DEVICE ENGINEERING FOR EFFICIENT DYE-SENSITIZED SOLAR CELLS

A DISSERTATION

SUBMITTED TO

THE DEPARTMENT OF APPLIED PHYSICS

AND THE COMMITTEE ON GRADUATE STUDIES

OF STANFORD UNIVERSITY

IN PARTIAL FULFILLMENT OF THE REQUIREMENTS

FOR THE DEGREE OF

DOCTOR OF PHILOSOPHY

George Yan Margulis

August 2013

http://creativecommons.org/licenses/by-nc/3.0/us/

This dissertation is online at: http://purl.stanford.edu/wm521zm2572

© 2013 by George Yan Margulis. All Rights Reserved.

Re-distributed by Stanford University under license with the author.

This work is licensed under a Creative Commons Attribution-Noncommercial 3.0 United States License.

ii

I certify that I have read this dissertation and that, in my opinion, it is fully adequatein scope and quality as a dissertation for the degree of Doctor of Philosophy.

Michael McGehee, Primary Adviser

I certify that I have read this dissertation and that, in my opinion, it is fully adequatein scope and quality as a dissertation for the degree of Doctor of Philosophy.

Ian Fisher, Co-Adviser

I certify that I have read this dissertation and that, in my opinion, it is fully adequatein scope and quality as a dissertation for the degree of Doctor of Philosophy.

Stacey Bent

Approved for the Stanford University Committee on Graduate Studies.

Patricia J. Gumport, Vice Provost for Graduate Education

This signature page was generated electronically upon submission of this dissertation in electronic format. An original signed hard copy of the signature page is on file inUniversity Archives.

iii

iv

Abstract

Dye-sensitized solar cells (DSCs) offer a variety of advantages to typical

silicon and thin film solar cells. And while the advantages of ease-of-processing

and fabrication from low-cost, earth-abundant materials make DSCs an attractive

technology, the efficiency of DSCs (13%) is still too low to compete with the current

inorganic incumbents. Hence, new ‘outside-of-the-box’ strategies must be used to

render DSCs competitive with current commercial technologies. This thesis describes

my work on identifying losses in DSCs and 2 strategies to improve the efficiency of

DSCs: the use of highly-soluble energy relay days to broaden the spectral response of

DSCs, and the fabrication of semi-transparent solid-state DSCs to help improve

the efficiency of inorganic devices in a tandem solar cell.

Solid-state dye-sensitized solar cells (ssDSCs) have historically lagged behind

their liquid-electrolyte counterparts in efficiency. To gain a better understanding of

why this is so, we have developed accurate internal quantum efficiency (IQE)

measurements for ssDSCs. By analyzing the IQE, it is found that while charge

collection is efficient in ssDSCs, often charge injection is not. This analysis also

shows that parasitic absorption by the Spiro-OMeTAD is an important loss

mechanism in ssDSCs and suggests that stronger absorbing sensitizers are the most

promising path to higher efficiencies.

In DSCs, the roles of absorbing light, injecting charge, and blocking

recombination are all given to the sensitizing dye, resulting in a myriad of design rules

for DSC sensitizers. An energy relay dye (ERD) is a second dye that helps relax these

design rules by providing complementary absorption and then transferring energy to

a sensitizing dye. However, such ERDs come with their own design rules, including

the need for high solubility for full light absorption, and high photoluminescence

for efficient energy transfer. We have designed and synthesized two such dyes, and

characterized them as ERDs in DSCs, yielding a 65% increase in efficiency.

Finally, even if DSCs are unable to reach efficiencies that render them

competitive against traditional inorganic solar cells, DSCs can be used in conjunction

with an inorganic solar cell in a hybrid tandem photovoltaic (HTPV). High open-

v

circuit voltages and cheap processing render DSCs attractive top cells in HTPVs, and

such devices can exceed efficiencies of 20%. However, in order to be used in HTPVs,

a DSC must be fabricated such that below bandgap light can pass through the

device and be absorbed by the inorganic bottom cell. Toward that end, we have

developed a transparent top contact for solid-state dye-sensitized solar cells that

renders ssDSCs attractive candidates for HTPVs.

vi

Acknowledgements

First and foremost, I would like to thank Professor Mike McGehee for the

opportunity to do my dissertation research in his group. His guidance and mentorship

were paramount in my development as a young scientist and he has taught me an

incredible amount about solar cells, materials, and experimental physics in general.

But additionally, Mike has helped me develop a variety of skills outside of science and

research – he has helped me become a more effective speaker, a better writer, and a

more capable teacher. I would like to thank him for his support (both financial and

otherwise) during my PhD work, and for helping me develop a diverse skill set as a

researcher. I am grateful to consider him a mentor and friend.

I would like to thank Professor Ian Fisher, who gave me my start as a scientist

during my junior year of college as an undergraduate researcher in his group. My

wonderful experiences in his lab prompted me to continue in graduate research work.

I would also like to thank Professors Stacey Bent and Dan Stack for being members of

my defense committee and for the guidance they have provided on our collaborations

with their students. Finally, I would like to thank Professor Michele Digonnet, the

final member of my thesis committee.

Two student mentors were particularly important in teaching me about solar

energy and materials during the start of my PhD research: Dr. Eric Hoke and Dr. Brian

Hardin. Their collaboration and guidance helped formulate much of the research

direction of this thesis, and I am particularly grateful for this. Additionally, I had the

pleasure of closely collaborating on multiple projects with Dr. Bogyu Lim and

Greyson Christoforo without whom I could not have done any of this thesis work.

Bogyu is responsible for all the synthesis of ERDs reported in Chapter 4, and Grey

developed the method of spray deposition of silver nanowires used in Chapter 5. I am

incredibly grateful for these partnerships. I would like acknowledge all my

collaborators, both at Stanford and abroad in EPFL and Munich. I would like to

personally thank every member of the McGehee group that I have had the pleasure of

interacting with – both for our discussions relating to science and discussions that

could not be further from our work. Additionally, I would like to thank the diverse

vii

group of friends and colleagues that I had at Stanford and in San Francisco, without

whom I don’t think I could have completed the endeavor of a PhD.

I am grateful for funding provided by the ABB Corporation through a Stanford

Graduate Fellowship in Science and Engineering, and the other sources of funding that

have allowed me to pursue my research: Office of Naval Research, TomKat Center for

Sustainable Energy and the Precourt Institute for Energy.

Finally, I would like to thank my family for standing behind me through my

graduate studies. While there may only be a small amount of them, the amount of

support that they have provided of the past 5 years has been incredible. The emphasis

on education, intellectual curiosity and problem solving by my parents and

grandparents is why I was inclined toward science and research, and I hope to pass

this legacy to my children some day. And last, but certainly not least, I would like to

thank Angela for her love and support these past two years.

viii

Dedication

To my father, Yan, from whom I learned that education is a lifelong journey

rather than just a means to an end.

ix

Table of Contents

List of Tables ..................................................................................................... xi

List of Figures ................................................................................................... xii

1 Introduction .................................................................................................. 1

1.1 Liquid Electrolyte Dye-Sensitized Solar Cells ..................................... 1

1.2 Solid-state Dye-Sensitized Solar Cells ................................................. 4

1.3 DSC Operation Principles ..................................................................... 6

1.4 An Abridged History of DSCs from 2008-2013 ................................... 7

2 Overview of Experimental Techniques ...................................................... 12

2.1 Current-Voltage Characteristics ......................................................... 12

2.2 External Quantum Efficiency and Absorption .................................... 14

2.3 Photoluminescence Measurements ..................................................... 16

2.4 Impedance Spectroscopy Measurements ............................................ 17

2.5 Layer Thickness Measurements .......................................................... 18

2.6 Other Measurements ........................................................................... 19

3 Internal Quantum Efficiency Measurements of ssDSCs ............................ 20

3.1 Measurement of IQE ........................................................................... 23

3.2 IQE for Z907 and TT1 Dyes ............................................................... 30

3.3 Effect of Coadsorbent on IQE ............................................................ 36

3.4 Quantification of Parasitic Absorption Losses From Modeling ......... 38

3.5 Conclusion .......................................................................................... 40

3.6 Experimental Details ........................................................................... 41

4 Highly Soluble Energy Relay Dyes ........................................................... 44

4.1 Dye Structure, Characterization and Förster Radius Calculations ..... 47

4.2 ERD DSC Characterization ................................................................ 50

4.3 ERD DSC EQE and Energy Transfer Efficiency ............................... 53

4.4 Introduction to Quenching and Analysis of ETE Losses .................... 56

4.5 Dynamic Quenching and Pore Size Dependence ............................... 58

4.6 ETE Losses Due to Sensitizing Dye Desorption ................................ 60

4.7 Static Quenching of ERDs .................................................................. 64

x

4.8 Conclusion .......................................................................................... 68

4.9 ERD Synthesis .................................................................................... 69

4.10 Experimental Details ....................................................................... 74

5 Silver Nanowire Electrodes for Semitransparent ssDSCs ......................... 77

5.1 Modeling ............................................................................................. 78

5.2 Transparent ssDSC Applications ........................................................ 84

5.3 Device Architecture ............................................................................ 86

5.4 Role of PEDOT:PSS Layer................................................................. 88

5.5 Electrode Characterization .................................................................. 92

5.6 Device Results .................................................................................... 93

5.7 Conclusions ......................................................................................... 97

5.8 Future Outlook .................................................................................... 98

5.9 Experimental Details ......................................................................... 100

6 Conclusions and Future Outlook .............................................................. 103

7 Copyright .................................................................................................. 105

8 References ................................................................................................ 106

xi

List of Tables

Table 3.1. Modeled Jsc of a 2-µm-thick Z907 ssDSC if various layers had no

parasitic absorption. Reference has no layers set to 0 absorption. ............. 39

Table 4.1. J-V characteristics for TT1 devices incorporating BL302 and BL315 as

an ERD. J-V curves are shown in Figure 4.6. ............................................ 51

Table 5.1. J-V characteristics for HelioVolt CIGS bottom cell used in HTPV

modeling. .................................................................................................... 80

Table 5.2. Photovoltaic figures of merit for best-performing semitransparent ssDSC

and reference device using an evaporated Ag electrode under simulated

AM 1.5G illumination. ............................................................................... 96

Table 5.3. Figures of merit of modeled PSSC-CIGS HTPV using a silver nanowire

electrode. The total device efficiency in a 4-terminal configuration is

19.2% .......................................................................................................... 99

xii

List of Figures

Figure 1.1. Schematic diagram of DSC depicting various layers. .................................. 2

Figure 1.2. a) Device architecture of typical ssDSC. b) Chemical structure of Spiro-

OMeTAD ...................................................................................................... 4

Figure 1.3. Energy level diagram of DSC. Energy levels are approximate and may

be modified with the use of different materials or interfacial dipoles. ......... 6

Figure 2.1. Typical J-V curve of a solar cell taken under illumination and in the

dark. The region of the curve where the max-power point (MPP), short-

circuit current (JSC), the open-circuit voltage (VOC), and series resistance

(RS) are measured are depicted on the curve. The ratio of the area of the

gray box to the product of JSC and VOC is the fill factor (FF). ................... 13

Figure 2.2. Schematic diagram of chopped EQE set-up. The photocurrent out of the

photodiode and DSC are typically measured by the lock-in after

amplification by a current-to-voltage transimpedance amplifier. The

reference photodiode is used to account for any fluctuations in intensity

of the monochromatic light source. ............................................................ 15

Figure 2.3. Schematic diagram of absorptance measurements using an integrating

sphere. ......................................................................................................... 16

Figure 3.1. EQE of ssDSC with no sensitizer. Photocurrent generated below 425

can be attributed to TiO2 absorption. Despite the absorption of spiro-

OMeTAD between 425 and 550 nm, there is almost no photoresponse in

this portion of the spectrum. ....................................................................... 21

Figure 3.2. a) Chemical Structure of Z907 dye. b) Chemical structure of TT1 dye. ... 22

Figure 3.3. Schematic diagram of ssDSC layers. ......................................................... 24

Figure 3.4. Representative cross-sectional SEM of ssDSC showing various device

layers. Layers visible (from top going downward): spiro-OMeTAD

overlayer, active layer, compact TiO2 layer, FTO. SEM images were

analyzed with ImageJ software. ................................................................. 26

Figure 3.5. Imaginary portion of index of refraction of Z907-dyed active layer,

TT1-dyed active layer, and un-dyed active layer as measured by optical

xiii

absorption. The imaginary component of the index of refraction, κ, is

related to the thin film absorption coefficient, α, by α=4πκ/λ, where λ is

the wavelength of light. .............................................................................. 27

Figure 3.6. Comparison of measured and modeled device absorptance using transfer

matrix modeling with no averaging. Total modeled parasitic absorptance

is shown as a dashed black line and the dye absorptance, or ABSmeasured,

DSC - ABSmodeled,parasitic is depicted by the dashed gray line. ........................ 28

Figure 3.7. a) Modeled absorptance for each layer of a 2.3-µm-thick Z907 ssDSC:

total device absorptance, active layer absorptance, FTO absorptance,

parasitic absorptance within the active layer, glass absorptance, and sum

of the absorptances of other layers (this corresponds to the TiO2 compact

layer, Spiro-OMeTAD overlayer and silver cathode). b) Comparison of

modeled and measured device absorptance. Total modeled parasitic

absorptance is shown as a dashed black line and the dye absorptance, or

ABSmeasured, DSC-ABSmodeled,parasitic is depicted by the dashed

gray line. This modeling is done using the averaging scheme described

in the text. ................................................................................................... 29

Figure 3.8. Dye absorptance as given by ABSmeasured, DSC - ABSmodeled,parasitic,

measured EQE, and calculated IQE plots for (a) 2.3-µm-thick Z907

device and (b) 2.2-µm-thick TT1 device. Dotted gray lines denote error

bars in IQE measurement based on a ±20% error in modeling parasitic

absorptance. Vertical black lines depict averaging range for calculating

a single IQE value for each device. ............................................................ 31

Figure 3.9. Dye absorptance as given by ABSmeasured, DSC - ABSmodeled,parasitic,

measured EQE, and calculated IQE plots for 2.3-µm-thick Z907 device

when the parasitic absorptance is modeled using a) no averaging, b) the

averaging scheme described in the text (this is the same as Figure 3.8a).

Dotted gray lines denote error bars in IQE measurement based on a

±20% error in modeling parasitic absorptance. Vertical black lines

xiv

depict averaging range for calculating a single IQE value for each

device. ......................................................................................................... 33

Figure 3.10. a) Modeled ssDSC absorptance using no averaging for a 2.3-µm-thick

Z907 device: total device absorptance, dye absorptance, parasitic

absorptance. b) IQE (green) as measured by dividing EQE by this

modeled dye absorptance (black). .............................................................. 34

Figure 3.11. IQE vs. active layer thickness for Z907 (black squares) and TT1 (gray

circles) ssDSCs. Error bars are calculated from the standard deviation of

the IQE in the measurement range summed in quadrature with the

average error in IQE caused by increasing/decreasing the parasitic

absorption by 20% (depicted by gray dotted lines previously). This error

metric takes into account uncertainty caused by large amounts of

parasitic absorption and not-flat IQEs caused by inaccuracies in

modeling and/or measurement. Error bars are larger for thinner devices

due to the relatively low dye absorptance in the ssDSC. The light gray

dashed lines depict the average IQE for Z907 and TT1 devices. ............... 36

Figure 3.12. a) Comparison of modeled and measured absorptance for high cheno

(60 mM) TT1-sensitized device with active layer thickness of 3.7 µm.

Total modeled parasitic absorptance is shown as a dashed black line and

ABSmeasured, DSC-ABSmodeled,parasitic is depicted by the dashed gray line.

The solid and dashed black lines overlap where all the absorptance is

due to non-photoactive materials (400-575 nm). b) IQE, EQE and

ABSmeasured, DSC-ABSmodeled,parasitic for same device. Dotted gray lines

denote error bars in IQE measurement based on a 20% error in modeling

parasitic absorptance. Vertical black lines depict the averaging range

used for calculating a single IQE value for each device. ........................... 37

Figure 4.1. Schematic diagram of sensitized TiO2 nanoparticles in a DSC utilizing

an ERD. ...................................................................................................... 46

Figure 4.2. Chemical structure of dyes. Energy relay dyes: (a) BL315, (b) BL302,

(c) DCM. Sensitizing dye: (d) TT1. .......................................................... 48

xv

Figure 4.3. Absorption and emission spectra of DCM (black), BL302 (red), and

BL315 (blue) dyes in CH2Cl2 (10-5

M concentration). ............................... 48

Figure 4.4. Relative PL emission of 1 mM DCM in acetonitrile and benzonitrile and

1 mM BL302 in benzonitrile excited at 514 nm by an argon ion laser. ..... 49

Figure 4.5. a) Relative PL emission of 1 mM BL315 and 1 mM BL302 in

benzonitrile excited at 514 nm by an Ar laser. b) Normalized absorption

of TT1. Molar extinction coefficient of TT1 at peak is 191500 M-1

cm-1

. . 50

Figure 4.6. J-V curves of TT1-sensitized DSCs with various concentrations of a)

BL302 and b) BL315 compared to a reference TT1 sensitized device

utilizing a benzonitrile-based electrolyte. ................................................... 51

Figure 4.7. Nyquist plots (taken at 0.8 V forward bias) and impedance fits (using

the circuit shown in Figure 4.8) for TT1 sensitized DSCs employing a

benzonitrile electrolyte a) without any ERD and b) with 180 mM

BL302. ........................................................................................................ 52

Figure 4.8. Equivalent circuit modeling the operation of a DSC in far forward bias. . 52

Figure 4.9. EQE of 6 µm-thick TT1-sensitized DSCs containing various amounts of

BL302 (a) and BL315 (b) in the electrolyte. .............................................. 55

Figure 4.10. Absorptance of FTO as measured with an integrating sphere to account

for scattering and reflection. ....................................................................... 55

Figure 4.11. EQE of TT1 Device with 180 mM BL302 and difference between

transmittance of TT1 sensitized substrate and FTO substrate (which is

the sensitizing dye absorptance). The EQESD and ABSSD are measured at

their peaks. There appears to be a slight redshift in the TT1 EQE vs. the

TT1 absorptance which can be attributed to the presence of solvent.

EQEERD is measured at 520 nm, as the electrolyte absorbs relatively

strongly below 500 nm. .............................................................................. 56

Figure 4.12. Time resolved PL spectrum of BL302 in 85:15

benzonitrile:valeronitrile. A fit to the linear portion of the decay (on a

log-linear scale) gives a PL lifetime of approximately 2.0 ns. ................... 57

xvi

Figure 4.13. Photoluminescence spectrum of 10 mM BL302 with electrolyte. 25%

and 50% corresponds to 25% and 50% of the concentrations of ions that

is used in the DSC electrolyte. As can be seen, there is no significant

change in shape of the photoluminescence spectrum – likely because any

aggregates (that may have a different PL spectrum) do not

photoluminesce. There is a slight redshift, but this may be caused by the

addition of ions to the solvent. ................................................................... 58

Figure 4.14. Experimental ETE compared to expected theoretical ETE for spherical

and cylindrical pore geometries. ................................................................. 59

Figure 4.15. a) Dynamic quenching of BL302 dye in 85:15

benzonitrile:valeronitrile mixture as measured by time-resolved

photoluminescence. τ is the Fluorescence lifetime and τ0 is the

fluorescence lifetime at 0 electrolyte concentration. b) Dynamic

quenching of BL302 due to high concentrations of dye in benzonitrile

solvent. Here, τ is the fluorescence lifetime and τ0 is the fluorescence

lifetime at 1 mM dye concentration. Multiplying the decreases in

lifetime together results in a total τ0/τ of 6.9. ............................................. 59

Figure 4.16. Absorption of electrolyte after equilibration of dye desorption for

benzonitrile and acetonitrile based electrolytes. Inset: experimental

schematic of measuring light absorption through the electrolyte. After

allowing dye desorption to equilibrate, the absorption of the dye in the

electrolyte is measured using a beam path shown by the arrows in the

inset. ............................................................................................................ 60

Figure 4.17. Schematic diagram of desorbed sensitizing dye a distance x from the

center of a spherical sensitizing-dye-lined pore. ........................................ 61

Figure 4.18. Fraction of ERD excitation lost as a function of the distance of the

desorbed dye from the center of the pore for a 17nm diameter pore with

1 sensitizing dye per nm2

surface coverage. When the dye is near the

center, the r6 nature of the FRET interaction causes a large amount of

ERDs in its vicinity to preferentially FRET to the desorbed sensitizer. .... 63

xvii

Figure 4.19. Comparison of steady-state photoluminescence quenching with

decrease in photoluminescence lifetime for 1 mM, 10 mM, and 27 mM

BL302 with varying concentrations of electrolyte. Solid and dashed grey

lines are linear fits of PL and τ, respectively. Note: the linear trend

continues for both PL and τ to 100% electrolyte concentration.

Electrolyte concentration (%) is the percentage of electrolyte

components relative to the standard electrolyte used in DSC devices. ...... 65

Figure 4.20. a) Comparison of time-resolved and steady-state quenching of 5 mM

DCM in acetonitrile electrolyte. b) Comparison of time-resolved and

steady-state quenching of 8.5 mM DCM in benzonitrile electrolyte.

Note: Electrolyte concentration (%) is the percentage of electrolyte

components relative to the standard electrolyte used in DSC devices. ...... 66

Figure 4.21. Comparison of time-resolved and steady-state quenching of BL302 in

benzonitrile with varying concentration. .................................................... 67

Figure 4.22. Synthetic scheme for BL302 .................................................................... 70

Figure 4.23. 1HNMR spectrum of BL302. ................................................................... 71

Figure 4.24. Synthetic Scheme for BL315 ................................................................... 71

Figure 4.25. 1HNMR spectrum of BL315. ................................................................... 73

Figure 4.26. Cyclic Voltammogram of BL302 (red), and BL315 (blue) dyes. From

the curves, it was found that HOMOBL302=5.12 eV, LUMOBL302=3.07

eV, HOMOBL315=5.05 eV, LUMOBL302=3.23 eV. ...................................... 74

Figure 5.1. a) Schematic of DSC in a 4-terminal HTPV configuration. b) Schematic

of DSC in a 2-terminal HTPV configuration. The DSC is built on the Si

or CIGS solar cell and may have an interfacial layer to make electrical

contact and/or planarize the inorganic solar cell. The arrows depict light

that is incident on each of the subcells of the tandem. ............................... 78

Figure 5.2. EQE of HelioVolt CIGS cell used in this section’s modelling along with

EQE of a DSC utilizing a dye with a 2.0 eV ‘bandgap.’ ............................ 80

Figure 5.3. Example transmittance of top DSC with a 2.0 eV ‘bandgap’ dye used in

our modeling of DSC-CIGS HTPV tandems. The lowered transmittance

xviii

in the red and near IR is due to the two transparent conductors while the

dye absorbs most of the light below 620 nm (2.0 eV). ............................... 81

Figure 5.4. Efficiency of modeled DSC-CIGS HTPV. Two-terminal devices use

only 1 transparent conductor in the modeling of the top device

transmittance. The efficiency of the bottom CIGS cell is depicted by the

dashed red line. In Equation 5.5, ELoss is assumed to be 0.8 eV. The red

dashed line represents the efficiency of the CIGS cell by itself. ................ 82

Figure 5.5. Efficiency of modeled DSC-CIGS HTPV. Two-terminal devices use

only 1 transparent conductor in the modeling of the top device

transmittance. The efficiency of the bottom CIGS cell is depicted by the

dashed red line. In Equation 5.5, ELoss is assumed to be 0.5 eV. Red

dashed line represents CIGS efficiency. ..................................................... 83

Figure 5.6. a) Schematic diagram of semitransparent ssDSC device. The device

consists of a 400-nm-thick F:SnO2 (FTO) layer, 100-nm-thick compact

TiO2 layer, 2-μm-thick dye-sensitized active layer, 200-nm-thick Spiro-

OMeTAD overlayer, approximately 85-nm-thick PEDOT:PSS layer and

solution deposited silver nanowires. b) SEM micrograph of

semitransparent ssDSC cross section at 20° angle of incidence. c) SEM

image of Ag NW/PEDOT:PSS electrode at normal incidence. d) SEM

image of the PEDOT:PSS/Ag NW composite electrode at 3° angle of

incidence, showing that the wires are embedded in the PEDOT:PSS

layer. ........................................................................................................... 87

Figure 5.7. a) Current-voltage characteristics of ssDSC sensitized with D35 with no

PEDOT:PSS interfacial layer. The J-V shows a distinctive ‘s-shape’

which causes a low fill factor (FF) and efficiency (Eff). b) Energy level

diagram of ssDSC. The work functions of the Spiro-OMeTAD,

PEDOT:PSS and Ag NWs was measured by PESA, while other energy

levels are approximate and shown for comparison. ................................... 89

Figure 5.8. PESA measurement of a) sprayed Ag NW film (from ethanol) on glass,

b) Spiro-OMeTAD on glass, and c) PEDOT:PSS (Clevios™ CPP-105)

xix

on glass. The work functions are measured to be 4.5 eV, 5.2 eV, and 5.0

eV, respectively. Red lines are fits to the baselines and sloped regions of

the curves. A power number of 0.5 was used for the metallic nanowires,

while 0.3 was used for organic materials. .................................................. 90

Figure 5.9. a) PESA measurement of sprayed Ag NW film (from ethanol) on glass

(same same as Figure 5.8) after 10 minutes of UV-ozone treatment (in 1-

minute intervals). The work function is measured to be 4.9 eV (up from

4.5 eV), and doesn’t display a second slope. b) Current-voltage

characteristics of semitransparent ssDSCs using Z907 dye without an

interfacial PEDOT:PSS layer (utilizing only a Ag NW mesh as top

electrode). J-V curves are shown after exposure to UV-ozone treatment

for a given period of time. The J-V curve begins with an s-shape

indicative of a barrier to charge transport, but shows rectifying J-V

characteristics typical of a solar cell after 6-14 minutes of UV-ozone

treatment. .................................................................................................... 90

Figure 5.10. Transmittance of 85 nm PEDOT:PSS film, Ag NW film and Ag

NW/PEDOT:PSS composite electrode. Inset: Glass slide with Ag

NW/PEDOT:PSS composite electrode on the bottom half of the glass

substrate. ..................................................................................................... 93

Figure 5.11. a) EQE of reference ssDSC device using an evaporated silver electrode

and EQE of semitransparent ssDSC (using Ag NW/PEDOT:PSS

electrode) illuminated from both the FTO and Ag NW electrodes. b)

Chemical structure of D35 dye used in device fabrication. c) Picture of

semitransparent ssDSC. The Ag NW/PEDOT:PSS electrode is barely

visible as a slightly darker square in the middle of the device. .................. 95

Figure 5.12 J-V curves of best semitransparent ssDSC and best reference device

using an evaporated silver electrode. The difference in current between

devices is only 0.3 mA/cm2, which is slightly less than typical between a

reference device and a semitransparent device. Device area was

xx

approximately 0.5 cm2 and was masked with a 0.2 cm

2 mask. Even for

small areas (0.1 cm2), references in our lab are less than 4.0% efficient. .. 95

Figure 5.13. Fractional spectral distribution of incident light upon the

semitransparent ssDSC. Shown is the fraction of photons transmitted

through the semitransparent ssDSC (all layers other than glass substrate,

including the FTO), the fraction absorbed by the semitransparent ssDSC,

and the fraction absorbed by the soda-lime glass substrate. The light

area at the top of the plot denotes reflected photons. Transmittance and

absorptance measurements were carried out using an integrating sphere

to account for scattering. ............................................................................ 97

Figure 5.14. Modeled fractional spectral distribution of record PSSC using the

silver nanowire electrode developed in this chapter, along with a CIGS

solar cell as the bottom cell in a HTPV. ..................................................... 99

1

1 Introduction

While in typical silicon and inorganic thin film solar cells, one semiconductor

is responsible for light absorption, electron transport and hole transport, the basic idea

behind dye-sensitized solar cells (DSCs) is the spatial separation of these processes

into 3 separate materials. DSCs typically utilize a dye to absorb light and then inject

electrons into a wide band gap metal-oxide semiconductor. Holes are transported to

the counter electrode by mass transport of ions within an electrolyte or by transport

through an organic hole transport material. The spatial separation of electrons and

holes in the DSC allows for the use of much less pure materials, such as

nanocrystalline metal-oxide semiconductors, than in typical silicon solar cells.

This use of nanoparticles by O’Regan and Grätzel1 was the research innovation

that spawned the field, as it increased the surface area of the metal-oxide

semiconductor, sensitizing dye loading and device absorption coefficient by a factor of

1000 from a that of a flat monolayer of dye. After this breakthrough in 1991, the field

has grown enormously, with a variety of research focusing on dyes, hole transport

materials, redox shuttles, metal oxide nanoparticles, new device architectures and

device physics. Rather than a cursory survey of this large body of research, this

section will focus on introducing the basic concepts of the structure and operation of

DSCs that are needed to explain this thesis work. At the end of the section, a short

subsection is contained on the timeline of breakthrough DSC research during the

period when this thesis work was carried out (2009-2013), which should serve to

motivate the dissertation.

1.1 Liquid Electrolyte Dye-Sensitized Solar Cells

The structure of the typical liquid-electrolyte dye-sensitized solar cell (or

simply dye-sensitized solar cell or Grätzel cell) is shown in Figure 1.1. The device is

typically built upon a glass substrate that is covered with a conductive layer of F:SnO2

(FTO) which serves as the transparent contact to the anode. On top of this layer, a

layer of mesoporous metal-oxide, typically TiO2 nanoparticles (typically 20 nm, but

nanoparticles as large as 400 nm are used for light scattering) are deposited from

2

solution, usually via screen printing or doctorblading. The paste is then sintered at

450 °C, removing any organic binders. Multiple layers of TiO2 may be deposited and

then sintered. TiO2 substrates are then immersed in TiCl4 solution overnight, which

serves to increase connectivity between nanoparticles, and decrease recombination.2,3

From here, the substrate is once again sintered at 450 °C and allowed to cool to room

temperature before being immersed in a dilute solution of sensitizing dye (typically

0.1-1 mM) for a period ranging from hours to days. A FTO counter electrode is

fabricated by drilling a 1 mm hole in the FTO, covering the FTO with a solution of

platinum salt (chloroplatinic acid hydrate) and then heating at 450 °C using a heat gun.

The sensitized titania film is then sealed to the counter electrode using a plastic spacer

(typically 25-μm-thick Surlyn®). Finally, the electrolyte containing the redox shuttle

and a variety of additives is filled into the device through the counter electrode hole

and the device is sealed with another layer of Surlyn®.

Figure 1.1. Schematic diagram of DSC depicting various layers.

DSCs have currently achieved record efficiencies of 12.3% in lab-scale

devices, with the record being achieved by a system utilizing a zinc-porphyrin dye,

YD2-o-C8.4 These results are very promising, but for widespread commercialization,

3

DSCs must achieve efficiencies of at least 18-20% on the lab scale, allowing for 15%

efficient DSC modules. Much of the improvement in DSC efficiency has come from

the optimization of the individual components of the device, particularly the invention

of new dyes and redox shuttles.

Over the years, each component of the DSC has investigated by a variety of

groups. Rather than going into detail on the many directions research has taken, I

would like to point the reader toward a variety of reviews and/or seminal papers on

each component of the DSC. Particularly interesting research efforts on the

photoanode have included investigating various metal oxides,5 optimizing metal oxide

nanostructures,6 and optimization of device thickness and porosity. Current record

results have been obtained using a 2-6 μm layer of small (typically 20 nm) titania

nanoparticles for high surface area and strong light absorption, along with a top

scattering layer of approximately 4 μm of 400 nm TiO2.

A huge variety of dyes have been investigated as sensitizers in DSCs, with a

few dyes performing well. To date, 3 types of dyes have achieved efficiencies of over

10% in DSCs: Ruthenium based dyes (N719 among others),7,8

organic dyes (Y123

among others),9 and the aforementioned zinc-porphyrin record holder (YD2 and YD2-

o-C8).4,10

Traditionally, high efficiency sensitizing dyes have been Ru-based, but in

the past few years, these new high efficiency organic and porphyrin dyes have been

developed. Some reviews exist on these three classes of dyes and structure-property

relationships in dye design.11,12

Until recently, nearly all DSC research utilized the traditional iodide-triiodide

redox shuttle for hole transport. An excellent overview written by Boschloo and

Hagfeldt exists on the properties of the I-/I3

- redox mediator.

13 The other components

of the electrolyte have been empirically found to decrease recombination or induce

favorable dipoles at the TiO2 interface14,15

and optimal concentrations depend on the

energetics and interactions of the specific sensitizing dye, TiO2 and redox couple.16

During the course of my thesis work, there have been significant breakthroughs in the

use of alternate redox couples,17,18

specifically cobalt polypyridyl complexes4,9,19

which are superior to the typical I-/I3

- due to the requirement of less driving force for

4

dye regeneration.20

Platinum is typically used to catalyze the reduction of I-, but other

catalysts have been developed, particularly for alternate redox couples.21

A variety of fantastic articles exist that provide an overall review of DSC

research. Two of my personal favorites are an older 2003 review by Grätzel22

and an

excellent 2010 review by Hagfeldt.23

Hardin et al. have also published a short but

valuable article describing recent (2010-2012) breakthroughs in the field.24

1.2 Solid-state Dye-Sensitized Solar Cells

Solid-state dye-sensitized solar cells (ssDSCs) were developed as a stable

alternative to traditional liquid electrolyte DSCs due to the corrosiveness of the I-/I3

-

redox couple. Rather than using mass transport of ions to transport holes to the

cathode, ssDSCs rely on transport through an infiltrated organic small molecule hole

transport material or HTM. The most widely used HTM is Spiro-OMeTAD (2,2’,7,7’-

tetrakis-(N,N-di-p-methoxyphenylamine)9,9’-spirobifluorene – shown in Figure 1.2b)

which has resulted in efficiencies as high as 7.2%.25

The structure of an ssDSC is

depicted in Figure 1.2a.

Figure 1.2. a) Device architecture of typical ssDSC. b) Chemical structure of Spiro-

OMeTAD

Fabrication of an ssDSC is very similar to a liquid DSC, except for a few

modifications. A 50-μm-thick layer of spray-pyrolyzed layer of TiO2 is deposited on

the FTO (from a solution of titanium diisopropoxide bis(acetylacetonate)). This layer

serves as a hole-blocking layer that stops the Spiro-OMeTAD from shorting to the

5

bottom FTO electrode. From here, the deposition of the mesoporous layer follows as

in the case of DSCs, except the optimum active layer thickness tends to be

approximately 2 μm of 20 nm TiO2 nanoparticles. Dye sensitization continues

similarly to liquid DSCs except the thinner TiO2 usually decreases the required time

for the sensitization process.26,27

Rather than sealing the cell and backfilling with

electrolyte, the TiO2 substrate is infiltrated with Spiro-OMeTAD by either spin-

coating or doctor-blading.28

This pore-filling process results in a pore-filling fraction

of approximately 65% for thin (between 1-um-thick and 4-μm-thick) films,29,30

allowing for relatively good charge transport and collection.31

For efficient ssDSCs,

the Spiro-OMeTAD must be doped, typically with a lithium salt, to increase material

conductivity.25,32,33

The top electrode for the ssDSC is composed of a evaporated

metal layer, typically gold or silver.

Research on ssDSCs has also gone in a variety of directions: including design

of new hole-transport materials, new dyes, and the various metal-oxide nanostructures.

However, best results for ssDSCs have been achieved using strongly absorbing D-π-A

dyes34,35

and Spiro-OMeTAD. A recent (2012) review of ssDSCs written by Hsu, et

al.36

gives a nice summary of progress in the field.

A giant leap in ssDSC research recently occurred in 2012 with the replacement

of the sensitizing dye by a hybrid perovskite absorber.37,38

Such materials had

previously been used in liquid DSCs without much success due to the dissolution of

the perovskite by the electrolyte.39,40

This problem doesn’t exist in ssDSCs and

perovskite sensitized solid-state solar cells have very quickly achieved efficiencies of

greater than 12%, with reports as high as 15% in early 2013.38,41

These efficiencies

have been typically achieved with CH3NH3PbClxI3-x, but other chemical structures

may also lead to high efficiencies in the near future. These devices have been called

meso-superstructured solar cells or MSSCs by Henry Snaith, and their structure is very

similar to ssDSCs save for the replacement of the sensitizing dye with perovskite and

a typically thinner active layer (typically under 1-μm-thick). Due to the high

efficiencies that these cells have achieved in such a short period of time, a lot of

6

ssDSC researchers have switched directions and began studying perovskite-sensitized

solar cells.

1.3 DSC Operation Principles

As mentioned previously, the idea behind the DSC is to separate the functions

of light absorption, electron transport, and hole transport spatially into 3 different

materials. In order for this to occur, an energetic driving force is needed for electron

injection into the titania and hole regeneration by the redox couple/HTM. The energy

level diagram for an ssDSC is shown in Figure 1.3. After light is absorbed the

sensitizer, the excited electron is injected into the titania conduction band, typically

requiring a driving force on the order of tenths of an electron volt.14,42

On the other

hand, regeneration by the I-/I3

- redox couple requires a driving force of at least 0.4 eV,

leading to a significant energy loss.13

Other redox couples based on ferrocene and

cobalt polypyridyl complexs have displayed a smaller minimum energy loss required

for efficient charge regeneration.18–20,43

As mentioned in passing previously, if the dye

LUMO and/or HOMO is at a level that does not allow for efficient electron injection

or charge regeneration, interfacial dipoles can be used to help correct the problem.44,45

Figure 1.3. Energy level diagram of DSC. Energy levels are approximate and may be

modified with the use of different materials or interfacial dipoles.

7

Once the electron and hole are injected into the TiO2 and HTM/redox couple,

respectively, the charges must diffuse out of the device to the electrodes. However,

during this time, charges can recombine – either across the TiO2 interface or with

excited charges on the sensitizing dye molecules – leading to an excitation loss.

Recombination in DSCs is typically probed using photocurrent and photovoltage

transients,46

or impedance spectroscopy.47

The voltage produced by a DSC is

determined by the offset between the Fermi level in the TiO2 and the redox potential

of the redox couple or the Fermi level in the Spiro. Once again, this energy difference

can be increased through the use of dipoles, but doing so can also shut off charge

injection and regeneration.

In many ways, the sensitizer is the key component of DSCs as it is the main

determinant of the efficiency of the device. Clearly, the dye structure determines the

absorption of the device which determines the maximum possible photocurrent the

device can generate. From there, the energy levels of the dye HOMO and LUMO,

along with their electronic interactions with the HTM/redox couple and TiO2,

respectively, determine the efficiency of electron injection and hole regeneration. The

dye also acts as a barrier at the interface between the hole transporting medium and the

electron transporting TiO2, helping stop recombination across this interface. Thus the

dye is responsible for increasing the lifetime of charge in the device which can

increase device photovoltage. Given the multitude of responsibilities of the dye in

DSCs, it shouldn’t be a surprise that a large amount of progress in the field can be

attributed to the creation of new dyes.24

1.4 An Abridged History of DSCs from 2008-2013

One of the most exciting aspects of my research has been the pace at which the

field has progressed during the past 5 years. In this short period of time, there have

been a variety of novel breakthroughs that have lead to new record efficiencies of

DSCs or ssDSCs. Additionally, 2008-2013 has been an extremely exciting time in

general for the solar industry, as the price of silicon photovoltaic modules dropped

from approximately $3.50 per watt peak ($/Wp) in 2008 to $0.64 per watt peak in Q1

8

of 2013.48

This precipitous decline in silicon solar module prices has led to a

continued increase in installation of photovoltaic modules, but has also affected the

direction that solar research, including DSC research, has taken. Hence, these changes

in the economics of solar energy and recent breakthroughs have really changed the

goals and direction of research in the field, having a pretty profound effect on my

research during my thesis. In this section, I would like to give a brief synopsis of what

I considered to be the biggest breakthroughs in DSCs during my thesis, and use this to

put a historical perspective on my work.

After DSCs were invented in 1991 by O’Regan and Grätzel,1 efficiencies came

up relatively quickly to the 10% mark in lab-scale devices, first hitting that efficiency

in 1993.8 While interest in the field continued to increase, with more and more

research groups working in DSCs, over the next 15 years the record efficiency of

DSCs increased to only approximately 11%. A couple reasons can be cited for the

slow progression. First, it is possible that early record efficiencies were mismeasured

due to researchers being unfamiliar with techniques required for accurate efficiency

measurement such as calculating spectral mismatch49

and accurately masking solar

cells.50

Secondly, much early research centered on Ru-based dyes. The absorption of

typical Ru sensitizers is very broad, but relatively weak, requiring thick (10 μm) active

layers for full absorption of incident light, which further dropped the photovoltage in

addition to energetic losses due to charge transfer.

Another serious problem with DSCs was the corrosive and volatile nature of

the electrolyte, which lead to stability and leakage issues. Because of this, Bach and

Grätzel developed the ssDSC in 1998, with Spiro-OMeTAD as the stable replacement

for the liquid electrolyte. Early progress on ssDSCs was relatively slow, with record

efficiencies near 4% in 2008.51

At the time, ssDSC research was focused on

understanding why ssDSCs lagged behind their liquid electrolyte counterparts, and

some of the bigger questions in the field were: does the diffusion length limit ssDSCs?

Is charge injection/regeneration different in ssDSCs and liquid DSCs? The 3rd

chapter

of this thesis on the internal quantum efficiency of ssDSCs investigates these

questions,52

and comes to the conclusion that the best way to increase the efficiency of

9

ssDSCs is to use more strongly absorbing D-π -A dyes. And indeed, in 2011, the

record ssDSC efficiency was set by such a dye,53

and this knowledge has help guide

our group’s research toward efficient (6.3%) D-π –A sensitizers.35

However, the rapid drop in Si solar module prices added a further urgency to

DSC and ssDSC research. When the price of Si modules was at approximately

$3.50/Wp, the main goal for researchers in alternative solar technologies was to simply

develop low cost solar cells. However, as the price of the module dropped, a

significant portion of the installed price of solar became the balance of systems (BOS)

cost. For researchers working on the solar cell itself, the only way to decrease the

BOS cost is to increase the efficiency: this allows less solar modules to be installed in

order to provide the same power output. As the price of Si solar modules dropped,

there began to be new pressure to bring the efficiency of DSCs above the 10-11% that

it had plateaued at in order to be more cost competitive with the increasingly cheap Si

technology.

To achieve this end, in the latter part of the 2000s decade, there was a push to

move away from these classic Ru-based dyes as a way to continue efficiency progess.

Significant progress came in 2010, when Bessho and Grätzel achieved 11% efficiency

using a zinc-porphyrin sensitizer.10

While this sensitizer absorbed fairly strongly in

two bands, it missed the green portion of the solar spectrum – realizing a new need for

a complementary absorber to break efficiencies of 12%. It is this need that was the

motivation for the work described in Chapter 4: Highly Soluble Energy Relay Dyes.54

If a sensitizer could be synthesized that had excellent injection efficiencies with low

energy loss, a record efficiency of 13% could be obtained using complementary

absorbers to help harvest the entire solar spectrum. However, cosensitization of a

second dye onto the TiO2 surface can cause a variety of problems, such as unfavorable

dipoles or increasing recombination. In addition, each cosensitized dye competes for

TiO2 adsorption sites. The concept of the energy relay dye or ERD by Hardin and

McGehee in 2009 helps ameliorate these issues.55

Indeed, other groups realized the

need for efficient complementary absorbers, and today the current record efficiency of

12.3% is held by a porphyrin dye cosensitized with an organic sensitizer (with the use

10

of a cobalt polypyridyl redox couple).4 Indeed, this brought the current record

efficiency close to Snaith’s estimate of the maximum attainable efficiency by DSCs of

13-14%.56

However, at the same time, the cost per watt of silicon modules had continued

to decrease to approximately $0.64/Wp in the first quarter of 2013. At these types of

prices and module efficiencies of as high as 15%, it would be difficult for DSCs to

compete even if they were able to obtain lab-scale efficiencies close to the theoretical

maximum of 13-14%. While DSCs alone could not achieve efficiencies over 20%,

DSCs possessed many properties that make them attractive as a top solar cell in a

tandem device: relatively high open-circuit voltages and near 100% internal quantum

efficiencies. Using a 15% Silicon or CuInGaSe (CIGS) bottom cell in such a hybrid

tandem photovoltaic (HTPV) in conjunction with a DSC top cell, efficiencies of

greater than 20% could be achieved.57

Such a device may have commercial potential,

and this has become the research direction for a variety of groups: working on semi-

transparent devices that can be used as top cells in such tandems. At the same time,

there was a massive breakthrough in the ssDSC field with the aforementioned use of

perovskite sensitizers.37,38

If an effective transparent top electrode could be made to

make such devices semi-transparent, they could be implemented in a HTPV with

efficiencies significantly exceeding 20%. Since their structure is nearly identical to

that of ssDSCs, this idea motivated the 5th chapter of my thesis: fabricating high

quality transparent electrodes for ssDSCs. 58

Rather than presenting my work based on the type of device (either solid-state

or liquid electrolyte), the work is presented chronologically following the evolution of

the goals of the DSC field. Despite the differences in the various projects I have

worked on, the goal has been the same: push DSCs toward higher efficiencies. As

touched upon previously, much of improvement in efficiency can be attributed to the

work of synthetic chemists on new sensitizers. In my opinion, the role of device

physicists/engineers is threefold: optimization and analysis of new synthesized

materials for DSCs, analysis of losses and understanding of how improvements can be

made, and design and fabrication of new device architectures that can help achieve

11

higher efficiencies. I have worked on all three these aspects of device engineering and

feel that this is the theme that connects my thesis work.

12

2 Overview of Experimental Techniques

This section provides an overview of experimental techniques used during my

research. Rather than go into detail about the makes and models of the equipment, I

will attempt to give an overview of the experimental techniques and the information

that can be gleaned from them. Additionally, I will try to touch on considerations that

must be done to use the technique correctly and avoid common pitfalls. For more

details about the exact equipment and measurement parameters used in each

experimental setup, each chapter contains an ‘Experimental Details’ section with the

specifics of the measurement used in that section.

2.1 Current-Voltage Characteristics

The simplest, and perhaps most important technique in characterizing solar

cells is taking current-density–voltage (J-V) characteristics of photovoltaics (often

referred to as simply current-voltage characteristics). While the main use tends to be

to measure the power-conversion efficiency of solar cells, careful analysis of J-V

curves can lead to a wealth of information. Additionally, despite being a simple

measurement, incorrect measurement of J-V curves has led to a variety of incorrect

reported power conversion efficiencies in literature. J-V measurements can be done

both under illumination and in the dark, with measurement under the AM1.5G

simulated spectrum used to report power conversion efficiency. The current is

typically measured while the voltage is swept, and the entire curve is divided by the

device area, resulting in a graph of current density vs. voltage (shown below in Figure

2.1). The power conversion efficiency of the solar cell biased at a given voltage is the

product of the voltage and current density at that point, with the max-power point

(MPP) being the point of maximum efficiency. The current-density at 0 V bias and

the voltage at 0 current-density are known as the short-circuit current (JSC), and the

open-circuit voltage (VOC), respectively and the fill factor (FF) is given by the

efficiency divided by the product of JSC and VOC and is a measure of how

“rectangular” the J-V curve is. Another common characteristic for solar cells is the

series resistance (RS), which is usually measured as the inverse slope of the dark curve

13

in far forward bias giving it units of resistance times area. High series resistance

decreases the slope of the J-V curve in forward bias decreasing the efficiency by

primarily decreasing the FF.

Figure 2.1. Typical J-V curve of a solar cell taken under illumination and in the dark.

The region of the curve where the max-power point (MPP), short-circuit current (JSC),

the open-circuit voltage (VOC), and series resistance (RS) are measured are depicted on

the curve. The ratio of the area of the gray box to the product of JSC and VOC is the fill

factor (FF).

While the measurement of J-V curves is relatively straightforward in principle,

it is not difficult to mismeasure cell efficiency and device characteristics. While solar

cells are measured under AM1.5G spectral illumination, most solar simulators display

a quite different spectrum. To correct for this, solar cell measurements must be done

accounting for this spectral mismatch, which requires knowledge of the external

quantum efficiency (EQE) of the solar cell being measured. A discussion with

application to organic solar cells was published by Yang Yang in 2006,49

and I will

refer the reader to that paper for the details. One main point is that the power

conversion efficiency of two solar cells with different EQE spectra must be measured

under different solar simulator calibrations. Another important issue is that the area of

the device must be properly calibrated. Many solar cells are tested with the area set to

the overlapping size of the electrodes, which can greatly over-measure the efficiency if

the material of the solar cell has enough conductivity to laterally collect charge from

14

regions with no electrodes. Hence, masking is often used to ensure the accuracy of

illumination area. However, using a mask can decrease the efficiency, particularly if

the mask area is significantly smaller than the device area. Henry Snaith has

published an article discussing the effects of masking on solar cell efficiency

measurements.50

2.2 External Quantum Efficiency and Absorption

Probably the second most common measurement of solar cells is external

quantum efficiency or EQE. The value of EQE is simply given by the number of

charges out of the device divided by the number of incident photons at a given

wavelength. EQE is measured by illuminating a solar cell with monochromatic light,

measuring the current out of the device and comparing to a calibrating photodiode of

known EQE. The EQE of many solar cells, including DSCs changes as a function of

the illumination intensity incident upon the device. Hence, for DSCs the EQE is

measured under illumination using a white light bias on the order of 0.1-1.0 suns with

the intensity calibrated by measurement of the solar cell JSC on a calibrated solar

simulator. In this setup, the monochromatic light must be chopped at a given

frequency and then measured using a lock-in amplifier (experimental setup depicted in

Figure 2.2). Here the chopping frequency has to be chosen carefully depending on the

rate of the electron transport processes in the solar cell. For ssDSCs, a chopping

frequency of 40 Hz or slower usually yields accurate EQE measurements, while DSCs

require a chopping frequency on the order of 2 Hz due to the long time required for

charge diffusion through the device. Slower chopping also requires long integration

time constants on the lock-in measurement, resulting in a long time required for the

EQE measurement.

15

Figure 2.2. Schematic diagram of chopped EQE set-up. The photocurrent out of the

photodiode and DSC are typically measured by the lock-in after amplification by a

current-to-voltage transimpedance amplifier. The reference photodiode is used to

account for any fluctuations in intensity of the monochromatic light source.

The white light bias can also be tuned and the EQE as a function of light

intensity can be measured. At low light intensities (<0.1 suns), the EQE of a DSC is

typically low due to the effects of electron trapping the TiO2 and the resulting charge

recombination. Typically, the EQE spectrum is relatively constant (as a function of

light intensity) from 0.1-1 sun, although it sometimes decreases slightly at higher

illumination due to the effects of bimolecular recombination. Since the EQE is the

photoresponse of the solar cell at a given wavelength, the integral of the EQE with the

AM1.5G solar spectrum should yield the short-circuit photocurrent of the device.

EQE measurements can also be extended to measuring EQE as a function of voltage

bias.

Absorption measurements of a device, film or solution can be done with a

standard UV-Vis spectrometer. Such a measurement measures the transmission of

light through the device and can be sensitive to an optical density (OD) of 4-5 or even

higher. This makes UV-Vis the method of choice when measuring relatively

concentrated dye solutions that scatter little light. However, UV-Vis measurements

only lead to accurate results when the reflection off of the device surface can be

accurately calibrated using a reference and the device does not scatter. For scattering

samples such as ssDSCs, DSCs and other films containing TiO2 nanoparticles, a more

accurate way to measure the fraction of absorbed photons or absorptance, is using an

16

integrating sphere. The experimental setup of an absorptance measurement using an

integrating sphere is shown in Figure 2.3. Care must be taken to ensure that the direct

reflection off the solar cell doesn’t escape the integrating sphere but scatters off the

sphere surface; thus the device must be tilted at a slight angle.

Figure 2.3. Schematic diagram of absorptance measurements using an integrating

sphere.

Additionally, by putting the same film in front of the integrating sphere the

fraction of transmitted photons or transmittance, can be measured. This allows for a

calculation of the reflectance by taking 1-transmittance-absortance. One point that

should be made about EQE and absorptance measurements on ssDSCs is that changes

in the angle of incidence of light can lead to a sprectral shift of the interference fringes

in the absortance and EQE measurements. Thus when trying to ensure alignment of

EQE and absorptance interference fringes, it is important to keep the angle of

incidence the same in both measurements.

2.3 Photoluminescence Measurements

Photoluminescence (PL) measurements can probe the quenching of excited

dyes in solution. Steady-state PL measurements are performed by illuminating the dye

solution with light and measuring the amount of photoluminescence from the solution

as a function of wavelength. This is typically done with monochromatic illumination,

and case must be taken to ensure that the incident light intensity is the same between

samples to allow for comparison. Additionally, the concentration and extinction

coefficient of the dye in solution must be considered: if the fraction of incident light

17

absorbed by solutions is different, the photoluminescence must be normalized by that

amount. Furthermore, if the absorption spectrum of the dye overlaps significantly

with the emission spectrum, PL measurements may lead to incorrect results due to the

effects of re-absorption of emitted light. Steady-state PL measurements described in

this work were done with illumination using a Ar ion laser, with lenses used to focus

light onto a detector. In this setup, the position of photoluminescence can significantly

affect the amount of PL signal collected in the detector and care must be taken to

ensure that when comparing different dyes and dye concentration the absorption is

such that the PL comes from exactly the same position.

Time-resolved photoluminescence measurements are performed with a time-

correlated single-photon counting (TCSPC) setup that measures the time between light

excitation and photoluminescence emission. The amount of photoluminescence

should decay exponentially with time, and by fitting the logarithm of the counts vs.

time, the time constant can be calculated. Care must be taken to ensure that the rise

time of the laser pulse incident on the solution is significantly longer than the time-

constant of the photoluminescence decay for accurate measurement

2.4 Impedance Spectroscopy Measurements

Impedance spectroscopy is a powerful technique that can be used to probe

nearly all processes that occur in a DSC.47,59

An impedance spectrum at a given

voltage bias consists of measuring the device’s impedance (both real and imaginary)

as a function of frequency. From here, the device is modeled as a circuit where each

of the elements corresponds to a certain process in the device. By fitting the model,

the time constants and efficiency of each process can be backed out. Rather than

diving into the various models of DSCs and which process corresponds to which

circuit element, I will mention some of the main pitfalls in interpreting impedance

spectroscopy data.

At many biases, there are multiple processes occurring at the same frequency

which often obscure each other in the impedance spectrum. For example, in DSCs

biased near 0 V the recombination resistance is so large that typically no feature in the

18

impedance spectrum can be seen corresponding to the diffusion of the redox shuttle

through the electrolyte. Thus it is important to know which physical processes can be

resolved at what bias and at what frequency. Additionally, the equivalent circuit

models used to DSCs are different depending on the bias which is applied to the

device, thus judicious care must be taken to fit the correct model. Impedance

spectroscopy can also be used to understand the processes occurring in ssDSCs,

although the models and fits tend to be poorer than in liquid electrolyte DSCs. While

the technique is extremely powerful and can extract many parameters including

diffusion lengths, recombination lifetimes, charge-transfer resistances, and charge

transport lifetimes, fitting the data can be difficult and requires a high level of

understanding of impedance spectroscopy theory for all but the simplest cases. An

excellent introduction to impedance spectroscopy of DSCs is contained in Chapter 12

of Dye Sensitized Solar Cells,60

written by Bisquert and Fabregat-Santiago.

2.5 Layer Thickness Measurements

It is often important to be able to measure the thickness of each of the layers of

a DSC or ssDSC. Two techniques are particularly useful in this respect: profilometry

and cross-sectional SEM microscopy. For a single layer, on a hard surface,

profilometry tends to be the easiest way to get an accurate measurement. However,

profilometry measurements have issues when the measured material is particularly

soft – resulting in scratching by the profilometer tip, or when the thickness of multiple

layers needs to be made out. For these situations, cross-sectional SEM microscopy is

typically used. In this technique, a device or film is broken with care taken not to

damage the film edge. The sample is then mounted at a 90 degree angle and can be

visualized using SEM microscopy. Cross-sectional SEM microscropy also has

limitations: for materials with similar elemental compositions, such as 2 organic

layers, it is often impossible to get any contrast between such layers. Additionally,

insulating material can be difficult to visualize due to charging effects.

19

2.6 Other Measurements

A variety of other measurements were conducted in the process of collecting

data for this thesis, many of them typical characterization techniques available in

laboratories. For these techniques, I won’t go into the details as they are relatively

standard.

20

3 Internal Quantum Efficiency Measurements of ssDSCs

Solid-state dye-sensitized solar cells were invented as a stable alternative to

DSCs which often employed the use of a volatile, often corrosive electrolyte.61

However, by 2008, the record efficiency of DSCs was above 11%,7 while the ssDSC

record stood near 4%.51

Certainly, the smaller thickness of ssDSCs (2-μm-thick vs.

10-μm-thick DSCs) caused a portion of the problem, and recent work in the McGehee

group had shown that pore-filling of Spiro-OMeTAD into the TiO2 pores was not a a

huge issue until thickness greater than 4 μm.30

An important question at the time was

what were the most significant sources of loss in performance that limited ssDSC

efficiency to less than half of that of DSCs? An important quantity in understanding

the performance of solar cells is the external quantum efficiency (EQE), defined as the

ratio of charge carriers collected divided by number of incident photons (as a function

of wavelength). The EQE of liquid DSCs is typically 85% at peak absorption, as

approximately 15% of incident photons are reflected or absorbed within the F:SnO2

(FTO) electrode, and the remainder are absorbed in the 10-µm-thick cells and then

converted to collected charge with an efficiency of near 100%.62

In contrast, the peak

EQE of ssDSCs is frequently only 30% to 75%.34,53

Because ssDSC devices optimize

at a thickness of approximately 2 µm, this lower EQE can stem from incomplete light

harvesting, competitive absorption by non-photoactive layers or electronic losses, such

as charge recombination or inefficient charge injection.63,64

It is important to

disentangle these absorption losses in the various layers of the ssDSC with electronic

losses for a full understanding of device operation.

Internal quantum efficiency (IQE) is a useful device metric which measures a

solar cell’s ability to convert photons absorbed within the active material into

electrons, and allows diagnosis of charge collection and absorption problems.62

In

ssDSCs, the notion of IQE may be a bit ambiguous, as the active layer consists of a

mesoporous network of TiO2 nanoparticles, an adsorbed sensitizing dye and

infilitrated HTM. However, since absorbed light by Spiro-OMeTAD does not lead to

photocurrent as shown in Figure 3.1, a natural definition for IQE in ssDSCs would be

electrons collected in the device divided by the photons absorbed by the sensitizing

21

dye within the active layer. Although UV photons absorbed by titania can generate a

modest amount of photocurrent, because the titania does not significantly absorb in the

visible, this definition of IQE quantifies the ability of the dye to convert visible light

into collected charge carriers. IQE is an important parameter for such considerations

as diagnosing current losses, current matching tandem devices and is used in the

characterization of liquid DSCs and organic solar cells.65–67

Figure 3.1. EQE of ssDSC with no sensitizer. Photocurrent generated below 425 can

be attributed to TiO2 absorption. Despite the absorption of spiro-OMeTAD between

425 and 550 nm, there is almost no photoresponse in this portion of the spectrum.

Calculation of the IQE requires both measurement of the EQE and knowledge

of the percentage of light (ie. absorptance) that is absorbed by the photoactive

material. In liquid DSCs, the dye and titania absorptance can be relatively accurately

measured by comparing the transmission through a sensitized titania DSC electrode

with the transmission through an unsensitized substrate. While this method neglects

some reflection and scattering effects, such errors tend to be relatively negligible in

typical DSCs with 10-µm-thick, strongly absorbing active layers. In contrast, ssDSCs

utilize a silver back contact which does not allow for such a simple measurement of

the photoactive layer absorptance. Thus, optical modeling based on measured indices

of refraction has been used to calculate the absorptance of the photoactive layer for

IQE measurements.68

However, accurate optical modeling of ssDSCs is difficult due

22

to the large number of layers, uncertainty in the indices of refraction, interference

effects and scattering. Large errors in optical modeling (20% or more) are propagated

and result in similarly large errors in internal quantum efficiency. It should also be

noted that these same problems exist for IQE measurements of other devices with

similar structures, such as ssDSC analogs with inorganic absorbers or meso-

superstructured solar cells (MSSCs).37,38,69–71

In this chapter, we use a combination of measurements and optical modeling

calculations to accurately determine what fraction of incident light is absorbed by the

dye and what fraction is absorbed parasitically by materials that do not generate

photocurrent, such as FTO and the HTM. This information allows for accurate

measurements of the IQE, from which we elucidate valuable information about

electronic losses within the device. The internal quantum efficiency of two common

sensitizing dyes was investigated: Z907, a broadly absorbing Ru-based dye, and TT1,

a Zinc-pthalocyanine-based red-absorbing sensitizer (Figure 3.2). 51,72

Both Z907 and

TT1 have shown high performance in liquid electrolyte DSCs, but TT1 shows

relatively poor performance in ssDSCs, with efficiencies around 1%. Z907, on the

other hand, was selected as it has been widely studied as a sensitizer in ssDSCs and

achieves efficiencies of approximately 4%.51

Additionally, both TT1 and Z907 based

devices show excellent reproducibility.

Figure 3.2. a) Chemical Structure of Z907 dye. b) Chemical structure of TT1 dye.

23

3.1 Measurement of IQE

A ssDSC can be viewed as a stack of 6 layers: a glass substrate, F:SnO2 (FTO)

layer, TiO2 compact layer, an active layer, Spiro-OMeTAD overlayer, and silver

electrode (device architecture shown in Figure 3.3). In addition to the parasitic

absorption in the glass, FTO, compact TiO2, Spiro-OMeTAD overlayer and silver

back contact, there is additional parasitic absorption in the active layer itself by

infiltrated Spiro-OMeTAD and titania. When light is incident on the ssDSC, a

fraction is absorbed by the dye, with the rest being reflected or lost by absorption

within these non-photoactive materials (termed the parasitic absorptance). In order to

accurately measure the fraction of light absorbed by the dye, a hybrid modeling-

experimental approach was used that has previously been applied to thin film organic

solar cells.67

This technique relies primarily on measurements of the device

absorptance, using optical modeling as only a small correction. This approach is

particularly well-suited to ssDSCs due to the difficultly of performing accurate optical

simulations of ssDSCs. First, the reflectance, Rdevice, of the entire ssDSC device is

measured using an integrating sphere to account for diffuse reflection due to

scattering. Since the silver back contact allows no transmission, the measured

absorptance of the device is given by ABSmeasured, DSC = (1- Rdevice). The absorptance of

each layer of the solid-state dye sensitized solar cell is then modeled using methods

discussed in subsequent sections, and the modeled parasitic absorptance is summed to

give ABSmodeled,parasitic. While in our experience, typical optical modeling can have an

error of 20% or more, primarily due to errors in index of refraction values,

ABSmodeled,parasitic is generally only 25% or less of ABSmeasured, DSC, leading to only a

small error in the total measurement of photons absorbed by the dye, which is given

by ABSmeasured, DSC-ABSmodeled,parasitic (for example 20%×25% error=5% error). The

IQE can then be calculated as

. (3.1)

24

Figure 3.3. Schematic diagram of ssDSC layers.

Optical modeling uses the optical properties of materials to calculate the

reflection and absorption of incident light within the ssDSC device stack. A

convenient method of modeling layered materials is the transfer matrix approach,

which has been previously applied to ssDSCs and organic solar cells, and has

calculation code readily available online.67,68,73

The inputs for transfer matrix

modeling (and many optical modeling methods) are the complex indices of refraction

(as a function of wavelength) and thicknesses for each layer in the device. Layer

thicknesses can be measured through cross sectional SEM microscopy (sample image

shown in Figure 3.4). Indices of refraction of thin films are typically measured

through variable angle spectroscopic ellipsometry (VASE), and indices of refraction

for all ssDSC layers are contained in literature.68

The index of refraction can be

written as a real portion, n, and a complex portion, κ, with κ related to the thin film

absorption coefficient, α, by 4πκ/λ. While VASE is a difficult measurement that

requires complex fitting, the absorption coefficient, α, and therefore κ, can also be

measured for strongly absorbing layers by simple absorption measurements. By

measuring absorptance of a thin film of a given thickness, x, the absorption

coefficient, α, can be calculated using Beer’s Law: the intensity of transmitted light, I,

is given by I=I0e-αx

, where I0 is the incident light intensity. In order to get more

accurate values of κ, it is ideal to average the absorption coefficient over multiple thin

film thicknesses to minimize measurement errors as well as errors due to optical

25

interference from reflections. While this measurement neglects reflections and

scattering, these are typically small (<10%) for thin films of the various layers in

ssDSCs. It was found that these measured complex index of refraction values lead to

much better correlation between the modeled and measured absorptance of dye-

sensitized films and devices than κ values from VASE. Errors in the real portions of

the indices of refraction, n, can also cause modeling errors, as they determine

reflections at the interfaces between layers. For ssDSCs, the mismatches between n at

layer interfaces tend to be relatively small, leading to small reflections (with the

exception being reflection off the silver back contact). Hence, any inaccuracy in the

value of n tends to cause less total error in calculating IQE than inaccuracy in the

measurement of κ. In fact, the same spectrum for n can be used for modeling the

active layer independent of which sensitizing dye was used, resulting in a negligible

error. For our modeling, the indices of refraction were taken from literature, except for

the active mesoporous layer, FTO, and glass, when the strong absorption of the ssDSC

layer allowed for a direct measurement of the absorption coefficient.68

26

Figure 3.4. Representative cross-sectional SEM of ssDSC showing various device

layers. Layers visible (from top going downward): spiro-OMeTAD overlayer, active

layer, compact TiO2 layer, FTO. SEM images were analyzed with ImageJ software.

As mentioned previously, the mesoporous active layer contains 3 principal

components: dye, titania and HTM, and it is important to decouple the parasitic

absorption of the HTM from the absorptance of the dye. The index of refraction of the

active layer can be written as nactive+iκactive, where the κactive can be split into the sum of

κdye and κparasitic. κactive can be measured from the absorption of the dye-sensitized

active layer, while κparasitic can be measured from the absorption of an unsensitized

active layer. Given the absorptance in the active layer, ABSactive, the parasitic

absorptance within the active layer can be calculated as (κparasitic/ κactive) ABSactive.

Measured κ values of the active layer with no sensitizer and with both the TT1 and

Z907 dyes are depicted in Figure 3.5; while the dye, titania and Spiro-OMeTAD all

absorb very strongly below 425 nm, the absorption above 425 nm is dominated by the

dye and a small but significant contribution by oxidized Spiro-OMeTAD.33,74

Simply

looking at the relative magnitudes of the imaginary portion of the index of refraction

suggests the parasitic absorption within the active layer is a significant loss.

27

Figure 3.5. Imaginary portion of index of refraction of Z907-dyed active layer, TT1-

dyed active layer, and un-dyed active layer as measured by optical absorption. The

imaginary component of the index of refraction, κ, is related to the thin film

absorption coefficient, α, by α=4πκ/λ, where λ is the wavelength of light.

Given the indices of refraction and layer thicknesses for each layer in the

ssDSC, transfer matrix modeling is used to calculate the absorptance for each layer.

In layered thin film devices reflections off the interfaces between layers can cause

constructive or destructive interference, resulting in an absorptance that ‘oscillates’ as

a function of wavelength. Optical modeling using the transfer matrix method assumes

each layer is of uniform thickness, which leads to very strong optical interference

fringes (Figure 3.6). The interference fringes in the modeled device absorptance are

much larger than those in the measured device absorptance. This can occur for a

variety of reasons: the device thickness varies throughout the illuminated area, there is

light scattering in the film, or the monochromator bandwidth is large during the EQE

measurement. The largest effect seems to be that the thickness of the ssDSC layers

(particularly the active layer) varies throughout the device, which partially averages

out the interference fringes expected by transfer matrix modeling. To account for this,

the absorptance resulting from transfer matrix modeling was averaged by varying the

active layer from +/ 5% of the measured thickness in 0.5% increments. Because of the

change in active layer thickness, the overlayer thickness was also changed by -0.3

times the change in active layer thickness (this is to account for additional Spiro-

28

OMeTAD, with an assumed 60% pore filling30

and 50% porosity of the active layer).

This overlayer compensation is under the assumption that the total amount of Spiro-

OMeTAD remains constant through the film. However, this compensation turns out

to be nearly negligible, as the absorption of the Spiro-OMeTAD overlayer is almost

negligible, while absorption of Spiro-OMeTAD within the active layer is fairly

significant. This can be explained by the fact that much of the Spiro-OMeTAD

oxidation occurs at the spiro-TiO2 interface, which cannot happen within the

overlayer.33

Additionally, a significant amount of light is absorbed in the active layer

before ever reaching the spiro-OMeTAD overlayer.

Figure 3.6. Comparison of measured and modeled device absorptance using transfer

matrix modeling with no averaging. Total modeled parasitic absorptance is shown as a

dashed black line and the dye absorptance, or ABSmeasured, DSC - ABSmodeled,parasitic is

depicted by the dashed gray line.

Modeled absorptance of each component of a Z907-sensitized DSC is shown

in Figure 3.7a and the total modeled device absorptance is compared to measured

ssDSC absorptance in Figure 3.7b. As can be seen, our modeling approach (utilizing

averaging) resulted in a smoother absorptance that matched experiment much better

than without averaging (Figure 3.6).

29

Figure 3.7. a) Modeled absorptance for each layer of a 2.3-µm-thick Z907 ssDSC:

total device absorptance, active layer absorptance, FTO absorptance, parasitic

absorptance within the active layer, glass absorptance, and sum of the absorptances of

other layers (this corresponds to the TiO2 compact layer, Spiro-OMeTAD overlayer

and silver cathode). b) Comparison of modeled and measured device absorptance.

Total modeled parasitic absorptance is shown as a dashed black line and the dye

absorptance, or ABSmeasured, DSC-ABSmodeled,parasitic is depicted by the dashed

gray line. This modeling is done using the averaging scheme described in the text.

In principle, a variety of optical simulation techniques actually results in

similar IQE values if accurate optical parameters are used in conjunction with the

hybrid measurement/modeling approach. The only issue is that poor modeling leads

to misalignment of the interference fringes between the EQE and dye absorptance,

30

which in turn leads to a more jagged IQE (as is discussed in the next section).

However, if calculating an average IQE over a certain wavelength regime, we will see

in section 3.2 that modeling without this averaging scheme also leads to relatively

accurate results. Indeed, using incoherent ray tracing to model the device (not shown)

also resulted in IQE values fairly close to those obtained when using transfer matrix

modeling or averaged transfer matrix modeling. This exemplifies the strength of this

hybrid approach to measuring IQE: errors in modeling do not substantially affect the

calculated IQE as the majority of the photoactive layer absorptance comes from

directly measuring the reflectance of the ssDSC device.

3.2 IQE for Z907 and TT1 Dyes

Once the absorptance of all the parasitic layers is calculated, it is

straightforward to apply Equation 3.1 to calculate the IQE. The dye absorptance, EQE

and calculated IQE for a Z907 ssDSC and a TT1 ssDSC are shown in Figure 3.8. Due

to uncertainty in the material optical parameters and layer thicknesses, the interference

fringes in the calculated absorption and measured EQE do not occur at the same

wavelengths. Consequently, the calculated IQE has a small oscillating component,

particularly for Z907 devices, as depicted in Figure 3.8a. In principle, the IQE of a

ssDSC should be constant with respect to wavelength, and if the modeling was perfect,

the absorptance of the dye should be proportional to the EQE times a constant value.

Using this hybrid approach allows for minimizing any errors such as interference peak

mismatches between the modeling and measurement to achieve as flat an IQE as

possible. In order to get a good measure of the exact IQE of each device, the

calculated IQE was averaged in the wavelength regime where the dye absorption was

highest (440-560 nm for Z907 dye and 620-705 nm for TT1 dye). Because the dye

absorbs the majority of the photons in these regimes, this allows for a minimization of

the errors caused by the difficulty of accurate optical modeling. To quantify the error

in the measurement, the modeled parasitic absorptance was changed by +/- 20% and

the resulting IQE is shown by the dotted lines (Figure 3.8). In the regime where the

dye absorbs strongly, errors in modeling lead to errors of 10% or less in IQE. As

31

shown in Figure 3.8b, for TT1-based ssDSCs, the IQE is almost completely flat in the

red portion of the spectrum where the dye absorption is strongest.

Figure 3.8. Dye absorptance as given by ABSmeasured, DSC - ABSmodeled,parasitic, measured

EQE, and calculated IQE plots for (a) 2.3-µm-thick Z907 device and (b) 2.2-µm-thick

TT1 device. Dotted gray lines denote error bars in IQE measurement based on a

±20% error in modeling parasitic absorptance. Vertical black lines depict averaging

range for calculating a single IQE value for each device.

At this point I would like to take a little time to compare the IQE results of our

approach with other approaches for modeling of ssDSCs and measurements of ssDSC

IQEs. As done in Figure 3.6, the absorptance of the device can be modeled without

averaging over thickness and then used to calculate the IQE. A comparison the IQE

32

for a 2.3-µm-thick Z907 device is shown using the averaged and not-averaged transfer

matrix modeling is shown in

Figure 3.9. As can be seen, the IQE when not utilizing averaging is extremely

jagged between 600-700nm due to misalignment of interference fringes between the

modeled dye absorptance and measured EQE. However, the IQE in the regime where

the due absorption is strongest (440-560 nm for Z907 dye) is about the same for both

modeling approaches. Indeed, averaging over this regime, the IQE for the thickness

averaged method is 0.97 versus 0.98 with no averaging. This is the main advantages

of the hybrid approach – errors in modeling do not affect the overall IQE very

significantly. The robustness of this hybrid modeling approach is essential for

modeling ssDSCs where uncertainties in optical parameters and layer thicknesses

make optical modeling challenging.

33

Figure 3.9. Dye absorptance as given by ABSmeasured, DSC - ABSmodeled,parasitic, measured

EQE, and calculated IQE plots for 2.3-µm-thick Z907 device when the parasitic

absorptance is modeled using a) no averaging, b) the averaging scheme described in

the text (this is the same as Figure 3.8a). Dotted gray lines denote error bars in IQE

measurement based on a ±20% error in modeling parasitic absorptance. Vertical black

lines depict averaging range for calculating a single IQE value for each device.

Additionally, we can compare this hybrid experimental and modeling approach

to another approach which relies on only using optical modeling to calculate the dye

absorptance.68

Shown below in Figure 3.10a is the modeled dye absorptance using the

transfer matrix modeling approach with no averaging. Dividing the EQE by this

calculated dye absorptance yields the very jagged IQE shown in Figure 3.10b.

However, when averaging the IQE between 440 and 560 nm, the average IQE is

34

102%, not too far from the value measured with the hybrid method. The particular

reason for this is the accuracy of our modeling: using absorption measurements to

calculate the imaginary portion of the index of refraction results in very good

agreement between modeled device absorptance and measured device absorptance (as

was shown in Figure 3.7b). In our experience, oftentimes modeling and measurement

do not agree so well, making results obtained through this methodology much worse,

while affecting IQE results from the hybrid measurement/modeling method

significantly less. Comparing the shape of the IQE from Figure 3.10b to the IQE in

Figure 3.8a shows there is still a big advantage to using the hybrid approach.

Figure 3.10. a) Modeled ssDSC absorptance using no averaging for a 2.3-µm-thick

Z907 device: total device absorptance, dye absorptance, parasitic absorptance. b) IQE

(green) as measured by dividing EQE by this modeled dye absorptance (black).

Our measurement of the internal quantum efficiency was used to measure the

IQE of ssDSCs over a variety of thicknesses. Figure 3.11 shows the IQE of Z907 and

TT1 ssDSCs for active layer thicknesses between 1 and 4 microns. In this thickness

range, the pore-filling fraction remains high enough to not limit performance, and due

to the large diffusion length at short circuit (significantly longer than the film

thickness), the IQE of both the Z907 and TT1 devices stays constant.29–31

Despite the

increase in thickness, the constant IQE with thickness suggests that charge collection

is not a problem in these devices even at thicknesses that are larger than the ideal

thickness of approximately 2 microns. The losses in efficiency, rather, come from the

decreasing voltage with thickness and decrease in fill factor.75

However, while the

35

IQE of Z907 stays high at 88%, the average IQE of the TT1 cells is only

approximately 36%, suggesting there is a significant electron or hole injection

problem with the TT1 dye into the TiO2 or Spiro-OMeTAD, respectively, leading to a

near 60% loss. While TT1 has near 100% IQE in liquid DSCs, it displays

significantly lower IQE in solid state devices. The electron injection of TT1 in liquid

DSCs is near unity, but it should be noted that the TiO2 conduction band energy levels

can be very different in solid and liquid devices due to the presence of additives such

as tert-butylpyridine and Li+ - leading to very different injection efficiencies.

14 It has

been seen that the lowest unoccuppied molecular orbital, or LUMO, of pthalocyanine

dyes is lower than that of N719 (a dye similar to Z907), which can lead to poor

electron injection efficiencies in liquid DSCs under certain electrolyte compositions.16

Thus, it is hypothesized that the TiO2 conduction band energy in solid-state devices is

at a level which allows for efficient electron injection from the Z907 dye to the titania,

but is too high for efficient TT1 electron injection. However, there are additional

factors such as dye aggregation which can affect dye electron and hole injection

efficiencies.

36

Figure 3.11. IQE vs. active layer thickness for Z907 (black squares) and TT1 (gray

circles) ssDSCs. Error bars are calculated from the standard deviation of the IQE in

the measurement range summed in quadrature with the average error in IQE caused by

increasing/decreasing the parasitic absorption by 20% (depicted by gray dotted lines

previously). This error metric takes into account uncertainty caused by large amounts

of parasitic absorption and not-flat IQEs caused by inaccuracies in modeling and/or

measurement. Error bars are larger for thinner devices due to the relatively low dye

absorptance in the ssDSC. The light gray dashed lines depict the average IQE for

Z907 and TT1 devices.

3.3 Effect of Coadsorbent on IQE

Many pthalocyanine dyes such as TT1 exhibit a tendency to aggregate due to

π-π interactions between macrocycles.76

A variety of strategies have been used to

suppress aggregation, such as the addition of bulky substituents and the addition of

coadsorbents.76,77

It has been reported that the addition of coadsorbents such as

chenodeoxycholic acid (cheno) can improve the efficiency of liquid DSCs through a

variety of ways, in particular, increasing charge injection efficiency by decreasing

aggregation.44,76,78

Coadsorbents, however, displace dye molecules on the titania

surface and consequently decrease dye loading. While the reduction in dye loading

does not severely decrease light absorption in 10-µm-thick liquid cells, it is

problematic in thinner ssDSCs, where there is a need to maximize dye coverage within

the limited thickness of the mesoporous titania layer. Coadsorbents have been

37

investigated in ssDSCs as a means to improving efficiency,79–81

but the effect of

coadsorbents on IQE has not been studied.

Figure 3.12. a) Comparison of modeled and measured absorptance for high cheno (60

mM) TT1-sensitized device with active layer thickness of 3.7 µm. Total modeled

parasitic absorptance is shown as a dashed black line and ABSmeasured, DSC-

ABSmodeled,parasitic is depicted by the dashed gray line. The solid and dashed black lines

overlap where all the absorptance is due to non-photoactive materials (400-575 nm).

b) IQE, EQE and ABSmeasured, DSC-ABSmodeled,parasitic for same device. Dotted gray lines

denote error bars in IQE measurement based on a 20% error in modeling parasitic

absorptance. Vertical black lines depict the averaging range used for calculating a

single IQE value for each device.

38

To investigate the effect of the suppression of aggregation on TT1 ssDSCs, the

cheno concentration was raised from 10 mM to 60 mM during dye sensitization,

which has been shown to significantly decrease aggregation.76

The additional cheno

adsorption results in significantly lowered dye loading and absorption, necessitating

the use of device active layers between 3.5-4 µm to ensure adequate dye absorption

for accurate IQE quantification. Device optical modeling and absorptance is shown in

Figure 3.12a. Despite the low absorption of the active layer, the absorptance of the

device at the dye’s peak wavelength of 690 nm is still 0.74. However, from optical

modeling, it can be seen that nearly half of this absorptance can be attributed to the

parasitic layers within the device. As shown in Figure 3.12b, this results in an average

IQE of 58% for the high cheno devices. Since cheno has been shown to raise the

conduction band of TiO2 (ie. lower electron affinity), 82

the increased internal

quantum efficiency is attributed to an improved injection efficiency caused by

suppression of dye aggregation rather than a shift in the TiO2 conduction band. While

an increase of 20% in IQE is significant, and deaggregation is an important

consideration for the improvement of dyes in DSCs, the IQE remains well below

100%, indicating that there are additional loss mechanisms such insufficient driving

force for charge injection from the excited dye into the TiO2 conduction band.

3.4 Quantification of Parasitic Absorption Losses From Modeling

Finally, the high IQE of Z907 also warrants discussion, as the peak EQE of the

same devices remains 60% or less, even for thicker films. Compared to many D-π-A

dyes being developed for solid state dye sensitized solar cells, Z907 is a relatively

weak absorber. However, both modeling and measurements suggest that the cell is

able to absorb nearly all of the light (85- 90%) at the dye’s absorption peak in

optimized 2-µm-thick devices. The low EQE is caused by the significant parasitic

absorption of Spiro-OMeTAD in the 450-550nm wavelength range, which can be

attributed to the oxidized form of Spiro-OMeTAD. 33

While undoped Spiro-

OMeTAD does not appreciably absorb visible light, oxidized Spiro-OMeTAD is

necessary to achieve sufficient conductivity to make efficient devices and minimize

39

series resistance losses.25,33

Although more strongly absorbing dyes are able to

outcompete Spiro-OMeTAD for light absorption in this wavelength regime, Z907-

based ssDSCs lose a very significant amount of light to the parasitic absorption in the

active layer itself. These losses in photocurrent due to parasitic absorption within the

active layer were quantified through optical modeling in order to understand the

potential for increasing device efficiency.

The optical modeling described in Section 3.1 allows for estimation of the

parasitic absorption losses in the active layer. In order to get an estimate of the

additional photocurrent generated in a device with no spiro-OMeTAD absorption, the

absorption coefficient (imaginary portion of the index of refraction) of the parasitic

absorption in the active layer was subtracted off of the total absorption coefficient of

the active layer. The device absorptance was then modeled to see the additional

absorptance in the active layer. This was done for Z907 devices, and assuming an IQE

of 90%, the modeled short-circuit photocurrent (JSC) of a 2-µm-thick ssDSC increases

nearly 14% from 7.32 to 8.49 mA/cm2, suggesting that a non-absorbing HTM can lead

to a significant increase in device performance. This loss turns out to be the dominant

parasitic absorption loss in the device, and is significantly more than losses incurred

do due to the parastic absorption of other layers such as FTO (shown in Table 3.1).

Table 3.1. Modeled Jsc of a 2-µm-thick Z907 ssDSC if various layers had no parasitic

absorption. Reference has no layers set to 0 absorption.

Layer JSC [mA/cm2] JSC Lost [mA/cm

2] % Loss

Reference 7.32 0 0

FTO 7.87 0.55 7.0

Active 8.49 1.17 13.8

Compact TiO2 7.46 0.16 1.9

Spiro-OMeTAD

Overlayer

7.36 0.04 0.5

40

This procedure can also be done using a hypothetical more strongly absorbing

dye. The absorption coefficient of a Z907-sensitized active layer was multiplied by a

factor of 10 and the modeling from the previous paragraph was repeated. Now the

reference device had an improved JSC of 15.98 mA/cm2 (primarily due to much

stronger absorption in the red). Upon removing the parasitic absorption of the active

layer, this number improves to 16.43 mA/cm2. This improvement is significantly

smaller than previously (only 2.8%), as the dye absorption is much stronger relative to

the parasitic absorption in the active layer. Thus an active layer absorption 10 times

that of Z907-sensitized ssDSCs seems like reasonable target (corresponding to an

absorption coefficient of 5-6 µm-1

) for ensuring losses due absorption by oxidized

Spiro OMeTAD are reasonably small.

These results explain why increasing the thickness of Z907 devices does not

bring the EQE above 60%: despite the short circuit charge collection efficiency

remaining constant, there is very little unabsorbed photon flux. Thus, techniques to

increase light harvesting such as light trapping83

or increasing device thickness would

have only a minimal effect below 550 nm wavelengths. Furthermore, any additional

absorption would be split between the parasitic absorption of Spiro-OMeTAD in the

active layer and the dye, leading to even less additional photocurrent. On the other

hand, D-π-A dyes have seen great success in DSCs with peak EQE’s approaching

80%, which can be understood through the high absorption coefficient allowing the

dye to outcompete Spiro-OMeTAD for absorption below 550 nm.34,53

Spiro-OMeTAD

has also been used as the hole transport material in MSSCs utilizing an inorganic

perovskite absorbing layer.37,38,69

The absorption coefficient reported for such devices

appears to be 5-10 times stronger than that of Z907-based DSCs and significantly

reduces parasitic absorption in the active layer.

3.5 Conclusion

While internal quantum efficiency is a particularly useful metric for the

analysis of photovoltaics, the difficulty in either modeling or measuring the

absorptance of the dye in a ssDSC device has made IQE a difficult quantity to

41

accurately measure in ssDSCs. Hence, a hybrid approach utilizing optical modeling

and absorption measurements approach is necessary for an accurate quantification of

the internal quantum efficiency of ssDSCs due to the multitude of materials and layers

in the device. This approach has been used to measure the IQE of ssDSCs using 2

sensitizers: TT1 and Z907, and elucidate interesting facts about their photovoltaic

performances. The IQE of Z907-based ssDSCs is calculated to be approximately

90%, suggesting that nearly all charge carriers generated in by the sensitizing dye

itself make it to the electrodes at short circuit. On the other hand, TT1-based ssDSCs

display significantly lower IQE’s despite having near unity IQE in typical liquid

electrolyte devices, due to low charge injection efficiency. Coadsorbents such as

chenodeoxycholic acid can be used to increase injection efficiency by 20% by

decreasing aggregation, but at the cost of dye absorptance due to competition for

adsorption onto the mesoporous titania. Finally, the parasitic absorption in the active

layer was found to actually be the largest optical loss in Z907-sensitized solar cells.

Increasing the absorption within the active layer is an important goal for increasing the

efficiency of solid-state dye sensitized solar cells, which are to be limited to 2 microns

due to charge transport. However, increasing device thickness and light trapping also

lead to increased Spiro-OMeTAD absorption and significant optical losses. Even

though the best performing dyes are able to convert photons to electrons at near-unity

yields, research into new less absorbing hole transport materials and strongly

absorbing dyes will be required to mitigate parasitic absorption losses and help push

ssDSCs to new record efficiencies. This understanding has helped direct some of the

other research that had been going on in our group, particularly our research into D-π-

A dyes for ssDSCs. Our best ssDSC efficiency is 6.3% achieved with a strongly

absorbing D-π-A dye coded WN3.1.35

Indeed, the current record-holding dye is

another similar strongly absorbing D-π-A dye, Y123.25

3.6 Experimental Details

Device Fabrication: TiO2 substrates were fabricated and sensitized with dye as

previously reported. 30

FTO substrates (TEC15, Hartford Glass Co.) were cleaned by

42

sonicating in detergent, acetone and isopropanol, with subsequent UV-ozone treatment

for 20 minutes. Approximately 50-100 nm of compact TiO2 was deposited using

spray pyrolysis of titanium diisopropoxide bis(acetylacetonate) (Aldrich 75 weight %

in isopropanol, diluted 10x with isopropanol). Films of varying titania thickness were

doctorbladed by using dyesol paste (NR-18T) diluted with terpinoel resulting in

nanoparticle films of thicknesses between 1 and 4 µm. Films were then sintered at

500° C for 30 minutes. Subsequently, titania films were then immersed in TiCl4

solution overnight and sintered once again at 500° C for 30 minutes. Titania

substrates were then sensitized by immersion for 18 hours in a 0.3 mM solution of

Z907 dye (Solaronix) in a 50:50 tert-butanol:acetonitrile or immersion for 4 hours in a

0.1 mM solution of TT1 dye in ethanol with 10 mM chenodeoxycholic acid.

Sprio-OMeTAD solution contained Spiro-OMeTAD (Luminescence

Technology corporation), tert-butylpyridine (4-tbp) and Lithium

bis(trifluoromethylsulfonyl)imide salt (Li-TFSI) (pre-solved in acetonitrile). Li-TSFI

solutions was first made by dissolving 170 mg/ml Li-TSFI in acetonitrile. Spiro

solution was made by taking a 1g Spiro-OMeTAD: 97 mL 4-tbp: 208 mL Li-TSFI

solution mixture dissolved in chlorobenzene (approximately 100-400 mg Spiro-

OMeTAD/mL chlorobenzene). The concentration of Spiro-OMeTAD was varied to

ensure adequate pore filling - which depends on the thickness of the TiO2 nanoparticle

film. The Spiro-OMeTAD solution was then infiltrated by spincoating as previously

reported, with increased Spiro-OMeTAD concentration to ensure a small

(approximately 200 nm) but visible overlayer and maximal pore-filling29,30

. Finally, a

200nm silver cathode was deposited by thermal evaporation at a pressure of

approximately 10-6

torr. All films and devices were subject to 15 minutes light

soaking before measurement. Films used for measuring κ values were made on plain

glass substrates with the film deposited with the same method as in actual ssDSC

device fabrication.

EQE and Absorption Measurements: External quantum efficiency

measurements were performed at a chopping rate of 40 Hz with a white light

illumination bias of approximately 0.1 suns applied using an incandescent bulb

43

powered by a DC voltage source. For the chopped EQE beam, a Newport Apex

monochromator illuminator was used in conjunction with a Princeton Instruments

monochromator and a filter wheel. The signal from the DSC was put through a

transimpedance amplifier and recorded on a Stanford Instruments lock-in amplifier.

The EQE Calibration was performed using a calibrated photodiode of known EQE.

The EQE beam was split with a 50:50 beam splitter into a 2nd

‘reference’ photodiode

that was used to correct for any fluctuations in the EQE beam source intensity.

Experimental set-up for absorptance measurements have also been reported.67

Device absorptance measurements were performed using the same light

source/monochromator as the EQE and measured using an integrating sphere with an

attached silicon photodiode. Care must be taken to ensure that the incidence angle of

the light should be as close to normal as possible, otherwise absorption and EQE

measurements can display misaligned interference effects.

Layer Thickness measurements: Device layer thicknesses were measured with

cross-section scanning electron microscopy using a FEI XL30 Sirion SEM and image

processing software (ImageJ).67

44

4 Highly Soluble Energy Relay Dyes

Although DSCs achieved efficiencies of 10%8 in only 2 years after invention

in 1991, 1 by 2009, record efficiencies were still under 12%.

84 While many Ru-based

dyes showed relatively weak but broad absorption between 400-800 nm, most other

dyes used in DSCs such as porphyrins and pthalocyanines10,72

had strong absorption

peaks but were unable to harvest the entire solar spectrum. It was clear that for these

dyes to be effective in a 12% or higher efficiency DSC, there needed to be another

complementary absorber that can harvest the remaining portion of the solar spectrum.

A variety of multiple dye solutions has been utilized to achieve strong and broad

absorption, including cosensitization,85,86

cosensitized energy transfer,87–89

dyadic

sensitizers,90,91

and energy relay dyes (ERDs).55,92–95

While my work has been

exclusively on the last method, ERDs, I will take a second to go through each of the

various methods.

Cosensitization is the process of sensitizing the TiO2 surface with two or more

dyes. While theoretically simple, cosensitization requires optimization of sensitizer

concentration in solution and TiO2 sensitizing time. 96

Additionally, having 2 dyes on

the surface can lead to energetic traps for electrons or holes, increased recombination

and unfavorable dipoles at the TiO2 interface. Another cosensitization method is

attaching a second dye to the surface that can either energy transfer or charge transfer

to the first sensitizer rather than injection charge itself.87–89

Finally, there have been

efforts in making 2-in-1 dyes that contain two linked chromophores called ‘dyadic

sensitizers.’ The chromophore that is attached closer to the TiO2 will inject directly

into the TiO2, while the more distant sensitizer will either transfer energy or charge to

the closer sensitizer. Some of the difficulties associated with the use of dyadic

sensitizers are synthesizing such complex molecules, sensitizing these large molecules

on the TiO2 surface and issues with the dyadic sensitizers packing well on the TiO2

surface.

As mentioned previously, in addition to functioning as the primary absorber in

the DSC, the sensitizing dye has a myriad of other responsibilities, including rapid

electron injection14,16

, efficient hole regeneration20,45

, and acting as an effective barrier

45

to recombination.97–99

The idea of the ERD is the split the functions of absorption with

those of charge injection and blocking recombination between the ERD and sensitizer.

ERDs are dissolved dyes within the electrolyte that, upon excitation, undergo Förster

resonance energy transfer (FRET) to the sensitizing dye (Figure 4.1).55,100

Hence,

using one or more ERDs to cover the solar spectrum in conjunction with an efficient,

highly performing sensitizing dye can be a strategy toward highly efficient DSCs.101

Additionally in contrast to cosensitization, there is little effort required to optimize a

DSC utilizing an ERD, as the ERD is simply added to the electrolyte with no other

changes in the device fabrication.

At this point, I would like to give a little background on the mechanism by

which energy is transferred between the ERD and sensitizer: Förster resonance energy

transfer or FRET.102

FRET is a dipole-dipole interaction between two chromophores

that can transfer energy over short distances (typically 1-5 nm). The rate of FRET is

determined by the ‘Förster radius’ or ‘FRET radius,’ R0, a characteristic length scale

given by,

, (4.1)

where QD is the photoluminescence (PL) quantum yield of the donor (ERD), κ2 is a

geometric term involving the orientation of dipoles (2/3 for randomly oriented dipoles

in the ERD case), n is the index of refraction of the medium, NA is Avogadro’s

number, fD(λ) is the normalized emission of the donor (ERD), and εA(λ) is the molar

extinction coefficient of the acceptor (sensitizer). From here, the rate of FRET can be

calculated as

, (4.2)

where τPL is the photoluminescence lifetime of the donor (ERD) and r is the distance

separating the two chromophores. From the rate of FRET, the efficiency of FRET can

be calculated by considering the rates of alternative pathways of ERD relaxation.

Additionally, in the case of the DSC, we are interested in the FRET from an excited

ERD to a multitude of sensitizers on the surface of TiO2 inside of a nanopore. Once

again, the FRET rate (Equation 4.2) can be used to calculate the FRET rate to each

46

sensitizer (based on the geometry of the pore) and calculate the energy transfer

efficiency (ETE) by comparing to the rates of other competing processes.103

Because

of the many chromophores on the surface of the pore, FRET in nanopores can occur

efficiently over much larger distances (10-20 nm) than between a single ERD and

sensitizer.

Previous work has shown that ERDs can transfer energy to the sensitizing dye

with near 100% efficiency.104

However, the poor solubility of ERDs has limited

performance due to the inability of the ERD to absorb most of the incident photons:

the previous largest enhancement due to an ERD in a DSC has been 28.5% due to low

EQE in the portion of the spectrum corresponding to ERD absorption. As the

thickness and porosity of typical nanoporous electrodes used in DSCs is

approximately 6 µm and 0.5, respectively, ERDs must be able to achieve an optical

density of at least 1 in an effective thickness of only 3 µm.

Figure 4.1. Schematic diagram of sensitized TiO2 nanoparticles in a DSC utilizing an

ERD.

47

While ERDs no longer perform the functions of efficiently injecting charge

and blocking recombination, they carry with them another set of design rules.

Effective long range FRET from ERD to sensitizing dye requires a large Förster radius

and excellent photoluminescence efficiency.103

In addition, the various ions in the

redox electrolyte can cause photoluminescence quenching of the excited ERD leading

to lost energy. Finally, the ERD must be extremely soluble in one of the various

solvents used in the electrolyte of DSCs.

4.1 Dye Structure, Characterization and Förster Radius Calculations

Two dyes were designed for use as ERDs in DSCs, coded BL302 and BL315,

with chemical structures depicted in Figure 4.2. The synthetic schemes for each dye is

given at the end of this chapter in Section 4.9. Both BL302 and BL305 are similar in

structure to the common laser dye DCM (4-(dicyanomethylene)-2-methyl-6-(p-

dimethylaminostyryl)-4H-pyran, also depicted in Figure 4.2c for comparison), which

has shown efficient energy transfer in DSCs.104

Through the use of additional alkyl

and alkoxy groups, both dyes display excellent solubilities of approximately 180 mM

in benzonitrile. BL302 has an absorption and photoluminescence spectrum nearly

identical to that of DCM, while BL315 shows a slightly redshifted absorption and

emission spectrum due to the insertion of a thiophene moiety into the dye (Figure 4.3).

48

Figure 4.2. Chemical structure of dyes. Energy relay dyes: (a) BL315, (b) BL302, (c)

DCM. Sensitizing dye: (d) TT1.

Figure 4.3. Absorption and emission spectra of DCM (black), BL302 (red), and

BL315 (blue) dyes in CH2Cl2 (10-5

M concentration).

49

Figure 4.4. Relative PL emission of 1 mM DCM in acetonitrile and benzonitrile and 1

mM BL302 in benzonitrile excited at 514 nm by an argon ion laser.

Both dyes show good complementary absorption with TT1 (Figure 4.3), a

strongly absorbing, zinc-pthalocyanine-based sensitizer (TT1 chemical structure

shown in Figure 4.2d).72,76

The FRET radius between each of these ERDs and TT1 can

be calculated by using Equation 4.1. It is known that DCM has a PL efficiency of

0.43 in methanol, 105

and it can be estimated that the PL quantum yield is similar in

acetonitrile. A comparison of DCM photoluminescence with that of BL302 (Figure

4.4) shows that while the spectrum looks relatively similar, the amount of

photoluminescence increases – resulting a PL efficiency estimate of 0.49 for BL302.

The emission spectrum of BL315 in benzonitrile is shown in Figure 4.5a, and by

integrating the emission of each and comparing the number of emitted photons, the

estimated photoluminescence quantum yield for BL315 is measured to be 0.28.

Clearly, there are a variety of errors in our measurement of the PL efficiency for these

dyes – the PL efficiency was measured by comparison to the PL efficiency of DCM,

and was done without a fluorometer typically used to get accurate PL efficiency

numbers. Still, these numbers shown be enough to get reasonable ballpark values for

the Förster radius. Given the PL efficiencies of our ERDs, the index of refraction of

benzonitrile of 1.5 and the extinction coefficient of TT1 shown in Figure 4.5b, the

FRET radius can be calculated for BL302 and BL315 to TT1, resulting in values of

5.84 nm and 4.92 nm, respectively. Once again, these values are only estimates, as the

50

true PL efficiency is difficult to measure, but an error of even 100% in some of the

factors (such as PL efficiency) doesn’t greatly alter the FRET radius, which goes as

QD1/6

. Given the various errors in the measurement, we only cite FRET of BL302 and

BL315 with TT1 to 1 significant digit – 5 nm and 6 nm respectively. Previously

performed calculations show that these FRET radii for BL302 and BL315 should

allow for high (greater than 95%) energy transfer in the 17-20 nm pores typically used

in DSCs.103

Figure 4.5. a) Relative PL emission of 1 mM BL315 and 1 mM BL302 in benzonitrile

excited at 514 nm by an Ar laser. b) Normalized absorption of TT1. Molar extinction

coefficient of TT1 at peak is 191500 M-1

cm-1

.

4.2 ERD DSC Characterization

TT1-sensitized DSCs were fabricated as previously reported, using varying

concentrations of BL302 or BL315 dissolved in a benzonitrile-based electrolyte. The

51

full device fabrication procedure and exact electrolyte composition is contained in the

Section 4.10. Device J-V characteristics are shown in Table 4.1 and J-V curves are

given in Figure 4.6.

Table 4.1. J-V characteristics for TT1 devices incorporating BL302 and BL315 as an

ERD. J-V curves are shown in Figure 4.6.

Device Jsc (mA/cm2) Voc (mV) FF Efficiency (%)

Reference 6.0 623 0.67 2.51

20 mM BL302 7.1 618 0.68 2.99

60 mM BL302 8.7 619 0.64 3.47

180 mM BL302 9.7 640 0.62 3.80

20 mM BL315 7.3 633 0.67 3.05

60 mM BL315 9.4 633 0.62 3.69

180 mM BL315 10.8 640 0.60 4.14

Figure 4.6. J-V curves of TT1-sensitized DSCs with various concentrations of a)

BL302 and b) BL315 compared to a reference TT1 sensitized device utilizing a

benzonitrile-based electrolyte.

As shown in Table 4.1, BL302 and BL315 both significantly improve the

efficiency of the reference device, particularly due to a strong increase in the short-

52

circuit photocurrent density (Jsc). Overall, BL315 results in a 65% increase in

efficiency of the TT1 DSC, while BL302 increases device performance by 51%.

While the fill factor (FF) does decrease slightly, this is not necessarily caused by the

ERD, as increasing the photocurrent of a solar cell typically lowers the fill factor.

Impedance spectroscopy was performed to investigate if the lowered fill factor was

caused by slower diffusion of the redox shuttle due to high ERD concentrations.

Impedance spectroscopy has been widely used to investigate the various charge-

transport and charge-transfer processes which take place in DSCs, including mass

transport of the electrolyte.47,59,106

In order to clearly see the part of the impedance

spectrum that corresponds to mass transport of the redox shuttle within the electrolyte,

the DSC impedance spectrum needs to be taken at far forward bias. Shown below in

Figure 4.7 are the Nyquist plots of TT1-sensitized DSCs with and without 180 mM

BL302, and their respective impedance fits. The equivalent circuit used for fitting is

shown in Figure 4.8 – such an equivalent circuit is typically used for fitting impedance

spectra of DSCs in far forward bias.

Figure 4.7. Nyquist plots (taken at 0.8 V forward bias) and impedance fits (using the

circuit shown in Figure 4.8) for TT1 sensitized DSCs employing a benzonitrile

electrolyte a) without any ERD and b) with 180 mM BL302.

Figure 4.8. Equivalent circuit modeling the operation of a DSC in far forward bias.

53

The element in the equivalent circuit (Figure 4.8) describing the diffusion of

the redox shuttle through the electrolyte is the warburg element, whose impedance is

given by

, (4.3)

where ω is the angular frequency and R and td are fit parameters. The best fit values of

(R, td) to Equation 4.3 were found to be (12.8 Ohms, 0.21 s) and (15 Ohms, 0.25 s) for

the no ERD and 180mM BL302 devices, respectively. As these values are very close,

it can be concluded that the addition of large amounts of ERD does not significantly

hinder mass transport of the I-/I3

- electrolyte. Impedance experiments were also

performed on TT1 sensitized cells with an acetonitrile electrolyte and fit to the same

model at a bias of 0.9 V (no feature due to diffusion was clearly visible at 0.8 V, and

the diffusion should be relatively independent of bias). This resulted in significantly

lower (R, td) values of (2.1 Ohms, 0.06 s), showing the effect of changing the

electrolyte solvent on mass transport of the redox couple is much more important than

the effects of adding high concentrations of ERD. As a side note, it can be pointed out

that the faster transport of the redox couple through an acetonitrile based electrolyte

than the benzonitrile electrolyte is the cause for the higher fill-factor seen in DSCs

with electrolytes using an acetonitrile solvent. These impedance measurements

suggest that increased ERD concentration doesn’t hinder mass transport of the I-/I3

-.

4.3 ERD DSC EQE and Energy Transfer Efficiency

The external quantum efficiency (EQE) of the DSCs is shown in Figure 4.9.

The superior photovoltaic performance of BL315 as compared to BL302 can be

attributed to its broader, more red-shifted absorption. However, despite the high dye

loading, the EQE in the ERD portion of the spectrum (450-550 nm) is still

significantly lower than the peak EQE of the TT1 sensitizing dye. Conversion of

incident photons into collected charges by an ERD is a three-step process, and the

EQE of the ERD (EQEERD) can be written as

, (4.4)

54

where ABSERD is the absorptance of the ERD within the mesoporous TiO2 layer

(fraction of photons absorbed by the ERD), ETE is the energy transfer efficiency of

the ERD to the sensitizing dye, and IQESD is the internal quantum efficiency (IQE) of

the sensitizer. By measuring the EQE of the device and the absorptances of the

sensitizing dye and the ERD, the energy transfer efficiency can be backed out. The

EQE and abssorptance of the ERD is measured in the green (520 nm), where the

sensitizer does not absorb strongly, but at higher wavelengths than the absorption of

the electrolyte. The IQE of the sensitizer is calculated in the red, where the ERD

doesn’t absorb, and to get an accurate measurement of the ETE, this is done at the

dye’s absorption peak – 690 nm.

The absorptance of the ERD can be estimated from Beer’s law:

, (4.5)

where TFTO is the transmission through the FTO electrode, ρ and x are the porosity and

thickness of the TiO2 mesoporous layer (approximately 0.5 and 6 µm, respectively), c

is the concentration of the ERD, and α is the molar extinction coefficient of the ERD

(30,000 M-1

cm-1

at 520 nm for BL302). TFTO can be measured by measuring the

absorptance of the FTO glass substrate (0.05 at 520 nm, shown in Figure 4.10) and

then accounting for 1 reflection off of the front glass/air interface of approximately 4%

for a total TFTO of 0.91 at 520 nm. For 180 mM BL302, ABSERD comes out to 0.89 at

520 nm, meaning nearly all available light transmitted through the FTO is absorbed by

the ERD. While it is possible that the concentration of the ERD within the TiO2 pores

is less than the average concentration within the electrolyte, the high absorptance of

the ERD means that even for a significant change in the ERD concentration, ABSERD

remain near 0.91. The flattening of the EQE in the ERD portion of the spectrum in

Figure 4.9 with increasing dye concentration suggests that indeed the absorption is

saturating.

55

Figure 4.9. EQE of 6 µm-thick TT1-sensitized DSCs containing various amounts of

BL302 (a) and BL315 (b) in the electrolyte.

Figure 4.10. Absorptance of FTO as measured with an integrating sphere to account

for scattering and reflection.

56

The IQE of the sensitizer can be computed by dividing the EQE of the

sensitizer (EQESD) by the percentage of photons absorbed by the sensitizer (ABSSD):

. (4.6)

Applying Equation 4.6 to the EQE and absorptance shown in Figure 4.11, the IQE of

the TT1 sensitizer was measured to be 85–90% for most devices. After calculating

IQESD and ABSERD, the ETE can be calculated from the measured energy relay dye

EQE and Equation 4.4. While previously ETE’s for DCM to TT1 have been shown to

be in excess of 90%,104

the average ETE for BL302 to TT1 is only approximately

70%, and the corresponding ETE for BL315 is approximately 67%.

Figure 4.11. EQE of TT1 Device with 180 mM BL302 and difference between

transmittance of TT1 sensitized substrate and FTO substrate (which is the sensitizing

dye absorptance). The EQESD and ABSSD are measured at their peaks. There appears

to be a slight redshift in the TT1 EQE vs. the TT1 absorptance which can be attributed

to the presence of solvent. EQEERD is measured at 520 nm, as the electrolyte absorbs

relatively strongly below 500 nm.

4.4 Introduction to Quenching and Analysis of ETE Losses

ETE losses can be caused by a variety of physical processes in the DSC, and

before analyzing which process contributes the most to the 30% losses seen with

BL302, it is instructive to give a brief overview of the ways ERD excitations can be

lost. FRET between two chromophores is a short scale interaction that can only occur

57

efficiently over distances of 1-10 nm. If the pore size is too large, then the ERD cannot

efficiently transfer energy to the sensitizing dye and the excitation is lost. The

theoretical energy transfer efficiency of ERDs in pores has been previously

simulated.103

Dynamic (collisional) quenching of the chromophore in solution is a

competing process with FRET, and the effects of dynamic quenching are included in

the calculations of ERD ETE as a function of pore size. On the other hand, static

quenching, such as quenching from forming a non-emissive complex in solution,

results in ‘dead dyes’ that cannot energy transfer to sensitizers. Thus, static quenching

results in a complete loss of excitation, rather than a competing rate process. Finally,

the ERD can energy transfer to a dye that is no longer attached to the TiO2 surface.

Such a desorbed dye will be unable to inject charge, and the excitation of the ERD will

be lost. Considering these three possibilities allows for an understanding of the

dominant loss mechanism of BL302 excitation in DSCs.

Figure 4.12. Time resolved PL spectrum of BL302 in 85:15 benzonitrile:valeronitrile.

A fit to the linear portion of the decay (on a log-linear scale) gives a PL lifetime of

approximately 2.0 ns.

Dynamic (collisional) quenching can be measured using time-resolved

photoluminescence measurements and manifests itself as a decrease in

photoluminescence lifetime. An example time-resolved PL measurement is shown in

Figure 4.12. To probe static quenching, the magnitude of the PL signal is examined

and the decrease in the total photoluminescence can give insight into the amount of

58

static quenching. A decrease in the steady-state photoluminescence can be caused by

either static or dynamic quenching, and the decrease in the PL signal of BL302 due to

the addition of ions contained in the electrolyte is shown in Figure 4.13.

Figure 4.13. Photoluminescence spectrum of 10 mM BL302 with electrolyte. 25%

and 50% corresponds to 25% and 50% of the concentrations of ions that is used in the

DSC electrolyte. As can be seen, there is no significant change in shape of the

photoluminescence spectrum – likely because any aggregates (that may have a

different PL spectrum) do not photoluminesce. There is a slight redshift, but this may

be caused by the addition of ions to the solvent.

4.5 Dynamic Quenching and Pore Size Dependence

As mentioned previously, energy transfer can be highly dependent on the TiO2

pore size. Devices were fabricated using substrates with 3 average pore sizes: 12 nm,

17 nm and 32 nm.4,107,108

Average ETE values for these devices are displayed in

Figure 4.14 along with theoretical ETE values. The energy transfer efficiency for an

ERD can be calculated based on the dynamic quenching rate, FRET radius, and the

sensitizing dye surface coverage for various pore geometries – see work of Hoke et

al.103

Based on the dynamic quenching rate of BL302 in the electrolyte, (Figure 4.15)

the calculated FRET radius of 6 nm, and estimated dye coverage of 1 nm-2

, the

simulated ETE can be calculated as shown in Figure 4.14. Even at a pore radius of 32

nm, the expected ETE inside a spherical pore is 94%; however, the experimentally

seen ETE remains significantly lower - relatively constant near approximately 70%.

59

This suggests that the pore size is not a significant factor in the incomplete energy

transfer efficiency and that ERDs are compatible with a variety of pore sizes,

including the larger nanopores implemented in current record devices.4

Figure 4.14. Experimental ETE compared to expected theoretical ETE for spherical

and cylindrical pore geometries.

Figure 4.15. a) Dynamic quenching of BL302 dye in 85:15 benzonitrile:valeronitrile

mixture as measured by time-resolved photoluminescence. τ is the Fluorescence

lifetime and τ0 is the fluorescence lifetime at 0 electrolyte concentration. b) Dynamic

quenching of BL302 due to high concentrations of dye in benzonitrile solvent. Here, τ

is the fluorescence lifetime and τ0 is the fluorescence lifetime at 1 mM dye

concentration. Multiplying the decreases in lifetime together results in a total τ0/τ of

6.9.

60

4.6 ETE Losses Due to Sensitizing Dye Desorption

Benzonitrile was used as an electrolyte solvent due to its relative stability, lower

vapor pressure and enhanced solubility of alkyl-substituted ERDs versus the

commonly used acetonitrile. Benzonitrile also enhances the solubility of the

sensitizing dye, since sensitizing dyes often use alkyl groups to help prevent

recombination at the titania interface.97–99

To quantify the amount of TT1 desorption, a

sandwich device was fabricated exactly like a DSC except substituting a plain glass

electrode for the Pt-counter-electrode (inset of Figure 4.16). The sandwich device was

then filled with benzonitrile or acetonitrile electrolyte and left for a week for the dye

desorption/adsorption processes to equilibrate. Over the course of a week, the

electrolyte turned a slightly blue color due to desorbed dyes from the TiO2 surface.

The absorption was measured through a region of the sandwich device that contained

no TiO2, but had electrolyte with desorbed dye electrode (inset of Figure 4.16). Using

the known width of the surlyn spacer, this gives an estimate of the desorbed dye

concentration within the electrolyte and hence the desorbed dye concentration within

the TiO2 pores.

Figure 4.16. Absorption of electrolyte after equilibration of dye desorption for

benzonitrile and acetonitrile based electrolytes. Inset: experimental schematic of

measuring light absorption through the electrolyte. After allowing dye desorption to

equilibrate, the absorption of the dye in the electrolyte is measured using a beam path

shown by the arrows in the inset.

61

As shown in Figure 4.16, only a small amount of sensitizing dye desorbs into

the acetonitrile electrolyte, while benzonitrile causes an approximately 6x increase in

concentration of TT1 within the electrolyte. At equilibrium the measured

concentration in the acetonitrile and benzonitrile electrolytes is 0.23 mM and

1.31 mM, respectively. It should be noted that these concentrations correspond to a

small fraction of the TiO2 adsorbed dyes and hence doesn’t greatly affect the density

of dyes on the surface. Assuming that this dye concentration is present within the

pores, this corresponds to approximately 2.0 dye molecules contained in a 17 nm

spherical pore for a DSC using benzonitrile electrolyte.

If a sensitizing dye desorbs from the TiO2, energy relay dyes can FRET to the

desorbed dye, resulting in lost photocurrent. Because of the r-6

dependence of the

FRET rate on the chromophore separation distance (r), ERDs within the neighborhood

of the desorbed dye can preferentially energy transfer to the desorbed dye. To model

this energy loss, a desorbed dye was placed a distance x from the center of a 17nm

pore (Figure 4.17). The ERD’s can be considered to be homogenously distributed in

order to calculate the average energy transfer efficiency to the desorbed dye.

Figure 4.17. Schematic diagram of desorbed sensitizing dye a distance x from the

center of a spherical sensitizing-dye-lined pore.

Consider a volume element, dV, of ERD at position (x1,y1,z1), the rate of

energy transfer from that dye element to the desorbed dye is given by

62

, (4.7)

where R0 is the calculated FRET radius, and is the photoluminescence lifetime.

The FRET rate of the same ERD volume element to the SD’s on the surface is given

by

, (4.8)

where CSD is the surface coverage of sensitizing dye, dS is the surface area element of

the pore, S is the pore surface and

is the distance of the ERD

volume element from the center of the pore. In spherical coordinates,

, and , and the surface area element is

given by (where r is simply the pore radius). Substituting and

integrating θ from 0 to π and φ from 0 to 2π gives the relative FRET rate of the ERD

volume element to the surface. From here, we can integrate the relative energy

transfer rates over the volume of the pore to find what percentage of the ERDs in the

pore energy transfer to the desorbed dye, LossERD,

(4.9)

It should be noted that exact knowledge of RFRET and is not necessary to

calculate to losses due to desorbed dyes, as these cancel in the calculation of LossERD

above. The percentage of ERD excitation lost, as a function of the distance of the

desorbed dye from the pore center, x, is shown in Figure 4.18. It should also be noted

that this calculation finds the percentage of ERD excitation lost by energy transfer to a

desorbed dye and doesn’t include other losses such as dynamic quenching. These can

be included as a position-independent rate in Equation 4.9. However, as was seen

previously in Section 4.5, the quenching rate is less than the FRET rate, resulting in

near 100% predicted ETE.

63

Figure 4.18. Fraction of ERD excitation lost as a function of the distance of the

desorbed dye from the center of the pore for a 17nm diameter pore with 1 sensitizing

dye per nm2

surface coverage. When the dye is near the center, the r6 nature of the

FRET interaction causes a large amount of ERDs in its vicinity to preferentially FRET

to the desorbed sensitizer.

As can be seen in Figure 4.18, the percent of excited ERD energy that is

transferred to the desorbed dye varies from 3.4% in the center of the pore down to

approximately 0.1% for a dye near the surface. Under the assumptions that desorbed

dyes will be homogenously distributed other than the outside 2nm of the pore (since

dyes on the surface take up physical volume) the ETE loss can be averaged over the

volume to get an average ETE loss of approximately 1.75%. If two dyes are desorbed,

then the total ETE loss would be slightly less than twice 1.75%, and would require

calculation of the FRET rates for both dyes and the surface. With many dyes, this

calculation because relatively intractable fairly quickly, and a monte carlo simulation

would be the easiest approach to estimating the ETE loss. However, for only a few

dyes, a reasonable estimate of the ETE loss is simply the number of dyes times the

average loss for one dye, as the dyes are unlikely to be in the same neighborhood of

eachother. Thus for 2.03 desorbed dyes per pore, an estimate of the ETE loss is

2.03*1.75%=3.6%. Thus, the approximate loss due to dye desorption in a 17 nm pore

is approximately 3.6%. While this loss is an important consideration, it does not

explain the entire 30% ETE loss seen in the BL302 devices.

64

4.7 Static Quenching of ERDs

FRET to an ERD occurs at a certain rate based on the distance between

chromophores and their FRET radius. In a DSC, there are competing rate processes,

such as dynamic (collisional) quenching of the excitation by the various ions in the

electrolyte. If the FRET rate is significantly faster than this dynamic quenching rate,

then the ETE can still approach unity. On the other hand, if the dye is statically

quenched, through a process such as forming a nonemissive complex in solution, the

complex no longer has an opportunity to FRET to a sensitizing dye on the TiO2

surface, and the excitation is lost.109

In order to investigate the losses due to of static

quenching of the ERD, steady-state and time-resolved photoluminescence (PL)

quenching experiments were performed. During a steady-state PL measurement, both

dynamic quenching and static quenching cause a decrease in the PL signal. However,

during time-resolved PL quenching measurements, a decrease in the PL lifetime can

only be caused by a change in the dynamic quenching rate – non-emissive statically-

quenched complexes simply do not photoluminesce. Thus by comparing the steady-

state PL quenching and decrease in PL lifetime, the amount of static and dynamic PL

quenching can be calculated.109

The steady state PL quenching should be given by

, (4.10)

where r is the fraction of dyes that are statically quenched, PL and PL0 are the

magnitudes of the steady state photoluminescence with and without addition of the

quenchers, respectively, and τ and τ0 are the PL lifetimes with and without quenchers,

respectively. A comparison of the time-resolved and steady-state PL measurements of

BL302 as a function of electrolyte concentration is shown in Figure 4.19.

65

Figure 4.19. Comparison of steady-state photoluminescence quenching with decrease

in photoluminescence lifetime for 1 mM, 10 mM, and 27 mM BL302 with varying

concentrations of electrolyte. Solid and dashed grey lines are linear fits of PL and τ,

respectively. Note: the linear trend continues for both PL and τ to 100% electrolyte

concentration. Electrolyte concentration (%) is the percentage of electrolyte

components relative to the standard electrolyte used in DSC devices.

As can be seen in Figure 4.19, there is significantly more steady-state PL

quenching than can be explained by the decrease in PL lifetime. While both PL0/PL

and τ 0/τ display linear trends, the decrease in the steady-state photoluminescence is

larger than the decrease in photoluminescence lifetimes, even at low electrolyte

concentrations. This data suggests that BL302 is being statically quenched, possibly

by forming complexes with components of the electrolyte. By measuring the

quenching at 100% electrolyte concentration for 10 mM BL302, PL0/PL and τ 0/τ were

found to be 3.0 and 2.0, respectively. Applying Equation 4.10, it is found that r, the

fraction of statically quenched dyes, is approximately 0.33 in the conditions present in

the DSC. While this number is larger than the 25-30% losses that are seen for the

BL302 devices, it is important to note that statically quenched dyes can still contribute

to photocurrent by direct injection. As has been previously seen,104

ERDs are able to

inject and generate photocurrent even in the absence of sensitizing dyes, albeit at a

much lower efficiency than by FRET. Thus, while a small portion of the statically

quenched dyes may directly inject and their excitations are not lost, static quenching

66

can explain the 26% of the ETE loss that is unaccounted for by energy transfer to

desorbed dyes.

Figure 4.20. a) Comparison of time-resolved and steady-state quenching of 5 mM

DCM in acetonitrile electrolyte. b) Comparison of time-resolved and steady-state

quenching of 8.5 mM DCM in benzonitrile electrolyte. Note: Electrolyte

concentration (%) is the percentage of electrolyte components relative to the standard

electrolyte used in DSC devices.

As shown in Figure 4.20, static quenching by the electrolyte does not occur for

DCM in either benzonitrile-based or acetonitrile-based electrolytes. The difference in

quenching between BL302 and DCM could originate from the difference in their

structure. Compared to DCM, BL302 has different length alkyl chains on the amine

group and a change from a methyl to a bulkier tert-butyl substituent group on the

pyran ring. This could lead to a different coordination with the electrolyte

components,110

possibly with formation of a non-emissive complex within the pores.

Another explanation could be that the increasing amount of ions with electrolyte

concentration causes the alkyl-substituted dye molecules to form complexes in

solution.

High concentrations of ERD can also cause both static and dynamic

photoluminescence quenching. Just as previously, the time-resolved and steady-state

quenching can be compared to deconvolute the amount of static and dynamic

quenching, as shown in Figure 4.21. The decrease in steady-state PL quenching is the

same as the decrease in the PL lifetime until 60 mM. However, at 180 mM, the

67

steady-state PL is quenched by a factor of 5.6, while the time-resolved PL goes down

by only a factor of 3.3. This means that significant number (approximately 41%) of

dyes are statically quenched. It is possible that this also contributes to the 30% ETE

loss seen with using 180 mM BL302 in the electrolyte. However, it is also possible

that high concentrations of ERD cause some ERD to crash out of solution and form

unsoluble aggregates. These aggregates would cause static quenching, but might not

be able to go inside the small pores of the DSC. Thus it cannot be said for certain how

much of a loss this static concentration actually causes.

Figure 4.21. Comparison of time-resolved and steady-state quenching of BL302 in

benzonitrile with varying concentration.

The ETE was measured for devices with 20 mM and 60 mM BL302, in exactly

the same way as previously: the absorptance of the ERD (ABSERD) estimated using

Beer’s law, and the ETE calculated from the EQEERD, IQESD and ABSERD. Since the

absorptance is not near 0.91 (the transmission of the FTO), it can depend relatively

strongly on the porosity of the substrates used. The average calculated ETE for 20

mM and 60 mM BL302 devices with 17 nm pores comes out to 76% for both

concentrations, when using a porosity of 0.5 to estimate the absorption. Even using a

porosity of 0.4 in the calculation of absorptance, a value lower than the porosity of the

substrate, the calculated ETE is below 90%. Hence, it stands to reason that the ETE is

still well below 100% even without the effects of dye concentration quenching, but

still possible that concentration quenching can be a factor in further lowering the ETE

68

slightly. This static concentration quenching may also be a cause of the ETE losses

seen with BL302 and BL315 in TT1 DSCs. However, due to the small size of the

pores, larger aggregates caused by lack of solubility may not be able to be formed

inside the pores, and it is also possible that any static concentration quenching does

not affect DSC performance.

Static quenching of the ERD (such as by complex formation) is responsible for

the majority of BL302 excitation losses. Losses due to desorbed dyes and the large

size of TiO2 nanopores should cause losses of less than 5% in these highly soluble

ERD systems, and if dyes can be designed that avoid static quenching, ERDs will be

able to achieve the dual goals of both greater than 95% light absorption and 95%

energy transfer efficiency.

4.8 Conclusion

While previously ERDs have been unable to achieve near 100% light

absorption within the pores of a DSC, BL302 and BL315 are able to absorb 97% of

the incident light. This results in a 65% increase in the efficiency of a TT1-based DSC,

the highest increase due to an ERD thus far reported. However, BL302 and BL315

only achieve an energy transfer efficiency of approximately 70%, despite the

theoretical ETE being very close to 100%. Of this 30% loss, approximately 3.6% can

be explained by energy transfer to desorbed sensitizing dyes, while the rest can be

explained by static quenching of the ERD. It is hypothesized that quenchers in the

electrolyte coordinate to alkyl groups on the soluble ERD, or the introduction of ions

causes these ERDs to aggregate. . Continuing to gain a better understanding of ERD

design rules is necessary for use of ERDs for complementary light harvesting in record

devices.

ERDs have drawn interest due to their ease of addition to a DSC, with

significant potential for improving the spectral response of the device. Other

strategies toward achieving broad absorption such as cosensitization have certain

drawbacks such as taking valuable ‘real estate’ on the TiO2 surface, and possibly

affecting charge injection and recombination in the device. However, one of the

69

biggest advantages of cosensitization is the myriad of dyes that have been synthesized

and used as sensitizers in DSCs, and it is relatively easy to find one with the correct

energetic and complementary absorption to be used as a cosensitizer. I think that this

is one of the reasons that cosensitization has been so effective: many of the research

groups working on DSCs have tens or hundreds of sensitizers available to them and

can simply try quite a few of them, while dyes that may be useable as ERDs need to be

characterized for absorption, photoluminescence, and solubility before they can even

be tried in a device. Indeed, the current world record DSC has been produced by

cosensitization,4 pushing the record efficiency to 12.3%.

According to Henry Snaith’s estimates of maximum attainable DSC

efficiency,56

this is getting close to the theoretical maximum of 14% or so, based on

the required energy losses involved in charge transfer. While this is a great efficiency

at the module scale that is competitive with current inorganic technologies, lab

efficiencies of 14% can only translate to module efficiencies of 10-12% due to losses

associated with larger areas, series resistances, shading from bus bars, etc. Thus in

order to achieve lab efficiencies of 20% or higher and be competitive with silicon and

thin film photovoltaics, a new strategy has to be employed. This is the impetus for

Chapter 5, incorporating DSCs and in particular, ssDSCs into devices that can achieve

efficiencies greater than 20%.

4.9 ERD Synthesis

Bogyu Lim designed and performed synthesis of ERDs, along with

characterization of intermediate and resulting compounds. This subsection contains

information on synthesis and characterization of the final and intermediate

compounds.

70

Figure 4.22. Synthetic scheme for BL302

Synthesis of 1: A mixture of 2-tert-butyl-6-methyl-4H-pyran-4-one (1.16 g, 7.00

mmol) and malononitrile (0.56g, 8.5 mmol) in 5 mL of acetic anhydride was stirred

and heated to 120 °C for 12 h under nitrogen atmosphere. The reaction mixture was

then quenched by deionized water and the solution was extracted with chloroform.

The organic layer was washed with brine solution followed by drying over anhydrous

MgSO4. Recrystallization from ethanol gave the product as peachy fibrous solid.

Yield: 97% (1.45 g). 1H NMR (CDCl3, 400 MHz, [ppm]): δ 6.56 (s, 1H), 6.55 (s, 1H),

2.34 (s, 3H), 1.30 (s, 9H); GC-MS: m/z 214, calcd 214.11.

Synthesis of BL302:111

A mixture of 1 (0.3 g, 1.40 mmol), diethylaminobenzaldehyde

(0.248 g, 1.40 mmol), and piperidine (0.20 mL, 2.00 mmol) was placed in a flask

containing dry acetonitrile (10 mL) under nitrogen atmosphere. The mixture was

heated to reflux for 24 hours. After cooling, the reaction was quenched by water and

was extracted with dichloromethane. The combined organic layer was dried over

anhydrous MgSO4 and evaporated under vacuum. The mixture was purified by silica

gel column chromatograph with dichloromethane as eluent. 0.335 g of the neon red

solid of product was collected (64% yield). 1H NMR (CDCl3, 400 MHz, [ppm]): δ

7.42 (d, 2H, J = 8.99 Hz), 7.35 (d, 1H, J = 15.83 Hz), 6.67 (d, 2H, J = 8.99 Hz), 6.58

(d, 1H, J = 2.08 Hz), 6.51 (d, 1H, J = 2.08 Hz), 6.48 (d, 2H, J = 15.83 Hz), 3.45 (m,

4H), 1.37 (s, 9H), 1.23 (m, 6H); LC-MS: m/z 374.2250, calcd 373.22. Anal. Calcd for

C24H27N3O: C, 77.18; H, 7.29; N, 11.25. Found: C, 77.6; H, 7.49; N, 11.24.

71

Figure 4.23. 1HNMR spectrum of BL302.

Figure 4.24. Synthetic Scheme for BL315

Synthesis of 2:112

A mixture of 4-bromoaniline (5.15 g, 30 mmol), 2-bromoethyl ethyl

ether (7.67 g, 68 mmol), K2CO3 (9.4 g, 68 mmol) and KI (0.63 g, 3.8 mmol) was

placed in a flask containing butanol (20 mL) under nitrogen atmosphere, and the

mixture was stirred under nitrogen at 120 °C for 6 d. The reaction was cooled to

room temperature and filtered through Celite® and the residue was washed with IPA.

The filtrate was concentrated under reduced pressure and the remaining oil was heated

with acetic anhydride (1.5 mL) in butanol (15 mL) at 120 °C for 15 min (the mono

alkylated by-product reacted to give the acetylated product.) The reaction was cooled

72

to room temperature and filtered through Celite® and the residue washed with IPA.

The filtrate was concentrated under reduced pressure and crude oil was purified by

silica gel column chromatograph with hexane:ethyl acetate (4:1) as the eluent. The

desired product was obtained as a slightly yellow viscous oil (8.145 g, 85%). 1H NMR

(CDCl3, 400 MHz, [ppm]): δ 7.26 (d, 2H, J = 9.12 Hz), 6.60 (d, 2H, J = 9.12 Hz), 3.57

(m, 8H), 3.49 (m, 4H), 1.19 (t, 6H, J = 14.02 Hz); GC-MS: m/z 315, calcd 315.08.

Synthesis of 3: A 25 ml flask was charged with 2 (2.05 g, 6.5 mmol), 2-thiophene

boronic acid (0.90 g, 7 mmol), Pd(PPh3)4 (0.15 g, 0.13 mmol) and toluene (10 mL),

and degassed with nitrogen for 15 min. 3.3 mL of 3M K2CO3 was added, and reaction

mixture was heated at 90 °C for 24 h and then cooled down at room temperature and

quenched with 2M HCl. The product was extracted with dichloromethane and the

combined organic layers were washed with plenty of water. The organic extracts were

then dried over anhydrous MgSO4, evaporated and purified with column

chromatography on silica gel with hexane:ethyl acetate (95:5) as the eluant to give

yellow viscous oil. Yield: 88% (1.82 g). 1H NMR (CDCl3, 400 MHz, [ppm]): δ 7.47

(d, 2H, J = 8.99 Hz), 7.13 (t, 2H, J = 10.70 Hz), 7.03 (dd, 1H, J = 8.89 Hz), 6.72 (d,

2H, J = 8.99 Hz), 3.60 (s, 8H), 3.53 (dd, 4H), 1.20 (t, 6H, J = 14.03 Hz); GC-MS: m/z

319.2, calcd 319.16.

Synthesis of 4: 0.17 mL (1.85 mmol) of phosphorus oxychloride (POCl3) was added to

N,N-dimethylformamide at 0 °C, and the solution was stirred for 30 min. 3 (0.50 g,

1.57 mmol) in 10 mL of dichloroethane was added to the above solution and stirred at

80 °C for 3 h. After cooling to room temperature, 1 M sodium hydroxide was added

and the mixture was stirred vigorously for 1 h for neutralization. The solution was

extracted with dichloromethane, and the combined organic extracts were washed with

brine and dried over MgSO4. After removal of the solvents under reduced pressure, the

residue was purified by flash column chromatography with dichloromethane to give

yellow viscous oil. Yield: 74% (0.405 g). 1H NMR (CDCl3, 400 MHz, [ppm]): δ 9.81

(s, 1H), 7.68 (d, 1H, J = 3.93 Hz), 7.54 (d, 2H, J = 8.82 Hz), 7.23 (d, 1H, J = 3.93 Hz),

73

6.74 (d, 2H, J = 8.82 Hz), 3.61 (s, 8H), 3.53 (dd, 4H, J = 20.93 Hz), 1.20 (t, 6H, J =

14.04 Hz); GC-MS: m/z 347, calcd 347.16.

Synthesis of BL315: A mixture of 1 (0.25 g, 1.17 mmol), 4 (0.40 g, 1.16 mmol), and

piperidine (0.17 mL, 1.70 mmol) was placed in a flask containing dry acetonitrile (10

mL) under nitrogen atmosphere. The mixture was heated to reflux for 24 hours. After

cooling, the reaction was quenched by pouring water and it was extracted with

dichloromethane. The combined organic layer was dried over anhydrous MgSO4 and

evaporated under vacuum. The mixture was purified by silica gel column

chromatograph with dichloromethane as eluent. The clay red solid of product was then

collected (0.469 g, 74% yield). 1H NMR (CDCl3, 400 MHz, [ppm]): δ 7.49 (m, 3H),

7.19 (d, 1H, J = 3.89 Hz), 7.13 (d, 1H, J = 3.89 Hz), 6.75 (d, 2H, J = 8.92 Hz), 6.61 (d,

1H, J = 1.99 Hz), 6.52 (d, 1H, J = 1.99 Hz), 6.43 (d, 1H J = 15.38 Hz), 3.61 (s, 8H),

3.53 (dd, 4H, J = 20.78 Hz), 1.37 (s, 9H), 1.22 (t, 6H, J = 14.15 Hz); LC-MS: m/z

544.2383, calcd 543.26. Anal. Calcd for C32H37N3O3S: C, 70.69; H, 6.86; N, 7.73; S,

5.90. Found: C, 70.43; H, 7.01; N, 7.67; S, 6.0

Figure 4.25. 1HNMR spectrum of BL315.

74

Figure 4.26. Cyclic Voltammogram of BL302 (red), and BL315 (blue) dyes. From the

curves, it was found that HOMOBL302=5.12 eV, LUMOBL302=3.07 eV,

HOMOBL315=5.05 eV, LUMOBL302=3.23 eV.

4.10 Experimental Details

Materials were purchased from commercial suppliers (Aldrich, Acros) and used as

received unless otherwise noted.

Device Fabrication: Nanoporous titania DSC substrates were prepared as previously

described in detail.113

Substrates were fabricated with 6 µm thick active layer made

with 20 nm particles and a peak pore diameter of 17 nm. Substrates utilizing

12nm107,108

and 32 nm average pore diameters were also fabricated as previously

described in the literature. Devices were fabricated by heating to 500 °C to remove

any water/organics, cooling back to room temperature, and then immediately

immersing in 0.1 mM TT1 solution in ethanol with 10 mM CDCA (chenodeoxycholic

acid). The devices were allowed to sit in solution for 4.5 hours and then were sealed

with a platanized FTO back electrode using a 25 nm surlyn spacer by heating and

pressing on a 125 °C hotplate. The platanized back electrode was made by drilling a 1

mm hole in FTO, then covering with a solution of chloroplatinic acid hydrate in

75

isopropyl alcohol and heating to 450 °C. The electrolyte was vacuum backfilled and

the entire DSC sealed with surlyn and a glass coverslip. The ‘benzonitrile-based’

electrolyte was composed of 0.6 M 1-methyl-3-propylimidazolium iodide, 0.04 M I2,

0.28 M 4-tert-butylpyridine, 0.025 M LiI and 0.05 M guanidinium thiocyanate in

85:15 benzontrile:valeronitrile. The ‘acetonitrile-based’ electrolyte was the same

components (0.6 M 1-methyl-3-propylimidazolium iodide, 0.04 M I2, 0.28 M 4-tert-

butylpyridine, 0.025 M LiI and 0.05 M guanidinium thiocyanate) in 85:15

acetontrile:valeronitrile.

Thickness measurements:TiO2 layer thicknesses were measured using a Veeco

Dektak3 ST surface profiler profilometer.

IV Measurements: J-V curves were taken with a Keithley 2400 sourcemeter, under

simulated AM 1.5G illumination with a Spectraphysics model 91160 solar simulator

which has been callibrated using a hamamatsu Si photodiode with KG5 filter. The

DSC was was masked with a 0.159 cm2 area machined mask during J-V

measurements.

EQE and absorptance measurements: EQE was measured at a chopping frequency of

approximately 2 Hz. For the chopped EQE beam, a Newport Apex monochromator

illuminator was focused on a Princeton Instruments monochromator and subsequently

put through a filter wheel. The signal from the DSC was amplified with a home-built

transimpedance amplifier and recorded on a Stanford Instruments lock-in amplifier.

The EQE was calibrated against a calibrated photodiode of known EQE. Additionally,

the EQE beam was split with a 50:50 beam splitter into a 2nd

‘reference’ photodiode

that was used to correct for any fluctuations in the illuminator source intensity.

Experimental set-up for absorptance measurements have also been reported.67

Device

absorptance measurements were performed using the same light

source/monochromator setup as the EQE and measured using an integrating sphere

with an attached silicon photodiode attached to a Kiethley sourcemeter. Care was

taken to ensure that the incidence angle of the light should be as close to normal as

possible.

76

Absorption measurements: Absorption measurements were made using a Cary 6000i

UV/Vis spectrophotometer.

Photoluminescence measurements: Steady-state photoluminescence measurements

were made using an Ar-ion laser for illumination at 488 nm. The resultant

photoluminescence was focused onto a monochromator and measured with a CCD,

correcting for the CCD response. Time resolved photoluminescence measurements

were performed using a Picoharp 300 time-correlated single photon counting system

with a Picoquant PDL 800-B pulsed laser diode driver with a Picoquant model LDH-

P-C-485 laser for 485 nm excitation with <1 ns rise time at a frequency of 10 Mhz,

detected with an avalance photodiode (PDM 100CT SPAD).

Impedance spectroscopy measurements: Impedance spectroscopy was performed with

a Biologic SA model VMP3 potentiostat. Fits were made using EC-Lab software.

77

5 Silver Nanowire Electrodes for Semitransparent ssDSCs

As of 2011, DSCs had achieved efficiencies of 12.3%, coming close to the

maximum attainable efficiency of 14%.56

However, due to the continue decrease in

price and increase in efficiency of traditional inorganic modules, the target efficiency

for DSCs to become cost competitive with their inorganic counterparts rose even

higher. And although it may be possible to achieve efficiencies of over 20% if the

required energy losses in DSCs were brought down, this is a difficult task. DSCs have

many of the properties that are required for a top cell in a two cell tandem

photovoltaic: transparency to low bandgap light, high open-circuit voltages and ease-

of-fabrication. Additionally, many of the solar cells in use today (in particular silicon

and CIGS cells) have bandgaps that are less than the ideal bandgap for a single

junction solar cells of 1.4 eV, and more in line with the ideal bandgap for the botton

cell in a tandem photovoltaic (1.0-1.1 eV). An idea pioneered by Beiley and McGehee

is that a organic or hybrid organic solar cell, such as a DSC or ssDSC, can be used as a

top cell in a hybrid tandem photovoltaic (HTPV) with a silicon or CIGS bottom cell.57

Such HTPV devices can achieve efficiencies well in excess of 20%, which can

significantly reduce installation and other balance of systems costs by requiring less

installation of photovoltaics to output the same power.

At this point, I would like to give a little background on tandem photovoltaics

and HTPV devices. A two subcell tandem photovoltaic utilizes a high bandgap

absorber that is able to get more energy per absorbed photon, and then a low second

lower bandgap absorber that is able to absorb longer wavelength photons. Tandem

photovoltaics are no longer subject to the Shockley-Queisser efficiency limit of

31%,114

and can surpass the maximum theoretical efficiency of a single junction

device. Two common configurations of a two cell tandem are shown in Figure 5.1: a

two terminal tandem, where the connection of the two subcells requires that the

currents in each subcell be matched, and a four-terminal tandem, where each of the

cells is operated independently.

78

Figure 5.1. a) Schematic of DSC in a 4-terminal HTPV configuration. b) Schematic of

DSC in a 2-terminal HTPV configuration. The DSC is built on the Si or CIGS solar

cell and may have an interfacial layer to make electrical contact and/or planarize the

inorganic solar cell. The arrows depict light that is incident on each of the subcells of

the tandem.

The 4-terminal architecture has the benefits that no current matching between

subcells is required. In a 2-terminal tandem, the device short-circuit current comes out

to be close to the value of the lower current subcell. On the other hand, a 4-terminal

tandem in the architecture shown in Figure 5.1a has 2-3 transparent conductors, which

can result in a significant amount of optical losses compared to the 2-terminal device

which can be fabricated with only one transparent conductor. Additional wiring may

be necessary with the 4-terminal device which also might add to the BOS costs. To

understand the types of efficiencies that might be achievable in a HTPV using a DSC

or ssDSC, the efficiencies of tandem devices was modeled following a similar

approach to the work of Beiley and McGehee.57

5.1 Modeling

The theoretical efficiency of a tandem photovoltaic can be modeled from the

transmittance and efficiency of the top cell along with the EQE of the bottom cell and

79

some simple assumptions the bottom device’s fill factor and open-circuit voltage. For

a given semitransparent top cell, the J-V characteristics of the device remain

unchanged when a bottom cell is introduced. However, the efficiency of the bottom

cell is significantly reduced as the light incident on the cell is filtered by the top cell

(seen in the schematic of Figure 5.1). Since the AM1.5G spectrum is incident on the

tandem cell, the transmitted spectrum (in units of # of photons/nm) to the bottom cell

is given by

, (5.1)

where TTop is the transmittance of the top subcell and AM15G(λ) is the AM1.5G

spectrum in units of in units of # of photons/nm. From here the short-circuit

photocurrent density of the bottom subcell can by calculated by integration:

, (5.2)

where EQEBottom is the EQE of the bottom subcell. The open-circuit voltage of the

bottom subcell can be approximated by the decrease in voltage of an ideal diode upon

decreasing the photocurrent:

, (5.3)

where k is Boltzmann’s constant, T is the temperature, n is the ideality factor (tends to

be 1-2), JSC,Bottom is the short-circuit photocurrent of the bottom cell under incident

AM1.5G spectrum and VOC,Bottom is the open-circuit voltage of the bottom cell under

incident AM1.5G spectrum. Equation 5.3 is an approximation, but typically only

leads to kT/q (~25 mV) difference between the VOC of the bottom subcell under direct

AM1.5G illumination and when used in a tandem. The fill factor of the bottom device

can be assumed to remain constant, a relatively good assumption under typical

circumstances. In a 4-terminal tandem configuration, the sum of the efficiencies of

each subcell is the efficiency of the total device. In a 2-terminal current-matched

tandem, the short-circuit photocurrent is the smaller of that of the two subcells, the

open-circuit voltage is the sum of the voltages of the subcells and the fill factor is

between the fill factor of the two subcells with their arithmetic mean a good

approximation.

80

The photovoltaic figures of merit for a variety of inorganic solar cells can be

found in literature, company releases, or modeled using PV simulators such as PC1D.

The overall conclusions of the modeling end up being the same in the case of a variety

of inorganic thin film and silicon solar cells, so in this section I will restrict to

modeling tandems employing a CIGS bottom cell for brevity. The PV parameters of

the CIGS bottom cell are given in Table 5.1 and are taken from an actual cell made by

HelioVolt. The EQE of the device is shown in Figure 5.2.

Table 5.1. J-V characteristics for HelioVolt CIGS bottom cell used in HTPV

modeling.

Device Jsc (mA/cm2) Voc (mV) FF Efficiency (%)

HelioVolt CIGS solar cell 33.0 631 0.72 15.0

Figure 5.2. EQE of HelioVolt CIGS cell used in this section’s modelling along with

EQE of a DSC utilizing a dye with a 2.0 eV ‘bandgap.’

From here we can model the efficiency of a HTPV employing a DSC top cell

and a CIGS bottom cell. The EQE of an efficient DSC is usually between 80-90% at

peak. For our modeling, we have assumed that the EQE of the DSC top cell is 85%

from 400 nm until the absorption onset of the dye, which occurs at the ‘bandgap’ of

the dye. The EQE of an ideal 2.0 eV ‘bandgap’ dye is also shown in Figure 5.2. A

81

DSC used as a top cell contains 2 transparent conductors so that sub-bandgap light can

pass into the CIGS bottom cell. Typical transparent conductors used in DSCs, such as

FTO, can achieve transmittances of greater than 90% while maintaining high enough

sheet resistance to be used in large scale DSC modules.115

Thus to include the effects

of transparent conductors, the DSC transmittance was modeled having 2 transparent

conductors with transmittances of 90% independent of wavelength. Other

assumptions on the transmittance are that no photons below 500 nm are transmitted

due to absorption by the electrolyte solution and that the IQE is 100% for the DSC (a

valid assumption for record devices). Thus, the total transmittance of the DSC top

device is given by:

, (5.4)

where TTCO,1 and TTCO,2 are the transmittances of the top and bottom transparent

conductors (both 0.9 in our modeling), Telec is the transmittance of the electrolyte

(assumed to be 0 below 500 nm and 1.0 above 500 nm) and EQEDSC is the DSC EQE

as previously discussed. Note that for a 2 terminal tandem there could be only 1

transparent conductor and TTCO,2 can be set to 1 in this case. Shown in Figure 5.3 is

our estimate of the transmittance of a DSC using a 2.0 eV ‘bandgap’ dye.

Figure 5.3. Example transmittance of top DSC with a 2.0 eV ‘bandgap’ dye used in

our modeling of DSC-CIGS HTPV tandems. The lowered transmittance in the red

and near IR is due to the two transparent conductors while the dye absorbs most of the

light below 620 nm (2.0 eV).

82

The open circuit voltage of the DSC top cell can be simply modeled as

, (5.5)

where EGap is the ‘bandgap’ of the dye, ELoss represents the loss in energy of the DSC

required for efficient charge injection and collection, and q is the unit of elementary

charge. In record DSCs, the difference between qVOC and the onset of absorption (the

‘bandgap’) is approximately 0.80 eV.4 Thus a reasonable estimate of ELoss for current

DSC technology is 0.8 eV but could be lower in the future. Finally the fill factor of

the top cell DSC was assumed to be 0.75, a number that has also been achieved in

record DSCs.4 Using these assumptions and Equations 5.1 – 5.5, the efficiency of a

DSC-CIGS HTPV can be modeled as a function of the DSC bandgap and is shown in

Figure 5.4.

Figure 5.4. Efficiency of modeled DSC-CIGS HTPV. Two-terminal devices use only

1 transparent conductor in the modeling of the top device transmittance. The

efficiency of the bottom CIGS cell is depicted by the dashed red line. In Equation 5.5,

ELoss is assumed to be 0.8 eV. The red dashed line represents the efficiency of the

CIGS cell by itself.

The modeling shows that a DSC-CIGS HTPV using a 15% efficient CIGS cell

can achieve efficiencies of nearly 20% under the very conservative assumptions of a

0.8 eV value for ELoss. The highest efficiencies are obtained for dye ‘bandgaps’ near

83

1.7 eV for a current matched device and approximately 1.9V for 4-terminal tandem.

Most dyes used in DSCs have their absorption onsets between 600 nm and 700 nm,

corresponding to a ‘bandgap’ of 1.65 eV-2.07 eV, making a DSC-CIGS HTPV

realizable using current highly performing dyes. Since typical DSCs are nearly always

made using 2 FTO-covered glass electrodes, it is relatively straightforward to utilize a

DSC in a tandem device, and in 2006 such a device was made, improving the

efficiency of a CIGS cell from 14% to over 15% in the tandem.116

Additionally, 2-

terminal, monolithically integrated DSC-CIGS tandems have been made, but the

corrosiveness of the electrolyte has limited stability.117

As mentioned in Section 1.2, an exciting recent development in the field is the

invention of perovskite-sensitized solar cells (PSSCs). These cells, using a very

similar architecture to ssDSCs have achieved efficiencies of greater than 12%.38

Additionally, the difference in energy between qVOC and the bandgap of the

perovskite sensitizer has seen to be as little as 0.5-0.6 eV.38,118

The estimated

efficiency of such a device is shown in Figure 5.5. Thus, if a semitransparent PSSC

can be realized, the efficiency of such a device in tandem with a CIGS cell could

exceed 23%, which would significantly help decrease the BOS costs associated with

such a system.

Figure 5.5. Efficiency of modeled DSC-CIGS HTPV. Two-terminal devices use only

1 transparent conductor in the modeling of the top device transmittance. The

efficiency of the bottom CIGS cell is depicted by the dashed red line. In Equation 5.5,

ELoss is assumed to be 0.5 eV. Red dashed line represents CIGS efficiency.

84

However, PSSCs cannot be used in an HTPV as easily as a DSC because they

utilize and opaque back electrode typically made of evaporated silver or gold. Thus, a

device-compatible transparent electrode must be developed for perovskite devices that

can be deposited without damaging the underlying device. This is one of the main

motivations for this chapter: developing an electrode for PSSCs. Since ssDSCs have

essentially the same structure as perovskite cells, we set out to understand how a high

quality transparent electrode could be deposited on ssDSCs.

5.2 Transparent ssDSC Applications

Developing a transparent top electrode to make a semitransparent ssDSC

allows for the application of ssDSCs as a large bandgap solar cell in a tandem

architecture. Since both ssDSCs and PSSCs have achieved open circuit voltages of

over 1 V, 11938,120

they are both attractive candidates for large bandgap absorbers to be

used in conjunction with a with low-bandgap absorber such as CIGS or Si.57

In such a

tandem device, an efficient semitransparent ssDSC must be fabricated with two

transparent electrodes, allowing for unabsorbed low energy photons to pass through

ssDSC and be absorbed in the lower bandgap solar cell.57,116,117,121–123

In addition to the main motivation of the use of a PSSC or ssDSC in a HTPV,

there are a variety of other applications of a semitransparent perovskite-solar cell or a

semitransparent ssDSC. Almost all PSSCs and ssDSCs are built on a FTO-glass

transparent bottom electrode and use an evaporated metal layer for the hole-collecting

top electrode. This architecture restricts ssDSCs and PSSCs to fabrication on glass

substrates, requires illumination from the bottom electrode and only allows for

fabrication of opaque solar cells. Development of a low-sheet resistance top electrode

that can be deposited without damaging the underlying ssDSC or PSSC removes these

restrictions and opens up the possibility of new device architectures. In a similar

architecture to a tandem device, upconverting materials can be used to expand the

photoresponse of a semitransparent solar cell by absorbing sub-bandgap light passing

through the device and re-emitting higher energy photons.124–126

By themselves,

85

semitransparent ssDSCs can be utilized in building-integrated photovoltaic windows.

Additionally, the development of a transparent top electrode allows for fabrication of

ssDSCs on opaque bottom substrates. Since ssDSCs require sintering at 450 °C, a

temperature too high for most plastics, it is difficult to build flexible ssDSCs using the

conventional architecture. With a transparent top electrode, ssDSCs can be roll-to-roll

fabricated on metal foils, allowing for easier deposition and processing, lower

materials costs and flexible devices.

Previous efforts in fabricating a transparent top contact for ssDSCs have

utilized a thin metal layer with sputtered tin-doped indium oxide (ITO),127

but this

electrode requires vacuum processing and has shown relatively high series resistance

and low transmittance since ITO must be deposited at a low temperature to avoid

damaging the organic hole transport material. Silver nanowires are a promising

candidate for a top electrode in ssDSCs due to their high transparency, low sheet

resistance and ability to be solution processed.128–133

Silver nanowire meshes have

achieved sheet resistances below 10 Ω/sq at transmittances of greater than 90% at 550

nm,134

figures of merit necessary for incorporation into large scale solar modules,

115,135 and have been incorporated in a variety of solar cells.

136–145 Typically, silver

nanowires are deposited on glass, receive post-treatment such as annealing to optimize

transmission and conductivity, and then are used as a bottom substrate for solar cell

fabrication. However, deposition of silver nanowires on top of an ssDSC requires

careful choice of deposition technique and parameters to not damage the underlying

device. As typical silver nanowire annealing temperatures (180 °C) cause rapid

ssDSC degradation, the deposited silver nanowires must obtain low sheet resistances

and high transmission without any annealing. Techniques such as fabricating the

nanowires on a separate substrate and then transferring146

using pressure and/or

temperature often results in breakage of the ssDSC substrate. Finally, the silver

nanowire electrode must make ohmic contact to the hole transport material, requiring

deposition and optimization of a interfacial layer with low contact resistance – once

again without damaging or dissolving the underlying device.

86

In this chapter, I describe the fabrication of efficient ssDSCs using a

completely solution-processed Ag NW/PEDOT:PSS electrode and analyze the optical

and electronic properties of the device. These devices show efficiencies similar to

those using an evaporated metal electrode, low series resistances, and allow for high

transmittance of below bandgap light. Such a transparent electrode should be

compatible with other solar cells employing similar device structures such as PSSCs.

5.3 Device Architecture

A schematic of the device architecture is shown in Figure 5.6, along with

scanning electron microscope (SEM) images of the device cross section. The structure

and fabrication procedure is exactly the same as a standard ssDSC except for the

omission of the evaporated silver top electrode. Instead of this electrode, a conductive

PEDOT:PSS layer (Clevios™ CPP-105D) is deposited by spin coating, followed by

spray deposition of Ag NWs (donated by Blue Nano Inc.) from methanol. The CPP-

105D formulation of PEDOT:PSS is used due to its ability to be deposited on

hydrophobic surfaces and its relatively high conductivity. The PEDOT:PSS

dispersion was diluted with isopropyl alcohol as it was found that the underlying

Spiro-OMeTAD layer is insoluble in most alcohols. Deposition of the PEDOT:PSS

layer required sonication of the dispersion before spin coating, or agglomeration of the

PEDOT:PSS in solution resulted in poor films. The CPP-105D formulation of

PEDOT:PSS was also chosen for its resistance to dissolution by solvents – once the

PEDOT:PSS layer was dried, it was not removed during subsequent nanowire

deposition. It is difficult to measure the thickness of the PEDOT:PSS layer on the

solar cell, as there is not enough contrast between the Spiro-OMETAD overlayer and

the PEDOT:PSS interfacial layer in the SEM image (Figure 5.6d). However, the spin

coating conditions used in making devices resulted in an 85-nm-thick PEDOT:PSS

layer on a glass substrate. As discussed in subsequent sections, the purpose of the

PEDOT:PSS layer is twofold: to allow for ohmic contact between the Spiro OMeTAD

overlayer and the nanowire electrode, and to decrease the series resistance caused by

lateral transport of charges between the nanowires.

87

Figure 5.6. a) Schematic diagram of semitransparent ssDSC device. The device

consists of a 400-nm-thick F:SnO2 (FTO) layer, 100-nm-thick compact TiO2 layer, 2-

μm-thick dye-sensitized active layer, 200-nm-thick Spiro-OMeTAD overlayer,

approximately 85-nm-thick PEDOT:PSS layer and solution deposited silver

nanowires. b) SEM micrograph of semitransparent ssDSC cross section at 20° angle

of incidence. c) SEM image of Ag NW/PEDOT:PSS electrode at normal incidence. d)

SEM image of the PEDOT:PSS/Ag NW composite electrode at 3° angle of incidence,

showing that the wires are embedded in the PEDOT:PSS layer.

Spray deposition of Ag NWs has been shown to produce uniform films of

silver nanowires.147

Here Ag NWs were deposited out of a methanol solution directly

onto the PEDOT:PSS layer using a custom-built spray deposition system with a

nitrogen gas driven atomizer nozzle. Solvent damage to the underlying Spiro-

OMETAD layer can be eliminated during wire deposition by carefully choosing key

deposition parameters (such as nozzle height, Ag NW solution flow rate and N2

pressure) to minimize the amount of liquid solvent that reaches the device surface.

Methanol was chosen as the solvent for the Ag NWs for its low boiling point - further

reducing the amount of solvent reaching the device. While Spiro-OMeTAD and

PEDOT:PSS are only minimally soluble in alcohols, solvent on the device may

remove additives in the Spiro-OMeTAD, causing damage to the device. Heating

during spray deposition was found to degrade the ssDSC and so spray deposition of

the Ag NWs was done at room temperature. Ag NW mesh density (and thus

transmission and conductivity) is easily controlled by varying the concentration of

wires in solution. As can be seen in Figure 5.6c, the silver nanowires form a sparse,

uniform, well-dispersed mesh with maximum wire-to-wire gaps on the order of 1-2

µm. At higher magnification, it can be seen that the spray-deposited wires actually

88

appear to embed into the PEDOT:PSS layer (Figure 5.6d), likely due to partial

solvation of the PEDOT:PSS layer during spray deposition.

5.4 Role of PEDOT:PSS Layer

The main purpose of the PEDOT:PSS interfacial layer is to ensure ohmic

contact between the Spiro-OMeTAD and silver nanowires. As shown in Figure 5.7a,

ssDSCs fabricated with no PEDOT:PSS layer display ‘s-shape’ current-voltage

characteristics, typically indicative of a barrier to charge transport.148

Photoelectron

spectroscopy in air (PESA) reveals that the work function of the as-sprayed nanowires

is 4.5 eV (Figure 5.8a), while the work function of Spiro-OMeTAD is 5.2 eV (ssDSC

energy level diagram depicted in Figure 5.7b). In addition, the PESA spectrum of a

sprayed Ag NW film (Figure 5.8a) displayed a two-slope spectrum indicative of a 2nd

material in addition to the silver; this material may be some sort of leftover surfactant,

although the wires were said to have little to no surfactant present when received from

Blue Nano Inc. It is hypothesized that this energy level discrepancy is responsible for

the energetic barrier seen in ssDSCs without an interfacial layer. The work function of

PEDOT:PSS was measured to be 5.0 eV, corresponding to a better energetic match

with Spiro-OMeTAD.

89

Figure 5.7. a) Current-voltage characteristics of ssDSC sensitized with D35 with no

PEDOT:PSS interfacial layer. The J-V shows a distinctive ‘s-shape’ which causes a

low fill factor (FF) and efficiency (Eff). b) Energy level diagram of ssDSC. The work

functions of the Spiro-OMeTAD, PEDOT:PSS and Ag NWs was measured by PESA,

while other energy levels are approximate and shown for comparison.

90

Figure 5.8. PESA measurement of a) sprayed Ag NW film (from ethanol) on glass, b)

Spiro-OMeTAD on glass, and c) PEDOT:PSS (Clevios™ CPP-105) on glass. The

work functions are measured to be 4.5 eV, 5.2 eV, and 5.0 eV, respectively. Red lines

are fits to the baselines and sloped regions of the curves. A power number of 0.5 was

used for the metallic nanowires, while 0.3 was used for organic materials.

Figure 5.9. a) PESA measurement of sprayed Ag NW film (from ethanol) on glass

(same same as Figure 5.8) after 10 minutes of UV-ozone treatment (in 1-minute

intervals). The work function is measured to be 4.9 eV (up from 4.5 eV), and doesn’t

display a second slope. b) Current-voltage characteristics of semitransparent ssDSCs

using Z907 dye without an interfacial PEDOT:PSS layer (utilizing only a Ag NW

mesh as top electrode). J-V curves are shown after exposure to UV-ozone treatment

for a given period of time. The J-V curve begins with an s-shape indicative of a

barrier to charge transport, but shows rectifying J-V characteristics typical of a solar

cell after 6-14 minutes of UV-ozone treatment.

91

While the as-sprayed wires showed a work function of only 4.5 eV, it was

found that application of UV-ozone treatment for a short period of time increased the

measured work function of the nanowire mesh. By applying UV-ozone for ten 1-

minute increments, it was seen that the work function of the mesh increases to 4.9 eV

(Figure 5.9a). The reason for this increase is relatively unclear, but it may be

conjectured that either surfactant is removed from the surface or a new material,

perhaps Ag2O, forms on the surface of the nanowires.

Applying UV-ozone treatment, in the same fashion, to ssDSCs utilizing only

sprayed Ag NWs as a top contact rectifies the s-kink that was seen upon spray-

depositing the wires (Figure 5.9b). Since UV-ozone treatment appears to modify the

work function of Ag NWs toward that of Spiro-OMeTAD, it stands to reason that this

effect is responsible for the better performance of UV-ozoned ssDSCs utilizing Ag

NWs as a top contact. This suggests that a good energetic match between the top

contact and Spiro-OMeTAD is necessary to avoid a barrier to charge transport and an

s-shaped J-V characteristic. Thus the disappearance of the ‘s-shape’ current-voltage

characteristics observed when using an interfacial PEDOT:PSS layer may be

attributed to the better work function match between PEDOT:PSS and Spiro-

OMeTAD than between Spiro-OMeTAD and the as-sprayed silver nanowires.

A second important function of the PEDOT:PSS layer is to facilitate lateral

charge transport in between the nanowires. While a 4-point probe measurement

quantifies the sheet resistance of charge transport through the nanowire mesh,

photogenerated charges in the gaps between wires must first travel laterally to a

nearby Ag NW before they can exit the device. As seen in Figure 5.6c, this distance

can be on the order of 500 nm, meaning for low conductivity materials, the additional

series resistance caused by lateral transport of charges to the nearest wire can be very

significant. For very low conductivity materials, such as zinc oxide nanoparticles, it

has even been seen that current is not collected far from the nanowires themselves,

resulting in ‘dead spots’ of low current.149

In the case of more moderate

conductivities, the resistance of lateral transport to the nanowire mesh results in

additional solar cell series resistance causing a loss in fill factor. In the regime where

92

the resistance of lateral transport does not cause ‘dead spots’, the photogenerated

current can be assumed to be constant, and the power loss due to lateral transport can

then be calculated for a given geometry of nanowires. For a 1D mesh of thin

nanowires each evenly spaced a distance w apart, the power lost per unit area can be

analytically calculated as

sRwJ 22

12

1, (5.6)

Where J is the photogenerated current density, and Rs is the sheet resistance of the

material the charges travel through (units of Ω/square). Equation 5.6 allows for an

estimate of the required interfacial material sheet resistance in order to have negligible

losses due to lateral charge transport to the nanowires. Using a photogenerated current

density, J, of 7 mA/cm2 and a distance, w, of 0.5 µm between nanowires, a sheet

resistance, Rs, of approximately 1×109 Ω/square is required to ensure resistive losses

are less than 0.1 mW/cm2 (or a 0.1% decrease in efficiency). This corresponds to the

sheet resistance of a 500-nm-thick layer of Spiro-OMeTAD (conductivity of 2×10-5

Ω-

1cm

-1)150

, hence if Spiro-OMeTAD made ohmic contact to the Ag NWs, the losses in

efficiency due to lateral transport would be relatively small: on the order of 0.1

mW/cm2. However, for sparser nanowire meshes or other hole transport materials

with lower conductivity (conducting polymers or small molecules with low doping

density)150,151

these losses could be significantly higher. Hence, adding a highly

conductive interfacial layer such as PEDOT:PSS (85 nm of Clevios™ CPP-105D

PEDOT:PSS was measured to have a sheet resistance on the order of 105 Ω/square)

will help eliminate any resistive losses due to lateral transport of photogenerated

charge to the nanowires.

5.5 Electrode Characterization

The transmittance of the PEDOT:PSS/Ag NW composite electrode is shown in Figure

5.10 along with an image of the electrode on glass. The electrode has a transmittance

of over 90% between 450 and 700 nm and remains above 86% out to 1100 nm. This

high transmittance is necessary for applications in tandem photovoltaics where the low

bandgap device may absorb photons until 1100-1200nm. The as-sprayed Ag NW

93

mesh shows a peak transmittance of approximately 94% at a sheet resistance of 18

Ohms/square, some of the best figures of merit reported for Ag NW electrodes, even

with no post-treatment such as annealing. Achieving such a low sheet resistance

without any annealing134,139

or additional post-processing152

is made possible by the

minimal amount of insulating surfactant on the surface of the silver nanowires in the

formulation received from Blue Nano Inc. This allows for fabrication of conductive

electrodes without damaging the underlying ssDSC with heat treatment typically

required to remove surfactant. The sheet resistance of the PEDOT:PSS/Ag NW

electrode was also measured when deposited on a ssDSC device (fabricated on glass

with no conductive oxide), and resulted in similar values to that of the bare wires on

glass (18 Ω/square). The density of wires can easily be adjusted to decrease the sheet

resistance at the cost of transmittance. The PEDOT:PSS decreases the transmittance

by only approximately 1-2% at 550nm, but absorbs slightly more in the near infrared,

causing a 4% drop in transmittance at 1100 nm.

Figure 5.10. Transmittance of 85 nm PEDOT:PSS film, Ag NW film and Ag

NW/PEDOT:PSS composite electrode. Inset: Glass slide with Ag NW/PEDOT:PSS

composite electrode on the bottom half of the glass substrate.

5.6 Device Results

Devices were fabricated using D35, a commercially available, strongly-

absorbing organic dye (chemical structure shown in Figure 5.11b).153

As seen in

94

Figure 5.11c, the device has a bright red color due to its absorption in the green

portion of the spectrum. Utilizing other color dyes can allow harvesting of other

portions of the solar spectrum, along with desirable aesthetics for applications such as

building-integrated photovoltaics. The current-voltage characteristics of the best-

performing devices are shown in Figure 5.12, and the figures of merit are summarized

in Table 5.2. While the open-circuit voltage (Voc) and fill factor (FF) of the

semitransparent ssDSC remains comparable to that of the reference device, the short-

circuit photocurrent density (Jsc) drops slightly. The overall efficiency of the

semitransparent device (3.6% under illumination from the FTO side) remained very

close to that of the reference with an evaporated electrode (3.7%). Since record

efficiencies of over 7% have been achieved with D-π-A dyes that have similar

absorption spectra to D35, semitransparent ssDSCs using record-efficiency dyes can

achieve efficiencies close to the 7% under FTO-side illumination.25

Furthermore,

PSSCs38,41

sensitized with a strongly-absorbing perovskite could approach efficiencies

of 12% using a PEDOT:PSS/Ag NW electrode , while still being transparent to sub-

bandgap photons. Such a device would be an ideal high energy absorber in a tandem

solar cell. This can be contrasted with organic solar cells, where the weaker

absorption of the device means that semitransparent cells achieve significantly lower

efficiencies than those with a reflective back electrode.138,140,144

As can be seen in

Figure 5.11a, the external quantum efficiency, or EQE, of the semitransparent device

is nearly as high as that of the reference, particularly at the dye’s peak absorption.

Additionally, it should be noted that these devices were fabricated with a moderate-

size active area (electrode area of approximately 0.5 cm2), and utilizing a spray-

deposited semitransparent electrode resulted in a higher device yield with less shorting

than in ssDSCs fabricated with a standard evaporated silver electrode. This is likely

due to PEDOT:PSS and Ag NWs being less likely to short the device through pinholes

than the silver evaporation. As shown in Table 1, the measured series resistances of

the J-V curves were nearly identical, showing that the additional series resistance of

the silver nanowire electrode is not a significant loss. The series resistance in these

95

ssDSCs is dominated by other components, such as the internal resistance of the

device and Spiro-OMeTAD overlayer.

Figure 5.11. a) EQE of reference ssDSC device using an evaporated silver electrode

and EQE of semitransparent ssDSC (using Ag NW/PEDOT:PSS electrode)

illuminated from both the FTO and Ag NW electrodes. b) Chemical structure of D35

dye used in device fabrication. c) Picture of semitransparent ssDSC. The Ag

NW/PEDOT:PSS electrode is barely visible as a slightly darker square in the middle

of the device.

Figure 5.12 J-V curves of best semitransparent ssDSC and best reference device using

an evaporated silver electrode. The difference in current between devices is only 0.3

mA/cm2, which is slightly less than typical between a reference device and a

semitransparent device. Device area was approximately 0.5 cm2 and was masked with

a 0.2 cm2 mask. Even for small areas (0.1 cm

2), references in our lab are less than

4.0% efficient.

96

Table 5.2. Photovoltaic figures of merit for best-performing semitransparent ssDSC

and reference device using an evaporated Ag electrode under simulated AM 1.5G

illumination.

Device JSC

(mA/cm2)

VOC (V) FF Eff

(%)

Series Resistance

(Ωcm2) [a]

Reference 7.6 0.82 0.60 3.7 13.8

NW device 7.2 0.82 0.61 3.6 15.1

[a] Based on a linear fit of the J-V curve in light at a forward bias of 0.91-1.00 V.

While semitransparent devices illuminated from the F:SnO2 (FTO) electrode

showed efficiencies comparable to those of reference devices, various device

architectures, such as ssDSCs on metal substrates, would require that incoming light

be incident through the Ag NW electrode. Because incoming light is filtered by the

Spiro-OMeTAD overlayer before it can be absorbed by the dye, the EQE drops to

nearly 0 below 425 nm where Spiro-OMeTAD is strongly absorbing (Figure 5.11a).

Furthermore, there is an additional loss in photoresponse between 425-550 nm, which

can be attributed to the parasitic absorption of the oxidized Spiro-OMeTAD species in

the overlayer.32,33,52,74

Beyond 550 nm, the EQE of the semitransparent ssDSC is the

same when illuminated from either side, due to the fact that the optical performance of

the Ag NW/PEDOT:PSS electrode is comparable to that of the FTO. Overall,

illumination from the Ag NW electrode side results in approximately a 20% drop in

integrated short-circuit photocurrent, primarily due to the effects of the Spiro-

OMeTAD overlayer.

One of the primary motivations for making semitransparent ssDSCs is for use

as a top cell in tandem devices with a silicon or copper indium gallium selenide

(CIGS) low bandgap bottom cell. In such a hybrid tandem photovoltaic (HTPV),57

the

incident light would first pass through the semitransparent cell, where the high energy

photons would be absorbed, and the remaining photon flux would be absorbed by the

low bandgap inorganic solar cell. In order to achieve an efficient HTPV device, the

semitransparent top cell must strongly absorb photons with energy above its bandgap,

97

but must show a transmittance of approximately 80% or more for lower energy

photons.57

The transmittance, absorptance, and reflectance of the semitransparent

ssDSC are quantified in Figure 5.13. As can be seen, a majority of the photons below

550 nm are absorbed within the device, while above the bandgap of the dye the

transmittance peaks at 74% at 820 nm, with an average transmittance of 67% between

700-1100nm. A significant amount of this loss in the infrared portion of the spectrum

can be attributed to the strong absorption of the soda-lime glass substrate, which

absorbs approximately 18% of incident 1100 nm photons. The actual absorptance of

the semitransparent ssDSC (every layer other than glass substrate) is only 10-22%

between 700-1100 nm, suggesting that with more transparent glass and an

antireflective coating, such a semitransparent cell could achieve the transmission

levels necessary for use in a HTPV.

Figure 5.13. Fractional spectral distribution of incident light upon the semitransparent

ssDSC. Shown is the fraction of photons transmitted through the semitransparent

ssDSC (all layers other than glass substrate, including the FTO), the fraction absorbed

by the semitransparent ssDSC, and the fraction absorbed by the soda-lime glass

substrate. The light area at the top of the plot denotes reflected photons.

Transmittance and absorptance measurements were carried out using an integrating

sphere to account for scattering.

5.7 Conclusions

98

Transparent top electrodes for ssDSCs have a variety of applications ranging

from tandem devices to photovoltaic windows. We have developed a method to spray

a silver nanowire/PEDOT:PSS electrode which achieves greater than 92% peak

transmittance without damaging the underlying ssDSC. The use of a PEDOT:PSS

interfacial layer allows the Ag NWs to make ohmic contact to the Spiro-OMeTAD

overlayer. Additionally, the PEDOT:PSS layer eliminates any contribution to the

series resistance caused by lateral transport of charges between nanowires.

Combining this electrode with a D35-sensitized ssDSC, we have been able to make an

aesthetically-appealing, transparent device with an efficiency of 3.6% - only slightly

less than the efficiency of standard D35 ssDSCs using an evaporated silver electrode.

The semitransparent device shows high transmission below the device bandgap, a

requirement for fabrication of efficient hybrid tandem photovoltaics. Development of

such highly transparent Ag NW-based electrodes for ssDSCs and PSSCs are an

important step in the realization of HTPV devices that can exceed 20% efficiency and

provide clean energy at competitive prices.

5.8 Future Outlook

Using the methodology of Section 1, the efficiency of a HTPV utilizing a

perovskite-sensitized solar cell as an absorber can be modeled. Using the absorption

and external quantum efficiency of the published record PSSC device, the absorption

and efficiency of the perovskite solar cell utilizing a silver nanowire electrode can be

modeled out.118

Because light is now only able to make one pass through the device,

the efficiency is lowered from 15% to 12.1% due to less absorption in the red portion

of the spectrum. The transmittance of a semi-transparent perovskite cell was then

calculated by: 1) adding another 10% absorptance to the total absorptance of the

modeled one-pass device through the material below 800 nm, and 2) taken to be the

transmittance of the semi-transparent device shown in Figure 5.13 whenever this

exceeds the absorptance of the modeled one-pass device. By ‘cobbling together’ the

modeled absorptance of the perovskite solar cell and the silver-nanowire electrode on

a ssDSC, we are able to understand how a perovskite device would currently perform

99

using the electrode that has already been developed. Modeling the transmittance and

the EQE of the underlying CIGS device results in the fractional spectral distribution

shown in Figure 5.14.

Figure 5.14. Modeled fractional spectral distribution of record PSSC using the silver

nanowire electrode developed in this chapter, along with a CIGS solar cell as the

bottom cell in a HTPV.

Table 5.3. Figures of merit of modeled PSSC-CIGS HTPV using a silver nanowire

electrode. The total device efficiency in a 4-terminal configuration is 19.2%

Device JSC

(mA/cm2)

VOC (V) FF Eff

(%)

PSSC 16.7 0.99 0.73 12.1

NW device 16.3 0.605 0.72 7.1

From here the transmittance can be used to calculate the incident spectal flux

on the bottom CIGS cell and calculate the expected efficiency of the tandem device,

resulting in a 4-terminal efficiency of 19.2% (Table 5.3). It should be noted that this is

the expected efficiency had such a device been made today, and with improvements to

both the electrode and PSSCs, even higher efficiencies can be achieved. While there

is still quite a bit of engineering required to make such a device, PSSC compatible

100

silver nanowire electrodes open up the possibility of greater than 20% efficiency

HTPVs.

5.9 Experimental Details

ssDSC Device Fabrication: FTO substrates (TEC15, Hartford Glass Co.) were

cleaned by sonication in Extran® detergent, acetone and isopropanol, with subsequent

UV-ozone treatment for 20 minutes. A compact TiO2 layer (50-100 nm) was

deposited using spray pyrolysis of titanium diisopropoxide bis(acetylacetonate)

(Aldrich 75 weight % in isopropanol, diluted 10x with isopropanol) on the FTO

substrate. The mesoporous titania layer was deposited by doctorblading titania

nanoparticles dispersed in paste. Dyesol paste (NR-18T) diluted with terpinoel 1:1 (by

weight) was used, resulting in nanoparticle films with thicknesses of approximately

2.2 µm. Films were subsequently sintered at 500° C for 30 minutes on a hotplate.

Titania nanoparticle films were then immersed in TiCl4 solution overnight and then

heated to 500° C for 30 minutes once again. The substrates were then sensitized by

immersion for 18 hours in a 0.2 mM solution of D35 dye (Dyenamo) in ethanol.

Spiro-OMeTAD solution contained tert-butylpyridine (4-tbp), Spiro-OMeTAD

(Luminescence Technology corporation), and Lithium

bis(trifluoromethylsulfonyl)imide salt (Li-TFSI) (pre-solved in acetonitrile in

170mg/mL). Spiro solution was made by taking a 1g Spiro-OMeTAD: 97 mL 4-tbp:

208 mL Li-TSFI solution mixture dissolved in chlorobenzene (with 225 mg Spiro-

OMeTAD/1 mL chlorobenzene). The Spiro-OMeTAD solution was then infiltrated

into the sensitized nanoparticle film by spincoating at 2000 rpm. After allowing the

device to dry for 18 hours in a dry air environment, the top electrode was deposited.

The electrode area was first masked off by polyimide tape and the tape removed after

electrode deposition. In the case of the evaporated silver reference device, a 200nm

silver cathode was deposited by thermal evaporation (10-6

torr pressure). In the case

of the semitransparent device, the electrode was deposited as described below.

Ag NW/PEDOT:PSS Electrode Fabrication: PEDOT:PSS from Heraeus (formulation

Clevios™ CPP-105D) was diluted 1:1 (by volume) with isopropyl alcohol and

101

deposited by spincoating at 4000 rpm on the device or film. The device/film was then

allowed to dry in a dry air environment for 5 hours. Silver nanowires, 10-30 μm long

and 30-50 nm in diameter were provided by Blue Nano Inc. in a 10mg/mL ethanol

solution. This solution was diluted to 5mg Ag/mL ethanol for long term storage and

further diluted with methanol to 0.11 mg/mL for spray deposition. 45 mL of this

solution was loaded into a Harvard apparatus PhD Ultra syringe pump and delivered to

a 1/4JN-SS+SU11DF-SS atomizing nozzle by Spray Systems Co. at a rate of 4

mL/min. Nitrogen gas at 25psi was delivered to the nozzle. The nozzle was positioned

110 mm above a motorized, computer controlled X-Y stage (Parker Hannafin

ProMECH LP28) onto which the ssDSC devices were taped. To ensure uniformity and

repeatability of the deposited electrodes, the X-Y stage moved the devices under the

nozzle at 20cm/s during the entire spray process such that the nozzle sprayed

uniformly over a 20x5.5cm area. This way 10 devices at a time received identical

nanowire coatings.

4-point Probe Measurements: Electrode sheet resistance was measured using an in-

line four-point probe with 1mm tip spacing by Jandel Engineering connected to a

Kiethley 2400 sourcemeter. Each measurement consisted of recording 100 pairs of

current and voltage data points between -30 and 30mV and checking that each pair

corresponded to the same sheet resistance. While the sheet resistance of the silver

nanowire mesh on glass was measured to be 18 Ω/square, the corresponding sheet

resistance of the composite PEDOT:PSS/Ag NW electrode on glass was unable to be

measured with a 4-point-probe (readings of 4-8 kΩ/square). However, 4-point-probe

measurements of the Ag NW/PEDOT:PSS electrode on a ssDSC (fabricated without

an FTO layer to avoid shorted measurement of the FTO sheet resistance) found sheet

resistances of approximately 18 Ω/square.

J-V Measurements: All devices were subject to 5 minutes light soaking under

simulated AM 1.5G spectrum before measurement. J-V curves were taken with a

Keithley 2400 sourcemeter, under simulated AM 1.5G illumination using a

Spectraphysics model 91160 solar simulator callibrated with a hamamatsu Si

102

photodiode with KG5 filter. The ssDSC was was masked with a 0.2 cm2 area

machined mask during J-V measurements to ensure accurate illumination area.

EQE, Transmittance, and Absorptance Measurements: External quantum efficiency

measurements were performed at a chopping rate of 40 Hz with a white light

illumination bias of approximately 0.4 suns applied using a white light LED array

powered by a DC power source. A Newport Apex monochromator illuminator was

used (in conjunction with a Princeton Instruments monochromator and a filter wheel)

to generate the monochromated, chopped signal. The current response of the ssDSC

was put through a 1000 Ω transimpedance amplifier and recorded using a Stanford

Instruments lock-in amplifier. Calibration of the EQE measurement was performed

using a calibrated photodiode of known EQE. Additionally, the monochromated EQE

signal was split with a 50:50 beam splitter into a 2nd

‘reference’ photodiode connected

to another Stanford Instruments lock-in amplifier that was used to correct for any

fluctuations in the EQE beam source intensity.

Absorptance and transmittance measurements were performed using the same

light source/monochromator as the EQE measurements and measured using an

integrating sphere with an attached silicon photodiode. For measuring transmittance,

the device/film was placed in front of the integrating sphere such that only light

passing through the device/film would be measured. For absorptance, the device was

placed in the center of the integrating sphere, allowing for a measurement of the sum

of the transmittance and reflectance. The reflectance was then calculated by 1-

(Transmittance+Absorptance).

SEM images: Device images were taken using a FEI XL30 Sirion SEM. The ssDSC

was cracked in half and placed on SEM sample mounts of various tilt angle (0°, 20°,

90°) using graphite paste.

PESA Measurements: Samples for PESA measurements were prepared using identical

spin coating or spraying conditions as during device fabrication but on a glass

substrate. PESA spectra were measured using a Riken Keiki AC-2 photoelectron

spectrometer.

103

6 Conclusions and Future Outlook

Over the past 5 years, the goal of researchers working on DSCs has moved

away from just trying to understand the mechanisms of such devices toward achieving

efficiencies that could render DSCs commercializable. This has been fueled by the

huge drop in prices of traditional silicon photovoltaic modules, and has added a sense

of urgency to research in the field. This sense of urgency has led to new creative

breakthroughs that have pushed DSC and ssDSC efficiencies forward: new redox

couples,4,9,18–20

new paradigms in dye design,4,35,154

and the recent invention of the

PSSC.38

My research has been aimed at developing understanding and device

architectures that can help push the efficiencies of DSCs and ssDSCs toward those

needed for commercialization.

The discovery of PSSCs in 2012 has been a huge turning point in the field,

with many research groups moving away from traditional DSCs and ssDSCs to the

newly discovered perovskite solar cells. With efficiencies already past 15% less than

a year later,118

it is not hard to suppose that such devices could reach efficiencies

approaching 20%, equal to the record lab attained efficiency of CIGS thin film solar

cells and surpassing that of CdTe. Such devices by themselves would be of interest to

industry and I would suspect that many companies will begin working on trying to

commercialize PSSC technology soon. In addition, using such cells in HTPVs can

lead to efficiencies near 25%, which would significantly decrease installation costs

and help make the technology competitive with traditional silicon solar cells. I

suspect that these will be the two most important directions of research going forward,

and are the most likely to lead to a commercialized product. As the solar industry is

quickly becoming larger and maturing, new technologies such as DSCs, PSSCs and

HTPVs need to quickly achieve efficiencies at least as high as more developed

inorganic solar cells. Otherwise, the advantages of incumbent technologies such as

reliability and economies of scale will prevent such new technologies from gaining a

foothold in the greater solar market, and relegate them to niche applications.

Finally, one of the most important research directions will be the

understanding of degradation mechanisms of PSSCs and DSCs in order to achieve the

104

necessary lifetimes (25+ years) needed for device commercialization. At this point,

there is very little data on the degradation of PSSCs and even less on the causes.

Research aimed at understanding device degradation and then improving lifetimes

through materials research will be paramount to the future success of such devices.

105

7 Copyright

I have previously published material from this thesis in peer-reviewed journals

and it is reproduced here with permission from the publishers. Chapter 3 was adapted

from a publication in Advanced Energy Materials,52

and reproduced here with

permission from Wiley and Sons. Chapter 4 is based on work published in Physical

Chemistry Chemical Physics,54

and reproduced by permission of the PCCP Ownership

Societies. Chapter 5 is reproduced from work published in Advanced Energy

Materials,58

with permission from Wiley and Sons.

106

8 References

1. O’Regan, B. & Grätzel, M. A low-cost, high-efficiency solar cell based on dye-

sensitized colloidal TiO2 films. Nature 353, 737–740 (1991).

2. O’Regan, B. C., Durrant, J. R., Sommeling, P. M. & Bakker, N. J. Influence of

the TiCl4 Treatment on Nanocrystalline TiO2 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).

3. Sommeling, P. M. et al. Influence of a TiCl4 post-treatment on nanocrystalline

TiO2 films in dye-sensitized solar cells. J. Phys. Chem. B 110, 19191–7 (2006).

4. Yella, A. et al. Porphyrin-sensitized solar cells with cobalt (II/III)-based redox

electrolyte exceed 12 percent efficiency. Science 334, 629–34 (2011).

5. Zhang, Q., Dandeneau, C. S., Zhou, X. & Cao, G. ZnO Nanostructures for Dye-

Sensitized Solar Cells. Adv. Mater. 21, 4087–4108 (2009).

6. Zhang, Q. & Cao, G. Nanostructured photoelectrodes for dye-sensitized solar

cells. Nano Today 6, 91–109 (2011).

7. Nazeeruddin, M. K. et al. Combined experimental and DFT-TDDFT

computational study of photoelectrochemical cell ruthenium sensitizers. J. Am.

Chem. Soc. 127, 16835–47 (2005).

8. Nazeeruddin, M. K. et al. Conversion of light to electricity by cis-X2bis(2,2’-

bipyridyl-4,4'-dicarboxylate)ruthenium(II) charge-transfer sensitizers (X = Cl-,

Br-, I-, CN-, and SCN-) on nanocrystalline titanium dioxide electrodes. J. Am.

Chem. Soc. 115, 6382–6390 (1993).

9. Yum, J.-H. et al. A cobalt complex redox shuttle for dye-sensitized solar cells

with high open-circuit potentials. Nat. Commun. 3, 631 (2012).

10. Bessho, T., Zakeeruddin, S. M., Yeh, C.-Y., Diau, E. W.-G. & Grätzel, M.

Highly efficient mesoscopic dye-sensitized solar cells based on donor-acceptor-

substituted porphyrins. Angew. Chem. Int. Edit. 49, 6646–9 (2010).

11. Li, L.-L. & Diau, E. W.-G. Porphyrin-sensitized solar cells. Chem. Soc. Rev. 42,

291–304 (2013).

107

12. Mishra, A., Fischer, M. K. R. & Bäuerle, P. Metal-free organic dyes for dye-

sensitized solar cells: from structure: property relationships to design rules.

Angew. Chem. Int. Edit. 48, 2474–99 (2009).

13. Boschloo, G. & Hagfeldt, A. Characteristics of the iodide/triiodide redox

mediator in dye-sensitized solar cells. Accounts Chem. Res. 42, 1819–26

(2009).

14. Koops, S. E., O’Regan, B. C., Barnes, P. R. F. & Durrant, J. R. Parameters

influencing the efficiency of electron injection in dye-sensitized solar cells. J.

Am. Chem. Soc. 131, 4808–18 (2009).

15. Raga, S. R., Barea, E. M. & Fabregat-Santiago, F. Analysis of the Origin of

Open Circuit Voltage in Dye Solar Cells. J. Phys. Chem. Lett. 3, 1629–1634

(2012).

16. Barea, E. M. et al. Energetic factors governing injection, regeneration and

recombination in dye solar cells with phthalocyanine sensitizers. Energ.

Environ. Sci. 3, 1985 (2010).

17. Cong, J., Yang, X., Kloo, L. & Sun, L. Iodine/iodide-free redox shuttles for

liquid electrolyte-based dye-sensitized solar cells. Energ. Environ. Sci. 5, 9180

(2012).

18. Daeneke, T. et al. High-efficiency dye-sensitized solar cells with ferrocene-

based electrolytes. Nature Chem. 3, 211–15 (2011).

19. Feldt, S. M. et al. Design of organic dyes and cobalt polypyridine redox

mediators for high-efficiency dye-sensitized solar cells. J. Am. Chem. Soc. 132,

16714–24 (2010).

20. Daeneke, T. et al. Dye regeneration kinetics in dye-sensitized solar cells. J. Am.

Chem. Soc. 134, 16925–8 (2012).

21. Wu, M. & Ma, T. Platinum-free catalysts as counter electrodes in dye-sensitized

solar cells. ChemSusChem 5, 1343–57 (2012).

22. Grätzel, M. Dye-sensitized solar cells. J. Photoch. Photobio. C 4, 145–153

(2003).

23. Hagfeldt, A., Boschloo, G., Sun, L., Kloo, L. & Pettersson, H. Dye-sensitized

solar cells. Chem. Rev. 110, 6595–663 (2010).

24. Hardin, B. E., Snaith, H. J. & McGehee, M. D. The renaissance of dye-

sensitized solar cells. Nat. Photonics 6, 162–169 (2012).

108

25. Burschka, J. et al. Tris(2-(1H-pyrazol-1-yl)pyridine)cobalt(III) as p-type dopant

for organic semiconductors and its application in highly efficient solid-state

dye-sensitized solar cells. J. Am. Chem. Soc. 133, 18042–5 (2011).

26. Harms, H. a, Tétreault, N., Gusak, V., Kasemo, B. & Grätzel, M. In situ

investigation of dye adsorption on TiO2 films using a quartz crystal

microbalance with a dissipation technique. Phys. Chem. Chem. Phys. 14, 9037–

40 (2012).

27. Gusak, V., Heiniger, L.-P., Graetzel, M., Langhammer, C. & Kasemo, B. Time-

resolved indirect nanoplasmonic sensing spectroscopy of dye molecule

interactions with dense and mesoporous TiO2 films. Nano Lett. 12, 2397–403

(2012).

28. Ding, I.-K. et al. Deposition of hole-transport materials in solid-state dye-

sensitized solar cells by doctor-blading. Org. Electron. 11, 1217–1222 (2010).

29. Docampo, P. et al. Pore Filling of Spiro-OMeTAD in Solid-State Dye-

Sensitized Solar Cells Determined Via Optical Reflectometry. Adv. Funct.

Mater. 22, 5010–5019 (2012).

30. Ding, I.-K. et al. Pore-Filling of Spiro-OMeTAD in Solid-State Dye Sensitized

Solar Cells: Quantification, Mechanism, and Consequences for Device

Performance. Adv. Funct. Mater. 19, 2431–2436 (2009).

31. Melas-Kyriazi, J. et al. The Effect of Hole Transport Material Pore Filling on

Photovoltaic Performance in Solid-State Dye-Sensitized Solar Cells. Adv.

Energy Mater. 1, 407–414 (2011).

32. Abate, A. et al. Lithium salts as “redox active” p-type dopants for organic

semiconductors and their impact in solid-state dye-sensitized solar cells. Phys.

Chem. Chem. Phys. 15, 2572–9 (2013).

33. Cappel, U. B., Daeneke, T. & Bach, U. Oxygen-induced doping of spiro-

MeOTAD in solid-state dye-sensitized solar cells and its impact on device

performance. Nano Lett. 12, 4925–31 (2012).

34. Dualeh, A. et al. Influence of Donor Groups of Organic D−π–A Dyes on Open-

Circuit Voltage in Solid-State Dye-Sensitized Solar Cells. J. Phys. Chem. C

116, 1572–1578 (2012).

35. Nguyen, W. H. et al. Molecular Engineering of Organic Dyes for Improved

Recombination Lifetime in Solid-State Dye-Sensitized Solar Cells. Chem.

Mater. 25, 1519–1525 (2013).

109

36. Hsu, C.-Y., Chen, Y.-C., Lin, R. Y.-Y., Ho, K.-C. & Lin, J. T. Solid-state dye-

sensitized solar cells based on spirofluorene (spiro-OMeTAD) and arylamines

as hole transporting materials. Phys. Chem. Chem. Phys. 14, 14099–109 (2012).

37. Kim, H.-S. et al. Lead Iodide Perovskite Sensitized All-Solid-State Submicron

Thin Film Mesoscopic Solar Cell with Efficiency Exceeding 9%. Sci. Rep. 2, 1–

7 (2012).

38. Lee, M. M., Teuscher, J., Miyasaka, T., Murakami, T. N. & Snaith, H. J.

Efficient Hybrid Solar Cells Based on Meso-Superstructured Organometal

Halide Perovskites. Science 338, 643–647 (2012).

39. Kojima, A., Teshima, K., Shirai, Y. & Miyasaka, T. Organometal halide

perovskites as visible-light sensitizers for photovoltaic cells. J. Am. Chem. Soc.

131, 6050–1 (2009).

40. Im, J.-H., Lee, C.-R., Lee, J.-W., Park, S.-W. & Park, N.-G. 6.5% Efficient

Perovskite Quantum-Dot-Sensitized Solar Cell. Nanoscale 3, 4088–93 (2011).

41. Noh, J. H., Im, S. H., Heo, J. H., Mandal, T. N. & Seok, S. Il. Chemical

management for colorful, efficient, and stable inorganic-organic hybrid

nanostructured solar cells. Nano Lett. 13, 1764–9 (2013).

42. Cai, J., Satoh, N. & Han, L. Injection Efficiency in Dye-Sensitized Solar Cells

within a Two-Band Model. J. Phys. Chem. C 115, 6033–6039 (2011).

43. Kashif, M. K. et al. A New Direction in Dye-Sensitized Solar Cells Redox

Mediator Development: In Situ Fine-Tuning of the Cobalt(II)/(III) Redox

Potential through Lewis Base Interactions. J. Am. Chem. Soc. 134, 16646–53

(2012).

44. Lim, J., Kwon, Y. S. & Park, T. Effect of coadsorbent properties on the

photovoltaic performance of dye-sensitized solar cells. Chem. Comm. 47, 4147–

9 (2011).

45. Meng, S. & Kaxiras, E. Electron and hole dynamics in dye-sensitized solar

cells: influencing factors and systematic trends. Nano Lett. 10, 1238–47 (2010).

46. Barnes, P. R. F. et al. Interpretation of optoelectronic transient and charge

extraction measurements in dye-sensitized solar cells. Adv. Mater. 25, 1881–

922 (2013).

47. Fabregat-Santiago, F., Garcia-Belmonte, G., Mora-Seró, I. & Bisquert, J.

Characterization of nanostructured hybrid and organic solar cells by impedance

spectroscopy. Phys. Chem. Chem. Phys. 13, 9083–118 (2011).

110

48. SEIA. U.S. Solar Market Insight Q1 2013. (2013).

49. Shrotriya, V. et al. Accurate Measurement and Characterization of Organic

Solar Cells. Adv. Funct. Mater. 16, 2016–2023 (2006).

50. Snaith, H. J. How should you measure your excitonic solar cells? Energ.

Environ. Sci. 5, 6513 (2012).

51. Schmidt-Mende, L., akeeruddin, S. M. & Grätzel, M. Efficiency improvement

in solid-state-dye-sensitized photovoltaics with an amphiphilic Ruthenium-dye.

Appl. Phys. Lett. 86, 013504 (2005).

52. Margulis, G. Y., Hardin, B. E., Ding, I.-K., Hoke, E. T. & McGehee, M. D.

Parasitic Absorption and Internal Quantum Efficiency Measurements of Solid-

State Dye Sensitized Solar Cells. Adv. Energy Mater. 3, 959–966 (2013).

53. Cai, N. et al. An organic D-π-A dye for record efficiency solid-state sensitized

heterojunction solar cells. Nano Lett. 11, 1452–6 (2011).

54. Margulis, G. Y. et al. Highly soluble energy relay dyes for dye-sensitized solar

cells. Phys. Chem. Chem. Phys. 15, 11306–12 (2013).

55. Hardin, B. E. et al. Increased light harvesting in dye-sensitized solar cells with

energy relay dyes. Nat. Photonics 3, 406–411 (2009).

56. Snaith, H. J. Estimating the Maximum Attainable Efficiency in Dye-Sensitized

Solar Cells. Adv. Funct. Mater. 20, 13–19 (2010).

57. Beiley, Z. M. & McGehee, M. D. Modeling low cost hybrid tandem

photovoltaics with the potential for efficiencies exceeding 20%. Energ.

Environ. Sci. 5, 9173 (2012).

58. Margulis, G. Y. et al. Spray Deposition of Silver Nanowire Electrodes for

Semitransparent Solid-State Dye-Sensitized Solar Cells. Adv. Energy Mater.

(2013). doi:10.1002/aenm.201300660

59. Wang, Q., Moser, J.-E. & Grätzel, M. Electrochemical impedance spectroscopic

analysis of dye-sensitized solar cells. J. Phys. Chem. B 109, 14945–53 (2005).

60. Kalyanasundaram, K. Dye-sensitized Solar Cells. (EFPL Press, 2010).

61. Bach, U. et al. Solid-state dye-sensitized mesoporous TiO2 solar cells with high

photon-to-electron conversion efficiencies. Nature 395, 583–585 (1998).

111

62. Wenger, S. et al. Coupled Optical and Electronic Modeling of Dye-Sensitized

Solar Cells for Steady-State Parameter Extraction. J. Phys. Chem. C 115,

10218–10229 (2011).

63. Snaith, H. J. et al. Charge collection and pore filling in solid-state dye-

sensitized solar cells. Nanotechnology 19, 424003 (2008).

64. Schmidt-Mende, L. & Grätzel, M. TiO2 pore-filling and its effect on the

efficiency of solid-state dye-sensitized solar cells. Thin Solid Films 500, 296–

301 (2006).

65. Yanagida, M., Onozawa-Komatsuzaki, N., Kurashige, M., Sayama, K. &

Sugihara, H. Optimization of tandem-structured dye-sensitized solar cell. Sol.

Energy Mater. Sol. Cells 94, 297–302 (2010).

66. Burkhard, G. F., Hoke, E. T., Scully, S. R. & McGehee, M. D. Incomplete

exciton harvesting from fullerenes in bulk heterojunction solar cells. Nano Lett.

9, 4037–41 (2009).

67. Burkhard, G. F., Hoke, E. T. & McGehee, M. D. Accounting for interference,

scattering, and electrode absorption to make accurate internal quantum

efficiency measurements in organic and other thin solar cells. Adv. Mater. 22,

3293–7 (2010).

68. Moul , A. J. et al. Optical description of solid-state dye-sensitized solar cells.

II. Device optical modeling with implications for improving efficiency. J. Appl.

Phys. 106, 073111 (2009).

69. Chung, I., Lee, B., He, J., Chang, R. P. H. & Kanatzidis, M. G. All-solid-state

dye-sensitized solar cells with high efficiency. Nature 485, 486–489 (2012).

70. Moon, S.-J. et al. Sb 2 S 3 -Based Mesoscopic Solar Cell using an Organic

Hole Conductor. J. Phys. Chem. Lett. 1, 1524–1527 (2010).

71. Sakamoto, H., Igarashi, S., Uchida, M., Niume, K. & Nagai, M. Highly efficient

all solid state dye-sensitized solar cells by the specific interaction of CuI with

NCS groups II. Enhancement of the photovoltaic characteristics. Org. Electron.

13, 514–518 (2012).

72. Cid, J.-J. et al. Molecular cosensitization for efficient panchromatic dye-

sensitized solar cells. Angew. Chem. Int. Edit. 46, 8358–62 (2007).

73. Pettersson, L. A. A., Roman, L. S. & Inganäs, O. Modeling photocurrent action

spectra of photovoltaic devices based on organic thin films. J. Appl. Phys. 86,

487 (1999).

112

74. Fantacci, S., De Angelis, F., Nazeeruddin, M. K. & Grätzel, M. Electronic and

Optical Properties of the Spiro-MeOTAD Hole Conductor in Its Neutral and

Oxidized Forms: A DFT/TDDFT Investigation. J. Phys. Chem. C 115, 23126–

23133 (2011).

75. Snaith, H., Schmidt-Mende, L., Grätzel, M. & Chiesa, M. Light intensity,

temperature, and thickness dependence of the open-circuit voltage in solid-state

dye-sensitized solar cells. Phys. Rev. B 74, 045306 (2006).

76. Yum, J.-H. et al. Effect of coadsorbent on the photovoltaic performance of zinc

pthalocyanine-sensitized solar cells. Langmuir 24, 5636–40 (2008).

77. Ragoussi, M.-E. et al. Carboxyethynyl anchoring ligands: a means to improving

the efficiency of phthalocyanine-sensitized solar cells. Angew. Chem. Int. Edit.

51, 4375–8 (2012).

78. Marinado, T. et al. Surface Molecular Quantification and Photoelectrochemical

Characterization of Mixed Organic Dye and Coadsorbent Layers on TiO 2 for

Dye-Sensitized Solar Cells. J. Phys. Chem. C 114, 11903–11910 (2010).

79. Wang, M. et al. Surface Design in Solid-State Dye Sensitized Solar Cells:

Effects of Zwitterionic Co-adsorbents on Photovoltaic Performance. Adv.

Funct. Mater. 19, 2163–2172 (2009).

80. Kwon, Y. S. et al. Reduced charge recombination by the formation of an

interlayer using a novel dendron coadsorbent in solid-state dye-sensitized solar

cells. RSC Adv. 2, 3467 (2012).

81. Krüger, J., Bach, U. & Grätzel, M. Modification of TiO2 Heterojunctions with

Benzoic Acid Derivatives in Hybrid Molecular Solid-State Devices. Adv.

Mater. 12, 447–451 (2000).

82. Neale, N. R., Kopidakis, N., van de Lagemaat, J., Grätzel, M. & Frank, A. J.

Effect of a coadsorbent on the performance of dye-sensitized TiO2 solar cells:

shielding versus band-edge movement. J. Phys. Chem. B 109, 23183–9 (2005).

83. Ding, I.-K. et al. Plasmonic Dye-Sensitized Solar Cells. Adv. Energy Mater. 1,

52–57 (2011).

84. Chen, C. et al. Highly efficient light-harvesting ruthenium sensitizer for thin-

film dye-sensitized solar cells. ACS nano 3, 3103–9 (2009).

85. Ozawa, H., Shimizu, R. & Arakawa, H. Significant improvement in the

conversion efficiency of black-dye-based dye-sensitized solar cells by

cosensitization with organic dye. RSC Adv. 2, 3198 (2012).

113

86. Jeong, N. C. et al. Effective panchromatic sensitization of electrochemical solar

cells: strategy and organizational rules for spatial separation of complementary

light harvesters on high-area photoelectrodes. J. Am. Chem. Soc. 134, 19820–7

(2012).

87. Hardin, B. E. et al. Energy and hole transfer between dyes attached to titania in

cosensitized dye-sensitized solar cells. J. Am. Chem. Soc. 133, 10662–7 (2011).

88. Pastore, M. & Angelis, F. De. First-Principles Computational Modeling of

Fluorescence Resonance Energy Transfer in Co-Sensitized Dye Solar Cells. J.

Phys. Chem. Lett. 3, 2146–2153 (2012).

89. Shrestha, M. et al. Dual Functionality of BODIPY Chromophore in Porphyrin-

Sensitized Nanocrystalline Solar Cells. J. Phys. Chem. C 116, 10451–10460

(2012).

90. Siegers, C. et al. A dyadic sensitizer for dye solar cells with high energy-

transfer efficiency in the device. ChemPhysChem 8, 1548–56 (2007).

91. Eichberger, R. et al. Charge separation dynamics at inorganic/organic

nanostructured hybrid photovoltaic interfaces. J. Photon. Energy 2, 021003

(2012).

92. Mor, G. K. et al. High-efficiency Förster resonance energy transfer in solid-

state dye sensitized solar cells. Nano Lett. 10, 2387–94 (2010).

93. Itzhakov, S. et al. Design Principles of FRET-Based Dye-Sensitized Solar Cells

with Buried Quantum Dot Donors. Adv. Energy Mater. 1, 626–633 (2011).

94. Unger, E. L. et al. Contribution from a hole-conducting dye to the photocurrent

in solid-state dye-sensitized solar cells. Phys. Chem. Chem. Phys. 13, 20172–7

(2011).

95. Humphry-Baker, N. et al. Time-evolution of poly(3-hexylthiophene) as an

energy relay dye in dye-sensitized solar cells. Nano Lett. 12, 634–9 (2012).

96. Wu, H.-P. et al. Molecular engineering of cocktail co-sensitization for efficient

panchromatic porphyrin-sensitized solar cells. Energ. Environ. Sci. 5, 9843

(2012).

97. Kroeze, J. E. et al. Alkyl Chain Barriers for Kinetic Optimization in Dye-

Sensitized Solar Cells. J. Am. Chem. Soc. 128, 16376–83 (2006).

114

98. Schmidt-mende, L. et al. Effect of Hydrocarbon Chain Length of Amphiphilic

Ruthenium Dyes on Solid-State Dye-Sensitized Photovoltaics. Nano Lett. 5,

1315–1320 (2005).

99. Li, R., Liu, J., Cai, N., Zhang, M. & Wang, P. Synchronously reduced surface

states, charge recombination, and light absorption length for high-performance

organic dye-sensitized solar cells. J. Phys. Chem. B 114, 4461–4 (2010).

100. Yum, J.-H. et al. Panchromatic response in solid-state dye-sensitized solar cells

containing phosphorescent energy relay dyes. Angew. Chem. Int. Edit. 48,

9277–80 (2009).

101. Yum, J.-H. et al. Incorporating multiple energy relay dyes in liquid dye-

sensitized solar cells. ChemPhysChem 12, 657–61 (2011).

102. Förster, T. Zwischenmolekulare Energiewanderung und Fluoreszenz. Annalen

der Physik 437, 55–75 (1948).

103. Hoke, E. T., Hardin, B. E. & McGehee, M. D. Modeling the efficiency of

Förster resonant energy transfer from energy relay dyes in dye-sensitized solar

cells. Opt. Express 18, 3893–904 (2010).

104. Hardin, B. E. et al. High excitation transfer efficiency from energy relay dyes in

dye-sensitized solar cells. Nano Lett. 10, 3077–83 (2010).

105. Drake, J. M., Lesiecki, M. L. & Camaioni, D. M. Photophysics and cis-trans

isomerization of DCM. Chem. Phys. Lett. 113, 530–534 (1985).

106. Fabregat-Santiago, F., Bisquert, J., Garcia-Belmonte, G., Boschloo, G. &

Hagfeldt, A. Influence of electrolyte in transport and recombination in dye-

sensitized solar cells studied by impedance spectroscopy. Sol. Energy Mater.

Sol. Cells 87, 117–131 (2005).

107. Szeifert, J. M. et al. “Brick and Mortar” Strategy for the Formation of Highly

Crystalline Mesoporous Titania Films from Nanocrystalline Building Blocks.

Chem. Mater. 21, 1260–1265 (2009).

108. Szeifert, J. M., Fattakhova-Rohlfing, D., Rathouský, J. & Bein, T. Multilayered

High Surface Area “Brick and Mortar” Mesoporous Titania Films as Efficient

Anodes in Dye-Sensitized Solar Cells. Chem. Mater. 24, 659–663 (2012).

109. Lakowicz, J. R. Principles of Fluorescence Spectroscopy. (Springer, 2006).

115

110. Li, X. et al. Measured binding coefficients for iodine and ruthenium dyes;

implications for recombination in dye sensitised solar cells. Phys. Chem. Chem.

Phys. 14, 15421–8 (2012).

111. Andreu, R. et al. New one- and two-dimensional 4H-pyranylidene NLO-phores.

Tetrahedron Lett. 50, 2920–2924 (2009).

112. Oosterbaan, W. D. et al. Photoinduced Charge Separation in Cyclohexylidene-

Based Donor−(σ-Bridge)−Acceptor Compounds − Building Blocks for

Materials. Eur. J. Org. Chem. 2003, 3117–3130 (2003).

113. 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).

114. Shockley, W. & Queisser, H. J. Detailed Balance Limit of Efficiency of p-n

Junction Solar Cells. J. Appl. Phys. 32, 510 (1961).

115. Rowell, M. W. & McGehee, M. D. Transparent electrode requirements for thin

film solar cell modules. Energ. Environ. Sci. 4, 131 (2011).

116. Liska, P. et al. Nanocrystalline dye-sensitized solar cell/copper indium gallium

selenide thin-film tandem showing greater than 15% conversion efficiency.

Appl. Phys. Lett. 88, 203103 (2006).

117. Wenger, S., Seyrling, S., Tiwari, A. N. & Grätzel, M. Fabrication and

performance of a monolithic dye-sensitized TiO2/Cu(In,Ga)Se2 thin film

tandem solar cell. Appl. Phys. Lett. 94, 173508 (2009).

118. Burschka, J. et al. Sequential deposition as a route to high-performance

perovskite-sensitized solar cells. Nature 499, 316–9 (2013).

119. Chen, P. et al. High open-circuit voltage solid-state dye-sensitized solar cells

with organic dye. Nano Lett. 9, 2487–92 (2009).

120. Edri, E., Kirmayer, S., Cahen, D. & Hodes, G. High Open-Circuit Voltage Solar

Cells Based on Organic–Inorganic Lead Bromide Perovskite. J. Phys. Chem.

Lett. 4, 897–902 (2013).

121. Ito, S. et al. High-voltage (1.8V) tandem solar cell system using a

GaAs/AlXGa(1−X)As graded solar cell and dye-sensitised solar cells with

organic dyes having different absorption spectra. Sol. Energy 85, 1220–1225

(2011).

116

122. Jeong, W.-S., Lee, J.-W., Jung, S., Yun, J. H. & Park, N.-G. Evaluation of

external quantum efficiency of a 12.35% tandem solar cell comprising dye-

sensitized and CIGS solar cells. Sol. Energy Mater. Sol. Cells 95, 3419–3423

(2011).

123. Tang, Z. et al. Semi-Transparent Tandem Organic Solar Cells with 90%

Internal Quantum Efficiency. Adv. Energy Mater. 2, 1467–1476 (2012).

124. Zou, W., Visser, C., Maduro, J. A., Pshenichnikov, M. S. & Hummelen, J. C.

Broadband dye-sensitized upconversion of near-infrared light. Nat. Photonics 6,

1–5 (2012).

125. Fischer, S. et al. Enhancement of silicon solar cell efficiency by upconversion:

Optical and electrical characterization. J. Appl. Phys. 108, 044912 (2010).

126. Huang, X., Han, S., Huang, W. & Liu, X. Enhancing solar cell efficiency: the

search for luminescent materials as spectral converters. Chem. Soc. Rev. 42,

173–201 (2013).

127. Chiang, Y.-F., Tsai, C.-H., Chen, P. & Guo, T.-F. Bifacial transparent solid-

state dye-sensitized solar cell with sputtered indium-tin-oxide counter electrode.

Sol. Energy 86, 1967–1972 (2012).

128. Lee, J.-Y., Connor, S. T., Cui, Y. & Peumans, P. Solution-processed metal

nanowire mesh transparent electrodes. Nano Lett. 8, 689–92 (2008).

129. De, S. et al. Silver Nanowire Networks as Flexible, Transparent, Conducting

Films: Extremely High DC to Optical Conductivity Ratios. ACS nano 3, 1767–

74 (2009).

130. Hu, L., Kim, H. S., Lee, J., Peumans, P. & Cui, Y. Scalable coating and

properties of transparent, flexible, silver nanowire electrodes. ACS nano 4,

2955–63 (2010).

131. Zhu, R. et al. Fused silver nanowires with metal oxide nanoparticles and

organic polymers for highly transparent conductors. ACS nano 5, 9877–82

(2011).

132. Gaynor, W., Burkhard, G. F., McGehee, M. D. & Peumans, P. Smooth

nanowire/polymer composite transparent electrodes. Adv. Mater. 23, 2905–10

(2011).

133. Kim, T. et al. Uniformly Interconnected Silver-Nanowire Networks for

Transparent Film Heaters. Adv. Funct. Mater. 23, 1250–1255 (2013).

117

134. Krantz, J., Richter, M., Spallek, S., Spiecker, E. & Brabec, C. J. Solution-

Processed Metallic Nanowire Electrodes as Indium Tin Oxide Replacement for

Thin-Film Solar Cells. Adv. Funct. Mater. 21, 4784–4787 (2011).

135. Servaites, J. D., Yeganeh, S., Marks, T. J. & Ratner, M. A. Efficiency

Enhancement in Organic Photovoltaic Cells: Consequences of Optimizing

Series Resistance. Adv. Funct. Mater. 20, 97–104 (2010).

136. Kim, A., Won, Y., Woo, K., Kim, C.-H. & Moon, J. Highly transparent low

resistance ZnO/Ag nanowire/ZnO composite electrode for thin film solar cells.

ACS nano 7, 1081–91 (2013).

137. Huang, J., Li, G. & Yang, Y. A Semi-transparent Plastic Solar Cell Fabricated

by a Lamination Process. Adv. Mater. 20, 415–419 (2008).

138. Krantz, J. et al. Spray-Coated Silver Nanowires as Top Electrode Layer in

Semitransparent P3HT:PCBM-Based Organic Solar Cell Devices. Adv. Funct.

Mater. 23, 1711–1717 (2013).

139. Gaynor, W., Lee, J.-Y. & Peumans, P. Fully solution-processed inverted

polymer solar cells with laminated nanowire electrodes. ACS nano 4, 30–4

(2010).

140. Lee, J.-Y., Connor, S. T., Cui, Y. & Peumans, P. Semitransparent organic

photovoltaic cells with laminated top electrode. Nano Lett. 10, 1276–9 (2010).

141. Leem, D.-S. et al. Efficient organic solar cells with solution-processed silver

nanowire electrodes. Adv. Mater. 23, 4371–5 (2011).

142. Colsmann, A. et al. Efficient Semi-Transparent Organic Solar Cells with Good

Transparency Color Perception and Rendering Properties. Adv. Energy Mater.

1, 599–603 (2011).

143. Chen, C. et al. Visibly transparent polymer solar cells produced by solution

processing. ACS nano 6, 7185–90 (2012).

144. Guo, F. et al. ITO-Free and Fully Solution-Processed Semitransparent Organic

Solar Cells with High Fill Factors. Adv. Energy Mater. (2013).

doi:10.1002/aenm.201300100

145. 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).

118

146. Hardin, B. E. et al. Laminating solution-processed silver nanowire mesh

electrodes onto solid-state dye-sensitized solar cells. Org. Electron. 12, 875–

879 (2011).

147. 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, 2621–8 (2011).

148. Tress, W., Pfuetzner, S., Leo, K. & Riede, M. Open circuit voltage and IV

curve shape of ZnPc:C60 solar cells with varied mixing ratio and hole transport

layer. J. Photon. Energy 1, 011114 (2011).

149. Morgenstern, F. S. F. et al. Ag-nanowire films coated with ZnO nanoparticles

as a transparent electrode for solar cells. Appl. Phys. Lett. 99, 183307 (2011).

150. Leijtens, T. et al. Hole transport materials with low glass transition

temperatures and high solubility for application in solid-state dye-sensitized

solar cells. ACS nano 6, 1455–62 (2012).

151. Zhang, W., Cheng, Y., Yin, X. & Liu, B. Solid-State Dye-Sensitized Solar Cells

with Conjugated Polymers as Hole-Transporting Materials. Macromol. Chem.

Phys. 212, 15–23 (2011).

152. Garnett, E. C. et al. Self-limited plasmonic welding of silver nanowire

junctions. Nat. Mater. 11, 241–9 (2012).

153. Jiang, X. et al. Highly Efficient Solid-State Dye-Sensitized Solar Cells Based

on Triphenylamine Dyes. Adv. Funct. Mater. 21, 2944–2952 (2011).

154. Cai, N. et al. An organic D-π-A dye for record efficiency solid-state sensitized

heterojunction solar cells. Nano Lett. 11, 1452–6 (2011).