling of the working electrode and low carrier mobility.
2.7 Counter Electrodes
The role of the counter electrode is to reintroduce the electrons extracted from the dye back
into the electrolyte after passing through the external circuit. The counter electrode is made of
19
a catalytic material which, since the introduction of the DSSC, has been predominantly
platinum. The reaction carried out is a variation on equation 2.2 which for the case of the
iodide/tri-iodide couple is
Equation 2.6
2.7.1 Platinum Electrodes
Platinum has been used in many electro-catalytic systems including the hydrogen and oxygen
evolution reactions and even the initial liquid junction solar cells. Platinum has been chosen
as a catalyst due to its chemical stability and excellent electrical and thermal conductivity.
Sputtered platinum on FTO has found application as the counter electrode for DSSCs
achieving excellent properties at a thickness range from 0.2 – 2 micron39
. This is not only due
to the electrical and catalytic properties of the platinum but also the reflectance of the film
allowing more light to be captured by the working electrode.
To lower the cost, thin film platinum must be considered. In addition thin films allow
transparency allowing illumination through the counter electrode or application of a reflective
material. These low thicknesses require mass per unit area of platinum in the range of 10-
100µg/cm2. This can be achieved by thermal deposition of chloroplatinic acid which is very
common. A deposition time dependence for sputtering of platinum by Mukherjee et al.
resulted in best performance at 50 nm Pt which was approximately 100µg/cm2 however
reasonable performance at a mass loading of 0.1 µg/cm2 resulted in approximately 78% that
of a film 1000 times the mass40
.
To further enhance platinum electrodes in thin films, exploration of the various
properties of nanostructured platinum was targeted. As mentioned previously with the
working electrodes, smaller particles lead to higher specific surface areas essential for
increasing the number of active sites available for reaction. Other nanostructures have also
been investigated to increase surface area of platinum in thin films41
.
Platinum, however, is a scarce metal and could become prohibitively expensive. Also
sputtering requires specialist equipment with high energy requirements and choloroplatinic
acid is not good for the nvironment. On top of this there some debate as to the stability of
platinum in the electrolyte42
. As such, much research has been undertaken to identify
alternatives.
20
2.7.2 Carbon Electrodes
An obvious option for the replacement of platinum is carbon, which is one of the most
abundant elements on the surface of the Earth. Carbon has been well studied and there are
many forms ready for application as counter electrodes in DSSCs: carbon black, activated
carbon, mesoporous carbon, graphite/ene and carbon nanotubes.
Carbon black is produced by the incomplete combustion of heavy hydrocarbons. It
has a favourable surface area-to-volume ratio which is desirable for catalytic activity.
Murakami et al. developed thick, carbon black-based counter electrodes in excess of 20
microns to achieve an efficiency of 9.1% which, while very high, did not include a platinum
based-cell as reference43
. However, Lin et al. produced thin, transparent, carbon black-based
DSSCs with an efficiency of 7.28% compared with the platinum based DSSCs of 7.12% with
the requirement of a high temperature anneal44
.
Mesoporous carbon is a synthesized carbon material with well-defined mesoscopic (2-
50nm) pores and high surface area making it a good candidate for counter electrode
materials. While higher surface areas should be desirable and can be achieved using smaller
pore diameters, the ability of the molecules of the electrolyte to diffuse into the pores should
be considered. Work by Ramasamay et al. demonstrates this where a high surface area (1400
m2/g)/low pore diameter (3 nm) counter electrode results in a lower efficiency of 6.75%
while a lower surface area (894 m2/g)/higher pore diameter (22 nm) results in a efficiency of
8.18% comparable to that of the platinum counter electrode with an efficiency of 8.85%45
.
Carbon can be found in the Earth’s crust in the form of graphite, a layered conductive
material. Much work has been done on mechanically and chemically exfoliated layers of
graphite to produce materials for use as counter electrodes due to the high conductivity and
surface area of such materials. Pristine graphite/ene displays electrochemical catalytic
activity at edge sites while the basal planes tend to be significantly more inert46–48
. Reports
of mechanically exfoliated graphite as counter electrodes show this to be a promising method
for producing viable solar cells. Veerappan et al. produced thick electrodes (6-9 microns)
resulting in a cell with an efficiency of 6.2%, comparable to that of their platinum cell at
6.8%49
. Kavan et al. improved on this by producing a thin, transparent graphene-based
21
counter electrode with an efficiency of 5% compared to the platinum cell of 6.89%. Thinner
films are preferable as they require less material, thus reducing the cost50,51
.
While graphite is useful as a catalytic material for DSSCs, functionalised graphene is
even more suitable. Graphite oxide is commonly prepared by Hummer’s method and can be
exfoliated to produce graphene oxide which is stable in water52,53
. This introduces oxide
functional groups into the basal plane of the graphene which are catalytically active. The
presence of these oxide functional groups also reduces the conductivity and thus reduction of
the graphene oxide is commonly performed. Once reduced, either thermally or chemically,
the graphene retains some functional groups while exhibiting suitable conductivities54,55
.
Graphene oxide has been integrated into DSSCs much more often than pristine graphene due
to the ease at which it can be dispersed in water.
Wan et al. produced thin films (~100nm) of chemically reduced graphene oxide to
operate as counter electrodes for DSSCs. These films, however, did not perform favourably
when compared to platinum demonstrating an efficiency of 0.74% compared to 3.7%56
. The
use of thicker films as done by Zhang et al., with a combination of chemical reduction and
thermal annealing, produced an efficiency of 6.81%, relatively close to that of the platinum at
7.59%57
. Further, work by Roy-Mayhew et al. produced equivalent performance with 6
microns of functionalized graphene with platinum (6.8%) in the tri-iodide couple and
superior performance in the cobalt and sulphur based-couples58
.
Substituting the oxide functional groups of graphene oxide with nitrogen functional
groups, as done by Hou et al., resulted in a higher efficiency of the N-doped reduced
graphene oxide (5.4%). This was greater than both the efficiencies of the undoped reduced
graphene oxide film (4.0%) and the platinum counter electrode (5.1%)59
.
While exfoliation is a common way of producing graphene in liquid phase, it can also
be chemically synthesised as demonstrated by Wang et al. It includes oxygen functionalities
and a three dimensional structure of the 20 micron film increases the surface area available
for reactions. This resulted in a performance efficiency of 7.8% with no comparison to
platinum60
.
Outside of the liquid phase, there have been reports of chemical vapour deposition
(CVD) produced graphene being used in DSSCs. Seo et al. use CVD to produce ‘graphene-
like’ films of thickness 500-600nm as a counter electrode with efficiency of 4.3% compared
22
to 5.9% for a platinum counter electrode61
. Pan et al. use CVD to produce vertically
orientated graphene with the advantage of having edges exposed to electrolyte and good
charge transport to the current collector at 30 micron thickness. This yields an efficiency of
7.63% relative to 8.48% for a platinum based cell62
.
Carbon nanotubes (CNTs) are another allotrope of carbon, although unlike graphite, it
does not occur abundantly in nature and as such needs to be synthesised. The synthesized
tubes can be dispersed in a liquid or paste for deposition. Ramasay et al. produced transparent
CNT films of various thicknesses by spray-coating the CNTs dispersed in ethanol. This
resulted in efficiencies of up to 7.59% with no platinum for comparison63
. Screen printed
electrodes produced by Lee et al resulted in a cell efficiency of 7.67% close to their 7.83%
platinum reference64
.
However, the electrodes produced by depositing CNTs in a dispersion or paste are
randomly aligned and require an additional processing step after synthesis. Nam et al.
produced CVD grown-CNTs directly on an FTO current collector. For comparison, a screen
printed CNT electrode of randomly aligned tubes with a similar thickness (500nm) was
fabricated. The directly synthesised electrode had an efficiency of 10.04% which was greater
than that of the screen printed electrode (8.03%) and even the platinum reference (8.80%)65
.
Doping of CNTs is another viable route to enhancing electro-catalytic activity. Lee et
al. doped vertically aligned CNTs with nitrogen. The nitrogen doped-CNTs (of length 20
microns) provided superior catalytic activity evident in the better current density
characteristics resulting in an efficiency of 7.04% compared to that of a platinum cell of
7.34%66
.
Outside basic carbon materials are a class of material called conductive polymers.
One such material is PEDOT, commonly produced as PEDOT:PSS for stability in solution.
Pringle et al. use electrodeposition of PEDOT directly onto FTO as a counter electrode.
Interestingly, no thickness dependence was observed in the range of thicknesses tested (0.03-
2 microns). This is useful for minimising required material to produce effective counter
electrodes. The PEDOT counter electrode resulted in cells with efficiency of 8%, in excess of
the platinum based cell of 7.9%67
.
Variations on PEDOT namely ProDOT and ProDOT-Et2 were compared by Lee et al.
The platinum reference cell used had an efficiency of 7.77%. Electrodepositing each polymer
23
resulted in films with varying degrees of surface roughness leading to better surface area with
increased roughness. PEDOT had the lowest surface area and demonstrated an efficiency of
3.93%. ProDOT and ProDOT-Et2 had greater surface areas resulting in efficiencies of 7.08%
and 7.88% respectively68
.
Polypyrrole is another conductive polymer utilized as a counter electrode in DSSCs.
Jeon et al. produced polypyrrole nanoparticles and deposited these on FTO as a counter
electrode. The initial electrodes, with no further treatment, gave efficiencies of 5.28% and
treatment in HCl vapour improved this to 6.83%. Further adjustments to the electrolyte
allowed the best possible efficiency of 7.73% with respect to an 8.2% efficiency platinum
reference cell69
.
Polyaniline (PANI) was used to create a partially transparent counter electrode by Tai
et al. via a dip-coating method to produce the PANI films on FTO. The transparency enabled
them to test the counter electrode by both front (photoanode) and back (counter electrode)
illumination. The back illumination had a relatively high efficiency of 4.26% when compared
to that of front illumination (6.54%) which itself nearly matched that of the platinum cell
(6.69%)70
.
2.7.3 Inorganic Electrodes
Transition metal compounds make up the bulk of research into alternative materials for
counter electrodes in DSSCs due to the wide variety in possible compounds. Extensive
research has been undertaken to identify the practicality of these materials. Besides the
multitude of transition metals, there are secondary elements in the compounds to alter the
properties of the material. As such there has been work done on carbides, nitrides, oxides and
chalcogenides.
Molybdenum (Mo2C) and tungsten carbides (WC) were assessed by Wu et al. as
alternatives to platinum. Purchased powders of particles less than 300nm were sprayed onto a
conductive FTO (Flourine doped Tin Oxide) substrate with sheet resistance of 15Ω/sq.
Compared to a platinum cell with 7.89% efficiency, these materials produced efficiencies of
5.35% (WC) and 5.7% (Mo2C)71
. Jang et al. assessed polymer assisted- and microwave
24
assisted-formation of tungsten carbide nanoparticles with diameters of 200 and 30nm
respectively. The smaller particles exhibited higher surface areas and a corresponding higher
efficiency of 7.01%, approximately 85% that of the platinum reference72
.
For nitrides, Li et al. used the nitridation of metal oxide precursors to produce
molybdenum nitride (MoN), tungsten nitride (WN), and iron nitride (Fe2N). While there were
differences in the morphologies of the particles, the thicknesses of each electrode was 13
microns. The efficiencies of the cells were 5.67% (MoN), 3.67% (WN) and 2.65% (Fe2N), as
such the closest to platinum (6.56%) was the MoN electrode73
. Vandium Nitride was
investigated by Wu et al. A variety of morphologies were fabricated by modifying the
synthesis parameters revealing an increase of efficiency with surface area. This resulted in an
efficiency of 7.29% compared to 7.68% for platinum in the tri-iodide couple while exhibiting
better performances in a thiolate/disulphide electrolyte74
.
To improve access of the electrolyte to the material, Song et al. compared MoN
nanorods and particles. The one-dimensional rods resulted in a more porous electrode
allowing for better diffusion of the electrolyte which was observed as a decrease in diffusion
resistance in the EIS spectra. The MoN particles displayed an efficiency of 6.48%. The
enhanced diffusion properties resulted in an efficiency of 7.29% for the MoN nanorods
compared to 7.42% efficiency for the platinum reference75
. Titanium Nitride (TiN) nanotube
arrays on a titanium foil current collector have also been demonstrated as a counter electrode
by Jiang et al. The electrode exceeded that of the platinum reference cell (7.45%) with an
efficiency of 7.73%76
.
Transition Metal Oxides have also been demonstrated as potential catalysts for
DSSCs. Various niobium oxides were synthesised and tested by Lin et al. Three crystalline
structures of Nb2O5 and one of NbO2 were analysed. The most promising candidate was the
NbO2 displaying an efficiency of 7.88% greater than the 7.65% demonstrated by the platinum
based-cell77
. Ruthenium Oxide nanocrystals synthesized by a hydrothermal method by Hou et
al. also displayed an efficiency (7.22%), greater than that of a platinum cell (7.17%)78
. A
broad study by Hou et al. based on DFT calculations of surface energies identified certain
materials within an ideal range of adsorption energies. Iron Oxide (α-Fe2O3) was one of the
candidates that displayed good values and as such was synthesised and tested. The counter
electrode material was screen printed and produced an efficiency of 6.96% comparable to the
7.32% efficiency of platinum79
.
25
One dimensional tungsten oxide (WO2) nanorods of diameter 20nm were synthesised
by Wu et al. and compared to WO3 particles ranging in size from 0.05 to 2 microns. In
addition to the higher surface area of the nanorods, the redox peaks as observed by cyclic
voltammetry of the nanorods were more similar to platinum than the particles indicating
more desirable electrochemical properties. As such, the nanorods outperformed the particles
with an efficiency of 7.25% to 4.67% compared with a platinum counter electrode with
efficiency of 7.57%80
.
Wu et al. produced a study of a wide variety of transition metal carbides, nitrides and
oxides of the transition metals titanium, zirconium, vanadium, niobium, chromium and
molybdenum. All materials were synthesised by a reaction of urea, an oxygen source, carbon
and nitrogen, and a metal chloride precursor. The efficiencies were compiled in a graph
(Figure 2.6) by the author. According to this study, vanadium based-compounds look to be
the best candidates for counter electrode materials81
.
Figure 2.6 Power Conversion efficiencies for various counter electrode materials.81
Transition metal chalcogendides (the chalcogen being sulphur, selenium and tellurium
in these cases) have found applications as counter electrode materials for DSSCs. Cobalt
sulphide, synthesised to have a honeycomb structure to maximise surface area, produced by
Lin et al. resulted in an efficiency of 6.01%, in excess of the 5.71% efficiency of the platinum
reference82
. Gong et al. produced micron size nickel selenide particles for use as counter
electrode material which also outperformed platinum with an efficiency of 8.69% to 8.04%83
.
26
A variety of molybdenum, tungsten and tantalum diselenide were synthesised and
tested by Guo et al. MoSe2 had the highest surface area and yielded an efficiency of 6.71%
while WSe2 had a marginally lower surface area but a superior efficiency of 7.48%.The
platinum reference in this work had an efficiency of 7.91%84
.
Further into the chalcogenides, Guo et al. also studied telluride-based, transition metal
compounds. In this case, cobalt and nickel telluride exhibited efficiencies of 6.92% and
7.21% respectively while the platinum cell reported an efficiency of 7.04%.
Many of these materials, especially the transition metal dichalcogenides (TMDs),
have 2 dimensional configurations leading to high theoretical surface areas. This should make
these materials ideal candidates for counter electrode materials. Cobalt sulphide synthesised
in 2 dimensional nanosheets by Tai et al. outperformed platinum by 6.39% to 6.06%. In
addition, the material was also highly transparent over the visible spectrum70
. MoS2 and WS2
films were tested as counter electrode materials by Wu et al. The performance of the
materials was similar with MoS2 producing an efficiency of 7.59% and WS2 producing an
efficiency of 7.73%, which was approximately that of the platinum cell with 7.64%
efficiency85
.
Tin sulphide (SnS) and tin disulphide (SnS2) nanosheets were the subject of a study
by Chen et al. The SnS composition proved to be the better of the two with an efficiency of
6.56% to 5.14%. One dimensional SnS nanowires were also tested and achieved a lower
efficiency of 5.00% providing evidence that the 2 dimensional configuration was superior86
.
This however is not always the case, a study by Lei et al. on monolayered, few-layered and
nanoparticulate MoS2 revealed that the nanoparticles exhibited superior performance under
illumination with the efficiency of the nanoparticles being 5.41% and the monolayered flakes
being 2.92%. This was explained by an increase in the diffusion impedance suggestive of
possible reaggregation preventing sufficient access of the electrolyte to the catalytic sites87
.
2.7.4 Composite and Hybrid Electrodes
The counter electrode materials discussed so far have been largely of a singular material.
However, there are strategies to enhance electro-catalytic activity by forming hybrids and
composites of two materials. For the purposes of this work, hybrids refers to materials
27
synthesised simultaneously, that form single structure and composites refers to the mixture of
already synthesised or produced materials. In these cases, one material usually provides
electro-catalytic activity while the other provides at least one other property such as
conductivity or porosity to the film.
Due to the diverse properties of the various allotropes and structures of carbon,
studies have been conducted into the possible synergies between alternate carbon materials.
Activated carbon has been demonstrated as a catalytic material for DSSCs due to its high
surface area. Graphene, while working as a catalyst at edge sites, has greater conductivity
enabling better charge transport to the current collector. Wu et al. produced activated
carbon/graphene composites using electrophoretic deposition on FTO. The cell using
activated carbon displayed an efficiency of only 6.66%. The activated carbon/graphene
composite formed by electrophoretic deposition achieved vertically aligned sheets in the
electrode resulted in an efficiency of 7.50%. However, an electrode formed by spin-coating
resulted in more horizontally orientated graphene sheets which led to a reduced efficiency of
5.99% due to less direct conductive paths and increased difficulty of diffusion88
.
Reduced graphene oxide (rGO) has been shown to be a sufficiently catalytic material.
However, due to the tendency of planar sheets to stack in solid films reducing surface area
and the anisotropy of conductivity, significantly lower in the out-of-plane direction, a
material that could increase porosity and conductivity would be an advantage. Carbon
nanotubes (CNTs) provide such an opportunity. Zhu et al. used electrophoretic deposition to
produce such counter electrodes containing various ratios of material. The best observed
efficiency (6.17%) was observed at 60%wt CNTs compared to a platinum cell with an
efficiency of 7.88%. Zheng et al. produced similar counter electrodes by gel coating and
achieved a higher efficiency than platinum (7.79%) with 20%wt CNTs (8.37%)89
. Battumur
et al. combined pristine graphene nanosheets with multi-walled carbon nanotubes to improve
the efficiency of the graphene-only cell achieving 4% compared to 5% for the platinum
based-cell90
.
Conductive polymers have also been used as both a conductive additive and a
provider of electro-catalytic sites in carbon material-based counter electrodes. Carbon
nanotubes were used to provide high surface areas and enhance the conductivity of PEDOT
with the hybrid material producing an efficiency of 4.62%, higher than that of either material
individually91
. Graphene was used to enhance the conductivity of a transparent PEDOT:PSS
28
film increasing the efficiency from 2.3% to 4.5% compared to the platinum reference
(6.3%)92
. PEDOT was also deposited on a pressed graphite substrate by Nagarajan et al. This
was compared to PEDOT on an FTO substrate to reveal that the effect of the graphite was
also providing catalytic activity on top of the role of a current collector. The efficiency of the
composite electrode was 5.78%, in excess of both the PEDOT on FTO counter electrode and
the platinum reference electrode93
. The performance of a polypyrrole counter electrode was
enhanced by the presence of 10%wt graphene quantum dots resulting in an efficiency of
5.27% compared to the platinum counter electrode of 6.02%94
.
A wide range of inorganic materials have demonstrated catalytic activity in DSSCs
but many of these particles are either semiconducting or insulating. Particles not in direct
contact with the current collector are less efficient due to poor charge transport. Conductive
additives, such as carbon nanotubes, graphene and conductive polymers, have the potential to
improve such counter electrodes.
Conductive carbon paste was used to provide conductivity to a wide range of
materials by Gao et al. The most promising of these is cadmium, in which the composite
counter electrode gave an efficiency of 6.71%, comparable with the platinum cell with an
efficiency of 7.06%95
. Wu et al. produced a range of inorganic materials as mentioned
previously. The most promising material, vanadium carbide, was added to mesoporous
carbon further improving the electrical and surface properties of the counter electrode81
.
MoS2 was directly synthesised on reactive sites of functionalised carbon nanotubes by
Yue et al. to improve upon the surface area and conductivity of a MoS2 counter electrode
(with initial efficiency of 5.42%). The hybrid counter electrode exhibited an efficiency of
7.92% greater than that of the platinum cell of 7.11%96
. MoN was produced on carbon
nanotubes using a similar method by Song et al. resulting in an efficiency of 6.74%
comparable to the platinum counter electrode (7.35%)97
.
Guo et al. used CNTs as a porous foundation for the growth of platinum
nanoparticles. The purpose of this counter electrode is not to replace platinum but to improve
upon the standard pyrolysis platinum counter electrode. The hybrid counter electrode resulted
in an efficiency of 7.69%, greater than that of the platinum only-electrode with an efficiency
of 6.31%98
.
29
The bulk of the graphene hybrids are created by a similar method to the CNT hybrids
using functional group sites to promote anchoring for the synthesised inorganic material. As
such, most of the work is done using graphene oxide and reduced graphene oxide. This has
also been done with MoS299,100
, Nickel Phosphide101
, Nickel Oxide102
and NiS2103
. The graph
below (Figure 2.7) illustrates the various efficiencies of the counter electrodes compared to
their platinum counterparts.
4 6 8
3
4
5
6
7
8
9
MoS2
NiP
NiO
NiS2% E
ffic
ien
cy C
E (
%)
% Efficiency Platinum (%)
Figure 2.7 Efficiency of various counter electrode materials synthetically grown on graphene compared to the platinum reference. Black line represents equivalent efficiency with
platinum.
Yue et al. used commercial MoS2 and graphene flakes to create a composite which
was printed with acetylene black and PVDF to form a counter electrode. This produced a
performance of 5.98% close to that of the platinum based cell of 6.23%104
. Cobalt sulfide,
previously mentioned as a good catalytic material, was also incorporated into CVD grown-
graphene on FTO by Das et al. The graphene on FTO alone had an efficiency of 1.27% but
the addition of the CoS increased the efficiency by more than a factor of 2 to 3.42%105
.
Platinum nanoparticles were deposited on graphene films by pulsed laser deposition
by Bajpai et al. The effect of the graphene increased the efficiency of the platinum compared
to platinum deposited onto FTO directly by the same method with the composite counter
electrode displaying an efficiency of 2.91% compared to 2.11% for the platinum alone106
.
30
Sudhajar et al. combined cobalt sulphide with PEDOT:PSS by dispersing CoS
nanoparticles in an aqueous PEDOT:PSS dispersion before spin-coating. This resulted in a
counter electrode with an efficiency of 5.4% compared to 6.1% for the platinum counter
electrode107
.
Yeh et al. produced a titanium nitride/PEDOT:PSS composite film by a similar
method on titanium foil. Various weight percentages of TiN nanoparticles were tested with
the 20%wt composite providing the best performance of 6.67%, slightly higher than the
efficiency of the sputtered platinum cell at 6.57%108
.
Polypyrrole was used in a metal-PPy-carbon composite by Liu et al. Cobalt, iron and
nickel were entrapped in the polypyrrole matrix and each showed improvement in the
catalytic effect of the counter electrode. Cobalt proved to be the most suitable metal for this
purpose with an efficiency of 7.64%109
.
Many of the above methods involve counter electrodes of high thickness, high
temperature processing or complicated methods for synthesis and deposition. Platinum
counter electrodes are usually produced in the thickness range of 20-100nm, via sputtering or
pyrolysis, involving temperatures in excess of 400⁰C and come from an expensive raw
material. The ideal replacement requires an up-scalable method of production combined with
reduced cost. The simplest way to achieve this is by reducing the amount of material
necessary and lowering processing temperatures. This work will resolve to provide some
manner of approaching these criteria.
31
3
Electrochemical Capacitors
3.1 Introduction
The generation of electricity from renewable sources can be infrequent and so to provide for
the constant demand for electrical energy by society, the excess must be stored for later use.
On top of this, portable devices have become essential to modern life and require appropriate
sources of both power and energy.
The charging and discharging of electrical energy for a range of processes have
different power requirements. Currently there are four devices for storing energy each with
different power and energy storage characteristics. These devices are the fuel cell, the battery,
the electrochemical capacitor, and the dielectric capacitor. A Ragone plot as shown in fig. 3.1
shows the energy and power per unit weight of some of these devices.
32
Figure 3.18 Ragone Plot including various devices including capacitors, electrochemical capacitors and a range of batteries110.
The electrochemical capacitor also known as the supercapacitor or ultracapacitor is
the main focus of this thesis. The electrochemical capacitor uses a similar method of energy
storage to the dielectric capacitor by charging surfaces but in construction it bares more
resemblance to a battery. This is due to the presence of two electrodes in an electrolyte. As
such the electrochemical capacitor occupies the space in the Ragone chart between these two
devices. With energy storage capability in the range of 1-10 Wh/kg and power storage in the
range of 500-10000 W/kg110–113
. In addition to the high power capability of electrochemical
capacitors they can be made of environmentally benign materials. The lifetime of
electrochemical capacitors in cycles lies between that of dielectric capacitor and that of a
battery which is in the range of thousands.
33
3.2 General Characteristics
A typical dielectric capacitor has two parallel plates separated by a dielectric. The
capacitance (C) is a property of the material system and given by the formula 3.1:
Equation 3.1
Where ε is the permittivity of the material between the electrodes, A is the common area of
the electrodes, d is the distance between the electrodes, Q is the charge stored on the
electrodes and ΔV is the potential difference between the electrodes. In a dielectric capacitor
the area corresponds to the geometrical area of the plates and the distance is on the order of
microns.
The electrochemical capacitor however uses a liquid-solid interface the solid being the
electrode material and the liquid being an electrolyte instead of a metal-dielectric interface.
This phenomenon was first observed by Helmholtz in 1853. Since then the field of
Figure 3.2 Structure of the Helmholtz Double Layer model114.
34
electrochemistry has gained understanding of this interface and the structure of this interface
is shown in fig. 3.2114
.
At the surface of the solid there exist molecules of the solvent and possibly some
specifically adsorbed ions. This is considered the inner Helmholtz plane. After the inner
Helmholtz plane solvated ions form the outer Helmholtz plane, this is the distance of closest
approach for the solvated ions and is approximately equivalent to twice the diameter of the
solvent molecule. Due to thermodynamic agitation these ions extend into the liquid which is
known as the diffuse layer. The charge density (σ) due to these ions is described as a sum of
both contributions of the ions as in equation 3.2:
Equation 3.2
Where the subscripts S, i and d stand for solution, inner Helmholtz plane and diffuse layer
respectively.
The nature of this interface affects the capacitance of the electrochemical capacitor
relative to the dielectric capacitor. Firstly, by reducing the distance between the interfaces to
the order of a nanometre about a factor of 1000 less than that of the dielectric capacitor.
Design of the electrode in electrochemical capacitors can result in high surface areas leading
to more area between the interfaces than the typical geometric area of the electrode. Current
commercial electrodes have surface areas in excess of 1000m2/g
115.
However due to the presence of two electrodes in the electrolyte the electrochemical
capacitor is in practice two capacitors in series. Considering two identical electrodes this
means an electrochemical capacitor has half the capacitance of a single electrode. The
dielectric constant of the electrochemical capacitor should be that of the solvent molecule.
However, due to the high electric fields and the compact nature of the Helmholtz layer the
dielectric constant of water near charged surfaces has been in the range of 5-25 as opposed to
approximately 80 for the bulk116–119
.
The two key metrics of electrochemical capacitors are the energy and power densities.
These are commonly reported in terms of mass but have also been reported in terms of
volume and area. When taking into account device construction volume is most likely to be
the governing limitation.
35
In the charged state the voltage across an electrochemical capacitor is across the
interfacial layer of ions on either electrode. As well as considering the electrodes, there is a
potential drop associated with the resistance of the cell. This resistance known as the
equivalent series resistance (ESR) is due to the charge transport properties of the electrodes
and the electrolyte. The maximum voltage of an electrochemical capacitor is dependent on
the breakdown potential of the solvent in the electrolyte causing electrolysis. This is
analogous to the breakdown of the dielectric in conventional capacitors. The potential
window for water based electrochemical capacitors is 1.2V while common organic based
commercial electrochemical capacitors have operation voltage windows of 2.7V120
.
The energy of an electrochemical capacitor (E) is governed by the capacitance of the
electrode material (C) and the voltage window over which it operates (ΔV). The formula
describing this relationship is as follows:
Equation 3.3
The maximum possible power based on a load with the same ESR of the electrochemical
capacitor is given by the formula:
Equation 3.4
It must be noted however that the loads electrochemical capacitors work with usually exceed
this resistance. However this is a good metric for comparing devices.
3.3 Non-Faradaic Electrodes
An ideally polarisable electrode is an electrode in which no charge transfer occurs between
the electrode and the electrolyte. While many electrode materials are not entirely ideally
polarisable the bulk of the capacitance comes from the electrical double layer. These
constitute the non-faradaic electrodes.
To operate as a high performance non-faradaic electrode in a super capacitor the
material needs to be reasonably conductive, have a high surface area, good chemical and
36
thermal stability, and cost effective. Various carbon structures have been identified as having
these desirable traits with activated carbons being popular in commercial devices115,121
.
Activated carbons are porous carbon architectures with many methods of fabrication
but essentially involving the combustion of organic compounds. Good activated carbons for
non-faradaic electrodes will have a combination of mesoporous (2-50nm) and microporous
(<2nm) pores115
. The micropores allow the best possible gravimetric capacitance of the
material, up to 15-20µF/cm2 for carbon
122, to be utilized using high surface areas of activated
carbons up to 2500m2/g
115,123,124. While the mesopores ensure the transport of the electrolyte
throughout the material. The micropores should be optimized to accommodate the ions and
solvent molecules of the electrolyte. If the pore is too small to accommodate the ion and the
shell of solvent molecules an increase in resistance occurs due to the energy required for the
removal of solvent molecules. A patent by Okamura et al. tuned the pore size of the positive
and negative electrodes accounting for the difference in cation and anion sizes to produce
enhanced capacitance at low internal resistance125
. Activated carbons have capacitances of
around 100-200 F/g in aqueous electrolytes115,122,123,126,127
.
The random array of micro and mesopores in activated carbons may not lead to the
optimum arrangement of the pores to facilitate charge transport and capacitance. Templating
the growth of carbon to can be achieved using nanostructures as the template and then
etching the template material once the carbon is grown. The mesopores created by the
template provide access to the micropores present in the carbon127–129
.
Conductive low dimensional allotropes of carbon, namely graphene and carbon
nanotubes, have excellent conductivity and high surface areas130
. Films of carbon nanotubes
have a mesoporous structure due to the disordered morphology of the film making them an
excellent material for non-faradaic electrodes131–135
with the possibility for not requiring a
current collector reducing the equivalent series resistance that occurs from the boundary
between a current collector and the active material. They also have excellent mechanical
properties allowing the fabrication of flexible devices136,137
. Capacitances of carbon
nanotubes have reached as high as 180F/g138
.
While graphene possesses superior theoretical surface areas and conductivity to
carbon nanotubes the tendency to restack in films reduces the surface area and prevents the
access of electrolyte. Thin films display capacitances in excess of 100F/g139,140
and thicker
37
films get similar capacitive performance by employing strategies to prevent re-aggregation of
the sheets maximizing the surfaces areas141–146
.
Due to the excellent conductive properties of these carbon nanotubes relative to
activated carbon it is common to see composite materials benefiting from the excellent
porosity and capacitance of the activated carbon. These composites exhibit lower equivalent
series resistance and as a result better power density147–149
.
Graphene and carbon nanotubes have also been used in concert. While graphene has
excellent conductivity it is highly anisotropic with reduced conductivity in the direction
perpendicular to the sheet. In addition the carbon nanotubes prevent stacking of the graphene
sheets allowing improved electrolyte penetration. The result is improvements in capacitance
and resistance improving both energy and power densities of the electrodes124,150,151
.
3.4 Faradaic Electrodes
Carbon electrodes based on non-faradaic electrical double layer capacitance are limited to
specific capacitances of 50µF/cm2
or less152
. For higher capacities faradaic processes must be
considered. A faradic process involves charge transfer between the electrode and the
electrolyte. Batteries operate under faradaic processes where the charge exchange is limited
by solid diffusion processes152,153
. The distinction between battery-like processes and
pseudocapacitance-like processes is that the charge transfer occurs at the surface of the
material producing an electrical response similar to that of a capacitor, hence
pseudocapacitance, while in batteries the ions have to diffuse into the bulk of the material and
the surface has less of a contribution. In general faradaic electrodes are not ideally capacitive
and have a current behaviour like:
Equation 3.5
Where k1 and k2 are rate constants of the various processes and v is the scan rate154
. Semi-
infinite linear diffusion as process associated with batteries results in a current dependent on
v1/2
while the capacitive processes ungoverned by diffusion have current dependent on v.
Pseudocapacitive materials show a combination of the processes.
38
The surface based reactions for charge storage can be broken down into various
mechanisms: underpotential deposition, surface redox reactions, intercalation and reversible
electrochemical doping of conducting polymers152–154
. These processes tend to cause
morphological changes in the charge/discharge cycle causing device lifetimes to be shorter
than those of the electrical double layer based devices.
These processes occur at the surfaces and thus the capacitance can be determined
from the percentage of surface sites that have undergone said process. The percent of surface
sites reacted for these processes is determined by the thermodynamic relation152
:
Equation 2.6
Where θ is the percentage of surface sites, K is the electrochemical equilibrium constant, C is
the concentration of the species involved in the charge transfer, V is the applied potential, F is
the faraday constant, R is the gas constant and T is temperature. Pseudocapacitance, Cϕ, is
given by the charge transferred multiplied by the derivative of surface coverage with respect
to potential as is described by the following formula152
:
Equation 2.7
Noble metals like platinum, rhodium or iridium can adsorb metal atoms to the surface
accompanied by a transfer of charge. The transfer of charge follows the reaction:
Equation 2.8
Where M is the metal ion in solution, S is the surface lattice site and z is the charge of the ion
in terms of the charge of an electron. For single state processes the capacitance associated
with such a reaction for example hydrogen adsorbed on platinum has a maximum value of
approximately 2200µF/cm2. However single state processes have narrow potential windows.
A system with more states represented by multiple charging peaks such as lead adsorbed on
Gold gives capacitance over approximately 0.6V. Due to the narrow potential windows and
expensive material like platinum and gold in addition to the difficulty of reaching surface
areas to be competitive with carbon, pseudocapacitors based on this process are less attractive
than others152
.
39
Redox active sites in materials can be accessed by ions on readily available surface
sites or intercalation sites within the material. This mechanism involves charge exchange
with a redox active site with variable oxidation state and an electrolyte ion. The redox
reaction is described by the equation:
Equation 2.9
Where Ox is the oxidized state, C is the cation in the electrolyte, Red is the reduced state and
z is the charge difference between the two states. Ruthenium oxide (RuO2) was the first
material in which surface redox pseudocapacitance was thoroughly investigated155
. Hydrous
RuO2.nH2O is commonly used as the electrode material. The redox reaction of hydrous RuO2
can be expressed as:
zyzxyx OHRuOzezHOHRuO
)( Equation 2.10
In the equation H represents the cation. RuO2 has multiple redox processes involving redox
states from Ru(II) to Ru(VI)156
, this and the stability over a 1.2V window in aqueous
electrolytes make for an excellent pseudocapacitve material. The maximum theoretical
capacitance of RuO2 is 1450F/g154
. Ruthenium however is very expensive and so other
materials with pseudocapacitive properties were identified for capacitor applications.
These materials can be divided by chemical composition and also the
intrinsic/extrinsic nature of the capacitance. Materials that have intrinsic pseudocapacitance
display capacitive properties regardless of the size or morphology of the material. Materials
with extrinsic pseudocapacitance, however, require modification. Namely, reduction of size
to access the capacitive like behaviour154
.
Transition metal oxides have a range of faradaic materials suitable for replacing RuO2
(MnO2157–159
, TiO2160
, Fe3O4161
, Nb2O5162
, WO3163
and LiCoO2162,164
). Of which MnO2 is the
most extensively studied due to its abundant availability and low impact on the environment.
MnO2 has a theoretical capacitance of 1370F/g over a potential window of 0.8V
corresponding to the redox reaction between the Mn(III) and Mn(IV) redox states158
.
Other transition metal compounds have also been identified as having significant
pseudocapacitance. Transition metal layered double hydroxides (Co(OH)2165
and Ni(OH)2166
)
has also displayed pseudocapacitance. Sulfides of transition metals like MoS2167
and CoS2168
40
are also pseudocapacitive. Nitrides such as vanadium nitride169
and carbides in the two
dimensional MXene variety (Ti3C2170
) have been reported as pseudocapacitive also.
Hetero atoms namely oxygen and nitrogen in normally non-faradaic carbon materials
are also potential redox active sites. In addition to adding pseudocapacitance these
heteroatoms can improve the wettability of the electrode materials171–173
.
Conducting polymers that undergo redox reactions which change the doping state of
the polymer chain have also been used in electrochemical capacitors111,174
. This class of
materials include PANI175
, PPy176
and PEDOT177
as the most common examples. These
materials provide good conductivity and flexibility on top of their pseudocapacitive
properties.
Many pseudocapacitve materials have low conductivity which has a severe impact on
the power of the device due to increasing the equivalent series resistance. In order to counter
this issue composite electrodes are used to make functional electrodes. Many electrodes are
constructed from pastes of the active material, carbon black (for conductivity) and PVDF (as
a binder)157,158,164,178
. Other conductive materials used for composite materials include carbon
nanotubes179–181
, graphene182–184
and conductive polymers185–187
. This approach is particularly
important as the film thickness increases as many of the high capacitance values are reported
for low-mass and thickness films on a current collector. Lee et al. produced 20 micron thick
films of only MnO2 resulting in a capacitance of 0.13 F/g and addition of carbon black
increased this by three orders of magnitude157
. While analysing ultra-thin films is good for
characterising a material it does not facilitate device applications.
On top of the series resistance faradaic materials exhibit a charge transfer resistance
due to the crossing of ions through the electrical double layer and interacting with the surface.
This manifests itself as an additional resistance component in the circuit in such a way that
faradaic materials generally have worse power densities than non-faradaic materials.
In addition the design of electrodes made from these pseudocapacitve materials must
take into account the requirement of high surface areas to maximize the area accessible by
electrolyte.
3.5 Electrolytes
41
The electrolyte contains the electrolyte ions and a solvent. As discussed earlier the potential
window of liquid electrolytes is limited by the electrolysis of the solvent. The maximum
potential windows are 1.2V for water, 2.5-2.8 for organic electrolytes and 3.4-3.7 for ionic
liquids120
. With commercial electrochemical capacitor manufacturers favouring organic
electrolytes. The maximum potential window is not the only important property of an
electrolyte however.
Figure 3.10 shows the many facets in which the electrolyte affects the performance of
a electrochemical capacitor. Ion properties such as conductivity, size and electrode material
interaction are important as are the solvent properties of viscosity and salt solubility. An
extensive review on this topic has been produced by Zhong et al120
.
Figure 3.9 Relationship between electrolye properties and electrochemical capacitor properties taken from reference120.
Solid and quasi-solid electrolytes offer the convenience of the absence of liquid. This
abolishes the possibility of leakage facilitating the integration of electrochemical capacitors
into wearable and flexible technologies. Most of the research in this field involves polymer
gels188,189
.
42
3.6 Device Design
Essentially all devices have two electrodes on current collectors, with a separator between
them and filled with an electrolyte as seen in figure 2.4. Due to the wide range of applications
for electrochemical capacitors there is no one size fits all style of device. Requirements for
different energy and power characteristics as well as physical properties like size, flexibility
and transparency govern the design of devices.
The thickness of the active electrode materials depends on whether the device is
designed for high power (around 10 microns) or high energy (around 200 microns)190
as
energy will increase with more capacitive material while the increase in thickness results in
charge transport issues. While most research focusing on material characterisation focuses on
the capacitance of the electrode material the overall energy and power densities with respect
to both mass and volume are affected by the other components of the cell. Minimizing the
contributions of these components is helpful to realizing effective devices.
Figure 3.10 Schematic of a electrochemical capacitor
Current collectors in commercial devices are usually metal foils such as aluminium or
copper in the range of 20 to 200 microns190–192
. These materials contribute little to the
capacitance of the device and can have thicknesses comparable to the active materials
especially in pseudocapacitve electrodes with thickness limited by conductivity. There is
also a transfer resistance associated with the current collector active material boundary192–194
.
An ideal solution would be to have the electrode material work as its own current collector.
43
The advantages are double, as it removes the contact resistance and increases the percentage
volume of active material in the device. This is often the case for ultra-thin film
electrochemical capacitors based on carbon nanotubes and conductive polymers139,195–197
.
Quintero et al. used carbon nanotubes as the current collector in a 200 micron thick
composite electrode with activated carbon198
.
Separator materials tend to be porous with low resistance to ion diffusion. The contribution of
separators to the equivalent series resistance is proportional to the separator thickness. To
reduce cost and volume and to improve power capability the separator should be as thin as
possible, however the minimum thickness is limited by the possibility of shorting due to free
electrode particles creating contact between electrodes. The materials for separators include
cellulose, glass fibre and various polymers. The thickness range in commercial
electrochemical capacitors is between 15 and 50 microns199
.
While sandwich style electrochemical capacitors have been fabricated, roll type
electrochemical capacitors are the industry norm. The roll type requires mechanical stability
and flexibility in all internal materials. To facilitate this there has been much focus of flexible
electrode materials. The electrical double layer materials, carbon nanotubes and graphene
were combined to make a flexible electrode with a capacitance of 0.1 mF/cm2 by Lu et al
200.
The addition of pseudocapacitive materials have been used to increase performance. Flexible
films using a carbon material and nanostructured MnO2201,202
have been demonstrated with
capacitances per unit area in excess of 1 F/cm2 up to 3.2 F/cm
2. A reduced graphene oxide/
Polyaniline composite electrode fabricated by Wu et al.203
reached a capacitance of 160
F/cm3.
For flexible devices a solid state and flexible electrolyte in conjunction with flexible
electrode materials is required. Carbon nanotubes on paper using an ionic liquid gel
electrolyte by Kang et al.204
and an activated carbon cloth with a polymer gel electrolyte by
Wang et al.205
produced flexible solid state capacitors with capacitances in the order of
10mF/cm2. Le et al.
206 used carbon nanotubes wrapped in a carbon fibre paper to create fibre-
like electrochemical capacitors with 86.8F/cm2. Better non-faradaic based devices using
polymer gel electrolytes include a cotton paper embedded with carbon nanotubes by Hu et al.
creating a 1.08 F device207
and a graphene hydrogel electrode by Xu et al.147
with a
capacitance of 402 mF/cm2.
44
Faradaic materials have also been integrated into all solid state electrochemical
capacitors. Peng et al. constructed an all solid state device using a graphene MnO2 hybrid
film of 40 nm thickness with a polymer gel electrolyte208
. The capacitance based on active
materials was 267 F/g. However low mass loading due to the low thickness of the films limits
device applications. Using carbon fibre as a current collector and mechanical support for
carbon nanoparticles and MnO2 and a polymer gel electrolyte Yuan et al.201
produced a
device with a capacitance of 109mF/cm2. Meng et al.
209 using carbon nanotubes and
polyaniline produced individual 30 micron electrodes of 0.8 F/cm2 incorporated into a device
with polymer gel electrolyte the capacitance was 37 F/cm3 taking into account all device
components.
Transparency is another ideal pursued by portable device designers. This requires
transparent energy and power storage also. King et al discussed the effect of percolation
theory on the production of transparent electrodes with carbon nanotubes195
. This resulted in
a capacitance of around 0.1 mF/cm2 at 90 % transmittance. Many transparent devices have
been fabricated based on carbon materials from carbon nanotubes, graphene and other carbon
materials136,197,210
. These devices range in transparencies of 50 to 70 percent and capacitances
from 0.1 to 1 mF/cm2. Liu et al.
70 used cobalt oxide to from a pseudo capacitive device with
51% transmittance at 550nm wavelength with a capacitance of 6.03 mF/cm2. Higgins and
Coleman produced thin film doped PEDOT:PSS to get capacitances as high as 1 mF/cm2 with
transparencies of 80% for a single electrode196
. In addition, these thin films have high
resistances and thus the dimensions of the device affect the capacitance per unit area.
Comparison of these devices however need to carefully done as internal resistances and
difference in voltage windows affect the energy and power densities.
A popular trend in electrochemical capacitor research has been to combine electrical
double layer electrodes with pseudocapacitor or even battery type electrodes in asymmetric
devices seeking to benefit from the high energy density of the faradaic electrode and the high
power density of the non-faradaic electrode112
. Due to the total capacitance of the electrodes
in series being 1/C = 1/Cf +1/Cnf the upper limit of the capacitance is determined by the lower
capacitance. While the increase in capacitance of the device will not reflect the superior
capacitance of the pseudocapacitance the two materials can have different ranges of working
potentials the device can have a larger potential window than the standard 1.2V associated
with aqueous symmetric devices211–213
. This has a significant impact on energy and power
densities.
45
For the production of commercial materials for electrochemical capacitors, methods
need to be compatible with scalable industrial processes. Doctor blading of pastes onto
substrates is a standard method for producing commercial devices199
. For scalable processes,
the prepared materials must be compatible with paste formation for doctor blading, screen
printing or drop casting. Inkjet214
printing or spray-coating215,216
are excellent methods for
producing thin film electrodes for transparent applications.
In this work PEDOT:PSS will be used to fabricate a range of charge storage devices
with a particular focus on the effect the electrode dimensions have on electrical and physical
transport properties.
46
4 Methods
In this thesis I fabricate a variety of DSSC counter electrodes and supercapacitor electrodes.
The aim of the work on DSSCs was to produce a counter electrode with comparable
performance to platinum with cheaper materials. For the supercapacitor electrodes the aim
was to analyse how the electrical properties associated with the materials used as well as
device dimensions impacted on electrode performance with a view to streamlining device
design.
4.1 Sample Preparation
In order to produce films for use in DSSCs and supercapacitiors raw materials need to be
processed in such a way that mechanically stable films can be reliably produced on a
substrate. The following section deals with methods pertaining to the formation of these
films.
4.1.1 Liquid Phase Exfoliation by Ultrasonication
With solid starting materials either naturally occurring or synthesized, the material is unlikely
to be in a state where the material will have the desired properties required for a device. For
examples, graphite consists of layers of graphene which has superior electrical properties and
carbon nanotubes tend to form bundles217
. Transition metal Dichalcogenides undergo band
structure changes with increasing degree of exfoliation218
.
To separate aggregated materials the attractive force between particles (usually Van
der Waals) must be overcome. The application of sonic energy in excess of 20kHz to liquid
solutions can be used to both separate these aggregated materials and disperse the particles in
a liquid. The process of dispersing materials in a liquid environment is referred to as liquid
phase exfoliation (LPE).
47
The device that provides this energy is commonly known as a sonicator. The sonicator
has an electronic signal generator that drives a piezoelectric converter to provide the
longitudinal mechanical vibrations of the probe. These vibrations can be on the order of 100
microns allowing high power densities of 100W/cm2 to be created.
When applied to powdered material in an appropriate solvent the vibration of the
probe creates waves in the liquid which result in the formation of cavitation bubbles.
Collapse of these cavitation bubbles result in high local temperatures and pressures resulting
in the formation of jets. These jets can drive aggregate particles apart themselves as well as
causing particle collisions which also assist disaggregation.
The use of sonication has been used to produce dispersion of carbon nanotubes, graphene
and other layered materials217,219–222
. This process does not produce monodisperse dispersions
of isolated nanoparticles. Usually a range of bundle thicknesses (for nanotubes) or layers for
two-dimensional stacked materials are observed. In addition differing degree of exfoliation
provided by sonication the particles dimensions are also affected by the process. Due to
‘sonication induced scission’ the covalent bonds of the material being sonicated can also be
broken221,223
. The average size of the particles decays as the inverse square root of sonication
time. This creates extra edge sites in the nanomaterial but otherwise does not cause the
material properties to change as basal plane defects would217
. Depending on the application,
the time of sonication can be adjusted to provide particles of suitable size distribution.
48
Figure 4.11 UP200S sonicator as used in the work in this thesis
Recently an alternative to sonication has been developed. Shear mixing has been
shown to generate sufficient shear forces to cause exfoliation of layered materials. The main
advantage of this is that it can be scaled up to larger volumes allowing a higher
throughput224,225
.
4.1.2 Stabilisation
Once the particles have dispersed by sonication or an alternative method they will re-
aggregate and sediment out of dispersion if they are not stabilized. One method of stabilizing
a dispersion is to choose an appropriate solvent. N-Methyl-2-pyrrolidone has been identified
as an excellent solvent for a wide range of materials217,221,226
due to its good matching of the
surface energy with the nanoparticles. The main disadvantage of the solvents which are
particularly good for stabilizing these nanoparticles tend to be toxic and have high boiling
points. This raises health and safety issues as well as difficulties in processing.
Analysis of the surface energies of solvents using Hansen and Hildebrand parameters
have been carried out in an effort to identify alternative low boiling point solvents. Isopropyl
Alcohol (IPA) with a low boiling point of 83⁰C has been used to disperse two dimensional
materials222,227
.
49
Water is cheap, non-toxic and has a relatively low boiling point. However water is
entirely unsuitable as a solvent due to poor surface energy matching. To produce stabilized
dispersions in water additives must be used to prevent re-aggregation of the nanoparticles.
Surfactants have been used to stabilize dispersions.
Surfactants can be split into ionic and non-ionic surfactants. Ionic surfactants, such as
sodium dodecyl sulphate (SDS) and sodium cholate (SC), bond to the surface of the
nanoparticle with a hydrophobic group and create a charged surface using the ionic tail
group. This charge surface acts as a repulsive layer between dispersed particles. The non-
ionic surfactants such as Triton X bind to the nanoparticles but rather than a charged tail the
steric hindrance of the tail groups prevent re-aggregation228
. Polymers have also been
reported to show this behaviour229
. Polyvinyl alcohol (PVA) in water has been used as a
stabilizing agent230
. As part of this thesis Ethyl Cellulose (EC) dissolved in IPA has been
used to form high concentration dispersions.
The use of additives raises challenges of their own however. The presence of these
surfactants can adversely affect the properties of the film and as such the use of surfactants
has to be administered with caution and may require further treatment steps to remove excess
material.
4.1.3 Centrifugation
Once the sonication process is completed there is a wide range of particle sizes present in the
dispersion. While the larger material in the dispersion will sediment out over time this
process is slow and does not offer a large degree of control over the particle size in the
dispersion. Centrifugation provides a way of removing unnecessary particles from a
dispersion quickly.
Centrifugation involves rotating the liquid sample about an axis at high speeds. The
centripetal force acts as an analogue to gravity with a much higher magnitude causing the
particles of larger sizes to sediment out at higher rates. Multiple steps can be used to isolate
particles in different size ranges231–233
. This is useful due to the suitability of different size
nanosheets to different applications. For example: Long lengths correspond to high aspect
ratios which are desirable for mechanical reinforcement applications and low lengths
correspond to an increased number of edge sites which can be of use in electrochemical
applications.
50
Determination of the size of particles can be done using atomic force microscopy
(AFM) or transmission electron microscopy (TEM). This requires gathering statistics over a
large number of flakes to provide accurate information. Recently use of various spectroscopic
methods have allowed the characterisation of flake dimensions of dispersions allowing for
quicker characterisation232,234
.
4.1.4 Film Formation
Many methods of preparing thin films were used over the course of this thesis. The
production of thin films from dispersion is vital to realising the applications of these
materials. The films used in this thesis range from less than 100 nanometres to in excess of
100 microns. The thickness of the film is an important property which is affected by charge
and mass transport properties in electrochemical applications. The use of liquid dispersion
allows the mixing of material systems provided the solvents are compatible to realise
composite films with a range of compositions. To produce these thin films in a wide range of
thicknesses multiple methods of production have been used.
Vacuum filtration allows the fabrication of uniform thin films by passing the
dispersion through a membrane with a pore size smaller than the size of the particles desired
in the film. The dispersion is pulled through the film using a negative pressure created by a
pump. The local pressure on the membrane is affected by the amount of material deposited
over a certain area allowing for film uniformity over the filtration area.
Figure 4.2 Vacuum Filtration Setup
51
By filtering various masses of dispersed particles films of thicknesses as low as 200nm up to
several microns can be fabricated, knowledge of dispersion concentration allows accurate
determination of film mass per unit area and thickness (provided the density of the material is
known). The films must then be transferred to a desired substrate for application and
measurement. This is an extra processing step. Also the membranes cannot be reused. These
issues, and the fact that the deposition area is limited to that of the filtration set up, limit the
industrialisation of this process. As such other methods of film formation were explored in
this thesis.
Drop-casting is the simplest method for creating films. It involves dropping the
dispersion on a substrate. The solvent can then be evaporated off under a variety of
conditions. Slow evaporation leads to reasonably uniform films but has the drawback of the
“coffee-ring” effect. This causes material to be deposited at the edges of the film. This
method is commonly used for the fabrication of freestanding polymer films. Due to the effect
of ambient conditions on dropcasting it lacks the reproducibility of other methods.
Spin-coating is a technique for creating uniform thin films on a substrate. The
dispersion is placed in the centre of the substrate while the substrate is stationary or rotating
at a low rate. The rate of rotation is then increased which allows the material to spread
causing a thin layer of the dispersion to coat the surface which can then quickly evaporate.
The thickness of the film can be altered by increasing the concertation of the dispersion (to
produce thicker films) and increasing rotation rate (to produce thinner films). One problem
arising from this method is that material can be lost by being flung from the substrate at high
speeds.
Spray-coating is a scalable method for producing films. It involves using an airbrush
gun and a driver gas to aerosolise the dispersion and drive it towards the substrate235
. A
needle controls the amount of dispersion to be aerosolised which can be used to adjust the
rate of deposition. The substrate is heated to a suitable temperature such that on contact the
droplets of dispersion evaporate quickly. This prevents the coffee stain effect and using a
robot to raster the airbrush gun across the substrate allows for excellent uniformity. The area
of the film is determined by the stage area which is easily scaled up which translates into
facile industrialisation.
52
Figure 4.3 Spray-coating onto substrate adapted from source235
Other methods of film deposition used in this work include doctor blading and screen
printing. These methods require a more viscous material to produce films. To increase
viscosity the concentration of the material can be increased as done in Chapter 8. While for
producing pastes of the required viscosity for screen-printing titanium dioxide nanoparticles
an additive ethyl-cellulose (EC), which can later be burned off is added13
.
Doctor blading involves the use of a blade to drag a paste across a substrate after
which excess solvent is evaporated. For supercapacitors the paste is commonly formed using
a slurry formed by mixing the active material with a conductive additive and a binding agent.
Doctor blading allows film formation of various thickness by changing the distance between
the blade and the substrate. While handheld doctor blade apparatus are common for
fabricating devices in the lab it is possible to scale up doctor blading for industrial processes.
Screen printing is similar to doctor blading in that a paste it pulled across a substrate
to produce a thin film. However in this case the material is dragged across the surface of a
screen with a mesh to allow the material to pass through. The thickness of the film is
determined by the thickness of the screen and the thickness and density of the fibres that
make up the mesh. An advantage of screen printing is that patterns with details in the range of
50 microns are possible and multiple layers can be printed allowing printable electronics to
be realised.
53
4.2 Sample Characterisation
Once the dispersions have been stabilised or the films been fabricated it is critical to
characterise them in order to ascertain the suitability of the material for various applications.
4.2.1 UV-Vis Specrophotometry
UV-Vis spectrophotometry uses light in a range of 200-800nm which is passed through a
sample. The intensity of the light source (I0) is then compared with the intensity at the
detector (I). Reduction in intensity is due to absorption and scattering of light by the sample.
Absorption occurs when light of energy in excess of the band gap causes the excitation of an
electron from one optical band to another. These can be associated with HOMO (highest
occupied molecular orbital) and LUMO (lowest unoccupied molecular orbital) for molecules
and the conduction and valence bands in solids. Scattering is caused by interaction of the
light with the material that causes the direction of the light to change. This is dependent on
particle size among other factors.
This method is used to characterise both films and dispersions in this work. For
characterising films the transparency was the property of interest. The transparency is the
ratio of the intensity at the detector to the intensity at the light source (T = I/I0). Transparency
is often used to characterise films for applications where high transparency is desirable.
For liquid samples such as dispersion transparency is not as useful for determining the
concentration. The property used in this case is the extinction which varies linearly with
concentrations within a certain concentration range. The relationship between the extinction
(E) and transparency is given by the following equation:
)log(TE Equation 4.1
The concentration is then related to the extinction by the Beer-Lambert law which states:
clE Equation 4.2
54
Where ε is the extinction coefficient, c is the concentration and l is the path length. For
measuring liquid samples the liquid is held in a cuvette of known path length. The
concentration can be found by filtering a known volume of dispersion. Then a dilution series
with known concentrations and extinctions can be used to calculate the extinction coefficient.
Once an extinction coefficient is known this enables quicker determination of the
concentration of subsequent samples. Of further interest the extinction can be broken into
absorption and scattering components. This can be of use to more accurately ascertain the
optical properties of the materials and has been used for determination of average particle
sizes of dispersions232,234
.
4.2.2 Profilometry
In this thesis to determine sample thicknesses contact profilometry was used. This involves
the dragging of a stylus across a surface. The stylus can move vertically using a feedback
loop in such a way that a constant force is maintained. The requirement to use force can be
destructive to some surfaces. The averaging of at least three measurements over the sample
gave the thicknesses for samples.
4.2.3 Electron Microscopy
To analyse topography of films and the dimensions of individual particles of a dispersion
optical microscopy is not sufficient. This is due to the resolution limitation which is
dependent on the wavelength of the particle probing the sample. High energy electrons have a
suitably low wavelength to probe the topology with high resolution.
Electrons are excited using either a thermal gun or a field effect gun and focused
using magnetic lenses and aperatures. Impact with the sample there are multiple interactions
that can happen. Electrons penetrate the surface and can be scattered. The incident electrons
can be scattered and can also scatter electrons in the sample. The result of removing
electrons from sample atoms causes the emission of both X-rays and Auger electrons236
.
Of primary use for determining topology are low energy secondary electrons as only
the electrons generated close to the surface can be detected. The angle of incidence of the
beam with the surface also effects the intensity of secondary electrons making them very
55
sensitive to topology. Secondary electrons are caused by the emission of valence electrons.
Back scattered electrons originate in the electron beam and tend to have higher energies than
secondary electrons. The back-scattering of electrons is dependent on the atomic number of
the sample nuclei and can be used to give compositional information. Auger electrons are
cause by the emission of an electron after another electron fills a hole in the core shells these
and X-rays can be used to give compositional information for atoms of low atomic number.
Transmitted electrons include electrons that have not interacted with the sample237
.
These pass through with no deflection from the beam path. And can be used to detect the
morphology of the sample. Scattered electrons can also be detected. Elastic scattered
electrons can be used to probe lattice properties and generate higher resolution images.
Inelastic scattered electrons can be analysed using electron energy loss spectroscopy (EELS)
to determine composition, bonding and valency of the sample. For transmission electron
microscopy the sample needs to be less than 100nm thick to allow sufficient transmission of
electrons for signal generation. In this thesis electron microscopy is done on flakes collected
on a grid. All transmission microscopy and data analysis was carried out by Andrew Harvey.
56
Figure 4.4 The incident electron beam creates multiple types of signals (A) transmitted electrons (B) Ineleastically scattered electrons (C) eleastically scattered electrons (D) back-
scattered electrons (E) secondary electrons (F) Auger electrons and (G) X-rays
4.2.4 Electrical Characterisation
To characterise the electrical properties of films such as sheet resistance and conductivity 4-
wire IV measurements are performed. This involves attaching four linear electrodes to a film.
A current is passed through the outer electrodes while the voltage drop is measured across the
two inner electrodes. The electrical characteristics of the system can then be analysed. The
advantage of the four wire setup is that it eliminates sources of error such as resistances
associated with electrode contact and the leads.
57
Figure 4.5 The 4-wire experimental setup. Current is passed from electrodes 1 to 4 and the voltage drop across electrodes 2 and 3 is measured.
The films analysed in this thesis display ohmic behaviour (V = IR). The resistance of the film
is determined and related to the sheet resistance by the length between the two inner
electrodes and the width of the film.
l
wRRs Equation 3.3
Where Rs is the sheet resistance, R is the resistance as measured from the four-wire
measurement and w and l are the width and length of the film respectively. Sheet resistance is
a measure used for films of uniform thickness and the unit is the Ohm per square (Ω/sq.).
Sheet resistance can also be used to determine the conductivity of a film as the relationship
between them is:
tRs
1 Equation 3.4
Where σ is film conductivity and t is the film thickness. Conductivity is particularly useful
when considering composites of materials with differing conductivity or very thin films
where conductivity varies with thickness195
.
58
4.2.5 Electrochemical Charaterisation
The properties of electrochemical solar cells and supercapacitors can be assessed using
various electrochemical techniques. The behaviour of electrodes in a solution under potential
can be used to assess charge and mass transport behaviours associated with chemical
reactions and processes.
To measure current and voltage characteristics of films in solution a potentiostat is
used. The potentiostat is capable of generating various voltage and current signals to probe
the properties of the film. In the simplest experimental setup two electrodes, a working
electrode and a counter electrode, can be used and the current-voltage characteristics can be
analysed. The working electrode is the film which is being examined. The counter electrode
is to complete the circuit as two electrodes are required. Common counter electrode materials
are graphite and platinum. In some cases a two electrode measurement with a symmetrical
configuration (working electrode and counter electrode are identical) can be used.
However, in the two electrode setup the potential across the cell is dependent on both
the working electrode and the counter electrode. While the behaviour of the working
electrode with us the fluctuation of potential of the counter electrode can obscure the actual
potential of the working electrode. To counter this a three electrode setup is used. The third
electrode called the reference electrode. The reference electrode is an electrode with a
electrochemical potential that is invariant in the voltage. In this thesis the silver-chloride
reference electrode is used though a range of reference electrodes exist. The silver chloride
electrode is a silver electrode coated in a porous silver chloride layer in a saturated sodium
chloride solution. The redox reaction associated with this system is:
ClsAgesAgCl )()( Equation 4.5
The Nerst equation114
can be used to determine the potential of the electrode.
Cl
aF
RTEE ln0 Equation 4.6
Where E is the electrode potential, E0 is the standard cell potential, R is the gas constant, T is
temperature and F is the faraday constant (at 25⁰C RT/F can be treated as 25.693mV) and aCl-
is the activity of the chloride ions in the solution. As the electrode is saturated with NaCl
59
solution the activity of the chloride is constant. This gives a reference electrode potential of
+0.197 vs the standard hydrogen electrode. Assuming the distance between the reference
electrode and the working electrode is sufficiently small the potential of the working
electrode is that between the working electrode and the reference.
Figure 4.6 The setup for a three electrode system. The current between the working electrode and the counter electrode is recorded and the potential difference between the
working electrode and reference electrode is recorded. Image adapted from source114
4.2.5.1 Voltammetry
Voltammetry involves the observation of the behaviour of current in electrochemical cells
over a range of potentials. In this thesis both linear sweep and cyclic voltammetry are
used. Linear Seep voltammetry uses the potentiostat to ramp the potential from one value
to another while cyclic voltammetry ramps the potential from one value to another and
back to the original potential often repeatedly. The decision to use linear sweep
voltammetry versus cyclic voltammetry is determined by the electrochemical system and
the information required from the system. If the reaction of interest is irreversible or no
more information of value can be obtained from performing cyclic voltammetry linear
sweep voltammetry may be used.
60
In this thesis linear sweeps are performed in a two electrode symmetrical setup. The
current and voltage characteristics of the symmetrical cell are described by the Tafel
equation:114
.
EnF
RTII
)ln()ln( 0 Equation 4.7
Where I is the current, n is the number of electrons exchanged in the reaction, F is the
Faraday constant, R is the gas constant, T is temperature and ΔE is overpotential. I0 is the
exchange current density which is characteristic of the electrode process related to the
rate constant and the concentrations of the electrolyte species. As such exchange current
can be used to assess the suitability of dye-sensitized solar cell electrodes. The exchange
current is achieved by extrapolating the linear region of the Tafel plot to zero
overpotential as shown in Fig 7.
Figure 4.7 Characteristic Tafel Plot with extrapolation of linear region to show exchange current
Cyclic voltammetry is used in this thesis to observe redox reaction in potential
counter electrodes for dye-sensitized solar cells and the charge storage in supercapacitor
electrodes. For a cyclic voltammetry experiments mentioned in this thesis the three
electrode setup was used. The phenomena observed by cyclic voltammetry in these cases
is the current flow associated with the formation of the electrical double layer described
in chapter 2 and peaks associated with redox reactions such as the one described in
chapter 1.
61
The electrical response due to the formation of a double layer is that of a capacitor. As
such it can be shown that the current under cyclic voltammetry can be described as:
RC
tsCI exp1 Equation 4.8
where I is the current, s is the scan rate, C is the capacitance of the double layer, t is time
and R is the resistance of the system. The current response of the double layer then looks
like Fig. 8(b).
The presence of redox active species can cause the appearance of peaks in the cyclic
voltammogram. These redox peaks appear on superimposed on the current associated
with the double-layer formation. As such if the double layer current is comparable to the
faradaic current then for current analysis the double layer current should be removed.
The current characteristics of a redox system is shown in Fig 4.9. For the purposes of this
thesis the analysis of the current was qualitative and more attention given to the voltages
at which the peaks appeared.
Figure 4.8 (a) The change of potential with time for a cyclic voltammetry cycle the slopes of the lines are the scan rate. (b) The current response due to double layer capacitance the maximum current is the product of the scan rate and the capacitance.114
The midpoint between the peaks is the standard potential of the redox couple. The distance
between the peaks is dependent on the rate constant of the reaction. A high rate constant
would correspond to a pair of peaks with a low potential difference, for the purposes of
62
testing counter electrodes for DSSC this is a suitable method for comparison of different
electrodes.
Figure 4.9 The current response due to faradic redox reactions. The peak is centred on E0 which is the standard potential of the redox couple114.
4.2.5.2 Impedance Spectroscopy
Impedance spectroscopy is a useful technique for analysing the variation of impedance with
frequency. Potentiostatic impedance spectroscopy involves oscillating a voltage at low
amplitude around a fixed voltage. The low amplitude is used in order to operate within a
linear region of the system which can then be examined using Ohm’s law. The voltage signal
looks like:
)sin(0 tVV Equation 4.9
where V is the Voltage, V0 is the amplitude of the oscillation, ω is frequency and t is time.
This produces a current response which looks like:
)sin(0 tII Equation 4.10
where I represents current. The new term, ϕ, is the phase which represent a delay in the
response of the system. An ideal resistor is completely in phase (0⁰) while a capacitor is
completely out of phase (90⁰). When Ohm’s law is applied the impedance is retrieved which
result in the impedance being expressed in terms of real and imaginary components.
63
When an impedance spectrum is recorded there are two common ways of viewing the
data. The Bode plot plots the phase and the magnitude of the impedance against frequency
while the Nyquist plot plots the imaginary component of the impedance against the real part
of the impedance.
Figure 4.10 Point in Nyquist plot represented by magnitude of impedance and phase. These quantities are plotted against frequency in a Bode plot114.
Usually analysis of impedance spectra involves using a circuit diagram to visualise
what the system looks like before fitting the data to the model to get values of the individual
components. These include resistor and capacitor components as well as components that
model ion diffusion in the electrolyte among others. Common models involve the
transmission line model for supercapacitor pores238
and physical dimensions196
and the
Randel’s circuit239
used to analyse charge transfer reactions such as the one in dye-sensitized
solar cells.
4.2.6 Solar Simulation
The test of solar cells requires the use of light to cause the cell to produce a voltage. For
experimental purposes it is unreliable and impractical to rely on the sun to provide the light
source. However since the light under which these cell will be exposed is the sun’s light a
simulation of the spectrum of the sun is necessary for testing.
Solar simulators provide a light supply approximating that of the sun. They employ a
light source which is commonly a zenon lamp but LED slight sources are also available. An
air mass filter is also used to simulate the absorption of light by the earth’s atmosphere. The
output of the lamp and filter of the simulator used in this thesis is shown in Fig 4.11.
64
Figure 4.11 the spectral irradiance of a solar simulator from manufacturer’s website
65
5
Materials
5.1 Graphene
Graphite is a layered material consisting on hexagonal lattices of sp2 hybridised carbon
atoms each about 0.142nm apart. The interlayer spacing is 0.335nm and the layers are
adhered to each other by Van Der Vaals forces. Individual layers were used to theoretically
examine the electronic properties of graphite but were thought to be unstable240
. These layers
were given the name ‘graphene’. As such the graphene lattice was used to describe the
formation of carbon nanotubes and fullerenes.
Figure 5.1 Graphene sheet and the construction of various carbon allotropes fullerene (green) nanotube (red) and graphite (blue) adapted from reference241
66
Monolayer graphene produced by mechanical exfoliation was first observed by
Novoselov & Geim in 2004 using scotch tape to mechanically isolate few layered to mono-
layered samples for electronic measurements. The few layered flakes with a thickness of less
than 3nm could be isolated with 10µm in the lateral dimensions242
.
These few layered graphenes were able to be transferred for a free standing
measurement of electronic properties such as charge carrier density and mobility of
approximately 1013
cm-2
and 10,000cm2V
-1s
-1 corresponding to sub-micrometre ballistic
transport. Free standing is critical for ascertaining the properties of extremely thin films as
substrate interaction can have an effect. On top of the excellent electrical properties, the
mechanical properties of graphene are phenomenal, making it one of the strongest known
materials243
. It is also very strongly absorbing of light for an atomically thin sample with a
transparency of 97.3%244
. Absorbance is related to transparency by a negative logarithmic
relationship as shown in equations 4.1 and 4.2.
With these excellent properties came the desire to produce graphene in reasonable
quantities to actualise devices. Unfortunately mechanical cleavage of graphite is not
compatible with industrial scaling. As such alternative methods to produce large quantities of
defect free graphene are required. The most popular methods are silicon carbide growth,
chemical vapour deposition and liquid phase exfoliation.
Silicon carbide growth involves heating a silicon carbide crystal in an ultra-high
vacuum to evaporate silicon on the surface leaving a layer of sp2 carbon atoms.
Approximately 30% of these carbon atoms are bonded covalently to silicon atoms in lower
layers degrading the graphene-like properties. However hydrogen intercalation can separate
the carbon from the substrate245
. Domain sizes and graphene properties vary depending on the
face of the crystal used with the carbide face having domains of 200nm compared to 100nm
of the silicon face. This results in better electronic properties for the carbide face246,247
. The
advantage of using silicon carbide as a substrate is that it is widely used in the electronics
industry.
Chemical Vapour Deposition (CVD) uses hydrocarbons such as methane, acetylene or
ethylene vapours at high temperatures on a variety of substrates. The most popular substrates
are metals, in particular copper248
, though semiconducting and insulating substrates have
been examined also with limited success. Particularly for deposition on metallic substrates,
the substrate needs to be removed for testing of electronic properties. This is done by
67
adhering a PMMA stamp over the graphene and etching away the metal in an acid249
. The
graphene can then be transferred to an arbitrary substrate. Domain size for CVD deposited
sample is also critical for producing the best possible electronic properties of the film250
. As
well as requiring good electronic properties, high area films are highly desirable. Bae et al.
produced 30 inch graphene by CVD using a roll to roll method which produced excellent
sheet resistances for the transparency251
.
Liquid phase exfoliation as it is applied in this thesis is described in the previous
chapter. The production of few layer and multilayer graphene in solution is a necessary step
for applications in printed electronics and for application in composites. However the lateral
dimension of the graphene layers is not significantly greater than a micron221
. Outside of the
methods included in this thesis there are other methods to exfoliate graphite in solution.
These are the oxidation of graphite to produce hydrophilic graphite oxide which produces
stable layers in aqueous media and graphite intercalation followed by exfoliation.
The oxidation of graphite into graphite oxide (GO) is commonly done by a modified
Hummers method52,252
. The graphite oxide can then be exfoliated using sonication, stirring or
thermal expansion. The graphene layers produced in this manner tend to have larger lateral
sizes than those produced using direct liquid exfoliation of graphite. These layers are heavily
oxidised and have much lower conductivity than graphene flakes54,55
. To combat this the
graphene oxide can be reduced chemically in solution with anhydrous hydrazine which also
prevents reaggregation55
. Thermal reduction in vacuum or under inert gas of films fabricated
from a GO dispersion is also possible54,253,254
.
Exfoliation of intercalated graphite compounds is facilitated by the increased distance
between the graphite layers due to the presence of the intercalation atoms/molecules. These
tend not to be stable in air and as such do not experience the popularity that the other liquid
phase methods enjoy255
.
The production of graphene at the large scale for applications can be split between the
formation of continuous graphene sheets and graphene flakes in dispersion. There is no one-
size-fits-all for application. Continuous graphene films see application as transparent and
flexible electrodes in a wide variety of devices. Doping of large area films can produce sheet
resistances of 30 Ω/sq. at 90% transparency251
. These continuous sheets can be used as
current collectors especially where transparency is required such as polymer and dye-
sensitized solar cells.
68
With excellent physical properties graphene is suitable for the reinforcement of
materials. Polyvinyl alcohol and polyurethane, among other polymers, can be cast from
composite dispersions of polymer stabilised graphene resulting in superior physical
properties230,256
. Similarly, conductive plastic composites can be made using a graphene filler
to provide conductivity257
.
High surface areas also require flakes of smaller sizes which can be produced by
liquid exfoliation. This is useful for applications where high surface areas are required, such
as supercapacitors. As such, graphene flakes have seen widespread application in thin film
and porous electrodes for supercapacitors139,140,145
and batteries258
. Also, due to the high
conductivity of graphene, the addition of graphene to less conductive but electrochemically
active materials to form enhanced composite electrodes is a common practice92,104,105,184,208
.
As well as having high surface areas the flakes also have edges. While the basal plane of
graphene is chemically inert, it has been shown that the edges are much more reactive46,48,259
.
This opens the possibility of using graphene as a catalyst and a sensor. While planar CVD
graphene can be used as a sensor, the presence of defects enhances the adsorption of gas,
allowing production of sensors from more economically produced graphene flakes260
. The
edges of graphene flakes also provide catalytic sites viable for the reduction of complexes in
DSSC electrolytes261
.
Graphene can also be functionalised, as demonstrated in the production of graphene
oxide. This functionalisation affects the chemical properties of the material as well as
improving the wettability of the surface262
. Graphene oxide has seen application in
supercapacitors due to the pseudocapacitance associated with these oxide groups as well as in
DSSCs. Beyond oxygen functionalisation, graphene can also be functionalised with nitrogen
groups. This can be achieved by introducing ammonia after the deposition of graphene by
CVD or additional chemistry on graphene oxide263,264
. There is a multitude of processes that
can be used to alter graphene and tune the properties of graphene for a range of
applications265
5.2 Carbon Nanotubes
The presence of carbon nanotubes (CNTs) were discovered among fullerenes by Iijima266
.
Fullerenes were produced by the arc discharge of graphite in the presence of a catalyst. In
69
comparison to the fullerenes the carbon nanotubes consisted of at least two concentric tubes
of rolled graphene. The spacing between layers is that of graphite. These tubes had high
aspect ratios with diameters of 4-40nm and lengths up to 1µm. The realisation of single
walled carbon nanotubes (SWCNTs) was reported in 1993 with a range of diameters from 0.4
to 3nm267
.
These tubes formed with a range of different chiralities. The chirality of a nanotube
corresponds to the orientation of the graphitic plane relative to the axis of the tube. The
chirality of the nanotube can be armchair, zigzag, or described by a vector of the lattice. The
chirality of the tube can affect properties including diameter and electronic properties with
one third of carbon nanotubes being conducting with the rest having a semiconductor
nature268
. Separating carbon nanotubes of different chiralities remains a challenge to
researchers.
The production of carbon nanotubes from the arc discharge method produces a large amount
of carbonaceous impurities and, as such, alternative methods to produce nanotubes in an
economically viable manner have been pursued. CVD is the most prominent method for mass
producing CNTs. Similar to the production of graphene the process involves the passing of
hydrocarbon vapours over a catalytic substrate. In the case of carbon nanotube formation, the
catalyst is in the form of particles for production of nanotubes, as opposed to being a flat foil
to promote growth of graphene269
. As such, a degree of control of length and diameter is
possible. The presence of a catalyst in the formation requires removal which can be achieved
by reflux in acids though this can reduce tube length and introduce surface groups on the
tube270,271
. As such, treatment of tubes should be done with great care.
Carbon nanotubes are available commercially in powder form, in which the tubes tend to
bundle via Van der Waals forces. To prepare nanotubes for applications they must be
individualised to access their excellent properties. Liquid phase exfoliation is an excellent
method for individualising tubes and allows for solution processing. Much of the work on
liquid exfoliation of graphene was a direct follow on from the dispersion of carbon nanotubes
as the material is similar. As such, dispersions of carbon nanotubes have been reported in
solvents272
, aqueous surfactant solutions273
and polymers274
including biopolymers such as
DNA275
.
Carbon nanotubes have a wide range of applications which can be achieved once exfoliated
in a liquid. Due to the excellent mechanical properties of carbon nanotubes the use a filler to
70
reinforce polymers is a common application276
. This can be done by direct exfoliation of
polymers and nanotubes in a solvent which is then cast and evaporated to form a film. For
other polymers melt processing is required to allow the introduction of nanotubes. The
creation of conductive composites is also possible277
.
Due to the excellent electrical properties of carbon nanotubes, networks of nanotubes can be
deposited from solution through a variety of methods to form conductive thin films. A carbon
nanotube randomly deposited from solution provides a conductive path across the film. The
variety of chiralities present as well as junctions between nanotubes means that the excellent
conductivities associated with individual metallic nanotubes cannot be realised, however
these films are still sufficiently conductive to be of interest. These networks are also
mechanically stable under bending and can be applied to flexible applications. Reducing
thickness leads to transparent electrodes, which could be of great importance in transparent
and flexible displays. At high transparencies however the percolation of the carbon nanotubes
has significant bearing on the film conductivity and presents challenges in producing films
that can achieve the requirements set by industry278
. Longer nanotubes are better for
conductivity and this is a problem as sonication induced scission occurs in dispersion
formation. To avoid this, dissolving nanotubes in chorosulfonic acid demonstrated by Mirri et
al. allowed a solution of carbon nanotubes with a length of 10µm to be produced. This
resulted in a sheet resistance of 140Ω/sq. at 88%279
. This is better than films produced from
sonicated dispersions278,280,281
. Nanotube films have been used in solar cells as transparent
current collectors compatible with the flexible organic solar cells.
Due to a high surface area, corresponding to one side of a graphene sheet (1315m2/g)
282
,carbon nanotubes see applications where high surface area is important. This is particularly
important for electrochemical systems such as supercapacitors134
and DSSCs64
, among others.
In many electrochemical systems carbon nanotubes are used as a conductive filler96,97,179,181
.
Due to a lower percolation threshold than graphene, effective conductivities can be reached at
lower volume fractions of nanotubes. This allows for more active material per electrode. As
such, composite electrodes have been produced for almost every electrochemical application.
The functionalisation of nanotubes can be achieved to allow dispersion in aqueous media and
to introduce other properties. The most common functionalization is oxidation of CNTs in
acid283
. The introduction of these oxygen functionalities allows hydrogen bonding, which
helps stabilize the nanotubes in water at the expense of the nanotubes electrical properties.
71
Many other functionalities exist and can be used to produce nanotubes effective in catalysis,
supercapacitors and batteries284
.
The application of nanotubes to electronics is an exciting field as the variety of electronic
properties available from nanotubes can allow nanotube-only logic elements. However, while
these applications have been developed, the reliably producing nanotubes of the correct
chirality in both sufficient amounts and with sufficient accuracy to enable wafer scale
development remains a significant challenge to researchers.
5.3 Molybdenum Disulfide (MoS2)
The discovery of the stability of graphene led to an increased interest in the possibility of
other two dimensional materials. Transition metal dichalcogenides (TMDs) were known to
have a structure of covalently bonded MX2 planes. Where the M represents the transition
metal and the X represents the chalcogenide. The planes have the transition metal layer
sandwiched between two layers of chalcogenide. These layers can have different symmetries
namely 1T, 2H and 3R. Monolayer TMD’s only have two symmetry groups 1T and 1H
reflecting the triangular and octahedral morphologies of the layers285
.
Figure 5.2 top-down, side and individual view of 1T TMD structure (left), Top-down, side and individual view of 1H TMD structure (right). Image adapted from source285.
72
Of primary interest in this thesis is Molybdenum Disulphide (MoS2). MoS2 in the 2H
symmetry is naturally occurring which is advantageous for liquid phase exfoliation as the
starting material does not need to be synthesised. MoS2 in the bulk state is an indirect-
bandgap semiconductor of 1.29eV. As the material approaches monolayer however, the band
structure undergoes changes such that it eventually becomes a direct-bandgap semiconductor
of 1.8eV286
. This leads to the possibility of using MoS2 in optoelectronic devices.
Naturally occurring MoS2 has long been subjected to intercalation chemistry and the
discovery of graphene led to renewed interest in these intercalation compounds287
. As such,
intercalation proved to be helpful in exfoliating this material in a liquid environment. The
stabilisation of the 1T phase is also possible using this method288
. The 1T symmetry displays
a higher conductivity than the 1H symmetry and as such is of interest for devices that require
high conductivities289
. Due to difficulties processing intercalated compounds, methods to
exfoliate MoS2 similar to the way graphene has been exfoliated have been developed224,290,291
.
In fact the surface tension range of solvents that exfoliate graphene and MoS2 overlap due to
the similar surface energies of the materials290,292
.
Liquid phase exfoliation, as mentioned earlier, is ideal where large quantities of
nanostructured material is required. Due to the ability to form intercalation compounds, MoS2
has been integrated into lithium ion batteries293
and supercapacitors167
. MoS2 is not as
chemically inert as graphene (though as with graphene the chemical activity derives from the
edges) and, as such, it has seen application in electrochemical catalytic systems, such as
DSSCs85
, the hydrogen evolution reaction294
and gas sensors295
. Reinforcement of polymers
is also possible but graphene is both cheaper and a better reinforcement material291
. Due to
the band-gap being in the visible spectrum MoS2 films can be used as light detectors due to
the photoconductivity effect296,297
.
MoS2 can also be synthesized via a solvothermal method. While reports of MoS2 nanosheets
produced by this method exist, they tend to have small lateral dimensions298,299
. As such, the
sheets would be equally well described as quantum dots. Much of the synthesis of MoS2
using the solvothermal route uses another material such as GO100
or TiO2300
to provide anchor
sites to promote growth.
73
For electronic applications, prototype devices can be fabricated using the scotch-tape method.
In order to provide suitable material of sufficient quantity, a synthetic method is required.
Physical and chemical vapour deposition is used to produce MoS2 from a variety of
Molybdenum precursors such as elemental molybdenum301
, (NH4)2MoS4302
and Molybdenum
Oxides296,303,304
. A traditional CVD process using vaporised MoCl5 with H2S has also
produced MoS2 on a variety of substrates305
. These materials can be used to fabricate
transistors302,303
, photo-transistors296
, photovoltaics306
and gas sensors301
.
5.4 Poly(3,4-ethylenedioxythiophene), PEDOT
Conductive polymers became the focus of intense research in the 1980s after the synthesis of
polyacetylene by Shirakawa et al307
. The potential for applications of these materials was
limited by the poor environmental stability and difficulties in processing and film formation.
Poly(3,4-ethylenedioxythiophene), or PEDOT, is a polymer that displays significant
conductivity and is sufficiently stable in air to be utilised in a range of applications.
The synthesis of PEDOT is commonly done using either an electrochemical polymerisation
or a chemical polymerisation of the EDOT monomer. Electrochemical polymerisation can be
used to form films on substrates308
. The requirement for the substrate as well as the scalability
of electrochemical deposition limits the practicality of electrophoretically produced PEDOT.
As such, for producing PEDOT in sufficient volume, chemical polymerisation is the most
efficient synthetic route.
Chemical polymerisation involves introducing the EDOT monomer to an oxidizing agent.
This is commonly either an Iron (III) compound (FeCl3 or Fe(OTs)3) or Na2S2O8309,310
.
PEDOT is initially formed in an undoped state which has a bandgap of 1.5eV and strong light
absorption in the visible range. Further doping shifts the maximum absorption into the infra-
red, producing a transparent blue material at low thicknesses and opening up the possibility
for transparent electrical applications311
. The conductivity of the undoped state is also
significantly lower (<10-5
S/cm)311
while doped PEDOT has a conductivity in the range of 1-
1000S/cm312,313
. The PEDOT formed by chemical polymerisation forms short chained
polymer cations with approximately 2 positive charges per 6 monomers.
74
Figure 5.3 Structure of PEDOT and PSS
This leads to the PEDOT being insoluble in practical solvents. To circumvent this, the
monomer/oxidant solution can be applied directly to the substrate314
or the oxidant can be
applied to the substrate and then be exposed to an EDOT vapour315
.
Liquid processing allows a range of deposition methods to be used and also the material to be
mass produced for commercial use. To produce a dispersed material, poly(strenesulfonate)
(PSS) is added to the reaction mixture316
. The deprotonated PSS molecule acts as a counter
ion to which the PEDOT can adhere to. PSS is water soluble and forms globules with
PEDOT on the inside and the PSS in contact with the water. It is worth noting that the
PEDOT polymer chain is much shorter than that of the PSS polymer chain, allowing the PSS
to envelope the PEDOT. This copolymer PEDOT:PSS is stable in water, with globules of size
less than 100nm and is commercially available from Heraeus GmBH with a PSS to PEDOT
mass ratio of 2.5317
.
75
Figure 5.4 PEDOT in a PSS chain as formed and stable in water (top) PEDOT is depicted as thicker lines while the PSS is longer and thinner, after secondary doping the PEDOT is
liberated from the PSS globule allowing for improved conductivity (bottom) image adapted from source313.
The resultant PEDOT:PSS films have low conductivities due to the presence of the insulating
PSS and the tendency for the PEDOT chains to be enclosed in the PSS chains. The
conductivity is usually lower than 1S/cm which is not suitable for many applications. To
further improve the electrical conductivity of these films, doping these films is attempted.
Since the PEDOT in these films is already in a doped state, this process is referred to as
secondary doping.
Addition of compounds to aqueous dispersions of PEDOT:PSS can be used to enhance the
conductivity of the resultant films. This is usually done by altering the aqueous medium in
such a way that the hydrophobic PEDOT can emerge from the PSS globule. Surfactants and
Ionic liquids313,318
have been used but the most effective additives tend to be high boiling
points solvents such as DMSO, DMF, THF319
, EG312
and NMP320
. These high boiling point
solvents alter the energetics of the liquid medium and as the water evaporates during film
formation the emergence of the PEDOT from the PSS becomes more energetically
favourable.
Treatment of films after deposition is an alternative method. The mechanism of these
treatments is a phase segregation of the PEDOT and the PSS.313
This can be achieved with
the high boiling point solvents mentioned previously as dispersion additives. Salt solutions
have also been used but the conductivity enhancement is modest and the migration of ions
presents an issue321
. Methanol and other alcohols have also been demonstrated as suitable
dopants of PEDOT:PSS films322
. A range of organic and inorganic acids have also been used
to enhance the conductivity323
. Repeated treatments in sulfuric acid at 160⁰C result in a
conductivity of 3065S/cm324
.
76
Due to the high transparency and conductivity of PEDOT thin films, the most obvious
application is as a transparent conductor. Formic acid treatment, as done by McCarthy et al
and used extensively in this thesis, produced a transparent conducting film with a sheet
resistance of 80Ω/sq. at a transparency of 90%, well within the requirements of industry325
.
The conductivity was approximately 900S/cm The sulfuric acid treatment mentioned
previously produced a film with 67Ω/sq. at 90% transparency324
These transparent films see application in solar cells such as polymer solar cells324
and
DSSCs67
. For DSSCs, this material can double as a current collector and a catalyst for the
electrolyte couple. In polymer solar cells, they can act as the current collector and a hole
injection layer. PEDOT has also found applications in many electrochemical devices. Even
had it not significant electrochemical activity, it would have found application as an
alternative to the insulating polymer binders used in many devices. The electrochemical
properties of PEDOT allow charge transfer to control the doping of the film. Doping of the
monomer site occurs according to the following equation:
CSSEDOTCeSSEDOT nn
0 Equation 5.1
Here SS is the monomer styrenesulfonate and C+ is the cation. The cation allows for charge
transfer between the electrode and the electrolyte allowing donation of an electron to the
PEDOT monomer. When the EDOT monomer carries no charge (as in the right hand side of
the equation) it is said to be undoped and is insulating310
. This allows PEDOT to be active as
a pseudocapacitor. The specific capacitance of PEDOT is over 100F/g and, as such, PEDOT
has been used as both the sole electrode material and in a variety of composites326
. The
undoped PEDOT film also absorbs more strongly in the visible and, as a result,
electrochromic devices can be fabricated using PEDOT311
.
77
6
Graphene Based DSSC Counter Electrodes
6.1 Introduction
In 1990 the Gratzel cell or Dye-Sensitized Solar Cell (DSSC) was established as a potential
electrochemical solar cell10
. While the efficiency was relatively low, this cell type displayed
considerable advantages in that it was potentially cheap to manufacture and did not require
the material purity associated with silicon based solar cells. Also the cells could be made in
normal ambient conditions and did not require clean room facilities. Dye-Sensitized Solar
Cells all employ a counter electrode which operates as a catalyst for the tri-iodide couple in
the following reaction327
:
IeI 323 Equation 6.1
As with many electrochemical reactions, platinum is the most commonly used and
effective catalysts for this reaction. However, for a cell type that stands out for its low cost
and ease of manufacture, the platinum counter electrode has always been problematic. The
high price of platinum has encouraged much research into alternative materials for use as
catalytic counter electrodes in DSSCs328
.
Much of this work has focused on producing carbon-based counter electrodes. This
interest is due to the high surface area and conductivity of the various forms of carbon such
as activated carbon and carbon aerogels329
, carbon nanotubes65
and graphene57,58,61,330–334
.
With graphene the catalytic activity is associated with active sites on the nanosheet edges47,48
.
The simplest way to maximise the overall catalytic performance of a graphene counter
electrode is to increase the number of catalytically active sites by either reducing the
nanosheet size (i.e. increasing total edge length) or increasing the electrode thickness. While
78
the second approach is relatively simple, above some critical thickness the overall catalytic
activity will saturate as the process becomes limited by mass and charge transport effects335
.
In addition, from a purely economic standpoint, one would like to use the minimum graphene
mass possible. The importance of thickness in these studies is often overlooked. While many
of these carbon electrodes have been in excess of a micron in thickness, Kavan et al. have
produced thin catalytic graphene films with high transparencies50
. However, what is really
needed is a comprehensive study on the dependence of DSSC cell performance on graphene
counter electrode thickness.
Addition of carbon nanotubes to enhance the efficiency of graphene counter
electrodes is a way of both increasing the internal porosity of the film allowing diffusion of
the electrolyte ions into the internal surface of the film and of improving the conductivity of
the film. This has been done with both reduced graphene oxide and pristine graphene with
reasonable effectiveness90,336
.
In addition to purely carbon counter electrodes, a number of other materials have been
studied. For example polymers, in particular PEDOT:PSS have been demonstrated as
catalysts337
, as well as transition metal compounds usually with oxygen, carbon, sulfur and
other chalcogenides. Many of these compounds are produced as particles with dimensions
ranging from a couple of hundred nanometers to microns81,83,84,338
. In addition, counter
electrodes from arrays of two dimensional (2D) transition metal compounds are being
explored such as molybdenum and tungsten sulfide85
, molybdenum selenide339
, tin sulfide86
and cobalt sulfide70
.
However, many of the more effective 2D catalysts are not good conductors of
electricity285
. This means the cell performance may be limited by the resistance of the counter
electrode, especially for thicker electrodes. To address this problem, a number of researchers
have demonstrated hybrid systems composed of a combination of conductive component,
usually a carbon allotrope or polymer, and an electrocatalytically active material89,92,105,340–342
.
Of particular interest are counter electrodes formed from MoS2/Graphene hybrids. Most of
the work done using MoS2 for counter electrodes has followed a synthetic route for the
production of material which is then combined with pristine graphene104,343
. Alternatively,
reduced graphene oxide has been used with the functional groups on Graphene Oxide acting
as a site for synthesizing the MoS299,100
. However, in many cases these processing routes are
overly complex and not in keeping with the goal of low cost, easily processed cells. This
79
could be addressed by using a simpler route to produce MoS2/graphene composite films. I
believe Liquid Phase Exfoliation (LPE) represents such a route.
Liquid Phase Exfoliation is a very simple and scalable method for exfoliating layered
crystals such as graphite, h-BN or MoS2 to give large quantities of nanosheets stably
suspended in appropriate liquids219,290,292
. The resultant nanosheets have been shown to have
virtually no basal plain defects introduced during the exfoliation process. The only defects are
due to edge sites which tend to be electrochemically active50,344
. Using these methods, it is
simple to produce mixed dispersions of different types of nanosheets (e.g. graphene and
MoS2) or even nanosheets and nanotubes179,290,345
. To our knowledge very little work has
been done on MoS2 counter electrodes prepared by liquid phase exfoliation. While this may
be due to the poor performance of MoS2 nanosheets relative to the smaller synthesized sheets
and nanoparticles87
, the inherent processability of LPE nanosheets should allow significant
performance enhancement via electrode optimisation.
In this work, we used LPE graphene nanosheet dispersions to produce counter
electrodes with a range of thicknesses. This allowed us to identify 400 nm as the optimum
electrode. We then further improved the counter electrode performance by adding small
amounts of carbon nanotubes or MoS2, resulting in a DSSCs with efficiency of up to 96% of
the efficiency of the equivalent platinum cell. With a significant contribution from the MoS2
flakes being due to the relative flake dimensions a further study on graphene flakes of
varying length was conducted to further probe the effect of flake size on efficiency.
6.2 Experimental Procedure
6.2.1 Materials
Graphite was purchased from Future Carbon GmBH. Molybdenum Disulfide, Ethyl Cellulose
and Isopropyl Alcohol was purchased from Sigma Aldrich. P3 Carbon Nanotubes were
purchased from Carbon Solutions Inc. Titania Pastes, a conductive substrate (fluorine doped
tin oxide on glass), Pt Counter Electrodes and Electrolyte Materials were purchased from
Dyesol.
6.2.2 Film Production
80
Dispersions of materials were prepared by probe sonicating, using a GEX600, 48 W, 24 kHz,
flat head probe, 8g of material in 80ml of 10mg/ml Ethyl Cellulose in Isopropyl Alcohol for
48 hours. Dispersions were then centrifuged in a Hettich Mikro 22R at 2000 rpm for 90
minutes. Concentrations were obtained by filtration of known volumes followed by washing
with isopropanol and obtaining the final mass.
Dropcast films were made by dropping known volumes of dispersion on cleaned 2cm ×
2.5cm FTO glass sheets. Different thickness were achieved by changing the deposited
volume. Having to make new films for each measurement resulted in a variation in the
thickness range for electrochemical testing and cell assembly. Spin coated films were
fabricated by droping 200 µL on a cleaned FTO glass and spinning at 1000 rpm. Different
thicknesses were achieved by changing concentration by dilution with isopropyl alcohol for
thinner films or by repeating the process for thicker films. Spraycoated films were fabricated
using a Harddner and Steenbeck Infinity airbrush spray system operated by a Janome
JR2300N robot. The dispersion was diluted to approximately 0.1mg/ml and sprayed onto
cleaned FTO glass at 130⁰C. Compositions were altered by mixing the two dispersions.
Once the films were deposited by dropcasting they were heated in air at 370⁰C for 20
minutes. Thicknesses were measured using a Dektak 3 Surface Profilometer.
Special case for flake size dependence:
Dispersions of materials were prepared by probe sonicating, using a GEX600, 48 W, 24 kHz,
flat head probe, 0.8g of material in 80ml of 1mg/ml Sodium Cholate in deionised water for 48
hours. Dispersions were then centrifuged in a Hettich Mikro 22R at a 12krpm for 90 minutes
the supernatant was collected for film production while the sediment was redispersed and
centrifuged at progressively lower spin rates to achieve different flake sizes. Concentrations
were obtained by filtration of known volumes followed by washing with deionised water and
obtaining the final mass.
The same mass of each dispersion was filtered and washed to obtain films of similar mass
(0.1mg/cm2) on ethyl cellulose membranes for comparison for the various flake sizes. The
films were then transferred to FTO glass sheets
81
0 1000 2000 3000
0
2000
4000
He
igh
t (n
m)
Distance (m)
(b)(a)
Figure 6.1 (a)dropcast graphene counter electrode (b) a sample profiometry scan exhibiting surface morphology
6.2.3 Electrochemical Charaterisation
All electrochemical measurements were carried out using a Gammry 3000 Potentiostat.
Electrodes were confined to an area of 1 cm2 using Surlyn 15µm sealant. Cyclic
Voltammograms were measured in a three electrode setup using a Pt counter electrode and an
Ag/AgCl reference electrode at a scan rate of 50mV/s from -0.2 to 1 V vs. Ag/AgCl. The
electrolyte used was 0.1 M LiClO4, 5 mM LiI, and 0.5 mM I2 in acetonitrile. For Linear
Sweep and EIS measurements symmetrical cells were prepared using the High Performance
Electrolyte provided by Dyesol. Linear Sweep scan range was from -1 to 1 V vs. Ag/AgCl.
EIS was measured at 0.5V vs. Ag/AgCl at 10mV amplitude from 1MHz to 0.1Hz.
6.2.4 Cell Production
FTO glass was cleaned by bath sonication for 15mins in Decon 90 solution, deionised water,
Acetone and stored in Isopropyl alcohol prior to use. FTO glass was immersed in 40 mM
solution of TiCl4 for 30 minutes at 70⁰C prior to screen printing 3 layers of approximately 4
µm of 90T transparent paste followed by 1 layer of WER2-O reflector paste. These were then
sintered in a furnace ramping up to temperatures of 500⁰C over 2 hours as described by Ito et
al.13
82
Figure 6.2 TiO2 photoanodes from left to right: As screen printed, the scattering layer side face up after 24 hours in dye, the transparent layer side face up after 24 hours in dye.
Cells were assembled as open cells separated by 15 nm Surlyn sealant. The working electrode
was immersed in the High Performance Electrolyte provided by Dyesol while the counter
electrode was clipped on to complete the cell. For comparison, the same working electrode
was used for a Platinum counter electrode and a Graphene/MoS2 counter electrode. Electrical
Characteristics of the cells were obtained using a Keithley 2400 Source Meter.
6.3 Results and Discussion
6.3.1 Graphene Film Thickness Dependence
0 300 600 900 12000
5
10
15
20
25
30<L> = 586nm
Count
Length (nm)
(c)(b)(a)
Figure 6.3(a) TEM of Typical Graphene Flake. Scale Bar = 100nm. (b) Histogram from TEM data (c) SEM of Graphene Film. Scale bar = 500nm
LPE was used to produce suspensions of graphene nanosheets in solutions of ethyl cellulose
in isopropyl alcohol (see methods). Shown in Fig 6.6(a) is a transmission electron
83
microscopy (TEM) image of a typical few-layer graphene nanosheet. Statistical TEM
analysis yielded a nanosheet size (L) histogram as shown in Fig 6.3(b) which gave an average
flake size of 586nm these were typically few layer flakes. The larger flake size was used as
there was difficulty in producing sufficient amounts of the smaller nanosheets. However, due
to there only being one centrifugation step the lower size nanosheets are present in this more
polydisperse sample. These nanosheet suspensions were formed into films by dropcasting
onto glass substrates. Shown in Fig 6.3(c) is a scanning electron microscope image of the
surface of such a film. This shows a surface morphology which is rough and porous,
properties which will facilitate electrolyte penetration and access of ions to catalytic sites in
the network interior. This makes these films suitable for use as counter electrodes.
In order to observe the electrochemical activity at different relative potentials, cyclic
voltammetry (CV) was performed for both a 400 nm thick graphene counter electrode
produced by the dropcast method and, for comparison, a platinum counter electrode in an
electrolyte composed of I3-/I
- species in acetonitrile. The pairs of peaks in Figure 6.4(a)
between 0 and 0.4 Volts vs. Ag/AgCl corresponds to the redox of I3-/I
- while the pairs of
peaks between 0.5 and 0.9V vs. Ag/AgCl are the redox couple of I2/I3-.81
In figure 6.4(a) the
voltage separation between the peaks for the graphene electrode is 0.17 V which is smaller
than the platinum separation of 0.24 V. This provides a slight performance advantage to the
graphene counter electrode. However the current densities at the graphene peaks associated
with the I3-/I
- couple are low relative to those of platinum, particularly for the reduction of I3
-,
which will have a negative impact on the graphene counter electrodes performance.
Dropcast graphene electrodes of various thickness (t=85-1100 nm) were fabricated into
symmetric cells using a high performance electrolyte provided by Dyesol and characterised
using linear sweep voltammograms (Figure 6.4(b)). These were compared to a platinum
electrode which is reflected as the shaded area. In each case, the exchange current density
was extracted by fitting the data around -1V to the Tafel equation114
. The exchange current
density (Figure 6.4(c)) increases linearly with electrode thickness from 2 mA/cm2 (t=85 nm)
to 19 mA/cm2 for a thickness of 1100nm. This is a reflection of the expected linear increase
in the number of catalytic sites with electrode thickness and indicates good electrolyte
penetration throughout the internal volume of the electrode. Importantly, by a thickness of
400 nm, the exchange current density surpasses that of platinum electrodes (~8.5 mA/cm2).
84
Electrochemical impedence spectroscopy (EIS) was performed also as shown in Figure 6.4(d)
for electrodes of varying graphene thickness as well as Pt. In all cases, the curves were
dominated by a single semicircle indicating a simplified Randel’s circuit239
, allowing us to
estimate the charge transfer resistance from the circle width along the ZReal axis. The charge
transfer resistances of all electrodes is plotted versus electrode thickness in Figure 6.4(e). We
find an inverse fall off with thickness from 19 (t=85 nm) to <1 for a thickness of
1100nm. This compares to a charge transfer resistance of ~6 Ω for the platinum electrode.
We find the charge transfer resistance to scale inversely with the exchange current density in
Figure 6.4(f) as expected from the low potential approximation to the Butler-Volmer
equation. This confirms the charge transfer resistance to be controlled by the number of
catalytically active sites.
85
-0.3 0.0 0.3 0.6 0.9 1.2-0.5
0.0
0.5
1.0
1.5
(f)(c)
(e)(b)
(d)
J (
mA
/cm
2)
Voltage (V vs. Ag/AgCl)
400nm Graphene
Pt
(a)
-1 0 1-6
-4
-2
Pt
85nm
215nm
400nm
1100nm Lo
g (
J/m
Acm
-2)
Voltage (V)
0 300 600 900 1200
0
5
10
15
20 Graphene
Pt
J0 (
mA
/cm
2)
Thickness, t (nm)
20 30 40 50 60 70
-2
0
2
4
6
8
10
12
14
16 Pt
85nm
215nm
400nm
1100nm
ZIm
(
)
ZRe
()
10 100 1000
1
10
100
Rct (
)
Thickness, t (nm)
0 1 2 3
0
50
100
Rct (
)
1/J0 (cm
2/mA)
Figure 6.4(a) Cyclic voltammograms measured for counter electrodes of both platinum and a 400 nm thick graphene film. (b) Tafel Plots measured for counter electrodes of platinum
(shaded area) and dropcast graphene films of varying thickness. (c) Exchange current density data (extracted from Tafel Plots) for graphene counter electrodes of a range of thickness. (d)
Electrochemical impedance spectra measured for counter electrodes of platinum and dropcast graphene films of varying thickness. (e) Charge transfer resistance plotted versus
film thickness. (f) Charge transfer resistance plotted against the inverse of the exchange current density. In (c), (e) and (f) the red square represents the datum for a Pt electrode.
86
0.0 0.2 0.4 0.60
-2
-4
-6
-8
-10
-12
-14
-16
Pt 5.52%
60nm Gra 2.72%
200nm Gra 3.82%
900nm Gra 4.28%
Curr
ent D
ensity (
mA
/cm
2)
Voltage (V)
Figure 6.5 IV Characteristics of DSSCs with graphene counter electrodes of different thickness. Derived parameters are shown in table 1.
We used graphene films of a range of thickness as counter electrodes in DSSCs (see
methods). Current density-voltage (J-V) curves for cells with different graphene electrode
thicknesses are shown in Fig 6.5. Shown for comparison is data for a cell with a standard Pt
counter electrode. It is clear from these curves that DSSC performance improves as the
graphene thickness is increased (device parameters are given in Table 6.1).
Sample t (nm) Jsc (mA/cm2) Voc (V) FF (%) Eff (%)
Graphene 60 11.66 0.626 37.3 2.72
Graphene 200 12.63 0.655 46.2 3.82
Graphene 900 12.59 0.659 51.6 4.28
Platinum n/a 14.06 0.653 60.1 5.52
Table 6.1: Table containing cell parameters for counter electrodes of varying thickness (t) -
short circuit current (Jsc), open circuit voltage (Voc), fill factor (FF) and efficiency (Eff)
To test this thickness dependence in more detail, we fabricated cells with graphene counter
electrodes which were deposited at a range of thicknesses by three different methods:
87
dropcasting (t = 60nm – 900nm), spin-coating (t = 15nm – 360nm) and spray-coating (t =
80nm – 420nm). We measured the J-V curves in all cases and extracted Jsc, Voc, FF and Eff .
70
80
90
100
(d)
(c)
(b)
Dropcast
Spin Coated
Spray Coated
Jsc/J
sc,P
t (%
)
(a)
0 200 400 600 800 100094
96
98
100
102
Vo
c/V
oc,P
t(%
)
50
60
70
80
90
100
FF
/FF
Pt (
%)
0 200 400 600 800 100040
50
60
70
80
Eff/E
ffP
t (%
)
Thickness (nm)
Figure 6.6(a) Short current density, (b) open circuit voltage, (c) fill factor and (d) efficiency for DSSCs fabricated with graphene counter electrodes plotted as a function of graphene film
thickness. Each parameter is expressed relative to that measured for an equivalent cell with a Platinum counter electrode.
The parameters are plotted against thickness in Fig 6.6(a)-(d). In all cases, the values are
normalised to the values measured for the Pt counter electrode. It is clear from the graphs that
there is little difference between dropcasting, spin-coating and spray-coating especially at the
higher range of thicknesses. For most samples it can be seen that the normalised short circuit
current density of the graphene electrodes is at 80-90% of the platinum electrode with little
dependence on thickness. Increasing the thickness of the graphene layer improves the
normalised open circuit voltage from 95% to 101% of platinum electrode. The most notable
increase however is in the normalised fill factor which increases from ~60% to ~85% as the
88
thickness increases from 15 to 900 nm. Primarily because of the increase in FF, the
normalised efficiency increases from ~50% to ~80% (relative to Pt) over the thickness range.
We note that there is a rapid increase in efficiency as the thickness increases to 400 nm, after
which minimal increases occur. Since the purpose of integrating graphene into the cell is to
be cost competitive, we suggest that 400 nm is the optimum counter electrode thickness.
6.3.2 Addition of Carbon Nanotubes
Carbon nanotube have previously been used to enhance the conductivity and porosity of
various graphene counter electrodes. The SEM image in figure 6.3(a) shows that the
graphene flakes are orientated close to parallel to the substrate plane. This means in the
vertical direction towards the current collector in the cell there are more junctions per micron
than there are parallel to the current collector. The carbon nanotubes can reduce this
anisotropy by weaving between the flakes facilitating contact to the current collector.
To observe the effect of addition of carbon nanotubes on the electrochemical properties of the
graphene film, EIS spectroscopy analysis was performed to extract estimates for the charge
transfer resistance as shown in figure 6.7.
10 20 30 40 500
5
10
15
20(b)
-Zim
()
Zreal
()
0% CNT
2%
5%
10%
15%
100%
(a)
0 20 40 60 80 100
4
6
8
10
12
14
16
Rct (
)
% weight CNT (%)
Figure 6.7 (a) Nyquist plot of impedance for graphene-CNT composite films (b) estimated charge transfer resistance of graphene-CNT composite films
The Nyquist plot in figure 6.7(a) shows a graphene plot similar to the characteristic Randle’s
circuit (black). The carbon nanotube only sample (navy) displays two characteristic semi-
circles imposed upon each other with the centres of the circles being displaced relative to
each other. The charge transfer resistance was estimated by taking the distance between the
start of the first semicircle and the end of the second semicircle. Even a small amount of
89
addition of the carbon nanotube results in a plot resembling the carbon nanotube plot more
closely.
There is minimal effect of the addition of carbon nanotubes on the series resistance of the
films which would be dominated by the current collector. The decrease in charge transfer
resistance as a result of the addition of carbon nanotubes is non-trivial with the charge
transfer resistance of the graphene only film being over 15Ω while addition of 15% carbon
nanotubes reduces this to 3Ω which is lower than that of the carbon nanotube only film of
approximately 6Ω. This could be due to the graphene having more available active sites than
the CNTs, as a result the improved conductivity of the CNTs provides better access to the
electrochemically active sites reducing the charge transfer resistance.
Due to the expense of carbon nanotubes, their addition at large amounts is not economically
feasible currently. To reduce the economic impact of introducing CNTs analysis of the cells a
carbon nanotube weight percent of 5% was chosen as this resulted in a significant decrease in
charge transfer resistance. Two 4µm thick films were compared in figure 6.8 one with 5%
carbon nanotube added and one without.
0.0 0.2 0.4 0.60
-2
-4
-6
-8
-10
-12
-14
Pt = 5.23%
Gra 4m =4.59%
Gra+ 5 CNT 4m =5.03%
Curr
ent D
ensity (
mA
/cm
2)
Voltage (V)
Figure 6.8 IV Characteristics of DSSCs with a graphene counter electrode with a thickness of 4μm (red) and a graphene counter electrode of same thickness with 5% weight carbon
nanotubes (blue) compared to a DSSC with a platinum counter electrode (black). Derived parameters are shown in table 3.
The characteristics of the cells show that there is an increase in short circuit current from
11.03 to 12.18mA/cm2. There is also a small increase of 0.01V in the open circuit voltage due
90
to the addition of the nanotubes. The fill factor is larger for the graphene only film but due to
the significantly higher current the efficiency of the film with the nanotubes is higher. The
efficiencies represented as percentages of the platinum cells are 87.6% for the graphene only
cell and 96.2% for the addition of the carbon nanotubes. This is very close to being
competitive with platinum. However, there is much more mass of graphene/CNT used and
with CNTs costing at least 3 times as much as platinum using conventional methods these
electrodes are not as economically viable.
Sample Jsc (mA/cm2) Voc (V) FF (%) Eff (%)
Graphene 11.03 0.62 67.0 4.58
Graphene:CNT 95:5 12.18 0.63 65.6 5.03
Platinum 13.29 0.6 65.6 5.23
Table 6.2: Table containing cell parameters for counter electrodes of graphene, graphene with
5%wt CNTs and Platinum - short circuit current (Jsc), open circuit voltage (Voc), fill factor
(FF) and efficiency (Eff)
6.3.3 Addition of MoS2
A number of papers have suggested MoS2 nanosheets to be effective catalysts for use in
DSSC counter electrodes with evidence suggesting that they are more effective than
graphene-based catalysts.328
However, networks of MoS2 nanosheets are poor electrical
conductors displaying in-plane conductivities227
of ~10-6
S/m and out of plane conductivities
which are three orders of magnitude lower.297
Thus we would expect counter electrodes
fabricated from networks of MoS2 nanosheets to be limited by their electrical properties. This
implies that mixtures of graphene and MoS2 nanosheets could make effective counter
electrodes. One would expect that while MoS2 might contribute more to the catalytic activity,
graphene would dominate the charge transport. Similar strategies have recently proved
successful for nanosheet-based supercapacitors,179
lithium ion batteries346
and hydrogen
evolution catalysts347
where addition of nanotubes has overcome resistance limitations,
boosting performance.
91
To test this, we prepared MoS2 nanosheets by liquid phase exfoliation. Shown in figure 6.9(a)
is a TEM image of a typical MoS2 nanosheet. Statistical analysis yielded histograms which
gave an average flake size of 156 nm (Fig 6.9(a)). Shown in figure 6.9(c) is an SEM of a
dropcast MoS2 film on glass. Clearly the network morphology is very similar to the graphene
networks.
The MoS2 and graphene suspensions were then mixed in a number of different ratios to give a
set of composite dispersions. These were then formed into MoS2/graphene composite films
by drop-casting for use as counter electrodes (thickness 400 nm in all cases).
0 200 400 6000
10
20
30
40
Count
Length (nm)
<L>= 156nm (c)(b)(a)
Figure 6.9 (a) TEM of Typical MoS2 Flake. Scale bar = 100nm. (b) MoS2 nanosheet size histogram extracted from TEM data. (c) SEM of MoS2 Film. Scale bar = 500nm
To investigate the effect of addition of MoS2 on the electrochemical properties of the
graphene film EIS spectroscopy analysis was performed to extract estimates for the charge
transfer resistance as shown in figure 6.10.
92
10 20 30 40 50 60 70-2
0
2
4
6
8
10
12
14 Gra
10% MoS2
75%MoS2
-Zim
g (
Zreal
(0 20 40 60 80
12
14
16
18
20
22
(b)
Rct (
)
% Weight MoS2 (%)
(a)
Figure 6.10 Charge transfer resistance of graphene-MoS2 composite films
The samples shown in figure 6.10(a) compare the Nyquist plots of a graphene only
symmetrical cell with those of a cell with electrodes with 10% and 75% weight MoS2
respectively. The MoS2 plot is not shown as the resistances were approximately two orders of
magnitude higher. It is sufficient to say that the graphene has a significant impact on the
charge resistance of the MoS2 as a result.
In figure 6.10(b) the charge transfer resistances are compared for films with different weight
percentages of MoS2. While initially the charge transfer resistance is higher than graphene for
small amounts of MoS2 it decreases for further addition of MoS2. 75% weight MoS2 gives a
charge transfer resistance of approximately 12Ω compared to 16Ω for the graphene electrode.
This could be due to the enhanced electrochemical activity of the semiconducting MoS2
benefiting from the enhanced charge transport in the composite provided by the graphene
To observe the effect of charge transfer resistance on cell performance JV curves were
measured for DSSCs prepared using such composite counter electrodes with a subset shown
in figure 6.11. It is clear from this data that the DSSC performance varies strongly with the
compositional balance of MoS2 and graphene (device parameters are given in table 6.3).
93
0.0 0.2 0.4 0.60
-2
-4
-6
-8
-10
-12
-14
Pt 4.40%
Gra 3.62%
90% Gra 4.35%
MoS2 2.53%
Curr
en
t D
en
sity (
mA
/cm
2)
Voltage (V)
Figure 6.11 IV Characteristics of DSSCs with MoS2/graphene composite counter electrodes of different compositions. Derived parameters are shown in table 2.
Sample Jsc (mA/cm2) Voc (V) FF (%) Eff (%)
Graphene 10.7 0.652 51.9 3.62
MoS2 9.14 0.589 47.0 2.53
90:10 Gra:MoS2 11.91 0.646 56.5 4.35
Platinum 13.39 0.657 50.0 4.40
Table 6.3: Table containing cell parameters for counter electrodes of varying composition -
short circuit current (Jsc), open circuit voltage (Voc), fill factor (FF) and efficiency (Eff)
To quantify this trend, the DSSC characteristics were extracted from the JV curves as
described above, normalised to the equivalent value measured for a Pt electrode-DSSC and
displayed in Fig 6.12 (a)-(d) as a function of graphene mass fraction. In all cases we find the
MoS2-only DSSCs to perform poorly with values of Jsc, Voc, FF, and Eff considerably lower
than the equivalent values for devices with graphene-only counter electrodes. The cumulative
effect of this can be seen in the normalised efficiency which was <45% for the MoS2-only
94
device compared to >75% for the graphene-only device. For both Jsc and Voc, the normalised
performance increased monotonically with graphene content in a manner roughly described
by the rule of mixtures (Figure 6.12 (a)-(b)). However, the normalised Fill Factor behaves
differently, increasing from ~75% for the MoS2-only device to ~115% for the counter
electrode containing 90% graphene before falling to approximately 80% for the graphene
only device (Figure 6.12 (c)). As a result of the behaviour of the FF, the normalised
efficiency also displays non-monotonic behaviour, rising from 40% to 95% for the counter
electrode containing 90% graphene before falling to <80% for the graphene only device
(Figure 6.12(d)). This data clearly shows that adding small amounts of MoS2 nanosheets
(~10wt%) to a graphene counter electrode can give a non-trivial increase in efficiency. With
regard to the measurements from EIS this seems to be at odds with the higher percentage
weight films displaying a lower charge transfer resistance.
We can understand these results in a phenomenological manner as follows. We assume that
MoS2 is a more effective catalyst than graphene. In the simplest case the efficiency of a
MoS2/graphene mixed electrode is given by the rule of mixtures:
)1(2 FMoSfGraphene MEffMEffEff Equation 6.2
where EffGraphene and EffMoS2 are the efficiency of graphene- and MoS2-only electrodes and
Mf is the mass fraction of graphene in the mixed electrode. Then the efficiency should fall on
adding graphene to an MoS2-only electrode. However, for electrodes dominated by MoS2, the
electrode resistance will be so high as to prevent the electrode operating to its full capability.
Adding graphene increases the electrode conductivity according to the percolation scaling
law179
which can be expressed approximately as
2
fM Equation 6.3
(here we set the percolation threshold to zero for simplicity and take the percolation exponent
as its universal value348
). By analogy to previous results for MnO2/nanotube based
supercapacitor electrodes, addition of the graphene should increase the effectiveness of the
electrode by minimising any resistance-based limitations. I propose the efficiency of the
95
60
70
80
90
(d)
(c)
(b)
Jsc/J
sc,P
t (%
)
(a)
0 25 50 75 10085
90
95
100
Vo
c/V
oc,P
t(%
)
70
80
90
100
110
120
FF
/FF
Pt (
%)
0 25 50 75 10030
40
50
60
70
80
90
100
110
Eff/E
ffP
t (%
)
Mass Fraction Graphene (%)
Figure 6.12(a) Short current density, (b) open circuit voltage, (c) fill factor and (d) efficiency for DSSCs fabricated with MoS2/graphene composite counter electrodes (t=400 nm) plotted
as a function of graphene mass fraction. Each parameter is expressed relative to that measured for an equivalent cell with a Platinum counter electrode. The dashed lines in (a)
and (b) represent rule of mixtures type behaviour. The line in (d) is a fit to equation 6.4.
composite electrode can be roughly modelled by modifying equation 6.2 by multiplying it by
equation 6.3 to crudely account for the effect of the graphene on the electrode conductivity.
Then we find:
)1(2
2
0 fMoSfGraphenef MEffMEffkMEffEff Equation 6.4
where k is a constant and Eff0 accounts for the fact that even in the absence of graphene, a
thin layer of MoS2 near the current collector can act as an effective catalyst without resistance
limitations. We have applied this expression to the data in figure 6.12(d), finding it to well-
match the overall behaviour of the data. This analysis implies Graphene
MoS
EffEff 2 ~6 and
0
2
EffEff MoS ~10. This suggests that while the MoS2 nanosheets used here may be a far better
catalyst than the graphene nanosheets, only about 10% of nanosheets in an MoS2-only
electrode are close enough to the current collector to be active.
96
These results clearly show that the MoS2 nanosheets studied here are better catalysts for the
tri-iodide reaction than the graphene sheets. In both cases, it is likely that the catalytically
active sites lie on the nanosheet edge.47,48,50
. This is certainly the case for MoS2 nanosheets
when catalysing the hydrogen evolution reaction344
and Ni(OH)2 nanosheets when catalysing
the oxygen evolution reaction.349
This means that two factors come into play, the site
catalytic activity, often expressed via the turn-over frequency, and the number of sites. While
the relative values of the former parameter are unknown, we can say something about the
relative active site densities for MoS2 and graphene. The number of edge catalytic sites per
unit volume of electrode scales inversely with nanosheet length.347
The mean lengths of the
nanosheets used here were 586 nm versus 156 nm for graphene and MoS2 respectively. This
means that, assuming all sites are active and roughly equally spaced, there are about four
times as many active sites per unit volume of MoS2 compared to graphene on a per unit
volume basis. Moreover, the analysis of McAteer et al.347
would suggest that the catalytic
activity should scale as the turnover number, R, divided by mean nanosheet length, <L>.
Assuming that the efficiency scales roughly linearly with catalytic activity would give
2
22
MoS
Graphene
Graphene
MoS
Graphene
MoS
L
L
R
R
Eff
Eff Equation 6.5
Then, using the values given above would suggest that 2 / ~1.5MoS GrapheneR R . This implies
that the MoS2 catalytic activity on a per site basis is only slightly larger than that of graphene.
In fact most of the difference between MoS2 and graphene efficiencies is due to the fact that
the MoS2 nanosheets used here are considerably smaller than the graphene ones.
6.3.4 Graphene Flake Size Dependence
Multiple works have suggested that the edges of the flake are responsible for the catalytic
activity of the graphene sheets in DSSCs47,48,51
. There is very little information on the effect
of the size of the flake on the efficiency of a cell. Ahmad et al use various graphene sources
with different surface areas and initial particle sizes which result in different efficiencies350
.
This surface area difference is probably due to the average number of layers of the graphene
particle provided and the particle sizes are in the range of 2-5µm. However since the
graphene is used in a composite with PEDOT:PSS the particle size effect of graphene only is
not investigated.
97
In this work graphene flakes of various average lengths were produced by exfoliation of
graphite in water and surfactant and filtered before transfer to FTO glass. The average length
of the flakes was determined statistically using transmission electron microscopy. The
histograms representing this analysis are shown in figure 6.13.
0
10
20
12krpm <L> = 94nm5krpm <L> = 129nm
2krpm <L> = 202nm
(d)(c)
(b)
Count
1krpm <L> = 446nm
(a)
0
10
20
30
0 200 400 600 800 10000
5
10
15
20
Count
Length (nm)
0 200 400 600 800 10000
5
10
15
20
25
30
Length (nm)
Figure 6.13 (a) Histogram of flake sizes for 1krpm sample with sample transmission electron microscopy image inset (b) histogram of flake size for 2krpm sample (c) histogram of flake
size of 5krpm sample (d) histogram of flake size of 12krpm sample
The length of the flakes decreases from 446nm to 202nm between the spin rates of 1 and
2krpm. To achieve smaller still flakes the spin rate was increased to 5krpm and subsequently
12krpm. The smallest average flake size achieved was 94nm. It is worth mentioning that the
relative concentrations of these dispersions decrease with decreasing flake size as
concentration is inversely proportional to the spin rate. This reduces the cost effectiveness of
the smaller flakes as the yield is significantly lower.
We used thin graphene films (0.1mg/cm2) as counter electrodes in DSSCs. This was done to
limit any possible saturation in efficiency that could be due to film thickness which will be
discussed later. Current density-voltage (J-V) curves for cells with different graphene flake
98
size electrodes are shown in Fig 6.14. Shown for comparison is data for a cell with a standard
Pt counter electrode. It is clear from these curves that DSSC performance improves as the
graphene flake size is decreased (device parameters are given in table 1).
0.0 0.2 0.4 0.60
2
4
6
8
10
12
14
16
18
Pt =6.7%
202nm =1.59%
129nm =2.44%
194nm =2.38%
Curr
ent D
ensity (
mA
/cm
2)
Voltage (V)
Figure 6.14 IV Characteristics of DSSCs with graphene counter electrodes of different flake lengths. Derived parameters are shown in table 1
From figure 6.14 there is a clear difference between the 2krpm sample and the 5krpm sample.
To analyse this difference the J-V characteristics ie. the short circuit density, Jsc, the open
circuit voltage, Voc, the fill factor, FF, and the efficiency, Eff ( /ff sc oc inE J V FF P , where Pin
is the input power) were extracted and represented as a percentage of the corresponding
platinum cell as shown in figure 6.15.
Sample <L> (nm) Jsc (mA/cm2) Voc (V) FF (%) Eff (%)
1krpm 446 9.64 0.626 17.9 1.09
2krpm 202 13.51 0.655 18.7 1.59
5krpm 129 14.73 0.659 25.7 2.44
12krpm 94 13.68 0.653 26.8 2.38
Table 6.4: Table containing cell parameters for counter electrodes of varying flake length
<L> - short circuit current (Jsc), open circuit voltage (Voc), fill factor (FF) and efficiency (Eff)
99
Below a certain flake size the short circuit current seems to be roughly invariant around 83%
that of the platinum cell. The highest flake size has a much lower current density at short
circuit of below 55% that of the platinum counter electrode. The open circuit voltage
increases steadily as flake length decreases. The Fill factor sees a dramatic increase between
a flake length of 200nm and 130nm increasing by one third, thereafter there is not as much
influence on the fill factor. The efficiency as a result of these properties ranges from 14% that
of the platinum counterpart at a flake length of 440nm to 35% at a flake length of 90nm.
While this increase is dramatic the efficiencies with respect to the platinum counterparts are
still very low. In order to pursue parity with platinum based DSSCs increasing the thickness
of the graphene film seems like a more viable strategy. This is because the current methods
cannot produce smaller flakes in sufficient quantities. However some forms of carbon black
have low particle sizes which may be a viable route to efficient thin counter electodes.
55
60
65
70
75
80
85
0.0025 0.005 0.0075 0.0191.5
92
92.5
93
93.5
94
94.5
28
30
32
34
36
38
40
42
44
46
(d)
(c)
(b)
Eff/E
ffP
t (%
)
Vo
c/V
oc,P
t(%
)
FF
/FF
Pt (
%)
(a)
0.0025 0.005 0.0075 0.01
15
20
25
30
35
Jsc/J
sc,P
t (%
)
<L>-1 (nm
-1)
Figure 6.15 (a) Short current density, (b) open circuit voltage, (c) fill factor and (d) efficiency for DSSCs fabricated with graphene counter electrodes plotted as a function of graphene flake length. Each parameter is expressed relative to that measured for an equivalent cell
with a Platinum counter electrode.
100
6.4 Conclusions
In this chapter we have analysed the dependence of the performance of dye sensitised
solar cells on the thickness of a graphene nanosheet counter electrode establishing an
optimum thickness of 400 nm. A thicker film was significantly enhanced by the addition of
carbon nanotubes. Using the optimised thickness of 400nm, we explored the effect of
changing the counter electrode composition by mixing the graphene with MoS2 nanosheets.
We find the performance to be optimised a composition of 90:10 graphene:MoS2 by mass. At
this composition, the efficiency was very close to that measured for a platinum counter
electrode. This data is consistent with the MoS2 nanosheets being somewhat better catalysts
but requiring the graphene nanosheets to render the electrode conductive. More detailed
analysis suggests the better performance of the MoS2 nanosheets to be mostly down to their
smaller size. A graphene flake size dependence was conducted which revealed a sub-linear
dependence on the inverse flake length. This implies that the relationship between edge sites
and efficiency is not as simple as initially thought.
Thin film composites using Graphene and MoS2 derived from liquid phase exfoliation
may represent a cheap alternative, using naturally occurring starting materials, to produce
counter electrodes comparable to those produced via synthetic routes. The potential for
upscaling in liquid phase exfoliation is an added benefit for driving down the price of counter
electrode materials in DSSCs.
101
7
Thickness Dependence of Capacitance of PEDOT:PSS
Supercapacitor Electrodes
7.1 Introduction
In this chapter, the characteristics of thick PEDOT:PSS films as supercapacitor electrodes are
investigated. Recent work has produced transparent supercapacitors using high conductivity
doped PEDOT:PSS films196
. Expanding on this work, we investigate thick, non-transparent
electrodes with a view to analysing the dependence of capacitance per unit area with
thickness.
Using the dimensions of commercial supercapacitor electrodes for energy storage (a
50 micron current collector and a capacitive layer in the order of 100s of microns) as a rough
template, the aim is to produce effective electrodes in that range or lower. The added
advantage of PEDOT:PSS is that it acts as a current collector itself, thus opening up the
possibility of achieving comparable energy storage at lower thicknesses increasing the
percentage mass of a device with capacitive properties.
Two deposition techniques will be assessed for effective deposition of the material
with an emphasis on scalable methods to facilitate production on an industrial scale.
Dropcasting is an obvious scalable method which can produce thick, free standing films in
moulds while doctor blading can be used to produce thinner films on a substrate.
For comparison with theory, we will work from thin electrodes up to higher
thicknesses to analyse difficulties that may arise due to limitations associated with electrical
resistance and, most importantly, diffusion. These limitations lead to a sub-linear dependence
of capacitance per unit area with thickness and as such require analysis to optimise devices.
102
This will be assessed using the rate dependence of the capacitor performance using
previously established theory196,351
.
For long charge/discharge times associated with low scan rates, we can assume that
capacitance will be close to theory defined by resistance limitations. As the scan rate
increases the proportion of the electrode accessible by the electrolyte ions may decrease
resulting in deviation from resistance limited-theory.
Using the knowledge gained from this study, the advantages of increasing the
thickness (increasing intrinsic capacitance per unit area, CA, and decreasing sheet resistance,
Rs) will be balanced against the main disadvantage (diffusion limitations). This will be used
to propose an optimum thickness for certain applications.
7.2: Experimental Procedure
7.2.1: Sample Preparation
Clevios PH1000 PEDOT:PSS in water dispersion was purchased from Hereus. 95% Formic
Acid was purchased from Sigma Aldrich.
To achieve films of thicknesses in the range of 1-20 microns, the doctor blading
technique was used. To produce a dispersion with suitable viscosity for this method, the
initial dispersion had to be concentrated using a Krosflo tangential flow filtration setup
(Figure 7.1). The initial concentration was approximately 1.1% by weight (w/w) and after
processing through the Krosflo this had increased to 2.3% w/w. The deposition was defined
by the height of heat resistant tape (of approximately 50 microns) adhered to a glass slide.
Concentrated dispersion was placed at the top of the slide and guided down the length by a
blade. The dispersion was then dried on a hotplate at 130⁰C for 10 minutes. Thickness control
was provided by multiple deposition steps and using additional layers of tape. The width of
the samples was in the 7-8 millimetre range.
103
Figure 7.1: KrosFlo tangential flow filtration system. Dispersion is pumped from resevoir right and passed through the filtration tube. Solvent passes through the filter into waste container
(right) while more concentrated dispersion returns to resevoir.
Increasing the thickness further using the doctor blading technique was prohibitively
time consuming so drop casting films in petri dishes using the initial dispersion produced
films in the 20-90 micron thickness range. To increase the thickness range, the concentrator
was used to increase the solution to approximately 2.3% w/w resulting in a film thickness of
200 microns.
Freeze dried films were also fabricated in an effort to increase the effectiveness of the
super capacitor electrodes by increasing porosity. Initial attempts to produce freeze dried
electrodes using the given dispersion yielded electrodes of unsuitable mechanical stability.
The concentrator produced a dispersion of 1.6% w/w before freeze drying. The thickness of
these films were in the range of 1-4 millimetres but, due to the sponge-like nature of the
films, high accuracy in thickness was difficult to obtain. Freeze drying was performed by Dr.
Zahra Gholmovand.
All films were then treated in formic acid by being submerged for 10 seconds
followed by drying in air.
104
7.2.2 Electrical Characterisation
Electrical characterisation was carried out via a four-wire measurement of the IV
characteristics of the films. This was done using a Keithley 2400 sourcemeter. The voltage
range was 0-0.1V. Connections were established using alcohol-based silver paint and silver
wires.
7.2.3 Electrochemical Characterisation
Electrochemical characterisation was carried out using a Gamry Ref 3000 potentiostat. All
measurements were taken in a three electrode setup using a silver/silver chloride (Ag/AgCl)
reference electrode and a graphite rod as the counter electrode. The electrolyte used was
0.5M Potassium Sulphate (K2SO4).
Cyclic voltammograms (CVs) were taken at a variety of scan rates from 0.02V/s to 2V/s. The
voltage window was between 0.1V versus Ag/AgCl and 0.8V versus Ag/AgCl. The electrode
was contacted with silver paint at the top which was insulted using varnish as shown in Fig
7.2(a & b). Potentiostatic EIS Spectra were taken at 0.4V versus Ag/AgCl using an AC
voltage perturbation of 10mV.
7.3 Results and Discussion
7.3.1 Electrical Properties
Films of thicknesses in the range of 1-20 microns were produced using doctor blading. A
characteristic sample prepared for measurement is shown in figure 7.2(a). For further
increases in thickness, drop-casting was used to produce samples similar to that shown in
figure 7.2(b).
105
Figure 7.2: Doped PEDOT:PSS electrodes for electrochemical testing. A thin 2.8mm sample (left) compared with a thicker 20mm sample (right) contacted at the top using silver paint.
The paint is insulated with varnish to prevent contact with the electrolyte.
The electrical characteristics of these films is critical in the absence of a current
collector normally present in supercapacitor electrode systems. From previous work, the
conductivity of the untreated films as prepared is of the order of 1 S/cm196,325
. Treating
PEDOT:PSS to improve the conductivity is critical to reduce resistance limitations in
devices. Extensive work has been done to improve PEDOT:PSS film conductivity, especially
for transparent applications. Excellent conductivities have been achieved using alcohols
(1362 S/cm)322
, ethylene glycol (1418 S/cm)352
, and sulfuric acid (3065 S/cm)324
to post-treat
the films. These methods improve the conductivity by removing some of the PSS and by
changing the conformation of the molecules in the film.
Using the same method as McCarty et al325
., which involves a simple one step dip of
the film in formic acid and allowing it to dry in air, the sheet resistance varies with thickness
as shown in figure 7.3. Sheet resistance, which ranges from approximately 2 Ω/sq for the
thinnest film measured (approximately 5 µm) to less than 0.1 Ω/sq for the thickest
(approximately 200 µm), should vary inversely with thickness
assuming a constant
conductivity. This is the case for the majority of the films which have conductivities close to
what would be expected for films with conductivity of 900 S/cm. This shows that the doping
of the films is effective up to thicknesses in excess of 100 microns.
106
10 100
0.1
1
Sh
ee
t R
esis
tan
ce
(
/sq
)
Thickness (m)
S=900S/cm
Figure 7.3: Sheet Resistance versus thickness of Formic Acid Treated PEDOT:PSS films.
For the thicker films, the sheet resistance is lower than the lowest PEDOT supercapacitor
electrodes as produced by Wang et al with a sheet resistance of 1.4 Ω/sq326
.
7.3.2 Cyclic Voltammetry of Thin Films
To ascertain the performance of the doctor-bladed and drop-cast thin films, the electrodes
were examined using cyclic voltammetry from 0.1 to 0.8V versus Ag/AgCl for scan rates in
the range of 0.02 V/s to 2 V/s. The samples were 11 mm in length and approximately 8 mm
in width.
Figure 7.4 displays the cyclic voltammograms of a thin 2.8 µm film (a) and a thick 40
µm film (b). The lowest and highest scan rates were 0.02 V/s (red) and 2 V/s (green)
respectively. Both films are approximately boxlike and symmetrical with the exception of the
high voltages in the anodic cycle and the low voltages in the cathodic cycle which is due to a
slow electrochemical process. The relative size of the voltammogram at 2 V/s to the
voltammogram at 0.02V/s is much smaller for the thicker film. This work seeks to describe
the effect of thickness on the characteristics of supercapacitors.
107
0.0 0.2 0.4 0.6 0.8
-200
-150
-100
-50
0
50
100
150
200
(b)
j/s (
F/m
2)
0.0 0.2 0.4 0.6 0.8
-4000
-2000
0
2000
4000
(a)
Figure 7.4: (a) Cyclic voltammogramms of a 2.8µm PEDOT:PSS film starting with scan rates
ranging from 0.02 V/s (red) to 2 V/s (green). (b) Cyclic Voltammogramms with same scan
rate range for a film of thickness 40 µm.
Previous work by Higgins and Coleman196
used the following equation to describe the
current density, j:
)exp(1
tCs
jA Equation 7.1
where CA is the intrinsic capacitance per unit area which would be achieved by charging
infinitely slowly, s is the scan rate, t is time and τ is the time constant of the capacitor.
Current density divided by scan rate (j/s) is useful as it allows comparison of cyclic
voltammograms over a range of scan rates. This equation assumes that the capacitor starts
with no charge on the electrodes. However, in cyclic voltammetry there is an initial charge on
the electrodes from the previous cycle. Figure 7.4 displays the whole of the cyclic
voltammogram such that the initial current or charge is clearly non-zero.
To account for this, Pell and Conway351
used the boundary conditions equating the
currents and the start of the anodic current (top part of the curve) and the cathodic current
(bottom part of the curve) at the initial voltage (in this case 0.1V) and the final voltage
(0.8V). This produced equation 7.2:
)exp(
)exp(1
21
t
sV
Cs
jA Equation 7.2
108
where ΔV is the voltage window of the cyclic voltammogram and all other variables have
been defined. The anodic part of the current has been plotted for films of various thicknesses
at different scan rates to illustrate the suitability of equations 7.1 and 7.2 to describe the
current in figure 7.5.
0.10 0.12 0.14 0.16 0.18 0.20
0
100
200
j/s (
F/m
2)
(d) 40m 0.5V/s(c) 40m 0.02V/s
(b) 2.8m 0.5V/s
Experimental
Higgins & Coleman
Pell & Conway
(a) 2.8m 0.02V/s
0.0 0.2 0.4 0.6 0.8
-100
0
100
0.1 0.2 0.3 0.4
-2000
0
2000
j/s (
F/m
2)
Voltage vs. Ag/AgCl (V)
0.0 0.2 0.4 0.6 0.8
-1000
0
1000
Voltage vs Ag/AgCl (V)
Figure 7.5: The experimental (black), as predicted by Higgins & Coleman (red) and as predicted by Pell & Conway (green) anodic current density divided by the scan rate for (a) a 2.8mm film at 0.02 V/s, (b) a 2.8mm film at 0.5 V/s, (c) a 40mm film at 0.02 V/s and (d) a
40mm film at 0.5 V/s
For the lower scan rate of 0.02 V/s, the earlier stage of the anodic current is where the
largest difference between the different models occur. With the exception of low thickness
and scan rate, the Pell & Conway model is the better model for fitting the anodic current
density curves due to the negative currents at 0.1 V vs Ag/AgCl.
109
In these models CA, the intrinsic capacitance per unit area, is a characteristic property
of the capacitor and should be invariant under scan rate and time constant. The fitting of the
anodic current density at different scan rates with the Pell & Conway model results in
different values for CA and time constant. This is illustrated in figure 7.6(a).
0.0 0.2 0.4 0.6 0.8
-1000
-500
0
500
1000
(b)
CEXP
= 1512F/m2 = 0.9s
CSIM
= 2074F/m2 = 2.44s
j/s (
F/m
2)
Voltage vs Ag/AgCl (V)
Thickness 40m
Scan Rate = 0.5V/s
(a)
0.01 0.1 1
800
1000
1200
1400
1600
1800
2000
2200
CA (
F/m
2)
Scan Rate (V/s)
1
2
3 C
A
(s
)
Figure 7.6: (a) The variance of intrinsic capacitance per unit area (black, left axis) and time
constant (red, right axis) with scan rate. (b) The experimental anodic current density with
fitting parameters at 0.5 V/s (black) simulated anodic current using the Pell & Conway CA
and time constant values for 0.02 V/s
This decay in CA and time constant may be due to the effect of diffusion. This would
result in only part of the capacitor being active at a given time due to the time it takes the
potassium ions to enter and leave the polymer matrix. Diffusion will be given a more in-depth
discussion later. Since time constant is the product of the capacitance and resistance, the
variance of the two properties is expected to be similar but this is not the case, as shown in
figure 7.6(a), due to the model making the best possible fit without being able to account for
diffusion.
Figure 7.6(b) compares the experimental anodic current density (red) with a simulated
anodic current density at 0.5 V/s (black). The experimental values for CA and time constant
were given by the Pell & Conway fitting the current density at 0.5V/s. The simulated anodic
current uses the Pell & Conway equation for current density but uses the values for CA and
110
time constant for the fit at 0.02 V/s the initial values which should be invariant with scan rate.
Despite the higher CA value for the simulated curve, the current density is lower due to the
higher time constant. This is an example of how diffusion can help produce higher current
densities and hence higher capacitances at higher rates than would be expected.
Plotting the scan rate dependence of the current density divided by the scan rate at
various voltages in the potential window can be done by substituting for time in equation 7.1
and 7.2 where t=(V-Vi)/s where V is the voltage of the electrode compared to the reference
electrode and Vi is the starting voltage of the scan. Figure 7.7 demonstrates the fitting of the
current density divided by the scan rate to the scan rate for; (a) a 2.8 µm film fitted using
equation 7.1, (b) a 40 µm film fitted using equation 7.1, (c) a 2.8 µm film fitted using
equation 7.2 and (d) a 40 µm film fitted using equation 7.2.
0.01 0.1 1-50
0
50
100
150
j/s (
F/m
2)
j/s (
F/m
2)
40m2.8m
(d)
40m
Scan Rate (V/s)
2.8m
0.01 0.1 1
-500
0
500
1000
1500
2000
2500
(b)
Scan Rate (V/s)
(a)
-50
0
50
100
150
0.2V
0.3V
0.4V
0.5V
0.6V
0.7V
-500
0
500
1000
1500
2000
2500
(c)
Figure 7.7 (a & c) Scan rate dependence of current density divided by scan rate for a 2.8 µm thick film at voltages from 0.2 to 0.7V vs. Ag/AgCl. (b & d) Same plot for 40µm thick film. (a
& b are fitted using equation 7.2 while (c & d) are fitted using equation 7.2
Due to the redox peak at the end of the anodic current, there is a contribution to the
current density at the higher voltages, particularly at low scan rates. This manifests as an
increase of the current density divided by the scan rate relative to that predicted by the
111
equations. Since the contribution of this peak is lower at higher scan rates, the fitting has been
constrained to the higher scan rates at the higher voltages.
The fitting using equation 7.1 (Figure 7.7(a & b)) is a particularly bad fit at the lower
voltages. This is due to the initial starting current being below zero. As the voltage increases,
however, the fit improves as can be seen in figure 7.6 where the fitting of the two models
converge at higher voltages. The fitting for equation 7.2 is clearly better thus validating the
equation as the appropriate model for this material system.
For the thinner films, the fits result in an intrinsic capacitance per unit area and time
constant that is constant. Compared to the thinner films, however, the fitting for the thicker
40 µm film produces lower intrinsic capacitances per unit area at the low voltages which can
again be attributed to diffusion as discussed previously.
The current density divided by the scan rate at 0.4 V has been chosen as the most
representative voltage for most films as it is sufficiently high for the effect of diffusion to be
reduced and below the onset of the redox peak. As such, the current density divided by the
scan rate at 0.4 V has been plotted against scan rate for various thicknesses to extract values
for the intrinsic capacitance per unit area and the time constant (figure 7.8).
0.01 0.1 1
j/s @
0.4
V (
F/m
2)
Scan Rate (V/s)
0.01 0.1 1
1
10
100
1000
10000
(b) 2.8m
40m
90m
Scan Rate (V/s)
(a)
Figure 7.8: (a) Scan Rate Dependence of the current density divided by scan rate for thicknesses of 2.8, 22 and 90 µm with fitting according to theory from Higgins (Equation 7.1). (b) Scan Rate Dependence of the capacitance per unit area for thicknesses of 2.8, 40
and 90 µm with fitting according to theory from Pell & Conway (Equation 7.2).
112
At low scan rates, the current density is larger for the thicker films. The effect of
increasing scan rate leads to a decrease in charge density which is more pronounced in the
thicker films. The thinnest film (thickness 2.8 µm) current density divided by scan rate at 2
V/s is approximately 60% that of the current density divided by the scan rate at 0.02 V/s. In
contrast, the thickest film retains less than 0.1% of the current density divided by scan rate at
0.02 V/s at 2 V/s.
The plots were fitted using the models used by Higgins & Coleman - equation 7.1 (a)
and Pell & Conway - equation 7.2 (b). The most noticeable difference between the fits occurs
at scan rates in excess of 0.1 V/s. After 0.1 V/s, the Pell & Conway model predicts lower
current densities than is experimentally observed while the Higgins & Coleman model
predicts higher current densities. This can be explained for the Higgins & Coleman model by
reviewing figure 7.5(d) and is due to the limitation of the model starting with a zero current
density. This phenomenon has been discussed previously for the Pell & Conway model and is
caused by diffusion.
The capacitance of the film can be extracted from the current density divided by the
scan rate by integrating equation 7.1 or 7.2 over the voltage range to get the charge stored and
then dividing by the voltage window. Substituting dV for sdt allows integration of equation
7.3
VV
V
i
i
dVs
j
VA
C 1 Equation 7.3
which gives the capacitance per unit area as equation 7.4 using the Higgins & Coleman
model:
)exp(11
s
VV
sC
A
CA Equation 7.4
The Pell & Conway model is represented as equation 7.5 which is determined by
substituting equation 7.2 for j/s into equation 7.3 and integrating:
)exp(1
)exp(1
21
s
V
sVV
sC
A
CA Equation 7.5
113
These give expressions for the experimentally observed capacitance per unit area in
terms of previously introduced variables. Figure 7.9 plots the experimental capacitance per
unit area against the scan rate for the same films in figure 7.8 and fits the data for (a) the
Higgins & Coleman Model – equation 7.4 and (b) the Pell & Conway model – equation 7.5.
0.01 0.1 1
10
100
1000
10000(b)
C/A
(F
/m2)
Scan Rate (V/s)
(a)
0.01 0.1 1
2.8m
40m
90m
Scan Rate (V/s)
Figure 7.9: (a) Scan rate dependence of the capacitance per unit area for thicknesses of 2.8, 22 and 90 µm with fitting according to theory from Higgins & Coleman. (b) Scan Rate
Dependence of the capacitance per unit area for thicknesses of 2.8, 22 and 90 µm with fitting according to theory from Pell & Conway.
Here the thicker films initially have a higher capacitance per unit area at low scan rates but
where the capacitance decreases as scan rate increases. The relative change in capacitance is
not as significant for thinner films. These figures differ from the figures in 7.8 as these data
are achieved over the entire voltage range as opposed to one point. Therefore, extracting
intrinsic capacitance per unit area and the time constant using this method is more accurate.
Interestingly, the values for intrinsic capacitance and time constant obtained using the Pell &
Conway fit for current density at 0.4V and capacitance per unit area agree strongly and are
shown in figure 7.10. However, this does not occur for the Higgins & Coleman model and as
such the values for capacitance per unit area represent those from fitting the data in figure
7.9(a).
The values for intrinsic capacitance per unit area and time constant are plotted as a
function of mass per unit area of the film in figure 7.10 (a) and (b) respectively. When plotted
against mass per unit area, the intrinsic capacitance per unit area follows a straight line. The
114
slope of this line yields an intrinsic capacitance per unit mass of 38.5 F/g for the Higgins &
Coleman Model and 33 F/g for the Pell and Conway model.
These capacitances per unit mass are lower than the capacitance per unit mass of
PEDOT films in literature which is in the range of 85 to 115 F/g326
. However, there is a
significant mass of PSS in the electrode (the ratio is PEDOT:PSS is 1:2.5 by mass according
to manufacturer specifications). While the doping of the film can remove some PSS, a
significant amount remains. Assuming minimal loss of PSS, the capacitances per unit mass of
the PEDOT-only component is 135 F/g for the Higgins and Coleman model and 116 F/g for
the Pell and Conway model. This suggests the Pell and Conway model is better as it is closer
to the range defined by literature. The formation of PEDOT for these electrodes involves
electrochemical polymerisation and deposition, which requires a conductive substrate and
investment of resources to gain suitably thick films. The advantage of the PEDOT:PSS is that
the catalyst molecules can be removed by a chemical step after polymerisation. This allows
for dispersion of the polymer particles in water which facilitates the simple formation of films
by the methods used in this work and also compatibility with processes like spray coating.
1 10 100
100
1000
10000(b)
CA (
F/m
2)
M/A (g/m2)
(a)
1 10 100
0.1
1
10
Higgins
Pell + Conway
(s)
M/A (g/m2)
Equation y = a + b*x
Weight No Weighting
Residual Sum of Squares
3.71451 0.57272
Pearson's r 0.98978 0.97893
Adj. R-Square 0.97713 0.95309
Value Standard Error
TauIntercept 0.37776 0.27325
Slope 0.06001 0.00306
CIntercept 0.19238 0.10729
Slope 0.01627 0.0012
Figure 7.10: (a) Intrinsic capacitance per unit area vs. mass per unit area as derived from fittings of the scan rate dependence of the current density divided by the scan rate and the
capacitance per unit area vs. the scan rate. (b) Time constant vs. mass per unit area similarly derived
The time constant is plotted against mass per unit area in figure 7.10(b). This
produces a straight line graph. The time constant is the product of the capacitance and
resistance of the system.
115
)( eelectrolytelectrode RRC Equation 7.6
Here C is the capacitance and R represents the resistance components due to the inherent
resistance of the electrode and the electrolyte. The capacitance is given simply by the product
of the intrinsic capacitance per unit area multiplied by the area (L×W, where L is length and
W is width). The resistance of the electrode is given by the sheet resistance (Rs) divided by
the electrode width and multiplied by the electrode length. The electrolyte resistance is given
by the inverse of the product of the electrolyte conductivity (GA) and the area through which
the current travels. The full expression of the time constant is shown in equation 7.7.
A
MM
A
SAG
tC
LC
LWGW
LRLWC
2
)1
( Equation 7.7
This can be further simplified by converting the intrinsic capacitance per unit area to
the intrinsic capacitance per unit mass (CA = CMρt where ρ is the density of the film and t is
the thickness). Rs the sheet resistance is also given by the inverse of the product of the
conductivity and the thickness ((σt)-1
).
This equation can be used with the equation of the line from the linear fit of the data.
This shows that the intercept is dependent on the conductivity of the film while the slope is
dependent on the conductivity of the electrolyte. The conductivities of the doped
PEDOT:PSS film extracted from the Higgins & Coleman model and the Pell & Conway
model are 176 and 297 S/cm respectively. However, there are large standard errors in the
intercept and as such this could partially explain the lower than expected value of 900 S/cm.
The slope is the intrinsic capacitance per unit mass divided by the electrolyte conductivity as
ρt is equivalent to the mass per unit area. This gives conductivities of the electrolyte for the
Higgins & Coleman model and the Pell & Conway model of 64 and 206 mS/cm respectively.
The electrical properties as determined by the Pell & Conway model are superior to those
from the Higgins and Coleman model in both the electrode and electrolyte conductivities.
The overall capacitance per unit area of the electrodes is important for integration into
devices. The capacitance per unit area of the electrodes as shown in figure 7.11. The
behaviour of the capacitance per unit area of these films can be fitted by inserting equation
7.7 into equations 7.4 and 7.5 using the values received for intrinsic capacitance per unit
mass, conductivity of the PEDOT:PSS film and the conductivity of the electrolyte.
116
The fits for both models are suitable at low scan rates of 0.02 V/s (a) and 0.1 V/s (b).
However, as the scan rate increases the models do not provide a good fit. The Higgins and
Coleman model in particular does not fit well after 30 g/m2 for 0.5 V/s (c) the source of this
deviation was initially assumed to be diffusion but is mainly due to the inability of the model
to account for negative current densities at higher scan rates. However, since the Pell &
Conway model fits better beyond 30g/m2 this shows the suitability of the model.
117
1 10 10010
100
1000
10000
(c)
(b)
C/A
H+C
P+C
C/A
(F
/m2)
(a)
1 10 10010
100
1000
10000
C/A
(F
/m2)
1 10 10010
100
C/A
(F
/m2)
M/A (g/m2)
Figure 7.11 Capacitance per unit area for (a) 0.02V/s (b) 0.1V/s and (c) 0.5V/s. Blue line fitted using Higgins and Coleman Model. Red line fitted using Pell and Conway model.
The maximum capacitance per unit area achieved is 6000 F/m2
at 0.02 V/s or 0.6
F/cm2
which is the preferred unit for devices. This is comparable with the capacitances in
literature for PEDOT and PEDOT:PSS films with the highest being 0.92 F/cm2. Due to the
118
relatively low sheet resistance and the ease of fabrication producing supercapacitor electrodes
following this method would be viable.
The analysis of the models show that the Pell & Conway is the better of the models. It
more accurately fits the current densities and hence the values for the capacitances and
conductivity using this model are more accurate than those achieved using the Higgins &
Coleman model. This is due to the negative current densities that arise as thickness and scan
rate increases. Both models fail to account for diffusion which leads to poor fitting of the
model at high scan rates.
7.3.3 Impedance Spectroscopy
Films were also tested using potentiostatic impedance spectroscopy with a voltage amplitude
of 10 mV around 0.4V vs. Ag/AgCl in a frequency range from 1 MHz to 0.1 Hz. Samples of
impedance spectra are given in figure 7.12 (a) and (b) for a thin 2.8 µm film and a thick 40
µm film respectively. This was used to analyse the characteristics of the electrode. The main
components for an electrode system such as the PEDOT:PSS electrode in the range of 1 to
10 15 20
0
100
200
High Frequency
-Zim
g (
)
Zreal
()
Low Frequency
6 7 8 9
0
2
4
6
8
10
12 (b) 40m
Zreal
()
(a) 2.8m
Figure 7.12: Impedance spectra of films with thickness of (a) 2.8 µm and (b) 40µm
200 microns are the resistive elements, the capacitive elements and the diffusive elements114
.
The impedance can be separated into real and imaginary components due to the phase
response of these elements. There is an obvious effect on both the real and imaginary
components of impedance with frequency.
119
The resistive elements are from the connections of the system, the sheet resistance of
the electrode and the resistance of the electrolyte. These resistances are independent of
frequency. The capacitive impedance is frequency dependent, completely out of phase with
the oscillating voltage and follows the equation:
CjZC
1 Equation 7.8
where j is the imaginary number and ω= 2πf with f being the frequency of oscillation in the
experiment. The diffusive element is determined by the Warburg impedance. The Warburg
impedance is frequency dependent with a phase of 45 degrees. It is described by the equation
)11(j
AZ W
W
Equation 7.9
where AW is the Warburg coefficient.
0 5 10
3
4
5
6
7
8
9
10
11
12
Zre
al @
f =
1M
Hz
Rs (sq)
Figure 7.13: Real Impedance at 1MHz vs the sheet resistance of the corresponding electrode.
A simple analysis of these components can be done by looking at regions of low and
high frequency. At high frequency, the imaginary components which are related inversely to
the frequency are minimized leaving the real components independent of frequency isolated.
This gives the resistances due to the electrolyte and the electrode and connections in the
120
experimental setup. Figure 7.13 displays the comparison of the real component of the
impedance at the highest frequency with the sheet resistance of the electrode.
Assuming only contributions due to the sheet resistance of the electrode to the
impedance at high frequencies a straight line through the origin would be expected. The line
in figure 7.13 has an intercept of 3.5 Ω. The contribution due to the electrolyte resistance as
determined in the previous section is in the order of 0.2 Ω, which is far below the value of the
intercept. This means that the rest of the impedance at high frequencies originates in the
connections of the experimental setup. This could also be the source of the scatter in this plot.
At low frequencies, the imaginary component of the impedance goes as 1/ω. This
corresponds to the capacitor element. The impedance at low frequencies is shown in figure
7.14(a)
0.1 1 10 100
0.01
0.1
1
10
100 2.8 m
20 m
40m
90m
1/f
C/A
(F
/m2)
-Zim
g (
)
Frequency (Hz)
0 50 100 150 200 250
0
2000
4000
6000
8000
(b)
EIS
CV
M/A (g/m2)
(a)
Figure 7.14: (a) Imaginary component of the impedance for the lower range of frequencies (b) Capacitance per unit area as obtained from Impedance spectra (black) and cyclic
voltammetry (red)
At the low end of the spectrum linear fits can be made with 1/f and the capacitance can then
be extracted from the slope using equation 7.8. Figure 7.14(b) shows the capacitance per unit
area as found from impedance spectroscopy (black) compared with the capacitance per unit
area as found from cyclic voltammetry at 0.02 V/s. The agreement is suitable for the lower
values of thickness however there is a significant divergence for the thickest film. This could
121
be due to the lower frequencies having a lower equivalent scan rate than 0.2 V/s which would
then result in a higher capacitance per unit area being obtained.
7.3.4 Analysis of Diffusion of Ions through Polymer Network
When designing an electrode for use in a commercial supercapacitor, attention must be paid
to the dimensions of the electrode. For low thicknesses, the energy capability per unit mass is
constant according to equation 2.3 capacitance increases approximately linearly as does mass.
This changes when the thickness, or as the scan rate gets too large, the capacitance starts to
drop, as described by equation 7.5. However, due to diffusion, equation 7.5 returns a lower
capacitance than is observed experimentally as indicated in figure 7.3 where the lower
capacitance and time constant due to diffusion produce a larger current density over the
voltage window and hence a higher capacitance.
By normalising the j/s of the experimental results with respect to that which would be
predicted by equation 7.2 and plotting it against sτ/ΔV, the behaviour of the electrodes in
reality (data points) can be compared to theory (line – Equation 7.2) as shown in figure
7.15(a). The same can be done for capacitance per unit area and equation 7.5. to produce
figure 7.15(b).
1E-3 0.01 0.1 1 10
1E-3
0.01
0.1
1 (b)
(J/s
)/C
A
s/V
1E-3 0.01 0.1 1 10
1E-3
0.01
0.1
1
(C
/A)/
CA
s/V
(a)
Figure 7.15: (a) Master curve of current density divided by scan rate for all samples as a function of the product of the scan rate and time constant divided by the voltage window.
122
The line is the plot of equation 7.2 (b) Master curve of capacitance per unit area as a function of same. The line is the plot of equation 7.5
All the experimental data coalesces into a master curve for direct comparison with
theory. There is a departure from the expected value, defined by the line in the graph, once
sτ/ΔV exceeds a certain value. This value is different for both j/s and C/A due to the fact that
the j/s data is taken at one voltage while the C/A is extracted by integrating j/s over the entire
voltage window. As a result, the point at which diffusion has an effect on the C/A (0.25)
occurs at a lower value of sτ/ΔV than it does for j/s (0.4).
Taking these values and using equation 7.7 for the time constant, the scan rate at
which diffusion has an effect for a given mass per unit area can be resolved. The black line in
figure 7.16(a) is the maximum scan rate as determined by the deviation in figure 7.15(a) and
the red line is the maximum limiting scan rate as defined by the deviation in figure 7.15(b).
The points in figure 7.16(a) are the point at which deviation occurred from the scan rate
dependence according to figures 7.8 and 7.9 with black and red points corresponding to j/s
and C/A respectively.
0 50 100 150 200 250 300
0.1
1
(b)
Lim
itin
g S
ca
n R
ate
(V
/s)
M/A (g/m2)
(a)
0 1x109
2x109
3x109
4x109
0.0
0.5
1.0
1.5
2.0
T-2 (m
-2)
Figure 7.16: (a) The black line represents the scan rate at which diffusion has an effect from the master curves of current density divided by scan rate(figure 7.15(a)). Black dots
represent scan rate point of deviation from individual fits (figure 7.6(a)). Red line represents the scan rate at which diffusion has an impact on capacitance per unit area (figure 7.15(b)).
Red dots represent the scan rate point of deviation from individual fits (figure 7.6(b)). (b) Point of scan rate dependence deviation as determined from fits of current density divided by
scan rate (red) and capacitance per unit area (black) as a function of the inverse thickness squared.
123
The source of the large error bars is due to the sampling of the electrode at different
scan rates. Within error, however, there is agreement between the points and the lines. The
point of deviation was plotted against the inverse square root of the thickness in figure
7.16(b). The relationship is approximately linear which can be explained using one
dimensional linear diffusion. For one dimensional linear diffusion the average distance the
ions will travel is:
Dtx Equation 7.10
where x represents distance, D represents the diffusion coefficient and t represents the time.
The relationship between time and scan rate has been discussed already. Substituting time
and rearranging the equation gives the relationship of electrode thickness to scan rate where
all variables have been previously defined with the exception of T which is the electrode
thickness. Γ has been chosen to avoid confusion with t for time in the previous equation.
)(2 iVV
Ds
Equation 7.11
The diffusion coefficient from the black line is 1210103.5 sm and the diffusion
coefficient corresponding to the red line is 1210103.6 sm . The Stokes-Einstein equation
provides a way to estimate the diffusion coefficient of the ions in a solution.
R
TkD B
6 Equation 7.12
where T is the temperature, η is the viscosity of the solvent and R is the radius of the ion.
Using the value for the radius of a potassium ion in water as calculated by Pau et al.353
, the
diffusion coefficient is 129106.1 sm . This indicates that the diffusion coefficient in the
film is close to that in water. The diffusion coefficient in a porous material is defined as the
diffusion coefficient in the solvent multiplied by the porosity of the film and divided by the
tortuosity of the pores. The diffusion coefficients as calculated from figure 7.16(b) are in the
range of 30-40% of the diffusion coefficient in water. This reduced rate of diffusion has a
significant impact on the rate dependence of these electrodes.
124
7.3.5 Limitations Due to Electrode Dimensions
The capacitance as determined by equation 7.5 starts to decrease before evidence of
diffusion. In figure 7.12(b), the capacitance per unit area is 0.3 times the intrinsic capacitance
per unit area when diffusion takes an effect. The decrease described by equation 7.5 starts to
become significant when sτ/ΔV is approximately 0.1 which is before the point at which
diffusion occurs. This means the electrical properties of the film have a bigger impact on the
decrease of capacitance with electrode dimensions and scan rate than diffusion. Equating 0.1
with sτ/ΔV and substituting in equation 7.3 for the time constant the maximum scan rate is
given by the following equation:
A
MM
G
tCLC
Vs
2max
10
Equation 7.13
Similarly, the maximum mass per unit area before the capacitance per unit area decreases as a
function of sτ/ΔV can be found by manipulating the formula to yield:
M
AM
Max C
GLC
s
V
AM
2
10 Equation 7.14
These equations are plotted in figure 7.17 to work as a model to define the maximum scan
rate (a) and maximum mass per unit area (b) before the capacitance per unit area deviates
significantly from its maximum possible value.
125
1 10 100
0.01
0.1
1(b)(a) L=1cm
L=2cm
L=4cm
L=8cm
sm
ax (V
/s)
M/A (g/m2)
0.01 0.1 1
1
10
100
M/A
ma
x (g/m
2)
s (V/s)
Figure 7.17: (a) maximum scan rate against thickness for various lengths of electrode (b) maximum mass per unit area against scan rate for various electrode lengths.
This graph works as a template for device design. Depending on the rate of charge required
for the device, one can choose an appropriate length and thickness for the electrodes. For
example, if the device will be operated at 0.1 V/s, a 2 cm long electrode with a mass per unit
area of approximately 60g/m2 would be most suitable as maximizing the product of length
and thickness is paramount to getting the highest capacitance per unit area. To find better
materials for supercapacitor electrodes, minimizing the time constant while keeping a high
capacitance is critical. This requires finding materials with higher conductivities for the
electrodes. Films of carbon nanotubes with a range of conductivities from 2000 S/cm for
liquid exfoliated films278,281
to over 12,000 S/cm354
for acid dissolved tubes which are higher
than that of the doped PEDOT:PSS in this work would be an ideal candidate. Especially with
the higher capacitance per unit mass reported for carbon nanotubes (in excess of 100 F/g).
However, carbon nanotubes are currently quite expensive and may only see application as a
conductive additive as a result.
126
7.3.6 Freeze Dried Foams
In order to improve performance per unit geometric area, especially at higher rates, freeze
dried films were fabricated. These foams allow electrolyte to enter the material more easily.
In figure 7.18, the left image shows an edge-on image of a freeze dried film of a similar mass
per unit area to the 90 µm sample. The right image in figure 7.18 shows a torn portion of an
edge revealing thin walls which form the pores.
Figure 7.18: The freeze dried film over 2mm thick similar mass to sample 90mm thick (left) Image of torn edge showing thin wall structure (right).
The porous nature of the electrode should allow quicker access of the electrolyte into
the electrode reducing the effects of diffusion. The freeze dried electrodes were tested by
cyclic voltammetry from 0.1 to 0.8 volts vs. Ag/AgCl with scan rates from 0.02 to 2 V/s. In
figure 7.19(a), cyclic voltammograms of a dropcast film of mass similar to that of the freeze
dried film (figure 19(b)) are displayed for comparison.
127
0.0 0.2 0.4 0.6 0.8
-5000
0
5000
j/s (
F/m
2)
(b) M/A = 92g/m2
Voltage vs. Ag/AgCl (V)
(a) M/A = 84g/m2
0.0 0.2 0.4 0.6 0.8
-5000
0
5000
Voltage vs. Ag/AgCl (V)
Figure 7.19: CVs of (a) a 84 g/m2 dropcast film and (b) a 92 g/m2 freeze dried film with scan rate range 0.02 V/s (red) to 2 V/s (green)
The red lines are the 0.02 V/s scan and the green lines are the 2 V/s scans. The
performance with increased scan rate for the freeze dried films is improved relative to the
dropcast film. This is more clearly visible if the current density divided by the scan rate and
capacitance per unit area are plotted against scan rate. In figure 7.20, these values are
normalised with respect to the value at 0.02 V/s for both the freeze dried film and the
dropcast counterpart for direct comparison.
0.01 0.1 1
1E-3
0.01
0.1
1
Freeze Dried 92g/m2
Dropcast 84g/m2
No
rma
lise
d j/s
@ 0
.4V
(F/m
2)
Scan Rate (V/s)
(a)
0.01 0.1 1
0.01
0.1
1
No
rma
lize
d C
/A
Scan Rate (V/s)
Figure 7.20: (a) Scan rate dependence of normalised (dq/dV)/A @ 0.4 V of freeze dried (black) and dropcast (red) films. (b) Scan rate dependence of normalized capacitance per unit area of freeze dried (black) and dropcast (red) films.
128
The current density at 2 V/s is approximately an order of magnitude higher for the
freeze dried film compared to the dropcast film. The difference in capacitance per unit area is
not as pronounced. However, consulting figure 7.11 for films in this mass range, diffusion
only has an effect close to 0.2 V/s. The effect of the improvement of freeze dried films are
evident immediately.
The fact that the improvement occurs earlier than expected for diffusion means there
is an improvement in either the intrinsic capacitance per unit area, which could be due to an
increase in surface area per unit mass of the material, or a reduction in the time constant
which would suggest the film allows better charge transport properties.
In order to find the intrinsic capacitance per unit area and time constant, the scan rate
dependence of capacitance per unit area was fitted according to equation 7.5.
0 20 40 60 80 100
0
1000
2000
3000
4000
(b)
CA (
F/m
2)
M/A (g/m2)
Equation y = a + b*x
Weight No Weighting
Residual Sum of Squares
619.70989
Pearson's r 0.9999
Adj. R-Square 0.9997
Value Standard Erro
CaIntercept 155.22804 25.04288
Slope 38.37995 0.38278
(a)
0 20 40 60 80 100
0.0
0.5
1.0
1.5
2.0
2.5
(s
)
M/A (g/m2)
Figure 7.21: (a) Intrinsic capacitance per unit area of the freeze dried films (black) and other films (blue) with a linear fit (red) to determin intrinsic capacitance per unit mass. (b) Time constant of freeze dried films (black) compared to other films (blue) with a linear fit (red).
Plotting the intrinsic capacitance per unit area and time constant against mass per unit
area, as shown in figure 7.21(a) and (b) respectively, the freeze dried films can be compared
with the films from the previous section. The capacitance per unit area of the freeze dried
films has a similar slope to the films from the previous section giving an intrinsic capacitance
per unit mass of 38 F/g. However, there is an appreciable intercept which may have arisen
due to experimental error.
129
The time constant has a much lower slope but a higher intercept. This suggests poorer
conductivity in the film but a better conductivity in the electrolyte possibly due to the
geometry of the electrode. From the intercept, the conductivity of the freeze dried film is 307
S/cm while the conductivity of the electrolyte is 0.42 S/cm2, almost twice as high compared
to the films in the previous section. This results in a time constant that is lower than that of
the dropcast films in the range of masses of the freeze dried films resulting in better rate
performance. The values extracted from the line-fits in figure 7.21 can be used to describe
the behaviour of capacitance per unit area with thickness by substituting for CA and τ in
equation 7.5.
1 10 100 100010
100
1000
10000
0.02V/s
0.1V/s
.5V/s
C/A
(F
/m2)
M/A (g/m2)
Figure 7.22: The dependence of capacitance per unit area with mass per unit area at 0.02 V/s (black), 0.1 V/s (red), 0.5 V/s (blue).
The model predicts the capacitance per unit area of the freeze dried films well,
particularly for the lower scan rates. However, the thickness of the freeze dried electrodes
compared to the dropcast electrodes results in a reduced volumetric capacitance which is
important for devices in which compact energy storage is required.
130
7.4 Conclusions
In this chapter, supercapacitor electrodes were fabricated by various casting methods and
subsequently doped in order to improve the conductivity of the electrode. A wide range of
thicknesses were produced with the aim of understanding the effects of thickness on
conductivity, capacitance and time constant.
Two models were used to describe the scan rate behaviour of capacitance. The
Higgins model fit the J/s-V curves well for low thicknesses and scan rates. However, as both
thickness and scan rate increased, the Pell and Conway model provided better insight into the
performance of the electrodes.
The effect of thickness on sheet resistance of the films was used to find the electrical
conductivity of the doped PEDOT:PSS films which was approximately 900 S/cm.
Comparing experimental data to theory, an intrinsic capacitance per unit mass of 34 F/g was
obtained. This capacitance was comparable to other work on PEDOT films accounting for
the mass of PSS in the film.
The electrical properties of the electrolyte, which were dependent on the electrode
thickness, were analysed by comparing the variation of the time constant, which is dependent
on the resistance of the system, with mass per unit area. Once the dependence of the time
constant was ascertained, the dependence of capacitance per unit area on mass per unit area
was compared to theory with good agreement. The maximum capacitance reached was 0.6
F/cm2
with a sheet resistance of less than 0.1 Ω/sq. The capacitance was comparable to other
work on PEDOT based capacitor electrodes, however, the sheet resistance was an order of
magnitude lower than similar electrodes326
.
Analysis of the theory led to establishing dimensional and rate-based limitations for
the design of the electrodes. This information is critical when designing devices for particular
applications as there is no ‘one size fits all’ for supercapacitors. The effect of diffusion on the
electrodes was found not to be as significant as that of the effect on the electrical properties
of the electrode and the electrolyte.
131
In order to improve charge transport of the electrolyte and limit the effect of diffusion,
the PEDOT:PSS dispersion was freeze dried. This led to an improved intrinsic capacitance
per unit mass of 38 F/g, coupled with an increase in the electrolyte conductivity by a factor
of two. These electrodes, however, were much less dense which would prove problematic in
compact device applications. Freeze drying a more concentrated dispersion may provide
suitable porosity while also keeping the film thickness in an appropriate range.
132
8
Transparent PEDOT:PSS Supercapacitors with Graphene
Current Collectors
8.1 Introduction
In this chapter the merits of adding a current collector to PEDOT:PSS transparent
supercapacitor electrodes will be discussed. Transparent supercapacitors with PEDOT:PSS
have been extensively investigated by Higgins and Coleman resulting in capacitances of 1
mF/cm2 at transparencies of 80 %
196.
Graphene has also seen application in transparent supercapacitors. 20 µF/cm2
is the
theoretical capacitance per unit area of graphitic carbon122
. The high surface area of graphene
allows a maximum capacitance per unit mass of 526 F/g. For transparent supercapacitors the
amount of graphene per unit area has to be constrained to limit light absorption. 4 layers of
graphene would have a transparency below 90 % and a mass per unit area of 3.8 x 10-8
g/cm2.
This leads to a theoretical capacitance per unit area in the order of tens of micro-farads per
centimetre squared.
Chen et al used wrinkled graphene as a stretchable flexible transparent electrode material
using a solid electrolyte. The wrinkling increases the mass per unit area and allows for
stretching. The total transparency of the device was 57 % and the capacitance was 5.8
µF/cm2.355
A combination of CVD graphene as a current collector and graphene oxide
quantum dots for enhanced capacitive properties, fabricated by Lee et al, reported a
capacitance of 9.09 µF/cm2 with an electrode transparency of 92.97 % at 550 nm
356. Yoo et al
report a high capacitance of 80 µF/cm2 for in plane graphene devices comprising of 2
graphene layers357
.
133
The low capacitances of graphene limit its application as the capacitive element in
supercapacitor electrodes. The excellent electrical and mechanical properties of graphene
make the use of graphene feasible as the current collector element of the electrode used in
conjunction with materials with higher capacitances with better transparencies.
Carbon nanotubes have also been demonstrated as a potential material in transparent
capacitors136,195,197,358
. As have composites with carbon nanotubes359–361
. Transition metal
oxides with high capacitances per unit mass have also been demonstrated in transparent
supercapacitors188,362
.
In this chapter PEDOT:PSS will be deposited on both PET (Polyethylene terephthalate) and
graphene substrates. The effect on the optoelectronic and capacitive properties of the
PEDOT:PSS film of the substrate will be investigated to determine the effectiveness of
graphene as a current collector for transparent PEDOT;PSS supercapacitor electrodes. The
dependence of these properties on the length of the electrode will also be investigated in
order optimize device design parameters.
8.2 Experimental Procedure
8.2.1 PEDOT:PSS and PEDOT:PSS on Graphene Film Preparation
Clevios PH1000 PEDOT:PSS dispersion was purchased from Hereus. 95% Formic Acid was
purchased from Sigma Aldrich. 4 layer CVD graphene on PET substrate was provided by
Strupinski et al363
.
The provided PEDOT:PSS solution was diluted to both 0.2 and 0.4 mg/ml. The method of
producing the film whether on a PET (Polyethylene terephthalate) substrate or a graphene-on-
PET substrate is the same as done by McCarthy et al325
. The substrate was heated to 130⁰C
and the diluted PEDOT:PSS solution was fed to a Sud-ko airbrush spray gun using a pipette
tube as a reservoir.
The material was deposited by spraying using a Janome JR2300N robot over an area of 5 by
5 cm2. The robot followed a raster pattern with line spacing of 0.5 cm with the airbrush
substrate distance being approximately 10 cm. The driving gas to the airbrush spray gun was
nitrogen with a pressure in the range of 45-55 psi.
134
Once the material was deposited the films were dipped in formic acid for 5 seconds to
improve the conductivity before being allowed to dry in air.
8.2.2 Optical and Electrical Characterisation
UV-Vis transmission spectra were taken in the wavelength range of 350 to 700 nm using a
Cary 600i spectrophotometer using a PET reference for comparison. The electrical properties
of the film were examined by running IV curves using the four wire method with a Keithley
2400 sourcemeter in the range of 0 to 1V. The film was contacted using an alcohol based
silver paint and silver wires.
8.2.3 Electrochemical Characterisation
Electrochemical characterisation was carried out using a Gammry Ref 3000 potentiostat. All
measurements were taken in a three electrode setup using a silver/silver chloride (Ag/AgCl)
reference electrode and a graphite rod as the counter electrode. The electrolyte used was
0.5M Potassium Sulphate (K2SO4).
Cyclic voltammograms (CVs) were taken at a variety of scan rates from 0.1 V/s to 5 V/s. The
voltage window was between 0.1 V and 0.8 V versus Ag/AgCl. The electrode was contacted
with silver paint at the top which was insulted using varnish. Potentiostatic EIS Spectra were
taken at 0.4 V versus Ag/AgCl using an AC voltage perturbation of 10 mV.
8.3 Results and Discussion
8.3.1 Optoelectronic Properties
Transparent films in a transparency range of approximately 95-80 % at 550 nm were
fabricated on both PET and graphene substrates. The most and least transparent materials
deposited on PET is shown in figure 8.1. At the top of the films, the silver paint for
135
contacting is visible while the varnish to protect the silver paint from the electrolyte is less
visible. Each sample was prepared by depositing a known volume of a known concentration
over 5 x 5 cm2. Then a doping step was undertaken using formic acid to improve the
conductivity of the film. As such, samples will be referenced by the concentration of
PEDOT:PSS in dispersion times the volume sprayed with a tag for substrate as follows
pt(concentration in mg/ml x 10)x(volume sprayed in deposition)-substrate. The tag for the
substrate is given as PET or Gra (for graphene).
Figure 8.1 Sprayed PEDOT:PSS on PET - pt2x4-PET(left) and pt4x6-PET (right) scale bar = 1cm
For transparent electronics good transparency across the visible spectrum is required. To
characterize further the transparency of our deposited PEDOT:PSS on both PET and
Graphene-on-PET, UV-Vis spectrophotometry was analysed in the range from 350 to 700nm.
In figure 8.2. the transparencies of the films are shown in the wavelength range. For low
thicknesses of PEDOT:PSS the transparency increases with decreasing wavelength with few
visible features. The thicker PEDOT:PSS films show a feature around 450-480 nm. The
graphene film which consists of 4 layers has a maximum transparency around 90 % at the
highest wavelength and this decreases steadily as the wavelength decreases.
136
400 500 600 70050
60
70
80
90
100
pt2x4
pt4x4
pt4x6
Gra
Gra+pt2x4
Gra+pt4x4
T (
%)
Wavelength (nm)
Figure 8.1 UV-Vis spectra for a range of thicknesses of PEDOT:PSS on PET and Graphene-on-PET
Theoretically, a single layer of graphene has a transparency of 97.7 %244,364
at 550 nm the
transparency of the 4-layer graphene as measured to be 89.4 %, which is lower than what
would be expected theoretically of 91%.
Once the viability of the material’s transparency has been determined, the electrical
properties must be analysed. ITO has been a leading transparent conductor for some time
with transparencies in excess of 90 % and sheet resistances in the range of 10s of Ω/sq281
.
However, it lacks the mechanical stability to be effective in flexible and stretchable devices,
displaying irreversible loss of conductivity under strains in excess of 1 %. The industry
standard requires a sheet resistance < 100 Ω/sq. and a transparency > 90 % at 550 nm.
To analyse the competitiveness of the PEDOT:PSS films, the graphene film, and the
PEDOT:PSS on graphene films, the transparency of the films was plotted against the sheet
resistance as calculated from a 4-wire measurement as shown in figure 8.3.
137
50 100 150 200 250 300 350 40076
78
80
82
84
86
88
90
92
94
96
PEDOT:PSS
Graphene
Graphene + PEDOT
Tra
nspare
ncy @
550nm
(%
)
Sheet Resistance (/sq)
Figure 8.3 Transparency of PEDOT:PSS, graphene and PEDOT:PSS-on-graphene films compared with sheet resistance. Lines correspond to equation 1.
The transparency of a conductive film at a certain wavelength (λ) can be expressed as a
function of sheet resistance according to equation 8.1.
2
0
2
0)(
21)(
21)(
sDC
op
opR
Zt
ZT
Equation 8.1
where Z0 is the impedance of free space (377 Ω), σop is the optical conductivity of the
material at a given wavelength, and t is the thickness of the film. Sheet resistance (Rs) is
defined as Rs = tDC
1where DC is the DC conductivity of the material. The formic acid
treated PEDOT:PSS samples fit well to a line (black) with a ratio of optical conductivity to
DC conductivity of 48. This ratio is the figure of merit used for transparent conductors, the
higher this figure of merit the more suitable the material is for a transparent conductor. The
blue line corresponds to a figure of merit of 40. The graphene sample provided had a sheet
resistance of 337 Ω/sq. and a transparency of 89.4 %. This is a significantly higher resistance
at comparable transparencies due to the high absorption of light that occurs in graphene
making graphene an inferior transparent conductor.
As far a transparent conductors go, the doped PEDOT:PSS meets the required industrial
criteria for transparent conductors. The criteria are met when the ratio of optical to DC
138
conductivity is 35 according to equation 8.1. The value achieved in this work is consistent
with the work of McCarthy et al. which gives an average figure of merit of 50.
Other materials which have been used to fabricate transparent conductors include carbon
nanotubes278,280,281
, graphene251,280,364
, metal nanowires215,216,365,366
and conductive
polymers367
.
Carbon nanotubes have been extensively tested as transparent conductors and have a figure of
merit of 13281
for pristine films. However, acid treatments or annealing can improve this to
25-35 which approaches industry requires. The junction resistance in CNT films can be
overcome with small amounts of graphene flakes improving the ratio by 40 %280
.
The limitations of graphene as a transparent conductor were reviewed by De & Coleman364
where a variety of methods of production were investigated and the figures of merit were
compared. The values lie in groups around 0.7, 4.5 and 11. This is lower than even pristine
carbon nanotube films. They stipulate that a sheet resistance of 10 Ω/sq. at a transparency of
90% can be reached using doping to increase the product of mobility and charge carrier
density. Bae et al. use a layer by layer method to stack 4 nitric acid doped graphene sheets to
achieve a sheet resistance of 30 Ω/sq. (an order of magnitude lower than the graphene in this
study) at 90% transmittance251
.
8.3.2 Scan Rate Dependence of Capacitance
To evaluate the energy storage performance of the transparent PEDOT:PSS films on both
PET and graphene of different lengths, the samples were tested using cyclic voltammetry.
The potential range was 0.1 to 0.8 V versus an Ag/AgCl reference electrode. The width of the
electrodes was 9 mm and the length range test was from 4.5 cm to 1 cm.
Figure 8.4 displays the cyclic voltammograms of a 3cm long film of sample pt2x4-PET (a)
and a 3cm long film of pt2x4-Gra (b). The lowest and highest scan rates were 0.1V/s (red)
and 5V/s (green) respectively. Both films are approximately boxlike, representing ideal
capacitor-like behaviour. The relative size of the voltammogram at 5V/s, compared to the
voltammogram at 0.1V/s, is smaller for the film without graphene as a current collector. The
capacitance values are also higher for sample 2x4-Gra. This can be explained by the decrease
139
in time constant due to a decrease in film resistance. The effect of which is described by
equation 7.1 in the previous chapter.
0.0 0.2 0.4 0.6 0.8-6
-4
-2
0
2
4
6
pt2x4-Gra 3cm
pt2x4-PET 3cm
0.0 0.2 0.4 0.6 0.8-6
-4
-2
0
2
4
6
0.0 0.2 0.4 0.6 0.8
-10
-8
-6
-4
-2
0
2
4
6
8
10
j/s (
F/m
2)
j/s (
F/m
2)
pt4x4-PET 3cmpt4x4-PET 3cm
Voltage vs. Ag/AgCl (V)
0.0 0.2 0.4 0.6 0.8
-12
-10
-8
-6
-4
-2
0
2
4
6
8
10
(d)(c)
(b)
Voltage vs. Ag/AgCl (V)
(a)
Figure 8.4 Cyclic voltammograms of (a) 0.2mg/ml PEDOT:PSS x 4ml on PET (b) 0.2mg/ml PEDOT:PSS x 4ml on CVD (4 layer) Graphene (c) 0.4mg/ml PEDOT:PSS x 4ml on PET (d)
0.4mg/ml PEDOT:PSS x 4ml on CVD (4 layer) Graphene. The red line is the voltammogram at 0.1V/s and the green voltammogram is the one at 5V/s
For figure 8.4(c) and (d) corresponding to 3cm in length films of samples pt4x4-PET and
pt4x4-Gra respectively are almost indistinguishable to the eye, with the pt4x4-Gra sample
having a marginally higher capacitance. This is possibly due to the relative resistance of the
graphene film (337 Ω/sq.) being significantly higher than that of the PEDOT:PSS doped film
(50 Ω/sq) compared to the pt2x4 samples where the sheet resistance exceeds 100 Ω/sq.
At first glance it looks like the addition of the current collector only provides significant
improvement on the performance when the resistance of the PEDOT:PSS film is high enough
to be comparable to the graphene film. To inspect the effect on the capacitance per unit area,
the cyclic voltammetry curves were integrated to give capacitances per unit area for the range
of scan rates tested. The integration of the curve can be expressed as equation 7.4.
140
0.1 1
1
4.2cm
3cm
2.5cm
2cm
1.5cm
1cm
pt4x4-Gra pt4x4-PET
pt2x4-Gra pt2x4-PET
C/A
(F
/m2)
0.1 1
1
(d)(c)
(b)(a)
0.1 1
2
4
6
8
10
C/A
(F
/m2)
Scan Rate (V/s)
0.1 1
2
4
6
8
10
Scan Rate (V/s)
Figure 8.5 Scan rate dependence for various lengths ranging from 1cm(magenta) to over 4cm (black) for (a) 0.2mg/ml PEDOT:PSS x 4ml on PET (b) 0.2mg/ml PEDOT:PSS x 4ml on CVD (4 layer) Graphene (c) 0.4mg/ml PEDOT:PSS x 4ml on PET (d) 0.4mg/ml PEDOT:PSS x 4ml on
CVD (4 layer) Graphene
In figure 8.5 the capacitance per unit area of various films as a function of scan rate is plotted.
(a) and (b) compare pt2x4-PET and 2x4-Gra samples respectively. The capacitances are
higher as was clear from the cyclic voltammogram as is the improved performance at higher
rates. Figures 8.5(c) and (d) compare pt4x4-PET and 4x4-Gra samples. The increased
capacitance per unit area is more obvious. The effect on scan rate dependence in particular
for the longer films (black) is also better for films with the graphene substrate provideding
larger C/A values for corresponding rates especially for the large scan rate..
To extract the characteristic features, the intrinsic capacitance per unit area, CA, and time
constant, τ, the scan rate dependence was fitted to equation 7.4. The fit is good across the
range of films returning values for CA that are within about 5% of the average.
141
8.3.3 Length Dependence of Capacitance
In Fig 8.5 the scan rate dependence for various lengths was shown. The fittings of these
capacitance per unit area curves resulted in CA and τ for each length of film. CA should be
invariant with length. However τ is dependent on the electrical properties of the film which
are determined by the dimensions of the film, as seen in equation 7.7. A modified version of
this equation can be used to extract values for sheet resistance and electrolyte conductivity.
A
AsA
G
CLRC 2 equation 8.2
Figure 8.6 shows the time constant as a function of film length squared for the films in figure
8.5. The time constant varies linearly for all electrodes with length squared. By fitting to a
straight line to the data, the sheet resistance can be obtained from the slope and the electrolyte
conductivity can be retrieved from the intercept.
0.0000 0.0005 0.0010 0.0015 0.00200.0
0.5
1.0
pt2x4-PET
pt2x4-Gra
pt4x4-PET
pt4x4-Gra
(s)
L2 (m
2)
Figure 8.6 The time constant of various films: 0.2mg/ml x 4ml PEDOT:PSS on PET, 0.2mg/ml x 4ml PEDOT:PSS on CVD Graphene. 0.4mg/ml x 4ml PEDOT:PSS on PET and 0.4mg/ml x 4ml
PEDOT:PSS on CVD Graphene.
The average electrolyte conductivity for all films analysed is 21 mS/cm2 which is close to the
values established in the previous chapter (64 mS/cm2). Of greater interest to this work is the
142
resistance of the films. The transparency of all films has been plotted against both CA and Rs
derived from the fitting of equation 7.4 to the scan rate dependence in figure 8.7.
2 4 6 8 10 12 147678808284868890929496
T (
%)
CA (F/m
2)
(a) (b)
0 25 50 75 100 125 1507678808284868890929496
Rs (/sq)
Figure 8.7 (a) Transparency against intrinsic capacitance per unit area for all films on PET substrate (black) and on graphene substrate extracted from fits to CV data (red) (b)
transparency against sheet resistance for all films on PET substrate (black solid) and on graphene substrate (red solid)determined from fitting the time constants of the films dependence on length compared with the corresponding values of sheet resistance as
determined by the 4-wire measurement (hollow crossed points)
In figure 8.7(a) the capacitance of the films using graphene as a substrate suffer from lower
transparencies for the same values of CA. Transparency can be fit to CA by modifying
equation 8.1. and letting the thickness of the electrode, t = CA/(CMρ). Where CM is the
intrinsic capacitance per unit mass (34,000 F/kg from previous chapter) and ρ is the film
density (1200 kg/m3).
2
0
2
0)(
21)(
21)(
A
M
op
op CC
Zt
ZT
Equation 8.3
This equation is a good fit to the data for transparent capacitors for the PEDOT:PSS-on-PET
samples. Due to the lack of contribution of capacitance from the graphene layer while still
contributing to the transparency, the fit is significantly worse for the PEDOT:PSS-on-
graphene. By multiplying equation 8.3 by the transparency of the graphene which is 89.4 %
(indicated by the red line) the fit improves to describe the transparency and capacitance well.
For homogenous systems, a figure of merit for transparent capacitors could be the ratio of
optical conductivity to the product of intrinsic capacitance per unit mass and density. This
figure of merit must be below 2.9x10-5
S.m2.F
-1 to achieve intrinsic capacitances of 1 mF/cm
2
143
at 90% transparency. The figure of merit for the PEDOT:PSS only samples is 4.55x10-5
S.m2.F
-1. While PEDOT:PSS is above this threshold, addition of nanoparticles with high
specific capacitance may be an avenue to reaching this number. This, in addition to the metric
for sheet resistance, can be used for design of transparent capacitor electrodes. Figure 8.7(b)
shows sheet resistances agrees well with those found using the 4-wire measurement.
Taking the values for CA and τ and applying them to equation 7.4 the equation can be
compared to the length dependence of the films. In figure 8.8 a sample CVs for various
lengths of films is provided. As length increases the capacitance per unit area decreases. This
is due to the increased resistance that charge has to pass through to the current collector. In
figure 8.8. a CV of the pt2x4-Gra film at different lengths shows a box-like voltammogram at
a length of 1cm(red line). As length increases, the voltammogram deviates from the original
boxlike shape. Since the current is normalized by area, the area of the curve in the CV for
greater lengths is clearly smaller.
0.0 0.2 0.4 0.6 0.8-6
-4
-2
0
2
4
6
j/s (
F/m
2)
Voltage vs. Ag/AgCl (V)
pt2x4-Gra
Figure 8.8 Cyclic voltammograms for the pt2x4-Gra for various lengths from 1cm (red) to 3cm (green)
The effect of increased length can be expressed by substituting for τ in equation 7.4 with
equation 8.2.
A
AsA
A
AsAA
G
CLRCs
VG
CLRC
V
sC
A
C
2
2 exp11 Equation 8.4
144
When the values for CA and τ derived from the fitting of the scan rate dependence curves in
figure 8.5. are substituted into this equation, the equation provides a good fit as shown in
figure 8.9.
0.01 0.02 0.03 0.04 0.05
1
C/A
(F
/m2)
0.01 0.02 0.03 0.04 0.05
1
0.01 0.02 0.03 0.04 0.05
1
C/A
(F
/m2)
Length (m)
pt2x4-PET
pt4x4-PET pt4x4-Gra
pt2x4-Gra
0.01 0.02 0.03 0.04
2
4
6
8
10
Length (m)
(a) (b)
(d)(c)
Figure 8.9 Length dependence for various scan rates for (a) 0.2mg/ml PEDOT:PSS x 4ml on PET (b) 0.2mg/ml PEDOT:PSS x 4ml on CVD Graphene (c) 0.4mg/ml PEDOT:PSS x 4ml on PET
(d) 0.4mg/ml PEDOT:PSS x 4ml on CVD Graphene
The comparison with similar amounts of PEDOT:PSS on PET and graphene show in
improvements of electrical and capacitance properties which have a more significant effect
on the time constant. This improvement is more pronounced for lower amounts of
PEDOT:PSS, as the sheet resistance of the graphene substrate is more comparable to that of
the PEDOT:PSS.
While the effect of the addition of the graphene current collector is of interest, direct
comparison between films of similar transparency is more important for this application.
There is a trade off with the addition of graphene as the graphene absorbs as much as the
pt4x4 around 550nm. This means that there is a significant amount of material (in terms of
relative absorbance) in the PEDOT:PSS-on-Graphene films that is effectively not
contributing to the capacitance. Consulting figure 8.7(a) shows a significant gap in intrinsic
145
capacitance per unit area. The films pt4x6-PET and pt4x4-Gra have similar transparencies
and so a direct comparison over length and scan rate is possible.
Figure 8.10 (a) Scan rate dependence of capacitance per unit area of two samples with similar transparency (approximately 79%): pt4x4-Gra(black) and pt4x6-PET(red). Dashed line is the difference between the capacitances according to equation x.4 and dotted line is the ratio of capacitances. (b) Length dependence of capacitance per unit area of two samples with similar transparency (approximately 79%): pt4x4-Gra(black) and pt4x6-PET(red). Dashed line is the difference between the capacitances according to equation x.4 and dotted line is the ratio of capacitances.
Figure 8.10. provides comparison between the pt4x4-Gra and the pt4x6-PET for both
different scan rates when the sample length is 3 cm (a), and different lengths when the scan
rate is 0.5 V/s (b). The fits over the testing range for both samples, show no signs of
converging, with the fits at the extrema seeming to become parallel.
For better comparison, the difference in capacitance per unit area (dashed line) and ratio of
capacitance (dotted line) of the fits were plotted. The ratio decreases slowly for both scan rate
and length. The point at which the ratio would equal one is much larger in both scan rate and
length for practical purposes.
However, adjusting the equation by decreasing the time constant by a factor of 2 (equivalent
to reducing the resistance by half), allows the ratio to equal one for scan rates of 1 V/s at a
length of 3 cm or less than 5 cm at 0.5 V/s. The reducing the resistance by half requires the
transparent conductor layer to have the same resistance as the PEDOT:PSS layer using the
parallel resistors rule. This is a resistance in the range of 35-40 Ω which is achievable for N-
doped graphene. In summary, the graphene used for this study is of insufficient conductivity
and transparency to make a meaningful impact on transparent capacitor performance.
0.1 11
10 pt4x4-Gra
pt4x6-PET
Difference
Ratio
C/A
(F
/m2)
Scan Rate (V/s)
0.01 0.02 0.03 0.04 0.051
10
(b)
Length (m)
(a)
146
8.3.4 Impedance Spectroscopy
To assess the impedance of the electrodes, impedance spectroscopy from 1 MHz to 0.1 Hz at
0.4 V with respect to Ag/AgCl with a perturbation amplitude of 10 mV was performed. The
Nyquist plot in figure 8.11(a) gives lines with two characteristic regions: one with a phase of
approximately 90⁰ and at low frequencies and one with a phase near 45⁰ corresponding to
high frequencies. The frequency at which the electrode changes phase from 90⁰ to 45⁰ as
frequency decreases is shown in figure 8.11(b) and (c). For the shortest film the transition
occurs around 10 Hz and for the longest film the transition occurs around 1.5 Hz.
-100 -50 0 50 100 150 200
0
100
200
300
400
500
-Zim
g (
)
Zreal
()
4.25cm
3cm
2.5cm
2cm
1.5cm
1cm20
40
6080
100
Zim
g (
)
Zre
al (
)
0.1 1 10 100
10
100
1000(c)
(b)
Frequency (Hz)
(a)
-1
Figure 8.11 (a) Nyquist plot of pt4x4-Gra sample for various lengths (b) The real component of the Nyquist Plot against frequency (c) The imaginary component of the Nyquist plot
against frequency.
Using a modified analytical solution to the transmission line model along the length of a thin
film capacitor196,238
the impedance can be described by the resistive and capacitive elements
as well as the device dimensions. The impedance below the transition frequency is:
LWC
iR
W
LZ
A
s
3
Equation 8.5
Where ω is the angular frequency in radians per second and all other variables have been
previously defined.
147
Extrapolating the linear portion below the transition frequency upward/downward gives the
real impedance of the system, which can be plotted against the ratio of length to width of the
electrode which as shown in figure 8.12(a) for various films.
1 2 3 4 5
0
100
200
300
pt2x4-PET
pt2x4-Gra
pt4x4-PET
pt4x4-Gra
pt4x6-PET
Zre
al (
)
L/W
0 100 200 300 400
76
78
80
82
84
86
88
90
92
94
96
PEDOT EIS
Graphene + PEDOT EIS
PEDOT:PSS 4-wire
Graphene + PEDOT 4-wire
(b)
T (
%)
Rs (/sq)
(a)
Figure 8.12(a) The Real Impedance component plotted against the ratio of length to width of various samples (b) The sheet resistance extracted from the fitting of previous figure(solid
points) compared with values obtained from the 4-wire measurement (crossed points).
The slope of these straight line graphs is equivalent to the sheet resistance divided by 3. In
figure 8.12(b) the sheet resistances found using the impedance data is compared to the sheet
resistances as found by the 4-wire measurement. There is not an agreement between the
impedance measurement and the four wire measurement.
Analysis of the low frequency region of the imaginary component of the impedance shows an
inverse relationship with the angular frequency. In figure 8.13(a) the low frequency points
were plotted against the inverse of the angular frequency and fitted to a straight line. The
capacitance of the films was extracted according to the imaginary part of equation 8.5 and
plotted against area for various films to get CA in figure 8.13(b). For comparison,
transparency was plotted against the CA obtained from the impedance data and the
voltammetry to reasonably good agreement.
148
0.2 0.4 0.6 0.8
100200300400500600700800900
10001100
4.5cm
3cm
2.5cm
2cm
1.5cm
1cm
-Zim
g (
)
1/ (s/rad)
0.0001 0.0002 0.0003 0.00040
1x10-3
2x10-3
3x10-3
4x10-3
5x10-3
6x10-3
pt2x4-PET
pt2x4-Gra
pt4x4-PET
pt4x4-Gra
pt4x6-PET
C (
F)
A (m2)
2 4 6 8 10 12 14 167678
8082
848688
9092
9496
PEDOT EIS
Gra + PEDOT EIS
PEDOT CV
Gra + PEDOT CV
T (
%)
CA (F/m
2)
Figure 8.13(a) Imaginary component of the impedance against the inverse angular frequencies in the low frequency linear region of a pt4x4-Gra film. (b) The capacitance of
various films plotted against area (c) a comparison of the transparency against the intrinsic capacitance per unit area obtained from impedance data (solid points) and voltammetry
data (crossed points)
149
8.4 Conclusions
In this chapter CVD graphene was used as a current collector for transparent supercapacitors.
The effect of the current collector was to reduce the sheet resistance of the electrode which
would also reduce the time constant. A reduced time constant would hopefully provide better
performances at higher scan rates and lengths.
The CVD graphene provided for this study had a sheet resistance of 337 Ω/sq. at
approximately 90 % transparency. This compares unfavourably with the formic acid doped
PEDOT:PSS which has a sheet resistance around 60 Ω/sq. for the same transparency.
The electrochemical characteristics of the films as supercapacitor electrodes were analysed
using cyclic voltammetry and impedance spectroscopy. The intrinsic capacitance per unit
area derived from the analysis of the data produced by both methods were in agreement.
As with sheet resistance, the intrinsic capacitance per unit area was lower for the films
employing a graphene current collector. This was expected as graphene has a capacitance per
unit area two orders of magnitude lower than that of PEDOT:PSS. However, analysis of the
time constants revealed that the ratio between the capacitance per unit area did not decrease
sufficiently with either length or scan rate to encourage implementation.
Were the current collector to have a comparable conductivity to the PEDOT:PSS, this could
result in capacitances that exceed the performance of PEDOT:PSS alone for scan rates over 1
V/s at an electrode length of 3 cm or for scan rates of 0.5 V/s for electrode lengths of 5 cm.
This is achievable with chemically doped graphene251
.
Another avenue is to employ other materials as the current collector such as carbon nanotubes
or metal nanowires. These materials have seen successful application as transparent
conductors and may be useful in transparent capacitors also215,280,366
.
150
9
Conclusions and Future Work
9.1 Conclusions
In this thesis various materials with dimensions in the nanoscale have been investigated for
use in electrochemical systems. For dye-sensitized solar cells the challenge of replacing the
expensive platinum in the catalytic counter electrode with graphene was thoroughly
investigated. The influence of size of the flakes proved to be less effective for improving the
efficiency than increasing the film thickness. However, creating films with efficiencies in
excess of 80% of that achievable with platinum remained elusive. As such, addition of other
materials into the graphene film to bridge the gap between graphene and platinum was
attempted. By enhancing conductivity, particularly in the vertical direction which was limited
due to the anisotropy of charge transport in graphene networks, with carbon nanotubes the
efficiency became comparable to platinum. The addition of a more catalytic material, MoS2,
also produced similar results with the advantage of the material being cheaper due to its
presence in nature. On investigation of the performance of the electrodes using percolation
theory it was revealed that while the edges of MoS2 are more catalytically active the main
advantage of using the MoS2 was that the nanosheets were on average smaller using the same
processing conditions. The smaller sheets in the lateral dimensions allowed for a higher
length to area ratio increasing the percentage mass of the particles contributing to the
catalytic activity of the material.
Supercapacitor electrode materials in the form of PEDOT:PSS films prepared by a
variety of methods and treated with formic acid to increase conductivity were alsostudied in
this thesis. Two models were compared to describe the effect of increasing thickness on the
capacitance per unit area. Both models returned capacitances per unit area within 10% of
151
each other allowing both models to be considered accurate for investigation of that property.
When looking at the current-voltage characteristics or the time constant however, the Pell &
Conway model which accounts for an initial charge on the electrode performs better than the
Higgins & Coleman model which assumes no charge initially. The role of diffusion affects
the capacitance per unit area at higher scan rates. Neither of the models account for it and
from this the diffusion coefficient was estimated to be sm /103.63.5 210 which is about
30-40% that of the ions in water. When accounting for time constant and scan rate the effect
of the electrical properties of the film with length and thickness were found to have a greater
effect on film performance than diffusion. For completeness freeze dried foam of the same
material were fabricated to both reduce the effects of diffusion and increase internal surface
area of the films. The benefits of freeze drying are modest and when compared to the
disadvantages like additional processing steps and electrode thickness lead to a conclusion of
this method not being viable for this material system.
Optically transparent supercapacitor electrodes represent a solution to energy storage
for transparent electronics. PEDOT:PSS has a combination of good electrical properties at
high transparencies that allow for application as transparent conductors. In addition
PEDOT:PSS also has an appreciable capacitance which led to it being demonstrated as a
transparent supercapacitor electrode material. In an attempt to improve the electrical
properties of the film a 4 layer sheet of graphene was used as a current collector. The effect
on performance of the addition of the graphene was negligible at comparable transparencies
of PEDOT:PSS only electrodes. In order to provide a significant improvement to these
electrodes the conductive layer needs to be at least of the same sheet resistance as the
capacitive layer. While this will lead to a lower capacitance at low rates the performance at
higher rates could exceed the capacitance of the PEDOT:PSS only electrode.
9.2 Future Work
9.2.1 DSSC Counter Electrodes
In this thesis graphene composite counter electrodes were fabricated which had a
performance almost as good as a platinum counter electrode (96% of Pt cell). These 400nm
152
thick films however were not optically transparent which limits the possible applications. The
reduction in size of the graphene particles resulted in an increase in efficiency. By reducing
the size of graphene further, it may be possible to manufacture transparent graphene counter
electrodes with comparable performance to platinum counter electrodes. Below a certain size
all of the atoms in the graphene sheet will be chemically active limiting the effect of size with
performance. Such a study would be of significant interest. Carbon black can be purchased
with particle sizes as low as 20nm. Comparison of carbon to the smallest possible graphene
(both by performance and economic viability) will determine the better carbon allotrope for
counter electrodes.
As far as other two dimensional materials are concerned the molybdenum and
tungsten dichalcogenides are the most well-known and easily dispersed. A direct comparison
of these materials would be instructive as well as identifying other materials such as gallium
selenide and other III-VI layered semiconductors. However, molybdenum disulfide has an
advantage in being available in nature while the other materials require synthesis. That and
the toxicity of the selenide and telluride elements may lead to molybdenum disulfide being
the best material in this class for many applications.
Size selection for the TMDs (Transition Metal Dichalcogenides) is another avenue for
increasing efficiency. In this thesis it has been shown that MoS2 is 1.5 times more active per
flake length. While reduction of sizes in graphene is difficult the TMDs are regularly
produced in smaller sizes as evidenced by work by Backes et al232–234
. By combining the
enhanced activity per unit length and minimizing the flake size to maximise edge length per
unit mass, reasonable efficiencies could be achieved with minimal material.
In this work the focus has primarily been on graphene due to the relative cheapness of
the starting material. Carbon nanotubes have much better electrical properties in composites
due to a lower percolation threshold. This leads to a lower loading requirement for CNTs
which combined with the possibility of price reduction in the future (due to improved
synthesis methods) could result in CNTs being the dominant conductive additive in films
with non-conductive electrochemically active materials. As such proper analysis of
percolation of CNT composites for counter electrodes should be a priority.
Eliminating the FTO (fluorine doped tin oxide) as well as the platinum is also a priority due
to the increasing price of indium. Much work has been done on replacing this material as
mentioned multiple times throughout this thesis. What is interesting is that many materials
153
used in the transparent conductive electrodes, for example: PEDOT:PSS, CNTs and
graphene, are also catalysts in counter electrodes for DSSCs. There has been considerable
work in replacing either the platinum or the FTO but there has been considerably less in
replacing both simultaneously. This represents a significant research interest.
9.2.2 Supercapacitor Electrode Materials
In this thesis a variety of thicknesses are used to produce supercapacitor electrodes with
different performances and potential applications. One of the issues with the PEDOT:PSS
electrodes is the relatively low capacitance per unit mass of 37F/g. This is much lower than
traditional carbon based double layer materials which can have capacitances in excess of
100F/g. One way of addressing this is the addition of a pseudo capacitive material as
discussed in chapter 2. One of the leading materials for this application is Ruthenium Oxide
RuO2. The use of RuO2 nanoparticles in a PEDOT:PSS matrix produced by spray coating
was done by Zhang et al368
. This material produced capacitances of 1.2mF/cm2 at 93%
transparency which is an improvement on work done by Higgins et al196
.
After receiving some of these RuO2 nanoparticles from Zhang attempts were made to
introduce these nanoparticles to the thicker PEDOT:PSS films. This was achieved by
dispersing the RuO2 nanoparticle powder in the aqueous PEDOT:PSS dispersion via
sonication. Films were produced by dropcasting. The resultant film had a different texture on
either side due to the RuO2 nanoparticles not being stable in the aqueous medium and
precipitating out of the dispersion. The capacitance of these films were measured and
compared to the standard PEDOT:PSS electrodes. Due to a limited amount of material only
two films have so far been measured. However, the results are encouraging with 20% by
weight of RuO2 nanoparticles showing an improvement of almost a factor of 5 as seen in
figure 9.1.
Moving forward, improving the dispersion of the RuO2 particles in the PEDOT:PSS
dispersion is a priority. Once the dispersion is stabilised, further investigation of the variation
of capacitance per unit area with composition and scan rate to optimize the energy storage
properties of the electrode will be conducted. The dependence of the electrical properties will
154
also be in need of investigation but the work by Zhang et al. provide the expected dependence
of conductivity with composition.
10
100
1000
10% RuO2
20% RuO2
PEDOT:PSS only
C/A
@ 5
0m
V/s
(m
F/c
m2)
Thickness (m)
10 100
Figure 9.1 Capacitance per unit area of PEDOT:PSS/RuO2 composite films measure from CVs at 50mV/s
As an alternative to adding nanoparticles to form composites PEDOT:PSS can be
used to synthesise pseudocapacitive nanoparticles. Liu et al. add MnO2 nanoparticles to
PEDOT nanowires by simply soaking the PEDOT in KMnO4 solution369
. The capacitance
was increased by a factor of 4 after a soaking time of 10 minutes. In an attempt to replicate
this on the spray deposited PEDOT:PSS resulted in a film of a slightly different colour and
comparable transparency compared to an untreated PEDOT:PSS sprayed film. Attempts at
electrochemical testing resulted in failure. This was possibly due to a change in conductivity
as a result of the reaction. This was avoided by using a gold current collector in the case of
Liu et al. To compare the PEDOT:PSS to the MnO2 decorated PEDOT:PSS the material was
sprayed onto an ITO (Indium Tin Oxide) substrate which would serve as the current
collector. We used a concentration of 25mM KMnO4 and a treatment time of up to 1 minute.
Figure 2 displays the scan rate dependence of capacitance per unit area for sprayed
PEDOT:PSS films of similar thickness for soak times of up to one minute. At the longest
soak time a capacitance improvement of a factor of 3 is possible. As scan rate increases the
improvement is less significant due to the lower rate capability of the pseudocapacitive
material and a possible decrease in sheet resistance.
155
To improve on this further work on the optical properties is necessary to investigate if
this material can be competitive with the PEDOT:PSS/RuO2 nanoparticle composites
produced by Zhang et al. Also Gui et al used KRuO4 in a similar method to that of Liu et al to
produce RuO2 nanoparticle decorated nanowires this would also be of interest to possible
transparent supercapacitors370
.
100 10000.5
1
1.5
2
2.5
33.5
4
0s
5s
10s
30s
60s
C/A
(m
F/c
m2)
Scan Rate (mV/s)
Figure 9.2 Scan rate dependence of capacitance per unit area of various PEDOT:PSS films on ITO following soaking in 25mM KMnO4 solution
Ideally to make flexible electrodes the ITO film would be removed and replaced with
a flexible transparent conductor such as carbon nanotubes or graphene (as shown in chapter
7). The limitation of the graphene in chapter 7 with sheet resistance in excess of 300 Ω/sq.
needs to be addressed. As mention previously the doped graphene produced by Bae et al with
a sheet resistance of 30 Ω/sq. at a transparency of 90% would be an ideal replacement for the
ITO251
.
156
10
Bibliography
1. Perez, R. Making the case for solar energy. Daylight&Architecture 9, 5 (2009).
2. Battaglia, C., Cuevas, A. & De Wolf, S. High-efficiency crystalline silicon solar cells:
status and perspectives. Energy Environ. Sci. 9, 1552–1576 (2016).
3. Namba, S. & Hishiki, Y. Color Sensitization of Zinc Oxide with Cyanine Dyes. J.
Phys. Chem. 770, 774–779 (1964).
4. Gerischer, H. The impact of semiconductors on the concepts of electrochemistry.
Electrochim. Acta 35, 1677–1699 (1990).
5. Kline, G., Kam, K., Canfield, D. & Parkinson, B. A. EFFICIENT AND STABLE
PHOTOELECTROCHEMICAL CELLS CONSTRUCTED WITH WSe 2 AND MoSe
2 PHOTOANODES. Sol. Energy Mater. 4, 301–308 (1981).
6. Parkinson, B. A., Heller, A. & Miller, B. Enhanced photoelectrochemical solar-energy
conversion by gallium arsenide surface modification. Appl. Phys. Lett. 33, 521–523
(1978).
7. Licht, S., Hodes, G., Tenne, R. & Manassen, J. A light-variation insensitive high
efficiency solar cell. Nature 326, 863–864 (1987).
8. Robbins, M. et al. Solar Conversion Efficiency of Pressure Sintered Cadmium
Selenide Liquid Junction Cells. Abstr. Pap. Am. Chem. Soc. 124, 195 (1977).
9. Parkinson, B. a., Furtak, T. E., Canfield, D., Kam, K.-K. & Kline, G. Evaluation and
reduction of efficiency losses at tungsten diselenide photoanodes. Faraday Discuss.
Chem. Soc. 70, 233 (1980).
10. O’Regan, B. & Grätzel, M. A low-cost, high-efficiency solar cell based on dye-
sensitized colloidal TiO2 films. Nature 353, 737–740 (1991).
11. Yella, a. et al. Porphyrin-Sensitized Solar Cells with Cobalt (II/III)-Based Redox
Electrolyte Exceed 12 Percent Efficiency. Science (80-. ). 334, 629–634 (2011).
12. Snaith, H. J. Estimating the maximum attainable efficiency in Dye-sensitized solar
cells. Adv. Funct. Mater. 20, 13–19 (2010).
13. Ito, S. et al. Fabrication of thin film dye sensitized solar cells with solar to electric
power conversion efficiency over 10%. Thin Solid Films 516, 4613–4619 (2008).
157
14. Chou, T. P., Zhang, Q., Russo, B., Fryxell, G. E. & Cao, G. Titania Particle Size Effect
on the Overall Performance of Dye-Sensitized Solar Cells. J. Phys. Chem. C 111,
6296–6302 (2007).
15. Yu, J., Fan, J. & Lv, K. Anatase TiO(2) nanosheets with exposed (001) facets:
improved photoelectric conversion efficiency in dye-sensitized solar cells. Nanoscale
2, 2144–2149 (2010).
16. Zhang, H. et al. Anatase TiO2 microspheres with exposed mirror-like plane {001}
facets for high performance dye-sensitized solar cells (DSSCs). Chem. Commun. 46,
8395–8397 (2010).
17. Wang, Y., Yang, W. & Shi, W. Preparation and Characterization of Anatase TiO 2
Nanosheets-Based Microspheres for Dye-Sensitized Solar Cells. Ind. Eng. Chem. Res.
50, 11982–11987 (2011).
18. Dembele, K. T. et al. Hybrid carbon nanotubes-TiO2 photoanodes for high efficiency
dye-sensitized solar cells. J. Phys. Chem. C 117, 14510–14517 (2013).
19. Hore, S., Vetter, C., Kern, R., Smit, H. & Hinsch, A. Influence of scattering layers on
efficiency of dye-sensitized solar cells. Sol. Energy Mater. Sol. Cells 90, 1176–1188
(2006).
20. Yu, I. G., Kim, Y. J., Kim, H. J., Lee, C. & Lee, W. I. Size-dependent light-scattering
effects of nanoporous TiO2 spheres in dye-sensitized solar cells. J. Mater. Chem. 21,
532 (2011).
21. O’Regan, B. C., Durrant, J. R., Sommeling, P. M. & Bakker, N. J. Influence of the
TiCl 4 Treatment on Nanocrystalline TiO 2 Films in Dye-Sensitized Solar Cells . 2 .
Charge Density , Band Edge Shifts , and Quantification of Recombination Losses at
Short Circuit. J. Phys. Chem. C 111, 14001–14010 (2007).
22. Mor, G., Shankar, K., Paulose, M., Varghese, O. K. & Grimes, C. A. Use of Highly-
Ordered TiO2 Nanotube Arrays in Dye-Sensitized Solar Cells. Nano Lett. 6, 215–218
(2006).
23. Sauvage, F. et al. Hierarchical TiO2 photoanode for dye-sensitized solar cells. Nano
Lett. 10, 2562–2567 (2010).
24. Tang, Y. et al. Incorporation of Graphenes in Nanostructured TiO 2 Films via
Molecular. ACS Nano 4, 3482–3488 (2010).
25. Wang, J. T. et al. Low-Temperature Processed Electron Collection Layers of
Graphene/ TiO 2 Nanocomposites in Thin Film Perovskite Solar Cells. (2014).
doi:10.1021/nl403997a
26. Jiang, C. Y., Sun, X. W., Lo, G. Q., Kwong, D. L. & Wang, J. X. Improved dye-
sensitized solar cells with a ZnO-nanoflower photoanode. Appl. Phys. Lett. 90, 14–17
(2007).
27. Wu, J.-J., Chen, G.-R., Yang, H.-H., Ku, C.-H. & Lai, J.-Y. Effects of dye adsorption
on the electron transport properties in ZnO-nanowire dye-sensitized solar cells. Appl.
Phys. Lett. 90, 213109 (2007).
158
28. Milan, R. et al. ZnO@SnO2 engineered composite photoanodes for dye sensitized
solar cells. Sci. Rep. 5, 14523 (2015).
29. Dong, Z. et al. Quintuple-shelled Sno2 hollow microspheres with superior light
scattering for high-performance dye-sensitized solar cells. Adv. Mater. 26, 905–909
(2014).
30. Prasittichai, C. & Hupp, J. T. Surface Modification of SnO 2 Photoelectrodes in Dye-
Sensitized Solar Cells: Significant Improvements in Photovoltage via Al 2 O 3 Atomic
Layer Deposition. J. Phys. Chem. Lett. 1, 1611–1615 (2010).
31. Ghosh, R. et al. Nanoforest Nb2O5 photoanodes for dye-sensitized solar cells by
pulsed laser deposition. ACS Appl. Mater. Interfaces 3, 3929–35 (2011).
32. Ou, J. Z. et al. Elevated temperature anodized Nb 2O 5: A photoanode material with
exceptionally large photoconversion efficiencies. ACS Nano 6, 4045–4053 (2012).
33. Li, N. et al. Dou ble-layered NiO photocathodes for p-Type DSSCs with record IPCE.
Adv. Mater. 22, 1759–1762 (2010).
34. Jena, A. et al. Dye Sensitized Solar Cells : A Review. Trans. Indian Ceram. Soc. 71,
1–16 (2012).
35. Rühle, S., Shalom, M. & Zaban, A. Quantum-dot-sensitized solar cells.
ChemPhysChem 11, 2290–2304 (2010).
36. Kawano, R. et al. High performance dye-sensitized solar cells using ionic liquids as
their electrolytes. J. Photochem. Photobiol. A Chem. 164, 87–92 (2004).
37. Wang, P. et al. A stable quasi-solid-state dye-sensitized solar cell with an amphiphilic
ruthenium sensitizer and polymer gel electrolyte. Nat. Mater. 2, 402–7 (2003).
38. Fabregat-Santiago, F. et al. Electron transport and recombination in solid-state dye
solar cell with spiro-OMeTAD as hole conductor. J. Am. Chem. Soc. 131, 558–562
(2009).
39. Calogero, G. et al. A new type of transparent and low cost counter-electrode based on
platinum nanoparticles for dye-sensitized solar cells. Energy Environ. Sci. 4, 1838
(2011).
40. Mukherjee, S. et al. Ultrafine sputter-deposited Pt nanoparticles for triiodide reduction
in dye-sensitized solar cells: impact of nanoparticle size, crystallinity and surface
coverage on catalytic activity. Nanotechnology 23, 485405 (2012).
41. Hsieh, T.-L. et al. A highly efficient dye-sensitized solar cell with a platinum
nanoflowers counter electrode. J. Mater. Chem. 22, 5550–5559 (2012).
42. Olsen, E., Hagen, G. & Eric Lindquist, S. Dissolution of platinum in methoxy
propionitrile containing LiI/I2. Sol. Energy Mater. Sol. Cells 63, 267–273 (2000).
43. Murakami, T. N. et al. Highly Efficient Dye-Sensitized Solar Cells Based on Carbon
Black Counter Electrodes. J. Electrochem. Soc. 153, A2255 (2006).
44. Lin, C. et al. Dye sensitized solar cells with carbon black as counter electrodes.
159
Japenese J. Apllied Physcs 55, 03CE01 (2016).
45. Ramasamy, E. & Lee, J. Large-pore sized mesoporous carbon electrocatalyst for
efficient dye-sensitized solar cells. Chem. Commun. (Camb). 46, 2136–2138 (2010).
46. Davies, T. J., Hyde, M. E. & Compton, R. G. Nanotrench arrays reveal insight into
graphite electrochemistry. Angew. Chemie - Int. Ed. 44, 5121–5126 (2005).
47. Banks, C. E., Davies, T. J., Wildgoose, G. G. & Compton, R. G. Electrocatalysis at
graphite and carbon nanotube modified electrodes: edge-plane sites and tube ends are
the reactive sites. Chem. Commun. (Camb). 829–841 (2005). doi:10.1039/b413177k
48. Yuan, W. et al. The edge- and basal-plane-specific electrochemistry of a single-layer
graphene sheet. Sci. Rep. 3, 2248 (2013).
49. Veerappan, G., Bojan, K. & Rhee, S. W. Sub-micrometer-sized graphite as a
conducting and catalytic counter electrode for dye-sensitized solar cells. ACS Appl.
Mater. Interfaces 3, 857–862 (2011).
50. Kavan, L., Yum, J. H. & Gra, M. Optically Transparent Cathode for Dye-Sensitized
Solar Cells Based on Graphene Nanoplatelets. ACS Nano 5, 165–172 (2011).
51. Kavan, L., Yum, J.-H. & Grätzel, M. Graphene Nanoplatelets Outperforming Platinum
as Electrocatalyst in Co-Bipyridine Mediated Dye-Sensitized Solar Cells. Nano Lett.
11, 5501–5506 (2011).
52. William S. Hummers, J. & Offeman, R. E. Preparation of Graphitic Oxide. J. Am.
Chem. Soc 80, 1339 (1958).
53. Stankovich, S. et al. Stable aqueous dispersions of graphitic nanoplatelets via the
reduction of exfoliated graphite oxide in the presence of poly(sodium 4-
styrenesulfonate). J. Mater. Chem. 16, 155–158 (2006).
54. Chen, C. M. et al. Annealing a graphene oxide film to produce a free standing high
conductive graphene film. Carbon N. Y. 50, 659–667 (2012).
55. Stankovich, S. et al. Synthesis of graphene-based nanosheets via chemical reduction of
exfoliated graphite oxide. Carbon N. Y. 45, 1558–1565 (2007).
56. Wan, L. et al. Room-temperature fabrication of graphene films on variable substrates
and its use as counter electrodes for dye-sensitized solar cells. Solid State Sci. 13, 468–
475 (2011).
57. Zhang, D. W. et al. Graphene-based counter electrode for dye-sensitized solar cells.
Carbon N. Y. 49, 5382–5388 (2011).
58. Roy-Mayhew, J. D., Boschloo, G., Hagfeldt, A. & Aksay, I. A. Functionalized
graphene sheets as a versatile replacement for platinum in dye-sensitized solar cells.
ACS Appl. Mater. Interfaces 4, 2794–2800 (2012).
59. Hou, S. et al. Nitrogen-doped graphene for dye-sensitized solar cells and the role of
nitrogen states in triiodide reduction. Energy Environ. Sci. 6, 3356 (2013).
60. Wang, H., Sun, K., Tao, F., Stacchiola, D. J. & Hu, Y. H. 3D honeycomb-like
160
structured graphene and its high efficiency as a counter-electrode catalyst for dye-
sensitized solar cells. Angew. Chemie - Int. Ed. 52, 9210–9214 (2013).
61. Seo, H. K., Song, M., Ameen, S., Akhtar, M. S. & Shin, H. S. New counter electrode
of hot filament chemical vapor deposited graphene thin film for dye sensitized solar
cell. Chem. Eng. J. 222, 464–471 (2013).
62. Pan, X. et al. Electrocatalytic properties of a vertically oriented graphene film and its
application as a catalytic counter electrode for dye-sensitized solar cells. J. Mater.
Chem. A 2, 12746–12753 (2014).
63. Ramasamy, E., Lee, W. J., Lee, D. Y. & Song, J. S. Spray coated multi-wall carbon
nanotube counter electrode for tri-iodide (I3-) reduction in dye-sensitized solar cells.
Electrochem. commun. 10, 1087–1089 (2008).
64. Lee, W. J., Ramasamy, E., Lee, D. Y. & Song, J. S. Efficient dye-sensitized solar cells
with catalytic multiwall carbon nanotube counter electrodes. ACS Appl Mater
Interfaces 1, 1145–1149 (2009).
65. Nam, J. G., Park, Y. J., Kim, B. S. & Lee, J. S. Enhancement of the efficiency of dye-
sensitized solar cell by utilizing carbon nanotube counter electrode. Scr. Mater. 62,
148–150 (2010).
66. Lee, K. S., Lee, W. J., Park, N.-G., Kim, S. O. & Park, J. H. Transferred vertically
aligned N-doped carbon nanotube arrays: use in dye-sensitized solar cells as counter
electrodes. Chem. Commun. (Camb). 47, 4264–4266 (2011).
67. Pringle, J. M., Armel, V. & MacFarlane, D. R. Electrodeposited PEDOT-on-plastic
cathodes for dye-sensitized solar cells. Chem. Commun. 46, 5367 (2010).
68. Lee, K.-M. et al. A high-performance counter electrode based on poly(3,4-
alkylenedioxythiophene) for dye-sensitized solar cells. J. Power Sources 188, 313–318
(2009).
69. Jeon, S. S., Kim, C., Ko, J. & Im, S. S. Spherical polypyrrole nanoparticles as a highly
efficient counter electrode for dye-sensitized solar cells. J. Mater. Chem. 21, 8146
(2011).
70. Tai, S. Y., Chang, C. F., Liu, W. C., Liao, J. H. & Lin, J. Y. Optically transparent
counter electrode for dye-sensitized solar cells based on cobalt sulfide nanosheet
arrays. Electrochim. Acta 107, 66–70 (2013).
71. Wu, M., Lin, X., Hagfeldt, A. & Ma, T. Low-cost molybdenum carbide and tungsten
carbide counter electrodes for dye-sensitized solar cells. Angew. Chemie - Int. Ed. 50,
3520–3524 (2011).
72. Jang, J. S., Ham, D. J., Ramasamy, E., Lee, J. & Lee, J. S. Platinum-free tungsten
carbides as an efficient counter electrode for dye sensitized solar cells. Chem.
Commun. (Camb). 46, 8600–8602 (2010).
73. Li, G. R., Song, J., Pan, G. L. & Gao, X. P. Highly Pt-like electrocatalytic activity of
transition metal nitrides for dye-sensitized solar cells. Energy Environ. Sci. 4, 1680
(2011).
161
74. Wu, M. et al. Synthesis of highly effective vanadium nitride (VN) peas as a counter
electrode catalyst in dye-sensitized solar cells. J. Phys. Chem. C 118, 12625–12631
(2014).
75. Song, J. et al. Enhancement of diffusion kinetics in porous MoN nanorods-based
counter electrode in a dye-sensitized solar cell. J. Mater. Chem. A 2, 10041 (2014).
76. Jiang, Q. W., Li, G. R. & Gao, X. P. Highly ordered TiN nanotube arrays as counter
electrodes for dye-sensitized solar cells. Chem. Commun. (Camb). 6720–2 (2009).
doi:10.1039/b912776c
77. Lin, X., Wu, M., Wang, Y., Hagfeldt, A. & Ma, T. Novel counter electrode catalysts of
niobium oxides supersede Pt for dye-sensitized solar cells. Chem. Commun. 47, 11489
(2011).
78. Hou, Y. et al. Highly electrocatalytic activity of RuO2 nanocrystals for triiodide
reduction in dye-sensitized solar cells. Small 10, 484–492 (2014).
79. Hou, Y. et al. Rational screening low-cost counter electrodes for dye-sensitized solar
cells. Nat. Commun. 4, 1583 (2013).
80. Wu, M., Lin, X., Hagfeldt, A. & Ma, T. A novel catalyst of WO2 nanorod for the
counter electrode of dye-sensitized solar cells. Chem. Commun. (Camb). 47, 4535–
4537 (2011).
81. Wu, M. et al. Economical Pt-Free Catalysts for Counter Electrodes of Dye- Sensitized
Solar Cells. J. Am. Chem. Soc. 134, 3419–3428 (2012).
82. Lin, J.-Y., Liao, J.-H. & Wei, T.-C. Honeycomb-like CoS Counter Electrodes for
Transparent Dye-Sensitized Solar Cells. Electrochem. Solid-State Lett. 14, D41 (2011).
83. Gong, F., Xu, X., Li, Z., Zhou, G. & Wang, Z.-S. NiSe2 as an efficient electrocatalyst
for a Pt-free counter electrode of dye-sensitized solar cells. Chem. Commun. (Camb).
49, 1437–9 (2013).
84. Guo, J. et al. Transition metal selenides as efficient counter-electrode materials for
dye-sensitized solar cells. Phys. Chem. Chem. Phys. 17, 28985–28992 (2015).
85. Wu, M. et al. Economical and effective sulfide catalysts for dye-sensitized solar cells
as counter electrodes. Phys. Chem. Chem. Phys. 13, 19298 (2011).
86. Chen, X., Hou, Y., Zhang, B., Yang, X. H. & Yang, H. G. Low-cost SnS(x) counter
electrodes for dye-sensitized solar cells. Chem. Commun. (Camb). 49, 5793–5 (2013).
87. Lei, B., Li, G. R. & Gao, X. P. Morphology dependence of molybdenum disulfide
transparent counter electrode in dye-sensitized solar cells. J. Mater. Chem. A 2, 3919
(2014).
88. Wu, M.-S. & Zheng, Y.-J. Electrophoresis of randomly and vertically embedded
graphene nanosheets in activated carbon film as a counter electrode for dye-sensitized
solar cells. Phys. Chem. Chem. Phys. 15, 1782–7 (2013).
89. Zhu, G. et al. Electrophoretic deposition of reduced graphene-carbon nanotubes
composite films as counter electrodes of dye-sensitized solar cells. J. Mater. Chem.
162
5621, 10288–1029148691 (2011).
90. Battumur, T. et al. Graphene/carbon nanotubes composites as a counter electrode for
dye-sensitized solar cells. Curr. Appl. Phys. 12, e49–e53 (2012).
91. Shin, H. J., Jeon, S. S. & Im, S. S. CNT/PEDOT core/shell nanostructures as a counter
electrode for dye-sensitized solar cells. Synth. Met. 161, 1284–1288 (2011).
92. Hong, W., Xu, Y., Lu, G., Li, C. & Shi, G. Transparent graphene/PEDOT-PSS
composite films as counter electrodes of dye-sensitized solar cells. Electrochem.
commun. 10, 1555–1558 (2008).
93. Nagarajan, S. et al. A PEDOT-reinforced exfoliated graphite composite as a Pt- and
TCO-free flexible counter electrode for polymer electrolyte dye-sensitized solar cells.
J. Mater. Chem. A 1, 1048–1054 (2013).
94. Chen, L. et al. Graphene quantum-dot-doped polypyrrole counter electrode for high-
performance dye-sensitized solar cells. ACS Appl. Mater. Interfaces 5, 2047–2052
(2013).
95. Gao, Y. et al. Improvement of adhesion of Pt-free counter electrodes for low-cost dye-
sensitized solar cells. J. Photochem. Photobiol. A Chem. 245, 66–71 (2012).
96. Yue, G., Zhang, W., Wu, J. & Jiang, Q. Glucose aided synthesis of molybdenum
sulfide/carbon nanotubes composites as counter electrode for high performance dye-
sensitized solar cells. Electrochim. Acta 112, 655–662 (2013).
97. Song, J., Li, G. R., Xiong, F. Y. & Gao, X. P. Synergistic effect of molybdenum
nitride and carbon nanotubes on electrocatalysis for dye-sensitized solar cells. J.
Mater. Chem. 22, 20580–20585 (2012).
98. Guo, J. W. et al. A sulfur-assisted strategy to decorate MWCNTs with highly
dispersed Pt nanoparticles for counter electrode in dye-sensitized solar cells. J. Mater.
Chem. A 1, 1982 (2013).
99. Lin, J.-Y., Chan, C.-Y. & Chou, S.-W. Electrophoretic deposition of transparent
MoS2-graphene nanosheet composite films as counter electrodes in dye-sensitized
solar cells. Chem. Commun. (Camb). 49, 1440–2 (2013).
100. Liu, C.-J. et al. Facile synthesis of MoS2/graphene nanocomposite with high catalytic
activity toward triiodide reduction in dye-sensitized solar cells. J. Mater. Chem. 22,
21057 (2012).
101. Dou, Y. Y., Li, G. R., Song, J. & Gao, X. P. Nickel phosphide-embedded graphene as
counter electrode for dye-sensitized solar cells. Phys. Chem. Chem. Phys. 14, 1339
(2012).
102. Bajpai, R., Roy, S., Koratkar, N. & Misra, D. S. NiO nanoparticles deposited on
graphene platelets as a cost-effective counter electrode in a dye sensitized solar cell.
Carbon N. Y. 56, 56–63 (2013).
103. Li, Z., Gong, F., Zhou, G. & Wang, Z.-S. NiS 2 /Reduced Graphene Oxide
Nanocomposites for Efficient Dye-Sensitized Solar Cells. J. Phys. Chem. C 117,
6561–6566 (2013).
163
104. Yue, G., Lin, J. Y., Tai, S. Y., Xiao, Y. & Wu, J. A catalytic composite film of
MoS2/graphene flake as a counter electrode for Pt-free dye-sensitized solar cells.
Electrochim. Acta 85, 162–168 (2012).
105. Das, S. et al. Synthesis of graphene-CoS electro-catalytic electrodes for dye sensitized
solar cells. Carbon N. Y. 50, 4815–4821 (2012).
106. Bajpai, R. et al. Graphene supported platinum nanoparticle counter-electrode for
enhanced performance of dye-sensitized solar cells. ACS Appl. Mater. Interfaces 3,
3884–9 (2011).
107. Sudhagar, P. et al. Synergistic catalytic effect of a composite (CoS/PEDOT:PSS)
counter electrode on triiodide reduction in dye-sensitized solar cells. ACS Appl. Mater.
Interfaces 3, 1838–1843 (2011).
108. Yeh, M.-H. et al. A composite catalytic film of PEDOT:PSS/TiN–NPs on a flexible
counter-electrode substrate for a dye-sensitized solar cell. J. Mater. Chem. 21, 19021
(2011).
109. Liu, G. et al. A class of carbon supported transition metal–nitrogen complex catalysts
for dye-sensitized solar cells. J. Mater. Chem. A 1, 1475 (2013).
110. Simon, P. & Gogotsi, Y. Materials for electrochemical capacitors. Nat. Mater. 7, 845–
54 (2008).
111. Zhang, Y. et al. Progress of electrochemical capacitor electrode materials: A review.
Int. J. Hydrogen Energy 34, 4889–4899 (2009).
112. Ko, R. & Carlen, M. Principles and applications of electrochemical capacitors.
Electrochim. Acta 45, 2483–2498 (2000).
113. González, A., Goikolea, E., Andoni, J. & Mysyk, R. Review on supercapacitors :
Technologies and materials. 58, 1189–1206 (2016).
114. Bard, A. J. & Faulkner, L. R. Electrochemical Methods: Fundamentals and
Applications. (Wiley, 2000).
115. Frackowiak, E. Carbon materials for supercapacitor application. Phys. Chem. Chem.
Phys. 9, 1774–1785 (2007).
116. Macdonald, J. R. & Barlow, C. A. Theory of Double‐Layer Differential Capacitance in
Electrolytes. J. Chem. Phys. 36, (1962).
117. Conway, B. E., Bockris, J. O. & Ammar, I. a. The dielectric constant of the solution in
the diffuse and Helmholtz double layers at a charged interface in aqueous solution.
Trans. Faraday Soc. 47, 756 (1951).
118. Dzubiella, J. & Hansen, J. P. Electric-field-controlled water and ion permeation of a
hydrophobic nanopore. J. Chem. Phys. 122, 1–14 (2005).
119. Teschke, O. & De Souza, E. F. Dielectric exchange: The key repulsive or attractive
transient forces between atomic force microscope tips and charged surfaces. Appl.
Phys. Lett. 74, 1755–1757 (1999).
164
120. Zhong, C. et al. A review of electrolyte materials and compositions for
electrochemical supercapacitors. Chem. Soc. Rev. 44, 7484–7539 (2015).
121. Nishino, A. Capacitors: operating principles, current market and technical trends. J.
Power Sources 60, 137–147 (1996).
122. Shi, H. Activated carbons and double layer capacitance. Electrochim. Acta 41, 1633–
1639 (1996).
123. Qu, D., Qu, D., Shi, H. & Shi, H. Studies of activated carbons used in double-layer
capacitors. Construction 99–107 (1998). doi:10.1016/S0378-7753(98)00038-X
124. Zhang, L. L., Zhou, R. & Zhao, X. S. Carbon-based materials as supercapacitor
electrodes. J. Mater. Chem. 38, 2520–2531 (2009).
125. Okamura, M. Electric double layer capacitor. (2000).
126. Hu, Z., Srinivasan, M. P. & Ni, Y. Preparation of Mesoporous High-Surface-Area
Activated Carbon. Adv. Mater. 12, 62–65 (2000).
127. Kyotani, T. Control of pore structure in carbon. Carbon N. Y. 38, 269–286 (2000).
128. Ryoo, R., Joo, S. H., Kruk, M. & Jaroniec, M. Ordered mesoporous carbons. Adv.
Mater. 13, 677–681 (2001).
129. Jurewicz, K. et al. Capacitance properties of ordered porous carbon materials prepared
by a templating procedure. J. Phys. Chem. Solids 65, 287–293 (2004).
130. Li, J., Cheng, X., Shashurin, A. & Keidar, M. Review of Electrochemical Capacitors
Based on Carbon Nanotubes and Graphene. Graphene 1, 1–13 (2012).
131. Du, C. & Pan, N. High power density supercapacitor electrodes of carbon nanotube
films by electrophoretic deposition. Nanotechnology 17, 5314–5318 (2006).
132. Gupta, V. & Miura, N. Polyaniline/single-wall carbon nanotube (PANI/SWCNT)
composites for high performance supercapacitors. Electrochim. Acta 52, 1721–1726
(2006).
133. Diederich, L. et al. Supercapacitors based on nanostructured carbon electrodes grown
by cluster-beam deposition. Appl. Phys. Lett. 75, 2662 (1999).
134. Ma, R. Z. et al. Study of electrochemical capacitors utilizing carbon nanotube
electrodes. J. Power Sources 84, 126–129 (1999).
135. Hu, L., Hecht, D. S. & Grüner, G. Carbon nanotube thin films: fabrication, properties,
and applications. Chem. Rev. 110, 5790–844 (2010).
136. Yuksel, R., Sarioba, Z., Cirpan, A., Hiralal, P. & Unalan, H. E. Transparent and
Flexible Supercapacitors with Single Walled Carbon Nanotube Thin Film Electrodes.
ACS Appl. Mater. Interfaces 15434–15439 (2014).
137. Choi, C. et al. Flexible supercapacitor made of carbon nanotube yarn with internal
pores. Adv. Mater. 26, 2059–2065 (2014).
138. An, K. H. et al. Electrochemical Properties of High-Power Supercapacitors Using
165
Single-Walled Carbon Nanotube Electrodes. Adv. Funct. Mater. 11, 387–392 (2001).
139. Yu, A., Roes, I., Davies, A. & Chen, Z. Ultrathin, transparent, and flexible graphene
films for supercapacitor application. Appl. Phys. Lett. 96, (2010).
140. Le, L. T., Ervin, M. H., Qiu, H., Fuchs, B. E. & Lee, W. Y. Graphene supercapacitor
electrodes fabricated by inkjet printing and thermal reduction of graphene oxide.
Electrochem. commun. 13, 355–358 (2011).
141. Zhang, L. L., Zhou, R. & Zhao, X. S. Graphene-based materials as supercapacitor
electrodes. Journal of Materials Chemistry 20, 5983 (2010).
142. Cao, X. et al. Preparation of novel 3D graphene networks for supercapacitor
applications. Small 7, 3163–3168 (2011).
143. Zhang, X. et al. Mechanically strong and highly conductive graphene aerogel and its
use as electrodes for electrochemical power sources. J. Mater. Chem. 21, 6494 (2011).
144. Ye, S., Feng, J. & Wu, P. Deposition of three-dimensional graphene aerogel on nickel
foam as a binder-free supercapacitor electrode. ACS Appl. Mater. Interfaces 5, 7122–
7129 (2013).
145. Liu, C., Yu, Z., Neff, D., Zhamu, A. & Jang, B. Z. Graphene-based supercapacitor
with an ultrahigh energy density. Nano Lett. 10, 4863–4868 (2010).
146. Stoller, M. D., Park, S., Zhu, Y., An, J. & Ruoff, R. S. Graphene-based ultracapacitors.
Nano Lett. 8, 3498–502 (2008).
147. Xu, G. et al. Binder-free activated carbon/carbon nanotube paper electrodes for use in
supercapacitors. Nano Res. 4, 870–881 (2011).
148. Noked, M., Okashy, S., Zimrin, T. & Aurbach, D. Composite carbon nanotube/carbon
electrodes for electrical double-layer super capacitors. Angew. Chemie - Int. Ed. 51,
1568–1571 (2012).
149. Taberna, P. L., Chevallier, G., Simon, P., D. & Aubert, T. Activated carbon-carbon
nanotube composite porous film for supercapacitor applications. Mater. Res. Bull. 41,
478–484 (2006).
150. Cheng, Q. et al. Graphene and carbon nanotube composite electrodes for
supercapacitors with ultra-high energy density. Phys. Chem. Chem. Phys. 13, 17615–
24 (2011).
151. Jung, N. et al. Synthesis of chemically bonded graphene/carbon nanotube composites
and their application in large volumetric capacitance supercapacitors. Adv. Mater. 25,
6854–6858 (2013).
152. Conway, B. E., Birss, V. & Wojtowicz, J. The role and utilization of
pseudocapacitance for energy storage by supercapacitors. J. Power Sources 66, 1–14
(1997).
153. Conway, B. E. & Pell, W. G. Double-layer and pseudocapacitance types of
electrochemical capacitors and their applications to the development of hybrid devices.
J. Solid State Electrochem. 7, 637–644 (2003).
166
154. Augustyn, V., Simon, P. & Dunn, B. Pseudocapacitive oxide materials for high-rate
electrochemical energy storage. Energy Environ. Sci. 7, 1597 (2014).
155. Trasatti, S. & Buzzanca, G. Ruthenium dioxide: A new interesting electrode material.
Solid state structure and electrochemical behaviour. J. Electroanal. Chem. Interfacial
Electrochem. 29, A1–A5 (1971).
156. Lee, C. Y. & Bond, A. M. Revelation of multiple underlying RuO2 redox processes
associated with pseudocapacitance and electrocatalysis. Langmuir 26, 16155–16162
(2010).
157. Lee, H. Y., Kim, S. W. & Lee, H. Y. Expansion of Active Site Area and Improvement
of Kinetic Reversibility in Electrochemical Pseudocapacitor Electrode. Electrochem.
Solid-State Lett. 4, A19 (2001).
158. Toupin, M., Brousse, T. & Bélanger, D. Charge storage mechanism of MnO2 electrode
used in aqueous electrochemical capacitor. Chem. Mater. 16, 3184–3190 (2004).
159. Kuo, S.-L. & Wu, N.-L. Investigation of Pseudocapacitive Charge-Storage Reaction of
MnO[sub 2]⋅nH[sub 2]O Supercapacitors in Aqueous Electrolytes. J. Electrochem.
Soc. 153, A1317 (2006).
160. Lu, X. et al. Hydrogenated TiO2 nanotube arrays for supercapacitors. Nano Lett 12,
1690–1696 (2012).
161. Wang, L. et al. Ultrasonic-assisted synthesis of amorphous Fe3O4 with a high specific
surface area and improved capacitance for supercapacitor. Powder Technol. 256, 499–
505 (2014).
162. Augustyn, V. et al. High-rate electrochemical energy storage through Li+ intercalation
pseudocapacitance. Nat. Mater. 12, 518–22 (2013).
163. Shinde, N. M. et al. Wet chemical synthesis of WO3 thin films for supercapacitor
application. Korean J. Chem. Eng. 32, 974–979 (2015).
164. Xu, Y. et al. Novel application of LiCoO<inf>2</inf> as a high-performance
candidate material for supercapacitor. J. Energy Chem. 24, 193–198 (2015).
165. Zhang, Y. Q., Xia, X. H., Kang, J. & Tu, J. P. Hydrothermal synthesized porous
Co(OH)2 nanoflake film for supercapacitor application. Chinese Sci. Bull. 57, 4215–
4219 (2012).
166. Xiong, X. et al. Three-dimensional ultrathin Ni(OH)2 nanosheets grown on nickel
foam for high-performance supercapacitors. Nano Energy 11, 154–161 (2015).
167. Acerce, M., Voiry, D. & Chhowalla, M. Metallic 1T phase MoS2 nanosheets as
supercapacitor electrode materials. Nat. Nanotechnol. 10, 313–8 (2015).
168. Zhang, L., Wu, H. Bin & Lou, X. W. (David). Unusual CoS2 ellipsoids with
anisotropic tube-like cavities and their application in supercapacitors. Chem. Commun.
48, 6912 (2012).
169. Zhou, X., Chen, H., Shu, D., He, C. & Nan, J. Study on the electrochemical behavior
of vanadium nitride as a promising supercapacitor material. J. Phys. Chem. Solids 70,
167
495–500 (2009).
170. Lukatskaya, M. R. et al. Cation Intercalation and High Volumetric Capacitance of
Two-Dimensional Titanium Carbide. Science (80-. ). 341, 1502–1505 (2013).
171. Hsieh, C. T. & Teng, H. Influence of oxygen treatment on electric double-layer
capacitance of activated carbon fabrics. Carbon N. Y. 40, 667–674 (2002).
172. Hulicova-Jurcakova, D., Seredych, M., Lu, G. Q. & Bandosz, T. J. Combined effect of
nitrogen- and oxygen-containing functional groups of microporous activated carbon on
its electrochemical performance in supercapacitors. Adv. Funct. Mater. 19, 438–447
(2009).
173. Jurewicz, K., Babe??, K., Pietrzak, R., Delpeux, S. & Wachowska, H. Capacitance
properties of multi-walled carbon nanotubes modified by activation and
ammoxidation. Carbon N. Y. 44, 2368–2375 (2006).
174. Snook, G. A., Kao, P. & Best, A. S. Conducting-polymer-based supercapacitor devices
and electrodes. J. Power Sources 196, 1–12 (2011).
175. Khdary, N. H., Abdesalam, M. E. & EL Enany, G. Mesoporous Polyaniline Films for
High Performance Supercapacitors. J. Electrochem. Soc. 161, G63–G68 (2014).
176. Dubal, D. P., Lee, S. H., Kim, J. G., Kim, W. B. & Lokhande, C. D. Porous
polypyrrole clusters prepared by electropolymerization for a high performance
supercapacitor. J. Mater. Chem. 22, 3044 (2012).
177. Wang, Z. et al. Solution-processed poly(3,4-ethylenedioxythiophene) nanocomposite
paper electrodes for high-capacitance flexible supercapacitors. J. Mater. Chem. A 4,
1714–1722 (2016).
178. Devaraj, S., Munichandraiah, N., Devaraj, S. & Munichandraiah, N. Effect of
Crystallographic Structure of MnO on Its Electrochemical Capacitance Properties
Effect of Crystallographic Structure of MnO 2 on Its Electrochemical Capacitance
Properties. 4406–4417 (2008). doi:10.1021/jp7108785
179. Higgins, T. M. et al. Effect of percolation on the capacitance of supercapacitor
electrodes prepared from composites of manganese dioxide nanoplatelets and carbon
nanotubes. ACS Nano 8, 9567–9579 (2014).
180. Sharma, A. K., Sharma, Y., Malhotra, R. & Sharma, J. K. Solvent tuned PANI-CNT
composites as advanced electrode materials for supercapacitor application. Adv. Mater.
Lett. 3, 82–86 (2012).
181. Sun, Z. et al. Microstructural and electrochemical characterization of RuO 2/CNT
composites synthesized in supercritical diethyl amine. Carbon N. Y. 44, 888–893
(2006).
182. Mendoza-Sánchez, B., Coelho, J., Pokle, A. & Nicolosi, V. A 2D graphene-manganese
oxide nanosheet hybrid synthesized by a single step liquid-phase co-exfoliation
method for supercapacitor applications. Electrochim. Acta 174, 696–705 (2015).
183. Yan, J. et al. Preparation of a graphene nanosheet/polyaniline composite with high
specific capacitance. Carbon N. Y. 48, 487–493 (2010).
168
184. Qian, Y., Lu, S. & Gao, F. Preparation of MnO2/graphene composite as electrode
material for supercapacitors. J. Mater. Sci. 46, 3517–3522 (2011).
185. Liu, R., Duay, J., Lane, T. & Bok Lee, S. Synthesis and characterization of
RuO2/poly(3{,}4-ethylenedioxythiophene) composite nanotubes for supercapacitors.
Phys. Chem. Chem. Phys. 12, 4309–4316 (2010).
186. Deshmukh, P. R., Bulakhe, R. N., Pusawale, S. N., Sartale, S. D. & Lokhande, C. D.
Polyaniline–RuO 2 composite for high performance supercapacitors: chemical
synthesis and properties. RSC Adv. 5, 28687–28695 (2015).
187. Liu, T., Shao, G., Ji, M. & Wang, G. Polyaniline/MnO2 composite with high
performance as supercapacitor electrode via pulse electrodeposition. Polym. Compos.
36, 113–120 (2015).
188. Nam, I., Park, S., Kim, G.-P., Park, J. & Yi, J. Transparent and ultra-bendable all-
solid-state supercapacitors without percolation problems. Chem. Sci. 4, 1663–1667
(2013).
189. Lu, X., Yu, M., Wang, G., Tong, Y. & Li, Y. Flexible solid-state supercapacitors:
design, fabrication and applications. Energy Environ. Sci. 7, 2160 (2014).
190. Stoller, M. D. & Ruoff, R. S. Best practice methods for determining an electrode
material’s performance for ultracapacitors. Energy Environ. Sci. 3, 1294 (2010).
191. Burke, A. Ultracapacitors: Why, how, and where is the technology. J. Power Sources
91, 37–50 (2000).
192. Portet, C., Taberna, P. L., Simon, P. & Flahaut, E. Modification of Al Current
Collector/Active Material Interface for Power Improvement of Electrochemical
Capacitor Electrodes. J. Electrochem. Soc. 153, A649 (2006).
193. Portet, C., Taberna, P. L., Simon, P., Flahaut, E. & Laberty-Robert, C. High power
density electrodes for Carbon supercapacitor applications. Electrochim. Acta 50, 4174–
4181 (2005).
194. Portet, C., Taberna, P. L., Simon, P. & Laberty-Robert, C. Modification of Al current
collector surface by sol-gel deposit for carbon-carbon supercapacitor applications.
Electrochim. Acta 49, 905–912 (2004).
195. King, P. J., Higgins, T. M., De, S., Nicolosi, N. & Coleman, J. N. Percolation effects in
supercapacitors with thin, transparent carbon nanotube electrodes. ACS Nano 6, 1732–
1741 (2012).
196. Higgins, T. M. & Coleman, J. N. Avoiding Resistance Limitations in High-
Performance Transparent Supercapacitor Electrodes Based on Large-Area, High-
Conductivity PEDOT:PSS Films. ACS Appl. Mater. Interfaces 7, 16495–16506
(2015).
197. Chen, T., Peng, H., Durstock, M. & Dai, L. High-performance transparent and
stretchable all-solid supercapacitors based on highly aligned carbon nanotube sheets.
Sci. Rep. 4, 3612 (2014).
198. Quintero, R. et al. Carbon nanotube 3D current collectors for lightweight, high
169
performance and low cost supercapacitor electrodes. RSC Adv. 4, 8230–8237 (2014).
199. Lu, M., Beguin, F. & Frackowiak, E. Supercapacitors: Materials, Systems and
Applications. (Wiley, 2013).
200. Lu, X. et al. A flexible graphene/multiwalled carbon nanotube film as a high
performance electrode material for supercapacitors. Electrochim. Acta 56, 5115–5121
(2011).
201. Yuan, L. et al. Flexible solid-state supercapacitors based on carbon nanoparticles/MnO
2 nanorods hybrid structure. ACS Nano 6, 656–661 (2012).
202. He, Y. et al. Freestanding three-dimensional graphene/Mno2 composite networks as
ultralight and flexible supercapacitor electrodes. ACS Nano 7, 174–182 (2013).
203. Wu, Q., Xu, Y., Yao, Z., Liu, A. & Shi, G. Supercapacitors Based on Flexible
Graphene/Polyaniline Nanofiber Composite Films. ACS Nano 4, 1963–1970 (2010).
204. Kang, Y. J., Chung, H., Han, C.-H. & Kim, W. All-solid-state flexible supercapacitors
based on papers coated with carbon nanotubes and ionic-liquid-based gel electrolytes.
Nanotechnology 23, 65401 (2012).
205. Wang, G. et al. Solid-state supercapacitor based on activated carbon cloths exhibits
excellent rate capability. Adv. Mater. 26, 2676–2682 (2014).
206. Le, V. T. et al. Coaxial Fiber Supercapacitor Using All-Carbon Material Electrodes.
5940–5947 (2013).
207. Hu, S., Rajamani, R. & Yu, X. Flexible solid-state paper based carbon nanotube
supercapacitor. Appl. Phys. Lett. 100, (2012).
208. Peng, L. et al. Ultrathin two-dimensional MnO2/graphene hybrid nanostructures for
high-performance, flexible planar supercapacitors. Nano Lett. 13, 2151–2157 (2013).
209. Meng, C., Liu, C., Chen, L., Hu, C. & Fan, S. Highly flexible and all-solid-state
paperlike polymer supercapacitors. Nano Lett. 10, 4025–4031 (2010).
210. Jung, H. Y., Karimi, M. B., Hahm, M. G., Ajayan, P. M. & Jung, Y. J. Transparent,
flexible supercapacitors from nano-engineered carbon films. Sci. Rep. 2, 1–5 (2012).
211. Singh, A. & Chandra, A. Significant Performance Enhancement in Asymmetric
Supercapacitors based on Metal Oxides, Carbon nanotubes and Neutral Aqueous
Electrolyte. Nat. Publ. Gr. 1–12 (2015). doi:10.1038/srep15551
212. Gao, H., Xiao, F., Ching, C. B. & Duan, H. High-performance asymmetric
supercapacitor based on graphene hydrogel and nanostructured MnO 2. ACS Appl.
Mater. Interfaces 4, 2801–2810 (2012).
213. Shen, J., Yang, C., Li, X. & Wang, G. High-performance asymmetric supercapacitor
based on nanoarchitectured polyaniline/graphene/carbon nanotube and activated
graphene electrodes. ACS Appl. Mater. Interfaces 5, 8467–8476 (2013).
214. Finn, D. J. et al. Inkjet deposition of liquid-exfoliated graphene and MoS2 nanosheets
for printed device applications. J. Mater. Chem. C 2, 925–932 (2014).
170
215. Scardaci, V., Coull, R., Lyons, P. E., Rickard, D. & Coleman, J. N. Spray deposition of
highly transparent, low-resistance networks of silver nanowires over large areas. Small
7, 2621–2628 (2011).
216. De, S. et al. Silver Nanowire Networks as Flexible ,. ACSNano 3, 1767–1774 (2009).
217. Coleman, J. N. Liquid-Phase Exfoliation of Nanotubes and Graphene. Adv. Funct.
Mater. 19, 3680–3695 (2009).
218. Wang, Q. H., Kalantar-Zadeh, K., Kis, A., Coleman, J. N. & Strano, M. S. Electronics
and optoelectronics of two-dimensional transition metal dichalcogenides. Nat.
Nanotechnol. 7, 699–712 (2012).
219. Nicolosi, V., Chhowalla, M., Kanatzidis, M. G., Strano, M. S. & Coleman, J. N.
Liquid Exfoliation of Layered Materials. Science (80-. ). 340, 1226419 (2013).
220. Harvey, A. et al. Preparation of gallium sulfide nanosheets by liquid exfoliation and
their application as hydrogen evolution catalysts. Chem. Mater. 27, 3483–3493 (2015).
221. Khan, U., O’Neill, A., Lotya, M., De, S. & Coleman, J. N. High-concentration solvent
exfoliation of graphene. Small 6, 864–871 (2010).
222. O’Neill, A., Khan, U., Nirmalraj, P. N., Boland, J. & Coleman, J. N. Graphene
dispersion and exfoliation in low boiling point solvents. J. Phys. Chem. C 115, 5422–
5428 (2011).
223. Hennrich, F. et al. The mechanism of cavitation-induced scission of single-walled
carbon nanotubes. J. Phys. Chem. B 111, 1932–1937 (2007).
224. Varrla, E. et al. Large-scale production of size-controlled MoS2 nanosheets by shear
exfoliation. Chem. Mater. 27, 1129–1139 (2015).
225. Paton, K. R. et al. Scalable production of large quantities of defect-free few-layer
graphene by shear exfoliation in liquids. Nat. Mater. 13, 624–630 (2014).
226. Bergin, S. D. et al. Multicomponent solubility parameters for single-walled carbon
nanotube-solvent mixtures. ACS Nano 3, 2340–2350 (2009).
227. Cunningham, G., Hanlon, D., McEvoy, N., Duesberg, G. S. & Coleman, J. N. Large
variations in both dark- and photoconductivity in nanosheet networks as nanomaterial
is varied from MoS2 to WTe2. Nanoscale 7, 198–208 (2015).
228. Smith, R. J., Lotya, M. & Coleman, J. N. The importance of repulsive potential
barriers for the dispersion of graphene using surfactants. New J. Phys. 12, (2010).
229. May, P., Khan, U., Hughes, J. M. & Coleman, J. N. Role of solubility parameters in
understanding the steric stabilization of exfoliated two-dimensional nanosheets by
adsorbed polymers. J. Phys. Chem. C 116, 11393–11400 (2012).
230. May, P., Khan, U., O’Neill, A. & Coleman, J. N. Approaching the theoretical limit for
reinforcing polymers with graphene. J. Mater. Chem. 22, 1278 (2012).
231. Khan, U. et al. Size selection of dispersed, exfoliated graphene flakes by controlled
centrifugation. Carbon N. Y. 50, 470–475 (2012).
171
232. Backes, C. et al. Edge and confinement effects allow in situ measurement of size and
thickness of liquid-exfoliated nanosheets. Nat. Commun. 5, 4576 (2014).
233. Backes, C. et al. Production of Highly Monolayer Enriched Dispersions of Liquid-
Exfoliated Nanosheets by Liquid Cascade Centrifugation. ACS Nano acsnano.5b07228
(2016). doi:10.1021/acsnano.5b07228
234. Backes, C. et al. Spectroscopic metrics allow in situ measurement of mean size and
thickness of liquid-exfoliated few-layer graphene nanosheets. Nanoscale 8, 4311–4323
(2016).
235. Hsu, H.-W. & Liu, C.-L. Spray-coating semiconducting conjugated polymers for
organic thin film transistor applications. RSC Adv. 4, 30145 (2014).
236. Vernon-Parry, K. D. Scanning electron microscopy: an introduction. III-Vs Rev. 13,
40–44 (2000).
237. Vernon-Parry, K. D. & Wright, A. C. TEM: An introduction. III-Vs Rev. 13, 36–40
(2000).
238. Bisquert, J. Theory of the impedance of electron diffusion and recombination in a thin
layer. J. Phys. Chem. B 106, 325–333 (2002).
239. Cooper, K. R. & Smith, M. Electrical test methods for on-line fuel cell ohmic
resistance measurement. J. Power Sources 160, 1088–1095 (2006).
240. Slonczewski, J. C. & Weiss, P. R. Band structure of graphite. Phys. Rev. 109, 272–279
(1958).
241. Novoselov, K. S. & Geim, A. K. The Rise of Graphene. Nat. Mater. 6, 183–191
(2007).
242. Novoselov, K. S. et al. Electric field effect in atomically thin carbon films. Science
(80-. ). 306, 666–669 (2004).
243. Lee, C., Wei, X., Kysar, J. W. & Hone, J. Measurement of Elastic Properties and
Intrinsic Strength of Monolayer Graphene. Science (80-. ). 321, 385–388 (2008).
244. Nair, R. R. et al. Fine Structure Constant Defines Visual Transperency of Graphene.
Science (80-. ). 320, 2008 (2008).
245. Riedl, C., Coletti, C., Iwasaki, T., Zakharov, A. A. & Starke, U. Quasi-free-standing
epitaxial graphene on SiC obtained by hydrogen intercalation. Phys. Rev. Lett. 103, 1–
4 (2009).
246. Emtsev, K. V., Speck, F., Seyller, T., Ley, L. & Riley, J. D. Interaction, growth, and
ordering of epitaxial graphene on SiC{0001} surfaces: A comparative photoelectron
spectroscopy study. Phys. Rev. B - Condens. Matter Mater. Phys. 77, 1–10 (2008).
247. de Heer, W. a. et al. Large area and structured epitaxial graphene produced by
confinement controlled sublimation of silicon carbide. Pnas 108, 16900–16905 (2011).
248. Li, X. et al. Large-area synthesis of high-quality and uniform graphene films on copper
foils. Science (80-. ). 324, 1312–1314 (2009).
172
249. Suk, J. W. et al. Transfer of CVD-grown monolayer graphene onto arbitrary
substrates. ACS Nano 5, 6916–6924 (2011).
250. Huang, P. Y. et al. Grains and grain boundaries in single-layer graphene atomic
patchwork quilts. Nature 469, 389–92 (2011).
251. Bae, S. et al. Roll-to-roll production of 30-inch graphene films for transparent
electrodes. Nat. Nanotechnol. 5, 574–578 (2010).
252. Chen, J., Yao, B., Li, C. & Shi, G. An improved Hummers method for eco-friendly
synthesis of graphene oxide. Carbon N. Y. 64, 225–229 (2013).
253. Gao, X., Jang, J. & Nagase, S. Hydrazine and Thermal Reduction of Graphene Oxide :
Reaction Mechanisms and Design. J. Phys. Chem. C 114, 832–842 (2010).
254. Zhu, Y. et al. Microwave assisted exfoliation and reduction of graphite oxide for
ultracapacitors. Carbon N. Y. 48, 2118–2122 (2010).
255. Viculis, L. M., Mack, J. J., Mayer, O. M., Hahn, H. T. & Kaner, R. B. Intercalation
and exfoliation routes to graphite nanoplatelets. J. Mater. Chem. 15, 974 (2005).
256. Khan, U., May, P., O’Neill, A. & Coleman, J. N. Development of stiff, strong, yet
tough composites by the addition of solvent exfoliated graphene to polyurethane.
Carbon N. Y. 48, 4035–4041 (2010).
257. Stankovich, S. et al. Graphene-based composite materials. Nature 442, 282–286
(2006).
258. Wang, G., Shen, X., Yao, J. & Park, J. Graphene nanosheets for enhanced lithium
storage in lithium ion batteries. Carbon N. Y. 47, 2049–2053 (2009).
259. Radovic, L. R. & Bockrath, B. On the chemical nature of graphene edges: Origin of
stability and potential for magnetism in carbon materials. J. Am. Chem. Soc. 127,
5917–5927 (2005).
260. Yuan, W. & Shi, G. Graphene-based gas sensors. J. Mater. Chem. A 1, 10078 (2013).
261. Roy-Mayhew, J. D. & Aksay, I. A. Graphene materials and their use in dye-sensitized
solar cells. Chem. Rev. 114, 6323–6348 (2014).
262. Wang, S., Zhang, Y., Abidi, N. & Cabrales, L. Wettability and surface free energy of
graphene films. Langmuir 25, 11078–11081 (2009).
263. Wang, H., Maiyalagan, T. & Wang, X. Review on recent progress in nitrogen-doped
graphene: Synthesis, characterization, and its potential applications. ACS Catal. 2,
781–794 (2012).
264. Wen, Z. et al. Crumpled nitrogen-doped graphene nanosheets with ultrahigh pore
volume for high-performance supercapacitor. Adv. Mater. 24, 5610–5616 (2012).
265. Kemp, K. C. et al. Functionalization of Graphene : Covalent and Non- Covalent
Approaches , Derivatives and Applications Functionalization of Graphene : Covalent
and Non-Covalent Approaches , Derivatives and Applications. (2012).
doi:10.1021/cr3000412
173
266. Iijima, S. Helical microtubules of graphitic carbon. Nature 354, 56–58 (1991).
267. Iijima, S. & Ichihashi, T. Single-shell carbon nanotubes of 1-nm diameter. Nature 363,
603–605 (1993).
268. Dresselhaus, M. S., Dresselhaus, G., Saito, R. & Jorio, A. Raman spectroscopy of
carbon nanotubes. Phys. Rep. 409, 47 (2005).
269. Moisala, a, Nasibulin, a G. & Kauppinen, E. I. The role of metal nanoparticles in the
catalytic production of single-walled carbon nanotubes - a review. J. Phys. - Condens.
Matter 15, S3011 (2003).
270. Hou, P. X., Liu, C. & Cheng, H. M. Purification of carbon nanotubes. Carbon N. Y. 46,
2003–2025 (2008).
271. Furtado, C. A. et al. Debundling and Dissolution of Single-Walled Carbon Nanotubes
in Amide Solvents. J. Am. Chem. Soc. 126, 6095–6105 (2004).
272. Bergin, S. D. et al. Towards solutions of single-walled carbon nanotubes in common
solvents. Adv. Mater. 20, 1876–1881 (2008).
273. Sun, Z. et al. Quantitative evaluation of surfactant-stabilized single-walled carbon
nanotubes: Dispersion quality and its correlation with zeta potential. J. Phys. Chem. C
112, 10692–10699 (2008).
274. Xie, X. L., Mai, Y. W. & Zhou, X. P. Dispersion and alignment of carbon nanotubes in
polymer matrix: A review. Mater. Sci. Eng. R Reports 49, 89–112 (2005).
275. Cathcart, H. et al. Spontaneous debundling of single-walled carbon nanotubes in
DNA-based dispersions. J. Phys. Chem. C 111, 66–74 (2007).
276. Coleman, J. N., Khan, U. & Gun’ko, Y. K. Mechanical reinforcement of polymers
using carbon nanotubes. Adv. Mater. 18, 689–706 (2006).
277. Aguilar, J. O., Bautista-Quijano, J. R. & Avilés, F. Influence of carbon nanotube
clustering on the electrical conductivity of polymer composite films. Express Polym.
Lett. 4, 292–299 (2010).
278. Scardaci, V., Coull, R. & Coleman, J. N. Very thin transparent, conductive carbon
nanotube films on flexible substrates. Appl. Phys. Lett. 97, (2010).
279. Mirri, F. et al. High-performance carbon nanotube transparent conductive films by
scalable dip coating. ACS Nano 6, 9737–9744 (2012).
280. King, P. J., Khan, U., Lotya, M., De, S. & Coleman, J. N. Improvement of transparent
conducting nanotube films by addition of small quantities of graphene. ACS Nano 4,
4238–4246 (2010).
281. Doherty, E. M. et al. The spatial uniformity and electromechanical stability of
transparent, conductive films of single walled nanotubes. Carbon N. Y. 47, 2466–2473
(2009).
282. Peigney, a., Laurent, C., Flahaut, E., Bacsa, R. R. & Rousset, a. Specific surface area
of carbon nanotubes and bundles of carbon nanotubes. Carbon N. Y. 39, 507–514
174
(2001).
283. Datsyuk, V. et al. Chemical oxidation of multiwalled carbon nanotubes. Carbon N. Y.
46, 833–840 (2008).
284. Hirsch, A. Functionalization of single-walled carbon nanotubes. Angew. Chemie - Int.
Ed. 41, 1853–1859 (2002).
285. Chhowalla, M. et al. The chemistry of two-dimensional layered transition metal
dichalcogenide nanosheets. Nat. Chem. 5, 263–275 (2013).
286. Mak, K. F., Lee, C., Hone, J., Shan, J. & Heinz, T. F. Atomically thin MoS2: A new
direct-gap semiconductor. Phys. Rev. Lett. 105, 2–5 (2010).
287. Woollam, J. A. Physics and Chemistry of MoS2 Intercalation Compounds. Mater. Sci.
Eng. 31, 289–295 (1977).
288. Py, M. A. & Haering, R. R. Structural destabilization induced by lithium intercalation
in MoS2 and related compounds. Can. J. Phys. 61, 76–84 (1983).
289. Bissessur, R., Kanatzidis, M. G., Schindler, J. L. & Kannewurf, C. R. Encapsulation of
polymers into MoS2 and metal to insulator transition in metastable MoS2 . J. Chem.
Soc. Chem. Commun. 1582 (1993). doi:10.1039/c39930001582
290. Coleman, J. N. et al. Two-Dimensional Nanosheets Produced by Liquid Exfoliation of
Layered Materials. Science (80-. ). 331, 568–571 (2011).
291. O’Neill, A., Khan, U. & Coleman, J. N. Preparation of high concentration dispersions
of exfoliated MoS 2 with increased flake size. Chem. Mater. 24, 2414–2421 (2012).
292. Hernandez, Y. et al. High-yield production of graphene by liquid-phase exfoliation of
graphite. Nat. Nanotechnol. 3, 563–568 (2008).
293. Stephenson, T., Li, Z., Olsen, B. & Mitlin, D. Lithium ion battery applications of
molybdenum disulfide (MoS 2 ) nanocomposites. Energy Environ. Sci. 7, 209–231
(2014).
294. Lukowski, M. A. et al. Enhanced hydrogen evolution catalysis from chemically
exfoliated metallic MoS2 nanosheets. J. Am. Chem. Soc. 135, 10274–10277 (2013).
295. Donarelli, M. et al. Response to NO2 and other gases of resistive chemically exfoliated
MoS2-based gas sensors. Sensors Actuators B Chem. 207, 602–613 (2015).
296. Zhang, W. et al. High-gain phototransistors based on a CVD MoS2 monolayer. Adv.
Mater. 25, 3456–3461 (2013).
297. Cunningham, G. et al. Photoconductivity of solution-processed MoS2 films. J. Mater.
Chem. C 1, 6899 (2013).
298. Berntsen, N. et al. A Solvothermal Route to High-Surface-Area Nanostructured MoS2.
Chem. Mater. 15, 4498–4502 (2003).
299. Feng, X. et al. Novel mixed-solvothermal synthesis of MoS2 nanosheets with
controllable morphologies. Cryst. Res. Technol. 48, 363–368 (2013).
175
300. Liu, Q. et al. One-step solvothermal synthesis of MoS2/TiO2 nanocomposites with
enhanced photocatalytic H2 production. J. Nanoparticle Res. 15, (2013).
301. Lee, K., Gatensby, R., McEvoy, N., Hallam, T. & Duesberg, G. S. High-performance
sensors based on molybdenum disulfide thin films. Adv. Mater. 25, 6699–6702 (2013).
302. Liu, K. K. et al. Growth of large-area and highly crystalline MoS 2 thin layers on
insulating substrates. Nano Lett. 12, 1538–1544 (2012).
303. Wang, X., Feng, H., Wu, Y. & Jiao, L. Controlled Synthesis of Highly Crystalline
MoS2 Flakes by Chemical Vapor Deposition. J Am Chem Soc 135, 5304–5307 (2013).
304. Amani, M. et al. Electrical performance of monolayer MoS2 field-effect transistors
prepared by chemical vapor deposition. Appl. Phys. Lett. 102, (2013).
305. Huang, C. C. et al. Scalable high-mobility MoS2 thin films fabricated by an
atmospheric pressure chemical vapor deposition process at ambient temperature.
Nanoscale 6, 12792–12797 (2014).
306. Choi, M. S. et al. Lateral MoS2 p-n junction formed by chemical doping for use in
high-performance optoelectronics. ACS Nano 8, 9332–9340 (2014).
307. Shirakawa, H., Louis, E. J., MacDiarmid, A. G., Chiang, C. K. & Heeger, A. J.
Synthesis of electrically conducting organic polymers: halogen derivatives of
polyacetylene, (CH). J. Chem. Soc. Chem. Commun. 578–580 (1977).
doi:10.1039/C39770000578
308. Ocampo, C., Oliver, R., Armelin, E., Alemán, C. & Estrany, F. Electrochemical
synthesis of poly(3,4-ethylenedioxythiophene) on steel electrodes: Properties and
characterization. J. Polym. Res. 13, 193–200 (2006).
309. Corradi, R. & Armes, S. P. Chemical synthesis of poly(3,4-ethylenedioxythiophene).
Synth. Met. 84, 453–454 (1997).
310. Lefebvre, M., Qi, Z., Rana, D. & Pickup, P. G. Chemical Synthesis, Characterization,
and Electrochemical Studies of Poly(3,4-ethylenedioxythiophene)/Poly(styrene-4-
sulfonate) Composites. Chem. Mater. 11, 262–268 (1999).
311. Pei, Q., Zuccarello, G., Ahlskog, M. & Ingan??s, O. Electrochromic and highly stable
poly(3,4-ethylenedioxythiophene) switches between opaque blue-black and transparent
sky blue. Polymer (Guildf). 35, 1347–1351 (1994).
312. Crispin, X. et al. The origin of the high conductivity of poly(3,4-
ethylenedioxythiophene)-poly(styrenesulfonate) (PEDOT- PSS) plastic electrodes.
Chem. Mater. 18, 4354–4360 (2006).
313. Ouyang, J. ‘secondary doping’ methods to significantly enhance the conductivity of
PEDOT:PSS for its application as transparent electrode of optoelectronic devices.
Displays 34, 423–436 (2013).
314. Anothumakkool, B., Soni, R., Bhange, S. N. & Kurungot, S. Novel scalable synthesis
of highly conducting and robust PEDOT paper for a high performance flexible solid
supercapacitor. Energy Environ. Sci. 8, 1339–1347 (2015).
176
315. Zuber, K., Fabretto, M., Hall, C. & Murphy, P. Improved PEDOT conductivity via
suppression of crystallite formation in Fe(III) tosylate during vapor phase
polymerization. Macromol. Rapid Commun. 29, 1503–1508 (2008).
316. Louwet, F. et al. PEDOT/PSS: Synthesis, characterization, properties and applications.
Synth. Met. 135–136, 115–117 (2003).
317. Shi, H., Liu, C., Jiang, Q. & Xu, J. Effective Approaches to Improve the Electrical
Conductivity of PEDOT:PSS: A Review. Adv. Electron. Mat. 1, n/a-n/a (2015).
318. Huang, J. et al. Investigation of the effects of doping and post-deposition treatments on
the conductivity, morphology, and work function of poly(3,4-
ethylenedioxythiophene)/poly(styrene sulfonate) films. Adv. Funct. Mater. 15, 290–
296 (2005).
319. Kim, J. Y., Jung, J. H., Lee, D. E. & Joo, J. Enhancement of electrical conductivity of
poly(3,4-ethylenedioxythiophene)/poly(4-styrenesulfonate) by a change of solvents.
Synth. Met. 126, 311–316 (2002).
320. M, F. Electrochemical and XPS Studies toward the Role of Monomeric and Polymeric
Sulfonate Counterions in the Synthesis, Composition, and Properties of Poly(3,4-
ethylenedioxythiophene). Macromolecules 36, 3337 (2003).
321. Xia, Y. & Ouyang, J. Anion effect on salt-induced conductivity enhancement of
poly(3,4-ethylenedioxythiophene):Poly(styrenesulfonate) films. Org. Electron.
physics, Mater. Appl. 11, 1129–1135 (2010).
322. Alemu, D., Wei, H.-Y., Ho, K.-C. & Chu, C.-W. Highly conductive PEDOT:PSS
electrode by simple film treatment with methanol for ITO-free polymer solar cells.
Energy Environ. Sci. 5, 9662 (2012).
323. Xia, Y. & Ouyang, J. Significant conductivity enhancement of conductive poly(3,4-
ethylenedioxythiophene): Poly(styrenesulfonate) films through a treatment with
organic carboxylic acids and inorganic acids. ACS Appl. Mater. Interfaces 2, 474–483
(2010).
324. Xia, Y., Sun, K. & Ouyang, J. Solution-processed metallic conducting polymer films
as transparent electrode of optoelectronic devices. Adv. Mater. 24, 2436–2440 (2012).
325. McCarthy, J., Hanley, C., Brennan, L., Lambertini, V. & Gun’ko, Y. Fabrication of
highly transparent and conducting PEDOT:PSS films using a formic acid treatment. J.
Mater. Chem. C 2, 764–770 (2014).
326. Wang, Z. et al. Solution-processed poly(3,4-ethylenedioxythiophene) nanocomposite
paper electrodes for high-capacitance flexible supercapacitors. J. Mater. Chem. A 4,
1714–1722 (2016).
327. Hagfeldt, A., Boschloo, G., Sun, L., Kloo, L. & Pettersson, H. Dye-sensitized solar
cells. Chem. Rev. 110, 6595–6663 (2010).
328. Thomas, S. et al. A review on counter electrode materials in dye-sensitized solar cells.
J. Mater. Chem. A Mater. Energy Sustain. 2, 4474–4490 (2014).
329. Zhao, B. et al. Flexible counter electrodes based on mesoporous carbon aerogel for
177
high-performance dye-sensitized solar cells. J. Phys. Chem. C 115, 22615–22621
(2011).
330. Choi, H., Kim, H., Hwang, S., Han, Y. & Jeon, M. Graphene counter electrodes for
dye-sensitized solar cells prepared by electrophoretic deposition. J. Mater. Chem. 21,
7548 (2011).
331. Hung, K. H. & Wang, H. W. A freeze-dried graphene counter electrode enhances the
performance of dye-sensitized solar cells. Thin Solid Films 550, 515–520 (2014).
332. Punckt, C., Roy-mayhew, J. D., Bozym, D. J., Punckt, C. & Aksay, I. A.
Functionalized Graphene as a Catalytic Counter Electrode in Dye-Sensitized Solar
Cells. ACS Nano 4, 6203–6211 (2010).
333. Wang, H. & Hu, Y. H. Graphene as a counter electrode material for dye-sensitized
solar cells. Energy Environ. Sci. 5, 8182 (2012).
334. Ferrari, A. C. et al. Science and technology roadmap for graphene, related two-
dimensional crystals, and hybrid systems. Nanoscale 7, 4598–4810 (2014).
335. Chen, D., Li, F. & Ray, A. K. External and internal mass transfer effect on
photocatalytic degradation. Catal. Today 66, 475–485 (2001).
336. Velten, J. et al. Carbon nanotube/graphene nanocomposite as efficient counter
electrodes in dye-sensitized solar cells. Nanotechnology 23, 85201 (2012).
337. Ellis, H. et al. PEDOT counter electrodes for dye-sensitized solar cells prepared by
aqueous micellar electrodeposition. Electrochim. Acta 107, 45–51 (2013).
338. Guo, J., Shi, Y., Chu, Y. & Ma, T. Highly efficient telluride electrocatalysts for use as
Pt-free counter electrodes in dye-sensitized solar cells. Chem. Commun. (Camb). 49,
10157–9 (2013).
339. Chen, H. et al. In situ growth of a MoSe2/Mo counter electrode for high efficiency
dye-sensitized solar cells. Chem. Commun. 50, 4475–4477 (2014).
340. Al-Mamun, M., Kim, J. Y., Sung, Y. E., Lee, J. J. & Kim, S. R. Pt and TCO free
hybrid bilayer silver nanowire-graphene counter electrode for dye-sensitized solar
cells. Chem. Phys. Lett. 561–562, 115–119 (2013).
341. Wang, G., Zhuo, S. & Xing, W. Graphene/polyaniline nanocomposite as counter
electrode of dye-sensitized solar cells. Mater. Lett. 69, 27–29 (2012).
342. Yeh, M. H. et al. A low-cost counter electrode of ITO glass coated with a
graphene/Nafion?? composite film for use in dye-sensitized solar cells. Carbon N. Y.
50, 4192–4202 (2012).
343. Lin, J.-Y. et al. Hydrothermal synthesis of graphene flake embedded nanosheet-like
molybdenum sulfide hybrids as counter electrode catalysts for dye-sensitized solar
cells. Mater. Chem. Phys. 143, 53–59 (2013).
344. Jaramillo, T. F. et al. Identification of Active Edge Sites for Electrochemical H2
Evolution from MoS2 Nanocatalysts. Science (80-. ). 317, 100–102 (2007).
178
345. Cunningham, G. et al. Percolation scaling in composites of exfoliated MoS2 filled
with nanotubes and graphene. Nanoscale 4, 6260–6264 (2012).
346. Wang, J.-Z. et al. Development of MoS 2 -CNT Composite Thin Film from Layered
MoS 2 for Lithium Batteries. Adv. Energy Mater. 3, 798–805 (2013).
347. McAteer, D. et al. Thickness Dependence and Percolation Scaling of Hydrogen
Production Rate in MoS2 Nanosheet and Nanosheet–Carbon Nanotube Composite
Catalytic Electrodes. ACS Nano 10, 672–683 (2016).
348. Stauffer, D. & Aharony, A. Introduction to Percolation Theory. (Oxford University
Press, New York, 1971).
349. Song, F. & Hu, X. Exfoliation of layered double hydroxides for enhanced oxygen
evolution catalysis. Nat. Commun. 5, 4477 (2014).
350. Ahmad, I., McCarthy, J. E., Baranov, A. & Gun'ko, Y. K. Development of
Graphene Nano-Platelet Based Counter Electrodes for Solar Cells. Materials (Basel).
8, 5284 (2015).
351. Pell, W. G. & Conway, B. E. Analysis of power limitations at porous supercapacitor
electrodes under cyclic voltammetry modulation and dc charge. J. Power Sources 96,
57–67 (2001).
352. Kim, Y. H. et al. Highly conductive PEDOT:PSS electrode with optimized solvent and
thermal post-treatment for ITO-free organic solar cells. Adv. Funct. Mater. 21, 1076–
1081 (2011).
353. Pau, P. C. F., Berg, J. O. & McMillan, W. G. Application of Stokes’ law to ions in
aqueous solution. J. Phys. Chem. 94, 2671–2679 (1990).
354. Hecht, D. S. et al. High conductivity transparent carbon nanotube films deposited from
superacid. Nanotechnology 22, 75201 (2011).
355. Chen, T., Xue, Y., Roy, A. K. & Dai, L. Transparent and stretchable high-performance
supercapacitors based on wrinkled graphene electrodes. ACS Nano 8, 1039–1046
(2014).
356. Lee, K. et al. Highly Transparent and Flexible Supercapacitors Using Graphene-
Graphene Quantum Dots Chelate. Nano Energy (2016).
doi:10.1016/j.nanoen.2016.06.030
357. Yoo,J.J.; Balakrishnan, K.; Huang, J.; Meunier, V.; Sumpter, B.G.; Strivastava, a.
Ultrathin Planar Graphene Supercapacitors. Nano Lett. 11, 1423–7 (2011).
358. Niu, Z. et al. A repeated halving approach to fabricate ultrathin single-walled carbon
nanotube films for transparent supercapacitors. Small 9, 518–524 (2013).
359. Gao, K. et al. Paper-based transparent flexible thin film supercapacitors. Nanoscale 5,
5307–5311 (2013).
360. Chen, P. C., Shen, G., Sukcharoenchoke, S. & Zhou, C. Flexible and transparent
supercapacitor based on In2 O 3 nanowire/carbon nanotube heterogeneous films. Appl.
Phys. Lett. 94, 2007–2010 (2009).
179
361. Ge, J., Cheng, G. & Chen, L. Transparent and flexible electrodes and supercapacitors
using polyaniline/single-walled carbon nanotube composite thin films. Nanoscale 3,
3084–8 (2011).
362. Liu, X. Y., Gao, Y. Q. & Yang, G. W. A flexible, transparent and super-long-life
supercapacitor based on ultrafine Co 3 O 4 nanocrystal electrodes. Nanoscale 4227–
4235 (2016). doi:10.1039/C5NR09145D
363. Strupinski, W. et al. Graphene epitaxy by chemical vapor deposition on SiC. Nano
Lett. 11, 1786–1791 (2011).
364. De, S. & Coleman, J. N. Are there fundamental limitations on the sheet resistance and
transmittance of thin graphene films? ACS Nano 4, 2713–2720 (2010).
365. De, S., King, P. J., Lyons, P. E., Khan, U. & Coleman, J. N. Size effects and the
problem with percolation in nanostructured transparent conductors. ACS Nano 4,
7064–7072 (2010).
366. Lyons, P. E. et al. High-performance transparent conductors from networks of gold
nanowires. J. Phys. Chem. Lett. 2, 3058–3062 (2011).
367. De, S. et al. Transparent, Flexible, and Highly Conductive Thin Films Based on
Polymer - Nanotube Composites. ACS Nano 3, 714–720 (2009).
368. Zhang, C. (John) et al. Highly Flexible and Transparent Solid-State Supercapacitors
Based on RuO2/PEDOT:PSS Conductive Ultrathin Films. Nano Energy (2016).
doi:10.1016/j.nanoen.2016.08.052
369. Liu, R., Duay, J. & Lee, S. B. Redox Exchange Induced MnO 2 Electrochemical
Energy Storage. ACS Nano 4, 4299–4307 (2010).
370. Gui, Z., Duay, J., Hu, J. & Lee, S. B. Redox-exchange induced heterogeneous RuO2-
conductive polymer nanowires. Phys. Chem. Chem. Phys. 16, 12332–40 (2014).